A Qualitative Approach to the Dimensions of Segregation among Pregnant Black Women

Abstract

This study compares and contrasts residents’ perceptions of segregation measures using qualitative and quantitative data. Most studies exploring racial residential segregation and health outcomes use large-scale, metropolitan-wide measures. As a result, we have limited understanding of racial residential segregation outside of Census data, particularly about the firsthand experiences of those living in segregated areas. The purpose of this study was to compare data from Census-based measures of racial residential segregation with qualitative descriptions of these same constructs by pregnant, Black women in two US cities. Using novel qualitative interview questions, we explored the dimensions of segregation and neighborhood racial distribution among a sample of 27 pregnant, Black women between April and November 2019. The participants included in this sample had perceptions about their neighborhood segregation and demographic composition that were often different from the data derived from existing residential US Census data. The differences between qualitative and quantitative measures and the possible reasons for the discordance suggest new approaches to measurement and new directions for the study of segregation and health.

Introduction

Racial residential segregation, the extent to which two groups live separately from one another, is one of the ways in which structural racism is operationalized. Racial residential segregation has been associated with preterm birth, increased cardiometabolic risk, pulmonary disease, and lung cancer among Black people living in the USA. [1–4] Racial residential segregation has roots in zoning restrictions, discriminatory housing laws, federal housing programs, and neighborhood covenants; all of which are aimed at creating and maintaining housing inequities among people living in the USA [5–7] Racial residential segregation is most widely measured using five dimensions (evenness, exposure, concentration, clustering, and centralization) that were first described and operationalized by Massey and Denton. [6] The two most common measures, evenness and exposure, refer to the proportion of racial majority and racial/ethnic minority group members and the potential contact between majority and racial/ethnic minority groups in the same place. Centralization and concentration are spatial dimensions that measure distances from the center of a city or town and from the nearest White neighborhood, respectively. The US Census has adopted Massey and Denton’s framework and has published data using the indices that measure evenness and exposures for several US cities. [8]

The extant literature linking racial residential segregation and adverse health outcomes that have used population-based studies have been valuable in establishing the relationship between racial residential segregation and adverse health outcomes. [1, 3, 4, 9] However, these studies have relied solely on Census-level measures over large geographic areas. While there has been qualitative work describing resident-defined neighborhood boundaries and activities, there is a lack of qualitative research that explicitly compares resident perceptions with the compositional measures as described by the widely used Massey and Denton framework. [10, 11]

Over the past two decades, the representativeness of Census-based measures of neighborhoods relative to actual urban experiences has been called into question. Residents’ perceptions of their neighborhoods may not be reflected by measures based on Census data. [11, 12] In addition, recent evidence suggests that younger residents spend the majority of their time outside of their home neighborhood. [13] These criticisms have given rise to activity space perspectives, which operationalize exposure profiles with GPS data, self-reported location data, or residents’ own definitions of their neighborhood. [14–16] Even those approaches draw, from the most part, only on quantitative data and analyses that do not consider to what extent everyday experiences align with these measures.

Furthermore, little has been published to describe how women who live in segregated areas perceive racial residential segregation and its link to the health inequities that they are experiencing first-hand. As a result, we have limited understanding of racial residential segregation outside of Census data, particularly about the firsthand experiences of those living in segregated areas. Qualitative data may provide a more localized meaning to the measures provided by the Census. The purpose of this study was to compare data from Census-based measures of racial residential segregation with qualitative descriptions of these same constructs by pregnant, Black women in two US cities.

Methods

Design

Data reported here are from a subsample of participants in an ongoing large mixed-methods study that examined the associations among social stressors (e.g., racial discrimination, neighborhood disorder) and risk for preterm birth among Black women from two different US cities (Detroit, MI, and Columbus, OH). Using a convergent, parallel design [17] for the parent study, participants completed questionnaires and participated in qualitative interviews. This paper reports on quantitative data from the US Census (segregation indices), demographic data, and qualitative data describing residential segregation.

