As concern grows over adolescent pregnancy and the spread of HIV and other sexually transmitted diseases (STDs) among young people, considerable research has focused on the sexual behavior of adolescents. Few studies, however, have addressed the accuracy of data used to measure sexual behavior. Having accurate measures of the extent of adolescent sexual activity is important to help identify populations in need of family planning and STD services and to assess the impact of such programs.
Sexual behavior is difficult to measure accurately, however, since self-reports serve as the only source of information; no independent record of sexual activity exists to serve as a "gold standard" of measurement. When respondents are asked about sensitive topics such as sexual experience, they might give what they consider to be socially desirable responses rather than accurate information. This type of misreporting could be more prevalent among adolescents who are not yet comfortable with their sexuality and who may be more sensitive than adults to the consequences of public knowledge of their sexual experience.1
Several U.S. studies have used different methodologies to determine the level of misreporting of sexual experience among adolescents. One longitudinal study of 11-16-year-olds in North Carolina, for example, asked respondents twice in a single questionnaire if they had experienced sexual intercourse. Between 4% and 8% of respondents were inconsistent in their reports of intercourse within the same survey round, and an additional 6% were inconsistent between surveys (i.e., they reported having had sexual intercourse by a certain age in Round 1, but reported no sexual activity at that age in Round 2).2 In another U.S. study, 7% of junior high school students admitted having lied about their sexual behavior in an earlier survey round; an additional 10% claimed that they had been truthful, but they could not accurately recall their earlier responses.3 Moreover, an eight-year longitudinal study found that 28-32% of U.S. adolescents reported an age at first intercourse that was inconsistent with the age they had given previously; white females and adolescents living in two-parent households were more likely than others to report their age at first intercourse consistently.4
However, no published research as yet has addressed the consistency of self-reported sexual activity among adolescents or adults in developing countries. This information is especially needed in countries where levels of adolescent activity are believed to be high, such as Jamaica: The majority of Jamaican 15-19-year-olds—59% of females and 75% of males—report that they have experienced sexual intercourse.5 Moreover, 40% of Jamaican women have been pregnant at least once before age 20, and 85% of these pregnancies are unplanned.
The purpose of this article is to assess the consistency of self-reported data from young adolescents in Jamaica regarding the initiation of sexual intercourse. We look at consistency in two ways. First, we examine the consistency among an individual's responses to multiple items asked on a single round's questionnaire. Second, we examine the agreement between responses collected from the same individual at three points in time. We then use multivariate analytic techniques to explore the factors that influence inconsistent reporting.
Data and Methods
The data come from a three-round longitudinal study conducted within the course of two years among young adolescents; the baseline survey was conducted among seventh graders aged 11-14 (mean age of 12). The survey was primarily designed to collect information about sexual decision-making and on adolescents' reproductive knowledge, attitudes and behavior. An additional objective of the longitudinal study was to evaluate the effect of a new family life education program,6 and the selection of the study participants grew out of this objective.*
The family life project was implemented in "all-age" schools (grades 1-9) and "new secondary" schools (grades seven through 11 or 12); students who attend these schools—about 60% of Jamaican youth aged 12-14—typically do not go on to receive a university education.7† The type of school one attends is closely associated with social class, with children from poorer socioeconomic backgrounds being considerably more likely than those from more affluent households to attend all-age and new secondary schools. Roughly one-half of the study participants attended five schools at which the specific family life education intervention was implemented; these schools were located in parishes across Jamaica and represented both urban and rural locations. The rest of the study participants attended five schools that offered other family life education curricula. Each of these comparison-group schools was located near an intervention school, and was similar in terms of academic caliber, size of student body and student demographics.
At each of the 10 selected schools, all seventh graders were asked to participate in the study. No student refused, but approximately 3% were absent when the baseline survey was conducted. All of the adolescents who participated came from low-income families (with nearly all parents working in the service sector or in unskilled or semi-skilled labor positions). Moreover, these students' academic performance was relatively poor. There were no significant baseline differences between the two groups of schools in terms of the adolescents' age, socioeconomic background, reproductive knowledge or sexual experience.
