THE EFFECTS OF FOREIGN RESIDENTS ON CRIME RATES AND LABOR MARKET OUTCOMES IN JAPAN

Transcription

1 THE EFFECTS OF FOREIGN RESIDENTS ON CRIME RATES AND LABOR MARKET OUTCOMES IN JAPAN A Thesis submitted to the Faculty of the Graduate School of Arts and Sciences of Georgetown University in partial fulfillment of the requirements for the degree of Master of Public Policy in Public Policy By Aki Horiuchi, B.A. Washington, DC April 10, 2012

2 Copyright 2012 by Aki Horiuchi All Rights Reserved ii

3 THE EFFECTS OF FOREIGN RESIDENTS ON CRIME RATES AND LABOR MARKET OUTCOMES IN JAPAN Aki Horiuchi, B.A. Thesis Advisor: Peter L. Hinrichs, Ph.D. ABSTRACT Using 2000, 2005 and 2010 Japanese prefecture-level data, I estimate the effects of the inflow of foreign residents in a prefecture on total crime rates, violent crime rates, and property crime rates in that prefecture. I also examine the relationship between the foreign resident share of the population in a prefecture and the wages and unemployment rates in that prefecture. I find that the percentage of foreign population has no effect on total crime rates or violent crime rates, though it has a significant adverse effect on property crime rates. I also find that the percentage of foreign residents has no effect on wages and unemployment rates in local labor markets. Nevertheless, the evidence is somewhat ambiguous. An array of other factors should be taken into account before forming concrete policy implications as a result of these findings. iii

4 The research and writing of this thesis is dedicated to everyone who helped along the way. Many thanks, Aki Horiuchi iv

5 TABLE OF CONTENTS I. Introduction... 1 II. Literature Review... 5 A. Research on the Relationship between Immigrants and Crime... 5 B. Research on the Impact of Immigrants on the Labor Market... 7 III. Data and Empirical Methods IV. Results A. Regressions of the Fraction of Foreigners on Crime Rates B. Regressions of the Fraction of Foreigners on the Labor Market Outcomes V. Policy Implications VI. Conclusion Tables References v

6 I. Introduction According to the Japanese Ministry of Justice s Immigration Bureau (2011a), at the end of 2010, the number of registered foreign residents in Japan stood at 2,134,151, accounting for approximately 1.67% of the total Japanese population. 1 Although the number of foreign residents in Japan had increased steadily over more than 30 years until 2008, it decreased slightly between 2008 and As a result of the prolonged stagnation of the Japanese economy and the massive earthquake and tsunami which hit Japan in March 2011, some foreign residents have left Japan, and it is unclear whether the number of foreigners residing in Japan will begin to rise again or continue to decline in the future. In today s globalized world, however, it is reasonable to assume that foreigners will continue to come to Japan. Currently, Japan s immigration policy accepts only skilled foreign workers for employment-based admissions. In addition to accepting skilled foreign workers, the government also accepts foreigners who have a familial relationship with Japanese citizens and foreign permanent residents, as well as foreign students. Basically, no limitation on activities including working is imposed upon family-sponsored foreign residents. The estimated number of unauthorized foreigners who stay in Japan beyond their permitted periods has consistently decreased since 1993 (Immigration Bureau, 2011a). According to the Immigration Bureau (2011a), this results from stricter immigration inspection and tougher law enforcement in regard to unlawful foreign workers and other policy changes. However, the Immigration Bureau estimates that there were still 90,000 to 100,000 unauthorized foreign residents in Japan as of January 1, 2011 (Immigration Bureau, 2011a). Addressing the 1 Foreign nationals who reside in Japan over 90 days, with some exceptions including diplomatic officials and members of U.S. military bases, are obliged to apply for Alien Registration. 1

7 issue of unauthorized foreign residents continues to be one of the main concerns for the Japanese government. While the number of non-citizens staying in Japan tends to increase, Japanese citizens attitudes toward foreign residents are divided. Today, in Japan, the birthrate is declining and the population is aging. Japan s total fertility rate (TFR), which was 2.13 births per woman in 1970, came down to reach 1.39 in Under such circumstances, some observers, especially in the business community, argue that in order to secure labor and to maintain international economic competitiveness, the government should accept more foreign workers. Among those in favor of accepting more foreign workers, there are differing opinions about whether the Japanese government should continue to focus on accepting skilled workers or accept more foreign workers comprehensively, including unskilled workers. In fact, the Japanese Ministry of Justice announced, in the Basic Plan for Immigration Control: 4th Edition (2010), that it would consider introducing a points-based admission system for highly skilled foreign workers. It appears that the Japanese government is moving toward promoting acceptance of skilled foreign workers, while it noted that it would stimulate and participate in the public debate over future national policies on acceptance of foreign nationals. In 2004, the Japanese Cabinet Office conducted a survey of public opinion on accepting foreign workers; 25.9% of the respondents answered that Japan should continue to accept only skilled foreign workers and not unskilled workers; 39% answered that Japan should accept lowskilled workers only for limited sectors which lack sufficient native labor; and 16.7% replied that Japan should accept low-skilled workers without such limitations. The survey also showed that common reasons for opposing acceptance of unskilled workers were that it might undermine 2

