Racial disparities in crime and wealth

This paper asks whether racial differences in labor income can simultaneously explain both the crime gap and the wealth gap between Black and White individuals in the United States. The authors build a large-scale overlapping generations (OLG) model in which property crime is endogenously determined — agents choose whether to steal alongside their consumption and savings decisions — while drug-related incarcerations are treated as exogenous, reflecting evidence that racial profiling distorts enforcement independently of offending behavior. The model is calibrated to match several well-documented racial disparities: Black individuals comprise 12.36% of the adult population but 33.8% of the incarcerated population; 42.7% of Black individuals fall in the bottom wealth quintile (below $3,400 in assets) versus 15.1% of White individuals; the median Black-White wealth gap is 89.5% (SCF 2019). Data sources include the Survey of Consumer Finances (SCF 2019), Uniform Crime Reports (UCR 1996–2011), NLSY79, PSID (1968–2021), and MORG (2000–2019).

The model incorporates four dimensions of labor market disparity between Black and White agents: educational attainment, unemployment risk and duration, age-earnings profiles, and idiosyncratic income shock processes. It also incorporates race-skill-specific survival probabilities (life expectancy at birth: 73 years for Black, 78 years for White), scarring effects from incarceration on future labor income, a progressive income tax, means-tested transfers, and accidental bequests distributed within race groups.

The benchmark model successfully replicates key data moments. Black individuals constitute 34.3% of the incarcerated population (data: 33.8%). The model-generated median wealth gap is 83.6% (data: 89.5%). The share of Black individuals in the bottom wealth quintile is 37.7% in the model versus 42.7% in the data. The model does not match the average wealth gap: the model-generated gap is 58.9% versus 84.4% in the SCF.

The main counterfactual experiments yield three findings. First, equalizing labor market conditions — particularly age-earnings profiles — is the dominant driver of both racial wealth and crime disparities. When all labor market conditions are equalized, the Black crime rate falls by 66.25% (from 11.97% to 4.04%), the median wealth gap declines by 69.6% (from 83.58% to 25.4%), and the share of Black individuals in the bottom wealth quintile falls from 37.73% to 20.75%. Equalizing age-earnings profiles alone accounts for the largest single-factor effect: the median wealth gap declines from 83.58% to 44.16% and the Black crime rate from 11.97% to 7.59%. The resource cost of equalizing age-earnings profiles is estimated at 3.29% of GDP for the No-HS group and 22.2% of GDP for the HS group.

Second, higher crime and incarceration rates among Black individuals do not significantly contribute to their lower wealth. When crime is entirely eradicated, the share of Black individuals in the bottom quintile barely moves (37.73% to 37.61%), and the median wealth gap falls only from 83.58% to 82.6%. The mechanism is that most crimes are committed by young, already-poor individuals who are not saving in any case; income loss during incarceration is not large enough to affect wealth accumulation meaningfully.

Third, equalizing life expectancy generates a 25.39% reduction in the median wealth gap and a 12.4% decline in the share of Black individuals in the bottom wealth quintile, with negligible effect on crime rates.

The paper also validates the model against Cesarini et al. (2023), who find a small, statistically insignificant effect of lottery wealth on criminal behavior in Sweden. The model replicates this finding: a $150,000 windfall reduces incarceration risk over seven years by 0.81 percentage points. The mechanism is that lottery winnings displace means-tested transfers, winnings gradually dissipate as low income persists, and individuals eventually return to poverty and resume criminal activity.

Q: What is the central research question and why does the paper treat property crime and drug crime differently? A: The paper asks whether racial labor income differences can simultaneously account for both crime and wealth disparities. Property crimes are modeled endogenously because offending behavior responds rationally to economic incentives. Drug crime incarcerations are exogenous to capture evidence that racial profiling in enforcement — rather than differential offending alone — drives racial disparities in drug arrests: Beck and Blumstein (2018) show differential offending explains only about 52% of the drug imprisonment gap, versus over 70% for overall imprisonment.

Q: What are the benchmark model’s key calibration targets and how well does it fit the data? A: The benchmark targets Black individuals as 34.3% of the incarcerated population (data: 33.8%), a median Black-White wealth gap of 83.6% (data: 89.5%), and 37.7% of Black individuals in the bottom wealth quintile (data: 42.7%). The model does not match the average wealth gap: the model-generated gap is 58.9% versus 84.4% in the SCF, which the authors acknowledge explicitly.

Q: What is the quantitative effect of equalizing all labor market conditions? A: Experiment 5 (equalize educational attainment, unemployment risk, and age-earnings profiles jointly) reduces the Black crime rate by 66.25% (from 11.97% to 4.04%), the median wealth gap by 69.6% (from 83.58% to 25.4%), and the share of Black individuals in the bottom quintile from 37.73% to 20.75%. Equalizing all factors including life expectancy drives the median wealth gap to 0%, with the bottom-quintile share for Black individuals at 19.31%.

Q: Which single labor market factor matters most for the wealth gap and crime rate? A: Equalizing age-earnings profiles (Experiment 3) is the single most important factor, reducing the median wealth gap from 83.58% to 44.16% and the Black crime rate from 11.97% to 7.59%. By contrast, equalizing educational attainment or unemployment risk each reduces the median wealth gap only to 76.71%, with smaller crime effects.

Q: Does education-group heterogeneity matter for interpreting the age-earnings equalization effect? A: Yes, substantially. Equalizing age-earnings profiles for the No-HS group reduces the Black crime rate by 21% with little effect on the median wealth gap. Equalizing profiles for the HS group reduces the median wealth gap by approximately 40% with a much smaller effect on crime rates. The earnings channel to crime operates primarily at the bottom of the education distribution, while the earnings channel to wealth accumulation operates more strongly in the high school group.

