Study measures gap between availability and delivery of “second chance” relief

Professor Colleen V. Chien of Santa Clara University has published a major empirical study in the Michigan Law Review that examines the gap between eligibility for and actual delivery of relief from contact with the criminal justice system, a construct she calls the “second chance gap.” (The term is defined with examples here.) Last week, Chien led a team of law students, researchers and data analysts from Santa Clara University in launching the Paper Prisons Initiative, a project that draws on her study’s methodology to estimate this gap for each state’s record relief laws.

During the current wave of criminal record reforms that began around 2013, every state legislature has taken steps to chip away at the negative effects of a record through authorizing or expanding expungement, sealing, and other forms of record relief. At the same time, it has become evident that bureaucratic and structural obstacles prevent many of these laws from achieving their full promise—particularly when they require a potential beneficiary to navigate a complex and burdensome judicial or administrative process.

Last June, Professors Sonja B. Starr and J.J. Prescott published the first broad-based empirical study of a state law limiting public access to criminal records, revealing that just 6.5% of those eligible for relief in Michigan successfully completed the application process within five years. This conclusion has given additional impetus to the movement to make record-sealing automatic: six states now authorize “clean slate” relief for a range of conviction records, 16 states do so for non-conviction records, and clean slate campaigns are underway in several additional states.

In “America’s Paper Prisons: The Second Chance Gap,”  Professor Chien provides a more general perspective on the gap between eligibility and delivery of second chance relief documented for Michigan by Starr and Prescott. In her article, Chien looks at three forms of relief from contact with the criminal justice system—record-clearing, restoration of voting rights, and shortening prison sentences. Based on her analysis, the paper concludes that in many cases only a small fraction of those eligible for relief (usually less than 10%) have received it. Most notably, she uses novel data collection and analysis to estimate that at least 20 to 30 million American adults have non-conviction records that appear to be clearable under existing law but that have not been cleared. In follow-up state-specific papers available on the Paper Prisons Initiative website, she and her team have estimated the second chance expungement gap in convictions relief for a number of states.


The first part of “America’s Paper Prisons” explores the reasons for the “second chance gap”:

In the same way that the accused remain innocent until proven guilty in the U.S. criminal justice system, many second chance programs require defendants to “prove” that they deserve second chances before awarding them. As such, getting one’s second chance through petition-based processes may include enduring a bureaucratic process, amassing information through a variety of sources, and being evaluated by an adjudicative or administrative body. The high cost of doing so in many cases may be insurmountable.

Chien divides the barriers that contribute to the second chance gap into three categories: administrative barriers, like informational and transactional costs; structural barriers, like a requirement to pay court debt and participate in a formal court hearing; and substantive barriers, like perceptions that the costs of the process outweigh its benefits.[1]


The second part of Chien’s paper estimates the second chance gap for an assortment of relief mechanisms, drawing on both original data collection and secondary sources. Chien uses two metrics: the “uptake gap,” which measures the share of individuals over time who are eligible and have not applied or not received relief; and the “current gap,” which looks at the share of individuals at the certain moment in time who are eligible but have not received relief. Her analysis shows, on a nearly uniform basis, that only a small percentage of eligible individuals obtain relief in petition-based programs.

The relief mechanisms she studies confer different benefits, are animated by differing policy objectives, and vary in eligibility criteria and administrative process. Still, a general pattern is clear across the following second chance gaps that Chien presents:

Early release of prisoners

  • During President Obama’s Clemency Initiative, only an estimated 3% of candidates who were eligible under the program’s criteria actually received clemency.[2]
  • Under the federal statute allowing release of federal prisoners for “extraordinary and compelling reasons” (a.k.a. “compassionate release”), just 2 federal prisoners were granted release during a 13-month period in 2015-16.[3]
  • Two California ballot measures, one allowing resentencing and reduction in classification for people convicted of certain theft and drug possession felony convictions (Prop. 47), and another allowing the same for certain marijuana convictions (Prop. 64), had 9% and 3% estimated uptake rates, respectively.

Restoration of voting rights

  • In 12 states that require people who have lost their voting rights due to a felony conviction to apply for restoration of their rights—either through a pardon application, judicial restoration procedure, or other administrative process—an estimated 12% of those who had completed their incarceration and supervision time had regained their voting rights over the course of various sampled time periods.[4]

Expungement or sealing of criminal records

  • Chien and researchers at Santa Clara and Columbia Universities provide an estimate of the national “current gap” in non-conviction relief. They created a novel dataset of 60,000 criminal histories of gig-economy jobseekers randomly chosen at the state level, with background checks between 2017-18, and analyzed them with reference to the non-conviction record-clearance laws of all 50 states. They conclude that an estimated 28 million American adults have non-conviction records that are either fully or partially clearable (state-by-state results are also provided in the appendices).
  • Estimated uptake rates for the expungement of eligible conviction records are generally low across several states, based on original analysis by Chien and the Paper Prisons Initiative. The following are uptake estimates for expungement of eligible conviction records: Washington (<3%); New York (<1%); North Carolina (~13%); Michigan (6.5%), Connecticut (~<3%); and for the expungement of either conviction or non-conviction records: Oregon (~6%), Colorado (5%), Delaware (6%), Iowa (~22%); Missouri (1%), Minnesota (~6%).[5]


The third part of the paper examines practices to narrow the second chance gap, with a menu of insights for those designing and implementing record reforms. Chien illustrates how small differences in how a law is drafted or implemented can lead to divergent outcomes, even when the burden of delivering relief is placed on the government. For example, Alaska and Vermont both have rules purporting to restrict the dissemination of non-conviction records after specified time periods. Alaska prohibits online publication of a case sixty days after all charges result in acquittal or non-dismissal. Vermont courts seal records 12 months after a dismissal or finding of no probable cause. But while Vermont non-conviction records almost never appear on background checks, Alaskan records do, likely owing to the Vermont state repository’s decision not to report information on non-convictions even before the 12-month waiting period has run.

