WPS6768 Policy Research Working Paper 6768 Economic Growth and Crime against Small and Medium Sized Enterprises in Developing Economies Asif Islam The World Bank Development Economics Global Indicators Group February 2014 Policy Research Working Paper 6768 Abstract Several studies have explored the relationship between economic growth is negatively associated with crime. economy-level crime rates or individual-level crime This relationship is stronger for small and medium firms and economic growth. However, few studies have than large firms. The study also explores several economy- examined the relationship between economic growth wide factors and their influence on the growth-crime and crime against firms. This study uses data for about relationship for small and medium enterprises. The 12,000 firms in 27 developing countries and finds that results are robust to various sensitivity checks. This paper is a product of the Global Indicators Group, Development Economics. It is part of a larger effort by the World Bank to provide open access to its research and make a contribution to development policy discussions around the world. Policy Research Working Papers are also posted on the Web at http://econ.worldbank.org. The author may be contacted at aislam@worldbank.org. The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. Produced by the Research Support Team Economic Growth and Crime against Small and Medium Sized Enterprises in Developing Economies Asif Islam Enterprise Analysis Unit Development Economics The World Bank, Washington DC, 20433 Email: asif.m.islam@gmail.com Tel: 651-246-4017 JEL Classification: K40, O10, O40, O50 Keywords: Crime, Firms, Economic Growth, Development Economic Growth and Crime against Small and Medium Sized Enterprises in Developing Economies 1 Introduction A distinction exists between crime experienced by firms and households. The former typically face both a higher risk and greater damage from crime than the latter (Hopkins 2002). Within the developing world context, rising crime experienced by firms can have direct consequences. The already vulnerable business climate in developing economies can further deteriorate. Crime may also retard the growth and performance of small and medium firms. Given that developing economies tend to be plagued by corruption and limited public resources to control crime, a key question is whether the burden of crime, especially for small and medium sized enterprises (SMEs), will be alleviated as developing economies grow. The relationship between crime experienced by firms and economic growth has been under-researched. While some studies analyze specific aspects of crime and corruption in SMEs such as tax payments (Baliamoune- Lutz and Garello, 2014) or how the quality of institutions shapes the performance of SMEs in developing countries (Aidis et al., 2012; LiPuma et al., 2013), this study fills the gap in the literature with a specific focus on how economic growth directly affects crime, such as robbery, vandalism, and theft, experienced by small and medium enterprises in developing economies. A fairly substantial theoretical and empirical literature already exists exploring the effect of economic growth or labor market conditions on crime (Fajnzylber et al, 2002; Donohue and Levitt, 2001; Gould et al, 2002; McDowall and Loftin, 2009; Bushway et al., 2012). Typically, most empirical studies on crime in developing economies either use cross-country crime data 2 from police reports or focus on household crime data in a specific city or country. While both types of studies are useful, they also face a few challenges. The limitation of using police reports is that given the low level of confidence in police especially in developing countries, crime tends to be under-reported in police reports (Soares, 2004). The limitation of the household crime studies is that they omit crime faced by firms, which tends to be more frequent and cause greater damage. These studies are also limited to the incidence of crime faced by households, which does not account for the financial costs incurred by the victim. Furthermore, although household level crime studies may have a lower degree of under-reporting of criminal activity than police reports, there is still the difficulty of drawing general policy inferences that are based on idiosyncratic country data. This is a particularly difficult obstacle to overcome as even if various household crime surveys in different countries were pooled together, there is still the issue of differences in sampling and methodology of how the surveys were undertaken. Finally, most of these household surveys are unable to address time invariant omitted variables as the data are typically available for one period of time. By using firm-level survey data that are comparable across economies as they use the same sampling methodology, we are able not only to explore the impact of per capita economic growth on crime, but also to overcome most of the obstacles faced by studies based on household surveys. This study contributes to the literature by making several improvements: (i) This study explores the effect of economic growth defined as real GDP per capita on both the incidence of crime and the burden of crime for firms in developing economies using firm surveys. The focus of this analysis is on small and medium sized firms but we also mention whether the results hold for a full sample of small, medium, and large firms. This is a vast improvement on the literature which 3 has mainly explored household-level crime. (ii) A unique firm level data set with about 12,000 firms in 27 developing countries maintained by the World Bank’s Enterprise Analysis unit (Enterprise Surveys) is used that follows a consistent methodology across surveys and therefore allows for cross comparisons between economies. This alleviates the external validity issue faced by most survey level studies on crime. Our results indicate that economic growth may actually reduce crime in developing countries. We find that an increase in real GDP per capita growth by 1 percent is associated with a 0.30 percent reduction in the losses due to crime for firms of all sizes in developing economies. More importantly, we find that this relationship is stronger for small and medium firms than large firms. That is, a 1 percent increase in real GDP per capita growth is associated with a 0.33 percent reduction in losses due to crime for small and medium firms as opposed to a 0.21 percent reduction for large firms. We also find that several macro-variables tend to amplify or reduce the effect of economic per capita growth on crime experienced by small and medium sized firms. Despite the improvements of this study over the literature in several aspects, there are some issues that need to be addressed which are typical of studies that are cross-country in nature. There is still the issue of time varying omitted variables, as well as the issue of reverse causality. Addressing these endogeneity issues is challenging given the data at hand, but we try and limit them in the following manner. We limit omitted variable bias by checking if our specification is robust to additional potential explanatory variables including ethnic fractionalization, security costs, corruption, human capital, government spending, and trade. In order to limit endogeneity, we also use the percent of land in the tropics and a malaria risk index as instruments and find that the results are retained, if not magnified. 