Does Court Efficiency have a Deterrent Effect on Crime? Evidence for Costa Rica

Does Court Efficiency have a Deterrent Effect on Crime? Evidence for Costa Rica Yuri Soares & Maria Micaela Sviatschi 1 This paper provides an empirical analysis of the impact of court efficiency on crime rates in Costa Rica. Using panel data on caseload clearance rates of first instance courts and crime rates for the period 2001-2007, we follow two identification strategies. First, we estimate a fixed effects model controlling for canton and year fixed effects. Second, we follow an instrumental variable approach that uses the rate of cases that are filed in the Courts of Appeal as exogenous variation for the first instance court efficiency. The main findings are that an increase in one percentage point of the court efficiency rate can reduce the number of crimes between 14 and 17 percent. This result is especially important in the context of Latin American countries, which account for the highest crime rates. 1Yuri Soares is Evaluation Economist at the Inter-American Development Bank s Office of Evaluation and Oversight (OVE). Maria Micaela Sviatschi is a PhD candidate in Economics at Columbia University.

I. Introduction In recent years, there has been growing awareness about the deterrent effect efficient judicial institutions can produce on crime. Theoretical papers have suggested how reductions of court trials and case backlogs can be effective in decreasing the number of crimes (Vereeck et al 2000; Torre, 2003; Listokin, 2004; Chappe, 2008). Most of these models are based on traditional crime theories (Beccaria, 1962; Bentham; 1963; Becker, 1968) that suppose that criminals internalize certain costs such as the probability of detention, punishment and conviction when they decide to commit a crime. They suggest that efficient courts can be important tools for increasing these probabilities and therefore the costs associated with committing a crime. In this paper, we provide an empirical analysis of the impact of court efficiency on crime rates in Costa Rica. In particular, we attempt to estimate the causal effect of court efficiency on crime rates, using detailed judicial caseload of first instance courts and criminal police data at the canton level from 2001 to 2007. We use caseload clearance rates as the measure of court efficiency. Following the theoretical literature, we expect to find a negative correlation between the efficiency of the courts and crime rates. In order to estimate a causal effect, we propose two different approaches. First, since there is variability over time and space of caseload and crime rates, we estimate a fixed effects model controlling for time and canton fixed effects. One important point to emphasize for identification is that in this case it is not necessary to control for judicial reforms or laws which may be simultaneously operating to improve efficiency and reduce crime rates, because most of them are national and therefore any effect arising from these laws would be captured by the time effect. 2 Second, we follow an instrumental variable approach based on Dalla Pellegrina (2006). We use the number of cases that are resorted to the Courts of Appeal as an instrument of the caseload clearance rates. The main findings are that court efficiency reduces the number of crimes between 14 and 17 percent. These results hold when testing different specifications. The validity of the causal interpretation of our estimates is tested restricting to the types of crimes where criminals might not follow the theory of rational choice (e.g. homicides and sexual offenses). It is found that while the efficiency of courts can be an important tool to decrease crimes where the logic of cost-benefit applies, it is uncorrelated with homicides and sexual offenses. These results are especially important in the case of Costa Rica where crime indicators have been increasing during the last decades. Juvenile delinquency rates and the average growth rate of domestic violence cases have increased in about 14.6 and 61.1 percent respectively during the period 1993-98. 3 In addition, the delinquency rate per 100,000 inhabitants has also increased in the last years (from 687.5 in 2002 to 1014.4 in 2007). This increase has been attributed to the lack of 2 Most of the changes were done at the court level and most of the cantons have courts that were reformed. 3 Inter-American Development Bank.

efficient judicial institutions able to investigate and solve penal processes in due time. In this sense, several efforts have been made by the government to improve the efficiency of courts in the last decade. One important reform was the Program for the Modernization of the Administration of Justice in 1996. The objective of this program was to improve the management of judicial offices and public services by building a modern system of administration that included the development of a computerized legal-data system. According to Soares et al. (2009), the reform was associated with an increase of 6 percent in clearance rates and with a reduction of 75 dollars per case disposed. This paper makes two important contributions. First, while there is an extensive theoretical literature that studies the causes and consequences of court inefficiencies, there is little empirical evidence that analyze the relationship between court delays and crime rates. The only empirical study found was that of Dalla Pellegrina (2006) that studies the impact of the length of trials on crime rates in Italy. Using an instrumental variables approach, the author finds that longer trials can reduce the deterrence effect of justice since criminals discount their punishment when there are trial delays. The paper suggests that a one-year delay can increase the number of thefts and robberies in about 18 and 23 percent respectively. Second, to the best of our knowledge there are no studies for Latin American countries, the region with the highest crime rates (see Table 1). According to Soares et al. (2007) mortality rates due to violence in Latin America are approximately 200 percent higher than in North America and Western Pacific, and 450 percent higher than Western Europe. In addition, surveys done by Latinobarometro in 2008 have shown that crime was considered the worst problem in the region, leaving behind unemployment and other economic problems. The rest of the paper proceeds as follows: Section II describes Costa Rica Judicial System and reforms, Section III presents the literature review, Section IV describes the data, Section V presents the estimation strategy, Section VI the results and Section VII some robustness checks. Finally, Section VII concludes. II. Costa Rica Judicial System The justice sector in Costa Rica is made up the Judiciary, the Ministry of Justice, The Office of the Attorney General of the Republic and the Community Defense Office. The Judiciary is composed of justices of peace, ordinary and appellate courts, and various chambers of the Supreme Court of Justice. The Supreme Court is the highest judicial body in Costa Rica, composed of 22 judges that are chosen by the Legislative assembly for an eight-year term. The appellate courts (tribunals) are divided in eight judicial districts: San José, Alajuela, Cartago, Heredia, Guanacaste, Puntarenas, Zona Sur and Zona Atlantica. The Courts of first judgment (juzgados) are the lower courts and are distributed over the 81 cantons. The Justices of Peace are local courts that handle minor cases (contravencionales y de menor cuantia) and are territorially spread over the districts in which vthe cantons are divided. The case of Costa Rica is especially important since, contrary to most Latin American countries, it has maintained a democratic tradition and a broad sense of social justice, providing its

