As Fathers and Felons: Explaining the Effects of Current and Recent Incarceration on Major Depression*

Abstract

Dramatic increases in the American imprisonment rate since the mid-1970s have important implications for the life chances of marginal men, including for their health. Although a large literature has considered the collateral consequences of incarceration on a variety of outcomes, studies concerned with health in particular have several limitations: most focus exclusively on physical health; those concerned with mental health only consider current incarceration or previous incarceration, but never both; none adjust for both observed and unobserved traits of prisoners; many fail to consider mechanisms; and virtually all neglect the role of family processes, thereby neglecting the social roles current and former prisoners inhabit. In this article, we use stress process theory to extend this research by considering the association between incarceration and major depression and some potential mechanisms explaining it. Results from the Fragile Families and Child Wellbeing Study (N = 3,107) show current and recent incarceration are substantially associated with the risk of major depression, suggesting both immediate and short-term implications. In addition, consistent with stress proliferation theory, the results show the well-known consequences of incarceration on socioeconomic status and family functioning partly explain these associations, suggesting the link between incarceration and mental health depends heavily on the consequences of incarceration on economic and social reintegration, not only the direct psychological consequences of confinement per se.

The American imprisonment rate has soared from approximately 100 per 100,000 individuals in the mid-1970s to approximately 500 per 100,000 individuals by the mid-2000s (Wakefield and Uggen 2010. Incarceration has broad implications for men’s subsequent life chances. A burgeoning literature considers the consequences of incarceration for the employment, family life, and civic engagement of formerly imprisoned men, almost always documenting negative consequences (Wakefield and Uggen 2010). Specifically, incarceration compromises labor market prospects (Pager 2003; Western 2006), destabilizes and diminishes the quality of romantic relationships (Braman 2004; Nurse 2002; Western 2006), and undermines participation in the political process (Uggen, Manza, and Thompson 2006). Recently, research has turned to the health consequences of incarceration, demonstrating associations between earlier incarceration and hypertension (Wang et al. 2009), functional limitations (Schnittker and John 2007), infectious and stress-related diseases (Massoglia 2008a), and poor self-rated health (Massoglia 2008b).

It is possible the consequences of incarceration are even stronger and more immediate for mental health than physical health, stemming from the “pains of imprisonment,” the psychological stresses associated with confinement (Haney 2006; Sykes 2007 [1958]). Life in the total institution of the prison may have lasting implications for psychological wellbeing, setting in motion a diminished coping capacity and chronic stress, and this stress may proliferate to other life domains both during and after release (Pearlin 1989; Pearlin, Aneshensel, and Leblanc 1997). Yet the consequences of incarceration on major depression may be different from those on physical health. For example, the immediate and short-term effects of incarceration may be more aligned in the case of mental illness than physical illness. In the case of physical illness, there is some evidence incarceration improves health while in prison, owing to some combination of better healthcare and a safer environment (Curtis 2011; Schnittker and John 2007; Spaulding et al. 2011). However, these apparent benefits are reversed sharply after release, increasing the risk of severe functional limitations (Schnittker and John 2007) and mortality (Spaulding et al. 2011). Although mental health care may also improve in prison (Wilper et al. 2009), better psychiatric treatment is perhaps insufficient to offset the conditions of confinement, separation, and regimentation. Indeed, relative to the association between incarceration and physical health, the association between incarceration and mental health may be driven to a larger degree by prison conditions instead of the social consequences of a prison sentence. In the case of physical illness, much evidence focuses on disadvantages that emerge after release, suggesting the negative effects of incarceration on health reflect the effects of economic hardship, family instability, discrimination, and stigma. Although the same may be true of mental health, much of the fairly small incarceration literature in this area focuses on confinement itself, following the early lead set by Goffman (1961) and Sykes (2007 [1958]).

In this article, we extend research on the incarceration-depression relationship (Schnittker, Massoglia, and Uggen 2012) by (1) simultaneously considering the consequences of current and recent incarceration for depression, (2) testing the indirect mechanisms through which incarceration compromises mental health, and (3) adjusting for fixed but unobserved traits of individuals, all of which represent substantial contributions to the literature on incarceration and mental health. Addressing these questions necessitates repeated measures of incarceration and depression, as well as a large sample of men experiencing incarceration, allowing us to evaluate the sensitivity of these relationships to a combination of observed and unobserved characteristics. It also requires repeated measures of the key mechanisms linking incarceration and depression, which are related to the labor market, romantic relationships, and parenting. Data from the Fragile Families and Child Wellbeing Study contain these features.

BACKGROUND

For a variety of reasons, major depression is an important addition to research on the consequences of incarceration and not merely another entry to research on its health consequences. Major depression is among the most common and severe psychiatric disorders in the United States (Kessler et al. 2003). Furthermore, major depression is chronic. Although its symptoms cycle over time and occasionally disappear completely, the initial onset of depression increases the risk of future episodes (Kendler, Thornton, and Gardner 2000). Depression is fundamentally psychological and somatic, but it has strong behavioral consequences. Indeed, major depression is a leading cause of disability and its influence on role impairment often exceeds that of common physical illnesses (Merikangas et al. 2007). Virtually all those suffering from major depression experience some resulting impairment (Kessler et al. 2003).

