Review of Religious Research

, 53:65

Measuring Religiousness Among Older African Americans: Exploring Race-of-Interviewer Effects

Authors

  • Christopher G. Ellison
    • University of Texas
    • Department of SociologyUniversity of Texas
  • Neal Krause
    • University of Michigan
Original Paper

DOI: 10.1007/s13644-011-0002-9

Cite this article as:
Ellison, C.G., McFarland, M.J. & Krause, N. Rev Relig Res (2011) 53: 65. doi:10.1007/s13644-011-0002-9

Abstract

In recent years a number of studies have explored possible sources of nonrandom error and response bias in survey data on religion. Building on a longstanding body of work in the social sciences, we examine a neglected issue in this domain: the potential for race-of-interviewer effects, specifically in African Americans’ self-reports of various dimensions of religiousness. After outlining two competing perspectives on this issue—which we term racial deference and racial solidarity—we test relevant hypotheses using data from the African American oversample of a nationwide study of older adults. Results indicate that older blacks tend to report higher levels of non-organizational religious practices and subjective religiousness when interviewed by whites. A number of implications and promising directions for future research are discussed.

Keywords

ReligionRaceMeasurementAfrican AmericansResponse biasRace-of-interviewer effects

Introduction

Sample surveys provide a vitally important source of data for the scientific study of religion. Although considerable care goes into developing concepts and crafting survey items, much less attention is given to wider social and contextual issues that may influence the way study participants respond to these items. Given widespread reliance on survey data, it is essential that scholars make every effort to “get it right,” i.e., to reduce nonrandom measurement error to the greatest extent possible.

Psychologists have long been interested in this issue, focusing on factors such as the tendency to provide socially desirable responses to survey items, or the tendency to respond indiscriminately in the affirmative to items tapping religiousness (Crowne and Marlowe 1960; Paulhus 1991). More recently, sociologists have turned their attention to possible measurement error and its implications. For example, Hadaway et al. (1993) set off a firestorm with papers that revealed the overreporting of church attendance, and subsequent studies identified a number of factors that may affect estimates of religious participation in the United States. Such influences include the mode of data collection; estimates may depend upon whether the information was gathered via face-to-face interviews, telephone surveys, or self-administered questionnaires (Presser and Stinson 1998; Woodberry 1998). In addition, religious attendance varies by seasonality and weather conditions (Iannaccone and Everton 2004), and in the short run religiousness may be affected by proximity to major events such as the 9/11 attacks (Uecker 2008). There is also evidence that respondents may reinterpret the meaning of survey items tapping religiousness (Hout and Greeley 1998), even seemingly unambiguous indicators of religiousness, such as frequency of attendance. Researchers have also investigated whether respondents’ tendencies to give socially desirable answers—along with other dispositional factors, such as conscientiousness or embarrassment—affect the empirical links between religion and the reporting of sensitive behaviors among adolescents (Regnerus and Smith 2005; Regnerus and Uecker 2007). Finally, Sherkat (2007) explored whether religion affects rates of cooperation and acquiescence in surveys, yielding a range of implications for inferences about religious effects on behaviors such as voting.

To our knowledge, however, there has been no significant attention to another key potential source of nonrandom measurement error in the study of religiousness: the race of the interviewer. This oversight is especially noteworthy because: (a) religious institutions, practices, and beliefs have played a central role in African American life, from the slave era through the present day (Lincoln and Mamiya 1990; Billingsley 1999); and (b) by virtually all conventional indicators, African Americans are more religious than whites (Taylor et al. 2004). These patterns are especially dramatic as older African Americans are likely the most religious segment of the population. Moreover, a substantial body of literature documents potent race-of-interviewer effects on political attitudes, particularly those involving attitudes toward whites, racial solidarity, and race-related policy preferences (Krysan and Couper 2003).

Our study contributes to the existing literature in several ways. We begin by outlining several theoretical reasons to anticipate race-of-interviewer effects on self-reported religiousness among older African Americans, focusing on two perspectives, which we term the racial deference and racial solidarity approaches. Relevant hypotheses are then tested using data from wave 1 of the Religion, Aging, and Health (hereafter RAH) project, a nationwide survey of older adults (ages 65 and over) in which African Americans were oversampled and in-person interviews were conducted (Krause 2002, 2008). After presenting the findings, we discuss their implications in terms of (a) future research at the nexus of race and religious life and (b) the use of survey data in the social scientific study of religion.

Theoretical and Empirical Background

The survey interview involves a process of interaction between interviewer and respondent. Answers to items may be a function not only of the questions themselves, but also the social context in which they exchange occurs, as well as a range of other psychosocial factors. Interviewer characteristics—such as experience, workload, and fatigue—may make a difference in the conduct and completion of interviews (Singer et al. 1983). Respondents’ ability to retrieve information from memory can also affect responses (Sudman et al. 1996). In addition, respondents may be more prone to give socially desirable answers (e.g., to erroneously report having voted in elections) when other persons besides the interviewer are present (Silver et al. 1986). Further, it has long been recognized that social factors such as the age and educational attainment of interviewer and respondent may influence response patterns, by shaping the rapport or distance between interviewer and respondent (e.g., Ehrlich and Riesman 1961).

The potential for nonrandom response error is especially acute when surveys deal with controversial topics, and respondents may alter their answers for reasons of impression management, or to maintain “polite conversation.” This issue has been explored by several researchers with regard to gender roles and gender equity issues (Kane and Macaulay 1993). For many of these issues, the gender of the interviewer does not appear to influence response patterns. At the same time, some studies suggest that both women and men are more inclined to endorse traditional gender role attitudes (e.g., regarding women’s labor force participation, childcare, and household division of labor) when interviewed by males, while both men and women tend to express greater support of women’s collective mobilization around issues of gender inequality to female interviewers, as compared to their male counterparts (Landis et al. 1973; Lueptow et al. 1990; Kane and Macaulay 1993).

