Money and crime: the relationship that wasn't

Poverty makes you... unable to afford expensive things

There must be 10,000s of studies and news articles reporting findings that wealth or income is negatively related to crime. In other words, poverty causes crime. The popular theory being that poor people simply can't afford food, so they steal to make ends meet (insert sob story here). A typical finding is something like this, based on Scottish data:

Sure enough, there is a clear gradient. More support for a causal role of income in reducing crime is that crime rates have been generally falling over the centuries as we got more wealthy and incomes rose. The relationship is certainly not perfect, e.g., there are sometimes major crime waves despite improvements in income, but the overall trend works out. Recall, wealth over time looks like this:

And the homicide rates something like this:

If we look at the values for some specific countries, we can see that (using 1825 as starting point because most data begins there):

• Spain: GDPpc rose from 1,698 to 31,497 USD, or 18.5 times. Homicide rate fell from 8.8/100k to 0.61, or 14.4 times.

• Italy: GDPpc rose from 2,703 to 34,364, or 12.7 times. Homicide rate fell from 8.0/100k to 0.56, or 14.3 times.

• Netherlands: GDPpc rose from 3,022 to 47,474, or 15.7 times. Homicide rate fell from 1.35/100k to 0.65, or 2.1 times.

Clearly, southern Europe saw a greater pacification process, but the trend is there in both cases.

Based on this kind of data, we might then be inclined to think that maybe higher income really does make people less inclined to commit crimes. But how can we be sure? The main issue with all social science is that people aren't assigned their circumstances at random, so the people who have more money, say, are in general quite different people than those with less. That is, unless you look at lotteries. I wrote a summary lottery studies back in 2020, so check that out. This post is about a new study which looked at crime, something not previously covered:

• Cesarini, D., Lindqvist, E., Östling, R., & Schroeder, C. (2023). Does Wealth Inhibit Criminal Behavior? Evidence from Swedish Lottery Winners and Their Children (No. w31962). National Bureau of Economic Research. Ungated PDF.

There is a well-established negative gradient between economic status and crime, but its underlying causal mechanisms are not well understood. We use data on four Swedish lotteries matched to data on criminal convictions to gauge the causal effect of financial windfalls on player’s own crime and their children’s delinquency. We estimate a positive but statistically insignificant effect of lottery wealth on players’ own conviction risk. Our estimates allow us to rule out effects one fifth as large as the cross-sectional gradient between income and crime. We also estimate a less precise null effect of parental lottery wealth on child delinquency.

Ouch! It seems that when money is literally randomly given to some people, and we follow them up to see if they commit fewer crimes, we find that don't, and not even for their children is there an effect. It's a direct disproof of this popular theory. The authors helpfully provide a summary of popular thinking:

Social scientists have proposed a range of explanations for the observed relationship between crime and economic status. A prominent class of theories in sociology emphasize that lack of economic resources may cause “strain” — anger, frustration, and resentment — and induce individuals to resort to crime to obtain what they cannot obtain through legal means (Merton 1938, Cloward & Ohlin 1960, Agnew 1992). A related literature argues that low economic status may lead to selection into geographic areas with less social control, increasing the propensity for criminal behavior (Shaw & McKay 1942, Sampson & Groves 1989). Common to these theories is the notion that lack of financial resources causes crime.

Economic theory predicts poor labor market conditions increase crime for economic gain (Ehrlich 1973, Sjoquist 1973, Block & Heineke 1975), but the effect of changes in unearned income and wealth is ambiguous. For example, crime can be increasing in wealth if individuals exhibit decreasing absolute risk aversion (Allingham & Sandmo 1972, Block & Heineke 1975), but may be decreasing if leisure from criminal activity is a normal good (Grogger 1998) or if the utility loss of imprisonment increases in wealth (Becker 1968). Economists have also highlighted that certain “consumption offenses” (Stigler 1970), such as illicit drug use, may be increasing in income.

What data did they use?

In this paper, we use data from four samples of Swedish lottery players matched with data on the universe of criminal convictions to investigate how positive wealth shocks affect criminal behavior. Matching adult players to their children, we also estimate the effect of parental wealth on child delinquency. A key advantage of our data is that we observe the factors conditional on which lottery wins are randomly assigned.

...

