1 | P a g e
The Determinants of Crime in the Post Soviet Union
and Eastern European Countries
By
Alex Kochlashvili
Department of Economics
University of Groningen
August 29
Supervisor:
Rob Alessie
2 | P a g e
Abstract
This paper studies the effect of probability of arrest, unemployment and average income level on different types of crime in the post Soviet Union and Eastern European countries during the period of 1990-2009. The sample covers the following 21 countries: Albania, Bulgaria, Croatia, Czech Republic, Hungary, Poland, Romania, Slovakia, Slovenia, Armenia, Azerbaijan, Belarus, Estonia, Georgia, Kazakhstan, Kyrgyzstan, Latvia, Lithuania, Moldova, Russian Federation, and Ukraine. The research question addressed in this paper is the following: How do the probability of arrest, unemployment and income level (police) affect the number of registered crime in the post Soviet Union and Eastern European countries. The following hypotheses are stated: 1)the probability of arrest has negative effect on crime;2) unemployment has a positive effect on crime;3) the effects of those explanatory variables are stronger on property crime than on violent crime;4) the effect of probability of arrest on crime differs between Soviet Union member and non member countries. In order to test these hypotheses and to answer the research questions, panel data models have been estimated. The estimation results show that probability of arrest has a negative effect on crime rate, unemployment has a positive effect on the crime rate, and this effect is higher in case of property crime than in case of violent crime.
3 | P a g e
1
Introduction
A crime can be defined as an act, which is forbidden according to public law or constitution of a country. The individuals, who violate the law are considered to be offenders and are liable to be punished according to the corresponding law. Among different types of crime, two main categories can be distinguished: a property crime and a violent crime. Property crime is committed, when an individual damages, destroys or steals any kind of property, which belongs to someone. This category includes robbery, theft, and burglary. Property crime is considered to be the most commonly committed crime in almost all countries (Hall, 1952). A crime is violent, when an individual’s act attempts to harm, threatens to harm or even conspires to harm someone else (Athens, 1992). This type of crime is offensive, which involves force or threat of force, such as rape, homicide, assault.
Individuals, who commit any kind of crime are supposed to be punished according to the corresponding law. A punishment level of offenders differs based on the category of crime they commit and based on the country or region, where it is committed. While an individual, who commits crime may have benefits from the action, the victim of the action or overall society will have losses. Thus, the issue of crime was and still remains as one of the core concepts, which needs to be eradicated.
4 | P a g e
any kind of crime, because expected cost in terms of reputation, loosing job and earnings in case of being caught will be higher than the expected cost of the individuals, who are not working. Similar logic may apply to the education and crime rate. As to the probability of arrest, it may be linked to the crime rate in a country in a sense that in the countries, where legal system or control is weak, and the probability of catching and punishing the offenders is low individuals may have more incentive to violate the law.
In the first half of the twentieth century, scholars were considering that a criminal behavior is like a social illness, see Merton (1938). Becker (1968) was the first, who studied the effect of the probability of being caught and nature of punishment on the individuals’ decision to participate in illegal activity and thus, to commit a crime. Becker’s model predicts a negative relationship between the probability of offender being arrested and punished and a number of committed crimes in a country. Many theoretical and empirical researches have been conducted in order to study the relationship between participation in illegal activities (committing crime) and a quality of the law system in the country or an economic status (employed/unemployed) of an individual in the labor market.
5 | P a g e
enforcement activity. Besides, the model links formally “the theory of participation in illegitimate activities with the general theory of occupational choices by presenting the offender's decision problem as one of the optimal allocation of resources under uncertainty of competing activities both inside and outside the market sector, rather than as a choice between mutually exclusive activities” (Ehrlich (1973)). In addition, the analysis distinguishes between deterrent and preventive effects of punishment by imprisonment on the rate of crime and gives the opportunity of determining empirically the preventive effect of the imprisonment on crime rate. The results of the estimation provide estimates of the effectiveness of law enforcement in deterring crime and reducing the social loss from crime. While theoretically the effect of unemployment on crime is obvious, empirical literature of early period does not show this trend and fails to support mentioned hypothesis and shows little and statistically insignificant effect on change of labor market opportunities on crime ( (Entorf and Spengler (2000)), (Chiricos (1987))).
6 | P a g e
opportunities in the country. Other factors such as education level, year and country fixed effects will be taken into account as well. Besides, this paper also controls for demographic variables, such as gender distribution, age distribution and crime rate in the country.
