The Relationship between the Business Cycle and Crime. A Literature Review

Tesis de Máster , 2018 , 90 Páginas , Calificación: 1,3

Economía de las empresas - Otros

From the industrialized countries to newly industrializing and developing countries of the world, violence and crime threaten social stability and are becoming major obstacles in development. It is a global phenomenon that results in the loss of property and/or lives, creates misery and is one of the most harmful social problems throughout the world. Several scholars estimate these costs with $160 billion up to $3,500 billion annually for the United States. Given these tremendous negative effects, it has become of central interest to policy makers and scholars of economics, sociology, and criminology. Understanding the determinant factors of crime is critical and provides a powerful tool to drastically reduce costs to society. Furthermore, there is no doubt that the business cycle has an omnipresent effect on economic activity, employment, individuals, and a variety of other social and economic variables.

List of Contents

ListofFigures

ListofTables

List of Abbreviations

List of Symbols

1. Introduction

2. Stylized Facts

3. The Business Cycle and Crime
3.1 Economics of Crime
3.1.1 Theoretical Frameworks
3.1.2The Argument of Cantor andLand(1985)
3.1.3 The Problem of Identification
3.2 Labor Market Opportunities and Crime
3.2.1 Simple Linear Regression Models
3.2.2 Instrumental Variable Approaches
3.2.3 Alternative Strategies
3.3Economic GrowthandCrime

4. Combining the Results

5. DiscussionandConclusion

References

List ofFigures

Figure 1: Property and Violent crime in the U.S., 1960-2016

Figure 2: Property and Violent crime per 100,000 population in the U.S., 1960-2016

Figure 3: Distribution of Property crime in the U.S., 2016

Figure 4: Distribution of Violent crime in the U.S., 2016

Figure 5: Distribution of Property and Violent crime per 100,000 population across U.S. states, 2016

Figure 6: Theft rate per 100,000 population in the European Union, 2015

Figure 7: Robbery rate per 100,000 population in the European Union, 2015

Figure 8: Unemployment Effect

Figure 9: Path Diagram of Structural Relationships and Reduced-Form Effects of Unemployment on Crime

Figure 10: Unemployment rate and Property crime rate per 100,000 population in the U.S., 1960-2016

Figure 11: Unemployment rate and Violent crime rate per 100,000 population in the U.S., 1960-2016

Figure 12: IRF of Auto theft to one-standard deviation Shock in Auto theft, Income, and Unemployment, 1964-2001

Figure 13: IRF ofFraud to one-standard deviation Shock in Auto theft, Income, and Unemployment, 1964-2001

List ofTables

Table 1: Regression results of State Unemployment and variables measuring State Demographic Structure on Property and Violent Crime, 1971-1999

Table 2: Regression results of State Unemployment on Crime subcategories, 1971-1999

Table 3: Regression results ofUnemployment and State Demographic variables on Property and Violent crime, 1974-2000

Table 4: Regression results of Contemporaneous and Lagged Unemployment on Property andViolentcrime, 1950-1990

Table 5: Regression results ofWages and Unemployment of non-college-educated males on Property and Violent crime, 1979-1997

Table 6: First-stage results of State Unemployment Rates on State Military Contracts per capita and State-Level Measure of Oil Costs

Table 7: OLS and 2SLS results of State Unemployment Rates on Crime, 1971-1999

Table 8: First-Stage results of changes in Real Exchange Rates on Unemployment

Table 9: First-Stage results of Oil Prices on Unemployment

Table 10: OLS and 2SLS results ofUnemployment on Property and Violent Crime, 1974­2000

Table 11: SinglelnstrumentalVariableresults ofUnemployment on Crime, 1974-2000 .

Table 12: OLS and 2SLS results ofWages and Unemployment of non-college-educated males on Property and Violent crime, 1979-1997

Table 13: OLS and 2SLS results ofUnemployment on Property crime, 1996-2003

Table 14: Regression results ofNYC Unemployment and Wages on Property and Violent crime, 1974-1999

Table 15: Regression results ofUnemployment and First-differenced Unemployment rates on Crime, 1978-2005

Table 16: Regression results ofUnemployment on Crime, 1975-2010

Table 17: Regression results ofEconomic variables on Homicide and Robbery Rates, 1970-1994

Table 18: Overview of the empirical findings of the effect ofUnemployment on Total crime and Property crime

Table 19: Overview of the empirical findings of the effect of unemployment on Violent crime

Table 20: Overview of the empirical findings of the effect ofWages and Income on Crime.