Recent research suggests that incarceration is associated with mood disorders including major depression (Schnittker, Massoglia, and Uggen 2012; also see Steadman et al. 2009), but it is not yet clear if current and former incarceration are differentially associated with depression and what might explain these associations. Understanding the potentially differential consequences of current and former incarceration is relevant to understanding what exactly is stressful about incarceration. Whereas the former studies often focus on the prison experience and the resulting psychological dysfunction, the latter focus on the diminished socioeconomic opportunities resulting from discrimination and the stigma of a prison record. Both processes are likely relevant for major depression among current and former inmates, but considering both processes simultaneously is important. For this task, stress process theory is particularly useful. Moreover, the situation of current and former inmates can help inform stress theory.

Stress Process Theory

Stress process theory suggests stressors emerge from the distinctive social contexts characterizing the lives of disadvantaged groups and that differential exposure to these stressors contributes to inequalities in physical and mental health (Pearlin 1989). Incarceration may be an important source of stress in this regard, given its concentration among minority, poorly educated men living in disadvantaged neighborhoods (Wakefield and Uggen 2010).

Moreover, incarceration may be an important source of stress proliferation. Stress proliferation is “the expansion or emergence of stressors within and beyond a situation whose stressfulness was initially more circumscribed” (Pearlin et al. 1997:223). Stress proliferation distinguishes primary and secondary stressors, while linking the two through a social process. According to this theory, a primary stressor (such as job loss) triggers the emergence of a series of secondary stressors (such as home foreclosure or divorce that result from job loss), which lead, in turn, to even more stressors over time (Pearlin 1989). Perhaps the best example of stress proliferation in the current literature is job loss, although incarceration likely has even stronger consequences than job loss because of the number of secondary stressors involved. For one, incarceration can be stigmatizing for individuals even long after release (Braman 2004; Goffman 1963). Furthermore, incarceration affects a variety of social roles at once, including men’s roles as employees, romantic partners, and fathers, and it does so both during incarceration and after release. Among other things, this idea suggests the pains of imprisonment may stem from the social consequences of incarceration—that is, from secondary stressors—as well as from the conditions of confinement itself. Even after release, stress proliferation is likely to occur and may perhaps grow. This point is often overlooked in research that focuses only on the primary stresses of incarceration itself, even though the primary stressors of incarceration are powerful.

Incarceration and Major Depression: Primary Stressors

The starting point of any discussion about the association between incarceration and depression is the experience of incarceration. A long line of research explores the psychological consequences of the incarceration experience, often under the more general concept of prisonization (Clemmer 1940; Goffman 1961; Sykes 2007 [1958]). Prisonization refers to how current inmates cope with their environments and highlights how most inmates do not develop psychiatric disorders (see Bonta and Gendreau 1990 for a skeptical view of the pains of imprisonment), but there are adverse consequences to even the most effective coping strategies, meaning prisonization is ultimately dysfunctional (Sykes 2007 [1958]; Zamble and Porpino 1988). Some argue the stress of incarceration emerges from the loss of liberty or from the isolation, confinement, and danger of the prison environment (Sykes 2007 [1958]). Under these circumstances, the most effective strategies for adjusting in the short term may be counterproductive in the long term, especially insofar as they undermine ordinary social interaction (Haney 2006). For example, many inmates adopt a “prison mask,” which involves suppressing weakness and emotional vulnerability in favor of an impassive, strong appearance (Toch and Adams 2002). Similarly, many inmates view others with distrust and suspicion, remaining vigilant to potential safety threats. These orientations allow inmates to adjust to the unique demands of prison life, where vulnerability is exploited and predation is common, but outside prison, where social reintegration is essential, they may be related to poor mental health.

Incarceration and Major Depression: Secondary Stressors

But consistent with stress proliferation theory, secondary stressors may play an even more important role in the link between incarceration and depression. Indeed, the stress of imprisonment reflects the loss of social roles as much as the conditions of confinement, so current and former inmates suffer in much the same way as those who simultaneously lose a job, a relationship, and a valuable role. Indeed, recent research reveals the multidimensional stress associated with incarceration, beginning with research on employment opportunities. A prison record harms the labor market prospects of former inmates. Men with a criminal record are less likely to receive job callbacks compared to their counterparts (Pager 2003) and earn less when employed (Western 2006). For former inmates, then, the consequences of incarceration on depression likely reflect at least two simultaneous influences: discrimination and job loss. Although partially attributable to former prisoners’ poor labor market prospects, incarceration is also independently associated with homelessness (Geller and Curtis 2011). Given that job loss (Burgard, Brand, and House 2007), discrimination (Kessler, Mickelson, and Williams 1999), and homelessness (Lee, Tyler, and Wright 2010) are all linked to depression, socioeconomic status likely plays an important role in the relationship between incarceration and depression.

Yet these economic factors likely are not the only secondary stressors. Former inmates have other social roles and obligations, and the impact of both current and recent incarceration on their capacity to be good romantic partners and fathers may be as strong as their impact on their likelihood of being well-paid employees. Most inmates are involved in romantic relationships at the time of their incarceration and are fathers (Mumola 2000). It is well established that incarceration substantially increases the risk of divorce and separation (Western 2006), which is linked to poor mental health (Williams 2003). But the spillover effects of incarceration may be even deeper than those implied by mere separation. Even when the relationship remains intact, ethnographic research suggests incarceration often diminishes relationship quality (Braman 2004; Nurse 2002; but see Comfort 2008), a well-known correlate of depression (Zlotnick et al. 2000).

