Running head: School District Quality and Crime 1 School District Quality and Crime: A Cross-Sectional Statistical Analysis Chelsea Paige Ringl Department of Sociology, Anthropology, Social Work, and Criminal Justice Oakland University Abstract This thesis explores the effects that school district quality has on crime rates in corresponding law enforcement zones in Michigan. I also analyze the additional factors of high school graduation rates, poverty, racial disadvantage, and population density, drawing on data compiled from the Michigan Department of Education, American Community Survey, US Census, and the Uniform Crime Reports. The ultimate aim is to determine how these five factors together affect criminal activity. Many studies have been conducted to research how these variables influence crime independently, but few have examined their collective effect. Additionally, focusing on the quality of a school district, as opposed to the quantity of education in the population offers a new method for analyzing how education impacts crime. This fresh perspective benefits law enforcement, school administration, and scholars in these fields by emphasizing the role that quality education plays in shaping the life course of students. The results of this analysis show that school district quality affects violent and property crime rates both directly, and through mechanisms involving the mediating variables of high school graduation and poverty rates. Based on these results, new crime prevention strategies can be explored to address the strong relationship between these key variables and crime rates.
School District Quality and Crime 2 School District Quality and Crime: A Cross-Sectional Statistical Analysis The areas of crime reduction and prevention have been focal points in scholarship and public policy for many decades. Social scientists, politicians, and countless interest groups continually investigate new potential causes of crime, along with inventive methods for preventing crime based on these causes. Some research addresses crime at the micro level with variables that influence an individual s propensity for crime, while others focus on causes of crime at the macro level that explain why crime rates vary from place to place. At the macro level we can examine structural inequalities and differences in descriptive population statistics at levels such as the city or state in order to draw conclusions on specific crime trends and correlating variables. The current analysis focuses on the macro-level variable of school district quality and the influence it has on crime rates, while accounting for multiple mediating and control variables. The goal of this research is to address the question of what direct and indirect effects quality education has on crime rates at the macro level. Current Research Robert Agnew s general strain theory is one lens through which the relationship between education and crime can be examined and understood. This theory suggests that there are three general types of strain which can lead to the commission of crime. The types of strain include the failure to achieve positively valued goals, the removal of positively valued stimuli, and the presentation of negatively valued stimuli (Agnew, 1992). Examples of these strains, respectively, can include the inability to graduate from high school or obtain a job, the repossession of a home due to bankruptcy, or racial discrimination in the workplace. According to this theory, strain causes individuals to experience negative affect which can result in a variety of criminal behaviors depending on the available opportunities (Agnew, 1992). At a macro level, poor quality education, low graduation rates, and high concentrations of impoverished racial minorities can result in cities or communities that are embedded with strain. These cities can therefore become breeding grounds for criminal activity. One method for reducing the crime rates in the city could be to raise the quality of the school district, which could improve graduation rates, lower poverty rates, and diminish strain in the community as a whole. Many scholars have examined the impact of education on crime rates at both the micro and macro levels. The majority of these studies focus on the issue of how crime rates are affected by the quantity of education in the population. Studies of self-report, arrest, and imprisonment data suggest a strong negative correlation between years of schooling and crime commission (Lochner, 2004; Lochner & Moretti, 2004; Machin, Marie, & Vujic, 2011). One
