Erhvervs-, Vækst- og Eksportudvalget 2016-17
ERU Alm.del Bilag 167
Offentligt
R
EPORT ON THE MONITORING EXERCISE
CARRIED OUT IN THE ONLINE HOTEL
BOOKING SECTOR BY
EU
COMPETITION
AUTHORITIES IN
2016
T
HE PARTICIPATING AUTHORITIES ARE THE
B
ELGIAN
, C
ZECH
,
F
RENCH
, G
ERMAN
, H
UNGARIAN
, I
RISH
, I
TALIAN
, D
UTCH
, S
WEDISH
AND
UK
NATIONAL COMPETITION AUTHORITIES AND
DG C
OMPETITION
 ERU, Alm.del - 2016-17 - Bilag 167: Notat om bookingportaler, fra erhvervsministeren
1745709_0002.png
Contents
Introduction
........................................................................................................................................... 4
1
Methodology ............................................................................................................................... 5
2
Key findings ................................................................................................................................ 6
3
Limitations of the results ............................................................................................................. 8
4
The results in full ....................................................................................................................... 10
4.1
Hotelier awareness of the changes to OTA parity clauses ................................................ 10
4.2
Room price differentiation between sales channels .......................................................... 10
4.3
Room availability differentiation between sales channels ................................................ 15
4.4
OTA commission rates ..................................................................................................... 17
4.5
Relative importance of hotels' main sales channels .......................................................... 18
4.6
Use of OTAs by hotels...................................................................................................... 19
4.7
OTA conversion rates ('look-to-book' ratios) ................................................................... 19
4.8
OTA preferred partner programs ...................................................................................... 20
4.9
Scope of parity clauses ..................................................................................................... 20
4.10 Hotel customer loyalty schemes ....................................................................................... 21
4.11 New entrants to the online hotel booking sector ............................................................... 21
4.12 Use of metasearch sites by hotels ..................................................................................... 22
4.13 OTA best price guarantees ................................................................................................ 22
4.14 Expected market developments ........................................................................................ 22
4.15 Other issues raised by stakeholders .................................................................................. 23
5
Submissions received from Booking.com and Expedia ............................................................ 23
5.1
Booking.com's submission ............................................................................................... 23
5.2
Expedia's submission ........................................................................................................ 25
Appendix 1: Econometric Analysis of Price Differentiation between OTAs
................................. 27
1. Introduction ............................................................................................................................... 27
2. The metasearch data .................................................................................................................. 27
2.1
The sample of hotels ......................................................................................................... 27
2.2
The data for hotels from the sample on price and other variables .................................... 28
3.
Analysis of the metasearch data ................................................................................................ 30
3.1
Definition of price differentiation ..................................................................................... 30
3.2
Preliminary information on effects ................................................................................... 30
3.3
The econometric model .................................................................................................... 32
3.4
Results
–
prices from all OTAs included .......................................................................... 34
3.5
Results
–
prices from Booking.com, Expedia and HRS only ........................................... 35
3.6
Remaining specifications .................................................................................................. 36
Insights from the data scraped from Booking, Expedia and HRS ............................................. 37
4.1
Data collection process ..................................................................................................... 37
4.2
Price differentiation and product differentiation............................................................... 38
4.
Appendix 2: Electronic survey of hotels
–
composition of samples and response rates
................ 42
2
 ERU, Alm.del - 2016-17 - Bilag 167: Notat om bookingportaler, fra erhvervsministeren
1745709_0003.png
List of Figures
Figure 1. Reasons for not price differentiating between OTAs ............................................................. 11
Figure 2. Reasons for price differentiating between OTAs ................................................................... 12
Figure 3. Reasons for not price differentiating in favour of an OTA vs hotel website ......................... 15
Figure 4. Reasons for not differentiating between OTAs for room availability .................................... 16
Figure 5. Dynamics of price/product differentiation between OTAs in the Participating Member States
and Canada ............................................................................................................................. 31
Figure 6. Price differentiation between OTAs in the scraped dataset .................................................. 39
Figure 7. Presence of room or condition differences given a price difference ..................................... 40
Figure 8. Presence of room or condition differences given no price difference ................................... 41
List of Tables
Table 1.
Table 2.
Table 3.
Table 4.
Share of bookings per sales channel, per hotel category and per year of observation ........... 19
Number of observations in the MSS data ............................................................................... 29
Results from the difference-in-differences model based on prices from all OTAs ................ 34
Results from the difference-in-differences model based on prices from Booking.com,
Expedia and HRS ................................................................................................................... 35
Table 5. Size of the scraped dataset...................................................................................................... 38
Table 6. Non-standard room type indications ...................................................................................... 39
3
 ERU, Alm.del - 2016-17 - Bilag 167: Notat om bookingportaler, fra erhvervsministeren
1745709_0004.png
Introduction
1) This report presents the results of a coordinated monitoring exercise carried out in the
online hotel booking sector by a group of eleven EU competition authorities in 2016.
1
The exercise was commissioned by the heads of the European Competition Network ('the
ECN') in November 2015. The purpose was to measure the effects of recent changes to
the parity clauses used by online travel agents ('OTAs') in their contracts with hotels. The
results of the exercise were taken into account by the heads of the ECN in their recent
discussion on future action in this sector. The conclusions of their discussion can be
found at:
http://ec.europa.eu/competition/antitrust/ECN_meeting_outcome_17022017.pdf.
2) By way of background, it should be recalled that since 2010 several national competition
authorities ('NCAs') have investigated OTA parity clauses,
2
and that these NCAs have
adopted differing approaches. Germany's Bundeskartellamt has pursued a prohibition
approach, whereas the French, Italian and Swedish NCAs pursued a commitments
approach.
3,4
The Bundeskartellamt prohibited the parity clause used by HRS (a major
German OTA) in December 2013. In April 2015, Booking.com committed to the French,
Italian and Swedish competition authorities to change its 'wide' parity clause to a 'narrow'
parity clause.
5
Booking decided to apply this change EU-wide from July 2015, and
Expedia also decided to apply a narrow parity clause EU-wide, from August 2015. In
December 2015, the Bundeskartellamt prohibited Booking.com's narrow parity clause in
Germany.
6,7
In addition to these antitrust measures, in France, in August 2015, the so-
1
The group consisted of the Belgian, Czech, French, German, Hungarian, Irish, Italian, Dutch, Swedish and UK
national competition authorities and DG Competition. In this report, it is referred to as 'the monitoring
working group'. The Austrian and Swiss NCAs also participated during the design phase of the monitoring
exercise.
Also known as 'Most Favoured Nation' or 'MFN' clauses.
The OFT (now CMA) also investigated the online hotel booking sector between 2010 and 2014, though its case
focussed not on OTA parity clauses but on restrictions on the ability of OTAs to offer discounted room prices.
The Irish NCA also subsequently accepted commitments from Booking.com modelled on those agreed by the
French, Italian and Swedish NCAs.
In brief, 'wide' parity clauses oblige the hotel to give the OTA the lowest room prices and best room
availability relative to all other sales channels, whereas 'narrow' parity clauses allow the hotel to offer lower
room prices and better room availability on other OTAs and on offline sales channels, but allow the OTA to
stop the hotel from publishing lower room prices on the hotel's own website.
Booking.com's appeal against the prohibition decision is pending before the German courts.
Expedia continues to apply its narrow parity clause in Germany. The Bundeskartellamt's investigation of
Expedia's clause continues.
2
3
4
5
6
7
4
 ERU, Alm.del - 2016-17 - Bilag 167: Notat om bookingportaler, fra erhvervsministeren
1745709_0005.png
called
Loi Macron
8
rendered null and void all OTA price parity clauses and, in Austria, in
November 2016, all OTA parity clauses were rendered null and void by an amendment to
the law on unfair competition.
9,10
1 Methodology
3) The monitoring exercise covered various aspects of the way hotels market and sell their
rooms, but focused on parameters which had been central to the theories of harm applied
by the NCAs which have investigated the use of parity clauses by OTAs. The theory of
harm for wide parity clauses in this sector is, first, that they lead to a softening of
competition between incumbent OTAs and, second, that they foreclose entry or
expansion by new or smaller OTAs. More specifically, the clauses reduce the incentives
for OTAs to compete on the conditions they offer to hotels, including the commission
rates they charge. Because a wide parity clause obliges the hotel to offer the same room
price on all the OTAs it uses, the hotel cannot 'reward' an OTA which charges a lower
commission rate, by giving it lower room prices, nor can it penalize an OTA which
charges a higher commission rate, by giving that OTA higher room prices. Because wide
parity clauses also apply to room availability, the same disincentives apply to competition
between OTAs on this parameter. The theory of harm for narrow parity clauses in this
sector is that they have the effect of preserving the restriction of competition caused by
wide parity, because they reduce the incentive for hotels to offer differing room prices on
different OTAs.
11
4) In light of these theories of harm, the monitoring exercise focused on:
i)
room price differentiation by hotels between sales channels;
ii)
room availability differentiation by hotels between sales channels, and
iii)
OTA commission rates.
12
8
Article 133 of the Loi n
o
2015-990 du 6 août 2015 pour la croissance, l'activité et l'égalité des chances
économiques
99. Bundesgesetz: Änderung des Bundesgesetzes gegen den unlauteren Wettbewerb 1984
–
UWG und des
Preisauszeichnungsgesetzes.
Similar legislation is being debated in the Italian parliament.
Under a narrow parity clause, if a hotel wishes to price differentiate between its OTA partners, the hotel is
obliged to offer a room price on its own website that is higher than the price it offers on at least one of the
OTAs it uses.
It was recognized that commission rates are only one of the parameters on which OTAs compete, however
they have the advantage of being objectively measurable.
9
10
11
12
5
 ERU, Alm.del - 2016-17 - Bilag 167: Notat om bookingportaler, fra erhvervsministeren
1745709_0006.png
5) The monitoring consisted of a uniform electronic questionnaire addressed to a sample of
16,000 hotels in the ten participating Member States.
