Multivariate analyses are likely to provide a better estimate of associations between poverty and crime, due to their ability to distill the proportion of reduced error due to poverty, after accounting for the influence of related factors. A more detailed evaluation of the various studies reveals further notable patterns consistent with the differential etiology of violence hypothesis. Steffensmeier and Haynie (2000) use a sophisticated analysis with many control variables in their models. They found strong relationships between male poverty and all three measures of violence: homicide, robbery and aggravated assault (p < .001). Poverty was also strongly associated with burglary, but it was not significantly associated with larceny. For females, poverty was significantly associated with homicide, robbery and aggravated assault (p < .001) but not significantly associated with burglary or larceny. The results were the same when the authors used an index of disadvantage. Controlling for ethnic heterogeneity, race variables, home ownership, occupied units, unemployment divorce, and population mobility, Hipp (2007) reported significant positive associations between poverty and aggravated assault, robbery and murder, but the relationships between poverty and burglary or motor vehicle theft were not statistically significant. The findings for average household income had less consistent findings, but these results may be due to a model specification problem. (The poverty variable was included in the same model.) Grant and Martinez (1997) used panel data to look at violent and property crime at the state level. They employed a measure of the “underclass” that consists of the poverty rate, percent Black, and percent female-headed households that they believe taps into economic deprivation and social isolation. This measure is positively, strongly, and significantly associated with violent crime but negatively associated with property crime.
Krivo and Peterson (1996) found that the operationalization of poverty or resource deprivation may affect one’s findings. Their findings are more nuanced because the authors employed separate measures for Columbus, OH, census tracts with low (< 20%), high (20-39%), and extreme (40% or more) levels of poverty or disadvantage. Thus, their design allows them to study not just the effect of poverty on crime, but rather the effects of different levels of poverty. While the authors do not report standardized coefficients (which prevents direct comparison of the effect sizes for violent vs. property crime), they emphasize that their measures of disadvantage “... have strikingly different effects for violent than for property crime” (p. 630). Their correlation matrix is instructive. While the high poverty variable is only moderately correlated with the property (r = 0.26) and violent (r = 0.27) crime rates, the correlation between the extreme poverty variable and violent crime (r = 0.62) was much greater than the correlation between extreme poverty and property crime rates (r = 0.33). This pattern of notably higher correlations with violent than property crime also held for the measure of extreme “disadvantage” (r = 0.717 vs. 0.377). The authors conclude that their multivariate analyses support the pattern established by the initial correlations; they state, “the role of extreme disadvantage is limited to its impact in producing especially heightened levels of violent (not property) crime” (p. 640).
It is too early to tell whether the differential impact of resource deprivation on violence will generalize across nations. All three of the studies that analyze data from outside the United States tend to find that indicators of poverty or resource deprivation are positively and significantly related to both violent and nonviolent offending rates (Fitzgerald, Wisener, & Savoie, 2004; Savoie, Bedard, & Collins, 2006; Weatherburn & Lind, 1997). Though the international observations raise the possibility that poverty may only be more strongly related to violent offending in America, it is too soon to draw any firm conclusions. This supposition would fit with Merton’s (1938) theoretical argument which implies that poverty-induced strain is likely to be most acute and most likely to provoke criminal behavior in a materialistic culture like that found in the United States.