However, the most durable line of research on nonrandom response error has centered on race-of-interviewer effects (Hatchett and Schuman 1975; Schaeffer 1980; Cotter et al. 1982; Tucker 1983). A vigorous research literature has probed the influence of race-of-interviewer effects for responses to a wide range of political and racial survey items. Taken together, studies indicate that African Americans are questioned by same-race interviewers, they tend to exhibit higher levels of group consciousness, skepticism of (or animosity toward) the sociopolitical status quo, and support for collective mobilization (e.g., backing the Rev. Jesse Jackson’s presidential campaign, starting a black political party). By contrast, those queried by whites tend to provide less liberal or militant answers to questions dealing with discrimination, feelings about whites, racial solidarity, and policy preferences involving racially-charged issues (Anderson et al. 1986, 1988a, b; Davis 1997a; Krysan and Couper 2003). Moreover, despite significant social change, such race-or-interviewer effects have a long legacy, persisting in survey-based studies for the past 50–60 years (Hyman et al. 1954; Krysan and Couper 2003).

Racial Deference

Early investigators found no significant race-of-interviewer effects on blacks’ responses to questions on topics that are non-political and lacked racial salience (Schuman and Converse 1971), and consequently systematic research on such issues ceased by the 1970s. Our goal is to reopen this line of inquiry, focusing on multiple facets of religiousness. Why might we expect to find race-of-interviewer effects on older African Americans’ reports of religiousness? One perspective in this literature focuses on what we might term racial deference (Davis 1997a, b). Briefly, this viewpoint acknowledges persistent tension and uncertainty in interactions between African Americans and whites, resulting largely from the long history and persistence of racial bigotry, demeaning stereotypes, and discrimination on the part of whites (e.g., Sniderman and Piazza 1993). Trend analyses reveal that, although whites’ attitudes toward African Americans have shifted over time, this reflects mainly processes of cohort succession (or replacement), with only limited evidence of individual-level (or intra-cohort) change in whites’ racial attitudes (e.g., Firebaugh and Davis 1988; Kluegel 1990). This is important in a study of older African Americans for two reasons: (a) it is older blacks who are likely to have experienced the cumulative effects of discrimination and racism; and (b) it is their contemporaries, i.e., older whites, who even today are likely to harbor the most negative views of persons of color (Davis et al. 2008).1 The deference perspective suggests that such baggage may well come into play when African Americans sit down with white interviewers. Thus, mixed-race interviews may be experienced as stressful, with expectations of awkwardness and possible (even if unintended) insensitivity. Studies reveal that experiences of perceived racial discrimination can evoke feelings of distress among African American adults (Bierman 2006; Ellison et al. 2008).

According to the racial deference perspective, African Americans tend to follow certain scripts in order to reduce tensions in their contacts with white strangers, and to get through the interview. As Davis (1997b) describes this process:

In its most basic form, the race of interviewer effects become a form of sophisticated “masking” in which African Americans become competent actors with an acute sense of what might satisfy the interviewer. Similar to other defensive postures, this scripted behavior among African Americans is fabricated to defend against the fear of total disintegration and loss of self (Davis 1997b, pp. 311–312).

Although Davis (1997a, b) connects this behavior to African Americans’ reports of political attitudes and activities,2 we suggest that it may also have relevance for the reporting of religiousness and religious attitudes. This may work in two ways. First, to counter widespread negative images of African Americans among whites (as lazy, dishonest, etc.; Sniderman and Piazza 1993), exaggerating their levels of religiousness may offer one approach to putting white interviewers at ease. In this way, African Americans may adjust their answers in order to be perceived as less threatening, more worthy, and to demonstrate their own “bona fides” as decent, honorable, good people. Certain facets of religiousness may offer especially promising currency in this regard. Indeed, survey interviewers have rated respondents with high personal or intrinsic religiousness as “nicer,” friendlier, and more cooperative than others (Morgan 1983; Ellison 1992; Brennan and London 2001). It is noteworthy that such patterns did not surface with regard to organizational religious practices, such as church attendance or other congregational participation, which may be more susceptible to extrinsic motivations, such as securing access to services, social networking, or status-seeking.

A second possible manifestation of “masking” could surface in older African Americans’ responses to questions at the nexus of race and religion. Briefly, the Black Church has occupied a singular place in the African American struggle for racial equality (Lincoln and Mamiya 1990; Billingsley 1999). As an institution, the Church provided organizational resources and symbolic leadership during the civil rights movement, a pattern which continues to the present (Harris 1999; McDaniel 2008). Moreover, mainstream African American theology that is (a) focused on liberation, linking the collective struggles of African Americans with scriptural narratives of freedom and justice, and (b) geared toward coping with the social and emotional consequences of marginality and exclusion. Awareness of these dual strands of the African American faith tradition may be particularly acute among older African Americans, many of whom may have been socialized during the heyday of the civil rights era (Ellison 1991). Previous research conducted from the deference perspective suggests that African Americans mute their authentic voice on matters of race and politics (Davis 1997a, b). With regard to the role of religion in race relations, older African Americans interviewed by whites may tend to emphasize more hopeful and conciliatory themes, while avoiding issues that could elicit negative feelings or exacerbate tension during the interview process. Thus, when queried by whites, African Americans may be more inclined to acknowledge racially unifying messages of the Christian faith, without stressing the importance of the Black Church in overcoming the sordid history of discrimination and exclusion that has been perpetrated by whites. Taken together, the arguments that make up this racial deference perspective suggest two study hypotheses:

H1

To the extent that the masking process is at work, older African Americans who are interviewed by whites will tend to report higher levels of non-organizational religiousness (e.g., prayer, other devotional practices) and subjective religiousness (e.g., intrinsic motivation, feelings of closeness to God)—but not necessarily organizational religious involvement (e.g., religious attendance)—than their counterparts who were interviewed by persons of their own race.