13
Written questionnaires were also
sent to a sample of OTAs, metasearch websites and large hotel chains.
14
The
questionnaires covered the period from January 2013 to June 2016,
15
with particular
focus on the period before and after the switch by Booking.com and Expedia from wide
to narrow parity clauses (mid-2015). The questionnaires were sent during July and
August 2016, and the replies were received by the end of September 2016.
6) In addition, hotel room price data was obtained from one or more major metasearch
websites and from the websites of Booking.com, Expedia and HRS.
16
2 Key findings
7) The monitoring exercise produced the following key findings, which are subject to the
important limitations set out in paragraph 15 below. These findings and the other results
are presented in more detail in Section 4.
Stakeholder awareness of the changes to OTA parity clauses
8) 47% of the hotels that responded to the electronic survey across the ten participating
Member States said that they did not know that Booking.com and Expedia had recently
changed or removed their parity clauses. This figure was lower in France and Germany
taken together (30%). Of those hotels that knew about the changes, the majority said they
had not acted upon them in any way.
Room price differentiation between OTAs
9) 79% of the hotels that responded to the electronic survey across the ten participating
Member States said that they had not price differentiated between OTAs in the period
13
The sample was selected to be representative of the general population of hotels listed on at least two major
OTAs in each Member State, in terms of number of rooms, star category and membership or not of a chain.
Replies were received from 1600 hotels, giving a response rate of around 12 percent. Details of the sample
used and a breakdown of the responses received for each Member State are given in Appendix 2 of this
report.
Questionnaires were sent to 20 OTAs, 19 hotel chains (including the ten largest chains in the EU) and 11
metasearch websites. Replies were received from 5 OTAs, 13 hotel chains and 7 metasearch sites.
This period captures the effects of the Bundeskartellamt's two prohibition decisions against OTA parity
clauses (HRS in December 2013 and Booking.com in December 2015), as well as the 'narrow parity'
commitments decisions of the French, Italian and Swedish NCAs addressed to Booking.com (April 2015,
implemented July 2015).
Only the metasearch data covered the period before and after the switch to narrow parity clauses.
14
15
16
6
 ERU, Alm.del - 2016-17 - Bilag 167: Notat om bookingportaler, fra erhvervsministeren
1745709_0007.png
since Booking.com and Expedia switched from wide to narrow parity clauses.
17
The
reasons most frequently given for not price differentiating were that the hotel saw no
reason to treat its OTA partners differently; the hotel's OTA contract did not allow it to
price differentiate; fear of penalization by OTAs to which the hotel did not give the
lowest price; the difficulty of managing different prices on different OTAs, and not
wanting the hotel's website to appear as more expensive than the OTAs.
18
10) For the 21% of respondents that did price differentiate between OTAs, the most frequent
reason given was to increase the hotel's visibility on a particular OTA (for example, its
display ranking). In France and Germany taken together, a higher share of respondents
(27%) said that they had price differentiated between OTAs, however this difference was
not confirmed by pricing data scraped by the monitoring working group from OTA
websites, which showed no significant variation between any of the participating Member
States.
19
11) In addition to the electronic survey of hotels, the monitoring working group analyzed
room price data from one or more major metasearch websites. The analysis suggests
that:
20
a)
the switch from wide to narrow parity clauses by Booking.com and Expedia
21
led to
an increase in room price differentiation between OTAs by hotels in eight of the ten
participating Member States;
22
b)
the switch from wide to narrow parity clauses and the entry into force of the
Loi
Macron
23
led to an increase in room price differentiation between OTAs by hotels in
France;
c)
the prohibition of Booking.com's narrow parity clause
24
led to an increase in room
price differentiation between OTAs by hotels in Germany.
25
17
Hotels were not asked whether they had price differentiated between OTAs in the period before Booking.com
and Expedia switched to narrow parity clauses because
(i)
the previous wide parity clauses prohibited this,
and
(ii)
the participating authorities had concerns about the reliability of replies from independent hoteliers in
respect of behaviour pre-dating 2015. However, evidence from the NCA antitrust cases suggests that many
hotels did not fully comply with their parity obligations under wide parity.
See paragraph 19) below.
See paragraph 19) below.
Difference-in-differences analysis of room price data from a major metasearch website.
Booking.com and Expedia implemented the switch to narrow parity clauses EU-wide in July and August
2015.
Belgium, Czech Republic, Germany, Hungary, Ireland, Italy, Sweden, United Kingdom.
The analysis does not distinguish between these two measures, as they took effect almost simultaneously
(July-August 2015).
18
19
20
21
22
23
7
 ERU, Alm.del - 2016-17 - Bilag 167: Notat om bookingportaler, fra erhvervsministeren
1745709_0008.png
Room availability differentiation between OTAs
12) 69% of the hotels that responded to the electronic survey across the ten participating
Member States said that they had not differentiated between OTAs for room availability
since the switch by Booking.com and Expedia to narrow parity clauses. The reason most
frequently given was that they saw no reason to treat their OTA partners differently. In
France and Germany taken together, slightly more hotels said that they had differentiated
between OTAs for room availability (37%, relative to 31% for all ten Member States).
Irrespective of Member State, more than 80% of hotels said that they had not changed
their behaviour as regards differentiating between OTAs for room availability since the
switch to narrow parity clauses.
OTA commission rates
13) 90% of hotels that responded to the electronic survey said that there had been no change
in the basic commission rate charged to them by OTAs in the period from July 2015 to
June 2016. According to evidence obtained from certain OTAs, the average
effective
rate
of commission
26
paid by hotels remained relatively stable or slightly decreased in the
period from January 2014 to June 2016.
27
14) Very few hotels said that they engaged in trade-offs whereby they grant OTAs more
favourable room prices or room availability in return for a lower commission rate.
28
3 Limitations of the results
15) The following caveats apply to the results of the monitoring exercise:
a)
in several Member States, the number of hoteliers that responded to the electronic
questionnaire was low and/or hotels belonging to chains were over-represented;
b)
inconsistencies in some of the replies suggest that hoteliers may not have fully
understood some of the questions;
c)
comparisons between Member States based on the replies to the electronic
questionnaire should be treated with caution, as they do not take into account
The Bundeskartellamt's prohibition decision was adopted in December 2015 and took effect in February 2016.
Various data limitations do not allow for a robust comparison of these results between Member States, as
explained in paragraph 22) below.
Basic commission plus optional additional commission paid in return for increased visibility or other benefits
See paragraphs 32) and 33) below.
3% of hotels that responded to the electronic survey across the ten participating Member States said that they
had traded lower room prices for a lower commission rate. Likewise, 3% of respondent hotels said that they
had traded better room availability for a lower commission rate.
24
25
26
27
28
8
 ERU, Alm.del - 2016-17 - Bilag 167: Notat om bookingportaler, fra erhvervsministeren
1745709_0009.png
possible composition effects (for example, the proportion of independent/chain
hotels differs between Member States);
d)
the results relating to room price differentiation are based on three sources: (i) replies
from hoteliers, (ii) room price data from one or more metasearch websites, and (iii)
room price data from OTA websites. As regards the replies from hoteliers, it should
be noted that these related to behaviour dating back twelve months, and that hoteliers
or their associations have been party to the national antitrust cases and/or to private
litigation relating to OTA parity clauses. As regards the metasearch pricing data, this
was the only data that covered the period before and after the switch to narrow parity
clauses, however the data does not distinguish between 'true' price differentiation and
differentiation resulting from differences between the products offered by the OTAs
on the metasearch website (for example, different categories of room, inclusion of
breakfast, differing cancellation rights);
e)
France and Germany cannot be considered as jurisdictions where OTAs no longer
apply parity clauses, since, in Germany, one of the big three OTAs (Expedia)
continues to apply a narrow parity clause
29
, and in France, there remain questions
about the scope and enforcement of the OTA provisions of the
Loi Macron.
30
Nor
should the other participating Member States be considered as jurisdictions where the
only OTA parity clauses still in force are narrow parity clauses: in these Member
States, smaller OTAs such as HRS continue to impose wide parity clauses on their
partner hotels;
f)
the monitoring exercise was carried out twelve months after Booking.com and
Expedia switched to narrow parity clauses and six months after the adoption of the
most recent prohibition decision in Germany (against Booking.com).
31
On the one
hand, this time span made data collection easier. On the other hand, it is possible that
the sector might not yet have fully adapted to the changes made to the major OTAs'
parity clauses. Furthermore, the terrorist attacks that occurred in France and Belgium
during the monitoring period may have affected the results relative to more typical
tourist seasons.
29
30
The German Bundeskartellamt's investigation of Expedia's narrow parity clause continues.
Article 133 of the
Loi Macron
renders null and void all OTA price parity clauses, but this provision relies on
private enforcement. Furthermore, it is unclear whether Article 133 applies to price parity agreed voluntarily
by hotels, for example as a condition of membership of an OTA's preferred partner program.
The German Bundeskartellamt's decision prohibiting HRS's parity clause was adopted in December 2013.
31
9
 ERU, Alm.del - 2016-17 - Bilag 167: Notat om bookingportaler, fra erhvervsministeren
1745709_0010.png
16) In the sections below, the results of the monitoring exercise are set out in greater detail.
4 The results in full
4.1
Hotelier awareness of the changes to OTA parity clauses
the recent changes to the parity clauses of the major OTAs (i.e. the switch from wide to
narrow parity clauses by Booking.com and Expedia and the prohibition of the parity
clauses of Booking.com and HRS in Germany), and whether hoteliers had changed their
behaviour as a result. Across the ten participating Member States, 47% of the hotels that
responded to the electronic survey said that they were not aware that certain OTAs had
changed or removed their parity clauses. Of those that were aware of the changes, 60%
said that they had not acted upon them in any way. One quarter of respondents said that
they had changed their behaviour in relation to room pricing and one tenth in relation to
room availability. In France and Germany, awareness of the changes to parity clauses was
higher (70%).