inhabitants with a reliable and credible judicial system. According to Kauffman s governance indicators, Costa Rica is among the highest ranked Latin American countries in the Rule of Law Index. 4 In addition, the Judiciary is composed of 17 judges per 100,000 inhabitants, placing the country among the highest in the region for this indicator. Despite the fact that Costa Rica has good governance indicators in comparison with other Latin American countries, its judicial system has a high incidence of backlogs and congestion in courts that impede effective access to justice. At the beginning of the 1990 s, an ordinary commercial or civil case could take up to seven years, and a simple executive judgment around three years. 5 The potential causes of this inefficiency were attributed to the centralization of the administrative operations in San Jose, the lack of technology for administering resources, the formalism of judicial procedures, the inefficiency of the judicial offices and the lack of knowledge on behalf of the judges. As consequence, Costa Rica embarked on a judicial reform process that had broad support from different political actors such as the state powers, the private sector and civil society. The objective of this reform was to enhance access to justice and its efficiency, and mostly consisted of the following elements: the introduction of oral processes in criminal cases, the implementation of new criminal procedure, the adoption of alternative dispute resolution mechanisms, the establishment of unique case numbers across different actors in the criminal system, the implementation of information technology and administrative changes in the management of courts, including the creation of mega-offices which centralized many back office and administrative functions. In 1973, the country adopted hybrid (oral and written) procedures in criminal cases. The new system divided the process into two stages: the instruction and the oral and public trial. This change in the judicial process from written to oral procedure promoted judicial transparency and efficiency. In 1998, a new Criminal Procedure Code was also implemented that restructured the administration of the Judiciary and the Public Ministry, transferring the functions of investigation and prosecution of crimes to the latter. Another central component of reform was the implementation of the Law for Alternative Conflict Resolution Mechanisms, which included arbitration and conciliation as alternative ways for solving conflicts. This law was especially important in the case of Costa Rica, where citizens traditionally resort to courts in order to solve many conflicts that do not need to be assessed by judges, causing a high rate of litigiousness, congestion and a negative public perception about the administration of justice. 4 The same results are found with the indexes: Voice and Accountability, Political and Stability, and Government Effectiveness. 5 Inter American Development Bank (1994). Proposal for a loan for the modernization of the administration of justice

In 1998, the Supreme Court approved the creation of mega-offices that consisted of the clustering of services for courts of the same instance into a single office. The mega-offices introduced many technological improvements, such as computers, Internet access, Intranet and centralized notification services. There was also a specialization of functions achieved by separating the offices of family, child support and domestic violence. Another significant change was the creation of a unique number identifying each civil and criminal case (Número Unico de Expediente); this same identifier is used across the different actors in the legal system, allowing the monitoring of court processes. III. Literature Review Based on orthodox theories of crime used in economics (Beccaria, 1962; Bentham; 1963; Becker, 1968; Gibbs, 1968), we suppose that criminals are rational individuals that will decide to commit a crime whenever the benefits associated with the crime are higher than the costs. In this paper, we will focus on the role of costs C,. i t First, following this theory, we suppose that costs are an increasing function of the probabilities of conviction ( Cn, ), detention ( D, ) and punishment ( P, ). 6 i t i t i t C F Cn, D, P ) (1) i, t ( i, t i, t i, t Second, we suppose that court efficiency can affect Cn i, t and D i, t since it might increase the certainty and reliability of the legal system. This assumption is consistent with the previous theoretical literature. For example, using an expected utility model, Torre (2003) has analyzed how trial delays can affect the decisions of the defendant and prosecutor and found that these delays might reduce the probability of conviction. In the same vein, Vereeck et al (2000) have argued that case backlogs can reduce the criminal likelihood of losing a trial since the quality of evidence deteriorates with the passing of time. Thus, they argue that court delays might undermine legal certainty. Another relevant factor that might be affected by the efficiency of courts is P,. If criminals discount the future, court delays might also reduce the value of punishments, undermining the crime deterrent effect. Listokin (2003) has suggested that the existence of efficient courts might increase the present value of punishments since they are able to reduce the lag between when the crime is committed and when the punishment is imposed. In this sense, a punishment imposed today has a larger deterrent and retributive effect than the same punishment imposed in the future. For example, in monetary terms the same fine today does not have the same value that the same i t 6 According to the ecological models there are other factors that might influence the probability of committing a crime such as education, genetics, environment and social variables (poverty, inequality, low government etc.)