8 public safety, that it might increase conflicts in local communities, and that it might have negative effects on labor opportunities among Japanese people. Actually, the opinion that more foreigners will jeopardize public safety in Japan and deprive Japanese citizens of employment opportunities has been the popular perception regarding immigration issues in Japan. Despite the fact that immigration issues caught the attention of the public and the media, little empirical research using Japanese data has been conducted on the effects of foreigners on public safety and labor market outcomes. In the United States, numerous empirical studies examine the relationships both between immigrants and crime, and between immigrants and the labor market. As for the impact of immigrants on public safety, many studies using U.S. data find that the influx of immigrants decreases crime rates rather than increasing them. On the other hand, it is difficult to evaluate the effects of immigrants on U.S. labor market outcomes among native workers. Researchers here have reached different conclusions and results are not as straightforward as those found in the crime rate studies. Yet it seems that many researchers conclude that immigrants have either a relatively small or no effect on labor market outcomes among American workers. This paper examines the effects of foreign residents on public safety and the labor market in Japan using Japanese data, but applying frameworks used in past studies in the United States. Given previous evidence in past U.S. research, one might expect the analysis to show that foreigners have no significant effects on public safety and the labor market in Japan. Analysis indicating that foreign residents do not jeopardize public safety and the labor market could help reduce general anxiety in Japan toward foreigners. On the other hand, a finding that the inflow of 3

9 foreigners does have adverse effects on public security and the labor market in Japan would imply that the government should be careful about opening the door for more foreign nationals. The paper uses prefecture panel data from official Japanese government sources on the number of foreign residents, crime rates, wages, and unemployment rates. I estimate models with and without fixed effects both in the crime rate analysis and the labor market analysis. The results show that controlling fixed effects matters, presumably because prefectural time-invariant characteristics are likely to bias the estimates on the foreigners share variable. It is common to investigate the impact of immigrants by classifying immigrants into different groups. Since data on educational attainment, work experience, or the length of residence of foreign residents are not available, I examine whether people from different countries of origin have different effects on public safety and labor markets. To do this, I use the share of foreigners from selected countries instead of the total foreign population. This draws a clearer picture of what is going on in terms of the link between foreign residents and crime rates and labor markets. My findings suggest that a 1 percentage point increase in the fraction of foreign residents in the total population increases property crime rates by approximately 62 percent, but the estimates of the effects of the fraction of foreign residents on crime rates are not so robust across different specifications. In addition, I find that the percentage of foreigners has no effect on wage and unemployment rates in local areas. Section II below reviews past research on this subject; Section III discusses data and empirical methods used in this study; Section IV presents the regression results; Section V discusses policy implications; and Section VI offers concluding remarks. 4

10 II. Literature Review A. Research on the Relationship between Immigrants and Crime Dozens of studies examine the relationship between immigration and crime in the United States, and these studies tend to suggest that increases in the number of immigrants are related to lower crime rates rather than higher crime rates. For example, Butcher and Piehl (1998b), using 1980 and 1990 Census data, examined whether immigrants were more likely than natives to be institutionalized. 2 Their paper shows that institutionalization rates are lower among immigrants than among natives, whether or not they control for demographic characteristics. It also indicates that more recent immigrants have less institutionalization probability than earlier immigrants. Butcher and Piehl (2007), drawing on 1980, 1990, and 2000 Census data, replicated their previous research and found similar results. Additionally, their findings suggest that the gap between immigrants and natives has widened over the last three decades. In addition to assessing the criminal propensities of immigrants themselves, researchers have explored the effect which immigration has on a community as a whole. Butcher and Piehl (1998a) using city-level Uniform Crime Reports (UCR) data and the Current Population Survey (CPS) during the 1980s, showed that the inflow of immigrants into a city does not affect the crime rates in that city, whether or not they control for city demographic characteristics. In addition, using individual-level data from the 1979 National Longitudinal Survey of Youth (NLSY), they found that young immigrants were less likely to commit crimes than young natives. Gonzales (2006), following Butcher and Piehl s macro-level approach and using 1990 and Butcher and Piehl (1998b) used data on institutionalization which included non-correctional facilities such as hospitals, nursing homes, and drug treatment centers along with correctional facilities. They focused their research on young adult males, for whom incarceration accounted for 70% of total institutionalization. 5