4.2
Room price differentiation between sales channels
17) The purpose of measuring this parameter was to find out whether hoteliers knew about
Room price differentiation between OTAs
18) The remedies adopted in the antitrust investigations of OTA parity clauses (resulting in
some OTAs switching to narrow parity clauses or complete prohibition of parity clauses)
were intended to promote competition between OTAs, by allowing hotels the possibility
to offer differing room prices to different OTAs (a practice which was prohibited by wide
parity clauses). The purpose of monitoring this parameter was therefore to assess the
extent to which hotels now price differentiate between their OTA partners and the reasons
why hotels do or do not do this.
32
19) Across the ten participating Member States, 79% of the hotels that responded to the
electronic survey said that they had not price differentiated between OTAs in the period
since Booking.com and Expedia switched from wide to narrow parity clauses. The
reasons most frequently given for not doing so were that the hotel saw no reason to treat
its OTA partners differently; the hotel's OTA contract did not allow it to price
differentiate; fear of penalization by OTAs to which the hotel did not give the lowest
price (for example, less favourable display or loss of preferred partner status); the
32
Although the remedies adopted by the NCAs were intended to enable hotels to price differentiate between
OTAs, an absence of price differentiation by hotels between OTAs does not, in itself, indicate a lack of
competition between OTAs.
10
 ERU, Alm.del - 2016-17 - Bilag 167: Notat om bookingportaler, fra erhvervsministeren
1745709_0011.png
 ERU, Alm.del - 2016-17 - Bilag 167: Notat om bookingportaler, fra erhvervsministeren
1745709_0012.png
 ERU, Alm.del - 2016-17 - Bilag 167: Notat om bookingportaler, fra erhvervsministeren
1745709_0013.png
showed a positive effect on price differentiation between OTAs in Germany following
the prohibition of Booking.com's narrow parity clause.
39
The effects are in general
statistically significant for the categories 'all hotels' and 'chain hotels', but not for
independent hotels. The results are robust to different choices of specification regarding
selection of OTAs and treatment of outliers.
22) The results of the difference-in-differences analysis should be treated with caution. First,
the metasearch data does not allow a distinction between 'true' price differentiation and
price differentiation caused by product differentiation. This may occur, for example,
where one OTA offers a non-refundable room without breakfast, whereas a second OTA
offers the same room including breakfast and/or the right to cancel, or even offers a
different category of room. Indeed, an analysis of pricing data scraped by the monitoring
working group from the websites of the big three OTAs indicated that in 45-55% of cases
where differing prices are observed, there is also a difference in room type and/or
booking conditions.
40
It should be noted that when Booking.com and Expedia switched
from wide to narrow parity clauses, the ability of hotels to product differentiate between
OTAs increased at the same time as their ability to price differentiate between OTAs.
41
As the relative importance of price and product differentiation may differ between
Member States and may have developed differently over time, the analysis does not allow
for robust before-and-after comparisons or comparisons between Member States. Second,
the magnitude of the effect observed in Germany after the prohibition of Booking.com's
narrow parity clause may be influenced by the fact - already mentioned - that one major
OTA (Expedia) continues to apply narrow parity clauses in Germany.
42
Room price differentiation between hotel websites and OTAs
23) The purpose of measuring this parameter was, first, to determine the extent to which
hotels which are not subject to narrow parity - whether as a result of antitrust prohibition
decisions or national legislation - publish lower room prices on their own website than
the prices they offer on OTAs. Second, in Member States where Booking.com and
39
The effect occurs after Booking.com actually disapplied its parity clause in February 2016, pursuant to the
Bundeskartellamt's prohibition decision of December 2015.
The methodology of this analysis is described in Section 4.1 of Appendix 1 of this report.
Narrow parity removes all parity obligations in respect of room availability: the hotel may offer more
favourable room availability to particular OTAs or on its own website.
The magnitude of the effects of both treatments measured in Germany (the switch to narrow parity and the
prohibition of Booking.com's narrow parity clause) may also be influenced by the prohibition of the parity
clause of the then largest OTA in Germany, HRS, in December 2013, i.e. before the start of the period for
which observations were obtained.
40
41
42
13
 ERU, Alm.del - 2016-17 - Bilag 167: Notat om bookingportaler, fra erhvervsministeren
1745709_0014.png
Expedia apply narrow parity clauses, this parameter indicates whether hotels comply with
their obligation not to undercut their OTA partners. It should be noted that hotels were
not asked whether they already price differentiated - either between OTAs or between
their website and OTAs - before Booking.com and Expedia switched from wide to
narrow parity clauses. Therefore, it is possible that some hotels already price
differentiated while they were subject to wide parity clauses, even though both these
forms of price differentiation were prohibited by wide parity clauses.
24) Across the ten participating Member States, 40% of the hotels that responded to the
electronic survey said that, in the period since Booking.com and Expedia switched from
wide to narrow parity clauses, they had undercut their OTA partners by publishing lower
room prices on their hotel website. 57% of these hotels did so most of the time.
43
Both of
these shares were higher for France and Germany taken together: 59% of hotels undercut
OTAs and 74% of these did so most of the time. In the eight participating Member States
where Booking.com and Expedia apply narrow parity clauses (Belgium, the Czech
Republic, Hungary, Ireland, Italy, the Netherlands, Sweden, and the UK), 35% of the
respondent hotels said that they undercut their OTA partners by publishing lower prices
on their hotel website (even though narrow parity does not allow this). 48% of these
hotels did so most of the time. The three most frequent reasons given by hotels for not
price differentiating in favour of their own website relative to OTAs were, in descending
order of frequency: fear of penalization by OTAs; the practice was not permitted by one
or more OTA partners, and such price differentiation was too difficult to manage.
25) Hotels were also asked whether, in the period since the implementation of narrow parity
clauses, they had offered lower room prices on at least one OTA relative to the prices
published on their hotel website.
44
Across the ten participating Member States, 80% of
hotels that responded to the electronic survey said that they had not done so. This share
was the same in France and Germany taken together. The reasons most frequently given
by respondent hoteliers were: that they did not want their hotel website to be more
expensive than an OTA; that they did not want to divert sales from the hotel's direct
43
44
In this report, "most of the time" means more than half of the time.
It should be recalled that, under a narrow parity clause, if a hotel wishes to price differentiate between its
OTA partners, the hotel is obliged to offer a room price on its own website that is higher than the price it
offers on at least one of the OTAs it uses. The purpose of asking this question was therefore again to test
hotels' willingness to make use of the greater flexibility afforded by narrow parity clauses compared to wide
parity clauses.
14
 ERU, Alm.del - 2016-17 - Bilag 167: Notat om bookingportaler, fra erhvervsministeren
1745709_0015.png
 ERU, Alm.del - 2016-17 - Bilag 167: Notat om bookingportaler, fra erhvervsministeren
1745709_0016.png
 ERU, Alm.del - 2016-17 - Bilag 167: Notat om bookingportaler, fra erhvervsministeren
1745709_0017.png
Room availability differentiation between hotel websites and OTAs
29) Both antitrust remedies (narrow parity and prohibition of parity clauses) also allow hotels
to differentiate in favour of their own website for room availability, a practice which was
prohibited by wide parity clauses.
48
For example, a hotel may reserve a certain category
of rooms to its own website, or offer no rooms on one or more OTAs, while still making
rooms available on its own website. The purpose of monitoring this parameter was
therefore to determine the extent to which hotels make use of this possibility. Across the
ten participating Member States, 30% of the hotels that responded to the electronic
survey said that for at least some of the time they offer no rooms on OTAs while still
offering rooms on their website. Again, this figure was higher in France and Germany
taken together (44%). However, more than three quarters of respondent hotels said that
they had not changed their practice in this respect since Booking.com and Expedia
switched to narrow parity clauses. Again, this suggests that some hotels did not comply
with the previous wide parity clause in respect of room availability.
49
30) As regards large hotel chains, almost none of the respondents to the monitoring
questionnaire said that they offer different room availability to different OTAs, and only
one third differentiate for room availability in favour of their own website. The main
reasons given were that they have no incentive to differentiate and that they feared
penalization by the OTAs (e.g. a reduction in visibility).
31) OTAs said that they use several means to incentivise hotels to give them favourable room
availability. These include membership of preferred partner programs, better visibility
(for example, availability is a factor in some OTAs' ranking algorithms) and the display
of 'quality' seals, which may also affect how hotels are filtered in the OTA's search
results.
4.4
OTA commission rates
was that these clauses reduce the incentive for OTAs to compete on the commission rates
they charge to hotels. While recognizing that OTA commission rates are likely to be
affected by a variety of factors, the monitoring working group assessed whether these
rates have changed in the period before and after the recent changes to the parity clauses
32) Part of the theory of harm applied in the national investigations of OTA parity clauses
48
In this respect, narrow parity treats room availability differently from room pricing: narrow parity clauses do
not allow the hotel to differentiate in favour of its own website for the room price.
Wide parity clauses obliged the hotel to observe room availability parity across all sales channels.
49
17
 ERU, Alm.del - 2016-17 - Bilag 167: Notat om bookingportaler, fra erhvervsministeren
1745709_0018.png
of the major OTAs. The basic commission rates of the three major OTAs range from ten
percent to above twenty percent. 90% of hotels that responded to the electronic survey
said that there had been no change in the basic commission rate charged to them by
OTAs in the period from July 2015 to June 2016. Depending on the OTA, basic
commission rates can vary between Member States and/or between cities or areas within
a Member State. The responses to the monitoring questionnaires indicate that OTAs are
paid solely on the basis of a share of each booking; they do not charge fixed fees. As
already stated, only 3% of the hotels that responded to the electronic survey said that they
had granted OTAs more favourable room prices in return for a lower commission rate in
the period since Booking.com and Expedia switched from wide to narrow parity clauses.
Likewise, 3% of respondent hotels said that they had traded better room availability in
return for a lower commission rate.
50
33) In addition to basic commission, some OTAs also charge hotels additional commission in
return for providing the hotel with optional additional services, such as better visibility on
the OTA's website or membership of a preferred partner program. According to evidence
obtained from certain OTAs, the average effective rates of commission paid by hotels
(basic commission plus optional additional commission) remained relatively stable or
slightly decreased in almost all participating Member States in the period from January
2014 to June 2016.
