Family is often a touchpoint in the national debate over the criminal justice system. To advocates of “criminal justice reform,” the harms to a child caused by sending his mother or father to prison often outweigh the benefits of incarceration. Proponents of this view point to evidence that children whose parents are incarcerated underperform their peers on a variety of indicators.
But the actual research is not so certain. As I’ve written previously for IFS, the aforementioned analyses do not control for the way in which the children of incarcerated parents may have worse outcomes independent of incarceration, e.g. because they were forced to live with the sort of parent who eventually went to prison. Studies that try to identify the causal effect of parental incarceration often estimate it has zero or small but positive effects on a variety of outcomes: academic achievement, future risk of incarceration, and even neighborhood quality. A recently released paper, based on a larger pool of American students than previous research, provides some nuance to the picture, but replicates the essential finding that we dramatically overestimate the harms of parental incarceration to children.
The paper is by NYU’s Arpit Gupta, an adjunct fellow at the Manhattan Institute,1 Christopher Hansman of Imperial College London, and Evan Riehl of Cornell. Their primary interest is in the “incarceration-achievement gradient”—the relationship between how much incarceration there is in a community and that community’s level of academic achievement. At a correlational level, unsurprisingly, the gradient is quite steep. But how much of that, they ask, is due to the effects of incarceration per se, as opposed to forces causing both incarceration and reduced achievement?
To answer their question, the authors exploit an expansive data set, combining public records on prisoners and school children in North Carolina to track what happens to a child when someone in their household is incarcerated (they can’t identify that person’s relationship to the child). To ascertain the causal effect of incarceration, they use a version of a now-standard empirical strategy, the “judge assignment” study, exploiting entry and exit of judges responsible for a county to capture changing average tendency to incarcerate in that county, and thereby changes in actual incarceration. Their estimates suggest that a one standard deviation increase in average judge stringency results in a 15 to 20% increase in annual active sentences.
The authors then look at how changes in county judge stringency—and therefore incarceration—affect achievement in that county. They estimate that a one standard deviation increase in stringency results in a statistically significant 0.024 to 0.036 standard deviation reduction for math scores, and a 0.014 to 0.020 standard deviation reduction in English scores.
Surprisingly, this effect persists even when only looking at students whose household members were not incarcerated—a suggestion that the effect is driven by “spillovers” to peers. Analyzing student-level data, the authors show that this is indeed the case, and note that the total contribution of family incarceration on county-level academic outcomes is driven, not by the student who lived with the incarcerated individual, but far more by the peers they indirectly effect, whose losses account for about 15% of the relationship between incarceration and achievement at the community level. This, the authors suggest, is probably due to the adverse effects of family incarceration on affected students’ classroom discipline, which, in turn, harms other students as well.
In short: incarceration of a household member leads to small reductions in both an affected child’s academic performance and his peers’ academic performance, which compounds at the county level to explain a meaningful share of the “incarceration-achievement gradient.” The paper’s authors are primarily interested in what this says about the relationship between incarceration and social mobility. But what can we learn from this paper about the effects of familial incarceration more generally?
As regards the aforementioned literature, it is true that Gupta, Hansman, and Riehl find a negative effect on academic performance, whereas other papers (most notably Norris, Peceno, and Weaver 2021) usually find no effect. That said, the effect identified by the new paper is remarkably small: a 15 to 20% increase in incarceration in a county cuts achievement by 1 to 4% of a standard deviation. Even at the community level, the causal effects of incarceration explain only about 15% of the relationship between it and achievement—a surprisingly small share given the alleged massive harms of incarceration to families. The paper also finds that familial incarceration reduces the future risk of offense, suggesting that (as Norris, Peceno, and Weaver find) all else equal, a family member going to prison deters rather than encourages future crime (although this finding in the new paper is not statistically significant).
Furthermore, the data are—of necessity—restricted to family members and children who are co-residents (because that is how the two data sets were joined). In other words, they tell us about the effects of incarcerating the restricted set of parents who a) live with their kids and b) commit crimes—a much smaller group, in all likelihood, than the number of individuals who commit crimes and have kids with someone who lives elsewhere. This is, as my colleague and IFS contributor Robert VerBruggen pointed out to me, a constraint of all prior research in this area, but it’s worth emphasizing here.
This is because, if this new research is evidence against incarceration on “interests of the child” grounds, it is only slightly so. In fact, what is most remarkable in these findings is that they explode the assumption that family incarceration is a harm that supports dramatic decarceration. As I argued previously for IFS, while intact households benefit most children, there are likely a group of parents—precisely those most likely to commit serious crimes—whose absence is preferable to their presence.
Furthermore, most of the negative effects on achievement come through an incarcerated child’s misbehavior in the classroom and the adverse effects on peers. To the extent that this is true, it suggests that the incarceration-achievement gradient can be mitigated as much through more effective classroom discipline strategies as it can through changing the incarcerated population.
Gupta, Hansman, and Riehl, to be sure, make a valuable contribution to the study of opportunity. It is worth underscoring, however, that their results can quite reasonably be read to reject, rather than support, the idea that parental incarceration is massively harmful—a finding well in line with prior research.
Charles Fain Lehman is a fellow at the Manhattan Institute and a Contributing Editor of City Journal.
1. Manhattan Institute is my employer. I discussed the paper with Arpit Gupta before writing this article.