11 Census and UCR data, found that a higher proportion of immigrants in a county is correlated with lower crime rates in that county. Her evidence, however, suggests that more recent immigrants who have stayed in the United States for less than five years are associated with higher crime rates while earlier immigrants are correlated with lower crime rates, which is contrary to Butcher and Piehl s findings. To address the limitations of previous cross-sectional studies which focus on the restricted geographical areas, Stowell, Messener, Mcgeever and Raffalovich (2009) explored the impact of immigration in metropolitan areas, using time-series annual data for 103 metropolitan areas from 1994 to Their time-series approach suggested that the influx of immigrants is associated with lower overall violent crime rates, robbery rates, and aggravated assault rates. Some scholars have focused on certain immigrant groups such as selected ethnic groups or unauthorized immigrants, assuming that different immigrant groups may affect crime differently. For example, Reid, Weiss, Adelman, and Jaret (2005), drawing on 2000 Census and UCR data, explored the effects of recent immigrants, Asian immigrants, Hispanic immigrants, and immigrants with limited English proficiency on homicide, robbery, burglary, and larceny. They found that the percent Asian foreign-born and the percent Hispanic foreign-born in a metropolitan area had no significant effect on that area s homicide rates, robbery rates, or burglary rates. But the percentage of the population who were born in Asian countries was negatively related to theft rates. Furthermore, Martinez and Lee (2000), looking descriptively at homicide victim rates for El Paso, Miami, and San Diego, 3 found that the Mariel Cubans raised Miami s homicide rates temporarily, but in the long run, these rates fell below 3 Martinez and Lee (2000) note that victim rates can be used as proxies for homicide offending rates because victim rates follow the same racial and ethnic outlines as offender rates. 6

12 national rates for all cities of Miami s size. They also found low Latino homicide rates in El Paso and San Diego, low Haitian homicide rates in Miami, and relatively low Asian homicide rates in San Diego. Coronado and Orrenius s study (2005) sheds more light on this issue by focusing on unauthorized immigrants. They analyzed how undocumented immigrants and border enforcement affected border crimes. The authors used monthly crime reports between 1991 and 2000 from 20 counties bordering Mexico. They used border patrol apprehensions as a proxy for undocumented immigration and included additional demographic and economic information. They found that unauthorized immigrants were positively related to violent crime rates, while they were not related to property crime rates. Given the fact that there were an estimated 11.2 million unauthorized immigrants in the United States in 2010 (Passel & Cohn, 2011), their study is noteworthy. In contrast to the United States, there is very little empirical work on the relationship between foreign immigrants and crime using Japanese data. One exception is Kobayashi (2010), who used Japanese data from 1995 to 2006 to estimate the effects of fraction of foreign residents in a prefecture on the prefecture s crime rates. Kobayashi concluded that a higher proportion of foreign residents is related to higher violent crime rates, but found no evidence that increases in foreign population affected total crime rates or property crime rates. B. Research on the Impact of Immigrants on the Labor Market Making use of various economic models and empirical methods, many scholars have studied the effect of immigration on the labor market opportunities of native workers in the 7

13 United States. While some of this literature (e.g., Borjas, 2003) estimates large negative impacts on competing native workers, most studies have found little or even no effect on the labor market outcomes of natives. For example, drawing on individual-level CPS data, Card (1990) examined the impact of the 1980 Mariel boatlift, as an unusual natural experiment, on the labor market in Miami. He found that the influx of Mariel Cuban immigrants had no significant effect on the wages or employment rates of less-skilled non-cuban workers or even the wages or employment rates of earlier Cuban immigrants. Altonji and Card (1991) used 1970 and 1980 Census data to estimate the effects of the immigrant fraction of the population in a city on labor market outcomes of the less-skilled natives. Their empirical results suggest that there is little evidence that an increase in the fraction of immigrants significantly affects the employment opportunities of less-skilled natives in a city, while the inflow of immigrants is negatively correlated with the wages of less-skilled natives. Later, Card (2005), using 2000 Census city-level data, updated his previous work and confirmed that an inflow of less-skilled immigrants had no impact on the wages of less-skilled natives and only a slightly negative effect on their employment. He determined that the inflow of less-skilled immigrants raised the relative supply of less-skilled workers in a city, but this excess supply of unskilled labor was likely to be absorbed by employers increasing the fraction of less-skilled workers within industries. Borjas (2003) argued that cross-section approaches which analyzed differences across local labor markets underestimated the true effect of immigration because this effect could spread across cities by inter-city mobility of people, goods, and services. To address this simultaneity effect, he developed a new approach which analyzed the effect of foreign-born 8

14 workers focusing on a national-level labor market. Using Census and CPS data, he aggregated individual workers into particular skill groups by educational attainment and work experience. He estimated the effect of the immigrant share of the labor force in a skill group on the labor market outcomes of natives in that group. Contrary to the findings of crosssectional studies, Borjas s time-series analysis estimated relatively large negative effects of immigration on the wages and employment opportunities of comparable native workers. Ottaviano and Peri (2008), however, argued that Borjas overestimated the adverse effect of immigration. They argued that Borjas s sizable negative impact was driven by his inappropriately small estimate of the substitution elasticity between high school dropouts and high school graduates and the lack of an assumption of capital adjustment in the short run. Using Census data and 2006 American Community Survey (ACS) data, they refined the national-level approach applied by Borjas by estimating the elasticities of substitution between high school dropouts and high school graduates and between natives and immigrants and taking into account the speed of capital adjustment. Analysis based on these estimates concluded that immigrants have small negative effects on less-skilled natives and average native wages in the short run, while they have small positive effects on less-skilled natives and average native wages in the long run. Few researchers have examined the impact of foreign residents on wages or employment in Japan. Using Japanese prefecture-level data for 1990, 1995, 2000, and 2005, Kayashima (2011) studied the effect of foreign workers on the wages of workers in Japan. Using two-stage least squares (2SLS) regression analysis, he found that the fraction of foreign workers was not related to the wages of workers in Japan. His study also showed that the rate of foreign skilled 9

