Association for Public Policy Analysis & Management

There is a growing literature that estimates causal effects of parental incarceration on child outcomes using variation in judge harshness (Bhuller, Dahl, Løken and Mogstad (2018) use Norwegian data, Arteaga (2018) uses Colombian data, Norris, Pecenco and Weaver (2018) use data from Ohio, Dobbie, Grönqvist, Niknami, Palme and Priks (2018) use Swedish data). Dacass and Gaulke (2019) find children are also impacted by sentence length using variation across states and time in Three Strikes laws.

Cunha and Heckman (2007) show that child development is nonlinear in age, and early childhood is especially important. Therefore, failure to account for different developmental periods may lead to imprecise estimates of the impact of parental incarceration. Arteaga (2018) finds the effects vary by age such that those aged 5-10 had smaller effects than those 0-5 or 10-15. Dacass and Gaulke (2019) find the negative impacts are especially prominent for kids who were age 9 when Three Strikes went into effect. In this paper we empirically test whether the effects of parental incarceration vary by age of the child using U.S. data. Given the literature on the importance of early childhood environments, we expect the impacts are largest for the younger children.

This growing literature has had to overcome empirical issues related to endogeneity given that incarceration is not random. Ewert, Sykes and Pettit (2014) and Wakefield and Wildeman (2013) document that incarceration mostly affects those families who are more disadvantaged. To address this concern, we follow Black, Breining, Figlio, Guryan, Karbownik, Nielson, Roth and Simonsen (2017) by implementing a difference-in-differences approach that compares outcomes across children within families that experience parental incarceration to outcomes across children within families that do not experience parental incarceration. If one were to compare outcomes of children within the same family and use a family fixed effect that would ignore the findings on birth-order effects. The Black et. al (2017) method will allow for a plausible causal estimate. We use restricted data from the National Longitudinal Study of Adolescent to Adult Health to estimate how the impact of parental incarceration varies by age of the child. The restricted data allows us to match siblings within the sample. We have information on whether the father (or mother) was incarcerated and the age of the child when the father (or mother) was first incarcerated. The data allows us to estimate impacts on educational outcomes, incarceration outcomes and labor market outcomes.

This research is important for public policy because if the impact on children varies by age this would suggest that certain age groups may benefit more from interventions when their parent is incarcerated.