Survey respondents report higher levels of risky sexual behavior when they answer questions anonymously in private polling booths using marked tokens than when they take part in face-to-face interviews, according to a study conducted in Cotonou, Benin.1 For example, the proportion of married men who reported ever having had sex with a female sex worker was 20% in face-to-face interviews, but 42% in polling booth surveys. Similarly, among married women, the proportion reporting ever having had extramarital sex was 5% in face-to-face interviews and 24% in polling booth surveys. Differences by interview type in the prevalence of sexual behaviors were greater than those of condom use and STI symptoms.

Although face-to-face interviews are widely used to collect data on sexual behavior, especially in populations with low literacy, the method is subject to social desirability bias—respondents may underreport socially proscribed behaviors or overreport those that are socially approved. Methods that improve confidentiality have been developed, but are more suitable for populations with high literacy rates. As part of an HIV-prevalence study conducted in Cotonou in 2008 among men aged 15–64 and women aged 15–49, researchers conducted face-to-face interviews with a random sample of 2,580 respondents and polling booth surveys with an independent random sample of 1,095 individuals from the same population. Dichotomous questions on HIV-related risk behaviors were read aloud to the polling booth survey participants, who then placed a token marked with a number corresponding to the question number into a box labeled "yes" or "no" (the tokens were placed to the side if the question was not applicable). To ensure the respondent’s anonymity, the box was shielded from the interviewer’s sight. The same questions were asked in the face-to-face interviews, along with questions on social and demographic characteristics. Pearson chi-square and Fisher’s exact test were used to compare the proportions of affirmative answers obtained by each survey method, stratified by gender and marital status, yielding four demographic groups—married men, unmarried men, married women and unmarried women.

The vast majority (90–96%) of eligible men and women participated in the two surveys. The median age of respondents interviewed face-to-face was 36 for married men, 22 for unmarried men, 30 for married women and 19 for unmarried women. In all four groups, the proportion of respondents who acknowledged having engaged in various risky sexual behaviors was generally greater in the polling booth group than in the face-to-face interview group. For example, among married men, those who answered questions in the polling booth were far more likely than their counterparts in the interview group to say that they had had extramarital sex in the past year (54% vs. 18%) or ever (67% vs. 59%), had had sex with a female sex worker in the past year (24% vs. 12%) or ever (42% vs. 20%), or had ever had anal sex with a woman (18% vs. 3%) or man (8% vs. 0%). Among married women, participants in the polling booth surveys were more likely than those in the face-to-face interviews to say that they had had extramarital sex in the past year (21% vs. 3%) or ever (24% vs. 5%), had been paid for sex in the past year (13% vs. 1%) or ever (17% vs. 2%), or had had anal sex (18% vs. 3%). Results for unmarried men and women were generally similar, with a few exceptions; for example, the proportion of unmarried men who reported having ever had sex with a woman, and the proportion of unmarried women who reported having ever had sex with a man, did not differ between the polling booth and face-to-face interview groups.

Findings for measures of condom use were less consistent. Among married men, those in the polling booth surveys were more likely than those in the face-to-face interviews to report having used a condom at last sex with their first spouse (25% vs. 7%) or with a female sex worker (73% vs. 56% of those who had had sex with a sex worker). On the other hand, among unmarried man, reported levels of condom use in various contexts were higher in face-to-face interviews than in polling booth responses; a possible reason for this finding, according to the authors, is that unmarried men were more inclined than their older, married counterparts to report condom use behavior in line with recent local interventions. Women in the polling booth samples were more likely than those in the interview samples to report condom use, with one exception: Among unmarried women, the prevalence of condom use at last paid sex did not differ by interview format.

In addition, respondents surveyed at polling booths were more likely than those interviewed face-to-face to report having had STI symptoms (genital ulcers or urethral or vaginal discharge) in the past year or having ever injected illicit drugs. And although they were generally less likely than their peers in the face-to-face interview group to say that they had heard of HIV or AIDS, they were more likely to have been tested.

The researchers acknowledge several limitations of the study: The participants were not randomly assigned to a specific survey type, and unlike face-to-face interviews, the polling booth survey method does not permit interaction between the participant and the interviewer (to clarify questions) or allow researchers to check for inconsistent responses. Moreover, the polling booth survey method allows only for collection of aggregated data. Despite these limitations, the researchers note that polling booth surveys "seemed to have increased respondents’ willingness to report stigmatized behaviours." The investigators conclude that polling booth surveys "are suitable to monitor reliable HIV/STI risk behaviours and … could be used to adjust for answers to sensitive behavioural questions in [face-to-face interviews]."—L. Melhado

REFERENCE

1. Béhanzin L et al., Assessment of HIV-related risky behavior: a comparative study of face-to-face interviews and polling booth surveys in the general population of Cotonou, Benin, Sexually Transmitted Infections, 2013, 89(7):595–601.