- Email: [email protected]
Crime, deterrence and unemployment in Greece: A panel data approach
The Social Science Journal 49 (2012) 167–174
Contents lists available at ScienceDirect
The Social Science Journal journal homepage: www.elsevier.com/locate/soscij
Crime, deterrence and unemployment in Greece: A panel data approach George Saridakis a,∗ , Hannes Spengler b a b
Loughborough University, School of Business and Economics, Ashby Road, LE11 3TU, Loughborough, Leicestershire, UK University of Applied Sciences Mainz, School of Business, Lucy-Hillebrand-Strasse 2, 55128 Mainz, Germany
a r t i c l e
i n f o
Article history: Received 5 August 2009 Received in revised form 9 August 2011 Accepted 10 August 2011 Available online 19 April 2012 Keywords: Crime Deterrence Unemployment Panel data GMM
a b s t r a c t This study empirically examines the relationship among crime, deterrence and unemployment in Greece. A regional dataset over the period 1991–1998 was collected and analysed. Our econometric methodology follows the Generalized Method of Moments (GMM) estimator applied to dynamic models of panel data. The results show that property crimes are significantly deterred by higher clear-up rates and that unemployment increases crime. For violent crimes, however, the effect of the clear-up rate and unemployment are found to be generally insignificant. Finally, our results may provide support to policy makers in forecasting criminal activity in the current economic downturn under a wave of harsh austerity measures, budget cuts and increased unemployment. © 2011 Western Social Science Association. Published by Elsevier Inc. All rights reserved.
1. Introduction Becker (1968) marked the beginning of attempts to apply economic models of rational decision-making to crime. Since then, a number of significant theoretical developments have been made and substantial empirical work has been published in academic journals on both economics and criminology. Most of the applied work, however, focuses on US crime statistics (prominent examples being the work by Ehrlich, 1973; Witte, 1980; Marvell & Moody, 1994; Levitt, 1996) and it is only recently that economics-of-crime models have been increasingly applied outside the USA and used to derive country-specific policy to reduce the rate of criminal behaviour. Examples of such work in Europe, for instance, can be found in Fougère, Kramarz, and Pouget (2006) for France, Buonanno and Montolio (2008) for Spain, Entorf and Spengler (2000)
∗ Corresponding author.. E-mail addresses: [email protected] (G. Saridakis), [email protected] (H. Spengler).
for Germany, Hale (1998) and Saridakis (2011) for Britain and Marselli and Vannini (1997) for Italy. In this context, this paper adds further evidence for Europe by examining the effect of two central variables in economics-of-crime research, these being the clear-up rate (as a proxy for the probability of getting caught) and the unemployment rate (as an inverse measure of legal income opportunities), on crime in Greece from 1991 to 1998. Indeed, this period was marked by a number of economic austerity and economic restructuring measures undertaken by the policy makers to meet the criteria for participation in the European Monetary Union (EMU) and which led to a significant increase in the country’s recorded unemployment from 7% to almost 11% (Papadopoulos, 2006) combined with a sharp increase in crime, especially in property crime (Table 1).1 In this paper we
1 Of course, there were other events, such as the significant increase in illegal immigration flows particularly from Balkan counties, and also from Middle-East and North Africa, which may be associated particularly with organised crime (e.g. drug related crimes increased 220% during the study period). Unfortunately, we do not have disaggregated data on
0362-3319/$ – see front matter © 2011 Western Social Science Association. Published by Elsevier Inc. All rights reserved. doi:10.1016/j.soscij.2011.08.005
168
G. Saridakis, H. Spengler / The Social Science Journal 49 (2012) 167–174
Table 1 Crime, deterrence and unemployment in Greece, 1991–1998. % Change in Recorded crime Property crime Breaking and entering Theft of motor cars Robbery Violent crime Murder Serious assault Rape Unemployment rate
Average Clear-up rate
Recorded crime
Clear-up rate
37 95 83
−33 −40 −29
365.1 116.5 15.6
13.4 12.8 31.1
50 0 0 57
−16 −4 12
2.7 67.3 2.4 8.9
74.9 95.5 69.5
Note: Recorded crime is per 100,000 inhabitants. Clear-up rate and unemployment are in %.
consider six major crime categories: breaking and entering, theft of motor cars, robbery, murder, serious assault and rape (the first three are property crimes, while the latter three are violent crimes2 ) and empirically examine these types of crime using Generalized Methods of Moments (GMM) to determine whether increasing unemployment and decreasing clear-up rates contributed to the increase in crime in the 1990s. Furthermore, focusing on unemployment and deterrence, our analysis may provide some useful information about the consequences of the recent Greek sovereign debt crisis on crime rates, since the extensive and unprecedented package of austerity measures are likely to increase, at least in the short and medium-run, unemployment (due to weak domestic demand) and reduce clear-up rates (perhaps due to budget cuts in law enforcement). Finally, our study contributes to the current debate regarding the relevance of deterrence and economic variables in explaining violent behaviour (Entorf & Spengler, 2000; Raphael & WinterEbmer, 2001; Saridakis, 2004), associations which for the most part are already well established for crimes against property. The outline of the paper is as follows. Section 2 provides an overview of the literature from which our hypotheses are derived. Section 3 describes the data. Section 4 presents the empirical framework. Section 5 is devoted to the presentation and discussion of the results. The last section concludes. 2. Background and hypothesis derivation There are a number of theories that may explain the causes of criminality from a sociological paradigm. Inter alia, there is Merton’s strain theory (Merton, 1938) which suggests that crime rates will tend to be higher
immigration and furthermore the size of illegal immigration is difficult to measure with accuracy. The 2001 census, however, conducted by the National Statistical Service of Greece suggests that immigrants represent around 7% of the total population, but it may rise up to 10% if full account of the illegal immigrants is taken into consideration (see Baldwin-Edwards, 2004). Also, unemployment rates for foreign-born individuals tend to be higher than for the native-born individuals (e.g. according to OECD statistics the unemployment rate for native-born and foreign-born males in 1995 were 6.1% and 14%, respectively). 2 Robbery is included in property crimes since it might be dominated by the desire to steal another person’s property (see Saridakis, 2004).
in societies where opportunities are most unequal. Alternatively, the social disorganization theory suggested by Shaw and McKay (1942) explains crime and delinquency rates by examining factors (low economic status, ethnic heterogeneity and residential mobility) that contribute to disruption of community social organizations. A third explanation is derived from Sutherland’s differential association/social learning theory (Sutherland, 1942) which mainly emphasizes that criminal behaviour is learned through interaction within the intimate family unit and close peers. Finally, the lifestyle/routine activity theory by Cohen and Felson (1979) represents the sociological interface of crime as opportunity. This theory suggests that crime will occur when there is a motivated person, a suitable target (potential loot or victim) and the absence of guardians capable of preventing the violation. The paper by Becker (1968) marked the beginning of attempts to apply economic models of rational decision making to crime. Becker suggested that individuals are rational utility maximisers who decide whether or not to engage in criminal activity by comparing the cost and benefits of crime. Becker’s paper, however, was not intended to present a new theory of criminal behaviour, but to analyse how to minimise the social loss of crime. To this end, Becker derived the supply of offence function which is “relating the number of offences by any person to his probability of conviction, to his punishment if convicted, and to other variables, such as the income available to him in legal and other illegal activities, the frequency of nuisance arrests, and his willingness to commit an illegal act” (Becker, 1968:177). Ehrlich (1973) extended the work of Becker by considering a time allocation model and for almost 30 years now in economics, the empirical work in crime has relied on the foundations of the Becker–Ehrlich model to inform statistical specifications. According to the basic economic crime model, individuals have to decide how to allocate their time (t) between legitimate (t − t2 ) and illegitimate (t2 ) activities. Based on the assumption that crime and legal employment are substitute activities it follows that the more time one devotes to a legitimate occupation, the less time one has to carry out criminal activity. The individual’s expected utility is a weighted average of his/her utility in the two alternative states of the world: (a) the individual is caught and punished and (b) the individual is not caught and punished.