To construct the estimation sample for adult players, we started with all winners and control individuals who were at least 18 and no older than 74 years of age in the year of the lottery draw. We then excluded observations who (i) had not been assigned to a cell, or had been assigned to a cell without any variation in the magnitude of the size of the prize won; (ii) lacked information about basic socio-economic characteristics measured in government registers or (iii) shared prizes in the Triss lottery. Imposing these restrictions leaves an estimation sample of 354,034 observations (280,783 individuals).

As with the adult sample, we exclude children not matched to a cell, or matched to a cell without prize variation and children whose parents shared a prize in the Triss lottery. We also restrict the sample to children whose parents were both alive the year before the lottery draw and for whom none of our basic socio-economic characteristics are missing in the registers. Imposing these restrictions, our intergenerational sample consists of 120,159 observations corresponding to 100,953 unique children of 60,074 lottery-playing parents (29,189 mothers and 30,885 fathers) who won a total of 69,264 prizes.

In other words, they used the great Swedish register data and the long-running government lotteries. Finally, something useful has come out of a government monopoly on gambling.

The main results look like this:

The black squares are from the lottery winners comparison, and the grey triangles are results from cross-sectional analysis of the general population. The X axis has the type of crime, and the Y value is the slope (gradient), that is, the strength of the relationship. As expected, in the general population looking between people, we see negative relationships (the main effect is about -4 on their somewhat difficult to interpret scale). But looking at the lottery winners, we see that the relationships are unexpectedly slightly positive, though they overlap with 0 in each case. In other words, there's no evidence at all that suddenly getting a lot of money makes you avoid or commit more crime. It doesn't matter about the type of crime either, violent, drug, traffic, or something else.

Likewise, for their sample of children of lottery winners, there are also no effects to be seen:

All p values are beyond .05 even before using multiple testing corrections.

The authors want to get their paper published, so they undersell it a bit:

Conclusions

We estimate a positive but statistically insignificant effect of lottery wealth on adults’ conviction risk. Though small protective effects of wealth cannot be ruled out, we can reject causal effects one fifth as large as the cross-sectional crime-income gradient in a representative sample. The results from our intergenerational analyses are less precise but allow us to rule out large effects of parental wealth in either direction.

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.

But in fact, there's quite a bit of supporting research in this area, much of which is not cited in the study. First, as mentioned above, there's a whole series of other lottery winner studies looking at benefits to oneself and one's children. The benefits are relatively few. People do become richer (duh, they don't spend all the money immediately) and somewhat happier (money buys some happiness), but it doesn't help one's children much. One might wonder why people even bother earning so much money in the first place.

Behavioral genetic support

Beyond lottery studies, behavioral genetic studies can offer us some insight. Cited in the new paper are the studies by Amir Sariaslan looking at siblings and crime. Full siblings have the same parents, but they don't have exactly the same childhood family income. The reason for this is that parental fortunes may change over time, so that one sibling grows up in relative affluence, and the other in more of a middle class position. The mount of overlap depends on the age gaps, and some children are 10+ years apart, so they did not share much of their childhoods together, and may not have lived the same place either. Using this kind of between-sibling within-family variation, one can see if family childhood income still relates to children's later outcomes. The work by Sariaslan and coauthors basically find that it doesn't work. Their most recent study from 2021 based on Finnish register data:

• Sariaslan, A., Mikkonen, J., Aaltonen, M., Hiilamo, H., Martikainen, P., & Fazel, S. (2021). No causal associations between childhood family income and subsequent psychiatric disorders, substance misuse and violent crime arrests: a nationwide Finnish study of> 650 000 individuals and their siblings. International journal of epidemiology, 50(5), 1628-1638.