7 | P a g e
This thesis addresses following research questions: How do the probability of arrest, unemployment and income level (police) affect on the number of registered crime in the post Soviet Union and Eastern European countries. Based on the theory of economics of crime with a special focus on the theory of Becker(1968) and Ehrlich (1973), following hypothesis are checked in case of mentioned countries:
1. A probability of arrest has a negative effect on the number of registered crime in the Post Soviet Union and Eastern European countries; Graph 1.a and graph 1.b reports preliminary prediction of the relationship between property crime and probability of arrest and a violent crime and a probability of arrest. The slope in both cases is negative but it is steeper in case of property crime than in case of violent crime.
Graph 1.a Property crime and probability of arrest Graph 1.bViolent crime and probability of arrest
8 | P a g e
3. The size of effect of unemployment on crime is higher on property crime rather than on violent crime in post Soviet Union and Eastern European countries; Graph 2.a and graph 2 b shows the relationship between unemployment and property crime and unemployment and violent crime correspondingly. The direction of relationship is not as obvious as in case of probability of arrest, but it seems to be positive in both cases. Many other factors which may have influence on the crime are not included in this graphs, thus doing regression analysis will give the opportunity to better explain determinants of crime.
Graph 2.a property crime and unemployment Graph 2.b property crime and unemployment
9 | P a g e
The research will rely on the panel database of 21 countries, which includes Albania, Bulgaria, Croatia, Czech Republic, Hungary, Poland, Romania, Slovakia, Slovenia, Armenia, Azerbaijan, Belarus, Estonia, Georgia, Kazakhstan, Kyrgyzstan, Latvia, Lithuania, Moldova, Russian Federation, and Ukraine. Country level panel data covering the period of 1990-2009 was used for the research.
In order to investigate the validity of the hypotheses, an econometric model of total crime as well as property and violent crime was estimated by the First Difference model. Mundlak (1978) test is used in order to select between RE and FE estimation models. Drukker (2003) test is conducted in order to test a serial correlation.
The structure of the paper is following: section 2 reviews a related literature, section 3 analyses the data used for the research, section 4 discusses the research methodology, section 5 reports the estimation results and discusses them, and the final part of the paper concludes.
2 Literature Review
2.1 The Economic Theory of Crime
10 | P a g e
activities theoretical as well as empirical research have been carried out by many scholars in the field of economics. The following section will review theoretical as well as empirical literature on the relationship of crime and socio economic variables, such as probability of being arrested and punished, wage, income inequality and unemployment.
The economics of crime began with Becker’s (1968) formal theory of crime, who stated that participating in criminal activity is determined “on the basis of a maximization problem in which agents have to compare costs and benefits of legal and illegal activities taking into account the probability of being arrested and punished and the expected returns from crime” (Becker, 1968). Based on this theory commitment of crime is associated with the “economic choice of rational agents” (Becker, (1968)). Becker developed a theoretical model, in order to answer the question, what is the optimal rate of offenses to be permitted and how many offenders should be left unpunished.
The paper develops the idea that an increase in a person's probability of being caught and the nature of the punishment generally decrease the number of crimes an individual commits. And the effect of change of probability of catching is higher than the change in the punishment on change of crime rate. Based on this assumption, Becker (1968) develops the following economic model:
(1)
where the number of offenses is the function of the probability that an individual, who commits crime will be caught , an offender will be punished or fined if caught and all other factors ), such as the income in legal and other illegal sectors, the frequency of arrests,
11 | P a g e
and thus reduces the crime rate. Becker (1968) argues that as fine is costless, it should be set at its highest value and the probability of detection and conviction should be used to match the fine in deterring individuals. In literature this result is known as “high-fine-low-probability result”.
Crime in this model includes not only murder, robbery and assault, but also tax evasion, “white color crimes” and other kind of violations of law (Becker, 1968). Becker’s model, where the decision of participating in a criminal activity is made based on the cost-benefit analyses of expected outcome, became the basis of the economics of crime.
Becker’s paper became the basis of researches aiming to study economic factors that determine individuals’ choice to participate in different kind of criminal activities. Later Ehrlich (1973) further develops the economic model of crime by studying the effect of the labor market, average wageand income inequality, which offenders face, on the crime rate. The author states that the criminals have a choice between legal and illegal sectors and they have to allocate resources optimally between these two opportunities, arguing that “for a given median income, income inequality can be an indicator of the differential between the payoffs of legal and illegal activities” (Ehrlich (1973)). Thus, while Becker was mostly concentrating on the effect of efficiency of law system on the crime rate, Ehrlich is mostly concentrating on the effect of labor market outcomes on the crime rate.