51
This was also true for France and Germany.
4.5
Relative importance of hotels' main sales channels
the parity clauses of the major OTAs have led to changes in the share of room sales made
by hotels through each of their main distribution channels (offline sales
52
, sales via
OTAs, hotel website sales).
35) According to the results from the electronic survey of hotels, offline sales still accounted
for the largest share of hotel room sales in 2016. Sales via OTAs accounted for the
second largest share and direct online sales (hotel website) took third place. Hotels
34) The purpose of monitoring this parameter was to determine whether the recent changes to
50
These results do not exclude the possibility that some hotels take into account already existing differences
between OTA commission rates when they make decisions on room pricing and room allocation between
OTAs.
Source: questionnaires to OTAs. As some OTAs charge different commission rates for different areas within
each Member State, effective commission rates can be influenced by changes in the composition of the hotels
listed on the OTA (for example, the share of partner hotels located in areas where the OTA applies
higher/lower commission rates).
Including telephone, email, walk-in, corporate and package bookings.
51
52
18
 ERU, Alm.del - 2016-17 - Bilag 167: Notat om bookingportaler, fra erhvervsministeren
1745709_0019.png
belonging to chains generally relied less on OTAs than was the case for independent
hotels (see Table 1).
Table 1. Share of sales per channel, per hotel category and per year of observation
All hotels
YEAR
Offline
2013
2014
2015
H1 2016
54
50
47
45
Chain hotels
Direct
online
13
13
14
14
Independent hotels
Direct
online
14
14
15
17
OTAs
33
36
40
41
Offline
58
55
52
48
OTAs
28
31
33
35
Offline
53
49
46
44
OTAs
35
38
41
42
Direct
online
12
13
13
13
Source:
Replies to the electronic hotel survey (all ten Member States).
36) In France and Germany taken together, the share of sales made offline in 2016 was higher
(52%) and the share of sales made via OTAs was lower (36%), compared to the ten
Member States as a whole.
37) Over the period covered by the monitoring exercise, the share of sales made offline
decreased by nine percentage points, mainly to the benefit of the OTAs (eight percentage
point increase), rather than hotels' direct online channel (increase of one percentage
point) (see Table 1). A similar trend was observed in France and Germany taken together.
4.6
Use of OTAs by hotels
OTAs generally and of particular OTAs has changed since the implementation of the
changes to the parity clauses of the major OTAs. According to the replies to the
electronic hotel survey, between 2014 and 2016, the number of hotels that use OTAs
increased in all the participating Member States. The most frequently used OTAs were
Booking.com, followed by Expedia and HRS. The main reason given by hotels for using
new OTAs was to reach new customers worldwide. Hotel chains generally negotiate
terms with the OTAs for the benefit of all the hotels in their chain.
4.7
OTA conversion rates ('look-to-book' ratios)
that either wide or narrow parity clauses are indispensable to prevent hotels from free
riding on OTA investments. Their argument is that, absent parity clauses, consumers will
use OTAs to search for and compare hotels, but will then book more cheaply on the
hotel's website, thereby depriving the OTA of commission revenue. OTA conversion
19
38) The purpose of monitoring this parameter was to determine whether hotels' usage of
39) In the context of the national investigations into OTA parity clauses, OTAs have argued
 ERU, Alm.del - 2016-17 - Bilag 167: Notat om bookingportaler, fra erhvervsministeren
1745709_0020.png
rates ('look-to-book' ratios) can be used as a measure of free-riding. The monitoring
working group therefore examined whether the conversion rates of the major OTAs have
changed following the recent changes to their parity clauses. An analysis of conversion
rate data provided by certain OTAs for a representative sub-sample of hotels in each
participating Member State showed no evidence of decreases in the OTAs' conversion
rates following the changes to OTA parity clauses. The results for France and Germany
did not differ materially from those for the other Member States.
4.8
OTA preferred partner programs
programs was to determine the share of hotels that belong to these voluntary programs
and the extent to which OTAs use these programs to impose either wide or narrow parity
obligations on member hotels. Two of the three large OTAs (Booking.com and HRS)
operate voluntary preferred partner or quality seal programs. On average, hotels that
belong to Booking.com's preferred program account for a small share of Booking's
partners but for a significant share of the reservations made on Booking.com. Across the
ten participating Member States, 30% of hotels that responded to the electronic survey
said that they belong to at least one OTA preferred partner or quality seal program,
though this share varies widely between Member States. Of the three large OTAs, only
Booking.com sometimes charges hotels extra commission for membership of its program
(up to [0%-5%]), depending on the Member State and the area).
53
Hotels that belong to
HRS's Top Quality Seal program must give HRS wide price parity.
4.9
Scope of parity clauses
certain major OTAs. The purpose of monitoring this parameter was to determine the
share of hotels that are still subject to a wide parity clause. Across the ten participating
Member States, one in five of the hotels that responded to the electronic survey said that
at least one OTA they dealt with obliged them by contract to give it price parity relative
to all other sales channels (wide parity). Among these hotels, most named Booking.com,
Expedia and HRS, in that order, as OTAs that imposed wide parity. This result is
somewhat surprising, since Booking.com and Expedia no longer apply wide parity in any
Member State. Possible explanations for the result could be that respondents did not
53
40) The purpose of monitoring the conditions and prevalence of OTA preferred partner
41) The national investigations into OTA parity clauses have been limited to the clauses of
Members of Booking.com's voluntary preferred partner programme must give Booking.com narrow parity,
even though the
Loi Macron
has rendered parity obligations in contracts between OTAs and hotels null and
void.
20
 ERU, Alm.del - 2016-17 - Bilag 167: Notat om bookingportaler, fra erhvervsministeren
understand the question; did not know about the changes to Booking.com and Expedia's
parity clauses; did not understand that narrow parity allows them to offer different room
prices on different OTAs, or felt compelled to continue to apply wide parity, for example
to avoid being penalized in terms of visibility by some OTAs.
4.10 Hotel customer loyalty schemes
42) The narrow parity obligation does not apply to room prices that hotels make available
through customer loyalty schemes, provided that the hotel does not publish discounted
prices online. As wide parity clauses were generally more restrictive in this respect, the
monitoring working group wished to determine whether the recent changes to the parity
clauses of the major OTAs had led to changes in hotels' use of customer loyalty schemes
and the volume of sales made through these schemes.
43) Across the ten participating Member States, 39% of the hotels that responded to the
electronic survey operate some form of customer loyalty scheme to offer lower prices or
more favorable conditions to their customers. This figure is higher for chain hotels (up to
62%), for large hotels (56%) and for hotels that have a star rating of four or five stars
(50%).
44) Loyalty schemes are more common in Member States where OTAs apply narrow parity
clauses (42% of hotels) than in Germany and France, where the parity clauses of some or
all OTAs have been prohibited or rendered void by legislation (29% of hotels).
Approximately half the respondents to the electronic survey reported that sales via their
loyalty scheme had increased in the twelve months following the switch by Booking.com
and Expedia to narrow parity clauses. A more mixed picture was reported by respondents
to the questionnaire sent to large hotel chains.
4.11 New entrants to the online hotel booking sector
45) Part of the theory of harm applied by the NCAs which have investigated OTA parity
clauses is that these clauses may foreclose the entry or expansion of new or smaller
OTAs. The monitoring working group therefore wished to determine whether new
players had entered the OTA sector since the implementation of the changes to the parity
clauses of the large OTAs. Some respondents to the monitoring questionnaires mentioned
various new players or the introduction of new strategies and technologies by existing
players, however the monitoring working group has no information on their market share.
These included the introduction of direct or 'instant' booking and the cost-per-acquisition
model by metasearch sites like TripAdvisor; technological developments relating to
21
 ERU, Alm.del - 2016-17 - Bilag 167: Notat om bookingportaler, fra erhvervsministeren
mobile searching and booking, and software tools allowing hoteliers to enhance their
websites and maximize direct bookings. The increasing presence of AirBnB and Google
was also mentioned.
4.12 Use of metasearch sites by hotels
46) The monitoring working group wished to determine whether the recent changes to the
parity clauses of the major OTAs have led to changes in the use of metasearch sites by
hotels, and to understand the extent to which metasearch websites constitute an
alternative to OTAs for hotels. Across the ten Member States, 30% of the hotels that
responded to the electronic survey contract directly with metasearch sites (34% in France
and Germany taken together). Direct contracts between metasearch websites and hotels
still generate a relatively small share (6%) of total bookings. However, both the share of
hotels that contract directly with metasearch websites and the share of bookings
generated by metasearch websites has increased since 2014, and these shares are larger
for chain and high category hotels. The metasearch websites used most frequently by
hotels are TripAdvisor, Trivago, Google and Kayak.
47) There has been a marked increase in OTA payments to metasearch websites and to
Google in the period monitored.
4.13 OTA best price guarantees
48) The monitoring working group wished to determine whether the implementation of
narrow parity and the prohibition/annulment of some or all OTA parity clauses had
affected the best price guarantees offered by OTAs to consumers. The replies to the
monitoring questionnaires indicate that most OTAs continue to offer consumers a best
price guarantee, promising to match - at the OTA's expense - lower room prices found on
other OTAs or, in some cases, on any other online sales channel. These guarantees
generally do not apply to prices offered to members of OTA or hotel loyalty schemes.
4.14 Expected market developments
49) Most of the large hotel chains that responded to the monitoring questionnaire expect the
large OTAs, particularly Booking.com and Expedia, to grow further and that new entrants
will face significant difficulties. The large hotel chains expect that they will need to
continue to invest heavily in their brands and loyalty schemes and some anticipate further
increases in the cost of advertising on search engines. Furthermore, they expect the
importance of metasearch sites to grow, if price differentiation increases.
22
 ERU, Alm.del - 2016-17 - Bilag 167: Notat om bookingportaler, fra erhvervsministeren
1745709_0023.png
4.15 Other issues raised by stakeholders
50) Many hotels mentioned measures taken by OTAs to
“penalize” unwanted behaviour
by
hotels - such as price and/or availability differentiation - without relying on parity
clauses.