G. Saridakis, H. Spengler / The Social Science Journal 49 (2012) 167–174
The expected utility is given by: EU = pU[W0 + W1 (t − t2 ) + W2 (t2 ) − F(t2 )] + (1 − p)U[(W0 + W1 (t − t2 ) + W2 (t2 )]
(1)
where p is the probability of being caught and punished, W0 is the individual’s (exogenous) wealth, W1 and W2 are the wealth from legitimate activity and crime, respectively. F is the monetary equivalent of the punishment. Crime occurs when the marginal expected returns from crime are greater than the marginal return in legitimate work,3 i.e. W2 (t2 ) − pF (t2 ) > W1 (t − t2 ). Thus, the basic crime theory4 predicts that sanctions and legitimate opportunities influence decisions to engage in crime. 2.1. Crime and deterrence Being caught and punished may deter offenders from committing a crime again, as well as deter future offenders who contemplate committing crimes. Ehrlich (1973) was the first economist who tested the crime deterrence model empirically using cross sectional data of US states from 1940, 1950 and 1960. Ehrlich’s econometric crossstate estimation showed a strong negative association between crime and criminal justice variables in 10 out of 14 crime categories. Interestingly, the absolute values of the estimated elasticities of violent crime with respect to deterrence variables were found not to be lower on average than those associated with crime against property. Hence, the implication appeared to be that law enforcement was not less effective in combating crimes of violence relative to crimes against property. However, a few years later, Wolpin (1978) using time-series data for the period 1894–1967 from England and Wales, found that the estimated coefficients of the probability variables devoted to deterrence were greater in magnitude and showed higher statistical significance for property crimes than crimes against the person. Since then, there has been a major research effort and also a major statistical development concerning potential simultaneity between deterrence variables and crime rates. The argument here is that while increases in the number of prisoners are likely to reduce crime, rising crime rates also translate into a larger prison population (resulting in a downward bias of the effect of deterrence on crime) or that a higher incidence of crime leads to lower clear-up rates due to the overload of the given and temporarily fixed police resources (resulting in an upward bias of the effect of deterrence on crime). For example, Levitt (1996) assessed the interaction problem between prisoner populations and US crime rates using state level data from the period 1971–1993. To account for the possible endogeneity
3
For further information and a formal proof see Pyle (1983). Since the Becker–Ehrlich model, significant theoretical and empirical developments have been made. For example, early economic studies applied a static framework to analyse criminal behaviour. However, there are several theoretical and empirical reasons, for example, through capital accumulation, peer group effects, improvement in fit when lagged dependent variables are included in the model, which suggest that criminal behaviour should be analysed in a dynamic framework (Witte and Witt, 2002). 4
169
between prison population size and crime rates, Levitt used the prison overcrowding litigation as an instrument for changes in prison population (assuming that overcrowding litigation influences prison population and has no direct effect on crime). Then, the effect of prison population on both property and violent crimes was estimated, but for the latter case the reported elasticities were found to be smaller in magnitude and to lack in statistical significance relative to property crimes. These earlier studies may provide evidence regarding the strong link between deterrence and property crime, but what can we expect to happen to violent crime rates when governments adopt costly deterrence-based policies by increasing their expenditures on prison, courts and law enforcement generally? Evidence from recent work is revealing. For example, Entorf and Spengler (2000) found weak effects of the clear-up rates on violent crime whereas they found strong support for the deterrence hypothesis for property crimes. Similarly, Cherry and List (2002) using county-level US panel data and considering a wide variety of criminal justice variables conclude that sanctions are less influential in deterring serious violent crimes, such as murder or rape. Saridakis (2004) using US time-series data found short-run effects of the prison population on murder and rape, but no long-run associations were established. Buonanno and Montolio (2008) using a short panel of Spanish provinces over the period 1993 and 1999 also failed to establish a significant association between clearance rate and crime against the person. We therefore hypothesise that: Hypothesis 1. Criminal justice variables are likely to deter property crimes. Hypothesis 2. Criminal justice variables are likely to have a weak or no association with violent crimes. 2.2. Crime and unemployment Nearly all economic (and non-economic) studies, consider unemployment as an important cause of crime. Unemployment reflects the (lack of) opportunity for participation in the legitimate job market and the acquisition of legal earnings (Freeman, 1999). The exclusion from legal income opportunities increases the expected returns from crime. The economic crime theory, as well as sociological theories of crime (Merton, 1938; Shaw & McKay, 1942), even though they may be based on different concepts and assumptions, suggest that the crime rates increase with increasing unemployment. However, those advocating opportunity theories suggest a negative association between unemployment and crime (Cohen & Felson, 1979). Although the data seem to support a positive link between unemployment and property crime the association between unemployment and violent crime is very diverse (Vold, Bernard, & Snipes, 2002). A substantial number of studies for the USA found that unemployment had a generally significant impact on property crime rates, but it had no significant impact on violent crime rates. Freeman and Rodgers (1999) found that young African American men in tight labour markets experienced a boost in employment and earnings during the 1990s.
170
G. Saridakis, H. Spengler / The Social Science Journal 49 (2012) 167–174
According to their analysis, a drop in US unemployment between 1992 and 1997 reduced youth (overall) crime. Levitt (2004), however, suggests that the decline in the US unemployment rate during the 1990s may explain the decline in property crime, but not changes in violent crime. Moreover, Saridakis (2004) used the duration of unemployment (mean of weeks unemployed) in his analysis of violent crime, but his results revealed no significant association between unemployment duration and violent crime. Consistent with this argument, studies for the UK suggest a marginal effect of unemployment on violent crime. For example, Carmichael and Ward (2001) using countylevel data for England and Wales suggest that neither adult nor youth unemployment had a significant effect on violent crime. Similarly, Elliot and Ellingworth (1996) examined the effect of unemployment on personal crime using the 1992 British Crime Survey (BCS) and found no significant relationship. Nevertheless, Field (1990) using time-series data for England and Wales from 1952 to 1987 failed to find any association between most types of crime and unemployment. Marris (2000) suggested that in order to understand the relationship between unemployment and crime, the distinction between involuntary and voluntary unemployment should be made. An insignificant association has also been reported between youth unemployment and violent crime in France (Fougère et al., 2006). Interestingly, some other studies for the United States (Greenberg, 2001; Raphael & WinterEbmer, 2001) and one for Germany (Entorf & Spengler, 2000) have reported a negative association between duration of unemployment or/and unemployment and serious violent crimes such as murder and/or rape. Weighing up the existing research on the unemployment–crime link leads us to our final two hypotheses: Hypothesis 3. Unemployment is likely to be positively associated with property crimes. Hypothesis 4. Unemployment is likely to have a weak or no effect on violent crimes. 3. Data The data set used in this paper is a panel of annual, regional-level (Nuts 2) observations,5 stretching from 1991 to 1998 (see Table 1). Reported crime data (per 100,000 inhabitants) and clear-up rates (%) were obtained directly from the collector of crime statistics in Greece and to our knowledge are not officially published in electronic or printed form.6 In contrast to cross-country studies/panels (Fajnzylber, Lederman, & Loayza, 2002), a regional panel
5 Greece is divided into 13 regional districts (population in 1000 is in parenthesis): Crete (553.1), South Aegean (263.6), North Aegean (186.8), Attica (3479.2), Peloponnese (650.7), Central Greece (637.0), West Greece (726.2), Ionian Islands (197.2), Epirus (360.2), Thessaly (740.6), West Macedonia (300.1), Central Macedonia (1759.3) and East Macedonia and Thrace (561.3). 6 The data were made available after an official data request from the German Federal Criminal Police Office (BKA) to Greek authorities to be analysed in a comparative European crime project by Entorf and Spengler (2002).
data set for one country overcomes or attenuates problems associated with differences in crime definitions, statistical data collection (usually there are no official differences in crime definitions and counting rules across different regions of the same country) and reporting propensities (as long as regional reporting propensities are constant over time differences between regions can be controlled by suitable estimation techniques, Section 4).7 Finally, data for the unemployment rate (%) were extracted from the Eurostat statistical database. 4. Empirical framework Based on the empirical deterrence literature, we can derive the following simple model: cit = ˇ0 + ˇ1 cit−1 + ˇ2 pit + ˇ3 uit + εit
(2)
εit = i + vit
(3)
where cit is the crime rate (per 100,000 inhabitants) by subcategory for administrative division i in year t. The lagged dependent variable reflects the tendency of individuals who are involved in criminal activity to continue with it even after the circumstances that led them to turn to crime have changed (Witt, Clarke, & Fielding, 1999). The measure pit is the clear-up rate, which is treated as endogenous. For example, this would be the case if a higher incidence of crime led to lower clear-up rates due to the overload of the given and temporarily fixed police resources. On the other hand, a feedback effect could be caused if the state responded to increasing crime rates by allocating more resources to the police, resulting in higher clear-up rates. The variable uit is the unemployment rate and is considered to be exogenous.8 All variables are measured in natural logarithms. Finally, i is the unobserved time-constant regionallevel effect which may be correlated with some of the regressors. For this reason, we estimate the following linear dynamic model: cit = ı1 cit−1 + ı2 pit + ı3 uit + vit
(4)
where the variables are first-differenced to eliminate time-invariant regional-level effects (i.e. regional heterogeneity). However, first-differencing introduces regressor–error correlation (endogeneity) since cit−1 = cit−1 − cit−2 is correlated with vit = vit − vit−1 because vit contains vit−1 which is correlated with cit−1 and also serial correlation in the error term (i.e. between vit = vit − vit−1 and vit−1 = vit−1 − vit−2 ). To deal with the variables cit−1 and pit , we have used the Generalized Methods of Moments (GMM) proposed by Arellano and Bond (1991). Baltagi (2008) provides a comprehensive overview of the Arellano and Bond estimator (pp. 149–155). Briefly,
7 Also, it surmounts statistical weaknesses of studies relying on national time-series crime data (see Levitt, 2001). 8 We have experimented by allowing unemployment to be endogenous in the model (i.e. crime participation may reduce the employability of previously convicted offenders and it may in turn contribute to observed unemployment – see Raphael & Winter-Ebmer, 2001). Generally, the results did not provide further insight into our analysis and, hence, are not reported here.