Since the schools included in the study were not randomly selected, the adolescents surveyed cannot be considered representative of all Jamaican young adolescents. However, they are unlikely to differ in meaningful ways from the larger population of 12-14-year-olds who attend all-age and new secondary schools. In addition, the selection of adolescents attending family life education classes is unlikely to bias the results, as the large majority of Jamaican schools offer such classes.
The baseline survey, administered in September 1995 when respondents were beginning seventh grade, included 942 young adolescents—487 females and 455 males. Approximately 92% of these respondents (n=869) completed Round 2 in June 1996, and 74% of the original sample (n=698) went on to complete Round 3 in June 1997. Attrition in the sample was not associated with socioeconomic status, gender, the specific school attended or urban-rural school location. Between Rounds 1 and 2, boys were significantly more likely than girls to be lost to follow-up (10% vs. 5%); there were no significant differences by gender in the proportions lost to follow-up between Rounds 2 and 3.
The survey questionnaire and its methods of administration were pretested among seventh graders at two Kingston schools that were similar in academic caliber and student demographic composition to the study schools. At the first pretest school, one group of students completed a self-administered questionnaire, and an interviewer guided a second group through the same questionnaire; this pretest determined that a large proportion of the students could not read and write well enough to complete self-administered questionnaires. Discussion sessions with small groups of students also revealed that many did not understand the wording of some questions. Thus, based on these pretest results, the questionnaire was shortened and skip patterns directing respondents to another question depending on their answer were omitted. The second pretest, conducted at a different Kingston school, resulted in only minor changes to the questionnaire.
Students completed the questionnaire in their classrooms in single-sex groups of 8-15 students. Due to the limited literacy of many students, an interviewer read aloud each question and its possible responses, while the students followed along on their questionnaire; most of the questions were closed-ended. Privacy was maintained within the group setting and respondents wrote their responses alone, without being observed by the interviewers or by other students. Empty desks were left between respondents, who used a blank sheets of paper to cover their responses.
In each survey round, a respondent was first asked directly if he or she had ever had sexual intercourse. (Interviewers defined sexual intercourse aloud in simple terms to convey heterosexual vaginal intercourse.) Respondents then answered five subsequent questions about their first sexual experience, providing information about their age at the time, their partner's age, the reasons why they first had sex, whether they used a method of family planning at that time and which method they used. As mentioned earlier, the deliberate omission of skip patterns meant that even respondents who had not yet experienced sexual intercourse responded to each of these questions, with "I have not had sex" being a possible response in each case.
We also collected data on background characteristics known to be related to adolescent sexual activity and shown in previous research to be correlates of inconsistent reporting. These included age, ever-use of alcohol and marijuana, household socioeconomic status, frequency of church attendance and whether the adolescent lived in a two-parent household. A socioeconomic index for this low-income sample was created based on four household amenities—electric light in the home, running water, a flush toilet and a gas or electric stove. (A household scored very low on the index if it had two or fewer of these amenities.)
Classifications of Consistency
We classified adolescents' reports of sexual activity as consistent or inconsistent, both within each survey round and between survey rounds. For example, we categorized adolescents' reports of sexual activity as consistent within each round if responses to the six questions on first sexual intercourse were in agreement; the reports were classified as inconsistent if these six answers did not agree.
We considered adolescents' overall reporting of their first sexual experience to be consistent if three conditions were met: responses were consistent within each survey round; a report of being sexually inexperienced never followed a report of having had sex; and age at first sex was reported consistently. (That is, age at first sexual experience was considered consistent if that age was within one year of the age at first intercourse given in a previous round; for adolescents who reported no sexual activity in a previous round, age at first sex was considered consistent if it was within one year of the adolescent's age at the previous round.)
Thus, adolescents' responses would be considered consistent overall under three circumstances: if they consistently reported no sexual activity; if they consistently reported experiencing first sexual intercourse before the baseline survey, with age at the time of that experience being reported consistently; or if they first experienced sexual intercourse during the study period, did not subsequently report having never had sex and reported the age at that first encounter consistently.
We coded a respondent as inconsistent overall if the following criteria were met. First, if a student was inconsistent in responses to the six questions in a single survey round, we automatically considered that student's responses to be inconsistent overall because we were unable to fully assess consistency between rounds. We also considered respondents to be inconsistent if they reported being sexually inexperienced after having previously reported sexual activity, or if the reported age at first sex was inconsistent between two or more survey rounds.