H2

Older African Americans interviewed by whites about religion and race relations may give less conflictual, more universalistic answers about the role of religion, e.g., emphasizing that all persons are spiritually equal regardless of skin color, as compared with those who are interviewed by persons of their own race.

Racial Solidarity

It is also possible that nonrandom measurement bias emerges when African Americans feel pressured by interviewers of the same race (Davis 1997a). Instead of providing excessively deferential answers to racially salient questions posed by white interviewers, African Americans may feel the need to exaggerate certain behaviors and attitudes in their interactions with members of their own race. This may be perceived as necessary in order to gain or maintain respect and to avoid being seen as insufficiently committed to African American culture, norms, or collective experiences of struggle. Numerous studies over the years find that African Americans report higher levels of racial solidarity, greater awareness of discrimination, more critical views of whites, and stronger preferences for policies such as racial preferences when interviewed by fellow African Americans than by whites (Anderson and Silver 1986; Anderson et al. 1988a, b; Krysan and Couper 2003). Such desires, conscious or unconscious, to express group solidarity could conceivably influence older African Americans’ self-reporting of religious involvement as well.

This may be the case for at least two reasons. First, in many areas, few other social organizations besides religious congregations existed. Therefore, church membership and attendance became important social norms (Ellison and Sherkat 1995; Hunt and Hunt 2000). Regular participation was an emblem of personal respectability and community standing, an expression of solidarity, and a gateway to other forms of social participation (Lewis 1955; Ross and Wheeler 1971). Individuals who did not join churches and attend services at least sporadically were often judged to be outside the mainstream of the African American community. A number of twentieth-century ethnographic studies conducted in the South concluded that there were social rewards for compliance with social norms of religious involvement, and sanctions on non-compliance (Johnson 1941; Lewis 1955). This line of argument has led some observers to characterize the Black Church as a “semi-involuntary institution,” particularly in the South (Nelsen et al. 1971; Ellison and Sherkat 1995).3

Further, the Black Church has played a vital role in the African American experience throughout history, due to its multi-functionality and symbolic centrality (Lincoln and Mamiya 1990). For much of African American history, religious institutions were arguably the only social sphere organized and operated by and for African Americans, largely beyond the control of whites (Billingsley 1999; Taylor et al. 2004). They stood at the center of numerous creative approaches to collective self-help and community uplift, from the earliest forms of burial insurance to the extension of educational opportunities to African Americans (Lincoln and Mamiya 1990). The Church has afforded a training ground for generations of local and national political leaders (e.g., Dr. Martin Luther King, Jr.), in addition to much of the spiritual and organizational energy for the civil rights movement (Morris 1996; Harris 1999) and subsequent political struggles at the local and national levels.

Taken together, these strands of thought suggest two additional study hypotheses:

H3

Older African Americans interviewed by persons of the same race will tend to report higher levels of organizational religious involvement (e.g., church attendance, congregational participation) than blacks interviewed by whites.

H4

Older African Americans interviewed by blacks about religion and race relations will tend to offer less accomodationist responses, e.g., emphasizing the distinctive role of the Black Church in dealing with discrimination and oppression perpetrated by whites, as compared with their counterparts interviewed by whites.

Data

The data that are analyzed below come from the RAH Survey (Krause 2008).4 The study population was defined as all household residents who are either African American or White, English-speaking, and 66 years of age or older at the time of the baseline survey. Geographically, the study population is restricted to eligible residents residing in the coterminous United States; residents of Alaska and Hawaii were excluded from this population due to the high cost of conducting in-person interviews, coupled with the comparatively low representation of older adults who are African American and/or Christian, in these states. Finally, the study was restricted to individuals who were either currently practicing Christians, people who were Christian in the past but no longer practice any religion, and individuals who were not affiliated with any faith at any point in their lifetime. The goal of the RAH Survey was to devise and test a comprehensive battery of items to assess religion in late life. Older people who were practicing a religion other than Christianity (e.g., Jews or Muslims) were excluded because it would have been too difficult to devise a comprehensive set of religion measures that would be suitable for persons of all faiths (Krause 2002).

The sampling frame for the RAH Survey consisted of all eligible persons contained in the Health Care Finance Administration (HCFA) Medicare Beneficiary Eligibility List (HCFA is now called the Center for Medicare and Medicaid Services—CMS). This list contains the name, address, sex, and race of virtually every older person in the United States. It should be emphasized that people are included in this list even if they are not receiving Social Security benefits. Even so, some older individuals are not included in this data base because they do not have a Social Security number. This may be due to factors such as undocumented immigration.

A five-step process was used to draw the sample. First, each year, researchers at HCFA draw a 5% sample of the names in their master file using a simple random sampling procedure. The sampled names include individuals who are at least 65 years of age. However, by the time the field period for the RAH Survey began, subjects in the 5% file were at least 66 years of age. It is for this reason that the study population was defined as including individuals who were 66 years of age or older. In the second step of the sampling procedure, the 5% file was split into two subfiles—one containing older Whites and one containing older Blacks. Each file was sorted separately by county, and then by ZIP codes within each county. In the third step, an nth interval was calculated for each subfile based on the total number of eligible records. Following a random start, 75 nth selections were made in each file. In the fourth step of the sampling strategy, primary sampling units (PSUs) were formed by selecting approximately 25 additional names above and 25 additional names below each case identified in step three. Finally, in the last stage, sampled persons within each PSU were recruited for an interview with the goal of obtaining approximately ten completed interviews per PSU.

