PDF to HTML - Convert PDF files to HTML files

Europaudvalget 2013-14
EUU Alm.del Bilag 180
Offentligt

20 December 2013

Report on appropriate uniform definitions of extremely

high quality liquid assets (extremely HQLA) and high

quality liquid assets (HQLA) and on operational

requirements for liquid assets under Article 509(3)

and (5) CRR

PDF to HTML - Convert PDF files to HTML files

Throughout the global financial crisis which began in mid-2007, many banks struggled to maintain

adequate liquidity. Unprecedented levels of liquidity support were required from central banks in order

to sustain the financial system and even with such extensive support a number of banks failed, were

forced into mergers or required resolution.

The crisis illustrated how quickly and severely liquidity risks can crystallise and certain sources of

funding can evaporate, compounding concerns related to the valuation of assets and capital

adequacy.

In particular, the main observable features of the crisis were the following:



The liquidity position of banks has been seriously hit;

Inappropriate funding structures and scarce liquidity buffers prevailed;

Liquidity stress situations have led, on occasions, to public interventions;

Liquidity stress situations have proved lasting over time.

This situation demonstrated a

prima facie

need for intervention through further liquidity regulation.

Some international regulatory steps on liquidity have been taken:

In December 2010, the Basel Committee on Banking Supervision (BCBS) announced the introduction

of a Liquidity Coverage Ratio (LCR) and Net Stable Funding Ratio (NSFR), to be put in place in 2015

and 2018 respectively (liquidity requirements being part of the Basel III new regulatory framework).

The LCR promotes the short-term resilience of a bank’s liquidity risk profile by ensuring that it has

sufficient high-quality liquid assets (HQLA) to survive a significant stress scenario lasting for one

month. It basically sets the minimum liquidity buffer to bridge liquidity mismatches for one month in a

crisis scenario. The NSFR has a time horizon of one year and is being developed to provide a

sustainable maturity structure of assets and liabilities.

In January 2013, the LCR was updated particularly with regard to the definition of HQLA and on the

implementation timetable (2015 – 2019), while the NSFR is still under development.

Further, regulatory steps on liquidity have been taken at European level:

In June 2013, the European Union published the Capital Requirements Regulation (EU) No 575/2013

(the CRR) and the Capital Requirements Directive 2013/36/EU (CRD IV) which seek to apply the

Basel III framework in the EU. They incorporate the cited liquidity coverage requirements, adapting

them to European specificities, and empower the European Commission to adopt a delegated act to

specify them in detail.

The CRR contains specific mandates for the EBA to develop draft regulatory or implementing technical

standards as well as guidelines and reports related to liquidity in order to enhance regulatory

harmonisation in Europe through the Single Rule Book.

PDF to HTML - Convert PDF files to HTML files

Article 509(3) CRR tasks the EBA with advising on appropriate uniform definitions of liquid assets for

the LCR buffer, and for this purpose defines two categories of transferable assets: assets of ‘extremely

high’ and of ‘high’ liquidity and credit quality. In addition, Article 509(5) CRR states that the EBA shall

report to the Commission on the operational requirements for the holdings of liquid assets. Both

mandates are the purpose of this report.

Furthermore the CRR also tasks the EBA with advising on the impact of the liquidity coverage

requirement, on the business and risk profile of institutions established in the Union, on the stability of

financial markets, on the economy and on the stability of the supply of bank lending

(please refer to a

separate EBA report on the impact assessment of the liquidity measures under Article 509(1) CRR).

These two reports must be submitted to the European Commission by 31 December 2013.

When drawing up this report, and as required by the CRR, the EBA has consulted the ECB and

ESMA. The EBA has also conducted other consultations. In particular, a public hearing, organised on

23 October 2013, a consultation of the EBA’s Banking Stakeholder Group (BSG), multiple bilateral

meetings with diverse stakeholders, as well as an informal roundtable with the banking industry

organised under the aegis of the BSG.

Regarding the consultation with the ECB and ESMA, this consultation has been formally organised

during the second week of November.

ESMA’s input has been provided to the European Commission as an accompanying document to this

report. In substance, ESMA´s feedback, from a different perspective than the supervisory purpose,

focused on two areas within its remit which could have cross effects with the EBA recommendations,

namely secondary markets and financial reporting. From the perspective of transparency of markets,

ESMA indicated that the categorisation of financial instruments as liquid or illiquid, probably

substantiated by an empirical analysis, will in all likelihood be one of the major tasks of ESMA when

designing the future Level 2 framework applying to pre- and post-trade transparency obligations under

the future Markets in Financial Instruments Regulation (MiFIR). The new MiFIR will likely impose

transparency obligations on assets on the basis of their consideration as liquid assets. As a

consequence liquid assets according to EBA´s recommendations could turn out not to be subject to

the most stringent market transparency obligations in so far as they may not fall under the same

definition of liquid assets. ESMA is of the view that this may hinder the marketability of liquid assets

under EBA´s recommendations. This does not necessarily mean that the instrument is not deemed

liquid in a wider sense or that one would not be able to quickly liquidate it in the secondary market or

would only be able to liquidate it with a large loss.

From the financial reporting perspective, ESMA has drawn the attention of the EBA to the way the

liquidity is assessed with regard to IFRS (International Financial Reporting Standards) mainly

regarding those assets measured at fair value. For financial reporting purposes market prices are

considered on the basis of the existence of an active market, which is often considered in conjunction

with the assessment of liquidity. In this regard the liquidity and credit quality are measured on a case

by case basis at the level of a particular financial instrument and not on the basis of a whole asset

class as it is the case of sovereign bonds in the EBA recommendations of liquid assets.

PDF to HTML - Convert PDF files to HTML files

The opinion from the ECB was not available at the time when the report was finalised. As soon as

available, it will be transmitted to the European Commission, together with an analysis from the EBA,

as far as necessary.

The opinion from the ECB was not available at the time when the report was finalised. As soon as

available, it will be transmitted to the European Commission, together with an analysis from the EBA,

as far as necessary.

This report is structured in six parts:



Detailed presentation of the mandate;

Methodology;

Executive summary;

Main findings;

Operational requirements;

Technical appendix: Annexes.

PDF to HTML - Convert PDF files to HTML files

List of abbreviations

ABS: Asset-backed securities

AFME: Association for Financial Markets in Europe

BCBS: Basel Committee on Banking Supervision

BGN: Bulgarian Lev

BIS: Bank for International Settlements

BSG: Banking Stakeholder Group

CCPs: Central Counterparty Clearing House

CDO: Collateralised Debt Obligation

CEBS: Committee of European Banking Supervisors

CIU: Collective investment undertaking

CLN: Credit linked note

CRD: Capital Requirements Directive

CRR: Capital Requirements Regulation

DP: Discussion Paper

EBA: European Banking Authority

EC: European Commission

ECAI: External Credit Assessment Institutions

ECB: European Central Bank

ECBC: European Covered Bond Council

EEA: European Economic Area

EFSF: European Financial Stability Facility

PDF to HTML - Convert PDF files to HTML files

EMIR: European Market Infrastructure Regulation

EMU: European Monetary Union

ES: Effective spread

ESM: European Stability Mechanism

ESMA: European Securities and Markets Authority

ETF: Exchange-traded fund

EU: European Union

EUR: euro

FINRA: Financial Industry Regulatory Authority

GBP: Great British Pound

GC: General Collateral

HQLA: High-quality liquid assets

ICMA: International Capital Market Association

ILR: Iliquidity ratio

IMF: International Monetary Fund

ISIN: International Securities Identification Number

ITS: Implementing Technical Standard

LCR: Liquidity Coverage Ratio

LTRO: Long-term Refinancing Operation

LTV: Loan-to-Value

MBS: Mortgage-backed security

MDB: Multilateral development banks

MiFID: Markets in Financial Instruments Directive

MSDC: Marginal Supply Demand Curve

PDF to HTML - Convert PDF files to HTML files

NDF: Non-Deliverable Forward

NSFR: Net Stable Funding Ratio

OTC: over the counter

PSE: Public Sector Entity

QIS: Quantitative Impact Study

QS: quoted spread

Repo: Repurchase Agreement

RMBS: Residential mortgage-backed security

SPV: Special Purpose Vehicle

STD: Standard Deviation

TTM: Time to maturity

UK: United Kingdom

US: United States

USD: US dollar

XLM: Xetra Liquidity Measure

WAL: weighted average life

PDF to HTML - Convert PDF files to HTML files

Detailed presentation of the mandate

In accordance with Article 509(3) CRR, the EBA shall report to the Commission on appropriate uniform

definitions of high and extremely high liquidity and credit quality of assets for the purpose of Article

416, and on appropriate haircuts for assets of high liquidity and credit quality.

The EBA shall in particular test the adequacy of the criteria below and the appropriate levels for such

definitions:

(a) minimum trade volume of the assets;

(b) minimum outstanding volume of the assets;

(d) credit quality steps referred to in Part Three, Title II, Chapter 2;

(e) proven record of price stability;

(f) average volume traded and average trade size;

(g) maximum bid/ask spread;

(h) remaining time to maturity;

(i) minimum turnover ratio.

Trilogue negotiations have added other categories of asset classes to be considered, in particular (i)

RMBS of high liquidity and credit quality, (ii) other categories of central bank eligible securities or

loans, for example local government bonds and commercial paper and (iii) other non-central bank

eligible but tradable assets, for example equities listed on a recognised exchange, gold, major index

linked equity instruments, guaranteed bonds, covered bonds, corporate bonds and funds based on the

above.

Based on different sources the analysis would need to be aligned to:

assess the appropriateness of the criteria (a)-(i),

set appropriate haircuts for different assets,

provide uniform definitions of high and extremely high liquidity and credit quality.

The report is due by 31 December 2013, after having consulted the ECB and ESMA.

In accordance with Article 509(5)(c) CRR the EBA shall report to the Commission on the operational

requirements for the holdings of liquid assets, as referred in points (b) to (f) of Article 417, in line with

international regulatory developments. The report is due by 31 January 2014.

Both mandates are fulfilled in this single report.

PDF to HTML - Convert PDF files to HTML files

2. Methodology – main features

The report follows the methodology previously published by the EBA in the ‘Discussion Paper on

Defining Liquid Assets in the LCR under the draft CRR’. As a first step, the EBA assessed a range of

asset classes against the fundamental definitions of liquid assets set out in Article 416(3) CRR. Then it

performed a detailed quantitative assessment of the liquidity of individual assets. The estimates of

various liquidity metrics from the quantitative data have been used to produce an ordinal ranking of the

relative liquidity of different asset classes. The relationship between the characteristics of specific

assets and these liquidity metrics has also been assessed to identify the characteristics that are of

particular importance to market liquidity. This evidence has been used to construct specific definitions

of the characteristics of assets that qualify them for consideration as potentially of high or extremely

high liquidity.

2.1 Principles underpinning this report

In the broadest terms, a liquid asset is one which can be converted into cash rapidly with little or no

loss of value. Although the liquidity of an asset depends on market conditions, the quantity to be

monetised and the timeframe considered, there are certain assets that are more likely to generate

funds without incurring large discounts in outright sale or repo markets.

The EBA’s task is to translate the general definition of liquidity, and the criteria and guidance for

classifying liquid assets in the CRR text, into a concrete definition of assets of high and extremely high

liquidity and credit quality based on objective criteria, which could serve as a basis for the

implementation of the LCR in the EU. The

principles

that underpin the EBA’s analysis are listed

below.



although the definitions are explicit, they are

formulated at the level of asset classes, and

not individual assets or ISINs.

It will remain the responsibility of individual banks to identify

that appropriate liquid assets to hold in order to comply with the LCR, working within the broad

asset class-level definitions proposed in this report.



The definition of liquidity

recognises that some asset classes are more liquid than others.

In compiling a comparative ranking of different asset classes, the EBA provides a tool to

distinguish highly - and extremely highly - liquid assets in EU regulation.

When considering the above-mentioned methodology, the following caveats must be taken into

consideration:



Only assets issued in EU currencies have been assessed.

The empirical analysis is

limited to assets denominated in EU currencies;



For some asset classes, conclusions on their eligibility for the liquidity buffer were determined

via a principles-based assessment, generally because of an absence of evidence on market

liquidity.

PDF to HTML - Convert PDF files to HTML files

2.2 Data: methodological overview

The analysis in the present report seeks to analyse the liquidity of a wide range of financial assets

traded in the EU, in order to classify these assets according to liquidity and credit quality. Our

methodology relies on data for the period 1 January 2008 to 30 June 2012. This time period has been

chosen in order to provide a thorough coverage of the liquidity characteristics of the chosen assets

during the financial crisis and the period thereafter.

The choice of assets to be included in this analysis is based on Article 509(3) CRR, which notes that

the liquidity of the following assets should be examined:



assets defined as liquid under Article 416 CRR, except assets referred to in paragraph 1(a)

and (c);

RMBS;

other categories of central bank eligible assets;

other non-central bank eligible but tradable assets, for example equities listed on a recognised

exchange and gold.

On this basis the following asset classes have been identified and data collected on them:



Government bonds;

Government guaranteed bonds

Bonds issued by local government;

Bonds issued by multinationals;

Bonds issued by Central Banks;

Corporate bonds;

Covered bonds;

Bonds issued by promotional banks;

Bank guaranteed bonds;

RMBS;

ABS;

Equities listed on a recognised exchange;

Gold.

Although Government bonds are not explicitly mentioned as an asset class to be analysed,

government bonds have been included to provide a benchmark for the analysis of the liquidity of other

asset classes.

In addition, data on repo activity for relevant asset classes has also been obtained.

The data used in this analysis is based on a number of sources, given the wide variety of asset

classes to be examined. As a result, the type of data and the frequency of observations also varies

across asset classes.

PDF to HTML - Convert PDF files to HTML files

Bond data

Data on government bonds, government guaranteed bonds, bonds issued by local government, bonds

issued by multinationals, bonds issued by central banks, corporate bonds, covered bonds, bonds

issued by promotional banks, bank guaranteed bonds, RMBS and ABS are taken from MiFID

transaction reports obtained from national supervisors in 27 European countries. The MiFID,a

directive which was implemented in November 2007, introduced a standardised set of requirements

across the EU for firms engaged in the provision of investment services to report trading activity for

any instrument admitted to trading on a regulated marked. The MiFID transaction reports therefore

constitute the most comprehensive data set on bond trades throughout the EU.

To verify how well the MiFID transaction reports cover the total EU bond market, the outstanding

amount of the bonds which are traded according to the MiFID data is compared to other sources of the

total outstanding amounts within the different categories of bonds. Chart 1 displays this comparison.

OUTSTANDING BONDS

Chart 1

Bn. EUR

10000

8000

6000

4000

2000

ECB, ECBC or AFM E

Note:

Source:

MiFID: Refers to the MiFID data set based on the securities present in the transactions reports.

MiFID reporting list: Refers to the full list of securities for which transaction reports are required according to the MiFID directive

ECB: General governments and corporates, excluding financials. The reason for the larger outstanding amounts in the MiFID data than in the ECB database is caused by differences in timing

and categorization.

ECBC: Covered bonds

AFME Securitisation Data Reports: ABS and RMBS

Using the other data sources (ECB, ECBC and AFME) as a reference point, the MiFID data seems to

cover the EU government and covered bond market very well in the period considered. The data

coverage is less extensive for corporate bonds. For RMBS and ABS, it is very low. Whereas the

coverage of the MiFID reporting list seems to cover the market for asset backed securities well, the

2009

General government

M iFID

2010

2011

2012

M iFID report ing list

2009

2010

Covered

2011

2012

Corporat e (non-f inancial)

2009

2010

2011

2012

2009

ABS and RM BS

2010

2011

2012

PDF to HTML - Convert PDF files to HTML files

RMBS and ABS transactions actually reported appear to include a much smaller set of instruments,

either due to lack of transactions or missing transaction reports from the financial institutions.

The MiFID reporting rules are primarily intended to be used to detect market abuse; the raw MiFID

data has therefore been cleaned to suit the needs of this analysis, as described in Annex 6. This

annex also describes the data variables from the MiFID data which are included in this analysis.

As the aim of the analysis is to determine the liquidity characteristics of the respective bonds in the

context of a bank’s ability to liquidate these assets in times of institution specific or market wide stress,

transactions with a nominal value of less than EUR 500,000 have been eliminated. This eliminates

transactions of retail size, while not significantly reducing the total turnover coverage of the data

sample. This is in line with standard practice in the academic literature and avoids distortions of some

of the liquidity metrics, which would occur if these small trades were included in the sample.

The table below shows the frequency of transactions for the different bond categories included in the

sample, before and after cleaning. It also shows the percentage of the total turnover in each bond

category, which has been dropped from the sample through cleaning. The relatively high percentage

of turnover that has been eliminated is to a large extent due to the exclusion of very large trades,

which are believed to be due to faulty reporting and not actual trades. As such, individual trades which

are stated to be of a size exceeding 10% of the total outstanding amount of an individual bond are

excluded. The drop in the number of observations is to a large extent due to the introduction of a

minimum transaction size, as noted above. However, this filtering step does not in itself change the

total included turnover volume significantly. Missing information regarding key variables and the

additional filtering steps noted in Annex 6 also contribute to the exclusion of a number of observations.

Bond type

Observations in

raw dataset

Observations in

filtered dataset

Percentage of

turnover

eliminated from

sample

41%

Number of

distinct assets

in filtered

dataset

3,562

Government bonds

incl. central bank

issued bonds and

bonds issued by

promotional banks

Government

guaranteed bonds

Local government

bonds

Multinational bonds

Corporate bonds

excl. financials

Covered bonds

Bank guaranteed

bonds

RMBS

Other ABS incl.

MBS

53,628,168

8,096,077

361,157

407,550

1,704,760

4,616,613

5,333,081

171,300

92,844

58,341

80,206

87,294

116,120

359,275

1,109,277

4,262

34,847

18,526

12%

46%

27%

37%

50%

47%

60%

93%

285

1,190

494

1,583

4,170

154

641

665

PDF to HTML - Convert PDF files to HTML files

It should be noted that the number of observations presented in the above table are the number of

transactions present in the dataset. For the cross asset class analysis, these transactions are

aggregated into observations at daily and monthly frequencies, in order to obtain comparable metrics

with the equity and gold data collected.

In addition to the transaction data obtained from the MiFID transaction reports, and the relevant

metadata also obtained from this source, a number of additional explanatory variables and

characteristics of the chosen assets and their markets were included in the analysis. These include in

particular credit ratings, time-to-maturity and information on regulatory regimes for covered bonds.

Equity data

Equity data was obtained from Bloomberg and Datastream. The data sample is at a daily frequency

and contains price and turnover data for equities included in the main equity indices of the European

countries for which bond data has been obtained. This allows the cross asset class analysis to be

based on the same sample jurisdictions and includes only constituents of major equity indices in each

chosen jurisdiction in the sample. Annex 7 contains further details of the data sample, cleaning

procedures and variables. Observations where the total turnover during one trading day is less than

EUR 500,000 have been eliminated. The reasoning behind this is the same as for bonds. The cut-off

level is however slightly milder than for bonds, in that observations where the daily volume is below

EUR 500,000 are eliminated whereas for bonds observations where the individual transaction size is

below EUR 500,000 are eliminated. This is due to the difference in data structure in the two samples.

In analysing the liquidity characteristics of equities, the EBA also looked at a number of explanatory

characteristics of the individual assets. These include the industry of the company issuing the stocks

and the total market capitalisation of the company. These characteristics may help explain the

variation in liquidity values within the equity asset class, and allow a possible split in the liquidity

degree of equities.

The equity dataset contains daily end of day prices and turnover for 844 distinct equities, resulting in

965,822 observations in total.

Gold data

Gold data refers to trading in physical gold for delivery. Price and spread data is obtained from

Bloomberg, which provides daily data, while volume data is estimated by the World Gold Council from

reports on gold clearing turnover on the London Bullion Market gathered from its members. The

measure is an annual turnover measure and Europe Economics (2013) states that this volume

measure is based on clearing numbers which are typically multiplied by a factor of 5-10 to indicate

gross trade volumes. Therefore the unadjusted measure presented in this report is likely to be a

conservative lower bound. The price data sample reflects the daily frequency; however there was no

daily volume data or outstanding amounts for the gold time series. This results in the inability to

compute some of the liquidity metrics in this paper for the gold asset class.

PDF to HTML - Convert PDF files to HTML files

2.3 Methodology for the analysis of liquidity

This section explains the empirical methodology used to perform the analysis of the liquidity of the

assets chosen in this report. The methodology follows a criteria approach where a common set of

liquidity metrics is computed across asset classes. An ordinal ranking of asset classes and subgroups

in terms of liquidity has been constructed on this basis.

The choice of metrics analysed in this report is in part driven by the mandate of Article 509(4) CRR,

which requires the EBA in particular to investigate the appropriateness of the following criteria in

defining the liquidity of assets, and to set appropriate levels for these measures:



minimum trade volume of the assets;

minimum outstanding volume of the assets;

transparent pricing and post-trade information;

credit quality steps referred to in Part Three, Title II, Chapter 2;

proven record of price stability;

average volume traded and average trade size;

maximum bid/ask spread;

(remaining time to maturity;

minimum turnover ratio.

These measures will enter the analysis at different stages, depending on whether they refer to actual

liquidity measures or can be shown to be characteristics that drive differences in liquidity between and

within asset classes.

The methodology of the paper is characterised by the following steps:



The first step of the analysis involves the assessment of a range of asset classes against the

fundamental definitions of liquidity included in the CRR. A number of additional liquidity

metrics widely used in academic analysis of this topic are also analysed. All these liquidity

metrics are computed for all the instruments in each of the asset classes. The enables a

comparison of asset classes at the aggregate level.



Although one asset class may perform better overall than another, this ranking does not

necessarily apply to all individual instruments within the asset classes. The second step of the

analysis therefore looks further at the characteristics that influence liquidity and credit quality

of assets within asset classes, and thus aims to qualify the results of the overall analysis of

liquidity at the asset class level. This entails testing whether easily observable explanatory

characteristics (e.g. ratings or coupon types) of individual securities within an asset class can

be used to predict their liquidity in quantitative terms. In addition to this quantitative

assessment, qualitative characteristics of assets and markets of relevance to an evaluation of

the inherent liquidity of assets are examined.



In the third step of the analysis, definitions are constructed that individual assets within an

asset class must fulfill in order to be included in the liquid asset buffer as assets of high or

extremely high liquidity and credit quality. An ordinal ranking of these asset classes is

PDF to HTML - Convert PDF files to HTML files

produced in order to determine which asset class falls into which liquidity category. The

observable characteristics (e.g. rating or issue size) are based on the previous step.



Lastly, appropriate haircuts are proposed for the assets found to be of high liquidity and credit

quality. These are based on evidence on historical price movements, while using the haircut

levels suggested by the Basel Committee as a lower bound.

2.4 Measuring liquidity

In the context of this analysis, assets may fundamentally be considered as liquid if they are easily and

immediately convertible into cash with little or no loss of value. Such a definition follows the spirit of the

LCR, which is intended to oblige institutions to hold liquid assets precisely so that they can turn these

into cash if they experience liquidity stress. Whereas such a theoretical definition is relatively easily

formulated, there is no one set way to measure such a definition of liquidity for an asset empirically.

Hence the present analysis is based on a number of different liquidity related measures, each showing

different aspects of the liquidity characteristics of an asset. The set of measures consists of price

impact measures, bid/ask spread measures, trading volume measures, zero-trading days and price

volatility measures. Each measure, or group of measures, provides an insight into the different aspects

of liquidity which when joined give a broad picture of the market liquidity of each asset. For some asset

categories, the measures may be supplemented with additional qualitative judgement. Annex 2

describes the measures in more detail in line with what was included in the ’Discussion

Paper on

Defining Liquid Assets in the LCR under the draft CRR’.

Price impact

The EBA has computed two different measures of price impact: the Amihud illiquidity ratio and an un-

scaled price impact measure. Over short periods, where no new information is available to the market,

the price impact of a given trade reflects the liquidity of the market. In a highly liquid market, individual

trades would not be expected to move the price of the asset and vice versa in illiquid markets.

The first price impact measure is the Amihud (2002) illiquidity ratio. This is defined as the absolute

value of asset return per unit traded. The ratio thereby gives the price impact per unit traded. The

higher the Amihud illiquidity ratio is, the less liquid the asset is deemed to be.

When calculating price impacts across trade sizes empirically, it transpires that price impact does not

depend appear to depend directly on the trade size. Therefore, a second price impact measure which

is not scaled by trade size is calculated. The somewhat surprising result that price impacts are

independent of trade size is supported by similar findings for the Danish bond market in Dick-Nielsen

et.al. (2012).

The EBA has computed the two price impact measures, both at transaction level and at for daily

frequency, using end of day prices. The price impact measures at transaction level give the purest

measure of liquidity, as there is less likelihood that the registered price movements are driven by other

information arriving on the market. However, the EBA did not have transaction level data for equities

and could thus only compute the measures at daily frequencies. In order to ensure using comparable

measures in the cross asset class analysis, the EBA thus also computed the price impact measures at

daily frequencies on the bond data.

PDF to HTML - Convert PDF files to HTML files

It is a requirement that an asset be traded on two consecutive days when computing daily returns in

the price impact measures. This minimises the likelihood that new information has arrived on the

market within the period where the price impact measure is computed.

Monthly price impact measures are defined by taking the monthly median of the daily measure for

each asset. When making cross asset class analysis, the average value across the relevant assets of

the monthly medians has been taken.

The Amihud measure is not computed for gold, as daily trading volume data for this asset class was

not available.

Bid/ask spread

Bid/ask spreads are a commonly used measure of liquidity. Tight bid/ask spreads indicate that one is

able to sell a portfolio with little loss of value. For bond data, the EBA does not have bid and ask price

series as such. Instead the Roll measure has been computed. Roll (1984) shows, that under certain

conditions, the relative effective bid/ask spread equals two times the square root of minus the

covariance between consecutive returns. The intuition is that the traded price of an asset bounces

back and forth between bid and ask prices. This will result in negative covariance between

consecutive returns, and higher bid/ask spreads lead to a higher negative covariance. Hence, the

more illiquid an asset is, reflected by a high bid/ask spread, the higher the Roll measure will be.

For equity data actual bid and ask end of day quotes are available, hence the bid/ask spread can be

computed directly on that basis. However, for some jurisdictions, the quality of the bid/ask spread data

is not satisfactory. Hence, the Roll bid/ask proxy for equities has also been computed. This also allows

for more direct comparability with the bond data.

As for the price impact measures, the Roll measure has been computed based on daily end of day

data. A monthly Roll measure has been computed by looking at the monthly covariance of the

computed daily returns.

Trading volume and turnover

High trading volume and turnover indicates high trading activity and the ability to liquidate sizeable

portfolios within a short time period. Although high trading volume and turnover indicate assets are

liquid, the opposite does not necessarily hold true. Some assets may be in high demand without being

traded much, and could easily be liquidated if needed.

Two measures of trading volume have been considered. Firstly, trading volume can be used as an

indicator for liquidity, as actively traded assets are considered to be more liquid than assets that are

traded to a lesser degree. Secondly, the EBA computed the turnover, which is given by the trading

volume divided by the outstanding amount of the relevant asset. The turnover ratio relates the traded

volume to the total outstanding amount, and thus the inverse of the turnover ratio can be interpreted

as the average holding time of the asset. That is to say, a turnover ratio of 0.5 for monthly data implies

an average holding period of 2 months.

PDF to HTML - Convert PDF files to HTML files

Both measures are computed on a monthly basis and when making cross asset class analysis, the

average value of the monthly measures across the relevant assets has been taken.

The trading volume and turnover measures are not computed for gold, as the EBA did not have daily

trading volume data for this asset class.

Zero-trading days

Assets with lower liquidity will tend to trade less frequently than assets with higher liquidity, hence

looking at the number of zero-trading days of an asset over a fixed period will also yield a proxy for the

liquidity of an asset. Although a high number of zero-trading days would in general be a sign of

illiquidity, this is not necessarily always the case for the same reasons as mentioned above. A monthly

zero-trading days measure has been computed by taking the percentage of trading days during a

month where an asset is not traded. A high value of this variable will indicate a relatively illiquid asset.

When making the cross asset class analysis, averages of the monthly zero-trading days across the

relevant assets have been taken.

Volatility

Volatility is relevant to this analysis in two respects. Firstly, the volatility of an asset’s return has been

shown to be linked to the ease and cost of liquidating that asset. High volatility implies high price

uncertainty, which in turn hampers liquidity. Secondly, high price volatility by definition equals large

changes in the value of an asset, thereby creating large swings in the liquidated value of that asset,

another aspect of liquidity which is relevant for this analysis. On this basis, the analysis includes a

measure of return volatility determined as the monthly standard deviation of daily returns. When

making the cross asset class analysis, the average value of the monthly measures across the relevant

assets has been taken.

In addition, the maximum absolute price change over a 30-day period has been computed. This is

suggested as a measure of price stability in the Basel Committee’s LCR proposal, and the maximum

price decline over a 30-day period is included as an asset characteristic requirement in the definition of

the liquidity buffer.

2.5 Cross asset class comparison

The cross asset class comparison is made by computing the previously noted liquidity measures and

aggregating to a monthly frequency. Time series for each asset class are computed by taking

averages across assets within the respective asset class.

In addition, probability distributions are produced for each asset class, based on the liquidity measures

computed for each asset in the relevant asset class. These probability distributions thus reflect

variation in liquidity measures both across assets within that asset class and across time.

Based on the time-series and probability distribution evidence a coarse ranking of asset classes by

liquidity levels can be produced. For this ranking, a correlation matrix of the different liquidity measures

is initially compiled, across asset classes. This is used in order to see whether a grouping of some of

the metrics into metrics describing similar liquidity features of the assets in our sample is possible.

PDF to HTML - Convert PDF files to HTML files

This step of the analysis results in a very coarse and initial liquidity ranking. It must be qualified further

by the within asset class analysis. In particular, our data sample is not balanced with respect to time

period coverage of individual asset classes, credit ratings, time to maturity or any other asset

characteristics, that may explain the liquidity of individual assets. If it is shown that credit quality is a

determinant of liquidity for individual assets, this would imply that the covered bond sample as a whole

may be found to have better liquidity characteristics than corporate bonds and government bonds,

simply because of this sample bias. In reality, corporate bonds and covered bonds of the same credit

quality might be equally liquid. Further analysis of liquidity measures for assets, taking into account

such explanatory variables, will be done in the next step, in order to give a final ordinal ranking. For

this reason, while the coarse ranking presented at this stage of the analysis may give an initial insight

into the relative liquidity of asset classes, it will not necessarily be fully coherent with the final asset

group liquidity ranking.

2.6 Within asset class analysis

The second step of the analysis involves looking further at the characteristics that determine liquidity

and credit quality within asset classes. In the second stage one looks at the liquidity measures defined

above at the individual asset level, and not aggregated to asset class level.

The EBA has identified the theoretically relevant explanatory characteristics of the liquidity and credit

quality of different asset class subgroups, and tested whether these do in fact help predict the liquidity

of asset subgroups.

The EBA used regression analysis to identify the explanatory ability of the different asset subgroup

characteristics on the different liquidity metrics defined. On the basis of this regression evidence, an

asset class has been divided into subgroups, based on the characteristics of the individual

observations. These buckets of assets are then used in the final step of the analysis, where an ordinal

ranking is produced.

2.7 Uniform definitions

Based on the quantitative analysis, the third step of the analysis is to create actual liquid asset

subgroups based on relevant asset characteristics. The asset characteristics are identified during the

within asset class analysis. These could e.g. include time to maturity for bonds. For each asset an

average value of each liquidity metric over time has then been computed. However, if an asset jumps

from one explanatory bucket to another over the life of the sample, then a separate average liquidity

measure is calculated for the asset in each of these buckets. If one assumes time to maturity is one of

the explanatory buckets, and one divides this into assets with time to maturity below 5 years and

above 5 years, then an asset that starts off with 6 years to maturity, will during our sample period jump

from the above 5 year bucket to the below 5 year bucket. Hence when computing average values over

time, one would include two values for this one asset: one including the average over the monthly

observations in the time period where the bond had a time to maturity above 5 years, and one over the

monthly observations during the time where the bond had a time to maturity below 5 years.

Once these average values of liquidity metrics and buckets are created and assigned, the EBA looked

at how well groups of assets within different buckets perform on the liquidity measures. In setting the

distinction between assets of extremely high liquidity and credit quality, and assets of high liquidity and

PDF to HTML - Convert PDF files to HTML files

credit quality, the EBA used government bonds as a benchmark. Hence the EBA picked a

homogeneous group of government bonds which, based on their liquidity metric values, can be viewed

as assets of extremely high and high liquidity and credit quality. These government bond buckets are

then compared to all the different asset buckets previously created, on the basis of asset class and

explanatory characteristics, and identify buckets which have liquidity metric distributions equal or close

to those of the government bond benchmark. The ordinal ranking is compiled on this basis and

aggregated into a basket of assets defining assets of extremely high liquidity and credit quality and a

basket of high liquidity and credit quality. Any asset buckets that have liquidity metric values below the

lower bound of assets with high liquidity and credit quality are excluded.

This step also involves some use of qualitative judgment. While the quantitative analysis performed

tells us something about observed market transactions, it does not necessarily tell us whether a given

asset would be easy to liquidate in the market. A high quality government bond may be held on the

book of a pension fund for several years, thus not showing any turnover. This bond would by most of

the quantitative metrics suggested previously in this paper be considered illiquid, however in reality it

would be very easy for the pension fund to liquidate this position if it so wished. While the main

element of this step is the empirical analysis, some qualitative judgment should also be taken into

account when making the final liquidity grouping and ordinal ranking of assets.

2.8 Haircuts

The final step of the analysis involves assessing appropriate haircuts for the asset classes defined as

of high liquidity and credit quality. This is done by calculating the 30-day price change metric for the

relevant asset groups. The haircut levels suggested by the Basel Committee (15% for covered,

corporate, sovereign and public sector securities in Level 2A, 25% for RMBS in Level 2B and 50% for

equities and corporate bonds rated between A+ and BBB-) should be used as a lower bound in the

setting of these haircuts.

PDF to HTML - Convert PDF files to HTML files

3. Executive Summary: assessment and policy recommendations

This report comprises the EBA’s response to two separate mandates given to it in the CRR text. It

meets the requirements of Article 509(3) by proposing uniform definitions of liquid assets eligible for

the LCR liquidity buffer in the EU, and recommending appropriate haircuts for the subset of the assets

defined as having high liquidity and credit quality. It meets the requirements of Article 509(5) by setting

out views on appropriate operational requirements that eligible liquid assets should also adhere to if

they are to be included in banks’ LCR buffers. The executive summary discusses the main findings of

the report in these two areas, starting with the work on uniform definitions of liquid assets.

The report sets out to devise uniform definitions of eligible liquid assets issued in the EU that:

(i)

(ii)

(iii)

(iv)

are grounded in observed market liquidity;

are easy to apply using public information;

provide a uniform level of liquidity across asset classes;

ensure that as low a proportion as possible of illiquid assets are inadvertently classified as

liquid;

(v)

attempt to achieve as broad a definition of liquid assets as possible, subject to satisfying

conditions (i) to (iv) above.

These definitions are founded on the analysis of representative data on the trading of relevant assets

in the outright sale and repurchase markets. MiFID transaction data is used for the bond asset

classes, Bloomberg and Datastream data for equities and Bloomberg and World Gold Council data for

gold. The data period extends from 1 January 2008 to 30 June 2012. Data on the repo market is less

detailed and for the purpose of this analysis is taken from regular ICMA European repo market

surveys and from surveys completed by European market participants, including Eurex. This data is

mainly used as supporting evidence, the bulk of our analysis relates to data on outright sale markets.

The data gathered on outright bond trades has been cleaned to ensure errors in reporting are

eliminated, including the removal of duplicate trades, inaccurate prices and trade sizes, smaller retail

trades, and transactions which do not represent a secondary market trade between distinct

counterparties. A sample of close to 10 million transactions covering more than 10,000 individual

securities remains after this process of data cleaning. Over 950,000 daily observations on 844 distinct

equities are also utilised. The liquidity of the following asset classes is assessed empirically:



government bonds;

government guaranteed bonds;

bonds issued by local government;

bonds issued by multinationals;

corporate bonds;

covered bonds;

bonds issued by promotional banks;

bank guaranteed bonds;

PDF to HTML - Convert PDF files to HTML files



RMBS;

ABS;

equities listed on a recognised exchange;

gold.

A range of liquidity metrics including those specified in the CRR mandate and others in common

usage in the academic literature have been calculated to enable the market liquidity of each asset

under investigation to be tracked through time.

The relative liquidity of the different asset classes covered in the report has been compared to provide

an initial insight into differences in market liquidity, and to help identify which metrics are, in

combination, sufficient to capture all facets of liquidity within our sample adequately. At an aggregate

level this analysis showed that sovereign bonds were the most liquid assets in the sample, followed

closely by covered bonds and some other forms of public sector securities. Corporate bonds were

shown to be moderately liquid, while ABS, gold and equity showed low levels of market liquidity on a

number of important metrics. However, this preliminary analysis did not attempt to control for

heterogeneities within asset classes, and was hence not sufficient in isolation for the purposes of

defining liquid assets.