Table 21: Overview of studies estimating the cost of crime to society

List of Abbreviations

Abbildung in dieser Leseprobe nicht enthalten

List of Symbols

Latin characters:

Abbildung in dieser Leseprobe nicht enthalten

Greek characters:

Abbildung in dieser Leseprobe nicht enthalten

1. Introduction

From the industrialized countries to newly industrializing and developing countries of the world, violence and crime threaten social stability and are becoming major obstacles in development. It is a global phenomenon that results in the loss of property and/or lives, creates misery and is one of the most harmful social problems throughout the world. Several scholars estimate these costs with $160 billion up to $3,500 billion annually for the United States (U.S.).^{1 2} Given these tremendous negative effects, it has become of central interest to policy makers and scholars of economics, sociology, and criminology. Understanding the determinant factors of crime is critical and provides a powerful tool to drastically reduce costs to society. Furthermore, there is no doubt that the business cycle has an omnipresent effect on economic activity, employment, individuals, and a variety of other social and economic variables. The National Bureau of Economic Research (NBER) Business Cycle Dating Committee states that:

“During a recession, a significant decline in economic activity spreads across the economy and can last from a few months to more than a year. Similarly, during an expansion, economic activity rises substantially, spreads across the economy, and usually lasts for several years." '

Thus, it is reasonable that the business cycle also influences criminal activity. Conventional belief is that crime is countercyclical: it increases during recessions and decreases during economic expansions. This stems from the idea that crime is a substitute for legitimate sources of income. In discussion of potential impacts of the economy on crime rates, many scholars and policy makers often consider labor market opportunities an important factor in the supply of criminal activities. Here, the unemployment rate is by far the economic indicator of choice.

In his seminal work, Becker (1968) is the first to incorporate economic facets into crime theory.³ He develops a theoretical framework that models the social cost of crime, the cost of apprehension and conviction, the supply of offenses, and punishments. Accordingly, the number of offenses committed by an individual is a function of his/her probability of conviction, his/her punishment, and other variables. Although it forms the groundwork in this area of research, this model mainly focuses on how the supply of offenses is affected by changes in the probability of being caught and the subsequent punishment(s). It says little about the link between labor market opportunities itself and crime. Ehrlich (1973) further develops Becker (1968)’s framework into a time allocation model where individuals choose between legal and illegal activities. These theoretical works have stimulated an extensive amount of scientific research attempting to establish a link between unemployment and crime. Yet, according to Chiricos (1987), who provides an early overview of 63 studies, analyses on this relationship are not clear-cut. Although some studies indicate a positive link, most of these suggest that the connection is weak or non-existent.⁴ Due to the development of unemployment and crime during the 1990s in the U.S, further attention was brought to a possible link. More recent research repeatedly demonstrates a significant positive relation for property crime: with unemployment, the opportunity cost of committing crime - namely, the legal wage - declines, which makes illegal income preferable. Accordingly, labor market opportunities affect the choice between legal and illegal income sources. Because violent crime is rarely economically motivated, a weak relationship between unemployment and violent crime can be expected.

The remainder of the paper is organized as follows: Chapter 2 gives a short overview on crime rates in the U.S. and Europe, Chapter 3 introduces underlying theoretical frameworks and core results using different methodological approaches and Chapter 4 combines the results. Chapter 5 discusses politico economic implications and concludes.

2. Stylized Facts

For this work it is important to understand the magnitude of crime and its tremendous effects on society. To do so, I will provide a brief overview of characteristic numbers for the U.S. and Europe.

First off, there is a consistent categorization of felonies: politics, criminologists, statistics, and scientific literature usually distinguish between property and violent crimes. The first are the more common type of criminal offense and do not involve any force or threat to the victim and range from lower-level criminal acts such as shoplifting to multi-million-dollar thefts. Property crimes are commonly composed of four offenses: burglary, larceny, theft, auto theft, and arson. Violent crimes are defined as offenses that involve threat or force and include murder, rape, robbery, and aggravated assault. These definitions are in line with the Federal Bureau of Investigation’s (FBI) yearly Uniform Crime Reporting (UCR) Program and represent the cornerstone of most empirical analysis. The UCR Program’s objective is to provide a reliable set of criminal justice statistics for law enforcement administration, operation, and management. This program is a nationwide, cooperative statistical effort of approximately 18,000 city, university and college, county, state, tribal, and federal law enforcement agencies voluntarily reporting data on crimes to the FBI. Figure 1 depicts the property and violent crimes in the U.S. for the period 1960-2016.

Figure 1: Property and Violent crime in the U.S., 1960-2016. Source: FBI Uniform Crime Reports (UCR).

Abbildung in dieser Leseprobe nicht enthalten

As is evident, total property crime began to rise in the mid-1960s and increased constantly until reaching its peak in the early 1990s. Since the mid-1990s, there is an ongoing decline, however, the level in 2016 is substantially higher than in I960. Violent crime behaves very similar to property crime: it increased during the mid-1960s, reached its peak in the early 1990s, and declined afterwards to a level in 2016 that is still four times higher than in I960. It is important to note that the upward trend might be a little deceptive, since there may have occurred changes in crime reporting in the meantime. Thus, differences between 2014 and I960 possibly overstate the true changes in total crimes committed.

Looking solely at the absolute numbers of crimes might be insufficient. Thus, Figure 2 depicts the property and violent crime rate per 100,000 population in the U.S. from I960 to 2016:

Figure 2: Property and Violent crime per 100,000 population in the U.S., 1960-2016. Source: FBI Uniform Crime Reports (UCR).

