REU, Alm.del - 2017-18 - Bilag 60: Henvendelse af 10/11-2017 fra Aarhus Universitet, Institut for Økonomi v/Anna Piil Damm vedr. "Ny forskning om virkninger af at sænke den kriminelle lavalder"

Retsudvalget 2017-18
REU Alm.del Bilag 60
Offentligt

Wor

King

Papers

Economics Working Papers

2017-10

Lowering the minimum age of criminal responsibility:

Consequences for juvenile crime and education

Anna Piil Damm, Britt Østergaard Larsen, Helena Skyt Nielsen and Marianne

Simonsen

REU, Alm.del - 2017-18 - Bilag 60: Henvendelse af 10/11-2017 fra Aarhus Universitet, Institut for Økonomi v/Anna Piil Damm vedr. "Ny forskning om virkninger af at sænke den kriminelle lavalder"

Lowering the minimum age of criminal responsibility:

Consequences for juvenile crime and education

Anna Piil Damm

[email protected]

Aarhus University

Britt Østergaard Larsen

[email protected]

The Danish Centre for

Applied Social Science

Helena Skyt Nielsen

[email protected]

Aarhus University

Marianne Simonsen

[email protected]

Aarhus University

October 2017

Abstract

This paper exploits a Danish policy reform combined with population-wide administrative registers

to investigate whether being above the minimum age of criminal responsibility deters juveniles from

crime. We study young individuals’ tendency to commit crime as well as their likelihood of

recidivism by exploiting police records on offenses committed by the population of children and

youth, including those below the minimum age of criminal responsibility. The reform lowered the

minimum age of criminal responsibility from 15 to 14 years. We find that the reform did not deter

14-year-olds from committing crime. Moreover, conditional on committing crime in the first place,

youths affected by the lower minimum age of criminal responsibility were

likely to recidivate

and less likely to be enrolled in the 9

grade, just as they have lower grades at the 9th grade exit

exam, conditional on participating. The latter results are consistent with labeling effects of processing

in the criminal justice system.

Keywords: Juvenile delinquency, sanctions, general deterrence, individual deterrence, labeling

effects

JEL-codes: K14, K42, I21

Acknowledgements:

We thank David P. Farrington, Britta Kyvsgaard, Olof Bäckman, and participants of the

8th Transatlantic Workshop on the Economics of Crime, 2016, the Stockholm Criminology Symposium 2015,

the Child Research Seminar, Aarhus University, and the 3rd Family and Education Workshop for their helpful

comments and discussions. We appreciate financial support from TrygFonden’s Centre for Child Research,

Aarhus University and research assistance from Kathrine Sørensen, Villiam Vellev and Iben Büchler Nielsen.

The usual disclaimer applies.

REU, Alm.del - 2017-18 - Bilag 60: Henvendelse af 10/11-2017 fra Aarhus Universitet, Institut for Økonomi v/Anna Piil Damm vedr. "Ny forskning om virkninger af at sænke den kriminelle lavalder"

1. Introduction

Youth crime has decreased considerably in recent years. This development is seen across a variety of

settings, including the US, the UK, and the Scandinavian countries (Danish Ministry of Justice, 2014,

2016). At the same time, many western countries have on-going political discussions about the ”right”

age-limits in the criminal justice system, and there are examples of both movements towards raising

the age boundaries (e.g. across US states

) and advocacies towards lowering the minimum age of

criminal responsibility (e.g. in Denmark). This paper exploits a “tough-on-crime” motivated

reduction in the age of criminal responsibility in Denmark during the summer of 2010 to investigate

how the risk of early, and formal, exposure to the justice system affects youth crime and education.

A vast literature studies the impact of tough, consequential sanctions, such as incarceration, on future

outcomes for adults, and finds mixed results.

A smaller literature considers the impact of juvenile

incarceration. Aizer and Doyle (2015) find detrimental effects on recidivism and education, whereas

Hjalmarsson (2009a) finds the opposite. Previous studies on age boundaries in the youth justice

system have largely focused on the age at which juvenile offenders transfer to the adult justice system

and results are mixed. There are both examples of studies finding no effects of the majority age (e.g.

Hjalmarsson 2009b; Loeffler and Chalfin 2017; Loeffler and Grundwald 2015) and studies

documenting general deterrent effects (Hansen and Waddell 2014; Lee and McCrary 2017; Levitt

1998).

Our study, in contrast, is concerned with less intrusive sanctions to juveniles at the margin of the

criminal justice system. We investigate the consequences of criminal justice system processing and

the receipt of a criminal record at an early age. Little is known about this area. Yet a rich literature

There are several states in the US, which have raised the age of majority up to eighteen, for example Illinois, Connecticut,

Mississippi, Massachusetts, and New Hampshire (Loeffler and Grundwald 2015).

The idea that the severity of punishment could deter crime at least dates back to Beccaria’s (1764) treatise and to

Bentham (1789).

Some studies find that incarceration is effective in terms of reducing recidivism (e.g. Bhuller et al. 2016; Di Tella and

Schargrodsky 2013; Landersø 2015), while other studies find the opposite (e.g. Bales and Piquero 2012; Cochran, Mears

and Bales 2014; Green and Winik 2010; Michel, Rosholm and Simonsen 2017; Nagin and Snodgras 2013). In addition,

results on the impact of incarceration on labour market outcomes are mixed. Some authors find beneficial effects in the

short- to medium run (e.g. Bhuller et al. 2016; Kling 2006; Landersø 2015) and suggest that this is due to rehabilitation

and prevention programs, while Michel, Rosholm and Simonsen (2017) find detrimental effects and Kling (2006) reports

fade out after 7-9 years. Drago, Galbiati and Vertova (2009) find general deterrence effects of longer sentences for a

sample of former inmates.

We use the term “general deterrence” to denote the general prevention of crime and the term “individual deterrence” to

denote discouragement of the individual from future criminal acts (like e.g. Bell, Jaitman and Machin 2014). The latter is

sometimes denoted “specific deterrence” (e.g. Di Telle and Schargrodsky, 2013)

REU, Alm.del - 2017-18 - Bilag 60: Henvendelse af 10/11-2017 fra Aarhus Universitet, Institut for Økonomi v/Anna Piil Damm vedr. "Ny forskning om virkninger af at sænke den kriminelle lavalder"

now acknowledges that childhood and early youth are particularly sensitive periods (e.g. Heckman,

2008) and emphasizes the importance of early interventions and prevention. Importantly, while IQ is

considered set after the first decade of life, behavior seems malleable at later ages; see e.g. Cunha,

Heckman, Lochner and Masterov (2006) and Cunha and Heckman (2008). Whether the presence of

official sanctions earlier on in life deters juvenile crime is unclear, as is whether experiencing formal

punishment earlier on in life reduces recidivism.

The Danish reform, exploited here, lowered the minimum age of criminal responsibility from age 15

to age 14. The political aim was clearly to deter juvenile crime.

The reform was communicated to

the public through substantial media coverage, especially in the year before the reform. In fact, the

term “minimum age of criminal responsibility” was mentioned more than 300 times in the national

newspapers during that year (see further details below). As a result, juveniles were likely to be aware

of the policy reform that changed the age-limit of prosecution, conviction, and sanctions of young

offenders in the Danish criminal justice system.

It is possible, therefore, that the reform could have

deterred 14-year-old youth from committing crime.

Our data are uniquely well suited to study this question. The longitudinal register data includes

information about all offences for which the Danish police have charged a juvenile offender or would

have charged a juvenile offender, in case the offender were above the minimum age of criminal

responsibility. We have information for the periods across which the minimum age of criminal

responsibility was lowered and analyze monthly records of reported crimes from November 1st 2008

to January 31st 2012. Via unique personal identifiers, we link these longitudinal data to background

information for children and their parents as well as to information about academic performance. Our

quasi-experimental strategy is to compare outcomes for 14-year-olds just before and after the change

in the minimum age of criminal responsibility, while carefully controlling for the downward trend in

youth crime during the observation period. Robustness analyses include estimation of placebo effects

for youths close to but above the minimum age of criminal responsibility throughout the same period

”I am convinced that lowering the age of criminal responsibility will prevent crime. It will force some youths to think

twice before committing crime”(in Danish: “Jeg er sikker på, at nedsættelse af den kriminelle lavalder vil have en

præventiv effekt. Det vil få nogle unge og børn til at tænke sig om en ekstra gang og lade være med at gå ud i kriminelle

handlinger.”),

Kim Andersen, legal affairs spokesperson, Liberal Party of Denmark (Venstre). (The 1st reading of the

bill,

Thursday

the

15th

April

2010:

http://www.ft.dk/samling/20091/lovforslag/l164/beh1-

75/2/forhandling.htm?startItem=#nav).

In studies exploiting policy reforms to investigate deterrent effects, an important intermediate outcome and indeed a

precursor to identifying deterrence is the extent to which potential offenders are aware that the policy has changed (Waldo

and Chiricos 1972; Nagin 1998; Apel 2013).

REU, Alm.del - 2017-18 - Bilag 60: Henvendelse af 10/11-2017 fra Aarhus Universitet, Institut for Økonomi v/Anna Piil Damm vedr. "Ny forskning om virkninger af at sænke den kriminelle lavalder"

and investigation of potential spillover effects on delinquency by youths close to but below the

minimum age of criminal responsibility in the same period. By doing both, we add to approaches

used by a number of recent papers to investigate general deterrence effects of the age of majority (Lee

and McCrary 2017; Hjalmarsson 2008; Hansen and Wadell 2014).

We find no evidence that the reform lowered the probability of committing crime among 14-year-

olds. Yet conditional on committing crime in the first place, youths were, in fact,

likely to

recidivate. Hence, we find no evidence that interactions with the criminal justice system and the

receipt of a criminal record at an early age induce less crime. We also find some evidence that

schooling outcomes, such as the tendency to be enrolled in regular 9

grade classrooms and language

arts exam grades, deteriorate more among offenders processed in the criminal justice system.

We structure the remainder of the paper as follows: Section 2 presents the background for the study,

Section 3 introduces the institutional details and discusses the policy reform, and Section 4 presents

our data. Section 5 estimates general deterrence effects of the minimum age of criminal responsibility,

while Section 6 considers individual deterrence effects on recidivism and schooling outcomes.

Finally, Section 7 concludes.

2. Background

2.1 Theoretical framework

In economics, criminal offending is regarded as a rational decision where the potential offender

commits a crime if expected benefits outweigh expected costs. Common for most theories in the field

are three behavioral predictions: crime is deterred by increasing the probability of being caught (due

to, for instance, policing intensity), increasing severity of punishment (in terms of time, income, or

psychic costs), or increasing opportunity costs (in terms of, for instance, labor market opportunities),

see review by Chalfin and McCrary (2017).

The seminal paper by Becker (1968) presents a static model where individuals face a gamble.

They

may either commit a crime and with probability

1-p

receive the benefit from crime, and with

probability

be caught and punished instead, or they may abstain from crime and derive a non-crime

See also Stigler (1970) and Polinsky and Shavell (1984).

REU, Alm.del - 2017-18 - Bilag 60: Henvendelse af 10/11-2017 fra Aarhus Universitet, Institut for Økonomi v/Anna Piil Damm vedr. "Ny forskning om virkninger af at sænke den kriminelle lavalder"

risk-free utility. In this set-up, the three predictions outlined above follow naturally. Ehrlich (1973)

focuses on the opportunity costs of legitimate and illegitimate activities as being important for how

potential offenders respond to incentives. Lee and McCrary (2017) develop a dynamic extension of

Becker’s model of crime, where individuals draw crime opportunities from a distribution of criminal

offers. Each period the individual decides whether to commit a crime, and if he commits a crime, he

is randomly apprehended. Their dynamic set-up emphasizes the potential importance of myopia and

the interplay between sentence length and degree of myopia.

Crime policies may affect criminal behavior through three main channels (corresponding to the three

predictions mentioned above): certainty of the punishment, i.e. the risk of detection, apprehension

and conviction given commission of crime; severity of the punishment, i.e. the onerousness of the

legal consequences given a sanction is imposed; and opportunity costs. The effects of such policies

rely crucially on the extent to which juveniles are aware of the changes in policies and whether they

subsequently adapt their expectations.

Crime policies may affect criminal behavior through the threat of punishment, i.e. general deterrence,

and due to the experience of punishment, i.e. individual deterrence. Incarceration will mechanically

reduce crime due to incapacitation, but the experience of punishment may also change the individual’s

future behavior by changing his information about the costs and benefits of criminal behavior (e.g.

deter the individual from future crime by instilling an understanding of the consequences/costs) or,

somewhat more speculative, his preferences for criminal relative to legal activities. The lowering of

the minimum age of criminal responsibility from 15 to 14 increases the severity of the expected

punishment of offenders aged 14. This threat may have a general deterrent effect on criminal behavior

among 14-year-olds. Moreover, the experience of punishment at the age of 14 may deter the

individual from future crime. The reform may also affect intertemporal choice. First, if crime requires

investments in crime-related human capital or learning by doing, then an increase in the expected

punishment of 14-year-old offenders may affect not only their current decision-making but also their

number of crimes committed in the future. Second, punishment itself may affect the returns to crime

versus legal activities. Serving time in a halfway house may facilitate the acquisition of criminal

human capital (creating spillovers or criminogenic effects using the term from criminology) and/or

See Polinsky and Shavell (1999) for another example of a dynamic model of crime in which the behavior of the offender

is sensitive to their time preferences.

REU, Alm.del - 2017-18 - Bilag 60: Henvendelse af 10/11-2017 fra Aarhus Universitet, Institut for Økonomi v/Anna Piil Damm vedr. "Ny forskning om virkninger af at sænke den kriminelle lavalder"

stigmatize the offender. Therefore, introducing more severe punishments to juvenile offenders may

reduce crime in the short run, but increase crime in the longer run.

Policies may affect individuals other than those initially targeted. In our case, the reform may also

have consequences for 13-year old offenders. If there are a limited number of criminal opportunities

that adolescents compete for, then offenders below the minimum age of criminal responsibility may

be substitutes for offenders above that age. Therefore, the reform may increase crime for adolescents

below the new minimum age of criminal responsibility. If, on the other hand, juveniles above the

minimum age of criminal responsibility serve as role models for juveniles below that age, then they

may be complements. If the reform is successful in deterring crime among 14-year olds, it may also

reduce crime among 13-year old individuals due to role model effects.

2.2 Prior evidence on the effects of sanctions on juvenile crime

While the literature on the effects of sanctions on adult crime is well developed,

research on the

effects of sanctions on juvenile crime is still limited. The literature falls into two broad strands: the

literature on general deterrence effects of changes in the severity (rather than certainty) of punishment

of juvenile crime and the literature on the effects of experiencing punishment for juvenile crime on

reoffending and education.

Previous empirical studies on general deterrence effects aim at measuring the effect of the risk of

juvenile incarceration. A number of empirical studies in criminology have estimated general

deterrence effects of transferring juveniles charged with serious crimes to adult court instead of

juvenile court handling using time-series data on arrest rates for US states (see e.g. Singer and

McDowall 1988; Jensen and Metsger 1994; Risler, Sweatman and Nackerud 1998; Steiner and

Wright 2006). They have exploited changes in the American system of juvenile justice which since

the mid-1970s has been moving away from its traditional rehabilitative orientation toward a model

based on the adult criminal justice system. Legislative waiver policies usually set a minimum age of

14 to 16 years for automatic transfer to criminal court (Feld 1987). The offenses covered by legislative

One strand of literature considers the effect of sentencing policy generally, or sentence enhancements on crime, see e.g.

Loftin, Heumann and McDowall (1983); Helland and Tabarrok (2007); Drago Galbiati and Vertova (2009); Buonnano

and Raphael (2013); Kilmer, Nicosia, Heaton and Midgette (2013); Bell, Jaitman and Machin (2014). Another strand of

literature investigates the effect of a capital punishment regime or the incidence of executions on murder, see e.g. Grogger

(1991); Cochran, Chamblin and Seth (1994); Donohue and Wolfers (2005); Land, Teske and Zhang (2009). For an

overview of this literature, see the recent literature reviews by Chalfin and McCrary (2017) and Nagin (2013).

REU, Alm.del - 2017-18 - Bilag 60: Henvendelse af 10/11-2017 fra Aarhus Universitet, Institut for Økonomi v/Anna Piil Damm vedr. "Ny forskning om virkninger af at sænke den kriminelle lavalder"

waiver policies are typically violent crimes or other felonies (Jensen and Metsger 1994). Deterrence

is one of the major arguments underlying this growth in the use of legislative waiver.

Singer and McDowall (1988) investigated the general deterrent effect of New York’s legislative

waiver statute enacted in 1978. This Juvenile Offender Law lowered the age at which a juvenile is

mandated to be tried in criminal court to either 13 or 14 years depending on the crime. This legislation

also mandated that penalties be similar to those for adults and required that all sentences be served in

secure facilities. Singer and McDowall (1988) assesses the effect of this legislation on serious juvenile

crimes. They concluded that the law had no effect on homicides, assaults, rapes, and arsons committed

by juveniles, but that it might have prevented robberies from increasing. In line with the findings by

Singer and McDowall, the studies by Jensen and Metsger (1994), Risler, Sweatman and Nackerud

(1998), and Steiner and Wright (2006) all find that file transfer laws have little or no effect on violent

juvenile crime.

More recently, economists have used a different approach to investigate the deterrence effect of tough

punishment of juvenile crime. They have investigated the deterrence effect of the laws that govern

the age of criminal majority, since these laws generate differences in juvenile and adult punishment

(Levitt 1998). Using US state-level panel data for the period 1978-93, Levitt (1998) finds that harsher

punishment for juveniles are strongly associated with lower rates of juvenile offending. Further

investigation suggests that the crime reduction comes from general deterrence and not incapacitation

or individual deterrence. The use of individual-level data allows for estimation of the effect of

deterrence without the potential confounding influence of incapacitation that necessarily arises in

aggregate data (Mocan and Rees 2005). Using micro-level data, more recent studies by economists

have exploited the idea that laws that govern the age of majority generate large discontinuities in the

sanctions faced by individual offenders when they cross the age threshold (Lee and McCrary 2017;

Hjalmarsson 2009b; Hansen and Wadell 2014). Despite the fact that the expected sentence length for

an adult offender is more than twice as long as that faced by a juvenile offender, these studies

surprisingly find little or only weak evidence of a deterrent effect. Few studies have used an

experimental design to evaluate the deterrence effect of formal sanctions of juvenile offenders.

A notable exception is the study by Piliavin, Gartner, Thornton and Matsueda (1986). Their evaluation of 5,050

participants from three distinct groups of persons at high risk of formal sanction (including youth) support the reward

component of the rational-choice model, but fail to support the cost or deterrent component, as measured by perceived

risks of formal sanctions.