15 workers in a prefecture had a positive relationship with the wages of unskilled workers in that prefecture. III. Data and Empirical Methods In this study, I pool 2000, 2005, and 2010 Japanese prefecture-level data. Since there are 47 prefectures in Japan, my sample has 141 observations. Data on the number of foreign residents and other basic demographic variables come from Population Censuses conducted by the Ministry of Internal Affairs and Communications (MIC); data on crime rates come from the National Policy Agency (NPA); data on unemployment rates come from the Labour Force Survey conducted by MIC; wage data come from the Basic Survey on Wage Structure conducted by the Ministry of Health, Labour and Welfare (MHLW); data on welfare recipients come from the Report on Social Welfare Administration and Services prepared by MHLW; and data on education attainment come from the School Basic Survey conducted by the Ministry of Education, Culture, Sports, Science and Technology (MEXT). Population Censuses are conducted every five years in Japan. This paper includes data from the latest Population Census conducted in 2010 and released in October Census data provided information on the total population, foreign population, mean age, and the gender ratio in each prefecture. The Census tries to enumerate every resident in Japan regardless of his or her nationality, and subjects are required to report by law. The limitation of using Census data, however, is that they miss most illegal residents. Even if responding to the survey is mandatory, many unauthorized foreigners do not respond because they are afraid of it being noticed that they are undocumented. This is always a problem for research on the effect of 10

16 immigration. We would like to know the impact of immigration including unlawful residents. However, it is impossible to know the exact number of illegal immigrants. Therefore, I simply use the number of foreign residents provided by Censuses. 4 I also use unemployment rates by prefecture provided by the Labour Force Survey. Yet the Labour Force Survey is not designed to collect data by prefecture and provides estimates by Regression Analysis of Time Series (RATS). These prefecture-level estimates are less accurate than unemployment rates at the national level. Criminal statistics in this study represent the number of criminal offenses in a prefecture which were reported to the police. 5 I use the numbers of offenses which were reported to the police rather than the number of arrests, because reported crimes seem to be more directly related to public safety. Violent crimes are defined as homicide, robbery, arson, rape, kidnapping, and indecent assault. Property crimes are defined as burglary, motor vehicle theft, bag-snatching, and pickpocketing. Thus, the data do not include some types of crime such as drug crimes or prostitution. If these crimes are committed disproportionately by foreigners, the absence of these data would lead to underestimation of the effects of foreigners on crime. The Basic Survey on Wage Structure provides prefectural mean wages. The 2000 data provide mean wages separately by gender, and I pooled these by-gender data to calculate mean wages regardless of gender. The Report on Social Welfare Administration and Services provides the annual number of residents receiving livelihood assistance in each prefecture. The limitation 4 Unauthorized foreign workers tend to concentrate on Kanto region (Tokyo, Kanagawa, Chiba, Saitama, Ibaraki, Gunma, Tochigi), which accounted for 69.4% of the total unauthorized foreign workers under removal proceedings in 2010 (Immigration Bureau, 2011a). 5 This excludes professional negligence resulting in injury or death and dangerous driving resulting in death or injury in traffic accidents. 11

17 of data on wages and unemployment is that they include foreign workers wages and unemployment rates. Since data focused on wages and unemployment rates among Japanese workers are not available by prefecture, I cannot separate the effects on labor market outcomes of Japanese workers and those of foreign workers. The School Basic Survey provides the fraction of high school graduates in a prefecture who entered college. The present analysis is divided in two parts. First, I analyze the effects of foreign residents on crime rates. Second, I examine the link between foreigners and wages and employment. In both analyses, I estimate ordinary least squares (OLS) models with and without fixed effects. I apply a fixed effects method to eliminate bias from omitted variables which are fixed over time. These omitted variables might be geographical factors, such as the ease of access for foreigners or local attitudes towards foreigners which tend to persist over time. In the analysis of the relationship between foreigners and crime, I estimate the model in the form: crime it = αforeign it + βx it + γ i + δ t + ε it (1) where crime it is the log of the total crime rate per 100,000 residents in a prefecture, the log of the violent crime rate per 100,000 residents in a prefecture, or the log of the property crime rate per 100,000 residents in a prefecture for prefecture i in year t. I use logarithms of crime rates instead of levels of crime rates to address two potential concerns. First, criminal offenses tend to be under-reported to the police. As Gould, Weinberg, and Mustard (2002) point out, the proportion of actual offenses reported depends on the type of crime and where it takes place. Ehrlich (1996) suggests that one approach to address this measurement error is to use logarithms of reported crime rates because this measure is likely to be proportional to true crime rates. Many studies have applied this method, and I take advantage of it too. foreign it is the percentage of foreign 12

