THE RELATION BETWEEN
CRIME AND UNEMPLOYMENT
IN THE NETHERLANDS
EBM877A20
Master’s Thesis
by
Nienke J.M. Oude Steenhof
s2171570
University of Groningen
Supervised by dr. Richard M. Jong A Pin January 16th, 2017
Abstract
This paper analyses the relation between crime rates in the Netherlands and unemployment benefits. To conduct the research, I use data on Dutch municipalities between 2007 and 2013. I form two models, a standard model and a model based on unemployment in the neighbouring municipalities. I find that the standard model shows a negative relation between crime and unemployment. The model based on the neighbouring municipalities shows a positive relation, in line with my hypothesis and supporting the claim that crime is not bounded to borders of a municipality. I find the crisis to have a negative effect on the crime rate.
2 1. INTRODUCTION
In 2005 the costs on society of crime in the Netherlands were estimated to be €20 billion (Groot et al., 2007). The Dutch government, as other governments around the world, aims to minimize the societal loss of crime (Becker, 1968). In 2010, the costs diminished to €17.7 billion (Moolenaar, Nauta, and Van Tulder, 2011). Crime rates reduced steadily in the Netherlands during the past decade, with a 5% decrease in 2015 compared to 2014 (CBS, 2016). To continue the decrease of crime and its societal costs, all aspects that determine crime and its related costs should be analysed. However, in the Netherlands, there is little, if no, research on the effect of unemployment on crime rates, even though there are studies in the field of criminology that link the two in the past, both theoretical and empirical (in other countries and regions) (Fleisher, 1966; Hooghe et. al, 2011; Crow et. al, 1989). Brenner (1976) formed the basis of many current researches, with an analysis on how several economic indicators can affect social wellbeing, for example health, education, and criminal aggression.
Unemployment is, of course, linked to the business cycle of the economy. In 2007 the economic crisis began; all over the world economies showed a declining trend. The Netherlands did not escape the crisis, and in 2009 this caused an enormous, exogenous rise in unemployment. Many companies filed for bankruptcy, or had to cut employment cost. This resulted in a nation-wide increase in unemployment from 4% to 5.4% (CBS/Statline, 2016). Of course, the increase differed per region.
3 I form two models, one standard model and a model that uses the average share of benefits in the neighbouring municipalities. In the standard model, I find a negative relation between the share of benefits and the total number of crimes. In the model with the average share of benefits of neighbouring municipalities, I do find a positive relation. This implies that crime is not bound to the borders of a municipality, and the share of benefits can influence crime rates in other municipalities.
In chapter 2, I discuss the theory on the matter. In chapter 3, the research framework and the data is introduced. The results are presented in chapter 4. Chapter 5 gives a conclusion, and in chapter 6 I present a discussion and ideas for future research.
2. THEORY
The topic of crime and unemployment is not new in the literature. For over a century, criminologists and economists researched a possible link between the two. In the 19th century,
Ferri (1881) published a paper where he introduced the link between economic wealth and crime.
The state of an economy and unemployment are known to have a considerable effect on many aspects of society, such as educational institutions and health care (Crow et. al., 1989). It would be illogical to think that crime and the justice system are not impacted. For example, Brenner (1976) provides an analysis on the relation between economic indicators and measurements of national wellbeing, such as physical health, mental health, and criminal aggression. He finds that of the economic indicators, unemployment has the most significant effect on all indicators of national wellbeing, including criminal aggression.
4 2.1 Motivated offenders
First, how does one find motivated offenders? Of course, unemployed people receive no income from legal employment, thus an illegal activity is preferred more readily. Fleisher (1966) reasons that a lower income reduces the expected cost of illegitimate activities, since the loss of lifetime utility from income in the case of conviction is lower. Furthermore, the opportunity cost of serving jail time is smaller. On the other hand, individuals with a low income expect relatively large payoffs from illegitimate activities compared to individuals with a high income, because they receive low payoffs from legitimate activities. Fleisher conducts a study in 101 cities in the United states on the relation between income and delinquency, where he finds that a 1% rise in income can reduce delinquencies with 2.5%. He also looks at the effect of a reduction in unemployment rate, where he finds that a 50% reduction in unemployment (in a region with a 10% unemployment rate) can reduce delinquencies with 10%. A Belgian study shows that poverty and inequality lead to higher crime rates (Hooghe et al., 2011). They find that poverty and inequality influence crime due to the increase of the feeling that people need to commit a crime. Low income, or low income compared to others, gives a feeling of unfairness, and a need to increase their own income, potentially through illegal activities.
In the Netherlands, unemployment does not mean that one loses his total income, since there is a social security system. The government provides two main benefits. Firstly, there are unemployment benefits (Werkloosheidswet, WW-uitkering). A person receives these the moment he becomes unemployed, and is 70% to 75% of his last earned salary. This benefit lasts between 3 and 34 months (UWV, 2016). After this period, one can receive social assistance (Bijstandsuitkering), which is 70% of the minimum wage for a single person (Rijksoverheid, 2016). These social benefits mean that people still have a legal income, and the loss in income is smoothened.