Interviewing for the Wave 1 survey took place in 2001. A total of 1,500 interviews were completed successfully. Older African Americans were oversampled so that sufficient statistical power would be available for exploring race differences in religion. More specifically, the final sample contained 748 older Whites and 752 older Black people. The overall response rate for the baseline survey was 62%.

Measures

It has long been recognized that religion is a complex and multidimensional phenomenon (Stark and Glock 1968; Idler et al. 2003). Although researchers remain far from agreement over the most appropriate strategies for conceptualizing and measuring religion, one popular approach—developed specifically with African Americans’ religious involvement in mind—distinguishes between organizational, non-organizational, and subjective religiousness (Levin et al. 1995). We loosely follow this approach in the analyses that follows. Specifically, we analyze scales measuring organizational religious practices, congregational activism, and non-organizational religious practices, as well as two factes of subjective religiousness: intrinsic religious motivation and closeness to God. In addition, the RAH includes a series of unique items posed only to African American respondents (Krause 2002). These items inquire about the connections between faith commitments and racial attitudes, and about the past and current role of the Black Church in African American life (Krause 2004).

Religious Measures

The RAH dataset includes a number of items tapping individuals’ involvement in religious groups. Based on the results of a principal components factor analysis, we created two scales.5 Organizational religious practices are measured in terms of the frequency with which individuals (a) attend religious services; (b) attend adult Sunday School of Bible study groups; and (c) participate in prayer groups that are not part of regular worship. For each of these items, response categories range from never (1) to several times a week (9). Our measure of organizational religious practices is the mean response to these three items (alpha = 0.78). Congregational activism is measured via responses to three items, tapping (a) the frequency with which respondents “do jobs around the church like preparing meals or doing yard work”; (b) the frequency with which they “spend time working in programs to help people in need at their church”; and (c) the number of leadership roles they occupy in their church (e.g., deacon, elder, lay pastor, church mother, choir director, chair of church committee, Sunday school teacher). Responses to (a) and (b) range from never (1) to more than once a week (6), while the third component is the sum of affirmative responses to individual items about church roles, ranging from 0 to 6. Our measure is the mean of these three items (alpha = 0.66).

Non-organizational religious practices are measured in several items, developed via the procedures described by Krause (2002). These specific items measure how often respondents: (a) pray alone; (b) watch or listen to religious talk shows or shows that report news from a Christian perspective; (c) read the Bible at home; (d) read religious literature other than the Bible at home; and (e) listen to religious music outside church. Response categories for each item range from never (1) to more than once a day (8). Individuals lacking valid responses to at least three of these items were dropped from the analyses. Our measure is the mean score on these items (alpha = 0.72).

To gauge subjective facets of religiousness, we consider scales tapping respondents’ feelings of closeness to God and intrinsic religious motivation. Feelings of closeness to God are measured via levels of agreement with the following three statements: (a) I have a close personal relationship with God. (b) I feel that God is right here with me in everyday life. (c) When I talk to God, I know he listens to me. Response categories range from strongly disagree (1) to strongly agree (4). These items were developed via the procedure described by Krause (2002). Our measure of perceived closeness to God is the mean score on these items; persons who are missing valid responses to more than one of the items are dropped from the analyses (alpha = 0.92).

Our measure of intrinsic religiousness is the respondents’ level of agreement with the following three items: (a) My faith shapes how I think and act each and every day. (b) I try hard to carry my religious beliefs over into all my other dealings in life. (c) My religious beliefs are what lie behind my whole approach to life. Responses range from strongly disagree (1) to strongly agree (4); individuals missing valid responses to more than one of these items are dropped from the analyses (alpha = 0.91). These items are the product of a working group convened by the Fetzer Institute and the National Institute on Aging (Idler et al. 2003).

Attitudes Regarding Religion and Race Relations

In addition to items tapping organizational and non-organizational religious behaviors and subjective facets of religiousness, the RAH data also includes information on African Americans’ views about the nexus of race and religion, based on a unique set of seven items developed for this purpose (Krause 2004). Based on results of a principal components factor analysis, which yielded a two-factor solution,6 we distinguish between items tapping a critical perspective on religion and race relations, and those suggesting a more deferential or conciliatory posture. Critical views are tapped via agreement with the following statements: (a) The church has sustained black people in the face of racial injustice. (b) The church is something black people can truly call their own. (c) I find strength knowing the church sustained blacks during slavery. (d) When I experience racial prejudice or racial discrimination, it is comforting to know that God is right there. (e) The church provides hope that the struggles of black people will someday be over. Responses range from strongly disagree (1) to strongly agree (4). Persons lacking valid responses to at least three of these items were dropped from the analyses. Our measure is the mean score on these items (alpha = 0.84). More deferential or conciliatory views on religion and race are measured with responses to the following two items: (a) God makes no distinction between black and white people. (b) My faith helps me to see how people of all races are equal in the sight of the Lord. Here again, responses range from strongly disagree (1) to strongly agree (4), and our measure is the mean score on these two items (alpha = 0.89, r = 0.80).