REU, Alm.del - 2017-18 - Bilag 60: Henvendelse af 10/11-2017 fra Aarhus Universitet, Institut for Økonomi v/Anna Piil Damm vedr. "Ny forskning om virkninger af at sænke den kriminelle lavalder"

The idea in our study is instead to exploit a change in the laws that govern the minimum age of

criminal responsibility for investigation of the deterrence effect of formal sanction of juvenile

offenders as these laws generate large discontinuities in the sanctions faced by young juvenile

offenders. We thereby widen the scope to the youth population at the fringes of the criminal justice

system. Besides the age group under consideration, our approach differs from the earlier studies on

the age of majority in two important ways. For one, we exploit detailed monthly criminal records for

a sample of individuals turning 14 before and after the reform. This allows us to handle potential

confounding age- and time trends in crime in our estimations. Second, earlier studies examine the

effects of differences in the severity of sanctions within the criminal justice system, namely between

the juvenile and adult criminal justice system, while we examine the differences in the severity of

sanction between the social system and criminal justice system.

The second broad strand of literature on deterrence of juvenile crime investigates the effects of

experience of juvenile punishment on reoffending, also referred to as individual deterrence (Bell,

Jaitman and Machin 2014) or specific deterrence by criminologists (Nagin 2013). Within this

literature, one line of studies explores the effects of experiencing soft sanctions, like having police

contact, being arrested and convicted in court for criminal offenses. Investigation of such effects

necessarily requires individual-level panel data and an empirical approach, which deals with negative

selection into arrests, prosecution and conviction. There is a large literature on this in criminology.

Most existing criminological studies attempt to identify a causal link by controlling for observed

individual characteristics (e.g. match individuals on self-reported delinquency). Such studies include

classic labeling studies using self-report data and official records (e.g. Farrington 1977; Bernburg and

Krohn 2003; Morris and Piquero 2013; Wiley, Slocum and Esbensen 2013; Liberman et al. 2014) and

studies of the effects of formal sanctions on education (see e.g. De Li 1999; Tanner, Davies and

O’Grady 1999; Bernburg and Krohn 2003; Sweeten 2006; Hjalmarsson 2008). The results of these

studies are consistent with the hypothesis that public labeling increases deviant behavior among youth

who are stopped or arrested by the police or processed in the criminal justice system. It is a challenge,

of course, to properly handle selection in this context. One example of a criminological study

applying more rigorous methods is Loeffler and Grundwald (2015) who investigate a reform in

Illinois that raised the maximum age of juvenile court from 17 to 18 for offenders who commit a

misdemeanor. Using a Difference-in-Difference approach with multiple control groups, they find

little evidence of an effect. Finally, the systematic review by Petrosino, Turpin-Petrosino and

Guckenburg (2010) examines 29 randomized controlled trials and quasi-experimental studies on the

REU, Alm.del - 2017-18 - Bilag 60: Henvendelse af 10/11-2017 fra Aarhus Universitet, Institut for Økonomi v/Anna Piil Damm vedr. "Ny forskning om virkninger af at sænke den kriminelle lavalder"

effects of experiencing traditional juvenile system processing on future criminal behavior. The review

finds no evidence that traditional juvenile system processing has crime control effects. In fact,

juvenile system processing increased subsequent delinquency compared to division to programs,

counselling, or doing ‘nothing‘.

Within the second broad strand of literature on deterrence of juvenile crime, another line of studies

investigates the effects of juvenile offenders experiencing tough sanctions, like juvenile incarceration.

The criminological study by Loughran, Mulvey, Schubert Fagan, Piquero and Losoya (2009)

estimates the effects of sentencing length for serious juvenile offenders and find null effect of

institutional placement on future rearrest rates and self-reported crime. Studies applying methods that

are more sophisticated also control for unobserved household fixed characteristics. Using this

approach, Hjalmarsson (2008) finds that incarcerated juveniles have lower propensity to be

reconvicted of crime. Besides unobserved heterogeneity, a second complicating factor in identifying

the causal effect of juvenile incarceration is that effects for juveniles on the margin of juvenile

incarceration may differ from the average juvenile, and the former group is most likely to be affected

by policy changes (Aizer and Doyle 2015). One way to deal with this is to use a regression-

discontinuity design exploiting sentencing rules to identify the impact of juvenile incarceration on

recidivism as done by Hjalmarsson (2009a). The study finds that at the margins where the sanctioning

becomes more severe, juveniles just above the threshold were found to be less likely to recidivate.

Aizer and Doyle (2015) instead exploit exogenous variation in juvenile detention stemming from the

random assignment of cases to judges who vary in their sentencing. With this strategy, they address

the issue of negative selection into juvenile incarceration and estimate effects for those at the margin

of incarceration where the judge assignment matters for the incarceration decision. Their findings

suggest that juvenile incarceration reduces the probability of high school completion and increases

the probability of adult incarceration.

Our study also adds to this small literature which exploits natural experiments like random assignment

of judges with different sentencing practices (e.g. Aizer and Doyle 2015) or social experiments (e.g.

Klein 1986) to estimate the effects of experiencing formal sanction on reoffending of juvenile

offenders.

REU, Alm.del - 2017-18 - Bilag 60: Henvendelse af 10/11-2017 fra Aarhus Universitet, Institut for Økonomi v/Anna Piil Damm vedr. "Ny forskning om virkninger af at sænke den kriminelle lavalder"

3. The minimum age of criminal responsibility and the reform

Denmark has no separate juvenile justice system. Juvenile offenders above the minimum age of

criminal responsibility are sentenced by the same criminal law and in the same courts as adult

offenders. Within this institutional setting, the minimum age of criminal responsibility is a significant

threshold for young offenders, as it not only defines whether the individual’s case is handled by the

social authorities or in the criminal justice system, but also whether they obtain a criminal record.

In comparison to the many legal systems with separate juvenile justice courts (e.g. in Scotland,

England, or states in the US), the Nordic countries have a higher minimum age-limit of criminal

responsibility (today 15 in all countries) and typically process criminal cases concerning children

under the age limit in the social service system. At the same time, it is the minimum age of criminal

responsibility (and not the age of majority) that in Denmark demarcates the line for prosecution in

the adult criminal justice system and for the receipt of a formal criminal record.

Processing of a juvenile suspect of a criminal offence differs significantly between offenders below

and offenders above the minimum age of criminal responsibility. The local social authorities in

Denmark handle cases with “as if charges” of a criminal offence

below

the minimum age of criminal

responsibility. The police can investigate the criminal act and detain the offender (if the conditions

of arrest are fulfilled), but the case is not presented before a judge and the offender is under normal

circumstances not entitled to legal representation. A social worker, who must be present during any

police interrogation, accompanies the offender and not a lawyer. Hence, it is solely up to the local

social authorities to settle the case and decide upon the enactment of further measures described in

the law of social services. Examples of consequences are participation in supervised activities,

support to the family and in the most severe cases out-of-home-placement in foster care or in an

institution,

see Kyvsgaard (2004).

Cases with juveniles

above

the minimum age of criminal responsibility are handled in the criminal

justice system under the same rules of investigation, prosecution, and court ruling as cases with

offenders over the majority age (eighteen). The types of sanctions are in most cases identical for adult

and juvenile offenders and range from fines, charge withdrawal with conditions and community

service to suspended and unsuspended prison sentences (Kyvsgaard 2004). The criminal law does

Juvenile offenders aged 10-17 who commit serious crimes are confined in the same highly secure institutions regardless

of whether they are under or above the minimum age of criminal responsibility.

REU, Alm.del - 2017-18 - Bilag 60: Henvendelse af 10/11-2017 fra Aarhus Universitet, Institut for Økonomi v/Anna Piil Damm vedr. "Ny forskning om virkninger af at sænke den kriminelle lavalder"

contain provisions to limit the severity of the sentencing of offenders aged 15 to 17 and some special

sanctions and measures to be used in cases with juvenile offenders.

In the relative few cases when

juvenile offenders are arrested, placed in custody, or sentenced to a prison term, they are typically

confined in secure institutions or under special circumstances in separate units in the adult prison

system (Clausen 2013).

Juvenile offenders above the minimum age of criminal responsibility who

are found guilty also receive a criminal record. The timeframe for which convictions remain on a

criminal record depends on the sentence type; records of fines and charge withdrawal with conditions

are erased after 1 year,

suspended sentences after 3 years, and prison sentences after 5 years.

Criminal offenses committed by 14-year-olds during the reform period are most frequently property

crimes (68 percent) such as shoplifting (26 percent), petty theft (9 percent), vandalism (9 percent) and

burglary (7 percent). Other offense types are violence (13 percent), traffic-related crimes (9 percent)

and weapon and drug offenses (6 percent). In 2011, 71 percent of the cases with conviction of a

juvenile offender aged 14 led to a fine, 18 percent to a suspended sentence and 2 percent to a prison

sentence (incl. youth sanctions) (Statistics Denmark, 2012). Therefore, our study of the impact of

severity of sanctions on juvenile offending is dominated by softer crimes leading to softer sanctions,

in particular fines.

The brief outline of the Danish criminal system shows significant differences in the sanctions of

juvenile offenders below or above the minimum age of criminal responsibility. This is the feature of

the criminal justice system that we exploit in this paper to investigate the impact of the severity of

softer sanctions on young people’s offending.

We exploit a reform of the criminal law changing the minimum age of criminal responsibility in

Denmark. From 1930 to 2010, the minimum age of criminal responsibility was 15 in Denmark. As of

July 1, 2010, the right-wing government changed the penal law and lowered the age of criminal

First, in the Danish legal system it is in general considered a mitigating circumstance that the offender is under eighteen

at time of the offense. Second, if a juvenile offender is already subject to measures authorized by social legislation for

children, charges can be withdrawn if he/she makes an unqualified confession. Third, offenders under the majority age

cannot be sentenced a lifetime conviction (Storgaard 2013).

On a given day, around 10 juvenile offenders above the minimum age of criminal responsibility serve in separate units

in the adult prison system (Clausen, 2013).

This category (fines and charge withdrawal with conditions) is the only one with a reduction in the duration of a criminal

record to offenders under the age of 18; otherwise, the terms are the same as for adult offenders. In all cases, the police

keep an official record of the criminal offenses for 10 years to be available under particular circumstances

(https://www.politi.dk, accessed on March 13, 2017).

REU, Alm.del - 2017-18 - Bilag 60: Henvendelse af 10/11-2017 fra Aarhus Universitet, Institut for Økonomi v/Anna Piil Damm vedr. "Ny forskning om virkninger af at sænke den kriminelle lavalder"

responsibility to 14. Within a timeframe of 20 months, however, the age limit was re-established at

15 by a new left-wing government as of March 1, 2012.

The government had set up a commission on juvenile crime who worked from December 2007 to

September 2009 to suggest how to target and improve interventions to reduce juvenile crime. The

commission explicitly recommended

not

reducing the minimum age of criminal responsibility.

However, on March 17, 2010, the government introduced a bill, which lowered the minimum age of

criminal responsibility, and the bill was passed on June 1, 2010. The bill received a lot of attention in

the news and in the public in terms of marches and happenings in the larger cities.

Figure 1 documents the media attention on the policy change by plotting the number of newspaper

articles mentioning the minimum age of criminal responsibility. The histogram shows that the

attention was at a maximum in October 2009 immediately after the commission on juvenile crime

published their final recommendations. The financing of the reform was part of the annual state

budget negotiations (Storgaard 2013). This suggests that the general public was well aware of the

changes taking place, and it also suggests that people may have anticipated a reduction in the age of

criminal responsibility already 6-12 months before the bill was formally passed by Parliament. On

the one hand, this figure supports the notion that juveniles knew of the policy change.

On the other

hand, it also indicates that strategic retiming could have affected crime prior to the reform due to

anticipation, and we investigate this in detail in our empirical analyses.

For details, see Danish Ministry of Justice (2009).

The political attention also resulted in a descriptive evaluation of the reform by the Danish Ministry of Justice (2015).

In case 13- and 14-year-olds or their parents do not follow the news, such information is also part of the school

curriculum in social studies in 8

and 9

grade. Furthermore, all police districts have an interdisciplinary framework for

prevention of juvenile crime involving the schools, the social services and the police (denoted SSP). This is a network of

relevant authorities who collaborate with the purpose of preventing juvenile crime in the local area, e.g. by visiting

schools. They are concerned with general, specific as well as individual-oriented policies and interventions.

REU, Alm.del - 2017-18 - Bilag 60: Henvendelse af 10/11-2017 fra Aarhus Universitet, Institut for Økonomi v/Anna Piil Damm vedr. "Ny forskning om virkninger af at sænke den kriminelle lavalder"

Figure 1. Number of national newspaper articles about the minimum age of criminal responsibility,

2008-2012.

160

140

120

100

Reform

July 1, 2010

Jul 2008

Jul 2009

Jul 2010

Jul 2011

Sep 2008

Sep 2009

Sep 2010

Sep 2011

Jul 2012

May 2009

May 2010

May 2011

May 2012

Mar 2008

Mar 2009

Mar 2010

Mar 2011

Nov 2008

Nov 2009

Nov 2010

Nov 2011

Mar 2012

Sep 2012

Jan 2008

May 08

Jan 2009

Jan 2010

Jan 2011

Jan 2012

Notes: The figure shows number of articles in 17 Danish national newspapers mentioning the term “minimum age of criminal

responsibility”, total number of articles is 694.

Data source: Constructed based on data from Infomedia.dk/mediearkivet/.

The policy reform constituted a natural experiment creating exogenous variation in the age limit for

prosecution, conviction, and sanctions (including a criminal record) of young offenders in the Danish

criminal justice system. The reform introduced a more severe punishment to a younger age group

from one day to the next as the age of criminal responsibility was lowered to 14. We study this change

in the risk of criminal justice involvement of juvenile offenders over time and investigate the potential

deterrent effect of the policy reform: Does lowering the age-limit from 15 to 14 deter 14-year-olds

from committing crimes?

A behavioral response relies on a general awareness of the change and a general understanding that

the severity of the punishment increases when offenders cross the minimum age of criminal

responsibility.

Nov 2012

REU, Alm.del - 2017-18 - Bilag 60: Henvendelse af 10/11-2017 fra Aarhus Universitet, Institut for Økonomi v/Anna Piil Damm vedr. "Ny forskning om virkninger af at sænke den kriminelle lavalder"

4. Data

Primary data sources and samples:

We extract our data from eight primary sources. Our key data set

is The Central Police Register that records individual charges (date of charge, date of committing the

offense, and type of offense) and convictions (date of conviction, verdict, and sentence) for the full

Danish population as well as all contacts of individuals and firms with the police. We first merge the

Central Police Register with the Population register containing individual demographic

characteristics such as birth date, gender, current residence, parent identifiers, household

composition, country of origin, and immigrant status (available 1980-2014). We next add a series of

other registers: the Education Register and Surveys, which records educational attainment (1980-

2014); the Lower Secondary School Grade register with information on grades obtained in the 9

grade (2002-2014), the income register (1980-2014); the occupation register (1980-2014); the

medical birth register that includes information about birth outcomes (1980-2010); the Psychiatric

Central Register that records ADHD diagnoses (1994-2010); and the Prescription Drug register

(1997-2010). See Table A1 for detailed information and primary data sources for each variable.

From these sources, we extract data for seven Danish cohorts, namely the 1993-1999 birth cohorts,

combine the information using unique personal identifiers, and compile it into a panel dataset with

monthly records of reported crimes from November 1st 2008 to January 31st 2012. The dataset

includes all children aged 13, 14, and 15 during this period. To ensure that we have available pre-

reform information about background characteristics across all individuals, we restrict the sample to

consist of individuals with Danish residence January 1st in the year they turn 10.

Table 1 lists the

covariates included in the analyses (see Table A2 in the appendix for means and standard deviations

of the covariates).

Criminal offenses and crime outcomes:

In our analysis, we measure individual criminal activity based

on charges for offenses against the criminal code, which are recorded in the charge registers from the

minimum age of criminal responsibility.

Charges are usually a predecessor of a conviction, i.e. court

rulings that the individual is guilty as charged.

In principle, a small share of these individuals may have left the country after their 10

birthday. As long as this

tendency is balanced with regards to exposure to the reform, our conclusions will be unaffected by emigration. It is, in

theory, possible that the reform will lead to more emigration; the appropriate way to interpret our estimates, then, is as

the total effect of the reform, including those effects stemming from out-migration.

We observe all charges even after the criminal record has been deleted from the individual’s file.

For a random 10 percent sample of Danes born in 1980 followed until age 21, 28% of charges led to a conviction

(Damm and Dustmann 2014). US studies on crime tend to measure individual crime by arrests. But arrests are uncommon

REU, Alm.del - 2017-18 - Bilag 60: Henvendelse af 10/11-2017 fra Aarhus Universitet, Institut for Økonomi v/Anna Piil Damm vedr. "Ny forskning om virkninger af at sænke den kriminelle lavalder"

In addition to recording arrests, charges and convictions, the Danish Police records all contacts with

individuals and firms. If an offense is committed by a person who is below the minimum age of

criminal responsibility, the police will record the offense as an ”as if charge”, meaning that the person

would have been charged with the offense, had the person reached the minimum age of criminal

responsibility. The Danish Police is required by law to register offenders below the minimum age of

criminal responsibility in the Central Police Register, if they violate the Penal Code Act, Weapons

Act or the Drugs Act.

Moreover, the Police reports crime detection rates for violations of the Penal

Code Act, Weapons Act or the Drugs Act on a quarterly basis as part of their quarterly publication of

key statistics. To achieve a high crime detection rate, the Police has an incentive to find and register

offenders of these laws, irrespective of whether the offenders are above or below the minimum age

of criminal responsibility. By combining the central police registers on charges and “as if charges”,

we can measure criminal offenses throughout the entire childhood and youth for each individual.

In the Central Police Registers, charges and “as if charges” are categorized into offenses against the

Penal Code Act (sexual assault, violent crime, crimes against property and other offenses against the

penal code), offenses against the Traffic Act, offenses against the Drugs Act, offenses against the

Weapons Act, and offenses against the Tax Acts or other special acts. Throughout our analyses, we

restrict the outcomes to offenses of the Penal Code Act. We plot the number of offenses for each birth

cohort in our sample in Figure A1. In our analysis we compare criminal behavior of children aged

14 before and after the policy reform because the policy reform of interest targeted this age group.

We use two measures of criminal behavior of children aged 14: an indicator for at least one “as if

charge” or charge of the penal code in a given month and an indicator for recidivism defined as re-

offending after first penal code “as if charge” or charge at age 14.