G. Saridakis, H. Spengler / The Social Science Journal 49 (2012) 167–174
the variables cit−1 and pit are instrumented with suitable lags of their own levels. First-differences of strictly exogenous covariates are also used to create moment conditions. u is modelled as strictly exogenous and contributes 1 instrument. The remaining instruments (42) come from the p-2 instruments available in periods p = 3–8. Furthermore, GMM takes account of the serial correlation in the same manner generalized least squares (GLS) takes care of non-independent and identically distributed (i.i.d.) errors. Since our model uses a double log functional form, the slope coefficients are elasticities. Thus, our estimates show the percentage change in the dependent variable (crime) if a specific explanatory variable changes by 1%, while the other explanatory variables are held constant. In our model, the estimated coefficients represent short-run effects (the immediate change of the dependent variable as a reaction to a change in an explanatory variable) but the long-run elasticities (i.e. the long-run change of the dependent variable across all periods as a reaction to a permanent change in an explanatory variable) can be easily obtained by dividing the estimated coefficients, excluding that of the lagged dependent variable, by one minus the coefficient of the lagged dependent variable (Buonanno and Montolio, 2008; Fajnzylber et al., 2002; Witt et al., 1999). 5. Empirical results Table 2 presents the one-step GMM results of the crime models for Greece. We first comment on the property crime results presented in columns 2–4. Clearly these columns show a strong negative effect of the clear-up rate on property offences (with the sole exception of motor-car theft) supporting Hypothesis 1. The insignificance of the short run effect in the regression for theft of motorcars might be a consequence of the typical car thief. Motor-car thieves can be considered as highly specialised and rewarded serial criminal offenders (presumably networked within groups of car racketeers), who have high opportunity costs for legal work and thus, being relatively immune against changes in deterrence, especially if the mean clear-up rate is low. The coefficient estimates for breaking and entering and robbery suggest long-run clear-up elasticities of −0.46 and −0.43, respectively.9 These estimates are similar to those reported for other European countries: see, for example, Buonanno and Montolio (2008) for Spain, Entorf and Spengler (2000) for Germany and Witt et al. (1999) for England and Wales. For the three individual property crimes, the estimated effects of unemployment are positive and significant. The GMM (short-run) estimates of the unemployment rate range from 0.236 to 0.866 suggesting long-run unemployment elasticities of 0.76, 0.81 and 1.45 for breaking and entering, theft of motor cars and robbery, respectively. In fact, panel data studies by Marselli and Vannini (1997) for Italy and Lin (2008) and Raphael and Winter-Ebmer (2001) for the USA also found significantly positive effects of unemployment on property crime rates. This result is reasonable and in line with the mainstream economic view of
9 The long-run clear-up elasticity for breaking and entering, for example, is calculated as follows: −0.274/(1 − 0.406) = −0.46.
171
criminality, suggesting that unemployed individuals who are excluded from legal income opportunities are, ceteris paribus, more likely to commit property crime than people who have a job and thus, providing strong support for Hypothesis 3. As expected from the theoretical discussion developed above, the results for the specific violent crimes presented in columns 5–8 are strikingly different from the results for property crime. The estimated deterrent effects of the clear-up rate are negative but statistically insignificant for all models supporting Hypothesis 2. A possible explanation for the failure of the clear-up rates in deterring these types of offences may be that violent crime is more often driven by impulsive actions (caused by hate, jealousy or anger) than by rational reckoning. Furthermore, unemployment is insignificant for murder and serious assault models, suggesting that unemployment does not affect the individual’s propensity towards violence providing support for Hypothesis 4. For rape, however, the estimated effect of unemployment is strongly negative. To resolve this (seemingly counterintuitive) result, we follow Raphael and WinterEbmer (2001) by separately identifying the effects of the unemployment rate among the offending and victimized populations.10 Accordingly, in the rape specification the general unemployment rate is substituted by gender specific unemployment rates. The results presented in the last column show that the coefficient for female unemployment is negative. This result can be explained within the theoretical framework developed by Cohen and Felson (1979). Thus, it can be argued that unemployed females are less exposed to the potential dangers that can arise during the time travelling to and returning from work and workinduced spare-time activities (and even while at work and work-induced spare time events). Female unemployment may also reduce the risk of child victimization committed by intimates or strangers when the parents are absent (Glaser & Rice, 1959). In contrast to that, the coefficient for male unemployment is positive with a significant effect consistent with Raphael and Winter-Ebmer (2001). This effect, however, disappears after controlling for income and demographic variables (see notes to Table 2) providing further support for Hypothesis 4.11 The Wald test suggests a rejection of the null hypothesis that all the coefficients are zero for the respective property crime and rape models. We find no significant evidence of serial correlation in the first-difference errors at order 2.12 The Sargan test, which comes from the one-
10 For rape, the greater part of the offending population is male and the victimised population is largely female (see Raphael & Winter-Ebmer, 2001, p. 277). 11 However, concerning the rape regression with gender-specific unemployment rates, a note of caution is in order. Due to the very high correlation of male and female unemployment rates (correlation = 0.8) we cannot rule out that the coefficients of these variables are artefacts resulting from “asking questions that may be too subtle for the available data to answer with any precision” (Wooldridge, 2009, p. 98). 12 We found evidence against the null hypothesis of zero autocorrelation in the first-differenced errors at order 1. However, this does not imply
172
Table 2 Crime equations – first difference GMM estimates. Property crimeb
Violent crimeb
Breaking and entering
Theft of motor cars
Robbery
Murder
Serious assault
Variable
GMM
GMM
GMM
GMM
log(ct−1 )a log(p)a log(u) log(ufemale ) log(umale ) Wald test Sargan test m1 m2 n
0.406** (0.098) −0.274** (0.114) 0.453** (0.167) – – 0 0.229 0.032 0.162 78
0.677** (0.123) −0.188 (0.146) 0.263** (0.118) – – 0 0.161 0.013 0.148 78
0.401* (0.235) −0.257* (0.136) 0.866** (0.118) – – 0 0.199 0.023 0.083* 78
0.083 (0.129) −0.128 (0.156) 0.189 (0.199) – – 0.746 0.619 0.017 0.521 76
Rape
Rape
GMM
GMM
GMM
0.057 (0.095) −0.273 (1.358) 0.025 (0.118) – – 0.888 0.005** 0.057 0.365 78
−0.050 (0.120) −0.108 (0.227) −0.319** (0.149) – – 0.002 0.121 0.042 0.670 49
0.018 (0.133) −0.149 (0.187) – −0.735** (0.227) 0.528** (0.194) 0.001 0.145 0.035 0.673 49
Note: Standard errors are robust to both heteroskedasticity and serial correlation and presented in parentheses. The p-value of the Wald test of a joint significance of all explanatory variables is reported. Sargan is a test of the overidentifying restrictions for the GMM estimators, p-value is reported. m1 and m2 are the tests of first-order and second-order serial correlation, asymptotically N(0,1), p-value is reported. GMM results are one-step estimates. a These variables are instrumented by lagged own values. b There are a number of other potential explanatory factors that are probably relevant to our analysis. Although data limitations generally frustrated this venture, we did collect data on gross domestic product (GDP) per capita, male population between 15 and 24 (in 1000), private vehicles (in 1000) and nights spent by non-residents in hotels and similar establishments. Including these additional variables does not generally alter our conclusions. We did find, however, a significant and positive effect of GDP per capita on breaking and entering and motor-car theft. Furthermore, the demographic and tourism variables show mixed results. Finally, controlling for GDP per capita and the young male population, we found that the male unemployment coefficient was not statistically significant anymore (full results are available upon request). * Significant at the 10% level. ** Significant at the 5% level.