School District Quality and Crime 3 study in England and Wales found that a one percent decrease in the percentage of the population with no educational qualifications, via an increase in the school leaving age, would result in approximately 40,000 fewer property crimes per year (Machin, Marie, & Vujic, 2011). Building on the connection between education and crime, the effect of high school graduation rates is also of interest. In 2012, 35% of inmates in the United States possessed a high school diploma compared to 82% of the general population (Deming, 2012). As a result of graduating from high school, a student s human capital is increased, which increases job opportunities, reduces strain, and also increases the risk and cost associated with crime commission (Lochner, 2004; Lochner & Moretti, 2004). These findings indicate that the social, economic, and physical benefits of education are astronomical and critical to crime prevention. Other environmental factors such as poverty and racial disadvantage have also been studied in relation to crime rates. A lack of social capital in poverty-stricken communities, potentially caused by a lack of quality education, can cause an increase in both unemployment and crime rates (Short, 1991; Piquero & Sealock, 2010). One analysis found that at the census tract level increased poverty rates have a strong positive correlation with aggravated assault, murder, robbery, burglary, and car theft (Hipp & Yates, 2011). Additionally, racial disadvantage plays a role in this relationship in multiple capacities, as African Americans and other racial minorities face high rates of institutional discrimination and other risk factors which cause strain and increase their propensity for crime (Duster, 1987; Short, 1991; Smith, Devine, & Sheley, 1992; McNulty, 1999; Piquero & Sealock, 2010). Population density is another factor that impacts crime rates at the macro level. In a study of Detroit neighborhoods, population density was found to have a significant negative effect on both property and violent crime rates (Raleigh & Galster, 2015). However, a positive relationship between population density and property crime rates has also been suggested by research. Based on general strain theory, it would be logical for areas with high population densities to have higher crime rates because of the lack of housing and employment opportunities, which may result in strain (Agnew, 1992). One study found that areas with high population density displayed higher rates of property crime because of the increased number of potential targets and the decreased likelihood of being arrested (Jabbar & Mohsin, 2013). Unlike previous studies, the current thesis synthesizes the five variables addressed above into one macrolevel, cross-sectional analysis. There is extensive research in each individual area however, there is a lack of research connecting all of the variables and comparing the strength of their relationships while simultaneously
School District Quality and Crime 4 differentiating the effects on violent and property crime rates. Instead of addressing risk factors for crime at the individual level, my concern is space, and the city-level structural conditions that can have the greatest impact on crime rates. Additionally, by focusing on the quality, as opposed to the quantity, of education in a city I offer a rational and unique perspective on the topic which has yet to be explored thoroughly. This research is critical to establishing functional crime prevention strategies that can benefit law enforcement agencies, the community, and those who inhabit it. Methodology To begin, I compiled data for a sample of 119 Michigan cities. I started by obtaining the 2012 Uniform Crime Reports for all available police departments in Michigan. These departments were generally associated with the largest cities in Michigan because many small police departments choose not to make their crime data available for the reports. Using information from the Michigan Department of Education, the United States Census, and the American Community Survey I then obtained data on individual school qualities, high school graduation rates, poverty, racial disadvantage, and population density for each corresponding city. Cities that lack data for any of the variables were excluded from the analysis. Measures The Federal Bureau of Investigation provides statistical data for violent and property crimes across the country through a yearly publication of the Uniform Crime Reports. Individual police departments voluntarily share information with the FBI on the number of crimes encountered during the year, which is then compiled to create the reports. Not every law enforcement jurisdiction provides data for the reports so the sample of cities in this thesis is based primarily on the availability of the crime data. The crime rates in these reports are calculated based on the number of incidents per 100,000 inhabitants. The violent crime rate includes measures of murder and nonnegligent manslaughter, forcible rape, robbery, and aggravated assault. Table 1 contains descriptive statistics for all variables in the analysis and shows the range of violent crime rates from 9.7 to 2,729.5, with an average of 417.743. The property crime rate includes measures of burglary, larceny-theft, motor vehicle theft, and arson. Table 1 also shows the range of property crime rates from 462.5 to 8,593.6, with an average of 2,798.696. I calculated the primary independent variable of school district quality based on the Top-to-Bottom Rankings of Michigan schools for the 2011-2012 school year, provided by the Michigan Department of Education. In this assessment each individual school is given a rating between 0 and 100 based on student performance in