fine in the future. In addition, the author has also argued that the efficiency of courts not only increases the value of punishments but also the respect and reliability of the justice system. Similarly, Darley et al. (2003) have pointed out that the delay between crime and punishment can undermine the deterrent effect. In summary, all of the most important factors that affect the costs associated with committing a crime are positively related to the efficiency levels of the courts. Cn D, P H( CourtEfficiency ) (2) i, t, i, t i, t i, t Therefore, crime rates are an increasing function of the efficiency levels of courts: Cr F H( CourtEfficiency )) (3) i, t ( i, t We could also argue that the decision to commit an illegal behavior is a function not of the actual court efficiency rates but of the history of them. That is, criminals internalize in their cost function past court behaviors since they have more information about past efficiency levels than of the actual ones. Indeed, since an improvement on the actual efficiency levels can be a permanent or transitory shock, the criminal might not be able to distinguish between both types and thus he might prefer to consider past information about efficiency levels. Formally, Cr F H( CourtEfficiency )) (4) i, t ( i, t 1 Notice that we are also assuming that criminals have some degree of knowledge regarding the probabilities of being caught, being prosecuted and convicted and serving time, as well as knowledge regarding the duration of criminal cases. This assumption might not be unrealistic since criminals are prone to live among other criminals and thus might have more information about the performance of the Judicial System (Krohn, 1986; Dalla Pellegrina, 2006). Cook (1979) has analyzed three important channels by which potential criminals can get information about the certainty and severity of punishment: the media, the visible presence of enforcers, and by personal experience and observation. He has noticed that since offenders accumulate personal experience during their criminal lives, they have a better perception about the effectiveness of the system. Furthermore, he has pointed out that there are peer effects among criminals, since they also obtain information about the probability of detention from their criminal friends. In this paper, we will provide an empirical assessment of function (4). While there are many theoretical models, there is little empirical evidence that studies the impact of court delays on crime-related outcomes (except for Dalla Pellegrina, 2006).

Most of these studies have found that an increase in the certainty and severity of punishments can decrease illegal behavior (Gibbs, 1968; Waldo et al, 1972; Pogarsky, 2002). Soares et al (2009) have also suggested that the presence of judicial institutions in rural areas of Peru may have reduced the incentives to generate conflict, such as those related to land expropriations and debt collection, by increasing the probability of being punished. In the same line, many studies have analyzed how police presence can reduce crime rates. While early studies found a positive correlation between both variables (Cameron, 1988), they did not take into account the reverse causality problem that arises when police and crime rates are simultaneously defined. In order to address this potential endogeneity problem, Di Tella et al. (2004) take advantage of a natural experiment in Buenos Aires. They use as an exogenous allocation of police the fact that all Jewish institutions were given police protection after a terrorist attack and find that police presence can have a large deterrent effect on crime. They show that city blocks that have police protection experienced 0.081 fewer car thefts than those that do not have. While there is a widespread belief that the severity and certainty of punishment are important to deter criminal acts, there is disagreement in the empirical literature regarding the magnitudes of both variables. Many authors have pointed out that certainty of punishment is more effective in deterring crime than the severity of punishment (Witte, 1983; Title, 1969; Jensen; 1969). IV. Data The data is comprised of three different sets: judicial office-level data on clearance caseload rates, police-level data on crime rates and household level data for control variables. We aggregate the data at the canton level covering the provinces of San Jose, Alajuela, Cartago and Heredia during the period 2001-07. The judicial office-level data comes from the Planning Department of the Judiciary Power (Departamento de Planificacion del Poder Judicial) and contains data on the number of disposed, filed and pending cases, number of sentences, expenditures of courts, and number of judges in every canton for the period 2001-07. This data is used to construct our court efficiency indicator: the caseload clearance rate. We define the caseload clearance rate as the ratio between the number of disposed cases and the sum of new cases filed, re-filed cases and pending cases from the past year. The police level data also comes from the Planning Department of the Judiciary (Departamento de Planificacion del Poder Judicial) and contains information on the number of filed crime cases and the number of police officers at the canton level for the provinces of San Jose, Alajuela, Cartago and Heredia during the period 2001-07. We distinguish between nine groups: crimes against life, property, family, freedom and public duties, public administration, public faith, sex offenses, violation of the Psychotropic Substances Act and regulatory violation. Household level data was obtained from Costa Rica Household Surveys for the period 2001-06 and contains data on population, unemployment rates and poverty levels.