G. Saridakis, H. Spengler / The Social Science Journal 49 (2012) 167–174
Type of crime
G. Saridakis, H. Spengler / The Social Science Journal 49 (2012) 167–174
step homoscedastic estimator,13 provides strong evidence in favour of the null hypothesis that the overidentifying restrictions are valid in all cases with the sole exception of serious assault.14 We can use our estimated model to make some simple predictions for the forthcoming crime burden of the country for 2011. As Greece currently faces its most serious economic crisis in modern times, changes in unemployment and clear up activity as a result of a weak aggregate demand and severe budget cuts are likely. These changes will in turn affect property crime and to a lesser extent violent crime. Eurostat reports an average annual unemployment rate for Greece of 12.6% in 2010 and the current issue of the Economist magazine (July 30th–August 5th) reports a rate of 15.8% for April 2011. If we assume that 15.8% will also be the annual average, unemployment would have risen by approximately 25% in 2011. Combining the above unemployment numbers with the significant short-run elasticities presented in Table 2 yields a rise in breaking and entering by more than 11%, a rise in theft of motor cars by more than 6.5% and a rise in robbery by more than 21.5%. Since crime is costly, Greek public wealth will be even more depressed. On the other hand, however, there will be a cost reduction as a consequence of a predicted fall in rapes by almost 8%. Though having substantially higher costs per offence than most property crimes, the decline in rape is unlikely to offset the additional costs of property crime, especially if we take into account that clear up rates will presumably fall in 2011 as a consequence of the overall austerity measures. If this is the case, the costs associated with property crime will be even higher. 6. Conclusions This paper empirically examines the effect of the clearup rate and the unemployment rate on crime in Greece. Using regional data from 1991 to 1998, which was a period characterised by austerity measures and economic restructuring in Greece, we estimated a linear dynamic panel data model based on the GMM estimator developed by Arellano and Bond (1991). We found that property crimes were significantly deterred by higher clear-up rates, while increased unemployment fostered criminal activity. This provides strong evidence for the relevance of the economic crime model to explain crimes with economic incentives. However, these associations could not be detected for violent crime. Only after employing gender-specific unemployment rates in the rape model we find evidence that male unemployment had a positive and significant effect, but the association became insignificant when other controls were
model misspecification because the first-differenced errors are serially correlated when the idiosyncratic errors are i.i.d. 13 The Sargan test cannot be computed when the robust option is specified since its asymptotic distribution is not known under the assumption of the robust model. 14 The Sargan test after a two-step estimator was used as an alternative. We found no evidence against the null hypothesis that the overidentifying restrictions are valid.
173
taken into consideration. The effect of female unemployment, however, turned out to be robustly negative and statistically significant. We suggest that the negative sign of the female unemployment coefficient in the rape equation may be better explained within the lifestyle/routine activity theory (Cohen & Felson, 1979), which represents the sociological interface of crime as opportunity. Putting our estimated model to practice we found that Greece will likely have to cope with much higher property crime in 2011. The fact that a decline in rape can be expected will not compensate for the expected rise in social costs due to property crime. This reasoning implies that crime does not work as an automatic stabilizer of the economy but rather as an automatic destabilizer. Thus good economic policy would also be good criminal policy. Our study also points toward future directions for research. As a consequence of the low explanatory power of the traditional explanatory variables motivated by economic crime theory to explain violent crime, we suggest that economists should start seeking models better suited to explaining violent behaviour. For example, violent crime may be better explained in the context of sociological and criminological theories of crime and therefore, theoretical integration may help better understand violent behaviour. A more extensive use of sociological theories must, however, be associated with a stronger focus on individual data because certain sociological crime theories cannot be sensibly investigated with macro data. Nevertheless, the availability of more informative macro panel data may also provide further insights, for example, on issues related to non-stationarity and cointegration of the crime, deterrence and socioeconomic variables usually explored within a time-series framework. Finally, sociologists and criminologists have frequently suggested that official data may be sparse and unreliable (especially for crimes such as rape) and individuals have little knowledge of the legal penalties that the criminal justice system assigns to various crimes. This may be overcome, for example, by regularly examining individuals’ perceptions of the criminal justice system and their criminal behaviour over time and on a larger scale.
References Arellano, M., & Bond, S. (1991). Some tests of specification for panel data: Monte Carlo evidence and an application to employment equation. Review of Economic Studies, 58(2), 277–297. Baldwin-Edwards, M. (2004). Statistical data on immigrants in Greece. Athens, Greece: Mediterranean Migration Observatory, Panteion University. Baltagi, B. (2008). Econometric analysis of panel data (4th ed.). Wiley. Becker, G. (1968). Crime and punishment: An economic approach. The Journal of Political Economy, 76(2), 169–217. Buonanno, P., & Montolio, D. (2008). Identifying the socioeconomic determinants of crime in Spanish provinces. International Review of Law and Economics, 28, 89–97. Carmichael, F., & Ward, R. (2001). Male unemployment and crime in England and Wales. Economic Letters, 73, 11–115. Cherry, T., & List, J. (2002). Aggregation bias in the economic model of crime. Economics Letters, 75(1), 81–86. Cohen, L., & Felson, M. (1979). Social change and crime rate trends: A routine activity approach. American Sociological Review, 44(4), 588–608.
174
G. Saridakis, H. Spengler / The Social Science Journal 49 (2012) 167–174
Ehrlich, I. (1973). Participation in illegitimate activities: A theoretical and empirical investigation. The Journal of Political Economy, 81(3), 521–565. Elliot, C., & Ellingworth, D. (1996). The relationship between unemployment and crime: A cross-sectional analysis employing the British Crime Survey 1992. International Journal of Manpower, 17, 81–88. Entorf, H., & Spengler, H. (2002). Crime in Europe: Causes and consequences. Berlin: Springer. Entorf, H., & Spengler, H. (2000). Socioeconomic and demographic factors of crime in Germany: Evidence from panel data of the German states. International Review of Law and Economics, 20, 75–106. Fajnzylber, P., Lederman, D., & Loayza, N. (2002). Inequality and violent crime. Journal of Law and Economics, 45(1), 1–40. Field, S. (1990). Trends in crime and their interpretation: A study of recorded crime in post-war England and Wales. In Home office research study 119. London: Home Office. Fougère, D., Kramarz, F., & Pouget, J. (2006). Youth unemployment and crime in France. CEPR Discussion Paper No. 5600. Freeman, R. (1999). The economics of crime. In Handbook of labour economics (pp. 3529–3571). Ashenfelter Card Editors 3C. Freeman, R., & Rodgers III, W., (1999). Area economic conditions and the labor market outcomes of young men in the 1990s expansion. National Bureau of Economic Research, Working Paper 7073. Greenberg, D. (2001). Time Series Analysis of Crime Rates. Journal of Quantitative Criminology, 17, 291–327. Glaser, D., & Rice, K. (1959). Crime, age, and employment. American Sociological Review, 24, 679–686. Hale, C. (1998). Crime and business cycle in post-war Britain revisited. British Journal of Criminology, 38(4), 681–698. Levitt, S. (2001). Alternative strategies for identifying the link between unemployment and crime. Journal of Quantitative Criminology, 17(4), 377–390. Levitt, S. (2004). Understanding why crime fell in the 1990s: Four factors that explain the decline and seven that do not. Journal of Economic Perspective, 18(1), 163–190. Levitt, S. (1996). The Effect of Prison Population Size on Crime Rates: Evidence from Prison Overcrowding Litigation. The Quarterly Journal of Economics, 111(2), 319–351. Lin, M.-J. (2008). Does unemployment increase crime? Evidence from U.S. data 1974–2000. Journal of Human Resources, 43(2), 413–436.
Marris, R. (2000). Survey of the literature on the criminoand economic factors influencing crime trends. logical www.volterra.co.uk/publicationpdf/VP 0011.pdf Marselli, R., & Vannini, M. (1997). Estimating a crime equation in the presence of organized crime: Evidence from Italy. International Review of Law and Economics, 17, 89–113. Marvell, T., & Moody, C. (1994). Prison population growth and crime reduction. Journal of Quantitative Criminology, 10(2), 109–140. Merton, R. K. (1938). Social structure and anomie. American Sociological Review, 3, 672–682. Papadopoulos, T. (2006). Support for the unemployed in a familistic welfare regime. In E. Mossialos, & M. Petmesidou (Eds.), Social policy developments in Greece. Aldershot: Ashgate. Pyle, D. J. (1983). The economic of crime and law enforcement. London: MacMillan. Raphael, S., & Winter-Ebmer, R. (2001). Identifying the effect of unemployment on crime. Journal of Law and Economics, XLIV, 259–283. Saridakis, G. (2011). Violent crime and incentives in the long-run: Evidence from England and Wales. Journal of Applied Statistics, 38(4), 647–660. Saridakis, G. (2004). Violent crime in the United States of America: A time-series analysis between 1960–2000. European Journal of Law and Economics, 18(2), 203–221. Shaw, C., & McKay, H. (1942/1969). Juvenile delinquency and urban areas (rev. ed.). Chicago, IL: University of Chicago Press. Sutherland, E. H. (1942/1973). Development of the theory. In Schuessler (Ed.), Edwin Sutherland on analysing crime (rev. ed., pp. 30–41). Chicago, IL: Chicago University Press. Vold, G., Bernard, T., & Snipes, J. (2002). Theoretical criminology (5th ed.). Oxford University Press. Witt, R., Clarke, A., & Fielding, N. (1999). Crime and economic activity: A panel data approach. British Journal of Criminology, 39(3), 391–400. Witte, & Witt. (2002). Crime causation: Economic theories (pp. 302–308). Encyclopaedia of crime and justice New York: Macmillan. Witte, A. (1980). Estimating the economic model of crime with individual data. Quarterly Journal of Economics, 16, 69–85. Wolpin, K. (1978). An economic analysis of crime and punishment in England and Wales, 1847–1967. The Journal of Political Economy, 86(5), 815–840. Wooldridge, J. M. (2009). Introductory econometrics: A modern approach (4th ed.). South Western.