Race of Interviewer

For the RAH project, personnel trained and supervised by the Harris Interactive organization conducted face-to-face interviews in the homes of the respondents. Approximately 37% of African Americans in the sample were interviewed by persons of their own race; the remainders were interviewed by whites. Although black and white interviewers were not assigned at random by Harris Interactive, there is no clear reason to anticipate bias in assignment of interviewers of different races.7

Demographic Control Variables

Our multivariate analyses include controls for the following factors: gender of respondent (1 = female, 0 = male); age (measured in years); education (years of formal school completed); marital status (1 = currently married, 0 = all others); region of residence (series of eight dummy variables). Our analyses also control for respondents’ self-rated health, ranging from excellent (1) to poor (5). Finally, financial strain is gauged via responses to the following question: “How much difficulty do you have inmeeting the monthly payments on you/your family’s bills?” Responses range from none (1) to a great deal (4). We employ this measure instead of a conventional indicator of household income because the latter may understate socioeconomic status among older adults, many of whom are no longer in the labor market, and who may or may not report earnings from Social Security, investments or pensions as income.

Results

Table 1 presents descriptive statistics on the variables used in our analyses. According to the figures on our religious measures, the average older African American in the RAH sample participates on organizational religious pursuits several times per year (4.54 on a 1–9 scale), participates in congregational activism once a month or less (1.34 on a 1–6 scale), and engages in non-organizational or devotional activities a few times per month (5.35 on a 1–8 scale). On average, respondents have relatively high scores on the two facets of subjective religiousness, intrinsic religious motivation (3.48 on a 1–4 scale) and perceived closeness to God (3.68 on a 1–4 scale). With respect to the nexus of religion and race relations, respondents tend to agree more strongly with the conciliatory or deferential items (3.57 on a 1–4 scale) than the more critical views (3.18 on a 1–4 scale).
Table 1

Descriptive statistics

 

N

Mean

SD

Range

White interviewer

734

0.63

0–1

Organizational religious practices

733

4.54

2.31

1–8

Congregational activism

602

1.34

0.89

0–5.67

Non-organizational religious practices

733

5.35

1.45

1–8

Intrinsic religious motivation

723

3.48

0.53

1–4

Closeness to God

729

3.68

0.45

1–4

Critical themes re: religion and race

716

3.18

0.57

1–4

Deferential themes re: religion and race

721

3.57

0.52

1–4

Female

734

0.65

0–1

Age

725

74.91

6.49

66–100

Education

712

10.05

3.35

2–22

Married

734

0.35

--

0–1

Financial strain

713

2.04

1.05

1–4

Self-rated health

733

2.58

0.85

1–4

New England

734

0.01

0–1

Mid-Atlantic

734

0.24

0–1

South Atlantic

734

0.36

0–1

Interior South

734

0.15

0–1

North Central

734

0.15

0–1

Upper Midwest

734

0.03

0–1

Mountain West

734

0.01

0–1

Pacific

734

0.06

0–1

In terms of demographic characteristics, the average older black respondent in the RAH study is approximately 75 years old, with 10 years of formal education completed. Nearly two-thirds (65%) of these respondents are female, and 35% are married. The regional patterns roughly parallel those of African Americans nationwide, with concentrations in the South Atlantic (36%), Mid-Atlantic (24%), interior southern (15%), and North Central (15%) regions. Finally, the average older black RAH respondent rates his/her physical health midway between “fair” and “good” (2.58 on a 1–4 scale) and reports “only a little” trouble meeting monthly financial obligations (2.04 on a 1–4 scale).

Next we estimate the net effects of the race of interviewer and covariates on each of the religious measures discussed earlier. Key findings, presented in Table 2, are quite straightforward. As anticipated by H1, older African Americans interviewed by whites report significantly higher levels of non-organizational religious practices (b = 0.345, p < 0.01), and higher levels of intrinsic religious motivation (b = 0.163, p < 0.001) and closeness to God (b = 113, p < 0.01), than their counterparts who were interviewed by African Americans. Consistent with H2, those older blacks who were interviewed by whites also express stronger support for conciliatory views regarding religion and race relations (b = 0.193, p < 0.001), as compared with those interviewed by fellow African Americans. These results are highly compatible with our expectation that African American elders would “don the mask” (in Davis’ [1997b] words) and present a more deferential face in order to reduce anxiety and defuse tensions in the interactions with white interviewers. By contrast, race of interviewer was unrelated to the frequency of organizational religious practices (b = 0.148, ns) or congregational activism (b = 0.034, ns), a finding that runs directly counter to H3. Finally, there is no association between the race of the interviewer and support for critical perspectives on the nexus of religion and race relations (b = 0.052, ns), contrary to H4.
Table 2

Race-of-interviewer on various dimensions of religiousness among older African Americans

 

Organizational religiosity

Congregational activism

Non-org religiosity

Closeness to God

Intrinsic religiosity

Deferential views

Critical views

White interviewer

0.148 (0.179)

0.034 (0.080)

0.345** (0.109)

0.113** (0.036)

0.165*** (0.042)

0.193*** (0.042)

0.052 (0.046)

Female

0.780*** (0.191)

−0.075 (0.087)

0.661*** (0.116)

0.174*** (0.038)

0.078+ (0.044)

0.093* (0.044)

0.086+ (0.049)

Age

−0.039** (0.014)

−0.019** (0.006)

−0.017* (0.008)

0.001 (0.028)

−0.001 (0.003)

−0.005 (0.003)

0.003 (0.004)

Education

0.060* (0.028)

0.009 (0.012)

0.049* (0.017)

0.008 (0.006)

0.007 (0.006)

0.001 (0.006)

−0.006 (0.007)

Financial strain

−0.002 (0.084)

−0.024 (0.037)

0.033 (0.051)

0.016 (0.017)

−0.007 (0.020)

−0.018 (0.020)

−0.015 (0.021)

Self-rated health

−0.434*** (0.104)

