Two bromides in discussions of crime are that poverty and inequality are important contributors. Thankfully, we can put aside speculative chatter and explore the issue in quantitative and scientific terms. Most of what I say on this issue will draw heavily from the fantastic dissection of it by The Alternative Hypothesis—Sean Last, in particular—so I extend one big hat tip in that direction (and I’d highly recommend bookmarking that site in general, as it is truly fantastic—although certainly not politically correct!).
First off, we can establish that poorer folks, relative to richer ones, are indeed more likely to be criminals. However, let’s look at the association specifically between poverty and inequality on the one hand, and crime on the other. The results of a large meta-analysis of 45 studies from the relevant peer-reviewed literature by Vieraitis can be seen in the charts below:
Results of an even larger meta-analysis on the associations between crime on the one hand, and median income, poverty, income inequality, and unemployment rates on the other, by Ellis, Beaver, and Wright can be seen below:
Chiricos’ meta-analysis of 288 studies looking at the correlations between crime and unemployment is depicted in this chart:
Now, let’s take a look at Pratt and Cullen’s meta-analysis of the important issue of effect size. In 153 studies, they found a mean effect size of poverty of .253, with 59% of all results being statistically significant. In 167 studies, the mean effect size of inequality on crime was .207, with 55% of the results being statistically significant. And in 204 studies, the mean effect size of unemployment on crime was .135, with 44% of those results being statistically significant. Nivette also performed a meta-analysis of 37 studies, finding the mean effect size of national wealth on crime to be -.055 (note the minus sign), and not statistically significant; the mean effect size of income inequality on crime to be between .224 to .416 (contingent on the manner in which income inequality was measured), with both values reaching statistical significance; and the mean effect size of unemployment on crime being .043, and not statistically significant (but note that this particular mean effect size was gathered from only four studies).
At this point, we would be remiss not to at least mention Hsieh and Pugh’s meta-analysis of 34 studies. Those researchers found positive correlations between poverty and violent crime (save for 2 of the 78 relationships examined), with an average correlation of .44. Oddly, they also found the same percentage of studies showing a correlation between inequality and crime to be the same—namely, .44. It is difficult to know what to make of their meta-analysis, but the larger, more recent ones that we surveyed above present a very different picture. And Hsieh and Pugh themselves note the inconsistent findings in literature reviews on the relation between economic variables and crime performed in the late 70s and early 80s. In any case, in view of the other meta-analyses, it stands to reason that Hsieh and Pugh’s meta-analysis is for some unknown reason(s) an aberration.
Before we proceed further, note that, conventionally speaking, .2 is considered the threshold for weak effects, .5 for moderate ones, and .8 for strong effects. Given this, the striking thing to observe among the results surveyed above is just how many of the mean effect sizes are weak or insignificant.
Moving on, let’s look at these relevant variables (poverty, income inequality, and unemployment) plotted across time, and in relation to crime.
(Poverty rates drawn from the United States Census Bureau; crime rates drawn from the Department of Justice’s Uniform Crime Reporting Program.)
Surprisingly, poverty and crime are actually negatively correlated longitudinally. In other words—and as counterintuitive as it will sound to many—as poverty rose, crime declined, and as poverty went down, the crime rate went up. So, for example, crime went down during the Great Depression—which is not what one would have predicted according to the thesis that poverty causes crime.
An analysis of 8 studies by Ellis, Beaver, and Wright also found inconsistent longitudinal relationships between the economy and unemployment on the one hand, and crime on the other:
Of 35 studies finding a longitudinal correlation between income inequality and crime, Rufrancos and colleagues found only 60% to be both positive and statistically significant.
The longitudinal data plotted in the following graphs also appears to be at loggerheads with the hypothesis that income inequality is causally implicated in crime.
(Data in the murder graph is taken from the FBI, the Office for National Statistics (UK), and the UN; data in the wages graph is taken from the US Bureau of Labor Statistics.)
As we can see, the crime rate in the US has ostensibly been moving downwards, while income inequality in the US has been rising over the same period of time. So the trend line for crime runs in the opposite direction predicted by the inequality-causes-crime thesis (given that income inequality has been widening over the same span of time).
Overall, the foregoing isn’t the picture one might expect to see if there were actually a strong and robust causal connection between these economic variables and crime. Indeed, even in so far as there are any correlations between crime and the various economic variables we have been considering, all of the usual caveats are in play, including the direction of causality. That is to say, even if, for argument’s sake, we were to grant a strong, robust correlation between crime and all or any of the economic variables we’ve considered, it would still be an open question as to whether any given economic variable in question was causally related to crime, or whether crime was casually related to the economic variable in question. In other words, in asking about the direction of causality here, we are effectively asking, ‘Does poverty/inequality/unemployment (at least partly) cause crime, or does crime (at least partly) cause poverty/inequality/unemployment?’ In sum, trying to ascertain the direction of causality, even on the assumption that there is some sort of causal connection, is quite tricky.
Now, perhaps we should remind ourselves that, given the weight of the evidence, the effect sizes of various economic variables on cross-sectional studies of criminality do not appear to be very large in the first place. And, as per usual, there is absolutely no guarantee that those effect sizes would not narrow or even disappear completely after controlling for various other variables—particularly the sorts that criminologists do not typically control for, such as IQ, or even molecular-genetic differences (i.e., allelic variants). (For instance, IQ is indeed associated with criminality—see here and here.)