Educational outcomes:

We study a series of educational outcomes

before age 17

including the

propensity to be enrolled in a regular 9

grade as opposed to special schools or boarding schools; the

propensity to participate in the 9

grade exam; and exam grades (standardized at the cohort level;

mean zero and standard deviation of one). We distinguish between exam grades in language arts and

math.

in Denmark. According to the Danish “Law on Administration of Justice” (Retsplejeloven. Article 755, part 1), the police

can arrest a person whom they have reason to suspect guilty of a criminal offence subject to public prosecution, but only

if an arrest is regarded as necessary in order to prevent further criminal offenses, ensure the subject’s presence for the

time being or to prevent his communication with other people. Further, an arrest should not be made if imprisonment

would be a disproportionate measure in regard of the nature of the offence or other circumstances.

BEK nr. 881 af 04/07/2014.

REU, Alm.del - 2017-18 - Bilag 60: Henvendelse af 10/11-2017 fra Aarhus Universitet, Institut for Økonomi v/Anna Piil Damm vedr. "Ny forskning om virkninger af at sænke den kriminelle lavalder"

Our analyses control for child background variables, child criminal history from age 10 to 13, and

parental background variables at age 9. See Table 1 for details.

Table 1. List of covariates from official administrative register data

Child background

Gender, age, ethnicity, birth weight, premature birth, ADHD (diagnosis

and use of prescription ADHD drugs) before age 9, use of other

psychotropic drugs before age 9.

Age of criminal onset, number of prior offenses, type of prior offenses.

Family structure (nuclear family, parent with new partner, single parent,

not living with parents), education level mother/father, income

mother/father, labor market position mother/father, convicted of criminal

offense mother/father, suspended or unsuspended prison sentence

mother/father, police district based on residence.

Child crime history

(age 10-13)

Family background

(child age 9)

5. General deterrence effects of the minimum age of criminal responsibility

The first goal of this paper is to estimate the effect of being above the minimum age of criminal

responsibility on the probability of committing crime. In doing so, we exploit the July 1, 2010 policy

reform that lowered the age of criminal responsibility from 15 to 14. Observations exposed to the

reform are individual-month observations for children who i) turned 14 in the months after the

introduction of the reform (post-reform sample) or ii) turned 14 up to 11 months before the reform

and, therefore experienced part of their 14

year after the reform (pre-/post reform sample), using

only their monthly observations

after

the reform. Observations

not

exposed to the reform include

children who i) experienced their entire 14

year in the months prior to the reform (post-reform

sample) and ii) children who had turned 14 prior up to 11 months before to the reform, using their

monthly observations

before

the reform (pre-/post reform sample). Since the reform was in place for

20 months, we use a bandwidth of 20 months around each side of the reform date.

Alternatively, one could have analyzed the effects of turning 14, using data after the introduction of the reform only,

which would be similar to the approach used in studies exploiting the laws that govern the age of majority (e.g.

Hjalmarsson 2009b; Levitt 1998). We prefer our strategy to this because the specific policy reform constituted a rather

salient change, the exact date of which could not be anticipated. Individuals can easily foresee their own birthday and

adjust their behavior in a period before the actual event. In addition, other things could change with birthday too. Figure

A2, however, shows descriptive evidence that turning 14 does not affect crime rates, neither before nor after the

introduction of the reform.

REU, Alm.del - 2017-18 - Bilag 60: Henvendelse af 10/11-2017 fra Aarhus Universitet, Institut for Økonomi v/Anna Piil Damm vedr. "Ny forskning om virkninger af at sænke den kriminelle lavalder"

Our empirical specification is the following:

(1)

where

month

is an outcome indicating whether child is charged with at least one penal code offense in

is an indicator of whether the reform is in place for child in month

first-degree polynomial in distance in time to the reform date that is allowed to differ on both sides

of the cutoff (to control for the downward trend in crime during our observation period);

consists

of age (measured in months) using a linear specification or indicators (to control for the crime-age

relationship) and calendar month indicators (to control for seasonal variation in juvenile crime),

whereas

consists of variables measured at a fixed point in time (child age 9): child characteristics

and parental background controls and police district fixed effects (to control for local variation in

police resources to apprehend and charge offenders and local variation in criminal and legal income

opportunities).

is an unobserved error term.

is the parameter of interest; it measures the effect

of the introduction of formal sanctions for crime committed by 14-year-old children.

Our primary sample consists of the population 14-year-olds in a window starting from 20 months

prior to the policy reform and ending 20 months after the introduction of the reform. We also

investigate potential spillovers of the reform to 13-year-olds using a similar observation window and

perform a placebo analysis using the sample of 15-year-olds.

The key identifying assumption in the current setup is that, except for the policy change, all other

factors, observed and unobserved, are continuous with respect to time. The assumption implies that

14-year-olds must not reschedule the timing of crime to take place immediately before the reform in

an attempt to avoid the new, stricter regime; see McCrary (2008). In the empirical analyses, we

investigate potential violation of this assumption by studying crime rates of 13- and 15-year-olds

around the reform. It also implies that the police must not change their policing or reporting strategies

from one day to the next because of the reform.

We check this assumption by looking at the distribution of reported crime across ages and across

types of penal code offenses before and after the reform. To further investigate this issue and perform

a first, descriptive, investigation of the impact of the reform, in Figure 2 we plot the distribution of

∙

is a

Our results are robust to using higher-order polynomials.

REU, Alm.del - 2017-18 - Bilag 60: Henvendelse af 10/11-2017 fra Aarhus Universitet, Institut for Økonomi v/Anna Piil Damm vedr. "Ny forskning om virkninger af at sænke den kriminelle lavalder"

monthly crime rates of 13-, 14- and 15-year-olds around the time of the introduction of the reform.

The monthly crime rate is defined as the number of individuals with a least one (as if) charge of the

Penal Code Act per individual in the age group. Over the reform period, we see a tendency for crime

rates to go down for all age groups, in line with international trends, but also an uptake in crime with

age. Crime rates increase over the summer following the reform but higher crime rates over the

summer is a general tendency present in all years (and for all age groups) and illustrates why our

formal analyses must account for calendar months. Importantly, we see no tendency for bunching of

crime of 14-year-olds just prior to the introduction of the reform; this would have indicated strategic

retiming of crime. This picture is also confirmed when we look at the reported offending rates in the

months around the 14th birthday for juveniles affected by the reform in Figure A2.

Another way to gauge the validity of the design is to look for discontinuities in observable

characteristics. We do this by comparing observable background characteristics of the sample of 14-

year-olds on either side of the introduction of the reform. As expected and shown in Table A3 and

Figure A3, we find no evidence of any economically significant differences. Furthermore, when we

look at the age-crime-curves for our estimation samples in Figure A1 there are no signs of the official

reporting of offenses changing with the minimum age of criminal responsibility (or the reform).

Table 2 shows the results from our formal analyses with gradually richer conditioning sets. All models

control for distance in time to the reform. We present the results for six different specifications. Each

specification contains more controls than the previous specification. Standard errors are clustered at

the individual level. Our preferred specification VI shows a small positive coefficient, corresponding

to 10% increase in the crime rate among 14-year-olds in July 2010. Yet the estimated effect is

miniscule and not significant at conventional levels. The findings suggest that there are no general

deterrent effect of lowering the minimum age of criminal responsibility. We can in fact rule out

general deterrent effects as small as -0.00012 (corresponding to less than 10% of the crime rate in the

relevant age group).

We have performed a range of robustness checks (see Table A4). Our baseline result for the preferred

specification is repeated at the top of the table (bandwidth July 1, 2010 +/-20 months) and robustness

checks follow in subsequent rows. The first set of checks are standard for the regression discontinuity

design: including an indicator variable for the cut-off month, including polynomials of the assignment

variable, extending and reducing bandwidth and employing a ‘donut hole’ strategy where June-

August are excluded from the analyses. In addition, we test whether our results are robust to inclusion

REU, Alm.del - 2017-18 - Bilag 60: Henvendelse af 10/11-2017 fra Aarhus Universitet, Institut for Økonomi v/Anna Piil Damm vedr. "Ny forskning om virkninger af at sænke den kriminelle lavalder"

of observations for children who turn 14 after the minimum age of criminal responsibility was re-

established at 15 (reform implemented 1

of March, 2012). Results are robust to these specification

changes. The second set of checks study potential announcement effects. First, we exclude individuals

turning 14 years between the announcement of the reform and the actual implementation of the

reform, and then we investigate whether the real response happened already when the media debate

took off on October 1

, 2009 (see Figure 1). We find no evidence that youths react to these

announcements.

Figure 2. Monthly prevalence rates for reported (as if) charges of penal code offenses for 13-,

14- and 15-year-olds

Data source: Administrative register data from Statistics Denmark and police records for birth cohorts 1993-1999.

Note: The figure shows the percentage of individuals in the age group with at least one (as if) charge for a penal code

offense in a given month. Population of 13-, 14- and 15-year-olds in Denmark 20 months (March 2008) pre reform of the

minimum age of criminal responsibility in Denmark (July 2010) to 19 months post reform (February 2012).

We proceed to investigate whether the reform had spillover effects on 13-year-olds, using the same

set of model specifications. Table 3 presents our results. We find no evidence that the reform had any

impact on the group of younger children. We similarly study whether the reform affected crime rates

among 15-year-olds. Since this group was above the minimum age of criminal responsibility both

REU, Alm.del - 2017-18 - Bilag 60: Henvendelse af 10/11-2017 fra Aarhus Universitet, Institut for Økonomi v/Anna Piil Damm vedr. "Ny forskning om virkninger af at sænke den kriminelle lavalder"

before and after the introduction of the reform, we consider this a placebo-type-analysis. As expected,

we find no evidence that crime rates of 15-year-olds changed because of the reform.

The upper panel of Table 4 next investigates whether effects vary by subgroups. Results are robust to

excluding children who committed crime prior to age 14, but the point estimate increases substantially

and becomes positive and statistically significant for individuals who committed crime prior to age

14, which point towards effects on recidivism. The lower panel of Table 4 distinguishes between

different types of penal code offenses, and reveals no significant effects.

As a final robustness analyses, we perform simpler OLS regressions where we pool information for

each individual into one single outcome (any offense/more than one offense before/at a given age).

Results are shown in Table A5 and support the finding reported in Table 2 that there is no general

deterrent effect of the reform on crime committed up until age 15. However, at the bottom of the

table, we see that individuals exposed to the reform (when they were 14 years old) have significant

higher

probability of a penal code offense at age 15. Table A6 reveals that this effect is also significant

when we single out individuals without prior offenses by age 14. We interpret this as a potential

indication of higher rates of reoffending among 14-year old offenders whose criminal cases were

handled by the criminal justice system rather than the social authorities due to the reform.

investigate recidivism in detail in the next section of the paper.

This is consistent with the small positive (though insignificant) point estimate of the reform effect in Table 2.

REU, Alm.del - 2017-18 - Bilag 60: Henvendelse af 10/11-2017 fra Aarhus Universitet, Institut for Økonomi v/Anna Piil Damm vedr. "Ny forskning om virkninger af at sænke den kriminelle lavalder"

Table 2. Effects of the reform on monthly reported offending rates (penal code offenses), population: 14-year-olds

Reform effects: July 2010 - February 2012

Number of months prior the reform (20-1, otherwise 0)

Number of months during the reform (1-19, otherwise 0)

Controls:

Age month specification

Calendar month dummies

Child background

Parents background

Child crime history

Police district fixed effects

Observations

Individuals

-0.00012

(0.00013)

0.00001

(0.00001)

-0.00001

(0.00001)

1,955,508

162,959

0.00003

(0.00013)

0.00002

(0.00001)

-0.00002

(0.00001)

Linear

1,955,508

162,959

III

0.00003

(0.00013)

0.00002

(0.00001)

-0.00002

(0.00001)

Dummies

1,955,508

162,959

0.00017

(0.00015)

0.00003

(0.00001)

-0.00003

(0.00001)

Dummies

Yes

1,955,508

162,959

0.00018

(0.00015)

0.00002

(0.00001)

-0.00002

(0.00001)

Dummies

Yes

1,955,508

162,959

0.00017

(0.00015)

0.00002

(0.00001)

-0.00002

(0.00001)

Dummies

Yes

1,955,508

162,959

Data source: Administrative register data from Statistics Denmark and Police records for birth cohorts 1993-1999.

Note: The table shows estimated effects of the reform lowering the minimum age of criminal responsibility (from July 2010 to February 2012) on the monthly reported

(as if) charges (penal code offenses) for the population of 14-year-olds from November 2008 to February 2012. The estimates are coefficients from linear panel models

on the probability of an (as if) charge in a given month and each column represents a gradually richer conditioning set. Standard errors are clustered at the individual

***

level and reported in parentheses,

< 0.10,

< 0.05,

< 0.01. Control variables: number of months relative to reform, age, calendar month, gender, ethnicity,

birth weight, parents’ income, occupation and education, family type (nuclear, single parent, new partner, child not living at home), child ADHD diagnosis, child using

prescriptive drugs, child and parents’ criminal history, police district.

REU, Alm.del - 2017-18 - Bilag 60: Henvendelse af 10/11-2017 fra Aarhus Universitet, Institut for Økonomi v/Anna Piil Damm vedr. "Ny forskning om virkninger af at sænke den kriminelle lavalder"

Table 3. Effects of the reform on monthly reported offending rates (penal code offenses), population: 13- and 15-year-olds

13-year-olds: reform effects

Observations

Individuals

15-year-olds: reform effects

Observations

Individuals

Controls:

Age month specification

Calendar month dummies

Child background

Parents background

Child crime history

Police district fixed effects

-0.00036

***

(0.00009)

1,903,320

158,610

-0.00053

***

(0.00015)

1,972,224

164,352

-0.00019

(0.00009)

1,903,320

158,610

-0.00043

***

(0.00016)

1,972,224

164,352

Linear

III

-0.00019

(0.00009)

1,903,320

158,610

-0.00043

***

(0.00016)

1,972,224

164,352

Dummies

-0.00014

(0.00011)

1,903,320

158,610

-0.00024

(0.00018)

1,972,224

164,352

Dummies

Yes

-0.00010

(0.00011)

1,903,320

158,610

-0.00023

(0.00018)

1,972,224

164,352

Dummies

Yes

-0.00010

(0.00011)

1,903,320

158,610

-0.00023

(0.00018)

1,972,224

164,352

Dummies

Yes

Data source: Administrative register data from Statistics Denmark and Police records for birth cohorts 1993-1999.

Note: The upper part of this table shows results from analyses of spill-over effects of the reform to 13-year-olds and the lower part shows results from placebo test of

reform effects to 15-year-olds. The estimates are coefficients from linear panel models on the probability of an (as if) charge in a given month for the population of 13-

or 15-year-olds from November 2008 to February 2012. Each column represents a gradually richer conditioning set. Standard errors are clustered at the individual level

< 0.01. Control variables: number of months relative to reform, age, calendar month, gender, ethnicity, birth

and reported in parentheses,

< 0.10,

< 0.05,

weight, parents’ income, occupation and education, family type (nuclear, single parent, new partner, child not living at home), child ADHD diagnosis, child using

prescriptive drugs, child and parents’ criminal history, police district.

***

REU, Alm.del - 2017-18 - Bilag 60: Henvendelse af 10/11-2017 fra Aarhus Universitet, Institut for Økonomi v/Anna Piil Damm vedr. "Ny forskning om virkninger af at sænke den kriminelle lavalder"

Table 4. Robustness analyses on monthly reported offending rates (penal code offenses), population: 14-year-olds

Baseline result:

reform effects to 14-year-olds (obs.= 1,955,508)

Different subpopulations:

Boys (obs.=1,004,484)

Girls (obs.=951,024)

Without prior offenses by age 14 (obs.=1,935,828)

With prior offenses by age 14 (obs.=19,680)

Different outcomes (subcategories of

penal code offenses):

Violent offense (obs.= 1,955,508)

Burglary (obs.= 1,955,508)

Shoplifting (obs.= 1,955,508)

Theft of vehicles (obs.= 1,955,508)

Vandalism (obs.= 1,955,508)

Controls:

Age month specification

Calendar month dummies

Child background

Parents background

Child crime history

Police district fixed effects

III

-0.00012 0.00003

0.00003

0.00017 0.00018 0.00017

(0.00013) (0.00013) (0.00013) (0.00015) (0.00015) (0.00015)

-0.00005

(0.00021)

-0.00018

(0.00015)

-0.00021*

(0.00012)

0.00878*

(0.00527)

-0.00003

(0.00005)

-0.00004

(0.00004)

-0.00008

(0.00007)

0.00000

(0.00005)

-0.00006

(0.00004)

0.00017

(0.00021)

-0.00011

(0.00015)

-0.00009

(0.00012)

0.01009*

(0.00522)

0.00002

(0.00005)

-0.00002

(0.00004)

-0.00004

(0.00007)

0.00003

(0.00005)

-0.00004

(0.00004)

Linear

0.00017

(0.00021)

-0.00011

(0.00015)

-0.00009

(0.00012)

0.01016*

(0.00524)

0.00002

(0.00005)

-0.00002

(0.00004)

-0.00004

(0.00007)

0.00003

(0.00005)

-0.00004

(0.00004)

Dummies

0.00025

(0.00024)

0.00009

(0.00016)

0.00004

(0.00013)

0.01137*

(0.00608)

0.00008

(0.00006)

-0.00001

(0.00004)

0.00005

(0.00008)

0.00001

(0.00005)

-0.00005

(0.00005)

0.00021

(0.00024)

0.00013

(0.00016)

0.00008

(0.00014)

0.00962

(0.00614)

0.00007

(0.00006)

0.00000

(0.00004)

0.00004

(0.00008)

0.00001

(0.00005)

-0.00006

(0.00005)

0.00021

(0.00024)

0.00013

(0.00016)

0.00006

(0.00015)

0.01157*

(0.00662)

0.00007

(0.00006)

0.00000

(0.00004)

0.00004

(0.00008)

0.00001

(0.00005)

-0.00006

(0.00005)

Data source: Administrative register data from Statistics Denmark and Police records for birth cohorts 1993-1999.

Dummies Dummies Dummies

Yes

REU, Alm.del - 2017-18 - Bilag 60: Henvendelse af 10/11-2017 fra Aarhus Universitet, Institut for Økonomi v/Anna Piil Damm vedr. "Ny forskning om virkninger af at sænke den kriminelle lavalder"

Note: The table shows result from robustness analyses including reform effects to different subpopulations of 14-year-olds and different outcomes (selected

subcategories of penal code offenses). The reported estimates are coefficients from linear panel models on the probability of an (as if) charge in a given month

from November 2008 to January 2012 and each column represents a gradually richer conditioning set. Standard errors are clustered at the individual level and

***

reported in parentheses,

< 0.10,

< 0.05,

< 0.01. Control variables: number of months relative to reform, age, calendar month, gender, ethnicity, birth

weight, parents’ income, occupation and education, family type (nuclear, single parent, new partner, child not living at home), child ADHD diagnosis, child using

prescriptive drugs, child and parents’ criminal history, police district.