2.2 Suitable targets
5 example for motor vehicles. This is explained by the fact that in economic good times, people acquire more goods, causing chances of profiting from crime to increase. Thus, during good times people and their possessions are a more suitable target for theft. In times of economic downfall, social crimes, such as vagrancy and drunkenness, increase. In this case, the loss of income due to economic downfall causes a fall on the social scale.
2.3 Absence of capable guardians
6 3. RESEARCH FRAMEWORK
3.1 Research question
The main question asked in this paper is: how does unemployment affect crime rates in Dutch municipalities? Besides the total effect, I also view different effects of several types of crime. I answer this question through empirical research, anecdotal evidence, and an analysis of the effect of the crisis.
3.2 Data
3.2.1 Data collection
To answer the research question, I collect panel data from 364 Dutch municipalities during a time span of 7 years (2007 to 2013). All data is collected from CBS/Statline. Crime is divided into different types of crime, besides the total number of crimes. Not all municipalities have complete data on the division of crimes, therefore the different crime types have fewer observations than the total number of crimes. Furthermore, I use registered crime (reported by citizens and government), meaning that there is possibly more crime, but those crimes are not registered. CBS created a classification system (Standaardclassificatie Misdrijven 20101) to
group different types of crimes. Table 1 (descriptive statistics) provides short descriptions of the classifications. Furthermore, I give an extensive division of the classifications in Appendix A.
In addition, I collect data on social benefits granted per municipality. For privacy reasons, CBS does not give the exact number of unemployed. Therefore, I use the number of citizens receiving unemployment benefits (WW-uitkering) and social assistance (Bijstandsuitkering), and combine them to form the total number of social benefits. The number of benefits is rounded off to tens by CBS.
My empirical research is based on the three pillars from the theory, namely the presence of motivated offenders, the presence of suitable targets, and the absence of capable guardians. These three pillars result in three variables. Of course, the share of the population that receives social benefits represents the group of motivated offenders. For the suitable
1 For more information regarding the classification, please visit
7 targets, I use the average housing value in a municipality. I use this variable instead of average income in the municipality for two reasons. First, it is less correlated to unemployment in the short-term, where average income is. Second, the data on housing value is complete, where the average income per municipality is not. Finally, to account for the absence of capable guardians I use the percentage of registered crimes that are solved. When many crimes are solved, the chance of being convicted for a crime are higher. The variable shows the capability of the police system. I use the percentage of solved crimes in the same year as the unemployment variable for two reasons. First, the police force can announce at the beginning of the year what their focus will be. Second, it allows for a short run effect. If I took the percentage of year t-1, criminals have the possibility to change their operations, and hide them better. If the chance of being convicted changes in year t, they do not have the possibility to change immediately.
Not all Dutch municipalities are included in the research. Due to municipal reclassifications, there have been some municipalities that do not exist anymore and municipalities that have been created in the selected period. Because the reclassifications may affect the characteristics of the municipalities, I have decided to omit the municipalities that have been submitted to a reclassification between 2007 and 2013. A change in characteristics can include changes in the structure of the public system. For example, the police force can work together, causing an exogenous shock in the percentage of solved crimes. Also, some municipalities are split up and divided between several others. Methods that calculate the numbers over the previous and future years divide the municipality on the basis of population or area. This, however, dismisses differences in neighbourhoods within municipalities, since there is no division on the basis of the socio-economic status of the neighbourhoods that are divided. Thus, there is no division between rich neighbourhoods with high employment and poor neighbourhoods with low employment. I found the possibility of changes in characteristics of municipalities more harmful for my conclusions than omitting 30 municipalities on a total of 394 municipalities.
3.2.2 Data limitations
8 citizens of the neighbouring municipality are more valuable, resulting in a higher expected pay-off. This limitation does not apply to all sorts of crime. For example, property crimes are likely to be affected, if a neighbouring municipality has higher value in their personal belongings, and the payoff is therefore higher. On the other hand, vandalism is less likely to be committed in another municipality, because the action is more emotional. To account for this limitation, I perform an extra regression, which uses the average share of benefits in the neighbouring municipalities in year t.
Second, it does not relate individuals to crimes. Thus, the research cannot prove that unemployed citizens commit more (or less) crimes. Only a relation between the share of unemployed and the number of crimes can be established. However, as my theory states, unemployed citizens function as motivated offenders to commit a crime.
Third, because there is no individual information, there is no possibility of learning the motivations of the criminal. This means that causality is more difficult to prove, because the research does not show whether the unemployment is in direct relationship to committing (or not committing) a crime.
Fourth, the dataset is incomplete for some types of crime. Most municipalities have not kept a complete record of the committed crimes, for which there is no clear reason. It could be that the number of crimes is too limited to keep track, however it could also be the case that the law enforcement in the municipality is not completely organized. Because of the many possible reasons why there is no record, it is not possible to state whether there is an influence of the lack of observations in these municipalities on the results.
Finally, not all crimes are registered. A recent research by the Dutch government shows that the registered crime rates display less than 20% of the total number of crimes committed. Many people do not report crime, for example bicycle theft or being threatened (Trouw, 2017).