54
The measures included "dimming" (removing pictures and other details of the
hotel) or "downgrading" the hotel (lowering its ranking) in the OTA's search results.
Other issues raised included allegedly excessive OTA commission rates; measures to
prevent direct contact between hotelier and customer; aggressive brand bidding by
OTAs
55
, and allegedly unfair or misleading OTA advertising practices.
5 Submissions received from Booking.com and Expedia
51) In addition to replying to questionnaires, Booking.com and Expedia made written
submissions to the monitoring working group.
5.1
Booking.com's submission
clauses. To support these arguments, it relied on
(i)
an analysis of metasearch data
relating to 40,000 of Booking.com's accommodation partners;
56
(ii)
the results of a Gfk
telephone survey of 3400 of its accommodation partners,
57
and
(iii)
internal data on the
commission rates it charges to hotels in each Member State.
53) According to the analysis of metasearch data:
a)
in [30%-40%] of the metasearch results analyzed, the room price displayed on
52) Booking.com submitted various arguments against the prohibition of narrow parity
Booking.com was [0%-5%] higher or lower than the price available on another OTA;
b)
for [40%-50%] of the accommodation partners covered by the analysis, this level of
price differentiation was observed in more than [40%-50%] of the searches
performed;
54
It should be noted that, in its April 2015 commitments to the French, Italian and Swedish competition
authorities, Booking.com committed to not apply certain measures likely to produce effects equivalent to wide
parity clauses, including linking commission rates to the observance of wide parity and directly linking
display ranking to the observance of wide parity.
For example, purchasing Google Adwords corresponding to the name or trademarks of the hotel, such that
the OTA's website is displayed before the hotel's website when consumers make a search for the name or
trade mark of the hotel.
Analysis of 43 million searches performed on TripAdvisor and Trivago in respect of 40,000 Booking.com
accommodation partners located in the ten participating Member States plus Austria and Switzerland
between 1 October 2015 and 31 July 2016. These include hotels and other types of holiday accommodation that
contract with Booking.com. (For room availability differentiation, the analysis was based on 55.5 million
searches on 48,000 accommodation partners in the same twelve Member States.)
Gfk survey commissioned by Booking.com of 3,400 of its accommodation partners in the ten participating
Member States plus Austria and Switzerland in June 2016.
55
56
57
23
 ERU, Alm.del - 2016-17 - Bilag 167: Notat om bookingportaler, fra erhvervsministeren
1745709_0024.png
c)
accommodation partners in Germany and France (where Booking.com no longer
applies a parity clause) did not price differentiate between OTAs more frequently
than accommodation partners in most of the other participating Member States
(where Booking.com applies a narrow parity clause);
58
d)
there was no significant change in the frequency of price differentiation between
OTAs by accommodation partners in Germany in the periods before and after the
prohibition of Booking.com's narrow parity clause;
e)
as regards room availability differentiation, in [5%-10%] of the metasearch results
analyzed, accommodation partners made rooms available on a rival OTA and/or the
accommodation's own website while offering no rooms on Booking.com.
54) According to the results of the Gfk telephone survey:
a)
[70%-80%] of accommodation partners said they do not price differentiate between
OTAs. The share of partners that said they did price differentiate between OTAs was
not higher in France or Germany than the average for the other Member States where
narrow parity clauses have not been prohibited/annulled by law;
59
b)
for partners that did not price differentiate between OTAs, the reasons most
commonly given were not expressly related to Booking.com's narrow parity clause.
55) According to Booking.com's internal data:
a)
the basic rates of commission charged by Booking.com in each Member State were
not affected by whether Booking.com applies narrow parity or no parity clause in the
Member State concerned.
56) The following points should be noted:
i)
the metasearch analysis did not examine whether room price differentiation by hotels
between OTAs increased following the recent changes to the parity clauses of
Booking.com and Expedia (narrow parity/prohibition/annulment) relative to the
preceding period, when wide parity clauses prohibited such behaviour;
ii)
it is not clear to what extent the metasearch analysis relates to room price
differentiation in the true sense, or price differentiation caused by differences in the
products offered (different types of room, conditions, etc.);
58
French and German accommodation partners did price differentiate more frequently than accommodation
partners in the UK and Austria (where Booking.com applies narrow parity).
Accommodation partners in all participating Member States plus Austria and Switzerland. Source: Gfk survey
59
24
 ERU, Alm.del - 2016-17 - Bilag 167: Notat om bookingportaler, fra erhvervsministeren
iii)
when comparing the frequency of price differentiation between Member States, the
analysis does not take into account country-specific factors other than the differences
in the terms of Booking.com's contracts, nor the continued application of Expedia's
narrow parity clause to some of Booking.com's partner hotels in Germany;
iv)
hotels which participated in the Gfk survey were informed that the survey was
performed on behalf of Booking.com: this may have influenced their replies.
5.2
Expedia's submission
degree of price differentiation applied by hotels. Expedia relies on an analysis of
29 million price comparisons performed in respect of 66,000 of its partner hotels in the
ten participating Member States plus Austria between August 2015 and October 2016.
58) According to this analysis:
a)
hotels subject to narrow parity clauses offered a room price on their website which
57) In its submission, Expedia argues that narrow parity clauses are not a key factor in the
was at least 2% higher than the price on Expedia’s website
in [10%-20%] of the price
comparisons analysed;
b)
hotels subject to narrow parity clauses offered a room price on Expedia which
differed by at least 2% from the price offered on Booking.com in [30%-40%] of the
comparisons analyzed;
c)
although price differentiation between OTAs in Germany did increase slightly
following the prohibition of Booking.com's narrow parity clause, the frequency of
price differentiation was lower in Germany ([20%-30%]) than the average for the
other participating Member States where Booking.com's narrow parity remains in
force ([30%-40%]), and a difference-in-differences analysis covering the period
before and after the German prohibition decision indicates that the increase in price
differentiation between OTAs in Germany was smaller than in other Member States.
59) The following points should be noted:
i)
Expedia did not examine whether room price differentiation by hotels between OTAs
increased following the recent changes to the parity clauses of Expedia and
Booking.com relative to the preceding period, when wide parity clauses prohibited
such behaviour;
ii)
Expedia recognizes that the situation in Germany is not 'clean' for comparison
purposes, as Expedia continues to apply a narrow parity clause in Germany.
25
 ERU, Alm.del - 2016-17 - Bilag 167: Notat om bookingportaler, fra erhvervsministeren
Although Expedia maintains that its share of OTA bookings in Germany is only
[10%-20%], this figure may under-represent the share of German hotels that contract
with Expedia, and whose pricing behaviour continues to be affected by Expedia's
narrow parity clause;
iii)
Expedia's submission relies on data scraped by data collection companies. Although
Expedia states that the data is subject to various checks to ensure accuracy, it is not
clear to what extent it shows room price differentiation in the true sense, as against
price differentiation caused by differences in the products offered (different types of
room, conditions, etc.).
26
 ERU, Alm.del - 2016-17 - Bilag 167: Notat om bookingportaler, fra erhvervsministeren
1745709_0027.png
Appendix 1:
Econometric Analysis of Price Differentiation between OTAs
1.
Introduction
This Appendix describes the econometric analysis of price differentiation between OTAs carried out
by the monitoring working group.
Section 2
describes the metasearch data used for the analysis.
Section 3
describes the analysis and the results.
Section 4
describes the data scraped by the
monitoring working group itself, focusing on the distinction between price and product
differentiation.
2.
The metasearch data
Metasearch websites provide consumers with the opportunity to compare hotel offerings across the
various distribution platforms, among which are OTAs. Some hotels also offer rooms directly on
metasearch websites, but they are currently a rather small minority, for which much less data on
prices is available. We therefore only use the metasearch data to study price differentiation
between
OTAs.
By way of summary, the dataset contains hotel prices quoted by OTAs on the metasearch
website(s) in question for a large sample of hotels for the Participating Member States
60
and Canada,
covering both the period before and after the changes in OTA parity clauses (February 2015 through
September 2016). Hotels in Canada are used as a control group, because we are not aware of any
changes in OTA parity clauses there. This data should therefore make it possible to study the effect of
the changes in OTA parity clauses on price differentiation between OTAs.
2.1 The sample of hotels
The monitoring working group defined a sample of hotels from each Participating Member State as
follows. First, the OTAs Booking.com, Expedia and HRS provided us with a list of hotels present on
their website, along with information on chain affiliation, star rating and size. Second, we removed
all hotels that were present on only one of these OTAs, as these hotels would not provide
information on price differentiation between OTAs. Finally, we drew a sample of hotels from this list.
The sample was stratified
61
to make it representative of the population with respect to hotel chain
60
61
Belgium, Czech Republic, France, Germany, Hungary, Ireland, Italy, The Netherlands, Sweden and the UK.
In fact, we used a method called Halton sequencing where we sort observations according to the chosen
variables and generate a sequence of indices spread evenly over the population. Observations corresponding
to the indices are chosen for the sample which is representative of the population with respect to the variables
chosen for stratification. (for details on how to generate the sequence see Halton, J. (1964), Algorithm 247:
Radical-inverse quasi-random point sequence, Association for Computing Machinery, p. 701)
27
 ERU, Alm.del - 2016-17 - Bilag 167: Notat om bookingportaler, fra erhvervsministeren
affiliation, star rating, and size. The sample was chosen such that we obtain ten percent of the hotels
on the OTA lists for each participating Member State.
For Canada, the monitoring working group could not rely on the lists of affiliated hotels provided by
OTAs. We therefore conducted the following exercise. First, we ran a Python script (as a scraping
tool) visiting the metasearch website(s) in question to collect the names of all hotels in all geographic
locations in Canada as defined on the metasearch website, along with information on chain affiliation
and star rating (information on the size of hotels was unavailable). Second, we drew a stratified
sample of hotels from this list. The stratification was done on the basis of chain affiliation and star
rating, which again ensures that we obtain a random sample of hotels on the metasearch website
that has sufficient observations for each possible combination of hotel affiliation and star rating. The
sample size is ten percent of the complete list of hotels.