Sample

Women were eligible to enroll in the parent study if they self-identified as African American or Black, had a singleton pregnancy, and were between 8 and 29 weeks gestation. Forty-five participants completed the qualitative interview for the parent study. After preliminary analysis of interview transcripts and quantitative data, cases with significant missing qualitative data (n = 9) or inadequate data in key areas (e.g., no data for key codes in the qualitative analysis) (n =8) were excluded from the final sample. One participant experienced a late-term fetal demise and was also excluded. The final sample for this study included 27 participants with complete qualitative data.

Variables and Instruments

Participants completed questionnaires that included socio-demographic characteristics. Census-level measures provided by the 2010 Census were used to describe racial residential segregation quantitatively by linking Census data to residential addresses for participants in the study.

• Exposure was measured using the Isolation Index. The Isolation Index measures how groups “experience” segregation through sharing or not sharing common areas.⁶ The Isolation Index percentage measures the potential for contact by minority groups with other groups, particularly White majority groups.

• Centralization measures how a group is located in relation to the center of an urban area. [6] Geocoded data from the Federal Financial Institutions Examinations Council (FFIEC) [18] and the Social Science Data Analysis Network [19] were used to describe demographic data for Census tracts based on the residential addresses provided by the study participants.

• Concentration is a measure of the relative physical space that a group occupies. [6] Concentration was measured as the distance from the residential address to the nearest predominantly White neighborhood tracts as determined by the Census using the mapping software. Concentration was also measured by self-report by asking participants how far they live from the nearest White neighborhood. Distances were calculated using the same self-report data and measures that the participants used in their responses (minutes, miles, walking, driving).

Census demographics with data about racial distribution (percent Black and percent White) for each participant’s address was provided by the FFIEC. [11]

Interview Guide

The qualitative interview guide included questions that were created de novo by the first author to match the dimensions of racial residential segregation as described by Massey and Denton. [11] Sample questions included “Thinking about your neighborhood or where you live, how often do you come into contact with people who aren’t Black?” which measured exposure. Table 1 shows the interview guide questions and index it represents.

Table 1.

Qualitative and quantitative segregation data sources

Dimension of segregation	Participant responses to the interview guide question	Census measures
Exposure	“Thinking about your neighborhood, how often do you come into contact with people who aren’t Black?”	Isolation Index
Centralization	“How far, in minutes or miles do you live from the heart or center of your town?”	Mapped distance to center of town
Concentration	“How far away is the nearest White neighborhood from where you live?”	Mapped distance to nearest White neighborhood as determined by Census tract

Procedures

Pregnant Black women were recruited to participate in the parent study from two urban medical centers in the Midwest. Institutional Review Board approval was obtained from universities at both sites. Women who met inclusion criteria were approached by research staff before or after prenatal appointments. Women who were interested in the study completed an informed consent process prior to study procedures. Participants completed questionnaires where residential addresses were collected. During consent, participants were invited to participate in qualitative interviews during their pregnancies. The qualitative interviews were conducted over the telephone with research assistants and were recorded digitally. Race concordance was not achieved during these interviews due to personnel availability. Interviews lasted from 20 to 65 min. Breaks during the interviews were offered, but not utilized. Participants were reimbursed with a $30 gift card for participating in the qualitative interview and $25 gift card for completing the questionnaire.

Data Management and Analysis

Addresses were located after the interviews using Google Maps and matched with map-linked data provided by the FFIEC.¹¹ Census demographics with data about racial distribution for the census tract where each participant’s address was located were recorded. The qualitative interviews were recorded digitally and transcribed verbatim by a secure third party. Transcripts were reviewed by the research assistant who conducted the interview. A code book was developed that corresponded with the interview guide and using interview data. Double coding was performed (EDM and LM) for 48% of the transcripts to ensure consistent applications of the codes and to enhance data quality using a process described by Guest et al. [20] Any discrepancies in coding were discussed until agreement was reached.

NVIVO software [21] was used for coding and deductive analysis was used for each dimension of segregation. These codes were then analyzed across and within individual cases. The qualitative data were compared against the demographic data and segregation indices as provided by the Census. Matrices were created using qualitative and quantitative data. Within-case analysis was conducted to assess congruence or discordance between qualitative and quantitative data for each case.