−0.162*** (0.047)

−0.194*** (0.063)

−0.040+ (0.021)

−0.034 (0.024)

−0.025 (0.024)

−0.040 (0.026)

Married

0.249 (0.190)

0.079 (0.084)

0.143 (0.116)

0.044 (0.038)

−0.019 (0.044)

−0.022 (0.044)

−0.093+ (0.049)

New England

−1.420 (2.214)

−0.839 (0.890)

−0.575 (1.348)

−0.048 (0.440)

−0.438 (0.512)

−0.542 (0.510)

0.085 (0.566)

Mid-Atlantic

−0.363 (0.226)

0.011 (0.082)

0.218 (0.138)

−0.016 (0.045)

−0.026 (0.052)

−0.044 (0.053)

0.006 (0.058)

Interior South

0.250 (0.258)

−0.009 (0.112)

0.227 (0.157)

0.010 (0.051)

0.133* (0.060)

0.155** (0.059)

0.251*** (0.066)

North Central

−0.178 (0.262)

−0.048 (0.114)

0.085 (0.160)

−0.071 (0.052)

−0.077 (0.061)

−0.090 (0.061)

−0.148* (0.068)

Upper Midwest

1.517** (0.542)

0.039 (0.225)

1.252*** (0.330)

0.179+ (0.108)

0.199 (0.125)

0.066 (0.125)

0.115 (0.139)

Mountain West

−0.460 (0.723)

−0.266 (0.324)

−0.546 (0.440)

−0.049 (0.144)

0.242 (0.167)

−0.347* (0.167)

0.068 (0.185)

Pacific

−0.787* (0.388)

−0.108 (0.166)

0.139 (0.237)

0.012 (0.079)

−0.003 (0.092)

−0.097 (0.091)

0.135 (0.101)

Adjusted R2

0.090

0.032

0.114

0.048

0.045

0.061

0.050

*** p < 0.001; ** p < 0.01; *p < 0.05, +p < 0.10

Supplementary Analyses

In ancillary analyses (not shown), we examined the possibility that race-of-interviewer effects may be stronger among certain segments of the older African American sample than among others. Specifically, we added a series of multiplicative interaction terms to the models displayed in Table 2, zero-centering continuous variables to minimize collinearity between raw and product terms (Aiken and West 1991). First, we considered the possibility that deferential tendencies may be more common among lower-SES respondents, who may have spent much of their lives in relatively subordinate social positions vis-a-vis whites. Because education is often regarded as a better indicator of SES among African Americans than income, due to the long history of employment discrimination against blacks, especially black men (e.g., Smith 1989), we added a race-of-interviewer × education interaction term to the models of each religious outcome. No significant findings emerged.

Second, it is entirely possible that the racial deference perspective outlined earlier is primarily relevant to African Americans who were socialized during the Jim Crow and civil rights eras, and the perceived need to engage in such behaviors may not pertain to younger blacks raised in the post-civil rights era. The RAH data do not allow us to test this hypothesis directly. However, we can explore whether there are age-related variations in race-of-interviewer effects within the RAH sample, by adding a race-of-interviewer × age interaction term to models of each religious outcome. Once again, no evidence of contingent effects surfaced.

Third, it is possible that, on average, African American men have more experience and skill in “donning the black mask” because they have long borne the brunt of negative stereotypes (e.g., violent or threatening, unintelligent) and interactions between black men and white women—who make up the majority of the interviewer pool for the RAH project—have long held particular salience and meaning in US race relations. To investigate possible gendered patterns, we added race-of-interviewer × gender interaction to models of each religious outcome. Significant gender interactions emerged for two outcomes: intrinsic religious motivation (p < 0.05) and non-organizational religious practices (p < 0.05). In each case, findings revealed that race-of-interviewer effects, interpreted here as support for the racial deference perspective, were stronger among older black men as compared with older black women.

Discussion

Our study has contributed to the recent growth of interest in potential response bias in the measurement of religiousness in survey data (Hadaway et al. 1993; Regnerus and Smith 2005; Regnerus and Uecker 2007; Sherkat 2007; Uecker 2008). Although various interviewer and respondent characteristics and other contextual factors have been considered, we are aware of no significant attention to possible race-of-interviewer effects on self-reports of religiousness among African Americans. As we noted earlier, this oversight is noteworthy given the exceptionally high levels of religiousness among older blacks, the distinctive role of religious institutions and practices throughout African American history, and the significant literature documenting race-of-interviewer effects on other, secular attitudes and behaviors. Most previous work on race-of-interviewer effects has centered on African Americans’ views on racial and political issues, such as reports of discrimination, attitudes toward whites and other groups, racial solidarity, and political activity (Hyman et al. 1954; Schaeffer 1980; Cotter et al. 1982; Tucker 1983; Anderson and Silver 1986; Anderson et al. 1988a, b; Davis 1997a, b; Krysan and Couper 2003).

This study extends the literature on race-of-interviewer effects by focusing on a neglected outcome, religion, in a sample of older blacks. We outlined two sets of arguments—termed racial deference and racial solidarity—each of which suggests reasons why African Americans’ reporting of some (but not all) specific facets of religiousness may be affected by the race of the interviewer. Overall, our findings are broadly consistent with the racial deference perspective. Although this perspective was common when race-of-interviewer effects first surfaced in surveys in the 1950s (e.g., Hyman et al. 1954), its popularity waned for several decades. Notions of racial deference may well have seemed implausible and obsolete to researchers during the years of civil rights mobilization, and also may have run against the grain of racial sensibilities during the 1970s and 1980s. However, the racial deference perspective has enjoyed a limited revival in recent years, due to the theoretical arguments and empirical findings of Davis (1997a, b). Briefly, the deference argument holds that due to their cumulative experiences with prejudice and discrimination, African Americans are (perhaps wisely) wary and cautious in their dealings with white strangers, including survey interviewers. To minimize anxiety and uncertainty, and to get through the interview, they may be inclined to give responses that reduce tension and put white interviewers at ease, while avoiding answers that may be offensive or confrontational. Davis (1997b) calls this process “donning the black mask.”

