Sources of crime in the state of Veracruz: The role of female labor force participation and wage inequality

Abstract

In recent years, crime has become a serious concern in Mexico as its increase has detrimentally affected government institutions and economic growth. There is considerable speculation among policy analysts about the causes of the increase in crime. Whereas some analysts attribute the increase to a rise in income inequality, others believe internal migration and a loss of morals are the roots of criminal behavior. This research shows that at least for the Mexican state of Veracruz, wage inequality and labor force participation have an important impact on crime. When gender is considered, however, the impact is more complicated than it seems. An increase in women's labor force participation decreases the overall number of alleged violent offenders. However, the number of alleged rapists and grievous bodily harm offenders increases as women's wage distribution improves. The results shed light on the gender dimensions of the economics of crime.

Keywords:

• Economics of crime
• gender wage inequality
• women's labor-force participation
• JEL Codes: J16, J31, O54

Acknowledgments

The authors would like to thank Jeff Brannon, Sean Flynn, Joan Friedman, and Manuel Reyes for help in editing the text. The authors would also like to thank in a very special way the staff at the State Population Council of Veracruz, in particular Angel F. Argüello Ortiz.

Notes

See Aleyda Aguirre 2005; Gabriela Romero, Laura Gómez, Josefina Quintero, Susana González, Raúl Llanos, and Mirna Servín 2004 for one example of the way this march captivated Mexico's attention, demonstrating the frustration many Mexicans felt with the increase in crime.

Most states in Mexico have fewer than 60 municipalities which makes regression analysis at the municipal level difficult. Veracruz not only has a large number of municipalities but the fact that they can be grouped into distinct regions allows us to control for geographic and cultural differences across the state.

Isaac Ehrlich (1996) includes a survey of these studies and research issues on the market for crime.

Steven Levitt (2004) found that the policing and incarceration strategies, and not the increase in legitimate labor market opportunities, caused the impressive drop in crime in the 1990s in the US.

An additional benefit of using labor force participation rates is that they can be used to keep track of both the “discouraged worker effect” and the “added worker effect” (Kye Woo Lee and Kisuk Cho 2005: 424). Discouraged workers are potential criminals; added workers are more than likely law-abiding women who are forced to enter the labor force during a time of crisis (Cerruti 2000; Parrado and Zenteño 2001). Accounting for both these categories would naturally improve our chances of correctly ascertaining the relationship, if any, between crime and the opportunities afforded by legal employment. Consequently, the following analysis uses labor force participation rather than unemployment rates.

We do not use convicted criminals because Mexico's judicial system is notoriously inefficient in apprehending criminals. Furthermore, the use of alleged offenders as opposed to sentenced offenders helps estimate the extent of criminal activity more accurately (Edward L. Glaeser, Bruce Sacerdote, and Jose A. Scheinkman 1996).

Kidnapping is an abduction of a person for the purpose of charging a ransom in money. It is therefore a specific form of abduction (René A. Jiménez Ornelas and Olga Islas de González Mariscal 2002).

The authors wish to thank one anonymous referee for the suggestion to concentrate only on violent crime, improving the focus of the paper.

Traditionally, crime rates are calculated per 100,000 people; many municipalities in Veracruz have fewer than 6,000 people. This is why we use the number of offenders per 1,000 people.

This measure is similar to the poverty index developed by the Social Security Administration in the US as a result of President Lyndon B. Johnson's “War on Poverty” (Social Security Administration n.d.).

The nine variables that make up the marginality index are:

	Illiterate persons 15 years of age and older
	Persons 15 years of age and older who have not completed primary education
	People living in households with cramped quarters
	People living in households without electricity
	People living in households without indoor plumbing
	People living in household with no running water
	People living in households with a dirt floor
	People living in areas with fewer than 5,000 residents
	People earning up to two times the minimum wage.

This last item is removed from the index when used in the regressions because of its high correlation with the labor market variables employed.

The five levels of marginality are: very low [−2.449, −1.281], low (−1.281, −0.697], medium (−0.697, −0.113], high (−0.113, 1.054], and very high (1.054, 3.390].

Unlike other statistical agencies, INEGI (2001e, 2001g) can collect data on religious beliefs.

Recently, there has been increasing tension between different religious groups in rural Mexico. Thus, there is also the possibility that religious affiliation actually increases the number of alleged violent offenders. Jean-Pierre Bastian (1997) provides more information on the effect of this religious transformation on Latin America.

The data on Labor Force and Working Age Population comes from INEGI (2001d, 2001f).

In the year 2000, the distribution of total income (labor and non-labor income) was a bit less skewed. The top household decile had 46 percent, and the bottom three household deciles had 3 percent of total income.

Minimum wage in Mexico varies by region and profession, ranging from 42 to 45 Mexican pesos a day or between US$5.87 and US$6.29 daily. These dollar amounts have been adjusted for purchasing power parity for the year 2000.

Examination of the data revealed that the number of alleged offenders as a function of the marginality index is U-shaped. Municipalities with very low and very high degrees of marginality have a higher number of alleged offenders than those with medium levels. This is why the dependent variable is a quadratic function of this index.

The Poisson regression is based on the probability distribution with the same name. This regression model is for count data (data that take on positive integer values) such as the number of specific alleged offenders.

The regressions consequently use a truncated marginality index consisting of eight instead of nine variables; this reduces the problem considerably.

Test results for spatial autocorrelation as well as other relevant diagnostics are available upon request.

The econometric specification of the spatial lag dependent variable regression is:

where W is a spatial weights matrix, WΔln(Offenders) represents the vector of spatial lags for the dependent variable, and ρ is a vector of spatial autoregressive coefficients (Anselin 1992).

The minimum for this parabola is achieved at the absolute value of b₁ over twice the value of b₂, |[b₁/2b₂]| (Wooldridge 2000). Thus, for alleged violent offenders this inflexion point is: |[0.064/(2 × (−0.13))]|=|[0.064/(−0.26)]|≈ 0.246.

Elster (1998) describes this evolution of feelings. For more on how rule-breakers become outcasts see Akerlof (1976).