> ## Content Index > Fetch the complete content index at: https://mike-morrison-public-safety-and-accountability.ghost.io/llms.txt > Use this file to discover other available public pages before exploring further. # The Limits of Actuarial Reform: What Predictive Modeling Has—and Has Not—Delivered - URL: https://mike-morrison-public-safety-and-accountability.ghost.io/the-limits-of-actuarial-reform-what-predictive-modeling-has-and-has-not-delivered/ - Published: 2026-09-04T01:11:43.000Z - Updated: 2026-09-04T01:16:56.000Z - Author: Michael D. Morrison - Tags: Bail, accountability, Bail Agents, Bail Bonds, Bail Enforcement, Bail Professional, Bail Recovery, Bail Reform, behind the scenes, Bounty Hunter, constitutional bail rights, court accountability, Court System, Criminal Justice, failure to appear, Government Accountability,, Legal Process, Mike Morrison Bail Bonding, pretrial release, Public Safety, victims rights ***For over a decade, the criminal justice system has been a testing ground for a compelling modern theory: that the foundational challenges of pretrial liberty and public safety could be optimized through data science.*** The high-water mark of this movement occurred when proponents began advocating for widespread adoption of the Public Safety Assessment (PSA), an actuarial risk tool developed with funding from Arnold Ventures. Promoted as a way to bring “Moneyball” analytics into the judiciary, the concept was compelling: by processing a limited set of historical factors through a standardized instrument, courts could reduce human bias, better assess pretrial risk, and reduce reliance on traditional commercial secured bail. The reform argument rested heavily on two propositions: that actuarial models could materially improve predictions about human behavior at initial appearance, and that nonfinancial government-supervised release should generally be regarded as less restrictive than financially secured release. More than a decade later, those propositions can be evaluated against actual experience. The evidence does not establish that actuarial assessment is useless in every jurisdiction, nor does it prove that every government pretrial program is ineffective. It does, however, show that predictive tools have not consistently delivered the transformative outcomes sometimes promised for them. At the same time, longstanding national data continue to demonstrate meaningful court-appearance and recovery advantages associated with commercial surety release. **Part I: The Empirical Reality of the Dane County PSA Study** One of the most rigorous tests of actuarial pretrial assessment came from the Harvard Law School Access to Justice Lab’s randomized controlled trial of the Public Safety Assessment–Decision Making Framework (PSA-DMF) in Dane County, Wisconsin.\[1\] Cases were randomized from mid-2017 through the end of 2019\. In half of the qualifying first-appearance hearings, commissioners and attorneys had access to the PSA-DMF; in the other half, they did not. After a two-year follow-up period, researchers compared the groups across failure to appear (FTA), new criminal activity, new violent criminal activity, pretrial incarceration days, and measures of racial fairness. The PSA did affect judicial decision-making. Commissioners made modest changes in the direction suggested by the risk assessment. Those changes, however, did not translate into statistically significant differences in the outcomes the reform was intended to improve. Researchers found no meaningful change in failures to appear, new criminal activity, violent criminal activity, incarceration days, or measures of racial fairness. That distinction matters. The study does not prove that every risk-assessment instrument fails in every jurisdiction. Implementation, local court practices, data quality, state law, and the characteristics of a particular population can all influence results. What the Dane County trial demonstrated is narrower and still significant: giving decision-makers access to this prominent actuarial tool changed some decisions, but did not produce the hoped-for improvements in measured criminal justice outcomes. The researchers identified two possible explanations. One was that the PSA did not sharply distinguish levels of actual risk within the Dane County population. The other was that differences in bail decisions themselves may have had limited effect on the misconduct being measured. The authors emphasized that neither explanation was certain. That caution should be respected just as seriously as the study’s null results. The lesson is not that mathematics has no place in criminal justice. It is that assigning a numerical risk score does not automatically produce better justice outcomes merely because the process appears more scientific. **Part II: The Misleading Simplicity of “Least Restrictive”** As actuarial tools have faced greater empirical scrutiny, much of the pretrial debate has increasingly focused on the idea of the least restrictive condition or combination of conditions necessary to assure appearance and, where authorized by law, protect public safety. That principle is important. The problem arises when “nonfinancial” becomes synonymous with “less restrictive” without examining what is actually required of the