However, we can attempt to zoom in on the question in a more fine-grained manner by considering one ingenious study that was conducted recently by Amir Sariaslan and colleagues, and which demonstrates how careful study design and controlling for the right variables can move us in the direction of ascertaining causality:
“In 2014 came the final nail in the coffin to the “poverty causes crime” thesis. A Swedish study conducted by Amir Sariaslan was published which—for the first time—tested directly whether growing up in poverty directly contributes to crime, or whether there are other factors about the kinds of families which tend to end up poor which also cause them to breed crime. What made Sariaslan’s study uniquely insightful was the decision to take families which rose out of poverty, and compare the lives of children born and raised within those families before their rise from poverty with the lives of children born and raised within those same families after their rise from poverty. The conclusion his research came to? “There were no associations between childhood family income and subsequent violent criminality and substance misuse once we had adjusted for unobserved familial risk factors.” Sariaslan’s study, in other words, had proven that growing up in poverty is not what creates one’s adult likelihood of committing violent crimes. Children who grow up in previously–poor families have exactly the same likelihood of committing crimes as children who actually grow up poor. The only conclusion we can soundly come to is that something else about poor families other than poverty itself must explain why their children go on to commit crimes.” [Source]
Such a finding seems to very tellingly count against the hypothesis that poverty is causally implicated in criminality. Assuming such a finding is robust, it is, needless to say, very revealing.
Relatedly, it might be alleged that various features of socioeconomically deprived neighborhoods are causally implicated in crime. Another study by Sariaslan and colleagues addressed this question, and their results interestingly did not bear out this hypothesis. Specifically, the researchers found that any association between neighborhood deprivation-related variables on the one hand, and crime (but also substance misuse) on the other, disappeared once both observed and unobserved familial confounders were controlled for. It is worth quoting their main findings at length:
“General neighbourhood effects are presented in Table 3. The crude models suggested that 12.2% and 4.2% of the variance in violent criminality and substance misuse, respectively, were attributable to the neighbourhood context. The adjusted models markedly reduced these effects to 1.8% and 1.9%, respectively, indicating that substantial proportions of the attributed variances came from characteristics of the individuals living in the neighbourhood contexts rather than from context-specific factors. In stark contrast, the family context proved to be highly influential, accounting for 30.1% and 22.8%, respectively, of the variances in violent criminality and substance abuse in the adjusted models.
The measure of neighbourhood deprivation was associated with the outcomes of both violent criminality and substance misuse in the total population sample (Table 4). An increase of 1 SD in the neighbourhood-deprivation score was associated with a 57% increase in the odds of being convicted of a violent offence. When we adjusted for observed confounders, the association was considerably attenuated (OR: 1.09; 95% CI: 1.06-1.12). In the final step, we adjusted for unobserved confounders within nuclear families and the association disappeared (OR 0.96; 95% CI 0.83–1.11). To obtain converging evidence about the validity of our results, we additionally studied the association within extended families among biological full cousins (N=169 254), and found that the results remained intact (OR 1.03; 95% CI 0.93–1.13).
An increase of 1 SD in the neighbourhood-deprivation score was associated with a 31% increase in the odds of engaging in substance misuse. The association disappeared, however, in the adjusted model (OR 0.98; 95% CI 0.96–1.01), indicating that the effect of the contextual exposure was confounded by family-level SES.”
In America, the discourse on gun control tends to be highly polarized and politicized. One idea that gets tossed around is the notion that gun homicides are more likely to occur to the extent that there are more guns around. And relatively few who opine on the matter, journalists and pundits included, seem to discuss it in quantitative and scientific terms. (Of course, this is not too surprising, given how seemingly rare quantitatively- and scientifically-informed discussions generally are in our society.)
First, let’s consider the oft-repeated assertion that local gun prevalence is somehow causally related to gun homicides. As it turns out, this is often a charge made by gun control proponents—roughly, the idea is that the more guns there are in a locale (i.e., at the municipal and state levels), the higher the gun homicide rate will be. Some will attempt to make a case for this hypothesis by marshalling data in its support, but without statistically controlling for gun suicides. Obviously, this is wrongheaded. However, looking at the data specifically on gun ownership rates and gun homicide rates, there does not even appear to be a correlation between the two variables.
(The data in this graph are taken from the Centers for Disease Control, both for state gun ownership rate and homicide rate—averaged across 3 years (2001, 2002, 2004); source.)
So, since there’s no correlation between the (state-level) gun ownership rate and the gun homicide rate, there does not appear to be a case for the causal connection between the two variables (as again, logically, there would minimally need to be a correlation for there to be a causal connection, realistically speaking).
Now, let’s zoom in a bit more and ask the following question. Is there any association between the state-level homicide rate and the state-level gun homicide rate? Well, the data can answer this question, too—and it appears that, Yes, there is a correlation, as the graph below shows.
(Data in this graph is taken from the Centers for Disease Control; source)
Since the gun homicide rate and the non-gun homicide rate are correlated, it suggests that there’s an underlying variable—or set of variables—causing both. And this is where things start to get interesting.
Overall, the usual suspects—poverty, income inequality, unemployment—do not appear to be what we should be looking at if we wish to uncover the causes of crime (or, to the extent that any of them are causally implicated, they are relatively marginal forces). And gun prevalence does not appear to be even correlated with homicide. What does cause crime, however, is, in my view, a more complicated and interesting question. Incidentally, I think it’s a question that can only be adequately answered with an explanatory framework whose core is decidedly evolutionary-psychological and genetic in its foundation. But that’s an extended excursion for another time.