A range of potentially useful explanatory characteristics was assessed using econometric analysis

performed within each distinct asset class. This enabled the key attributes suitable for use in a set of

definitions to be identified and their relationship with liquidity to be better understood. In general, for

sovereign and public sector debt, corporate and covered bonds and all types of ABS, credit rating,

time to maturity and issue size were found to be significant determinants of liquidity in each asset

class. In addition, in the covered bond asset class, indicators of the strength of the covered bond

regulatory regime were also found to be useful determinants of market liquidity.

Uniform definitions of liquid assets were finalised via a detailed analysis of all possible combinations of

definitions using the key explanatory variables identified for each asset class. Definitions that reliably

identify liquid assets within each asset class with a low error rate, while ensuring maximum possible

coverage of the set of liquid assets in that asset class, were identified via examinations of plots of the

distribution of type I and type II errors for each possible definition. Once a subset of definitions with the

highest informational content was identified for each asset class, uniformity of liquidity across asset

classes was assured via selecting appropriate definitions with reference to the typical liquidity of the

set of sovereign bonds allocated to the extremely highly and highly liquid categories. Some tolerance

was allowed through the requirement to be as liquid as the relevant sovereign bonds at each liquidity

level being applied in a flexible manner.

For some asset classes conclusions on their suitability for the liquidity buffer were determined via a

principles-based assessment, generally because of an absence of evidence on market liquidity.

Data on observed 30-day price moves for the assets defined as of high liquidity and credit quality were

assessed to ascertain whether the minimum proposed haircuts applied to these assets in the Basel

text were sufficient to guard against likely fluctuations in market prices

This report combines the EBA’s responses to the mandates in CRR Articles 509 (3) and 509 (5).

PDF to HTML - Convert PDF files to HTML files

In the report an empirical and qualitative analysis of the suitability of a range of asset classes for

designation as assets of high or extremely high liquidity and credit quality is performed. A set of

uniform definitions is proposed for this purpose. These definitions are designed to ensure that banks in

the EU can have a high degree of certainty that the assets they hold in their LCR liquidity buffer will be

reliably liquid in a combined idiosyncratic and market-wide stress.

The report concludes that the minimum haircuts set for assets of high liquidity and credit quality in the

Basel LCR text would be sufficient for the assets defined in that category in this analysis.

The report also builds on the CRR text to make specific recommendations about how operational

requirements for assets in the LCR liquidity buffer should be designed.

The EBA recommends that the analysis performed in this report be repeated on a regular basis to

ensure that the uniform definitions provided for EU assets remain appropriate over time.

3.1 Empirical conclusions on the definitions of liquid assets

The empirical approach of the report leads to the following conclusions:

Definitions of assets of extremely high liquidity and credit quality

EEA sovereign bonds issued in domestic currency, rated ECAI 1, with a minimum issue size of

EUR 250 million (or the local currency equivalent).

EEA covered bonds rated ECAI 1 with a minimum issue size of EUR 500 million (or the local currency

equivalent) and subject to additional conditions relating to the regulations governing the covered bond

structure (as set out below).

Note that this analysis has not sought to assess the liquidity of central bank reserves and holdings of

notes and coin, which are assumed to be liquid assets by definition.

Definitions of assets of high liquidity and credit quality

Sovereign bonds issued in domestic currency rated ECAI 2 or above, of minimum issue size

EUR 100 million (or the local currency equivalent).

Covered bonds rated ECAI 1 of minimum issue size EUR 250 million (or the local currency

equivalent).

Corporate bonds rated ECAI 4 or better, of minimum issue size EUR 250 million (or the local currency

equivalent) and a maximum time to maturity of 10 years.

RMBS rated ECAI 1 of minimum issue size EUR 100 million (or the local currency equivalent) and a

maximum time to maturity of 5 years subject to the additional regulatory requirements listed below.

Bonds issued by supranational institutions in EEA currencies, rated ECAI 1, of minimum issue size

EUR 250 million (or the local currency equivalent).

PDF to HTML - Convert PDF files to HTML files

Bonds issued by local government institutions in EEA currencies, rated ECAI 2 or above, of minimum

issue size EUR 250 million (or the local currency equivalent) and a maximum time to maturity of 10

years.

Additional regulatory requirements for covered bonds:

The empirical investigation of covered bonds has shown that the liquidity characteristics of covered

bonds differ across differing covered bond regulations. On this basis, covered bonds admitted to Level

1, should be issued under stringent covered bond regulation, which is designed to reduce the credit

risk faced by bond holders and provide standardized and transparent information to prospective

issuers. In practice our tests have shown, that covered bonds issued under regimes for which all of the

following applies qualify as Level 1 assets:



the covered bonds are eligible for the treatment set out in Article 129(4) or (5) CRR or bonds

referred to in Article 52 (4) of Directive 2009/65/EC;



a special licence is required from the local supervisor for the issuing of covered bonds. In

order to obtain this license, the issuer must respect stricter risk management requirements

compared to those applying to general banking supervision regimes.



Additional supervisory reporting or transparency/disclosure duties are applied to the covered

bond issuer concerning the cover pool, which go beyond the requirements of regular banking

supervision regimes.



Mandatory minimum overcollateralisation is required.

Exposure to market and liquidity risk is reduced by law or contract.

In the event that an issuer becomes insolvent, covered bondholders are protected against

claims from other creditors via preferential treatment in the creditor ranking and as regards

access to the assets in the underlying cover pool.



Special LTV limits are used for calculating collateralisation rates for the cover pool.

Additional regulatory requirements for RMBS:



All assets backing an RMBS must be first-lien residential mortgages.

Only senior tranches can be included.

The time to maturity of the security shall be calculated using either the first call date of the

RMBS or the lowest of either the pricing prepayment assumption as defined in the offering

circular or 20% conditional prepayment rate (CPR).



For securities which have been issued and outstanding for more than one year and which do

not include a call option, the prepayment speed (CPR) to be used when calculating the WAL

shall be the arithmetic average of:

PDF to HTML - Convert PDF files to HTML files

a. the historical CPR observed since issuance (the lifetime CPR);

b. the lower of the (i) the pricing CPR as defined in the offering circular or (ii) the 20% CPR.

Assets found to be insufficiently liquid

Based on the empirical and principles-based analysis in this report, there is insufficient evidence of

market liquidity to propose the following asset classes as assets of high liquidity and credit quality:



equities;

gold;

ABS not backed by residential mortgages;

credit claims;

securities issued by financial institutions;

central bank securities;

bank-issued government guaranteed bonds;

bonds issued by promotional banks.

3.2 Conclusion on operational requirements of liquid assets

The analysis of operational requirements to be applied to assets deemed eligible for the LCR liquidity

buffer is founded on the requirements set out in the CRR - Article 417(b) to (f) - and in the Basel LCR

text according to the mandate. In general it concludes the requirements are well founded and will

ensure a high degree of certainty that a bank’s liquid assets will serve their purpose in a stress

situation. The EBA does however suggest clarification or review of the conditions outlined below.

Legal and practical availability

The concept of availability is fundamental to operational requirements for liquid assets, since it

determines the practical usability of the liquidity generating capacity of liquid assets. It is therefore

important that any aspect of availability be considered under a liquidity perspective which, in some

cases, might differ from a solvency perspective. This is especially true when applying the concept of

unencumbrance, where recent reporting requirements take a solvency perspective and focus on the

availability for unsecured debitors in the event of default, instead of on short-term liquidity generating

capacity.

The EBA recommends that for the purpose of liquidity reporting, the notion of unencumbrance follow

that of paragraph 31 of Basel III: Liquidity Coverage Ratio and liquidity monitoring tools, January 2013

which states that

PDF to HTML - Convert PDF files to HTML files

“Unencumbered”

means free of legal, regulatory, contractual or other restrictions on the ability of the

bank to liquidate, sell, transfer, or assign the asset”.

The EBA further recommends that

any

liquid asset or other integrant of the liquidity buffer

held

in a

country with transfer restrictions should only be used to cover outflows in that respective country. Any

excess holding of liquid assets should not be admissible for use to cover outflows in other countries.

To restrict the application of this requirement to assets under Article 416(1)(c) CRR is not compatible

with a prudential approach.

Likewise the EBA recommends that any liquid asset denominated in a non-convertible currency should

only be used to cover outflows in that currency, regardless of whether the currency risk has effectively

been hedged or not.

Control by a liquidity management function

The EBA recommends that the requirement not explicitly demand that liquid assets be held by a

specific function (e.g. treasury) or that liquid assets be held in a separate book, but that institutions

have policies and procedures in place that establish the proper authority of the liquidity management

function and ensure sufficient operational capacity.

As different liquidity management models co-exist in the EU, the interpretation of the criteria should

reflect and allow this diversity.

Test sales

The EBA recommends that supervisors apply a principles-based approach to test sales, considering

the normal trading activity of an institution and the costs of test sales. In line with the CEBS Guidelines

on Liquidity Buffers and Survival Periods (Annex 2), based on the proportionality principle, smaller

banks which access markets through another institution, will, in most cases, not have to be active in

several advanced money and capital markets.

In order to achieve regulatory transparency and preclude any circumventing of the operational

requirements by institutions a harmonised definition of “liquidity risk that an institution incurs in a

Member State or a third country” should be prepared and rules governing the treatment of assets that

fall under two or more definitions of eligible assets.

Not used in ongoing operations

The EBA recommends allowing the use of liquid assets in hedging or trading strategies as long as

they can easily be monetised without conflicting with existing risk limits that control for market, credit

and counterparty risk.

To ensure consistency with Article 417(c) CRR, treasury should be interpreted as “liquidity

management function”.

PDF to HTML - Convert PDF files to HTML files

Currency consistency

A perfect match of currency consistency of liquid assets and net outflows is not required according to

the general liquidity requirement under Article 412(1) CRR (or Basel III). Until a specific currency

requirement is specified in detail, institutions should monitor risk, test the effectiveness of their hedging

strategies and develop contingency plans for stressed conditions.

3.3 EBA recommendations

The EBA is of view that the empirical conclusions on definitions of transferable assets of high

and extremely high liquidity and credit quality included in the report item 3.1 should be

supplemented by a qualitative/expert judgment substantiated basically on the supervisory

advice to stress the great importance attached by the EBA to the alignment of the EU

legislative framework with the international standards defined by the Basel Committee on

Banking Supervision.

Transferable assets of extremely high liquidity and credit quality:

Particularly the EBA recommends to consider

all sort of bonds issued or guaranteed by EEA

Sovereigns and EEA Central Banks in the domestic currency and also those issued or

guaranteed by Supranational Institutions

(the Bank for International Settlements, the International

Monetary Fund, the Commission, multilateral development banks, the European Financial Stability

Facility and the European Stability Mechanism)

as transferable assets of extremely high liquidity

and credit quality:

Although the empirical analysis shows some degree of differentiation in the liquidity features of

different sovereign bonds, there are strong supervisory arguments for not discriminating between

various Member State sovereigns. In particular, it should be taken into consideration the eventual and

likely damage that an exclusion of some sovereign bonds would create in the European market by

triggering incentives to invest in domestic sovereign, thus reinforcing the fragmentation of the Single

Market and the sovereigns-banks loop.

The analysis shows that some covered bonds display an excellent liquidity based on the available

data, which reflects the European covered bond market. Nevertheless, two thirds of the observations

come from markets that did not experience a real estate crisis. There are doubts as to whether the

findings of the current analysis are sufficient to justify a deviation from the international standards and

the inclusion of some covered bonds in the category of extremely high quality liquid assets the

characteristic of which is to allow unlimited recourse to such instruments to cover for liquidity

requirements.

Transferable assets of high liquidity and credit quality:

The EBA recommends following the empirical definitions of

transferable assets of high liquidity and

credit quality

regarding covered bonds, corporate bonds, RMBS and bonds issued by local

government institutions. Furthermore the EBA recommends considering as HQLA common equity

PDF to HTML - Convert PDF files to HTML files

shares in accordance with the requirements established in par. 54 (c) of the text of Basel III: The

Liquidity Coverage Ratio and liquidity risk monitoring tools, Jan. 2013 .

Equities show high volume based metrics and are found to be very active as collateral in terms of

repoability. The international regulatory framework foresees this asset class under the application of

high haircuts.

From the empirical analysis

some whole asset classes do no turn out to meet the empirical

requirements neither to be deemed as extremely HQLA nor HQLA.

In EBA’s view there are not

sufficiently robust qualitative arguments to upgrade the following: Gold, Asset Backed Securities not

backed by residential mortgages, Credit Claims, Securities issued by Financial Institutions (unless

they are guaranteed by Sovereigns) and bonds issued by promotional banks.

On the other hand since the data used for the analysis is based on EEA assets these

recommendations would be applicable for EEA assets. The EBA recommends for non EEA assets to

apply the regulation of the corresponding local jurisdictions in this respect.

Finally the EBA recommends following the conclusions reached in the report on operational

requirements of liquid assets and on haircuts for the assets of high liquidity and credit quality.

“Common equity shares that satisfy all of the following conditions may be included in Level 2B, subject to a 50%

haircut:



not issued by a financial institution or any of its affiliated entities;



exchange traded and centrally cleared;



a constituent of the major stock index in the home jurisdiction or where the liquidity risk is taken, as

decided by the supervisor in the jurisdiction where the index is located;



denominated in the domestic currency of a bank’s home jurisdiction or in the currency of the jurisdiction

where a bank’s liquidity risk is taken;



traded in large, deep and active repo or cash markets characterised by a low level of concentration; and



have a proven record as a reliable source of liquidity in the markets (repo or sale) even during stressed

market conditions, ie a maximum decline of share price not exceeding 40% or increase in haircut not

exceeding 40 percentage points over a 30-day period during a relevant period of significant liquidity.

PDF to HTML - Convert PDF files to HTML files

Contents

1. Detailed presentation of the mandate .................................................................................................. 8

2. Methodology – main features .............................................................................................................. 9

2.1 Principles underpinning this report ............................................................................................... 9

2.2 Data: methodological overview ................................................................................................... 10

2.3 Methodology for the analysis of liquidity ................................................................................... 14

2.4 Measuring liquidity ..................................................................................................................... 15

2.5 Cross asset class comparison ....................................................................................................... 17

2.6 Within asset class analysis .......................................................................................................... 18

2.7 Uniform definitions ..................................................................................................................... 18

2.8 Haircuts ....................................................................................................................................... 19

3. Executive Summary: assessment and policy recommendations........................................................ 20

3.1 Empirical conclusions on the definitions of liquid assets............................................................ 22

3.2 Conclusion on operational requirements of liquid assets ............................................................ 24

3.3 EBA recommendations................................................................................................................ 26

Contents ................................................................................................................................................. 28

4. Main findings .................................................................................................................................... 30

4.1 Quantitative analysis ....................................................................................................................... 30

4.1.1 Quantitative analysis of market liquidity across asset classes .................................................. 30

4.1.2 Quantitative analysis of market liquidity within asset classes ................................................. 38

4.1.3 Assessing appropriate uniform definitions ............................................................................... 51

4.1.4 Computation of Haircuts .......................................................................................................... 55

4.2. Principles-based analysis of asset classes................................................................................... 55

4.2.1 Discussion of the rationale backing the principles ................................................................... 56

4.2.2 Mapping between asset classes and principles ......................................................................... 60

4.2.3 Conclusions of the principles-based analysis ........................................................................... 66

4.3. Quantitative analysis of repo market data ................................................................................... 67

4.3.1 Summary of the European repo market .................................................................................... 67

4.3.2 The EU repo market during the financial crisis ........................................................................ 68

4.3.3 Developments in the EU repo market ....................................................................................... 69

4.3.4 Data description ........................................................................................................................ 69

4.3.5 Data analysis for the EU repo market ....................................................................................... 70

4.3.6 Summary of the EU repo market .............................................................................................. 84

5. Operational requirements of liquid assets ......................................................................................... 85

5.1 Availability .................................................................................................................................. 86

5.2 Control by a liquidity management function ............................................................................... 91

5.3 Test sales option .......................................................................................................................... 92

5.4 Not used in ongoing operations ................................................................................................... 93

5.5 Currency Consistency .................................................................................................................. 95

PDF to HTML - Convert PDF files to HTML files

Annex 1: Relevant CRR Articles on uniform definitions of liquid assets ......................................... 96

Annex 2: Literature on market liquidity .......................................................................................... 102

Annex 3: Literature on the liquidity of different asset classes ........................................................ 109

Annex 4: Survey of liquidity metrics .............................................................................................. 115

Annex 5: Responses to the EBA Discussion Paper on the methodology to be used in this report . 121

Annex 6: Bond data description ...................................................................................................... 126

Annex 7: Equity data description .................................................................................................... 128

Annex 8: Definition and implementation of liquidity metrics ........................................................ 134

Annex 9: Time series for cross asset class analysis ........................................................................ 137

Annex 10: Probability distributions for cross asset class analysis .................................................. 143

Annex 11: ECAI rating mapping..................................................................................................... 148

Annex 12: Relevant CRR articles on operational requirements ...................................................... 149

Annex 13: Operational Requirements as set out in “Basel III: The Liquidity Coverage Ratio and the

liquidity risk monitoring tools”, Jan 2013 BIS ................................................................................ 151

Annex 14: Guideline on Liquidity Buffers & Survival Periods, CEBS 9 December 2009 ............. 154

References ....................................................................................................................................... 156

PDF to HTML - Convert PDF files to HTML files

4. Main findings

4.1. Quantitative analysis

4.1.1 Quantitative analysis of market liquidity across asset classes

The first stage in our quantitative analysis is a comparison of liquidity measures across asset classes

at the aggregate level. This enables us to assess some basic stylised facts about the different liquidity

properties of the asset classes under investigation and to observe the properties of the liquidity metrics

used in our analysis. This stage will provide the platform for a more detailed examination of the

liquidity variation within asset classes on the basis of the relevant liquidity characteristics (section

4.2.2). Finally, our findings can be combined into an analysis of appropriate definitions across asset

classes, based on the relevant liquidity characteristics (section 4.2.3).

This section presents cross asset class results for all the liquidity metrics analysed in this report. In the

following subsections, the results for each measure are given in tables summarising the average and

variation for across assets over the considered sample period. In Annexes 9 and 10, detailed results

are shown in charts of time series and probability distributions taking into account variations in the

metrics over time and within asset classes.

The original sample of data presented in section 3 is used to compute monthly observations for all the

liquidity metrics for each asset in the sample. Not all assets have observations for all metrics at all

points in time however. In some cases, this is simply because the asset in question did not exist for

some of the time period used in this analysis, i.e. it was either not issued or it had matured. For

equities, it may also be the case that these only enter our sample for a sub-period, as they were not a

part of the chosen equity indices for the full sample period.

Another important reason for missing values is that some metrics require actual trading to take place

in order to be calculated. As many of the individual assets in our sample are not traded every day or

even every month, there will not be a full time series of liquidity metrics for these assets. This should

be kept in mind when viewing time series and probability distribution evidence of asset class

performance. On this basis, tables showing the frequencies of observations and average number of

observations of each metric for each asset within an individual asset class are presented at the

beginning of each metric section.

Price impact

The two price impact measures computed are the Amihud illiquidity ratio and a pure price impact

measure. The price impact measures show how much a given trade moves the market price. The

intuition is that for highly liquid assets, individual trades without informational value will not move the

price of the asset. For non-liquid assets, the lack of market depth will result in larger price movements.

The Amihud illiquidity ratio scales the price impact by the trade volume, resulting in a measure of the

price impact per euro traded, whereas the second price impact measure is unscaled.

As is evident from the table below, government bonds generally have the lowest Amihud illiquidity

measure, followed closely by covered bonds. The variation in the Amihud ratio within the government

and covered bond samples are of approximately the same magnitude, and the time series and

PDF to HTML - Convert PDF files to HTML files

probability distribution evidence shown in Annexes 9 and 10 underline the view of two very similar

series, with the level of the government bond series being slightly below the covered bond series.

Corporate bonds and ABS show average values of the Amihud ratio that are relatively close. However,

volatility within the individual asset categories is somewhat higher for ABS than for corporate bonds in

general. Time-series evidence shown in Annex 9 gives a similar picture, with a volatile series for ABS

in particular. This is in part due to the relatively low number of observations available for this asset

class, as noted in the table below. The low number of observations is in itself the result of relatively

sparse trading in this asset class, supporting the evidence of fairly poor liquidity performance of this

asset class on this measure. The Amihud ratio of corporate bonds appears to spike by an equivalent

magnitude at the height of the financial crisis, with the ABS series spiking even further. The Amihud

illiquidity ratio of equities has by far the highest value. This would in principle imply that equities are

the least liquid asset class based on the Amihud ratio, as the price impact per unit traded is the

highest. However, we must be aware of the limitations of using a metric of this kind, which is based on

daily returns, to compute liquidity effects. For equity markets in particular, the high frequency of trading

and informational trading, imply that the Amihud metric is more an indicator of price volatility in general

than of the liquidity impact of trades.

Asset class

Amihud illiquidity

ratio

Government

bonds*

Covered bonds

Corporate bonds

(excl. financials)

ABS (incl. MBS

and RMBS)

Equities

Gold

(%-points

0.059

(0.1965

0.090

(0.205)

0.188

(0.303)

0.208

(0.591)

0.406

(0.886)

per million euro)

0.328

(0.548)

0.265

(0.434)

0.370

(0.457)

0.773

(1.495)

1.554

(0,953)

0.746

(0.322)

Note: Average measure over the sample period and standard deviations in brackets.

23,852**

43.5**

2,316

3.8

14,075

12.8

29,831

13.1

51,164

Price impact (in

%-points)

Number

observations

Average

per asset

12.7

number

observations



*Includes bonds issued by central governments, bonds issued by local governments, bonds

guaranteed by central or local governments, bonds issued by multinationals, bonds issued by

promotional banks and bonds issued by central banks.

**Based on Amihud

For the unscaled price impact measure, the picture is slightly different than for the Amihud ratio. Here,

covered bonds generally show the lowest values, both on average and from time-series and

probability distribution evidence. The price impact measure, on average, peaks to the same value for

PDF to HTML - Convert PDF files to HTML files

covered bonds and government bonds during the financial crisis in 2008-2009, while government

bonds show higher price impact values on average in 2011-2012 than covered bonds.

The performance of corporate bonds is better than ABS. Time series evidence (annex 9) also shows a

very large peak for the ABS series during the financial crisis, indicating relatively large sensitivity of

this asset class to market unrest. Gold performs at a similar level to ABS.

Equities in general have higher values for the price impact measure than bonds, indicating relatively

high day-to-day price volatility. Similar to evidence on the Amihud illiquidity ratio, to what extent this

can be seen as an indicator of low liquidity or a higher frequency of informational trading in equities

than bonds, is unclear. The relatively large number of observations for the measures per asset for

equities indicates that these assets are traded more frequently than bonds.

Bid/ask spread

As we do not have actual bid/ask spread data for the bonds in our sample, we proxy the relative

spread by the Roll measure. As we require at least 6 trades in a given asset over a one month period

in order to compute a Roll measure, the number of observations for the Roll measure is somewhat

lower than for some of the simpler metrics. This also means that the least frequently traded assets in

our sample are not included in the Roll measure calculations. However, this does not appear to

change the relative performance of the different asset classes significantly compared to other liquidity

measures.

Asset class

Roll measure ( in %-

points)

Government bonds*

Covered bonds

Corporate bonds (excl.

financials)

ABS (incl. MBS and

RMBS)

Equities

Gold

0.630

(1.230)

0.535

(0.839)

0.617

(0.864)

1.081

(2.283)

1.313

(2.506)

3.938

(24,184)

Note: Average measure over the sample period and standard deviations in brackets.

*Includes bonds issued by central governments, bonds issued by local governments, bonds

guaranteed by central or local governments, bonds issued by multinationals, bonds issued by

promotional banks and bonds issued by central banks.

23,084

44.7

555

3.0

10,043

12.9

21,414

13.6

Number

observations

40,962

Average

number

observations per asset

13.4

The lowest average Roll measure is observed for covered bonds, indicating that this asset class on

average trades at the lowest relative bid/ask spreads. Covered bonds in general perform slightly better

than government bonds on the Roll measure, with the time-series for this asset class lying just below

PDF to HTML - Convert PDF files to HTML files

the government bond series. In addition, the two series diverge somewhat in the 2011-2012 time

period, with the Roll value of government bonds increasing slightly. This indicates that while bid/ask

spreads of government bonds were negatively affected by the market unrest in 2011-2012, covered

bond spreads did not react.

The Roll value of corporate bonds is typically in the same range as for government and covered

bonds. However time-series evidence shows a spike in this series during the financial crisis, which is

not as prevalent in the government and covered bonds asset classes. This indicates higher sensitivity

to stress for the liquidity of this asset class. Equities generally have a somewhat higher Roll measure

value than the bond asset classes, except for ABS, which again show a rather volatile path. This is

particularly the case in the first part of the sample, where very few observations are available for this

asset class, as very few assets in this asset class are traded frequently enough for the Roll measure to

be computed. Gold has by far the highest bid-ask spread on the Roll measure.

Trading volume and turnover

The trading volume and turnover ratio measures give an indication of the ease with which larger

amounts of a given asset can be liquidated in the market and to what extent the outstanding amount in

individual assets is traded.

The table below looks at average values of monthly trading volume per ISIN. It is evident that the

average trading volume per ISIN is far larger for government bonds than for any other asset class.

This indicates a better ability to liquidate large positions in this asset class than in the other bond asset

classes. Equities have the second highest value, followed by covered bonds. Corporate bonds and

ABS generally show a somewhat lower level. The time series for all asset classes are relatively stable

over our time sample.

Asset class

Average

ISIN (mill. euro)

3,596

(17,128)

155

(827)

Corporate bonds (excl.

financials)

ABS (incl. MBS and

RMBS)

Equities

(58)

(65)

753

(1,711)

Note: Average measure over the sample period and standard deviations in brackets.

*Includes bonds issued by central governments, bonds issued by local governments, bonds

guaranteed by central or local governments, bonds issued by multinationals, bonds issued by

promotional banks and bonds issued by central banks.

The trading volume statistic shows how large amounts of a given asset are traded on average over a

given month. This gives a picture of liquidity, in that larger trading amounts indicate the ability to

21,020

40.9

8,674

8.5

24,974

16.4

trading

Number

observations

63,202

51,227

Average

number

volume per month per

Government bonds*

Covered bonds

observations per asset

13.2

13.8

PDF to HTML - Convert PDF files to HTML files

actually trade such large amounts. Naturally, these values will be different across jurisdictions, with

larger jurisdictions typically showing larger average trading volumes. This is particularly the case for

government bonds, as is also evident from the relatively high standard deviation noted in the table.

However, this does not indicate that those assets traded in smaller jurisdictions are not liquid per

definition. Instead it could be said that a lower bound of trading volume indicates liquidity, and above

this the actual trading volume is more reflective of market conventions and market structures. This is

also the case when looking at the cross asset class figures. In this paper we are looking at liquidity in

the context of the ability of financial institutions to liquidate their liquidity buffer assets, so the average

trading volume should be seen in this context.

Asset class

Turnover

month

Government bonds*

Covered bonds

Corporate bonds (excl.

financials)

ABS (incl. MBS and

RMBS)

Equities

0.131

(0.489)

0.059

(0.290)

0.026

(0.073)

0.022

(0.090)

0.071

(0.085)

Note: Average measure over the sample period and standard deviations in brackets.

*Includes bonds issued by central governments, bonds issued by local governments, bonds

guaranteed by central or local governments, bonds issued by multinationals, bonds issued by

promotional banks and bonds issued by central banks.

As a supplement to the trading volume static, the turnover ratio per month indicates how large a

proportion of the outstanding amount of a given asset is traded over this period. This allows the trading

volume to be scaled relative to the actual size of outstanding assets. For government bonds, on

average 15% of the outstanding amount in any individual series is traded each month. This bond asset

class has the largest turnover ratio, followed by equities and covered bonds. The ratio remained

relatively stable over the time period we sampled for all asset classes, although increasing somewhat

for ABS and corporate bonds over the sample period.

Zero-trading days

The zero-trading days measure shows how frequently assets in a given asset class are traded. It

shows the proportion of days during a month, where a given asset was not traded. The higher the

value of the measure, the less frequently assets in this asset class are traded. As can be seen from

the table below, gold and equities are traded virtually every day. In actuality, gold is traded every day

and the slight positive value of this measure is due to the construct of the measure, which just

assumes an average of 21 trading days per month. Hence, in months with more trading days than this

number of trading days, gold will appear to have some non-trading days even though this is not the

case. Across asset classes however, this bias should be of the same magnitude, hence a comparison

of the values over asset classes is not biased.

26,733

51.9

47,718

46.6

46,748

30.4

125,225

33.2

ratio

per

Number

observations

122,697

Average

number

observations per asset

25.2

PDF to HTML - Convert PDF files to HTML files

Asset class

Government bonds*

Covered bonds

Corporate bonds (excl.

financials)

ABS (incl. RMBS)

Equities

Gold

Zero-trading days

0.792

(0.334)

0.884

(0.218)

0.863

(0.202)

0.985

(0.048)

0.319

(0.428)

0.004

(0.013)

Number of observations

166,026

135,587

48,056

70,287

32,346

Average

number

28.3

32.9

30.5

48.1

observations per asset

Note: Average measure over the sample period and standard deviations in brackets.

*Includes bonds issued by central governments, bonds issued by local governments, bonds

guaranteed by central or local governments, bonds issued by multinationals, bonds issued by

promotional banks and bonds issued by central banks.

Amongst the bond asset classes, government bonds are the most frequently traded. This is in line with

the evidence from trading volume and turnover statistics discussed previously. On average, covered

bonds and corporate bonds show the same level of non-trading frequency. However time-series

evidence shows that this average reflects a relatively stable level for covered bonds, while the zero-

trading days measure has fallen continuously over the sample period for corporate bonds. Probability

distributions show that for all bond asset classes, half of the zero-trading days measures reflect assets

that are traded very infrequently, with a zero-trading days measure of one. Hence even for

government bonds, a relatively large proportion of the bonds in our sample are not traded frequently.

This does not necessarily indicate that these bonds could not be traded, but merely that they are not

traded in our sample. Government bonds have the largest probability mass at values of the zero-

trading days measure below 0.5, i.e. bonds that are traded on average every other day. The corporate

bonds and covered bond asset classes also contain assets that are traded at a relatively high

frequency, but the proportion is generally smaller than for government bonds. To what extent this is

driven by differences in the composition of assets within the different asset classes, i.e. with a larger

proportion of high quality assets in one asset class versus another, will be investigated further in the

within asset class analysis of this paper. The proportion of ABS displaying a high frequency of trading

is significantly smaller than for the other asset classes examined.

Price volatility

In addition to the many liquidity measures investigated above, we also look at measures of price

volatility. Price volatility indicates both the ease with which an asset can be liquidated, and the

certainty the holder has of the value at which the asset will be liquidated. In the context of the LCR, it

is important for financial institutions both to be able to liquidate their liquid assets portfolios and to

have relative certainty of the value at which they may be liquidated. Price volatility can to some extent

be compensated for by setting higher haircuts, on the value at which the assets are included in the

liquidity buffer. This will be discussed further later in this analysis.

PDF to HTML - Convert PDF files to HTML files

Asset class

Government bonds

Covered bonds

Corporate bonds (excl.

financials)

ABS (incl. RMBS)

Equities

Gold

Return volatility

1.001

(1,701)

0.725

(0.972)

0.904

(1.184)

2,927

(6.106)

3.294

(6,097)

4.042

(20,582)

Number of observations

65,389

42,049

19,779

4,991

24,452

Average

number

14.2

14.5

15.8

5.8

44.1

observations per asset

Note: Average measure over the sample period and standard deviations in brackets.

*Includes bonds issued by central governments, bonds issued by local governments, bonds

guaranteed by central or local governments, issued by multinationals, bonds issued by promotional

banks and bonds issued by central banks.

The table above shows that covered bonds on average are the least volatile asset category.

Government bonds show a slightly higher level of price volatility along with corporate bonds. The same

picture is observed when looking at time series evidence, as presented in Annex 9. From this it is also

evident that both corporate bonds and government bonds show some spikes in volatility around the

financial crisis in 2008-2009 and again in 2011-2012, while covered bonds remain relatively stable

through these periods. Probability distributions (Annex 10) underscore this picture, with covered bonds

having the largest mass of observations with low return volatility. The probability distribution of

government bonds is similar to that of financial bonds, while ABS and equities have the smallest

probability mass with low return volatility. The price volatility of gold is higher than that observed for

both ABS and equities.

The Basel Committee’s LCR proposal includes limitations on the maximum price drop of an asset over

a 30 day period. For this reason we also look at maximum absolute price changes over a 30-day

period. This metric not only contains downward price movement, as the Basel requirement is linked to,

but gives a picture of the size of movements both upwards and downwards over a 30-day period, i.e. a

more volatility linked measure. If using this as an actual liquidity criterion, one would limit this to

downward movements.

Asset class

Maximum price change

within

Government bonds

Covered bonds*

Corporate bonds (excl.

financials)

30-day

3.143

(6,282)

1.830

(2.030)

2.807

(4.746)

26,908

17.1

61,950

15.0

period

86,969

(%-points)

14.8

Number of observations

Average

number

observations per asset

PDF to HTML - Convert PDF files to HTML files

ABS (incl. MBS and

RMBS)

Equities

Gold

5.830

(10.757)

17.112

(12.785)

9.115

(4.314)

13,424

25,719

9.2

42.9

Note: Average measure over the sample period and standard deviations in brackets.

*Includes bonds issued by central governments, bonds issued by local governments, bonds

guaranteed by central or local governments, bonds issued by multinationals, bonds issued by

promotional banks and bonds issued by central banks.

The table above shows that, as was the case for the return volatility measure, covered bonds have the

lowest 30-day price change on average. This is followed by corporate bonds and government bonds.

Gold has the second highest 30-day price change measure while equities show a significantly higher

value of 30-day price change than any of the other asset classes.

The time series evidence presented in Annex 9 shows a very stable level of the maximum price

change measure for covered bonds, with a very small spike during the 2008-2009 financial crisis. This

is supported by probability distribution evidence (Annex 10), which also shows that covered bonds

have the largest mass of low price-change observations. Government bonds and corporate bonds

show a somewhat more volatile pattern, with spikes both during the 2008-2009 and 2011-2012

periods. ABS generally have a higher level of 30-day price change than the other bond asset classes.

Time series evidence shows that this is particularly the case during the period 2008-2010, while in the

later years of our sample, this asset class performs in a manner equivalent to government bonds and

corporate bonds. The ABS asset class covers a relatively diverse subset of assets however. Further

analysis within this asset category will therefore have to be conducted to determine the drivers of this

time-series development. Probability distribution evidence similarly shows a large spread in the

probability mass for ABS.

Ranking across measures

On the basis of the evidence presented above, an initial coarse liquidity ranking of asset classes is

compiled. A ranking of the five asset classes for which all metrics are calculated is produced for each

metric. These rankings are then averaged, in order to give an overall ranking. The initial metric

rankings are made on the basis of the mean and standard deviation evidence presented in the tables

in the previous sections, the time-series evidence presented in Annex 9 and probability distribution

evidence presented in Annex 10.

Overall, government bonds and covered bonds obtain a similar ranking. Government bonds generally

perform better than covered bonds on volume and trading frequency measures, while covered bonds

perform better on measures related to the volatility of prices and bid/ask spreads.

Equities are the most frequently traded asset class. However, they perform relatively poorly on price

volatility related measures. Corporate bonds perform moderately on most measures, aside from those

relating to volume, where they do poorly. ABS are weak across the board. Gold is not included in this

table as not all measures can be calculated, however it shows a similar range of performance as

equities, with strong trading volumes, but a significant price volatility across all measures.

PDF to HTML - Convert PDF files to HTML files

Amihud

Price

impact

Roll

Trading

volume

Turnover

ratio

Zero

trading

days

Return

volatility

30-day

price

change

Average

ranking

2.00

3.00

Government

bonds

Covered bonds

Corporate

bonds

financials)

ABS

RMBS)

Equities

(incl.

(excl.

4.38

3.63

This ranking presented above allows liquidity levels to only differ by asset class. Hence, asset class is

in principle viewed as the only explanatory characteristic for liquidity. Clearly, this is an over-

simplification of reality. As the table below demonstrates, there is significant variation in the

composition of each asset class along the dimensions of credit rating, time to maturity and issue size.

For example, the government bond asset class is issued in much larger sizes than the other asset

classes, while ABS have the lowest issue size and longest time to maturity. Corporate bonds have the

lowest mean credit rating while covered bonds are almost exclusively rated as ECAI level 1. All these

characteristics, which may drive liquidity to a higher degree than just the asset class, will be looked

into in the following step of the analysis. A final ranking of different asset groups according to liquidity

will be produced on the basis of these additional characteristics.

