Abstract
Theories of stress and strain, which emphasize the concentration of social stressors among vulnerable groups, suggest that police contact—the most common type of criminal justice contact—can have deleterious health consequences. Research documents a relationship between police contact and adverse health, but less is known about the mental health consequences of police stops among adolescents. I examined this with data from the Fragile Families and Child Wellbeing Study (N = 3,437), a longitudinal survey of individuals born around the turn of the 21st century and followed over a 15-year period. First, personal police contact and vicarious but not personal police contact (both compared to no police contact) are positively associated with depressive symptoms, net of characteristics associated with police contact (including prior mental health, delinquency, and impulsivity). Second, more intrusive police contact (such as stops that include frisks or searches) is positively associated with depressive symptoms. Third, the association between police contact and depressive symptoms is concentrated among girls and Blacks. Given the concentration of police contact among already vulnerable adolescents living in highly surveilled and disadvantaged neighborhoods, those same adolescents at greatest risk of health impairments, police contact may exacerbate population health disparities.
Introduction
The criminal justice system—with its attendant consequences for individuals and families—is a critical mechanism of health disparities in the United States (Kirk and Wakefield 2018; Massoglia and Pridemore 2015; Turney and Wakefield 2019; Wildeman, Goldman, and Turney 2018). Its precipitous growth over the past half century means that a considerable number of individuals have some exposure to the criminal justice system. Police contact is the most common form of exposure, with an estimated 31 million individuals experiencing a stop or another form of involuntary contact annually (Langton and Durose 2016). Police contact is common in adolescence, a pivotal life course period that sets in motion adult life trajectories (Crosnoe and Johnson 2011). Research documents that, among 15-year-olds born in urban areas, 27 percent of them report being stopped by the police and an additional 53 percent report witnessing a police stop or knowing someone stopped by the police (Geller and Fagan 2019).
The stress process perspective, which emphasizes the concentration of social stressors among vulnerable groups such as race/ethnic minorities and those living in poverty, suggests that police contact is an adverse experience that can have deleterious consequences for health (Geller 2018; Pearlin 1989; Pearlin et al. 1981; Turney and Wakefield 2019). The stress process perspective also highlights how stressors can proliferate, from the individual experiencing the stressor to those connected to them, indicating that vicarious police contact can also have deleterious consequences for health (Pearlin, Aneshensel, and LeBlanc 1997). General strain theory, similarly highlighting the vulnerability of particular groups, also suggests that personal or vicarious strains can lead to negative emotions such as depressive symptoms (Agnew 2006). Indeed, a burgeoning literature documents a relationship between police contact and mental health impairments among adolescents, many of whom are experiencing criminal justice contact for the first time (Del Toro et al. 2019; Rios 2011; Stuart 2016). There are opportunities to build upon this research by considering the consequences of both personal and vicarious police contact for adolescent mental health (McFarland, Geller, and McFarland 2019), by examining the relationship between intrusive police contact and mental health among adolescents (Jackson et al. 2019), and by documenting variation in responses to personal and vicarious police contact across gender, race/ethnicity, and poverty status (Baćak and Nowotny 2020).
In this article, I use data from the Fragile Families and Child Wellbeing Study, a longitudinal survey of children born in urban areas around the turn of the 21st century, to examine the relationship between adolescent police contact and depressive symptoms, a distinct indicator of mental health that may facilitate criminal activity (Agnew 2006) and is linked to educational, economic, and health outcomes in adulthood (Fletcher 2010, 2013; Meadows, Brown, and Elder 2006). First, I examine the consequences of personal and vicarious police contact for depressive symptoms, adjusting for a large number of characteristics that may facilitate exposure to a personal or vicarious police stop (including prior mental health, delinquency, and impulsivity) and adjusting for other forms of criminal justice contact such as arrest and incarceration. Second, I examine the relationship between intrusive police contact (such as stops that include frisks or searches) and depressive symptoms. Third, I investigate heterogeneity in the relationship between police contact and depressive symptoms by gender, race/ethnicity, and poverty status. These analyses build on the burgeoning literature that describes adversities experienced by adolescents exposed to police contact by considering both personal and vicarious contact for mental health, by considering the intrusiveness of police contact, and by considering heterogeneity in the consequences of police contact.
Background
Linking Police Contact and Depressive Symptoms
Consequences of Personal Police Contact
The stress process perspective, which highlights the deleterious consequences of stressors for mental health, provides a theoretical lens for understanding how personal police contact may increase depressive symptoms among adolescents (Pearlin 1989; Pearlin et al. 1981). Similarly, general strain theory posits that strains, such as personal police contact, can lead to negative emotions including depressive symptoms (and, further, that these negative emotions can trigger criminal activity) (Agnew 2006). Indeed, personal police contact, a usually involuntary form of legal social control, can be a stressor or a strain (Geller 2018). Given the often traumatic and upsetting nature of this stressor, personal police contact may facilitate immediate depressive symptoms in adolescents. Police contact can involve invasive searches of one’s body or belongings, physical violence, and emotional degradation (Epp, Maynard-Moody, and Haider-Markel 2014; Rios 2011; Shedd 2015; Stuart 2016). Indeed, among urban adolescents, 34 percent of those stopped by the police report being searched and 38 percent report being frisked during their most memorable personal stop (Jackson et al. 2019).
The stressor or strain of personal police contact may also facilitate depressive symptoms among adolescents in ways that are less immediate. Feelings of sadness or anger may stem from police contact, especially if the adolescent perceives the contact to be unfair or discriminatory, and these feelings may linger (Stuart 2016). The emotionally exhausting nature of the experience may also linger. Adolescents may ruminate over the experience, may spend time deliberating about disclosing the stop, and may anticipate and experience negative sanctions from parents, peers, or teachers, all of which could increase depressive symptoms. Depressive symptoms may also emerge if adolescents experience fear about future police contact, the intrusiveness of future police contact, and the potential perceived consequences of future police contact such as stigma or other criminal justice sanctions.
Consistent with the stress process perspective and general strain theory, a burgeoning literature indicates that police contact is associated with mental health impairments, including greater depressive symptoms (Baćak and Nowotny 2020), anxiety (Geller 2018; Geller et al. 2014), posttraumatic stress (Geller 2018; Geller et al. 2014), shorter telomere length (McFarland et al. 2018), emotional distress (Jackson et al. 2019), and psychological stress (Del Toro et al. 2019). Personal police contact is also associated with lower self-rated health among adolescents (McFarland et al. 2019). Furthermore, the intrusiveness of police contact is associated with greater emotional distress and posttraumatic stress among adolescents (Jackson et al. 2019). Other research shows that exposure to police victimization—defined as physical violence, sexual violence, psychological violence, and neglect—is associated with psychological distress and depression, suicide attempts, and subthreshold psychotic experiences (DeVlyder, Cogburn et al. 2017; DeVlyder, Frey et al. 2017; DeVylder, Oh et al. 2017; Sewell and Jefferson 2016; Sewell, Jefferson, and Lee 2016).
Consequences of Vicarious Police Contact
The stress process perspective, as well as general strain theory, also provides a theoretical lens for understanding how vicarious police contact may increase depressive symptoms in adolescence. Stressors and strains are contagious across individuals, with those experienced by one individual proliferating to impair mental health among those connected to them (Barr, Simons, Simons, Beach, and Philibert 2018; Pearlin et al. 1997). This suggests that vicarious police contact, via witnessing a police stop or knowing someone who experienced a police stop, may increase depressive symptoms among adolescents. These vicarious experiences, similar to personal experiences, may be traumatic. They may also facilitate immediate or lingering feelings of sadness, fear, or anger. Adolescents may reflect on the vicarious contact, particularly if it is perceived to be discriminatory, and may also envision their own police contact and its consequences. All of this may enable depressive symptoms. General strain theory suggests that vicariously experienced strains are generally less consequential than personally experienced strains (Agnew 2006).
Existing research shows that vicarious police contact is associated with health impairments. One recent study finds that vicarious police contact—defined as witnessing a police stop or knowing someone who experienced a police stop—is associated with lower self-rated health among adolescents, particularly when the police contact was perceived to be procedurally unjust (McFarland et al. 2019). Other research documents that neighborhood-level police stops, another way individuals can vicariously experience police contact, is associated with psychological distress among men (Sewell et al. 2016; also see McFarland et al. 2018). Neighborhood-level police stops are also associated with physical health outcomes including fair/poor health, diabetes, asthma, and overweight/obese (Sewell and Jefferson 2016).
Selection into Personal and Vicarious Police Contact
The stress process perspective and general strain theory, as well as empirical findings from prior research, suggest that both personal and vicarious police contact will be positively associated with depressive symptoms. The stress process perspective, though, also highlights the interconnected nature of stressors (Pearlin 1989), and accordingly, adolescents who experience police contact are more likely than their counterparts to experience additional stressors that make them vulnerable to mental health impairments. There are differences in the demographic, socioeconomic, and family characteristics of adolescents who do and do not experience police contact (Geller 2018). Therefore, though police contact may increase depressive symptoms among adolescents, it is also possible these associations result from characteristics that increase the probability of experiencing police contact. For example, engagement in delinquent behavior may lead to both police contact and depressive symptoms (Wiesner and Kim 2006). Similarly, parental contact with the criminal justice system is linked to both police contact and depressive symptoms (Foster and Hagan 2013). Accounting for these and other potentially spurious adolescent, family, and neighborhood factors is important to isolate the relationship between adolescent police contact and depressive symptoms.
Heterogeneity in the Relationship between Police Contact and Depressive Symptoms
Both exposure to police contact and reactions to police contact may not be uniform across social groups. Instead, the stress process perspective posits that exposure to and reactions to stressors are contingent on one’s social position, with disadvantaged social groups most likely to experience stressors such as police contact (Pearlin 1989). General strain theory also posits that individual characteristics may condition the relationship between strains and negative emotions (Agnew 2006). The consequences of personal and vicarious police contact for depressive symptoms among adolescents may vary by gender, race/ethnicity, and poverty.
First, the relationship between adolescent police contact and depressive symptoms may vary by adolescent gender, as males and females may manifest distress differently (Aneshensel, Rutter, and Lachenbruch 1991; Longest and Thoits 2012). Police contact is more common among boys than girls (Geller 2018), but these differences in exposure to police contact may or may not translate into differences in the consequences of police contact. On the one hand, boys may be more susceptible than girls to the deleterious consequences of police contact. Feelings of sadness, fear, or anger stemming from police contact may be especially common among boys, who may especially anticipate future police contact and other forms of criminal justice contact such as arrest and incarceration (Pearlin and Bierman 2013). On the other hand, girls may be more susceptible than boys to the deleterious consequences of police contact. The less frequent exposure to police contact may mean that contact is especially troubling when it occurs among girls, as they may experience more stigma and accompanying distress. Existing research provides evidence that the association between police contact and depressive symptoms may be concentrated among men, as neighborhood-level police frisks in New York City are associated with greater psychological distress among men but not women (Sewell et al. 2016).
Second, the relationship between adolescent police contact and depressive symptoms may vary by race/ethnicity. Adolescents of color have higher rates of police contact than their White counterparts (Geller 2018), though it remains less clear if police contact will be more or less consequential for depressive symptoms among these adolescents. On the one hand, the less frequent—and potentially less anticipated—exposure to police contact among Whites may mean that, when it does occur, its unexpected nature may especially engender feelings of distress. On the other hand, adolescents of color are more likely than White adolescents to experience additional adversities such as family instability, discrimination, parental incarceration, and residence in a disadvantaged neighborhood; the compounding nature of stressors may be especially consequential. There are other reasons to expect the relationship between police contact and depressive symptoms to be magnified among adolescents of color. Qualitative research, for example, shows that people of color often experience police contact as dehumanizing, which may engender depressive symptoms (Rios 2011). Other research shows that a history of unfair treatment by the criminal justice system magnifies the deleterious effects of police contact among people of color (Slocum and Wiley 2018); among adolescents, this history of exposure to the criminal justice system may include their own exposure but also the exposure of their parents or siblings. Further, as invasive police contact is more common among people of color than their White counterparts, this may be one reason to expect the relationship between police contact and depressive symptoms to be concentrated among Black and Hispanic adolescents (Kahn et al. 2016). Recent empirical evidence finds that the deleterious health consequences of both personal and vicarious police contact are largest among Blacks (Baćak and Nowotny 2020; McFarland et al. 2019; Sewell and Jefferson 2016).
Third, the relationship between adolescent police contact and depressive symptoms may vary by poverty status. It remains an open question, however, as to whether the relationship between police contact and depressive symptoms will be more consequential for adolescents living in households at or below the poverty line or adolescents living in households above the poverty line. The compounded nature of adversities may make police contact especially consequential for adolescents living in poverty (Ferraro and Kelley-Moore 2003). Alternatively, the less anticipated nature of police contact among those not living in poverty might be especially detrimental (Wheaton 1982).
Methods
Data
To estimate the relationship between police contact and depressive symptoms in adolescence, I used data from the Fragile Families and Child Wellbeing Study, a population-based cohort of 4,898 children born to mostly unmarried parents around the turn of the 21st century (Reichman et al. 2001). Researchers sampled births in hospitals across 20 U.S. cities, interviewing children’s mothers and fathers after the birth. Parents were interviewed an additional five times (when their children were ages 1, 3, 5, 9, and 15) and children were interviewed at ages 9 and 15 (with the most recent interviews occurring between 2014 and 2017). These data provide an excellent opportunity to understand the association between police contact and depressive symptoms among adolescents, as the latest wave was collected during an era of proactive policing (Kubrin et al. 2010).
The analytic sample includes 3,437 adolescents. I first removed the 1,453 observations in which the adolescent did not participate in the 15-year survey and then removed the additional 7 observations missing data on the dependent variable, depressive symptoms. I examined demographic differences between the analytic sample and the baseline sample, finding few statistically significant observed differences. Adolescents in the analytic sample, compared to adolescents in the baseline sample, are significantly more likely to have Black mothers (50 percent compared to 48 percent, p < .05), less likely to have Hispanic mothers (25 percent compared to 27 percent, p < .01), and less likely to have foreign-born mothers (13 percent compared to 17 percent, p < .001). They are more likely to have mothers with at least a high school education (68 percent compared to 65 percent, p < .01). Other characteristics—including parents’ baseline relationship status, poverty, and neighborhood disadvantage—are similar between those in the analytic and baseline samples.
Missing data are uncommon among variables reported by mothers and adolescents (for example, <1 percent of observations are missing values on adolescent police contact) but more common among variables reported by fathers (as survey attrition was more common among fathers). I preserved missing covariate values with multiple imputation, pooling results across 20 imputed data sets (Allison 2001).
Measures
Depressive Symptoms
The measure of depressive symptoms, ascertained at the 15-year survey, is an average of responses to the following five statements about feelings in the past 4 weeks (1 = strongly disagree to 4 = strongly agree): (1) I feel I cannot shake off the blues, even with help from my family and friends; (2) I feel sad; (3) I feel happy (reverse coded); (4) I feel life is not worth living; and (5) I feel depressed. These items are drawn from a modified version of the Center for Epidemiologic Studies Depression Scale (CES-D) (Radloff 1977; α = .76).
Police Contact
Police contact is measured with mutually exclusive binary variables reported by adolescents at the 15-year survey: personal police contact, vicarious but not personal police contact, and no police contact (reference category). Personal police contact is measured by adolescent reports of ever being stopped by the police on the street, at school, in a car, or some other place. Vicarious police contact is measured by adolescent reports of seeing someone stopped by the police and adolescent reports of knowing anyone stopped by the police. Adolescents reporting both personal and vicarious contact are coded as having experienced personal contact (as only 1 percent of adolescents report personal but not vicarious contact).
Additionally, adolescents with either personal or vicarious police contact were asked to report on the following six types of intrusiveness of their most memorable police stop (1 = yes, 0 = no): (1) officer engaged in frisk or pat down, (2) officer searched bags or pockets, (3) officer used harsh language, (4) officer used racial slurs, (5) officer threatened physical force, and (6) officer used physical force. For each of the six types of intrusiveness, I created a series of mutually exclusive variables: personal stop with intrusiveness, personal stop without intrusiveness, and no personal contact (reference category). I also created a summary indicator ranging from 0 to 6.
Control Variables
The analyses adjust for an array of adolescent, parent, and neighborhood characteristics to isolate the relationship between police contact and depressive symptoms. All control variables, unless otherwise noted, are measured prior to the 15-year survey (and, therefore, prior to measures of adolescent police contact or depressive symptoms). These comprise mother’s and father’s demographic characteristics including race/ethnicity (White [non-Hispanic], Black [non-Hispanic], Hispanic, other race [non-Hispanic]), immigrant status, age, family structure in childhood (1 = lived with both biological parents), relationship status (married, cohabiting, nonresidential, or no relationship), repartnership status, relationship quality (1 = poor to 5 = excellent), and number of children. These comprise mother’s and father’s socioeconomic characteristics including educational attainment (less than high school, high school, more than high school), employment, income-to-poverty ratio, and material hardship (a sum of 11 items about hardship experienced in the past year, such as “moved in with other people even for a little while because of financial problems” and “evicted from home or apartment for not paying rent or mortgage”). They comprise mother’s and father’s health characteristics including depression (1 = major depressive disorder in the past year, measured by the Composite International Diagnostic Instrument–Short Form [CIDI-SF]; Kessler et al. 1998), fair/poor overall health, heavy drinking (1 = parent had four or more drinks in one sitting in past month), and illicit drug use (1 = parent reported using drugs without a doctor’s prescription or for a longer period than prescribed in the past month). They comprise mother’s and father’s neighborhood characteristics including disadvantage (an average of the following standardized census tract characteristics: percent without college degree, percent unemployed, percent below poverty line, percent receiving public assistance), race/ethnic composition (individual indicators of the following census tract characteristics: percent White, percent Black, percent Hispanic), social control (an average of five items [1 = very unlikely to 5 = very likely] including “neighbors would do something or get involved if children were skipping school and hanging out on the street” and “neighbors would do something or get involved if a fight broke out in front of the house or building”), and social cohesion (an average of five items [1 = strongly disagree to 4 = strongly agree] including “people around here are willing to help their neighbors” and “this is a close-knit neighborhood”). I also adjust for mother’s and father’s characteristics especially associated with adolescent police contact including cognitive ability (Wechsler 1981), impulsivity (an average of six items [1 = strongly agree to 4 = strongly disagree] including “I often say what comes into my head without thinking” and “I often say/do things without considering the consequences”; Dickman 1990), police contact (1 = parent ever stopped by the police), and incarceration (1 = parent ever incarcerated).
Adolescent characteristics include gender, age (measured at the 15-year survey), low birth weight, and prior mental health (measured by mother’s reports of internalizing behaviors from the Child Behavior Checklist [Achenbach 1992]). Adolescent characteristics also include delinquency (a sum of 17 items, measured at the 9-year survey, including “purposely damaged or destroyed property that wasn’t yours”; “secretly taken a sip of wine, beer, or liquor”; and “smoked marijuana, grass, pot, weed”; Maumary-Gremaud 2000) and impulsivity (measured at the 15-year survey but considered a stable characteristic; Dickman 1990). Finally, some analyses adjust for adolescent characteristics that could have occurred after police contact (therefore providing a conservative test of the relationship between police contact and depressive symptoms). These include a continuous variable indicating peer delinquency (an average of 11 items ranging from 1 = never to 3 = often) and binary variables indicating a police officer is stationed at the adolescent’s school, ever smoked cigarettes, ever drank alcohol, ever used drugs, ever arrested, and ever incarcerated (all measured at the 15-year survey). The latter two variables, though likely adjusting for experiences occurring after police contact, ensure the observed relationship between police contact and depressive symptoms is independent of other forms of criminal justice contact that are associated with mental health (Porter 2019).
Statistical Analyses
Following an examination of mean differences in depressive symptoms across the three groups of exposure to police contact (personal police contact, vicarious but not personal police contact, no contact), the multivariate analyses occur in three stages. First, I use ordinary least squares (OLS) regression models to estimate adolescent depressive symptoms as a function of personal police contact and vicarious but not personal police contact. Model 1 estimates the unadjusted association. Model 2 adjusts for all control variables. This model ensures proper time ordering between depressive symptoms (measured at the 15-year survey), exposure to police contact (measured at the 15-year survey), and the control variables (measured at or prior to the 9-year survey). A minority of adolescents (n = 39) report their first police contact occurring at age 8; supplemental analyses excluding these observations produce similar conclusions as those presented. Model 3, a conservative model, further adjusts for adolescent characteristics measured at the 15-year survey that could have resulted from police contact (e.g., peer delinquency, arrest). These and all subsequent models adjust for city fixed effects, as policing practices may vary across urban cities (Geller 2018).
Second, I use OLS regression models to consider the association between the intrusiveness of personal police contact and depressive symptoms among adolescents. These models adjust for all control variables.
Third, I use OLS regression models to estimate subgroup analyses of the association between police contact and depressive symptoms across gender (boys and girls), race/ethnicity (White, Black, and Hispanic), and poverty status (at or below the poverty line and above the poverty line, measured at the 9-year survey). The subgroup analyses by race/ethnicity refer to mother-reported race/ethnicity, but supplemental analyses using adolescent-reported race/ethnicity produce similar findings. All subgroup analyses adjust for all control variables. I examine statistically significant differences across groups (Paternoster et al. 1998).
Analyses that employ inverse probability of treatment weighting for multiple treatments produce estimates that are consistent in the size and magnitude of those presented here (available upon request). Given that these inverse probability of treatment weighted models are similarly limited as the analyses presented here (that is, they are unable to account for unobservable characteristics that might render the relationship between police contact and depressive symptoms spurious), I present the OLS regression models for ease of interpretation.
Sample Description
Table 1 presents descriptive statistics. On average, adolescents reported 1.597 depressive symptoms (on a scale ranging from 1 = strongly disagree to 4 = strongly agree). More than three-quarters of adolescents reported police contact, with 26.9 percent reporting personal contact and an additional 51.5 percent reporting vicarious but not personal police contact. Frisks and searches were the most common type of intrusion reported, with 9.2 percent and 10.2 percent of adolescents in the analytic sample reporting a personal stop that included a frisk and search, respectively.
Descriptive Statistics for Variables Used in Analyses.
Note. b = measured at baseline survey, y1 = measured at 1-year survey, y3 = measured at 3-year survey, y5 = measured at 5-year survey, y9 = measured at 9-year survey, y15 = measured at 15-year survey.
The majority of adolescents have parents (78.2 percent of mothers and 81.0 percent of fathers) who are racial/ethnic minorities. The majority of adolescents (61.3 percent) have biological parents who are not in a residential relationship. About 17.7 percent of mothers and 16.1 percent of fathers report depression. About 21.6 percent of mothers report ever being stopped by the police, and 10.2 percent of mothers report ever being incarcerated. This is true of 60.6 percent and 49.8 percent of fathers, respectively. Male adolescents comprise slightly more than half (51.5 percent) of the sample, and on average, adolescents are 15.6 years old at the latest survey wave.
Results
Depressive Symptoms, by Police Contact
Figure 1 presents means of depressive symptoms by police contact (personal, vicarious but not personal, and neither personal nor vicarious), first for the full sample and then by gender, race/ethnic, and poverty subgroups. Turning first to the means for the full sample, adolescents with personal police contact report the most depressive symptoms (1.693), followed by adolescents with vicarious but not personal contact (1.602) and adolescents with neither personal nor vicarious contact (1.465). The differences between all groups are statistically significant (p < .001). These patterns exist for boys and girls; for White, Black, and Hispanic adolescents; and for adolescents living in households with incomes at or below the poverty line and adolescents living in households above the poverty line.

Means of depressive symptoms, by exposure to police contact.
These descriptives also show that the level of depressive symptoms varies across subgroups. For example, girls with personal police contact have more depressive symptoms than boys with personal police contact (1.870 compared to 1.616, p < .001), and girls with vicarious but not personal police contact have higher depressive symptoms than boys with vicarious but not personal contact (1.662 compared to 1.527, p < .001). Of those with personal police contact, depressive symptoms are highest for Hispanics (1.769, compared to 1.699 for Whites [ns] and 1.666 for Blacks [p < .05]), and of those with vicarious but not personal police contact, depressive symptoms are highest for Blacks (1.631, compared to 1.602 for Hispanics [ns] and 1.539 for Whites [p < .05]). Across all three types of police contact, adolescents in poverty report more depressive symptoms than adolescents not in poverty (for example, 1.750 compared to 1.651 among those with personal police contact, p < .05).
Estimating Depressive Symptoms as a Function of Police Contact
Table 2 presents results from OLS regression models that estimate adolescent depressive symptoms as a function of police contact. Model 1, the unadjusted model, shows that both personal (b = 0.227, p < .001) and vicarious but not personal (b = 0.137, p < .001) police contact are associated with greater depressive symptoms. In Model 2, which adjusts for an array of adolescent and parent characteristics associated with police contact and depressive symptoms, the associations are reduced in magnitude but remain statistically significant. Both personal (b = 0.141, p < .01) and vicarious but not personal (b = 0.079, p < .01) police contact are associated with greater depressive symptoms. The coefficients are not statistically different from one another (p = 0.113), suggesting that both personal contact and vicarious but not personal contact are similarly associated with depressive symptoms. In Model 3, which further adjusts for adolescent characteristics at the 15-year survey, both personal (b = 0.079, p < .05) and vicarious but not personal (b = 0.052, p < .05) police contact are associated with greater depressive symptoms. The coefficients for personal contact and vicarious but not personal contact are not statistically different from one another (p = 0.535). In this conservative model, these coefficients translate into one-eighth and one-tenth of a standard deviation of the dependent variable, respectively. Also, in this conservative model, the personal contact and vicarious but not personal contact coefficients are, respectively, about one-half and one-third as large as the adolescent gender coefficient, which is noteworthy given the established association between gender and depressive symptoms among adolescents (Avison and McAlpine 1992).
OLS Regression Models Estimating Adolescent Depressive Symptoms as a Function of Personal and Vicarious Police Contact.
Note. Model 1 presents the unadjusted association. Model 2 includes all control variables from Table 1 except for the following adolescent characteristics measured at the 15-year survey: peer delinquency, police officer stationed at school, ever smoked cigarettes, ever drank alcohol, ever used drugs, ever arrested, and ever incarcerated. Model 3 includes all control variables from Table 1 including the adolescent characteristics measured at the 15-year survey. Coefficients are unstandardized. All models adjust for city fixed effects. OLS = ordinary least squares.
p < .05. **p < .01. ***p < .001.
Supplemental Analyses
The measure of vicarious but not personal police contact combines adolescents who witnessed a police stop (50 percent of the analytic sample) and adolescents who know someone stopped by the police (55 percent of the analytic sample [descriptives not shown]). Supplemental analyses show that, net of all control variables, both witnessing a police stop (b = 0.085, p < .01) and knowing someone stopped by the police (b = 0.055, p < .05) are positively associated with depressive symptoms. Furthermore, adolescents who report knowing someone stopped by the police were asked to identify their relationship to that individual(s), with 9 percent reporting a parent’s stop, 5 percent reporting a sibling’s stop, 33 percent reporting a friend’s stop, 4 percent reporting a neighbor’s stop, and 19 percent reporting a stop of someone else. Supplemental analyses show that the police contact of friends (b = 0.045, p < .05)—but not that of parents (b = 0.022, ns), siblings (b = 0.024, ns), neighbors (b = 0.002, ns), or other individuals (b = 0.030, ns)—is positively associated with depressive symptoms.
Estimating Depressive Symptoms as a Function of Intrusiveness of Police Contact
Table 3 presents results of OLS regression models that estimate adolescent depressive symptoms as a function of the intrusiveness of personal police contact. Three key findings emerge. First, both intrusive and nonintrusive personal contact (compared to no contact) are associated with greater depressive symptoms. For example, compared to adolescents who report no contact, adolescents who experience a stop with a frisk (b = 0.117, p < .05) and adolescents who experience a stop without a frisk (b = 0.067, p < .10) have greater depressive symptoms. Similarly, compared to adolescents who report no contact, adolescents who experience a stop with harsh language (b = 0.131, p < .05) and adolescents who report a stop without harsh language (b = 0.069, p < .10) have greater depressive symptoms. Second, across most forms of intrusiveness, the magnitude of the intrusive contact coefficient is larger than the magnitude of the nonintrusive contact coefficient (though the coefficients, across all forms of intrusiveness, are not statistically different from one another). Third, a greater level of intrusiveness is positively associated with depressive symptoms (b = 0.032, p < .001).
OLS Regression Models Estimating Adolescent Depressive Symptoms as a Function of Intrusiveness of Police Contact.
Note. Each type of intrusion estimated with a separate regression model. All models adjust for variables in Model 2 of Table 2. Coefficients are unstandardized. All models adjust for city fixed effects. OLS = ordinary least squares.
^p < .10. *p < .05. **p < .01. ***p < .001.
Considering Variation in Relationship between Police Contact and Depressive Symptoms
Table 4 presents results by adolescent gender, race/ethnicity, and poverty status. First, the magnitude of the association between police contact and depressive symptoms is larger among girls than among boys. Personal contact is associated with a 0.203-point increase in depressive symptoms among girls (p < .01) and a 0.095-point increase in depressive symptoms among boys (p < .05). Similarly, vicarious but not personal contact is associated with a 0.090-point increase in depressive symptoms among girls (p < .01) and a 0.070-point increase in depressive symptoms among boys (p < .05). The differences across groups in the relationship between personal police contact and depressive symptoms is not statistically significant (z = −1.56).
OLS Regression Models Estimating Adolescent Depressive Symptoms as a Function of Personal and Vicarious Police Contact, by Gender, Race/Ethnicity, and Poverty Status.
Note. All models adjust for variables in Model 2 of Table 2. Coefficients are unstandardized. All models adjust for city fixed effects. OLS = ordinary least squares.
^p < .10. *p < .05. **p < .01.
Second, the association between police contact and depressive symptoms is concentrated among Blacks. Personal contact is associated with a 0.161-point increase in depressive symptoms among Blacks (p < .01), a 0.124-point increase in depressive symptoms among Hispanics (ns), and a 0.065-point increase in depressive symptoms among Whites (ns). Similarly, vicarious but not personal contact is associated with a 0.146-point increase in depressive symptoms among Blacks (p < .01), a 0.018-point increase in depressive symptoms among Hispanics (ns), and a 0.026-point decrease in depressive symptoms among Whites (ns). The differences in the relationship between vicarious police contact and depressive symptoms are statistically significant when comparing Whites and Blacks (z = 2.45) and when comparing Whites and Hispanics (z = 2.40).
Finally, personal contact is associated with a 0.165-point increase in depressive symptoms among adolescents in poverty (p < .05) and a 0.132-point increase in depressive symptoms among adolescents not in poverty (p < .01). Vicarious but not personal contact is associated with a 0.131-point increase in depressive symptoms among adolescents in poverty (p < .01) and a 0.057-point increase in depressive symptoms among adolescents not in poverty (p < .10). The differences between groups are not statistically significant.
Supplemental Analyses
The association between intrusiveness of police contact and adolescent depressive symptoms may vary by race/ethnicity, gender, and poverty status. Appendix Table 1 presents descriptive statistics of stop intrusiveness across these demographic groups. Findings highlight that intrusive police stops are concentrated among boys. Compared to girls, boys are six times as likely to experience a stop with a frisk (15.5 percent compared to 2.6 percent); five times as likely to experience a stop with a search (16.4 percent compared to 3.5 percent); four times as likely to experience a stop with a slur (3.2 percent compared to 0.9 percent), threat (5.9 percent compared to 1.4 percent), or force (4.9 percent compared to 1.2 percent); and three times as likely to experience a stop with harsh language (8.1 percent compared to 2.7 percent). Findings also highlight that intrusive police stops are concentrated among Blacks and among adolescents living in households at or below the poverty line. For example, Blacks are three times more likely than Whites (12.6 percent compared to 3.9 percent) and two times more likely than Hispanics (12.6 percent compared to 7.3 percent) to experience a stop with a frisk.
Appendix Table 2 presents results from regression models that estimate the association between intrusive police contact and depressive symptoms for each of the demographic subgroups. Few coefficients reach statistical significance, due to the small sample sizes (and the relatively few number of adolescents within each subgroup who experience intrusive stops), but the patterns across subgroups mirror the patterns for the full sample; both intrusive and nonintrusive personal contact (compared to no contact) are positively associated with depressive symptoms, and the magnitude of the intrusive contact coefficients is often larger than the magnitude of the nonintrusive contact coefficients.
Discussion
In this paper, I used data from the Fragile Families and Child Wellbeing Study, a longitudinal survey of youth last interviewed during the peak of the proactive policing era (Kubrin et al. 2010), to investigate the relationship between personal and vicarious police contact and depressive symptoms in adolescence. Results suggest three primary conclusions about the role of police contact as a social determinant of mental health among adolescents.
First, both personal police contact and vicarious but not personal police contact is associated with greater depressive symptoms among adolescents. These findings are consistent with the stress process perspective, which posits that stressors, concentrated among vulnerable and disadvantaged groups, have deleterious consequences for health (Pearlin 1989; Pearlin et al. 1981). These findings are also consistent with general strain theory, which posits that negative emotions stem from socially patterned strains (Agnew 2006). Indeed, police contact is a consequential stressor and strain not uncommonly experienced by adolescents (Geller 2018). There are several mechanisms through which personal police contact could trigger or intensify depressive symptoms. Personal police contact can involve invasive searches, emotional degradation, and physical violence (Brunson and Weitzer 2009; Rios 2011; Shedd 2015). Personal police contact can be traumatic and can trigger anticipation of future criminal justice contact and its attendant consequences (Geller et al. 2014). It may also reduce personal efficacy and decrease coping resources, both of which could exacerbate the relationship between police contact and depressive symptoms (Thoits 1995). Furthermore, vicarious police contact, via witnessing a stop or knowing someone stopped, could also trigger or intensify depressive symptoms. Indeed, the mental health consequences of vicarious police contact is consistent with the stress process perspective’s proposition that stressors can have contagious or proliferating consequences (Pearlin et al. 1997; also see Agnew 2006). Examining the processes underlying adolescent police contact and depressive symptoms is an important direction for future research.
Second, although both intrusive and nonintrusive police contact are similarly consequential, a higher level of intrusiveness is positively associated with depressive symptoms. These findings dovetail with prior research documenting that intrusiveness is positively associated with anxiety and posttraumatic stress disorder (PTSD) and that intrusiveness is positively associated with emotional distress (Geller 2018; Jackson et al. 2019). These findings also dovetail with research showing that exposure to police victimization is harmful to mental health (Sewell et al. 2016; Sewell and Jefferson 2016). The positive association between intrusiveness and depressive symptoms also provide suggestive evidence that intrusiveness may be a mechanism through which police contact increases depressive symptoms among adolescents.
Third, the association between police contact and depressive symptoms is concentrated among girls (compared to boys) and among Blacks (compared to Whites and Hispanics). The stress process perspective and general strain theory are nonspecific in nature, but these findings suggest the importance of considering how stressors and social position interact in predicting adolescent health (Agnew 2006). The stress process perspective posits that social groups most commonly exposed to a stressor, such as police contact, will experience the most deleterious consequences of the stressor (Pearlin 1989). This perspective is supported in the findings about race/ethnic heterogeneity in consequences, as Blacks are most commonly exposed to police contact and experience the largest consequences of police contact. This finding is also consistent with prior research on young adults (Baćak and Nowotny 2020). The stress process perspective is not supported in the findings about gender heterogeneity in consequences, as boys are more commonly exposed to police contact but girls suffer the most deleterious consequences. The larger consequences for girls may emerge from the fact that girls generally exhibit more depressive symptoms than boys, and it may be they are more reactive in this way to stressors such as police contact (Leadbeater et al. 1999). Importantly, given that males and females often manifest distress differently in response to the same type of stress (with males being more likely to exhibit substance use or abuse symptoms, for example), it is critical to consider multiple mental health outcomes in future research (Aneshensel et al. 1991).
These results have implications for research and practice. They suggest that adolescents who experience police contact, either themselves or vicariously, are a vulnerable population at risk of depressive symptoms. Researchers should continue to interrogate how the criminal justice system affects adolescent health (and which groups of adolescents are most vulnerable to the negative repercussions). Pediatricians, especially those in urban areas, may consider asking adolescents about their interactions with the police and pay particular attention to the mental health concerns of these adolescents. Additionally, given the associations between police contact and depressive symptoms are larger among girls than boys, pediatricians should not assume that girls are not experiencing police contact.
Limitations
The analyses used the best available data to understand the association between police contact and depressive symptoms in adolescence, but limitations exist. First, the Fragile Families data comprise adolescents born in urban areas, and therefore, the analyses omit adolescents born in rural areas. The consequences of police contact may vary across urban and rural areas. Second, adolescents are asked to report on the intrusiveness of their most memorable (personal or vicarious) police stop. I examine intrusiveness only among adolescents reporting personal police contact, but it is possible that one’s most memorable stop was a vicarious stop. Future data collection efforts should distinguish between the intrusiveness of personal and vicarious police contact. Third, though the analyses adjust for an array of characteristics (including delinquency, impulsivity, prior mental health, and parental criminal justice contact), the observational data necessarily preclude causal conclusions.
Conclusions
Taken together, these findings extend our understanding of how the criminal justice system contributes to social disparities in health by considering police contact, which stands in contrast to the majority of research on the criminal justice system that instead examines incarceration (Turney and Wakefield 2019). Specifically, the findings about the consequences of personal and vicarious police contact and depressive symptoms extend prior research on police contact in three important ways (Baćak and Nowotny 2020; DeVylder, Oh et al. 2017; Geller 2018; Geller et al. 2014; McFarland et al. 2019; Sewell et al. 2016; Sewell and Jefferson 2016). First, they extend prior research by examining mental health in adolescence, an important life course period when health trajectories are set in motion (Crosnoe and Johnson 2011). Second, they show that vicarious contact, in addition to personal contact, has deleterious consequences for mental health, complementing and extending prior research that examines how neighborhood-level vicarious contact is associated with physical and mental health (McFarland et al. 2018; Sewell et al. 2016; Sewell and Jefferson 2016). Third, the consideration of heterogeneity in the stop experience (via intrusiveness) and heterogeneity in the consequences of police contact by gender, race/ethnicity, and poverty status also complements and extends prior research in this area (Baćak and Nowotny 2020; Geller 2018; McFarland et al. 2018; Sewell et al. 2016). Given how strains can exacerbate crime through facilitating negative emotions (such as depressive symptoms), police contact may indirectly trigger criminal activity (Agnew 2006). Further, given the concentration of police contact among already vulnerable adolescents living in highly surveilled and disadvantaged neighborhoods, police contact may exacerbate inequalities in adolescent health (Turner 2013).
Footnotes
Appendix
OLS Regression Models Estimating Adolescent Depressive Symptoms as a Function of Intrusiveness of Police Contact, by Gender, Race/Ethnicity, and Poverty Status.
| Gender | Race/Ethnicity | Poverty Status | ||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Boys | Girls | Whites | Blacks | Hispanics | At or Below Poverty Line | Above Poverty Line | ||||||||
| b | SE | b | SE | b | SE | b | SE | b | SE | b | SE | b | SE | |
| Police contact (reference = no contact) | ||||||||||||||
| Stop with frisk | 0.090 | (0.049)^ | 0.141 | (0.079)^ | -0.014 | (0.189) | 0.087 | (0.054) | 0.177 | (0.127) | 0.113 | (0.055)^ | 0.118 | (0.059)^ |
| Stop without frisk | 0.010 | (0.041) | 0.138 | (0.046)** | 0.107 | (0.088) | 0.037 | (0.036) | 0.073 | (0.075) | 0.033 | (0.051) | 0.082 | (0.034)* |
| Police contact (reference = no contact) | ||||||||||||||
| Stop with search | 0.075 | (0.049) | 0.032 | (0.085) | 0.012 | (0.187) | 0.052 | (0.059) | 0.103 | (0.131) | 0.065 | (0.065) | 0.099 | (0.052)^ |
| Stop with no search | 0.020 | (0.045) | 0.167 | (0.051) | 0.108 | (0.086) | 0.061 | (0.042) | 0.108 | (0.063) | 0.066 | (0.055) | 0.090 | (0.034)* |
| Police contact (reference = no contact) | ||||||||||||||
| Stop with harsh language | 0.059 | (0.049) | 0.315 | (0.128)* | -0.023 | (0.138) | 0.071 | (0.068) | 0.351 | (0.150)* | 0.062 | (0.077) | 0.177 | (0.074)* |
| Stop without harsh language | 0.033 | (0.041) | 0.105 | (0.052)^ | 0.095 | (0.092) | 0.053 | (0.041) | 0.040 | (0.069) | 0.062 | (0.048) | 0.074 | (0.035)* |
| Police contact (reference = no contact) | ||||||||||||||
| Stop with racial slurs | 0.121 | (0.122) | 0.296 | (0.133)* | 0.046 | (0.017) | 0.196 | (0.113) | 0.176 | (0.195) | 0.112 | (0.129) | 0.214 | (0.124) |
| Stop with no racial slurs | 0.037 | (0.036) | 0.130 | (0.053)* | 0.092 | (0.092) | 0.046 | (0.039) | 0.098 | (0.079) | 0.063 | (0.043) | 0.085 | (0.034)* |
| Police contact (reference = no contact) | ||||||||||||||
| Stop with threat of force | -0.028 | (0.061) | 0.131 | (0.117) | -0.204 | (0.130) | 0.035 | (0.061) | -0.021 | (0.199) | 0.005 | (0.067) | 0.026 | (0.075) |
| Stop with no threat of force | 0.057 | (0.043) | 0.139 | (0.049)* | 0.110 | (0.093) | 0.062 | (0.042) | 0.120 | (0.076) | 0.080 | (0.051) | 0.100 | (0.035)* |
| Police contact (reference = no contact) | ||||||||||||||
| Stop with force | 0.092 | (0.065) | 0.231 | (0.179) | -0.005 | (0.127) | 0.057 | (0.072) | 0.375 | (0.206)^ | 0.104 | (0.079) | 0.166 | (0.111) |
| Stop with no force | 0.037 | (0.043) | 0.132 | (0.049)* | 0.093 | (0.092) | 0.058 | (0.039) | 0.075 | (0.070) | 0.059 | (0.048) | 0.087 | (0.035)* |
| Intrusion index | 0.031 | (0.010)** | 0.032 | (0.009)*** | 0.024 | (0.026) | 0.034 | (0.009)** | 0.028 | (0.012)* | 0.020 | (0.013) | 0.037 | (0.010)** |
| N | 1,771 | 1,666 | 748 | 1,723 | 845 | 1,236 | 2,201 | |||||||
Note. Each type of intrusion estimated with a separate regression model. All models adjust for variables in Model 2 of Table 2. Coefficients are unstandardized. All models adjust for city fixed effects. OLS = ordinary least squares.
^p < .10. *p < .05. **p < .01. ***p < .001.