Table 2 presents summary statistics for the judicial indicators. We have a total of 280 observations and find variability across cantons and time for these outcomes. Table 3 presents summary statistics on the different types of crime. There are a total of 280 observations covering 40 cantons and variability across cantons and time. According to the crime shares in Costa Rica for the period 2001-2007, 77 percent of crimes were against property. These include thefts, swindles, robberies, frauds, usurpations and other crimes against property. Table 4 presents the evolution of crime rates per 100,000 inhabitants. Between the years 2001 and 2007, there was an increase in the number of total crimes. Most of this increase is due to crimes against property (6 percent) and crimes against family (15 percent). Table 5 presents the crime rates by court efficiency quantiles. Results suggest that those cantons that are at the higher distribution of efficiency present lower crime rates. If we compare the crime rates between those cantons that are in the lowest quantile to those that in the first, we find a difference of about 42 percent in the crime rates for cantons that have the more efficient courts. Nevertheless, it is important to notice that the bivariate relationship seen in Table 5 does not take into account the possible issues with endogeneity of court efficiency. These issues are detailed and addressed in the next section. V. Identification Strategy A. Fixed Effects Model The purpose of this study is to estimate the causal effect of court efficiency on crime rates at the canton level. Since this is not an experimental design, we have to address several endogeneity problems. First, we could have a reverse causality problem if court inefficiency is the result of a large number of criminal cases taken to them. Second, different particular characteristics in the cantons may have affected the crime rates and the productivity of courts. It could be the case that some cantons, for example those with more corrupt judges have higher crime rates and also more inefficient courts. 7 In this sense, correlation between court efficiency and certain factors such as corruption that influence the crime rates can lead to biased estimates. The same could also happen with the number of judges per canton since this variable might affect both crime and caseload clearance rates. In order to control for these potential biases that might confound our identification, we use panel data and estimate a fixed effects model. This specification eliminates potential biases that could come from those observable and non-observable characteristics that vary between cantons but are fixed over time, such as number of courts and police stations, sentencing procedures, geographical characteristics, etc. The model includes year-fixed effects that control for potential 7 Buscaglia et al (1999) have suggested that court delays might allow court personnel to ask for bribes in order to expedite the procedure.

common shocks in a given year and canton-fixed effects that control for all differences between cantons that are time invariant. Formally we estimate the following equation: Crime Rate i, t Clearance Ratei,t-1 X i, t Province j * t i t i, t (1) where Crime Rate i, t is the crime rate for canton i and year t; Clearance Rate i,t- 1 is the caseload clearance rate for court i and year t ; X i, t includes those variables that vary across cantons and time; is the time effect common to all the cantons; i is a canton fixed effect; Province j * t is an interaction of the province fixed effect and unobserved variation within province over time. The t the time fixed effect that controls for any X i, t vector includes variables like number of judges, police officers and other court officers, expenditures per court, poverty levels, unemployment rates and the number of sentences. We do not include a proxy for corruption since we can argue that although the corruption rates might differ between cantons, they might also be fixed in time as many institution related variables. The number of police officers, judges and court officers gives an idea of the expected probability of detention of criminals (Cameron, 1998; Levitt, 1995; Di Tella et al. 2004). One could argue that it is not the effect of efficiency which reduces crime but the effect of enforcers such as police officers and judges. These variables can be positively correlated with the efficiency of courts and negatively correlated with the crime rates. Poverty levels and the unemployment rates are used as proxies for macroeconomic conditions. First, cantons with lower poverty levels may tend to have better institutions and hence more effective courts. In addition, they may also tend to have lower crime rates. Ludwig et al (2001) have suggested that teenagers living in neighborhoods with lower poverty levels are less prone to commit crimes. In the same line, Land et al. (1990) and Kelly (2000) have found that income inequality has an important effect on violent crime. Second, several studies have also shown that higher unemployment rates are positively correlated with higher crime rates (Freeman, 1999; Raphael et al, 2001). It is important to note that by including this variable we are also indirectly controlling for educational attainment since it is correlated with unemployment. Finally, we also include the expenditures per court since we could argue that the most efficient cantons are the ones that spend more resources in courts. Many studies, such as Contini et al. (2007) have suggested that one of the most common factors attributed to the excessive duration of trials or inefficiencies is the lack of resources. The error is a canton time-varying error that is generally assumed to be independent across time and space; however, as the analysis uses panel data, the errors could be correlated across time in the same canton. In the case of a positive correlation, the standard errors could be computed smaller and the null hypothesis could be over rejected. To avoid potential biases in their