3.2.3 Descriptive statistics
9 municipality is more than 54. Some types of crimes have no registrations in some municipalities (but are tracked in the data), for example traffic crimes and drug offences. Also, solved crimes vary heavily. The share of the population receiving benefits also varies widely, namely from 0.66% in Rozendaal in 2012 to 8.86% in Rotterdam in 2013. This shows a strong difference between municipalities when looking at unemployment levels.
When looking at the geographical division of crime in the Netherlands, it is clear that in the western and southern regions of the Netherlands total crime rates are higher than northern and eastern regions. This trend is mainly due to property crimes (type 1) and traffic crimes (type 5). Other crimes show weaker regional differences in data. Large cities often have high crime rates per 1,000 citizens, compared to more rural areas.
Table 1: Descriptive statistics
Variable Mean Std.Dev. Min Max N n Explanation
crime total 54.57 20.64 16.20 142.20 2539 363 Total crimes, per 1,000 citizens crime 1 30.43 13.56 7.50 95.20 2502 362 Property crimes, per 1,000 citizens
crime 2 10.68 3.87 3.70 48.20 2163 341 Vandalism, crimes public order, per 1,000 citizens
crime 3 6.21 2.24 2.40 14.00 1461 249 Violence and sexual crime, per 1,000 citizens crime 4 0.69 0.54 0.00 1.80 214 69 Crimes code of criminal law, other than property,
vandalism and violence, per 1,000 citizens crime 5 7.95 2.78 0.00 22.30 1959 310 Traffic crime, per 1,000 citizens
crime 6 1.85 1.90 0.00 10.70 282 80 Drug offences, per 1,000 citizens crime 7 0.30 0.74 0.00 4.70 235 106 Weapon crime, per 1,000 citizens crime 9 0.08 0.33 0.00 2.50 145 87 Other crimes, per 1,000 citizens
share benefits 2.77 1.24 0.66 8.86 2538 364 Share of population receiving unemployment benefits or social security, January
share benefitsn 2.78 0.62 1.10 6.28 2466 356 Average share of population neighbouring
municipalities receiving social benefits, January value 250.61 61.71 129.00 636.00 2539 364 Average value of residences, in €1,000
10 3.3 Anecdotal evidence
In this paragraph, I look at some events that caused a sudden shock in unemployment, and the reaction of crime to these shocks.
Zalco (Zeeland Aluminium Company) and Thermpos, Vlissingen
In 2011, Zalco filed for bankruptcy. With 500 employees, Zalco was a large employer within the municipality of Vlissingen. In 2012, the second big company in Vlissingen went bankrupt, Thermpos (phosphorus factory), having 475 employees (Business Insider, 2015). The number of citizens receiving benefits rose from 720 to 920 between 2011 and 2012, and on to 1,140 in 2013. Crime rates did not increase, they continued the steady decrease that was achieved in earlier years, from 77.7 in 2011 to 71.1 in 2012 and 66.7 in 2013, similar to the steady decrease of crime in the Netherlands in total (CBS, 2016).
Groothuis Woningbouw B.V., Tubbergen
Groothuis was a big player on the regional construction market. With its bankruptcy in 2012, 110 employees suddenly lost their jobs (Business Insider, 2012). At beginning of 2013, the unemployment benefits rose in Tubbergen by 160 citizens receiving benefits, a 43.2% increase. Between 2011 and 2012, there has been a large decrease in crimes committed per 1,000 citizens (from 25 to 21.5). However, in 2013 the number of crimes increased by 1.5 per 1,000 citizens. Through the years, there were changes in crime of this size, so it is difficult to attribute this increase to unemployment caused by the bankruptcy of Groothuis Woningbouw. Licom N.V., several municipalities in Limburg
11 Concluding, the events mentioned all show no changes in crime rates that can be accounted to the shock in unemployment. Therefore, these local cases do not prove a relation between unemployment and crime rates.
3.4 Empirical analysis
I do a Woolridge autocorrelation test, with the purpose of using the most appropriate model for my eventual analysis of the data. I assume that a dynamic panel data approach is most fit for the data, for example an Arellano-Bond model. Furthermore, the Arellano-Bond model is a good fit for my panel data where there is a large N (2539 observations for the total number of crimes), and a small T (7 years) (Roodman, 2006).
The results show that autocorrelation is present, thus I use the Arellano-Bond model 3.5 Models
To conduct the research, I form two models, namely a standard model, and a model taking the average of the observations of neighbouring municipalities. Because I have panel data, and want to leave room for dynamics that arise from autocorrelation, I use the Arellano-Bond estimation approach, with one lag, crimex,it-1. This method differs from static panel data
models, because it uses instruments to prevent endogeneity and it corrects for autocorrelation by estimating a variance matrix. I add a lag, because I expect crime in period t to be dependent on crime in period t-1. An environment of high crime, invites others to commit crimes. Furthermore, a criminal can increase the number of crimes committed if he was successful.