2.2 The Empirical Studies
12 | P a g e
opportunities has significant effect in determining participation in a criminal activity. It appears that crime is positively related to both the average income and the percentage of families that are below this level. Property crime is also positively related to income inequality. However, these results are not statistically significant and the estimated effects of unemployment on crime are not stable across regressions (Ehrlich, 1973). Unemployment has less important role in determining crime rate compared to income level and its distribution.
In the early literature there is a gap between economic theory of crime and empirical research, as some studies do not find significant relationship between labour market and crime rate, ordo not show the same results as economic theory of study was predicting (Freeman (1983), Chiricos (1987)). The results in more recent empirical researches, where the researchers include economic, sociological, regional control variables, appears to be more similar to theory. Papps and Winkelman (2002) study the effect of unemployment and average income, on the crime rate in sixteen regions of New Zeland. They use panel data for the period between 1984 and 1996, which give the opportunity to control specific effects characterizing to this area. The estimation results show that, unemployment has significant effect on certain type of crime, but its effect on overall crime, without distinguishing separate types of crime, appears to be insignificant.
13 | P a g e
crime. Chiu and Madden (1998) studies the effect of income distribution on the crime rate. Based on the estimation results the authors conclude that higher the income inequality higher the rate of participation in the illegal activities, especially in property crime.
Gould, Weinberg, and Mustar (2002) study the effect of labor market opportunities on crime rate for unskilled individuals in the USA. The effect of both wage and unemployment is observed on the variety of violent and property crime, using an instrumental variable approach. The state level panel data set used for the estimation covers the period between 1979 -1997. Three kinds of analyses are carried out. In the first case panel regression is run using country and time fixed effects based on annual country level data in order to explain a country level crime rate for unskilled men. The second analyses explains ten years change in country crime rate by ten years change of unemployment rate for unskilled men measured at metropolitan level. The third analyses uses an individual level data in order to explain an individual criminal activity, controlling individual’s characteristics such as education, family background, ability, by the labor market opportunity. The estimation shows that the relationship between labor market opportunity and crime rate for this sample is significant. The decrease in wages explains more than 50% of variation of the property and violent crime rates. The analyses shows that the effect of wage is also significant and plays an important role in explaining change in crime rate in this sample.
14 | P a g e
static as well as dynamic econometric models are used. The estimation results confirm Becker-Ehrlich hypothesis regarding the property crime, that labor market outcome effects more on property crime rather than violent crime. The relationship between income inequality and crime is positive, but the effect of these independent variables on dependent variable is higher in case of property crime rather than in case of violent crime. The effect of overall unemployment is ambiguous, but being young and unemployed increases the probability of committing crime. Besides economic factors, demographic factors also play an important role in explaining crime rate in case of Germany, as more urbanized the region higher is the rate of crime in the region. Raphael and Winter-Ember (2001), using the US panel data for the period of 1970 – 1993, observe that unemployment has significant effect not only on property crime, but also on violent crime. The estimation results show that 2% decrease of unemployment predicts the 13 % decrease of theft and 14 % decrease of rape and robbery crime.
15 | P a g e
9% and its affect on the violent crime is also negative and equals 12%. In the long run the corresponding affect is higher. Based on the empirical results the authors conclude that criminal justice system plays an important role in decreasing the crime rate in the mentioned country. Similar results are shown for the case of USA (Grogger, 1991). Increase in the probability of arrest and the time spent in prison has negative effect on the probability of participation in the criminal activity, but the size of effect differs, the change in the probability of arrest tends to be more effective than the change in the duration of imprisonment.
3 Data
3.1 data description
In order to study the determinants of the crime in the Post Soviet Union and Eastern European countries, the country level panel data set is constructed for the period between 1990- 2009. The dataset includes the variables of the following 21 countries: Albania, Bulgaria, Croatia, Czech Republic, Hungary, Poland, Romania, Slovakia, Slovenia, Armenia, Azerbaijan, Belarus, Estonia, Georgia, Kazakhstan, Kyrgyzstan, Latvia, Lithuania, Moldova, Russian Federation and Ukraine.
16 | P a g e
mentioned factors explain why it is interesting to make observation on those countries in this time period.