Abbildung in dieser Leseprobe nicht enthalten

On closer inspection, it becomes obvious that property and violent crime differ slightly after 1970: property crime peaked in 1980, while the violent crime rate reached its maximum in 1991. Both rates increased during the latter half of the 1980s, one less than the other. In the mid-1990s property and violent crime rates declined until the present. In 2016 there were an estimated 1.25 million violent crimes and 7.92 million property crimes in the U.S., which translates to 386 violent and 2,451 property crimes per 100,000 inhabitants.⁵ Figures 3 and 4 display the distribution of those crimes.

Figure 3: Distribution of Property crime in the U.S., 2016. Source: FBI Uniform Crime Reports (UCR).

Abbildung in dieser Leseprobe nicht enthalten

Figure 4: Distribution of Violent crime in the U.S., 2016. Source: FBI Uniform Crime Report (UCR).

Abbildung in dieser Leseprobe nicht enthalten

Larceny-theft makes up almost three-fourths of total property crime while aggravated assault is responsible for two-thirds of violent crime.⁶ Overall, property crime decreased by 1.3 percent compared to2015 figures, while violent crime rose by 4.1 percent.

Focusing on state-level data, Figure 5 depicts the distribution of property and violent crime rates across all states in the U.S.in2016:

Figure 5: Distribution of Property and Violent crime per 100,000 population across U.S. states, 2016. Source: FBI Uniform Crime Report (UCR).

Abbildung in dieser Leseprobe nicht enthalten

Apparently, there are some heavy fluctuations: property crime ranges from 1,512.9 in New Hampshire to 3937.1 in New Mexico and violent crime from 123.8 in Maine to 804.2 per 100,000 population in Alaska. This suggests, looking solely at the aggregate level in empirical analysis might be deceptive, since crime rates seem to vary across time and space.

Furthermore, there are well-documented criticisms of the UCR data that must be considered when looking at these data for any purpose. First off, the FBI does only report the aggregate counts of Part I crimes (see above) and Part II crimes (including simple assault, fraud, prostitution, and driving under influence (DUI)) are not considered. Secondly, UCR reporting requires the use of a hierarchical coding system: if there occur two or more crimes simultaneously, only the most severe one will be counted. Thus, the number of reported crimes is potentially underestimated. Thirdly, and most importantly, the UCR data only contains those crimes reported or known to law enforcement as opposed to all felonies that actually occurred. Fortunately, the U.S. Department of Justice administers an additional statistical program to measure the magnitude, nature, and impact of crime: the National Crime Victimization Survey (NCVS). Two times a year, the U.S. Census Bureau interviews approximately 43,000 households (about 76,000 people) of persons age 12 or older and collects information on crimes suffered by individuals and households, whether or not those crimes were reported to law enforcement.⁷ This survey - started in 1973 and redesigned in 1993 - provides a detailed picture of the identical subset of crimes as the UCR.⁸ It includes demographic information about victims, offenders, and the crimes (time, place, use of weapons, nature of injury, economic consequences), which are collected in the Crime Incident Report (CIR).⁹ According to the NCVS, there were 15.9 million property (36 percent reported to the police) and 5.8 million violent (42 percent reported to the police) victimizations in the U.S in 2O16.¹⁰

UCR and NCVS measure an overlapping, but nonidentical set of offenses. Both exhibit drastic differences in methodology and definitions such that estimates from these data sources do not parallel each other.¹¹ Therefore, both statistics should not be compared but should be viewed as complementary sources. By understanding the strengths and limitations of each program, it becomes clear that together they provide a more comprehensive picture of criminal activity in the U.S than either could produce alone.

For the European Union¹², approximately 1.5 million violent and 10.5 million property crimes - 305.8 violent and 2,086.9 property crimes per 100,000 inhabitants - were recorded in 2015.¹³ Property crime decreased by 3.4 percent and violent crime increased by 1.7 percent compared to 2014.¹⁴ Again, international data average across all national fluctuations, removing potentially useful further insights. Therefore, Figures 6 and 7 representatively display the theft and robbery rates per 100,000 population in 2015 in more detail:

Figure 6: Theft rate per 100,000 population in the European Union, 2015. Source: Eurostat.

Abbildung in dieser Leseprobe nicht enthalten

Figure 7: Robbery rate per 100,000 population in the European Union, 2015. Source: Eurostat.

Abbildung in dieser Leseprobe nicht enthalten

Again, it becomes obvious that there exists a tremendous degree of variability at the national level. The theft rate fluctuates from 78.8 in Cyprus to 3815.5 per 100,000 population in Sweden. Averaging across all EU28 countries yields 1379.1 per 100,000 inhabitants which reduces the information content drastically. The same holds for the robbery rate, where Belgium exhibits with 195.6 approximately 24-times the rate of Cyprus (8.0). For Germany, according to the Polizeiliche Kriminalstatistik (PKS) of the Bundeskriminalamt (Criminal Police Office), there were approximately 200,000 violent and 2.4 million property crimes in 2016.¹⁵ This equals 243.9 violent and 2,926.8 property crimes per 100,000 inhabitants. Overall violent crime increased by 6.7 percent and property crime decreased by 4.4 percent compared to2015 figures.