Bond Type

1.GOVERNMENT

2.COVERED

3.ABS

4.CORPORATE

Rating (ECAI)

Mean

STD

1.2

1.0

1.7

2.5

0.7

0.1

1.1

1.0

Time to maturity (year)

Mean

STD

7.4

6.1

7.5

5.4

8.3

7.6

5.5

5.7

Issue size (EUR mill.)

Mean

STD

2,980

571

133

485

6,460

100

267

467

4.1.2 Quantitative analysis of market liquidity within asset classes

This section presents the analysis of market liquidity within the asset classes. The objectives are to

identify the characteristics of the assets being examined that are good proxies for market liquidity and

hence can be used to design reliable uniform definitions of liquid assets. For each asset class a set of

theoretically relevant explanatory characteristics of the liquidity and credit quality has been identified

based on the list of asset characteristics suggested in the CRR and in the EBA working paper. In this

section we test econometrically which of the characteristics of the individual securities within each

asset class are useful predictors of their observed market liquidity.

The analysis is based on pooled regressions in which the dependent variables are the monthly

averages of the liquidity metrics identified in the first part of the analysis for each security included in

the analysis. Some of the explanatory variables are common across asset classes, whereas others will

PDF to HTML - Convert PDF files to HTML files

vary depending on the relative importance of the specific characteristics for each asset class and the

availability of data.

There are a number of characteristics that from a theoretical point of view are thought to affect the

liquidity of all asset classes:



Remaining time to maturity. This captures the possibility that market liquidity is affected by a

bond’s time to maturity. We expect to observe higher price volatility for bonds having a long

remaining time to maturity due to their duration, which makes them more exposed to interest

rate risk. We divide the remaining time to maturity into four baskets, where remaining time to

maturity of below 1 year is the base scenario compared to remaining times to maturity of

between 1 and 5 years, between 5 and 10 years and above 10 years.



Issue size. Smaller issue sizes could potentially trade with lower volumes or be held by a few

investors for longer periods of time. This could imply less transparent pricing and thus reduce

liquidity. A narrow investor base and possible exclusion of institutional investors due to limited

volume could also add to the risk of the asset turning illiquid. The variable is based on the

issue size for each ISIN extracted from Bloomberg. We use the issue size at the extraction

point, and if unavailable, the issuance size. The differences between these are not expected

to alter the overall conclusions. The issue size is divided into five groups: less than EUR 10

million (the base scenario), between EUR 10 and 100 million, between EUR 100 and 250

million, between EUR 250 and 500 million and above EUR 500 million. For assets

denominated in other currencies, the market exchange rates are used for the conversion to

euros.



Credit quality. Bonds with higher credit quality are expected to be more liquid than bonds with

lower credit quality, simply because higher-rated bonds provide less risk and more

transparency for investors. The quality of the underlying assets decreases the price volatility of

bonds with higher credit quality relative to lower rated bonds. Furthermore, these bonds are

expected to broadly out-perform in periods with stressed markets, as flight to quality is

generally observed. The credit quality variable is created as a homogenous mapping of ratings

into a preliminary version of the ECAI credit quality notation. A table of the mapping is shown

in Annex 11. The base scenario is ECAI = 2, and we then estimate the effect of bonds which

are ECAI = 1, 3, 4, 5 and 6.

In addition, dummy variables for time periods and countries are included in the analyses in order to

control for time and country effects.

Public sector Bonds

Within the category of public sector bonds, we test the relative liquidity of the following five

subcategories: 1) bonds issued by central or local governments, 2) bonds guaranteed by central and

local governments, 3) bonds issued by central banks, 4) bonds issued by multinationals and 5) bonds

issued by promotional banks. Bonds issued by governments are on average much larger in terms of

amounts outstanding and also have longer maturities than the other subcategories. Multinational

bonds are second, whereas bonds issued by central banks have the lowest average size. Government

PDF to HTML - Convert PDF files to HTML files

bonds also have the highest average credit rating, followed by government guaranteed bonds and

local government bonds.

In the following two tables the mean and standard deviations of liquidity measures for all public sector

bonds and, specifically, for government bonds are given.

Public sector bonds

Mean

Amihud

Price impact

Roll

Return volatility

Price change

Trading volume (mill. EUR)

Turnover ratio

Zero-trading

Government bonds

Mean

Amihud

Price impact

Roll

Return volatility

Price change

Trading volume (mill. EUR)

Turnover ratio

Zero-trading

0.027

0.321

0.631

0,809

4.000

6,910,000

6.910

0,338

0,365

Standard deviation

0.117

0.454

1.224

1.268

9.086

24,050,000

24.050

0,839

0,417

Frequency

19848

19808

19003

20850

21881

21817

26782

27123

0.039

0.305

0.612

0.757

3.003

4,437,000

0.157

0.686

Standard deviation

0.141

0.446

1.184

1.223

7.333

19,500,000

0.575

0.408

Frequency

24882

24845

22092

29093

35977

34171

62079

62562

The explanatory characteristics in the analysis of the public sector bond category are given by credit

ratings, remaining time to maturity and outstanding amount. In addition, the coupon type and the

currency denomination are added as explanatory variables in the analysis. The results are given in the

table below:

PDF to HTML - Convert PDF files to HTML files

Liquidity characteristics of public-sector bonds

Price Volatility Measures

Amihud

TTM btw 1 and 5 years

TTM btw 5 and 10 years

TTM above 10 years

ECAI 1

ECAI 3

ECAI 4

ECAI 5

ECAI 6

Issue size btw. 10-100 Mio. EUR

Issue size btw. 100-250 Mio. EUR

Issue size btw. 250-500 Mio. EUR

Issue size of 500 Mio. EUR above

Fixed coupon

Denominated in euro

Government guarenteed

Promotional banks

Local govenment

Multinationals

Central bank

Constant

R-squared

Number of observations

0.018***

(8.68)

Volume measures

0.764***

(5.93)

Price impact

0.115***

(19.24)

Roll

0.142***

(6.84)

Return vol. Price change Trading Vol. Turnover ratio Zero-trading

0.250***

(14.64)

1.20E+03***

(11.16)

0.038***

(7.54)

-0.011***

(-3.34)

0.032***

(13.12)

0.282***

(40.65)

0.389***

(15.56)

0.658***

(31.01)

2.305***

(16.01)

2.84E+03***

(9.85)

0.072***

(9.39)

-0.007**

(-1.96)

0.039***

(13.53)

0.529***

(58.76)

0.580***

(20.56)

1.031***

(41.38)

4.335***

(23.21)

1.05E+03***

(2.92)

0.045***

(5.04)

-0.020***

(-4.72)

-0.029***

(-5.18)

-0.243***

(-16.63)

-0.240*** -0.391***

(-6.32)

(-11.21)

-1.835***

(-15.32)

2.95E+03***

(9.71)

0.081***

(11.87)

-0.037***

(-7.87)

0.112***

(4.18)

0.360***

(6.72)

0.695***

(5.15)

0.557***

(5.62)

2.305**

(2.48)

-1.56E+03***

(-4.41)

-0.130***

(-13.28)

0.040***

(5.51)

0.081***

(7.51)

0.405***

(13.96)

0.675***

(9.28)

0.862***

(11.85)

5.382***

(11.59)

-4.52E+03***

(-19.08)

-0.381***

(-24.53)

0.047***

(4.29)

0.238***

(6.44)

0.605***

(9.77)

1.041***

(6.87)

1.214***

(8.42)

5.672***

(9.32)

-6.59E+03***

(-12.9)

-0.512***

(-24.34)

0.201***

(8.63)

0.437***

(11.99)

1.729***

(19.65)

2.666***

(10.75)

4.708***

(19.85)

35.453***

(10.81)

-6.50E+03***

(-16.92)

-0.498***

(-24.3)

0.425***

(25.54)

0.154***

(3.77)

0.084

(1.47)

0.152

(0.91)

0.300***

(2.7)

0.017

(0.1)

5.37E+02***

(3.86)

0.021***

(3.58)

-0.047***

(-8.47)

0.021

(0.88)

-0.006

(-0.12)

-0.145

(-0.91)

0.199**

(2.56)

-0.145

(-0.96)

4.50E+02***

(3.55)

0.003

(0.55)

-0.036***

(-6.54)

0.005

(0.23)

-0.060

(-1.34)

-0.095

(-0.71)

0.015

(0.2)

-0.270*

(-1.79)

-5.51E+00

(-0.04)

0.030***

(5.07)

-0.057***

(-9.89)

-0.027

(-1.4)

-0.103***

(-2.72)

-0.186*

(-1.65)

-0.296***

(-4.83)

-0.288*

(-1.78)

1.34E+02

(1.12)

0.125***

(21.36)

-0.365***

(-63.97)

0.005

(1.53)

0.042***

(5.05)

0.190***

(7.67)

0.195***

(8.32)

0.682***

(4.38)

1.22E+03***

(16.42)

0.004

(1.54)

-0.088***

(-30.33)

-0.027***

(-4.65)

-0.034***

(-3.09)

0.002

(0.08)

0.003

(0.13)

0.477***

(7.77)

1.49E+03***

(18.85)

0.052***

(18.82)

-0.207***

(-59.54)

0.053***

(12.47)

0.031***

(2.95)

0.104***

(3.77)

0.103***

(4.22)

-0.549***

(-8.93)

-2.79E+03***

(-33.67)

-0.113***

(-35.6)

0.371***

(101.78)

0.020*

(1.95)

0.007

(0.22)

0.314***

(2.87)

0.266**

(2.1)

-0.034

(-0.19)

-6.46E+03***

(-12.43)

-0.229***

(-12.72)

0.163***

(18.41)

0.077***

(9.73)

0.063***

(4.46)

0.273***

(6.58)

0.129***

(5.07)

-0.688***

(-15.47)

-2.41E+03***

(-22.89)

-0.091***

(-23.71)

0.391***

(105.16)

0.037***

(3.02)

-0.010

(-0.65)

-0.105**

(-2.38)

-0.009

(-0.12)

-0.165*

(-1.64)

-1.23E+03***

(-6.81)

-0.004

(-0.65)

0.134***

(13.9)

0.061**

(2.26)

-0.066

(-1.42)

0.000***

()

0.085

(0.76)

-2.011***

(-7.67)

-3.60E+04***

(-21.88)

-1.204***

(-26.6)

0.849***

(86.83)

0.131***

(5.34)

0.337***

(6.87)

1.180***

(6.87)

2.449***

(14.12)

0.446*

(1.92)

-3.60E+03***

(-4.1)

-0.113***

(-3.69)

1.293***

(89.38)

0.193

24.603

0.407

24.563

0.150

21.950

0.298

28.579

0.324

34.753

0.187

33.156

0.240

57.592

0.658

58.052

Note: TTM: time to maturity. Baseline scenario: TTM below 1 year, ECAI 2, issue size below

EUR 10 million and bonds issued by the government.

As expected, the results across bonds of different

credit qualities

suggest that for all liquidity

measures, lower rated bonds are less liquid than the highest rated bonds (ECAI 1).

With regard to

maturity,

we observe that bonds with a shorter remaining maturity are more liquid than

bonds with a longer maturity for all price based liquidity measures. On the other hand, trading volume

and frequency seem to be higher for maturities above 1 year, compared to maturities of less than a

year. In particular, trading activity is highest for bonds with a time to maturity between 5 and 10 years.

PDF to HTML - Convert PDF files to HTML files

With regard to

outstanding amounts,

the issue size does not seem to matter significantly for price

volatility, except for very high issue sizes of above EUR 500 million for which price volatility is relatively

lower. The volume based liquidity measures indicate that larger issue sizes imply higher trading

volume, turnover ratio and number of trading days, even though the results are mixed.

Finally, results for

coupon types

are mixed. Fixed coupon bonds are less liquid according to the price

measures, but more liquid according to the volume based liquidity measures. With regard to the

currency denomination,

bonds denominated in EUR in general appear more liquid.

Turning to the different

public bond subcategories,

the results generally suggest that public bonds

issued by other entities than governments are less liquid. Firstly, the trading activity and volume is less

for the other bonds. Secondly, price volatility seems to be higher for non-government issued bonds,

except for bonds issued by multinationals and central banks. For these entities, price volatility is not

significantly different than for government bonds.

We repeat the empirical assessment, for government bonds only, in the table below. The results

largely mirror the evidence obtained when including other sub-categories which is not surprising given

that government bonds dominate the sample. Only the importance of bond issue size has changed

and shows some rather mixed results for government bonds. The results are similar for the price

volatility measures but the trading volume is lower for bonds with an issue size between EUR 100 and

500 million than for issue sizes below EUR 10 million, which is rather puzzling.

PDF to HTML - Convert PDF files to HTML files

Liquidity characteristics of government bonds

Price Volatility Measures

Amihud

TTM btw 1 and 5 years

TTM btw 5 and 10 years

TTM above 10 years

ECAI 1

ECAI 3

ECAI 4

ECAI 5

ECAI 6

Issue size btw. 10-100 Mio. EUR

Issue size btw. 100-250 Mio. EUR

Issue size btw. 250-500 Mio. EUR

Issue size of 500 Mio. EUR above

Coupon

EUR

Constant

R-squared

Number of observations

0.016***

(9.23)

Volume measures

0.895***

(3.7)

Price impact

0.115***

(16.88)

Roll

0.137***

(5.86)

Return vol. Price change Trading Vol. Turnover ratio Zero-trading

0.263***

(12.03)

1.94E+03***

(10.81)

0.092***

(7.77)

-0.007

(-1.11)

0.022***

(10.59)

0.281***

(36.88)

0.380***

(13.68)

0.619***

(24.32)

2.550***

(10.3)

4.46E+03***

(10.54)

0.182***

(11.14)

-0.040***

(-6.24)

0.028***

(12.33)

0.534***

(56.58)

0.567***

(18.93)

0.980***

(35.61)

4.566***

(16.18)

1.74E+03***

(4.29)

0.110***

(6.93)

-0.029***

(-4.4)

-0.016***

(-3.74)

-0.232***

(-16.6)

-0.235*** -0.390***

(-5.9)

(-11.06)

-2.082***

(-11.09)

4.68E+03***

(8.18)

0.228***

(11.68)

-0.009

(-1.06)

0.040***

(4.23)

0.313***

(7.72)

0.515***

(4.2)

0.466***

(6.22)

5.160***

(2.63)

-3.99E+03***

(-5.13)

-0.262***

(-12.4)

0.040***

(2.97)

0.060***

(7.14)

0.390***

(13.96)

0.659***

(9.05)

0.793***

(12.09)

5.247***

(10.96)

-5.44E+03***

(-17.12)

-0.469***

(-25.03)

0.079***

(6.29)

0.225***

(6.13)

0.591***

(9.63)

1.011***

(6.68)

1.167***

(8.13)

5.927***

(9.58)

-7.32E+03***

(-11.51)

-0.553***

(-20.61)

0.196***

(8.23)

0.423***

(11.66)

1.705***

(19.39)

2.677***

(10.77)

4.669***

(19.79)

35.294***

(10.76)

-8.25E+03***

(-13.9)

-0.599***

(-22.51)

0.436***

(25.1)

0.389**

(1.97)

0.696*

(1.71)

-0.331

(-0.28)

0.482

(0.7)

0.554

(0.89)

-8.11E+02

(-1.06)

-0.068**

(-2.51)

-0.208***

(-15.9)

0.071

(0.79)

-0.037

(-0.18)

-0.077

(-0.05)

0.089

(0.25)

-0.256

(-0.47)

-4.04E+03***

(-6.39)

-0.316***

(-12.6)

-0.047***

(-4.39)

0.094

(0.91)

-0.120

(-0.63)

-0.984

(-1.25)

-0.229

(-0.75)

-0.251

(-0.63)

-4.24E+03***

(-7.16)

-0.018

(-0.88)

-0.062***

(-6.69)

-0.077

(-1.04)

-0.332***

(-2.57)

-1.198*** -1.043***

(-2.79)

(-4.71)

-0.362

(-0.99)

1.40E+03***

(3.93)

0.280***

(15.8)

-0.648***

(-83.97)

-0.012***

(-4.12)

0.020**

(2.24)

0.175***

(6.33)

0.115***

(4.22)

0.824***

(2.61)

2.80E+03***

(20.86)

0.082***

(10.49)

-0.126***

(-18.09)

-0.024***

(-2.76)

-0.021

(-1.26)

0.033

(0.7)

0.043

(1.42)

0.829***

(9.01)

2.01E+03***

(14.07)

0.054***

(7.54)

-0.336***

(-44.94)

0.177**

(2.31)

0.553***

(4.11)

2.182***

(4.84)

3.242***

(11.05)

0.372

(0.82)

-9.84E+03***

(-7.57)

-0.318***

(-4.86)

1.677***

(64.04)

0.2415

19.810

0.444

19.768

0.148

18.976

0.323

20.776

0.309

21.727

0.172

21.690

0.228

26.458

0.529

26.799

Note: TTM: time to maturity.

Baseline scenario: TTM below 1 year, ECAI 2, issue size below EUR 10 million.

Corporate bonds

The analysis of the liquidity of corporate bonds follows the same method as for government bonds.

The table below gives the mean and standard deviations of liquidity measures for corporate bonds.

Corporate bonds

Mean

Amihud

Price impact

Roll

Return volatility

Price change

Trading volume (mill. EUR)

Turnover ratio

Zero-trading

0.163

0.337

0.581

0.814

2.703

37,302

0.035

0.782

Standard deviation

0.266

0.410

0.844

1.068

3.669

65,324

0.072

0.236

Frequency

8,323

8,330

6,381

11,056

13,679

13,022

18,278

18,369

PDF to HTML - Convert PDF files to HTML files

The explanatory characteristics in the analysis of corporate bond liquidity are given by credit ratings,

remaining time to maturity and outstanding amount. In addition, the coupon type is added as

explanatory variable. The results are given in the table below:

Liquidity characteristics of corporate bonds

Price Volatility Measures

Amihud

TTM btw 1 and 5 years

TTM btw 5 and 10 years

TTM above 10 years

ECAI 1

ECAI 3

ECAI 4

ECAI 5

ECAI 6

Issue size btw. 10-100 Mio. EUR

Issue size btw. 100-250 Mio. EUR

Issue size btw. 250-500 Mio. EUR

Issue size of 500 Mio. EUR above

Fixed coupon

Constant

R-squared

Number of observations

0.067***

(6.01)

Volume measures

1.355***

(22.1)

Price impact

0.127***

(6.92)

Roll

0.229***

(5.95)

Return vol. Price change Trading Vol. Turnover ratio Zero-trading

0.356***

(11.8)

6.395***

(4.57)

0.007***

(6.26)

-0.043***

(-10.6)

0.125***

(10.81)

0.285***

(14.93)

0.373***

(9.63)

0.709***

(22.19)

2.622***

(37.36)

10.902***

(6.18)

0.012***

(8.3)

-0.028***

(-6.06)

0.182***

(12.57)

0.461***

(19.31)

0.619***

(12.34)

1.151***

(27.35)

3.689***

(31.23)

7.255***

(3.67)

0.005***

(3.13)

0.022***

(4.35)

-0.025***

(-2.72)

-0.026*

(-1.91)

-0.076**

(-2.11)

-0.045

(-1.5)

-0.299***

(-5)

3.269**

(2.11)

-0.005***

(-3.51)

0.021***

(4.95)

-0.002

(-0.23)

0.057***

(5.9)

0.037

(1.45)

0.124***

(5.72)

0.845***

(12.76)

13.933***

(11.86)

0.019***

(17.03)

-0.064***

(-20)

0.054***

(4.81)

0.161***

(9.3)

0.128***

(3.16)

0.241***

(7.03)

1.848***

(15.58)

16.339***

(4.08)

0.021***

(6.15)

-0.069***

(-10.95)

0.155***

(5.49)

0.481***

(9.65)

0.597***

(5.34)

0.943***

(10.05)

5.625***

(13.9)

15.023***

(4.07)

0.008***

(3.05)

0.004

(0.46)

0.379**

(2.55)

0.627***

(4.38)

-0.081

(-0.3)

1.385***

(3.74)

5.082***

(5.14)

0.246

(0.06)

-0.007***

(-2.76)

0.037***

(3.1)

0.085

(1.46)

0.228

(1.42)

1.182

(1.09)

0.489***

(2.89)

-0.265

(-1.21)

-19.920***

(-10.2)

-0.015***

(-4.39)

0.135***

(25.94)

0.071*

(1.8)

0.020

(0.46)

-0.029

(-0.28)

0.168**

(1.99)

-0.335

(-1.6)

-14.842***

(-10.55)

-0.021***

(-11.79)

0.138***

(34.02)

0.037**

(2.07)

0.049**

(2.14)

0.048

(0.73)

0.132**

(2.56)

0.126

(0.94)

-8.784***

(-7.36)

-0.018***

(-10.39)

0.100***

(24.11)

-0.041***

(-5.18)

-0.013

(-1.18)

-0.012

(-0.44)

-0.098***

(-3.93)

0.098

(1.35)

21.186***

(17.01)

0.004**

(2.35)

-0.115***

(-31.63)

0.035**

(2.47)

0.058***

(2.9)

0.007

(0.14)

0.165***

(3.44)

0.154

(0.96)

-2.846**

(-2.24)

0.000

(-0.16)

-0.011***

(-2.75)

0.137**

(2.41)

0.046

(0.78)

1.652***

(5.92)

3.088***

(9.05)

1.706***

(4.56)

-20.040***

(-5.57)

-0.015***

(-4.35)

1.069***

(86.56)

0.121

8.255

0.224

8.263

0.110

6.326

0.254

10.971

0.220

13.565

0.124

12.918

0.130

18.107

0.433

18.198

Note: TTM: time to maturity.Baseline scenario: TTM below 1 year, ECAI 2, issue size below EUR 10

million.

Time to maturity

broadly seems to have the same effect on corporate bonds as on bonds issued by

the public sector. This implies higher price volatility for bonds with a longer time to maturity, but on the

other hand also with a higher trading volume and frequency.

Turning to

credit quality,

price volatility in general is lower for bonds with an ECAI 1 rating and higher

for lower ratings compared to bonds rated ECAI 2. Credit quality has a less clear effect on trading

volume and frequency with both high and low ratings resulting in higher trading volumes.

The outstanding amount

does not significantly influence the price impact or the bid/ask spread. Only

the return volatility seems to increase when the issue size is above EUR 10 million, but volatility does

not seem to increase further as issue sizes grow. The results for the volume measures are somewhat

puzzling, as trading volume and frequency seem to decrease with increasing issue sizes.

Results for

coupon types

show that price volatility is higher and trading volume lower for fixed coupon

bonds. On the other hand, the trading frequency seems to be higher.

PDF to HTML - Convert PDF files to HTML files

Covered bonds

The characteristics that are important for the liquidity of covered bonds are attempted to be identified.

Firstly, an overview of the liquidity measures for covered bonds is given in the table below:

Covered bonds

Mean

Amihud

Price impact

Roll

Return volatility

Price change

Trading volume (mill. EUR)

Turnover ratio

Zero-trading

0.081

0.237

0.487

0.663

1.763

211,000

0.083

0.863

Standard deviation

0.192

0.387

0.782

0.895

1.945

1,052,000

1.052

0.366

0.245

Frequency

19,340

19,358

14,461

26,432

38,196

30,391

70,510

78,366

The common characteristics are included in the analysis: time to maturity, issue size and credit quality.

Two additional characteristics are also taken into account:



Credit risk regulation. This variable captures whether regulation, which reduces the credit risk

of bond investors, is present. Clear regulatory regimes are expected to affect the liquidity of

covered bonds positively as risk is decreased and transparency is increased for the investor.

The variable has been constructed through an analysis of different covered bond schemes.

We create a binary variable for each ISIN with a value of 1 if the scheme satisfies the five

requirements listed below, and 0 otherwise.

Are there any special LTV limits used solely for calculating collateralisation rates for

the cover pool?

Is mandatory minimum over-collateralisation required?

Is exposure to market risk (e.g. interest rates, currency risks) required to be mitigated

by law or contract?

Is exposure to liquidity risk required to be mitigated by law or contract?

How are covered bondholders protected against claims from other creditors if the

issuer becomes insolvent?



Transparency. This variable covers the availability of information to investors and the level of

standardisation of the issue. Information, transparency and standardisation are expected to

increase the liquidity of the assets as they reduce uncertainty for investors and broaden the

investor base. We create a binary variable based on the criteria listed below. All four criteria

have to be fulfilled to obtain the value 1, otherwise the value is 0.

Is a special licence required for the issuing of covered bonds?

Does a covered bond issuer have special reporting duties to the supervision authority

concerning covered bonds and the cover pool, which go beyond the requirements of

the regular banking supervision regime?

Covered bonds eligible for the treatment set out in Article 129(4) or (5) CRR or bonds

referred to in Article 52(4) of Directive 2009/65/EC

Does the covered bond fulfill the criteria of Article 52(4) of Directive 2009/65/EC?

PDF to HTML - Convert PDF files to HTML files

Does the covered bond legislation completely fall within the criteria of Article 129(4)

and (5) CRR?

The results of the regression analysis are given in the table below:

Liquidity characteristics of covered bonds

Price Volatility Measures

Amihud

TTM btw 1 and 5 years

TTM btw 5 and 10 years

TTM above 10 years

ECAI 1

ECAI 3

Issue size btw. 10-100 Mio. EUR

Issue size btw. 100-250 Mio. EUR

Issue size btw. 250-500 Mio. EUR

Issue size of 500 Mio. EUR above

Credit Risk Reg.

Transparency

Constant

R-squared

Number of observations

0.025***

(8.63)

Volume measures

0.640***

(24.27)

Price impact

0.076***

(12.18)

Roll

0.130***

(10.26)

Return vol. Price change Trading Vol. Turnover ratio Zero-trading

0.204***

(19.73)

93.416***

(3.26)

0.017***

(9.12)

0.012***

(5.73)

0.061***

(12.37)

0.194***

(18.93)

0.283***

(12.53)

0.551***

(31.96)

1.497***

(44.62)

-72.126***

(-2.85)

0.014***

(7.39)

0.032***

(13.85)

0.043***

(9.24)

0.199***

(21.08)

0.399***

(18.78)

0.595***

(35.64)

1.938***

(59.28)

21.279

(0.84)

0.241***

(39.38)

-0.101***

(-33.19)

-0.110***

(-5.63)

-0.325***

(-8.67)

-0.363*** -0.531***

(-5.17)

(-9.1)

-1.682***

(-11.9)

186.333***

(7.43)

0.031***

(5.06)

0.002

(0.29)

0.264**

(2.05)

0.278

(1.2)

1.796***

(3.7)

0.780

(1.53)

2.871*

(1.94)

78.760

(0.95)

0.017

(0.41)

-0.047

(-1.49)

0.063***

(8.24)

0.012

(0.79)

0.176***

(4.28)

0.288***

(10.1)

0.020

(0.61)

-18.555***

(-2.78)

0.058***

(9.34)

0.157***

(52.25)

0.056***

(7.96)

0.023*

(1.92)

0.093***

(2.78)

0.232***

(9.85)

0.179***

(6.68)

-20.383***

(-3.75)

0.047***

(9.11)

0.141***

(47.69)

0.023***

(3.29)

0.011

(0.89)

0.100***

(2.84)

0.051**

(2.33)

-0.071**

(-2.41)

47.029***

(7.27)

0.079***

(13.59)

0.077***

(20.89)

-0.002

(-0.75)

0.005

(0.72)

-0.017

(-1.33)

-0.045***

(-3.88)

0.217***

(8.76)

436.054***

(27.53)

0.155***

(27.21)

-0.185***

(-49.94)

-0.066***

(-13.88)

-0.163***

(-16.21)

-0.262*** -0.318***

(-12.42)

(-19.79)

-0.503***

(-19.04)

354.410***

(23.98)

0.076***

(37.64)

-0.095***

(-45.96)

-0.046***

(-6.69)

-0.140***

(-10.11)

-0.304*** -0.301***

(-9.95)

(-13.45)

-0.992***

(-22.31)

73.223***

(10.78)

0.009***

(2.86)

0.002

(0.48)

0.228***

(10.58)

0.661***

(16.27)

1.729***

(15.93)

2.588***

(25.96)

3.159***

(20.04)

-563.212***

(-9.42)

-0.112***

(-5.82)

0.888***

(78.98)

0.080

19.340

0.138

19.358

0.116

14.461

0.197

26.432

0.247

38.196

0.050

30.391

0.082

70.510

0.310

78.366

Note: TTM: time to maturity.

Baseline scenario: TTM below 1 year, ECAI 2, issue size below EUR 10 million.

Overall, the results show that the liquidity of covered bonds depends on all the selected characteristics

and the estimated coefficients are in general as expected.

All measures indicate that the price volatility of covered bonds increases with a

bond’s time to

maturity.

Turning to the measures of trading volumes, the results are mixed. The turnover ratio seems

to increase with maturity, while the results for trading volume and trading days is less clear. It should

be noted that bonds with a lower trading volume are not necessarily less liquid. Some of the bonds

with less trading activity may be bought and held to maturity, implying low actual trading, but they

could possibly be liquidated without any problem if a need arose.

High credit quality

has a significant effect on liquidity according to the results. Bonds with an ECAI 1

rating have lower price volatility, reflecting the stabilising effect of higher credit quality on prices. High

credit quality also implies higher trading volumes and turnover.

PDF to HTML - Convert PDF files to HTML files

The price measures indicate that price volatility in general increases with

issue size,

except for issue

sizes of above EUR 500 million. Whilst turnover is higher for higher issue sizes, the effect of issue size

on trading volume and trading days is mixed.

Credit risk regulation and transparency

have explanatory power for almost all measures. The

estimated negative coefficients on all price measures reflect as expected that a regulatory regime,

which reduces credit risk and enhances transparency and standardisation of bond issues dampen

price volatility. Furthermore, the regulatory measures also enhance trading volume, turnover and the

number of trading days.

ABS including RMBS

In the analysis of the liquidity of ABS, we perform an analysis of the total ABS market and of RMBS in

particular. An overview of the liquidity measures for ABS and for the subgroup of RMBS is given in the

table below:

ABS, total

Mean

Amihud

Price impact

Roll

Return volatility

Price change

Trading volume (mill. EUR)

Turnover ratio

Zero-trading

0.207

0.756

1.105

2.768

5.579

30.2

0.243

0.982

Standard deviation

0.600

1.465

2.291

5.835

1.015

67.0

0.098

0.051

Frequency

1921

1920

470

4162

10810

6939

35144

50255

RMBS

Mean

Amihud

Price impact

Roll

0.152

0.552

0.936

2.230

5.139

34.5

0.034

0.979

Standard deviation

0.477

1.054

2.086

5.562

1.135

7.3

0.123

0.058

Frequency

1388

1387

398

2906

6767

4301

19572

29360

Return volatility

Price change

Trading volume (mill. EUR)

Turnover ratio

Zero-trading

In line with the other within asset class analyses, we consider the following variables in the attempt to

explain the variation in liquidity within the ABS asset class: credit quality, issue size and time to

maturity. In addition, given that the ABS class is fairly heterogeneous, we include the subtypes RMBS,

MBS and a dummy variable for the senior tranches of ABS.

PDF to HTML - Convert PDF files to HTML files

The

time to maturity

seems to have a negative impact on liquidity, as both price volatility increases

and trading activity decreases with time to maturity. For maturity above 10 years, the price impact and

bid/ask spread is unaffected, but longer maturity still seems to imply higher price volatility and lower

trading activity.

The results indicate that

credit quality

has important effects on the liquidity of ABS. ABS with an ECAI

1 rating seem to have the highest liquidity, whereas liquidity decreases for lower rating categories.

As regards

issue size,

we observe that as expected, trading activity and volume increase with volume

size. On the other hand, the price impact and return volatility is higher for larger issue sizes.

With respect to the split by

subtype,

the baseline is other ABS and we test for the impact of assets

being RMBS and MBS. In general, the subtype seems to have some significance, in particular in the

price change and trading activity measures.

Tranche seniority

does not seem to play an important role in price volatility, but trading volume and

frequency are significantly lower for senior tranches.

PDF to HTML - Convert PDF files to HTML files

Liquidity characteristics of ABS

Price Volatility Measures

Amihud

TTM btw 1 and 5 years

TTM btw 5 and 10 years

TTM above 10 years

ECAI 1

ECAI 3

ECAI 4

ECAI 5

ECAI 6

Issue size btw. 10-100 Mio. EUR

Issue size btw. 100-250 Mio. EUR

Issue size btw. 250-500 Mio. EUR

Issue size of 500 Mio. EUR above

RMBS

MBS

ABS senior

Constant

R-squared

Number of observations

0.074***

(2.72)

Volume measures

2.012***

(7.44)

Price impact

0.242***

(4.09)

Roll

0.230

(1.59)

Return vol. Price change Trading Vol. Turnover ratio Zero-trading

0.610**

(2.36)

-0.122

(-0.02)

0.004

(1.58)

0.001

(0.63)

0.150***

(3.93)

0.623***

(5.84)

0.905**

(2.54)

1.086***

(3.43)

3.092***

(9.53)

-12.886***

(-2.62)

-0.009***

(-3.68)

0.008***

(6.23)

0.151

(1.4)

0.165

(0.82)

0.248

(0.45)

4.434***

(5.2)

3.648***

(7.29)

-12.694***

(-2.66)

-0.021***

(-9.04)

0.014***

(10.88)

0.122**

(2.14)

-0.053

(-0.3)

0.158

(0.29)

-0.771*

(-1.72)

-3.230***

(-6.27)

-1.153

(-0.58)

-0.002

(-1.54)

-0.004***

(-6.84)

0.372***

(4.46)

0.584***

(3.19)

0.319

(0.73)

2.216***

(3.87)

4.463***

(6.23)

-0.085

(-0.05)

0.002

(1.11)

-0.002***

(-3.53)

0.373

(1.33)

0.152

(0.54)

-1.676**

(-2.13)

2.978**

(2.22)

6.131***

(3.86)

-0.984

(-0.24)

-0.011***

(-6.86)

0.003***

(4.99)

1.011

(0.97)

1.526

(1.25)

2.935

(1.45)

5.398***

(4.22)

-0.988

(-0.28)

-0.014***

(-7.09)

0.005***

(4.93)

0.687**

(1.98)

2.396***

(3.33)

0.734

(0.4)

2.768

(1.19)

11.148***

(6.34)

5.038*

(1.65)

-0.006***

(-2.61)

0.004***

(4.31)

0.424***

(6.57)

0.725***

(5.22)

1.052*

(1.8)

2.930***

(6.89)

4.646***

(14.44)

-5.205

(-0.82)

0.012***

(9.55)

0.010***

(16.83)

0.179***

(5.06)

0.542***

(5.53)

0.830*

(1.92)

1.057***

(3.98)

2.724***

(12.56)

11.619*

(1.82)

0.033***

(16.37)

-0.006***

(-6.1)

0.054

(1.55)

0.379***

(3.77)

0.077

(0.21)

0.749***

(2.85)

3.005***

(15.44)

38.580***

(5.8)

0.043***

(16.49)

-0.021***

(-15.62)

0.031

(1.13)

0.169*

(1.89)

-0.230

(-0.65)

0.396*

(1.75)

2.965***

(16.36)

54.865***

(7.65)

0.023***

(11.68)

-0.022***

(-16.23)

-0.070

(-0.98)

0.183

(1.19)

1.464**

(2.04)

-0.492

(-0.35)

-1.188*

(-1.79)

-3.096

(-0.2)

-0.031**

(-2.26)

0.031***

(8.21)

-0.058

(-0.85)

0.205*

(1.82)

1.254*

(1.95)

0.069

(0.05)

-1.656***

(-2.91)

3.775

(0.43)

-0.047***

(-3.43)

0.047***

(13.27)

-0.014

(-0.43)

-0.007

(-0.09)

-0.179

(-0.93)

-0.298

(-1.44)

-0.266

(-1.47)

-16.270***

(-7.3)

-0.017***

(-11.07)

0.007***

(9.58)

-0.402***

(-3.62)

-0.720

(-1.38)

2.700**

(2.38)

11.904***

(3.01)

0.943

(1.01)

1.195

(0.11)

0.013

(0.94)

1.004***

(360.47)

0.210

1.903

0.297

1.902

0.378

467

0.236

4.134

0.204

10.752

0.149

6.937

0.064

35.134

0.122

50.043

Note: TTM: time to maturity.

Baseline scenario: TTM below 1 year, ECAI 2, issue size below EUR 10 million.

The results of the separate analysis of RMBS are given in the table below. It indicates that the

characteristics in general have the same impact on liquidity of RMBS as on ABS, which is not

surprising given the significant share of RMBS in the sample.