estimation, standard errors are clustered at the canton level, allowing an arbitrary covariance structure within cantons over time. 8 This is necessary since we do not observe variability in court efficiency within each canton. B. Instrumental Variables Despite the fact that in the fixed effects model we address many of the potential biases controlling for many factors that are fixed in time and vary between cantons, there are many unobservable factors such as the ability of judges that might vary across cantons and across time that can still affect our estimation. To address this issue, we use an instrumental variable approach in order to capture exogenous variability in the court efficiency variable. According to the theory, a proper instrument should satisfy two important assumptions. First, it should not be correlated with the error term and second, it should be significantly correlated with the efficiency of courts. Following Dalla Pellegrina (2007), we instrument the caseload clearance rates with the rate of cases filed in the courts of appeal. We define the probability to appeal a case as the ratio between the number of cases filed in the courts of appeal and the number of cases filed in first instances courts. According to national data in 2008 from 1390 crime cases per 100,000 inhabitants, approximately 50 percent are taken into first instances courts, from which 9 percent are taken to courts of appeal. 9 The mechanisms of transmission between caseload clearance rates and cases taken to the appeal courts are the following. First, if courts have a low clearance rate, the duration of cases is likely higher, since they only finish a small proportion of filed cases. As a consequence of this delay, the quality of evidence may deteriorate (Vereeck et al (2000) and Chappe (2008)) and therefore outcomes may not fulfill plaintiffs expectations. This in turn will lead to a higher probability of appeal. Second, court inefficiency could signal uncertainty about probability of specific outcomes in the case. This increase in uncertainty in turn may lead parties who are losing a case to believe that they would be more successful in an appeal. In this sense, the rate of cases filed in the Courts of Appeal is correlated with crime rates only through its relationship with the efficiency of first instances courts. On the other hand, the appeals rate might not affect the decision to commit a crime since under second instances defendants are already suffering the legal consequences for their crime. In this sense, the proportion of cases that are taken to the courts of appeal does not affect the probability of conviction, detention and punishments through other means that are not the efficiency rates of first instances courts. 8 See Bertrand et al. (2004). 9 In order to calculate these statistics, we compare the number of filed cases in courts to the number of crime cases reported to the police since we do not have a follow up on the crime cases that are taken to courts, nor the ones that are prosecuted.

VI. Results Column A in Table 6.1 presents the results for the basic model including only the clearance caseload rates, the fixed effects for each canton, and the year dummies. Cantons with higher levels of court productivity experience a reduction in crime rates. However, this reduction is not statistically significant. One concern regarding the fixed effects model is that some characteristics that vary across time and across cantons can be correlated with crime rates and court efficiency levels. To address this issue, column B includes a set of socioeconomic characteristics such as the unemployment rate, poverty variables, and province time trends. Once we include these controls, the estimated impact is increased and statistically significant. We find that an increase in one percentage point of the caseload clearance rate is associated with a reduction of about 14 percent in crime rates. Note that the population is not included as a control variable since all the variables are calculated on a per capita basis. Table 6.2 presents the results of the instrumental variable approach. The first observation is that that the parameter estimates for caseload clearance rates are remarkably stable, both across FE and IV specifications, and, more importantly, between specifications. Estimates for the model with police controls which is our preferred specification show a reduction in crime on the order of 1.3 percent for each 10 percentage point increase in the caseload clearance rate in the case of the IV specification. This reduction is also of 1.3 percent in the case of the FE specification. Also note that the introduction of the number of police slightly reduces the clearance parameter. This is to be expected, since characteristics of the police can arguably have a larger impact on deterrence than the efficiency of courts. Overall, these results suggest that court efficiency can be an important tool to decrease the crime rates. The results from the first stage of the instrumental variable model provide evidence that the efficiency of courts is significantly correlated (at one percent level) with the proportion of cases that are taken to the Appeal Courts. Indeed the sign of the correlation is consistent with the model presented in the previous section. An increase in the number of filed cases in the Appeal Courts is associated with the inefficiency of first instance courts. As a robustness check, we also follow the same identification strategy presented in section V but instead of using the past clearance rates, we use the actual caseload clearance rate. We find no significant change in the estimated effect suggesting that our previous estimates are correct. Indeed, we also estimate the fixed effect model including both the actual clearance rates and the ones of the year before. While we find that past clearance rates have a negative impact of 12 percent on crime rates, we find no impact of the actual clearance rates. This might support our argument that criminals have better information about past rates than of the actual ones. 10 10 Results are available upon request.