The consequences of incarceration on romantic relationships are also vital for the psychological wellbeing of these men because they influence interactions with their children. Although some former inmates may be ineffective parents, evidence also suggests inmates express concern about their children and attempt to provide support. During their prison sentence, for example, many incarcerated fathers see a reunion with their children as something to look forward to and, upon release, many reflect on their prison sentence as an opportunity to become a better father (Nurse 2002). But, on average, the strength of their good intentions is overwhelmed by the reality of the numerous barriers they face. Even under the most generous visitation policies, currently incarcerated men have little opportunity to interact with their children (Braman 2004; Nurse 2002). And, after release, fathers are less likely to spend time with children than never incarcerated fathers (Swisher and Waller 2008) and the quality of father-child interaction diminishes (Nurse 2002), which may be associated with depression (Davis et al. 2009).

Selection versus Causation

Before considering which of these mechanisms are in play, we first must acknowledge that the incarceration-depression relationship may be driven by selection into incarceration rather than incarceration itself. For one, many psychiatric disorders predate incarceration (Schnittker et al. 2012) and, thus, many inmates and former inmates would experience depression regardless of whether they were incarcerated or not. Additionally, incarcerated men, compared to their non-incarcerated counterparts, experience difficulties in employment and family life prior to incarceration. To this end, before testing the various mechanisms linking current and recent incarceration and depression—the key contribution of this paper—we must ensure the starting relationship is robust by adjusting for both observed and unobserved differences between the men who eventually experience incarceration and those who do not. After establishing a baseline relationship, it is then possible to evaluate mechanisms.

DATA AND METHODS

Data

We use data from the Fragile Families and Child Wellbeing Study, a longitudinal survey of 4,898 mostly unmarried parents of children born in urban areas between 1998 and 2000 (Reichman et al. 2001). The sampling frame included hospitals in 20 U.S. cities with populations greater than 200,000. Initial interviews were conducted with mothers in hospitals shortly after the birth and with fathers at the same time or as soon as possible thereafter. Parents were then interviewed by telephone approximately one, three, and five years after the birth. Nearly four-fifths (78%) of fathers participated in the baseline interview and, of these, 69%, 67%, and 64% completed the one-, three-, and five-year surveys, respectively (Bendheim-Thoman Center for Research on Child Wellbeing 2008).

The Fragile Families data provide a unique opportunity. First, and perhaps most importantly, they include a large number of ever-incarcerated men. Because the sample over-represents unmarried parents, many men in the sample are minorities, do not have education beyond high school, and reside in areas of concentrated disadvantage (McLanahan 2009), all of which are correlated with incarceration (Wakefield and Uggen 2010). Additionally, the data allow us to examine immediate and short-term consequences of incarceration on depression. Finally, the data include a wealth of information about fathers, making it possible to adjust for pre-existing differences between fathers who have and have not experienced incarceration and to examine mechanisms through which incarceration may lead to depression.

Our analytic sample comprises 3,107 fathers. We made efforts to preserve as many respondents as possible. We first dropped the 1,738 observations in which the father did not participate in the five-year survey. We dropped an additional 13 observations missing information on depression at the five-year survey and an additional 39 observations missing data on incarceration. Few observations were missing data on covariates, and we used multiple imputation to preserve these observations, producing 20 data sets (Royston 2007). There are some differences between fathers in the baseline survey and fathers in the analytic sample, with fathers in the analytic sample less likely to be racial minorities (81%, compared to 78%), more likely to have education beyond high school (35%, compared to 32%), and more likely to be married to the focal child’s mother (29%, compared to 24%).

Measures

Depression.

Our main dependent variable, major depression at the five-year survey, comes from fathers’ responses to the Composite International Diagnostic Interview Short Form (CIDI-SF) Version 1.0 (Kessler et al. 1998; see also Kessler and Ustun 2004). Fathers were asked if, at some time during the past year, they had feelings of depression or were unable to enjoy normally pleasurable things. Those who reported at least one of these conditions most of the day, every day, for a two-week period were asked additional questions about the same period, and those who answered affirmatively to three or more of these additional questions are considered depressed. Although the CIDI-SF provides a reliable indicator of major depression for use in general surveys, it is limited by its dichotomous nature (Link 2002).

Incarceration.

Our two measures of primary stressors are current incarceration and recent incarceration. Fathers experienced current incarceration if they were in prison or jail at the five-year interview.² Fathers experienced recent incarceration if they were incarcerated at the three-year survey or between the three- and five-year surveys. In addition to these explanatory variables, we control for prior incarceration, an indicator the father was ever incarcerated before the three-year survey (including prior to the birth of the focal child), to adjust for selection into incarceration.¹ We rely only on fathers’ reports of current incarceration, as time differences between the mothers’ and fathers’ interviews could result in conflicting yet accurate reports (i.e., the father was incarcerated when the mother was interviewed but not when he was interviewed). For recent and prior incarceration, we rely on both mothers’ and fathers’ reports of fathers’ incarceration, and assume the father was incarcerated if at least one report is affirmative. Importantly, these three measures of incarceration are distinct but not mutually exclusive, and are consistent with recent research using these data (Geller et al. 2012; Wildeman, Schnittker, and Turney 2012). Indeed, of the fathers who experienced prior incarceration, 10% were currently incarcerated and 29% were recently incarcerated.

Despite their advantages, our incarceration measures are somewhat imprecise with respect to the timing, duration, and type of incarceration. With respect to timing, we can precisely measure the timing of current incarceration, as only fathers interviewed in prison or jail are considered currently incarcerated. The timing of recent incarceration is somewhat less precise, as we know only that fathers have been incarcerated during the past two years. Despite this limitation, given all instances of recent incarceration occurred prior to the measurement of our dependent variable, it is unlikely the measurement of recent incarceration introduces great imprecision into our estimation strategy. The timing of prior incarceration is much less precise, as we only know that the father had ever been incarcerated before the three-year interview. Given this imprecision, prior incarceration is not a key explanatory variable and is instead a control variable.