Contents lists available at ScienceDirect
The Social Science Journal journal homepage: www.elsevier.com/locate/soscij
Crime, deterrence and unemployment in Greece: A panel data approach George Saridakis a,∗ , Hannes Spengler b a b
Loughborough University, School of Business and Economics, Ashby Road, LE11 3TU, Loughborough, Leicestershire, UK University of Applied Sciences Mainz, School of Business, Lucy-Hillebrand-Strasse 2, 55128 Mainz, Germany
a r t i c l e
i n f o
Article history: Received 5 August 2009 Received in revised form 9 August 2011 Accepted 10 August 2011 Available online 19 April 2012 Keywords: Crime Deterrence Unemployment Panel data GMM
a b s t r a c t This study empirically examines the relationship among crime, deterrence and unemployment in Greece. A regional dataset over the period 1991–1998 was collected and analysed. Our econometric methodology follows the Generalized Method of Moments (GMM) estimator applied to dynamic models of panel data. The results show that property crimes are significantly deterred by higher clear-up rates and that unemployment increases crime. For violent crimes, however, the effect of the clear-up rate and unemployment are found to be generally insignificant. Finally, our results may provide support to policy makers in forecasting criminal activity in the current economic downturn under a wave of harsh austerity measures, budget cuts and increased unemployment. © 2011 Western Social Science Association. Published by Elsevier Inc. All rights reserved.
1. Introduction Becker (1968) marked the beginning of attempts to apply economic models of rational decision-making to crime. Since then, a number of significant theoretical developments have been made and substantial empirical work has been published in academic journals on both economics and criminology. Most of the applied work, however, focuses on US crime statistics (prominent examples being the work by Ehrlich, 1973; Witte, 1980; Marvell & Moody, 1994; Levitt, 1996) and it is only recently that economics-of-crime models have been increasingly applied outside the USA and used to derive country-specific policy to reduce the rate of criminal behaviour. Examples of such work in Europe, for instance, can be found in Fougère, Kramarz, and Pouget (2006) for France, Buonanno and Montolio (2008) for Spain, Entorf and Spengler (2000)
∗ Corresponding author.. E-mail addresses: [email protected] (G. Saridakis), [email protected] (H. Spengler).
for Germany, Hale (1998) and Saridakis (2011) for Britain and Marselli and Vannini (1997) for Italy. In this context, this paper adds further evidence for Europe by examining the effect of two central variables in economics-of-crime research, these being the clear-up rate (as a proxy for the probability of getting caught) and the unemployment rate (as an inverse measure of legal income opportunities), on crime in Greece from 1991 to 1998. Indeed, this period was marked by a number of economic austerity and economic restructuring measures undertaken by the policy makers to meet the criteria for participation in the European Monetary Union (EMU) and which led to a significant increase in the country’s recorded unemployment from 7% to almost 11% (Papadopoulos, 2006) combined with a sharp increase in crime, especially in property crime (Table 1).1 In this paper we
1 Of course, there were other events, such as the significant increase in illegal immigration flows particularly from Balkan counties, and also from Middle-East and North Africa, which may be associated particularly with organised crime (e.g. drug related crimes increased 220% during the study period). Unfortunately, we do not have disaggregated data on
0362-3319/$ – see front matter © 2011 Western Social Science Association. Published by Elsevier Inc. All rights reserved. doi:10.1016/j.soscij.2011.08.005
168
G. Saridakis, H. Spengler / The Social Science Journal 49 (2012) 167–174
Table 1 Crime, deterrence and unemployment in Greece, 1991–1998. % Change in Recorded crime Property crime Breaking and entering Theft of motor cars Robbery Violent crime Murder Serious assault Rape Unemployment rate
Average Clear-up rate
Recorded crime
Clear-up rate
37 95 83
−33 −40 −29
365.1 116.5 15.6
13.4 12.8 31.1
50 0 0 57
−16 −4 12
2.7 67.3 2.4 8.9
74.9 95.5 69.5
Note: Recorded crime is per 100,000 inhabitants. Clear-up rate and unemployment are in %.
consider six major crime categories: breaking and entering, theft of motor cars, robbery, murder, serious assault and rape (the first three are property crimes, while the latter three are violent crimes2 ) and empirically examine these types of crime using Generalized Methods of Moments (GMM) to determine whether increasing unemployment and decreasing clear-up rates contributed to the increase in crime in the 1990s. Furthermore, focusing on unemployment and deterrence, our analysis may provide some useful information about the consequences of the recent Greek sovereign debt crisis on crime rates, since the extensive and unprecedented package of austerity measures are likely to increase, at least in the short and medium-run, unemployment (due to weak domestic demand) and reduce clear-up rates (perhaps due to budget cuts in law enforcement). Finally, our study contributes to the current debate regarding the relevance of deterrence and economic variables in explaining violent behaviour (Entorf & Spengler, 2000; Raphael & WinterEbmer, 2001; Saridakis, 2004), associations which for the most part are already well established for crimes against property. The outline of the paper is as follows. Section 2 provides an overview of the literature from which our hypotheses are derived. Section 3 describes the data. Section 4 presents the empirical framework. Section 5 is devoted to the presentation and discussion of the results. The last section concludes. 2. Background and hypothesis derivation There are a number of theories that may explain the causes of criminality from a sociological paradigm. Inter alia, there is Merton’s strain theory (Merton, 1938) which suggests that crime rates will tend to be higher
immigration and furthermore the size of illegal immigration is difficult to measure with accuracy. The 2001 census, however, conducted by the National Statistical Service of Greece suggests that immigrants represent around 7% of the total population, but it may rise up to 10% if full account of the illegal immigrants is taken into consideration (see Baldwin-Edwards, 2004). Also, unemployment rates for foreign-born individuals tend to be higher than for the native-born individuals (e.g. according to OECD statistics the unemployment rate for native-born and foreign-born males in 1995 were 6.1% and 14%, respectively). 2 Robbery is included in property crimes since it might be dominated by the desire to steal another person’s property (see Saridakis, 2004).
in societies where opportunities are most unequal. Alternatively, the social disorganization theory suggested by Shaw and McKay (1942) explains crime and delinquency rates by examining factors (low economic status, ethnic heterogeneity and residential mobility) that contribute to disruption of community social organizations. A third explanation is derived from Sutherland’s differential association/social learning theory (Sutherland, 1942) which mainly emphasizes that criminal behaviour is learned through interaction within the intimate family unit and close peers. Finally, the lifestyle/routine activity theory by Cohen and Felson (1979) represents the sociological interface of crime as opportunity. This theory suggests that crime will occur when there is a motivated person, a suitable target (potential loot or victim) and the absence of guardians capable of preventing the violation. The paper by Becker (1968) marked the beginning of attempts to apply economic models of rational decision making to crime. Becker suggested that individuals are rational utility maximisers who decide whether or not to engage in criminal activity by comparing the cost and benefits of crime. Becker’s paper, however, was not intended to present a new theory of criminal behaviour, but to analyse how to minimise the social loss of crime. To this end, Becker derived the supply of offence function which is “relating the number of offences by any person to his probability of conviction, to his punishment if convicted, and to other variables, such as the income available to him in legal and other illegal activities, the frequency of nuisance arrests, and his willingness to commit an illegal act” (Becker, 1968:177). Ehrlich (1973) extended the work of Becker by considering a time allocation model and for almost 30 years now in economics, the empirical work in crime has relied on the foundations of the Becker–Ehrlich model to inform statistical specifications. According to the basic economic crime model, individuals have to decide how to allocate their time (t) between legitimate (t − t2 ) and illegitimate (t2 ) activities. Based on the assumption that crime and legal employment are substitute activities it follows that the more time one devotes to a legitimate occupation, the less time one has to carry out criminal activity. The individual’s expected utility is a weighted average of his/her utility in the two alternative states of the world: (a) the individual is caught and punished and (b) the individual is not caught and punished.