4 Given the data limitations, it is impossible to completely justify a causal link. Thus, at worst, this study quantifies the correlation between economic growth and crime against firms, and then explores this relationship further in terms of several socio-economic and firm characteristics. Our results are robust to several other sensitivity checks. Mechanisms Crime can take many forms, with different implications. For this study we have data available for property crime, which mainly includes theft, robbery, vandalism and arson. Several studies have explored the mechanisms of the effect on economic growth and crime. In general, economic growth increases job market opportunities, consequently raising the opportunity cost of individuals engaging in illegitimate activities (Becker, 1968). Along similar lines, per capita economic growth may increase employment, further decreasing the rewards for criminal activity. Generally, unemployment has been empirically associated with increases in criminal activity (Machin and Meghir, 2000; Donohue and Levitt, 2001). Furthermore economic growth may increase tax revenues resulting in higher government budgets for crime prevention (Cook, 2010; Levitt, 2004). Both mechanisms propose an inverse relationship between economic growth and crime. However, economic growth also serves as an indicator of increasing prosperity and thus the effect on crime may depend on the level of risk aversion (Ehrlich, 1973). Increasing prosperity may increase the number of options encouraging crime-prone behavior such as individuals leaving their houses more often, increasing alcohol consumption, or owning more cars. Furthermore economic growth can increase the reward for committing a crime as the society gets wealthier. Finally legal institutions, being typically slow to adapt, may find it difficult to deal with rapid economic growth and thus provide more leeway for criminals to 5 escape punishment, therefore reducing the costs of engaging in illegitimate activities. In general, empirical findings have indicated a negative relationship between economic growth and crime (Fajnzylber et al, 2002; Bourguignon, 1999; Dutta, 2009). A few mechanisms translate well with regards to crime against firms. Economic growth resulting in increasing economic opportunities of individuals makes firms less susceptible to theft, and rising government spending on crime prevention may complement a firm’s security measures. However, in terms of risk prone behavior induced by economic growth, it is unlikely that alcohol consumption or increasing outings by individuals is highly correlated with crime against firms, although there may be increases in firm spending on expensive property across all sizes of firms that may attract criminal activity. Economic growth may also increase the number of firms of various sizes, therefore increasing criminal activity given the larger number of options available to criminals. It is expected that the rate at which legal institutions adapt to increasing economic growth will determine the level of crime against firms. A fast growing economy may not have the ability to cope with rapidly increasing crime. Furthermore, economic growth may allow firms to spend more on security measures, thus reducing the losses due to crime. However, we do find later in this study that controlling for security costs does not alter our main results. The relationship between economic growth and crime experienced by firms may also depend on the size of the firm. On one hand, economic growth may increase the ability of larger firms to better protect themselves from theft and vandalism than small and medium firms which may lead criminals to target small and medium enterprises. A key presumption of this mechanism is that the benefits of economic growth are largely absorbed by large firms. On the other hand, 6 economic growth can increase the opportunities for small and medium firms to expand relative to large firms that may have reached their optimal size. Such an expansion may allow small and medium firms to protect themselves better, thereby reducing the degree of crime they face. Given the wealth of theoretical mechanisms available in the literature, the objective of this study is to provide empirical verification of the effect of per capita economic growth on crime experienced by firms. Depending on which mechanisms dominate, we may have a positive or negative association between crime and economic growth. We structure the rest of the paper as follows. Section 2 describes the data, section 3 provides the estimation and results, and sections 4, 5, and 6 provide instrumental variable estimations, robustness checks, and conclusions, respectively. 2 Data and Main Variables The data for firm level variables are collected by the World Bank’s Enterprise Surveys (ES). The Enterprise Surveys use standard survey instruments to collect firm-level data on a country's business environment from business owners and top managers. The surveys cover a broad range of topics including access to finance, corruption, infrastructure, crime, competition, labor, obstacles to growth and performance measures. The survey is designed to be representative of a country’s private non-agricultural economy and only registered firms with at least five employees are included in the sample. The data consist of a random sample of 12,000 firms across 27 developing countries in different regions stratified by firm size, location, and sector using a consistent methodology. The survey year ranges between 2007 and 2009. Police data are 7 obtained from the United Nations Survey on Crime Trends and the Operations of Criminal Justice Systems. This police data set consists of an unbalanced panel covering years between 2003 and 2008 for both developing and developed economies. Due to the unbalanced nature of the data set, only about 39 countries match between the ES firm level data and the UN country level police data. Accounting for issues such as missing questions and incorrect data, the sample size reduces to a cross-section of 27 developing economies. Details of countries in the main sample and their respective survey years can be found in the first column of table A3 in the appendix. Data sources and descriptions of the variables can be found in table A1, with summary statistics in table A2. 2.1 Dependent Variable The main dependent variable utilized is annual losses due to crime as a percentage of annual sales. This variable is derived from the survey question: “In fiscal year [insert fiscal year], what are the estimated losses as a result of theft, robbery, vandalism or arson that occurred on the establishment’s premises calculated as a percent of annual sales?” In some instances the total value of losses is reported, and in that case the share of losses due to crime over total sales is manually calculated. Crime losses as a percentage of sales capture the intensity of crime. We make no distinction between a firm that has experienced no crime and a firm that has experienced crime but incurred no losses. Both firms get a zero value for the dependent variable. This adjustment may not affect our estimations much since we retain the same results when we examine the impact of per capita economic growth on the incidence of crime (see next paragraph). The crime losses variable averages 0.72% in the sample with a standard deviation of 8 3.9%. Using country averages across all firms, Azerbaijan has the lowest amount of crime losses at 0.20% of sales, while El Salvador has the highest with 1.73%. We also estimate a model where the dependent variable is a dummy variable indicating whether or not a firm has experienced crime in the last fiscal year. The survey question is phrased as follows: “In fiscal year [insert last complete fiscal year], has this establishment experienced losses as a result of theft, robbery, vandalism or arson?” This estimation elucidates the relationship between economic growth and the incidence of crime as opposed to the burden of crime. 