Results

Sample Characteristics

The sample for this study consisted of 27 participants between the ages of 18 and 37 years. Women participated in qualitative interviews at 12–29 weeks gestation. The interviews were completed between April 2019 and November 2019. Nineteen participants (70%) reported being never married. Fourteen participants (52%) had an annual income of less than $10,000. Fourteen participants (52%) had a high school diploma or GED as the highest level of education. Table 2 reports sample characteristics.

Table 2.

Demographic characteristics of the participants (N=27)

Characteristics	n	Percent
Maternal age (years)
18–24	13	48.1
25–30	6	22.2
31–35	6	22.2
36–45	2	7.4
Annual household income
Less than $10,000	14	51.9
$10,000–$19,999	4	14.8
$20,000–$29,999	6	22.2
$30,000–$39,999	2	7.4
$40,000–$59,999	1	3.7
Relationship/habitation status
Divorced	1	3.7
Living with partner	4	14.8
Married	3	11.1
Never married	19	70.4
Education level
Less than high school	2	7.4
Graduated high school or GED	14	51.9
Vocational/technical training	1	3.7
Bachelor’s degree	1	3.7
Some college	9	33.3
Gestational age at interview
12 – 19 weeks	19	70
20 – 29 weeks	8	30

Racial Distribution

Using the same interview question “What is the racial mix of your neighborhood,” responses were compared with 2019 Census-based racial distribution for the participant’s Census tract. Fourteen of the 25 (56%, data missing for two) participants responded to this question with different information compared to the Census. A participant in Columbus lived in a Census tract that was 65% Black and 18% White. She described her neighborhood as being “Mainly all African American.” In Columbus, another participant lived in a Census Tract that was 35% Black and 52% White. She described the racial distribution of her neighborhood: “There’s not, I mean, there, there is…White, but just not a lot, I guess, compared to others like Black and stuff. Um, I guess it’s like probably…I don’t know. It’s not a lot of White people.” Another participant in Detroit lived in a Census tract that was 95% Black answered that her neighborhood was “mostly Whites.”

Exposure

The 2010 Census Isolation Index was 15% in Detroit and 47% in Columbus. This means that residents from Detroit have a 15% chance of contact with other races and residents from Columbus have a 47% chance of contact with other races. Exposure was measured qualitatively with an interview guide question about contact: “Thinking about your neighborhood, how often do you come into contact with people who aren’t Black?” Seventeen of 25 (68%, data missing for two) participants responded differently from the Isolation Index for their metropolitan area. Some participants responded to this question with a frequency of contact including “not often” or “every day,” or “hardly never.” Others were more detailed. A participant from Detroit described:

Um…not direct contact. Uh, I see other races, like we have a, um, a lot of White people over there, but I don’t necessarily come like as far as having conversation with. I might see them like during passing with either going to the store, or, um, they’re either driving or walking or something like that, but never any direct contact.

Another participant in Columbus gave examples of contact with the people she sees in her neighborhood:

I wouldn’t necessarily say come into contact with people that aren’t Black, I would say like I do see a couple Caucasian people, a couple Latino people—I’m a Latino also... across from us we have a Caucasian family. And then down the street is a Latino family. So we see them when they walking to the store, but everyone don’t talk, so they just walk past and no one speaks to no one.

Centralization

Centralization, or the distance to the city center, was explored using the question “How far, in minutes or miles do you live from the heart or center of your town?” The majority of the participants answered this question with a number of minutes or miles. Thirteen out of 22 (59%, data missing for five) participants responded that the distance from their city center was different from the mapped distance. For example, a participant from metropolitan Detroit lived about a 13-min drive away from the center of her city according to mapping software. She responded to the question about this distance: “To drive? Uh, probably like 45 minutes, if I’m going downtown, like 30, 45 minutes.” Others were less specific and said “Not that far” or “I’m pretty much the center of it.” A participant in Columbus lived about 9 miles away from the center of her city and responded, “Right down the street.” Another participant in Detroit lived about a 13-min drive from the city center and responded that she lived about a “10-15-minute drive” away from the city center.