3.5.1 Standard model
The research will be performed with all crime types. The model in equation 1 adjusts for different types of crimes.
crimex,it = ρ1*crimex,it-1 + β1*share benefitsit + β2*valueit
+ β3*crimex solvedit + β4-9*year dummies + μi + υit
12 In equation 1, crimex,it is the number of crimes of type x per 1,000 citizens in municipality i at
time t. Similarly, crimex,it-1 is the number of crimes at time t-1. The variables share benefitsit,
crimex solvedit, and valueit represent respectively the share of citizens receiving social
benefits, the percentage of solved crimes, and the average housing value in municipality i at time t. The year dummies are 1 in the represented year (2007-2013) and 0 otherwise. Finally, μi stands for the fixed effects per municipality i, and υit are the idiosyncratic shocks in
municipality i at time t. Because I include these two terms, there is no further need to implement other variables to define the difference between the municipalities, because leave room for differences between municipalities, both static differences and changes per year. 3.5.2 Neighbouring municipalities
As mentioned in section 3.2.2 (Data limitations), crimes are not bound to a municipality. Therefore, it is interesting to look at the effect when we take crime and housing value of the neighbouring municipalities. For this reason, I form an extra model as specified in equation 2.
crimex,it = ρ1*crimex,it-1 + β1*share benefitsnt + β2*valueit
+ β3* crimex solvedit + β4-9*year dummies + μi + υit
(2)
In equation 2, share benefitsnt represents the average share of benefits in the neighbouring
municipalities n at time t. 3.6 Hypothesis
3.6.1 Standard model
On the basis of the previously mentioned literature, I expect there to be a positive relation within my dataset on the relation between unemployment and crime. I hypothesize that the strength of the relation differs across different types of crimes, however the total effect will be positive. Thus, this implies the following:
H0: β1=0 (3)
13 I expect the strongest effects in the cases of property crimes, and vandalism and crimes on public order. Property crimes are a rational choice if one loses his income, and needs to acquire extra income. Vandalism and public order are expected to be higher due to the emotional aspect of these crimes. Unemployed people are more likely to find themselves to be aggressive due to the fall on the social scale.
3.6.2 Neighbouring municipalities
In the case of neighbouring municipalities, I expect again a rise in crime when the average share of benefits increases. Thus, the hypothesis is equal to equations 3 and 4. I expect this because crime is not bound to the borders of a municipality. The hypothesis is therefore the same as in the standard model. I expect the strongest effect for crimes of type 1. Since I think property crimes are most rational, a criminal is more rational in choosing a location to commit the crime. This causes them to consider a different location more readily.
3.6.3 Effects of the crisis
I also perform an analyses on the effect of unemployment due to the crisis on crime rates. Similar to the standard model, I expect the exogenous to have a positive effect on crime rates. 3.7 Specification choices
14 4. RESULTS
This chapter is divided into five paragraphs. First, I look at the standard model with the total number of crimes. Second, I look at the results of the standard model for the different types of crimes. Third, I will look at the model taking neighbouring municipalities into account from the view of the total number of crimes. Fourth, I again look at the different types, this time for the neighbouring model. Finally, I analyse the effect of unemployment caused by the crisis on crime.
Table 2: Results standard model, total crime Variable crime total crime totalt-1 0.496*** (0.083) share benefits -0.952* (0.530) value -0.054*** (0.018) crime total solved -0.213***
(0.036) yr08 4.492*** (1.012) yr09 5.356*** (1.006) yr10 3.702*** (0.754) yr11 4.086*** (0.565) yr12 1.720*** (0.474) # instruments 47 # observations 1804 # groups 363 Sargan test chi2 prob>chi2 50.33 (0.087)
The standard errors are given between the brackets. Note:
15 4.1 Standard model total number of crimes
In table 2, I give the results of the standard model on the total number of crimes. The main coefficient I am interested in, is of course the share of benefits. The coefficient of -0.952 is significant at the 10% level, but not moving in the direction that I expect in my hypothesis, as it is negatively related to the number of crimes committed per 1,000 citizens. An explanation is that unemployment also decreases the number of people that function as suitable targets for crimes, as their possibility to acquire property decreases.
Both the value of housing and the percentage of solved crimes are significant at the 1% level. The coefficient for the value of housing is -0.054, meaning that for a rise in the value of housing of €1,000, crimes decrease by 0.05, which is a small change in crimes. This provides the opposite result compared to the theory, which states that the presence of suitable targets increases crime rates. However, the value of housing can also function as a proxy for the socio-economic state of a municipality. The percentage of solved crimes moves as the theory expects. If the percentage of solved crimes increases by 1%, the total number of crimes per 1,000 citizens decreases with over 0.2 crimes. This impact of chances of getting caught is in line with the theory. It seems like a small impact, but if the probability of conviction rises with 10%, the number of crime decreases with 2 crimes per 1,000 citizens. When looking at the mean number of crimes, 54, this means a decrease of over 4%. This supports the claims that the probability of conviction is important in reducing the number of crimes. Finally, the lagged variable for crime is very significant (at the 1% level), and positive. This is a logical relation, as it means that when there was an environment of high crime rates in the previous year, it is likely that the current environment still has high crime rates. Also the coefficients for the year dummies are very significant, all at the 1% level. They differ per year, and do not show an increasing or decreasing trend, but they are all positive.
