INTERVIEWER EFFECTS AND SOCIAL DESIRABILITY BIAS

The literature on interviewer effects suggests that the race and gender of an interviewer can systematically and significantly affect survey responses, biasing outcomes in a number of ways. Interviewer race has been found to impact attitudinal, behavioral, and even political-knowledge indicators, especially on questions addressing issues related to race (Schaeffer Reference Schaeffer1980; Cotter, Cohan, and Coulter Reference Cotter, Cohen and Coulter1982; Davis and Silver Reference Davis and Silver2003). Regarding the gender of the interviewer, a number of studies show that respondents provide more progressive and egalitarian answers on gender-sensitive questions when interviewed by women (Lueptow, Moser, and Pendleton Reference Lueptow, Moser and Pendleton1990; Kane and Macaulay Reference Kane and Macaulay1993; Flores-Marcias and Lawson 2008; Benstead Reference Benstead2010). Both male and female respondents are also more willing to report details on sensitive topics, such as sexual activity, to female interviewers (Abramson and Handsschumacher 1978; Hansen and Schuldt Reference Hansen and Schuldt1982; Darrow et al. 1986; Catania et al. Reference Catania, Binson, Canchola, Pollack and Huack1996).

In explaining interviewer effects, scholars believe respondents may mask their true feelings in an effort to please the interviewer or to appear to hold socially-approved norms (Goffman 1963). Social conformity and desirability bias generally refer to the tendency of respondents to provide responses that they believe will be viewed favorably by others and avoid embarrassment or criticism. Past social and psychological studies show that individuals will sometimes — consciously, or subconsciously — follow the dominant norms or opinion of the society they are in rather than maintaining their initial stance on a variety of issues (Asch 1951). Such conformity is thought to be encouraged by informal social sanctions or rewards that are frequently related to emotions such as shame and pride (Goffman 1967; Scheff Reference Scheff1988). The culture in which individuals grow up may be responsible for this conforming behavior, as determinants of shame and pride are based on culture-specific expectations (Bond and Smith Reference Bond and Smith1996). The impact of these social sanctions is, therefore, different from culture to culture, and studies have found evidence of such desirability effects that reflect the norms of different settings (Zerbe and Paulhus Reference Zerbe and Paulhus1987; Neto Reference Neto1995; Kondo et al. Reference Kondo, Saito, Deguchi, Hirayama and Acar2010).

Survey data have given us insight into societal attitudes and interpersonal interactions by revealing under what conditions respondents will intentionally and inaccurately present themselves in socially admirable ways (Evans, Hansen, and Mittlemark Reference Evans, Hansen and Mittlemark1977; Himmelfarb and Lickteig Reference Himmelfarb and Lickteig1982; Paulhus Reference Paulhus1984; Palvos 1972; Sigall and Page Reference Sigall and Page1971; Warner Reference Warner1965). Respondents representing a variety of races, for example, seek to avoid answering questions in ways that might offend the interviewer and, as a result, respondents typically express warmer feelings and higher levels of trust toward people that share the interviewer's race (Schuman and Converse Reference Schuman and Converse1971; Anderson, Silver, and Abramson Reference Anderson, Silver and Abramson1988; Davis Reference Davis1997a; Reference Davis1997b; Krysan and Couper Reference Krysan and Couper2003). Among members of the same racial group, respondents may seek to avoid insulting an interviewer to a greater extent than an interviewer of another race as a result of in-group loyalty. Social deference theory also suggests that a desire to minimize the social distance between two strangers may lead to responses that complement the interviewer's perceived social group and in some cases demean the respondent's own social group (Lenski and Leggett Reference Lenski and Leggett1960; Williams Reference Williams1964; Reese et al. Reference Reese, Danielson, Shomaker, Chang and Hsu1986; Kane and Macaulay Reference Kane and Macaulay1993; Webster Reference Webster1996; Davis Reference Davis1997a).