PDF to HTML - Convert PDF files to HTML files

Liquidity characteristics of RMBS

Price Volatility Measures

Amihud

TTM btw 1 and 5 years

TTM btw 5 and 10 years

TTM above 10 years

ECAI 1

ECAI 3

ECAI 4

ECAI 5

ECAI 6

Issue size btw. 10-100 Mio. EUR

Issue size btw. 100-250 Mio. EUR

Issue size btw. 250-500 Mio. EUR

Issue size of 500 Mio. EUR above

ABS senior

Constant

R-squared

Number of observations

0.065***

(2.79)

Volume measures

1.310***

(3.73)

Price impact

0.204***

(4.47)

Roll

0.180

(1.09)

Return vol. Price change Trading Vol. Turnover ratio Zero-trading

0.221

(0.68)

3.369

(0.56)

0.015***

(4.83)

-0.001

(-0.6)

0.112***

(4.09)

0.572***

(6.59)

0.964**

(2.38)

0.983***

(2.62)

2.932***

(6.57)

-6.640

(-1.07)

0.002

(0.5)

0.006***

(3.48)

0.044

(0.84)

0.086

(0.73)

0.012

(0.02)

2.402***

(3.08)

2.135***

(3.79)

-10.249*

(-1.69)

-0.016***

(-5.89)

0.016***

(9.19)

0.135**

(2.04)

-0.032

(-0.17)

0.184

(0.29)

-0.031

(-0.05)

-3.034***

(-3.91)

-2.882

(-1.21)

-0.009***

(-3.61)

-0.002***

(-2.81)

0.412***

(4.46)

0.544***

(2.9)

0.296

(0.59)

2.164***

(3.43)

4.606***

(4.73)

-0.852

(-0.47)

0.001

(0.42)

-0.001

(-1.57)

0.207**

(2.04)

0.205

(0.56)

4.177**

(2.08)

4.681**

(1.99)

-13.571***

(-3.48)

-0.025***

(-8.95)

0.007***

(7.09)

1.064

(1.5)

2.193***

(3.72)

1.283*

(1.91)

6.914**

(2.45)

3.735**

(2.21)

-12.039

(-1.48)

-0.033***

(-10.47)

0.015***

(11.64)

7.287**

(2)

9.549***

(3.74)

-1.865

(-0.55)

-0.020***

(-5.12)

0.011***

(8.97)

0.411***

(5.61)

0.785***

(5.16)

3.793***

(7.66)

6.321***

(14.1)

-12.722

(-1.37)

0.010***

(4.85)

0.017***

(17.81)

0.111***

(3.94)

0.558***

(5.91)

1.166**

(2.28)

1.451***

(4.98)

3.625***

(12.58)

7.308

(0.78)

0.035***

(12.46)

-0.002*

(-1.79)

0.009

(0.36)

0.283***

(3.65)

0.426

(1.05)

0.690**

(2.5)

2.904***

(10.11)

46.365***

(4.75)

0.064***

(14.29)

-0.031***

(-12.89)

0.025

(1.17)

0.174***

(2.57)

0.083

(0.21)

0.482**

(2.18)

2.031***

(9.28)

56.013***

(5.19)

0.016***

(5.04)

-0.026***

(-10.8)

0.013

(0.48)

0.037

(0.62)

-0.133

(-0.7)

-0.192

(-0.96)

-0.736***

(-3.42)

-12.979***

(-5.73)

-0.015***

(-7.23)

0.005***

(5.93)

-0.300

(-0.98)

0.652

(0.44)

0.415

(0.47)

16.110***

(2.81)

9.698

(1.29)

-5.239

(-0.46)

-0.007

(-1.53)

0.996***

(529.89)

0.203

1.370

0.297

1.369

0.391

395

0.260

2.878

0.213

6.711

0.177

4.301

0.083

19.572

0.163

29.158

Note: TTM: time to maturity.

Baseline scenario: TTM below 1 year, ECAI 2, issue size below EUR 10 million.

Equities

In the following, we attempt to identify the characteristics that are important for the liquidity of equities.

Firstly, an overview of the liquidity measures for equities is given in the table below:

Equities

Mean

Amihud

Price impact

Roll

Return volatility

Price change

Trading Volume (mill. EUR)

Turnover ratio

Zero trading

0.492

1.587

1.351

3.411

17.530

753.7

0.071

0.340

Standard deviation

0.954

0.969

2.519

6.099

13.125

1711.2

0.085

0.432

Frequency

19356

19367

18619

19981

21287

21020

26733

PDF to HTML - Convert PDF files to HTML files

The characteristics included in the equity analysis naturally differ from those used in the analysis of the

different bond categories. For equities we still include time dummies and dummies for issue size,

although with a slightly different categorisation than for the bond categories, in that the smallest bucket

includes issues of up to EUR 100 million. In addition, we include industry sector dummies. The

hypothesis is that there may be differences in particular in price volatility measures, as some sectors

are typically characterised by higher earnings volatility for the corporations within certain sectors. The

base case sector is industrials.

Price Volatility Measures

Amihud

Issue size btw. 100-250 Mio. EUR

Issue size btw. 250-500 Mio. EUR

Issue size EUR500M or above

Basic materials

Communications

Concumer, cyclical

Consumer, non-cyclical

Diversified

Energy

Technology

Utilities

Constant

R-squared

Number of observations

-0.534***

(-2.62)

Volume measures

Turnover ratio Zero-trading

0.014***

(11.57)

Price impact Roll

-0.333*

(-1.94)

(-2.7)

Return vol. Price change Trading Vol.

-6.912***

(-5.53)

-3.283*** -7.177***

(-6.19)

14.682

(0.87)

-0.110***

(-25.38)

-1.285***

(-6.62)

-0.616***

(-3.76)

-3.714*** -9.153***

(-3.08)

(-8.03)

-6.896***

(-5.71)

3.688

(0.22)

0.035***

(21.69)

-0.360***

(-45.25)

-2.823***

(-14.84)

-1.215***

(-7.58)

-4.472*** -10.977*** -10.993***

(-3.72)

(-9.69)

(-9.49)

812.353***

(42.58)

0.091***

(97.65)

-0.908***

(-460.71)

-0.036**

(-1.99)

0.022

(1.14)

-0.071

(-1.38)

0.047

(0.51)

0.844***

(3.08)

184.698***

(6.64)

0.005***

(2.96)

0.037***

(7.84)

-0.049***

(-2.74)

-0.308***

(-15.94)

-0.108**

(-2.04)

-0.315***

(-2.99)

-2.166***

(-7.77)

613.861***

(8.74)

-0.023***

(-15.58)

0.066***

(12.43)

0.080***

(4.05)

-0.043**

(-2.31)

-0.045

(-0.91)

-0.006

(-0.05)

-0.278

(-1.06)

160.585***

(6.62)

-0.005***

(-2.77)

0.062***

(14.19)

-0.009

(-0.53)

-0.421***

(-26.12)

-0.097**

(-2.18)

-0.418***

(-4.35)

-4.301***

(-19.43)

18.877

(0.98)

-0.026***

(-23.87)

0.047***

(14.75)

0.368***

(4.47)

-0.052

(-0.93)

-0.142

(-1.06)

0.399

(1.49)

-0.267

(-0.27)

-463.400***

(-22.77)

-0.048***

(-18.81)

0.120***

(8.55)

0.103***

(4.89)

-0.031

(-1.53)

0.134**

(2.13)

0.177*

(1.66)

0.517*

(1.87)

457.728***

(8.3)

-0.014***

(-6.72)

0.078***

(12.62)

-0.247***

(-17.66)

-0.041

(-1.24)

-0.014

(-0.17)

-0.033

(-0.23)

0.769

(1.44)

459.653***

(9.21)

0.048***

(12.94)

-0.012***

(-3.37)

-0.074***

(-3.78)

-0.424***

(-21.22)

-0.238*** -0.494***

(-4.12)

(-4.7)

-4.833***

(-19)

642.557***

(11.55)

-0.040***

(-23.37)

0.145***

(18.71)

3.202***

(16.4)

3.446***

(20.85)

17.020*** 45.673***

(12.82)

(26.3)

32.789***

(22.7)

602.753***

(4.09)

0.061***

(11.94)

0.911***

(90.79)

0.385

19.356

0.415

19.367

0.372

18.619

0.529

19.981

0.305

21.287

0.069

21.020

0.263

26.733

0.744

26.733

Issue size generally has a significant impact on the liquidity metrics for equities, with larger issuances

showing better liquidity values both on price and trading activity related measures. As regards sector,

the evidence is more mixed. Consumer non-cyclical, communications and utilities generally perform

well, with low price volatility related measures relative to other sectors.

Gold

For gold we naturally do not perform a within asset class analysis, as this only consists of one single

asset. The cross asset analysis in section 4.2.1 generally showed this asset class to have relatively

high price volatility, but we are not able to compute comparable trading activity measures as we do not

have daily turnover data for this asset class, although we have evidence that this asset class shows

continuous trading at relatively high volumes.

4.1.3 Assessing appropriate uniform definitions

The approach adopted to finalise uniform definitions of liquid assets builds on the within asset class

analysis described in section 4.2.2. This gives a basis for creating a set of possible definitions based

PDF to HTML - Convert PDF files to HTML files

on explanatory variables that have been identified as having useful predictive power for market

liquidity in each asset class.

An important objective of this stage of the analysis is to ensure that definitions for each asset class

reliably identify securities that are liquid, without either incorrectly defining an illiquid security as liquid

or inefficiently defining liquid securities as illiquid. This is achieved within each asset class by setting a

common classification of a liquid asset that can be applied across asset classes, and classifying each

security on the basis of our empirical analysis. For this stage of the analysis we do not assess liquidity

on the full range of metrics analysed previously, but rather select a single volume-based metric

(turnover ratio) and a single price-based metric (return volatility) and use these in combination to

assess whether an asset is empirically liquid.

Once each asset is classified, we can assess for each possible operational definition how many assets

it captures, how many of those assets defined as liquid were actually illiquid (a Type I error), and also

how many empirically liquid assets are not captured in the definition (a Type II error). Then we can

compare all feasible definitions within an asset class, based on a combination of the factors found to

have explanatory power in section 4.2.2, and assess which definitions perform best at minimising the

rate of Type I and II errors. The chart below illustrates a typical set of results for sovereign bonds. In

this plot, each blue dot represents an operational definition, i.e. a definition which is based on easily

observable characteristics (for instance: bonds issued by sovereign governments, above a given issue

size, rated no less than ECAI 2, with fixed coupons). There is a very large number of such definitions

because each of the criteria can be set at different thresholds, and these choices compound each

other. Type I errors are plotted on the x-axis, on the y-axis 1-Type II errors are displayed, effectively

capturing the fraction of empirically liquid assets successfully identified by each definition. For each

definition, the ‘type I error’ figure is the fraction of bonds in the sample which are objectively illiquid

(based on the liquidity metrics from our analysis) and yet would be defined as liquid were we to adopt

the definition being tested (represented by the blue dot). The ‘type II error’ figure is the fraction of

bonds in the sample which are objectively liquid (based on our analysis) and yet would be defined

illiquid were we to adopt that definition.

In this graph, the successful definitions are those closest to the upper left corner of the chart. It is clear

immediately that most definitions are strictly dominated by a rival definition, that is to say, there exists

another definition that can capture at least as many liquid assets (i.e. the Type II error is no higher)

while capturing fewer illiquid assets (i.e. with a lower Type I error). These definitions should clearly be

disregarded. However, there is still a whole ‘frontier’ of definitions that are undominated (roughly

speaking, the ones standing at the north-eastern edge of the set of blue dots). This frontier goes from

(0,0) to (1,1). It shows the trade-off which exists between Type I and Type II errors: a definition which

excludes all assets would guarantee 0% Type I but would result in 100% Type II errors, while a

definition which accepts all bonds in the sample would achieve 0% Type II but result in 100% Type I

error rate. In this diagram we judged that the point furthest to the left on the 0.7 line on the y-axis,

represented the best trade-off between Type I and Type II errors. This dot corresponds to the

definition we retained for sovereign bonds (sovereign bonds, ECAI 1, a minimum issue size of EUR

250 million).

PDF to HTML - Convert PDF files to HTML files

Plot of Type I and II errors for plausible sovereign bond definitions

0.9

0.8

1- Type II errors %

0.7

0.6

0.5

0.4

0.3

0.2

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

Type 1 Errors %

We performed a similar process for all asset classes for which we had explanatory variables to test

(i.e. not including gold and equities, which are treated as single asset classes with no additional

explanatory characteristics). We used the same thresholds (the one we had used for the sovereign

bonds) for turnover ratio and for return volatility to classify observations as objectively liquid and

objectively illiquid. For each asset class in turn, we then assessed whether there were any definitions

which would select a set of observations which would be comparable to the set of sovereign bond

observations which meets the definition we had selected. However, we did not require that the sets

thus selected should be at precisely the same level, but we allowed for some tolerance. For each

possible definition, we checked whether it would select a subsample which matched the median on the

selected subsample of sovereigns for one of the metrics (either turnover ratio or price volatility) and for

the other metric (turnover ratio or price volatility) we simply required that the security at the 75th

percentile for the turnover ratio measure or the 25th percentile for the price volatility measure be able

to match the median security in the sovereign bond subsample. For asset classes where it was

impossible to find a definition that would achieve this, we considered that these asset classes could

not be eligible for the ‘extremely highly liquid’ level. For the asset class (covered bonds) where there

were definitions that would achieve this, we created the same graph as above, and picked the best

definition in terms of Type I and Type II errors.

Despite introducing these tolerances, the only other asset class with a comparable level of market

liquidity was the covered bond class (as we might have expected from our aggregate comparisons of

asset classes in section 4.2.1). Here, we had identified a tightly defined set of covered bonds with

large issue sizes and meeting the requirements for enhanced transparency and credit risk

management as important determinants of liquidity in section 4.2.2.

PDF to HTML - Convert PDF files to HTML files

The same process was then adopted for the definition of assets of high liquidity and credit quality,

omitting all those assets identified as extremely highly liquid, finding the best performing sovereign

bond definition based on Type I and II errors for the remaining set of assets, then comparing the

relative liquidity of this set to the best sets chosen in the other asset classes. As the liquidity of the

remaining sovereigns was significantly lower, on this stage of the analysis we found definitions in a

number of asset classes which were of a comparable level of liquidity and hence were deemed to

qualify as assets of high liquidity and credit quality. The tables below set out data on the observed

market liquidity of the assets defined as liquid in each asset class under this methodology,

demonstrating that this method achieved a relatively uniform outcome across asset classes

Neither

equities or gold were found to qualify as assets of high liquidity and credit quality as they failed on the

price volatility criteria.

Assets defined as assets of extremely high liquidity and credit quality.

Mean of return

volatility

EEA

sovereign

bonds, ECAI

1, min. issue

size of EUR

250M

EEA covered

bonds, ECAI

1, min. issue

size

EUR500M

0.6

Median return

volatility

0.4

Mean turnover

ratio

0.44

Median

turnover ratio

0.17

75th percentile

turnover ratio

0.43

0.4

0.3

0.16

0.05

0.17

Assets defined as assets of high liquidity and credit quality.

Mean of return

volatility

EEA sovereign

bonds, ECAI 1-

2, min. issue

size of EUR

100M

EEA covered

bonds, ECAI 1,

min. issue size

EUR250M

Corporate

bonds, ECAI 1-

4, min. issue

size

EUR250M,

max time to

0.9

Median return

volatility

0.6

Mean turnover

ratio

0.22

Median

turnover ratio

0.07

75th percentile

turnover ratio

0.27

0.8

0.6

0.09

0.03

0.07

0.9

0.7

0.08

0.04

0.09

PDF to HTML - Convert PDF files to HTML files

maturity 10

years

RMBS, ECAI

1, min. issue

EUR 100M,

max time to

maturity 5

years

Bonds issued

supranationals,

ECAI 1, min.

issue size EUR

250M

Local

government

issued bonds,

ECAI 1-2, min.

issue size EUR

250M, max.

time to maturity

10 years

1.6

0.5

0.06

0.03

0.08

0.7

0.4

0.06

0.02

0.08

0.3

0.9

0.07

0.05

0.08

4.1.4 Computation of Haircuts

Based on the empirical analysis contained in this report, no additional haircuts are warranted for the

assets of high liquidity and credit quality, over and above the minimum levels specified in the Basel

LCR text.

4.2. Principles-based analysis of asset classes

Article 416 describes in general terms the composition of the buffer of the LCR. Whereas Article

509(3) CRR mandates the EBA to report to the Commission a uniform definition of transferable assets

that are of extremely high liquidity and credit quality (Article 416(1)(b)) and of transferable assets that

are of high liquidity and credit quality (Article 416(1)(d)), the CRR itself introduces a set of principles

limiting the scope of the eligible assets.

In accordance with to Article 416(1) CRR, assets need to be

transferable.

Article 416(2) restricts the range of admissible issuers: they should not be issued by a credit institution

or by any of the following: (i) an investment firm; (ii) an insurance undertaking; (iii) a financial holding

company; (iv) a mixed financial holding company; (v) any other entity that performs one or more of the

activities listed in Annex I to Directive 2013/36/EU as its main business. Exceptions are made for

covered bonds, asset backed instruments demonstrated to be of the highest credit quality and bonds

issued by banks set up by a Member State, guaranteed by a Member State or issued by promotional

banks.

PDF to HTML - Convert PDF files to HTML files

Article 416(3) sets out a set of further conditions that assets eligible for the buffer must meet:

a. they should be unencumbered or stand available within collateral pools to be used for the

obtaining of additional funding under committed but not yet funded credit lines available to the

institution;

b. they should not be issued by the institution itself or its parent or subsidiary institutions or

another subsidiary of its parent institutions or parent financial holding company;

their price should generally be agreed upon by market participants and able to be easily

observed in the market, or their price can be determined by a formula that is easy to calculate

based on publicly available inputs and does not depend on strong assumptions as is typically

the case for structured or exotic products;

d. they should be eligible as collateral for standard liquidity operations of a central bank in a

Member State or if the liquid assets are held to meet liquidity outflows in the currency of a third

country, of the central bank of that third country;

e. they are listed on a recognised exchange or they are tradable on active outright sale or via a

simple repurchase agreement on approved repurchase markets. These criteria must be

assessed separately for each market.

While conditions a and b refer to the eligibility of a single asset within the buffer, conditions c to e are

conditions that apply to the asset class itself.

In the following section we will assess each of the asset classes considered by the report in its

empirical analysis against the conditions contained in the aforementioned articles:

1. transferability;

2. admissibleness of the issuer;

3. transparency of pricing;

4. central bank eligibility;

5. listing on a recognized stock exchange or tradability on active outright sale or repo markets.

4.2.1 Discussion of the rationale backing the principles

1. Transferability

Article 416(1) requires assets to be transferable. Assets constituting the buffer of the LCR should

enable banks to generate means of payment at short notice in order to meet liquidity outflows. To this

end banks need to be able to exchange these assets against means of payment. Assets that cannot

be transferred cannot in consequence be exchanged against means of payment and are therefore

inadequate to insure banks against liquidity risk.

PDF to HTML - Convert PDF files to HTML files

2. Admissibleness of the issuer

Article 416(2) excludes assets issued by credit institutions or other financials. Exceptions are made for

covered bonds, asset backed instruments demonstrated to be of the highest credit quality and bonds

issued by banks set up by a Member State, guaranteed by a Member State or issued by promotional

banks.

Since the LCR is designed to protect banks against systemic stress, it is of the utmost importance that

the ratio itself avoid features of pro-cyclicality and wrong way risk. When faced with liquidity stress, a

bank is permitted to use the buffer and liquidate the assets it holds either by outright sale or in repo

transactions. A fire sale of uncovered bonds issued by financials will inevitably lead to a decline in

prices, thus undermining the capacity of banks to refinance themselves on the capital markets and

generating pro-cyclical effects.

If eligible, credit institutions could issue these bonds to each other, increasing the interconnectedness

in the system and thus systemic risk. This argument underlines the importance of the

concept of

outside liquidity.

This concept emphasises that assets which are intended to serve as a reliable

source of liquidity in stress times cannot be solely linked to issuers and investors from the financial

system.

3. Transparency of Pricing

The CRR requirement of pricing transparency relates to the Basel HQLA characteristic of “ease and

certainty of valuation”. The Basel framework argues that “an asset´s liquidity increases if market

participants are more likely to agree on its valuation. Assets with more standardised, homogeneous

and simple structures tend to be more fungible, promoting liquidity. The pricing formula of a high-

quality liquid asset must be easy to calculate and not depend on strong assumptions. Inputs into the

pricing formula must also be publicly available. In practice, this should rule out the inclusion of most

structured or exotic products.” (Basel 2013, paragraph 24) The rationale behind this assumption is that

assets that form part of the buffer might need to be liquidated in stress periods and at short notice.

Transparent pricing facilitates price discovery under stressed conditions.

4. Central bank eligibility

Liquid assets should be eligible collateral for standard liquidity operations of the central bank of the

country in question. This does not imply that an institution actually does need to have direct access to

these central bank operations.

The general outset of the LCR underlines the importance of banks´ reliance on private markets to

increase their resilience against liquidity stress. Policymakers have identified issues of moral hazard

when banks rely too heavily on central bank refinancing in stressed periods. Central banks have

generally been perceived as an insurance against liquidity risk, limiting the scope of action of central

banks in a crisis.

PDF to HTML - Convert PDF files to HTML files

However, central banks are lenders of last resort. If private markets fail, it should be possible to

pledge eligible assets with the central bank to obtain liquid funds. Therefore this criterion can be

described as backstop criterion for cases of severe market failure.

Given the varying scope of central bank eligibility within the European Union as shown in the following

graph, the compulsory condition of central bank eligibility will lead to different definitions of liquid

assets in the different currency zones of the European Union. Policymakers should balance the

desired effect of central bank eligibility and the undesired effect of an uneven playing field within

Europe when reconsidering the requirement of central bank eligibility. In the Basel framework the

condition of central bank eligibility is phrased with “ideally” (Basel 2013, paragraph 23).

PDF to HTML - Convert PDF files to HTML files

List of central bank eligible assets per country:

PDF to HTML - Convert PDF files to HTML files

5. Listing on a recognised stock exchange or tradability on active outright sale or repo

markets

The existence of private markets for liquid assets is a prerequisite for them being able to be liquidated.

These markets need to be “active and sizeable” in Basel terminology.

The CRR requires these criteria to be “assessed separately for each market”. Since the focus of the

LCR lies in the liquidation of the buffer on private markets, it is crucial that these markets exist. Listing

on a recognised stock exchange can be interpreted as evidence that there is a market place for the

asset. Otherwise assets should be tradable on active outright sale markets. As long as there is a

possibility to sell the assets in a reasonably short time, the criterion should be met. Therefore for those

assets for which we have sufficient data to perform a quantitative analysis, judging whether this

criterion has been met is primarily an empirical question.

Taking into consideration repo markets reflects the circumstance that repo markets can substitute

outright sale markets as generators of liquidity. For some markets this is especially true under

stressed conditions.

It should therefore be ensured that these private markets work under stressed conditions. This implies

a high diversity of market participants, a robust infrastructure and ongoing trading.

4.2.2 Mapping between asset classes and principles

1. Asset classes considered by the empirical analysis:



Government Bonds

Transferability.

Government bonds are transferrable

Admissibleness of the issuer.

An exemption from the exclusion of the issuer is made for bonds issued

by credit institutions that are set up by a Member State central or regional government and that

government has an obligation to protect the economic basis of the institution and maintain its viability

throughout its lifetime or the asset is explicitly guaranteed by that government.

Transparency of pricing.

Government bonds are standardised and have simple structures. Valuation

does not depend on strong assumptions.

Central bank eligibility.

State government bonds are central bank eligible for standard liquidity

operations in all countries where these exist. Bonds issued by regional governments are not accepted

in Bulgaria, the Czech Republic, Iceland, Latvia, Lithuania and Romania. Securities guaranteed by the

central government are not admissible in Bulgaria, the Czech Republic, Poland and Romania.

PDF to HTML - Convert PDF files to HTML files

Listing on a recognised stock exchange or tradability on active outright sale or repo markets.

In many

countries, government bonds are listed on a recognised stock exchange. The results of the

quantitative analysis show that government bonds are traded on active markets. Alongside cash,

government bonds are the most frequently used collateral in repo transactions.



Corporate Bonds

Transferability.

Corporate bonds are transferrable.

Admissibleness of the issuer.

Corporate bonds issued by a credit institution, an investment firm, an

insurance undertaking, a financial holding company, a mixed financial holding company or any other

entity that performs one or more of the activities listed in Annex I of Directive 2013/36/EU as its main

business are not admissible as liquid assets under the LCR buffer.

Transparency of pricing.

Corporate bonds are standardised and have simple structures.

Central bank eligibility.

Corporate bonds are central bank eligible for standard liquidity operations in

many countries. They are not eligible in Bulgaria, the Czech Republic, Denmark, Iceland, Lithuania

and Romania.

Listing on a recognised stock exchange or tradability on active outright sale or repo markets.

Most

corporate bonds are listed on a recognised stock exchange. The results of the empirical analysis show

that most corporate bonds are traded on active markets. Corporate bonds are less frequently used in

repo transactions, although they are eligible collateral for some repo pools comprising central bank

eligible assets.



Covered Bonds

Transferability.

Covered bonds are transferrable.

Admissibleness of the issuer.

Covered bonds eligible for the treatment set out in Article 129(4) or (5)

CRR, or bonds referred to in Article 52(4) of Directive 2009/65/EC, are exempt from the exceptions

concerning the admissibleness of the issuer. The perceived rationale is that given the collateralisation

of covered bonds, these are protected from the effects of wrong way risk. Self-issued covered bonds

remain excluded.

Transparency of pricing.

The structure and design of covered bonds is standardised in national

covered bond regulation. These may involve differing degrees of standardisation with respect to the

structure and transparency requirements of the bonds and special regulation for the issuer.

PDF to HTML - Convert PDF files to HTML files

Central bank eligibility.

Covered bonds are central bank eligible for standard liquidity operations in

many countries. They are not eligible in Bulgaria, the Czech Republic, Iceland, Lithuania and

Romania.

Listing on a recognised stock exchange or tradability on active outright sale or repo markets.

Most

covered bonds are listed on a recognised stock exchange. The results of the empirical analysis show

that most covered bonds are traded on active markets. Covered bonds are less frequently used in

repo transactions, although they are eligible collateral for some repo pools comprising central bank

eligible assets. Covered bonds which are privately placed are not eligible for the liquidity buffer under

the LCR.



Bonds issued by promotional banks

Transferability.

Bonds issued by promotional banks are transferrable.

Admissibleness of the issuer.

Bonds issued by promotional banks are exempt from the conditions

limiting the admissibleness of the issuer. The rationale resides in their direct link to state or regional

governments and their role in sponsoring the economy. Bonds issued by promotional banks may also

have a state or regional government guarantee.

Transparency of pricing.

Bonds issued by promotional banks are standardised and have simple

structures.

Central bank eligibility.

Bonds issued by promotional banks that do not have a state guarantee are not

accepted in Bulgaria, the Czech Republic, Denmark, Iceland, Lithuania and Romania. Otherwise, the

requirements for state guaranteed bonds apply.

Listing on a recognised stock exchange or tradability on active outright sale or repo markets.

Most

bonds issued by promotional banks are listed on a recognised stock exchange. The results of the

quantitative analysis show that most bonds issued by promotional banks are traded on active markets.

Bonds issued by promotional banks are less frequently used in repo transactions, although they are

eligible collateral for some repo pools comprising central bank eligible assets.



ABS including RMBS

Transferability.

ABS are transferrable.

Admissibleness of the issuer.

ABS if demonstrated to be of the highest credit quality as established by

the EBA pursuant to the criteria in Article 509(3), (4) and (5) are exempt from the exceptions

concerning the admissibleness of the issuer. This criterion should therefore not be binding within the

report. ABS which are self-issued are not eligible as liquid assets under the LCR buffer.

PDF to HTML - Convert PDF files to HTML files

Transparency of pricing.

The structure and design of ABS are not subject to national regulation.

Generally, these products are bespoke. Industry initiatives have been launched to increase the

transparency of ABS.

Central bank eligibility.

ABS are not eligible for standard central bank operations in Bulgaria, the

Czech Republic, Denmark, Hungary, Iceland, Latvia, Lithuania, Poland and Romania. In Sweden they

are eligible only until 31 December 2013.

Listing on a recognised stock exchange or tradability on active outright sale or repo markets.

Most

ABS are listed on a recognised stock exchange. The results of the empirical analysis show that the

liquidity of ABS markets is highly heterogeneous. ABS are less frequently used in repo transactions,

although they are eligible collateral for some repo pools comprising central bank eligible assets. ABS

which are privately placed are not eligible for the liquidity buffer under the LCR.



Equities

Transferability.

Equities are transferrable.

Admissibleness of the issuer.

Equities issued by a credit institution, an investment firm, an insurance

undertaking, a financial holding company, a mixed financial holding company or any other entity that

performs one or more of the activities listed in Annex I of Directive 2013/36/EU as its main business

are not admissible as liquid assets under the LCR buffer.

Transparency of pricing.

Equities are standardised and have simple structures. Equities have high

price volatility.

Central bank eligibility.

Equities in general are not central bank eligible.

Listing on a recognised stock exchange or tradability on active outright sale or repo markets.

Equities

are listed on a recognised stock exchange. The results of the quantitative analysis show that equities

are traded on active markets and show high trading volumes and turnover rates.



Gold

Transferability.

Gold is transferrable.

Admissibleness of the issuer.

As a physical asset, the criteria that refer to the admissibleness of the

issuer do not apply.

Transparency of pricing.

Gold and gold contracts are standardized and have simple structures.

Central bank eligibility.

Gold is not eligible as collateral for standard central bank operations in the EU.

PDF to HTML - Convert PDF files to HTML files

Listing on a recognised stock exchange or tradability on active outright sale or repo markets.

Gold is

not listed on stock exchanges. It is traded in outright sale markets. Gold is not found to be used as

collateral in our survey of repo transactions.

2. Asset classes not considered by the empirical analysis due to lack of data (a detailed

assessment is required)



Credit Claims

Transferability.

The secondary market for credit claims is limited.

Admissibleness of the issuer.

Credit claims are claims generated by credit institutions, collateralised

however by claims against a private or corporate customers.

Transparency of pricing.

Credit claims are claims that banks hold against customers based on an

individual loan agreement. Therefore there is no standardised or uniform documentation. Credit claims

lack external prices and are generally not rated. The price of credit claims is not generally agreed

upon by market participants. At the same time, it may be argued that credit claims can be turned into

liquid assets by including them in securitisations, which may subsequently be issued in the primary

and traded in the secondary market. A drawback is that the securitisation process may need

considerable time and may take too long to obtain liquidity within the 30 day time limit set out for

liquidity coverage ratio, without accepting large discounts or haircuts. Another consideration may be

that under distressed market conditions the liquidity of credit claims may quickly vanish, being highly

dependent on the ease of valuation and market conditions. Overall, the lack of liquidity limits the

usefulness of credit claims in the liquidity coverage requirement set out under Article 412 CRR.

Central bank eligibility.

The Eurosystem provides credit against a set of non-marketable asset

including credit claims subject to eligibility criteria . Under the Eurosystem’s general framework, credit

claims to debtors or guarantors from the public sector, non-financial corporations and international and

supranational institutions which are established in the euro area, subject to high credit standards and

denominated in euros are eligible as collateral . The reasons for using credit claims by the Eurosystem

are twofold. First, credit claims were already used as eligible collateral by a number of Eurosystem

members. Second, the Eurosystem aims at implementing monetary policy with a wide set of collateral.

As of Q2 2013, credit claims used as collateral in Eurosystems operations accounted for EUR 452.5

billion or 19% of all used collateral . The inclusion of credit claims as eligible LCR assets would thus

significantly widen the definition of extremely high and high liquid assets and undermine the objective

of the LCR.

See Article 6.2.2 in the Guideline of the ECB of 20 September 2011 on monetary policy instruments and

procedures of the Eurosystem (recast) (ECB/2011/14).

In addition, retail mortgage backed debt instruments and fixed term deposits from eligible counterparties are also

eligible as non-marketable assets.

N.B.: the volume and share of non-marketable assets pledged as collateral in the Eurosystem’s operations has

gone up significantly since the onset of the financial crisis. In 2006, the corresponding share was 4 percent.

PDF to HTML - Convert PDF files to HTML files

Listing on a recognised stock exchange or tradability on active outright sale or repo markets.

The lack

of available evidence for the quantitative analysis reflects the lack of secondary market trading activity

in these assets. Credit claims cannot be liquidated directly in secondary markets.



Major index linked equity instruments: Exchange traded funds (ETFs)

Exchange traded funds are funds traded on the stock exchange investing in all kinds of assets. What

started as a simple plain vanilla replication of a main stock market index, which was passively

managed, may now also be an actively managed synthetic ETF where the assets are not held outright

but where the ETF attempts to replicate index performance via swap contracts. In addition, ETFs

holding equity shares outright obtain a significant part of their income from securities lending giving

rise to counterparty credit risk concerns. From a macroprudential perspective, a sudden withdrawal by

ETF investors may negatively impact the market liquidity of assets held by ETFs . As a result, equity

ETFs besides being volatile may not exhibit the desired characteristics for the liquidity coverage

requirement.

Transferability.

ETFs can be transferred on private markets. While low transaction and management

costs permit the efficient liquidation of ETF shares to ETF investors, the underlying assets of ETFs

typically give rise to high volatility

Admissibleness of the issuer.

ETFs are issued either by financial institutions or SPVs. Especially

synthetic ETFs which only replicate the index are highly dependent on the ability of the credit

institution to honour its swap contracts with the emitting SPV.

Transparency of pricing:.The

price of the ETF should mirror the price of the index it represents.

Central bank eligibility:

ETFs are not central bank eligible

Listing on a recognised stock exchange or active outright sale or repo markets.

The assets of equity

ETFs are by definition listed on a stock exchange. They are not usually used as collateral in repo

transactions.



Mutual funds (CIUs)

For the purposes of this assessment we restrict the assessment of mutual funds to CIUs mentioned in

the CRR reporting requirement.

As per Article 416(6) CRR, shares in CIUs may be treated as liquid assets provided that a CIU (apart

from derivatives to mitigate interest rate or credit or currency risk), only invests in liquid assets eligible

for the LCR. The CIU is therefore considered to be a liquid asset depending on the assets it invests in.

The above-mentioned principle-based criteria refer only to specific assets. It seems however useful to

assess whether CIUs have the liquidity generating capacity that is needed when allowing CIUs to form

See Financial Stability Board (2011), Potential financial stability issues arising from recent trends in exchange

traded funds.

PDF to HTML - Convert PDF files to HTML files

part of the buffer. This refers mainly to the smoothness and timeliness of them being liquidated. An

institution should therefore be able to liquidate shares of CIUs at short notice in private markets or

return their shares to the asset management company. In the former case, liquidation should be

possible in active outright or repo markets. In the latter case, the decision should lay exclusively with

the bank returning the CIU. Moreover, consideration should also be given to the specific inclusion of

shares in CIUs in the LCR. Shares of a CIU investing entirely in assets of extremely high liquidity may

be considered as eligible in Level 1, while CIUs investing in assets corresponding to the definition of

high liquidity could be included in Level 2. In order to facilitate the classification of CIU shares, CIUs

will need to be transparent with regard to their investment portfolios. Further consideration will be

needed for cash held by CIUs as deposits with banks with regard to their inclusion in the LCR.

Summary Table: Mapping of assets to principles for acceptable liquid assets:

Transferability

Issuer

Pricing

Central

eligibility

Bank

Listing

active outright

sale or repo

markets

Government

bonds

Covered

bonds

Corporate

bonds

Bonds issued

by promotional

banks

ABS

RMBS

Non-financial

Equities

Gold

Credit Claims

eligible

CBs

ETF

CIU

with

incl.



not applicable



not applicable



or X*



or X*



or X*



not applicable



not applicable



or X*



or X*



not applicable

*Depends on the regime of the national central bank

4.2.3 Conclusions of the principles-based analysis

The conclusions that can be drawn from the principles-based analysis are as follows:

PDF to HTML - Convert PDF files to HTML files

Central bank eligibility as collateral for standard operations differs substantially between the currency

zones in the European Union. Policymakers should balance the desired effect of central bank eligibility

and the undesired effect of an uneven playing field within Europe when reconsidering the requirement

for central bank eligibility. In the Basel framework the condition of central bank eligibility is prefaced

with “ideally” (Basel 2013, paragraph 23).

The principles-based analysis of the different asset classes shows the following result:



Government, corporate and covered bonds all pass the principles-based assessment at an asset

class level. However, the latter two are not eligible as collateral in standard central bank

operations in some countries within the European Union.



ABS show possible deficiencies in the transparency of pricing and trading.

Equities and gold both fail the central bank eligibility criteria.

A principles-based assessment indicates that credit claims and ETFs should not be considered as

eligible as they fail on multiple criteria.

This analysis at an asset class level should not exempt institutions from applying the principles to

individual assets, as for some markets and/or assets this general validation may not hold true. It

remains the responsibility of individual single bank to judge whether a single asset would qualify to be

eligible for the LCR-buffer.

4.3. Quantitative analysis of repo market data

Article 417(b) CRR specifies that institutions may only report as liquid assets those that, inter alia, are

legally and practically readily available at any time during the next 30 days to be liquidated via outright

sale or via a simple repurchase agreement on approved repurchase markets in order to meet

obligations coming due.