2.2 Explanatory Variable The main variable of interest is annual real GDP per capita growth rate, which is available for the years the firm surveys were undertaken. The sample mean for the rate of economic growth is 2.66% and the standard deviation is 4.2%. Belarus and Romania are the fastest growing economies in the sample with growth rates of 10.2% and 9.6% respectively. The country level average of crime losses and economic growth rates are presented in table A4 for each country in the appendix. We also use 5 year averages (2002-2007) of real GDP per capita growth and find that the results are retained. 2.3 Other Explanatory Variables We control for several firm level and country level variables. The degree of crime a firm faces may depend on its size, the sector it belongs to, and its locale (Amin, 2009; Glaeser and Sacerdote, 1999). The costs and benefits of criminal activity targeting firms may vary depending 9 on sector or size of the firm. Big cities tend to provide higher pecuniary benefits for criminals and also criminal activity may involve lower costs due to the low probability of arrest and higher probability of anonymity. We control for firm size using dummy variables for small and medium firms. A small firm is defined as a firm with fewer than 20 employees, while a medium firm has between 20 and 99 workers. Of course, when we examine the effect of economic growth on crime for small and medium firms, we only control for small firms using a dummy variable as we restrict the sample to only small and medium firms. We control for sector or industry by using a dummy variable for manufacturing firms. A priori it is not clear whether a manufacturing firm should have higher crime losses with respect to other sectors. We include a city size dummy variable that takes the value of 1 if the city has a population of 250,000 or greater, or is a capital city, and 0 otherwise. There is also evidence that female owned households as well as female owned firms have a positive association with higher crime rates (Glaeser and Sacerdote, 1999; Islam, 2013; Kelly 2000). We also include a dummy variable for whether a firm has at least one female owner. All these variables are from the Enterprise Survey’s data set. We omit security costs from the base estimations due to concerns of simultaneity bias between crime losses and security costs, however later in the robustness section we do add it to the base specification and we find that the base estimation results are unaffected. At the country level we control for the number of police per 100,000 population. The variable is obtained from the United Nations Survey on Crime Trends and the Operations of Criminal Justice Systems. The variable is typically lagged by 1 year to reduce potential simultaneity bias between crime losses and police presence. Due to data constraints, for some countries the variable is lagged by more than one year. The specific details of the number of lags are available 10 in table A3 in the appendix. We also control the level of development of the economy using real GNI per capita. Inequality is controlled for using the country-wide Gini index. Both variables are prominent in the literature (Fajnzylber et al., 2002; Soares, 2004; Demombynes and Ozler, 2005). Increasing inequality typically increases the proportion of poor who may engage in criminal activity as well as the greater reward for engaging in criminal activity. For cases where data for the exact date are unavailable, we use data for the closest date available. The exact years of data used for GNI per capita and the Gini index are presented in table A3 in the appendix. We finally control for country size using the total population of the country. These data are available from the World Bank’s World Development Indicators. Data source and description of the variable can be found in table A1, with summary statistics in table A2. 3. Estimation We estimate the following equation (1) using OLS when the dependent variable is crime losses over sales, and Logit when the dependent variable is a dummy variable indicating whether a firm experienced crime in the last fiscal year. (1) crimeij= β1GDPgrj + β 2 Policelag j + β3 Femij + β 4GNIcap j + β5GINI j + β 6 Population j + β 7 Smallij + β8 Mediumij + β9 LargeCityij + β10 Manfij + ε ij Where crime is the losses due to crime as a % of sales or a dummy variable representing whether a firm experienced crime in the last fiscal year, GDPgr is the real GDP per capita growth rate, Policelag is the lagged number of police per 100,000 population, Fem is a dummy representing female ownership, GNIcap is the real Gross National Income per capita, GINI is the Gini coefficient, Population is the total population of the economy, Small and Medium are firm size 11 dummies, LargeCity is a dummy variable for cities with population of 250,000 and greater, or capital cities, and finally Manf is a dummy variable for manufacturing firms. When we restrict the sample to only small and medium firms, we omit the medium firm size dummy variable. There is, of course, some concern about whether this specification adequately captures the economic growth-crime relationship. We include several other control variables, including security costs, as a specification test in the robustness section and find that they do not affect the results. All estimates are based on standard errors clustered at the country level. In the later sections we add additional variables and interact them with the variable of interest to elucidate several relationships between these variables, economic growth, and crime faced by small and medium sized firms. The usual econometric issues of endogeneity and omitted variable bias are of concern. Reverse causality is an issue given that higher crime can lead to slower economic growth. We also have the issue of omitted variable bias as economic growth may be correlated with several factors. Both issues are challenging given data limitations. We address this by using various checks as presented in the robustness section including instrumental variable estimation and panel estimations using a panel data set. 3.1 Base Regression Results The base results are presented in table 1. Column1 indicates that firms in faster growing economies are less likely to experience crime. Column 2 shows that an increase in GDP per capita growth by 1 percent is associated with a 0.30 percent reduction in the losses due to crime 12 as a percentage of sales. 