Table 1 describes the variables, which are used in research and show also the source of the information:
Table 1: A data description
Variable Names Data description Sources
Crime Total reported number of crimes per 100000 inhabitants in each sampled countries;
United Nations Office on Drugs and Crime Property Sum of the registered number of committed
robbery, theft, burglary per 100 000 inhabitants in each country;
United Nations Office on Drugs and Crime
Violent Sum of the registered number of committed homicide, assault and rape per 100 000 inhabitants
in each country;
United Nations Office on Drugs and Crime
Unem Annual registered level of unemployment per 100000 inhabitants from labor force in each
sampled countries;
The National Statistics offices
Prob Ratio of number of the offenders to the total number of registered crime;
The UN Office on Drugs and Crime; UNICEFE TransMONEE database Income Average monthly income level in real terms. The
average income level of each country was converted into US dollars using corresponding years exchange rate;
17 | P a g e
Educ Number of students per 100 000 inhabitants in each country;
UNESCO Institute for Statistics (UIS) Gender Number of female population per 100 000
inhabitants in the corresponding country;
The world bank
Pop Number of 18-59 years old individuals per 100 000 inhabitants in each country;
The World Bank
Information about each category of crime rate in each country was collected from United Nations Office on Drugs and Crime and shows the number of registered crimes in each country during each year. Total crime includes all registered crime types, but some specific crime types like trafficking, illegal crossing of borders, taking bribes and drug or car accidents related crimes are not included separately in the data as they are not considered to be violent or property type crimes. The panel data was unbalanced as some data was missing for several countries and in order to avoid the missing data problems some missing values were imputed by linear interpolation. The maximum time gap when linear interpretation was used is three years. Fortunately all data values were available for explanatory variables. Besides, in order for the crime rate to be comparable for all sample countries, registered crime numbers for whole population was replaced by registered crimes per 100 000 inhabitants for each country.
18 | P a g e
Table 2 presents the descriptive statistics of the variables, with mean, standard deviation, minimum and maximum values:
Table 2: Descriptive statistics of the variables
Descriptive statistics
Variable Name Mean Standard deviation Minimum Maximum
Total Crime 1728.0 1217.6 138.9 5850.3 Property Crime 1027.45 758.68 21.98 3706.49 Violent Crime 56.62 48.72 5.22 256.1 Unemployment 4333.1 2206.5 40.5 11474.1 Probability of Arrest 0.33 0.17 0.03 1.14 Income 353.66 349.24 6.06 1966.67 Education 3158.53 1419.35 687.09 6675.39 Gender 52110.79 1143.37 48713.92 54099.78 Population 18-59 57408.48 2917.00 48257.02 64252.75
Table 2 reports descriptive statistics for the independent and explanatory variables of each regression. The mean value of total property crime is bigger than mean value of total violent crime. Probability of arrest varies over countries a lot, with having the lowest probabilities. As can be seen from the table, unemployment level also significantly varies over countries.
19 | P a g e
time. One can see that Estonia, Hungary and Czech Republic have highest total property crime rates. But if one looks on the homicide numbers, will see that mean value for Russian Federation is at least twice larger than that for almost all other countries. For other types of violent crime, Czech Republic shows highest assault rates and interestingly, data shows that in Estonia mean value of registered rape was enormously higher than for other countries in the sample.
Table 3: Crime Distribution over Countries
Country Property Crime Robbery Theft Burglary Violent Crime Homicide Assault Rape
Albania 73.2 7.6 49.1 16.5 21.8 9.9 9.9 2.0 Armenia 121.4 7.4 97.7 16.3 12.8 3.6 8.0 1.2 Azerbaijan 51.2 3.3 41.5 6.4 11.1 4.0 5.9 1.2 Belarus 994.4 64.1 703.4 226.9 42.5 9.5 26.4 6.6 Bulgaria 1360.1 55.1 877.2 427.8 50.8 6.0 33.9 10.9 Croat ia 942.5 18.8 606.4 317.3 44.1 7.0 30.6 6.4 Czech Republic 1987.0 46.3 1366.9 573.9 200.6 2.4 184.9 13.4 Est onia 2670.6 180.6 1483.7 1006.3 87.0 13.6 25.5 47.9 Georgia 270.3 16.5 192.8 61.0 20.0 7.8 10.2 2.0 Hungary 2210.4 30.4 1589.8 590.3 119.5 3.7 109.4 6.4 Kazakhst an 492.5 63.1 396.5 32.9 48.1 13.6 24.4 10.1 Kyrgyzstan 392.6 33.7 298.4 60.4 28.8 9.5 14.9 4.4 Latvia 1363.2 90.4 1084.9 187.8 48.4 9.0 34.5 4.9 Lithuania 1355.7 91.5 1053.4 210.7 36.5 10.1 21.2 5.2 Moldova 543.3 50.4 385.3 107.6 49.2 8.9 32.2 8.0 Poland 1312.5 84.9 569.8 657.7 71.2 2.6 62.0 6.6 Romania 478.2 15.6 360.5 102.1 42.3 2.7 33.7 5.9 Russia 1195.4 126.8 902.1 166.4 54.8 18.7 26.7 9.5 Slovakia 1156.2 26.7 561.2 568.2 73.6 2.2 67.6 3.8 Slovenia 1530.2 19.3 1096.3 414.6 118.1 3.7 104.6 9.8 Ukraine 541.2 61.4 373.9 105.8 24.5 7.9 12.8 3.8
Note:All crime rates are defined by reported incidents per 100 000 residents; Numbers are averaged over the sample period; Property crime is sum of Robbery, Theft and Burglary;Violent Crime is sum of Homicide, Assault and rape
20 | P a g e
statistics for almost all types of crime. Estonia and Hungary dominate other countries in registered rates of theft, and Estonia has highest burglary rates as well. Table shows that Estonia reports highest rates in this sample of countries in almost all types of crime. When one looks at these numbers, some caveats should be kept in mind. First, one should remember that these are registered crime numbers and not the numbers of all committed crime.