18 residents in prefecture i in year t. The parameter of interest is α, which shows the effects of foreigners on crime rates. X it indexes a vector of control variables including log total population, population density (residents per km 2 ), mean age, percentage of females, percentage of high school graduates entering college, percentage of welfare recipients, log hourly wage, and unemployment rate. Since the variables of education, welfare, hourly wage, and unemployment rate may be endogenous, I also run regressions excluding these four variables to examine how the coefficient estimates of interest change. In the models with fixed effects, I include a set of prefecture dummies and year dummies. ε it is the error term. Since the data I use are prefectural averages, I will look at how regression results change if I weight the prefectures by population size. In addition, I estimate models which replace the percentage of the total foreign population with the rates for residents from five different countries which ranked in the top five in arrest cases among foreign residents in Japan in In the analysis of the link between foreigners and wages and employment, I estimate models of the form: y it = αforeign it + βx it + γ i + δ t + ε it (2) Here, y it indicates the log of the average hourly wage or the unemployment rate for prefecture i in year t. foreign it is a percentage of foreign residents in prefecture i in year t. I use the percentage of foreign residents rather than the percentage of foreign workers because this makes the results from the second model comparable to those in the first model. α, the parameter of interest, indicates the effects of foreigners on wages and unemployment rates. X it is a vector of control variables including log total population, population density, mean age, percentage of 6 This ranking excludes the number of arrest cases of permanent residents, members of U.S. military bases, and those with their residence statuses unknown. 13

19 females, percentage of high school graduates entering college, percentage of welfare recipients, log hourly wages, and unemployment rate. 7 As I do in the first analysis, I run additional regressions without potential endogenous variables and see how estimates vary when weighting prefectural data by population. When I estimate models with fixed effects, I include a full set of prefecture dummies and year dummies. ε it denotes the error term. In addition, I estimate models which replace the percentage of total foreign population with the proportions of residents from five different countries. These countries accounted for the five largest populations of foreign residents in Japan in IV. Results A. Regressions of the Fraction of Foreigners on Crime Rates Table 1 shows summary statistics of the 2000, 2005, and 2010 data, providing prefecturelevel mean values for each variable. Crime rates refer to the number of criminal offenses which were recorded by the police per 100,000 residents in a prefecture. The table indicates that the log of the total crime rate declined from 2000 through While there is no consistent tendency in violent crimes, the log of the property crime rate decreased from 5.4 to 4.7 over that decade. Although not shown in Table 1, the mean prefectural property crime rate per 100,000 residents fell from to Hourly wages are shown deflated by the Japanese consumer price index (CPI). At a glance, there is no significant change in hourly wages and unemployment rates over the period. The percentage of foreign residents, the most important variable in this paper, has increased consistently, although the change has been gradual and varies across immigrant group. 7 I drop log of hourly wages and unemployment rates when I use these variables as dependent variables. 14

20 The table shows that the percentage of Chinese residents increased from 2000 to On the other hand, the share of Koreans fell during this period. The mean age of the total population also continues to rise. The table also indicates that the proportions of high school graduates entering college and welfare recipients have been rising over this time period. 8 First, I run regressions of total, violent, and property crimes on the percentage of foreigners. Table 2 presents the coefficient estimates and standard errors for each variable, the number of observations, and the R-squared from regressions that use the total crime rate as the left-hand side variable. Column (1) shows the result of a baseline bivariate regression. Column (2) indicates the result of a regression controlling for demographic variables including log total population, population density, mean age, and the fraction of females. Column (3) shows the percentage of high school graduates entering college, the percentage of welfare recipients, log hourly wages, and unemployment rate, as well as the demographic variables. Column (4), (5), and (6) provide the estimates with fixed effects. While the percentage of the foreign population is highly statistically significant in Column (1) and (2), it falls to insignificance when education, welfare, hourly wages, and unemployment variables are taken into consideration. When I use fixed effects models, the percentage of foreigners is not statistically significant at conventional levels in any specification. This may suggest that the fraction of foreign residents has no impact on total crime rates after controlling for unobservable prefectural characteristics which do not change over time. These 8 Here, welfare refers to the Japanese Public Assistance System. This system, which is funded by the national and local governments jointly, provides poor households with a minimum standard of living expenses. Practically, some foreign residents also receive this assistance, since there seem to be no clear and consistent regulations. According to the 2010 Report on Social Welfare Administration and Services, during 2010, a total of 827,584 foreign nationals received public assistance. Foreigners accounted for 3.5% of all recipients. 15