Differential social status and power is also believed to impact interview response bias (Davis 1997; Kane and Macaulay Reference Kane and Macaulay1993; Lenski and Leggett Reference Lenski and Leggett1960; Reese et al Reference Reese, Danielson, Shomaker, Chang and Hsu1986; Webster Reference Webster1996; Williams Reference Williams1964). Some respondents, particularly those in lower status social groups, might perceive an enumerator's higher social status. Respondents would act deferentially toward the enumerator by expressing opinions that complement the interviewer's social group or demean the respondents own social group. Davis for instance highlights African American respondents' reluctance to reveal their true political beliefs and “disassociate themselves from black issues, and alternatively appear more docile and accommodating” (Davis Reference Davis1997a, 320). Thus, respondents who are members of groups that have been subjected to discrimination in the past, such as African-Americans or women, might be especially likely to express views that are respectful to high-status interviewers' social groups, such as white males.

Converse considers such findings in survey research as a useful “indicator of evolving areas of interpersonal tension…and deserve to be treated as a fact of social life and not merely as an artifact of the survey interview” (Converse 1971, 45). In this study, we use face-to-face surveys as an opportunity to examine interreligious interactions in Egypt and the influence of perceived interviewer religiosity on expressions of attitudes related to religious and cultural values. While public opinion experts recognize the importance of race and gender in social desirability effects, little scholarly work has sought to address the unique types of interviewer effects that might influence survey responses in the religious context. Past studies show that weekly church attendance, a positive value in American society (Groves Reference Groves1989), is grossly over reported in standard surveys in the United States (Hadaway, Marler, and Chaves Reference Hadaway, Marler and Chaves1993; Chaves and Cavendish Reference Chaves and Cavendish1994; Smith Reference Smith, Gilbert, Fiske and Lindzey1998); however, few studies have explored social desirability bias within the context of interfaith relations.

Our study considers social desirability effects in the context of Egypt, a country with two primary religious groups where Muslims vastly outnumber Christians. Specifically this article considers the effect of interviewer dress, particularly whether or not the interviewer dons the hijab, or Islamic headscarf, on questions related to personal religiosity and religious cultural norms. Understanding potential desirability bias in interfaith contexts will be increasingly important for sound research as investigation of religious attitudes increases. Major research projects that conduct polls in the Islamic world, including the World Values Survey and the Pew Global Attitudes project, seek to measure worldwide attitudes on a variety of subjects, including personal religiosity and religious adherence in the context of sectarian tension. Failure to account for the impact of non-verbal cues — like religious dress — may create biases in assessment both over time and in terms of cross-sectional variation.Footnote ¹

INTERVIEWER DRESS AS A FORM OF NON-VERBAL COMMUNICATION

In contemporary Egypt, one observes considerable variety in terms of the style of dress worn by women, ranging from completely Westernized apparel to both traditional clothing and more modern Islamic garb. Previous research suggests that Islamic dress, particularly donning the veil, provides an external signal of Muslim identity and religious observance. Many scholars believe the choice of dress by women in part reflects different interpretations of the Koranic instructions for appropriate Islamic attire as well as varying levels of personal piety.

The veil serves as an effective signal of piety, particularly since a woman who chooses to wear the Islamic headscarf may incur a cost for her action (Chehabi Reference Chehabi1993; Carvalho Reference Carvalho2013; Patel Reference Patel2012). Patel (Reference Patel2012) argues that personal religiosity isa trait only known by an individual and is, thus, not directly observed by onlookers. He finds that proper Islamic dress allows women to convey information about themselves to others, and that more pious women adopt more conservative clothes and behaviors to differentiate themselves from less pious women. If veiling increases the perceived religiosity of survey enumerators, what is the cognitive impact of such a signal on respondent response?

Expressing socially desirable responses as a result of the perceived religiosity of a survey enumerator may be related to processes of strategic self-presentation of identities or impression management.Footnote ² Theory and research on impression management suggest that people monitor the image that they provide of themselves to the audiences they address (Baumeister Reference Baumeister1982; Bischoping and Schuman Reference Bischoping and Schuman1992; Leary and Kowalski Reference Leary and Kowalski1990; Schlenker Reference Schlenker1980). Such behavior is motivated, in part, by a desire to act in a manner consistent with social expectations, lest one face emotions of shame, embarrassment, or risk losing the power associated with adhering to such standards. Individuals will often modify or change their image so that it corresponds to social values, assuming the individual has a belief about what the audience values.