Concluding, it is important to note that the above approximations for each region are not comparable, since there are differences in definitions of specific crimes, crime categories, and crime reporting. Nevertheless, all figures tell the same story: crime is an existent problem all over the world and responsible for enormous (economic) costs on society.

3. The Business Cycle and Crime

With these dimensions, it is unsurprising that there is a multitude of empirical work that tries to uncover the determinants of crime. The question of whether economic conditions or deterrence policies are more effective tools of crime control has also become an important political issue. A convincing explanation of crime can ideally account for all variation across time and space. Using annual time-series data for four types of Index Crimes between 1933 and 1982, Cook and Zarkin (1985) analyze the business cycle and its impact on various crimes. Employing a nonparametric test based on changes in criminal activity, they find that auto theft is strongly procyclical, burglary and robbery are countercyclical and that homicide rates are not affected. Bushway et al. (2012) replicate and extend the analysis with 26 years of additional data (1933-2008) and confirm that economic contractions indeed increase burglary and robbery rates. Thus, it is tempting to suggest that big macroeconomic factors explain criminal behavior. Furthermore, defining a recession as “a significant decline in economic activity spread across the economy" leaves open the possibility that any number of combinations of economic variables may be affected. In addition, it implies that no two recessions necessarily are the same, indicating that some economic variables may be more affected during some recessions than during others.

3.1 Economics of Crime

Should I mug you or should I burgle your home? Should I cheat on my income tax or should I sell or buy illegal drugs? These are only some of the questions that motivate policy makers and researchers to investigate the economics of crime. Since crime is undoubtfully connected to risk, the attitudes of individuals towards risky behavior are critical in decision-making. Thus, criminal activities are subject to strategic gaming by the police, criminal individuals, and the public, per the Prisoner’s Dilemma: individuals choose between criminal actions and legitimate work based on the expected return to those acts. According to this, I provide a brief theoretical overview of the seminal work of Becker (1969) and Ehrlich (1973). Afterwards I present the argument of Cantor and Land (1985) who argue that earlier results on the relationship between unemployment and crime are ambiguous. Furthermore, I will discuss other factors affecting crime that must be included in econometric analyses. Finally, I will present the core results for the relationship between the business cycle and illegitimate activity, evaluate the methodological approach, and reflect on the data employed. Since unemployment is by far the economic indicator of choice, it forms the better part of this work.

3.1.1 Theoretical Frameworks

The first economic approach to crime, developed by Becker (1968), assumes that individuals are opportunistic beings who compare the expected return when using their time for criminal actions with the expected benefits when spending their time for other activities (e.g., working in the labor market). If the expected return for committing crime is higher than for working in the labor market, individuals will engage in criminal activity. Thus, the sum of all criminal actions of an individual can be modeled as a function of the probability of being arrested, the associated stiffness of sanctions, the expected benefits, the expected return to legal activities, and some other variables.¹⁶ In this case, the expected returns to legal activities are incorporated as opportunity costs in the crime supply function. Therefore, higher expected benefits of legitimate occupations lead to higher opportunity costs and, thus, reduce the number of criminal acts of an individual. In contrast, lower returns to labor decrease the opportunity costs and consequently increase the sum of criminal offenses (Becker, 1968, p. 9). The expected benefits of work within the legal sector mainly depends on individual skills, abilities, working experience, and on the general economic conditions. In this sense, unemployment seems the obvious choice as an indicator for the health of the economy: if unemployment is high, the expected return for working in the labor market is low for many agents. As a result, individuals try to compensate the lower benefits to legitimate work by increasing their supply of criminal actions. Ehrlich (1973, pp. 524-529) further develops this framework into a one-period uncertainty model where individuals allocate their time between legal and illegal activities.

According to this basic economic crime model, individuals allocate their time (t) between illegitimate (tt) and legitimate (t — tt) activities. Here, legal employment and crime are assumed to be substitutes. The individual’s expected benefit is a weighted average of his/her utility in the two alternative states of the world: (i) being caught and getting punished; (ii) not being caught and not getting punished. Equation (1) depicts the expected utility of an individual:

Abbildung in dieser Leseprobe nicht enthalten

This implies that if W'-ff,) — pF'(ti) > W'jft — fy), then the marginal expected returns from crime are greater than the marginal benefits of legitimate work and individuals engage in criminal activities.

Raphael and Winter-Ebmer (2001) incorporate these theoretical approaches into a modified labor-leisure model in a similar way to Grogger (1999).¹⁷ In this framework, labor and crime are not assumed to be substitutes since crimes are not exclusively committed by unemployed individuals.¹⁸ Figure 1 presents the model, where an individual has a fixed amount of working time 0/1 that he/she allocates between legal and illegal income:

Figure 8: Unemployment Effect. Source: Raphael and Winter-Ebmer (2001, p. 263); own representation􀍘􀀃