G. Saridakis, H. Spengler / The Social Science Journal 49 (2012) 167–174
The expected utility is given by: EU = pU[W0 + W1 (t − t2 ) + W2 (t2 ) − F(t2 )] + (1 − p)U[(W0 + W1 (t − t2 ) + W2 (t2 )]
(1)
where p is the probability of being caught and punished, W0 is the individual’s (exogenous) wealth, W1 and W2 are the wealth from legitimate activity and crime, respectively. F is the monetary equivalent of the punishment. Crime occurs when the marginal expected returns from crime are greater than the marginal return in legitimate work,3 i.e. W2 (t2 ) − pF (t2 ) > W1 (t − t2 ). Thus, the basic crime theory4 predicts that sanctions and legitimate opportunities influence decisions to engage in crime. 2.1. Crime and deterrence Being caught and punished may deter offenders from committing a crime again, as well as deter future offenders who contemplate committing crimes. Ehrlich (1973) was the first economist who tested the crime deterrence model empirically using cross sectional data of US states from 1940, 1950 and 1960. Ehrlich’s econometric crossstate estimation showed a strong negative association between crime and criminal justice variables in 10 out of 14 crime categories. Interestingly, the absolute values of the estimated elasticities of violent crime with respect to deterrence variables were found not to be lower on average than those associated with crime against property. Hence, the implication appeared to be that law enforcement was not less effective in combating crimes of violence relative to crimes against property. However, a few years later, Wolpin (1978) using time-series data for the period 1894–1967 from England and Wales, found that the estimated coefficients of the probability variables devoted to deterrence were greater in magnitude and showed higher statistical significance for property crimes than crimes against the person. Since then, there has been a major research effort and also a major statistical development concerning potential simultaneity between deterrence variables and crime rates. The argument here is that while increases in the number of prisoners are likely to reduce crime, rising crime rates also translate into a larger prison population (resulting in a downward bias of the effect of deterrence on crime) or that a higher incidence of crime leads to lower clear-up rates due to the overload of the given and temporarily fixed police resources (resulting in an upward bias of the effect of deterrence on crime). For example, Levitt (1996) assessed the interaction problem between prisoner populations and US crime rates using state level data from the period 1971–1993. To account for the possible endogeneity
3
For further information and a formal proof see Pyle (1983). Since the Becker–Ehrlich model, significant theoretical and empirical developments have been made. For example, early economic studies applied a static framework to analyse criminal behaviour. However, there are several theoretical and empirical reasons, for example, through capital accumulation, peer group effects, improvement in fit when lagged dependent variables are included in the model, which suggest that criminal behaviour should be analysed in a dynamic framework (Witte and Witt, 2002). 4
169
between prison population size and crime rates, Levitt used the prison overcrowding litigation as an instrument for changes in prison population (assuming that overcrowding litigation influences prison population and has no direct effect on crime). Then, the effect of prison population on both property and violent crimes was estimated, but for the latter case the reported elasticities were found to be smaller in magnitude and to lack in statistical significance relative to property crimes. These earlier studies may provide evidence regarding the strong link between deterrence and property crime, but what can we expect to happen to violent crime rates when governments adopt costly deterrence-based policies by increasing their expenditures on prison, courts and law enforcement generally? Evidence from recent work is revealing. For example, Entorf and Spengler (2000) found weak effects of the clear-up rates on violent crime whereas they found strong support for the deterrence hypothesis for property crimes. Similarly, Cherry and List (2002) using county-level US panel data and considering a wide variety of criminal justice variables conclude that sanctions are less influential in deterring serious violent crimes, such as murder or rape. Saridakis (2004) using US time-series data found short-run effects of the prison population on murder and rape, but no long-run associations were established. Buonanno and Montolio (2008) using a short panel of Spanish provinces over the period 1993 and 1999 also failed to establish a significant association between clearance rate and crime against the person. We therefore hypothesise that: Hypothesis 1. Criminal justice variables are likely to deter property crimes. Hypothesis 2. Criminal justice variables are likely to have a weak or no association with violent crimes. 2.2. Crime and unemployment Nearly all economic (and non-economic) studies, consider unemployment as an important cause of crime. Unemployment reflects the (lack of) opportunity for participation in the legitimate job market and the acquisition of legal earnings (Freeman, 1999). The exclusion from legal income opportunities increases the expected returns from crime. The economic crime theory, as well as sociological theories of crime (Merton, 1938; Shaw & McKay, 1942), even though they may be based on different concepts and assumptions, suggest that the crime rates increase with increasing unemployment. However, those advocating opportunity theories suggest a negative association between unemployment and crime (Cohen & Felson, 1979). Although the data seem to support a positive link between unemployment and property crime the association between unemployment and violent crime is very diverse (Vold, Bernard, & Snipes, 2002). A substantial number of studies for the USA found that unemployment had a generally significant impact on property crime rates, but it had no significant impact on violent crime rates. Freeman and Rodgers (1999) found that young African American men in tight labour markets experienced a boost in employment and earnings during the 1990s.
170
G. Saridakis, H. Spengler / The Social Science Journal 49 (2012) 167–174
According to their analysis, a drop in US unemployment between 1992 and 1997 reduced youth (overall) crime. Levitt (2004), however, suggests that the decline in the US unemployment rate during the 1990s may explain the decline in property crime, but not changes in violent crime. Moreover, Saridakis (2004) used the duration of unemployment (mean of weeks unemployed) in his analysis of violent crime, but his results revealed no significant association between unemployment duration and violent crime. Consistent with this argument, studies for the UK suggest a marginal effect of unemployment on violent crime. For example, Carmichael and Ward (2001) using countylevel data for England and Wales suggest that neither adult nor youth unemployment had a significant effect on violent crime. Similarly, Elliot and Ellingworth (1996) examined the effect of unemployment on personal crime using the 1992 British Crime Survey (BCS) and found no significant relationship. Nevertheless, Field (1990) using time-series data for England and Wales from 1952 to 1987 failed to find any association between most types of crime and unemployment. Marris (2000) suggested that in order to understand the relationship between unemployment and crime, the distinction between involuntary and voluntary unemployment should be made. An insignificant association has also been reported between youth unemployment and violent crime in France (Fougère et al., 2006). Interestingly, some other studies for the United States (Greenberg, 2001; Raphael & WinterEbmer, 2001) and one for Germany (Entorf & Spengler, 2000) have reported a negative association between duration of unemployment or/and unemployment and serious violent crimes such as murder and/or rape. Weighing up the existing research on the unemployment–crime link leads us to our final two hypotheses: Hypothesis 3. Unemployment is likely to be positively associated with property crimes. Hypothesis 4. Unemployment is likely to have a weak or no effect on violent crimes. 3. Data The data set used in this paper is a panel of annual, regional-level (Nuts 2) observations,5 stretching from 1991 to 1998 (see Table 1). Reported crime data (per 100,000 inhabitants) and clear-up rates (%) were obtained directly from the collector of crime statistics in Greece and to our knowledge are not officially published in electronic or printed form.6 In contrast to cross-country studies/panels (Fajnzylber, Lederman, & Loayza, 2002), a regional panel
5 Greece is divided into 13 regional districts (population in 1000 is in parenthesis): Crete (553.1), South Aegean (263.6), North Aegean (186.8), Attica (3479.2), Peloponnese (650.7), Central Greece (637.0), West Greece (726.2), Ionian Islands (197.2), Epirus (360.2), Thessaly (740.6), West Macedonia (300.1), Central Macedonia (1759.3) and East Macedonia and Thrace (561.3). 6 The data were made available after an official data request from the German Federal Criminal Police Office (BKA) to Greek authorities to be analysed in a comparative European crime project by Entorf and Spengler (2002).
data set for one country overcomes or attenuates problems associated with differences in crime definitions, statistical data collection (usually there are no official differences in crime definitions and counting rules across different regions of the same country) and reporting propensities (as long as regional reporting propensities are constant over time differences between regions can be controlled by suitable estimation techniques, Section 4).7 Finally, data for the unemployment rate (%) were extracted from the Eurostat statistical database. 4. Empirical framework Based on the empirical deterrence literature, we can derive the following simple model: cit = ˇ0 + ˇ1 cit−1 + ˇ2 pit + ˇ3 uit + εit
(2)
εit = i + vit
(3)
where cit is the crime rate (per 100,000 inhabitants) by subcategory for administrative division i in year t. The lagged dependent variable reflects the tendency of individuals who are involved in criminal activity to continue with it even after the circumstances that led them to turn to crime have changed (Witt, Clarke, & Fielding, 1999). The measure pit is the clear-up rate, which is treated as endogenous. For example, this would be the case if a higher incidence of crime led to lower clear-up rates due to the overload of the given and temporarily fixed police resources. On the other hand, a feedback effect could be caused if the state responded to increasing crime rates by allocating more resources to the police, resulting in higher clear-up rates. The variable uit is the unemployment rate and is considered to be exogenous.8 All variables are measured in natural logarithms. Finally, i is the unobserved time-constant regionallevel effect which may be correlated with some of the regressors. For this reason, we estimate the following linear dynamic model: cit = ı1 cit−1 + ı2 pit + ı3 uit + vit
(4)
where the variables are first-differenced to eliminate time-invariant regional-level effects (i.e. regional heterogeneity). However, first-differencing introduces regressor–error correlation (endogeneity) since cit−1 = cit−1 − cit−2 is correlated with vit = vit − vit−1 because vit contains vit−1 which is correlated with cit−1 and also serial correlation in the error term (i.e. between vit = vit − vit−1 and vit−1 = vit−1 − vit−2 ). To deal with the variables cit−1 and pit , we have used the Generalized Methods of Moments (GMM) proposed by Arellano and Bond (1991). Baltagi (2008) provides a comprehensive overview of the Arellano and Bond estimator (pp. 149–155). Briefly,
7 Also, it surmounts statistical weaknesses of studies relying on national time-series crime data (see Levitt, 2001). 8 We have experimented by allowing unemployment to be endogenous in the model (i.e. crime participation may reduce the employability of previously convicted offenders and it may in turn contribute to observed unemployment – see Raphael & Winter-Ebmer, 2001). Generally, the results did not provide further insight into our analysis and, hence, are not reported here.