4.2 Standard model different crime types
16 increase in benefits leads to a decrease of these crimes. There is no strong explanation for the decrease of crimes of type 2 when there is an increase in unemployment. It can be that people rationally think of the consequences, for example receiving a fine. Because of a lower income, a fine is an extra high burden. To avoid receiving a fine, people may not commit these crimes. Traffic crimes have a more logical explanation, as one spends less time in traffic when unemployed, decreasing the chance of committing a traffic crime. Furthermore, drug offences can by explained through the demand for drugs. As one loses income, they have less to spend on luxury goods. Drugs are often a luxury, and expensive, good. When people have to cut
Table 3: Results standard model, different crime types
Variable crime 1 crime 2 crime 3 crime 4 crime 5 crime 6 crime 7 crime 9 crimex,t-1 0.609*** (0.086) -0.112* (0.066) 0.329*** (0.110) -0.385*** (0.114) 0.190** (0.083) 0.247*** (0.032) -0.096 (0.213) 0.280 (0.424) share benefits -0.172 (0.371) -0.617*** (0.226) -0.062 (0.103) 0.039 (0.059) -0.336** (0.143) -0.086** (0.041) 0.053 (0.051) -0.070 (0.047) value -0.018 (0.013) -0.006 (0.009) -0.012** (0.005) 0.001 (0.002) -0.024*** (0.005) 0.025*** (0.006) 0.001 (0.001) 0.018*** (0.004) crimex solved -0.144*** (0.026) -0.034*** (0.011) -0.000 (0.003) -0.240 (0.451) 0.018*** (0.005) 0.002 (0.003) 0.002 (0.006) yr08 -0.472 (0.328) 0.966*** (0.149) -0.128 (0.118) 0.190* (0.098) 0.377*** (0.080) -0.058 (0.730) -0.495*** (0.118) yr09 0.901*** (0.284) 0.198*** (0.075) -0.063 (0.131) -0.007 (0.091) -0.035 (0.094) -0.537*** (0.136) yr10 -1.239*** (0.215) -0.277** (0.109) -0.065 (0.070) 0.452*** (0.134) -0.275*** (0.057) -0.026 (0.027) -0.540*** (0.109) yr11 0.369* (0.203) -1.894*** (0.281) -0.294** (0.141) -0.017 (0.074) 0.174 (0.144) -0.233*** (0.040) 0.024 (0.046) -0.341*** (0.069) yr12 -0.383* (0.223) -2.846*** (0.325) -0.383** (0.154) 0.008 (0.056) -1.060*** (0.166) -0.157*** (0.047) yr13 -1.048** (0.429) -3.852*** (0.483) -0.873*** (0.224) -1.123*** (0.290) 0.085 (0.053) -0.080 (0.061) # instruments 47 48 47 48 48 49 40 15 # observations 1769 1465 947 86 1331 135 60 15 # groups 358 317 214 23 280 38 18 8 Sargan test chi2 prob>chi2 33.80 0.664 41.84 0.348 46.80 0.155 37.82 0.524 54.42 0.051 22.29 0.989 11.98 0.999 15.00 0.020 The standard errors are given between the brackets.
Note:
17 spending, due to a loss of income, they decrease the spending on luxury goods such as drugs. The decrease in demand causes it to be less interesting to supply drugs. This means that the number of drug crimes decreases. The other crime types vary in sign, but are all insignificant.
The coefficients for the average housing value in a municipality differ in sign and significance. It is negative and significant with crimes of type 3 (violence and sexual crime) at the 5% level, and type 5 (traffic crimes) at the 1% level. This can, again, be due socio-economic state of a municipality. The coefficients related to crimes of type 6 (drug offences), and type 9 (other crimes) are positive and significant at the 1% level. The relation of drug offences has the same reason as the share of benefits, namely the availability of funds (of which the value of housing is a proxy) to buy drugs, thus increasing demand. The relation to other crimes can be explained by the function of a socio-economic proxy of the variable for housing value. Furthermore, I find that the coefficients for the percentages of crimes solved are significant for crimes of type 1, 2 and 5. At crimes of type 1 (property crime) and type 2 (vandalism and public order) the coefficients are negative, as the theory predicts. The coefficients are relatively large, similar to the overall results. For crimes of type 1, a 10% increase in solved crimes leads to a decrease of 1.4 crimes per 1,000 citizens in a municipality. This is a 4% decrease compared to the mean of 30 crimes per 1,000 citizens of type 1. Type 2 crimes decrease by 0.3 if solved crimes increase by 10%. This means a 3% decrease compared to the mean of 10 crimes per 1,000 citizens. The opposite is the case for crimes of type 5 (traffic crimes). The coefficient for the percentage of crimes solved is positive. This is not in line with theory that crime decreases when chances of being convicted increase. The possible reason is that at the moment that there is an increase in these crimes, the police and justice system move more resources towards the investigation of these crimes, causing a higher percentage of solved crimes. This implies reverse causality.
4.3 Neighbouring model total number of crimes
18 locate their crimes in a neighbouring municipality. The other variables are similar to the standard model in sign and significance.