1 Both results are significant at 1%. We repeat the estimations in column 1 and 2 with the restriction that the sample includes only small and medium firms. As shown in columns 3 and 4 table 1, the results are largely the same. The sign and significance of the estimation results in columns 1, 2, 3 and 4 of table 1 are unchanged when all controls are excluded. A 1 percent increase in economic growth is associated with a reduction in crime loss as a percentage of sales by 0.24 percent with a 1 % level of significance for the whole sample of small, medium and large firms when all controls are excluded from the estimation. When restricting the sample to just small and medium firms, the corresponding figure is a 0.25 percent reduction in crime loss as a percentage of sales. These results indicate that economic growth may be a deterrent for criminal activity against firms. One reason could be that economic growth increases the opportunity cost of criminal activities leading to less crime. This is consistent with studies for individual level crime (Bourguignon, 1999; Fajnzylber et al., 2002). A few other results stand out. Small and medium firms are less likely to experience crime than large firms, but face higher losses due to crime as a percentage of sales, which has been also found in previous studies (Amin, 2009). Firms with female owners are both more likely to experience crime, and also face a higher percentage of losses due to crime over sales. A couple of results retain significance but switch signs when comparing the impact on likelihood of experiencing crime (incidence of crime) and the losses due to crime as a percentage of sales (burden of crime). GNI per capita is positively associated with the likelihood of experiencing crime while negatively associated with the losses due to crime as a percentage of sales. That is, with higher income, crime against firms may become more widespread but also more petty. 1 We replicated the estimations using a 5 year average of real GDP per capita growth. The results were retained and are available upon request. 13 Income per capita has generally shown non robust results with regards to individual level crime in the literature (Fajnzylber et al., 2002; Soares, 2004). 2 The size of the country in terms of population is positively associated with losses due to crime as a percentage of sales, but no significant association with the incidence of crime. Thus, in larger economies, firms may experience higher losses due to crime. We now consider how several country and firm-level factors may influence the relationship between economic growth and the burden of crime on firms, particularly for small and medium firms. We examine the strength of the relationship between economic growth and crime and how it depends on factors such as firm size, police, female ownership and management, governance and voter turnouts in elections. The effect of economic growth on crime may differ depending on the size of the firm, as the ability for firms to protect themselves from criminal activity and their attractiveness as potential targets of criminal activity vary by firm size. Economic growth may not only increase labor market opportunities, but may also capture the quality of legal and public institutions (Lederman et al., 2002). The effects of police and governance on crime have been explored in the literature, and thus we expect the relationship between economic growth and crime may be strengthened or weakened by these factors. We also use voter turnout as a proxy for social organization, and thus we are able to examine whether the effect of economic growth on crime losses is strengthened by social organization, or weakened by its presence. Finally, given the robust positive relationship between female ownership and management and crime, which is consistent with empirical relationship between female headed households and crime, we examine whether economic growth weakens or strengthens this relationship. We do 2 Due to the lack of continuous data for GNI per capita for some countries, we use the GDP growth rate instead of the GNI growth rate as our indicator for economic growth. This is typical in the literature (Fajnzylber et al, 2002). 14 also comment on whether these relationships hold when considering the whole sample instead of the restricted small and medium firm sample, and by and large this seems to be the case. We use the OLS estimation results in table 1, column 4 as the base results for the small and medium firm sample as with OLS estimations there is a more straightforward interpretation with regards to the magnitude of the effects of interaction variables. However, it is worth noting that both the Logit and OLS estimations provide qualitatively similar results for economic growth 3.2 Small and Medium Firms and per Capita Economic Growth Interaction Column 1 of table 2 presents the estimation results of the interaction between small and medium firms and economic growth using the full sample of small, medium, and large firms. The interaction is negative, implying that economic growth reduces crime losses for small and medium firms to a greater extent than large firms. A one percent increase in per capital economic growth is associated with a 0.33 percent decline in crime losses for small and medium firms, statistically significant at the 1% level. The corresponding figure for large firms is 0.21 percent, statistically significant at the 5% level. The implication is that an increase in per capita economic growth is far more beneficial for small and medium firms than large firm with regards to crime. Thus pro-growth policies may be one way of insulating small and medium firms from crime. 3.3 Number of Police and per Capita Economic Growth Interaction The interaction term between the number of police and economic growth is positive and significant at the 5% level of significance for the sample of small and medium sized firms. The overall association of economic growth with crime losses for small and medium sized firms is still negative at the sample mean number of police after interacting economic growth with police. 15 However, looking at the extremes of the sample, economic growth has no significant association with crime losses at the sample maximum number of police, but has a large and significant negative association at the sample minimum number of police. The results are presented in column 2 of table 2. The effect of increasing economic growth by 1 percent is associated with a 0.27 percent reduction in crime losses at the sample mean number of police and 0.48 percent reduction in crime losses at the lowest number of police in the sample, both results statistically significant at 1%. The number of police turning point, after which the association between economic growth and crime losses is insignificant, in the sample is 430 per 100,000 of population, which is at the 88th percentile of the sample. These results are retained when we use the full sample of small, medium, and large firms. The implication of these results may indicate that a larger police force may be a substitute for the economic growth when it comes to reducing crime against firms. Thus during times of low economic growth and potentially high levels of crime, expanding the police force may be a viable alternative. 3.4 Female Ownership, Female Management and per Capita Economic Growth Interaction The interaction term between female ownership and management and economic growth is positive and statistically significant at 10% for the sample of small and medium sized firms. Economic growth is associated with a reduction in crime losses whether or not a firm has a female or male owner, but the magnitude of the association is larger if the firm has a female owner, or both a female manager and at least one female owner as shown in columns 3 and 4 of table 2. The results indicate that a 1 percent increase in economic growth is associated with a 0.34 percent reduction in the dependent variable for female owned firms in contrast to 0.24 percent reduction in crime losses for firms that do not have a female owner. The magnitude is 16 even greater for firms with both a female owner and female manager at 0.44 percent reduction in contrast to 0.24 percent reduction for firms that have all male owners and male managers. These results are largely retained when using the full sample that includes large firms. An interpretation of this result is that economic growth may benefit the more vulnerable or less well off in a society and female owned or managed firms are less susceptible to crime. 