G. Saridakis, H. Spengler / The Social Science Journal 49 (2012) 167–174
the variables cit−1 and pit are instrumented with suitable lags of their own levels. First-differences of strictly exogenous covariates are also used to create moment conditions. u is modelled as strictly exogenous and contributes 1 instrument. The remaining instruments (42) come from the p-2 instruments available in periods p = 3–8. Furthermore, GMM takes account of the serial correlation in the same manner generalized least squares (GLS) takes care of non-independent and identically distributed (i.i.d.) errors. Since our model uses a double log functional form, the slope coefficients are elasticities. Thus, our estimates show the percentage change in the dependent variable (crime) if a specific explanatory variable changes by 1%, while the other explanatory variables are held constant. In our model, the estimated coefficients represent short-run effects (the immediate change of the dependent variable as a reaction to a change in an explanatory variable) but the long-run elasticities (i.e. the long-run change of the dependent variable across all periods as a reaction to a permanent change in an explanatory variable) can be easily obtained by dividing the estimated coefficients, excluding that of the lagged dependent variable, by one minus the coefficient of the lagged dependent variable (Buonanno and Montolio, 2008; Fajnzylber et al., 2002; Witt et al., 1999). 5. Empirical results Table 2 presents the one-step GMM results of the crime models for Greece. We first comment on the property crime results presented in columns 2–4. Clearly these columns show a strong negative effect of the clear-up rate on property offences (with the sole exception of motor-car theft) supporting Hypothesis 1. The insignificance of the short run effect in the regression for theft of motorcars might be a consequence of the typical car thief. Motor-car thieves can be considered as highly specialised and rewarded serial criminal offenders (presumably networked within groups of car racketeers), who have high opportunity costs for legal work and thus, being relatively immune against changes in deterrence, especially if the mean clear-up rate is low. The coefficient estimates for breaking and entering and robbery suggest long-run clear-up elasticities of −0.46 and −0.43, respectively.9 These estimates are similar to those reported for other European countries: see, for example, Buonanno and Montolio (2008) for Spain, Entorf and Spengler (2000) for Germany and Witt et al. (1999) for England and Wales. For the three individual property crimes, the estimated effects of unemployment are positive and significant. The GMM (short-run) estimates of the unemployment rate range from 0.236 to 0.866 suggesting long-run unemployment elasticities of 0.76, 0.81 and 1.45 for breaking and entering, theft of motor cars and robbery, respectively. In fact, panel data studies by Marselli and Vannini (1997) for Italy and Lin (2008) and Raphael and Winter-Ebmer (2001) for the USA also found significantly positive effects of unemployment on property crime rates. This result is reasonable and in line with the mainstream economic view of
9 The long-run clear-up elasticity for breaking and entering, for example, is calculated as follows: −0.274/(1 − 0.406) = −0.46.
171
criminality, suggesting that unemployed individuals who are excluded from legal income opportunities are, ceteris paribus, more likely to commit property crime than people who have a job and thus, providing strong support for Hypothesis 3. As expected from the theoretical discussion developed above, the results for the specific violent crimes presented in columns 5–8 are strikingly different from the results for property crime. The estimated deterrent effects of the clear-up rate are negative but statistically insignificant for all models supporting Hypothesis 2. A possible explanation for the failure of the clear-up rates in deterring these types of offences may be that violent crime is more often driven by impulsive actions (caused by hate, jealousy or anger) than by rational reckoning. Furthermore, unemployment is insignificant for murder and serious assault models, suggesting that unemployment does not affect the individual’s propensity towards violence providing support for Hypothesis 4. For rape, however, the estimated effect of unemployment is strongly negative. To resolve this (seemingly counterintuitive) result, we follow Raphael and WinterEbmer (2001) by separately identifying the effects of the unemployment rate among the offending and victimized populations.10 Accordingly, in the rape specification the general unemployment rate is substituted by gender specific unemployment rates. The results presented in the last column show that the coefficient for female unemployment is negative. This result can be explained within the theoretical framework developed by Cohen and Felson (1979). Thus, it can be argued that unemployed females are less exposed to the potential dangers that can arise during the time travelling to and returning from work and workinduced spare-time activities (and even while at work and work-induced spare time events). Female unemployment may also reduce the risk of child victimization committed by intimates or strangers when the parents are absent (Glaser & Rice, 1959). In contrast to that, the coefficient for male unemployment is positive with a significant effect consistent with Raphael and Winter-Ebmer (2001). This effect, however, disappears after controlling for income and demographic variables (see notes to Table 2) providing further support for Hypothesis 4.11 The Wald test suggests a rejection of the null hypothesis that all the coefficients are zero for the respective property crime and rape models. We find no significant evidence of serial correlation in the first-difference errors at order 2.12 The Sargan test, which comes from the one-
10 For rape, the greater part of the offending population is male and the victimised population is largely female (see Raphael & Winter-Ebmer, 2001, p. 277). 11 However, concerning the rape regression with gender-specific unemployment rates, a note of caution is in order. Due to the very high correlation of male and female unemployment rates (correlation = 0.8) we cannot rule out that the coefficients of these variables are artefacts resulting from “asking questions that may be too subtle for the available data to answer with any precision” (Wooldridge, 2009, p. 98). 12 We found evidence against the null hypothesis of zero autocorrelation in the first-differenced errors at order 1. However, this does not imply
172
Table 2 Crime equations – first difference GMM estimates. Property crimeb
Violent crimeb
Breaking and entering
Theft of motor cars
Robbery
Murder
Serious assault
Variable
GMM
GMM
GMM
GMM
log(ct−1 )a log(p)a log(u) log(ufemale ) log(umale ) Wald test Sargan test m1 m2 n
0.406** (0.098) −0.274** (0.114) 0.453** (0.167) – – 0 0.229 0.032 0.162 78
0.677** (0.123) −0.188 (0.146) 0.263** (0.118) – – 0 0.161 0.013 0.148 78
0.401* (0.235) −0.257* (0.136) 0.866** (0.118) – – 0 0.199 0.023 0.083* 78
0.083 (0.129) −0.128 (0.156) 0.189 (0.199) – – 0.746 0.619 0.017 0.521 76
Rape
Rape
GMM
GMM
GMM
0.057 (0.095) −0.273 (1.358) 0.025 (0.118) – – 0.888 0.005** 0.057 0.365 78
−0.050 (0.120) −0.108 (0.227) −0.319** (0.149) – – 0.002 0.121 0.042 0.670 49
0.018 (0.133) −0.149 (0.187) – −0.735** (0.227) 0.528** (0.194) 0.001 0.145 0.035 0.673 49
Note: Standard errors are robust to both heteroskedasticity and serial correlation and presented in parentheses. The p-value of the Wald test of a joint significance of all explanatory variables is reported. Sargan is a test of the overidentifying restrictions for the GMM estimators, p-value is reported. m1 and m2 are the tests of first-order and second-order serial correlation, asymptotically N(0,1), p-value is reported. GMM results are one-step estimates. a These variables are instrumented by lagged own values. b There are a number of other potential explanatory factors that are probably relevant to our analysis. Although data limitations generally frustrated this venture, we did collect data on gross domestic product (GDP) per capita, male population between 15 and 24 (in 1000), private vehicles (in 1000) and nights spent by non-residents in hotels and similar establishments. Including these additional variables does not generally alter our conclusions. We did find, however, a significant and positive effect of GDP per capita on breaking and entering and motor-car theft. Furthermore, the demographic and tourism variables show mixed results. Finally, controlling for GDP per capita and the young male population, we found that the male unemployment coefficient was not statistically significant anymore (full results are available upon request). * Significant at the 10% level. ** Significant at the 5% level.