School District Quality and Crime 5 various academic topics, graduation rates, and overall improvement from previous years. For each law enforcement zone, I identified and averaged the ratings of all of the schools in the school district(s) serving the area to arrive at one rating for each law enforcement zone. Table 1 displays the range of calculated school district qualities from 4.7 to 92.6, with an average rating of 48.545. The American Community Survey from 2011 provides information on the percent of people over 25 years old who have graduated from high school. This variable is treated as a mediating variable between school district quality and crime rates. Theoretically, if the school district quality is higher there should be higher graduation rates which could translate into greater economic opportunity and less strain-driven crime. As seen in Table 1, this percentage ranges from 60.0% to 97.8%, with an average of 88.736%. The American Community Survey from 2011 also provides information on the poverty rates for each city in the sample. Poverty rate refers to the percentage of individuals and families that fall below the federally-dictated poverty thresholds. These thresholds vary based on factors such as the number of children in the family or the age of the individual. This variable is also treated as a mediating variable between school district quality and crime rates because higher quality school districts should reduce poverty rates and thus, reduce criminal activity. Table 1 indicates that the poverty rates range from 1.7% to 48.1%, with the average rate being 15.814%. I included racial disadvantage as a control variable in this analysis because there is evidence to suggest that concentrations of minority races can be correlated with crime rates. Structural inequalities can severely limit opportunities for minorities, causing economic and social strain, which can result in crime commission (Duster, 1987; Smith, Devine, & Sheley, 1992; Piquero & Sealock, 2010). I used racial distribution data from the 2010 US Census in this thesis to represent racial disadvantage. For each city, the percent white is subtracted from 100 to calculate the percent minority. In this sample of cities, Table 1 shows that the percent minority ranges from 3.8% to 97.1%, and the average is 24.062%. I also included population density, or number of people per unit of area, as a control variable because densely populated areas can have greater risk for criminal activity, specifically if the area is impoverished. In densely populated areas, economic and social opportunities can be limited which can result in strain and crime commission. Although there is evidence to suggest that the effect of population density can be negative at times, my theoretical perspective suggests that the relationship will be positive. The 2010 US Census provides the population
School District Quality and Crime 6 density per square mile of land area for each city in the sample. Table 1 displays the range of 305.4 to 10,751.0, with an average of 2,695.916. Analytic Strategy Following the compilation of data, I performed multiple statistical analyses using the SPSS software. I obtained descriptive statistics and a correlations matrix using the software and have included them as Tables 1 and 2. Subsequently, I ran three statistical regression models that gain complexity as each variable is added. I ran each model twice, once with violent crime rate as the dependent variable and again with property crime rate. The first model examines the relationship between school district quality and crime rates, controlling for racial disadvantage and population density. The second model adds high school graduation rates to the equation as a mediating variable. The final model adds poverty rates to complete the analysis. I examine the strength of each relationship as the models gain complexity and identify the variables with the strongest correlations. Results Descriptive Statistics To begin, the descriptive statistics for violent and property crime rates (Table 1) are critical to understanding the results from the three models. The violent crime rate has a much smaller range (9.7 to 2,729.5) than the property crime rate (462.5 to 8,593.6). This is important to note when examining the results of the models because there is much more variance to account for with property crime rates than there is for violent crime rates. Thus, it should be expected that the independent variables will account for a smaller percentage of the property crime rates variance. Correlations The correlations matrix (Table 2) shows the strength of the bivariate relationships between all variables in the analysis. In every case, except for population density, the correlations are significant at the.01 alpha level. The high correlations between school district quality and the mediating variables of percent graduated high school (.653) and percent poverty (-.568) support the theoretical mechanisms suggested in previous research. Interestingly, school district quality has a stronger correlation to violent crime rate (-.615) than to property crime rate (-.588). A similar pattern is found with all of the other variables in relation to violent and property crime rates. Additionally, the correlations between percent graduated high school, percent poverty, and percent minority are high enough to merit caution. Since these are all independent variables the issue of collinearity must be addressed to ensure the validity of the results.