Table 4: Results model neighbouring municipalities, total crime Variable crime total
crime totalt-1 0.351*** (0.090) share benefitsnt 0.642*** (0.216) value -0.051** (0.020) crime total solved -0.210***
(0.037) yr08 1.636*** (0.421) yr09 2.318*** (0.304) yr10 yr11 0.009 (0.343) yr12 -2.524*** (0.430) yr13 -5.242*** (0.710) # instruments 51 # observations 1748 # groups 356 Sargan test chi2 prob>chi2 51.60 0.147
The standard errors are given between the brackets. Note:
19 4.4 Neighbouring model different crime types
In table 5, I give the results of the model on neighbouring municipalities on the different crime types. Again, I investigate the coefficient of the average share of benefits in the neighbouring municipalities. The coefficient related to crimes of type 1 (property crimes) is positive and significant at the 1% level. This supports my hypothesis. Furthermore, it supports that crimes such as theft are likely to be committed in a neighbouring municipality, as it can be more attractive to commit a crime in another municipality. The coefficients of crimes of
Table 5: Results model neighbouring municipalities, different crime types
Variable crime 1 crime 2 crime 3 crime 4 crime 5 crime 6 crime 7 crime 9 crimex,t-1 0.650*** (0.060) 0.097 (0.101) 0.244** (0.102) -0.322** (0.143) -0.033 (0.073) 0.294*** (0.027) 0.032 (0.249) share benefitsnt 0.510*** (0.179) 0.159 (0.103) 0.038 (0.038) -0.005 (0.009) 0.057 (0.066) -0.010 (0.047) -0.016 (0.039) -0.256*** (0.069) value -0.019 (0.014) -0.004 (0.008) -0.014*** (0.005) 0.002 (0.002) -0.016*** (0.006) 0.027*** (0.006) -0.000 (0.002) 0.005 (0.006) crimex solved -0.146*** (0.027) -0.021* (0.012) -0.002 (0.003) 0.004** (0.002) 0.013*** (0.005) -0.001 (0.002) -0.000 (0.005) yr08 -0.306 (0.261) 2.369*** (0.227) -0.194*** (0.034) 0.697*** (0.115) -0.128* (0.074) yr09 0.968*** (0.223) 1.483*** (0.149) 0.202*** (0.711) -0.140** (0.065) -0.296*** (0.101) 0.258*** (0.069) -0.027 (0.074) 0.325*** (0.109) yr10 -0.344*** (0.080) -0.180*** (0.048) -0.244* (0.133) -0.039 (0.027) 0.041 (0.138) yr11 0.376* (0.209) -0.257 (0.182) -0.383*** (0.098) -0.069 (0.057) -0.494*** (0.110) -0.050 (0.049) 0.046 (0.032) 0.229*** (0.045) yr12 -0.492** (0.242) -1.237*** (0.237) -0.518*** (0.096) -0.001 (0.030) -1.700*** (0.107) 0.304*** (0.074) yr13 -1.140*** (0.351) -2.449*** (0.345) -1.104*** (0.134) -2.230*** (0.184) 0.228** (0.116) -0.052* (0.030) 0.367*** (0.084) # instruments 51 50 51 48 51 47 36 12 # observations 1721 1429 924 79 1297 127 49 12 # groups 352 314 212 22 277 37 15 6 Sargan test chi2 prob>chi2 35.78 0.739 28.93 0.922 42.80 0.437 40.03 0.424 45.89 0.314 21.51 0.986 10.29 0.998 12.00 0.007 The standard errors are given between the brackets.
Note:
20 type 2 (vandalism and public order), type 3 (violence and sexual violence), and type 5 (traffic crimes) are positive, but not significant. Crimes of type 9 (other crimes) are negative and significant. This can be due to the presence of economic crimes, which are more often committed in a work environment, and thus linked to the number of employed in a municipality. The other variables show no interesting differences in signs and significance. 4.5 Effects of the economic crisis
In the late 2000s, the crisis hit, where one can find the highest increase in unemployment in 2009. To check for an effect in crime due to this increase, I have performed an analysis where I compare the years 2007/2008 (before the crisis) and 2010/2011 (crisis). I excluded 2009, because I consider that year as a ‘shock’. In the analysis, I look at the difference before and after this shock. This analysis has the form of equation 5.
crimex,it = β1*share benefitsit + β2* valueit + β3*crimex solvedit
+ β4*crisis + β5* crisis*share benefitsit
(5)
In equation 5, crisis is a dummy variable with a value 1 for 2010 and 2011 and 0 otherwise and crisis*share benefitsit is the interaction term between the dummy for the crisis years and
the share of benefits. In table 6, I present find the results of these regressions.
As can be seen in table 6, with most regressions the coefficients of the variables crisis and the interaction term crisis*share benefitsit are both negative and significant for the total
21
Table 6: Results analysis crisis
Regression crisis crisis*share benefitsit
crime total -3.80*** (0.73) -1.21*** (0.22) crime 1 0.13 (0.53) -0.73*** (0.16) crime 2 -2.20*** (0.27) -0.46*** (0.08) crime 3 -0.76*** (0.15) -0.10** (0.04) crime 4 -0.17** (0.07) 0.03** (0.02) crime 5 -1.39*** (0.11) 0.11** (0.05) crime 6 -0.41 (0.40) -0.08 (0.09) crime 7 0.13 (0.10) -0.02 (0.03) crime 9 0.01 (0.02) -0.00 (0.01) The standard errors are given between the brackets.