VII. Heterogeneous Crimes According to the transmission mechanism of court efficiency on crime rates, criminals internalize the higher costs associated with the efficiency of courts and thus following the theory of rational choice they decide to commit a crime when the benefits associated with it are higher than the cost. Nevertheless, it is important to mention that this argument might differ by type of crime. For some crimes, criminals might not act as rational decision-makers that decide to engage in illegal activity according to the expected utility associated with it. For example, in the case of homicides and sexual offenses it is difficult to argue that individuals might internalize this increase in the probability of being punished when they decide to commit a crime. Several studies have suggested that while property crimes might be lead by rational factors such as the desire for monetary gains and might be planned in advance, homicides and sexual offenses might be the result of emotional factors or special circumstances that affect the criminal (Borkowitz, 1993, Buvinic et al. 1999; Conroy, 2010). To address this issue, the fixed effects model presented in equation (1) is estimated for homicides and sexual offenses. In addition to check robustness of the results, the instrumental variable approach is also followed. If the estimation presented in the previous section is correct, the efficiency of courts should not operate affecting crimes such as homicides and sexual offenses. 11 Results from Table 7 suggest that efficient courts are not able to deter certain types of crimes that might be also the product of irrational factors. This suggests that any other plausible explanation for our main results can be eliminated and strengthens the causal interpretation. IX. Conclusions This paper suggests that increasing the efficiency of courts can be an important tool to decrease crime rates, supporting the theoretical literature. Since potential criminals are aware of the efficiency level of courts and the higher costs associated with them, they may decide not to perform illegal activities. The logic behind this mechanism is that when punishments are certain to happen and are imposed soon after the criminal act, individuals are more likely to be deterred from committing crimes. Using a combination of methods, we found that cantons with more efficient courts experienced, on average, a reduction of between 14 and 19 percent in crime rates. This result is consistent with the existent theoretical and empirical literature. This result is especially important in the context of Latin American countries, where crime is one of the most important public policy problems. In particular, this result is relevant in the case of Costa Rica, where crime rates have been increasing in the last several years (7.3 percent). 11 Previous results do not change if we exclude from the sample this type of cases. This could be due to the fact that most of the frequent crimes are against property where the deterrent logic makes sense.

Many factors suggest that the relation between the efficiency of courts and crime rates might be causal. First, the model includes canton, province and time fixed effects and the conclusions are robust to the inclusion of variables that may affect the probability of committing a crime and may also influence the efficiency of courts. Results are robust to controlling for differences in the province time trends. Second, results do not change substantially when we employ the instrumental variables approach. Third, we find no significant change in the estimated effect when we include the caseload clearance rates of the previous year instead of the current rates. Fourth, we demonstrate that the efficiency of courts has not affected those crimes that might be the result of irrational factors. These results shed light on a number of important policy debates regarding the effective tools to reduce crime rates. This paper suggests that increasing the efficiency of courts can have a significant effect on crime. Furthermore, Chappe (2008) has pointed out that there are many other social costs apart from crime that can be reduced by increasing the efficiency of courts such as: the lack of compensation for injured parties, the dearth of accurate decisions in individual cases, the deterrence of individuals who bring cases to the court, the decrease in the reliability and credibility of the justice system, the fall in the net value of punishments and the deterioration in the quality of evidence. In this sense, more attention ought to be paid to policies regarding court efficiency since these policies may have highly beneficial effects.

Tables Table 1. Homicide Rates per 100,000 inhabitants in the World. Average for 1990s Mortality due to Violence Latin America & Caribbean 21.8 North America 6.5 Western Europe 4 Form. Communist 17.2 Western Pacific 7.8 Source: Soares et al (2007) Table 2. Court Summary Statistics Mean Std. Dev. Min Max Pending Cases 7907.593 27122.66 87 181246 Filed Cases 13883.73 41874.42 240 312686 Re-filed Cases 319.2321 1167.157 0 9249 Disposed Cases 13019.1 38465.34 241 295645 Clearance Rate.5822525.0814032.3223887.806082 Note: Each mean was calculated taking into account the whole period (2001-2007). Table 3. Crime Summary Statistics at the Canton Level Std. Variable Obs. Mean Dev. Min Max Total Crimes 280 830.41 493.09 132.20 2848.34 Crimes Against Property 280 638.77 495.60 32.08 4286.66 Crimes Against Life 276 58.95 57.00 1.92 693.53 Psychotropic Substances Act violation 208 20.14 27.48 1.23 191.55 Sexual Offenses 260 26.80 19.57 1.89 204.76 Law Violation 212 16.96 22.03 1.23 156.22 Crimes Against Family 124 3.69 3.11 0.40 21.91 Crimes Against Freedom 172 6.45 6.16 0.77 52.64 Crimes Against the Public Admin. 219 16.84 16.93 1.31 110.76 Crimes Against Public Faith 214 29.49 40.95 0.45 274.11 Note: Each mean was calculated taking into account the whole period (2001-2007).