School District Quality and Crime 7 Regression Models Table 3 shows the results from the three regression models addressing violent crime rates. The unstandardized coefficients (b), standardized coefficients (β), and standard error (SE) are represented in the table. Model 1 examines the effect that school district quality has on violent crime rates, controlling for percent minority and population density. According to the unstandardized coefficient for school district quality, a one point increase in the rating of the school district would result in 8.132 fewer violent crimes per 100,000 people. In terms of the standardized coefficient, one standard deviation increase in school district quality would lead to a.380 standard deviation decrease in violent crime rates. Throughout the analyses school district quality remains significant at the.001 alpha level, showing a strong connection with violent crime rates regardless of the impact made by the other independent variables. Model 1 also indicates that a one percent increase in the percent minority in a city would increase violent crime by 13.240 incidents per 100,000 people, which is significant at the.001 alpha level. Population density, however, never reaches a level of significance in any of the violent crime rate models and has a relatively small impact overall. In Model 1 these three variables account for 58.6% of the variance in in violent crime rates. In Model 2, the percent of the population who have graduated high school is added to the equation and has a significant effect, but only at the.05 alpha level. The unstandardized coefficient shows that a one percent increase in the percentage of high school graduates in a city reduces violent crime occurrences by 15.487 per 100,000 people. In this model the effect that a one point increase in school district rating has on violent crime rates is diminished to a 6.147 incident reduction per 100,000 people. The coefficient for school district quality is lower here because the effect of school district quality on violent crime rates is partially mediated by the percent graduated high school variable, as expected. A similar mechanism is at play for the percent minority variable, which also has a slightly smaller coefficient after adding the percent graduated high school variable. In Model 3, including percent poverty results in a staggering decrease in the effect made by the percent graduated high school variable, dropping it to a level of insignificance, while percent poverty shows significance at the.05 alpha level. The unstandardized coefficient indicates that a one percent increase in the percentage of people below the poverty line will result in 10.619 more violent crimes per 100,000 people. Again, the coefficients for school district quality and percent minority are reduced in this model. These results imply that the effects that school district quality and percent minority have on violent crime rates are mediated by both percent graduated high school
School District Quality and Crime 8 and percent poverty. In the final model, when all five independent variables are included in the analysis 62.4% of the variance in violent crime rates is accounted for. The results for the models addressing property crime rates are displayed in Table 4. As in Table 3, the unstandardized coefficients (b), standardized coefficients (β), and standard error (SE) are represented in Table 4. The coefficients in these models are significantly larger than those in the violent crime rate models because of the drastically larger range of property crime rates. Model 1 indicates that a one point increase in school district rating results in a reduction of 28.339 incidents of property crime per 100,000 people. Again, school district quality remains significant at the.001 alpha level across the three models. In Model 1, percent minority is also significant at the.001 alpha level and a one percent increase in this variable reduces property crime by 29.856 incidents per 100,000 people. In this model, population density has a minimal, statistically insignificant effect on property crime rates. These three variables account for 46.4% of the property crime rate variance in this model. Model 2 includes the percent graduated high school variable, which does not reach a level of significance. The unstandardized coefficient shows that a one percent increase in the percentage of high school graduates will diminish the number of property crimes by 32.234 incidents per 100,000 people. Similar to the violent crime rate models, the addition of this variable reduces the coefficients for both school district quality and percent minority, indicating that these variables also impact property crime rates partially through the percent graduated high school variable. Interestingly, in Model 2 the coefficients for population density increase and reach the.05 alpha level of significance. Here, a one unit increase in the population density reduces property crime by.142 incidents per 100,000 people. Percent poverty is included in Model 3 and is significant at the.05 alpha level. In this model, a one percent increase in the percentage of people living in poverty increases property crime by 32.454 incidents per 100,000 people. As seen in the violent crime rate models, the coefficients for school district quality, percent minority, and percent graduated high school are reduced when percent poverty is added to the equation. In Model 3, a one point increase in school district rating will diminish property crime by 22.468 incidents per 100,000 people. The effect of percent graduated high school is cut dramatically and a one percent increase only prevents 7.326 incidents of property crime per 100,000 people. Percent minority also has a weakened effect which is now only significant at the.01 alpha level. Finally, the population density variable remains significant at the.05 alpha level and reduces property crime incidents by.135 per 100,000 people for every one unit increase in population density. The drastic