Background Childhood family income has been shown to be associated with later psychiatric disorders, substance misuse and violent crime, but the consistency, strength and causal nature of these associations remain unclear.
Methods We conducted a nationwide cohort and co-sibling study of 650 680 individuals (426 886 siblings) born in Finland between 1986 and 1996 to re-examine these associations by accounting for unmeasured confounders shared between siblings. The participants were followed up from their 15th birthday until they either migrated, died, met criteria for the outcome of interest or reached the end of the study period (31 December 2017 or 31 December 2018 for substance misuse). The associations were adjusted for sex, birth year and birth order, and expressed as adjusted hazard ratios (aHRs). The outcomes included a diagnosis of a severe mental illness (schizophrenia-spectrum disorders or bipolar disorder), depression and anxiety. Substance misuse (e.g. medication prescription, hospitalization or death due to a substance use disorder or arrest for drug-related crime) and violent crime arrests were also examined. Stratified Cox regression models accounted for unmeasured confounders shared between differentially exposed siblings.
Results For each $15 000 increase in family income at age 15 years, the risks of the outcomes were reduced by between 9% in severe mental illness (aHR = 0.91; 95% confidence interval: 0.90–0.92) and 23% in violent crime arrests (aHR = 0.77; 0.76–0.78). These associations were fully attenuated in the sibling-comparison models (aHR range: 0.99–1.00). Sensitivity analyses confirmed the latter findings.
Conclusions Associations between childhood family income and subsequent risks for psychiatric disorders, substance misuse and violent crime arrest were not consistent with a causal interpretation.

And the 2013 one based on Swedish register data:

• Sariaslan, A., Larsson, H., D'Onofrio, B., Långström, N., & Lichtenstein, P. (2014). Childhood family income, adolescent violent criminality and substance misuse: quasi-experimental total population study. The British Journal of Psychiatry, 205(4), 286-290.

Background Low socioeconomic status in childhood is a well-known predictor of subsequent criminal and substance misuse behaviours but the causal mechanisms are questioned.

Aims To investigate whether childhood family income predicts subsequent violent criminality and substance misuse and whether the associations are in turn explained by unobserved familial risk factors.

Method Nationwide Swedish quasi-experimental, family-based study following cohorts born 1989–1993 (ntotal = 526 167, ncousins = 262 267, nsiblings = 216 424) between the ages of 15 and 21 years.

Results Children of parents in the lowest income quintile experienced a seven-fold increased hazard rate (HR) of being convicted of violent criminality compared with peers in the highest quintile (HR = 6.78, 95% CI 6.23–7.38). This association was entirely accounted for by unobserved familial risk factors (HR = 0.95, 95% CI 0.44–2.03). Similar pattern of effects was found for substance misuse.

Conclusions There were no associations between childhood family income and subsequent violent criminality and substance misuse once we had adjusted for unobserved familial risk factors.

But they also went further and looked at within-person designs (see also the summary post from 2020). In this case, we look at the same person over time. Sometimes they are relatively wealthy (good job), sometimes they are not (lost the job). Do people commit fewer crimes in their relatively wealthier parts of their lives? Not so say Sariaslan and colleagues in 2017 who looked at which neighborhoods people lived in:

• Sariaslan, A., Larsson, H., Lichtenstein, P., & Fazel, S. (2017). Neighborhood influences on violent reoffending risk in released prisoners diagnosed with psychotic disorders. Schizophrenia bulletin, 43(5), 1011-1020.

Released prisoners diagnosed with psychotic disorders have elevated rates of violent reoffending risk and their exposure to adverse neighborhood environments may contribute to this risk. We identified all released sentenced prisoners in Sweden between 2003 and 2013 (n = 47226) and followed them up for a median period of 4.4 years. We identified prisoners who had ever been diagnosed with a psychotic disorder (n = 3782) or prescribed antipsychotics (n = 7366). We examined 3 neighborhood characteristics: income, proportion of welfare recipients, and crime rate. By fitting generalized mixed-effects and negative binomial regression models and adopting within-individual designs that controlled for all time-invariant unmeasured confounders within each individual, we estimated neighborhood intraclass correlations (ICCs) and associations between specific neighborhood characteristics and violent reoffending. Neighborhood factors explained 13.5% (95% CI: 10.9%; 16.6%) of the violent reoffending risk among released prisoners diagnosed with psychotic disorders. This contrasted with 4.3% (95% CI: 3.7%; 4.9%) in all released prisoners. However, after controlling for unmeasured confounding, these estimates were not statistically significant (ICC_{psychotic disorders} = 0.9%; 95% CI: −0.8%; 2.3%; ICC_{all prisoners} = 0.3%; 95% CI: −0.02%; 0.6%). Similarly, none of the within-individual correlations between the specific neighborhood factors and violent reoffending were significantly different from zero. We found consistent results when we investigated prisoners with other psychiatric and substance use disorders. These findings suggest that placing released prisoners with psychotic disorders in less deprived neighborhoods might not reduce their violent reoffending risk, which may also apply to other psychiatric disorders. The assessment, treatment, and community linkage of high-risk prisoners as a strategy to reduce reoffending needs further research.