3.5 Governance and per Capita Economic Growth Interaction We use ICRG’s Quality of government indicator as a measure of governance. The interaction term between Quality of Government and Economic growth is positive and statistically significant at 5 % as shown in column 1 of table 3. Economic growth has a negative and significant association with crime losses experienced by small and medium sized firms at the sample mean of the governance indicator. However, this relationship turns insignificant at the sample maximum of the governance indicator. At the sample minimum level of governance, the association of economic growth with crime losses has a larger magnitude while retaining significance, when compared to the mean level of governance. A 1 percent increase in economic growth is associated with an approximately 0.28 and 0.43 percent reduction in crime losses for the sample mean and minimum quality of governance respectively. These results are largely retained when extending the sample to large firms. One interpretation is that better governance is a substitute for the crime-reducing effect of economic growth. 3.6 Voter Turnout and per Capita Economic Growth Interaction Social disorganization theory indicates that factors that diminish the effectiveness of informal social controls increase criminal activity (Kelly, 2000). Here we consider one indicator of social 17 disorganization – voter turnout in elections. The interaction term between voter turnout and economic growth is positive and statistically significant at 1% as shown in column 2 of table 3. An increase in economic growth is significantly associated with a decrease in crime losses experienced by small and medium sized firms at the mean level of voter turnout. However, this relationship loses significance at the sample minimum of voter turnout, but actually has a larger magnitude and retains significance at the sample maximum of voter turnout. A 1 percent increase in economic growth is associated with a 0.23 percent reduction in crime losses at the sample mean level of voter turnout and a 0.59 percent reduction in crime at the minimum level of voter turnout, both results statistically significant at 1%. The turning point where the effect of an increase in the economic growth ceases to be significant is a voter turnout above 0.8, which is at the 88th percentile of the firm sample. These results are largely retained when the sample is extended to large firms. This result indicates that a higher voter turnout, possibly indicating greater social organization, is a substitute for the effect of economic growth on crime. Two major concerns about the estimation results are that the relationship between economic growth and losses due to crime are reverse causality and omitted variable bias. Essentially, an increase in losses due to crime can deter economic growth which would imply that economic growth is correlated with the stochastic error term, hence biasing the estimates. Similarly, economic growth may be capturing something else such as institutions that are not included in the estimations, thus producing biased and inconsistent estimates. We mitigate this possibility by using instrumental variable estimations and subjecting the base estimates to several robustness checks. 18 4. Instrumental Variables We use two instruments for economic growth: the % of land in tropics for each country, and malaria risk index, which is the proportion of each country’s population that live with risk of malaria transmission in 1994 developed by Sachs and Gallup (2004). Both these variables have been found to be an important determinant of economic growth (Sachs et al, 1998). We report the results in table 4 for the sample of small and medium sized firms. The first stage estimation results presented in column 2 shows that while % of land in tropics is significant, malaria ecology is not. We include the latter variable in order to pass the test of overidentifying restrictions. The overall results are retained without the malaria risk instrument. The second stage estimation results are presented in column 1 of table 4. The coefficient of real GDP per capita growth using instrumental variables estimation method retains the sign and significance of the base estimations. The magnitude increases slightly from -0.08 in the base estimations to -0.14 in the estimations using the instrumental variables. As indicated in the bottom of table 4, the Sargen-Hansen test of overidentifying restrictions is not rejected for all conventional levels of significance. Thus we cannot reject the null that all the instruments are valid. We also reject that the estimation is underidentified at 5% level of significance. We replicate the instrumental variable estimations in table 4 by extending the sample to large firms. The results are largely retained. However, the results should be treated with caution. It is possible that the instruments may not satisfy the exclusion restriction. Geographical factors such as % of land in tropics may be correlated with institutional factors or other unobservables that may be correlated with crime. 19 Thus the potential endogeneity of our instruments cannot be ruled out completely, and these results should be interpreted as a robustness check. 5. Robustness We check for robustness in terms of model specification. Several studies have shown that demographics and human capital (Kovandzic and Sloan, 2002; Kelly, 2000; Di Tella and Schargrodsky, 2004; Usher, 1997), corruption (Gaviria, 2002), fractionalization (Fajnzylber et al, 2000), trade (Ghosh et al., 2011), prison population (Kovandzic and Sloan, 2002; levitt, 2004), and government spending (Naidoo, 2006) are factors that influence crime. We add sets of variables that proxy for the aforementioned factors and present the results for small and medium sized firms in table A5. We also include firm level variables such as total employees and security costs as a % of total sales and check if they affect our estimation of interest. We also control for inflation, using it as an indicator of macroeconomic stability. As shown in table A5, including these sets of variables does not improve the goodness of fit, the explanatory power of the model, or eliminate the statistical significance of the variable of interest. There is a marginal improvement in the goodness of fit when considering a continuous measure of firm size and security spending. However, the significance of the coefficient of interest is unaffected. We also worry that extreme observations in the sample may be driving the results. Thus we omit the top 1%, bottom 1 %, and top and bottom 1% observations of losses due to crime as well economic growth, to see if our results are retained. As indicated in table A6, the significance of our results is unaffected. 20 Finally, we worry that our results may be dominated by certain countries. Thus we drop one country at a go from the sample of small and medium sized firms and see if the coefficient of economic growth retains the signs and statistical significance. As indicated by figure A1, the results are not dominated by any particular country in the sample as we retain the sign and significance at 5%. 