In line with this, feedback during the public consultation clearly pointed

towards the importance of the repo market in any analysis to provide uniform definitions of assets of

high and extremely high liquidity and credit quality.

While sections 4.3.1 to 4.3.3 present qualitative summaries of the European repo market as well as its

development during recent years, sections 4.3.4 and 4.3.5 present a survey-based analysis.

4.3.1 Summary of the European repo market

The European repo market grew steadily before the crisis and remained within stable relatively

bounds after December 2009, it is currently experiencing expansion which may be partly due to LTRO

repayments.

As the latest ICMA European Market Survey (June 2013) shows, cross-border repo transactions

continue to dominate, although their share fell from 50.5% in December 2012 to 48.2%, while

PDF to HTML - Convert PDF files to HTML files

domestic and anonymous transactions rose to 30.7% and 21.1% respectively. CCPs retain an

important role in the European repo market, with a share of approximately 30% of all deals, although

market sentiment is good and risk aversion should not be high. At the same time, the tri-party segment

stayed below 10%. European repo transactions are mainly conducted in euros (about 65%), with U.S.

dollars and British pounds following (about 15% and 11%). German, British and French collateral

dominates (22%, 12% and 12%), although British government securities exhibited a significant decline

and

Pfandbriefe

halved their share. In tri-party transactions, government securities play the most

important role (38%), followed by equity (21%) and corporate bonds (14%). Haircuts narrowed slightly

for a number of collateral types in 2013 and vary between 2.2% for public agencies/sub-national to

9.5% for commercial mortgage-backed. The haircuts of lower quality collateral had increased in 2012.

Generally the maturities of the contracts are short, 57% of them maturing in less than a month, up

from 50.5% in December 2012, probably a sign of migration from the LTRO towards market funding.

The degree of concentration is on a downward sloping trend.

4.3.2 The EU repo market during the financial crisis

During the recent financial turmoil the terms for repo transactions tightened and the absolute volume

of repos generally declined, a noteworthy exception being the German Eurex repo market and its

EURO GC (General Collateral) pooling segment. However, because market participants experienced

a loss of confidence in their counterparties’ creditworthiness, the importance of secured lending

increased, which translated into superior relative volumes of repos, compared to the rest of the money

markets.

To begin with, the average repo maturity shortened significantly throughout the world, overnight repos

gaining an increasingly important share of the market, while Germany remained the exception and

experienced longer maturities due to the reduced credit risk on the CCP-cleared market.

With regard to collateral, cross-border use of it increased, reflecting the shortage of high quality

collateral. Interbank repos were mainly based on government securities, while ABS and non-

marketable securities were used with the central banks. Markets that allowed re-use of collateral seem

to have performed better (e.g. Luxembourg). In addition, the liquidation of collateral posed, at times,

problems of a legal and operational nature.

Margins and haircuts increased in many markets, such as the US and the UK, but not in Germany,

Japan or Switzerland. Nevertheless, settlement fails generally did not increase, excepting the US, UK

and Japan, which saw an increase in settlement fails especially during the period around the default of

Lehman Brothers. The cost of failing was however low.

Central banks played an important role during the crisis by actively providing funding in the repo

market and broadening the list of central bank eligible collateral. At the same time, the liquidity

facilities they offered, in combination with a low interest rate environment, further contributed to a

reduction in repo transactions in some markets.

PDF to HTML - Convert PDF files to HTML files

Consistent with the rise in risk aversion, credit risk mitigants proved to be especially valuable and

CCP-cleared repos actually increased in volume in some markets or decreased considerably less than

non-CCP cleared ones. Furthermore, although tri-party repo seems to have remained a reliable

source of funding for some institutions, the tri-party repo volume generally decreased, but to a lesser

extent than for bilateral transactions.

Overall, high credit risk aversion, combined with legal and operational issues related to the liquidation

of collateral and the lack of transparency in some repo markets led to a broad deleveraging of market

participants. Credit lines have been significantly reduced and counterparty limits tightened. High

quality collateral became scarce, while interest rates plummeted. Repo maturities decreased as

margins and haircuts increased. Under such conditions, repo activity could only shrink in nominal

terms.

4.3.3 Developments in the EU repo market

The GC pooling segment is based on standardized baskets of collateral and saw an increase in

popularity over the past years. It is also attractive because it makes use of a third party, which is often

a central clearing counterparty. A good example of such a market is the Eurex GC pooling segment,

which increased in volume by more than 10% per year even during the financial turmoil. An important

benefit of trading over it is the reusability of collateral for other money market transactions; in

particular, the participants have also the opportunity of refinancing within the framework of European

Central Bank open market operations. Nevertheless, it provides anonymity and eliminates credit or

security allocation issues through its CCP.

With regards to the tri-party segment of the repo market, it is much larger in the United States

compared to Europe, amounting to as much as two thirds of total transactions. In Europe, not only

does the tri-party segment represent approximately 10% of the market, but it is on a downward trend

and it currently experiencing a low. Possible reasons for such behavior are the fact that tri-party

agents do not absorb any of the risks associated with the transactions and, at the same time, 3-year

LTROs are competing with tri-party repo financing.

4.3.4 Data description

For the quantitative analysis of the European repo market, we use information from a survey-based

data collection. Our data contains information per asset class and collateral type from June 2008 to

June 2013. More specifically, the dataset includes information on the total amounts and percentages

outstanding as well as the average haircuts of all asset classes, and is further broken down by

currency (EUR, GBP, other) in addition to credit quality steps.

The following asset classes are included: 1) EEA sovereign debt; 2) Non-EEA sovereign debt; 3) EEA public

agency/Non-central government PSEs; 4) Non-EEA public agency / Non-central government PSEs; 5) BIS,

IMF, EC and MDB debt; 6) ESM and EFSF debt; 7) Financial corporate debt; 8) Non-financial corporate debt;

9) Covered bonds; 10) ABS; 11) RMBS; 12) Other MBS; 13) CDOs; 14) CLNs and 15) Equities.

PDF to HTML - Convert PDF files to HTML files

As is usually the case with surveys, our data are subject to potential measurement biases. While we

cannot do this with absolute certainty, we verified the data as far as possible via cross-comparisons

with other data sources. On top of this general problem, we could only collect detailed information

from 2008 to 2013, implying that our time series might not be fully representative. Importantly,

however, our dataset includes a few months before the failure of Lehman Brothers and is therefore

sufficient for our purposes. A third shortcoming of our data is that for some elements of the analysis it

only includes tri-party repos cleared by a limited number of large counterparts. Each EMU country

however is represented in every period of our data. Data included in the analysis does not measure

the value of repos transacted with central banks as part of refinancing operations.

Despite these shortcomings, our data can be considered to be sufficiently representative of the

European market and will facilitate a comparison across asset classes at a given point in time and a

comparison across time periods of the individual asset classes. Still, the analysis of the repo markets

can only be seen as supplementary to the main data analysis and is not a stand-alone exercise.

4.3.5 Data analysis for the EU repo market

Total value outstanding

Figure 1 shows the total value at a given point in time for repos and reverse repos outstanding at

European institutions covered by the survey. It highlights a decline in repo business in 2008, in the

middle of the financial crisis, and a steady increase in repos outstanding from that point, back to pre-

crisis levels. The amount outstanding reached a recent low of EUR 4.6 trillion in Dec 2008 and a high

of EUR 6.9 trillion in June 2010.

Figure 1 – Total value outstanding 2004 - 2013

PDF to HTML - Convert PDF files to HTML files

Maturity analysis

The maturity profiles of repo transactions have changed over time. Historically, repo maturities have

been relatively short with over 60% of outstandings having a maturity of 3 months or less throughout

the period 2001 – 2013. Between 2007 and 2009 there was a marked shortening of maturities. By

December 2009, over 40% of repos outstanding had a maturity of 7 days or less. Over the same

period there was a corresponding a drop in outstandings with maturities of over 1 month. In recent

years there has been an increase in open maturity repos and forward-forward repos. The changes to

maturity profiles can be seen in Table 1.

Table 1 – Maturity analysis 2007 - 2012

Dec-07

1 day

2-7days

7 days

1m o

1 m o

3 m o

6 m o

6m o

<= 12

m o

more than

12m o

forw ard-forw ard repos

open

14%

17%

23%

20%

Dec-08

18%

17%

20%

19%

Dec-09

22%

18%

23%

15%

11%

Dec-10

21%

19%

23%

15%

Dec-11

16%

17%

13%

10%

Dec-12

17%

16%

17%

16%

13%

Jun-13

17%

16%

17%

16%

13%

Figure 2 – Maturity analysis 2007 - 2013

PDF to HTML - Convert PDF files to HTML files

Breakdown by collateral type

Figure 3 gives the breakdown by collateral type of the main fixed income asset classes. Sovereign

debt is the most commonly used collateral in repo transactions, accounting for approximately 39% of

all outstanding repos followed by public agency/PSE and supranational debt. Covered bonds account

for 12% and corporate bonds 10%, however the majority of this debt is financial corporate debt. While

ABS make up 4% of the total outstanding, this is comprised of all types of ABS, including RMBS,

CMBS, CDO and CLNs.

Figure 3 – % Outstanding by fixed income collateral type as at June 2013

The type of collateral used in fixed income repo transactions has changed significantly since the onset

of the financial crisis. In 2008, corporate debt and ABS would have accounted for over 59% of all fixed

income repos. This combined total has dropped to 13% in 2013. Sovereign, public agency/PSE,

supranational debt and covered bonds have gone from making up 33% of the total outstanding in

2008 to 74% in 2013. Figure 4 emphasises this change over a relatively short time period.

PDF to HTML - Convert PDF files to HTML files

Figure 4 – % Outstanding by fixed income collateral type 2008 - 2013

When Equity is included in the analysis as a type of collateral used in tri-party repo transactions and

compared against fixed income collateral types, it is shown to make up an increasing amount of the

total collateral outstanding. Figure 5 highlights the increase in the use of equity compared to fixed

income collateral from 2009 to 2013.

It should be noted that while the volume outstanding of equity collateral is relatively high, it does not

necessarily signify a corresponding increase in the use of the asset in money market transactions,

rather, it highlights that there has been an increase in the amount of collateral used in equity market

transactions relative to fixed income repo transactions.

Figure 5 – % Outstanding - fixed income & equity - 2009 - 2013

PDF to HTML - Convert PDF files to HTML files

Analysis by collateral type

The following section examines the change over time of the use of the main asset classes as collateral

and gives a further breakdown of the percentage outstanding in asset subclasses and includes detail

on the relative use of collateral with credit quality as a determining factor.

Sovereign debt

By June 2008, there was a large decrease in the total repos outstanding which reflects a reduction of

business transacted through repos. Between December 2008 and June 2009 there was another

significant shift. The repo market itself remained relatively stagnant, yet dealers moved away from

using financial corporate debt and ABS as collateral and relied more on sovereign bonds. As the

market picked up in 2010 this trend continued with the percentage of sovereign debt to total collateral

outstanding remaining relatively constant. This can be seen in Figure 6. There has been a slight

reduction in the use of sovereign debt since 2011, which coincides with a relative increase in the use

of public agency and supranational debt.

The amount of non-Euro EEA sovereign debt used for collateral relative to total repos outstanding has

remained relatively constant throughout the period, fluctuating between 2% - 4% of total collateral

outstanding.

Figure 6 – Sovereign debt as a % of fixed income repos outstanding 2008 - 2013

Public agency & non-central government public sector entities

The amount of public agency and non-central government PSE debt used as collateral was relatively

constant between June 2008 and June 2011. Starting in December 2011 there was a large increase in

the use of these bonds as collateral. In a six month period, the amount used as collateral doubled. The

amount relative to total fixed income collateral outstanding went from 10% in June 2011 to 24% in

June 2013.

PDF to HTML - Convert PDF files to HTML files

Figure 7 – Public agency & non-central govt. PSE debt as a % of fixed income repos outstanding 2008

- 2013

Supranational debt

Supranational debt used as collateral, i.e. debt issued by the BIS, the IMF, the Commission,

multilateral development banks, the ESM and the EFSF has increased from close to 0% of all

outstanding collateral in 2008 to 12% of all fixed income collateral traded in 2013. These bond types

became more popular from 2009 as dealers reduced their holdings in other types of collateral. EFSF

debt was first issued in 2011; ESM and EFSF debt now account for circa 5% of all outstanding fixed

income collateral or 42% of the supranational debt used as collateral.

PDF to HTML - Convert PDF files to HTML files

Figure 8 – Supranational debt as a % of fixed income repos outstanding 2008 - 2013

Corporate debt

The amount of corporate debt used as collateral in repo transactions has decreased considerably

since 2008. Taken together, financial and non-financial corporate debt accounted for over 43% of

outstanding repo collateral in 2008. This has since dropped to 10%.

Figure 9 – Corporate debt as a % of fixed income repos outstanding 2008 - 2013

When financial corporate debt is analysed by credit quality steps 1 to 3, as in Figure 10, the decrease

in outstandings across all credit quality steps can be seen. In 2008, financial corporate debt at credit

PDF to HTML - Convert PDF files to HTML files

quality step 1 made up the larger proportion of all outstanding financial corporate debt collateral. In

2013 however, this split is less pronounced, repo outstandings across all credit quality steps are

relatively low. The decrease in all credit quality steps is indicative of a reduction in the use of financial

corporate debt as collateral, rather than simply debt, moving to a lower credit quality step following a

downgrade.

Figure 10 – Financial corporate debt by credit quality step as a % of fixed income repos outstanding

2008 - 2013

When non-financial corporate debt is analysed by credit quality step, this downward trend is again

prevalent. Apart from a possible outlier in Dec 2009, Figure 11 shows that there has been a steady

decrease in the use of non-financial corporate debt as collateral. It accounted for 9% of all fixed

income outstandings in 2008, but is now at 1% across all credit quality steps.

PDF to HTML - Convert PDF files to HTML files

Figure 11 – Non-financial corporate debt by credit quality step as a % of fixed income repos

outstanding 2008 - 2013

Covered bonds

The use of covered bonds as repo collateral has generally increased over the time period of the

analysis. There was a decrease in the use of covered bonds between 2008 and 2010. While total

collateral outstanding also decreased in this period, the proportion of covered bonds used relative to

sovereign/public agency/supranational debt fell slightly. Until 2012 it was mostly covered bonds at

credit quality step 1 that were used in this category. In 2012 and 2013 there was an increase in the

use of covered bonds at credit quality step 2. Covered bonds used as repo collateral have

predominately been denominated in EUR.

PDF to HTML - Convert PDF files to HTML files

Figure 12 – Covered bonds by credit quality step as a % of fixed income repos outstanding 2008 -

2013

Asset backed securities

When all classes of ABS are analysed collectively or separately, the result is the same. There has

been a marked decrease in the use of these asset types as collateral in repo transactions. The

amount outstanding of this asset category has gone from 16% in 2008 to 3.7% in 2013. In 2008 RMBS

and other MBS made up the bulk of total ABS outstanding. While the outstandings of all types of ABS

fell, the decrease in the use of RMBS has been less pronounced. Figure 13 shows the split between

types of ABS from 2008 to 2013.

PDF to HTML - Convert PDF files to HTML files

Figure 13 – ABS by credit quality step as a % of fixed income repos outstanding 2008 - 2013

When RMBS as a sub category of asset is analysed, as indicated in Figure 14, apart from possible

outliers in 2010, between 2008 and 2012 there was not a significant fall in the use of these assets.

2013 to-date has seen the largest drop in outstandings of RMBS as collateral; there was however an

increase in the use of RMBS at credit quality step 2 in 2012 and 2013.

PDF to HTML - Convert PDF files to HTML files

Figure 14 – RMBS by credit quality step as a % of fixed income repos outstanding 2008 - 2013

The decrease in the use of ABS as collateral for repo transactions is emphasised when the use of

other MBS is analysed as in Figure 15. Commercial MBS are included in this category. Securities with

credit quality step 1 made up the bulk of other MBS that were used as collateral in 2008. Virtually no

other MBS of either credit quality step 1 or 2 were used as collateral in repo transactions in 2013.

PDF to HTML - Convert PDF files to HTML files

Figure 15 – Other MBS by credit quality step as a % of fixed income repos outstanding 2008 - 2013

Equity

Figure 16 highlights that the use of equities in repo transactions increased between 2009 and 2012. It

should be noted that this analysis does not differentiate between the markets that the repo

transactions are used for, i.e. the majority of these transactions may relate to equity market positions

rather than use as a liquidity generating instrument.

PDF to HTML - Convert PDF files to HTML files

Figure 16 – Equity as a % of total repos outstanding 2009 - 2013

Analysis by average haircut

Average haircuts applied to the different collateral types used in repo/reverse repo transactions were

analysed. The same type of survey data was used for this analysis as that discussed in section 4.4.4.

It should be noted that this survey is relatively new for institutions and may be less reliable than the

survey on repo volumes outstanding.

In general, average haircuts for RMBS, corporate and covered bonds are greater than those for

sovereign, public agency and supranational debt. Average haircuts for RMBS, covered and sovereign

bonds increased since 2008 while those for AAA corporate debt decreased. Average haircuts for

sovereign debt in 2013 are greater than those for public agency and supranational. The opposite was

the case in 2008.

PDF to HTML - Convert PDF files to HTML files

Figure 17 – Repo haircut by collateral type 2008 - 2013

4.3.6 Summary of the EU repo market

The presence of an active repo market increases the market liquidity of a particular asset class. The

volume of repo/reverse repo transactions fell during the recent financial crisis however it has

recovered somewhat in recent years, indicating that during a stressed situation, as in outright sales,

liquidity related to repo markets may lessen.

The relative composition of the type of collateral used in repo transactions has changed significantly

since 2008, with sovereign, public agency, supranational and covered debt replacing corporate debt

and ABS as the main types of collateral used. This relative composition is still changing. In the period

immediately following the financial crisis, sovereign debt became the main type of collateral used.

While this remains the case, the proportion of public agency and supranational debt as collateral is

increasing markedly.

In relation to the individual asset classes, it can be said that in general, there is an active repo market

for sovereign, public agency, supranational and covered bonds however the same cannot be said for

corporate bonds and ABS, as the percentage of these asset types used has fallen significantly since

the onset of the financial crisis.

PDF to HTML - Convert PDF files to HTML files

5. Operational requirements of liquid assets

This section of the report is written in response to the mandate presented to the EBA in Article

509(5)(c)CRR that “the EBA shall report to the Commission on the operational requirements for the

holdings of liquid assets, as referred in points (b) to (f) of Article 417, in line with international

regulatory developments”.

Independent of the eligibility in principle of the assets covered by Article 416(1),

an institution’s

holdings of liquid assets have to meet certain operational requirements so that these assets can

effectively be considered as part of the buffer under the Liquidity Coverage Requirement as specified

in Article 412 CRR.

These operational requirements are set out in Articles 416 and 417 CRR. The EBA is mandated to

produce a report deliberating on the conditions set out in Article 417(b) to (f). The diversification

requirement of Article 417(a) is not part of the mandate for this report and will therefore not be

considered further.

This report intends to reflect the rationale of each requirement and deliberates on their practical

implications. The report furthermore highlights cases in which the rationale of the requirements seems

weak, and contains recommendations for future review.

The range of liquid assets as set out in Article 416(1) CRR includes cash (notes and coins) and

exposures to central banks (a), transferrable assets ((b) to (d)), standby credit facilities granted by

central banks (e) and legal or statutory minimum deposits with the central credit institution and other

statutory or contractually available liquid funding from the credit institution (f). Shares or units in CIUs

may be treated as liquid assets up to an absolute amount of EUR 500 million in the portfolio of liquid

assets in each institution as long as it only invests in liquid assets (Article 416 (6) CRR).

Since these different forms of liquid assets have different characteristics, the interpretation of the

operational requirements will have to be made taking these special characteristics into account.

Special consideration will be given to the application to CIUs, where the operational requirements do

not only apply to the liquid assets contained in a CIU, but should also address the liquidity generating

capacity of a CIU itself, i.e. apply to the shares or units in the CIU.

According to the mandate of the report in Article 509(5) CRR, the report is to be made “in line with

international regulatory developments”. Reference is therefore made to “Basel III: The Liquidity

Coverage Ratio and the liquidity risk monitoring tools”, Jan 2013 BIS and the CEBS “Guideline on

Liquidity Buffers & Survival Periods”, 9 December 2009.

The Basel QIS exercises have highlighted that liquid assets “trapped” in legal entities and liquid assets

outside the control of the treasury or in trading positions accounted for most of the liquid assets

excluded from the buffer due to operational requirements.

PDF to HTML - Convert PDF files to HTML files

As shown by the Basel QIS exercises, it is important to ensure a harmonised understanding of

operational requirements in order to avoid differences in the strictness of application between banks.

To this end, this report is intended to contribute to a common understanding of the required

operational requirements and minimise the scope for deviations in interpretation.

For banks to determine whether any assets should be excluded from the liquidity buffer due to

operational requirements, they accordingly need to have the ability to assess these assets on an

ongoing basis.

The next sections of this report consider the relevant CRR operational requirements in sequence.

5.1. Availability

As per Article 417(b) CRR, the holdings of liquid assets need to be legally and practically available at

any time during the next 30 days to be liquidated via outright sale or via simple repurchase agreement

on approved repurchase markets in order to meet obligations coming due.

Rationale

“Legally and practically available”

The liquid assets that form part of the liquidity buffer under the LCR are intended to be effectively

monetised to meet liquidity outflows in the 30 days of the scope of the LCR. Therefore it is necessary

that an institution have unlimited access and power of disposal to meet this end.

An asset is

legally

available if it is:



unencumbered in the sense that there are no legal, regulatory or contractual impediments to sell or

lend the asset;

there are no transfer restrictions in the country where the asset is held;

the asset is denominated in a convertible currency.

An asset is

practically

available if:



there are no regulatory, tax or accounting impediments to effectively monetise the position in liquid

assets;

the legal entity in which the assets are being held has direct market access or the assets can be

freely and timely transferred to another legal entity that has market access;

PDF to HTML - Convert PDF files to HTML files



any excess liquid assets that are taken into account at consolidated level must be freely

transferable between legal entities. Excess liquid assets are those that are not needed at the single

entity level to cover LCR-minimum requirements;



the liquidity management function has a timely overview over the amount, composition and location

of liquid assets;

the liquidity management function has full control over the holdings of liquid assets. Procedures

and systems are installed allowing immediate monetization assets when required. (Further laid out

in Article 417(c)).



Monetisation must be executable in the standard settlement period for the asset class in the

relevant jurisdiction.

Liquid assets are not used in ongoing operations (further laid out in Article 417(e)).

As regards CIUs, the liquidity generating capacity of CIUs is highly dependent on the legal structure of

a CIU, the allowed form of liquidation and the time needed to actually generate liquid means.

Institutions and supervisors should ensure that, under normal and stressed conditions, the generation

of liquidity from CIUs:



falls under the sole decision of the institution;

is unrestricted; and

can be executed without delay.

The thorough assessment of these characteristics should be linked to the proportion of the liquidity

buffer held in the form of CIUs. If CIUs account for the major part of the liquidity buffer of an institution,

requirements on availability should be particularly strict, to ensure the usability of the buffer.

Liquid assets must be available at any time during the next 30 days. Concerning secured lending, the

CRR assumes prolongation of secured lending coming due within 30 days and collateralised by HQLA

(Article 422 CRR). Therefore assets received from reverse repos and other securities financing

transactions maturing within the next 30 days that are held within an institution and where the

institution has full rights of rehypothecation may be considered available during the next 30 days

(however an outflow has to be taken into account under Article 422).

Encumbrance

The availability of assets is linked to the concept of encumbrance.

PDF to HTML - Convert PDF files to HTML files

Article 416(3)(a) addresses the need of HQLA to be unencumbered explicitly: “They are

unencumbered or stand available within collateral pools to be used for the obtaining of additional

funding under committed but not yet funded credit lines available to the institution”.

The CRR does not provide a definition of encumbrance, but the EBA in its draft ITS on asset

encumbrance reporting under Article 100 CRR offers the following draft definition of encumbrance:

“For

the purpose of this Regulation an asset is considered encumbered if it has been pledged or if it is

subject to any form of arrangement to secure, collateralise or credit enhance any transaction from

which it cannot be freely withdrawn.”

However, for the purpose of liquidity provision it is important that the definition of encumbrance not be

based on solvency considerations, but on liquidity considerations (within a time horizon of 30 days).

The practical interpretation of the concept of encumbrance in the context of liquidity reporting should

therefore differ from the interpretation of encumbrance applied for the reporting on asset

encumbrance, which focuses solely on solvency.

The Basel definition of encumbrance/unencumbrance (Basel III: Liquidity coverage ratio and liquidity

monitoring tools, January 2013 – paragraph 31) captures the essential features of liquidity, by focusing

on the possible impediments to the provision of liquidity.

“Unencumbered”

means free of legal, regulatory, contractual or other restrictions on the ability of the

bank to liquidate, sell, transfer, or assign the asset”

Paragraph 31 of the Basel III text further states:

“An

asset in the stock should not be pledged (either explicitly or implicitly) to secure, collateralise or

credit-enhance any transaction, nor be designated to cover operational costs (such as rents and

salaries). Assets received in reverse repo and securities financing transactions that are held at the

bank, have not been rehypothecated, and are legally and contractually available for the bank's use

can be considered as part of the stock of HQLA. In addition, assets which qualify for the stock of

HQLA that have been pre-positioned or deposited with, or pledged to, the central bank or a public

sector entity (PSE) but have not been used to generate liquidity may be included in the stock.

(Footnote: If a bank has deposited, pre-positioned or pledged Level 1, Level 2 and other assets in a

collateral pool and no specific securities are assigned as collateral for any transactions, it may assume

that assets are encumbered in order of increasing liquidity value in the LCR, ie assets ineligible for the

stock of HQLA are assigned first, followed by Level 2B assets, then Level 2A and finally Level 1. This

determination must be made in compliance with any requirements, such as concentration or

diversification, of the central bank or PSE.)”

The Basel definition recognises that an asset may be bound for some specific purpose but still be a

source of liquidity. For instance, liquid assets received in a reverse repo would be considered

encumbered under a solvency perspective, but not in general under the Basel definition. If an

institution has full disposal of the assets within the remaining time of the reverse contract, these assets

PDF to HTML - Convert PDF files to HTML files

may be counted as part of the liquidity buffer. A further requirement would of course be that the

institution has the ability to sell or rehypothecate the asset and can do so in compliance with existing

regulatory requirements.

As another example, cash received from customers and earmarked to cover future claims are likely to

be considered as encumbered under a solvency perspective. However, if these means are invested in

liquid assets precisely with the aim of meeting such future claims, they would count as part of the

liquidity buffer under the Basel definition - again provided that an institution has the ability to sell or

rehypothecate the asset and can do so in compliance with existing regulatory requirements.

Transfer restrictions and non-convertible currencies

Article 417(b) furthermore states that “liquid assets referred to in point (c) of Article 416(1) which are

held in third countries where there are transfer restrictions or which are denominated in non-

convertible currencies shall be considered available only to the extent that they correspond to outflows

in the third country or currency in question, unless the institution can demonstrate to the competent

authorities that it has hedged the ensuing currency risk appropriately.

Transfer restrictions

This requirement refers to liquid assets under Article 416(1)(c)

held

in third countries (i.e. non-EU

countries) where there are transfer restrictions. In this case the asset should only be intended to cover

outflows in that third country.

Transfer restrictions are existing restrictions imposed under applicable laws, regulations and

supervisory requirements. Accordingly, it would be appropriate for institutions to have processes in

place to capture all liquidity transfer restrictions to the extent practicable, and to monitor the rules and

regulations in the jurisdictions in which a group operates and assess their liquidity implications for the

group as a whole.

In the case of Article 416(1)(c) the restrictions set up by the CRR are twofold: government bonds and

other liquid assets mentioned under letter (i)



should only be held to cover liquidity outflows in the country of their issuance and

should not held outside this country if there are transfer restrictions.

However, it is plausible that

any

integrant of the liquidity buffer, not just assets referred to in Article

416(1)(c),

held

in countries with transfer restrictions is not legally and practically available to be used

to cover outflows outside the respective country. Excess holdings of any liquid asset that is held in a

country with transfer restrictions cannot by definition be used to cover liquidity outflows in another

location. The requirement should therefore apply to any liquid asset coming under the scope of the

definition in Article 416.

PDF to HTML - Convert PDF files to HTML files

Non-convertible currencies

Liquid assets under Article 416(1)(c) which are denominated in non-convertible currencies will be

considered available only to the extent that they correspond to outflows in the currency in question,

unless an institution can demonstrate to the competent authorities that it has appropriately hedged the

ensuing currency risk.

If liquidated, the value of an asset denominated in a non-convertible currency cannot be converted

freely into a means of payment in another currency, since non-convertible currencies are not traded on

foreign exchange markets or only traded in limited amounts. Hedging the currency risk by means of a

non-deliverable forward (NDF) foreign exchange transaction eliminates market risk. However, the

nominal corresponding to the value of the asset is not converted into another currency. The asset is

therefore not suitable to cover liquidity outflows in other currencies.

As with the previous condition, it is plausible that this condition should apply to

any

liquid asset

denominated in a non-convertible currency and not be restricted to assets under Article 416(1)(c). The

use of the proceeds from HQLA in a non-convertible currency is restricted to covering outflows in that

respective currency.

Conclusion

The concept of availability is fundamental to operational requirements for liquid assets, since it

determines the practical usability of the liquidity generating capacity of liquid assets. It is therefore

important that any aspect of availability be considered under the point of view of liquidity, which, in

some cases might differ from availability considered from the point of view of solvency. This is

especially true when applying the concept of unencumbrance, where recent reporting requirements

take a solvency perspective and focus on the availability for unsecured debitors in the event of default,

instead of on short-term liquidity generating capacity.

We recommend that for the purpose of liquidity reporting, the notion of unencumbrance follow that of

paragraph 31 of Basel III: Liquidity Coverage Ratio and liquidity monitoring tools, January 2013 which

states that

“Unencumbered”

means free of legal, regulatory, contractual or other restrictions on the ability of the

bank to liquidate, sell, transfer, or assign the asset”.

We further recommend that

any

liquid asset or other integrant of the liquidity buffer

held

in a country

with transfer restrictions should only be used to cover outflows in that respective country. Any excess

holding of liquid assets should not be admissible for use to cover outflows in other countries. To

restrict the application of this requirement to assets under Article 416(1)(c) is not compatible with a

prudential approach.

PDF to HTML - Convert PDF files to HTML files

Likewise we recommend that any liquid asset denominated in a non-convertible currency should only

be used to cover outflows in that currency, regardless of whether the currency risk has effectively

been hedged or not.

5.2. Control by a liquidity management function

Article 417(c) CRR requires that holdings of HQLA be controlled by a liquidity management function.

Rationale

In order to ensure effective monetisation in case liquidity outflows have to be met through the

liquidation of liquid assets, these need to be under the control of the liquidity management function.

The function should have the continuous authority, and legal and operational capability, to monetise

any asset in the stock in case it is needed and the proceeds should be available to the function (Basel

2013, paragraph 33). The operational requirements should not impose a requirement for a

segregated, ring-fenced liquidity pool.

Control over the buffer of liquid assets can be demonstrated in 2 ways:



“either by maintaining liquid assets in a separate pool under the sole control of the liquidity

management function”;

or “by demonstrating that the function can monetize the assets at any point in the 30-day stress

period and that the proceeds of doing so are available throughout the 30-day stress period without

directly conflicting with a stated business or risk management strategy”. In this case, paragraphs

Furthermore, the liquidity management function could also be defined as the one identified by internal

methodologies described in Article 86(4) CRD, as in charge of the identification, measurement,

management and monitoring of funding positions. This would ensure that this is in line with the liquidity

management’s internal structures of each firm or liquidity subgroup.

Conclusion

We recommend that the requirement not explicitly demand liquid assets to be held by a specific

function (e.g. treasury) or that liquid assets be held in a separate book, but rather that institutions

should have policies and procedures in place that establish the proper authority of the liquidity

management function and ensure sufficient operational capacity.

As different liquidity management models co-exist in the EU, the interpretation of the criteria should

reflect and allow this diversity.

PDF to HTML - Convert PDF files to HTML files

5.3.Test sales option

Article 417(d) requires that a portion of the liquid assets except those referred to in points (a), (c) and

(e) of Article 416(1) be periodically and at least annually liquidated via outright sale or via simple

repurchase agreements on an approved repurchase market for the following purposes:

(i) to test the access to the market for these assets;

(ii) to test the effectiveness of its processes for the liquidation of assets;

(iii) to test the usability of the assets;

(iv) to minimise the risk of negative signalling during a period of stress.

Rationale

It is important that banks regularly test the effectiveness of their internal processes and functions to

ensure timely liquidation of assets. It is equally important to accustom markets to the fact that an

institution trades on the market so that there will be no immediate feedback relating any asset sales to

a perceived liquidity stress of the institution.

The CEBS guideline points out: “Firms should seek to be active on a regular basis in each market in

which they hold assets for liquidity purposes. Accessing the market regularly will help to reduce the

potential stigma of firms suddenly accessing markets, alerting other firms to the fact that they may be

under liquidity pressure (in turn, causing more investors to withdraw funds, thereby accentuating the

liquidity pressure).”

Assets under Article 416(1) (a), (c) and (e) are not subject to the test sales (or repos) under the CRR

provision. In the case of letters (a) and (e) the rationale is clearly intuitive, since both letters address

directly available central bank liquidity.

Literal application of the requirement to assets under Article 416 (1) (f) is not possible, since deposits

or available liquid funding from the central credit institution cannot be sold. However, following the

rationale of the requirement, the availability of this kind of liquidity should also be tested.

The rationale of excluding (c) from test sales could lie in the restricted scope of application of the

liquidity under letter (i). Assets qualifying under Article 416(1)(c) are eligible to the buffer

independently of their empirical liquidity because they may only be used under the LCR reporting

requirement “if the institution incurs a liquidity risk in that Member State or third country that it covers

by holding those liquid assets”. However there is no definition of “liquidity risk that an institution incurs

in a Member State or a third country” so it will be difficult to check that these assets are used only to

cover such risk exposure. There are also no rules on how assets that do conform both to points (b)

and (c) (or (e) and (c)) should be reported, meaning that imposing more precise requirements might

be extremely difficult to monitor, as the institutions can game the requirements simply by regularly

PDF to HTML - Convert PDF files to HTML files

changing the classifications of assets . It should in any case be part of a prudent approach of

institutions to closely monitor changes in the liquidity generating capacity of these assets.

Concerns have been raised concerning the costs of test sales. Since bonds reported under Article

416(1)(c) are exempted from test sales, banks might preferably want to classify them as such, even in

cases where they would also be eligible under Article 416 (1)(b). Furthermore, in the absence of a

definition and a reporting duty of “liquidity risk in that Member state or third country” adequate

supervision is not possible.

A further concern has been raised that banks might manipulate test sales (or test repos) to each other

to report liquid markets in cases where this liquidity might be in doubt.

Both concerns address the ability of banks to arbitrage test sales either by circumvention of the

requirement or by manipulation of the results. These concerns should be brought to the attention of

supervisors.

Conclusion

We recommend that supervisors apply a principles-based approach to test sales, considering the

normal trading activity of an institution and the costs of test sales. In line with the CEBS Guidelines on

Liquidity Buffers and Survival Periods (Annex 2), based on the proportionality principle, smaller banks

which access markets through another institution, will, in most cases, not have to be active in several

advanced money and capital markets.

In order to achieve regulatory transparency and preclude circumventing the operational requirements

by institutions a harmonised definition of “liquidity risk that an institution incurs in a Member State or a

third country” should be also prepared as well as rules governing the treatment of assets that fall

under two or more definitions of eligible assets.

5.4 Not used in ongoing operations

Article 417 (e) states that price risks associated with the assets may be hedged, but that the liquid

assets are subject to appropriate internal arrangements that ensure that they are readily available to

the

treasury

when needed and especially that they are not used in other ongoing operations,

including:

(i) hedging or other trading strategies;

(ii) providing credit enhancements in structured transactions;

(iii) covering operational costs.

The example of such rules might be treatment of bonds in held to maturity portfolio, where mark-to-market valuation is not

required, for accounting purposes.

PDF to HTML - Convert PDF files to HTML files

Rationale

Article 417(e) addresses the use of liquid assets in the context of hedging operations. Price risks

associated with the assets may be hedged, however institutions should take into account the cash

outflow that would arise if the hedge were to be closed out early or apply adequate haircuts.

As for the inclusion of liquid assets in hedging strategies, these may, at the sale of the asset, create

an open risk position for an institution. Hedging strategies may therefore lead to practical impediments

to the monetisation of the assets through outright sale. However, this is not the case for assets that

are monetised using a repo transaction, if the availability of liquid repo markets for the asset is

guaranteed. The same consideration is valid for the use of liquid assets in trading strategies.