Concentration

Concentration, or the distance to the closest predominantly White neighborhood, was explored with the question, “How far away is the nearest White neighborhood from where you live?” Again, the majority of the participants answered this question quantitatively with responses like “about seven or eight minutes maybe.” Thirteen of 23 (57%, data missing for four) participants felt that the distance to the nearest White neighborhood was different from the mapped distance. Four participants lived in predominately White neighborhoods. Nine of the 13 (69%) overestimated the distance to the nearest White neighborhood. Others gave answers that described their relationship to the nearest White neighborhood instead of the distance. A participant from Detroit who lived about 1.4 miles away from the nearest predominantly White Census tract said, “I don’t even know. I don’t remember the last time I seen a predominantly White neighborhood.” Another participant in Columbus who lived in a predominantly White Census tract answered, “Um, I don’t know. I don’t know too much about this neighborhood.” A participant in Detroit who lived about a 4-min drive away from the nearest White neighborhood according to the Census said that she lives “A 20-minute drive” away. Another participant in Columbus who lived in a predominantly Black neighborhood said, “I don’t even know how to answer that.”

Discussion

Metropolitan-wide level segregation indices have been used to examine the relationship between racial residential segregation and adverse health outcomes. [2, 3, 22] A qualitative perspective on racial residential segregation, as described here, provides complementary methodology for assessing this relationship on a deeper, more individualized level. To our knowledge, this is the first study to contrast qualitative descriptions of the dimensions of segregation with the long-standing measures created by Massey and Denton and used by many public health researchers. [3, 6, 23, 24] The participants included in this sample had perceptions about their neighborhood segregation and demographic composition that were often different from the data derived from existing residential US Census data. The differences between qualitative and quantitative measures and the possible reasons for the discordance, discussed below, suggest new approaches to measurement and new directions for the study of segregation and health.

The Census-derived segregation exposure measures movement necessary to achieve integration and potential contact between groups and contact and provides a metropolitan area-level assessment of segregation in the USA. Measurement at this population-level scale may explain the discordance between the participants’ qualitative perceptions and the exposure segregation index. These segregation indices may not resonate locally as reflected by the considerable lack of alignment in our sample, particularly for exposure. In contrast with segregation indices, sociodemographic data (e.g., racial distribution) are readily available at the smallest level from the US Census and provide the most localized comparison for each participant and a unit closer in scale to the “neighborhood.” Therefore, even when tied to a smaller and more local unit of geography, objective data is aligned with what is experienced and reported by the pregnant Black participants in our study across two cities.

We were able to assess centralization (where participants lived in relation to city centers) and concentration (where participants lived in relation to White neighborhoods) using mapping software with participants’ addresses. Qualitative descriptions of centralization and concentration, as opposed to the objective segregation indices, may offer insight into the beliefs of Black women about where they reside geographically. These two dimensions may also help us understand how racial residential segregation may increase risk adverse health outcomes. Many of the participants had very different views of concentration and centralization compared with data from mapping software and Census. For example, some participants (n=13, 59%) had different perspectives of how far they lived from the central downtown or from a White neighborhood. Women may limit how they access resources before, during, and after pregnancy across the wider community if they perceive the distances from their homes to city centers or White areas to be much farther away than is the case. For example, in Detroit, the majority of the major healthcare facilities are located in the city center. The potential gap between the perceived geographic distances between city centers and one’s neighborhood may discourage women from keeping prenatal appointments or asking for transportation from family or friend support systems, as the distance is perceived to be too remote.

Determining why the gap between the perceived and mapped distances exists may increase comprehension of how geography affects health and lead to more effective strategies for engagement among pregnant Black women. The participants who felt that the geographic distances were of a much greater distance away from the mapped distances may have other reasons, beyond perceived distance, to feel alienated from that physical space. A particular neighborhood may be perceived as more distant because of the ways that it was historically made inaccessible for certain groups due to housing inequality. Access to transportation, as noted above, may also influence these perceptions. Objective measures of proximity do not give insight into the individual’s relationship to that place. Examining both subjective and objective measures together draws a clearer picture of how pregnant Black women see themselves situated within those geographic contexts and provides a deeper, albeit beginning, understanding of their perspectives.