21 fixed characteristics may include geographical factors such as the ease of access for foreigners or general attitudes toward non-citizens in local communities. In addition, Gould et al. (2002) suggest that different local authorities have different ways to collect and report crime data, and a fixed effects model could eliminate these differences in reporting data across regions. Table 3 presents the effects of foreign residents on the violent crime rate. The pattern seen in Table 2 persists. Fixed effects estimates are statistically insignificant across specifications. Table 4 reports the estimates of the effects on property crime rates. Contrary to the results provided in Table 2 and 3, the fixed effects estimates on foreigners are statistically significant. Even in the fully specified model controlling for education, welfare, hourly wages and unemployment variables, the estimate is still statistically significant. Thus, a 1 percentage point increase in the fraction of foreign residents is estimated to increase property crime rate by approximately 62 percent. This effect seems to be large given the magnitudes of estimates shown in other regressions. Therefore, we should be wary in interpreting this result. The magnitude appears larger than it really is because the percentage of foreign population in Japan is quite small. Since the registered foreigners share of the total Japanese population is approximately 1.67%, a 1 percentage point increase would represent a sizable increase. These findings are in sharp contrast to existing U.S. studies. Many previous researchers there conclude that immigration is associated with lower crime, or at least has no impact on crime rates. My findings also differ from Kobayashi s results. Kobayashi (2010) found that, in Japan, increases in the number of foreign residents are correlated with higher violent crime rates, but not with total crime and property crime rates. 16

22 Many factors might explain these differences. For example, the characteristics of incoming foreigners might be different in the United States and in Japan. Researchers in the United States often argue that immigration is related to lower crime rates because immigrants are self-selected. They come to the United States to work and want to keep working without being deported. Therefore, immigrants have an incentive not to commit crimes. Also, U.S. law presumes that all non-citizens who come to the United States are permanent residents (Aleinikoff, Martin, Motomura, & Fullerton, 2011). On the other hand, in Japan, foreign residents are not necessarily immigrants as described in the United States. Foreigners in Japan are presumed to be temporary residents, although they can get permanent residence after a reasonable period. Therefore, there might be less keeping these temporary residents in Japan from committing crimes compared to immigrants in the United States. In addition, most U.S. studies on immigration do not distinguish immigrants (permanent residents) and non-immigrants (temporary residents), 9 but this difference in a basic presumption about newcomers might contribute to different behaviors among these newcomers in the United States and Japan. 10 For example, the immigrant presumption in the United States might encourage newcomers to integrate into U.S. society. Furthermore, existing U.S. studies tend to include naturalized citizens in the immigrant share. On the other hand, Japanese data 9 I assume this because the U.S. Census uses the term foreign born rather than immigrant, and foreign born includes naturalized U.S. citizens, lawful permanent residents, temporary migrants, refugees, and unauthorized immigrants. Also many past studies (e.g., Altonji & Card, 1991; Borjas, 2003; Butcher & Piehl, 1998a, 2007; Gonzales, 2006; Ottaviano & Peri, 2008; Stowell et al., 2009) explicitly note that immigrants in their studies refer to foreign-born residents. 10 Passel and Cohn (2011) estimated that there were 12.4 million legal permanent resident aliens and 1.7 million legal temporary migrants in the United States in Thus, the temporary residents share of legal foreign residents was approximately 88% in the United States. On the other hand, in Japan, there were 399,106 registered permanent residents at the end of 2010, including special permanent residents. Permanent residents accounted for approximately 45% in Japan. 17

23 differentiate natives and foreigners based on their nationalities. This measurement difference might drive differences between my findings and most U.S. findings. In addition, social infrastructure might help to explain differences between the United States and Japan. The United States is a country of immigrants. It has more experience and stored knowhow for accepting newcomers. There may be more widespread social services available to immigrants or informal community-level assistance than in Japan. On the other hand, Japan, which is surrounded by the sea and is a historically closed society, is not a country which has been accepting towards immigrants. Foreigners residing in Japan may face more difficulties in integrating into the existing society or a stronger backlash among Japanese neighbors. These might push foreign people away from the mainstream and encourage them to become involved in criminal activities. There are no strong hypotheses about what causes the differences between Kobahashi s findings and mine. One possibility, however, may be timing. He explored the time period from 1995 through In contrast, I focus on the last decade. The characteristics of foreign residents or their surroundings might vary between these two time periods. Another hypothesis is that the foreign population shown in the Census data is different from the population measured by the registered alien statistics which Kobayashi used. These two surveys target almost the same range of foreigners, but there are differences in the characteristics of the datasets. For example, the Census tries to capture actual resident conditions at the time of the survey, while the alien registration data tend to reflect official addresses regardless of physical whereabouts I use census data for the foreign population partly because of data availability issues. In fact, the number of foreign residents from the registered alien statistics in 2010 provided by the Japanese Ministry of Justice was larger than that of the Census (2,134,151 in the alien registration and 1,648,037 in Census). This might imply that the 18