Additionally, our incarceration measures are limited because we cannot distinguish between the duration and type of incarceration. It is likely the effects of incarceration lasting one night differ from the effects of incarceration lasting 10 years, and it is possible the effects of being in prison differ from the effects of being in jail. Although our data do not allow us to address either limitation, we find it most plausible our measure of incarceration is driven mostly by short but not incredibly short jail stays. In these data, incarcerated men have been convicted for a range of crimes (Wildeman 2010:291) and few men are incarcerated at both the three- and five-year surveys. Thus, we are unlikely picking up a large number of prison sentences. Likewise, we are unlikely picking up many short jail stays—those under a week or two—because it would be difficult for survey researchers to obtain new contact information so quickly. Because jail time is far more common than prison time, our results apply to a broader group than would data mostly capturing prison incarceration.

Controls.

The multivariate analyses control for characteristics associated with both incarceration and depression. The following paternal characteristics are measured at baseline: race (white, Black, Hispanic, and other race), immigrant status, age, education (less than high school diploma, high school diploma or GED, some college, and college), number of children in the household, and self-rated health (1 = poor to 5 = excellent). We also control for familial history of major depression (dummy variable indicating one of the father’s biological parents experienced a two-week period of feeling depressed, down in the dumps, or blue), measured at the three-year survey.

We also control extensively for socioeconomic status and family functioning (measured at the three-year survey): employment, income-to-poverty ratio, homelessness, relationship status and quality, shared responsibility in parenting, and perceptions of self as a father. A dummy variable indicates the father worked in the prior week. Income-to-poverty ratio is the ratio of total household income to official poverty threshold established by the U.S. Census Bureau. Fathers are considered homeless if one of the following conditions are met: they reported living in temporary housing, a group shelter, or on the street at the time of the interview; they reported staying somewhere not intended for regular housing for at least one night in the past year; or they reported living with friends or family without paying rent at the time of the interview. Relationship status with the child’s mother is as follows: married, cohabiting, in a non-residential romantic relationship, and not romantically involved. Relationship quality is based on reports of his relationship with the mother (1 = poor to 5 = excellent). A dummy variable indicates the father is in a relationship with a new partner. Shared responsibility in parenting comprises the average of mothers’ responses about how often the father assisted with things such as looking after the child and running errands (1 = never to 4 = often). Finally, fathers were asked to rate how they feel about themselves as a father (1 = not a very good father to 4 = an excellent father).

In addition, we control for three paternal characteristics that may account for selection into incarceration: impulsivity, domestic violence, and drug or alcohol use. Impulsivity is measured with an abbreviated version of Dickman’s (1990) impulsivity scale (α = .84). Domestic violence is measured by mothers’ reports that the father hit, slapped, or kicked her at the one- or three-year survey. Drug or alcohol problem is measured by mothers’ and/or fathers’ reports, at the one- or three-year survey, that drugs or alcohol interfered with his work or made it difficult to get a job or get along with friends or family.

Mechanisms.

We examine two sets of secondary stressors, which we regard as potential mechanisms linking incarceration and depression, both of which change over time: socioeconomic status and family functioning. Socioeconomic status is represented by changes in employment, income-to-poverty ratio, and homelessness between the three- and five-year surveys. Family functioning is measured with a dummy variable indicating the father separated from the child’s mother between the three- and five-year surveys, as well as by changes in relationship quality with the child’s mother, shared responsibility in parenting, and perceptions of self as a father between the three- and five-year surveys. Although many of these mechanisms are correlated with each other, as one would expect if stress proliferation were operating, they are far from perfectly correlated and multicollinearity tests found no evidence the coefficients were estimated imprecisely when included simultaneously.

Sample Description

Table 1 presents descriptive statistics of all variables. Both depression and incarceration are common among fathers. About 17% of fathers at the three-year survey and 12% of fathers at the five-year survey reported depression. The prevalence of depression among fathers in this sample is greater than the prevalence of men in the general population (7.7%) (Kessler et al. 1993; also see Kessler et al. 2003). These discrepancies may be because the sample over-represents nonmarital births and, therefore, economically disadvantaged parents who may be more susceptible to episodes of depression (Kessler et al. 2003). Fully 41% of fathers were ever incarcerated, with 6% of fathers experiencing current incarceration, 13% experiencing recent incarceration, and 38% experiencing prior incarceration.

Table 1.