Table 4. Crime Rates per 100,000 Inhabitants 2001 2002 2003 2004 2005 2006 2007 % Total Crimes 780.35 766.99 846.36 848.35 866.15 864.05 841.08 7.78 Crimes Against Property 614.86 541.82 691.75 594.24 722.96 656.15 649.60 5.65 Crimes Against Life 55.42 53.66 80.36 58.31 59.48 55.64 50.79-8.35 Psychotropic Substances Act violation 28.39 23.07 28.37 20.72 17.68 13.97 10.68-62.38 Sexual Offenses 26.37 30.27 34.24 29.54 22.32 19.35 25.21-4.42 Law Violation 21.98 13.35 22.19 16.58 17.54 16.68 11.52-47.59 Crimes Against Family 3.40 3.23 3.62 3.43 4.31 3.92 3.91 14.97 Crimes Against Freedom 6.97 8.12 6.67 5.84 7.39 4.93 5.64-19.14 Crimes Against Public Faith 22.83 31.62 50.61 39.39 25.59 13.27 16.00-29.90 Court Efficiency Table 5. Crime Rates per Efficiency Quantiles Crimes Against Crimes Total Crimes Property Against Life Sexual Offenses Crimes Against Public Faith First Quantile 1049 825 76 29 34 Second Quantile 900 670 63 30 32 Third Quantile 807 634 60 30 32 Fourth Quantile 608 459 32 19 22 Note: Each mean was calculated taken into account the period 2001-2007

Table 6.1 Court Efficiency and Crime Rates (A) (B) (C) FE FE FE Caseload clearance rate -40.31631-102.27456** -104.64207*** (32.65263) (37.784) (36.920) % in crime rate 12 13 Percentage of households without income Percentage of households in extreme poverty Percentage of households with unsatisfied basic needs -988.50593-1,117.35260 (853.577) (941.169) -350.24856-369.42081 (405.040) (410.799) -95.36101-181.93079 (237.856) (244.636) Unemployment rate -1,036.66196-986.69309 (1,093.890) (1,233.465) Unemployment rate^2 949.47623 643.79367 (1,443.853) (1,558.789) Sentence rate -0.00034-0.00038 (0.001) (0.001) Court expenditure 6.94040 20.07500 (22.524) (55.019) Number of judges, police officers and other personnel -0.32855 (2.094) Canton Fixed Effects Yes Yes Yes Year Fixed Effects Yes Yes Yes Province Trends No Yes Yes Province Fixed Effects No Yes Yes Constant 932.7158 1,397.67705*** 1,306.43250*** (73.10723) (251.175) (342.262) Observations 239 189 179 R-squared 0.18 0.406 0.413 Number of Cantons 40 38 36 Note: Standard errors clustered at the canton level are in braces. * Statistically different from zero at the.1 level of significance. ** Statistically different from zero at the.05 level of significance. *** Statistically different from zero at the.01 level of significance.

Table 6.2 Court Efficiency and Crime Rates (D) (E) IV IV Caseload clearance rate -136.55536*** -104.86055** (48.596) (39.779) % in crime rate 17 13 Percentage of households without income Percentage of households in extreme poverty Percentage of households with unsatisfied basic needs 1,007.50113-362.64679 (2,451.700) (1,530.939) -95.24132 239.77853 (386.579) (278.774) 755.56324 845.71044* (685.588) (497.066) Unemployment rate -5,907.61584-2,703.93940 (3,947.771) (2,662.240) Unemployment rate^2 9,515.21885* 4,678.58837 (5,471.289) (3,772.803) Sentence rate 0.01104*** 0.00795** (0.004) (0.003) Court expenditure 115.88575*** 42.70948 (23.858) (35.367) Number of judges, police officers and other personnel 1.04279** (0.420) Canton Fixed Effects No Yes Year Fixed Effects Yes Yes Province Trends Yes Yes Province Fixed Effects Yes Yes Constant 1,660.85319** 1,051.65169* (816.481) (530.594) Observations 189 179 R-squared 0.623 0.774 Number of Cantons 38 36 Note: Standard errors clustered at the canton level are in braces. * Statistically different from zero at the.1 level of significance. ** Statistically different from zero at the.05 level of significance. *** Statistically different from zero at the.01 level of significance.

Table 6.3 First Stage Results (1) Appeal rate -53.444*** (9.628) Sentences rate 0.00008** (0.000) Court expenditure 0.19631 (0.233) Number of judges, police officers and other -0.00375* personnel (0.002) Percentage of households without income -18.94658 (15.276) Percentage of households in extreme poverty Percentage of households with unsatisfied basic needs -2.19763 (2.039) 8.83292** (3.771) Unemployment rate -20.29347 (19.617) Unemployment rate^2 29.69578 (24.998) Constant 3.95180 (4.063) Observations 180 R-squared 0.517 Note: Standard errors clustered at the canton level are in braces. * Statistically different from zero at the.1 level of significance. ** Statistically different from zero at the.05 level of significance. *** Statistically different from zero at the.01 level of significance.

Table 7. Deterrent Effect (A) 2001-2007 Mean Crime Rate per 100,000 (B) Estimated Impact Coefficients of the FE Model (C) Estimated Impact Coefficients of the IV Model Homicides 12.20-3.06032 1.487747 (3.499403) (1.783337) Sexual Offenses 26.50 5.191448-1.522643 (6.5934) (1.82414) Note: Standard errors clustered at the canton level are in parentheses. * Statistically different from zero at the.1 level of significance. ** Statistically different from zero at the.05 level of significance. *** Statistically different from zero at the.01 level of significance.