6. Conclusion This paper contributes to the literature by examining the relationship between economic growth and crime against firms, a fairly under-researched area especially with regards to small and medium sized firms. The paper finds a negative relationship between firm losses due to crime and economic growth. We find that an increase in real GDP per capita growth by 1 percent is associated with a 0.30 percent reduction in the losses due to crime as a percentage of total sales experienced by firms. This figure is larger for small and medium firms (0.33 percent) than larger firms (0.21 percent). The suggested mechanism for this effect is that economic growth increases opportunities elsewhere and thus increasing the opportunity cost of crime. Furthermore, economic growth may result in small and medium firms growing faster and increasing performance, thus allowing them to better protect themselves from criminal activity. The results of this study are consistent with the literature that finds similar beneficial effects of economic growth on crime experienced by households. An important policy implication is that per capita economic growth maybe an effective tool for policy makers to deal with crime targeting businesses, especially small and medium firms. Policies that promote growth may be an alternative to costly measures that improve the 21 deterrence of criminal activity. Pro-growth policies may have the dual effect of benefiting both the private sector directly as well as indirectly by reducing the degree of criminal activity they face. It is also important to note that a large proportion of firms are in the informal sector in developing economies. While this study has focused on formal firms, a natural extension would be to examine the informal sector. 22 TABLE 1: GDP GROWTH AND CRIME AGAINST FIRMS Logit: OLS: % of Logit: OLS: % of Losses Experienced Losses Due to Experienced Crime Due to Crime Over Crime Crime Over Sales Sales Full Sample Small and Medium Firms sample 1 2 3 4 Real GDP per Capita Growth -0.0941*** -0.0808*** -0.0914*** -0.0833*** [0.0226] [0.0208] [0.0224] [0.0201] Police per 100,000 persons lagged -0.0008 -0.0011* -0.0006 -0.0010* [0.0009] [0.0005] [0.0008] [0.0006] 0.1357** 0.1181* 0.092 0.1815** Firm with Female Owner [0.0621] [0.0632] [0.0579] [0.0834] Real GNI per capita in 100s 0.0040** -0.0058*** 0.0035** -0.0063*** [0.0018] [0.0012] [0.0016] [0.0014] GINI -0.0023 -0.0065 -0.0011 -0.0019 [0.0121] [0.0113] [0.0115] [0.0113] Population in Millions, Total -0.004 0.0034* -0.0012 0.0038* [0.0035] [0.0017] [0.0032] [0.0020] Small firm dummy variable -0.8113*** 0.4023*** -0.4262*** 0.2446*** [0.1138] [0.0940] [0.0557] [0.0846] Medium firm dummy variable -0.3551*** 0.1664** [0.0820] [0.0658] Large City or Capital dummy variable 0.0157 0.0331 0.0012 0.0826 [0.0657] [0.0953] [0.0728] [0.0974] Manufacturing firm dummy variable -0.6394*** -0.2238** -0.4717*** -0.1835* [0.0741] [0.0813] [0.0667] [0.1032] Number of Countries 27 27 27 27 Number of Observations 12801 12274 9482 9160 * significant at 10%; ** significant at 5%; *** significant at 1%, Standard errors in brackets clustered at the country level TABLE 2: GDP GROWTH AND CRIME AGAINST FIRMS – SME, POLICE, AND FEMALE MANAGER INTERACTIONS % of Losses Due % of Losses Due % of Losses Due % of Losses Dependent variable to Crime Over to Crime Over to Crime Over Due to Crime Sales Sales Sales Over Sales Full Sample Small and Medium Firms sample 1 2 3 4 Real GDP per Capita Growth -0.0563** -0.1699*** -0.0713*** -0.0699*** [0.0258] [0.0479] [0.0200] [0.0215] Real GDP per Capita Growth x Small and -0.0319* Medium Firms [0.0166] Real GDP per Capita Growth x lagged Police 0.0003** [0.0001] Real GDP per Capita Growth x Female -0.0280* Owner [0.0145] Real GDP per Capita Growth x firm with -0.0591* female owner and manager [0.0293] Firm with female owner and manager 0.4792** [0.1940] Police per 100,000 persons -0.0011** -0.0020** -0.0010* -0.0009* [0.0005] [0.0009] [0.0006] [0.0005] Firm with Female Owner 0.1161* 0.1674* 0.2575** [0.0617] [0.0837] [0.1055] Real GNI per capita (in 100s) -0.0057*** -0.0065*** -0.0063*** -0.0060*** [0.0012] [0.0015] [0.0014] [0.0015] GINI -0.0064 -0.0019 -0.0017 0.0007 [0.0114] [0.0115] [0.0112] [0.0108] Population in Millions, Total 0.0034* 0.0029 0.0039* 0.0040** [0.0017] [0.0021] [0.0020] [0.0019] Small and Medium firm dummy variable 0.3622*** [0.0871] Small firm dummy variable 0.2409*** 0.2422*** 0.2240*** [0.0850] [0.0851] [0.0774] Large City or Capital dummy variable 0.0255 0.0944 0.0811 0.0828 [0.0934] [0.0957] [0.0972] [0.0966] Manufacturing firm dummy variable -0.2414*** -0.1922* -0.1838* -0.1750* [0.0815] [0.1032] [0.1029] [0.1015] Number of Countries 27 27 27 27 Number of Observations 12274 9160 9160 9237 * significant at 10%; ** significant at 5%; *** significant at 1%, Standard errors in brackets clustered at the country level 24 TABLE 3: GDP GROWTH AND CRIME AGAINST FIRMS – QUALITY OF GOVERNANCE AND VOTER TURNOUT INTERACTIONS % of Losses Due to Crime Over % of Losses Due to Crime Over Dependent variable Sales Sales Small and Medium Firms sample 1 2 Real GDP per Capita Growth -0.2469*** -0.3573*** [0.0742] [0.0941] Real GDP per Capita Growth x Quality of 0.2995** Governance [0.1338] GDP per Capita Growth 0.4022*** x Voter Turnout [0.1417] Quality of Governance -2.1963** [0.8490] Voter Turnout -1.8393** [0.7621] Police per 100,000 persons -0.0018*** -0.0010* [0.0006] [0.0006] Firm with Female Owner 0.1739* 0.1736* [0.0855] [0.0849] Real GNI per capita (in 100s) -0.0033* -0.0046*** [0.0018] [0.0016] GINI -0.0192 0.0049 [0.0132] [0.0099] Population in Millions, Total 0.0032* 0.0040** [0.0017] [0.0019] Small firm dummy variable 0.2432** 0.2455*** [0.0938] [0.0844] Large City or Capital dummy variable 0.1042 0.0859 [0.1047] [0.0970] Manufacturing firm dummy variable -0.1662 -0.1717 [0.1198] [0.1071] Number of Countries 24 27 Number of Observations 8299 9160 * significant at 10%; ** significant at 5%; *** significant at 1%, Standard errors in brackets clustered at the country level 25 TABLE 4: GDP GROWTH AND CRIME AGAINST FIRMS – INSTRUMENTAL VARIABLES % of Losses Due to Crime Real GDP per capita Growth Over Sales First Stage Results IV Estimates Second Stage Results Small and Medium Firms sample 1 2 -0.1438*** Real GDP per Capita Growth [0.0406] Police per 100,000 persons lagged -0.0012* -0.0073 [0.0006] [0.0051] 0.1615** -0.1774 Firm with Female Owner [0.0782] [0.1701] -1.2454 Malaria Ecology Index [1.8219] % of land in the Tropics -3.6897*** [0.9537] Real GNI per capita in 100s -0.0090*** -0.0597*** [0.0020] [0.0188] GINI -0.0269 -0.2665*** [0.0180] [0.0757] Population in Millions, Total 0.0051*** 0.0353** [0.0016] [0.0163] Small firm dummy variable 0.2501*** 0.0628 [0.0838] [0.0763] Large City or Capital dummy variable 0.0757 -0.0129 [0.1017] [0.2594] Manufacturing firm dummy variable -0.1834* 0.0797 [0.1013] [0.1029] Number of Countries 27 27 Number of Observations 9160 9160 Instruments Malaria Ecology Index, % of land in the Tropics Under-identification test (Kleibergen-Paap LM statistic) P- 0.0210 value: Hansen J statistic (over-identification test of all instruments) 0.1155 p-value: * significant at 10%; ** significant at 5%; *** significant at 1%, Standard errors in brackets clustered at the country level 26 References Aidis, R., S. Estrin, and T. M. Mickiewicz (2012). “Size Matters: Entrepreneurial Entry and Government.” Small Business Economics 38 (1): 119-139. Alesina, A., A. Devleeschauwer, W. Easterly, S. Kurlat, R. Wacziarg (2003). “Fractionalization” Journal of Economic Growth 8(2), 155–194 Amin, M. (2009). “Crime, Security and Firms in Latin America” (Short Note). Available at: http://works.bepress.com/mohammad_amin/11 (accessed October 31, 2011). Baliamoune-Lutz, M. and P. Garello (2014). “Tax Structure and Entrepreneurship.” Small Business Economics 42(1): 165-190. Barro, R., J-W. Lee. (2010). “A New Data Set of Educational Attainment in the World”, 1950- 2010. NBER Working Paper No. 15902. Becker, G. (1968). “Crime and Punishment: An Economic Approach.” Journal of Political Economy 76, 169–217. Bourguignon, F. (1999). “Crime, Violence and Inequitable Development. Delta, Paris. manuscript. Bushway, S., P. J. Cook, and M. Phillips (2012), "The Overall Effect of the Business Cycle on Crime" German Economic Review 13(4): 436-4446 Cook, P. J. (2010). “Property Crime Yes violence No.” Criminology and Public Policy, 9(4), 693–697. Demombynes, G., B. Ozler. (2005). "Crime and local inequality in South Africa," Journal of Development Economics, Elsevier, vol. 76(2), pages 265-292. Di Tella, R., E. Schargrodsky. (2004). “Do Police Reduce Crime? Estimates Using the Allocation of Police Forces After a Terrorist Attack.” The American Economic Review 94, 115– 133. Donohue, J., S. Levitt. (2001). “Legalized Abortion and Crime.” Quarterly Journal of Economics. 