Means and Standard Deviations of Variables Used in Analyses

	Full Sample		Ever Incarcerated		Never Incarcerated
	Mean	SD	Mean	SD	Mean	SD
Major depression (y3, percentage)	16.8		22.5		12.9		***
Major depression (y5, percentage)	11.5		16.3		8.1		***
Current incarceration (y5, percentage)a	5.7		13.8		.0
Recent incarceration (y3, y5, percentage)a	13.3		32.2		.0
Prior incarceration (b, y1, y3, percentage)	38.0		92.1		.0
Race (b)
White (percentage)	20.8		11.2		27.5		***
Black (percentage)	48.6		61.4		39.6		***
Hispanic (percentage)	26.4		24.0		28.1		*
Other race (percentage)	4.2		3.4		4.8
Foreign-born (b, percentage)	15.8		7.2		21.9		***
Age (b)	28.12	(7.28)	26.38	(6.82)	29.34	(7.34)	***
Education (b)
Less than high school (percentage)	30.9		41.2		23.5		***
High school diploma or GED (percentage)	34.7		41.0		30.4		***
Postsecondary education (percentage)	34.4		17.8		46.1		***
Number of children in household (b)	1.00	(1.20)	1.06	(1.26)	.96	(1.15)	**
Self-rated health (b)	3.98	(.93)	3.92	(.97)	4.03	(.90)	**
Parent experienced depression (y3, percentage)	31.6		35.4		28.9		***
Impulsivity (y1)	2.06	(.94)	2.20	(.99)	1.97	(.89)	***
Domestic violence (y1, y3, percentage)	7.9		14.3		3.4		***
Drug or alcohol abuse (y1, y3, percentage)	16.4		28.1		8.2		***
Employed (y3, percentage)	79.4		67.5		87.7		***
Income-to-poverty ratio (y3)	2.75	(3.37)	2.03	(3.07)	3.24	(3.47)	***
Homeless (y3, percentage)	4.8		7.5		3.0		***
Relationship status with child’s mother (y3)
Married (percentage)	37.2		17.2		51.4		***
Cohabiting (percentage)	21.9		23.0		21.1
Nonresidential relationship (percentage)	6.1		8.1		4.6		***
Not in a relationship (percentage)	34.8		51.7		22.9		***
In a relationship with another partner (y3, percentage)	14.8		22.2		9.6		***
Relationship quality with child’s mother (y3)	3.56	(1.30)	3.18	(1.35)	3.82	(1.19)	***
Shared responsibility in parenting (y3)	2.85	(1.08)	2.48	(1.16)	3.11	(.93)	***
Perception of self as a father (y3)	3.15	(.85)	2.99	(.89)	3.26	(.79)	***
Change in employment (y5–y3)	.00	(.45)	.00	(.54)	.00	(.37)
Change in income-to-poverty ratio (y5–y3)	.11	(2.82)	.01	(2.80)	.18	(2.84)	*
Change in homelessness (y5–y3)	−.01	(.27)	−.01	(.34)	−.01	(.22)
New separation from child’s mother (y3, y5, percentage)	10.8		13.8		8.6		***
Change in relationship quality with child’s mother (y5–y3)	−.06	(1.22)	−.09	(1.36)	−.04	(1.10)
Change in shared responsibility in parenting (y5–y3)	−.13	(.86)	−.20	(.99)	−.07	(.75)	***
Change in perception of self as father (y5–y3)	−.05	(.80)	−.07	(.88)	−.05	(.74)
n	3,107		1,284		1,823

Note: Asterisks are for two-sided significance tests comparing ever-incarcerated fathers to never-incarcerated fathers. b = baseline survey; y1 = one-year survey; y3 = three-year survey; y5 = five-year survey.

^aCurrent incarceration includes fathers in prison or jail at the five-year survey. Recent incarceration includes fathers in prison or jail at the three-year survey or between the three- and five-year surveys. Prior incarceration includes fathers in prison or jail at any point before the three-year survey (including prior to the birth of their child).

^{^}p < .10.

^*p < .05.

^**p < .01.

^***p < .001.

These descriptive statistics also demonstrate substantial differences among ever- and never-incarcerated fathers. Importantly, ever-incarcerated fathers report more depression at the five-year survey (16%, compared to 8%). Ever-incarcerated fathers are more likely to be Black and less likely to be white, Hispanic, or foreign-born. They are younger and have lower educational attainment. At the three-year survey, ever-incarcerated fathers have lower economic wellbeing, are less likely to be married to the child’s mother, have lower quality relationships with the child’s mother, and have lower shared responsibility in parenting.

Analytic Strategy

Logistic regression models.

The multivariate analyses proceed in two parts. In the first stage, presented in Table 3, we use logistic regression models to estimate depression as a function of our primary stressor, incarceration.³ All models include the key explanatory variables, current and recent incarceration. In the first and all subsequent models, we control for prior incarceration. The second model adjusts for variables that precede current and recent incarceration: race, immigrant status, age, education, number of children in the household, self-rated health, depression in a parent, impulsivity, domestic violence, and drug or alcohol abuse, and depression measured at the three-year survey. Model 3 adds the following additional variables measured at the three-year survey (and, thus, also prior to incarceration): employment, income-to-poverty ratio, homelessness, relationship status with child’s mother, relationship quality with child’s mother, new partner, shared responsibility in parenting, and perceptions of self as a father.

Table 3.

