There is no current consensus among researchers or policy makers regarding a definition of a person who has been through an incarceration experience. As most of the analysis in this area has been compiled from surveys that collect data for purposes other than capturing information on criminal justice contact or incarceration experience, the questions posed to respondents are not uniform across surveys. Because of these different measures, there is substantial variation in estimates of the variables of interest.
Chris Wildeman (Cornell University) began by discussing a figure that depicts—in a simplified way—the path of criminal justice contact and all of the potential outcomes that could occur after commission of the crime, flowing through arrest, police contact, charges, bail, conviction, imprisonment, and parole in the “most involved” scenario (see Figure 2-1).
One can use the figure to help in thinking about the intervention points that can be considered for exploring the causal mechanisms at different levels of involvement, Wildeman said. Criminal justice and inmate surveys fall short of capturing all these variations (as they focus on certain sub-populations of the criminal justice community at a certain stage in their experiences), while national non-institutionalized population-based health surveys do not provide information on the prevalence of the health conditions in the incarcerated population. Wildeman noted that it is important to consider not only the variations within the different types of criminal justice contact but also variations across them. He also suggested that although
one’s first inclination may be to study the causal relationships between criminal justice involvement and health, the criminal justice contact might also serve as a risk indicator. If a person has been arrested, for example, what implications does that lend toward his or her likelihood for certain health outcomes?
Wildeman pointed out that there is limited information available on the health of adults who have experienced the post-conviction stages of criminal justice contact—incarceration, probation, and parole—such as the mortality of inmates and the physical and mental health of inmates. And there is even less information available about people on probation, parole, or no longer under any kind of supervision. In terms of thinking about the connections between criminal justice contact and health, the research community has not gathered many data on the contact points in the “middle” of the figure. And when there are health and criminal justice linkage data for the bookends, they are often collected at a high level and therefore not able to be disaggregated by type of institution or any other degree of contact specificity.
When comparing rates of criminal justice contact within and across subgroups, one would look for disparities between those groups. Wildeman cautioned, however, that the nature of the criminal justice contact questions that are asked could actually influence the relative size of the disparities. He sees this often in his own research, which focuses on racial disparities in the criminal justice system. Crime type, conviction type, and transition probability (likelihood of conviction after arrest) are examples for which one might see varying degrees of racial disparity depending on the measurement variable used: one example is drug conviction rates, which might yield higher racial disparities than homicide conviction rates. The alternative is to look at aggregate totals instead of breakdowns by type, but the tradeoff is loss of the ability to examine the disparities by subgroup.
On the basis of the existing literature and analyses of the effects of criminal justice contact on health, Wildeman said he surmises that the empirical models explain, at best, that 40 percent of the association is causal. The criminal justice mechanisms and the ways in which they affect health differ at every contact level and can also differ within the contact level, as he had noted. The mechanisms driving the relationship between arrest and health, for instance, could be very different from the mechanisms driving the relationship between imprisonment and health. And the context of imprisonment—such as solitary confinement, differing security levels, and distance from family—can all potentially affect the extent to which any mechanism is enabled.
Wildeman noted that measurement becomes even trickier when considering how the contact affects an individual’s family and community due to vague measures of criminal justice contact and the omission of
specific variables in regression models. He suggested that criminal justice contact questions in national health surveys can be modified to get a better handle on the measure, with the added benefit that the surveys already have information on covariates to test for causal mechanisms of criminal justice contact on health. To achieve those objectives, Wildeman proposed five question topics for inclusion on a nationally representative population health survey:
- self-reported criminal activity;
- whether someone has ever been arrested; if yes,
- ever incarcerated; if yes,
- type of crime for which incarcerated; and
- type of facility in which incarcerated.
Wildeman noted that the likelihood of being able to include all five of these questions was slim, but that delving into the type of incarceration could also prove incredibly useful.
John Laub (University of Maryland) asked Wildeman why he placed emphasis on crime and conviction type, especially when 90 percent of cases are resolved by plea bargains, which means that the crime for which a person is punished may not accurately reflect the original act. Wildeman agreed that because of plea bargains, the information that is obtained might create ambiguity for researchers when trying to identify the conviction and its prevalence.