4 Methodology
21 | P a g e
variables in the model is that, demographic composition of the country may be correlated with the main explanatory variables of the model, such as unemployment and average income level, thus, if those variables are excluded from the model, the results may suffer omitted variable bias.
As it is discussed in the data section, the data on the different types of crimes, such as homicide, assault, rape, burglary, robbery, theft is available. These different types of crime are grouped into two types of crime: property crime and violent crime. A property crime is sum of the number of theft, burglary and robbery crime values. A violent crime is the sum of the number of rape, homicide and assault crime.
Hypothesis 4 stated that that an unemployment level and a probability of arrest have different impact on crime rate in the post Soviet Union countries than in the Eastern European countries. For this reason the dummy variable USSR, which gets the value one for all post Soviet Union countries during the transition five year period from 1990 till 1994 was created.
In order to explain the determinants of the crime rate in the mentioned countries, the following equation is built:
= ) + ) + + + + (2)
Where crime is the function of unemployment level, probability of arrest, education, average income level, gender and age distribution of the country at time denotes the time effect which is same in all countries but differs in time and during estimation it can be controlled by adding time dummies. The unobserved country specific effect constant over time is captured by
. is unobserved error term for country at time . The expectation of the error term, which is
22 | P a g e
effect, is zero and this assumption of the strict exogeneity is given by the following equation :
(3)
The variables are transformed into log variables, by using the logarithms. Presenting the variables in logarithms gives the opportunity to interpret the effect of independent variables on the dependent variable in terms of percentage changes, which appears to be more convenient for the explanation. Thus, it includes multiple observations for each country over time. In order to control the country specific unobserved effects ( ), first of all the models with random effects (RE) and fixed (FE) effects are analyzed. In the RE model, the country specific effect is a random variable that is assumed not to be correlated with the explanatory variables of the model, while in the FE model the country specific effect is allowed to be correlated with the explanatory variables. The Mundlak type specification test is used in order to select among those two models.
One more issue, which needs to be addressed during the panel data estimation, is the problem of serial correlation. The presence of a high serial correlation in the error term may bias the estimates of standard error. In order to avoid mentioned problem, the serial correlation was tested after the running the FE regression. This test is based on Drukker(2003), who shows that the test has good size and power properties in reasonably sized samples.
Thus, after transforming the variables into logs and including fixed effects, the following equation is estimated:
Δ = Δ + Δ Δ Δ Δ + Δ + Δ +Δ (4)
23 | P a g e
unemployment level, probability of arrest, education level, gender, population and age distribution between time and in the country . . . Similar model is used for studying disaggregate crime, where in case of property crime, the dependent variable will be Δ and in case of violent crime the dependent variable will be Δ .
To sum up, this section explained the methodology which is used to study the determinants of the overall crime rate as well as disaggregated crime rates in the post Soviet Union and the Eastern European countries during the period of 1990 – 2009. The model includes fixed effects in order to catch the possibility of different effects of explanatory variables on the crime rate in case of Soviet Union member and non Soviet Union member countries.
5 Results and Discussion
The following section reports and analyses of the estimation of the equation and equation 3 for the case of overall crime as well as property and violent crime separately. The section is divided into two parts. The sub section 5.1 reports the estimation results of equation 4, and the sub section 5.2 reports the estimation result of the same equation , which includes the USSR dummy and analyses the difference between the Soviet Union member and non member countries.