The Basel framework explicitly allows for assets to be used in hedging strategies (paragraph 34 of

Basel III: Liquidity Coverage Ratio and liquidity monitoring tools, January 2013:

A bank is permitted to hedge the market risk associated with ownership of the stock of HQLA and still

include the assets in the stock. If it chooses to hedge the market risk, the bank should take into

account (in the market value applied to each asset) the cash outflow that would arise if the hedge

were to be closed out early (in the event of the asset being sold).

The same rationale would apply to liquid assets involved in trading strategies.

Liquid assets should not be used to provide credit enhancements in structured transactions and

should not be considered part of the liquidity buffer, if this means the assets cannot be freely

liquidated. There are two ways in which liquid assets can be used in credit enhancement techniques:



liquid assets can be directly used as overcollateralization in which case they are pledged and no

longer freely available to the institution for use outside the structure;

an institution incurs a payment obligation (e.g. a standby letter of credit) which takes the form of a

guarantee.

In the first case, liquid assets are not eligible because they are not available for liquidation. In the

second case, the institution incurs a contingent liquidity outflow (which might well materialise under the

LCR stress scenario) that it covers with the holding of liquid assets. In both cases, the liquid assets in

question are no longer freely available to the institution to cover other outflows.

Liquid assets should not be used to cover operational costs. Operational costs are not considered to

be an outflow relevant for the LCR stress scenario in Article 420 CRR. This is because operational

cost are not typically stress sensitive, may not in all cases lead to an liquidity outflow (e.g. in the case

of salaries paid on accounts held with the institution) or their amount is negligible with a view to the

overall liquidity risk of an institution.

PDF to HTML - Convert PDF files to HTML files

However, if an institution decides to make provisions for the liquidity outflows that may arise from its

operational costs in the form of holdings of liquid assets, these are no longer available to cover other

outflows considered under the LCR stress scenario.

Conclusion

We recommend allowing the use of liquid assets in hedging or trading strategies as long as they can

easily be monetised without conflicting with existing risk limits that control for market, credit and

counterparty risk.

To ensure consistency with Article 417(c), treasury should be interpreted as “liquidity management

function”.

5.5 Currency Consistency

Article 417 (f) requires the denomination of the liquid assets to be consistent with the distribution by

currency of liquidity outflows after the deduction of inflows.

Rationale

Since liquid assets need to be monetised to meet liquidity outflows in a specific currency, it is

mandatory that these liquid assets be able generate liquidity in the respective currency. This is

particularly true for non-convertible currencies. However, even with currencies that are regularly

traded, it should not be taken for granted that foreign exchange markets will remain accessible at any

time during the stressed conditions reflected in the LCR. The reliance on foreign exchange markets

during stressed conditions may, furthermore, expose an institution to significant market risk, which

could sharply widen existing mismatched positions.

Institutions should therefore monitor the currency mismatch they incur in different currencies.

Instruments to this end are the LCR and NSFR reporting by significant currency, as well as the

reporting of the additional monitoring metrics by significant currency. Institutions should test the

effectiveness of existing currency risk hedging strategies under stress.

Conclusion

A perfect match of currency consistency of liquid assets and net outflows is not required according to

the general liquidity requirement under

article 412.1

(or Basel III). Until a specific currency requirement

is specified in detail, institutions should monitor risk, test the effectiveness of their hedging strategies

and develop contingency plans for stressed conditions.

PDF to HTML - Convert PDF files to HTML files

Annex 1: Relevant CRR Articles on uniform definitions of liquid assets

Recital 100

Institutions should hold a diversified buffer of liquid assets that they can use to cover liquidity needs in

a short term liquidity stress. As it is not possible to know ex ante with certainty which specific assets

within each asset class might be subject to shocks ex post, it is appropriate to promote a diversified

and high quality liquidity buffer consisting of different asset categories. A concentration of assets and

overreliance on market liquidity creates systemic risk to the financial sector and should be avoided. A

broad set of quality assets should therefore be taken into consideration during an initial observation

period which will be used for the development of a definition of a liquidity coverage requirement. When

making a uniform definition of liquid assets at least government bonds, and covered bonds traded on

transparent markets with an ongoing turnover would be expected to be considered assets of extremely

high liquidity and credit quality. It would also be appropriate that assets corresponding to Article 416(1)

(a) to (c) should be included in the buffer without limitations. When institutions use the liquidity stock,

they should put in place a plan to restore their holdings of liquid assets and competent authorities

should ensure the adequacy of the plan and its implementation.

Article 416: Reporting on liquid assets

Institutions shall report the following as liquid assets unless excluded by paragraph 2 and only

if the liquid assets fulfil the conditions in paragraph 3:

(a)

cash and exposures to central banks to the extent that these exposures can be withdrawn at

any time in times of stress. As regards deposits held with central banks, the competent authority and

the central bank shall aim at reaching a common understanding regarding the extent to which

minimum reserves can be withdrawn in times of stress;

(b)

(c)

(i)

other transferable assets that are of extremely high liquidity and credit quality;

transferable assets representing claims on or guaranteed by:

the central government of a Member State, a region with fiscal autonomy to raise and collect

taxes, or of a third country in the domestic currency of the central or regional government, if the

institution incurs a liquidity risk in that Member State or third country that it covers by holding those

liquid assets;

(ii)

central banks and non-central government public sector entities in the domestic currency of

the central bank and the public sector entity;

(iii)

the Bank for International Settlements, the International Monetary Fund, the Commission and

multilateral development banks;

PDF to HTML - Convert PDF files to HTML files

(iv)

(d)

(e)

the European Financial Stability Facility and the European Stability Mechanism;

transferable assets that are of high liquidity and credit quality;

standby credit facilities granted by central banks within the scope of monetary policy to the

extent that these facilities are not collateralised by liquid assets and excluding emergency liquidity

assistance;

(f)

if the credit institution belongs to a network in accordance with legal or statutory provisions,

the legal or statutory minimum deposits with the central credit institution and other statutory or

contractually available liquid funding from the central credit institution or institutions that are members

of the network referred to in Article 113(7), or eligible for the waiver provided in Article 10, to the extent

that this funding is not collateralised by liquid assets.

Pending specification of a uniform definition in accordance with Article 460 of high and extremely high

liquidity and credit quality, institutions shall identify themselves in a given currency transferable assets

that are respectively of high or extremely high liquidity and credit quality. Pending specification of a

uniform definition, competent authorities may, taking into account the criteria listed in Article 509(3),

(4) and (5) provide general guidance that institutions shall follow in identifying assets of high and

extremely high liquidity and credit quality. In the absence of such guidance, institutions shall use

transparent and objective criteria to this end, including some or all of the criteria listed in Article 509(3),

(4) and (5).

(a)

(i)

The following shall not be considered liquid assets:

assets that are issued by a credit institution unless they fulfil one of the following conditions:

they are bonds eligible for the treatment set out in Article 129(4) or (5) or asset backed

instruments if demonstrated to be of the highest credit quality as established by the EBA pursuant to

the criteria in Article 509(3), (4) and (5);

(ii)

they are bonds as referred to in Article 52(4) of Directive 2009/65/EC other than those referred

to in point (i) of this point;

(iii)

the credit institution has been set up by a Member State central or regional government and

that government has an obligation to protect the economic basis of the institution and maintain its

viability throughout its lifetime; or the asset is explicitly guaranteed by that government; or at

least 90 % of the loans granted by the institution are directly or indirectly guaranteed by that

government and the asset is predominantly used to fund promotional loans granted on a

non-competitive, not for profit basis in order to promote that government's public policy objectives;

(b)

assets that are provided as collateral to the institution under reverse repo and securities

financing transactions and that are held by the institution only as a credit risk mitigant and that are not

legally and contractually available for use by the institution;

PDF to HTML - Convert PDF files to HTML files

(c)

(i)

(ii)

(iii)

(iv)

(v)

assets issued by any of the following:

an investment firm;

an insurance undertaking;

a financial holding company;

a mixed financial holding company;

any other entity that performs one or more of the activities listed in Annex I to

Directive 2013/…/EU as its main business.

In accordance with paragraph 1, institutions shall report assets that fulfil the following

conditions as liquid assets:

(a)

they are unencumbered or stand available within collateral pools to be used for the obtaining

of additional funding under committed but not yet funded credit lines available to the institution;

(b)

they are not issued by the institution itself or its parent or subsidiary institutions or another

subsidiary of its parent institutions or parent financial holding company;

(c)

their price is generally agreed upon by markets participants and can easily be observed in the

market, or their price can be determined by a formula that is easy to calculate based on publicly

available inputs and does not depend on strong assumptions as is typically the case for structured or

exotic products;

(d)

they are eligible collateral for standard liquidity operations of a central bank in a Member State

or if the liquid assets are held to meet liquidity outflows in the currency of a third country, of the central

bank of that third country;

(e)

they are listed on a recognised exchange or they are tradable on active outright sale or via a

simple repurchase agreement on approved repurchase markets. These criteria shall be assessed

separately for each market.

The conditions referred to in points (c), (d) and (e) of the first subparagraph shall not apply to the

assets referred to in point (e) of paragraph 1.

The condition referred to in point (d) of the first subparagraph shall not apply in the case of liquid

assets held to meet liquidity outflows in a currency in which there is an extremely narrow definition of

central bank eligibility. In the case of liquid assets denominated in currencies of third countries, this



OJ: Please insert serial number of the Directive contained in document PE-CONS 15/13 (2011/0203 (COD)).

PDF to HTML - Convert PDF files to HTML files

exception shall apply and only apply if the competent authorities of the third country apply the same or

an equivalent exception.

Notwithstanding the provisions of paragraphs 1, 2 and 3, pending the specification of a binding

liquidity requirement in accordance with Article 460 and in accordance with the second subparagraph

of paragraph 1 of this Article, institutions shall report on:

(a)

and (5);

(b)

other central bank eligible and tradable assets such as asset backed instruments of the

other non-central bank eligible but tradable assets such as equities and gold based on

transparent and objective criteria, including some or all of the criteria listed in Article 509(3), (4)

highest credit quality as established by EBA pursuant to the criteria in Article 509(3), (4) and (5);

(c)

other central bank eligible but non-tradable assets such as credit claims as established by the

EBA pursuant to the criteria in Article 509(3), (4) and (5).

The EBA shall develop draft implementing technical standards listing the currencies which

meet the conditions referred to in the third subparagraph of paragraph 3.

EBA shall submit those draft implementing technical standards to the Commission by 31 March 2014.

Power is conferred on the Commission to adopt the implementing technical standards referred to in

the first subparagraph in accordance with Article 15 of Regulation (EU) No 1093/2010.

Before the entry into force of the technical standards referred to in the third subparagraph, institutions

may continue to apply the treatment set out in the second subparagraph of paragraph 3, where the

competent authorities have applied that treatment before ….

Shares or units in CIUs may be treated as liquid assets up to an absolute amount of

EUR 500 million in the portfolio of liquid assets of each institution provided that the requirements in

Article 132(3) are met and that the CIU, apart from derivatives to mitigate interest rate or credit or

currency risk, only invests in liquid assets as referred to in paragraph 1 of this Article.

The use or potential use by a CIU of derivative instruments to hedge risks of permitted investments

shall not prevent that CIU from being eligible. Where the value of the shares or units of the CIU is not

regularly marked to market by the third parties referred to in points (a) and (b) of Article 418(4) and the

competent authority is not satisfied that an institution has developed robust methodologies and

processes for such valuation as referred to in the first sentence of Article 418(4), shares or units in that

CIU shall not be treated as liquid assets.



OJ: Please insert the date of application of this Regulation.

PDF to HTML - Convert PDF files to HTML files

Where a liquid asset ceases to be eligible in the stock of liquid assets, an institution may

nevertheless continue to consider it a liquid asset for an additional period of 30 calendar days. Where

a liquid asset in a CIU ceases to be eligible for the treatment set out in paragraph 6, the shares or

units in the CIU may nevertheless be considered a liquid asset for an additional period of 30 days

provided that those assets do not exceed 10% of the CIU's overall assets.

Article 509: Liquidity requirements

(...)

The EBA shall, after consulting ESMA and the ECB, by 31 December 2013, report to the

Commission on appropriate uniform definitions of high and of extremely high liquidity and credit quality

of transferable assets for the purposes of Article 416 and appropriate haircuts for assets that would

qualify as liquid assets for the purposes of Article 416, with the exception of assets referred to in

points (a), (b) and (c) of Article 416(1).

The European Parliament and the Council shall be given the opportunity to state their views on that

report.

The report referred to in the first subparagraph shall also consider:

(a)

other categories of assets, in particular residential mortgage-backed securities of high liquidity

and credit quality;

(b)

other categories of central bank eligible securities or loans, for example local government

bonds and commercial paper; and

(c)

other non-central bank eligible but tradable assets, for example equities listed on a recognised

exchange, gold, major index linked equity instruments, guaranteed bonds, covered bonds, corporate

bonds and funds based on those assets.

The report referred to in paragraph 3 shall consider whether, and if, to what extent standby

credit facilities referred to in point (e) of Article 416(1) should be included as liquid assets in light of

international development and taking into account European specificities, including the way monetary

policy is performed in the Union.

The EBA shall in particular test the adequacy of the following criteria and the appropriate levels for

such definitions:

(a)

(b)

(c)

minimum trade volume of the assets;

minimum outstanding volume of the assets;

transparent pricing and post-trade information;

100

PDF to HTML - Convert PDF files to HTML files

(d)

(e)

(f)

(g)

(h)

(i)

credit quality steps referred to in Part Three, Title II, Chapter 2;

proven record of price stability;

average volume traded and average trade size;

maximum bid/ask spread;

remaining time to maturity;

minimum turnover ratio.

(...)

101

PDF to HTML - Convert PDF files to HTML files

Annex 2: Literature on market liquidity

This section provides a definition of liquidity and summarises the literature on liquidity measures and

methods to guide and provide grounds for the liquidity measures intended to be adopted in the EBA

methodology. Due to the breadth of this literature, the overview needs to be limited and focuses

mainly on empirical methods and metrics that are of direct relevance to the project.

The first section below provides a definition of market liquidity, the second section surveys the most

relevant literature on liquidity and liquidity measurement. The last section provides a summary of

empirical results regarding actual liquidity estimates, and measures applied, for different asset

classes.

Defining liquidity

When attempting to measure and analyse liquidity, one must confront a basic question: how should

liquidity be defined? At the general level, it is a simple concept. However, the exact meaning of

liquidity is far from apparent, and how one defines it has implications for how one goes about

measuring it. Liquidity also means different things to different traders, depending on their

characteristics and typical trading needs. An early general definition of liquidity can be found in

Keynes (1930), who considers an asset as more liquid

‘if it is more certainly realisable at short notice

without loss’.

As defined by O’Hara (1995), ‘a

liquid market is one in which buyers and sellers can

trade into and out of positions quickly without having large price effects’.

Implicit in these definitions is

the notion that a liquid market has the ability to absorb large liquidity demands without generating

excess volatility. The absorption capacity of markets is of particular importance in the case of LCR, as

a funding shock to banks would potentially trigger a joint liquidation of LCR assets by the banking

sector. Thus, the EBA will focus on the absorption capacity of markets and measure liquidity

accordingly.

To provide this absorption capacity, markets rely on some form of liquidity suppliers that are willing to

take the opposite side in transactions initiated by traders that demand liquidity. Who these liquidity

suppliers are differs across market structures. For example, quote-driven markets (dealer markets)

rely on intermediaries such as designated market makers, specialists or dealers that continuously post

bid and ask quotes at which they are committed to buy and sell a specified quantity of an asset. In

pure limit order markets, on the other hand, there are no formal intermediaries, and liquidity is

provided through limit orders submitted by the traders themselves (the limit order book). Hybrid

markets (e.g. the New York Stock Exchange) rely on liquidity provision by both designated market

makers (former specialists) and limit order traders, while non-continuous markets (e.g. call markets)

concentrate all trading interest at discrete points in time. The EBA will be attentive to the details of the

market structure.

Traders that demand liquidity are typically impatient and willing to pay a cost to liquidate or build a

position quickly, while patient traders are willing to supply liquidity to the more impatient traders. Most

liquidity measures attempt to measure the cost associated with demanding liquidity, or analogously

the compensation required to supply liquidity. As such, variables that measure for example the

102

PDF to HTML - Convert PDF files to HTML files

frequency of trading activity may not in isolation be appropriate liquidity measures as e.g. a high

frequency of trading activity does not necessarily map into low implicit costs associated with executing

these trades. To obtain a direct measure of liquidity one would need trading activity to be translated

into a measure of costs or examined in conjunction with other liquidity measures. This point has been

made by e.g. Aitkin and Commerton-Forde (2003), and more recently by Vayanos and Wang (2012).

The above description does not provide much guidance with respect to how exactly liquidity should be

measured. For this purpose, Harris (1990) attempts to provide a more operational definition of market

liquidity and proposes that liquidity can be defined through four interrelated dimensions:

width, depth,

immediacy and resiliency.



width

reflects the cost of demanding liquidity (as opposed to supplying it) and is typically captured

by the size of the bid/ask spread. While the spread measures the cost incurred by consuming

liquidity immediately (‘crossing the spread’), it does not capture the quantity that can be

transacted at the best quotes.



Depth

refers to the quantity of liquidity supplied, and is typically measured by the volume offered

at the bid and ask quotes.



Immediacy

refers to how quickly a large trading need can be accomplished. In most continuous

markets small orders are typically executed immediately. However, large orders (e.g. by

institutional traders) may take time to execute without incurring large costs. In an illiquid market, a

large trader may incur significant delay costs if prices move during the time it takes to build the

position. This cost is commonly referred to as implementation shortfall (see e.g. Perold, 1988).

Immediacy may also reflect search costs, which can be significant in e.g. OTC markets (see e.g.

Duffie et al, 2007).



Resiliency

refers to the time it takes for a price to return to its pre-trade equilibrium level after a

large (uninformed) order consumes liquidity. While resiliency was initially defined with respect to

prices, resiliency can also be defined as the speed at which liquidity (quotes and depth)

replenishes to equilibrium levels after a large liquidity shock.

An important insight from the above discussion is that several measures are typically needed to obtain

an accurate picture of an asset’s market liquidity. For example, a tight bid/ask spread is not enough to

define an asset as extremely liquid unless a large quantity can be transacted at the best quotes

relatively quickly. For example, large traders that split up their orders into smaller chunks to minimise

transaction costs, resiliency and immediacy are also very important. In other words, while the

dimensions proposed by Harris (1990) are closely related, individual liquidity metrics are typically only

able to capture some facets of liquidity. Hence a combination of liquidity measures that capture

different aspects is required to get a complete picture. The EBA plans to use a variety of measures in

its analysis.

103

PDF to HTML - Convert PDF files to HTML files

Measures of market liquidity

When structuring the vast literature on market liquidity, the liquidity dimensions proposed by Harris

(1990) are useful as a general framework to think about different liquidity measures. However, since

many measures and methods aim at capturing several aspects of liquidity simultaneously, the EBA will

mainly use his definitions to interpret the measures proposed in the literature.

There are several recent papers that provide a comprehensive literature overview. For example a

recent paper by Vayanos and Wang (2012) gives a detailed overview of both the theoretical and

empirical literature on market liquidity. In this overview, however, the focus is mainly on the empirical

literature with particular focus on liquidity measures that are applicable to this project. It is also

important that the measures be relatively simple and transparent as they will be applied to a broad

range of assets and markets. However, the accuracy and (proven) quality of a measure should not be

sacrificed for simplicity.

Another important aspect is the distinction between liquidity measures that rely on high frequency

transaction and/or order level data versus measures that rely on low frequency data (e.g. daily data).

While the EBA intends to use detailed transaction data (MiFID data) in the final report, it will also cover

measures that are tailored to be estimated using daily data, since such measures will be useful for

evaluating the liquidity of asset classes for which the EBA does not intend to gather transaction level

data. Finally, it should be noted that the major part of research on liquidity measurement historically

has focused on equity markets due to the superior data availability. However, more recent studies on

the liquidity of other asset classes (e.g. corporate bonds) typically apply the same measures initially

developed for equity markets. In instances where there are concerns regarding the applicability of a

measure to a specific asset class or market structure, those will be explicitly noted.

Spread measures

The most widely applied proxy for liquidity is the bid/ask spread. As noted above, spread measures

relate to Harris’ (1990) width dimension of liquidity. A large part of the theoretical market

microstructure literature focuses on understanding why a positive bid/ask spread appears in

equilibrium and which factors give rise to cross-sectional differences in spreads. Overall, these studies

have established that a spread consists of three main cost components; inventory costs, order

processing costs and adverse selection costs (see e.g. O’Hara, 1995). In dealer markets, these costs

are incurred by the market makers who typically set a zero-profit (competitive) spread that exactly

covers these costs. An important part of both the theoretical and empirical literature focuses on the

adverse selection cost component of the spread, which is defined by the market-makers’ expected

loss to privately informed traders (see Glosten and Milgrom, 1985).

In the empirical literature several spread measures are commonly applied that capture different

aspects of transaction costs. The most basic spread measure is the quoted spread (

simply measured as the difference between the highest bid quote

associated with transaction (or quote update) ,

), which is

and the lowest ask quote

. While this is a direct measure of

the roundtrip cost associated with trading a

small

quantity of an asset, it is commonly expressed

104

PDF to HTML - Convert PDF files to HTML files

relative to the bid/ask midpoint price such that the proportional spread (

, where

) can be expressed as

Since trades in dealer- or OTC markets can occur inside the prevailing best quotes, the effective

spread might provide a more accurate measure of the spread. The effective spread (

transaction

is typically measured as,

where

) for

is a signing variable that takes

the value 1 if the trade is initiated by a buyer and -1 if the trade is initiated by a seller. In many

transaction datasets, however, there is no indicator variable (

explicitly stating whether a trade was

occurs above (buyer

initiated by a seller or a buyer. In those cases, the standard method used in the literature is the Lee

and Ready (1991) algorithm. This algorithm simply checks whether transaction

initiated) or below (seller initiated) the bid/ask midpoint prevailing just before the trade occurs. In the

cases where the trade happens exactly at the prevailing midpoint, the standard procedure is to check

the transaction price against the midpoint for longer lags. Ellis et al. (2000) examine the accuracy of

the Lee and Ready (1991) algorithm and show that it is able to sign about 76% of transactions

correctly.

An issue not discussed above is that intraday measures of the spread are also commonly volume

weighted or time weighted. By weighting the individual spread observations by the total volume traded,

one would obtain a spread measure that reflects more accurately the actual costs realised by traders.

The time weighted spread is typically calculated by weighting each spread observation by its intraday

duration, which gives a more representative spread measure if there are long periods with the same

spread. Calculating a time weighted spread, however, typically requires order-level data.

In cases when one do not directly observe the bid and ask quotes the literature suggests several

implicit spread estimators. The best known estimator is proposed by Roll (1984) that exploits the fact

that trades typically occur at, or within, the bid and ask quotes. As such it is closely related to the

effective spread measure discussed earlier. The main idea is that transaction prices tend to ‘bounce’

between the (unobserved) bid quote and ask quote as buyer and seller initiated trades execute

against these quotes. This generates a negative autocorrelation between consecutive transaction

prices (and returns), which can be used to map out the size of the spread. The Roll (1984) implicit

spread estimator for a time window is simply estimated as,

There is also a growing part of the literature that examines low frequency estimators of the spread that

only require daily data. These studies are very useful as they offer methods that make it possible to

construct long liquidity time series for large samples of individual assets. There are several recent

measures that are shown to be very good when evaluated against actual high frequency spread

measures. One early estimator is the LOT measure proposed by Lesmond, Ogden, and Trzcinka

(1999) that exploits the information inherent in zero-return days to estimate the effective transaction

costs for any asset. The basic idea is that assets with large spreads require larger information value

for new information to be traded into the price. Hence, if the value of new information is less than the

costs of trading, prices will not move. They show that their measure is highly correlated with spread

measures calculated using actual quote observations. While the LOT measure requires relatively little

data, it needs to be estimated by maximum likelihood, and more importantly requires the return on a

105

, where

PDF to HTML - Convert PDF files to HTML files

market index. While the requirement of having the return on a market index is unproblematic for

equities, for other asset classes and markets the availability of high quality benchmarks are not readily

available.

To solve this issue, Fong, Holden and Trzcinka (2011) propose a simplified and less data intensive

version of the LOT measure, which they coin FHT. They evaluate the power of this measure against

other low frequency and high frequency proxies estimated for more than 18 000 stocks listed on 43

different exchanges around the world, covering a period of more than ten years. The measure is

shown to strongly dominate prior percentage cost proxies, and is also highly correlated with various

price impact measures. The measure is an analytic measure that requires only daily return

observations (i.e. not a market index) and the fraction of zero return days relative to total trading days

for the asset. The measure can be calculated for period

as a fraction of total trading days during the period; and

normal distribution.

Depth and price impact measures

As the spread measures typically measure the implicit cost of trading a small quantity at the best

prevailing quotes, Harris (1990) proposes a second important aspect of liquidity;

depth.

In recent years

the availability of order-level data from limit order markets has made it possible to measure depth

more accurately. However, direct depth measures typically involve massive amounts of data,

especially due to the tremendous growth in high frequency trading activity recent years. The EBA will

focus on metrics applied in the literature that are tailored to provide implicit measures of depth.

A key element in asymmetric information models is that trades convey information, and the quicker

prices reflect new information (private and public) the more efficient are prices. Typically, the speed of

price discovery, and hence price efficiency, is closely linked to liquidity. As discussed above,

Lesmond, Ogden, and Trzcinka (1999) suggest that assets that have higher transaction costs also

have less informationally efficient prices. An important distinction is between informed and uninformed

trades, and is a key to measuring the price impact of trades. Informed trades should adjust prices

permanently

to new equilibrium levels while uninformed trades caused by e.g. traders receiving

idiosyncratic shocks that trigger trades should generate temporary price impacts where prices quickly

revert to the pre-trade equilibrium price. This distinction forms the basis for several empirical models of

price discovery that aim at quantifying price impacts. One important contribution in this regard is

Hasbrouck (1991) who proposes that the interactions of trades and quote revisions be modelled as a

vector autoregressive system. Without going into the details of his model, the main results can be

summarized as follows; (i) a trade’s full price impact is not instantaneous; (ii) the impact is a positive

and concave function of the trade size; (iii) large trades widens the spread and produce larger price

impacts; and finally (iv) information asymmetries are more significant for smaller firms. In a more

recent paper Hasbrouck (2009) examines low frequency proxies of effective cost estimators for the US

equity markets, where he proposes an intraday price impact measure. He proposes that the

representative price impact (

of trading in a stock can be estimated as,

where

is the aggregate signed square-root dollar volume during window , measured as

106

as,

where

the standard deviation of the asset’s daily returns over the period;

is the number of zero return days

is the inverse function of the cumulative

PDF to HTML - Convert PDF files to HTML files

where

is trading volume (in currency) executed during window . The length of the aggregation

window, , used in Hasbrouck (2009) is 5 minutes.

There are also several low frequency price impact measures that are typically applied when

researchers use daily data. One widely applied measure is the Amihud (2002) illiquidity ratio (ILR).

While ILR can be applied to intraday data, it is typically considered a low frequency measure of

illiquidity. The ILR is simply calculated as

the asset over time window ,

number of windows

factor of

where

is the absolute return of

just reflect the

is the currency volume traded over window , and

(which could be number of hours or days). The ILR is typically scaled up by a

for practical purposes. Intuitively, the ILR capture how much the price of an asset moves

(in either direction) per currency unit of trade. On average, the ILR measure reflect how sensitive the

price is to trade volume, and assets with a high ILR are less liquid (have greater price impact) than

assets with a low ILR. One issue with the ILR is that as the denominator gets close to zero, the ILR

measure goes to infinity. Hence, especially when applying this measure to less traded stocks, or at the

intraday frequency, one may experience large outliers.

Price and liquidity reversal

Another dimension that Harris (1990) proposes as an important characteristic of liquid assets is how

quickly the price reverts back to the equilibrium price after large liquidity trades. As discussed earlier,

this may also be defined as how quickly liquidity supply replenishes after a large liquidity demand has

been filled. To measure the resiliency of an asset is relatively challenging, and the literature is

relatively scarce. The papers that attempt to estimate the resiliency are typically focusing on electronic

limit order markets. Degryse et al. (2005) studies the resiliency of stocks traded on the Paris Bourse

by examining the price impact of aggressive order flow (marketable limit orders that demand liquidity)

and how prices recover after aggressive trades. They find that the Paris Bourse stocks recover

quickly, within just a few quote updates. In another paper Large (2007) measures resiliency as how

long it takes for liquidity to replenish after aggressive trades. Moreover, he studies the time it takes for

the limit order book to return to its normal shape after large trades. As this estimator requires detailed

order level data, the EBA does not go into the details as it will not be applicable for this paper.

However, other measures are available. One example is Pastor and Stambaugh (2003), who propose

that stock market liquidity can be measured by regressing next period returns for an asset on the

signed current period return multiplied by the currency volume of trades in the asset. The authors

obtain a monthly measure (

, where

traded in asset

on day

from the regression,

on day

in month ;

is the currency volume

is the excess return on asset

in month . While their measure is essentially a measure of price impact, it

has a flavour of resiliency as it instead of measuring the direct price impact of trades it captures how

much the price reverts to get back to its equilibrium level. It should be noted however that the authors

themselves warn that their measure seems too noisy to be useful at the individual asset level due to

large sampling errors, and is most applicable to measure market-wide liquidity for an asset class.

107

PDF to HTML - Convert PDF files to HTML files

Other methods

There is also a part of the literature on market liquidity measurement that proposes that one looks at

common (latent) factors across different measures of liquidity. As such, these methods attempt to

reduce the universe of liquidity metrics across several liquidity dimensions to a few metrics.

Several papers study whether liquidity risk is systematic and whether that systematic risk is priced in

equity markets. Among these are studies by Chordia, Roll, and Subrahmanyam (2000), and

Hasbrouck and Seppi (2001), which demonstrate that liquidity has a common systematic factor. One

important contribution to this literature is by Korajczyk and Sadka (2007), who propose using principal

component analysis to extract common liquidity factors from a large set of different liquidity measures.

In their paper they extract three common factors from a set of liquidity measures containing ILR,

turnover, quoted spread (QS) effective spread (ES), four price impact measures and a measure of

order imbalance measured as the ratio of the net sum of signed trading volume over the month scaled

by the number of shares outstanding. They show that shocks to assets' liquidity have a common

component across measures which account for most of the explained variation in individual liquidity

measures. The main purpose of their analysis is to examine whether liquidity is a priced risk factor in

stock market returns, and find that the across-measure systematic liquidity factor is priced. A recent

application of principal component methods to the US corporate bond market is in Dick-Nielsen,

Feldhutter and Lando (2012), which is discussed below.

108

PDF to HTML - Convert PDF files to HTML files

Annex 3: Literature on the liquidity of different asset classes

The discussion in Annex 2 mapped out some common liquidity measures applied in the literature.

While several of these measures were initially tailored to and assessed for equity markets, they have

also been applied to other asset classes. It is very useful to obtain an overview of which measures

have been applied successfully to various asset classes. It will also be useful to secure an overview of

the typical liquidity estimates obtained for different asset classes, both as a benchmark against which

the results obtained can be compared, and to examine how different asset classes have been ranked

relative to each other for other markets and other periods. Note that there are several limitations in the

literature, largely due to data availability. While studies for equity markets exist for a wide range of

markets, a large part of the literature on corporate and government bonds, as well as for other asset

classes, concentrates on US data.

Corporate bonds

Historically, corporate bonds have largely been traded over the counter, and data availability has been

relatively scarce. There are, however, several studies on the liquidity of the corporate bond market in

the US that use different low frequency data sources. In a recent paper, Chen et al. (2007), examine

various liquidity measures for a sample of more than 4000 US corporate bonds spanning a wide range

of bonds from investment grade to speculative categories. Their main data sources are Bloomberg

and Datastream, which they use to calculate several liquidity measures for individual bonds, and

groups of bonds. The first measure they calculate is the proportional bid/ask spread (using Bloomberg

data). Due to the low frequency (quarterly) nature of their data, they also apply the spread proxies

discussed earlier in this section. The two additional measures they calculate are the percentage zero-

return days measure (Zeros), and the Lesmond et al. (1999) LOT. Overall, they show that their

estimates of liquidity costs increase as credit rating worsens. With respect to proportional spreads,

they report average spreads of 0.25% for AAA, to 0.31% for BBB and 0.77% for CCC to D for bonds

with maturities between 1 and 7 years. For the LOT and Zeros measure, they estimate a significantly

larger difference in liquidity costs between investment grade and speculative grade bonds.

In the US, the Financial Industry Regulatory Authority’s (FINRA) TRACE is responsible for operating

the reporting and dissemination facility for over-the-counter corporate bond trades. The data starts in

2002, and the reporting requirements were gradually increased until 2005 when about 99% of all

public transactions are captured. This dataset has been applied in several more recent studies

examining the liquidity of the US corporate bond market.

One important paper that exploits the TRACE transactions data is Bao et al. (2011) who stud the US

corporate bond market for the period 2003 through 2007. They show that the illiquidity of US corporate

bonds is substantial and significantly greater than can be explained by bid/ask spreads. Furthermore,

they establish a strong link between bond illiquidity and bond prices and show that changes in market

wide illiquidity can very often explain variation in the yield spreads of high-rated bonds.

The paper also shows that the liquidity risk component overshadows the credit risk component. The

illiquidity measure they apply is calculated as;

where

, and prices

109

PDF to HTML - Convert PDF files to HTML files

are in natural logarithms. While their measure looks very similar to the Roll (1984) measure, the

economic intuition is different. While the Roll (1984) estimator exploits the negative autocorrelation in

trade prices caused by the bid ask bounce, the measure by Bao et al (2011) is motivated by the fact

that the transitory impact of liquidity should be uncorrelated with the fundamental value of the asset.

Hence, their measure is closely related to the Pastor and Stambaugh (2003) measure discussed

earlier, as it measures the illiquidity as the size of price reversals. More illiquid stocks are expected to

have a stronger

negative

covariance of consecutive price changes (greater reversals), which would

imply a larger positive

given their specification of

above. In terms of magnitude, their estimate of

has a cross-sectional average of 0.58 using the full time-series sample.

To compare their price reversal measure against other metrics, Bao et al. (2011) also calculate the

Roll (1984) measure for their sample bonds. Furthermore, they also compare their results with another

paper by Edwards, Harris, and Piwowar (2007) who uses a more detailed version of the TRACE

dataset which also includes information on the side on which a dealer participated. This makes them

able to determine whether a trade was initiated by a buyer or seller, such that they can directly

measure the effective spread. Bao et al. (2011) compare their Roll measure with the effective spreads

in Edwards, Harris, and Piwowar (2007) for various trade sizes. Interestingly, in both studies the

implicit and effective spreads decline with trade size. E.g. for trade sizes in the range of $7.5k to $15k,

the average implicit spread is 1.98% and the effective spread is 1.42%, while for trade sizes above

$750k, the implicit spread is 0.52% and the effective spread is 0.18%.

The most extensive study of the liquidity of the European corporate bond market is a commissioned

report produced by Biais et al. (2006) for the City of London on the transparency, liquidity and

efficiency of the European corporate bond market. This report surveys the empirical literature on the

liquidity of corporate bonds and also provides empirical evidence for the European market using

International Index Company (IIC) data containing daily end-of-day bid and ask quotes in the iBoxx

index. They examine euro- and sterling-denominated corporate bonds for the period 2003-2005, and

the report contains a wide range of statistics across various characteristics. While it is not clear how

representative their results are for today’s market, they find that, after controlling for sample period

and credit risk, the effective spreads in euro-denominated bonds are lower than their US counterparts.

Moreover, they find that the effective half spread for trades above one million Euros is 0.049%. This is

significantly lower than the results in Goldstein et al. (2005) for the US market, who find that the half

spread is 0.22% for transactions above one million USD. Interestingly, the spread estimates for the US

corporate bond market are similar to the results in Chen et al. (2007) discussed earlier. Biais et al.

(2006) suggest some explanations for the difference between the US and European bond markets.

Since the advent of the euro, the European bond market has become integrated with investors from all

European countries trading in the same market. This large pool of buyers attracts sell-side

intermediation. As such, they argue that their results suggest that the supply of liquidity in the euro-

denominated bond market is rather competitive, which drives spreads down.

Since this study is about identifying which assets have the ability to retain their liquidity during stressful

periods, a recent paper by Dick-Nielsen, Feldhutter and Lando (2012) is particularly enlightening. They

perform an extensive study of the corporate bond market liquidity in the US for the period 2005 to

2009. For this purpose, they calculate several liquidity measures including an Amihud (2002) ILR,

110

PDF to HTML - Convert PDF files to HTML files

zero-trading days, and the Roll (1984) implicit spread measure among other things. They also extract

principal components from a set of eight different liquidity and activity measures. Their overall results

suggest that bond market liquidity worsened dramatically during the subprime crisis. They also find

that their common liquidity factor derived from a principal component analysis is superior to earlier

measures when it comes to explaining variations in yield spreads.