School District Quality and Crime 9 changes in these coefficients indicate that poverty is a key mediating variable when assessing the effect that school district quality has on property crime rates. When all variables are included in Model 3, 49.3% of the variance in property crime rates is accounted for. Discussion The results of this analysis are consistent with existing research on the subject of education and crime. The regression models for both violent and property crime rates show that school district quality has a statistically significant negative effect, even when accounting for mediating and control variables. The models also suggest that the effect of school district quality works partially through multiple mediating variables including percent graduated high school and percent poverty. A possible mechanism to explain these findings is that a high quality school district can result in high graduation rates which offer students the opportunity for economic success. At the macro level, this can lower poverty rates and ultimately reduce both violent and property crime rates through a reduction in strain. The results of this analysis offer support for this theoretical mechanism while also suggesting that school district quality has an independent effect on both violent and property crime rates. The effect that the percent poverty variable has on the percent graduated high school variable is also of interest. When percent poverty is added to the final regression model for violent crime rates, the significance level of percent graduated high school falls dramatically. Logically, a mechanism through which high school graduation affects crime is through avoiding the strain of poverty, so when poverty is added as a separate variable the direct effect of high school graduation on crime is diminished. A similar effect is seen in the property crime rates model with respect to percent minority. Percent minority is significant at the.001 alpha level until percent poverty is added, causing the significance to drop to the.01 alpha level. Because of the high correlation between these two variables in this analysis and the structural connections between race and poverty in the real world it is difficult for the software to disentangle the effects that these two variables have on crime rates. Simply being a part of a racial minority does not cause someone to commit a crime, but the straining conditions that those minorities are forced to live under, such as poverty, can result in an inclination towards criminal behavior (Duster, 1987; Short, 1991; Smith, Devine, & Sheley, 1992; McNulty, 1999; Piquero & Sealock, 2010). This analysis supports this point by showing that as the impact of poverty is accounted for in the model, the impact of percent minority is diminished. Theoretically, the variables that have been included in this analysis should have a stronger connection to the property crime rates because they deal mostly with economic strain and human capital (Agnew, 1992; Piquero &
School District Quality and Crime 10 Sealock, 2010). While there are many personal factors such as mental illness, childhood abuse, or drug use that can lead to violent crime, property crime tends to stem from a lack of economic stability. Since low quality education can result in lack of employment opportunities and low human capital, it is reasonable to assume that the variables included in this analysis would have a stronger impact on property crime rates than violent crime rates. In this analysis, the opposite effect is observed. More of the variance in violent crime rates (62.4%) is accounted for by the five independent and control variables than the property crime rates (49.3%). This phenomenon is based on the staggering difference in the ranges of violent and property crime rates. Considering the fact that the range of violent crime rates is approximately three times smaller than the range of property crime rates, it is logical for less of the variance in property crime rates to be accounted for, regardless of the nature of the independent and mediating variables. Additionally, this explains why the correlations between the independent variables and violent crime rates are slightly stronger than the correlations with property crime rates. Limitations One factor which could limit the validity of this analysis is collinearity. Collinearity or multicollinearity in multiple regression analyses refers to high correlations between two or more independent variables which can skew the results of the regressions (Lewis-Beck, 1980). There are multiple methods used for detecting the presence of collinearity including measures of variance inflation factors (VIF) and tolerance, which are used in this thesis. However, there are discrepancies throughout the literature in deciding at what point these measures indicate a true problem because it is imperative that they are viewed within the context of one s own analysis. For example, O Brien (2007) states that, Not uncommonly a VIF of 10 or even one as low as 4 (equivalent to a tolerance level of 0.10 or 0.25) have been used as rules of thumb to indicate excessive or serious multi-collinearity (p. 674). The Statistical Consulting Group at UCLA also cites a VIF of 10 as indicating an issue with collinearity (Chen, Ender, Mitchell, & Wells, 2003). In the regression models for this thesis, the highest VIF encountered is 2.696 in Model 3 for both violent and property crime rates. Despite the high correlations between independent variables seen in Table 2, the measures of collinearity never become high enough to merit adjustment of the models. Collinearity is present in the models so the results may be slightly skewed but in this case the effect of collinearity was negligible. However, in similar future investigations issues resulting from collinearity must be accounted for. A second limitation for nearly all research addressing crime is omitted variable bias. There are many reasons why it is impossible to include every variable which affects crime. One reason is that some variables are not
