Cash bail punishes poverty, not flight risk

The claim

The cash bail system is defended as a mechanism for ensuring court appearance: defendants who post money bail have a financial incentive to appear, and those who pose genuine flight risk or danger to the community can be detained through appropriately high bail amounts. On this account, the system is a rough but workable proxy for risk — those who fail to appear or reoffend lose their deposit, and those who are genuinely dangerous are priced out of release. The system’s defenders acknowledge it is imperfect but argue that the alternative — categorical release with no financial skin in the game — would produce worse appearance rates and more pretrial crime. The problem, on this framing, is not structural but actuarial: we need better risk instruments, not elimination of bail.

The mechanism

The theory of cash bail rests on a straightforward incentive argument: a financial stake creates a credible commitment to appear. The mechanism breaks in at least three places.

First, the stake is not symmetrical across income levels. A $500 bail amount is functionally zero for a high-income defendant and functionally permanent detention for someone earning $25,000 a year. The same dollar figure produces vastly different behavioral incentives and vastly different deprivations of liberty. If the mechanism were working as designed, bail amounts would be calibrated to individual income — they almost never are.

Second, the mechanism connecting bail to risk is weak in practice. Judges set bail primarily by offense category and prior record, not by individualized risk assessment. Empirical studies using algorithmic risk tools — which assess actual behavioral predictors of appearance — find that they outperform money bail at predicting court appearance at every wealth level, and find no meaningful correlation between a defendant’s ability to pay a given bail amount and their underlying flight risk.

Third, and most importantly, the mechanism produces severe punishment independent of guilt. Pretrial detention — the consequence of failing to post bail — exposes defendants to job loss, housing loss, family separation, and intense pressure to plead guilty regardless of actual guilt. The system therefore punishes poverty at the pretrial stage, before any adjudication of facts.

The evidence

Pretrial detention causes convictions, not merely predicts them. The most rigorous causal evidence comes from Dobbie, Goldin, and Yang’s 2018 study in the American Economic Review. Using the random assignment of defendants to more or less bail-lenient judges as an instrumental variable, they isolate the causal effect of pretrial detention from the selection effect (that detained defendants may be more culpable on average). They find that pretrial detention increases the probability of conviction by 13 percentage points and increases sentence length by 42%. These effects persist when controlling for charge severity and criminal history. The mechanism appears to be plea pressure: detained defendants plead guilty at substantially higher rates, and do so faster, than equivalent defendants who were released. A defendant sitting in jail, losing their job, and facing weeks until trial has strong incentives to accept a guilty plea even when innocent.

Bail amounts predict poverty, not risk. Megan Stevenson’s 2018 analysis of Philadelphia and Miami bail records finds that the strongest predictor of whether a defendant is detained pretrial is not their assessed risk score but the bail amount set by the judge — and that identical risk-score defendants with different bail amounts have sharply different detention rates. The study finds that Black defendants receive bail amounts approximately 35% higher than white defendants with equivalent charges and criminal histories, meaning that racial wealth gaps compound the poverty penalty: race determines both the bail amount set and the likelihood of being able to post it.

Risk-assessment tools decouple appearance from wealth. Arnold, Dobbie, and Yang’s work on the Arnold Foundation’s Public Safety Assessment tool demonstrates that algorithmic risk assessment — using factors such as prior failure-to-appear, age, and charge type, but explicitly excluding wealth — achieves court-appearance rates comparable to or better than money bail systems. The New Jersey courts’ adoption of this tool in their 2017 bail reform produced a 44% reduction in pretrial detention with no statistically significant increase in failure-to-appear rates or new criminal activity, according to the New Jersey Courts’ own annual reporting. Washington D.C. has operated a near-universal non-monetary pretrial release system for decades; its appearance rates consistently exceed the national average.

Scale of the problem. On any given day in the United States, roughly 450,000 people are held in local jails without a conviction — approximately two-thirds of the total jail population, according to Bureau of Justice Statistics data. Many are held not because a judge found them dangerous or likely to flee, but because they could not post bail amounts of a few hundred dollars. The median bail amount for felony defendants in state courts is approximately $10,000; fewer than half of defendants can post it. The pretrial detention population is therefore substantially a poverty-sorted subset of the arrested population, not a risk-sorted one.

Downstream consequences compound the penalty. Beyond the conviction and sentencing effects documented by Dobbie et al., pretrial detention produces cascading material harms. Studies of employment records find that detained defendants lose jobs at rates far exceeding those of released defendants charged with identical offenses. Housing instability, child welfare involvement, and loss of public benefits all follow detention. The criminological literature on these collateral consequences documents a feedback loop: detention produces the very instability — job loss, housing loss — that genuine risk-assessment tools identify as predictors of future court non-appearance. The system thus creates the risk it purports to manage.