In our data, older African Americans interviewed by whites consistently report higher levels of non-organizational religious behaviors (e.g., private prayer, listening to religious music, reading religious literature) and they also score higher on several subjective aspects of religiousness (e.g., intrinsic motivation, feelings of closeness to God), compared to older African Americans interviewed by persons of their same race. Further, those interviewed by whites score significantly higher on items tapping hopeful or non-confrontational views of the role of religion in race relations. Such patterns do not surface with regard to measures of organizational religious practices (e.g., church attendance or other congregational participation), or on items tapping more critical or potentially confrontational beliefs about the role of religion in African American life. Thus, it seems that older African Americans interviewed by whites may be especially inclined to show that they are worthy, upright, and non-threatening by: (a) emphasizing their personal piety, close relationship with God, and authentic, internalized spirituality; and (b) emphasizing God’s love for all people while not dwelling on potentially controversial themes, such as the persistence of racism and discrimination, or the distinctive character of African American religious institutions and practices.

As Davis (1997a, b) has pointed out, African American survey respondents may also feel pressured by black interviewers. In particular, these respondents may be at pains not to appear to be “sell-outs,” or to be uncommitted to the collective struggles of African Americans. Building on this logic, we outlined an alternative, racial solidarity perspective. Briefly, the Black Church has played a distinctive role throughout African American history and particularly in the civil rights movement and political mobilization for racial justice. If the racial solidarity perspective is accurate, then older African Americans might have been prone to exaggerate their involvement in religious organizations, through attending services regularly, participating in Sunday School or Bible study groups, and engaging in other service or volunteer work around the church. To the contrary, however, there is a consistent pattern of null effects in this area. Nor do we find any evidence that older African Americans with same-race interviewers expressed greater sympathy for the statements about the unique role or exclusivity of the Black Church, or the ongoing function of African American religion as a balm when coping with racism, than their counterparts who were interviewed by whites.

Although self-reports of certain behaviors (e.g., voting) can be compared with validated records to establish their accuracy (e.g., Anderson and Silver 1986), this is obviously not feasible with attitudes or subjective perceptions. Thus, as Schuman and Converse (1971) observed during the heyday of this line of research, it may be useful to regard evidence of race-of-interviewer effects partly as a barometer of racial attitudes rather than assuming that one set of responses is entirely “true” and the other is totally “false” (Krysan and Couper 2003). In this case, our results are consistent with two main conclusions: First, significant numbers of older African Americans are wary and may experience tension when dealing with white strangers. This is likely to reflect their own experiences, and their keen sensitivity to our difficult history—and continued challenges—with respect to racial equality in the US. Second, older black survey respondents may perceive private religious piety and internalized religious values as signifiers of decency and moral character, as indicators of one’s bona fides as a person of good will. This also underscores to the importance of religiousness and spirituality in the individual lives and cultural identities of many older blacks (Ellison 1992; Taylor et al. 2004).

This study is characterized by several limitations. First, like most studies of race-of-interviewer effects (e.g., Anderson et al. 1988a, b); we use data that were collected mainly for a different purpose. Therefore, interviewers were assigned largely based on logistical considerations, and were not assigned completely at random, as would be the case in an experimental study. Although there is no clear reason to anticipate bias in the assignment of black versus white interviewers, and the inclusion of standard control variables (including region) makes little differences in most of our major findings, it would be desirable to conduct this work using a pure experimental design. Second, this study has centered on a particular cohort of older African Americans. Respondents are ages 66 and over, with a mean age of approximately 75. Thus, it is possible that the racial deference perspective is especially germane to this group because of (a) the distinctively harsh racial attitudes of their white contemporaries, and (b) the fact that they have experienced a lifetime of cumulative prejudice and discrimination in their dealings with whites. Future studies are needed to explore variations in these patterns of race-of-interviewer effects on self-reported religiousness across various cohorts of African Americans.

Third, the relatively small sample size prevented us from carefully considering subgroup variations in race-of-interviewer effects.8 There are hints in these data that possible response bias was greater on the part of older African American men interviewed by white women. This pattern makes sense in light of the historical sensitivity of relations between black men and white women, and the views of whites from older generations (especially visible in the South) regarding racial threat, the sexualization of black males, and miscegenation. Indeed, some of the most vicious acts of cruelty perpetrated by whites against African American males centered on the perceived need to protect white women. Although the possibility of a gendered pattern of response bias would be fascinating to explore in greater detail, the vast majority (90%) of RAH interviewers were female, thus limiting the statistical power of analyses aimed at clarifying this issue. A sample with a larger number and/or higher proportion of male interviewers would be needed in order to investigate possible gendered patterns of racial deference. The same issue of statistical power also precludes a more detailed examination of regional differences (see footnote 3). Fourth, although it lies well beyond the scope of the present paper, the next logical step in this agenda is to examine what (if any) substantive difference race-of-interviewer effects make in studies linking religiousness with various outcome of interest. While the analyses provided here suggest that response bias may affect the mean level of certain facets of religiousness among older African Americans, it does not necessarily follow that the relationship between religiousness and measures of other study outcomes are biased. Does taking this possible source of bias into account weaken or strengthen our estimates of the role of religiousness in mental and physical health, racial attitudes and political orientations, civic engagement, and other crucial domains of individual and collective experience among African Americans? The following example suggests the potential importance of the findings reported here: To the extent that many surveys employ significant numbers of white interviews to query African Americans, they may overstate the private and subjective religiousness (e.g., frequency of prayer, intrinsic motivation, closeness to God) within this population. As a consequence, studies could underestimate association between private religiousness and well-being, because the variance of the former could be artificially reduced due to a ceiling effect.