24 In Tables 5, 6, and 7, to explore whether different ethnicities have different effects on crime rates, I use the fractions of foreign residents from five countries (China, Brazil, Vietnam, Korea, and Philippines) instead of the fraction of total foreigners. These countries ranked in the top five in arrest cases among foreigners in Japan in Table 5 provides estimates of the effects on the total crime rate. In Column (1), the percentage of the population that is Chinese is negatively correlated with total crime rate, and the percentages of the population that are Korean and Philippine are positively correlated. However, when I control for additional explanatory variables, only the coefficient estimate on the Korean population is consistently highly statistically significant. None of the fixed effects estimates are statistically significant. Table 6 explores the relationship between foreign residents from the five leading countries and violent crime. The results are similar to those in Table 5. Without controlling for fixed effects, however, none of the five nationality variables are consistently statistically significant across Column (1) to Column (3). In using fixed effects models and including all explanatory variables, no coefficient estimates on the five nationality variables are statistically significant. Table 7, which shows the outputs from regressions that use the property crime rate as the left-hand side variable, also follows a similar pattern to those seen in Table 5 and 6. However, Column (6) indicates that a 1 percentage point increase in the share of the population that is Philippine is estimated to increase the property crime rate by percent. The magnitude becomes so large because the fraction of the population that is Philippine itself is small. This Census does not show as accurate numbers as alien registration. Or, some registered foreigners might have left Japan temporarily at the time of the Census. 19

25 result is slightly unexpected. There seems to be no persuasive explanation for why only Philippine residents are correlated with higher property crime rates. As Becker (1968) suggests, economists usually argue that a person commits crime when the utility he or she gets by engaging criminal activities exceeds the utility he or she gets from returns to other legal activities. That is, criminal activities are substitute for legal labor market opportunities. Data on earnings by nationality are not available, but 2005 Census data indicate that the employment rate of the Philippine population is almost comparable to the employment rates of other nationalities. These data alone do not show that the Philippine population has poor labor market outcomes. Whether there are actual relationships between the fraction of the Philippine population in the total population and property crimes and whether other factors such as economic conditions explain this result remain open questions pending further research which looks at the demographic composition, income levels, education levels, and residence status distribution of Philippine residents. 12 B. Regressions of the Fraction of Foreigners on the Labor Market Outcomes In Tables 8 through 11, I present the estimates of the effects of the percentage of foreign residents on wages and unemployment rates. As I did in the earlier part of this paper, I start with a baseline model only including hourly wages and the fraction of foreign residents; then add total population, population density, mean age, and gender ratio; and then go on to include 12 Evidence indicates that young males are more likely to engage in criminal activities. However, the 2010 registered alien registration statistics show that the female ratio of Philippine residents (78%) is much higher compared to other leading immigrant population groups (whose female population ratios range from 46% to 58%). 20

26 educational attainment, welfare recipient rate, hourly wages, and unemployment. Columns (4), (5), and (6) refer to the results using fixed effects models. Table 8 shows somewhat interesting results. The estimates on foreigners from pooled OLS models are all highly statistically significant, and indicate that the percentage of foreign residents is positively correlated with local hourly wages, regardless of including additional control variables, though the inclusion of more explanatory variables attenuates the size of the effects. However, controlling for fixed effects dramatically changes the results. With fixed effects, the estimates on the share of foreigners lose statistical significance. This might suggest that prefectural characteristics which are fixed over time explain much of the variation in wages. Table 9 reports the results from regressions of foreigners on unemployment rates. Only Column (2) shows statistical significance, and it indicates that a higher fraction of foreign residents is related to lower unemployment rates. Nevertheless, the fixed effects estimates are not statistically significant. Overall, these findings suggest that an increase in foreign residents has no significant impact on mean hourly wages and unemployment rates in local labor markets. This finding is consistent with the evidence of many previous studies. Table 10 and 11 present the estimates from regressions using five country variables (China, Korea, Brazil, Philippines, and the United States) instead of the total foreign population. 13 Column (3) in Table 10 indicates that a higher Korean population rate is correlated with higher hourly wages, but the magnitude is quite small. Also, adding fixed effects makes all the estimates statistically insignificant. 13 These five countries ranked in the top five in the population by nationality in 2010 Census. The 2010 registered alien statistics shows that the five top countries were China, Korea, Brazil, Philippines, and Peru; the United States comes next. 21

27 Table 11 provides some interesting results. In pooled OLS models, increases in the percentages of the population that are Chinese and Brazilian lead to lower unemployment rates. In sharp contrast, increases in the American population rate are related to higher unemployment rates. This suggests that American residents are competitive with the existing workers. In addition, a higher fraction of Koreans is correlated with higher unemployment rates. Korean residents are different from other nationals in that some of them are granted special permanent residence. This special permanent residence is established by a law covering Korean people who lived in Japan at the end of World War II and their descendants. These special permanent residents are treated less restrictively in many ways compared to other foreign residents, and therefore are more integrated into Japanese society. This might be the reason why the Korean population is likely to be substitutable for the existing workers. In my preferred specification with fixed effects and all control variables, only the estimate on the Brazilian population rate exhibits statistical significance. A 1 percentage point increase in the Brazilian share is estimated to decrease the unemployment rate in a given prefecture by percentage point. This might result from the fact that the Brazilian population includes many Japanese Brazilians, descendants of Japanese people who immigrated to Brazil. In general, these Japanese Brazilians are granted resident status based on their blood relationships rather than their employment relationships. Since they enter Japan with long-term resident status, their problems, especially insufficient education for Brazilian children and high unemployment among Brazilian workers, often attract media attention. 14 These Japanese 14 According to the 2010 alien registration statistics, 33.6% of the registered Brazilians were long-term status holders in By contrast, the fraction of total foreign residents having long-term resident status was 9.1%. 22