In order to compare the Fixed effect model to the Random Effect model under the null hypothesis that the country specific effects are not correlated with the other explanatory
24 | P a g e
that if there is a serial correlation in the error term the results of the test can be misleading. In order to avoid this problem instead of Housman test Mundlak (1978) test will be conducted. Mundlak test allows possible correlation between the explanatory variables and the country
specific individual effects. The Mundlak (1978) estimation method for total crime gives low value which is close to zero. Thus, the results of the Mundlak (1978) test for the
total crime denies possibility of using RE model. Thus, the FE model is preferred in this case. Serial correlation is checked by conducting Drukker (2003) test. The test statistics of the Drukker test are high and .These results indicates a strong serial correlation in the error term. Thus the null hypothesis about no serial correlation should be rejected. There is an evidence of the disturbances following the AR (1) process. Because of this, it is more appropriate to use the first difference (FD) model. The estimated model shows less serial correlation, the fixed country specific effects will be cancelled out. In the first difference (FD) model will be used cluster-panel robust standard errors to mop up the serial correlation, moreover time-constant unobserved heterogeneity is not a problem anymore.
5.1 Determinants of crime using the econometric specification
25 | P a g e
Table 4: Crime
Variables D.Ln crime D.Lnproperty D.Ln Violent
D.Lnunem 0.100** (0.035) 0.096*** (0.031) 0.068 (0.064) D.Lnprob -0.451*** (0.062) -0.426*** (0.082) -0.223** (0.106) D.Lneduc 0.144* (0.083) 0.261 (0.233) -0.032 (0.218) D.Lnincome 0.026 (0.027) -0.063 (0.044) 0.059 (0.078) D.Lngender -5.821 (6.656) 1.952 (11.190) -21.910 (16.420) D.Lnpop 0.014 (1.174) 0.550 (1.657) 0.345 (2.060) Observations Rsq 297 0.404 272 0.221 257 0.109
note: *, **, *** indicate the significance at 10%, 5%, 1%; Panel robust standard errors
in parentheses; Year dummies are included; regressions are without the constant.
26 | P a g e
statistically significant: the coefficient of education is positive and significant at 10% significance level only in case of total crime.
As it is discussed in the previous sections, the aim of this research is to study the effect of unemployment and a probability of arrest on different type of crime. The results report that 1% increase in the percentage change (compared to previous year) in the probability of arrest decreases the percentage change in total crime by 0.45% and accordingly this number is 0.43 for the total property crime and 0.22 for total violent crime. These results support the hypotheses stated in the introduction of this paper. It seems that all individuals before committing some kind of property crime analyze expected consequences and plan their future actions rationally. As they seem to be afraid of going to the jail the number of individuals, who commit the crime is less when the police works efficiently.
27 | P a g e
As it was assumed based on the theory the size of effects of explanatory variables are expected to be higher in case of property crime than in case of violent crime. The size of the estimated coefficients and their level of significance mainly support the expectations stated in the previous sections of the paper and support the idea that the determinants of the crime in terms of the economic incentives cannot be applied to the cases of violent crime. Level of unemployment is not a statistically significant determinant of the total violent crime. This result seems logical in light of the theory; if an unemployed individual wants to get some economic benefit and is ready to commit a crime, will most prob ably commit a property crime rather than homicide or assault.
5.2 USSR/Transition Countries/Periods
As it is stated in the previous sections, the explanatory variables such as unemployment level and probability of arrest may have different effects on crime rate in case of Soviet Union member countries and in case of non Soviet Union member countries. Thus, to study this difference, the interaction terms of USSR dummy with unemployment level and interaction term of USSR dummy with probability of arrest is added to the first equation and the equation 4 is estimated.
Table 5: The effects of unemployment level and probability of arrest on crime in USSR member and non USSR member countries.
Variables D.Ln crime D.Lnproperty D.Ln violent
28 | P a g e D.Lneduc 0.122 (0.087) 0.246 (0.243) -0.041 (0.221) D.Lnincome 0.020 (0.030) -0.059 (0.057) 0.051 (0.080) D.Lngender D.Lnpop -2.133 (6.370) -0.092 (1.175) 3.126 (11.370) 0.470 (1.724) -19.970 (18.190) 0.351 (2.090) USSR*D.lnunemp 0.183* (0.091) 0.410*** (0.046) 0.255 (0.152) USSR*Dlnarrest -0.430*** (0.136) -0.703** (0.293) -0.355 (0.295) Observations Rsq 297 0.409 272 0.225 257 0.109
note: *, **, *** indicate the significance at 10%, 5%, 1%; Panel robust standard errors in parentheses; Year dummies are included; regressions are without the constant.
29 | P a g e
prospects at that time and might have felt more pressured because of general hardship or instability in the country and that could have led some to commit more illegal activities.
30 | P a g e
more cautious and risk averse and their mentality is different, this argumentation helps to understand why in the Soviet Union countries probability of arrest has much stronger influence on crime.