Logistic Regression Models Estimating Depression as a Function of Incarceration

	Logistic Regression Models								Fixed-Effect Logistic Regression Model
	M1		M2		M3		M4		M5
Current incarceration (y5)	.70	(.15)***	.60	(.22)**	.56	(.21)**	.49	(.24)*	.77	(.32)*
Recent incarceration (y3, y5)	.61	(.14)***	.44	(.16)**	.49	(.18)**	.27	(.22)	.35	(.20)^
Prior incarceration (y1, y3)	.32	(.11)**	.03	(.14)	−.05	(.14)
Race (b)
White (reference)
Black			−.14	(.20)	−.29	(.21)	−.31	(.34)
Hispanic			.01	(.19)	−.11	(.21)	−.16	(.32)
Other race			−.06	(.31)	−.24	(.30)	−.02	(.51)
Foreign-born (b)			−.37	(.17)*	−.26	(.01)	−.75	(.02)
Age (b)			−.01	(.01)	.00	(.18)	−.02	(.54)	−.32	(.05)***
Education (b)
Less than high school (reference)
High school diploma or GED			.06	(.13)	.03	(.13)	.05	(.24)	−.31	(.95)
Postsecondary education			−.11	(.14)	−.09	(.14)	−.21	(.24)	−.91	(.78)
Number of children in household (b)			.03	(.05)	.05	(.05)	.12	(.08)	−.18	(.07)*
Self-rated health (b)			−.16	(.08)*	−.13	(.07)^	−.06	(.09)	−.38	(.09)***
Parent experienced depression (y3)			.72	(.15)***	.71	(.15)***	.86	(.22)***
Impulsivity (y1)			.13	(.08)	.11	(.08)	.16	(.11)
Domestic violence (y1, y3)			.30	(.18)	.06	(.20)	.22	(.22)
Drug or alcohol abuse (y1, y3)			.31	(.21)	.23	(.21)	.12	(.25)
Depression (y3)			1.05	(.16)***	.97	(.15)***	.58	(.19)**
Employed (y3)					−.11	(.16)	−.04	(.22)
Income-to-poverty ratio (y3)					.06	(.02)	.02	(.03)
Homeless (y3)					.23	(.22)**	.63	(.31)*
Relationship status with child’s mother (y3)
Married (reference)
Cohabiting					.08	(.17)	.29	(.38)
Nonresidential relationship					−.38	(.31)	.27	(.49)
Not in a relationship					.10	(.20)	.45	(.47)
In a relationship with another partner (y3)					.00	(.18)	−.08	(.30)
Relationship quality with child’s mother (y3)					−.22	(.06)***	−.13	(.07)^
Shared responsibility in parenting (y3)					.03	(.08)	.05	(.10)
Perception of self as a father (y3)					−.07	(.08)	−.06	(.11)
Constant	−2.70		−2.50		−1.72		−2.24
Pseudo R-squared	.05		.13		.15		.14
n	3,107		3,107		3,107		1,181		620

Note: All models include city fixed-effects and use robust standard errors. b = baseline survey; y1 = one-year survey; y3 = three-year survey; y5 = five-year survey.

^{^}p < .10.

^*p < .05.

^**p < .01.

^***p < .001.

We then take two steps to diminish unobserved heterogeneity. First, in Model 4, we consider the full logistic regression model for a subsample of fathers at risk of incarceration, those who have experienced prior incarceration. Although this reduces the threat of unobserved heterogeneity, there are trade-offs to limiting the sample entirely to previously incarcerated men. Estimation within this sample may eliminate some unobserved heterogeneity and, thus, produce smaller coefficients than estimation within the full sample, but, to the extent that some coping or adaptation occurs during second or higher order incarcerations, it may also produce estimates that are smaller than true estimates. In effect, by limiting the sample to previously incarcerated men, we are estimating the association between an additional incarceration and depression, which may be sharply diminished in effect from the initial incarceration. As little research examines the relationship between incarceration and depression, it is difficult to adjudicate the relative magnitude of these influences, but readers should recognize this model is not necessarily a better approximation of the average association between incarceration and depression and, indeed, there are reasons to expect it is a lower bound. Second, in Model 5, we estimate a fixed-effects model considering the association between changes in incarceration and depression between the three- and five-year surveys. This model provides substantial leverage with respect to unobserved heterogeneity. Taken together, these assorted strategies provide a comprehensive portrait of the incarceration-depression relationship.

Logistic regression models, with mechanisms.

The second analytic stage is presented in Table 4. Here we use logistic regression models to examine how secondary stressors, changes in socioeconomic status and family functioning, explain the association between incarceration and depression. In these analyses, we extend Model 3 from Table 3. Model 1 includes changes in socioeconomic status, Model 2 includes changes in family functioning, and Model 3 includes both sets of models. In Model 4, we again restrict the sample to fathers who experienced prior incarceration. Because observations were drawn from 20 cities, we use clustered standard errors and include city fixed-effects in all logistic regression models. We present coefficients and standard errors in Tables 3 and 4 but, to better understand the magnitude of effects, discuss results in terms of coefficients and odds ratios.

Table 4.

Logistic Regression Models Estimating Depression as a Function of Incarceration, with Mechanisms

	M1		M2		M3		M4
Current incarceration (y5)	.39	(.21)	.47	(.20)*	.32	(.22)	.41	(.23)
Recent incarceration (y3, y5)	.44	(.19)*	.35	(.19)	.33	(.19)	.19	(.27)
Change in employment (y5–y3)	−.55	(.16)**			−.50	(.17)**	−.13	(.20)
Change in income-to-poverty ratio (y5–y3)	−.04	(.03)			−.04	(.03)	−.13	(.06)*
Change in homelessness (y5–y3)	.74	(.26)**			.67	(.25)**	1.31	(.38)**
New separation from child’s mother (y3, y5)			.29	(.20)	.32	(.22)	.37	(.39)
Change in relationship quality with child’s mother (y5–y3)			−.35	(.08)***	−.34	(.08)***	−.21	(.11)*
Change in shared responsibility in parenting (y5–y3)			−.03	(.08)	−.02	(.08)	.07	(.13)
Change in perception of self as father (y5–y3)			−.14	(.09)	−.13	(.09)	−.18	(.12)
Constant	−1.48		−1.17		−.97		−1.98
Pseudo R-squared	.16		.17		.18		.18
n	3,107		3,107		3,107		1,181

Note: All models include all covariates from Model 3 of Table 3. All models include city fixed-effects and use robust standard errors. y3 = three-year survey; y5 = five-year survey.

^*p < .05.

^**p < .01.

^***p < .001.