Houweling, Mentink and Vorst (2005) take a different approach by studying bond- and issuer-specific

characteristics that are related to liquidity. Motivated by the fact that for corporate bonds, most

transactions occur on the over-the-counter market, direct liquidity measures (based on transaction

data) are often not reliable and difficult to obtain, and researchers need to rely on other proxies of

liquidity. They examine a set of bond and issuer specific characteristics that have been used to proxy

for corporate bond liquidity and liquidity risk in other papers. The measures they examine are (i)

issued amount, (ii) whether a bond is listed on a stock exchange, (iii) whether a bond is denominated

in euros or in one of the legacy currencies, (iv) if a bond is on-the-run, (v) age, (vi) missing prices (as a

proxy for no-trades), (vii) yield volatility, (viii) the number of contributors (proxy for competition) and (ix)

yield dispersion. Their null hypothesis that liquidity risk is not priced in their sample of euro corporate

bonds is rejected for eight out of nine liquidity proxies. The only proxy that they do not find support for

relating to liquidity risk is the number of contributors. Overall, they find a significant liquidity premium,

ranging from 13 to 23 basis points. Another important contribution of their paper is that the vast

majority of empirical papers on sovereign and corporate bond liquidity studied data from the United

States, and relatively little is known about the extent to which these results apply to the euro market.

Although their paper does not give any direct measures on corporate bond liquidity, this is one of the

first papers that study the effects of corporate bond liquidity on bond yields in the European markets.

Government bonds

Also for government securities, most empirical studies focus on the US market. A much cited study on

the liquidity of the US treasury market is Fleming (2003). His paper estimates and evaluates a wide

range of liquidity measures using high frequency data from the interdealer market (GovPX data). The

measures examined are trading volume, trading frequency, bid/ask spreads, quote sizes, trade sizes

and price impact coefficients. He compares various measures against each other to provide insights

as to how liquidity can be best measured and tracked. One main result in this regard is that the bid/ask

spread is found to be a good measure of treasury market liquidity and is highly correlated with a more

sophisticated price impact measure (the coefficient from a regression of price changes on signed

volume and trades). Interestingly, he finds that trading volume and trading activity in particular are

found to be weak proxies for market liquidity, as both high and low levels of trading activity are

associated with periods of poor liquidity. This is consistent with the results and arguments in Aitken

and Comerton-Forde (2003) for the equity market.

For US government bonds, a recent paper by Goyenko, Subrahmanyam and Ukhov (2011) examines

how liquidity varies across different maturities and also between on-the-run and off-the-run bonds.

Their sample period spans more than 35 years covering the period from November 1967 – December

2005. They use data from the Center for Research in Security Prices (CRSP) daily Treasury Quotes

file containing bid and ask prices for Treasury fixed income securities for maturities ranging from 3

111

PDF to HTML - Convert PDF files to HTML files

months up to 30 years. The main liquidity measure they apply is the proportional quoted bid/ask

spread. Their main results are that spreads increases in recessions across all maturities, both for on-

and off-the-run bonds. However, they find that the increase in spreads is especially large for long term

bonds. They argue that these results suggest that investors shift into short term bonds during

recessions. The average proportional spread for medium term US government off-the-run bonds is

0.11%, which is much lower than what is found for US corporate bonds in e.g. Bao et al. (2011)

reported earlier. When looking separately at US recession periods and non-recession periods (as

classified by NBER), their spread estimates are 0.149% during recessions and 0.101% during non-

recessions. For short term bonds (Treasury bills) the whole sample average proportional spread is

0.025%, with an average spread of 0.049% during recessions and 0.002% during non-recession

periods.

For the European market a similar commissioned report to the one produced by Biais et al. (2006) on

the corporate bond market was produced for the European government bond market. This report was

produced by Dunne et al. (2006). The empirical analysis in the report uses several data sources. For

the euro-denominated European government bond market they use detailed data for the limit order

book and transactions from the MTS trading platform covering selected months over the period 2003

through 2005. They also apply various datasets for the US treasury market (e.g. Cantor Market Data,

BrokerTec and GovPX) to compare the liquidity of the European versus the US government bond

market. While they also examine a wide range of other characteristics, the liquidity variables they

calculate are the effective spread, the steepness of the order book, trade size, liquidity (depth) at the

best bid and ask quotes and depth at the best three quotes. Overall they find that the median effective

spreads for the European MTS sample is significantly higher than for the US sample. They suggest

that this can be explained by the relatively small size of issuance, the fragmented nature of the euro-

denominated markets, and the fact that there are fewer primary dealers providing liquidity across a

larger number of issues. For short maturity bonds, the median effective spread for European issues is

more than double that of US counterparts. These results are also largely mirrored for medium and

long maturities.

Covered bonds

The literature studying the liquidity of covered bonds is much scarcer than for the above asset classes.

However, a recent study by Dick-Nielsen, Gyntelberg and Sangill (2012) examines the secondary

market liquidity of government and covered bonds in Denmark before, during and after the 2008

financial crisis. They apply a liquidity measure that is closely related to the Amihud (2002) ILR

measure, but modified in a manner demonstrated by Dick-Nielsen, Feldhutter and Lando (2012) to be

a good proxy for US corporate bond liquidity. The modified ILR measure is calculated as,

, where

refers to the ith transaction on day

in bond . Hence, the measure

captures how much the price moves in either direction in response to a trade. The study finds that the

liquidity of both government and covered bonds worsened during the crisis period. Whilst government

bonds outperformed covered bonds before the crisis, the liquidity of the two instruments was broadly

similar during the crisis. Therefore the liquidity of covered bonds worsened less than government

bonds during the crisis, although overall liquidity conditions were similar across the two markets.

112

PDF to HTML - Convert PDF files to HTML files

Equities

The literature on liquidity is most developed for equities markets due to the superior data availability

both with respect to readily available very detailed intraday order level data for many markets, the time

series length of the data, and also the large theoretical equity market microstructure literature. O’Hara

(1995) provides a study of the theoretical microstructure literature while Madhavan (2000), Biais,

Glosten and Spatt (2005) and Vayanos and Wang (2012) survey both the theoretical and empirical

literature. Most, if not all, market liquidity measures that are applied to other asset classes were

initially developed for equities. The literature on equity market liquidity is large and we limit the review

to some key articles. The literature can generally be divided in two main parts: (a) studies that

examine how liquidity can be measured, and (b) studies on the asset pricing implications of liquidity

and liquidity risk (applying various liquidity measures). As the first part of the literature is the one most

relevant for this paper, we will disregard the second part (asset pricing) of the literature, and focus on

the more recent studies.

A much cited paper that that examines the development in US equity market liquidity using a century

of data is Jones (2002). He compiles a time series of bid/ask spreads on Dow Jones Stocks covering

the years 1900 through 2000. He finds that the average quoted and proportional bid/ask spreads have

decreased over the sample period, a result also documented in Hasbrouck (2009) among others. The

proportional spread estimates in Jones (2002) suggest that the average proportional spread has fallen

from around 0.7% in the early 1900s to about 0.2% in 2000. Interestingly, he toofinds that the annual

share turnover was much higher early in the sample (about 200% in 1900) than in the more recent

part of the sample (50% in 1990), although there is an increasing trend from the 1970s. Another result

in Jones (2002) is that equity market liquidity has a cyclical component and worsens in periods of

market stress. The results in Næs, Skjeltorp and Ødegaard (2011) strongly suggest that equity market

liquidity in the US (1947-2008) and Norway (1980-2008) has a cyclical component related to the

business cycle. Moreover a worsening of equity market liquidity is found to be a strong and stable

predictor of NBER recession periods and the business cycle. The business cycle component of market

liquidity is most pronounced for stocks with low market capitalisations. Using detailed equity ownership

data, Næs et al. (2011) also find that investors shift their portfolios both within equities and out of

equities during economic downturns, and that this is consistent with a flight to quality. The liquidity

measures applied in the above studies are quoted and proportional spreads, the Hasbrouck (2009)

effective cost measure, the Amihud (2002) illiquidity ratio (ILR) and the Lesmond et. al (1999)

measure (LOT).

More recent studies on equity market liquidity use detailed high frequency order book data to provide

more accurate measures of liquidity. Also, a large part of this recent empirical literature tries to identify

the low-frequency measures that best proxy for their high-frequency counterparts, and also to analyse

the liquidity of markets outside the US. A recent paper by Fong, Holden and Trzcinka (2011) examines

the liquidity of more than 18 000 stocks listed on 43 exchanges around the world. They examine 19

low frequency measures over a sample period from 1997 through 2007. They find that overall, both

price impacts and effective spreads have fallen across most markets over their sample period. An

important contribution in Fong et al. (2011) is that they propose a new low frequency proxy for

113

PDF to HTML - Convert PDF files to HTML files

effective spread (FHT) which is both simple to calculate and is shown to be superior to most other

percent cost proxies.

Commodities

Since gold is defined as a potential LCR eligible asset class in Art. 481 paragraph 2 of the most recent

CRR, it is useful to review the literature assessing the market liquidity of commodities markets. While

this literature is very scarce, a recent paper by Marshall, Nguyen and Visaltanachoti (2011) provides a

broad analysis applying various liquidity measures to a wide range of commodities, including gold.

They also evaluate which liquidity measure best captures the liquidity of commodities markets by

comparing high frequency benchmarks with low frequency proxies similar to Goyenko et al. (2009)

and Fong et.al (2010) for equities. Their study examines commodity futures covering six energy

commodities, eight agricultural commodities, three livestock commodities, five industrial metals and

two precious metals. These are the twenty-four commodities that comprise the S&P Goldman Sachs

Commodity Index (S&P GSCI). Their main data source is the Thomson Reuters Tick History database

covering both open-outcry and electronic trading. They also examine daily commodity data obtained

from Thomson Reuters Datastream. The paper examines a total of 17 low frequency (daily data)

liquidity proxies and three high frequency benchmarks (tick data). The main result in their study is that

the Amihud (2002) ILR measure has the highest correlation with the high frequency benchmarks

across all low-frequency liquidity measures. Interestingly, their results find support for earlier findings

by Ferguson and Mann (2001) that the Roll (1984) effective spread estimator is a poor proxy for

commodity liquidity. With respect to the liquidity of different commodities, their results strongly suggest

that precious metals (gold and silver) are a highly liquid commodity class. Bid/ask spread measures,

price impact measures and resiliency measures show that gold is consistently the most liquid

commodity.

114

PDF to HTML - Convert PDF files to HTML files

Annex 4: Survey of liquidity metrics

Metrics

Trading volume

Formula

Concept



Input

Trades sizes

(



)

�½



�½

Sum of the number of securities traded

during a time interval

∆t.

Turnover

(



)

�½



�½



Sum of the number of trade prices (trade

size * trade price) during a time interval

∆t.

Sum of best ask and bid volumes in the

order book at time

Sum of the natural logarithm of the best

ask and of the best bid volumes in the

order book at time

Arithmetic average of the product of the

best ask price and its respective volume

and the best bid price and its respective

volume in the order book at time

It counts the number of trades during

time interval ∆t.

It counts the number of orders inserted

into the order book during time interval

∆t.

Trades sizes and prices

Depth or Quantity depth

(

)

�½

ask

(

)



bid

(

)

log

(

)

�½

ln(

ask

(

) *

bid

(

))

Best ask

volumes

Best ask

volumes

and

bid

Log depth

and

bid

Dollar depth

(

)

�½

ask

(

) *

ask

(

)



bid

(

) *

bid

(

)

Best ask and bid prices

and volumes

Number of transactions per

time unit

Number of orders per time

unit

(



)

Order book

(



)

Order book

115

PDF to HTML - Convert PDF files to HTML files

Absolute spread (or dollar

spread)

(

)

�½

ask

(

)



bid

(

)

log

(

)

�½



ask

(

)



bid

(

)



Difference between the lowest ask and

the highest bid prices in the order book

at time

Natural logarithm of absolute spread at

time

Lowest ask and the

highest bid prices at

time

Lowest ask and the

highest bid prices at

time

Lowest ask and the

highest bid prices at

time

Lowest ask and the

highest bid prices and

the last paid price before

time

/ Order book

Lowest ask and the

highest bid prices at

time

Lowest ask and the

highest bid prices and

the last paid price before

time

/ Order book

Lowest ask and the

highest bid prices and

the last paid price before

time

/ Order book

Log absolute spread

Relative

proportional

spread calculated with mid

price

Relative spread calculated

with last price



(

)



bid

(

)



�½

2 *



ask

(

)



bid

(

)



(

)

�½

ask



ask

(

)



bid

(

)



mid

(

)

(

)

Absolute spread at time

divided by the

mid-price at time



(

)



bid

(

)



(

)

�½

ask

(

)

Absolute spread at time

divided by the

last paid price before time

Relative spread of log prices



(

)



SRelLog

(

)

�½



ask





(

)





bid



Difference of natural logarithms of the

lowest ask and of the highest bid prices

in the order book at time

Difference between the last paid price

before time

and the mid-price at time

Effective spread



(

)



bid

(

)



Seff

(

)

�½

(

)



mid

(

)

�½

(

)

 

ask









(

)



bid

(

)



(

)

 

ask



(

)



mid

(

)





Seff RelMid(t)

�½

mid

(

)



ask

(

)



bid

(

)











Relative effective spread

calculated with mid price

Effective spread at time

divided by the

mid-price at time

116

PDF to HTML - Convert PDF files to HTML files

Relative effective spread

calculated with last price



(

)



bid

(

)



(

)

 

ask



(

)



mid

(

)





Seff RelLast(t)

�½

(

)

(

)

Effective spread at time

divided by the

last price at time

Lowest ask and the

highest bid prices and

the last paid price before

time

/ Order book

Lowest ask and highest

bid prices and their

respective volumes at

time

Trades sizes and prices

and asset return during

time interval ∆t.

Trades sizes and prices

and asset return during

time interval ∆t and total

number

securities

available and owned by

the firm

Best bid and the best

ask volumes and trades

sizes and prices.

Order book

Quote slope



ask

(

)



bid

(

)



(

)

(

)

�½

log

(

)



ln(

ask

(

))



ln(

bid

(

))



(



)

(

)

�½

(



)

Absolute spread divided by the Log

depth.

Liquidity ratio 1 (Baker

(1996)) (or Amivest liquidity

ratio)

Liquidity ratio 2 (Ranaldo

(2000))

Turnover divided by the return during

time interval ∆t.

(

)

(

)

�½

(



)

LR1 divided by the difference between

the total number of securities and the

number owned by the firm during time

interval ∆t.

Order ratio

(

)



ask

(

)

(

)

�½

bid

(



)

Difference between the best bid and the

best ask volumes at time

scaled by the

turnover during time interval

Difference between the ask and the bid

price in the order book at time

for a

certain turnover

to be generated.

This represents the number of securities

that must be traded to move the price

a certain amount from the quote

Market impact

(

)

�½

ask

(

)



bid

(

)

Depth for price impact

ask

(

ask

(

)



mid

(

))

�½

ask

(

ask

(

)



mid

(

))

bid

(

bid

(

)



mid

(

))

�½

bid

(

bid

(

)



mid

(

))

Order book

117

PDF to HTML - Convert PDF files to HTML files

midpoint.

Price impact









ask

(

) *

ask

(

)



ask

(

)

�½



�½





(

) *

mid

(

)

















bid

(

) *

bid

(

)



bid

(

)

�½ 



�½





(

) *

mid

(

)









with

PI is calculated on the demand and

supply curve. A market order of size

executed at

different prices with

securities trading at price p

and

Order book

(

)

�½



(

)

�½

(

)

�½



(

)

�½

Xetra

Liquidity

Measure

(XLM) (weighted

spread

liquidity measure)







ask

(

) *

(

)









bid

(

) *

(

)





The XLM is a weighted spread liquidity





measure. It was developed by Deutsche

XLM

(

mid

)

�½ 



Boerse Group for its electronic platform

mid





Xetra. The Xetra platform calculates





automatically XLM based on the Xetra

electronic book which includes the

iceberg

orders,

trades

submitted

invisible.

The XLM is the average limit-order-

volume weighted price of all limit orders,

which are required for transacting a

Measure provided by

Xetra for the Xetra

trading platform but can

be calculated from the

order book (including

iceberg orders)

118

PDF to HTML - Convert PDF files to HTML files

specific size, relative to the mid price.

The XLM can be seen as the cost of

immediate execution of a specific size

(n*p

mid

)

compared to the mid price.

The XLM is very similar to the Price

Impact.

Marginal Supply

Curve (MSDC)

Demand

The Marginal Supply-Demand Curve (MSDC) is defined formally

as follows:

An asset A is a good traded in the market with prices given by a

function



(

) :



\ {0}

 

called the Marginal Supply-Demand Curve (MSDC) which

satisfies:

m(x

)≥m(x

i+1

) if x

i+1

m(x)

is cadlag (i.e.

lim

y↓x

m(y)=m(x))

for

x<0

and ladcag (i.e.

lim

y↑x

m(y)=m(x))

for

x>0

Let

m =m(0 )

be the

best bid

and

m =m(0 )

the

best ask.

The

quantity

δm= m -m ≥0

is called the

bid/ask spread.

A necessary assumption is the no arbitrage requirement that any

bid be lower than any ask prices. Given this assumption, the

MSDC is decreasing by construction.

Let

represent the prices and

∆x

represent the corresponding maximum

sizes (expressed in number of contracts

x).

For instance, suppose that we have

to sell a given number of contracts

of a

certain asset. We list all the bid prices in

a decreasing order vector, which lists

the prices from the most convenient to

the least convenient from our sell point

of view. So we will index the quotes

accordingly (m

≥ m

if i ≤ j).

We will firstly

exploit the highest bids by selling

amounts

∆z

≤ ∆x

until

∑

∆z

= z.

The net

proceed from our sales will be

P=∑

∆z

For ask prices, the reasoning

is symmetrical and the best quotes will

be the lowest.

We can build the MSDC

x→m(x)

collecting all the prices

m(x)

available for

trades of

contracts and sorting them

by decreasing order in

We obtain a

Order book

119

PDF to HTML - Convert PDF files to HTML files

single decreasing curve with the best

ask quotes for buying

contracts in the

interval

x є [-z,0[

and the best bid quotes

for selling

contracts in the interval

x є

]0,z].

The MSDC does not only include the

bid/ask spread (tightness) information

but also the bid/ask depth (depth)

information and the demand and supply

information (resiliency).

120

PDF to HTML - Convert PDF files to HTML files

Annex 5: Responses to the EBA Discussion Paper on the methodology to be used in this

report

Between 21 February 2013 and 21 April 2013 the EBA conducted a public consultation on the

methodology it proposes to use to identify a uniform definition of liquid assets in the LCR under the

CRR.

Thirty-four responses have been received by the EBA (23 responses attributable to the banking

sector, 5 from other financial institutions and 6 responses from other sectors).

Answers to the questions

Q1. Given the difficulties with obtaining transactional data outlined here, do you think a data

sample cover 2008-2012 is sufficient for this analysis? Would you see merit in extending the

sample in those countries where more data is available?



There is broad agreement about the data sample proposed. Some responses recognise that a

wider data sample may have some practical benefits, but data would be difficult to obtain

(MiFIDMiFiD data is not possible to obtain). Some respondents disagree with the proposed

sample and ask for a longer period (at least 10 years, or more than an economic cycle).



There is ample support for the inclusion of data for repo markets.

Some respondents have raised the issue of including assets denominated in non-European

currencies in the analysis.

Q2. Do you have additional data sources to suggest? Specifically, can you suggest a source of

repo data and gold that would fit our needs?

The most commonly suggested additional data sources are:



data providers (Bloomberg, Reuters, etc.);

custodians;

clearing houses;

The REPO council annual survey.

Some respondents stress the advantages of the use of publicly available date, so that banks can

duplicate the EBA´s assessment.

121

PDF to HTML - Convert PDF files to HTML files

Q3. Do you agree with the list of liquidity metrics under consideration to be used in the EBA

assessment, as mentioned in this section and Annex 5? Can you suggest further metrics the

EBA should make use of, where information would be available?



The list of metrics is considered sufficiently broad. The main concern pointed out in some

responses is calibration of the individual metrics, since the consultation paper is not explicit about

the weighting and combination of the individual metrics. Several responses are concerned that

some assets, considered as liquid, may fail some of the individual metrics (failing a litmus test,

thresholds). In this context, the concern is raised that the use of several metrics additively may

lead to an overly restrictive definition of liquid assets.



Several respondents point to the fact that overreliance on quantitative liquidity metrics might lead to

a liquidity definition biased towards big markets and will not take the rather qualitative liquidity

features of small markets sufficiently into account.



There was moreover a concern that since the EBA proposes to establish a relative ranking of

assets the final cut off between “extremely liquid”, “highly liquid” and “ineligible” would be arbitrary,

since no general definition of liquidity is provided.

Q4. Do you agree with the list of explanatory characteristics whose linkage to liquidity it is

proposed to be tested in the EBA assessment? Can you suggest further characteristics the

EBA should assess?

There is broad support for the explanatory characteristics proposed. However, some concerns, listed

below, are indicated in the responses.



Practicality for market participants and the ability to measure such a characteristic objectively is not

evident.

There were concerns about the use of these metrics in the analysis.

Many of the explanatory characteristics (e.g. the large number of market makers, being traded via

additional platforms and markets, the wide range of potential buyers) cannot be handled in

quantitative terms, since they are qualitative in nature.



The criterion of “remaining time to maturity” considered inappropriate for short term securities and

liquid assets monetised via repo.

Eligibility in the main clearing houses (i.e. GC pooling) should be considered as an additional

qualitative characteristic.

The EBA should give consideration to the inclusion of an asset as a constituent of major market

indices or the presence of active derivatives and index markets for the asset (e.g. Futures, CDX,

ITRAXX etc)

122

PDF to HTML - Convert PDF files to HTML files

Q5. Do you agree with the methodology proposed? Do you have alternative approaches that

might be used?

There is general acknowledgement of the difficulties faced by the EBA with producing a uniform

definition of liquid assets. Most respondents agree with the proposed methodology and advocate a

simple, harmonised definition of liquid assets. The definition should be easy to use by banks. As for

differentiation within asset classes and across countries, views are split, with some respondents

asking for a harmonised approach and others arguing in favour of differentiation between currencies

and markets.

The fact that the EBA will define eligible asset classes is expressively welcomed by respondents, as

banks lack the (data) resources to conduct a comprehensive liquidity analysis on their own. Doubts

remain however, as to what extent banks will have to assess the liquidity of the single selected asset.

In this context the wish for an ISIN list of liquid assets is repeatedly mentioned.

A major flaw is seen in EBA focusing only on transactional data as opposed to repos.

One alternative repeatedly proposed is ECB-eligibility as only and sufficient criterion for liquidity

(collateral eligibility in repo markets for equities). This approach is justified by the current role of the

central bank and the definition of the LCR stress scenario of being not only idiosyncratic but also

systemic.

As a variation, ECB-eligibility is proposed as the “standard approach” whereas sophisticated banks

should be allowed to use their own definitions of liquid assets.

As another alternative, respondents propose a more qualitative approach, where entities analyse the

liquidity of assets on their own on the basis of criteria and thresholds proposed by EBA.

Other main findings

Periodic review/update of the definition of liquid assets is suggested, as the liquidity of asset classes

and/or the appropriateness of liquidity metrics may change over time.

Respondents ask for increased transparency on the scorecard model used by the EBA, i.e. the

envisaged combination of the outcome of the application of different liquidity metrics. This

transparency should allow banks to duplicate the identification process of liquid assets, to anticipate

changes in the eligibility and avoid cliff effects.

The proposed approach focuses primarily on transactional data as a basis for the liquidity definition of

an asset. That would not appropriately reflect the possibility of liquidating a high credit quality asset

which is not traded actively in the market simply because of the fact that it is kept by their current

holders as a high credit quality asset stock. “Structural bid” should therefore be considered as a

correction factor in the analysis.

123

PDF to HTML - Convert PDF files to HTML files



The impact of the definition of the liquidity of assets itself on the liquidity of the respective assets

should be taken into account (liquidity is not determined entirely endogenously). The EBA

should also take into account the expected effect of other regulatory initiatives (financial

transaction tax, Solvency II, EMIR).



The EBA´s analysis focuses on assets denominated in European currencies. Internationally

active banks will require guidance on the definition of HQLA denominated in other currencies,

most importantly the US Dollar.



Some respondents refer to the possible outcome of the EBA´s analysis and argue in favour of

the inclusion of specific assets (notably RMBS and ABS). One respondent suggested a further

consultation on the uniform definition of liquid assets itself. For the sake of ensuring a global

level playing field, other respondents express their expectation that the EBA´s definition should

be in line with the one established by the BCBS.

Proposed Feedback

The EBA proposes to address the issues listed below.



The EBA currently stepping up efforts to include liquidity information from repo markets.

Whilst producing a relative ranking of the liquidity of different asset classes, the quantitative

analysis needs to be flanked by qualitative information. Defining the cut-off between eligible and

ineligible asset classes needs to be based to a certain extent on expert judgement and may be

somewhat subjective. The EBA agrees on the need to depict the rationale of the analysis in a

transparent way and will point out the stage at which expert judgements need to be made or

political decisions taken.



The need for qualitative judgement does also apply to the incorporation of a forward looking

approach into the analysis, taking into account other regulatory initiatives implemented in

parallel, such as financial transaction tax.



In the spirit of the single rule book, the EBA will present a uniform definition of HQLA applicable

throughout Europe. The qualitative metrics used should allow application in all markets,

independently of their size.



As for the liquidity assessment of specific asset classes, this subject was not within the scope of

the discussion paper, which focused strictly on the methodology and not on any analytical

outcome. At this juncture the outcome of the analysis cannot be prejudged.



The scope of the asset classes subject to the assessment is defined by the CRR. Government

bonds fall within this scope as they might be either admissible as HQLA in their own right if

covering a liquidity risk in the country issuing a bond, as well as thanks to their liquidity features.

124

PDF to HTML - Convert PDF files to HTML files

Furthermore, when producing a relative ranking of asset classes, government bonds might

serve as a benchmark.



Concerning the interdependency between the definition of HQLA and monetary policy, the EBA

will conduct its analysis of the liquidity of assets independently of these considerations, which

will be a subject of the overall impact assessment of the LCR. However, central bank eligibility

is acknowledged to be a factor influencing the market liquidity of an asset as an explanatory

characteristic and is currently a necessary condition in the CRR for HQLA eligibility.



The definition of HQLA is based on market liquidity. The systemic component of LCR stress and

the role of central bank liquidity are taken into account in the design of the LCR, which allows

recognition of exposures to central banks and standby credit facilities of central banks as part of

the enumerator of the LCR.



Due to reasons of practicability and data restrictions, the quantitative assessment conducted by

the EBA will focus on European currencies only.

Within the framework of asset classes proposed which will be proposed by the EBA, the

selection of the individual ISIN remains the responsibility of the bank.

Under the CRR, the EBA has been charged with providing a one-off report on a uniform

definition of liquid assets. If there is a need to adapt this definition to a changing market or

regulatory environment (that might be unveiled by the annual overall impact assessment of the

LCR) we believe the European legislator will duly react.

125

PDF to HTML - Convert PDF files to HTML files

Annex 6: Bond data description

The major source of bond data for this analysis is transaction data for EU bond markets. The data

gathered encompass outright sales trades on government bonds, corporate bonds, covered bonds,

bonds issued by promotional banks, RMBS and ABS. The data source is data collected by national

supervisors on the basis of the reporting requirements on security trades due to the MiFID directive.

Data has been submitted from 27 countries covering trades in the period from 1 January 2008 to

30 June 30 2012.

Cleaning and filtering the data

The purpose of this step is to filter out erroneous observations, and also to obtain the most uniform

data set across the contributing countries. This ensures that the level of quality both within and

between countries is similar and thus permits a comparative analysis of the metrics. The filtering

therefore deals with both erroneous reporting and the inherent nature of the data. Some filters are

therefore absolute thresholds, while others try to consider specific market characteristics. The specific

filtering methods and thresholds are presented below.

Price filters

Bonds should trade close to 100. Hence, reported prices that deviate from 100 are considered

erroneous, even though some variation should be allowed as credit rating or changing interest rate

levels can create substantial deviations from the issue price. Therefore all transactions not priced

between 5 and 500 are eliminated.

For many countries there are several bonds that trade at prices around 1. These observations are

assumed to be percentage notations. Hence, to include these trades, all prices between 0.7 and 1.3

are multiplied by 100.

The next step is to eliminate outliers within an individual ISIN code. These outliers are for the most

part caused by a defaulted trade price (100) or simply a mistake made by the reporter. This problem

is alleviated by removing observations that range outside two standard deviations from the mean

price, measured over the entire sample for a specific bond. This method allows for larger single price

jumps in volatile bonds than less volatile instruments, and the filter is unlikely to remove any single

price jumps that are not erroneous. Finally, bonds that do not have any variation in the price are also

deleted using this filter. This is done because some bonds are reported only with default prices (100).

In addition, this filter also removes bonds in which only one trade is registered. A limit of a 100 bps

price change between trades has been implemented within each instrument. These trades are either

errors or the price change reflects new information, as opposed to liquidity characteristics. As a last

price filter, any prices within an instrument that are in multiples of 10 compared to the lowest price are

adjusted. That is to say, if a bond is traded at both 30 and 300, the large price is divided by 10.

126

PDF to HTML - Convert PDF files to HTML files

Trade size filters

Initially all trades with the same time and price are merged, as other investors will not be able to

differentiate between them, and therefore the price information should only be used once.

There are a substantial amount of retail trades in the data (small trade value). As these trades do not

carry any liquidity information, and because there seems to be a correlation between retail trades and

bad reporting in general, all trade sizes less than EUR 500 000 are deleted. This eliminates the

majority of bad reporting, meanwhile preserving a maximum of information. With respect to large trade

sizes, all trades that are greater than ten percent of the outstanding amounts are removed. This

removes all upward errors in the reporting.

Time and instrument code filters

As the time stamp is essential in removing repos and redundant trades, all trades without useful time

information are removed. Furthermore all transactions in invalid ISIN-codes are removed.

Repo filters

Market participants should in principle exclude repo transactions in MiFID trade reporting. Repos are

nonetheless present in the data, and should thus be removed. Repos are identified as trades between

the same entities, within a day with the same quantity, and as a buy and a sell respectively. Most

repos are reported within one or two minutes of each other, however. Using the day reference point

does not change results, and allows for some lag in reporting.

Redundancy filter

If both parties in a trade are MiFID reporters, a trade will figure twice in the data. The redundant trades

are removed by identifying trades with exactly the same price and quantity, reported as a buy and a

sell between the parties. If two such trades are identified, only the buy side is kept in the data.

127

PDF to HTML - Convert PDF files to HTML files

Annex 7: Equity data description

The source of equity data for this analysis is Bloomberg. For each of the countries represented in our

bond sample from the EU and the EEA, we pick out equities included in a representative major stock

index of that country. The indices included are presented below.

We use the daily closing price, daily trading volume, outstanding amount and industry sector variables

from Bloomberg for our analysis. The data cover the full sample period from 1 January 2008 to

30 June 2012. We eliminate observations where the daily trading volume is below EUR 500 000. This

is in order to eliminate retail size trades and ensure comparability of the liquidity metrics with the

analysed bond data. No further cleaning or filtering of the data is performed.

Country

ISO

Full Name

Description

# Members

Austria

AUSTRIAN TRADED ATX

INDX

The Austrian Traded Index is a capitalisation-

weighted index of the most heavily traded

stocks on the Vienna Stock Exchange. The

equities use free-float adjusted shares in the

index calculation. The index has a base level

of 1000 as of 2 January 1991.

The BEL 20 Index is a modified capitalisation-

weighted index of the 20 most capitalised and

liquid Belgian stocks that are traded on the

Brussels Stock Exchange. The equities use

free-float shares in the index calculation.

The index was developed with a base value

of 1,000 as of 1 January 1991.

Cyprus General Market Index (CSE)

Belgium

BEL 20 INDEX

Eurozone

Cyprus

GENERAL MARKET INDEX

CSE

OMX TALLINN OMXT

Estonia

OMX Tallinn (OMXT)is a capitalization

weighted chain-linked total return index which

includes all the shares listed on the Main &

Secondary lists on the Tallinn Stock Exchange.

The aim of the index is to reflect the current

status & changes on the Tallinn market. The

base date is 3 June 1996, with a value of 100.

The OMX Helsinki 25 Index is a modified-

capitalisation weighted index that consists of

the 25 most traded series on Helsinki Stock

Exchange's Main List. The index is used as a

benchmark for the Finnish Market. The index

calculation includes free float factors and each

company in the index is limited to a weight of

128

Finland

OMX HELSINKI 25 INDEX

PDF to HTML - Convert PDF files to HTML files

10%.

France

CAC 40 INDEX

The CAC 40, the most widely-used indicator of

the Paris market, reflects the performance of

the 40 largest equities listed in France,

measured by free-float market capitalisation

and liquidity. The index was developed with a

base level of 1,000 as of 31 December 1987.

The German Stock Index is a total return index

of 30 selected German blue chip stocks traded

on the Frankfurt Stock Exchange. The equities

use free-float shares in the index

calculation. The DAX has a base value of

1,000 as of 31 December 1987. As of June 18,

1999 only XETRA equity prices are used to

calculate all DAX indices.

The German DAX Mid-Cap Index is a total rate

of return index of 50 mid-cap issues that rank

below the DAX. The MDAX exclusively tracks

issues from the various traditional sectors. The

index was developed with a base value of

1,000 as of December 31, 1987.

The Athens Stock Exchange General Index is

a capitalisation-weighted index of Greek

stocks listed on the Athens Stock Exchange.

The index was developed with a base value of

100 as of December 31, 1980.

The ISEQ Overall Index is a capitalisation-

weighted index of all Official list equities in the

Irish Stock Exchange but excludes UK

registered companies. The index has a base

value of 1000 as of January 4, 1988.

The Index will consist of the 40 most liquid and

capitalised stocks listed on the Borsa Italiana.

In the Foreign shares will be eligible for

inclusion in theFTSE MIB Index. Secondary

lines will not be eligible for inclusion. The

calculation and methodology will be

unchanged from the S&P MIB Index.

The Luxembourg LuxX Index is a weighted

index of the most capitalised (by free float) and

liquid Luxembourg stocks. The index was

developed with a base value of 1,000 as of

Germany

DAX INDEX

Germany

MDax Index

Greece

Athex Composite Share Pr

Ireland

IRISH OVERALL INDEX

Italy

FTSE MIB INDEX

Luxembo

urg

LUXEMBOURG LuxX

INDEX

129

PDF to HTML - Convert PDF files to HTML files

January 4, 1999.

Malta

MALTA STOCK

EXCHANGE IND

The Malta Stock Exchange (MSE) index is a

capitalisation weighted index

encompassing all shares traded on the Stock

Exchange of Malta. Index = Current Market

Value of all shares listed. The index was

created on December 27, 1995 with a base

value of 1000. From May 19, 1998 the index

has been calculated on a daily basis.

The AEX-Index is a free-float adjusted market

capitalisation weighted index of the leading

Dutch stocks traded on the Amsterdam

Exchange. The index was adjusted to the

Dutch Guilder fixing rate. The old value as of

31 December 1998 was 1186.38 and the new

value at start of trading on 1/4/99 was 538.36,

after

conversion. HP and GP can be

adjusted back to Dutch Guilders by typing

NLG.

The Portugal PSI 20 Index is a capitalisation

weighted index of the top 20 stocks listed on

the Lisbon Stock Exchange. The equities use

free-float shares in the index calculation. The

index was developed with a base value of

3,000 as of December 31, 1992.

The Slovak share index (SAX) is a capital-

weighted total return index that compares the

market capitalisation of a selected set of

shares with the market capitalisation of the

same set of shares as of a given reference

day. The index was developed with a base

level of 100 of September 14, 1993.

SBI TOP is the Slovenian blue chip index. It is

a free-float capitalisation weighted index

comprising the most liquid shares traded on

the Ljubljana Stock Exchange. Each stock's

weighting is capped at 30%. The index was

developed with a base level of 1 000 as of 31st

March 2006.

The IBEX 35 is the official index of the Spanish

continuous Market. The index is comprised of

the 35 most liquid stocks traded on the

Continuous market. It is calculated,

supervised and published by the Sociedad de

130

Netherlan

AEX-Index

Portugal

PSI 20 INDEX

Slovakia

SLOVAK SHARE INDEX

Slovenia

Slovenian Blue Chip Idx

Spain

IBEX 35 INDEX

PDF to HTML - Convert PDF files to HTML files

Bolsas. The equities use free-float shares in

the index calculation. The index was created

with a base level of 3000 as of December 29,

1989.

Bulgaria

SOFIX INDEX

The Bulgaria Stock Exchange Sofix Index is a

free-float market capitalisation weighted index

representing the most liquid companies listed

on the exchange. The market capitalisation of

each company should not be less than BGN

50 million.

The PX index is the official index of the Prague

Stock Exchange. The index was calculated for

the first time on March 20, 2006 when it

replaced the PX50 and PX-D indices. The

index took over the historical values of the

PX50 index. The starting date was April 5,

1994 with a base of 1 000 points. As of 24 Sep

2012, its composition fully reflects the free float

of members due to methodology changes.