We investigated the predictors of inconsistency in self-reporting (dependent variable) by estimating a logistic regression model. The independent variables in the analysis included gender, socioeconomic status, ever-use of alcohol, frequency of church attendance, whether the adolescent lived in a two-parent home and whether he or she attended a school in the intervention group or the comparison group. We dropped 14 of the 698 observations from the multivariate analysis because of missing data on one or more of these variables.
Adolescents' reports of their alcohol use, socioeconomic status and household composition varied little over the study period; we use Round 3 data for these characteristics, since our measures of overall consistency include reports up through the final round. Moreover, adolescents probably reported the factors comprising the socioeconomic index more accurately at older ages. Age and marijuana use were also considered as independent variables; however, we did not include them in the final model because there was very little variance in age (since all students entered the study in seventh grade) and because only a small percentage had ever used marijuana.
The 698 adolescents who completed all three surveys were 12.1 years old, on average, at Round 1; they were aged 12.8 by the time they completed Round 2 nine months later; and they were 13.8 years old by Round 3. At baseline, 42% of these students had tried alcohol, and 48% had done so by Round 3 (Table 1). Few students had ever smoked marijuana—6% at the time of the baseline survey, a proportion that had increased by four percentage points by Round 3. In all three rounds, the proportions reporting alcohol and marijuana use were consistently higher among boys than among girls. While all adolescent respondents were from families of low and lower-middle socioeconomic backgrounds, about 47% (in Round 3) lived in very low socioeconomic status households. At the time of Round 3, most adolescents lived either with both parents (35%) or with just their mother (39%).
Within each survey round, 95-100% of adolescents responded consistently to questions about their first sexual intercourse (Table 1). Most girls consistently reported that they had never had sex (86-95%), while the majority of boys consistently reported that they had done so (60-75%). The proportion of respondents who were inconsistent was highest in Round 1 (5% overall—4% of girls and 6% of boys), when the adolescents were youngest and their literacy skills were probably the least developed. The proportions inconsistent were also slightly higher among boys than girls in Round 3 (1.8% vs. 1.6%), but the seemingly reverse situation at Round 2—0.3% among boys and 0.5% among girls—reflects very small absolute numbers (just two girls and one boy).
We next assessed the extent of agreement over the multiple rounds of the survey (Table 2). As expected, the proportion of adolescents who were inconsistent between survey rounds was much higher than the proportion who were inconsistent within a single round (37% vs. 5% or less). Boys and girls exhibited very different patterns of overall consistency in their responses.
For example, 65% of boys reported their sexual experience inconsistently. Eight percent were inconsistent within a single round, while 16% reported their sexual experience inconsistently between rounds (that is, they reported no sexual experience after previously having claimed to have had some). An additional 30% consistently claimed sexual experience in all three rounds, but reported their age at first sex inconsistently. Finally, the remaining 11% consistently reported that they first had intercourse after the study began, but then inconsistently gave their age at first sex. Thus, among boys, inconsistent reporting of age at first intercourse accounted for a large portion of overall inconsistency.
Of the 35% of boys who were consistent in their reporting, 10% consistently reported no sexual experience, and the remaining 25% consistently reported having had intercourse (8% after the study began and 16% before Round 1).
Among the girls, just 12% reported their sexual experience inconsistently, with most of these girls being inconsistent with their responses within a single round (6%) or claiming to have never had sex after having reported sexual activity in a previous round (4%). The remaining 2% was attributable to girls who began sexual activity during the study period, but reported their age at first sex inconsistently. No girl who claimed to be sexually experienced at each survey round inconsistently reported her age at the time of her first sexual experience.
The overall proportion of girls who were consistent in their reported sexual activity—88%—is primarily made up of those who consistently reported no sexual activity (79%), with the remaining 9% giving consistent responses about their first sexual experience (1% who had had sex before the study began, plus 8% who initiated sexual activity during the study period).
