Introduction

Access to sexual knowledge is considered a human right (WHO, 2006). The right to sexual knowledge also poses a means of realizing other human rights, such as the right to health, right to information, and sexual and reproductive rights (Muñoz Villalobos, 2010). Sexual knowledge is delivered in formal and non-formal settings, both of which can be gender-specific. In the WHO European Region, remarkable progress in the level of sexual knowledge has been achieved since 2000 (Michielsen & Ivanova, 2022). Gender equity, including reducing gender imbalances, is considered one of the principles for successful program interventions in sexual education (WHO, 2006).

Adolescent sexual activity remains important for public health policies and can be shaped by them (Cowan, 2002). Adolescent sexual activity can carry various risks, including physical, emotional, legal, and economic aspects. In Poland, the country from which the data analyzed in the article originate, and internationally, it has been found that the prevalence of unintended pregnancy is mostly higher among adolescents than in other age groups (Jarząbek-Bielecka et al., 2015; Kost et al., 2017; Rizvi et al., 2019). In Poland, most adolescent fathers were found not to live with the mother of their child (Ratajczak, 2016) and most of their marriages end in divorce (Jarząbek-Bielecka et al., 2015). The majority of adolescent fathers reported financial difficulties, with taking on temporary employment as a potential solution to the problem (Ratajczak, 2016). Moreover, adolescents have been indicated to be biologically more susceptible to sexually transmitted infections and to face multiple obstacles in accessing confidential health care services (Ljubojević & Lipozenčić, 2010). It should also be acknowledged that adolescents constitute a small fraction of those with sexually transmitted diseases (Niedźwiedzka-Stadnik & Zakrzewska, 2021).

A substantial and continuously growing body of literature examines the relationship between sexual education and sexual activity. Far fewer authors have investigated the relationship between sexual knowledge and sexual activity. The United Nations Expert Paper by Kirby (2011) reviewed 73 studies focusing on sexual education and the timing of initial sexual intercourse and concluded that 48 studies found no significant relationship, while the rest of the studies documented that participants in sexual education programs revealed delayed sexual intercourse. Among 38 studies focusing on sexual education and the frequency of sexual intercourse, 25 studies indicated no significant relationship, 12 studies revealed that participants in sexual education programs reported a lower frequency of sexual intercourse, and one study documented a higher frequency of sexual intercourse among them (Kirby, 2011). Among 41 studies focusing on sexual education and the number of sexual partners, 25 studies indicated no significant relationship, while the rest of the studies documented a lower number of sexual partners among sexual education program participants (Kirby, 2011). In previous studies in the field, sexual education has been broadly measured by a single item, namely, a binary variable, e.g., participation in formal sexual education. Most noteworthy recent studies on sexual activity and sexual knowledge sensu stricto are summarized in the next paragraph. They were not reviewed by Kirby (2011).

Andersson-Ellström et al. (1996) found knowledge of sexually transmitted diseases to be positively related to sexual experience and the number of sexual partners among adolescent women. Chao et al. (2010) demonstrated that sexual knowledge was positively associated with the frequency of sexual activity among young adult male military conscripts. Part et al. (2008) indicated a positive relationship between sexual knowledge and sexual activity among school pupils aged 14−16 years. Most recently, a good understanding of sexual concepts was identified by Osadolor et al. (2022) to be negatively related to sexual initiation based on a combined sample of adolescent men and women. The results of the four studies demonstrate that testing for the nature (i.e., sign) of the effect of sexual knowledge on sexual activity has given inconsistent results; however, a positive relationship has been predominant. The abovementioned studies by Andersson-Ellström et al. (1996), Chao et al. (2010), Part et al. (2008), and Osadolor et al. (2022) did not use a latent variable approach.

Analyses of boys’ first sexual experiences are strongly underrepresented relative to analyses of girls’ first sexual experiences (Ott et al., 2012). However, some studies examined gender differences in the relationship between sexual activity and sexual education or knowledge. For example, Cueto and Leon (2016) found that among adolescent men, unlike among adolescent women, sexual knowledge was positively associated with having had sexual intercourse. Adolescent men attending school who had received sexual education were found to be more likely to use birth control the first time that they had sexual intercourse than those who had not received sexual education; however, this association was not observed in female counterparts (Mueller et al., 2008). Among young women, AIDS education was shown to be positively related to the likelihood of condom use during intercourse, while this relationship was not significant among young men (Tremblay & Ling, 2005). Aarons et al. (2000) revealed that participation in reproductive health classes postponed sexual initiation among female high school students, while no changes were observed among their male counterparts. The results of the four studies demonstrate that testing for the nature (i.e., sign) of the moderating effect of gender has given inconsistent results.