REU, Alm.del - 2017-18 - Bilag 60: Henvendelse af 10/11-2017 fra Aarhus Universitet, Institut for Økonomi v/Anna Piil Damm vedr. "Ny forskning om virkninger af at sænke den kriminelle lavalder"

6. Effects on recidivism: Individual deterrence?

The analysis above showed little effects of the reform on general deterrence. Yet if anything, there

was actually a slight increase in crime rates. That new offenders drive these results would be against

theoretical predictions since the reform uniformly increased the costs of committing crime. It may be,

however, that the observed slight increase in crime rates among the 14-year-olds is due to higher

reoffending rates. We therefore investigate whether the reform lowering the age limit for processing

juveniles in the Danish criminal justice system affects young offenders’ recidivism.

In theory, reform effects on recidivism may be either negative or positive.

Interactions with the

criminal justice system at an earlier age may affect subsequent offending if such interactions change

the offender‘s perception of the costs and benefits of crime (Becker 1968; Nagin, Cullen, and Jonson

2009). These experiences may influence either the expected probability of apprehension or the

expected severity of punishments and thereby deter young people from committing criminal

offenses.

Labeling theories in criminology argue that involvement with the criminal justice system

can have a negative influence on offenders’ future outcomes. First, interactions with the official

system leading to public labeling of the young offender as deviant can cause him/her to change self-

perception that in turn raise the likelihood of engaging in deviant behaviors (Lemert 1951). Second,

institutions’ responses to the official label may reduce offenders’ opportunities in both the education

system and labor market and thereby increase the risk of delinquency (Paternoster and Iovanni 1989;

Sampson and Laub 1997). Third, the labeling of an offender can influence peer interactions by

enhancing the risk of social exclusion from conventional groups and increasing contact with deviant

groups (Lemert 1951). All three mechanisms imply that juveniles’ involvement with the legal system

is likely to increase the likelihood of subsequent criminal behavior. The net effect of juveniles’

interactions with the criminal justice system on subsequent crime may be either negative or positive

depending on which mechanisms dominate.

For this analysis, we select the 14-year-olds who were (as if) charged for a violation of the penal code

at age 14, and estimate OLS regressions of reoffending 3, 6, 9, 12, 15 and 18 months after the first

offense. We estimate the following regression equation:

(2)

In criminology, the negative effect is denoted “specific deterrence”, whereas the positive effect is denoted “labeling”.

Hjalmarsson (2009b) found evidence of changes in the perceived severity of punishment related to the age of majority,

but not for offenders who had been arrested prior to reaching the age limit.

REU, Alm.del - 2017-18 - Bilag 60: Henvendelse af 10/11-2017 fra Aarhus Universitet, Institut for Økonomi v/Anna Piil Damm vedr. "Ny forskning om virkninger af at sænke den kriminelle lavalder"

where

is the relevant outcome (indicators for whether child

is charged with a new penal code

is an

indicator for whether the first offense of child

took

place pre- or post-reform and thus whether the

offense within 3, 6, 9, 12, 15, or 18 months along with educational outcomes).

sanction was determined in the social service system or in the criminal justice system;

variables, and police district fixed effects. Again,

is the parameter of interest. Here,

is a list of

control variables: calendar month of first offense, offense age and type, child and family background

measures the

effect of committing crime under the new, stricter regime on subsequent crime and educational

outcomes. Standard errors are robust and clustered at the police district level. As a robustness check,

we estimate simple Cox Proportional Hazard models of the time to next penal code offense. As an

additional outcome, we also study educational achievement before age 17 measured by: school

enrollment, exam participation, grades in language arts and math and school type.

Our estimation sample consists of the population of 14-year-olds (as if) charged with a penal code

offense, who turned 14 twenty to thirteen months prior to the policy reform and who turned 14 in the

first eight months after the introduction of the reform. In other words, we compare 14-year-old

offenders who have their 14th year just prior to the reform with 14-year-old offenders who have their

14th year during the reform period. We restrict the sample to offenders with an (as if) charge of the

penal code. For one, these are the offenses with potential detrimental effects on future outcomes

because they result in a criminal record if the offender is found guilty. Second, we found no general

deterrent effects in terms of reduction of such offenses and this limits the risk of selection bias. We

compare observable background characteristics of the sample of 14-year-old offenders who were (as

if) charged on either side of the reform. Table A7 shows only a few significant differences. However,

they few significant differences suggest that the post-reform group is slightly negatively selected in

terms of criminal history of parents, educational attainment of the mother and use of ADHD medicine.

In view of the downward trend in crime over the observation period seen in Figures 2 and A1, this is

not a surprising finding. We include a rich set of control variables in Eq. 2 to take account of potential

selection between the pre- and post-reform group..

Table 5 shows recidivism rates at varying points in time after the first offense at age 14 and compares

offenders pre-reform who had their criminal case handled by the social authorities and offenders post-

reform who had their criminal case processed in the criminal justice system. There is a clear tendency

that 14-year-olds affected by the reform recidivates faster than 14-year-olds who committed their

REU, Alm.del - 2017-18 - Bilag 60: Henvendelse af 10/11-2017 fra Aarhus Universitet, Institut for Økonomi v/Anna Piil Damm vedr. "Ny forskning om virkninger af at sænke den kriminelle lavalder"

offense prior to the reform. The post-reform offenders have five-percentage points higher recidivism

rates after 12 to 18 months.

Table 5. Recidivism rates (penal code offenses) 14-year-old penal code offenders

Pre-reform

(obs.=893)

0.17

0.22

0.26

0.29

0.32

0.36

Post-reform

(obs.=676)

0.18

0.24

0.29

0.34

0.38

0.41

Difference

and t-test

0.01

0.02

0.04

0.05*

Data source: Administrative register data from Statistics Denmark and Police records for 14-year-old penal code

offenders.

Note: The table shows recidivism rates for 14-year-old penal code offenders with their 14

year pre- or post-reform and

the last column shows the difference between the two groups and t-test of difference in means * p<0.05, ** p<0.01, ***

p<0.001.

Recidivism after:

3 months

6 months

9 months

12 months

15 months

18 months

Table 6 shows the formal results. Specification I mirrors the difference in the raw means seen in Table

5, whereas specifications II to IV gradually include a richer conditioning set. When all control

variables are added in our preferred specification IV, we find that the probability of reoffending is

around 4-percentage point higher for post-reform offenders compared to pre-reform offenders, and

this gap is statistically significant 9 to 18 months after the first offense. This gap corresponds to 10

percent higher recidivism after 18 months for post-reform offenders. These results are robust to using

the Cox Proportional Hazard model (see Table A8 in Appendix).

The results show that interaction with the criminal justice system at 14

increases

reoffending. There

are no individual deterrent effects of offenders experiencing formal sanctions in the criminal justice

system when compared to having the case handled in the social system. Two potential explanations

of these effects are, on one hand, that the perceived costs of crime are reduced (or benefits increased)

after first interaction with the criminal justice system. In that, the offender experiences softer

sanctions than he expected. On the other hand, an explanation may be that the official sanctions create

a labeling effect by which the offenders experience social exclusion and/or institutional responses to

the public label that enhance subsequent offending.

REU, Alm.del - 2017-18 - Bilag 60: Henvendelse af 10/11-2017 fra Aarhus Universitet, Institut for Økonomi v/Anna Piil Damm vedr. "Ny forskning om virkninger af at sænke den kriminelle lavalder"

As mentioned earlier, there is a downward trend in juvenile crime over the relevant period (see Figure

2). Therefore, one concern may be whether a similar trend influences recidivism. In Table A9, we

estimate the time (or cohort) trend in recidivism based on our pre-reform sample of 14-year-old

offenders. We run OLS regressions of recidivism 3 to 18 months after the first offense on a linear

birth month variable. The trend estimates are small in magnitude and not statistically significant.

However, the point estimates in specification I are negative and may suggest a weak downwards trend

in recidivism, which could contaminate the estimated reform effect, which is not separately identified

from other time or cohort influences. In Table A10, we re-estimate the coefficients in Table 6

including the time (or cohort) trend as a control variable, allowing this trend to counteract the effect

of the reform. In this case, the parameters of main interest are roughly three times larger, although all

the trend estimates are insignificant. Hence, we regard our main results in Table 6 as conservative

estimates of the impact of processing 14-year-old offenders in the criminal justice system.

This is consistent with the conclusions by Danish Ministry of Justice (2014, 2016).

REU, Alm.del - 2017-18 - Bilag 60: Henvendelse af 10/11-2017 fra Aarhus Universitet, Institut for Økonomi v/Anna Piil Damm vedr. "Ny forskning om virkninger af at sænke den kriminelle lavalder"

Table 6. Effects of reform on recidivism (penal code offenses), population: 14-year-old penal

code offenders

Effects of reform on recidivism 3 months

0.0121

(0.0239)

0.0224

(0.0208)

0.0391

(0.0172)

0.0535

(0.0192)

0.0525

(0.0213)

0.0507

(0.0215)

1,569

0.0228

(0.0219)

0.0336

(0.0182)

0.0495

***

(0.0161)

0.0673

***

(0.0183)

0.0663

***

(0.0199)

0.0650

***

(0.0208)

Yes

1,569

III

0.0198

(0.0292)

0.0300

(0.0207)

0.0365

(0.0165)

0.0450

***

(0.0132)

0.0414

(0.0158)

0.0375

(0.0131)

Yes

1,569

0.0204

(0.0286)

0.0305

(0.0206)

0.0377

(0.0164)

0.0471

***

(0.0125)

0.0428**

(0.0148)

0.0388***

(0.0122)

Yes

1,569

Effects of reform on recidivism 6 months

Effects of reform on recidivism 9 months

Effects of reform on recidivism 12 months

Effects of reform on recidivism 15 months

Effects of reform on recidivism 18 months

Controls:

Calendar month dummies

Offense age and type

Child background variables

Family background variables

Police district fixed effects

Observations

Data source: Administrative register data from Statistics Denmark and Police records for 14-year-old penal code offenders

pre- and post-reform.

Note: The reported estimates are coefficients from OLS regressions, rows represent separate regression models on the

probability of recidivism at 3, 6, 9, 12, 15 and 18 months and each column represents a gradually richer conditioning set.

Robust standard errors are clustered at police district and reported in parentheses, * p < 0.10, ** p < 0.05, *** p < 0.01.

Control variables: offense month, offense age, type of the offense, child gender, ethnicity, birth weight, child ADHD

diagnose, child using prescriptive drugs, child’s criminal history (age 10-13), parents’ income, occupation and education,

family type, parents’ criminal history, police district.

We perform a range of heterogeneity analyses, all reported in Table A11: We study effects on 14-

year-old offenders with all types of offenses, divide the sample by gender and by prior reported

offenses, exclude violent offenders and offenders who commit vandalism. Overall, the results from

the robustness analyses are very similar to the reported findings. The background to and findings of

two of these robustness checks are relevant to highlight. First, violent offenses are among the most

severe criminal charges and could lead to an unsuspended prison sentence or to placement in a secured

REU, Alm.del - 2017-18 - Bilag 60: Henvendelse af 10/11-2017 fra Aarhus Universitet, Institut for Økonomi v/Anna Piil Damm vedr. "Ny forskning om virkninger af at sænke den kriminelle lavalder"

institution. Hence, because juveniles who commit a criminal offense while being

under

the age of

criminal responsibility have their case handled by the social authorities, the offenders with the most

serious crimes (334 individuals in our sample) risk placement in secured institution with no upper

time limit. To ensure that this potential incapacitation effect does not influence our results we run an

analysis where the population is restricted to juveniles with non-violent offenses. The results are very

similar to the results reported in Table 6. Second, we distinguish between different sub-groups

defined by prior offenses. The effect size is negative and insignificant for individuals who have a

criminal history already at ages 10-13. This is consistent with Hjalmarsson (2009b) who found

evidence of changes in the perceived severity of punishment related to the age of majority, but not

for offenders who had been arrested prior to reaching the age-limit.

The above analysis suggests that exposure to the formal criminal justice system increases recidivism,

which could indicate some type of labeling effect. The next step is to explore to what extent these

detrimental effects of the reform are also seen for educational outcomes. Table 7 shows that 14-year-

old offenders charged in the criminal justice system are less likely to be enrolled in 9

grade, less

likely to be enrolled in the ordinary lower secondary school (in 9

grade) and more likely to attend

boarding schools when compared to 14-year-old offenders who committed their offense prior to the

reform. Furthermore, offenders affected by the reform have lower grades conditional on participating

in the 9

grade exit exam (participation rates are not statistically significantly different). Hence, the

results show both higher recidivism rates and poorer educational outcomes for 14-year-old offenders

charged in the criminal justice system during the reform period.

Recidivism within 18 months and educational achievement before age 17 are closely related, but it is

not possible to disentangle the causal chain in this study. Instead, we present raw means of educational

outcomes separately for 14-year-old offenders with and without recidivism within 18 months in order

to get a sense of the mechanisms, at least descriptively (see Table A12). Unsurprisingly, individuals

who recidivate generally have much poorer educational outcomes than those who do not recidivate.

In addition, the group of recidivists affected by the reform are seven percentage points less likely to

be enrolled in 9

grade. This suggests that - although the initial crime and subsequent sanction most

often was relatively soft - the fact that the case was processed in the criminal justice system means

that they are more likely to enter an unfortunate life course. On the other hand, the group of 14-year-

old offenders who did not recidivate are equally likely to be enrolled in 9

grade, no matter whether

the offense was committed before or after the reform. Instead, the type of school is different;

REU, Alm.del - 2017-18 - Bilag 60: Henvendelse af 10/11-2017 fra Aarhus Universitet, Institut for Økonomi v/Anna Piil Damm vedr. "Ny forskning om virkninger af at sænke den kriminelle lavalder"

individuals who are affected by the reform are seven percentage points less likely to be enrolled in

ordinary secondary school and correspondingly more likely to attend boarding schools. This suggests

that some offenders are moved to another school environment with different codes of conduct, other

adults and peers, and this may explain why no further offenses are registered within 18 months.

However, these individuals do have lower grades (conditional on participating) compared to their

counterparts before the reform.

REU, Alm.del - 2017-18 - Bilag 60: Henvendelse af 10/11-2017 fra Aarhus Universitet, Institut for Økonomi v/Anna Piil Damm vedr. "Ny forskning om virkninger af at sænke den kriminelle lavalder"

Table 7. Effects of reform on educational achievement, population: 14-year-old penal code

offenders

Outcome

Enrolled in 9th grade

Exam participation (0/1)

Type of school (if enrolled 9

grade):

Ordinary lower secondary schools

Boarding schools

Schools for children with special needs

-0.0448

***

(0.0153)

0.0109

(0.0232)

-0.0408

(0.0258)

0.0428

(0.0179)

0.0075

(0.0122)

-0.0579

(0.0491)

-0.1270

***

(0.0424)

-0.0512

(0.0536)

-0.1032

(0.0520)

-0.0514

***

(0.0155)

0.0034

(0.0232)

-0.0452

(0.0261)

0.0402

(0.0178)

0.0084

(0.0125)

-0.0651

(0.0484)

-0.1396

***

(0.0424)

-0.0597

(0.0541)

-0.1045

(0.0528)

Yes

III

Observations

1,569

-0.0481

***

-0.0470

***

(0.0172) (0.0170)

0.0216

(0.0244)

-0.0379

(0.0285)

0.0457

(0.0198)

0.0030

(0.0133)

-0.0353

(0.0487)

-0.1015

(0.0422)

-0.0211

(0.0541)

-0.0561

(0.0540)

Yes

0.0223

(0.0245)

-0.0388

(0.0284)

0.0478

(0.0197)

0.0014

(0.0134)

-0.0366

(0.0493)

-0.1017

(0.0425)

-0.0117

(0.0544)

-0.0581

(0.0540)

Yes

1,421

Grades (conditional on participation):

Language Arts (teacher assessment)

Language Arts (exit exam)

Math (teacher assessment)

Math (exit exam)

Controls:

Calendar month dummies

Offense age and type

Child background variables

Family background variables

Police district fixed effects

1,016

1,021

1,008

996

Data source: Administrative register data from Statistics Denmark and Police records for 14- year-old penal code

offenders pre- and post-reform.

Note: The reported estimates are coefficients from OLS regressions, rows represent separate regression models with

different educational outcomes (conditional on enrolment/participation) before age of 17 and each column represents a

gradually richer conditioning set. Robust standard errors are reported in parentheses and * p < 0.10, ** p < 0.05, *** p <

0.01. Control variables: offense month, offense age, type of the offense, child gender, ethnicity, child’s birth weight, child

ADHD diagnosis, child using prescriptive drugs, child’s criminal history (ages 10-13), parents’ income, occupation and

education, family type, parents’ criminal history, police district.

REU, Alm.del - 2017-18 - Bilag 60: Henvendelse af 10/11-2017 fra Aarhus Universitet, Institut for Økonomi v/Anna Piil Damm vedr. "Ny forskning om virkninger af at sænke den kriminelle lavalder"

7. Conclusion and discussions

This study uses Danish administrative data to estimate the consequences of lowering the minimum

age of criminal responsibility on crime rates. We also investigate effects on recidivism and

educational outcomes. To do this, we exploit a reform in Denmark that lowered the minimum age of

criminal responsibility from 15 to 14. Much against the political intentions, we find no evidence that

the reform lowered crime rates. If anything, we detect a small uptick in crime among 14-year-olds

after the implementation of the reform. We also see that youth who committed crime during the

reform period and processed in the criminal justice system were more likely to recidivate at any given

point in time just as they experienced worse schooling outcomes than 14-year-old offenders processed

in the social system did.

The findings from the study suggest that the Danish policy reform that introduced more severe

punishments by lowering the age-limit of criminal responsibility did not have the intended deterrent

effects on criminal behaviors among 14-year-olds. These results from analyses of juveniles at the

fringes of the criminal justice systems coincides with the findings from previous studies of age-limits

within the criminal justice system where

“[the] literature around the age of criminal majority

produces little evidence of deterrence among young offenders.”(Chalfin

and McCrary, 2017:30).