2.2
The data for hotels from the sample on price and other variables
For every hotel in the sample, the metasearch website(s) provided us with all the price data shown to
website visitors when they searched on the website. For convenience, we only used search results
shown in the months of February, April, June, September and November in 2015, and January,
March, May, July and (partially) September 2016. For each hotel, multiple price quotes are provided,
so we can assess whether hotels appearing on multiple OTAs publish the same offering on each OTA.
Importantly, the metasearch website(s) in question aim to yield results to consumers that are
comparable. Therefore, only the lowest possible price quote from a given OTA is reported in
response to consumer searches. However, the metasearch website(s) stated that they cannot
guarantee that the prices quoted by different OTAs for a particular hotel relate to the exact same
product on each OTA. For example, it may happen that for a particular hotel and room type listed on
OTA
x
breakfast is included whereas it is not included on OTA
y.
Other examples are possible
differences in room type
superior or lu e , i lusio of WiFi, a d a ellatio
o ditio s. The
implication is that observed differences in prices may actually be driven by differences in the
product. In other words, each observation reflects both possible price and product differentiation
between OTAs. We will come back to the possible implications of this feature of the dataset in the
next section, where the analysis is presented. For ease of exposition, in the text we often refer to
price differentiation, even though observed price differences may be partly or completely caused by
differences in the product. Section 4 of this appendix tries to draw careful inferences concerning the
possi le e te t of produ t differe tiatio a d pure pri e differe tiatio .
Ne t, the data is
erged ith a list of hotels attri utes hai affiliatio , star rati g a d size so that
we can take these into account in the analysis. Further, we exclude certain OTAs for which there are
28
 ERU, Alm.del - 2016-17 - Bilag 167: Notat om bookingportaler, fra erhvervsministeren
1745709_0029.png
indications that they operate a wholesale model, and delete prices which deviate by more than 50%
from the mean price for each search, in order to remove outliers and possibly faulty observations
62
.
Only the data points with more than one price quote are kept in order to study price differentiation
and we restrict the observations only to the default searches for 2 persons and 1 room, to avoid
finding spurious relationships caused by inconsistent treatment of non-standard searches.
Finally, in some cases we see multiple price quotes per OTA for a single combination of search
characteristics.
63
These observations result from identical searches carried out on the same day,
which can happen if a consumer checks the offers multiple times, or if a hotel appears in the search
results of multiple consumers who are interested in a stay of the same length for the same number
of people in the same area. As we cannot distinguish which prices belong to which search, we select
the lowest price per OTA for further analysis.
64
Table 2 shows the size of the final dataset with
respect to all OTAs and with respect to only Booking.com, Expedia and HRS. Given that these three
OTAs tend to be the largest OTAs in the participating Member States, we explore further below
whether there is a material variation in the extent of price differentiation when considering only
these three OTAs compared to considering data for all OTAs in the dataset.
Table 2. Number of observations in the metasearch data
All OTAs
Country
Belgium
Czech Republic
France
Germany
Hungary
Ireland
Italy
Netherlands
Sweden
United Kingdom
Booking, Expedia, HRS
#hotels
84
120
964
1,042
56
46
1,133
139
77
382
#hotels
84
124
974
1,063
61
48
1,159
140
78
390
#price info shown
86,236
81,129
798,567
312,083
62,093
105,473
1,230,265
211,724
36,901
787,126
#price info shown
72,854
57,687
601,275
241,294
50,375
81,988
926,775
169,504
30,898
594,951
EU
Canada
4,121
487
3,711,597
836,775
4,043
406
2,827,601
644,383
Total
4,608
4,548,372
4,449
3,471,984
62
Altering the 50% threshold does not affect the results, as this step only leads to deletion of extreme
observations which are rare in the dataset.
The available characteristics are: date of search, hotel, check-in date, check-out date
Selecting the highest prices does not affect the results.
63
64
29
 ERU, Alm.del - 2016-17 - Bilag 167: Notat om bookingportaler, fra erhvervsministeren
1745709_0030.png
3.
3.1
Analysis of the metasearch data
Definition of price differentiation
We defined price differentiation as the case where at least one OTA price differs by more than 5
percent from one other OTA price. We chose 5% because it is a wide enough margin to avoid basing
conclusions on irrelevant small differences due to e.g. rounding or exchange rate differences. We
checked our results using a margin of 2%, and the results were similar.
Price differentiation is thus a binary variable. The disadvantage of our approach is that it does not
allow for measuring the extent of price differentiation. On the other hand, using indicators that
measure the extent of price differentiation would be more sensitive to large price differences caused
by product differentiation or errors. Finally, as price differentiation was not allowed at all by the wide
parity clauses, it is of particular interest to see whether more hotels differentiate more often.
However, one should recall that the price differentiation displayed on the metasearch website may
be (partly) driven by product differentiation, as explained in Section 2.2.
3.2
Preliminary information on effects
Figure 5 gives a first impression of the development of price differentiation between OTAs
–
as
observed in the dataset
–
within the Participating Member States over time
65
. However, it should be
noted that the data shows a combination of product and price differences. The vertical bars
represent the moments where changes were made to OTA parity clauses. In July/August 2015, both
Booking.com and Expedia switched from
ide
to
arro
parity clauses throughout the EEA. At the
same time, the Loi Macron was implemented in France, which annuls any kind of OTA price parity
clause. In February 2016, the decision by the German competition authority (Bundeskartellamt) to
prohibit Booking.com from using any kind of parity clause came into effect. It should be noted that
the Bundeskartellamt already prohibited HRS from using parity clauses in 2013, which is before the
time period analysed here. This potentially affects the results found for Germany. Due to the
preceding prohibition decision against HRS and the ongoing proceedings against Expedia, one cannot
interpret the Booking.com decision of the Bundeskartellamt as a unitary change from wide to narrow
parity clauses in Germany.
65
Technically, the lines are local polynomial regressions with respect to time that connect the data points in a
smooth way.
30
 ERU, Alm.del - 2016-17 - Bilag 167: Notat om bookingportaler, fra erhvervsministeren
1745709_0031.png
Figure 5. Dynamics of price/product differentiation between OTAs in the Participating
Member States and Canada
Figure 5 suggests that price differentiation and/or product differentiation were present already
before the changes in the parity clauses, and that the development of price/product differentiation
differs between the Participating Member States. There are no sudden changes in the level of
price/product differentiation around the dates of interest for any Participating Member State. One
may however note that price differentiation in Canada seems to be following a downward trend
starting from the summer of 2015. The two OTA selections (the big three OTAs and all OTAs) differ
31
 ERU, Alm.del - 2016-17 - Bilag 167: Notat om bookingportaler, fra erhvervsministeren
1745709_0032.png
clearly in the level of offer differentiation and in the size of the gap that appears between the
Participating Member States and Canada. The difference in price/product differentiation levels is to
be expected, since it follows from our definition of price/product differentiation that the likelihood of
price/product differentiation is potentially higher as more OTAs are included in the analysis.
3.3
The econometric model
To analyse the development of price differentiation in greater detail, we estimate a linear model for
the binary choice between price differentiating and offering the same price through all OTA partners.
To measure the effects of the changes in parity clauses, we employ a difference-in-differences
approach. This means that we compare the difference between the trend of price differentiation
within the Participating Member States before and after the changes, with the trend of price
differentiation in Canada (where parity clauses remained unchanged) over the same period. If it is
the case that the trends are the same in the Participating Member States and Canada, this would
suggest that the changes to parity clauses in Europe had little effect on the trend of price
differentiation.
In the model we control for:
the number of days between the search date and check-in date
length of stay
weekend stays, and
the number of OTA prices available
Because we have time series data on a sample of hotels, we can run a fixed-effects model on the
level of hotels. This allows us to control for any unobserved characteristics of the hotels that may
drive their decision to differentiate prices. For example, skills and beliefs of the hotel manager are
likely to directly impact price differentiation by the hotel. Hence the model controls for many more
factors that may influence the decision to differentiate prices than the variables that we can actually
observe.
Formally, the model is written as:
���� =
+ ′�� + ′�� + ′�� + ��′ + ��′ + ,
where
PD
is the price differentiation indicator,
X
is a vector that contains the 4 control variables
listed above,
H
is a vector of dummies for each individual hotel (and hence captures the fixed effects
due to unobserved hotel characteristics),
M
is a vector of monthly dummy variables for the time of
search
66
capturing the overall time trend,
S
is a vector of dummy variables for months of stay
67
M indicates in which month the search was carried out (February 2015
–
September 2016)
66
32
 ERU, Alm.del - 2016-17 - Bilag 167: Notat om bookingportaler, fra erhvervsministeren
1745709_0033.png
accounting for eventual seasonality,
T
is a vector of dummy variables that equal 1 if some kind of
change to the parity clause is in effect in a particular country (thus
T
captures all the treatments), and
The vector of coefficients
��
contains the coefficients of interest. This indicates whether price
is the error term.
differentiation in the given country is lower or higher given that the parity clause differs in some way
from the wide parity clause, and controlling for other factors. Note that the treatment variables
T
include different treatments, namely:
1)
the switch from wide to
narrow parity clauses
by Booking.com and Expedia in all
Participating Member States in July/August 2015,
2)
the entry into force of the
Loi Macron annulling all OTA price parity clauses in France,
which
coincided with the switch from wide to
narrow parity clauses
by Booking.com and Expedia in
France in July/August 2015, and
3)
the Bundeskartellamt's decision against Booking.com prohibiting Booking.com from using
any kind of parity clause as of February 2016.
However, caution should be applied when comparing the effects of the treatments between Member
States. The reason
is that the o ser ed le el of pri e differe tiatio
a
e true pri e
differentiation or (partly) driven by product differentiation (see below, Section 4). The relative
importance of these two factors may differ between Member States, however. And more
importantly, they may have developed differently over time in the different Member States. This
makes it difficult to conclude that one of the treatments has a stronger or weaker effect on
pure
price differentiation. Regarding the comparison to Canada, it is useful to note that the change by
Booking.com and Expedia to narrow parity in Europe relaxed both price and availability parity, and
this is therefore likely to increase both product and price differentiation relative to Canada.