References Beccaria, C. (1963). Essays on crimes and punishments. Bentham, J. (1962). The Works of Jeremy Bentham. New York: Russel and Russell. Becker, G. (1968). "Crime and Punishment: An Economic Approach". The Journal of Political Economy. Vol. 76, pp. 169 217. Buscaglia, E and Dakolias, M. (1999). An Analysis of the Causes of Corruption in the Judiciary. World Bank publication. Buscaglia, E. (2001). Investigating the Links between Access to justice and Governance Factors: An Objective Indicator s Approach. Paper presented at the World Bank Conference on Justice, St. Petersburg, Russia. Buscaglia, E. and Dakolias, M. (1999). Comparative International Study of Court Performance. World Bank publication. Chappe, N. (2008). Court Congestion and the Decision to Litigate or Settle. SSRN working paper. Chiricos, T.G. (1987). Rates of Crime and Unemployment: An Analysis of Aggregate Research Evidence. Social Problems, Vol. 34, No. 2, pp. 187-212. Chiricos, T. G. and Waldo, G. P. (1972). Perceived Penal Sanction and Self Reported Criminality: A Neglected Approach to Deterrence Research. Social Problems, Vol. 19, pp. 522-540. Dalla Pellegrina, L. (2006). Courts Delays and Crime Deterrence (An Application to Crimes Against Property in Italy). SSRN working paper. Darley M. J., Robinson P. H., (2003). The Role of Deterrence in the Formulation of Criminal Law Rules: At Its Worst When Doing Its Best. Georgetown Law Journal. Fenn, P. and Rickman N. (1999). Delay and Settlement in Litigation. The Economic Journal, Vol. 109, No. 457, pp. 476-491.

Gibbs, J. P. (1968). Crime, Punishment, and Deterrence. Southwestern Social Science Quaterly, Vol. 48, pp. 515-530. Grasmick, H. G. and Bryjak G. J. (1980). The Deterrent Effect of Perceived Severity of Punishment. Social Forces, Vol. 59, No. 2, pp. 471-491. Inter American Development Bank (1994). Proposal for a loan for the modernization of the administration of justice. Inter American Development Bank (2003). Informe de terminación de operaciones. Costa Rica: Programa de Modernización Administración de Justicia. Jensen, G. F. (1969). Crime Doesn t Pay: Correlates of a Shared Misunderstanding. Social Problems, Vol. 17, pp. 189-201. Kessler, D. (1996). Institutional Causes of Delay in the Settlement of Legal Disputes. Stanford University and National Bureau of Economic Research, Vol. 12, No. 2. Krohn, M.D. (1986). The Web of Conformity: A Network Approach to the Explanation of Delinquent Behavior. Social Problems, Vol. 33, No. 6, Special Theory Issue, pp. S81-S93. Listokin, Y. (2004). Crime and (with a Lag) Punishment: Equitable Sentencing and the Implications of Discounting. SSRN working paper. Muhl M. and Vereeck, L. (2000). An Economic Theory of Court Delay. European Journal of Law and Economics, Vol. 10, No. 3, pp. 243-268. Nagin, D. and Pogarsky, G. (2003). An Experimental Investigation of Deterrence: Cheating, Selfservicing Bias and Impulsivity. Criminology, Vol. 41, No. 1, pp. 167-194. Pogarsky, G. (2002). Identifying deterrable offenders: implications for research on deterrence. Justice Quarterly, Vol. 19, No. 3, pp. 431-452. Raphael S. and Winter-Ebmer R. (2001). Identifying the Effect of Unemployment on Crime. Journal of Law and Economics, Vol. 44, pp. 259. 283. Soares, R. and Naritomi, J. (2007). Understanding high crime rates in Latin America: the role of social and policy factors. Mimeo, PUC-Rio.

Tittle, C. R. (1969). Crime Rates and Legal Sanctions. Social Problems, Vol. 16, No. 4, pp. 409-423. Torre A. (2003). The Impact of Court Delays on the Prosecutor and the Defendant: An Economic Analysis. European Journal of Law and Economics, Vol. 16, No. 1, pp. 91-111. Torre (2008). Socially Optimal Criminal Court Waiting Times: A Partial Investigation. Australian Economic Papers. Urban, Lynn S. (2009). Identifying Deterrable Offenders in a Sample of Active Juvenile Offenders. The Southwest Journal of Criminal Justice, Vol. 6(1). Witte, A. D. (1983). Economic Theories. Encyclopedia of Crime and Justice, Vol. 1. Wright, B. R.E., Caspi, A., Moffitt, T.E. and Paternoster, R. (2004). Does the perceived risk of punishment deter criminally prone individuals? Rational choice, self-control, and crime. Journal of Research in Crime & Delinquency, Vol.41, No. 2, pp. 180-214.