Several theoretical explanations prevail for why this policy did not reduce crime rates among

juveniles. First, preconditions are that the change in the criminal law was known to the population of

14-year-olds and that this awareness of the reform increased their perception of the severity of the

punishments for criminal behavior. We show that the reform of the minimum age of criminal

responsibility was widely debated in the mass media; however, we cannot be sure that the information

about the reform and its consequences also reached the 14-year-olds at that time. Moreover, the

prevalence rates among 14-year-olds are low and therefore only few individuals in the cohorts have

experience with the consequences of the reform, either directly or indirectly (e.g. through friends or

classmates punished in the criminal justice system) (Stafford and Warr 1998). It is possible that legal

reforms like ours actually affect the perceptions of the severity of the expected punishment more

among law-abiding juveniles who would not have committed criminal acts in the first place. Second,

the police must have enforced the new law so that the perceived certainty of the punishment (e.g. risk

of apprehension or sanctions) was not at the same time reduced. We find no indications of the Danish

police changing their enforcements of the criminal law for example by being more reluctant to book

14-year-olds during the reform. Third, the perceived severity of the punishments is only one element

REU, Alm.del - 2017-18 - Bilag 60: Henvendelse af 10/11-2017 fra Aarhus Universitet, Institut for Økonomi v/Anna Piil Damm vedr. "Ny forskning om virkninger af at sænke den kriminelle lavalder"

in the decision making, so even though potential offenders are aware of the reform they may decide

to engage in crime anyway if the expected benefits (e.g. money, thrills or peer approval) exceed the

expected costs.

The results from the analyses of recidivism show that penal code offenders who were affected by the

reform and processed in the criminal justice system at the age of 14 have higher recidivism rates. This

finding has several possible explanations where the first one derives from the deterrence perspective.

Within this theoretical work the experience with the criminal justice system can change the offenders’

information of the perceived severity of the punishment and can lead to an increase in subsequent

offending if the perceived cost is reduced, for example by the experienced sanction being more lenient

than expected. Second, the results can also be explained with reference to labeling theories, which

state that the public label of criminal justice system interactions can increase criminal behavior among

juveniles. The official status as ‘criminal’ can increase future offending by formal (exclusion

conventional opportunities) or informal reactions (exclusion from non-deviant groups) and the

change of self-perception.

The latter theoretical explanation is substantiated by the results from analyses of their educational

outcomes, which show that 14-year-olds processed in the criminal justice system are less likely to be

enrolled in the 9th grade and have lower grades at exit exam, conditional on participating. These

results coincides with previous studies that found negative effects of court appearance, arrest and

incarceration to juveniles’ educational attainments (e.g. Aizer and Doyle 2015; Hjalmarsson 2008;

Sweeten 2006). Moreover, Aizer and Doyle (2015) find substantial effects of juvenile incarceration

on high school completion, which relates to lower likelihood of ever returning to school after

imprisonment and higher likelihood of receiving a classification of emotional or behavioral disorders

among those who do return to the school system.

Finally, the 14-year-olds who committed a penal code offense will receive a criminal record for one

to five years, depending on the sanction. If convicted, the information will be retained in the ‘system’

for ten years with an official record, which could influence future meetings with the police and as

well as decisions on whether or not to press charges (Petrosino, Turpin-Petrosino, and Guckenburg,

2010). In itself, this could increase the

reported

recidivism rates in the group affected by the reform.

Moreover, the results show higher enrollment rates in boarding schools, which also can be influenced

by different official reactions to offenders with and without an official label, as boarding schools are

used for both preventive measures and placements by the social authorities.

REU, Alm.del - 2017-18 - Bilag 60: Henvendelse af 10/11-2017 fra Aarhus Universitet, Institut for Økonomi v/Anna Piil Damm vedr. "Ny forskning om virkninger af at sænke den kriminelle lavalder"

Of course, our analyses only provide evidence of the total effects of the policy reform; it remains to

be studied which exact mechanisms can explain the findings.

REU, Alm.del - 2017-18 - Bilag 60: Henvendelse af 10/11-2017 fra Aarhus Universitet, Institut for Økonomi v/Anna Piil Damm vedr. "Ny forskning om virkninger af at sænke den kriminelle lavalder"

Literature

Aizer, A. and J. J. Doyle (2015). Juvenile incarceration, human capital and future crime: Evidence

from randomly-assigned judges.

Quarterly Journal of Economics,

130 (2), 759-803.

Apel, R. (2013). Sanctions, Perceptions and Crime: Implications for Criminal Deterrence.

Journal of

Quantitative Criminology,

29(1): 101-135.

Bales, W. and A. Piquero (2012). Assessing the impact of imprisonment on recidivism.

Journal of

Experimental Criminology,

8(1), 71-101.

Beccaria, C. (1764).

On Crime and Punishments.

Trans. Henry Paolucci. Englewood Cliffs. NJ:

Prentice Hall.

Becker, G. (1968). Crime and punishment: An economic approach.

Journal of Political Economy,

76, 169-217.

BEK nr. 881 af 04/07/2014. Bekendtgørelse om behandling af personoplysninger i Det Centrale

Kriminalregister.

https://www.retsinformation.dk/Forms/R0710.aspx?id=164093

Bell, B., Jaitman, L., and S. Machin (2014). Crime Deterrence: Evidence from the London 2011 Riots,

Economic Journal,

124 (May), 480-506.

Bentham, J. (1789). The Principles of Morals and Legislation. Buffalo, NY: Prometheus Books,

reprint.

Bernburg, J. G. and M. D. Krohn (2003). Labeling, life chances, and adult crime: the direct and

indirect effects of official intervention in adolescence on crime in early adulthood.

Criminology,

vol.

41, 1287-1318.

Buonnano, P. and S. Raphael (2013). Incarceration and Incapacitation: Evidence from the 2006 Italian

Collective Pardon.

American Economic Review,

103(6), 2437-2465.

Bhuller, M., Dahl, G. B., Mogstad, M., and K. V. Løken (2016). Incarceration, Recidivism and

Employment.

NBER DP

no. 22648.

Chalfin, A. and J. McCrary (2017). Criminal Deterrence: A Review of the Literature.

Journal of

Economic Literature,

55(1), 5-48.

REU, Alm.del - 2017-18 - Bilag 60: Henvendelse af 10/11-2017 fra Aarhus Universitet, Institut for Økonomi v/Anna Piil Damm vedr. "Ny forskning om virkninger af at sænke den kriminelle lavalder"

Clausen, S. (2013).

Klientundersøgelsen 2011. Delrapport om unge klienter.

Direktoratet for

Kriminalforsorgen, København.

Cochran, J.K., Chamblin, M.B., and M. Seth (1994). Deterrence or Brutalization? An Impact

Assessment of Oklahoma’s Return to Capital Punishment.

Criminology,

32(1), 107-134.

Cochran, J., Mears, D., and W. Bales (2014). Assessing the effectiveness of correctional sanctions.

Journal of Quantitative Criminology,

30(2), 317-347.

Corman, H. and H. N. Mocan (2000). A Time-Series Analysis of Crime, Deterrence and Drug Abuse

in New York City.

American Economic Review,

90(3), 584-604.

Cunha, F., Heckman, J. J.. Lochner, L. and D. V. Masterov (2006). Interpreting the Evidence on Life

Cycle Skill Formation, In: E. Hanushek and F. Welch, Editor(s),

Handbook of the Economics of

Education,

Vol. 1, 697-812. Elsevier.

Cunha, F. and J. J. Heckman (2008). Formulating, identifying, and Estimating the Technology for the

Formation of Skills.

Journal of Human Resources

43(3), 738-782.

Damm, A. P. and C. Dustmann, “Does Growing Up in a High Crime Neighborhood affect Youth Criminal

Behavior?”

American Economic Review

2014, 104(6): 1806-1832.

Danish Ministry of Justice (2009). Measures to combat juvenile crime. (In Danish: Indsatsen mod

ungdomskriminalitet). White Paper no. 1508/2009 from the Commission on Juvenile Crime.

Danish Ministry of Justice (2014). Udviklingen i børne- og ungdomskriminalitet 2001-2013.

Justitsministeriets Forskningskontor.

Danish Ministry of Justice (2015). Udviklingen i børne- og ungdomskriminalitet 2001-2014.

Justitsministeriets Forskningskontor.

Danish Ministry of Justice (2016). Udviklingen i børne- og ungdomskriminalitet 2006-2015.

Justitsministeriets Forskningskontor.

De Li, S. (1999). Legal Sanctions and Youth’s Status Achievement: A Longitudinal Study.

Justice

Quarterly,

16, 377-401.

Di Tella, R. and E. Schargrodsky (2013). Criminal recidivism after prison and electronic monitoring.

Journal of Political Economy,

121(1), 28-73.

REU, Alm.del - 2017-18 - Bilag 60: Henvendelse af 10/11-2017 fra Aarhus Universitet, Institut for Økonomi v/Anna Piil Damm vedr. "Ny forskning om virkninger af at sænke den kriminelle lavalder"

Donohue, J.J. and J. Wolfers (2005). Uses and Abuses of Empirical Evidence in the Death Penalty

Debate.

Stanford Law Review,

58, 791-846.

Drago, F., Galbiati, R., and P. Vertova, (2009). The Deterrence Effect of Prison: Evidence from a

Natural Experiment."

Journal of Political Economy,

107 (2), 257-280.

Ehrlich, I. (1973). Participation in Illegitimate Activities: A Theoretical and Empirical Investigation.

Journal of Political Economy,

81(3), 521-565.

Farrington, D.P. (1977). The Effects of Public Labelling.

British Journal of Criminology.

17(2), 112-

115.

Feld, B. (1987). The Juvenile Court Meets the Principle of the Offense: Legislative Changes in

Juvenile Waiver Statutes.

Journal of Criminal Law and Criminology,

78, 471-533.

Green, D. P. and D. Winik (2010). Using random judge assignments to estimate the effects of

incarceration and probation on recidivism among drug offenders.

Criminology,

48(2), 357-388.

Grogger, J. (1991). Certainty versus Severity in Punishment.

Economic Inquiry,

29(2), 297-309.

Hansen, B. and G.R. Waddell (2014). Walk Like a Man: Do Juvenile Offenders Respond to Being

Tried as Adults? Unpublished Manuscript.

Heckman, J. J. (2008). Schools, Skills, and Synapses.

Economic Inquiry

46, 289–324

Helland, E. and A. Tabarrok (2007). Does Three Strikes Deter? A Nonparametric Estimation.

Journal

of Human Resources,

42, 309-330.

Hjalmarsson, R. (2008). Criminal Justice Involvement and high school completion.

Journal of Urban

Economics,

63(2), 613-630.

Hjalmarsson, R. (2009a). Juvenile Jails: A Path to the Straight and Narrow or Hardened Criminality?.

Journal of Law and Economics

52(4), 779-809.

Hjalmarsson, R. (2009b). Crime and Expected Punishment: Changes in Perceptions at the Age of

Criminal Majority.

American Law and Economics Review

11(1), 209–248.

Jensen, E.L. and Metsger, L. K. (1994). A Test Deterrent Effect of Legislative Waiver on Violent

Juvenile Crime.

Crime & Delinquency,

40(1), 96-104.

REU, Alm.del - 2017-18 - Bilag 60: Henvendelse af 10/11-2017 fra Aarhus Universitet, Institut for Økonomi v/Anna Piil Damm vedr. "Ny forskning om virkninger af at sænke den kriminelle lavalder"

Kilmer, B., Nicosia, N., Heaton, P., and G. Midgette (2013). Efficacy of Frequent Monitoring With

Swift, Certain and Modest Sanctions for Violations: Insights from South Dakota’s 24/7 Sobriety

Project.

American Journal of Public Health,

103(1), 37-43.

Klein, M.W. (1986). Labeling theory and delinquency policy: An experimental test.

Criminal Justice

and Behavior,

13(1), 47-79.

Kling, J. R. (2006). Incarceration Length, Employment, and Earnings.

American Economic Review,

96(3), 863-876.

Kyvsgaard, B. (2004). Youth Justice in Denmark.

Crime and Justice

31, Youth Crime and Youth

Justice: Comparative and Cross-National Perspectives, 349-390.

Land, K.C., Teske, R.H., and H. Zhang (2009). The Short-Term Effects of Executions on Homicides:

Deterrence, Displacement or Both?

Criminology,

47(4), 1009-1043.

Landersø, R. (2015), Does Incarceration Length Affect Labor Market Outcomes?

Journal of Law and

Economics,

58, 205-234.

Lee, D. and J. McCrary. (2017). The Deterrence Effect of Prison: Dynamic Theory and Evidence.

Advances in Econometrics,

38, 73-146.

Lemert, E. (1951).

Social pathology.

New York: McGraw-Hill.

Levitt, S. (1998). Juvenille Crime and Punishment.

Journal of Political Economy,

106 (6), 1156-

1185.

Liberman, A. M., Kirk, D. S., and K. Kim. (2014). Labeling effects of first juvenile arrests: secondary

deviance and secondary sanctioning.

Criminology,

52, 345-370.

Loeffler, C.E. and B. Grundwald (2015). Decriminalizing Delinquency: The Effect of Raising the

Age of Majority on Juvenile Recidivism.

Journal of Legal Studies,

44(2), 361-388.

Loeffler, C. E. and A. Chalfin (2017). Estimating the Crime Effects of Raising the Age of Majority.

Criminology & Public Policy,

16, 45–71.

Loftin, C., Heumann, M., and D. McDowall (1983). Mandatory Sentencing and Firearms Violence:

Evaluating an Alternative to Gun Control.

Law and Society Review,

17, 287-318.

REU, Alm.del - 2017-18 - Bilag 60: Henvendelse af 10/11-2017 fra Aarhus Universitet, Institut for Økonomi v/Anna Piil Damm vedr. "Ny forskning om virkninger af at sænke den kriminelle lavalder"

Loughran, T.A., Mulvey, E.P., Schubert, C.A., Fagan, J., Piquero, A.R., and S.H. Losoya (2009).

Estimating a Dose-Response Relationship between Length of Stay and Future Recidivism in Serious

Juvenile Offenders.

Criminology,

47(3), 699-740.

McCrary, J. (2008). Manipulation of the Running Variable in the Regression Discontinuity Design:

A Density Test.

Journal of Econometrics,

142.

Michel, B., Rosholm, M., and M. Simonsen (2017). Measuring the Impact of Incarceration on Life

Trajectories, In: Michel, B. “Essays on the Economics of Crime and Development Economics”, PhD

Dissertation 2017-11, Aarhus BSS, Ch. 1.

Morris, R. G. and A. R. Piquero (2013). For Whom Do Sanctions Deter and Label?

Justice Quarterly,

30, 837-868.

Nagin, D. (1998). Criminal deterrence research at the outset of the twenty-first century. In

Crime and

Justice: An Annual Review of Research,

23, 1-42, (ed.) M. Tonry, Chicago: Univ. Chicago Press.

Nagin, D. (2013). Deterrence: A Review of the Evidence by Criminologists for Economists.

Annual

Review of Economics,

5, 83-105.

Nagin, D. S., Cullen, F. T., and C.L. Jonson (2009). Imprisonment and Reoffending. In

Crime and

Justice: A Review of Research,

38, 115-200, (ed.) M. Tonry, University of Chicago Press.

Nagin, D. S. and G. M. Snodgrass (2013). The effect of incarceration on re-offending: Evidence from

a natural experiment in Pennsylvania.

Journal of Quantitative Criminology,

29(4), 601-642.

Paternoster, R. and L. Iovanni (1989). The labeling perspective and delinquency: An elaboration of

the theory and an assessment of the evidence.

Justice Quarterly,

6, 359–394.

Petrosino, A., Turpin-Petrosino, C., and S. Guckenburg (2010). Formal System Processing of

Juveniles: Effects on Delinquency. Campbell Systematic Reviews, 2010:1

Piliavin, I., Gartner, R., Thornton, C., and R. Matsueda (1986). Crime, Deterrence, and Rational

Choice.

American Sociological Review,

51(1), 101-119.

Polinsky, A.M. and S. Shavell (1984). The Optimal Use of Fines and Imprisonment.

Journal of Public

Economics,

24 (June), 89-99.

REU, Alm.del - 2017-18 - Bilag 60: Henvendelse af 10/11-2017 fra Aarhus Universitet, Institut for Økonomi v/Anna Piil Damm vedr. "Ny forskning om virkninger af at sænke den kriminelle lavalder"

Polinsky, A.M. and S. Shavell (1999). On the disutility and discounting of imprisonment and the

theory of deterrence.

Journal of Legal Studies,

28, 1-16.

Risler, E. A., Sweatman T., and L. Nackerud (1998). Evaluating the Georgia Legislative Waiver's

Effectiveness in Deterring Juvenile Crime.

Research on Social Work Practice.

8(6), 657-667.

Sampson, R. J. and J. H. Laub (1997). A life-course theory of cumulative disadvantage and the

stability of delinquency. In

Developmental theories of crime and delinquency.

(ed.) T. P. Thornberry,

New Brunswick: Transaction.

Singer, S. I. and D. McDowall (1988). Criminalizing Delinquency: The Deterrent Effects the New

York Juvenile Offender Law.

Law & Society Review.

22(3), 521-536.

Statistics Denmark (2012). Criminality 2011. Danmarks Statistik.

Stafford, M. C. and M. Warr (1993). A reconceptualization of general and specific deterrence.

Journal of Research in Crime & Delinquency,

30, 123-135.

Steiner, B. and Wright, E. (2006). Assessing the Relative Effects of State Direct File Waiver Laws

on Violent Juvenile Crime: Deterrence or Irrelevance?

The Journal of Criminal Law and

Criminology,

96 (4), 1451-1477.

Stigler, G. J. (1970). The Optimum Enforcement of Laws.

Journal of Political Economy,

(May/June), 526-36.

Storgaard, A. (2013). Unge lovovertræderes strafansvar, retssikkerhed og lavalder. In

Argumentation

i nordisk straffrätt.

(ed.) K. Nordlöf. Stockholm: Nordstedts, 200-215.

Sweeten, G. (2006). Who Will Graduate? Disruption of High School Education by Arrest and Court

Involvement,

Justice Quarterly,

23, 462-480.

Tanner, J., Davies, S., and B. O’Grady (1999). Whatever Happened to Yesterday’s Rebels?

Longitudinal Effects of Youth Delinquency on Education and Employment.

Social Problems,

46,

250-274.

Waldo, G.P. and T.G. Chiricos (1972). Perceived Penal Sanction and Self-Reported Criminality: A

Neglected Approach to Deterrence Research.

Social Problems,

19(4), 522-540.

REU, Alm.del - 2017-18 - Bilag 60: Henvendelse af 10/11-2017 fra Aarhus Universitet, Institut for Økonomi v/Anna Piil Damm vedr. "Ny forskning om virkninger af at sænke den kriminelle lavalder"

Wiley, S. A., Slocum, L. A., and F.-A. Esbensen (2013). The unintended consequences of being

stopped or arrested: An exploration of the labeling mechanisms through which police contact leads

to subsequent delinquency

Criminology,

vol. 51, 927- 966.