School District Quality and Crime 11 measurable and/or data for these variables is not available. For example, it would be extremely difficult to measure strain as dictated by general strain theory because the number of factors which can cause strain in an individual is so vast and what causes strain for one person may be seen as positive to another (Agnew, 1992). Another reason that omitted variable bias is unavoidable is the impact on the correlations between independent variables. If twenty variables which affect crime are included in the same analysis it would be impossible to differentiate their effects and the results would be highly skewed due to collinearity. A final reason that accounting for all related variables is impossible is that the academic community has yet to exhaust the possibility of variables that affect crime. The macro level of analysis involves complex, inter-connected variables which all may have an impact on crime through various mechanisms so no study could possibly address every variable in a valid, understandable fashion. Implications for Future Research and Public Policy There are many avenues through which future research can expand on the current analysis. One option is to make the school district quality variable more inclusive by adding additional factors, such as the ratio of students to teachers. This would give a more accurate picture of the quality of education being received by the students. Also, including private institutions in the analysis would offer a better representation of the quality of education in the area as a whole. If private schools are utilized by a sizable percentage of the population in an area and offer higher quality education then including them in an analysis would yield different results than simply focusing on public schools. Additionally, examining how these variables impact crime rates in different environments would be a useful expansion to this thesis. Comparing urban areas to rural areas, or high population cities to low population cities would offer insight into the significance of the social environment. It would be interesting to see if school district quality has a more or less significant effect on crime rates in areas where poverty and/or minority populations are low. It might also be beneficial to take samples of cities in different regions of the United States outside of Michigan and compare the results from running the same models. In terms of policy implications the result of this thesis are clear. It will take efforts from multiple organizations working together to affect these crime-conducive variables in order to diminish crime rates. Educational organizations, social welfare groups, and law enforcement agencies can all have an impact on crime prevention because the variables which predict crime can be addressed on multiple fronts. For example, research suggests that economic opportunity and stability are major factors impacting the crime rates in an area and there are
School District Quality and Crime 12 many avenues for addressing the variable. Programs can be implemented to increase the quality of the school district through hiring more qualified teachers, offering tutoring centers, and increasing the number of extracurricular activities available to students. By increasing the quality of education that students receive it is more likely that they will graduate from school and have opportunities for economic success. Economic stability can also be affected through welfare programs and community programs that focus on employment training and acquisition. These community-based programs would also increase the levels of social support and social organization, which were not included in this analysis but can have an effect on crime rates. The results of this thesis are also beneficial to law enforcement agencies and organizations that work to control crime directly. When developing new crime prevention strategies it is essential to understand what factors have the strongest correlation to crime rates so resources are not wasted. Based on this analysis, school district quality and percent poverty have the greatest overall impact on crime rates. This implies that police should focus their efforts in areas with high poverty rates and/or low quality schools. Instead of trying to increase police presence in the entire city, for example, patrolling these high-risk areas will be more effective at reducing the crime rates while saving the department time, money, and resources. By expanding our knowledge on the causes of crime adjustments and additions can also be made to existing programs to focus on the most pertinent issues and increase effectiveness and efficiency. Conclusion Academia may never reach a point where all crime can be understood through simple mechanisms. However, this thesis is one example of how many variables can be compiled into a complex analysis in order to gain a more inclusive understanding of the variance in both violent and property crime rates. This analysis shows how school district quality and poverty rates have the highest correlations to both violent and property crime rates and has offered various implications for future research and public policy. Increasing our understanding of the mechanisms which impact crime is critical to the improvement of society as a whole, even in areas which do not suffer from high crime rates. Also, reducing crime rates at the city level is one step in the right direction to addressing large scale problems such as prison overcrowding and mass incarceration at the national level. It is imperative that we continue to research and develop crime prevention strategies that address a variety of social variables and mechanisms in order to make progress in addressing these critical social problems.