This study also included income and welfare support:

Model 2 are those using the within-person design ("Model I: Crude between-estimate; Model II: Within-individual estimate, adjusted for age.").

And there's a Finnish, independent replication as well:

• Airaksinen, J., Aaltonen, M., Tarkiainen, L., Martikainen, P., & Latvala, A. (2021). Associations of neighborhood disadvantage and offender concentration with criminal behavior: Between-within analysis in Finnish registry data. Journal of criminal justice, 74, 101813.

The association between neighborhood disadvantage and crime has been extensively studied, but most studies have relied on cross-sectional data and have been unable to separate potential effects of the neighborhood from selection effects. We examined how neighborhood disadvantage and offender concentration are associated with criminal behavior while accounting for selection effects due to unobserved time-invariant characteristics of the individuals. We used a registry-based longitudinal dataset that included all children aged 0–14 living in Finland at the end of year 2000 with follow-up until the end of 2017 for criminal offences committed at ages 18–31 years (n = 510,189). Using multilevel logistic regression with a between-within approach we examined whether neighborhoods differed in criminal behavior and whether within-individual changes in neighborhood disadvantage and offender concentration were associated with within-individual changes in criminal behavior. Our results indicated strong associations of most measures of neighborhood disadvantage and offender concentration with criminal behavior between individuals. The within-individual estimates accounting for selection related to unobserved individual characteristics were mostly non-significant with the exception of higher neighborhood disadvantage being associated with increased risk for violent crimes. Our findings suggest that criminal behavior is better explained by individual characteristics than by causal effects of neighborhoods.

Finally, it should be mentioned that there are quite a lot of studies of the heritability of criminal or antisocial behavior itself. It goes without saying that if criminal behavior is mainly due to genetics, then (childhood) family incomes, neighborhood disadvantage and whatever cannot be so important causal factors. Studying crime is, however, difficult due to measurement issues. If research relies on self-report, you have self-report bias, lying, and faulty memories. If you use court records, people will claim the courts and the police are biased against this or that demographic, and in any case, most crimes aren't caught or punished, so the data will be noisy in any case. There is only one really good study of the heritability of antisocial behavior more generally which is this one (see the 2019 post for details):

• Baker, L. A., Jacobson, K. C., Raine, A., Lozano, D. I., & Bezdjian, S. (2007). Genetic and environmental bases of childhood antisocial behavior: a multi-informant twin study. Journal of abnormal psychology, 116(2), 219.

Genetic and environmental influences on childhood antisocial and aggressive behavior (ASB) during childhood were examined in 9- to 10-year-old twins, using a multi-informant approach. The sample (605 families of twins or triplets) was socioeconomically and ethnically diverse, representative of the culturally diverse urban population in Southern California. Measures of ASB included symptom counts for conduct disorder, ratings of aggression, delinquency, and psychopathic traits obtained through child self-reports, teacher, and caregiver ratings. Multivariate analysis revealed a common ASB factor across informants that was strongly heritable (heritability was .96), highlighting the importance of a broad, general measure obtained from multiple sources as a plausible construct for future investigations of specific genetic mechanisms in ASB. The best fitting multivariate model required informant-specific genetic, environmental, and rater effects for variation in observed ASB measures. The results suggest that parent, children, and teachers have only a partly “shared view” and that the additional factors that influence the “rater-specific” view of the child’s antisocial behavior vary for different informants. This is the first study to demonstrate strong heritable effects on ASB in ethnically and economically diverse samples.

In other words, they relied on multiple observers for each child, and they modeled the overall view of a given child in a twin design. Doing this removes most of the measurement problems discussed above and gave a shocking value of 96% heritability. This is probably too high, but I would not be surprised if later sophisticated studies also found values above 60%.

Conclusions

All in all, we can be fairly confident that lack of money, broadly speaking, is not an important cause of crime. Whatever theories ascribe this power to money, need to look somewhere else. That somewhere else is mainly genetics. It is not only genetics, but genetics is the starting point of serious discussion. The various crime waves probably don't relate much to genetics (e.g. 1980s US crime wave and related drug abuse), but the long-term historical trend probably does (pacification by execution).