Predictors of Inconsistent Reporting
The logistic regression results confirmed the gender differences in consistency indicated in the bivariate data: Boys were 14 times as likely as girls be inconsistent in their self-reports of sexual experience (Table 3). Adolescents who lived in the lowest socioeconomic-status households were about 49% more likely than those in other households to be inconsistent in their responses. Attending a school in the family life education intervention project under study was also significantly associated with inconsistent reporting of sexual activity (odds ratio, 1.61), but none of the other independent variables were associated with the likelihood of giving inconsistent answers about sexual activity.
Discussion and Conclusions
Most previous research assessed consistency in self-reports of sexual behavior by repeating a question within a single survey or by asking whether the respondent had experienced sexual intercourse in two or more surveys. This study builds on this approach by also using the respondents' reported age at first sexual intercourse to evaluate consistency. In doing so, the study revealed a much higher rate of inconsistent responses among boys than would have been indicated if we had assessed consistency using reports of sexual experience only. It should be noted, however, that relatively few females were even "eligible" to be considered inconsistent in their responses because most of them reported that they had never had sexual intercourse (and therefore would be unable to inconsistently report their age at first intercourse). In other words, those adolescents who claimed that they were sexually experienced (generally boys) had to verify the truth of their responses on more occasions than adolescents who said they had never had sex.
Our study found a substantial amount of inconsistent reporting of sexual activity among young adolescents in Jamaica. Although the vast majority of adolescents reported their sexual experience status consistently within a given round, about one in 10 girls and nearly two in three boys responded inconsistently when we examined the agreement of responses between rounds and the consistency of the reported age at first intercourse. Among girls, most of the overall inconsistency resulted from inconsistent responses within a single round, and from reporting no sexual experience after previously reporting sexual activity. While much of the overall inconsistency among boys also stemmed from differing reports of sexual experience between rounds, the largest portion of their inconsistency was attributable to giving inconsistent ages at the first sexual experience (despite the fact that boys were consistent between rounds in reporting that they had ever had sex).
The multivariate analysis indicated that boys were much more likely than girls to report their sexual experience inconsistently. Interestingly, the multivariate analysis also indicated that adolescents whose school was part of the family life education program being evaluated were also more likely than those in the comparison group to inconsistently report their sexual activity. The intervention may have influenced young adolescents' attitudes toward sexual activity, or there may have been baseline differences in the two populations that this study did not detect.
Our findings raise questions about the limitations of relying on self-reported data, particularly from boys, to identify sexually active adolescents, to estimate the prevalence of sexual activity among adolescents and to evaluate the impact of interventions designed to delay first intercourse. We advise caution in using self-reports of sexual behavior data for these purposes, particularly among adolescent populations. Moreover, boys' greater proclivity for inconsistent reporting should be kept in mind in discussions of gender differences in sexual activity. Indeed, qualitative data collected from these same young adolescents indicate that in the Jamaican sociocultural context, boys feel considerable pressure to be sexually active, while girls who become sexually active suffer damage to their reputations.8
Our results do not speak to the validity of adolescents' reports of sexual activity; some adolescents who reported sexual activity consistently may not have reported it accurately. Whether they misrepresented their actual experience deliberately or inadvertently, they did so in a consistent manner. Nor can the study findings tell us if inconsistent reporting indicates inaccurate recall of events or deliberate misreporting. Inconsistency is likely related to the sensitive nature of the questions asked; these young adolescents were asked to share highly personal information during a difficult period of their lives. First sexual intercourse is an important personal milestone, and the age at which that milestone occurs is unlikely to be forgotten. Interestingly, none of the responses on using alcohol were inconsistent: Not a single adolescent reported never having drunk alcohol after previously reporting having done so (data not shown). This suggests that these adolescents are capable of consistently reporting a behavior other than sexual activity.
The study considered only a small number of characteristics associated with adolescents' sexual activity and their self-reported acknowledgement of that activity. We could not measure factors such as adolescents' perceptions of pressure to remain sexually inexperienced or to engage in sexual activity, and the level of perceived confidentiality of the survey responses; these may have affected adolescents' willingness to respond honestly. Further research is needed on other factors that might influence adolescents' reports of sexual behavior and on the most effective ways to collect such information. Obtaining qualitative feedback from the study participants themselves on the honesty of their responses would also be useful, as would information on their level of confidence in the confidentiality of their responses and on their perception of the interviewers' views of adolescent sexuality.