116:2, pp. 379 –420. Dutta, M., H. Zakir. (2009). “Determinants of Crime Rates: Crime Deterrence and Growth in Post-liberalized India.” MPRA Paper No. 14478. Ehrlich, I. (1973). “Participation in Illegitimate Activities: A Theoretical and Empirical Investigation.” Journal of Political Economy 81, 521–565. 27 Fanjzylber, P., D. Lederman, N. Loayza. (2002). “Inequality and Violent Crime.” Journal of Law and Economics 45(1), 1–39. Fanjzylber, P., D. Lederman, N. Loayza. (2000). Crime and Victimization: An Economic Perspective. Economia 1 (1), 219– 278 (Fall). Gaviria, A. (2002).” Assessing the Effects of Corruption and Crime on Firm performance: Evidence from Latin America.” Emerging Markets Review (3), 245-268. Glaeser, Edward L., Bruce Sacerdote (1999). Why is There More Crime in Cities. Journal of Political Economy 107(6), 225-258. Ghosh, A., P. E. Robertson, M-C Robitaille. (2011). “Crime, Factor Abundance and Globalization: Evidence from Cross-country Panel data.” Unpublished manuscript. Downloaded October 31, 2011 http://economics.adelaide.edu.au/events/ATW2011/Robertson.pdf Gould, E. D., B. A. Weinberg, D. B. Mustard (2002), ‘Crime Rates and Local Labor Market Opportunities in the United States: 1979–1997.’ Review of Economics and Statistics 84, 45–61 Hopkins, M. (2002). “Crimes against Businesses: The Way Forward for Future Research.” British Journal of Criminology, 42(4): 782-797. Islam, A.(2013), "Is There a Gender Bias in Crime Against Firms in Developing Economics?" Women's Studies International Forum 37(2013): 1-15 Kelly, M. (2000). “Inequality and Crime.” Review of Economics and Statistics 82 (4), 540– 554 (November). Kovandzic, T. V., J. J. Sloan. (2002). “Police Levels and Crime Rates Revisited - A county-level analysis from Florida (1980-1998).” Journal of Criminal Justice 30, 65-76. Lederman, D., Laoyza, N., and A. M. Menendez,(2002). “Violent Crime: Does Social Capital Matter? Economic Development and Cultural Change 50(3), 509-539 (April 2002). Levitt, S. (2004). “Understanding why crime fell in the 1990s: four factors that explain the decline and six that do not.” Journal of Economic Perspectives 18, 163–190. Lipuma, A. J., S. L. Newbert, and J. P. Doh (2013). “The Effect of Institutional Quality on Firm Export Performance in Emerging Economies: A Contingency Model of Firm Age and Size.” Small Business Economic 40 (4): 817-841. Machin, S., C. Meghir (2004). "Crime and Economic Incentives." Journal of Human Resources, 39(4): 958-979 28 McDowall, D. and C. Loftin (2009), “Do United States City Crime Rates Follow a National Trend? The Influence of Nationwide Conditions on Local Crime Patterns”, Journal of Quantitative Criminology 25, 307–324 Naidoo, J. (2006). “The Impact of HIV/AIDS on Crime in South Africa,” http://www.sarpn.org.za, date accessed on 31 October 2011. Sachs, J., J. L. Gallup. (2004). “The Economic Burden of Malaria.” The Supplement to The American Journal of Tropical Medicine & Hygiene, vol. 64, no. 1, 2, pp 85-96. Soares, R. R. (2004). “Development, crime and punishment: accounting for the international differences in crime rates.” Journal of Development Economics 73(1), 155–184. Sachs, J., J. L. Gallup, A.. Mellinger. (1998). “Geography and Economic Development,” in Pleskovic, Boris and Joseph E. Stiglitz (eds.), Annual World Bank Conference on Development Economics 1998 (April), The World Bank: Washington, DC. Usher, D. (1997). “Education as a Deterrent to Crime,” Canadian Journal of Economics 30, 367- 84 29 APPENDIX TABLE A1: VARIABLE DEFINITIONS Variable Definition Data Source Response to the Question: Enterprise Surveys, World Bank “In fiscal year [insert fiscal year], what are the estimated losses as a result of theft, robbery, Losses Due to Crime (% of vandalism or arson that occurred on sales) establishment’s premises calculated as a percent of annual sales” For cases where actual values were reported, % of sales was calculated. Response to the Question: Enterprise Surveys, World Bank “In fiscal year [insert last complete fiscal year], has Experienced Crime this establishment experienced losses as a result of theft, robbery, vandalism or arson?” Real GDP per Capita Growth Rate, constant 2000 Word Development Indicators (WDI), Real GDP per Capita Growth USD World Bank Police per 100,000 persons lagged. Due to data United Nations Survey on Crime Trends Police per 100,000 persons constraints, for some countries the lag is not and the Operations of Criminal Justice lagged exactly lagged by year. Specifics of all lag years can Systems be found in the appendix. Yes Response to Question: Enterprise Surveys, World Bank Firm with Female Owner "Are any of the owners female?" Yes Response to Questions: Enterprise Surveys, World Bank Female Owner and Manager “Is the Top Manager female?” and "Are any of the owners female?" Gross National Income per Capita, Constant Word Development Indicators (WDI), Real GNI per capita (in 100s) 2000 USD World Bank Gini index of 0 represents perfect equality, while WDI, World Bank, Development Research GINI an index of 100 implies perfect inequality. Group Word Development Indicators (WDI), Population, Total in millions Total Population World Bank Small firms Dummy is 1 if firm is small (<20) Enterprise Surveys, World Bank Medium firms Dummy is 1 if firm is medium (20-99) Enterprise Surveys, World Bank Dummy is 1 if city is either the capital or has more Enterprise Surveys, World Bank Large City or Capital than 250,000 population Manufacturing Dummy is 1 for manufacturing firms Enterprise Surveys, World Bank Mean value of the ICRG governance variables International Country Risk Guide – The “Corruption”, “Law and Order”, and PRS Group Quality of Governance “Bureaucracy Quality”, scaled 0-1. Higher values indicate better quality of government. 1990-2007 average used. IDEA: International Institute for Turnout in parliamentary elections measured as Democracy and Electoral Assistance Voter Turnout the total number of votes cast divided by the http://www.idea.int/vt/index.cfm number of registered voters. Index of Democracy (Polity 2). Score between -10 and 10 that indicate how democratic a country. Polity IV, Polity 2 Values increase with greater democracy. 1990- http://www.systemicpeace.org/polity/poli 2007 average used. ty4.htm 30 Average Years of Schooling of Population over Barro and Lee (2010) Years of Schooling 15. 1990-2007 average used. Corruption Perception Index: 10 point scale Transparency International Corruption where higher values indicate less corruption. 