Conclusion
This research studies the determinants of crime in the Eastern European and post Soviet Union countries, concentrating on the effect of probability of arrest, unemployment and income on the crime rate in the mentioned countries. The sample of the study covers the following 21 countries: Albania, Bulgaria, Croatia, Czech Republic, Hungary, Poland, Romania, Slovakia, Slovenia, Armenia, Azerbaijan, Belarus, Estonia, Georgia, Kazakhstan, Kyrgyzstan, Latvia, Lithuania, Moldova, Russian Federation, and Ukraine. The data is country level unbalanced panel and covers the period of 1990-2009.
The paper addresses the following research question: how do the probability arrest, unemployment level and income affect the crime in the Easter European and the post Soviet Union countries. The hypothesis presented in the beginning of this paper states that the probability of arrest has negative effect on the crime, unemployment has positive effect on the crime, but the size of the effect is higher in case of property crime than in case of violent crime and that the size of the effect differs between the Soviet Union member and non member countries.
31 | P a g e
significant effect on the crime rate. The effect size is comparatively higher in case of property crime than in case of other types of crime. Thus, it can be concluded that the efficiency of police and criminal justice system in the country can contribute much to decrease of crime rate. It can be concluded that from the policy point of view, investing in the restructuring and advancing police system in the countries, which face high crime rate should be a rational decision.
Another determinant of crime, which was studied in this paper, is the unemployment level in the country. The coefficient of unemployment appears to be significant and positive. This means that a higher unemployment level is associated with the higher crime rate in the country. As the findings show, the effect size of unemployment on crime is smaller than the effect size of probability of arrest.
The research distinguished the effect of mentioned explanatory variables on the violent crime and on the property crime. The results report that the effects of the economic factors are higher in case of property crime compared to the violent crime.
The First Difference model of crime is also estimated in order to study the difference of the effect of explanatory variables in case of the Soviet Union member countries and non member countries. The estimation of the model shows that probability of arrest and also unemployment level have higher effect on the crime rate in a sample of the Soviet Union countries compared to the countries which are not Soviet Union members. Increasing the risk of arrest reduces property crime significantly while the effect seems to be weaker in case of violent crime.
32 | P a g e
case of violent crime. The probability of arrest has much stronger influence on crime rate in the Soviet Union member countries than in the countries which are not Soviet Union members.
Doing deeper analyses was not possible due to the data limitation. For the future research adding the duration of an average sentencing for each type of crime can be considered. Also finding the proxy for the benefit from crime will yield interesting results.
Appendix
a)Limittations
33 | P a g e
Table 6: Missing values
Country Total Crime Robbery Theft Burglary Homicide Assault Rap e
Albania 1990-1991 1990-1994 1990-1994 No No 1990-1994 1990-1994 Armenia No No 1990-1994 1990-1994 No No No Czech Republic No 1990-1994 No No No 1990-1994 1990-1994 Est onia No 1990-1994 No 1990-1994 No No No Georgia No No No 1990-1997 No No No Kazakhst an No No No No No 2001-2009 2001-2009 Latvia No No No No No 2003-2009 2003-2009 Lithuania No No No No No 2003-2009 2003-2009 Romania No No No 2003-2009 No No No Slovakia No No No 2003-2009 No 1990-1997 2003-2009 Ukraine No No 1991-1994 No No No No
Note: "No" means no missing values for that specific type of crime; one can see only those countries for which I lack information for at least one type of crime;
b) Probability of arrest and unemployment rate
34 | P a g e
Data on unemployment rate are collected from each countries statistical departments and using number of total labor force and population (from World Bank) unemployment rate is converted to number of unemployed individuals in per 100 000 individuals from labor force.
c) Crime Definitions
All crime types which are dependent variables in this study show number of registered crimes per 100,000 people. The data for all crime types are collected from United Nations Office on Drugs and Crime.
Homicide is a legal category to define lethal interpersonal violence; that is, unlawful death purposefully inflicted on a person by another person.
Assault is defined as a physical attack against the body of another person resulting in serious bodily injury
Rape was defined as sexual intercourse without valid consent.
Robbery is defined as the theft of property by force or threat of force.
Burglary was described to mean “to gain access to a closed part of a building or other premises by use of force with the intent to steal goods”.
35 | P a g e
References
Altindag, Duha Tore. "Crime and Unemployment: Evidence from Europe." Department of Economics, Louisiana State University (2009).
Athens, Lonnie H.. The creation of dangerous violent criminals: University of Illinois Press, 1992.
Becker, Gary S. "Crime and punishment." Journal of Political Economy 76(2). (1968): 169-217. Blau, Judith.R, and Peter.M Blau. "the Cost of Inequality: Metropolitan Structure and Violent
Crime." American Sociological Review 47 (1982): 114-129.
Buonanno, Paolo . "The Socioeconomic Determinants of Crime. A Review of the Literature." University of Milano-Bicocca, Department of Economics (2003).
