Presumably, one of the purposes of prison is to change or ‘rehabilitate’ inmates so that they do not reoffend – to reduce recidivism, in policy language. The reality is, many formerly incarcerated people do reoffend and many return to prison. Recidivism rates, therefore, are an important set of data for criminal justice policymakers. What’s more, internationally-funded projects addressing crime and violence are focusing more on the prison sector, especially on building education programs and alternatives to incarceration.
This attention is much overdue: prisons are the least-resourced component of citizen security projects with international funding in the region, by far. These projects often import standard monitoring and evaluation tools, which demand clear, quantitative indicators. Recidivism is an important and commonly used indicator – but it is complex and requires a careful approach.
It seems obvious that any project that can claim a reduction in recidivism should be called a success, and that the opposite would signal a problem. But simple statements of recidivism outcomes without a clear definition or a specific context can lead to misleading interpretations, especially when making comparisons across years or across countries.
To examine this issue more concretely, we looked at recent corrections-sector projects implemented with international funding in Central America and the Caribbean, including IDB-funded loans.
Varying Definitions of Recidivism
The projects reviewed share similar objectives: reducing youth and/or adult perpetration of violent crime. Some specify the objective of reducing recidivism among juveniles and/or adults who are already incarcerated. These are clearly two separate objectives, as violent crime rates are not the same as recidivism rates. Violent crime rates typically draw on police data. In contrast, the definitions of recidivism vary substantially and lack detail. Some projects state that recidivism refers to re-offending, but do not specify whether this is tracked by arrest, conviction, incarceration, or something else. Others state that recidivism refers to re-admission to a detention facility, but do not specify charge or reason for detention. Still others define recidivism as a re-conviction within a given period of time after release. One project defines it as readmission three times or more over a lifetime.
So, even in a small set of projects with relatively similar designs, the definitions of recidivism fall into at least three different categories: re-arrest, re-conviction, and re-admission. None of these are wrong approaches; there is no “ideal” way of measuring recidivism, though some methods have more limitations than others. The main challenge arises when comparisons of indicators do not take into account the differences in these definitions and in the context within which these are measured.
Using re-arrest as a proxy for recidivism tends to over-count actual reoffending. Because arrest patterns often reflect shifting police tactics more than actual crimes, arrest rates are high for young men in “high risk” neighborhoods, and relatively low for less visible offenses, such as domestic violence.
Conversely, using re-conviction as the definition can under-count actual reoffending. Re-conviction may be the best measure in theory, because it provides the justice system’s confirmation that the accused did commit the offense. But, conviction rates are low in Latin American and the Caribbean. For example, in Latin America, convictions occur only in 24% of reported homicide cases, compared to a global average of 43% and an 81% rate in Europe.
Finally, using re-admission to prison as the definition may over-count “re-offending,” since many people are re-admitted on pretrial detention, for unpaid fines, or for technical violations. Asking merely whether or not it is a person’s second time through the prison door does not distinguish whether the first incarceration was a “real” offense. Also, some systems do not integrate data across facilities, and so may not count a second admission to a different facility as re-admission.
Varying Sources
Most of the projects do not explain the data source or methodology for recidivism figures, such as whether data are limited to a particular facility or population. What’s more, some do not take into account the difference between self-reported recidivism data (asking inmates whether they had been previously incarcerated) versus data from official records taken at different stages of incarceration. Finally, the sampling approach also varies: some projects track a sample of people who went through a certain program, some select a sample representative of the population, and some use records of the full population. The demographics and criminal justice history of the people in a sample influence their likely recidivism patterns, and so should be clearly set out. Again, all of these sources are valid, but comparing data with different types of sources can be misleading.
Another aspect of defining data sources relates to whether there is a control group or not. General institutional-level recidivism data provide a window on the overall performance of the system, but evaluations of programs often compare the group that participated in the intervention to a control group that did not. Evaluations also compare recidivism rates of those who completed a program tothose who could not or did not complete for various reasons. Any comparison of recidivism data should explain the justification for the comparison and the design of the two comparison groups.