REU, Alm.del - 2017-18 - Bilag 60: Henvendelse af 10/11-2017 fra Aarhus Universitet, Institut for Økonomi v/Anna Piil Damm vedr. "Ny forskning om virkninger af at sænke den kriminelle lavalder"

Appendix

Figure A1. Annual number of (as if) charges of penal code offenses per 10,000 by cohorts 1993

to 1999 over the age-interval 10-16

Data source: Administrative register data from Statistics Denmark and Police records for birth cohorts 1993-1999.

Note: The figure shows the crime-age curves for seven Danish birth cohorts born in 1993 to 1999 based on the annual number of (as

if) charges for a penal code offense per 10,000 in the cohort for age groups 10 to 16. The administrative records of (as if) charges for

these birth cohorts include the years 2002 to 2013 and the crime-age curves are therefore end at ages 13 to 15 for the three youngest

cohorts.

REU, Alm.del - 2017-18 - Bilag 60: Henvendelse af 10/11-2017 fra Aarhus Universitet, Institut for Økonomi v/Anna Piil Damm vedr. "Ny forskning om virkninger af at sænke den kriminelle lavalder"

Figure A2. Number of (as if) charges of penal code offenses per 10,000 by month ages 13 to 15

for children who have their 14

year pre or post reform

Data source: Administrative register data from Statistics Denmark and Police records for birth cohorts 1993-1999.

Note: The figure shows scatterplots of the number of reported (as if) charges of the penal code per 10,000 in a given month between

age 13 to 15 for children who have their 14

year prior to the reform of the minimum age of criminal responsibility and children who

have their 14

year during the reform. The solid grey lines are a local linear trend lines and the dashed lines indicate 95% confidence

intervals.

REU, Alm.del - 2017-18 - Bilag 60: Henvendelse af 10/11-2017 fra Aarhus Universitet, Institut for Økonomi v/Anna Piil Damm vedr. "Ny forskning om virkninger af at sænke den kriminelle lavalder"

Figure A3. Mean of selected control variable by distance to/from reform to 14

birthday

Note: The figure shows scatterplots of selected variables by distance to/from reform date to 14

birthday, e.g. at the vertical line at 0,

the plot shows mean of variable for individuals turning 14 years on July 1, 2010. The solid line is a local polynomial smoothed line

and the corresponding dashed lines indicate 95% confidence intervals.

Data source: Administrative register data from Statistics Denmark and police records for birth cohorts 1993-1999.

REU, Alm.del - 2017-18 - Bilag 60: Henvendelse af 10/11-2017 fra Aarhus Universitet, Institut for Økonomi v/Anna Piil Damm vedr. "Ny forskning om virkninger af at sænke den kriminelle lavalder"

Figure A4. Survival plots: time to recidivism (penal code offenses) for 14- year-old penal code

offenders who have their 14

year pre or post reform

Data source: Administrative register data from Statistics Denmark and Police records for 14- year-old penal code offenders.

Note: The figure shows the survival probability from first reported penal code offense at the age of 14 and the number of days to

recidivism to a new penal code offense within the first 18 months. The control group consist of 14- year-old penal code offenders who

have their 14

year prior to the reform of the minimum age of criminal responsibility and the treatment group consist of 14- year-old

penal code offenders who have their 14

year during the reform.

REU, Alm.del - 2017-18 - Bilag 60: Henvendelse af 10/11-2017 fra Aarhus Universitet, Institut for Økonomi v/Anna Piil Damm vedr. "Ny forskning om virkninger af at sænke den kriminelle lavalder"

Table A1.A: Variable definitions and primary data sources: Individual characteristics.

Variable

Female

Age

Native Dane

Western immigrant

Non-western immigrant

Birth weight

Definition

Dummy equals one if the child is female.

Age based on birthday and measured in months.

Dummy equals one if the child is native Dane.

Dummy equals one if the child is 1

or 2

generation

immigrant from a western country.

Dummy equals one if the child is 1

or 2

generation

immigrant from a non-western country.

Weight of the child at birth in grams.

Primary data source

Population register, DST.

Medical Birth Register,

DST.

Birth weight < 1500 g.

Dummy equals one if the birth weight of the child is

Medical Birth Register,

DST.

less than 1500 grams.

Birth weight < 2500 g.

Dummy equals one if the birth weight of the child is

Medical Birth Register,

DST.

less than 2500 grams,

Born premature

The length of the pregnancy in weeks.

Medical Birth Register,

Born extremely premature Dummy equals one if the pregnancy is shorter than 31 DST

Medical Birth Register,

DST.

weeks.

Dummy equals one if the child has been diagnosed

Psychiatric Central

ADHD diagnose

with ADHD at before his/her 10

birthday.

Dummy equals one if the child has been prescribed

Ritalin

Prescriptions of

Medicinal Products, DST.

Ritalin (age 0-9).

Dummy equals one if the child has been prescribed

Ritalin (180 dd min. 1

year)

Medicinal Products, DST.

Ritalin min. 180 accumulated daily doses) in a

minimum of 1 year (age 0-9).

Other psychotropic drugs Dummy equals one if the child has prescriptions of

Medicinal Products, DST.

other psychotropic drugs (N-group: N05, N06 (excl.

N06BA04) and N07) before his/her 10

birthday.

Central Police Register.

Prior offense age 10-13

Dummy equals one if the child has at least one as if

charge of a criminal offense before his/her 14

birthday.

Number of prior offenses Number of prior as if charges age 10-13.

Central Police Register.

Prior violence

Dummy for as if charge(s) of violent offense age 10-

Central Police Register.

Prior burglary

Dummy for as if charge(s) of burglary age 10-13.

Central Police Register.

Prior shoplifting

Dummy for as if charge(s) of shoplifting age 10-13.

Central Police Register.

Prior vehicle theft

Dummy for as if charge(s) of vehicle theft age 10-13. Central Police Register.

Prior other theft

Dummy for as if charge(s) of another theft age 10-13. Central Police Register.

Prior robbery

Dummy for as if charge(s) of robbery age 10-13.

Central Police Register.

Prior vandalism

Dummy for as if charge(s) of vandalism age 10-13.

Central Police Register.

Prior other property off.

Dummy for as if charge(s) of other property age 10-

Central Police Register.

Prior traffic offense

Dummy for as if charge(s) of traffic offense age 10-

Prior drug offense

Dummy for as if charge(s) of drug offense age 10-13. Central Police Register.

Central Police Register.

Prior other offense type

Dummy for as if charge(s) of other offenses age 10-

13 of the first reported offense age 10-13.

Criminal debut age

Age

Central Police Register.

REU, Alm.del - 2017-18 - Bilag 60: Henvendelse af 10/11-2017 fra Aarhus Universitet, Institut for Økonomi v/Anna Piil Damm vedr. "Ny forskning om virkninger af at sænke den kriminelle lavalder"

Table A1.B: Variable definitions and primary data sources: Family characteristics.

Variable

Nuclear family

Parent and new partner

Single parent

Not living with parents

Mother's annual income

Father's annual income

Mother working

Definition

Dummy for living in a two-parent household at age 9.

Dummy for living with one parent and his/her new

partner at age 9.

Dummy for living in a single-parent household at age

Dummy equals one if the child is not living with any

of the parents at age 9.

Mother's annual income in the year the child is age 9,

deflated to 2004 prices and measured in 1000 DKK.

Father’s annual income in the year the child is age 9,

deflated to 2004 prices and measured in 1000 DKK.

Dummy equals one if mother is working in the year

the child is age 9.

Primary data source

Population register, DST.

Income Register, DST.

Integrated Database for

Labor Market Research,

DST.

Father working

Dummy equals one if father is working in the year the Integrated Database for

child is age 9.

Labor Market Research,

DST.

Mother primary and

Dummy equals one if mother has primary or

Education Register (annual

secondary school

secondary school as the highest education in the year registrations), DST

child is age 9.

Mother vocational

Dummy equals one if mother has vocational education Education Register (annual

education

as the highest education in the year child is age 9.

registrations), DST

Mother general upper

Dummy equals one if mother has general upper

Education Register (annual

secondary education

secondary education as the highest education in the

registrations), DST

year the child is age 9.

Education Register (annual

Mother short cycle higher Dummy equals one if mother has short cycle higher

education

education as the highest education in the year the child registrations), DST

is age 9.

Mother medium cycle

Dummy equals one if mother has medium cycle

Education Register (annual

higher education

higher education as the highest education in the year

registrations), DST

the child is age 9.

Education Register (annual

Mother long cycle higher Dummy equals one if mother has long cycle higher

education

education as the highest education in the year the child registrations), DST

is age 9.

Father primary and

Dummy equals one if father has primary or secondary Education Register (annual

secondary school

school as the highest education in the year the child is registrations), DST

age 9.

Father vocational

Dummy equals one if father has vocational education Education Register (annual

education

as the highest education in the year the child is age 9. registrations), DST

Education Register (annual

Father general upper

Dummy equals one if father has general upper

secondary education

secondary education as the highest education in the

registrations), DST

year the child is age 9.

Father short cycle higher Dummy equals one if father has short cycle higher

Education Register (annual

education

education as the highest education in the year the child registrations), DST

is age 9.

Father medium cycle

Dummy equals one if father has medium cycle higher Education Register (annual

higher education

education as the highest education in the year the child registrations), DST

is age 9.

Father long cycle higher

Dummy equals one if father has long cycle higher

Education Register (annual

education

education as the highest education in the year the child registrations), DST

is age 9.

REU, Alm.del - 2017-18 - Bilag 60: Henvendelse af 10/11-2017 fra Aarhus Universitet, Institut for Økonomi v/Anna Piil Damm vedr. "Ny forskning om virkninger af at sænke den kriminelle lavalder"

Variable

Mother convicted of

criminal offense

Mother convicted of

prison sentence

Definition

Primary data source

Crime Statistics Register

(Dispositions), DST.

Crime Statistics Register

(Dispositions), DST.

Crime Statistics Register

(Dispositions), DST.

Crime Statistics Register

(Dispositions), DST.

Medical Birth Register, DST

Dummy equals 1 if mother is convicted and

found guilty of criminal offense (any type).

Dummy equals 1 if mother is convicted and

found guilty of a suspended or unsuspended

prison sentence.

Father convicted of

Dummy equals 1 if father is convicted and

criminal offense

found guilty of criminal offense (any type).

Father convicted of prison Dummy equals 1 if father is convicted and

sentence

found guilty of a suspended or unsuspended

prison sentence.

Missing identifier(mother) Dummy equals one if the identifier for the

mother is missing in the Medical Birth

Missing identifier(father)

Missing register data

(mother)

Missing register data

(father)

Dummy equals one if the identifier for the

father is missing in the Medical Birth Register.

Medical Birth Register, DST

Income Register, DST.

Dummy equals one if the income information

for the mother is missing in the year the child is

Dummy equals one if the income information

for the father is missing in the year the child is

Income Register, DST.

REU, Alm.del - 2017-18 - Bilag 60: Henvendelse af 10/11-2017 fra Aarhus Universitet, Institut for Økonomi v/Anna Piil Damm vedr. "Ny forskning om virkninger af at sænke den kriminelle lavalder"

Table A1.C: Variable definitions and primary data sources: specific to analysis of deterrence effects

Variable

Calendar month

Months prior to reform

Months during reform

Reform month

Distance to reform 1

(quadratic)

Distance to reform 2

(quadratic)

Distance to reform 1

(cubic)

Distance to reform 2

(cubic)

Outcome variables:

Monthly offending rates

(all penal code offenses)

Monthly offending rates

(violent offenses)

Monthly offending rates

(burglary)

Monthly offending rates

(shoplifting)

Monthly offending rates

(theft of vehicles)

Monthly offending rates

(vandalism)

Any penal code offense

before 14

Definition

Dummies for the calendar month (1-12) in each

year.

Number of months prior to July 2010 (20-1,

otherwise 0).

Number of months after July 2010 (1-19,

otherwise 0).

Dummy equals one if the month is July 2010.

Number of months prior to July 2010 (second-

degree polynomial)

Number of months after July 2010 (second-degree

polynomial)

Number of months prior to and after July 2010

(third-degree polynomial)

Number of months after July 2010 (third-degree

polynomial)

Dummy equals one if the individual has at least

one (as if) charge of a penal code offense in a

given month (November 2008 to January 2012).

Central Police Register, Crime

Statistics Register (Charges), DST.

Central Police Register, Crime

Statistics Register (Charges), DST.

Central Police Register, Crime

Statistics Register (Charges), DST.

Central Police Register, Crime

Statistics Register (Charges), DST.

Central Police Register, Crime

Statistics Register (Charges), DST.

Central Police Register, Crime

Statistics Register (Charges), DST.

Central Police Register, Crime

Statistics Register (Charges), DST.

Central Police Register, Crime

Statistics Register (Charges), DST.

Central Police Register, Crime

Statistics Register (Charges), DST.

Central Police Register, Crime

Statistics Register (Charges), DST.

Central Police Register, Crime

Statistics Register (Charges), DST.

Primary data source

Dummy equals one if the individual has at least

one (as if) charge of a violent offense in a given

Dummy equals one if the individual has at least

one (as if) charge of burglary in a given month.

Dummy equals one if the individual has at least

one (as if) charge of shoplifting in a given month.

Dummy equals one if the individual has at least

one (as if) charge of theft of vehicles in a given

Dummy equals one if the individual has at least

one (as if) charge of vandalism in a given month.

Dummy equals one if the individual has at least

one (as if) charge of a penal code offense before

his/her 14

birthday.

More than one penal code Dummy equals one if the individual has more than

one (as if) charge of a penal code offense before

offense before 14

his/her 14

birthday.

Any penal code offense at

More than one penal code

offense at 14

Any penal code offense

before 15

Dummy equals one if the individual has at least

one (as if) charge of a penal code offense in the

Dummy equals one if the individual has more than

one (as if) charge of a penal code offense in the

Dummy equals one if the individual has at least

one (as if) charge of a penal code offense before

his/her 15

birthday.

More than one penal code Dummy equals one if the individual has more than Central Police Register, Crime

offense before 15

one (as if) charge of a penal code offense before

Statistics Register (Charges), DST.

his/her 15 birthday.

Any penal code offense at

More than one penal code

offense at 15

Dummy equals one if the individual has at least

one (as if) charge of a penal code offense in the

Dummy equals one if the individual has more than

one (as if) charge of a penal code offense in the

Central Police Register, Crime

Statistics Register (Charges), DST.

Central Police Register, Crime

Statistics Register (Charges), DST.

REU, Alm.del - 2017-18 - Bilag 60: Henvendelse af 10/11-2017 fra Aarhus Universitet, Institut for Økonomi v/Anna Piil Damm vedr. "Ny forskning om virkninger af at sænke den kriminelle lavalder"

Table A1.D: Variable definitions and primary data sources: specific to analysis of recidivism

Variable

Offense age

Offense month

Offense 14: violence

Offense 14: burglary

Offense 14: shoplifting

Offense 14: vehicle theft

Offense 14: other theft

Offense 14: robbery

Offense 14: vandalism

Definition

Dummies for offense age (14.0 to 15.0) when first

penal code offense was committed at age 14.

Primary data source

Central Police Register, Crime

Statistics Register (Charges), DST.

Dummies for the calendar month (1-12) of the first Central Police Register, Crime

penal code offense committed at age 14.

Statistics Register (Charges), DST.

Dummy equals one if the offender is (as if)

charged of a violent offense at age 14.

Dummy equals one if the offender is (as if)

charged of burglary at age 14.

Dummy equals one if the offender is (as if)

charged of shoplifting at age 14.

Dummy equals one if the offender is (as if)

charged of vehicle theft at age 14.

Dummy equals one if the offender is (as if)

charged of other types of theft at age 14.

Dummy equals one if the offender is (as if)

charged of robbery at age 14.

Dummy equals one if the offender is (as if)

charged of vandalism at age 14.

Central Police Register, Crime

Statistics Register (Charges), DST.

Central Police Register, Crime

Statistics Register (Charges), DST.

Central Police Register, Crime

Statistics Register (Charges), DST.

Central Police Register, Crime

Statistics Register (Charges), DST.

Central Police Register, Crime

Statistics Register (Charges), DST.

Central Police Register, Crime

Statistics Register (Charges), DST.

Central Police Register, Crime

Statistics Register (Charges), DST.

Central Police Register, Crime

Statistics Register (Charges), DST.

Central Police Register, Crime

Statistics Register (Charges), DST.

Central Police Register, Crime

Statistics Register (Charges), DST.

Central Police Register, Crime

Statistics Register (Charges), DST.

Central Police Register, Crime

Statistics Register (Charges), DST.

Central Police Register, Crime

Statistics Register (Charges), DST.

Central Police Register, Crime

Statistics Register (Charges), DST.

Central Police Register, Crime

Statistics Register (Charges), DST.

Offense 14: other property Dummy equals one if the offender is (as if)

offense

charged of other types of a property offense at age

Offense 14: other offense Dummy equals one if the offender is (as if)

types

charged of other offense types at age 14.

Outcome variables:

Recidivism 3 months

Dummy equals one if the offender recidivates to a

penal code offense within 3 months from first

offense age 14.

Dummy equals one if the offender recidivates to a

penal code offense within 6 months from first

offense age 14.

Dummy equals one if the offender recidivates to a

penal code offense within 9 months from first

offense age 14.

Dummy equals one if the offender recidivates to a

penal code offense within 12 months from first

offense age 14.

Dummy equals one if the offender recidivates to a

penal code offense within 15 months from first

offense age 14.

Dummy equals one if the offender recidivates to a

penal code offense within 18 months from first

offense age 14.

Recidivism 6 months

Recidivism 9 months

Recidivism 12 months

Recidivism 15 months

Recidivism 18 months

REU, Alm.del - 2017-18 - Bilag 60: Henvendelse af 10/11-2017 fra Aarhus Universitet, Institut for Økonomi v/Anna Piil Damm vedr. "Ny forskning om virkninger af at sænke den kriminelle lavalder"

Variable

Time to recidivism

Definition

The number of days from the first penal code

offense committed at age 14 to recidivism to a

penal code offense within 18 months.

The number of months from 14

birthday to

reform month (July 2010).

Dummy equals one if the offender has been

enrolled in 9th grade before his/her 17

birthday.

Dummy equals one if the offender has participated

in 9th grade exam (in one or more subject(s))

before his/her 17

birthday.

Conditional on enrollment in 9

grade dummy

equals one if the offender was enrolled at an

Conditional onlenrollment in 9

grade dummy

equals one if the offender was enrolled at a

b di

Conditionalhon lenrollment in 9

grade dummy

equals one if the offender was enrolled at a school

for children with special needs.