RESULTS

Depression and Secondary Stressors, by Incarceration

In Table 2, we present selected descriptive statistics for depression and our potential secondary stressors, by incarceration. About 25% of currently incarcerated fathers report depression at the five-year survey, compared to 11% of fathers not currently incarcerated. Similarly, 21% of recently incarcerated fathers and 10% of not recently incarcerated fathers report depression. Not surprisingly, currently and recently incarcerated fathers are more likely than their counterparts to report a change in employment status. Similarly, currently and recently incarcerated fathers, compared to their counterparts, are more likely to report a new separation from the child’s mother and a change in relationship quality.

Table 2.

Means of Key Variables, by Current and Recent Incarceration

	Current Incarceration		Recent Incarceration
	Yes	No	Yes	No
Depression (y5, percentage)	25.4	10.6***	21.3	9.9***
Change in employment (y5–y3)	−.21	.01***	.05	−.01**
Change in income-to-poverty ratio (y5–y3)	−.20	.13^	−.06	.14
Change in homelessness (y5–y3)	−.04	−.01	−.01	−.01
New separation from child’s mother (y3, y5, percentage)	17.5	10.4**	21.6	9.1***
Change in relationship quality with child’s mother (y5–y3)	−.20	−.05**	−.23	−.03***
Change in shared responsibility in parenting (y5–y3)	−.33	−.11**	−.30	−.10***
Change in perception of self as father (y5–y3)	−.13	−.05	−.06	−.05
n	177	2,930	413	2,694

Note: Asterisks are for two-sided significance tests that compare currently incarcerated fathers to not currently incarcerated fathers and recently incarcerated fathers to not recently incarcerated fathers. y3 = three-year survey; y5 = five-year survey.

^{^}p < .10.

^*p < .05.

^**p < .01.

^***p < .001.

Estimating the Association between Incarceration and Depression

Logistic regression models.

We turn first to logistic regression models estimating depression as a function of our primary stressor, incarceration. According to Model 1 of Table 3, fathers experiencing current or recent incarceration, controlling only for prior incarceration, are more likely to report depression than those not currently or recently incarcerated. Currently incarcerated fathers have about twice the odds (e^0.699=2.01) of reporting depression than their non-incarcerated counterparts, adjusting for prior incarceration (p<.001), a magnitude similar to the general association between stress and major depression (Kessler et al. 1999). Recently incarcerated fathers have 1.84 times the odds (e^0.611) of reporting depression (p<.001). Fathers who experienced prior incarceration are also more likely to report depression, although the magnitude and statistical significance of the coefficient is smaller than that of current or recent incarceration (b=0.324, OR=1.38, p<.01).

Both current and recent incarceration remain associated with depression after adjusting for observed individual-level characteristics. Adjusting for paternal demographic characteristics, family history of depression, and additional characteristics related to social selection diminishes the coefficients for current and recent incarceration (Model 2). These coefficients are further diminished after adjusting for socioeconomic status and family functioning (Model 3). This final model shows currently incarcerated fathers have 1.76 times the odds (e^0.563) of reporting depression (p<.01) and recently incarcerated fathers have 1.63 times the odds (e^0.490) of reporting depression (p<.01) than their counterparts. The current incarceration and recent incarceration coefficients are not statistically different from one another.⁴

In Model 4, we limit the sample to fathers who experienced prior incarceration. The incarceration coefficient in this model is similar to the coefficient estimated using the full sample. Net of a wide array of individual-level characteristics, currently incarcerated fathers have 1.64 times the odds (e^0.494) of reporting depression as their counterparts (p<.05). Recently incarcerated fathers are not more likely to report depression, providing some evidence that first-time incarceration has an especially strong relationship with depression. Finally, in Model 5, using fixed-effects, the coefficient remains statistically significant, despite the relatively few individuals who report changes in both incarceration and depression between the three- and five-year surveys. According to this model, changes in current incarceration (b=0.770, OR=2.16, p<.05) and recent incarceration (b=0.352, OR=1.42, p<.10) are associated with changes in depression.

Explaining the Association between Incarceration and Depression

The above models establish that current and recent incarceration are associated with depression among fathers. These analyses, however, have not considered the secondary stressors that may explain the relationship between incarceration and depression, which we consider in Table 4. In Model 1, we adjust for changes in socioeconomic status between the three- and five-year surveys. We include all three indicators of socioeconomic status simultaneously in the model, as a chi-square test revealed joint significance (F=16.82, p<.001). Taking into account socioeconomic status reduces the magnitude (by 31% from Model 3 in Table 3) and statistical significance of current incarceration, suggesting economic hardship may be a secondary stressor. Among the socioeconomic variables, change in employment status, not income-to-poverty ratio or homelessness, most strongly diminishes the current incarceration coefficient. In many ways, these results reflect the realities of incarceration, as these fathers often lose their job but not their residence. Though economic factors explain a moderate proportion of the current incarceration coefficient, these factors explain only 11% of the recent incarceration coefficient, suggesting other factors may be at play.

In Model 2, we adjust for four indicators of family functioning. As in the prior model, we test the joint significance of these variables (F=51.68, p<.001). The current incarceration coefficient is reduced by 17%, and the recent incarceration coefficient is reduced by 28% and to statistical insignificance. Experiencing a recent separation from the child’s mother independently explains 23% of the recent incarceration coefficient, though relationship quality and shared responsibility in parenting also explain a moderate portion of this coefficient (18% and 21%, respectively).⁵

Model 3 considers the contributions of socioeconomic status and family functioning, and shows the current and recent incarceration coefficients are reduced by 44% and 32%, respectively, both to statistical insignificance. Thus, after adjusting for changes in socioeconomic status and family functioning, currently and recently incarcerated fathers are not significantly more likely than their counterparts to report depression. Similar to Table 3, we also present results for fathers who experienced prior incarceration (Model 4), which provides results consistent with those for the full sample.