International comparison. The United States is one of only two countries in the world — the Philippines is the other — that allows commercial for-profit bail bonding, in which private companies post bail for a non-refundable fee (typically 10%). The UK, Germany, France, the Netherlands, and Canada all use non-monetary pretrial release conditions: release on recognizance, reporting requirements, electronic monitoring, or in serious cases preventive detention orders subject to judicial review on risk grounds rather than financial ability. These systems achieve court-appearance rates comparable to the United States without exposing defendants to commercial exploitation or detaining them for inability to pay.

Who benefits

The commercial bail bond industry — concentrated among a handful of large insurers including Tokio Marine (which underwrites the largest US bail bond surety, Lexington National) and Markel Corporation — collects an estimated $2 billion annually in non-refundable premiums. The industry spends heavily through trade associations including the American Bail Coalition to oppose reform legislation in state legislatures, framing any move toward non-monetary release as a public safety threat. Private prison and jail contractors including CoreCivic and GEO Group benefit directly from pretrial detention population, as county governments pay per-diem rates for beds; their lobbying disclosures reflect consistent opposition to decarceration legislation including bail reform. Prosecutors benefit procedurally: detained defendants are more likely to accept plea offers quickly, which reduces trial caseloads and increases conviction statistics. District attorney associations in several states have formally opposed bail reform on grounds that will raise failure-to-appear rates — claims that subsequent research in New Jersey and elsewhere has not borne out.

The counter

The steelman case for cash bail rests on three genuine observations. First, failure-to-appear rates are real, costly, and fall on victims and courts, not only on defendants. Even if money bail is an imperfect predictor of appearance, some financial stake may have some marginal deterrent effect at the margin for some defendants — though this argument struggles against the New Jersey evidence. Second, risk-assessment algorithms are not neutral: they encode historical data in which policing patterns are racially skewed, meaning that a tool trained on prior arrests may perpetuate racial disparities in detention decisions through a different mechanism. This is a genuine concern in the algorithmic fairness literature, and researchers including Dressel and Farid have raised questions about the accuracy of proprietary risk tools. Third, preventive detention in cases of genuine danger is practiced in all peer countries — the difference is that it is done explicitly, on articulable risk grounds, not covertly through bail-setting. Critics of reform who conflate “eliminating cash bail” with “releasing everyone” are addressing a strawman, but the underlying concern about high-risk defendants requires a genuine legislative answer: a transparent, rights-respecting preventive detention process, which most bail reform proposals include.

The evidence is genuinely contested at the margins regarding the magnitude of appearance-rate effects and the optimal design of risk tools. It is not meaningfully contested that the current system detains hundreds of thousands of people primarily because they are poor.

References

Dobbie, W., Goldin, J., & Yang, C. S. (2018). The effects of pretrial detention on conviction, future crime, and employment: Evidence from randomly assigned judges. American Economic Review, 108(2), 201–240. https://doi.org/10.1257/aer.20161503

Stevenson, M. (2018). Distortion of justice: How the inability to pay bail affects case outcomes. Journal of Law, Economics, and Organization, 34(4), 511–542. https://doi.org/10.1093/jleo/ewy019

Arnold, D., Dobbie, W., & Yang, C. S. (2018). Racial bias in bail decisions. Quarterly Journal of Economics, 133(4), 1885–1932. https://doi.org/10.1093/qje/qjy012

Schlesinger, T. (2005). Racial and ethnic disparity in pretrial criminal processing. Justice Quarterly, 22(2), 170–192. https://doi.org/10.1080/07418820500089929

Gupta, A., Hansman, C., & Frenchman, E. (2016). The heavy costs of high bail: Evidence from judge randomization. Journal of Legal Studies, 45(2), 471–505. https://doi.org/10.1086/688907

Lowenkamp, C. T., VanNostrand, M., & Holsinger, A. (2013). Investigating the impact of pretrial detention on sentencing outcomes. Laura and John Arnold Foundation.

New Jersey Courts. (2018). New Jersey judiciary pretrial release report: 2017 annual report. Administrative Office of the Courts.

Dressel, J., & Farid, H. (2018). The accuracy, fairness, and limits of predicting recidivism. Science Advances, 4(1), eaao5580. https://doi.org/10.1126/sciadv.aao5580

Goldkamp, J. S., & Gottfredson, M. R. (1985). Policy guidelines for bail: An experiment in court reform. Temple University Press.

Digard, L., & Swavola, E. (2019). Justice denied: The harmful and lasting effects of pretrial detention. Vera Institute of Justice.