Average of the final teacher assessments in

language arts (reading, writing, spelling, oral

presentation and order) in 9

grade. The test scores

are standardized with national average for that

specific test (e.g. reading) in a given school year.

Primary data source

Central Police Register, Crime

Statistics Register (Charges), DST.

Population register, DST.

Education Register (all entries),

DST.

Education Register (grades lower

secondary education), DST.

Education Register (all entries and

institution register), DST.

Education Register (all entries and

institution register), DST.

Education Register (all entries and

institution register), DST.

Education Register (grades lower

secondary education), DST.

Birth month

Enrolled in 9th grade

Exam participation

Ordinary schools

Boarding schools

Schools special needs

Language Arts (teacher

assessment)

Average of the exit exams in language arts

(reading, writing, spelling, oral presentation and

order) in 9

grade. The test scores are standardized

with national average for that specific test in a

Math (teacher assessment) Average of the final teacher assessments in math

(written test in skills, problem solving and oral

presentation) in 9

grade. The test scores are

standardized with national average for that specific

Math (exit exam)

Average of the exit exams in math (written test in

skills, and problem solving) in 9

grade. The test

scores are standardized with national average for

that specific test in a given school year.

Language Arts (exit

exam)

Education Register (grades lower

secondary education), DST.

Education Register (grades lower

secondary education), DST.

Education Register (grades lower

secondary education), DST.

REU, Alm.del - 2017-18 - Bilag 60: Henvendelse af 10/11-2017 fra Aarhus Universitet, Institut for Økonomi v/Anna Piil Damm vedr. "Ny forskning om virkninger af at sænke den kriminelle lavalder"

Table A2. Summary statistics of sample of 14-year-olds used in analysis: Means (std. dev.).

162,959

154,029

142,675

162,959

1,640

162,959

161,321

156,438

162,161

159,788

Mean

0.51

0.49

0.90

0.01

0.08

3488.96

0.01

0.05

0.04

0.01

0.00

0.04

0.01

1.55

0.14

0.06

0.41

0.08

0.02

0.24

0.08

0.01

0.00

0.04

11.26

0.72

0.09

0.18

0.01

193.58

297.99

0.79

0.86

0.50

0.30

0.07

0.28

592.60

0.08

0.21

0.19

0.09

0.08

0.06

0.19

0.10

1.52

0.35

0.24

0.49

0.27

0.15

0.43

0.27

0.12

0.03

0.19

0.78

0.45

0.28

0.38

0.09

142.90

247.71

0.41

0.34

Child

characteristics

Criminal

history

(age 10-13)

Family

characteristics

(age 9)

Male

Female

Native Dane

Western immigrant 1st & 2nd generation

Non-western immigrant 1st & 2nd generation

Birth weight

Birth weight under 1500 g.

Birth weight under 2500 g.

Born premature

Born extremely premature

ADHD diagnose (age 0-9)

Use of Ritalin (age 0-9)

Use of Ritalin (180 dd min. 1 year) (age 0-9)

Use of other psychotropic drugs (age 0-9)

Charged of an offense before age 14

Number of prior charges

Charged of a violent offense

Charged of burglary

Charged of shoplifting

Charged of vehicle theft

Charged of theft

Charged of robbery

Charged of vandalism

Charged of other property offenses

Charged of a traffic offense

Charged of a drug offense

Charged of another offense

Criminal debut age

Nuclear family

Parent and new partner

Single parent

Not living parents

Mother's annual income(1000 DKK, deflated)

Father's annual income(1000 DKK, deflated)

Mother working

Father working

REU, Alm.del - 2017-18 - Bilag 60: Henvendelse af 10/11-2017 fra Aarhus Universitet, Institut for Økonomi v/Anna Piil Damm vedr. "Ny forskning om virkninger af at sænke den kriminelle lavalder"

Family

characteristics

(age 9)

Primary and secondary school - mother

Vocational education - mother

General upp. secondary edu. - mother

Short cycle higher edu. - mother

Medium cycle higher edu. - mother

Long cycle higher edu. - mother

Primary and secondary school - father

Vocational education - father

General upp. secondary edu. - father

Short cycle higher edu. - father

Medium cycle higher edu. - father

Long cycle higher edu. - father

Mother convicted of criminal offense

Mother convicted of prison or suspended prison sentence

Father convicted of criminal offense

Father convicted of prison or suspended prison sentence

Missing identifier(mother)

Missing identifier(father)

Missing register data child age 9(mother)

Missing register data child age 9(father)

Post-reform (age 14 after reform)

Pre-/post reform (age 14 prior to and after reform)

Pre-reform (age 14 prior to reform)

Observations

162,959

Mean

0.24

0.38

0.07

0.04

0.19

0.08

0.25

0.40

0.05

0.08

0.10

0.01

0.00

0.05

0.02

0.00

0.02

0.01

0.04

0.28

0.43

0.29

0.44

0.48

0.25

0.20

0.39

0.26

0.45

0.49

0.22

0.26

0.30

0.29

0.12

0.06

0.21

0.14

0.08

0.17

0.23

0.27

0.45

0.49

0.45

Data source: Administrative register data from Statistics Denmark and Police records for birth cohorts 1993-1999.

Note: The table shows summary statistics for the population of 14-year-olds from November 2008 to February 2012.

REU, Alm.del - 2017-18 - Bilag 60: Henvendelse af 10/11-2017 fra Aarhus Universitet, Institut for Økonomi v/Anna Piil Damm vedr. "Ny forskning om virkninger af at sænke den kriminelle lavalder"

Table A3. Summary statistics of sample used in analysis: Means, Differences and t-tests.

Pre-

reform

(1)

Mean

0.51

0.49

0.90

0.00

0.08

3490.90

0.01

0.05

0.03

0.01

0.00

0.03

0.01

1.63

0.10

0.06

0.39

0.08

0.02

0.26

0.08

0.02

0.00

0.05

11.29

0.72

0.09

0.18

0.01

Pre-

/post-

reform

(2)

Mean

0.52

0.48

0.90

0.01

0.08

3485.91

0.01

0.05

0.04

0.01

0.00

0.04

0.01

1.55

0.16

0.07

0.40

0.08

0.02

0.24

0.09

0.02

0.00

0.03

11.23

0.72

0.09

0.18

0.01

Post-

reform

(3)

Mean

0.51

0.49

0.90

0.01

0.09

3491.59

0.01

0.05

0.04

0.01

0.00

0.04

0.01

1.42

0.16

0.04

0.45

0.09

0.07

0.03

0.22

0.06

0.01

0.00

0.02

11.26

0.72

0.09

0.18

0.01

Child

characteristics

Male

Female

Native Dane

Western immigrant 1st & 2nd generation

Non-western immigrant 1st & 2nd generation

Birth weight

Birth weight under 1500 g.

Birth weight under 2500 g.

Born premature

Born extremely premature

ADHD diagnose (age 0-9)

Use of Ritalin (age 0-9)

Use of Ritalin (180 dd min. 1 year) (age 0-9)

Difference Difference

& t-test

(1-3)

(2-3)

-0.00

0.00

-0.00

-0.69

0.00

-0.00

-0.01

***

-0.00

***

-0.00

-0.01

***

0.00

***

0.20

-0.06

0.01

-0.06

-0.01

0.01

-0.01

0.04

0.02

0.01

-0.00

0.03

0.00

-0.00

0.01

-0.01

0.00

-0.00

-5.68

0.00

-0.00

0.00

-0.00

0.00

0.13

-0.01

0.03

-0.05

-0.01

0.01

-0.02

0.02

0.03

0.01

-0.00

0.01

-0.03

-0.00

0.00

-0.00

Criminal

history

(age 10-13)

Use of other psychotropic drugs (age 0-9)

Charged of an offense before age 14

Number of prior charges

Charged of a violent offense

Charged of burglary

Charged of shoplifting

Charged of vehicle theft

Charged of theft

Charged of robbery

Charged of vandalism

Charged of other property offenses

Charged of a traffic offense

Charged of a drug offense

Charged of another offense

Criminal debut age

Family

characteristics

(age 9)

Nuclear family

Parent and new partner

Single parent

Not living parents

REU, Alm.del - 2017-18 - Bilag 60: Henvendelse af 10/11-2017 fra Aarhus Universitet, Institut for Økonomi v/Anna Piil Damm vedr. "Ny forskning om virkninger af at sænke den kriminelle lavalder"

(1-3)

(2-3)

Mean

Family

Mother's annual income(1000 DKK, deflated) 193.17

193.18

194.64

-1.47

-1.46

characteristics

-3.32

Father's annual income(1000 DKK, deflated)

294.04

298.30

301.61

-7.58

***

(age 9)

-0.01

***

Mother working

0.79

0.80

-0.01

***

-0.01

***

Father working

0.86

0.87

-0.01

***

0.01

***

Primary and secondary school - mother

0.24

0.23

0.01

***

Vocational education - mother

0.38

0.37

0.01

0.00

General upp. secondary edu. - mother

0.07

0.00

Short cycle higher edu. - mother

0.04

0.05

-0.00

Medium cycle higher edu. - mother

0.19

-0.00

-0.01

***

Long cycle higher edu. - mother

0.07

0.08

-0.01

***

0.01

Primary and secondary school - father

0.26

0.25

0.01

Vocational education - father

0.40

0.00

-0.00

General upp. secondary edu. - father

0.05

-0.00

Short cycle higher edu. - father

0.08

-0.00

Medium cycle higher edu. - father

0.11

0.10

0.00

-0.00

Long cycle higher edu. - father

0.09

0.10

-0.01

***

-0.00

***

Mother convicted of criminal offense

0.01

0.02

-0.01

***

Mother convicted of prison or suspended

prison sentence

0.00

0.01

-0.00

***

-0.00

***

-0.01

***

Father convicted of criminal offense

0.04

0.05

0.06

-0.02

***

Father convicted of prison or suspended prison

sentence

0.01

0.02

-0.01

***

-0.01

***

Missing identifier(mother)

0.00

-0.00

Missing identifier(father)

0.02

0.00

-0.00

Missing register data child age 9(mother)

0.01

0.00

Missing register data child age 9(father)

0.04

0.00

-0.00

Observations

47,441

69,785

45,733

Data source: Administrative register data from Statistics Denmark and Police records for birth cohorts 1993-1999.

Note: The table shows summary statistics for the population of 14-year-olds from November 2008 to February 2012,

divided into three groups according to whether they have their 14th year pre-reform, pre-and post reform or post reform.

The last two column show the differences between the groups and t-tests of difference in means * p<0.05, ** p<0.01, ***

p<0.001.

Pre-

reform

(1)

Pre-

/post-

reform

(2)

Post-

reform

(3)

Difference Difference

and t-test and t-test

REU, Alm.del - 2017-18 - Bilag 60: Henvendelse af 10/11-2017 fra Aarhus Universitet, Institut for Økonomi v/Anna Piil Damm vedr. "Ny forskning om virkninger af at sænke den kriminelle lavalder"

Table A4. Robustness analyses: Effects of the reform on monthly reported offending rates

(penal code offenses), population: 14-year-olds

Baseline result:

Reform effect July 2010-February 2012

(obs.= 1,955,508)

Different reform specifications:

Including dummy variable for reform month (obs.= 1,955,508)

Distance to reform (quadratic) (obs.= 1,955,508)

Distance to reform (quadratic and cubic) (obs.= 1,955,508)

Adding 14-year-olds after the minimum age of criminal responsibility

was re-established at 15 (14

birthday after 1 March, 2012) (obs.=3,449,100)

Excluding 14-year-olds closest to the reform (14

birthday June, July and August

2010) (obs.=1,742,040)

Excluding 14-year-olds furthest away from the reform (14

birthday November-

December 2008 and January-February 2012) (obs.=1,816,104)

Announcements effects:

Excluding 14-year-olds born 17 March -30 June 1996 (obs.= 1,715,076)

Effects of media debate October 2009 - June 2010 (obs.=1,030,623)

Controls:

Age month specification

Calendar month dummies

Child background

Parents background

Child crime history

Police district fixed effects

0.00017

(0.00015)

0.00017

(0.00016)

-0.00025

(0.00027)

-0.00002

(0.00035)

0.00020

(0.00013)

0.00016

(0.00015)

0.00010

(0.00016)

0.00014

(0.00016)

-0.00068

(0.00270)

Dummies

Yes

Data source: Administrative register data from Statistics Denmark and Police records for birth cohorts 1993-1999.

Note: The table shows result from robustness analyses to test of different model assumptions and specifications (e.g.

higher-order polynomials in distance to reform, bandwidth definitions and announcements effects). The reported

estimates are coefficients from linear panel models on the probability of an (as if) charge in a given month from November

2008 to January 2012. Standard errors are clustered at the individual level and reported in parentheses,

< 0.10,

***

0.05,

< 0.01. Control variables: number of months relative to reform, age, calendar month, gender, ethnicity, birth

weight, parents’ income, occupation and education, family type (nuclear, single parent, new partner, child not living at

home), child ADHD diagnosis, child using prescriptive drugs, child and parents’ criminal history, police district.

REU, Alm.del - 2017-18 - Bilag 60: Henvendelse af 10/11-2017 fra Aarhus Universitet, Institut for Økonomi v/Anna Piil Damm vedr. "Ny forskning om virkninger af at sænke den kriminelle lavalder"

Table A5. OLS-models: Effects of the reform (penal code offenses), population: 14-year-olds

Outcomes (obs.=93,174):

Any penal code offense before 14

More than one penal code offense before 14

Any penal code offense at 14

More than one penal code offense at 14

Any penal code offense before 15

More than one penal code offense before 15

Any penal code offense at 15

More than one penal code offense at 15

Controls:

Age specification

Calendar month dummies

Birth month relative to reform

Child background

Parents background

Police districts fixed effects

-0.00534

***

(0.00092)

-0.00161

***

(0.00050)

-0.00396

***

(0.00084)

-0.00056

(0.00047)

-0.00861

***

(0.00117)

-0.00246

***

(0.00070)

-0.00349

***

(0.00090)

-0.00051

(0.00055)

0.00458

(0.00416)

0.00123

(0.00222)

-0.00128

(0.00379)

-0.00071

(0.00212)

0.00224

(0.00526)

0.00000

(0.00311)

0.00913

(0.00397)

0.00213

(0.00245)

Yes

III

-0.00165

(0.00353)

-0.00101

(0.00216)

-0.00236

(0.00412)

-0.00176

(0.00233)

-0.00326

(0.00498)

-0.00297

(0.00316)

0.00731

(0.00439)

0.00094

(0.00271)

Yes

-0.00176

(0.00352)

-0.00104

(0.00216)

-0.00248

(0.00412)

-0.00178

(0.00233)

-0.00350

(0.00498)

-0.00302

(0.00316)

0.00731

(0.00439)

0.00095

(0.00270)

Yes

Data source: Administrative register data from Statistics Denmark and Police records for birth cohorts 1993-1999.

Note: The reported estimates are coefficients from OLS regressions for the population of 14-year-olds pre-reform (14

birthday November 2008 to June 2010) and post-reform (14

birthday July 2010 to February 2012). Rows represent

separate regression models with different outcome specifications and each column represents a gradually richer

conditioning set. Robust standard errors are reported in parentheses and * p < 0.10, ** p < 0.05, *** p < 0.01. Control

variables: number of months relative to reform, gender, ethnicity, birth weight, parents’ income, occupation and

education, family type (nuclear, single parent, new partner, child not living at home), child ADHD diagnosis, child using

prescriptive drugs, child and parents’ criminal history, police district.

REU, Alm.del - 2017-18 - Bilag 60: Henvendelse af 10/11-2017 fra Aarhus Universitet, Institut for Økonomi v/Anna Piil Damm vedr. "Ny forskning om virkninger af at sænke den kriminelle lavalder"

Table A6. OLS-models: Effects of the reform (penal code offenses), population: 14-year-olds

sub group

Subsample: without prior

offense by age 14

Outcomes:

Any penal code offense before 14

More than one penal code offense before 14

Any penal code offense at 14

More than one penal code offense at 14

Any penal code offense before 15

More than one penal code offense before 15

Any penal code offense at 15

More than one penal code offense at 15

Observations

Controls:

Age specification

Calendar month dummies

Birth month relative to reform

Child background

Parents background

Police districts fixed effects

-0.00352

***

(0.00067)

-0.00071

(0.00034)

-0.00375

***

(0.00080)

-0.00044

(0.00043)

-0.00670

***

(0.00099)

-0.00128

(0.00057)

-0.00312

***

(0.00087)

-0.00020

(0.00051)

92,249

-0.00226

(0.00352)

-0.00160

(0.00172)

-0.00198

(0.00396)

-0.00316

(0.00216)

-0.00357

(0.00501)

-0.00308

(0.00285)

0.00742

(0.00426)

0.00236

(0.00255)

92,249

Yes

Subsample: with prior

offense by age 14

0.03392

***

(0.01024)

-0.01492

(0.03130)

0.01747

(0.02641)

0.00756

(0.01938)

0.02056

(0.00902)

-0.02833

(0.03258)

-0.00131

(0.02606)

-0.00786

(0.02063)

925

0.03066

(0.04962)

-0.06129

(0.15166)

-0.05281

(0.13083)

0.11815

(0.09359)

-0.01196

(0.03831)

-0.09227

(0.15790)

0.03291

(0.12481)

-0.10778

(0.09886)

925

Yes

Data source: Administrative register data from Statistics Denmark and Police records for birth cohorts 1993-1999.

Note: The reported estimates are coefficients from OLS regressions. The sample of 14-year-olds pre-reform (14

birthday

November 2008 to June 2010) and post-reform (14

birthday July 2010 to February 2012)(obs.=93,174) is dived into two

groups according to their criminal history before age 14. Robust standard errors are reported in parentheses and * p <

0.10, ** p < 0.05, *** p < 0.01. Control variables: number of months relative to reform, gender, ethnicity, birth weight,

parents’ income, occupation and education, family type (nuclear, single parent, new partner, child not living at home),

child ADHD diagnosis, child using prescriptive drugs, child and parents’ criminal history, police district.

REU, Alm.del - 2017-18 - Bilag 60: Henvendelse af 10/11-2017 fra Aarhus Universitet, Institut for Økonomi v/Anna Piil Damm vedr. "Ny forskning om virkninger af at sænke den kriminelle lavalder"

Table A7. Summary statistics of sample used in analysis of recidivism: Means and t-tests.