Note:
*significant at the 10%-level **significant at the 5%-level ***significant at the 1%-level
5. CONCLUSION
Governments take many actions to diminish the number of crimes. However, there has not been an investigation on the effects that unemployment has on crime in the Netherlands, despite many international papers. In this paper, I investigated the relation between the two in Dutch municipalities.
The research is based on the model of Cohen and Felson (1979). They base crime on three pillars: the presence motivated offenders, the presence of suitable targets, and the absence of capable guardians. I analysed the three using three different variables. Motivated offenders are represented by the share of unemployed citizens of the population in a municipality. Suitable targets are represented by the average housing value of the municipality. The absence of capable guardians is measured with the percentage of solved crimes.
22 in the neighbouring municipalities, because crime is not limited to the borders of a municipality. I expected the coefficient to be positive in both models.
The results show that the coefficient for the share of benefits is significant in the standard model for the total number of crimes committed, but the movement is opposite of my hypothesis. I conclude that there is no evidence in the data that there is a positive relation between crime and unemployment in the Netherlands. I account the negative coefficient to the decrease in the suitability as a target of the unemployed citizens. Also in the cases for different crime types, I find difference in significance, but the significant coefficients are again negative.
The results of my second model are significant and positive, and support my hypothesis. This is the case for both the total number of crimes and property crimes. I conclude that crime is indeed not bound to a municipality, and unemployment in neighbouring municipalities can cause an increase in crime in a municipality.
The other two variables, value and the percentage of solved crimes, differ in significance. Value is often significant, however very small. Also, the movement in relation to the total number of crime is opposite as the theory predicts, since crime decreases as the value of housing increases. The percentage of solved crimes is very significant and similar to my theory, as high chances of being convicted for a crime decrease the total number of crimes. This is the case for the total number of crimes and for the two largest subgroups. I find the effect to be relatively large. It emphasizes the importance of the probability of conviction. In smaller subgroups, the effect is opposite, which may be due to reverse causality.
I also conduct a research on the effect of unemployment caused by the economic crisis on the crime rate. I observe a negative relation between unemployment due to the crisis and crime rates.
Concluding, there is no proof for a positive relation between unemployment and crime within a municipality. There is, however, a positive relation between unemployment in neighbouring municipalities and crime in a municipality. In the following chapter, I discuss the research and possibilities for future research.
6. DISCUSSION AND FUTURE RESEARCH
23 social benefits and the volume of crime, without taking the committers into account. Secondly, the crimes are attributed to the municipality in which they were committed, and not the municipality of the criminal. If a criminal would do a cost-benefit analysis, they logically want to look at a neighbouring municipality, where the belongings of its citizens are higher, and there is more to gain from crime. This is in line with my results Finally, there is no personal record, in which the motivation behind a crime is explained. This could lead to further insights on the (causal) relationship between unemployment and crime. Instruments that can be used are the length of unemployment and last salary received before unemployment.
The allocation of social benefits in the Dutch system makes a direct relation less plausible, because the need to provide an income is less urgent. If one loses their job, the government will provide him with some benefits which eases the loss of income.
For future research, I advise to conduct an individual research on criminals, where motivational theories can be used. Furthermore, length of unemployment and change in salary are interesting instruments to use. This creates the possibility to conduct a research that can analyse causality.
REFERENCES
Becker, G.S. (1968), Crime and punishment: an economic approach. Journal of Political Economy.
Brenner, H.M. (1976), Estimating the social costs of national economic policy: Implications for mental and physical health and criminal aggression. Paper No. 5, Joint Economic Committee, Congress of the United States, Washington, D.C.: U.S. Government Printing Office.
Business Insider (2012), Bouwbedrijf Groothuis failliet; 110 werknemers op straat. Available at: https://www.businessinsider.nl/bouwbedrijf-groothuis-failliet-110-werknemers-op-straat/. [Accessed 1 December 2016].
24 CBS (2010), Standaardclassificatie misdrijven 2010. Available at:
https://www.cbs.nl/nl-nl/onze-diensten/methoden/classificaties/standaardclassificatie-misdrijven-2010. [Accessed 28 September 2016].
CBS (2016), Geregistreerde criminaliteit daalt met 5 procent. Available at: https://www.cbs.nl/nl-nl/nieuws/2016/11/geregistreerde-criminaliteit-daalt-met-5-procent. [Accessed 16 November 2016].
CBS/Statline (2016), CBS Statline. Available at: statline.cbs.nl. [Accessed 27 September 2016].
Cohen, L.E., and M. Felson (1979), Social change and crime rate trends: a routine activity approach. American Sociological Review 44:588-607
Crow, I., P. Richardson, C. Riddington, F. Simon (1989), Unemployment, crime and offenders. National Association for the Care and Resettlement of Offenders.