David Cantor (University of Maryland) and Daniel Nagin (Carnegie Mellon University) both commented that learning the nature of the conviction was an important step in determining an appropriate response. Amanda Geller (New York University) agreed that crime type was important, particularly when trying to distinguish between the effects of criminal justice treatment and engagement in criminal justice activity. She also noted that attempting to distinguish the committed crime from the crime to which a person entered a guilty plea will introduce a significant degree of endogeneity.
Wendy Manning (Bowling Green State University) discussed the key findings from a 2012 forum and study, as well as a subsequent survey that was birthed at the forum.
The National Center for Marriage and Family Research (NCMFR) 2012 forum included the following topics: assessing existing measures of incarceration in household surveys; adapting measures for household surveys; recommendations for future data collection efforts; and linking
administrative records data to household surveys. At the NCMFR forum, three surveys that collect data on incarceration experience were compared: the National Survey of Family Growth (NSFG), the 1997 National Longitudinal Survey of Youth, and the National Longitudinal Study of Adolescent to Adult Health. The NCMFR study focused on data collected for male respondents who reported incarceration by age 25. The study panel looked at differences in question content and mode of administration, and weighed the benefits and disadvantages of asking a single question about incarceration experience in comparison with asking multiple questions. The study also pointed to significant variations in reported rates of incarceration, based on the structure of the question(s) about incarceration.
Manning explained that the data disparities among the three surveys motivated the panel to attempt to develop new methods of criminal justice experience measurement that would provide a more accurate picture of how these experiences affect children and families. The panel created an online pilot survey, the Survey of Criminal Justice Experience (SCJE), with items determined by the forum discussions on the most important questions to ask on population surveys for information about criminal justice involvement.
The online survey was administered to 3,200 adults aged 18 to 64, and it had a 62 percent completion rate.1 The questions were divided into two categories—criminal justice experience and criminal justice supervision—and asked about these phenomena in terms of adult (since age 18) and juvenile (prior to age 18) involvement. The “experience” questions asked about arrest and conviction, and the “supervision” questions asked about probation, jail, and prison. Approximately 20 percent of the respondents reported having had criminal justice experience. Of those, Manning said, about 10 percent pled guilty to or were convicted of a crime. The differences by race, gender, and age range were within expected ranges, with men having higher involvement than women and African Americans having higher involvement than whites and Hispanics.
Manning and her team compared these data to NSFG data for respondents aged 18-44: the rates of criminal justice experience were higher on the NSFG. She said this may be due in part to the way in which the NSFG question is worded, which might encourage respondents to recall any criminal justice experience and not just experience that yielded supervision. Manning suggested to the workshop participants that questions being
1 The survey was administered by GfK, a market research organization, in May 2013 using its online research panel. Panel members were recruited initially through landline telephone using the random-digit dialing (RDD) sampling method. In response to the growing number of cell-phone-only households that are outside the RDD frame, in 2009, GfK switched to the probability-based sampling of addresses from the U.S. Postal Service’s Delivery Sequence File (Brown and Manning, 2013).
considered for a survey should be cognitively tested and that researchers should continue to do comparative work to gain a better understanding of what data are being captured across surveys.
Rashida Dorsey (U.S. Department of Health and Human Services) asked what criteria were used in selection of questions for the SCJE and whether or not any cognitive testing had been done. Manning responded that the committee compiled a list of criminal justice experience questions currently asked on national surveys and pared the list on the basis of insight and information obtained and exchanged at the NCMFR forum. The committee then circulated that short list to the researchers who attended the forum and asked for their feedback. Manning reiterated her opinion that any subsequent criminal justice experience question design for a survey should include a round of cognitive testing. Dorsey then asked whether Manning had any history on the pool of existing questions with which the researchers started, to see, for example, if they had been cognitively tested or gone through any prior iterations. Manning said that there was minimal historical information on the origin of the questions on those national surveys.