28 Brazilians tend to be low-skilled workers who might not be substitutable for existing local workers. Finally, Table 12 summarizes coefficient estimates from different specifications. For example, I separate my sample into eastern and western prefectures to see whether geography affects my results. 15 Interestingly, the fixed effects estimate on property crime rates in the eastern sample is statistically insignificant, while that of western sample is highly statistically significant. A possible explanation is that the characteristics of foreigners might be different according to the part of Japan in which they choose to reside. In particular, to test whether Tokyo, which is economically vibrant and is likely to attract more foreign residents both legal and illegal, might change the results, I run regressions excluding Tokyo from the sample. Yet the results do not significantly change. In addition, since I use prefecture averages, I weight regressions by population to determine whether weighting will change the results. Generally, weighting does not have significant effects on the results. However, the fixed effects estimate on property crime rates increases as a result of weighting from to That is, a 1 percentage point increase in the percentage of foreign residents is now estimated to increase property crime rate by 69.8 percent. V. Policy Implications 15 Eastern prefectures include Hokkaido, Aomori, Iwate, Miyagi, Akita, Yamagata, Fukushima, Ibaraki, Tochigi, Gunma, Saitama, Chiba, Tokyo, Kanagawa, Niigata, Nagano, Yamanashi, and Shizuoka. Western prefectures include Toyama, Ishikawa, Fukui, Gifu, Aichi, Mie, Shiga, Kyoto, Osaka, Hyogo, Nara, Wakayama, Tottori, Shimane, Okayama, Hiroshima, Yamaguchi, Tokushima, Kagawa, Ehime, Kochi, Fukuoka, Saga, Nagasaki, Kumamoto, Oita, Miyazaki, Kagoshima, and Okinawa. 23

29 Results of the present study show that the proportion of foreign residents in a prefecture is positively correlated with the property crime rate in that prefecture. Then, should the Japanese government curb the further influx of foreigners? Saying yes would be too hasty a conclusion. These results also imply that other factors such as educational attainment, welfare recipient rates, wages, and unemployment rates might explain the variation in crime rates. Furthermore, although my findings suggest that there is a positive correlation between the percentage of foreign residents and the property crime rate, we cannot be certain whether there is a causal relationship between them. Additionally, the findings are not consistent with previous evidence using Japanese data. Therefore, in order to evaluate whether an increase in the percentage of foreign residents is likely to cause an increase in crime rates, we need to look at accumulated findings which future research could provide. In particular, we need to know whether different kinds of foreigners with different skills or different legal statuses have different effects. We also need to take many other considerations into account when we consider the immigration debate. From an economic perspective, accepting more foreigners to address the problem of ongoing population decline might be necessary for economic growth. And introducing highly skilled workers might drive technological progress. At the same time, we need to think about the fiscal impact of foreign residents. We need to consider the cost of social services for foreigners such as education, welfare, and health care, and the benefit of tax revenues they provide. How much importance we give to homogeneity in Japanese society also matters. As the United States has put value on diversity, historically, homogeneity has been highly valued in Japan. In addition, we need to consider the extent of social integration, such as respect for 24

30 Japanese values, we would expect newcomers to achieve. Immigration policies should not be discussed without considering these social values. Policy choices could differ depending on which of these issues people prioritize. If the difference between my findings and most U.S. findings is attributed to the different presumptions toward newcomers, my findings imply that accepting foreigners as immigrants rather than as non-immigrants might raise their incentive not to engage in crime. Also, this might shed light on the question of how to treat unauthorized foreign residents. Whether unauthorized foreign residents lead to higher crime rates and if so, what causes the higher crime rates among them, should be analyzed. 16 If weak personal ties to a host country induce non-citizens to engage in crimes and better integration into the domestic society leads to lower crime rates, granting amnesty to unauthorized foreigners might be a policy option. 17 My findings also suggest that an increase in the proportion of the population that is Philippine is associated with an increase in property crime rates. Here, I apply the same logic as I did above. We need to consider an array of other issues before we take a stand. Moreover, setting different ethnic qualifications for admission would run counter to general historical trends and be morally untenable. Selection criteria should primarily depend on individual attributes such as age, education, and income levels. Obviously, what type of people come to host countries matters. In fact, even the United States has long cared about the composition of immigrants. It sets per- 16 The effect of unauthorized immigrants is unclear. On one hand, they may be more likely to be involved in crimes. They might be more crime-prone, or their marginalized circumstance might encourage them to commit crime. On the other hand, they might avoid contact with law enforcement agencies even for minor offenses so that their illegal statuses are not revealed (Butcher & Piehl, 2007). In fact, NPA reported (2011) that illegal residents share of total arrested foreigners accounted for only 7% in Legalization raises other questions, for example, whether it might jeopardize the principle of the rule of law and the integrity of the nation s immigration system, and attract further unauthorized immigrants. However, these questions are beyond the scope of this paper. 25