In the case of interviewer effects, the use of impression management is determined by the relation between the social identity of the respondent and that of the audience. Impression management involves both a cognitive and strategic factor. The cognitive factor refers to the aspect of social identity that is made cognitively salient to the respondent in the process of the interview. The strategic factor refers to the desire to express identity in a contextually appropriate fashion, and implies sensitivity to the identity of the audience and the audience's relative power or ability to impose social sanctions (Reicher, Spears, and Postmes Reference Reicher, Spears and Postmes1995; Spears and Lea 1994).

The presence of a veiled enumerator arouses expectations related to the respondents' beliefs about religion, sentiments toward the religion of the enumerator, and what they know about interactions between those who hold the enumerator's religious identity and those who hold the respondent's religious identity. For example, a Muslim respondent will reference what she believes about Muslims and Islamic norms as well as what she associates with interpersonal interactions between Muslims. A Christian respondent will similarly reference her beliefs about Muslims and Muslim-Christian interactions in Egypt. This cognitive process then informs the strategic response provided by the interviewee to express her identity in a way deemed agreeable to the enumerator. If an individual has an aversion to being viewed as impious or antagonistic to the religion of the enumerator, she may opt to “edit” her response (Tourangeau, Rips and Rasinksi Reference Tourangeau, Rips and Rasinski2000).

As a Muslim, being perceived as pious and less susceptible to secular influences can be considered socially desirable as it may elicit numerous social and economic benefits. Piety is an important attribute in Muslim societies, as it typically indicates good moral character, trustworthiness, and honesty, leading to a good reputation in social interactions. Reputation is critical as individuals and families frequently rely on informal networks to secure employment, improve desirability in the marriage market, obtain subsidized goods and services, and access informal savings and credit networks (Singerman Reference Singerman1995; Hoodfar Reference Hoodfar1997; Patel Reference Patel2012). Similar to interviewer effects relating to race and gender, respondents may seek to minimize social distance between themselves and the veiled enumerator whom they perceive to be religious, and desire to appear to subscribe to culturally desirable norms.

For Christians, pressure to conform to dominant societal norms may be even more intense. Christians and Muslims in Egypt have been integrated for centuries; however, the relationship between the two religious groups has varied historically. Religious identity and ethnic consciousness have grown in salience in Muslim and Christian communities of Egypt since the 1970s (Sedraa Reference Sedraa1999). Coptic Christians are believed to make up between 8 and 10% of the country's total population. Indeed, Egypt is home to the largest number of Orthodox Christians in the Middle East. During the second half of the 20th century, the Christian community's condition declined and many Christians faced religious violence with the resurgence of militant Islamic groups with violence peaking in the 1990s. According to Zeidanab (Reference Zeidanab1999), Copts have increasingly suffered from persecution and discrimination because of the changing terms of “national identity” in Egypt; while Coptic Egyptians were core to the national project during the heyday of Arab nationalism, national identity in Egypt is increasingly conflated with Muslim identity. With the increasing Islamization of the Muslim community and concomitant decline in secularism, Muslims are increasingly likely to see Copts as “the other,” and vice versa. Communal differences have been the source of tension (Makari Reference Makari2007) particularly since associational life in Egypt tends to revolve around religion and religious community. In some cases, Christian students have been coerced to wear the Islamic headscarf; anti-Coptic books and pamphlets continue to be in widespread circulation (Ibrahim Reference Ibrahim1996).

The literature on interviewer effects would suggest that the enumerator's perceived religion and degree of religiosity may interact with the respondent's religion, thereby influencing the responses by Muslims and Christians in different ways. As a result, we expect the signal of the veil to trigger a desire to self-present as more pious, on average, for Muslim respondents out of a desire to affirm their membership in the dominant group and establish a common degree of religiosity with the enumerator. Conversely, we expect Christian respondents to deemphasize their religious identity to avoid appearing antagonistic to the dominant group to which the enumerator appears to belong. This expectation is consistent with research finding that people at times opt to dissociate themselves from socially disadvantaged groups (Ellemers, Spears and Doosje Reference Ellemers, Spears and Doosje1999). Christians, therefore, may respond to items related to religiosity by expressing lower levels of piety, or less adherence to Christianity, when interviewedby a veiled woman in order to minimize social distance.