Abbildung in dieser Leseprobe nicht enthalten

The length AB displays the income when individuals are neither undertaking legal nor illegal activities and represents the so-called reservation wage. Looking at the diagram A, the income from legitimate work is depicted by the straight line CD, whereas the curved line illustrates the income from illegitimate occupations. As apparent, the slope of this curve is declining which means that the marginal return to income-generating crimes is high in the beginning and then decreases with additional crimes committed. These diminishing expected benefits implicate that the probability of getting arrested increases with each additional felony. Further, according to the “rational offender” assumption of the authors, individuals will first undertake income-generating illegal activities which promise high expected returns. Consequently, when expected benefits of illegal acts are still higher than those of legal work, also less promising crimes are committed. Time is only allocated to legal activities when the marginal expected return to labor market work is higher than the expected benefits of crime (Raphael and Winter-Ebmer, 2001, p. 263). In diagram A the returns to lawbreaking acts are higher than the benefits of legal activities until the point C. This implies, that A — t0 hours of available working time are spent on felonies that generate income. Point C represents the situation where the marginal returns to legal and illegal activities are equal. After that point onwards, individuals will devote all the remaining working time to legal labor market work. Given these circumstances, individuals in diagram A will choose a mix of income-generating illegal activities and legitimate work. The budget constraint is then given by the distance ABCD (Raphael and Winter-Ebmer, 2001, p. 263). For diagram B, the straight line BC illustrates the income of legal working time and EID depicts the expected benefits of income-generating offenses. As is obvious, the marginal expected return to lawbreaking activities never exceeds the returns to labor market work. Consequently, individuals will devote all their available working time to legitimate work and the budget line is thus given by ABC.

With these two simple models, the effect of sudden unemployment of individuals can be explained. Accordingly, with ajob loss the potential wages earned in the labor market drop to zero. Consequently, the marginal return to legitimate work will always be lower than the expected return to income-generating crimes. For diagram A, this would imply that the new budget constraint is then depicted by ABCE (Raphael and Winter-Ebmer, 2001, pp. 263­264). In general, if and how the total time devoted to illegal activities is affected by sudden unemployment, depends on the preferences of the respective individual, the previous amount of time devoted to income-generating felonies, and the reservation wage. Individuals that initially choose a mix between both options of income generation, unemployment leads to more time spent on crimes. For those of which choose only to devote their working time to illegitimate income, a theoretical loss of their job has no effect on their previous behavior: after getting unemployed they still use all their available time for income-generating offenses. For individuals that only work in the legal sector and previously did not engage in any crimes, two potential outcomes are possible. First, assuming the marginal return on the first hour devoted to crime is lower than the reservation wage, even after the sudden job loss they do not undertake illegitimate activities. Second, if the reservation wage is lower than the marginal returns to income-generating felonies, they become criminal. Accordingly, they commit crimes until the marginal benefits equal their reservation wage (Raphael and Winter- Ebmer, 2001, p. 263).

Concluding, the theoretical concept defines four possible types of individuals for who a suddenjob loss has different consequences. Two of these types are not affected at all: in both cases - employment and unemployment - those individuals do not engage in criminal activities. In contrast, for the other two types sudden unemployment increases the amount of working time devoted to income-generating felonies. This implies that an increase in unemployment should be accompanied by an increase in crimes with a pecuniary background. However, according to the model, the overall effect might not be distinct: since we have four types of individuals, the quantity of each type matters. If the bigger part of the individuals in a society are those who are not affected by sudden unemployment, the aggregate effect on crime may be negligible.

Besides the model by Raphael and Winter-Ebmer (2001) there are also other theoretical innovations. Machin and Meghir (2004), for example, use a similar theoretical approach and also come to the conclusion that returns to crime and to legal work matter in the supply of criminal activities since they compete for the individual’s leisure time. Furthermore, the benefit system that determines transfers to low income or unemployed individuals plays an important role. Burdett et al. (2004) use an equ ilibrium search model to allow crime, unemployment, and inequality to be endogenous. They show how the possibility of criminal activities can raise wage inequality and how the possibility of crime can naturally generate multiple equilibria. Accordingly, there can be two levels of unemployment, inequality, and crime for two identical neighborhoods. In a similar approach, Huang et al. (2004) also find multiple steady-state equilibria with high crime and unemployment being correlated.

3.1.2 The Argument of Cantor and Land (1985)

As already mentioned in Chapter 1, the early research was dominated by criminology and sociology and conventional belief was that the relationship between unemployment and crime is positive and that it is stronger for property crime than for violent crime. However, Chiricos (1987) finds mixed results and states that the evidence is not clear-cut. Interest in the topic was rejuvenated by the publication of the theoretical (and empirical) work by Cantor and Land (1985). Accordingly, the mixed results and partial null findings of prior research result from the missing distinction between two paths through which unemployment possibly affects crime.

