Why Do Poor People Commit More Crime?

It’s well known that people with lower incomes commit more crime. Call this the cross-sectional result. But why? One set of explanations suggests that it’s precisely the lack of financial resources that causes crime. Crudely put, maybe poorer people commit crime to get money. Or, poorer people face greater strains–anger, frustration, resentment–which leads them to lash out or poorer people live in communities that are less integrated and well-policed or poorer people have access to worse medical care or education and so forth and that leads to more crime. These theories all imply that giving people money will reduce their crime rate.

A different set of theories suggests that the negative correlation between income and crime (more income, less crime) is not causal but is caused by a third variable correlated with both income and crime. For example, higher IQ or greater conscientiousness could increase income while also reducing crime. These theories imply that giving people money will not reduce their crime rate.

The two theories can be distinguished by an experiment that randomly allocates money. In a remarkable paper, Cesarini, Lindqvist, Ostling and Schroder report on the results of just such an experiment in Sweden.

Cesarini et al. look at Swedes who win the lottery and they compare their subsequent crime rates to similar non-winners. The basic result is that, if anything, there is a slight increase in crime from winning the lottery but more importantly the authors can statistically reject that the bulk of the cross-sectional result is causal. In other words, since randomly increasing a person’s income does not reduce their crime rate, the first set of theories are falsified.

A couple of notes. First, you might object that lottery players are not a random sample. A substantial part of Cesarini et al.’s lottery data, however, comes from prize linked savings accounts, savings accounts that pay big prizes in return for lower interest payments. Prize linked savings accounts are common in Sweden and about 50% of Swedes have a PLS account. Thus, lottery players in Sweden look quite representative of the population. Second, Cesarini et al. have data on some 280 thousand lottery winners and they have the universe of criminal convictions; that is any conviction of an individual aged 15 or higher from 1975-2017. Wow! Third, a few people might object that the correlation we observe is between convictions and income and perhaps convictions don’t reflect actual crime. I don’t think that is plausible for a variety of reasons but the authors also find no statistically significant evidence that wealth reduces the probability one is suspect in a crime investigation (god bless the Swedes for extreme data collection). Fourth, the analysis was preregistered and corrections are made for multiple hypothesis testing. I do worry somewhat that the lottery winnings, most of which are on the order of 20k or less are not large enough and I wish the authors had said more about their size relative to cross sectional differences. Overall, however, this looks to be a very credible paper.

In their most important result, shown below, Cesarini et al. convert lottery wins to equivalent permanent income shocks (using a 2% interest rate over 20 years) to causally estimate the effect of permanent income shocks on crime (solid squares below) and they compare with the cross-sectional results for lottery players in their sample (circle) or similar people in Sweden (triangle). The cross-sectional results are all negative and different from zero. The causal lottery results are mostly positive, but none reject zero. In other words, randomly increasing people’s income does not reduce their crime rate. Thus, the negative correlation between income and crime must be due to a third variable. As the authors summarize rather modestly:

Although our results should not be casually extrapolated to other countries or segments of the population, Sweden is not distinguished by particularly low crime rates relative to comparable countries, and the crime rate in our sample of lottery players is only slightly lower than in the Swedish population at large. Additionally, there is a strong, negative cross-sectional relationship between crime and income, both in our sample of Swedish lottery players and in our representative sample. Our results therefore challenge the view that the relationship between crime and economic status reflects a causal effect of financial resources on adult offending.