1995- ww.transparency.org 2009 average used. Probability that two randomly selected people Alesina, Devleeschauwer, Easterly, Kurlat, Ethnic Fractionalization from a given country will not belong to the same and Wacziarg (2003) ethnic group Probability that two randomly selected people Alesina, Devleeschauwer, Easterly, Kurlat, Language Fractionalization from a given country do not speak the same and Wacziarg (2003) language Probability that two randomly selected people Alesina, Devleeschauwer, Easterly, Kurlat, Religious Fractionalization from a given country do not belong to the same and Wacziarg (2003) religion Response to Question: Enterprise Surveys, World Bank “At the end of fiscal year [insert last complete Total Employees fiscal year], how many permanent, full-time employees did this establishment employ?” Security Costs as a % of sales Enterprise Surveys, World Bank Proportion of Female WDI, World Bank Population, female (% of total) Population Population in urban agglomerations of more than WDI, World Bank one million is the percentage of a country's Percentage of Population in population living in metropolitan areas that in Urban Agglomerates 2000 had a population of more than one million people. WDI, World Bank Population Density People per sq. km of land area Proportion of Population WDI, World Bank Population ages 15-64 (% of total) between 15 and 64 Life Expectancy WDI, World Bank Government spending on Includes spending on police, fire protection Government Financial Statistics (GFS), Public Order and Safety over services, law courts, and prisons International Monetary Fund total Government spending Inflation WDI, World Bank Prisons refer to “Prisons, Penal Institutions or United Nations Survey on Crime Trends Correctional Institutions” which means all public and the Operations of Criminal Justice Adult Prison Capacity and privately financed institutions where persons Systems are deprived of their liberty. United Nations Survey on Crime Trends Judges Number of judges per 1000 of population and the Operations of Criminal Justice Systems Trade Exports plus imports as a % of GDP WDI, World Bank Proportion of each country’s population that live Sachs and Gallup (2004) Malaria Risk with risk of malaria transmission in 1994 % of land in Tropics Sachs et. al. (1998) 31 TABLE A2: SUMMARY STATISTICS FOR FULL SAMPLE INCLUDING SMALL, MEDIUM, AND LARGE FIRMS Variable Mean Std. Dev. Min Max Data Unit Losses Due to Crime (% of sales) 0.716 3.942 0.000 100.000 Firm Experienced Crime 0.232 0.000 1.000 Firm GDP per Capita Growth 2.656 4.213 -5.529 10.192 Country Police per 100,000 persons lagged 287.894 108.187 90.110 480.013 Country Firm with Female Owner 0.419 0.000 1.000 Firm Female Owner and Manager 0.152 0.000 1.000 Firm Real GNI per capita (in 100s) 37.345 31.014 2.257 141.816 Country GINI 39.826 7.768 25.810 52.330 Country Population, Total in millions 28.615 27.065 1.353 141.816 Country Small firms 0.369 0.000 1.000 Firm Medium firms 0.378 0.000 1.000 Firm Large City or Capital 0.627 0.000 1.000 Firm Manufacturing 0.537 0.000 1.000 Firm Quality of Governance 0.525 0.101 0.351 0.791 Country Voter Turnout 0.718 0.089 0.460 0.893 Country Polity 2 6.153 3.926 -5.647 10.000 Country Years of Schooling 8.066 1.816 2.850 11.689 Country Corruption 3.369 0.960 1.982 6.091 Country Ethnic Fractionalization 0.383 0.173 0.118 0.663 Country Language Fractionalization 0.372 0.254 0.030 0.836 Country Total Employees 121.246 478.223 1.000 20843 Firm Life Expectancy 69.407 7.440 44.966 78.314 Country Proportion of Female Population 50.957 1.339 48.750 54.020 Country Percentage of Population in Urban Country Agglomerates 21.339 10.624 4.406 39.027 Population Density 88.485 71.842 1.610 290.871 Country Proportion of Population between Country 15 and 64 63.549 5.393 47.959 70.721 Government Spending on Public Country Order & Safety over Total Spending 0.057 0.026 0.015 0.119 Inflation 9.488 7.816 2.358 50.705 Country Security Costs as a % of sales 1.514 5.373 0.000 384.615 Firm Judges per 1000 of Population 13.166 11.808 0.760 47.020 Country Adult Prison Capacity 41384 50603 1017 160327 Country Trade as a % of GDP 84.496 36.049 37.354 182.512 Country Malaria Risk Index 0.259 0.377 0.000 1.000 Country % of land in tropics 0.370 0.454 0.000 1.000 Country 32 TABLE A3: POLICE, GINI, AND GNI PER CAPITA DATA AND SURVEY YEAR Country Survey Year Police per 100,000 persons GINI GNI per Capita lagged Year Argentina 2009 2008 2009 2009 Azerbaijan 2008 2006 2008 2008 Belarus 2007 2004 2007 2004 Bosnia and Herzegovina 2008 2007 2007 2007 Costa Rica 2009 2006 2009 2009 Czech Republic 2008 2007 1996 2008 Ecuador 2009 2006 2009 2009 El Salvador 2009 2006 2007 2009 Estonia 2008 2007 2004 2008 Fyr Macedonia 2008 2006 2008 2005 Hungary 2008 2007 2007 2008 Kazakhstan 2008 2007 2007 2008 Latvia 2008 2007 2008 2008 Lithuania 2008 2007 2008 2008 Moldova 2008 2007 2008 2008 Mongolia 2008 2004 2008 2000 Nepal 2008 2006 2004 2000 Nicaragua 2009 2006 2005 2009 Paraguay 2009 2006 2008 2009 Peru 2009 2004 2009 2009 Philippines 2008 2007 2006 2008 Poland 2008 2007 2008 2008 Romania 2008 2007 2008 2008 Slovak Republic 2008 2007 1996 2008 Slovenia 2008 2007 2004 2008 Turkey 2007 2006 2008 2007 Ukraine 2007 2006 2008 2007 33 TABLE A4: CRIME AND GROWTH - COUNTRY AVERAGES Country % of Losses Due to Crime Over Sales Real GDP per Capita Growth Argentina 0.63 -0.13 Azerbaijan 0.20 9.54 Belarus 0.72 10.19 Bosnia and Herzegovina 0.44 5.57 Costa Rica 0.54 -2.79 Czech Republic 0.48 1.58 Ecuador 1.11 -0.70 El Salvador 1.73 -4.00 Estonia 1.69 -4.99 Fyr Macedonia 0.50 4.76 Hungary 0.25 1.00 Kazakhstan 0.60 2.05 Latvia 0.42 -3.82 Lithuania 0.43 3.46 Moldova 0.54 7.97 Mongolia 0.52 7.67 Nepal 0.87 4.18 Nicaragua 1.68 -2.76 Paraguay 1.54 -5.53 Peru 0.64 -0.27 Philippines 1.32 2.27 Poland 0.45 5.11 Romania 0.36 9.59 Slovak Republic 0.64 5.99 Slovenia 0.26 3.33 Turkey 0.38 3.36 Ukraine 0.45 8.55 34 TABLE A5: ROBUSTNESS – ADDED CONTROLS Coefficient of Real GDP per Capita Growth Adjusted R Squared Small and Medium Firms sample BASE -0.0833*** 0.01 [0.0201] Fractionalization -0.0641*** 0.01 Ethnic, Language, Religion [0.0209] Employees and Security -0.0840*** 0.02 Total Employees, Security Costs as [0.0195] a % of sales Corruption and Democracy -0.0879*** 0.01 Corruption (Transparency [0.0203] International), Polity 2 Score Population and Demographics -0.0576** 0.01 Proportion of Female Population, [0.0231] Population Density, Proportion of Population between 15 and 64 Human Capital -0.0870*** 0.01 Life Expectancy, Years of [0.0302] Schooling Government Spending and -0.0563* 0.01 Inflation [0.0309] Government Spending on Public Order & Safety, Inflation Judges and Prison Capacity -0.0758*** 0.01 Judges per 1000 of Population, [0.0240] Adult Prison Capacity Trade -0.0779*** 0.01 Trade as a % of GDP [0.0213] 35 TABLE A6: ROBUSTNESS – EXTREME OBSERVATION DOMINANCE Coefficient of Real GDP per Small and Medium Firms sample Capita Growth Dropping Extreme Crime Loss Observations Bottom 1% -0.083*** [0.020] Top 1% -0.034*** [0.011] Top and Bottom 1% -0.034*** [0.011] Dropping Extreme Real GDP per Capita Growth Observations Bottom 1% -0.078*** [0.021] Top 1% -0.086*** [0.020] Top and Bottom 1% -0.081*** [0.021] FIGURE A1: COUNTRY DOMINANCE – COEFFICIENT OF ECONOMIC GROWTH USING SAMPLE OF SMALL AND MEDIUM SIZED FIRMS Country Dominance Country Dropped 0 r1 r2 r3 r4 r5 r6 r7 r8 r9 r10 r11 r12 r13 r14 r15 r16 r17 r18 r19 r20 r21 r22 r23 r24 r25 r26 r27 -0.02 -0.04 coef -0.06 -95% -0.08 95% -0.1 -0.12 -0.14 36