Q: Why does crime have so little effect on the wealth distribution? A: Criminals are predominantly young and already-poor individuals who are not accumulating savings. Because these individuals have minimal assets and rely heavily on means-tested transfers for consumption, the income loss during incarceration does not reduce their wealth meaningfully. When crime is completely eradicated, the share of Black individuals in the bottom quintile falls only from 37.73% to 37.61% and the median wealth gap declines from 83.58% to only 82.6%.

Q: What is the effect of eliminating drug-related incarcerations on the Black wealth distribution? A: Experiment 4 (eliminating drug crime incarcerations) reduces the share of Black individuals in the bottom quintile only slightly, from 37.73% to 37.30%. Eliminating the scarring effect of all incarcerations likewise has negligible effects on the bottom-quintile share (37.53% versus 37.73% in the benchmark) and the zero-assets share (32.56% versus 33.24%). Neither the direct incarceration penalty nor its labor market scarring meaningfully affects wealth accumulation.

Q: What happens to crime and wealth when the property crime clearance rate changes? A: Doubling the clearance rate from 17.2% to 34.4% reduces the Black crime rate from 11.97% to 1.72% and the White rate from 3.05% to 0.52%, with minimal change in the wealth distribution (Blacks in bottom quintile: 37.83%). Halving the clearance rate to 8.6% more than doubles Black crime to 27.53% and White crime to 9.55%, and increases the share of Black individuals in the bottom quintile by about 11% to 42.01%. This asymmetry — crime reduction barely helps wealth but crime increase does hurt — is consistent with the poverty-trap mechanism.

Q: How does the model validate against the Cesarini et al. (2023) Swedish lottery study? A: Cesarini et al. find a small, statistically insignificant negative effect of a $150,000 lottery windfall on conviction rates. The model replicates this: simulating 34,709 individuals per skill-race group, a $150,000 windfall reduces incarceration risk over the following seven years by 0.81 percentage points. When the authors use model-generated property crime records rather than incarceration records as the dependent variable, they find a statistically significant effect more than twice as large, suggesting incarceration data systematically understates the crime-reducing effect of wealth shocks.

Q: What is the mechanism by which lottery winnings have minimal persistent effects on crime? A: Lottery winners in the model are disproportionately drawn from low-income, low-wealth individuals who also receive means-tested transfers. After winning, these individuals lose transfer eligibility, so winnings substitute for lost transfers rather than being invested. With income levels remaining low, winnings dissipate over time, individuals return to poverty, and resume criminal activity. Larger lottery prizes extend the crime-free interval but do not permanently alter behavior.

Q: What is the role of life expectancy differences in racial wealth and crime gaps? A: Equalizing survival probabilities generates a 25.39% reduction in the median wealth gap and a 12.4% reduction in the share of Black individuals in the bottom quintile, with virtually no change in crime rates. The channel operates through savings incentives: a shorter expected lifetime (73 years for Black versus 78 for White) reduces the return to wealth accumulation independently of income.

Q: What are the fiscal resource requirements implied by the income equalization experiments? A: Implementing equalized age-earnings profiles for the No-HS group would require resources equal to 3.29% of total GDP, while equalization for the HS group would require 22.2% of GDP. These figures reflect the scale of redistribution needed to close earnings profiles and serve as a benchmark for assessing policy feasibility.

Q: How does incarceration scarring affect lifetime income in the benchmark, and how does this validate against external data? A: A Black high school graduate who experiences at least one incarceration earns 16.8% less over his lifetime than one who is never incarcerated; for White high school graduates the gap is 28.7%. Gordon et al. (2023) report corresponding empirical estimates of 18.6% for Black and 32.7% for White high school graduates, closely validating the model’s scarring calibration.

Endogenous property crime: A rational choice by working-age agents who weigh the expected gain from stealing (fraction γ = 6.4% of average labor income y) against the probability of apprehension (clearance rate πa = 17.2%), the loss of means-tested transfers, scarring of future labor income, and the minimum consumption floor in jail. Retired agents face no such choice.

Exogenous drug incarceration: Incarceration for drug possession modeled as an exogenous shock with race-age-specific probabilities, not responsive to individual optimization, capturing the possibility that racial profiling in enforcement generates disparities in drug arrests independently of offending behavior.

Scarring effect: Post-incarceration labor income penalty modeled as a higher probability of drawing a lower idiosyncratic income shock state upon labor market re-entry, calibrated so the model reproduces lifetime income gaps between ever-incarcerated and never-incarcerated individuals by race-skill group (18.6% for Black HS, 32.7% for White HS per Gordon et al. 2023).

Age-earnings profile (ε^{i,ζ}_j): The deterministic, skill-race-age-specific component of labor income estimated from PSID data for each of six race-education groups. The gap between Black and White age-earnings profiles is identified as the dominant driver of both the racial wealth gap and racial crime disparities, accounting for the largest single-factor reduction in both outcomes across all counterfactual experiments.

Means-tested transfer floor: A consumption support program that fills the gap between an agent’s post-tax income plus assets and a minimum threshold κ (5.8% of average net tax income and assets). This transfer is a critical mechanism linking wealth shocks to crime: lottery winnings and other wealth gains displace transfer eligibility, causing winnings to be consumed rather than saved, and eventually exhausted.

Median wealth gap: The percentage difference between median White and median Black wealth — 89.5% in the 2019 SCF, 83.6% in the benchmark model — used as the primary scalar summary of racial wealth disparity, chosen because the model does not match the average wealth gap (model: 58.9%, data: 84.4%).

Victimization probability (πv(Y)): A step-wise decreasing function of taxable income capturing spatial concentration of property crime in low-income neighborhoods; in equilibrium this equals the aggregate property crime rate χp, ensuring market clearing in the crime sector and implying that poorer agents face higher victimization risk.