(i.e. estimate of
��),
we are not concerned about problems relating to the chosen functional form and
Finally, as we are mainly interested in the effect of the changes in the degree of price differentiation
the binary nature of the dependent variable. The difference-in-differences model is based on
comparisons of means and as such it is perfectly suitable also for binary variables. Using non-linear
binary choice models such as probit or logit would lead to a rather cumbersome interpretation of the
coefficients and/or marginal effects of the treatment. The linear probability model is therefore
suitable for our purposes.
67
S indicates for which month of stay (January
–
December) a consumer searched.
33
 ERU, Alm.del - 2016-17 - Bilag 167: Notat om bookingportaler, fra erhvervsministeren
1745709_0034.png
3.4
Results
–
prices from all OTAs included
Table 3 reports the parameter estimates from the model applied to all hotels, chains and non-chains.
For clarity, narrow parity treatments are defined as dummy variables equal to 1 for a period from
July 2015 onwards in
all Participating Member States
(including France and Germany). In France, the
treatment coincides with the entry into force of the
Loi Macron.
The treatment measuring the effect
of the prohibition of all parity clauses of Booking.com in Germany is defined as an indicator with
value one for the period starting February 2016 and as such it measures an effect
in addition to the
narrow parity treatment.
Table 3. Results from the difference-in-differences model based on prices from all
OTAs
All
coef
Narrow parity
Belgium
Czech Republic
Germany
Hungary
Ireland
Italy
Netherlands
Sweden
United Kingdom
Loi Macron/narrow parity (F)
No parity by Booking.com (DE)
Controls
# OTAs
# days until stay/100
weekend
intercept
Month of stay dummies
Month of search dummies
Length of stay dummies
Hotel fixed effects
R
2
F-statistic (p-value)
# hotels
# observations
0.054***
0.007**
0.002
0.154***
YES
YES
YES
YES
0.083
131.9 (0.000)
4,608
4,548,372
(59.5)
(3.2)
(1.3)
(16.6)
0.054***
0.002
0.005**
0.167***
YES
YES
YES
YES
0.092
84.8 (0.000)
1,834
2,538,662
(43.8)
(0.8)
(2.8)
(14.5)
0.055***
0.013***
-0.003
0.136***
YES
YES
YES
YES
0.076
73.8 (0.000)
2,774
2,009,710
(42.4)
(3.9)
(1.6)
(9.6)
0.071*
0.141***
0.102***
0.082*
0.118**
0.155***
0.030
0.126***
0.114***
0.087***
0.035**
(2.2)
(4.6)
(4.6)
(2.3)
(3.3)
(7.3)
(0.8)
(4.9)
(4.7)
(4.2)
(3.3)
0.100*
0.153***
0.139***
0.060*
0.167***
0.195***
0.054
0.175***
0.149***
0.134***
0.045**
(2.3)
(4.5)
(5.9)
(2.6)
(3.5)
(7.6)
(1.2)
(6.0)
(5.9)
(6.4)
(2.7)
0.014
0.084
0.036
0.069
0.036
0.080*
0.009
0.041
0.065
0.007
0.021
(0.3)
(1.8)
(0.2)
(1.3)
(0.7)
(2.2)
(0.2)
(0.9)
(1.4)
(0.2)
(1.6)
Chains
t-stat
coef
t-stat
Non-chains
coef
t-stat
*,**,*** denote statistical significance at: * 5%, ** 1% and *** 0.1% significance level (the t-statistic is based
on cluster robust standard errors)
34
 ERU, Alm.del - 2016-17 - Bilag 167: Notat om bookingportaler, fra erhvervsministeren
1745709_0035.png
We can note three main results:
1. the coefficients for treatment effects of the switch to narrow parity by Booking.com and
Expedia are positive and statistically significant at 5% significance level for all Participating
Member States except The Netherlands. This includes also France where the treatment
consisted of the switch to narrow parity by Booking.com and Expedia and the entry into force
of the
Loi Macron;
2. it appears that the effects are driven by chain hotels. For independent hotels only, there is no
statistically significant effect of any treatment, except in Italy;
3. the results for the Bundeskartellamt's Booking.com decision exhibit a positive significant
effe t, suggesti g that the prohi itio of Booki g. o
observed price differentiation between OTAs.
s parit
lauses led to a i rease i
3.5
Results
–
prices from Booking.com, Expedia and HRS only
The results from an alternative specification focusing on price differentiation among the three major
OTAs are presented in Table 4.
Table 4.
Results from the difference-in-differences model based on prices from
Booking.com, Expedia and HRS
All
coef
Narrow parity
Belgium
Czech Republic
Germany
Hungary
Ireland
Italy
Netherlands
Sweden
United Kingdom
Loi Macron/narrow parity (F)
No parity by Booking.com (DE)
Controls
# OTAs
# days until stay/100
weekend
intercept
Month of stay dummies
Month of search dummies
0.001*
-0.006*
0.008***
0.196***
YES
YES
(2.0)
(2.2)
(7.2)
(16.7)
0.002*
-0.007
0.010***
0.174***
YES
YES
(2.2)
(1.9)
(6.2)
(12.5)
0.001
-0.005
0.007***
0.226***
YES
YES
(0.7)
(1.2)
(3.7)
(11.6)
0.105**
0.101**
0.066*
0.081*
0.124
0.118***
0.028
0.097*
0.050
0.056*
0.041**
(2.8)
(3.0)
(2.6)
(2.2)
(1.9)
(4.8)
(0.4)
(2.2)
(1.9)
(2.3)
(2.8)
0.169***
0.148**
0.113***
0.101**
0.198*
0.157***
0.051
0.173***
0.086**
0.091***
0.071**
(4.0)
(2.9)
(4.1)
(3.1)
(2.3)
(5.2)
(0.6)
(5.5)
(3.2)
(3.7)
(3.3)
0.003
0.032
-0.011
0.036
-0.002
0.050
0.006
-0.048
-0.002
-0.000
-0.003
(0.1)
(0.6)
(0.2)
(0.5)
(0.0)
(1.2)
(0.1)
(0.4)
(0.0)
(0.0)
(0.2)
Chains
t-stat coef
t-stat coef
Non-chains
t-stat
35
 ERU, Alm.del - 2016-17 - Bilag 167: Notat om bookingportaler, fra erhvervsministeren
1745709_0036.png
Length of stay dummies
Hotel fixed effects
R
2
F-statistic (p-value)
# hotels
# observations
YES
YES
0.021
25.5 (0.000)
4,449
3,471,984
YES
YES
0.022
16.8 (0.000)
1,754
1,980,271
YES
YES
0.023
12.6 (0.000)
2,695
1,491,713
*,**,*** denote statistical significance at: * 5%, ** 1% and *** 0.1% significance level (the t-statistic is based
on cluster robust standard errors)
While the magnitudes of the effects are different from the previous specification for some
Participating Member States, we observe estimates suggesting the same effects as the specification
considering pricing information from all OTAs. That is, all treatments appear to have a significant
positive effect on observed price differentiation. Although the t-statistics of the parameter estimates
are below the threshold corresponding to the 5% significance level for Ireland and the United
Kingdom, we do not consider this to be of major importance, as it is only marginally below this
threshold. The estimated effects for the narrow parity treatment in Germany and in the UK appear to
be most affected by the different subset of OTAs included in the analysis.
3.6
Remaining specifications
The above versions of the model rest on the assumption of a common trend in price differentiation
for all countries, both treated (i.e. Participating Member States) and untreated (i.e. Canada).
Although Figure 5 shows that the trends in the pre-treatment period are not wildly different for most
countries, we may wish to add country-specific linear trends to check the robustness of the results.
The problem of using additional trends is that a treatment effect is only identified (i.e. disentangled
from the trend) if there is a significant break in the time-series. This does not seem to be the case
judging from Figure 5 and it is rather unrealistic to expect for treatments of this nature, where we
expect the effects to be steady, rather than immediate. Moreover, we have carried out a statistical
test for the common trend assumption which does not invalidate the assumption.
68
The robustness of the results from the difference-in-differences analysis was further tested by
making alternative choices such as:
different sets of OTAs (e.g. OTAs listing more than 1000 hotels, largest website per OTA
corporate group);
treatment of outliers (all observations, 20% deviation from mean);
2% threshold for price differentiation.
68
See for how to test for the common trend assumption e.g. Autor, David H. "Outsourcing at will: The
contribution of unjust dismissal doctrine to the growth of employment outsourcing."
Journal of labor economics
21.1 (2003): 1-42.
36
 ERU, Alm.del - 2016-17 - Bilag 167: Notat om bookingportaler, fra erhvervsministeren
1745709_0037.png
All mentioned choices and their combinations lead to similar results.
4.
Insights from the data scraped from Booking, Expedia and HRS
he k o the
etasear h data. ‘elati e to the
metasearch data, the
The monitoring working group also scraped data from hotel room offerings on OTAs. This was done
in order to perform
a sa it
scraped data has the advantage that it contains slightly more information on the product being
offered, such as whether breakfast is included. It is therefore possible to obtain more insight into the
amount of
price
differentiation between OTAs as distinguished from
product
differentiation between
OTAs. The downside is that this data is not historical, and so we cannot use it to estimate the effect
from the treatments to parity clauses. The next two sub-sections describe the data collection
procedure and the results, respectively.