The individual-level personal perspectives reported here contribute new data about how these dimensions are experienced by people who live in neighborhoods in real-time. The differences between these descriptions of the experiences of living in racially segregated areas and Census measures illustrate the limitations of “objective” data. When Census data indicate that a person has a 15% chance of contact with someone from another group and people living in that area describe more frequent contact, it suggests a disconnect in these supposedly analogous constructs. The individual experiences described here did not always map onto the objective data for the community in which the women resided. Research that reports on relationships between dimensions of segregation and adverse health outcomes may be making assumptions about a construct that does not reflect the experience of individuals. [3, 22] For example, if a certain dimension of segregation is associated with increased or decreased rates of preterm birth, but the index that measures that dimension does not accurately reflect the experience of the pregnant women who experience preterm birth in that area, we must delve further into the meaning of these constructs. Location-based interventions aimed at this particular group of people would also benefit from this more in-depth information. New measures and qualitative interviews are needed.

Limitations

This study was not without limitations. Interviewers in this study were not matched with participants in terms of race. This may have resulted in participants being less forthcoming during their interviews. However, as the interviews were conducted over the phone, the partial anonymity (faces, not voices) could have helped the participants feel more open. More research would be needed to determine the full impact of the race discordant and phone interview processes. The interviews lasted from 20 to 60 min; thus, there was a great deal of variation in terms of depth.

The development of the segregation interview questions by one researcher, rather than using methods such as expert panels, focus groups, or interviews of community members, may also have limited the validity and reliability of the data. To our knowledge, this was the first qualitative study to interview participants using questions derived from Massey and Denton concepts of the dimensions of segregation. Overall, the interview guide questions appeared to work well based on the flow of the interview and the richness of responses. However, retrospectively, questions in the interview guide could have been more refined. For example, the city center was not defined when asking the centralization question. The researcher and/or interviewer may have considered one area to be central while the participant had another location in mind. Precise landmarks should be used in the future to determine actual distances. It also may help to know if the woman had a driver’s license; a lack of regular driving may in fact skew the woman’s perceptions of distance. Based on participant responses, some of the terms used for these questions could be clarified. For example, questions about “contact” could be refined to address specific forms of contact including talking or waving to neighbors or social support.

It could also be argued that some of the constructs we sought to compare are not measuring the same underlying construct and that the geographic unit may have constrained the comparisons. The segregation dimension exposure is measured over large geographic areas like cities and metropolitan areas. Relationships on the smaller scales that were explored in this study may not be appropriate to compare with the larger scale measures. Furthermore, the Isolation Index may not reflect street to street differences in Census tract data. [25] While the dimensions as described by Massey and Denton have been the standard since their inception, new measures for isolation have also been proposed. [26, 27] Further refinement of these new approaches to the qualitative measures of racial residential segregation would be helpful for future work.

Another segregation dimension that may have suffered “in translation” was concentration. For example, cities with long histories of racial residential segregation with stark contrasts in demographics on a street by street basis may have been considered by some participants to be off limits and therefore not on the “radar” as a nearby White area. [7] The women in the sample may also have different perceptions about what constitutes a “White neighborhood” because of demographic differences over time. What they may consider a predominately White neighborhood currently may be influenced by what they considered White when they were growing up. In addition, the data could be skewed by intergenerational perspectives of neighborhoods that were transmitted to the young women.

Conclusion

This is the first study to compare individual perspectives of the dimensions of segregation with those provided by the US Census. These perspectives among the participants were sometimes different when compared to the measures described by the Census. The comparisons between the two methods can help researchers better understand each concept; the methodology also offers a more robust understanding of the way that individuals perceive their environment and its relationship with health. When situated among women’s experiences, light will be shed on the importance of racial residential segregation and its impact on health outcomes. Future work in this area will contribute more when these segregation concepts are not treated as objective “factors” but rather the actual experiences from the voices of women who experience racial residential segregation in their daily lives.