defendant. A release mechanism’s restrictiveness cannot be determined solely by whether money changes hands. It should be evaluated by examining the complete package of financial obligations, surveillance, reporting requirements, geographic restrictions, treatment mandates, and other conditions imposed on the individual. Government pretrial supervision can, depending on the jurisdiction and the case, include: • Continuous GPS monitoring through an electronic ankle monitor. • Curfews and geographic travel restrictions. • Required reporting to pretrial services officers or case managers. • Drug or alcohol testing. • Counseling or treatment requirements. • Other court-ordered monitoring intended to manage perceived risk. When imposed extensively, these conditions can place significant day-to-day restrictions on individuals who have not been convicted. Commercial surety operates through a different structure. The bail agent’s principal financial interest is tied to the defendant’s obligation to appear and the potential forfeiture of the bond. The relationship is private, but that does not make it costless. Defendants or their families may pay nonrefundable premiums, provide collateral, comply with contractual reporting requirements, or accept other lawful conditions. Those burdens can be substantial, particularly for households with limited financial resources. That acknowledgment strengthens rather than weakens the comparison. Commercial surety should not automatically be described as less restrictive simply because it is private, just as government release should not automatically be described as less restrictive because it is nonfinancial. The relevant question is what each arrangement actually requires of the defendant. For some defendants, government-supervised release may impose very little beyond reminders and periodic contact. For others, months of electronic monitoring, testing, travel restrictions, and compulsory reporting may intrude significantly into employment and family life. Likewise, the affordability and contractual requirements of a commercial bond vary by case. “Least restrictive” should therefore describe an individualized outcome—not an ideological preference for one release category over another. **Part III: The Performance Record of Commercial Surety** When pretrial policy is evaluated, theory must eventually be compared with measurable performance. **The Baseline Appearance Advantage** Bureau of Justice Statistics research tracking felony defendants in the nation’s largest counties has historically shown strong appearance performance among defendants released on commercial surety bonds.\[2\] In one widely cited national analysis, defendants released through commercial surety had an observed 18% failure-to-appear rate. Property bonds produced a lower observed rate within a much smaller category, but commercial surety performed better than several heavily used alternatives, including release on recognizance and unsecured bonds. The historical nature of those data should be acknowledged. They do not describe every modern jurisdiction, and pretrial systems have changed substantially since the underlying cases were collected. But they remain important national evidence concerning how different release mechanisms have performed. **Performance After Statistical Controls** One longstanding criticism of raw surety performance data is selection: commercial agents have discretion over which bonds they are willing to write. That criticism deserves consideration. Researchers have therefore attempted to account statistically for differences between defendants. Cohen and Reaves reported that after controlling for factors including defendant characteristics, criminal history, and charge severity, the predicted failure-to-appear probability remained lower for surety release than for recognizance or unsecured release.\[3\] The estimated probabilities were 20% for surety, 24% for Release on Recognizance, and 28% for unsecured release. Statistical controls cannot eliminate every possible unmeasured difference, but the results indicate that the observed appearance advantage cannot simply be dismissed as the product of raw defendant selection. **Dallas County and Matched Defendants** More targeted research has reached similar conclusions within particular jurisdictions. A 2017 peer-reviewed study by Stephen Clipper, Robert Morris, and Amanda Russell-Kaplan examined 29,416 defendants booked into the Dallas County jail and used propensity-score matching to compare similarly situated defendants released through different mechanisms.