This section includes a useful discussion of how to draft eligibility criteria to allow for implementation at scale, including avoiding “hard-to-ascertain criteria” that present administrability problems. For instance, determining whether fines and fees have been paid is a constant challenge, very specific offenses grade and classification requirements are not always recorded, and specific disposition criteria are vulnerable to changes in definition and “unclean” (inaccurate or incomplete) data (Table 5).


Chien’s final section discusses several open policy and research questions concerning access to relief (the inequities of debt-related barriers to automated relief[6]; the challenges of unclean data and ambiguous dispositions) and the effect of relief (the persistence of sealed records on the internet and in electronic databases[7]; the lack of notice to beneficiaries of automatic relief).

Chien also tackles a concern at the intersection of automatic record relief and racial discrimination in hiring. Studies have found that “Ban the Box” policies, which prohibit criminal history inquiries until later in the hiring process, can lead employers—in the absence of this information—to discriminate against young men of color on the assumption that they have criminal records. Professors Jennifer Doleac and Sarah Lageson, who conducted some of these studies, argued last year that a similar harmful effect might result from broad-based automatic sealing laws. Here, Chien distinguishes automatic sealing from Ban the Box by arguing that because automatic sealing does not apply to all records, but only to a subset of records that meet specific eligibility criteria, employers “won’t know what they are missing,” and as a result will be less likely to replace individualized determinations with generalized discrimination. However, Chien acknowledges the need for more data and research, and that implementation matters.

Another open question not addressed here is the extent to which access barriers can be diminished short of a full “clean slate” solution. While there is growing support for automatic relief, especially when it comes to restoring voting rights and clearing many criminal records, it seems that individualized review is likely to remain the norm in the near term for questions like early release from prison or sealing a serious felony conviction. In many cases, full discretionary review may be replaced with a presumption in favor of relief, which can be overcome following official objection. Moreover, even when a discretionary decision is required, barriers can be minimized by such measures as assigning administrative tasks to the government and providing counsel to those who need it (two key provisions in Virginia’s “transformative” new record-sealing legislation).

Overall, Prof. Chien’s study makes wide-ranging contributions to the field of criminal record relief, bringing to the fore the urgency and challenges of narrowing the second chance gap.

CCRC was pleased to assist with Chien’s analysis of state laws for expunging non-conviction records.  We are also a partner on the new Paper Prisons Initiative, which “conducts empirical research to draw attention to the tens of millions of Americans burdened by contact with the criminal justice system despite being eligible for relief from this contact,” and which has already produced reports estimating the “second chance gap” for record relief in 10 states.

[1] By comparison, Starr and Prescott attribute the low uptake of Michigan expungements to six likely factors: (1) lack of information; (2) administrative hassle and time constraints; (3) fees and costs; (4) distrust and fear of the criminal justice system; (5) lack of counsel; and (6) insufficient motivation.

[2] Chien cites a study of the Initiative by the U.S. Sentencing Commission that attributed at least some part of this gap to President Obama’s decision to apply the criteria announced at the outset of the Initiative more flexibly, and to other administrative issues. In the end, only 5% of those who received commutations satisfied all of the announced criteria. See Glenn Schmitt et al, U.S. Sent’g Comm’n, An Analysis of the Implementation of the 2014 Clemency Initiative (2017),

[3] Chien notes that the requirement that the Bureau of Prisons support a prisoner’s petition for compassionate release was repealed by the First Step Act in 2018, resulting in a substantial increase in petitions granted by federal courts. In 2020 health risks due to COVID-19 led to many more grants.

[4] Chien notes that this estimated percentage may be low because it does not account for the small subset of offenses ineligible for reenfranchisement or for felony-reenfranchisement waiting times. It also does not appear to account for people who have completed incarceration and supervision but remain ineligible to restore their voting rights due to outstanding court debt, in at least some of the states in question (i.e., Alabama, Arizona, Tennessee).

[5] The uptake estimates are derived from the Paper Prisons Initiative’s multi-year comparison of estimated eligible records with estimated expungement rates (except the Michigan estimate, which is based on the Prescott and Starr empirical study). These figures, provided by Chien, were updated from the time of her paper’s publication to take into account a November 2020 SEARCH report, and are available on the Paper Prisons website,, within the individual state expungement or sealing gap reports. See Becki Goggins et al.; Survey of State Criminal History Information Systems, 2020: A Criminal Justice Information Policy Report, SEARCH (2020), available at

[6] Notably, Pennsylvania enacted legislation last year eliminating unpaid fines and fees (excluding restitution and a filing fee) as barriers to existing automated and petition-based sealing.

[7] Virginia’s newly enrolled record-sealing legislation includes a provision requiring that private companies that buy and sell criminal records to routinely delete sealed records (this is also the current practice of Pennsylvania state courts). The bill creates a private right of action for individuals against companies that refuse to do so.