The first path is the so-called opportunity effect. With a recession come fewer jobs, fewer hours worked, and less time spent in job-related and leisure travel. In turn, the number of suitable targets in a society decreases since individuals spend more time in their residences and residential neighborhoods (the so-called “guardianship” increases). This results in an overall decrease in target vulnerability because of a reduction in the circulation of people and property (Cantor and Land, 1985, p. 320). Conversely, in better economic times incomes are higher and more high value goods are consumed which increases the attractiveness of targets. Furthermore, the circulation of individuals (and their valuable property) increases and, consequently, the likelihood of being victimized rises. Therefore, the authors expect the opportunity effect to be contemporaneous (Cantor and Land, 1985, p. 332). In contrast, similar to Becker (1968)’s expectations, the motivation hypothesis implies that a decrease in viable economic prospects will increase the incentive to engage in crime. However, the motivation to commit crimes may be lagged as, for example, newly unemployed persons usually are covered by unemployment benefits and may receive additional financial support by family, friends, and charitable organizations. In this spirit, it is implausible to assume an immediate motivation effect (Cantor and Land, 1985, p. 322). As support structures (and possible psychological support) gradually deteriorate over time, the unemployed become more highly motivated to engage in criminal acts (Phillip and Land, 2012, p. 683). Furthermore, the motivational effect can be expected to be of relatively short duration since many unemployed individuals will be rehired or recover economically within two years after a recession begins (Cantor and Land, 1985, p. 322). According to this belief, Cantor and Land argue that the level of unemployment in the present must be compared to the one in the previous year. The authors bring forward that “if the former is higher than the latter, and if this change in the level of employment results in a positive motivational impact on crime, then this should produce upward fluctuations in crime rules' (1985, p. 322). In contrast, dropping unemployment rates should induce downward fluctuations.

Figure 9 depicts the argument by Cantor and Land (1985, p. 321) in a graphical way. The upper half incorporates the negative impact of unemployment on criminal opportunity and the bottom half contains the positive criminal motivation effect. In addition, other exogenous factors (e.g., lifestyle, changes in age structure, etc.) are included that may also influence opportunity and motivation.

Figure 9: Path Diagram of Structural Relationships and Reduced-Form Effects of Unemployment on Crime.

Source: Cantor and Land (1985, p.321); own representation.

Abbildung in dieser Leseprobe nicht enthalten

The authors argue that only the reduced form of the relationship - indicated by the dashed arrows - is typically estimated in empirical research (and not the full structure which is represented by solid arrows). However, they appreciate the fact that:

“ [...] the data necessary to estimate the full structural model simply do not exist. Because of obvious measurement problems, there are no national-sample survey, which contain the questions necessary to measure individuals' level of motivation to commit criminal acts, nor is the data available to assess levels of criminal opportunities'" (1985, p. 321).

Thus, when only measuring the reduced form of the relationship, the result may be suppressed in absolute value towards zero, explaining the often weak findings for the unemployment-crime relationship. Furthermore, the effects are expected to work differently based on the type of crime: the motivation effect is more important for property crime while the opportunity effect is relevant for both property and violent crimes. Consequently, when comparing findings from different analyses it is reasonable that one of the estimates is positive while the other is negative, depending on the relative weight of the two counterbalancing effects in the respective data employed.

Modelling the argument of Cantor and Land (1985, 2001) in a simple time series dynamic regression would take the form of the following equation:

Abbildung in dieser Leseprobe nicht enthalten

Accordingly, the coefficients of is expected to be negative (representing the opportunity effect) and of fi2 is expected to be of positive sign (referring to the motivation effect). Although their economic model is accepted among other scholars, their econometric specification claiming to capture their argument is heavy under criticism (see, e.g., Cantor and Land, 2001; Greenberg, 2001; Levitt, 2001; Paternoster and Bushway, 2001; Kleck and Chiricos, 2002). One of these major critiques concerns the time series properties of their model and the other is the use of a differenced unemployment rate to capture the lagged effect of criminal motivation. Greenberg (2001, p. 302), among others, argues that to capture the long-run relationship a cointegration model is indispensable since differencing discards information about the long-run trend of the time series.

Concluding, there are four hypotheses that define the argument of Cantor and Land (1985):¹⁹

Hypothesis 1: The level of unemployment (criminal opportunity) is expected to have a negative effect on crime rates.
Hypothesis 2: The change in the level of unemployment (criminal motivation) is expected to positively affect criminal activities.
Hypothesis 3: In sum, the overall unemployment-crime relationship is not expected to be strong and to have a statistically significant effect on crime due to those two countervailing effects.
Hypothesis 4: Due to the nature of the motivational effect, stronger effects are expected for crimes with pecuniary background.

3.1.3 The Problem ofldentification

Before proceeding to the core part of this work, it is important to stress the problems with estimating the effect of economic activity on crime. To reliably identify an isolated effect, all determinants that affect the supply of criminal acts must be included. To the extent that variables not included in the regression are correlated with the dependent and independent variable, the estimates of the elasticities of interest will be biased. The key assumption that must hold is the so-called conditional mean independence. According to this, the causal variable of interest and the error term are uncorrelated once all control variables are controlled for. Therefore, controls that clean out the correlation between the variable of interest and the error term are essential. Fortunately, with increasing data collection and computing capacity over the last few decades and especially the increased use of panel data estimation techniques, the scope of omitted variables markedly reduced. The latter allows to control for time and area fixed effects and area-specific time trends. By including an extensive array of economic, social, demographic, and criminological factors, and using panel data strategies, more recent studies net out both observed and unobserved differences across regions.