Beyond simple expressions of religiosity however, there is reason to expect the enumerator's perceived religion may impact socio-cultural responses as well. With approximately 90% of Egyptians identifying as Muslims, Islamic cultural norms have a large impact on Muslims and the Christian minority alike. We should expect Muslims to express stronger adherence to cultural norms associated with Islam when interviewed by a veiled enumerator in an effort to align with desirable Islamic social standards. Despite being of a different religion, Christians, too, may be more likely to express greater observance to Islamic cultural norms due to their deferential social status in Egypt and — perhaps unconscious — desire to adhere to the local society's practices, which happen to be Islamic. This study explores the effect of an explicitly Muslim female enumerator in the Egyptian societal conditions on responses given by Muslim and Christian women.

This study relates to existing work on the impact of gender and dress on attitudes in the Islamic world. Benstead (Reference Benstead2010) shows that both male and female respondents in Morocco reported less progressive attitudes toward women's participation in politics to female interviewers wearing the hijab, but more progressive attitudes towards equality in personal status laws to female interviewers, regardless of dress. Our study is distinct from Benstead (Reference Benstead2010) in a number of ways. Benstead (Reference Benstead2010) focuses primarily on attitudes toward gender equality although she also finds an effect of female interviewers wearing the hijab on reported religiosity and support for democracy; this article addresses issues related to levels of religious conviction and adherence to Islamic cultural norms. The number of interviews conducted by female interviewers wearing a hijab in Benstead (Reference Benstead2010) was relatively small (N = 65) and was assessed as only a partially successful randomization of housing listings in a nationally-representative survey of political attitudes; the data analyzed in this article represent a much larger number of interviews completed by veiled interviewers (N = 582 veiled, i.e., treatment; N = 586 unveiled, i.e., control) and we assess our effort at random assignment achieved successful balance on the dimensions of age, religion, education and socioeconomic status across treatment and control groups. Finally, Benstead (Reference Benstead2010) also deals with the impact of interviewer gender and the hijab on question response and item-missing data for respondents of both genders, while this article narrows in on the impact of Islamic dress on a sample of women from two ethno-religious groups.

SURVEY QUESTIONS

This article seeks to analyze the effect of having an enumerator who wears a hijab — or simple Islamic headscarf — on responses to questions related to religiosity and religion-related cultural norms. To operationalize these concepts, this study analyzes responses to three items that are argued by various scholars to capture piety and adherence to perceived Islamic cultural norms in Egypt.Footnote ⁶

The first item is straightforward in measuring the degree to which respondents indicate that they incorporate their religion into their personal, everyday lives. During the length of the survey, respondents were asked the following question:

To this question, Muslims indicated how observant they were of Islam and Christians on how observant they were of Christianity.

The second item seeks to measure adherence to a cultural norm that is associated with Islam in Egyptian society, female genital mutilation, also known as female circumcision. Practiced by both Muslim and Christians, circumcision involves the partial or complete removal of the external female genitalia, and is typically performed on girls between the ages of eight and 14, before a girl reaches puberty (Assad Reference Assad1980).

Although most scholars consider circumcision in Egypt to be a social custom rather than a religious one, the practice is increasingly thought of as part of an Islamic tradition.Footnote ⁷ Heba El-Kholy (Reference El-Kholy2002), in her anthropologic study of low-income women in Cairo, notes that female circumcision is thought to be legitimized by various religious precedents and that many Muslim women believe it is a good tradition required by their religion. Indeed, religious Muslims have become the most vocal advocates of female circumcision (Moore et al. Reference Moore, Randolph, Toubia and Kirberger1997; Allam et al. Reference Allam, de Irala-Estevez, Fernandez-Crehuet Navajas, Castillo, Hoashi, Pankovich and Liceaga2001; Boyle, Songora, and Foss Reference Boyle, Songora and Foss2001; Yount Reference Yount2004), even in the face of government efforts to criminalize the practice.Footnote ⁸ At the same time, Christians are becoming increasingly less likely to circumcise their daughters (Yount Reference Yount2004). In a sample of married women in Minya governorate, Christian women are having fewer daughters circumcised, less extensive forms of cutting are used and the practice is seen as less beneficial than among their Muslim neighbors.Footnote ⁹