Even though recidivism data is sometimes cited comparatively across countries, there is no single data warehouse that lays out these distinctions in what’s behind the data. Thus, it is difficult to draw conclusions and policy recommendations. Based on the limited information available through open-source policy and government reports, re-admission to the prison seems to be the most commonly-used definition. It is the most concrete and the easiest measure to track within a single institution.
Potential Unintended Consequences of Comparing Recidivism Numbers
These diverse approaches lead to recidivism “figures” that can become almost meaningless out of their own context. The projects examined in the study do not have final numbers for these indicators, but the potential misinterpretations are evident in recidivism figures listed in other recent government and policy reports. For example, reported figures range from 46% in Uruguay (2010 prisoner census: self-reported readmission) and 59% in Guatemala (2010-2013, government report on a sample of inmates) to 24% in the Bahamas (government reported figures for 2013), 46% in Argentina (2010 prisoner census: self-reported readmission with conviction), 15% in Belize (2006-2012 government report: recorded readmission), and 11% in El Salvador (2015 prisoner census: self-reported readmission).
This comparison does not necessarily tell us much about whether the prison system in each country is performing well or is more or less effective at rehabilitating inmates. Countries that impose harsher sentences on more people, even for lesser crimes, such as much of Central America under mano dura hardline law enforcement tactics, are likely to have lower recidivism simply because more people are behind bars for longer, without a chance to reoffend. This is not necessarily “success” for the justice system. Conversely, a country that makes a clear policy decision to build alternatives to incarceration for certain types of defendants may actually see an increase in recidivism, as the people who remain in the prison system (and are counted in recidivism statistics) are the more serious, difficult inmates (more likely to reoffend). In this case, reducing overall incarceration may be a positive move for justice policy more broadly. Increases in recidivism rates are always a cause for concern and scrutiny. However, if rising recidivism rates are due mainly to desirable system-level policy changes, they should be explained in this context – and compared to a similar population in prior years – rather than interpreted simply as a “negative” outcome.
Implications and Recommendations
Recidivism data are a crucial tool for tracking the consequences of many aspects of the justice system, especially prisons. Yet, in the countries where data is most needed to guide policy, it is scarce, opaque, and inconsistent. Projects involving recidivism data should invest resources in building more robust data and in ensuring that the data parameters are clear and appropriate.
The Urban Institute recommends some ways to improve recidivism data and also proposes alternative indicators to measure prison performance in the US; these lessons could be adapted for countries in Latin America and the Caribbean. First, data systems could track desistance from crime – that is, not reoffending – rather than just the first ‘failure’ of reoffending. Second, tracking the severity of the subsequent offense gives more nuance to recidivism data. Third, tracking specific behavior changes that are supposed to change due to a certain treatment programs generate additional indicators. Fourth, even basic re-arrest or re-conviction data should break down numbers by the risk level and supervision type (parole, probation, etc.) of the individual. An in-depth comparative study on measuring recidivism in some Latin American countries, published by the University of Chile’s Center on Citizen Security Studies, raises some similar concerns and recommendations.
In short, as the standard requirements of project results frameworks and the complex context of prison systems converge, improving recidivism data is a priority. Simple indicators are tempting, but can conceal more than they reveal. Recidivism rates dropping and rising may not tell the story we assume at first glance. When presenting recidivism indicators, we must be able to clearly explain exactly what changed, how we know, and why it happened.
By Jennifer Peirce and Lina Marmolejo
Peirce and Marmolejo have been consultants for the IDB. Peirce is a PhD student in Criminology at John Jay College of Criminal Justice and a Pierre Elliott Trudeau Foundation Scholar. Marmolejo is a PhD student in Criminology at the George Mason University.
Most of our blog posts are in Spanish but occasionally we publish texts in English. Go here if you wish to subscribe to Sin Miedos
Leave a Reply