DISCUSSION

In this paper, we use data from the Fragile Families and Child Wellbeing Study to explore the association between incarceration and depression. In addition to exploring both current and recent incarceration, a strength of our study, we also adjust for observed and unobserved influences, consider a variety of mechanisms cutting across different domains, and focus on family processes to a much greater extent than previous theory and research. We find incarceration has both immediate and enduring consequences for major depression among fathers. These associations persist in fixed-effect logistic regression models.

We further find evidence indicating these results reflect stress proliferation. Incarceration represents a fundamental shift in the life course, leading to economic insecurity and labor market detachment (Pager 2003; Western 2006), disruptions and tensions in romantic relationships (Western 2006), and strained relationships with children (Swisher and Waller 2008), both when behind bars and after release. No one mechanism is capable of explaining the entire association. Furthermore, secondary stressors may ultimately be nearly as important as primary stressors in the sense that what happens as a result of incarceration is equally important as the incarceration experience. The conditions of incarceration itself—including confinement, regimentation, and violence—may be directly linked to depression (Sykes 2007 [1958]), and we cannot discount the importance of prison experiences for depression. Indeed, prison experiences (or other indicators we cannot measure with our data, such as stigma) may account for the remaining associations between incarceration and depression. For all these stressors, stigma likely serves as an important underlying process linking primary and secondary stressors. Stigma underlies the discrimination former inmates experience in the labor market, of course, but the effects of stigma on depression are probably not limited to those domains where other people are being explicitly prejudicial. As in the case of those with psychiatric disorders, for example, former inmates might behave in ways that increase depression even if they forestall discrimination, as when they isolate themselves from others, associate only with other former inmates, or try to hide their status (Link et al. 1989). In this way, the expectation of rejection—of which former inmates are well aware—can be as consequential as actual rejection.

Our results highlight the centrality of considering both primary and secondary stressors in understanding the stress process associated with incarceration, but also highlight distinctions among the secondary stressors and the importance of considering a variety of different types of secondary stressors. For instance, socioeconomic factors explain a large fraction of the association between current incarceration and depression, which is consistent with research highlighting the centrality of incarceration’s effects on economic wellbeing (Western 2006). At least with respect to depression, the role impairments associated with incarceration may be nearly as relevant as the stress of imprisonment itself. Socioeconomic status, however, does little to explain the effects of recent incarceration, which is explained better by romantic relationships and parenting. In both cases, our results highlight the importance of understanding secondary stressors. Inmates and former inmates are embedded in social relationships with their current romantic partners, mothers of their children, and their children, and these relationships are critical to understanding the effects of incarceration. Incarceration is strongly associated with major depression, but this association stems in part from role-related stressors that can affect anyone and speak to the stress process more generally.

Limitations

Several limitations exist. First, the Fragile Families sample is limited to fathers with young children. The consequences of incarceration for depression may differ for men without young children. Given that many prisoners have young children (Mumola 2000), understanding the collateral consequences of incarceration for inmates with children should closely approximate those found for the average inmate, but future research should consider if incarceration differentially affects the mental health of men with and without children. The association between incarceration and depression among childless men may be smaller insofar as the scope of their obligations is narrower. Additionally, attrition among fathers means our analytic sample is selective of more advantaged fathers, and it is possible that our estimates are biased as a result of such attrition.

Additionally, although we distinguish between current and recent incarceration, incarceration experiences are sufficiently complex that we cannot disentangle them all. We do not, for example, have reliable measures of incarceration duration, though stress proliferation theory suggests the effects of a primary stressor become more pronounced as length of exposure increases (Pearlin 1989). Similarly, we cannot distinguish between prisons and jails, although presumably the stress associated with these environments differs. If the stress of jails is less than the stress of prisons, our results understate the impact of incarceration insofar as the two groups are both coded as incarcerated. This does not render the distinction irrelevant, but does suggest that understanding the consequences of incarceration requires a better understanding of the conditions surrounding jail and prison incarceration, and we see considering these differences to be one of, if not the, key tasks for future research in this area. Other related conditions presumably matter as well. We do not have information regarding, for example, experiences surrounding the arrest, other interactions with the criminal justice system, interactions with other inmates, and the frequency of visitation from family members. Using indicators of current and recent incarceration, we have established a positive relationship between incarceration and depression, but in order to address these effects it is necessary to discern the best targets for intervention or policy, regarding, for example, prosecution, prison administration, or reintegration services. Considering the degree to which the conditions of confinement interact with individual-level and familial-level characteristics to explain stress process is another crucial goal for future research. On a related note, depression is only one measure of psychological distress and it is possible incarceration exerts a different influence on other measures of distress such as physical violence (Umberson, Williams, and Anderson 2002). Exploring outcomes of this sort is important, especially given the relationship between these outcomes and criminal behavior.

Conclusion

It has become increasingly clear that incarceration has broad collateral consequences for former inmates. Our study highlights strong connections between these collateral consequences and depression, and shows the mechanisms that produce them. Incarceration experiences per se appear not to be the sole drivers of these effects. Indeed, we find the primary stress of incarceration is partially a reflection of how incarceration undermines employment and families, that is, the social life of current and former inmates. For those interested in the wellbeing of current and former inmates, this finding reminds us that it is important to think of inmates not only as felons, but as fathers. By the same token, it is important not only to think of how inmates adjust to prison life, but how these adjustments relate to their lives outside of prison. The effects of incarceration depend heavily on the challenges awaiting inmates upon release.