Ehrlich, I. (1973), Participation in illegitimate activities: an economic analysis. Journal of Political Economy.
Ferri, E. (1881), New horizons of criminal law and penal procedure. Turin.
Fleisher, B.M. (1966), The effect of income on delinquency. Economic Review, 56 (1).
Groot, I, T. de Hoop, A. Houkes, D. Sikkel (2007), De kosten van criminaliteit: Een onderzoek naar de kosten van criminaliteit voor tien verschillende delicttypen. SEO Economisch onderzoek, SEO Rapport 971, mei 2007.
25 Moolenaar, D.E.G., B. Nauta, F.P. van Tulder (2011), Kosten van criminaliteit. Criminaliteit
en Rechtshandhaving 2011, WODC, 251-281.
Rijksoverheid (2016), Bijstand (Onderwerp, Rijksoverheid.nl). Available at:
https://www.rijksoverheid.nl/onderwerpen/bijstand. [Accessed 1 December 2016]. Roodman, D. (2006), How to do xtabond2: An introduction to “difference” and “system”
GMM in Stata. The Stata Journal, Volume 9, Number 1, pp. 86-136, November 2006. Trouw (2017), ‘Nederlandse politie ziet de meeste misdaad niet’. Available at:
http://www.trouw.nl/tr/nl/39681/nbsp/article/detail/4447567/2017/01/13/Nederlandse-politie-ziet-de-meeste-misdaad-niet.dhtml. [Accessed 13 January 2017].
United Nations Social Defence Research Institute (1976), Economic crises and crime. Rome, UNSDRI.
UWV (2016), Kan ik een WW-uitkering krijgen? Available at:
26 Appendix A: Crime classifications
CBS created a classification system for all types of crime (CBS, 2010). Below, I present the classification groups and their subgroups. Also, I give the articles of the Penal Code, the Traffic Law, and the Opium Law to which they relate. The terms are translated from the official terms in Dutch2.
1 Property crimes Penal Code Article 1.1 Theft, embezzlement and burglary 310-312, 321-325
1.1.1 Theft and burglary with violence 312 1.1.2 Simple theft 310 1.1.3 Qualified theft 311 1.1.4 Embezzlement 321-325 1.2 Deception 326-338 1.2.1 Fraud 326 1.2.2 Absconding 326a 1.2.3 Fraud (other) 326b-338 1.3 Forgery crimes 208-214, 216-223, 225-234 1.3.1 Coin Crime 208-214
1.3.2 Falsification of seals and marks 216-223 1.3.3 Forgery 225-234 1.4 Fencing 416-417bis 1.5 Extortion and blackmail 317-318 1.6 Bank break 340-343
1.7 Laundering 420bis-420quater 1.8 Property crime (other) 314-315, 344-348 2 Vandalism and public order Penal Code Article 2.1 Destruction and injury 350-352
2.2 Public order offence 131-136, 138-151c 2.2.1 Public violence 141
2.2.2 Trespassing 138 2.2.3 Breach of the peace 139
2 For the Dutch terms of the classifications, please visit
27 2.2.4 Computer intrusion 138ab, 138b
2.2.5 Discrimination 137c-137g
2.2.6 Public order offence (other) 131-137, 138a, 139-140, 142-151c 2.3 Arson, explosion 157-158
2.4 Public authorities offence 177-182, 184-206 2.4.1 Failure to follow official order 184
2.4.2 Rebellion 180-182 2.4.3 False declaration 188 2.4.4 Smuggling 197a 2.4.5 Stay unwanted foreigner 197
2.4.6 Public authority offence (other) 177-179, 185-187, 189-196, 197b-206 3 Violence and sexual offences Penal Code Article
3.1 Abuse 300-306 3.2 Threats and stalking 284a-285b
3.2.1 Threat 284a-285a 3.2.2 Stalking 285b 3.3 Sexual offence 239-250 3.3.1 Sexual assault 246 3.3.2 Rape 242 3.3.3 Desecration of accountability 239 3.3.4 Fornication with minor 248a-248e 3.3.5 Pornography 240-240b 3.3.6 Fornication with abuse of authority 249
3.3.7 Sexual offence (other) 243-245, 247-248, 250 3.4 Life felony 287-296
3.5 Custody, kidnapping 282-282a 3.6 Trafficking 273f
3.7 Violence offence (other) 274-281, 307-308 4 Crime code of criminal law (other)
5 Traffic offences Traffic Law Article (WVW) 5.1 Abandoned place accident 7
28 5.4 Driving while driving ban 162
5.5 Conducting false registration number 41 5.6 Joyriding 11 5.7 Refusing breathalyser, blood test 163
5.8 Traffic offence (other) 6, 51, 61, 74, 114, 138 6 Drug crimes Opium Act Article 6.1 Hard drugs 2
6.2 Soft drugs 3 7 (Fire) weapon offences
9 Crimes other laws
9.1 Economic law crimes (WED) 9.1.1 Environment crimes
9.1.2 Other economic law crimes
9.2 Military crimes 96-166 (WvMSr) 9.3 Crimes other laws (other)