Although more common among low-income woman than in the upper classes, circumcision is said to be significant in Egyptian society as it relates to protecting female modesty and chastity, traits highly valued in Islam, which may improve a young woman's reputation and marriage prospects (Assad Reference Assad1980; Singerman Reference Singerman1995). Indeed, one of the four Sunni schools of religious law — the Shafi'i school — considers the trimming of the clitoral hood mandatory (Antes Reference Antes1990).Footnote ¹⁰ Since circumcision is nearly universal among women of reproductive age in Egypt (El-Zanaty and Way Reference El-Zanaty and Way2006), we operationalize adherence to this religiously-related cultural practice by asking each respondent the following question about her daughters:

Finally, we utilize a third item intended to measure what some perceive as an Islamic cultural norm related to a Muslim woman's role in her marriage. Within the Koran, there exist verses (ayah) that are interpreted by some Muslim scholars as suggesting that in an Islamic marital contract, a woman has a religious duty to provide sexual satisfaction to her husband. Barlas (Reference Barlas2002, 161) writes, “from these ayah many Muslims infer that a wife is her husband's sexual property and that he has the right to have sex with her as and when he pleases (without her consent).” Mernissi (Reference Mernissi1987) further suggests that female disobedience and the refusal to obey a husband's request, particularly for sex, is interpreted by many Muslim clerics as a serious offense. Refusing sex is thought by many to be a religious sin, which often results in wife-beating (El-Kholy Reference El-Kholy2002). For example, a 2004 poll by the Egyptian National Population Center found that 70% of married women believed that a husband was justified in beating his wife for her refusal to have sex.Footnote ¹¹ While this study does not assess the extent to which this belief is held among women in the sample, we believe that a reluctance to report episodes of forced sex to an interviewer offers another window into the effect of having a veiled versus unveiled enumerator on a respondent's answers to survey questions. In other words, a woman concerned about being perceived as pious (as a result of being assigned a veiled interviewer) may be less inclined to report forced sex. Considering this framework, item three measures a woman's desire to appear to adhere to a cultural expectation of providing sexual satisfaction to her husband.

Together these three items aim to capture diverse aspects of perceived piety and adherence to religion-related cultural norms among respondents. We have different expectations regarding how Christians and Muslims will respond to being assigned a veiled survey enumerator. We expect that assignment to the “treatment” category (i.e., assignment of a veiled interviewer) will lead Muslim women to say that they are more religiously adherent, circumcising a larger percentage of their daughters, and subject to fewer incidents of forced marital sex. We also expect that being interviewed by a veiled enumerator will lead Christian women to say that they are less religiously adherent and circumcising a larger percentage of their daughters. For Christian respondents, the impact of veiling on willingness to admit to being forced to have sex within the marriage is more ambiguous since there is no expectation Christians would adhere to a religious practice justified by clerics on Koranic grounds. Yet if we believe that Christian women are more likely to offer survey responses in line with what may be perceived as majority-held (i.e., Muslim) cultural values, we should expect fewer reports of forced sex even among Christian women who are assigned a veiledenumerator.

COMPARABILITY OF TREATMENT AND CONTROL

This study was designed for respondents to be randomly assigned either a veiled or unveiled enumerator to conduct the face-to-face survey. In most cases, enumerators would alternate either between wearing a head scarf between interviews or sets of interviews. Four out of 16 survey enumerators preferred to remain veiled or unveiled for all of their interviews, however. These four enumerators were responsible for interviewing about 30% of the respondents.Footnote ¹² Since enumerators were randomly assigned to housing blocs, we do not anticipate significant bias as a result of enumerator non-compliance.

This assessment is supported by examination of the success of randomization over interviewer dress. Randomization successfully balances on the dimensions of age, whether or not the respondent attended school, and socioeconomic status, proxied by whether or not the respondent has satellite television. Notably, whether or not an enumerator was veiled or not appears to have had no effect on her ability to recruit Christian and Muslim respondents as the treatment and control groups were balanced on the distribution of religious denomination. The balance across treatment and control is particularly important over religious identification since it would be problematic if veiled enumerators were unable to solicit Christian respondents (or unveiled enumerators were not able to solicit Muslim respondents). This does not appear to be the case. In addition, there is no indication that veiled and unveiled enumerators were differentially attracting more religious versus less religious respondents. Previous research suggests that education and income are robust predictors of individual-level women's religiosity in Egypt and elsewhere in the Islamic world (Blaydes and Linzer Reference Blaydes and Linzer2008); t-statistics for the mean differences confirm that the mean differences of all covariates between the treated and control-group units — including on education and income proxies — are not statistically significant.