Pre-reform

Obs. Mean

893

0.65

893

0.35

893

0.77

893

0.01

893

0.22

809

3455.77

809

0.00

809

0.04

799

0.03

799

0.00

893

0.02

893

0.01

893

0.00

893

0.04

893

0.48

893

0.13

893

0.03

893

0.01

878

130.21

817

202.33

888

0.57

854

0.69

893

0.47

893

0.31

893

0.07

893

0.02

893

0.11

893

0.02

893

0.43

893

0.33

893

0.06

893

0.04

893

0.05

893

0.04

893

0.03

893

0.01

0.09

0.04

Post-reform

Difference

Obs. Mean

and t-test

676

0.64

0.01

676

0.36

-0.01

676

0.75

0.02

676

0.01

-0.00

676

0.23

-0.01

610

3489.50

-33.72

610

0.00

610

0.04

-0.00

462

0.04

-0.01

462

0.00

676

0.02

-0.00

676

0.03

-0.02

***

676

0.02

-0.02

676

0.03

0.01

676

0.47

0.01

676

0.14

-0.01

676

0.04

-0.01

676

0.01

0.00

660

113.04

17.17

613

189.88

12.45

668

0.56

0.01

645

0.67

0.02

676

0.48

-0.01

676

0.33

-0.02

676

0.05

0.01

676

0.03

-0.00

676

0.08

0.03

676

0.01

0.00

676

0.44

-0.01

676

0.34

-0.01

676

0.05

0.00

676

0.06

-0.01

676

0.04

0.01

676

0.03

0.02

676

0.06

-0.02*

676

0.03

0.15

0.07

-0.02**

-0.06***

-0.03**

Child

characteristics

Family

characteristics

(age 9)

Male

Female

Native Dane

Western immigrant 1st & 2nd generation

Non-western immigrant 1st & 2nd generation

Birth weight

Birth weight under 1500 g.

Birth weight under 2500 g.

Born premature

Born extremely premature

ADHD diagnose (age 0-9)

Use of Ritalin (age 0-9)

Use of Ritalin (180 dd min. 1 year) (age 0-9)

Use of other psychotropic drugs (age 0-9)

Nuclear family

Parent and new partner

Single parent

Not living with parents

Mother's annual income(1000 DKK, deflated)

Father's annual income(1000 DKK, deflated)

Mother working

Father working

Primary and secondary school - mother

Vocational education - mother

General upp. secondary edu. - mother

Short cycle higher edu. - mother

Medium cycle higher edu. - mother

Long cycle higher edu. - mother

Primary and secondary school - father

Vocational education - father

General upp. secondary edu. - father

Short cycle higher edu. - father

Medium cycle higher edu. - father

Long cycle higher edu. - father

Mother convicted of criminal offense

Mother convicted of prison or suspended prison

sentence

Father convicted of criminal offense

Father convicted of prison or suspended prison

sentence

REU, Alm.del - 2017-18 - Bilag 60: Henvendelse af 10/11-2017 fra Aarhus Universitet, Institut for Økonomi v/Anna Piil Damm vedr. "Ny forskning om virkninger af at sænke den kriminelle lavalder"

Pre-reform

Post-reform

Difference

Obs. Mean

and t-test

Missing identifier(mother)

893

0.01

676

0.01

-0.01

Missing identifier(father)

893

0.04

676

0.05

-0.00

Missing register data child age 9(mother)

893

0.02

676

0.02

-0.01

Missing register data child age 9(father)

893

0.09

676

0.09

-0.01

Crime history

Charged of an offense before age 14

893

0.11

676

0.12

-0.01

(age 10-13)

Number of prior charges

2.01

2.20

-0.18

Charged of a violent offense

0.14

0.20

-0.05

Charged of burglary

0.03

0.07

-0.04

Charged of shoplifting

0.35

0.38

-0.03

Charged of vehicle theft

0.15

0.23

-0.08

Charged of theft

0.14

0.10

0.05

Charged of robbery

0.08

0.12

-0.04

Charged of vandalism

0.29

0.24

0.04

Charged of other property offenses

0.09

0.10

-0.01

Charged of a traffic offense

0.02

-0.00

Charged of a drug offense

0.01

-0.00

Charged of another offense

0.09

0.04

0.06

Criminal debut age

11.22

11.34

-0.12

Offense at 14

Offense_age

893

14.50

676

14.49

0.01

Offense age 14.0

893

0.04

676

0.05

-0.02

Offense age 14.1

893

0.12

676

0.09

0.02

Offense age 14.2

893

0.10

676

0.12

-0.02

Offense age 14.3

893

0.10

676

0.11

-0.01

Offense age 14.4

893

0.11

676

0.08

0.03

Offense age 14.5

893

0.10

676

0.07

0.02

Offense age 14.6

893

0.10

676

0.13

-0.03

Offense age 14.7

893

0.10

676

0.09

0.01

Offense age 14.8

893

0.10

676

0.09

0.00

Offense age 14.9

893

0.09

676

0.11

-0.02

Offense age 15.0

893

0.06

676

0.05

0.01

Charged of a violent offense

893

0.17

676

0.18

-0.01

Charged of burglary

893

0.06

676

0.05

0.01

Charged of shoplifting

893

0.33

676

0.37

-0.03

Charged of vehicle theft

893

0.11

676

0.09

0.02

Charged of other theft

893

0.08

676

0.08

0.00

Charged of robbery

893

0.04

676

0.03

0.01

Charged of vandalism

893

0.13

676

0.11

0.02

Charged of other property offenses

893

0.05

676

0.06

-0.00

Charged of other types of offenses

893

0.01

676

0.03

-0.01

Data source: Administrative register data from Statistics Denmark and Police records for 14- year-old penal code

offenders pre- and post-reform.

Note: The table shows summary statistics for the population of 14- year-old penal code offenders, divided into two groups

according to whether they have their 14th year pre- or post-reform. The last column shows the difference between the

groups and t-tests of difference in means * p<0.05, ** p<0.01, *** p<0.001.

REU, Alm.del - 2017-18 - Bilag 60: Henvendelse af 10/11-2017 fra Aarhus Universitet, Institut for Økonomi v/Anna Piil Damm vedr. "Ny forskning om virkninger af at sænke den kriminelle lavalder"

Table A8. Effects of reform on time to recidivism (

penal code offenses

), population: 14-year-old

penal code offenders

Effects of reform

Controls:

Calendar month dummies

Offense age and type

Child background variables

Family background variables

Observations:

1.1735

(0.0961)

1,569

1.2447

***

(0.1038)

Yes

1,569

III

1.1802

(0.1062)

Yes

1,569

1.1574

(0.1060)

Yes

1,569

Data source: Administrative register data from Statistics Denmark and Police records for 14- year-old penal code

offenders pre- and post-reform.

Note: The table shows results from supplementary analyses and the reported estimates are hazard ratios of recidivism

within 18 months from Cox regressions for 14- year-old penal code offenders pre- and post-reform. * p < 0.10, ** p <

0.05, *** p < 0.01. Control variables: offense month, offense age, type of the offense, gender, ethnicity, birth weight,

child ADHD diagnosis, child using prescriptive drugs, child’s criminal history (ages 10-13), parents’ income, occupation

and education, family type and parents’ criminal history.

REU, Alm.del - 2017-18 - Bilag 60: Henvendelse af 10/11-2017 fra Aarhus Universitet, Institut for Økonomi v/Anna Piil Damm vedr. "Ny forskning om virkninger af at sænke den kriminelle lavalder"

Table A9. Time trend in recidivism rates: Effects of birth month on recidivism (

penal code offenses

population: 14-year-old penal code offenders pre-reform

Effects of birth month on recidivism 3 months

III

-0.0003

(0.0042)

-0.0020

(0.0037)

-0.0017

(0.0040)

-0.0008

(0.0041)

-0.0025

(0.0031)

-0.0014

(0.0032)

1,383

-0.0005

(0.0042)

-0.0020

(0.0034)

-0.0018

(0.0035)

-0.0009

(0.0038)

-0.0025

(0.0033)

-0.0015

(0.0033)

Yes

1,383

0.0001

(0.0047)

-0.0011

(0.0043)

-0.0009

(0.0044)

-0.0004

(0.0044)

-0.0017

(0.0039)

-0.0011

(0.0040)

Yes

1,383

0.0004

(0.0046)

-0.0008

(0.0043)

-0.0006

(0.0043)

-0.0000

(0.0043)

-0.0013

(0.0039)

-0.0008

(0.0039)

Yes

1,383

Effects of birth month on recidivism 6 months

Effects of birth month on recidivism 9 months

Effects of birth month on recidivism 12 months

Effects of birth month on recidivism 15 months

Effects of birth month on recidivism 18 months

Controls:

Calendar month dummies

Offense age and type

Child background variables

Family background variables

Police district fixed effects

Observations:

Data source: Administrative register data from Statistics Denmark and Police records for 14- year-old penal code

offenders pre- and post-reform.

Note: The table shows results from supplementary analyses of time trends in recidivism rates. The sample includes 14-

year-old penal code offenders pre-reform from March 2008 to June 2010. The reported estimates are coefficients from

OLS regressions of the effects of birth month on the probability of recidivism at 3, 6, 9, 12, 15 and 18 months and each

column represents a gradually richer conditioning set. Robust standard errors are clustered at police district and reported

in parentheses, * p < 0.10, ** p < 0.05, *** p < 0.01. Control variables: offense month, offense age, type of the offense,

child gender, ethnicity, birth weight, child ADHD diagnose, child using prescriptive drugs, child’s criminal history (age

10-13), parents’ income, occupation and education, family type, parents’ criminal history, police district.

REU, Alm.del - 2017-18 - Bilag 60: Henvendelse af 10/11-2017 fra Aarhus Universitet, Institut for Økonomi v/Anna Piil Damm vedr. "Ny forskning om virkninger af at sænke den kriminelle lavalder"

Table A10. Effects of reform on recidivism (

penal code offenses

), population: 14-year-old penal code

offenders

Without cohort

corrections

With cohort

corrections

3 months

Effect of reform on recidivism

Birth month to reform month

0.0204

(0.0286)

0.1197

(0.0988)

-0.0052

(0.0051)

0.0973

(0.1202)

-0.0035

(0.0061)

0.0740

(0.1101)

-0.0019

(0.0058)

0.0824

(0.1051)

-0.0019

(0.0051)

0.0966

(0.0828)

-0.0028

(0.0041)

0.1285

(0.0853)

-0.0047

(0.0043)

1,569

Yes

6 months

Effect of reform on recidivism

Birth month to reform month

0.0305

(0.0206)

9 months

Effect of reform on recidivism

Birth month to reform month

0.0377

(0.0164)

12 months

Effect of reform on recidivism

Birth month to reform month

0.0471

***

(0.0125)

15 months

Effect of reform on recidivism

Birth month to reform month

0.0428

(0.0148)

18 months

Effect of reform on recidivism

Birth month to reform month

0.0388

***

(0.0122)

Data source: Administrative register data from Statistics Denmark and Police records for 14- year-old penal code

offenders pre- and post-reform.

Note: The reported estimates are coefficients from OLS regressions, rows represent separate regression models on the

probability of recidivism at 3, 6, 9, 12, 15 and 18 months and the two columns represent models without/with cohort

correction. Robust standard errors are clustered at police district and reported in parentheses, * p < 0.10, ** p < 0.05, ***

p < 0.01. Control variables: offense month, offense age, type of the offense, child gender, ethnicity, birth weight, child

ADHD diagnose, child using prescriptive drugs, child’s criminal history (age 10-13), parents’ income, occupation and

education, family type, parents’ crime history, police district.

Observations

Controls:

Calendar month dummies(offense)

Offense age and type

Child background variables

Family background variables

Police district fixed effects

1,569

Yes

REU, Alm.del - 2017-18 - Bilag 60: Henvendelse af 10/11-2017 fra Aarhus Universitet, Institut for Økonomi v/Anna Piil Damm vedr. "Ny forskning om virkninger af at sænke den kriminelle lavalder"

Table A11. Robustness analyses: Effects of reform on recidivism (

penal code offenses)

, population: 14-

year-old penal code offenders

3 months 6 months 9 months 12 months 15 months 18 months

Different subpopulations:

Pre-/post reform same calendar months 0.0200

(obs.=1,597)

(0.0212)

14-year-old offenders

(all types of crimes obs.=1,716)

With prior offenses by age 14

(obs.=180)

Without prior offenses by age 14

(obs.=1,389)

Boys (obs.=1,018)

0.0096

(0.0213)

0.0105

(0.0821)

0.0259

(0.0225)

-0.0055

(0.0385)

0.0804

(0.0305)

0.0469

(0.0360)

0.0193

(0.0242)

Yes

0.0203

(0.0171)

0.0092

(0.0205)

0.0592

(0.0821)

0.0319

(0.0168)

0.0139

(0.0326)

0.0715

(0.0377)

0.0522

(0.0253)

0.0245

(0.0176)

Yes

0.0255

(0.0184)

0.0167

(0.0168)

-0.0128

(0.1121)

0.0347

(0.0140)

0.0291

(0.0154)

0.0042

(0.0974)

0.0315

(0.0188)

0.0249

(0.0154)

0.0506

(0.0680)

0.0343

(0.0189)

0.0237

(0.0136)

0.0421

(0.0702)

0.0453

***

0.0590

***

0.0548

***

0.0475

***

(0.0126) (0.0135) (0.0145) (0.0146)

0.0182

(0.0294)

0.0839

(0.0389)

0.0534

(0.0222)

0.0338

(0.0156)

Yes

0.0283

(0.0257)

0.0835

(0.0429)

0.0609

(0.0225)

0.0498

***

(0.0140)

Yes

0.0305

(0.0252)

0.0771

(0.0388)

0.0559

(0.0218)

0.0427

(0.0198)

Yes

0.0361

(0.0191)

0.0618

(0.0397)

0.0589

(0.0223)

0.0455

(0.0192)

Yes

Girls (obs.=551)

Excluding offenders with violent off.

(obs.=1,243)

Excluding offenders with vandalism

(obs.=1,378)

Controls:

Calendar month dummies

Offense age and type

Child background variables

Family background variables

Police district fixed effects

Data source: Administrative register data from Statistics Denmark and Police records for 14- year-old penal code

offenders pre- and post-reform.

Note: The table shows result from robustness analyses with different (sub-) populations of 14-year-old penal code

offenders. Each column represents separate regression models on the probability of recidivism at 3, 6, 9, 12, 15 and 18

months including all control variable. Robust standard errors are clustered at police district and showed in parentheses, *

p < 0.10, ** p < 0.05, *** p < 0.01. Control variables: offense month, offense age, type of the offense, gender, ethnicity,

birth weight, child ADHD diagnosis, child using prescriptive drugs, child’s criminal history (ages 10-13), parents’

income, occupation and education, family type, parents’ criminal history, police district.

REU, Alm.del - 2017-18 - Bilag 60: Henvendelse af 10/11-2017 fra Aarhus Universitet, Institut for Økonomi v/Anna Piil Damm vedr. "Ny forskning om virkninger af at sænke den kriminelle lavalder"

Table A12. Educational mean outcomes for

14-year-old penal code offenders by sub group

Subsample: with recidivism

within 18 months

Pre-

Post-

reform reform

Diff.

Outcome:

Enrolled in 9th grade

Exam participation (0/1)

Type of school (if enrolled 9

grade):

Ordinary lower secondary schools

Boarding schools

Schools for children with special needs

Production colleges

Treatment institutions

Grades (conditional on participation):

Language Arts (teacher assessment)

Language Arts (exit exam)

Math (teacher assessment)

Math (exit exam)

0.87

0.58

0.56

0.09

0.10

0.14

0.11

-1.13

-0.91

-1.23

-1.04

0.80

0.62

0.56

0.10

0.09

0.14

0.11

-1.04

-0.99

-1.20

-1.14

-0.07

0.03

0.00

0.01

-0.01

0.00

-0.01

0.09

-0.08

0.03

-0.10

Subsample: without

recidivism within 18 months

Pre-

Post-

reform

Diff.

0.96

0.77

0.72

0.11

0.02

0.08

0.06

-0.57

-0.49

-0.74

-0.71

0.95

0.79

0.66

0.18

0.03

0.06

0.07

-0.65

-0.60

-0.78

-0.01

0.02

-0.06

0.07

0.01

-0.03

0.00

-0.08

-0.11

-0.04

-0.06

Data source: Administrative register data from Statistics Denmark and Police records for 14- year-old penal code

offenders pre- and post-reform.

Note: The table shows means for different educational outcomes (conditional on enrolment/participation) before age of

17 for the sample 14- year-old penal code offenders pre- and post-reform (obs.=1,569). The sample is divided into two

subgroups according to whether they recidivate or not within 18 months. The two columns labelled “Diff.” show the

differences between the pre and post reform groups within each sub group, and t-tests of difference in means * p<0.05,

** p<0.01, *** p<0.001.

REU, Alm.del - 2017-18 - Bilag 60: Henvendelse af 10/11-2017 fra Aarhus Universitet, Institut for Økonomi v/Anna Piil Damm vedr. "Ny forskning om virkninger af at sænke den kriminelle lavalder"

Economics Working Papers

2016-10:

John Kennes and Daniel le Maire: On the equivalence of buyer and

seller proposals within canonical matching and pricing environments

Ritwik Banerjee, Nabanita Datta Gupta and Marie Claire Villeval: The

Spillover Effects of Affirmative Action on Competitiveness and

Unethical Behavior

Rasmus Landersø, Helena Skyt Nielsen and Marianne Simonsen: How

Going to School Affects the Family

Leslie S. Stratton, Nabanita Datta Gupta, David Reimer and Anders

Holm: Modeling Enrollment in and Completion of Vocational

Education: the role of cognitive and non-cognitive skills by program

type

Nabanita Datta Gupta, Anton Nielsson and Abdu Kedir Seid: Short-

and Long-Term Effects of Adolescent Alcohol Access: Evidence from

Denmark

Michael Koch and Marcel Smolka: Foreign Ownership and Skill-biased

Technological Change

Mette Trier Damgaard and Helena Skyt Nielsen: Nudging in

education: A survey

Alexander K. Koch and Julia Nafziger: Motivational Goal Bracketing:

An Experiment

Ina C. Jäkel Allan Sørensen: Exporter Price Premia?

Marianne Simonsen, Lars Skipper and Niels Skipper: Piling Pills?

Forward-Looking Behavior and Stockpiling of Prescription Drugs

Federico Ciliberto and Ina C. Jäkel: Superstar Exporters: An

Empirical Investigation of Strategic Interactions in Danish Export

Markets

Anna Piil Damm, Britt Østergaard Larsen, Helena Skyt Nielsen and

Marianne Simonsen: Lowering the minimum age of criminal

responsibility: Consequences for juvenile crime and education

2016-11:

2017-01:

2017-02:

2017-03:

2017-04:

2017-05:

2017-06:

2017-07:

2017-08:

2017-09:

2017-10: