Racial Differences in Activity Space Exposures and Everyday Perceptions of Safety Among Urban Youth

Abstract

Purpose:

The everyday experience of safety promotes health and successful development during adolescence. To date, few studies have examined racial variation in the spatial determinants of in-the-moment perceived safety.

Methods:

Drawing on data from the Columbus, OH-based Adolescent Health and Development in Context (AHDC) study (N=1405), we consider the influence of intra-individual variability in GPS-based exposure to both high proportion white urban neighborhoods and neighborhood violence for the everyday location-based safety perceptions of Black and white youth (ages 11 to 17) as captured by ecological momentary assessment.

Results:

Exposure to higher area-level violence reduces youths’ safety perceptions. Momentary exposure to residentially white-dominated neighborhoods also reduces perceived safety, but only for Black youth who spend more time, on average, in white areas. In contrast, we observe some limited evidence that white youth perceive greater safety when in white neighborhoods if they spend more time in white neighborhoods on average.

Discussion:

These findings point to the need for greater attention to in situ experiences in understanding the origins of racial disparities in health and wellbeing. For Black youth, a restricted focus on the consequences of residing in Black segregated neighborhoods may obscure potentially health consequential exposures beyond these areas.

INTRODUCTION

Among youth, the state of safety enables the pursuit and optimal experience of everyday activities that foster successful adolescent development [1]. In contrast, occurrences of compromised safety may threaten mental and physical health [2,3], reduce the benefits of educational experiences [4], and encourage a defensive and mistrustful orientation toward everyday interaction [5]. Evidence suggests that even the anticipation of unsafe encounters can have significant long-term consequences for physical and mental health [6].

The global concept of safety includes both physical safety (absence of the threat of victimization and injury) and psychological safety (absence of the threat of social exclusion). A substantial literature has focused on safety and associated fear of crime, emphasizing physical victimization threat, mostly in residential neighborhood contexts [7]. Neighborhood-based fear of crime has been linked with reduced physical activity [8], physical health [9], mental health [10], and school functioning [11] among youth.

Psychological safety experiences, however, may also be relevant for understanding global perceptions of safety. Research on psychological safety has examined the beneficial role of belief that participation in a given setting such as a workplace or school (e.g., through voicing one’s opinions) will not be met with sanction or rejection [12]. We extend this concept to participation in activity space settings – the collection of locations youth encounter on a routine basis – more broadly, with psychological safety capturing the experience of trust that one’s presence at a given location will not result in unprovoked social exclusionary responses. We expect that youth incorporate judgments about both physical and psychological safety in arriving at perceptions of safety in the moment [13].

Black American youth face unique and potentially pervasive challenges to their basic sense of safety in everyday life. The historic and ongoing residential racial segregation characterizing a majority of large-scale US metro areas contributes to the concentration of neighborhood poverty and associated diminished social organization and elevated violence. Youth residing in Black segregated neighborhoods are at increased risk of exposure to violence [14] and likely experience lower levels of perceived everyday safety as a result. Yet, evidence also points to the considerable complexity of everyday activity spaces [15,16]. In particular, emerging mobility data indicate that Black youth spend non-trivial amounts of time in racially mixed and even white-dominated neighborhoods [17]. Black youth may experience compromised safety when navigating white areas due to the risk, and anticipation, of racist interactions [18,19]. The higher probability of such encounters in white-dominated areas may elicit feelings of both physical and psychological unsafety. In the former case, the risk of being targeted by police and the potential for escalation of encounters for Black youth may induce uncertainty with respect to safety [20]. Experiences of psychological unsafety may also be elevated for Black youth in white neighborhoods due to the ongoing potential for racism-related exclusionary encounters. Extant research documents the range of negative experiences faced by Black people when navigating high proportion white neighborhoods, at the hands of both law enforcement [21] and local residents [18]. This approach thus points to the unsafety consequences of time in both higher violence and white-dominated areas as significant sources of everyday compromised safety for Black youth.

We address the unexplored consequences of exposure to white neighborhoods for real-time experiences of safety among urban Black youth. We also consider the extent to which overall familiarity with white-dominated neighborhoods captured by time spent in such contexts conditions the influence of in-the-moment white neighborhood exposure on perceptions of safety. The classic intergroup contact hypothesis [22] and related research on the benefits of familiarity in the context of conventional routines would expect that more frequent exposure to white neighborhoods would enhance familiarity and diminish feelings of uncertainty when Black youth encounter white neighborhoods (the familiarity reduces threat hypothesis). A less optimistic but plausible hypothesis would emphasize the role of frequent encounters with white neighborhoods in familiarizing Black youth with the specific nature of the threats to which they are vulnerable. More time in white neighborhoods will likely lead to cumulative experiences of both direct and vicarious exposure to instances of discrimination, surveillance, targeting by police and other authorities, micro-aggression and other exclusionary processes that lead Black youth to anticipate reduced safety in the white urban areas they encounter (the familiarity reveals threat hypothesis) [23]. For instance, in an analysis of Black families residing in suburban white neighborhoods, Posey-Maddox [24] found that Black caregivers expressed initial optimism about their new residential contexts but became increasingly frustrated over time as negative race-based experiences accumulated.

Empirical investigation of the contextual factors that shape in situ safety perceptions has been limited. Important qualitative work has addressed the salience of ambient violence in shaping mobility patterns and the challenges of navigating areas of differing racial composition for urban Black youth [25]. In addition, a substantial body of research has examined fear of crime; however, these studies have largely drawn on social survey reports of fear in residential neighborhoods [26] or qualitative interview data on fear of crime at particular locations [27]. An emerging literature has employed mobile app-based approaches to measuring perceptions of safety and fear of crime in situ. These studies provide preliminary data indicating the feasibility of collecting real-time reports on safety but have typically relied on small samples and do not link independent information on the geographic contexts of reports to explain variability in perceptions [28,29].

This study considers hypotheses regarding the association between within-person variation in exposure to violence and residentially white spaces and in-the-moment perceptions of safety using extensive smartphone-based geographically-explicit ecological momentary assessment (GEMA) data [30] from the Adolescent Health and Development in Context study (AHDC) – a representative study of 1,405 youth aged 11 to 17 and their caregivers in Franklin County, OH. In regression models for perceived safety, we control for all between-respondent time-invariant characteristics with person fixed effects.

MATERIALS AND METHODS

AHDC is a longitudinal, representative study of the sociospatial contexts of adolescent development (N = 1405). The Institutional Review Board of Ohio State University approved the AHDC study design, consent process, and data collection. The study included urban and suburban areas within Interstate 270 – the Franklin County outer belt, including the majority of the city of Columbus and numerous suburbs (67% non-Hispanic white, 21% non-Hispanic Black, 5% Hispanic, and 4% Asian, with a poverty rate of 18%, according to the 2009–2013 American Community Survey (ACS) estimates – just prior to the AHDC 2014–16 field period). The sampling frame was based on a combination of a vendor-provided list of households with high probability of meeting eligibility criteria and data from public school districts represented in the study area. Among contacted, eligible households, one randomly selected youth aged 11–17 and one primary caregiver (English speaking) were recruited to participate in the study. The American Association for Public Opinion Research (AAPOR) Cooperation Rate 1 – the percent of completed interviews out of contacted households that were determined to be eligible – is 88%. The AAPOR Response Rate 3 for the sample is 21.3%, which is the percent of completed interviews out of the number of estimated eligible households in the sampling frame.¹

First, Entrance Surveys were conducted separately with the focal youth and their caregiver covering a range of topics including demographic and socioeconomic background; household composition, family structure and marital status; employment and income; health, social support, and alcohol/substance use; youth relationships; youth risk and pro-social behaviors; and the coordinates of places youth regularly go. This was followed by a seven-day smartphone-based GEMA [30] period for the youth, combining GPS tracking and brief phone-based questionnaires to examine youth perceptions, behaviors, and activity space locations in real time across the study week. During the in-home interview, the interviewer provided a smartphone to the youth with instructions to carry the phone continuously over the course of the week. The GPS feature of the study design facilitated collection of data on the locations at which youth spend time through continuous tracking. At the end of the GEMA week, each adolescent participant and the interviewer reviewed the activity space data together using an interactive recall-aided space-time budget [33] (see Appendix Section 1).

The EMA component of the study involved administration of five randomly timed mini-surveys per day over the course of the week. The EMA averages approximately 4.5 minutes in length and asks the respondent questions about their current location, activity, presence of network partners, spatial and social characteristics of the setting, and risk behaviors. Youth were given 20 minutes to acknowledge the prompt and an additional 20 minutes to complete the questionnaire once logged on. Youth were asked to report on their experiences at the time of the prompt rather than their current location. Youth participants received a monetary incentive at the completion of 50% and 75% of the EMAs administered over the course of the week. EMA prompts were not administered during school hours on school days. The EMA response rate was 53%.

Dependent Variable

Our dependent variable is youths’ in-the-moment perceptions of current location safety. On each EMA, youth were asked if they felt the location they were in was a safe place to be (1 = “strongly disagree,” to 5 = “strongly agree”). The analyses presented below use a dichotomized safety outcome where “strongly agree” = 1 (71% of all safety reports) and all other responses = 0. We employ this approach under the assumption that any uncertainty about safety in the respondent’s current setting may generate stress and potential obstacles to the pursuit of goals. In the appendix, we also consider models for safety reports in their original 1–5 metric. To ensure that results are not biased due to missingness on EMAs, we conducted sensitivity analyses using multiple imputation and inverse probability of treatment weights for EMA safety report missingness. Results from both these sets of analyses yield substantive conclusions consistent with those based on our presented models. We also considered the link between average reports of safety perceptions for routine locations reported on the entrance survey (strongly associated with EMA reports of safety) and missingness, finding no evidence of an association. These results are discussed in section 2 of the appendix.

Independent Variables

We include both Black and white youth in the analyses considering the latter in order to capture contrasts in the effect of neighborhood racial composition on perceived safety across race. In-the-moment measures of activity space racial composition and SES are based on Franklin County Census block group characteristics corresponding to the youth’s location at the time of the EMA response. Accordingly, activity space measures are time-varying. Block group characteristics are constructed using 2009–2013 ACS data [34]. We measure exposure to white-dominated spaces using a binary indicator of whether the block group residential population is at least 70% non-Hispanic white. Block group socioeconomic status is the mean of six z-score standardized items: (1) poverty rate, (2) unemployment rate, (3) the percent of households that are female-headed, (4) the percent of households with an annual income greater than $50,000, (5) the percent of adults ages 25 or older with a college degree, and (6) the percent of adults with a professional occupation. By using both positive and negative items, we capture variation at both the high and low ends of socioeconomic status; higher values indicate higher block group level socioeconomic status. At the EMA level, we control for whether the self-reported location at the time of the EMA was within 30 meters of home and whether it was given on a school day (vs. weekend/holiday/not in session). Hours since 12am is a continuous measure of time of day, as the number of hours that have elapsed since midnight at the time of the EMA prompt, ranging from 6.67–22.47. The quadratic of time and interactions between these time variables with dichotomous school day are also controlled.

In-the-moment violent crime rate is based on the recent violent crime rate of the block group corresponding to a respondent’s location at the time of their EMA response. Violent crimes include incidents of homicide, robbery, aggravated assault, and rape in the Ohio Incident-Based Reporting System from 2014–2016. Incidents that occurred during the field period were geocoded to Franklin County block groups based on spatial coordinates. For each block group j and day t combination, we generated a cumulative count of the number of incidents in the entire t-180 day period, which is then expressed in analyses as the rate of violent crime incidents in the 180-day period per 1000 residents of the block group. For a given EMA response i on day t, we match the crime rate of the block group j in which the respondent is located with the relevant 180-day violent crime rate.

Individual-level activity space %white is calculated by aggregating exposure data from the recall-aided space-time budget information provided by the youth over five days of the EMA/GPS week [33]. We calculate individual-level mean exposure to block group %white across all locations visited within Ohio for the week, weighted by time spent in minutes at each location during waking hours, including time spent at home.

Analytic Strategy and Sample

To test our hypotheses regarding how within-person change in activity space exposures relates to change in perceptions of safety, we employ regression models with person fixed effects. We feature results from linear probability models for “strongly agree” reports that the current location is safe, based on all non-missing EMA reports. In section 2 of the appendix, we present results when using alternative measurement strategies for the dependent and focal independent variables, different modeling strategies, and when accounting for missingness in EMA safety perceptions.

Our analytic sample is restricted to non-missing EMA reports of perceptions of safety by Black and white respondents (total EMA n=17,536; Respondent n=1,139) matched to block groups in the Franklin County study area with non-missing racial composition and violent crime rates (EMA n=16,712; Respondent n=1,135; see Appendix for more details on block group matching), and reported by a youth who completed the interactive recall-aided space-time budget and for whom respondent-level average %white measure could be created (EMA n=16,379; Respondent n=1,103). Because our analyses focus on within-person variability in safety perceptions, our analytic sample declines further to exclude respondents with perfectly consistent reports (e.g., always “strongly agree” safe), which yields varying sample sizes across specific analyses [35]. Our final analytic sample for models presented in the main results includes 10,799 safety reports from 658 individuals.

RESULTS

Descriptive statistics for all study variables are displayed in Table 1, stratified by respondent race. Figure 1 displays the distribution of the number of respondents by respondent race and average exposure to block group proportion white, and Figure 2 displays the number of EMA safety reports by these variables, illustrating that a substantial number of safety perception reports are from Black youth who spend a significant proportion of their time in high proportion white areas.

Table 1.

Descriptive Statistics for Study Variables by Respondent Race.

	White Youth (n=363)			Black Youth (n=295)
	EMA			EMA
	n	Mean	SD	n	Mean	SD	Min.	Max.
EMA Safety Perceptions	5876			4923
Strongly Disagree		0.01	-		0.01	-	0	1
Disagree		0.01	-		0.01	-	0	1
Neither Agree nor Disagree		0.06	-		0.09	-	0	1
Agree		0.23	-		0.22	-	0	1
Strongly Agree		0.70	-		0.66	-	0	1
Within 30 meters of home	5876	0.72	-	4923	0.68	-	0	1
Hours since 12am (6–22.5)	5876	16.24	4.43	4923	16.05	4.59	6.67	22.47
School day (vs. non-school day, holiday)	5876	0.47	-	4923	0.50	-	0	1
BG SES (z-scores)	5876	0.63	0.96	4923	−0.66	0.76	−2.44	1.99
BG violent crime rates, last 180 days	5876	1.77	3.06	4923	5.25	6.38	0.00	66.67
BG >=70% white	5876	0.77	-	4923	0.17	-	0	1
Youth-level week mean %white	5876	0.78	0.17	4923	0.37	0.21	0.02	0.93

Note: EMA = Ecological Momentary Assessment; BG = Block Group; SES = Socioeconomic Status.

Figure 1.

Count of Respondents by Race and Youth-level Mean Exposure to Census Block Group Proportion White.

Figure 2.

Count of EMA Safety Reports by Race and Youth-level Mean Exposure to Block group Proportion White.

Table 2 displays results from linear regression models for dichotomized safety perceptions where “strongly agree” safe = 1 and all other responses = 0 (i.e., linear probability models). Model coefficients are interpreted as the change in probability of strongly agreeing the current location is safe associated with a unit increase in a given covariate. The first model includes only race as an independent variable, indicating a small Black-white difference (b = −.042, p < .1) in the outcome. Model 2 adds all EMA-level variables and person fixed effects to adjust for all respondent-level differences. In this model, time since midnight and its quadratic are controlled, as is a binary indicator for whether the EMA was reported on a school day. The time-of-day variables are also interacted with the school day indicator to adjust for potential differences in the association between time and safety perceptions on school vs. non-school days. Youth responding to an EMA within 30 meters of the home have about a 14% greater probability of strongly agreeing that they feel safe in the moment (b = .137, p < .001). On average, a 100 percent increase in the block group-level violent crime rate is associated with a 18% decrease in the probability of the respondent strongly agreeing that they feel safe (b = −.018, p < .05). In a model not shown, the interaction between crime and Black race was also assessed, but was statistically nonsignificant (Black*violent crime b = .002, p = .866). The coefficient for block group SES is positive and significant (b = .036, p < .05) and there is no association between being in a white-dominated block group (“BG >= 70% white”) and the outcome. To assess if the association between in-the-moment presence in a white-dominated block group and safety varies by race, Model 3 adds an interaction term between this measure and respondent race, but still finds no association with the outcome (i.e., a statistically nonsignificant association with the outcome for both Black and white respondents).

Table 2.

Linear Regression Models for EMA Strongly Agree Safe vs. All Other Reports.

	Strongly Agree Safe	Strongly Agree Safe	Strongly Agree Safe	Strongly Agree Safe
Respondent Black	−0.042+ (0.023)
Within 30 meters of home		0.137*** (0.014)	0.135*** (0.014)	0.133*** (0.014)
Hours since 12am (6–22.5)		−0.019* (0.008)	−0.019* (0.008)	−0.019* (0.008)
Hours since 12am * Hours since 12am		0.001+ (0.000)	0.001+ (0.000)	0.001+ (0.000)
School day (vs. non-school day, holiday)		−0.207** (0.078)	−0.207** (0.078)	−0.212** (0.078)
School day * Hours since 12am		0.016 (0.012)	0.016 (0.012)	0.017 (0.012)
School day * Hours since 12am * Hours since 12am		−0.000 (0.000)	−0.000 (0.000)	−0.000 (0.000)
BG SES		0.036* (0.015)	0.038** (0.014)	0.032* (0.014)
ln(BG violent crime rates, last 180 days)		−0.018* (0.008)	−0.017* (0.007)	−0.017* (0.007)
BG >=70% white		0.002 (0.025)	0.025 (0.032)	−0.227+ (0.120)
Black * BG >=70% white			−0.056 (0.046)	0.375** (0.143)
BG >=70% white * week mean %white				0.367* (0.170)
Black * BG >=70% white * week mean %white				−0.762** (0.233)
Intercept	0.693*** (0.015)	0.752*** (0.063)	0.747*** (0.063)	0.731*** (0.064)
Respondent Fixed Effects		✓	✓	✓
Respondent n	658	658	658	658
EMA n	10799	10799	10799	10799

^***p < 0.001

^**p < 0.01

^*p < 0.05

⁺p < 0.1.

Note: EMA = Ecological Momentary Assessment; BG = Block Group; SES = Socioeconomic Status.

Finally, Model 4 tests our central hypotheses by adding the interactions between white-dominated block group * respondent average exposure to proportion white areas (“week mean %white”), and the interaction between this term and respondent race. To facilitate the interpretation of the statistically significant three-way interaction efficiently, Figure 3 displays the estimated average marginal effects (i.e., coefficients) of being in a white-dominated block group on the probability of a strongly agree EMA safety report as a function of the time-weighted average exposure to proportion white over 5 days of the study week for Black and white youth, with corresponding 95% confidence bands for these marginal relationships. For Black youth with a time-weighted average proportion white exposure greater than 37%, the association between in-the-moment presence in a white-dominated block group and expected perceived safety is negative. For example, among Black youth who are exposed, on average, to 70% white neighborhoods, being in a white-dominated block group is associated with an expected 13% decrease in the probability of strongly agreeing that they are in a safe place. In contrast, among white youth who are exposed, on average, to 85% white neighborhoods, being in a white-dominated block group is associated with an increase in the probability of strongly agreeing that they feel safe.

Figure 3.

Average Marginal Effects of Block Group >=70% White on EMA Strongly Agree Safe by Respondent Race.

We include EMAs within 30 meters of the home in the presented results under the assumption that neighborhood conditions may spillover into the home setting affecting safety perceptions. However, we also considered the extent to which the associations held when youth were outdoors (generally), under the assumption that unsafety effects of exposure to white neighborhoods for Black youth should also be observed under these conditions. Specifically, we restricted the analysis to EMAs in which the youth reported being “outdoors” at the time of the prompt. Average marginal effects indicate that Black youth who spend time in whiter neighborhoods on average experience lower in-the-moment safety when outdoors in white neighborhoods (see Appendix – SI Figure 9). Compared with models including all time, the observed unsafety effect increases at a somewhat greater rate with increasing average exposure to white neighborhoods (e.g., at 70% average exposure to white neighborhoods, being outdoors in a white neighborhood in-the-moment reduces the probability of reporting strong agreement that the location is safe by roughly 26%). We also considered the potential for gender interactions in the observed association. Although the estimated negative coefficients for the effect of in-the-moment presence in a white neighborhood on perceived safety increase in magnitude at a greater rate for boys than girls with increasing average exposure to white neighborhoods, we found no evidence of a statistically significant interaction by gender (see Appendix – SI Figures 12/13).

DISCUSSION

Understanding sources of reduced safety perceptions in urban environments is critical to addressing race-based disparities. We examined the role of both exposure to neighborhood poverty and violence as well as white-dominated areas in shaping perceptions of safety during Black youths’ routines. Drawing on linked GPS and ecological momentary assessments of perceived safety, we employed a fixed effects approach to assess intraindividual variation in safety reports by location characteristics.

Analyses offered evidence that both Black and white youth experience declines in safety when spending time in higher violence areas, consistent with evidence linking residential crime rates and both safety and fear reports. We found additional evidence that Black youth report lower levels of perceived safety when in residentially white-dominated (>70% white) neighborhoods. However, this effect was only evident for Black youth who spent a substantial amount of waking time in white neighborhoods. The finding is consistent with the proposed familiarity reveals threat hypothesis – for Black adolescents, more time spent in white neighborhoods is likely to result in the direct or vicarious experience of race-based scrutiny, monitoring, targeting, and microaggressions, amplifying perceptions of unsafety. In contrast, we found no evidence in support of the familiarity reduces threat hypothesis (based on the expectation that more time spent in white areas promotes trust and comfort navigating such spaces).

The observed pattern does not yield optimism regarding the prospects for successful day-to-day mobility-related integration of Black youth into white-dominated – and often resource-rich – areas. The achievement of primary goals – such as schooling, work, and other institutional involvements – may be constrained to the extent that lack of security produces stress, limits trust, and interferes with developmentally appropriate risk-taking in such contexts. Moreover, the experience of unsafety may have greater negative impact for Black youth particularly when rooted in the pervasive threat of race-based targeting [36]. Evidence suggests, for instance, that EMA-based perceptions of unsafety are linked with higher levels of physiological stress as captured by hair cortisol concentration for Black, but not white youth [37]. Racism-based “hypervigilance” – cognitively and emotionally burdensome attention to dress, comportment, and affect to ensure a non-threatening presence among Black people encountering white spaces – has also been linked with compromised health [38]. Less consistently, we find that white adolescents experienced greater safety when in white residential contexts as the average level of exposure to white areas increased.

Our analyses are not without limitations. Our sample was drawn from the Columbus, OH area only, limiting generalizability. Large-scale, representative smartphone-based data collection efforts also face complications associated with EMA-level missingness. Although we found few systematic predictors of missingness other than timing during the GEMA week, the potential for nonresponse bias nevertheless remains a concern. Relatedly, we were unable to collect EMA data during school hours. In addition, our data do not directly address youths’ own assessments of the reasons they experience variation in perceived safety, limiting our ability to corroborate the mechanisms explaining the associations of racial composition and violence with perceived safety. Finally, the extent to which the effects of residential white composition are mediated or moderated by the racial composition of the ambient environment in the moment remains an important question. Future research probing these mechanisms – through both GEMA-based and qualitative approaches – is needed to understand the impact of sociospatial contexts on youths’ safety perceptions.

These analyses point to the need for greater attention to in situ experiences in understanding the origins of racial disparities in wellbeing. GEMA-based data collection strategies offer the potential to reveal mechanisms of racial disparity previously obscured due to the inability to capture lived experience more precisely. More broadly, in combination with insights on the nature of everyday mobility processes, these methods hold great promise to enrich understanding of both the challenges and benefits of the urban experience.

IMPLICATIONS AND CONTRIBUTION.

To date, few studies have examined racial variation in the spatial determinants of in-the-moment safety. Drawing on real-time, smartphone-based reports from 11- to 17-year-olds, we find that Black youth perceive lower safety when in white-dominated areas, but only Black youth that spend a substantial amount of time in white areas.

Abbreviations:

AHDC

Adolescent Health and Development in Context

AAPOR

American Association for Public Opinion Research

ACS

American Community Survey

BG

Census Block Group

EMA

ecological momentary assessment

GEMA

geographically-explicit ecological momentary assessment

GPS

Global Positioning System