\[4\] Commercial-bond defendants were less likely to experience failure to appear leading to bond forfeiture than comparable defendants released through the other mechanisms examined. The pattern generally persisted across felony and misdemeanor cases and several offense categories. The authors were appropriately cautious. They did not argue that commercial bail should universally replace every other form of release, and they warned that the findings from Dallas County might not be generalizable to every jurisdiction. That limitation should accompany the result. The evidence nevertheless demonstrates something important: when researchers attempted to address the selection problem directly by comparing similar defendants within one large jurisdiction, commercial surety retained a measurable appearance advantage. **Resolving the Fugitive Window** Court appearance is only one part of pretrial accountability. Another is what happens after a defendant fails to appear. When a defendant released without private surety absconds, responsibility for executing the warrant generally rests with publicly funded law-enforcement agencies operating under competing demands for personnel and resources. Commercial surety creates an additional enforcement incentive. Because the bond may be forfeited, the surety has a direct financial interest in locating and returning the defendant within the period permitted by law. Eric Helland and Alexander Tabarrok examined this distinction in their study published in the Journal of Law and Economics.\[5\] Their research compared failure-to-appear, fugitive, and capture outcomes under public and private enforcement structures and found meaningful advantages associated with the private incentives created by commercial bail. Bureau of Justice Statistics data have similarly shown lower long-term fugitive percentages among surety defendants than among defendants released through several nonfinancial mechanisms. The important policy point is not that public law enforcement ignores outstanding warrants—police agencies work diligently within their resource limits. The difference is structural: commercial surety creates a separate, parallel private financial incentive to pursue the defendant, active alongside and supplementing the public enforcement system. **Conclusion: Returning to Defensible Pretrial Policy** The history of pretrial law has always involved competing concerns: liberty before conviction, court appearance, public safety, fairness, and the proper limits of government authority. No modern release mechanism resolves every one of those concerns perfectly. Some jurisdictions have reported success using court-date reminders, transportation assistance, carefully targeted supervision, and other interventions that help defendants appear without intensive restrictions—narrowing appearance gaps particularly where the alternative would otherwise be pure release on recognizance or unassisted release. Commercial surety also imposes costs that cannot be ignored, particularly where financial conditions contribute to unnecessary detention of people who genuinely cannot afford release. But those concerns do not erase the empirical record supporting surety performance, any more than Dane County’s results establish that every risk-assessment system everywhere must fail. The broader lesson is one of policy discipline. Algorithms should not be accepted merely because they appear objective. Government supervision should not be called less restrictive merely because no commercial premium is involved. Commercial surety should not be defended merely because it is traditional. Each mechanism should be judged by what it actually does: how effectively it secures appearance, protects legitimate public-safety interests, preserves liberty, distributes financial and supervisory burdens, resolves failures to appear, and uses public resources. The evidence supports commercial secured surety as an effective component of that system. It also supports continued scrutiny of every competing model—including our own. That is the standard pretrial policy deserves. **Notes & References** *\[1\] Access to Justice Lab, Harvard Law School. (2024). Randomized Control Trial Evaluation of the Implementation of the PSA-DMF in Dane County, WI: Abbreviated Summary of Final Report.* *\[2\] Cohen, T. H., & Reaves, B. A. (2007). Pretrial Release of Felony Defendants in State Courts. Bureau of Justice Statistics Special Report, NCJ 214994.* *\[3\] Cohen, T. H., & Reaves, B. A. (2007). Pretrial Release of Felony Defendants in State Courts. Bureau of Justice Statistics Special Report, NCJ 214994.* *\[4\] Clipper, S. J., Morris, R. G., & Russell-Kaplan, A. (2017). “The Link Between Bond Forfeiture and Pretrial Release Mechanism: The Case of Dallas County, Texas.” PLOS ONE, 12(8), e0182772.* *\[5\] Helland, E., & Tabarrok, A. (2004). “The Fugitive Apprehension Incentive: An Empirical Analysis of Commercial Bail and Pretrial Release.” Journal of Law and Economics, 47(1), 93–122.* **ABOUT:** Mike Morrison has spent 35 years working as a Professional Bail Agent in Mississippi, building a career grounded in firsthand experience with defendants, families, courts, law enforcement and the day-to-day realities of the criminal justice system. His commitment to education has led to recognition as a bail trainer by the Mississippi Judicial College and invitations to speak before judges, justice court clerks, prosecutors and other criminal justice professionals, including as an invited guest speaker for the Mississippi Attorney General’s Office. For nearly two years, Morrison has worked to develop the Master Class Bail Agent Series with one goal: to elevate the knowledge, professionalism and value of the modern Bail Agent. The series draws from decades of practical experience and focuses on giving Bail Agents a deeper understanding of the laws, procedures and responsibilities that allow them to better serve their clients, their courts and their profession.