The first variable that seems an obvious choice is education. With a higher level of educational attainment, the expected (future) legal earnings and employment opportunities are higher which decreases their willingness for criminal acts. The poorer the labor market opportunities become, the higher the incentives of engaging in crime might be. However, education may also foster crime: certain types of white collar crime (e.g., embezzlement, fraud) may increase with education if they sufficiently reward skills learned in school. Therefore, the overall effect of education on crime might be ambiguous. Secondly, cultural characteristics (e.g., religion), gender, household structure (e.g., single parent household) and urbanization are factors that might affect individuals’ decision. For the latter, social interactions seem to be important determinants: given the same expected benefits for crime, individuals may be more likely to engage in criminal acts if their peers also do so. In turn, your decision also impacts their behavior (Glaeser et al., 1996, p. 509; Freeman, 1999, p. 3549). Thirdly, crime may very well be age specific: Greenberg (2001, p. 309) states that increases in the murder rate in the mid-late-1980s were restricted to mainly individuals younger than 25. During apprenticeship financial resources are scarce and individuals are tremendously dependent on parental income transfers. If they cannot provide a sufficient amount of resources, young individuals may have a higher probability of engaging in criminal acts. Following from this, an aggregate analysis might be misleading if there is no distinction between age categories. Besides that, the individual’s past criminal record might play an important role. It affects the decision to commit a crime in several ways: (i) convicts tend to be stigmatized in the legal labor market and therefore have diminished employment opportunities and consequently less expected income; (ii) the costs of committing crime possibly reduce over time (“learning by doing”); (iii) once entered the realm of criminality, individuals tend to have a reduced moral threshold. Furthermore, the level of economic growth is important since it creates attractive opportunities for the legal employment sector but also improves the monetary value of potential loot. This implies that the effect ofbetter economic health on the individual’s incentives to commit crime is ambiguous. Corresponding to this, the effect of income inequality then depends on the relative income position of the respective individuals. Obviously, increasing inequality will have no effect on the richer part of the society while for the low-income earners the engagement in crime may become more attractive (Fajnzylber et al., 2002, p. 1328). During a recession, individuals will also consume less. If alcohol, drugs, and guns are assumed to be normal goods, consumption of these will also decrease. However, if these goods induce criminal behavior, procyclical consumption will result in procyclical variations in some crimes.²⁰ Consequently, if the consumption of drugs and alcohol is negatively correlated with labor market opportunities and has a positive impact on crime, omission of these variables from the estimation will downward bias the estimates of the relationship between labor market opportunities and criminal behavior (Raphael and Winter-Ebmer, 2001, p. 265). Furthermore, countries that are drug producers (e.g., Columbia) or located to high drug consumption centers (e.g., Mexico in the relation to the U.S.) exhibit tremendous criminal opportunities. These do not only consist of drug production/trafficking, but also involve elements of violence and corruption (Fajnzylber et al., 2002, p. 1329). In conjunction with this argument, the police force and the judicial system are other important variables: the relative strength of these two forces define the probability of being caught and the severity of punishment, potentially reducing the incentives to commit crime.²¹

In addition to the problems associated with omitted variables, there is the possibility of reverse causation.²² Accordingly, the crime level might itself constrict employment growth and contribute to regional unemployment levels such that the estimates for the effect of labor market opportunities on crime will not reflect the true relationship. As mentioned earlier, criminal activity may have a significant impact on subsequent legal employment and earnings: convicted offenders experience reduced earnings through the increased difficulties in obtaining employment, loss of professional licenses and potential denial in obtaining new licenses, and reputational effects that scare away employers, among other mechanisms (Raphael and Winter-Ebmer, 2001, p. 261; Mustard, 2010, p. 13). Furthermore, areas with high crime rates might influence the decisions of firms to relocate or expand, which in turn affects the overall employment opportunities in the respective area (Mustard, 2010, p. 14). Additionally, GDP can affect crime and vice versa; they can further affect each other both in a positive and negative way.

In sum, by including causal variables and other variables that, while themselves not causal, are sufficiently correlated with omitted causal factors (e.g., socioeconomic, and demographic variables), omitted variable bias can be drastically reduced and the isolated estimate of the coefficient of interest is more reliable and the risk of spurious correlations becomes smaller. An additional possible fix to these endogeneity concerns is the use of instrumental techniques. Nevertheless, this listing of potential controls and other factors affecting crime is not all-embracing but embodies the fact that a large array of different variables are necessary in statistical analyses.

3.2 Labor Market Opportunities and Crime

The first, and by far the most used measure oflabor market prospects of potential criminals is the unemployment rate. Figures 10 and 11 plot the unemployment and crime rates for the U.S. from I960 to 2016:

Figure 10: Unemployment rate and Property crime rate per 100,000 population in the U.S., 1960-2016. Sources: FBI Uniform Crime Report (UCR), Labor Force Statistics from the Current Population Survey (CPS).

Abbildung in dieser Leseprobe nicht enthalten

Figure 11: Unemployment rate and Violent crime rate per 100,000 population in the U.S., 1960-2016. Source: FBI Uniform Crime Report (UCR), Labor Force Statistics from the Current Population Survey (CPS).

