Income inequality and crime across nations reexamined

Open Access
Pare, Paul-Philippe Pare
Graduate Program:
Crime, Law and Justice
Doctor of Philosophy
Document Type:
Date of Defense:
June 26, 2006
Committee Members:
  • Richard B Felson, Committee Chair
  • D Wayne Osgood, Committee Member
  • George Farkas, Committee Member
  • C Shannon Stokes, Committee Member
  • Criminology
  • homicide
  • ICVS
  • income inequality
  • poverty
Many scholars believe that income inequality is a major causal factor affecting the homicide rate across nations. Indeed, many empirical studies have found a positive relationship between income inequality, as measured by the Gini coefficient, and the homicide rate. These prior studies, however, have some limitations that cast doubt on the relationship between income inequality and homicide. The current dissertation, divided in three research papers, addresses these limitations. The first paper presents a cross-national analysis based on a sample of 63 nations that isolates the effects of income inequality and poverty on the homicide rate. The evidence indicates that income inequality is unrelated to the homicide rate, once poverty is adequately measured. Poverty is positively associated with the homicide rate. The analysis also suggests that the negative relationship between social welfare and the homicide rate observed in previous studies is attributable to poverty reduction, not income inequality reduction. The second paper presents a re-analysis of the International Crime Victim Survey data. It is based on a sample of 122,491 respondents across 28 nations and tests whether relationships between income inequality and physical assault, robbery, burglary, and theft are observed when some limitations of prior studies are addressed. These limitations include (1) the measurement of poverty, (2) the measurement of crime, (3) model specification in aggregate vs. multilevel analyses, (4) measurement error with capital-city samples, (5) and the analysis of sub-samples of nations instead of the full sample of available nations. The evidence indicates that nations’ levels of income inequality are not associated with the frequency of any of these crimes. Nations with greater levels of poverty, however, have more frequent robberies, burglaries, and thefts, but not physical assaults. The third paper utilizes Heckman selection regressions to detect and correct sample selection biases in cross-national analyses of the WHO homicide data and the ICVS data. The results indicate that while the level of development and the level of socio-political freedom are strongly associated with the probability that nations will be included in the substantive samples, these selection effects do not produce substantive biases.