Despite these limitations, we believe this study has extended our knowledge of response bias in survey-based self-reports of religiousness. It appears that some questions about religion may be more sensitive and contextually-influenced, at least among older African Americans, than one might have assumed, and that it also matters who is asking those questions. The phenomenon of response bias in this area may be more complex than we have previously recognized. Nevertheless, given the widespread use of sample surveys in the scientific study of religion, it is vitally important to gain a better grasp of the influence that dynamics of the interview process may have on the reporting of religious behaviors, beliefs, and experiences. Further research along the lines discussed above can enhance our understanding of these neglected issues.

Footnotes
1

To explore this issue further, we conducted a brief analysis of GSS data from 2002, the year closest to that of the RAH survey. We compared the racial attitudes of whites ages 65 and over with those of their younger counterparts. The percentage of these persons indicating they would oppose or strongly oppose a close family member marrying a black person was: 58.5% among those 65 and over, as compared with 48.1% (55–64), 31.9% (45–54), 23.5% (35–44), and 18.1% (34 or younger). Other items were examined, including whites willingness to live in close proximity to African Americans, and whites stereotyping of African Americans (i.e., lazy, unintelligent). The results showed there were substantial age/cohort differences with those 65+ holding the most prejudiced views. Such patterns suggest that older African Americans have ample reason to be wary in their dealings with white strangers, given the persistent prejudice on the part of their white counterparts.

 
2

This line of argument is reminiscent of earlier arguments by Carr (1971), who argued that rates of acquiescence by lower-SES blacks to some survey items, such as those developed by Srole to tap “anomie,” are disproportionately high among lower-SES African Americans as compared with other social groups. Citing Ralph Ellison’s (1947) Invisible Man, Carr (1971) concluded that such acquiescence (or yes-saying) “has proved to be, historically, an effective rational tactic for survival” (p. 291) and that “acquiescence can be seen more as a tactic used in social interaction rather than a deep-rooted personality syndrome” (p. 292).

 
3

Although most of the work on the “semi-involuntary institution” has centered on the Black Church as an institutional presence in African American communities, Sherkat and Cunningham (1998) noted that non-organizational religious practices often take place in social settings, e.g., prayer and religious media consumption with family and friends. For that reason, they argue that non-conformity with broader norms about these behaviors can carry negative social sanctions, just as it can when individuals opt out of participation in congregational life. Thus, Sherkat and Cunningham (1998) contend that the semi-involuntary thesis is also germane to private religious consumption. This might suggest that African Americans interviewed by persons of their same race could feel pressured to overreport their non-organizational religious practices (and perhaps even their subjective religious perceptions) as well.

 
4

The RAH sample was originally designed to consist of equal numbers of African American and white elders. We dropped a small number of respondents who were queried by interviewers from a racial/ethnic background besides black or white (e.g., Latino, Asian). Missing cases on individual items were handled via listwise deletion, which allows the effective N in our analyses to vary somewhat across specific dependent variables.

 
5

Briefly, RAH items on organizational involvement and congregational activities were subjected to an exploratory principal components factor analysis using the FACTOR procedure in SAS version 9.1. Two clear factors emerged with eigenvalues greater than 1.5. Three items (attending services, attending Sunday School or Bible study groups, participating in prayer groups) loaded on the first factor at greater than 0.60, with loadings on the second factor at 0.20 or less. These three items were retained as the components of the organizational religious practices scale. Three different items (work around the church, working on church programs, leadership roles) loaded on the second factor at greater than 0.45, with similarly minimal loadings on the second factor. These latter three items were used to construct the congregational activism scale.

 
6

The seven items developed by Krause (2004) to measure attitudes regarding religion and race relations were subjected to an exploratory principal components factor analysis using the FACTOR procedure in SAS version 9.1. One again, two clear factors emerged with eigenvalues greater than 1.5. Five items (enumerated in the measures section of the article) loaded on the first factor at 0.50 or greater, with loadings on the second factor of 0.15 or less. This set of items was used to construct the scale of critical views. The remaining two items loaded on a second factor at greater than 0.70, loading on the first factor at less than 0.25. Therefore, the remaining two items were used to create our measure of deferential views on religion and race.

 
7

Several studies have maintained that the “semi-involuntary institution” thesis is particularly relevant to African American life in the South (Ellison and Sherkat 1995; Sherkat and Cunningham 1998; Hunt and Larry 2000). If that is the case, then one might expect for the patterns anticipated by H3 and H4 to be stronger among southern African Americans, as compared with those residing elsewhere. Although the best and most straightforward test of this idea would involve adding cross-product interaction terms (region × race of interviewer) to the OLS models in Table 2, unfortunately that is not feasible, due to a dearth of African American interviewers in some regions. Nevertheless, simple regional comparisons of the bivariate associations between race of interviewer and (a) organizational religiousness and (b) critical responses to items on religion and race relations suggest no support for this version of the “racial solidarity” perspective.

 
8

Although some studies over the years have focused on race-of-interviewer effects among both African Americans and whites, and it would be interesting to do so with regard to religion, that is not feasible with the RAH data. All but 12 of the 750 white respondents were interviewed by whites.

 

Copyright information

© Religious Research Association, Inc. 2011