G. Saridakis, H. Spengler / The Social Science Journal 49 (2012) 167–174
Type of crime
G. Saridakis, H. Spengler / The Social Science Journal 49 (2012) 167–174
step homoscedastic estimator,13 provides strong evidence in favour of the null hypothesis that the overidentifying restrictions are valid in all cases with the sole exception of serious assault.14 We can use our estimated model to make some simple predictions for the forthcoming crime burden of the country for 2011. As Greece currently faces its most serious economic crisis in modern times, changes in unemployment and clear up activity as a result of a weak aggregate demand and severe budget cuts are likely. These changes will in turn affect property crime and to a lesser extent violent crime. Eurostat reports an average annual unemployment rate for Greece of 12.6% in 2010 and the current issue of the Economist magazine (July 30th–August 5th) reports a rate of 15.8% for April 2011. If we assume that 15.8% will also be the annual average, unemployment would have risen by approximately 25% in 2011. Combining the above unemployment numbers with the significant short-run elasticities presented in Table 2 yields a rise in breaking and entering by more than 11%, a rise in theft of motor cars by more than 6.5% and a rise in robbery by more than 21.5%. Since crime is costly, Greek public wealth will be even more depressed. On the other hand, however, there will be a cost reduction as a consequence of a predicted fall in rapes by almost 8%. Though having substantially higher costs per offence than most property crimes, the decline in rape is unlikely to offset the additional costs of property crime, especially if we take into account that clear up rates will presumably fall in 2011 as a consequence of the overall austerity measures. If this is the case, the costs associated with property crime will be even higher. 6. Conclusions This paper empirically examines the effect of the clearup rate and the unemployment rate on crime in Greece. Using regional data from 1991 to 1998, which was a period characterised by austerity measures and economic restructuring in Greece, we estimated a linear dynamic panel data model based on the GMM estimator developed by Arellano and Bond (1991). We found that property crimes were significantly deterred by higher clear-up rates, while increased unemployment fostered criminal activity. This provides strong evidence for the relevance of the economic crime model to explain crimes with economic incentives. However, these associations could not be detected for violent crime. Only after employing gender-specific unemployment rates in the rape model we find evidence that male unemployment had a positive and significant effect, but the association became insignificant when other controls were
model misspecification because the first-differenced errors are serially correlated when the idiosyncratic errors are i.i.d. 13 The Sargan test cannot be computed when the robust option is specified since its asymptotic distribution is not known under the assumption of the robust model. 14 The Sargan test after a two-step estimator was used as an alternative. We found no evidence against the null hypothesis that the overidentifying restrictions are valid.
173
taken into consideration. The effect of female unemployment, however, turned out to be robustly negative and statistically significant. We suggest that the negative sign of the female unemployment coefficient in the rape equation may be better explained within the lifestyle/routine activity theory (Cohen & Felson, 1979), which represents the sociological interface of crime as opportunity. Putting our estimated model to practice we found that Greece will likely have to cope with much higher property crime in 2011. The fact that a decline in rape can be expected will not compensate for the expected rise in social costs due to property crime. This reasoning implies that crime does not work as an automatic stabilizer of the economy but rather as an automatic destabilizer. Thus good economic policy would also be good criminal policy. Our study also points toward future directions for research. As a consequence of the low explanatory power of the traditional explanatory variables motivated by economic crime theory to explain violent crime, we suggest that economists should start seeking models better suited to explaining violent behaviour. For example, violent crime may be better explained in the context of sociological and criminological theories of crime and therefore, theoretical integration may help better understand violent behaviour. A more extensive use of sociological theories must, however, be associated with a stronger focus on individual data because certain sociological crime theories cannot be sensibly investigated with macro data. Nevertheless, the availability of more informative macro panel data may also provide further insights, for example, on issues related to non-stationarity and cointegration of the crime, deterrence and socioeconomic variables usually explored within a time-series framework. Finally, sociologists and criminologists have frequently suggested that official data may be sparse and unreliable (especially for crimes such as rape) and individuals have little knowledge of the legal penalties that the criminal justice system assigns to various crimes. This may be overcome, for example, by regularly examining individuals’ perceptions of the criminal justice system and their criminal behaviour over time and on a larger scale.
References Arellano, M., & Bond, S. (1991). Some tests of specification for panel data: Monte Carlo evidence and an application to employment equation. Review of Economic Studies, 58(2), 277–297. Baldwin-Edwards, M. (2004). Statistical data on immigrants in Greece. Athens, Greece: Mediterranean Migration Observatory, Panteion University. Baltagi, B. (2008). Econometric analysis of panel data (4th ed.). Wiley. Becker, G. (1968). Crime and punishment: An economic approach. The Journal of Political Economy, 76(2), 169–217. Buonanno, P., & Montolio, D. (2008). Identifying the socioeconomic determinants of crime in Spanish provinces. International Review of Law and Economics, 28, 89–97. Carmichael, F., & Ward, R. (2001). Male unemployment and crime in England and Wales. Economic Letters, 73, 11–115. Cherry, T., & List, J. (2002). Aggregation bias in the economic model of crime. Economics Letters, 75(1), 81–86. Cohen, L., & Felson, M. (1979). Social change and crime rate trends: A routine activity approach. American Sociological Review, 44(4), 588–608.
174
G. Saridakis, H. Spengler / The Social Science Journal 49 (2012) 167–174
Ehrlich, I. (1973). Participation in illegitimate activities: A theoretical and empirical investigation. The Journal of Political Economy, 81(3), 521–565. Elliot, C., & Ellingworth, D. (1996). The relationship between unemployment and crime: A cross-sectional analysis employing the British Crime Survey 1992. International Journal of Manpower, 17, 81–88. Entorf, H., & Spengler, H. (2002). Crime in Europe: Causes and consequences. Berlin: Springer. Entorf, H., & Spengler, H. (2000). Socioeconomic and demographic factors of crime in Germany: Evidence from panel data of the German states. International Review of Law and Economics, 20, 75–106. Fajnzylber, P., Lederman, D., & Loayza, N. (2002). Inequality and violent crime. Journal of Law and Economics, 45(1), 1–40. Field, S. (1990). Trends in crime and their interpretation: A study of recorded crime in post-war England and Wales. In Home office research study 119. London: Home Office. Fougère, D., Kramarz, F., & Pouget, J. (2006). Youth unemployment and crime in France. CEPR Discussion Paper No. 5600. Freeman, R. (1999). The economics of crime. In Handbook of labour economics (pp. 3529–3571). Ashenfelter Card Editors 3C. Freeman, R., & Rodgers III, W., (1999). Area economic conditions and the labor market outcomes of young men in the 1990s expansion. National Bureau of Economic Research, Working Paper 7073. Greenberg, D. (2001). Time Series Analysis of Crime Rates. Journal of Quantitative Criminology, 17, 291–327. Glaser, D., & Rice, K. (1959). Crime, age, and employment. American Sociological Review, 24, 679–686. Hale, C. (1998). Crime and business cycle in post-war Britain revisited. British Journal of Criminology, 38(4), 681–698. Levitt, S. (2001). Alternative strategies for identifying the link between unemployment and crime. Journal of Quantitative Criminology, 17(4), 377–390. Levitt, S. (2004). Understanding why crime fell in the 1990s: Four factors that explain the decline and seven that do not. Journal of Economic Perspective, 18(1), 163–190. Levitt, S. (1996). The Effect of Prison Population Size on Crime Rates: Evidence from Prison Overcrowding Litigation. The Quarterly Journal of Economics, 111(2), 319–351. Lin, M.-J. (2008). Does unemployment increase crime? Evidence from U.S. data 1974–2000. Journal of Human Resources, 43(2), 413–436.
Marris, R. (2000). Survey of the literature on the criminoand economic factors influencing crime trends. logical www.volterra.co.uk/publicationpdf/VP 0011.pdf Marselli, R., & Vannini, M. (1997). Estimating a crime equation in the presence of organized crime: Evidence from Italy. International Review of Law and Economics, 17, 89–113. Marvell, T., & Moody, C. (1994). Prison population growth and crime reduction. Journal of Quantitative Criminology, 10(2), 109–140. Merton, R. K. (1938). Social structure and anomie. American Sociological Review, 3, 672–682. Papadopoulos, T. (2006). Support for the unemployed in a familistic welfare regime. In E. Mossialos, & M. Petmesidou (Eds.), Social policy developments in Greece. Aldershot: Ashgate. Pyle, D. J. (1983). The economic of crime and law enforcement. London: MacMillan. Raphael, S., & Winter-Ebmer, R. (2001). Identifying the effect of unemployment on crime. Journal of Law and Economics, XLIV, 259–283. Saridakis, G. (2011). Violent crime and incentives in the long-run: Evidence from England and Wales. Journal of Applied Statistics, 38(4), 647–660. Saridakis, G. (2004). Violent crime in the United States of America: A time-series analysis between 1960–2000. European Journal of Law and Economics, 18(2), 203–221. Shaw, C., & McKay, H. (1942/1969). Juvenile delinquency and urban areas (rev. ed.). Chicago, IL: University of Chicago Press. Sutherland, E. H. (1942/1973). Development of the theory. In Schuessler (Ed.), Edwin Sutherland on analysing crime (rev. ed., pp. 30–41). Chicago, IL: Chicago University Press. Vold, G., Bernard, T., & Snipes, J. (2002). Theoretical criminology (5th ed.). Oxford University Press. Witt, R., Clarke, A., & Fielding, N. (1999). Crime and economic activity: A panel data approach. British Journal of Criminology, 39(3), 391–400. Witte, & Witt. (2002). Crime causation: Economic theories (pp. 302–308). Encyclopaedia of crime and justice New York: Macmillan. Witte, A. (1980). Estimating the economic model of crime with individual data. Quarterly Journal of Economics, 16, 69–85. Wolpin, K. (1978). An economic analysis of crime and punishment in England and Wales, 1847–1967. The Journal of Political Economy, 86(5), 815–840. Wooldridge, J. M. (2009). Introductory econometrics: A modern approach (4th ed.). South Western.