The US is now in its third decade of a devastating drug crisis, with roughly 92,000 drug deaths in 2020 alone. Between 1999 and 2020, over 932,000 people died of drug poisonings—about the same number of US soldiers that died in all armed conflicts from the Civil War through the present day (about 990,000).1 As the drug crisis has developed, research has documented the impact of the epidemic on employment (Aliprantis and Schweitzer, 2018; Harris et al., 2020; Currie et al., 2019; Powell, 2021a), crime (Doleac and Mukherjee, 2022; Szalavitz and Rigg, 2017; Dave et al., 2021), infant health (Ziedan and Kaestner, 2020), marriage (Kaestner and Ziedan, 2020), enrollment in disability insurance programs (Cutler et al., 2017), and intimate partner violence (Stone and Rothman, 2019).

An important feature of the crisis is that it has primarily impacted people in early adulthood through mid-life. As Figure 1 shows, the drug death rate (deaths/100,000) between 1999 and 2019 was highest for those between the ages of 25 and 54—an age range in which many will be parents of young children and adolescents. In this paper, we examine how this drug crisis among adults has affected the living arrangements of children. Drug use could separate a child from one or both parents through several pathways. First, a parent who is using drugs could leave the household voluntarily, or be asked to leave by a family member. The parent could be absent due to enrollment in a substance abuse treatment program, or due to incarceration because of a crime related to the abuse. If the drug use results in child abuse or neglect, the child welfare system could become involved, increasing the chance of a court-ordered placement with another relative or a foster family. The child could also move into the home of a family friend or relative to protect them from the consequences of the parent's use. Finally, a child's parent could have been one of the many people who have died because of the crisis.

There is abundant anecdotal evidence in the popular press which suggests that many children are indeed experiencing these kinds of events. A 2016 article in the Wall Street Journal reported that “Social workers say the scale of the trouble exceeds anything they saw during the crack-cocaine or methamphetamine crises of previous decades” and quoted a child welfare worker who warned that “honestly, if something doesn't happen with this addiction crisis, we can lose a generation of kids” (Whalen, 2016). Outlets including Vox, NBC, PBS, CNN, and the Associated Press have noted that grandparents are increasingly assuming parenting responsibilities because of the opioid crisis—especially in states that have been hard-hit.2 National statistics support the observations in these stories—between 1980 and 2018, the fraction of children living in a household headed by a grandparent more than doubled from 3.7 to 8.3 percent.3

In recent years, researchers have documented a relationship between the drug crisis and foster care. The number of children entering or living in foster care and the fraction of removals from the home related to drugs have increased (Meinhofer and Anglero-Diaz, 2019). Counties with higher opioid poisoning death rates or drug prescriptions have been shown to have more foster care entries and child removals (Radel et al., 2018; Quast, 2018; Quast et al., 2018; Quast et al., 2019; Bullinger and Ward, 2021). While some work has shown that state-mandated prescription drug monitoring programs have worked to reduce child welfare removals (Gihleb et al., 2020; Bullinger and Ward, 2021), other work suggests that recent opioid supply restrictions may have increased child maltreatment (Evans et al., 2022).

Our paper contributes to this nascent literature in two key ways. First, as we describe in detail below, previous estimates of the relationship between the severity of the crisis and child living arrangements potentially suffer from both omitted variables bias and reverse causality. Changes in the economy or in the strength of institutions could affect both drug abuse rates and child living arrangements, while a separation from one's children could increase either the motivation for using drugs or the number of opportunities to do so. To overcome these challenges, we use an instrumental variables strategy that exploits variation across states and over time in children's exposure to triplicate prescription laws. Alpert et al. (2022) show that because the producers of the most frequently abused prescription opioid marketed the drug less aggressively in states with these laws, the drug crisis was less severe. We use this variation in an instrumental variables framework to estimate the effect of the drug crisis on children's living arrangements.

Second, while the previous literature has focused on the relationship between the crisis and foster care admissions, we consider its effects on parental absence and on the likelihood of living in a household headed by someone other than a parent. These are important outcomes to consider, as many of the pathways we describe above would not necessarily result in a formal foster care placement. Indeed, our results reveal that moves into these more informal arrangements are much more common than moves into the foster care system. For example, at the end of 2015, there were 135,000 children in the foster care system with parental drug abuse listed as a contributing reason for removal; this is an upper bound on the number of removals due to the opioid crisis as it includes all drugs as well as children who would not have been removed on the basis of the drug abuse alone.4 Meanwhile, we estimate that in that same year, 1.5 million children aged 0-16 were living away from a mother or a father, and just over 270,000 were living away from both parents as a direct result of the crisis. About 941,000 children are in a household with a non-parent as the household head, with 43% of these living in a household headed by a grandparent. We also provide evidence that hundreds of thousands of children are living in households with fewer resources as a result of the crisis.

To conduct this analysis, we use data from the March Annual Social and Economic Supplement (ASEC) to the monthly Current Population Survey (CPS) from 1990 through 2015. We construct age-state-year cell rates of different living arrangements for children aged 0-16, and pair these with estimates of children's exposure to the crisis that are constructed from the National Center for Health Statistics Multiple Cause of Death (MCOD) data from 1973 to 2015. Our measure of exposure to the crisis is the cumulative drug-related death rate for likely parents over the child's life; this measure accounts for the fact that a transition to the child's current living arrangement could have occurred at any point in the child's life, and older children have had more exposure to the crisis than younger children. While our measure is based on drug deaths, the death of a parent is not the only way that a child could experience a change in their living arrangements. We view our measure as a proxy for how severe the crisis has been in their state over their lifetime. In OLS models with state, age, and year fixed-effects, we document a strong correlation between this measure and child living arrangements. Our 2SLS estimates suggest that the relationship is causal—OLS and 2SLS results are similar in magnitude and we fail to reject the null hypothesis that the 2SLS estimates are equal to their OLS counterparts for most outcomes. We show that greater exposure to the crisis increases the chance that a child's mother or father is absent from the household and it increases the likelihood that he or she lives in a household headed by a grandparent.

It is important to note that the variation we exploit corresponds to a particular counterfactual: a world without the drug crisis. While having a child live away from a parent who is using drugs could be a constrained optimal decision conditional on being affected by the drug crisis, this does not imply that the child is better off than she would have been in the absence of the crisis. In the penultimate section of the paper, we discuss research from across the social sciences which suggests that the changes in living arrangements that we document will have harmful long-term consequences for a great many of these children.