School District Quality and Crime 13 Bibliography Agnew, R. (1992). Foundation for a general strain theory of crime and delinquency. Criminology, 30(1), 47-87. Chen, X., Ender, P., Mitchell, M., & Wells, C. (2003). Regression diagnostics. In Regression with SPSS (chapter 2). Retrieved from http://www.ats.ucla.edu/stat/spss/webbooks/reg/default.htm. Deming, D. J. (2012). Does school choice reduce crime? Evidence from North Carolina. Education Next, 12(2), 70-76. Duster, T. (1987). Crime, youth unemployment, and the black urban underclass. Crime & Delinquency, 33(2), 300-316. Hipp, J. R., & Yates, D. K. (2011). Ghettos, thresholds, and crime: Does concentrated poverty really have an accelerating increasing effect on crime? Criminology, 49(4), 955-990. Jabbar, S. M., & Mohsin, H. M. (2013). Economics of property crime in Panjab. The Pakistan Development Review, 52(3), 221-233. Lewis-Beck, M. S. (1980). Applied regression: An introduction. (Sage University paper series on quantitative applications in the social sciences, series no. 07-022). Newbury Park, CA: Sage. Lochner, L. (2004). Education, work, and crime: A human capital approach. International Economic Review, 45(3), 811-843. Lochner, L., & Moretti, E. (2004). The effect of education on crime: Evidence from prison inmates, arrests, and selfreports. The American Economic Review, 94(1), 155-189. Machin, S., Marie, O., & Vujic, S. (2011). The crime reducing effect of education. The Economic Review, 121(552), 463-484. McNulty, T. L. (1999). The residential process and the ecological concentration of race, poverty, and violent crime in New York City. Sociological Focus, 32(1), 25-42. O Brien, R. M. (2007). A caution regarding rules of thumb for variance inflation factors. Quality & Quantity, 41, 673-690. Piquero, N. L., & Sealock, M. D. (2010). Race, crime, and general strain theory. Youth Violence and Juvenile Justice, 8(3), 170-186. Raleigh, E., & Galster, G. (2015). Neighborhood disinvestment, abandonment, and crime dynamics. Journal of Urban Affairs, 37(4), 367-396. Short, J. F., Jr. (1991). Poverty, ethnicity, and crime: Change and continuity in U.S. cities. Journal of Research in Crime and Delinquency, 28(4), 501-518. Smith, M. D., Devine, J. A., Sheley, J. F. (1992). Crime and unemployment: Effects across age and race categories. Sociological Perspectives, 35(4), 551-572.
School District Quality and Crime 14 Table 1. Descriptive statistics Variable Mean Standard deviation Min Max Dependent variables Violent crime rate 417.743 504.290 9.7 2,729.5 Property crime rate 2,798.696 1,541.084 462.5 8,593.6 Independent variables School district quality 48.545 23.572 4.7 92.6 Percent graduated high school 88.736 6.300 60.0 97.8 Percent poverty 15.814 10.848 1.7 48.1 Control variables Percent minority 24.062 19.890 3.8 97.1 Population density 2,695.916 1,701.654 305.4 10,751.0 * p <.05; ** p <.01 (two-tailed tests) Table 2. Correlations 1. 2. 3. 4. 5. 6. 7. 1. Violent crime rate 1.000.714** -.615** -.607**.673**.689**.148 2. Property crime rate 1.000 -.588** -.508**.583**.559**.047 3. School district quality 1.000.653** -.568** -.470** -.204* 4. Percent graduated high school 1.000 -.706** -.553** -.369** 5. Percent poverty 1.000.649**.235* 6. Percent minority 1.000.230* 7. Population density 1.000 * p <.05; ** p <.01 (two-tailed tests)
School District Quality and Crime 15 Table 3. Violent crime rate regression models Variable/Measure Model 1 Model 2 Model 3 b b b β β β (SE) (SE) (SE) School district quality -8.132*** -6.147*** -5.577*** -.380 -.287 -.261 1.463 1.695 1.673 Percent graduated high school -15.487* -7.337 -.193 -.092 6.995 7.582 Percent poverty 10.619*.228 4.261 Percent minority 13.240*** 11.866*** 9.801***.522.468.387 1.744 1.824 1.967 Population density -.015 -.027 -.024 -.050 -.090 -.082.018.019.018 R².586.603.624 * p <.05; ** p <.01; *** p <.001 (two-tailed tests) Table 4. Property crime rate regression models Variable/Measure Model 1 Model 2 Model 3 b b b β β β (SE) (SE) (SE) School district quality -28.339*** -24.207*** -22.468*** -.433 -.370 -.344 5.087 5.976 5.939 Percent graduated high school -32.234-7.326 -.132 -.030 24.661 26.912 Percent poverty 32.454*.228 15.123 Percent minority 29.856*** 26.996*** 20.685**.385.348.267 6.065 6.430 6.981 Population density -.118 -.142* -.135* -.130 -.157 -.149.064.066.066 R².464.472.493 * p <.05; ** p <.01; *** p <.001 (two-tailed tests)