Chiricos, Theodore G. "Rates of Crime and Unemployment: An Analysis of Aggregate Research Evidence ." Social Problems 34, no. 2 (1987): 187-212.
Chiu, W. Henry, and Paul Madden. "Burglary and income inequality." Journal of Public Economics 69 (1998): 123-141.
Corman, Hope, and Naci Mocan. "A Time-Series Analysis of Crime, Deterrence, and Drug Abuse in New York City." The American Economic Review 90, no. 3 (2000): 584-604. Cornwell, Christopher , and William N. Trumbull. "Estimating the Economic Model of Crime
with Panel Data." The Review of Economics and Statistics 76, no. 2 (1994): 360-366. Doyle, Joanne M. , Ehsan Ahmed, and Robert N. Horn. "The Effects of Labor Markets and
36 | P a g e
Drukker, David M. . "Testing for serial correlation in linear panel-data models." The Stata Journal 3, no. 2 (2003): 168-177.
Edmark, karin. "Unemployment and Crime: Is There a Connection?." The Scandinavian Journal of Economics 107 (2005): 353-373.
Ehrlich, Isaac. "Participation in Illegitimate Activities: A Theoretical and Empirical Investigation.” Journal of Political Economy 81, no. 3 (1973): 521-565.
Entorf, Horst , and Hannes Spengler. "Socioeconomic and demographic factors of crime in Germany Evidence from panel data of the German states." International Review of Law and Economics 20 (2000): 75-106.
Fowles, Richard , and Mary Merva. "Wage Inequality and Criminal Activity: An Extreme Bounds Analysis for the United States, 1975 to 1990." Criminology 34, no. 2 (1996): 163-182.
Fajnzylber, Pablo , Daniel Lederman, and Norman Loayza. "What causes violent crime?." European Economic Review 46(7) (2002): 1323-1357.
Fougre, Denis , Francis Kramarz, and Julien Pouget . "Youth unemployment and crime in France." Journal of the European Economic Association 7(5) (2009): 909-938.
Freeman, R. B . "Crime and Unemployment." Crime and Public Policy, Institute for Contemporary Studies Press, San Francisco, CA (1983).
Gould, Eric D. , Bruce A. Weinberg, and David B. Mustard. "Crime rates and local labor market opportunities in the United States: 1979-1997." The Review of Economics and Statistics 84, no. 1 (2002): 45-61.
37 | P a g e
Grogger, Jeffrey T. . "Market Wages and Youth Crime." Journal of Labor Economics 16, no. 4 (1998): 756-91.
Hall, Jerome. Theft, law, and society. 2d ed. Indianapolis: Bobbs-Merrill, 1952.
Kelaher , Richard , and Vasilis Kelaher . "Crime and Punishment Revisited." University of Sydney, Faculty of Economics and Business (2011).
Lee , Kangoh . "Unemployment and Crime." Department of Economics San Diego State University (2009).
Levitt, Steven . "The Effect of Prison Population Size on Crime Rates: Evidence from Prison Overcrowding Litigation." Quarterly Journal of Economics 111, no. 2 (1996): 319-351. Lochner, Lance , and Enrico Moretti. "The Effect of Education on Crime: Evidence from Prison
Inmates, Arrests, and Self-Reports ." American Economic Review 94, no. 1 (2004): 155-189.
Merton, Robert. "Social Structure and Anomie." American Sociological Review 3, no. 5 (1938): 672-682.
Mover, M . "Will the Recession Spark a Crime Wave."
http://www.scienti_camerican.com/article.cfm?id=do-recessions-increase-crime (2009). Mundlak, Yair. "On the pooling of time series and cross section data." Econometrica 46 (1978):
69-85.
Nilsson, Anna, and Jonas Agell. "Crime, unemployment and labor market programs in turbulent times." Institute for Evaluation of Labour Market and Education Policy (2003).
38 | P a g e
Raphael, Steven, and Rudolf Winter-Ember. "Identifying the Effect of Unemployment on Crime." Journal of Law and Economics 44 (2001): 259-283.
Sampson, Robert J. , and Jacqueline Cohen. "Deterrent Effects of the Police on Crime: A Replication and Theoretical Extension." Law and Society Review 22, no. 1 (1988): 163-190.
Walmsley, Roy . "Global incarceration and prison trends." Forum on Crime and Society 3, no. 1 (2003): 65-78.
Wan, Wai-Yin , Steve Moffatt , Craig Jones , and Don Weatherburn. "The effect of arrest and imprisonment on crime." Crime & Justice 158, no. 1 (2012): 1-19.