Abbildung in dieser Leseprobe nicht enthalten

Looking at these rates, it seems intuitive to choose unemployment as an indicator for labor market opportunities. Both crime and unemployment fluctuate over time and thus may be intercorrelated. During the 1960s unemployment and property crime behaved countercyclical; while the unemployment rate decreased, there was an observed increase in the property crime rate. However, from the 1970s until the mid-2000s both rates show somewhat similar behavior: when unemployment increased (decreased), property crime also increased (decreased). Especially the peaks in 1971, 1975, 1982, and 1992 were accompanied by peaks in property crime. Between 2007 and 2010 there was an increase in the unemployment rate, while property crime per 100,000 population decreased. After 2010 both rates declined, again pointing in the same direction. For violent crime there is a similar picture, although not as clear.

Obviously, this simple graphical illustration is not adequate to investigate the relationship between the business cycle and illegitimate activities. There’s much more to it, than just visually comparing two variables over time. Nevertheless, there is no doubt that unemployment also has a range of negative effects on individuals, society, and economy. Unemployed people are confronted with psychological problems, ill health, and social distance. Furthermore, with rising unemployment there are also nascent problems for society and economy: higher expenses for social security, lower tax income, reduction of domestic demand, and political instability. Unsurprisingly, scientific literature and especially political debates consider unemployment an important factor in the supply of criminal activities.

[...]

---

¹ The President’s Commission on Law Enforcement (1967); Miller et al. (1996); Anderson (1999); Chalfin (2013); Anderson (2011). These numbers were adjusted to 2018 dollars using the Consumer Price Index (CPI).

² http://www.nber.org/cycles/general_statement.html (accessed June 22, 2018).

³ Before, the research on crime was dominated by sociologists and criminologists.

⁴ Less than half of the 63 studies reviewed indicate a significant positive relation. However, he also finds that the level of aggregation and the data employed in the analyses heavily influence the results.

⁵ Complete crime statistics for 2017 will be released in September 2018. At this moment there only exists a preliminary report that covers January to June of 2017.

⁶ According to the FBI, arson is only included in trend, clearance, and arrest tables, but not in any estimated volume data.

⁷ Households stay in the representative sample for 3 years. New households rotate into the sample on an ongoing basis. Furthermore, the NCVS estimates the proportion of each crime type reported to law enforcement, and it summarizes the reasons that victims give for reporting or not reporting.

⁸ It does not measure homicide or commercial crimes such as burglaries of stores. Furthermore, rape is measured for both sexes, while the UCR only measures the crime against women.

⁹ Offender demographics are based on the victim’s perceptions.

¹⁰ Source: Bureau of Justice Statistics - NCVS Victimization Analysis Tool (NVAT).

¹¹ The UCR includes homicide and commercial crimes while the NCVS does not. The NCVS includes sexual assault, while the UCR excludes it. Beyond that, the NCVS estimates are based on a representative sample of the U.S., while the UCR estimates are based on counts of crimes reported by approximately 18,000 city, university and college, county, state, tribal, and federal law enforcement agencies. Finally, the NCVS excludes crime against children age 11 or younger while these are included in the UCR. Though, restricting the NCVS data to serious violence reported to police only, yields similar results to the UCR reports.

¹² Austria, Belgium, Bulgaria, Croatia, Cyprus, Czech Republic, Denmark, Estonia, Finland, France, Germany, Greece, Hungary, Ireland, Italy, Latvia, Lithuania, Luxembourg, Malta, Netherlands, Poland, Portugal, Romania, Slovakia, Slovenia, Spain, Sweden, United Kingdom (consisting of England & Wales, Scotland, and Northern Ireland).

¹³ Source: Eurostat.

¹⁴ Crime statistics for 2016 will be released later in 2018.

¹⁵ Source: Polizeliche Kriminalstatistik (2016).

¹⁶ Those variables are excluded because they are not relevant for this investigation.

¹⁷ In contrast, the original labor-leisure model optimizes the allocation of available time (24 hours) between working in the labor market and leisure time.

¹⁸ See also Grogger (1998) and Fagan and Freeman (1999), among others.

¹⁹ See also Cantor and Land (2001) and Phillips and Land (2012).

²⁰ However, the argument can be made that the result of drugs on crime is mixed. For example, Fagan (1990, p. 225) finds that drugs such as marijuana are more likely to reduce aggressive behavior and ultimately criminal activity.

²¹ In criminology this is labeled as the “crime-deterrence-effect”. However, this effect might suffer from measurement error (higher numbers of police may mean increasing reporting) and reverse causation (increasing crime rates may induce citizens to hire more police).

²² Also, often referred to as simultaneity bias.

• Universidad: University of Hohenheim
• Curso: Markets and Consumption
• Calificación: 1,3
• Autor: Felix Müller (Autor)
• Año de publicación: 2018
• Páginas: 90
• No. de catálogo: V1034826
• ISBN (Ebook): 9783346442352
• ISBN (Libro): 9783346442369
• Idioma: Inglés
• Etiqueta: relationship, business, cycle, crime, literature, review