4.1
Data collection process
To obtain detailed data at room level, we ran a Python script acting as a virtual consumer which
saves prices from the websites of Booking.com, Expedia, and HRS. The procedure is designed as
follows:
save all hotels from top 20 destinations on Booking.com for each Participating Member State
(Canada is excluded);
look for hotels in the same destinations on Expedia and HRS;
save URLs of the hotel pages along the way;
hotels from the three websites are matched based on names and city (with manual quality
control checks);
for each hotel, price data on room level is gathered from all three OTAs, on the basis of a one
night stay for two people, for a night one month ahead of the scraping date
69
;
for each hotel, prices from different OTAs are scraped immediately one after another to
avoid any differentiation arising from prices changing over time);
attributes of the offer are saved along with the price for breakfast, if available. The attributes
are breakfast inclusion, free cancellation and the name of the room. While there are other
characteristics (for example, spa access), they are not recorded in a sufficiently similar way
across platforms to be able to scrape them accurately;
we ran three rounds of the scraping exercise, on September 6
th
and 20
th
and October 3
rd
,
2016.
Table 5 provides information on the size of the dataset.
69
Only one stay date has been used, due to time constraints. A stay date one month ahead was used in order to
avoid looking at too many fully-booked hotels.
37
 ERU, Alm.del - 2016-17 - Bilag 167: Notat om bookingportaler, fra erhvervsministeren
1745709_0038.png
Table 5. Size of the scraped dataset
Member State
Belgium
Czech Republic
France
Germany
Hungary
Ireland
Italy
Netherlands
Sweden
United Kingdom
#hotels
487
656
2,490
2,225
334
300
1,983
646
421
1,513
#searches with
price available
1,326
1,656
6,538
4,983
872
808
4,866
1,670
1,100
4,072
Total
11,059
27,891
4.2
Price differentiation and product differentiation
Within this set of selected prices, price differentiation between OTAs is shown in Figure 6, where the
entire bar for a Participating Member State shows the share of cases where prima facie price
differentiation is observed. So, for example, apparent price differentiation was observed in
approximately 40% of searches across the selected hotels in the Czech Republic. Because we
collected data in September and October 2016, it was not possible to check the levels of price
differentiation when wide parity clauses were still in place.
As a first observation, the level of apparent price differentiation is roughly similar to that in the
metasearch dataset. It should be noted that we use the same definition for price differentiation here
as in the previous section. However, since we have many price quotes for each hotel from each of
the three OTAs (depending on room type, conditions, etc.), we compute price differentiation only
over
the set of lowest price quotes
from each OTA for a given hotel. This is also how the metasearch
website constructs its search results for consumers: it selects from each OTA the lowest price quote
for a particular hotel in the search results. Hence, although this way of selecting prices might well
lead to the comparison of diverging products, it is an appropriate way to check the metasearch data.
Using the data on room type difference and condition differences in the scraped data, we separate
pure pri e differe tiatio
et ee OTAs fro
price differences due to differences in room type and
from price differences due to differences in conditions (namely breakfast and cancellation policies).
So, for example, of the 40% of searches which gave rise to apparent price differentiation in the Czech
Republic, less than half of these searches (and 17% of total searches in the Czech republic) appear to
38
 ERU, Alm.del - 2016-17 - Bilag 167: Notat om bookingportaler, fra erhvervsministeren
1745709_0039.png
represe t pure pri e differe tiatio , rather tha a
i of pri e differentiation
and differences in the
type of room offered or differences in the conditions of the offer.
Figure 6. Price differentiation between OTAs in the scraped dataset
To identify differences between room types across OTAs, we identified the most frequent strings in
room names used on OTAs and checked for their presence on all OTAs where a hotel is listed. Table 6
lists the recognised strings.
Table 6. Non-standard room type indications
70
Budget
Budget
Economy
Klein (=small)
Voordelig (=cheap)
Superior
Superior
Comfort
Other
(De)luxe; Executive; Royal;
Uitzicht (=view); Familie (family);
Studio; Appartement; VIP;
Premium; Suite;
Privilege
For 'budget' and 'superior' type rooms, rooms are considered identical only when, on all OTAs, their
name contains strings belonging to the same group in the table above. For other types of rooms,
rooms are only considered identical when, on all OTAs, the room name contains identical strings
from the category 'other' in the above table. The above approach likely misses some inconsistencies
70
We scraped from Dutch language websites; we did this because scraping from English language websites
would yield prices in pounds sterling or American dollars by default. We have given relevant translations
where necessary.
39
 ERU, Alm.del - 2016-17 - Bilag 167: Notat om bookingportaler, fra erhvervsministeren
1745709_0040.png
between room types with atypical names
71
, but it also probably marks some identical rooms as
different when the name differs but the room is the same. For example, a room may be called
Ro al
on one OTA and
Delu e
on another. As the majority of the observations use the standard
names, we believe that the extent of the unavoidable error is unlikely to influence the results
–
especially if we look at differences between Member States, as the same approach is applied to all
Member States.
With regard to booking conditions, we identify offers with free breakfast and free cancellation and
hence, we are able to correct for these factors that may drive price differences. Of course, there are
other factors that may drive price differences, but these two seem to be relatively important. For
example, free WiFi is something that most hotels offer to all guests or to no guests, and is unlikely to
be a differentiator between OTAs. Nevertheless, it should be noted that our control for product
differences is not perfect.
The resulting breakdown of the sources of differences in prices is depicted in Figure
7.
This Figure is identical to Figure 6, except that each bar has been normalised to 100 percent, and
should be interpreted as follows.
Figure 7. Presence of room or condition differences given a price difference
As stated above, for each hotel in a particular Member State we have selected the lowest price quote
from each OTA. Then, we checked whether there was differentiation between these price quotes.
71
For example, we miss differences such as room with whirlpool bathtub and standard room .
40
 ERU, Alm.del - 2016-17 - Bilag 167: Notat om bookingportaler, fra erhvervsministeren
1745709_0041.png
Next, we checked if there were differences in the room type for these price quotes, and we checked
if there were differences in conditions for these price quotes. So, for Belgium, for example, when
price differentiation occurs for a given hotel, in about 55 percent of the cases this difference
cannot
be explained by differences in room type or differences in conditions (breakfast or cancellation). In
about ten percent of the cases, the price difference may be explained by differences in the
conditions, and in about 35 percent of the cases the price differences may be explained by
differences in room type.
Figure 7 shows that differences in conditions are an infrequent driver of price differentiation. Room
type differentiation is more common. Price differentiation cannot be explained by differences in
room type or conditions in about 40-60 percent of the cases.
Another challenge with our data is the possibility of seeing the same price for different rooms or for
rooms with different conditions. Therefore, it is important to check whether in the cases where we
do not observe a price difference, the room type and conditions are the same. This is shown in Figure
8.
Figure 8. Presence of room or condition differences given no price difference
As can be seen in Figure 8, in most cases when there is no price differentiation, there are no
differences in room type or conditions either. The only notable exception here is the Netherlands,
where in forty percent of the cases when there is no price difference, there is a difference in room
type, conditions, or both.
41
 ERU, Alm.del - 2016-17 - Bilag 167: Notat om bookingportaler, fra erhvervsministeren
1745709_0042.png
Appendix 2:
Electronic survey of hotels
–
composition of samples and response rates
Country
Variable
Part of chain
Value
YES
NO
2 or less
3
4
5
Average
Median
Sample
28%
72%
18%
52%
27%
2%
53
30
966
Replies
28%
72%
20%
54%
24%
1%
55
27
86
Country
Variable
Part of chain
Value
YES
NO
2 or less
3
4
5
Average
Median
Sample
49%
51%
35%
47%
16%
2%
48
36
3080
Replies
36%
64%
29%
51%
17%
3%
49
30
93
BELGIUM
Star rating
FRANCE
Star rating
Number of rooms
Number of hotels
Number of rooms
Number of hotels
Country
Variable
Part of chain
Value
YES
NO
2 or less
3
4
5
Average
Median
Sample
10%
90%
3%
51%
42%
4%
47
28
352
Replies
19%
81%
2%
46%
46%
6%
56
32
289
Country
Variable
Part of chain
Value
YES
NO
2 or less
3
4
5
Average
Median
Sample
17%
83%
13%
59%
26%
2%
53
31
1210
Replies
16%
84%
4%
60%
33%
3%
59
33
1113
CZECH
REPUBLIC
Star rating
GERMANY
Star rating
Number of rooms
Number of hotels
Number of rooms
Number of hotels
42
 ERU, Alm.del - 2016-17 - Bilag 167: Notat om bookingportaler, fra erhvervsministeren
1745709_0043.png
Country
Variable
Part of chain
Value
YES
NO
2 or less
3
4
5
Average
Median
Sample
15%
85%
6%
51%
40%
3%
64
37
694
Replies
32%
68%
5%
31%
64%
0%
102
65
78
Country
Variable
Part of chain
Value
YES
NO
2 or less
3
4
5
Average
Median
Sample
35%
65%
13%
46%
39%
2%
63
36
1630
Replies
32%
68%
10%
46%
42%
2%
70
32
121
HUNGARY
Star rating
NETHERLANDS
Star rating
Number of rooms
Number of hotels
Number of rooms
Number of hotels
Country
Variable
Part of chain
Value
YES
NO
2 or less
3
4
5
Average
Median
Sample
31%
69%
6%
46%
42%
6%
80
67
595
Replies
59%
41%
0%
44%
37%
19%
126
103
35
Country
Variable
Part of chain
Value
YES
NO
2 or less
3
4
5
Average
Median
Sample
44%
56%
7%
46%
45%
2%
91
62
999
Replies
64%
36%
1%
24%
72%
3%
127
117
116
IRELAND
Star rating
SWEDEN
Star rating
Number of rooms
Number of hotels
Number of rooms
Number of hotels
Country
Variable
Part of chain
Value
YES
NO
2 or less
3
4
5
Average
Median
Sample
9%
91%
16%
48%
33%
3%
44
30
4290
Replies
10%
90%
10%
53%
32%
4%
46
31
374
Country
Variable
Part of chain
Value
YES
NO
2 or less
3
4
5
Average
Median
Sample
47%
53%
10%
51%
36%
4%
74
47
2638
Replies
66%
34%
4%
33%
56%
8%
213
89
137
ITALY
Star rating
UNITED
KINGDOM
Star rating
Number of rooms
Number of hotels
Number of rooms
Number of hotels
43