The OMX Copenhagen 20 index consists of

the 20 most actively traded shares on the

Copenhagen Stock Exchange. OMXC20 is a

tradable index on which futures and options

are issued. The composition of the OMXC20

index is revised twice a year.The OMXC20 is a

market weighted price index. The base date for

this index is July 3, 1989, with a base value of

100.

The Budapest Stock Exchange Index is a

capitalisation-weighted index adjusted for free

float. The index tracks the daily price only

performance of large, actively traded shares

on the Budapest Stock Exchange. The shares

account for 58% of domestic equity market

capitalisation. The index has a base value of

1 000 points as of January 2, 1991 and is a

total return index.

OMX Riga (OMXR) is an all-share index

consisting of all the shares listed on the main

and secondary lists on the Riga Stock

Exchange in Latvia with the exception of

companies where a single shareholder

controls at least 90% of the outstanding

shares. The aim of the index is to reflect the

current status & changes on the Riga market.

The base date is 1999/12/31, with a value of

131

Czech

Republic

PRAGUE STOCK EXCH

INDEX

Denmark

OMX COPENHAGEN 20

INDEX

Hungary

BUDAPEST STOCK EXCH

INDX

Latvia

OMX RIGA OMXR

PDF to HTML - Convert PDF files to HTML files

100.

Lithuania

OMX VILNIUS OMXV

OMX Vilnius (OMXV) is a total return index

which includes all the shares listed on the main

and secondary lists on the Vilnius Stock

Exchange. The aim of the index is to reflect the

current status &changes on the Vilnius market.

The base date is December 31, 1999, with a

value of 100.

The Polish Traded Index (PTX) is a

capitalisation-weighted price index which

contains the most attractive Polish blue chip

stocks. The index is calculated by Wiener

Börse and is part of the CECE Index Family.

The index had a base value of 612.12 index

points on January 04, 1999.

Bucharest Exchange Trading Index (BET) is a

capitalisation-weighted index, comprised of the

most liquid 10 stocks listed on the BSE tier 1.

The index is a price index and was developed

with a base value of 1000 as of Sepember 22,

1997.

The OMX Stockholm 30 Index is a

capitalization-weighted index of the 30 stocks

that have the largest volume of the trading on

the Stockholm Stock Exchange. The equities

use free-float shares in the index calculation.

The index was developed with a base level of

125 as of September 30, 1986. ** Effective on

April 27, 1998 there was a 4-1 split of the index

value.

The FTSE 100 Index is a capitalisation-

weighted index of the 100 most highly

capitalised companies traded on the London

Stock Exchange. The equities use an

investibility weighting in the index calculation.

The index was developed with a base level of

1 000 as of January 3, 1984. * Please see the

UKEDA 100 Index and the FTPTP 100 Index

for the official FTSE 100 index dividend yield

and P/E ratio. *

The OMX Iceland 6 index consists of the six

most actively traded shares on the Nasdaq

OMX Iceland exchange. The limited number of

132

Poland

POLISH TRADED INDEX

EUR

Romania

BUCHAREST BET INDEX

Sweden

OMX STOCKHOLM 30

INDEX

United

Kingdom

FTSE 100 INDEX

101

Iceland

OMX Iceland 6 PI

PDF to HTML - Convert PDF files to HTML files

constituents guarantees that all the underlying

shares of the index have excellent liquidity,

which results in an index that is highly suitable

as underlying for derivatives products.

Norway

OBX STOCK INDEX

The OBX Index is a capitalization-weighted

index of the largest companies traded on the

Oslo Stock Exchange. The equities use free-

float shares in the index

calculation.

* Note that effective April 21 2006, the OBX

Stock Index became a total return index and

had a price history split by a factor of 4.

The CROBEX is a capitalisation-weighted

index, capped at a maximum 20% weighting of

the index capitalisation. The index was

designed to measure the price movements of

shares listed on the Zagreb Stock Exchange.

The CROBEX was developed with a base level

of 1 000 for the base period beginning

September 1, 1997.

Croatia

CROATIA ZAGREB

CROBEX

133

PDF to HTML - Convert PDF files to HTML files

Annex 8: Definition and implementation of liquidity metrics

In this annex we describe in detail how the liquidity measures presented in section 4.2.1 are defined

and how the measures are calculated empirically.

The Amihud illiquidity ratio

The daily Amihud illiquidity measure for asset

for day

is given as the ratio of the daily absolute log

return on asset

to the trading volume for asset

of that day.

In order to compute the Amihud measure we require a positive trading volume on the two adjacent

trading days over which the log return is calculated. This is to ensure that the Amihud measure relates

to liquidity effects and not general price impacting market information.

Monthly illiquidity ratios are created by taking the median of the daily Amihud measure of each ISIN

over a calendar month. For cross asset class comparisons, averages of the monthly Amihud measure

are taken across assets within the respective asset class after removing the upper 0.05 percentile

within each asset class, which means 99.5% of data is used. This ensures that we capture the general

characteristics of the data, and that the conclusions are not based on outliers. The results presented in

this paper are scaled to percentage points.

Price impact

The daily price impact measure for asset

for day

is given as the absolute log return between two

adjacent days:

In order to compute the price impact measure we require a positive trading volume on the two

adjacent trading days over which the log return is calculated.

Monthly price impact measures are created by taking the median of the daily price impact measure of

each ISIN over a calendar month. For cross asset class comparisons, averages of the monthly price

impact measure are taken across assets within the respective asset class after removing the upper

0.05 percentile within each asset class, which means 99.5% of data is used. This ensures we capture

the general characteristics of data, and that the conclusions are not based on outliers. The results

presented in this paper are scaled to percentage points.

134

PDF to HTML - Convert PDF files to HTML files

The Roll measure

Roll (1984) shows that under certain conditions, the percentage bid/ask spread equals two times the

square root of minus the covariance between consecutive returns:

Where

is the time period over which the measure is calculated. We compute a daily Roll measure,

using a 21 trading day rolling window in the computation of the covariance. We require at least 6

transactions within the 21 day window in order to compute the Roll measure. This is to ensure a

reasonably stable covariance computation. In cases where a positive covariance is found, the Roll

measure is set to zero. We define a monthly Roll measure by taking the median of the daily Roll

measures for each ISIN over a calendar month. For cross asset class comparisons, averages of the

monthly Roll measures are taken across assets within the respective asset class after removing the

upper 0.05 percentile within each asset class, which means 99.5% of data is used. This ensures we

capture the general characteristics of data, and that the conclusions are not based on outliers. The

results presented in this paper are scaled to percentage points.

Trading volume

The trading volume metric is calculated as the sum of the EUR value Q of each trade x per ISIN i in

month t. X denotes the total number of trades for ISIN i in month t.

For cross asset class comparisons, averages of the trading volume measures are taken across assets

within the respective asset class

Turnover ratio

The turnover ratio shows the trading volume compared to outstanding amount for a given month. It is

calculated as volume per month t, divided by the outstanding amount for each ISIN i:

Then the average is taken for each asset class. This ensures that highly traded instruments do not

weigh more than other instruments in the final ratio. The outstanding amount is taken at the time of

extraction from Bloomberg. If unavailable, the issued amount is used instead. For a small subset of

the instruments neither variable is available, and these instruments have been left out of the turnover

ratio calculation.

135

PDF to HTML - Convert PDF files to HTML files

Zero-trading days

The zero-trading days metric is given as the percentage of days during a month where an asset is not

traded. Zero trading is calculated as the inverse of the number of days with a recorded trade, divided

by the number of trading days in a month t for each ISIN i. We set the number of trading days to 21 for

all periods, which is assumed to be representative across the different markets and countries:

The zero-trading day metrics for each asset class is calculated as the average of all instruments in

that class.

Volatility

The volatility metric is calculated as the standard deviation of the log returns r of each ISIN i within a

month t. The number of trades per ISIN in each month is denoted T.

Where x denotes each specific trade for an ISIN i in each month t. The number of trades for an ISIN in

a month is denoted X. For cross asset class comparison we take the average for each metrics within

the asset class. The results printed in the paper are scaled to percentage points.

30-day price change

The 30-day price change measure is defined as the maximum observed absolute daily log return over

a 30-day period

for t=1 through 21, where t is trading days. The measure is computed over rolling 30 calendar day

periods, so that a daily measure is established. A monthly measure is computed by taking the

maximum of the daily measures within a calendar month for each ISIN. For cross asset class

comparisons, averages of the monthly measures are taken across assets within the respective asset

class. The results presented in this paper are scaled to percentage points.

136

PDF to HTML - Convert PDF files to HTML files

0,5

1,5

2,5

AMIHUD ILLIQUIDITY RATIO

Annex 9: Time series for cross asset class analysis

1.GOVERN

2.COVERE

3.ABS

4.CORPOR

5.FINANC

EQUITY

200801

200802

200803

200804

200805

200806

200807

200808

200809

200810

200811

200812

200901

200902

200903

200904

200905

200906

200907

200908

200909

200910

200911

200912

201001

201002

201003

201004

201005

201006

201007

201008

201009

201010

201011

201012

201101

201102

201103

201104

201105

201106

201107

201108

201109

201110

201111

201112

201201

201202

201203

201204

201205

201206

137

PDF to HTML - Convert PDF files to HTML files

0,02

0,1

0,04

0,06

0,08

ROLL

PRICE IMPACT

0,12

1.GOVERN

1.GOVERNMENT

2.COVERE

3.ABS

4.CORPOR

5.FINANC

EQUITY

GOLD

2.COVERED

3.ABS

4.CORPORATE

5.FINANCIAL

EQUITY

GOLD

200801

200802

200803

200804

200805

200806

200807

200808

200809

200810

200811

200812

200901

200902

200903

200904

200905

200906

200907

200908

200909

200910

200911

200912

201001

201002

201003

201004

201005

201006

201007

201008

201009

201010

201011

201012

201101

201102

201103

201104

201105

201106

201107

201108

201109

201110

201111

201112

201201

201202

201203

201204

201205

201206

200801

200802

200803

200804

200805

200806

200807

200808

200809

200810

200811

200812

200901

200902

200903

200904

200905

200906

200907

200908

200909

200910

200911

200912

201001

201002

201003

201004

201005

201006

201007

201008

201009

201010

201011

201012

201101

201102

201103

201104

201105

201106

201107

201108

201109

201110

201111

201112

201201

201202

201203

201204

201205

201206

138

PDF to HTML - Convert PDF files to HTML files

Average volume per ISIN

Millioner

6.000

5.000

4.000

3.000

2.000

1.000

1.GOVERN

2.COVERE

3.ABS

4.CORPOR

5.FINANC

EQUITY

139

PDF to HTML - Convert PDF files to HTML files

0,1

0,2

0,3

0,4

0,5

0,6

0,7

0,8

0,9

0,05

0,15

0,1

0,2

0,25

ZERO TRADING

TURNOVER RATIO

1.GOVERN

2.COVERE

3.ABS

4.CORPOR

5.FINANC

EQUITY

2.COVERE

3.ABS

4.CORPOR

5.FINANC

EQUITY

GOLD

200801

200802

200803

200804

200805

200806

200807

200808

200809

200810

200811

200812

200901

200902

200903

200904

200905

200906

200907

200908

200909

200910

200911

200912

201001

201002

201003

201004

201005

201006

201007

201008

201009

201010

201011

201012

201101

201102

201103

201104

201105

201106

201107

201108

201109

201110

201111

201112

201201

201202

201203

201204

201205

201206

200801

200802

200803

200804

200805

200806

200807

200808

200809

200810

200811

200812

200901

200902

200903

200904

200905

200906

200907

200908

200909

200910

200911

200912

201001

201002

201003

201004

201005

201006

201007

201008

201009

201010

201011

201012

201101

201102

201103

201104

201105

201106

201107

201108

201109

201110

201111

201112

201201

201202

201203

201204

201205

201206

140

PDF to HTML - Convert PDF files to HTML files

PRICE VOLATILITY

1.GOVERN

2.COVERE

3.ABS

4.CORPOR

5.FINANC

EQUITY

GOLD

200802

200803

200804

200805

200806

200807

200808

200809

200810

200811

200812

200901

200902

200903

200904

200905

200906

200907

200908

200909

200910

200911

200912

201001

201002

201003

201004

201005

201006

201007

201008

201009

201010

201011

201012

201101

201102

201103

201104

201105

201106

201107

201108

201109

201110

201111

201112

201201

201202

201203

201204

201205

201206

141

PDF to HTML - Convert PDF files to HTML files

PRICE CHANGE

1.GOVERN

2.COVERE

3.ABS

4.CORPOR

5.FINANC

EQUITY

GOLD

200802

200803

200804

200805

200806

200807

200808

200809

200810

200811

200812

200901

200902

200903

200904

200905

200906

200907

200908

200909

200910

200911

200912

201001

201002

201003

201004

201005

201006

201007

201008

201009

201010

201011

201012

201101

201102

201103

201104

201105

201106

201107

201108

201109

201110

201111

201112

201201

201202

201203

201204

201205

201206

142

PDF to HTML - Convert PDF files to HTML files

Annex 10: Probability distributions for cross asset class analysis

AMIHUD ILLIQUIDITY

RATIO

PRICE IMPACT

143

PDF to HTML - Convert PDF files to HTML files

ROLL

RETURN VOLATILITY

144

PDF to HTML - Convert PDF files to HTML files

TURNOVER RATIO

145

PDF to HTML - Convert PDF files to HTML files

ZERO TRADING

PRICECHANGE

146

PDF to HTML - Convert PDF files to HTML files

VOLUME

147

PDF to HTML - Convert PDF files to HTML files

Annex 11: ECAI rating mapping

One of the explanatory characteristics used in the analysis of bond asset classes is their ECAI rating.

For the purposes of this analysis, this variable is formed as a direct mapping of bond credit ratings.

The table below shows how the credit rating of individual bonds has been mapped into one of the 6

ECAI buckets.

ECAI

Moody’s Rating

Aaa to Aa3

A1 to A3

Baa1 to Baa3

Ba1 to Ba3

B1 to B3

Caa1 and below

S&P/Fitch Rating

AAA to AA-

A+ to A-

BBB+ to BBB-

BB+ to BB-

B+ to B-

CCC+ and below

148

PDF to HTML - Convert PDF files to HTML files

Annex 12: Relevant CRR articles on operational requirements

Article 417: Operational requirements for holdings of liquid assets

The institution shall only report as liquid assets those holdings of liquid assets that meet the following

conditions:

a) they are appropriately diversified. Diversification is not required in terms of assets

corresponding to points (a), (b) and (c) of Article 416(1);

b) they are legally and practically readily available at any time during the next 30 days to be

liquidated via outright sale or via a simple repurchase agreement on approved repurchase

markets in order to meet obligations coming due. Liquid assets referred to in point (c) of Article

416(1) which are held in third countries where there are transfer restrictions or which are

denominated in non-convertible currencies shall be considered available only to the extent

that they correspond to outflows in the third country or currency in question, unless the

institution can demonstrate to the competent authorities that it has appropriately hedged the

ensuing currency risk;

c) the liquid assets are controlled by a liquidity management function;

d) a portion of the liquid assets except those referred to in points (a), (c) and (e) of Article 416(1)

is periodically and at least annually liquidated via outright sale or via simple repurchase

agreements on an approved repurchase market for the following purposes:

(i) to test the access to the market for these assets;

(ii) to test the effectiveness of its processes for the liquidation of assets;

(iii) to test the usability of the assets;

(iv) to minimise the risk of negative signaling during a period of stress.

e) price risks associated with the assets may be hedged but the liquid assets are subject to

appropriate internal arrangements that ensure that they are readily available to the treasury

when needed and especially that they are not used in other ongoing operations, including:

(i) hedging or other trading strategies;

(ii) providing credit enhancements in structured transactions;

(iii) covering operational costs.

The denomination of the liquid assets is consistent with the distribution by currency of liquidity outflows

after the deduction of inflows.

149

PDF to HTML - Convert PDF files to HTML files

Article 509: Liquidity requirements

(...)

By 31 January 2014, the EBA shall furthermore report on the following:

a) uniform

definitions

high

and

extremely

high

liquidity

and

credit

quality;

b) the possible unintended consequences of the definition of liquid assets on the conduct of

monetary policy operation and the extent to which:

(i) a list of liquid assets that is disconnected from the list of central bank eligible assets may

incentivise institutions to submit eligible assets which are not included in the definition of

liquid assets in refinancing operations;

(ii) regulation of liquidity may disincentivise institutions to lend or borrow on the unsecured

money market and whether this may lead to question the targeting of EONIA in monetary

policy implementation;

(iii) the introduction of the liquidity coverage requirement may make it more difficult for central

banks to ensure price stability by using the existing monetary policy framework and

instruments;

the operational requirements for the holdings of liquid assets, as referred in points (b) to (f) of

Article 417, in line with international regulatory developments.

150

PDF to HTML - Convert PDF files to HTML files

Annex 13: Operational Requirements as set out in “Basel III: The Liquidity Coverage Ratio

and the liquidity risk monitoring tools”, Jan 2013 BIS

2. Operational requirements

28. All assets in the stock of HQLA are subject to the following operational requirements. The purpose

of the operational requirements is to recognise that not all assets outlined in paragraphs 49-54 that

meet the asset class, risk-weighting and credit-rating criteria should be eligible for the stock as there

are other operational restrictions on the availability of HQLA that can prevent timely monetisation

during a stress period.

29. These operational requirements are designed to ensure that the stock of HQLA is managed in

such a way that the bank can, and is able to demonstrate that it can, immediately use the stock of

assets as a source of contingent funds that is available for the bank to convert into cash through

outright sale or repo, to fill funding gaps between cash inflows and outflows at any time during the 30-

day stress period, with no restriction on the use of the liquidity generated.

30. A bank should periodically monetise a representative proportion of the assets in the stock through

repo or outright sale, in order to test its access to the market, the effectiveness of its processes for

monetisation, the availability of the assets, and to minimise the risk of negative signalling during a

period of actual stress.

31. All assets in the stock should be unencumbered. “Unencumbered” means free of legal, regulatory,

contractual or other restrictions on the ability of the bank to liquidate, sell, transfer, or assign the asset.

An asset in the stock should not be pledged (either explicitly or implicitly) to secure, collateralise or

credit-enhance any transaction, nor be designated to cover operational costs (such as rents and

salaries). Assets received in reverse repo and securities financing transactions that are held at the

bank, have not been rehypothecated, and are legally and contractually available for the bank's use

can be considered as part of the stock of HQLA. In addition, assets which qualify for the stock of

HQLA that have been pre-positioned or deposited with, or pledged to, the central bank or a public

sector entity (PSE) but have not been used to generate liquidity may be included in the stock.

32. A bank should exclude from the stock those assets that, although meeting the definition of

“unencumbered” specified in paragraph 31, the bank would not have the operational capability to

monetise to meet outflows during the stress period. Operational capability to monetise assets requires

having procedures and appropriate systems in place, including providing the function identified in

paragraph 33 with access to all necessary information to execute monetisation of any asset at any

time. Monetisation of the asset must be executable, from an operational perspective, in the standard

settlement period for theasset class in the relevant jurisdiction.

If a bank has deposited, pre-positioned or pledged Level 1, Level 2 and other assets in a collateral pool and no

specific securities are assigned as collateral for any transactions, it may assume that assets are encumbered in

order of increasing liquidity value in the LCR, ie assets ineligible for the stock of HQLA are assigned first, followed

by Level 2B assets, then Level 2A and finally Level 1. This determination must be made in compliance with any

requirements, such as concentration or diversification, of the central bank or PSE.

151

PDF to HTML - Convert PDF files to HTML files

33. The stock should be under the control of the function charged with managing the liquidity of the

bank (eg the treasurer), meaning the function has the continuous authority, and legal and operational

capability, to monetise any asset in the stock. Control must be evidenced either by maintaining assets

in a separate pool managed by the function with the sole intent for use as a source of contingent

funds, or by demonstrating that the function can monetise the asset at any point in the 30-day stress

period and that the proceeds of doing so are available to the function throughout the 30-day stress

period without directly conflicting with a stated business or risk management strategy. For example, an

asset should not be included in the stock if the sale of that asset, without replacement throughout the

30-day period, would remove a hedge that would create an open risk position in excess of internal

limits.

34. A bank is permitted to hedge the market risk associated with ownership of the stock of HQLA and

still include the assets in the stock. If it chooses to hedge the market risk, the bank should take into

account (in the market value applied to each asset) the cash outflow that would arise if the hedge were

to be closed out early (in the event of the asset being sold).

35. In accordance with Principle 9 of the Sound Principles a bank “should monitor the legal entity and

physical location where collateral is held and how it may be mobilised in a timely manner”. Specifically,

it should have a policy in place that identifies legal entities, geographical locations, currencies and

specific custodial or bank accounts where HQLA are held. In addition, the bank should determine

whether any such assets should be excluded for operational reasons and therefore, have the ability to

determine the composition of its stock on a daily basis.

36. As noted in paragraphs 171 and 172, qualifying HQLA that are held to meet statutory liquidity

requirements at the legal entity or sub-consolidated level (where applicable) may only be included in

the stock at the consolidated level to the extent that the related risks (as measured by the legal entity’s

or sub-consolidated group’s net cash outflows in the LCR) are also reflected in the consolidated LCR.

Any surplus of HQLA held at the legal entity can only be included in the consolidated stock if those

assets would also be freely available to the consolidated (parent) entity in times of stress.

37. In assessing whether assets are freely transferable for regulatory purposes, banks should be

aware that assets may not be freely available to the consolidated entity due to regulatory, legal, tax,

accounting or other impediments. Assets held in legal entities without market access should only be

included to the extent that they can be freely transferred to other entities that could monetise the

assets.

38. In certain jurisdictions, large, deep and active repo markets do not exist for eligible asset classes,

and therefore such assets are likely to be monetised through outright sale. In these circumstances, a

bank should exclude from the stock of HQLA those assets where there are impediments to sale, such

as large fire-sale discounts which would cause it to breach minimum solvency requirements, or

requirements to hold such assets, including, but not limited to, statutory minimum inventory

requirements for market making.

152

PDF to HTML - Convert PDF files to HTML files

39. Banks should not include in the stock of HQLA any assets, or liquidity generated from assets, they

have received under right of rehypothecation, if the beneficial owner has the contractual right to

withdraw those assets during the 30-day stress period.

40. Assets received as collateral for derivatives transactions that are not segregated and are legally

able to be rehypothecated may be included in the stock of HQLA provided that the bank records an

appropriate outflow for the associated risks as set out in paragraph 116.

41. As stated in Principle 8 of the Sound Principles, a bank should actively manage its intraday

liquidity positions and risks to meet payment and settlement obligations on a timely basis under both

normal and stressed conditions and thus contribute to the smooth functioning of payment and

settlement systems. Banks and regulators should be aware that the LCR stress scenario does not

cover expected or unexpected intraday liquidity needs.

42. While the LCR is expected to be met and reported in a single currency, banks are expected to be

able to meet their liquidity needs in each currency and maintain HQLA consistent with the distribution

of their liquidity needs by currency. The bank should be able to use the stock to generate liquidity in

the currency and jurisdiction in which the net cash outflows arise. As such, the LCR by currency is

expected to be monitored and reported to allow the bank and its supervisor to track any potential

currency mismatch issues that could arise, as outlined in Part 2. In managing foreign exchange

liquidity risk, the bank should take into account the risk that its ability to swap currencies and access

the relevant foreign exchange markets may erode rapidly under stressed conditions. It should be

aware that sudden, adverse exchange rate movements could sharply widen existing mismatched

positions and alter the effectiveness of any foreign exchange hedges in place.

43. In order to mitigate cliff effects that could arise, if an eligible liquid asset became ineligible (eg due

to rating downgrade), a bank is permitted to keep such assets in its stock of liquid assets for an

additional 30 calendar days. This would allow the bank additional time to adjust its stock as needed or

replace the asset.

Refer to paragraph 146 for the appropriate treatment if the contractual withdrawal of such assets would lead to

a short position (eg because the bank had used the assets in longer-term securities financing transactions).

153

PDF to HTML - Convert PDF files to HTML files

Annex 14: Guideline on Liquidity Buffers & Survival Periods, CEBS 9 December 2009

(…)

Guideline 5 – Credit institutions need to manage their stocks of liquid assets to ensure, to the

maximum extent possible, that they will be available in times of stress. They should avoid

holding large concentrations of single securities and there should be no legal, regulatory, or

operational impediments to using these assets.

67. Depending on the structure of the asset, issuer-specific factors (such as the issuer’s credit

quality), issuance-specific factors (such as the maturity and size of the issuance), and

institutional factors (such as whether the asset is traded in centralised markets or over the

counter and whether it has a diversified investor base) can be important factors in determining

the liquidity of asset classes and whether they will remain liquid in times of stress. Investors

are more likely to regard an asset as a safe haven when the issuer’s credit quality is high; the

issuance is large; it is actively traded in organised markets and it has a diversified investor

base.

68. Concentrations of particular securities should be avoided, as a market breakdown for these

asset types could severely damage the institution’s funding capacity. Banks should seek to

diversify, for example, by issuer, maturity, and currency. The need to diversify holdings of

assets becomes greater as the liquidity of the asset becomes lower (as indicated by the above

factors). For example, it is more important to diversify a portfolio of high-quality corporate

bonds than a portfolio of high-quality government bonds. Attempts to liquidate large

concentrated positions of less liquid assets could trigger illiquidity in the market itself, with

declines in market prices (fire sales), which may force other institutions to take write-downs on

similar assets that they hold. That, in turn, could weaken the liquidity position of other banks,

prompting further asset sales and an evaporation of market liquidity, adversely affecting the

financial system as a whole.

Firms should seek to be active on a regular basis in each market in which they hold assets for

liquidity purposes. Accessing the market regularly will help to reduce the potential stigma of

firms suddenly accessing markets, alerting other firms to the fact that they may be under

liquidity pressure (in turn, causing more investors to withdraw funds, thereby accentuating the

liquidity pressure) .

In addition, as there may be legal or cross-border regulatory constraints that restrict firms’

ability to use their buffer of liquid assets at particular times, or for particular purposes, firms

should also ensure that they are aware of the specific constraints that apply in particular

jurisdictions.

To use certain funding markets (e.g., repo or securitisation), banks need to have well-

established platforms that allow them to raise more funds promptly. Setting up arrangements

from scratch typically requires significant due diligence and thus time. If such operational

Based on the proportionality principle, smaller banks which access markets through another institution, will, in

most cases, not have to be active in several advanced money and capital markets.

154

PDF to HTML - Convert PDF files to HTML files

arrangements are not in place as a matter of normal business, rapid access in stressed times

should not be relied upon.

The specification of the liquidity buffer (type and amount of assets) should also be driven by

the degree to which legal entities should be self-sufficient in terms of liquidity, taking into

account intra-group dependencies and the extent to which liquidity should be allocated to

different currencies because of potential disruptions in swap markets, etc.

Guideline 6 – The location and size of liquidity buffers within a banking group should

adequately reflect the structure and activities of the group in order to minimize the effects of

possible legal, regulatory or operational impediments to using the assets of the buffer.

The buffer should differentiate between currencies, and should reflect legal entity specificities

where appropriate, especially with regard to intra-group exposures. Determining the adequate

location and size of buffers for legal entities, jurisdictions, and regions should be responsive to

individual needs and situations. In general, several drivers of the decision process can be

identified, such as operational risk considerations, the degree of centralisation of liquidity

management, jurisdictional specificities in terms of winding up directives, deposit guarantee

schemes and local regulatory requirements, different treatment of branches and subsidiaries,

and differences in local business models, time zones and access to capital markets. A final

decision should be made and applied through the dialogue between the group and its home

and host supervisors.

There is no single model for the organisation of liquidity management: they range from fully

centralised management to the fully decentralised independent local management of liquidity.

Centralised management of the buffers may be acceptable once it has been established that

there are no impediments to the transfer of liquidity within the group and the relevant

regulators are satisfied that the ability to move funds between entities would be resilient in a

stress situation .

69. As a general principle, when an entity responsible for liquidity management has a material

holding of a currency, it, by implication, has a material level of liquidity risk in this currency and

should hold a buffer for it. The holding of several buffers may impose additional costs on

banks, but it addresses the risk of potential disruptions in the foreign exchange market that

may impair the ability to convert across currencies.

See CEBS’ Technical Advice to the European Commission on Liquidity Risk Management, September 2008,

paragraphs 94-96, for a discussion on the complexities that may arise in a banking group using centralised

liquidity management.

155

PDF to HTML - Convert PDF files to HTML files

References

Aitken, M. and Carole Comerton-Forde, 2003, ‘How should liquidity be measured?’, Pacific-Basin

Finance Journal, 11, pp. 45-59.

Amihud, Y. and Haim Mendelson, 1991, ‘Liquidity, maturity, and the yield on US treasury securities’,

Journal of Finance, 46, pp.479-486.

Amihud, Y., 2002, ‘Illiquidity and stock returns: cross-section and time-series effects’, Journal of

Financial Markets, 5, pp. 31-56.

Amihud, Y., Haim Mendelson, and Lasse Pedersen, 2005, ‘Liquidity and Asset Prices’, Foundations

and Trends in Finance, 1, pp. 269-364.

Bao, J., Jun Pan and Jiang Wang, 2011, ‘The Illiquidity of Corporate Bonds’, 66, The Journal of

Finance, pp. 911-946.

Biais, B., Fany Declerck, James Dow, Richard Portes, Ernst-Ludwig vin Thadden, 2006, ‘European

Corporate Bond Markets: transparency, liquidity, efficiency’, Centre for Economic Policy Research

(CEPR) http://faculty.london.edu/rportes/TT%20CorporateFULL2.pdf

Biais, B., Larry Glosten and Chester Spatt, 2005, ‘Market microstructure: A survey of

microfoundations, empirical results, and policy implications.’, Journal of Financial Markets, 8, pp.217-

264.

Chen, L., D.A. Lesmond and Jason Wei, 2007, ‘Corporate Yield Spreads and Bond Liquidity’, Journal

of Finance, 62, pp.119-149.

Chordia, T., Richard Roll, Subrahmanyam, A., 2000, ‘Commonality in liquidity’. Journal of Financial

Economics, 56, pp. 3–28.

Chordia, T., Asani Sarkar and Avanidhar Subrahmanyam, 2005, ‘An Empirical Analysis of Stock and

Bond Market Liquidity’, Review of Financial Studies, 18, pp. 85-129.

Degryse, H., Frank De Jong, M. V. Ravenswaaij, and Gunter Wuyts, 2005, ‘Aggressive orders and the

resiliency of a limit order market.’, Review of Finance, 9, pp. 201–242.

Dick-Nielsen, J., P. Feldhutter, and D. Lando, 2012, ‘Corporate bond liquidity before and after the on-

set of the subprime crisis’, Journal of Financial Economics, 103, 471-492.

Dick-Nielsen, J., Jacob Gyntelberg, and Thomas Sangill, 2012, ‘Liquidity in Government versus

Covered Bond Markets’, Danmarks Nationalbank Working Papers, n.83.

Dufour, A. and Robert F. Engle, 2000, ‘Time and the Price Impact of a Trade’, 55, pp. 2467-2498.

156

PDF to HTML - Convert PDF files to HTML files

Duffie, D., Nicolae Gârleanu and Lasse Heje Pedersen, 2007, ‘Valuation in Over-the-Counter

Markets’, Review of Financial Studies, 20, pp.1865-1900.

Dunne, P., Michael Moore and Richard Portes, 2006, ‘European Government Bond Markets:

transparency, liquidity, efficiency’, Center for Economic Policy Research (CEPR).

Edwards, A. K., Lawrence E. Harris and Michael S. Piwowar, 207, ‘Corporate Bond Market

Transaction Costs and Transparency’, LXII, pp. 1421-1451.

Ellis, K., Roni Michaely and Maureen O'Hara, 2000, ‘The Accuracy of Trade Classification Rules:

Evidence from Nasdaq’, Journal of Financial and Quantitative Analysis, 35, pp 529-551.

Europe Economics (2013), ‘The liquidity of gold under proposed EBA liquidity tests’.

http://www.eba.europa.eu/documents/10180/42030/World-Gold-Council_1.pdf.

Ferguson, M. F. and Steven C. Mann, 2001, ‘Execution Costs and Their Intraday Variation in Futures

Markets’, Journal of Business, 74, pp. 125-160.

Fong, K., Craig W. Holden, Charles A. Trzcinka, 2011, ‘What Are The Best Liquidity Proxies For

Global Research?’, Working Paper, Available at SSRN: http://ssrn.com/abstract=1558447.

Fleming, M. 2003, ‘Measuring Treasury Market Liquidity’, FRBNY Economic Policy Review.

http://www.newyorkfed.org/research/epr/03v09n3/0309flem.pdf

Glosten, L. R. and Paul R. Milgrom, 1985, ‘Bid, Ask and Transaction Prices in a Specialist Market with

Heterogenously Informed Traders’, Journal of Financial Economics, 14, 1985, pp. 71-100.

Goldstein, M. A., Edith S. Hotchkiss and Erik R. Sirri, 2007, ‘Transparency and Liquidity: A Controlled

Experiment on Corporate Bonds’, Review of Financial Studies, 20, 235-273.

Goyenko, R., Avanidhar Subrahmanyam, and Andrey Ukhov, 2011, ‘The Term Structure of Bond

Market Liquidity and Its Implications for Expected Bond Returns’, Journal of Financial and Quantitative

Analysis, 46, 111-139.

Harris, L., 1990, Liquidity, ‘Trading Rules, and Electronic Trading Systems’, New York University

Salomon Center Monograph Series in Finance, Monograph # 1990-4.

Hasbrouck, J., 1991, ‘Measuring the Information Content of Stock Trades’, Journal of Finance, 46,

pp.179-207.

Hasbrouck, J., 2009, ‘Trading Costs and Returns for U.S. Equities: Estimating Effective Costs from

Daily Data’, 2009, 64, pp. 1445-1477.

Hasbrouck, J., Duane J. Seppi, 2001, ‘Common factors in prices, order flows, and liquidity. Journal of

Financial Economics, 59, pp. 383–411.

157

PDF to HTML - Convert PDF files to HTML files

Houweling, P., Albert Mentink and Ton Vorst, 2005, ‘Comparing possible proxies of corporate bond

liquidity’, Journal of Banking and Finance, 29, pp. 1331-1358.

Huang R. D., and H. R. Stoll, 1997, ‘The components of the bid/ask spread: A general approach’, The

Review of Financial Studies 10, pp. 995-1034.

Keynes, J. M., 1930, Treatise on Money, Vol. II, p. 67 (Macmillan, London).

Korajczyk, R. A. and Ronnie Sadka, 2008, ‘Pricing the commonality across alternative measures of

liquidity’, Journal of Financial Economics, 87, pp. 45-72.

Large, J., 2007, ‘Measuring the Resiliency of an Electronic Limit Order Book’, Journal of Financial

Markets, 10, pp 1-25.

Lee, C. M.C. and Mark J. Ready, 1991, ‘Inferring trade direction from intraday data’,

Journal of Finance 46, 733-747.

Lesmond, D.,Ogden,J.,Trzcinka,C.,1999. ‘A new estimate of transaction costs.’, Review of Financial

Studies, 12, pp. 1113–1141.

Marshall, B. R., Nhut H. Nguyen and Nuttawat Visaltanachoti, 2013, ‘Commodity Liquidity

Measurement and Transaction Costs’. The Review of Financial Studies, 25, pp. 599–638.

Næs, R., Johannes A. Skjeltorp and Bernt Arne Ødegaard, 2011, ‘Stock Market Liquidity and the

Business Cycle’, Journal of Finance, 66, pp.139-176.

O’Hara, M., 1995, ‘Market Microstructure Theory’, Blackwell Publishers, Cambridge.

Pastor, L. and Robert F. Stambaugh, 2003, ‘Liquidity Risk and Expected Stock Returns,’ Journal of

Political Economy, 111, pp. 642–685.

Perold, Andre F., 1988, ‘The implementation shortfall: paper vs. reality’, Journal of Portfolio

Management, 14, pp. 4–9.

Schultz, P., 2001, ‘Corporate Bond Trading Costs and Practice: A Peek behind the

Curtain’, The Journal of Finance, 56, pp 677-698.

Silber, W., 1991, ‘Discounts on restricted stock: The impact of Liquidity on Stock Prices’, Financial

Analyst Journal, pp. 66-64.

Goyenko, R., Avanidhar Subrahmanyam and Andrey Ukhov, 2011, ‘The Terms Structure of Bond

Market Liquidity and Its Implications for Expected Returns’, Journal of Financial and Quantitative

Analysis, 46, pp. 111-139.

158

PDF to HTML - Convert PDF files to HTML files

Vayanos, D. and Jiang Wang, 2012, ‘Market Liquidity: Theory and Empirical Evidence’, Handbook of

the Economics of Finance (forthcoming).

Wagner, W., 2011, ‘Systemic Liquidation Risk and the Diversity–Diversification Trade-Off.’,

Journal of Finance, 66: 1141–1175

The

159