This study hypothesizes that sexual knowledge is protective against adolescent sexual activity measured by the age of sexual initiation, the number of sexual partners, and the time since the last sexual intercourse (H1). This direction of the relationship is consistent with broad evidence in previous studies (Kirby, 2011). Following this, gender is hypothesized to moderate the relationship between sexual knowledge level and adolescent sexual activity (H2). The conceptual study by Peterson et al. (1994) called attention to the importance of considering gender as a moderator in predicting the age at sexual intercourse.

The hypothesized existence of the moderating effect of gender requires some clarification. Gender differences in the relationship between sexual knowledge level and sexual activity are likely to reflect the motives for sexual activity of adolescent men and women. The stylized facts given by WHO (2011) indicate that traditionally adolescent men are more motivated than adolescent women to have sexual intercourse by curiosity, physical gratification, and because “friends are doing it,” while adolescent women are more likely than adolescent men to report the motives of love, relationship enhancement, obligation to the boy or man, and his promises of gifts or money. Similarly, empirical data from Poland revealed that physical gratification and curiosity are considerably more important motives for sexual initiation for adolescent men than adolescent women (Izdebski, 2012). The strength of these motivations is likely to vary in their resilience to the acquisition of sexual knowledge.

Building on the previous point, the gender difference in the strength of the social pressure, including the motive of “friends are doing it,” stems from traditional sexual roles of male dominance and female passivity (Kiefer & Sanchez, 2007). Social pressure and expectations for gender roles and masculine scripts have been shown to be linked to the overestimation of biological sexual drive by men. Reduced male libido has been stigmatized (Mark & Lasslo, 2018; Nimbi et al., 2020; van Anders, 2012).

The novelty of the current study lies in analyzing the moderating effect of gender on the relationship between sexual knowledge level and sexual activity using a latent variable approach. Knowledge is an example of a latent variable and should be measured using a set of indicators (Rindskopf, 1984). The considered control variables of the educational levels of parents or guardians, grade point average, and the material situation of the household are well-established predictors of sexual activity (França & Frio, 2018; Lee et al., 2018; Reis et al., 2023).

Data and Method

Sample

Data were derived from the quantitative study by the Educational Research Institute in Warsaw entitled “Opinions and expectations of young adults and parents of school-age children on education on psychosexual development and sexuality.” The design of the entire study has been previously described (Bulkowski et al., 2015b). Briefly, individuals born between 1 November 1995 and 31 October 1996 were identified from the Universal Electronic System for Registration of the Population (PESEL). Two thousand four hundred adolescents were selected using two-stage random sampling. The response rate was 51.9%: the total effective sample included 625 adolescent men and 621 adolescent women (hereafter referred to as boys and girls).

Data were collected between 23 February and 13 March 2015, meaning that all participants were aged 18−19 years. In the weighted sample, 91% of boys and 89% of girls declared that they had been in love. Significantly more girls than boys, 54% as opposed to 42%, declared being in a long-term relationship. Furthermore, 88% of girls and 91% of boys were living with their parent(s), while 5% of girls and only 1% of boys were living with their partner. In contrast, more boys than girls, 61% as opposed to 54%, reported having had sexual intercourse. Only 45% of boys reported that they used a condom during their last sexual intercourse. Most frequently, friends and peers, family, and the sexual education teacher (in Poland, literally the family life education teacher) were considered by adolescents to be among the main three sources of information about human sexuality.

Additionally, the “Discussion” section of the current paper provides some figures regarding parents of school-age children. Notably, the sample of parents was independent from the sample of adolescents. In simpler terms, regrettably, the sample of parents included individuals who were not the parents of the adolescents included in the study. The parents of school-age children answered questions about one selected child. In this part of the survey, 4800 individuals were selected using two-stage random sampling and the abovementioned PESEL. The response rate was 52.6%, and the total effective sample included 2526 parents.

Procedure

All presumptive participants received a letter of invitation followed by personal contact. Data were entered anonymously by respondents using computer-assisted self-administered interviews. Only same-sex assistants were used to help each respondent with the completion of the questionnaire, if needed. Verbal informed consent was obtained prior to the interview. The place of living of 49% of the survey participants was visited at least twice due to their absences. E-mail and telephone support during the time the survey was deployed (Bulkowski et al., 2015b). The study protocol was approved by the Ethical Committee at the Educational Research Institute.

Analysis

Partial least squares path modeling was used to test the proposed hypotheses. It allows for estimating complex models including manifest and latent variables. The latent variables represent unobserved (not directly measured) concepts. Five multiple-item latent variables were included in the path model: sexual activity, sexual knowledge level, reproductive system knowledge level, sexual intercourse knowledge level, and the educational levels of parents or guardians. Sexual activity was measured through reflective indicators. The reflective measurement model assumes that the latent construct causes variation in the observed measures. Other constructs were modeled using formative indicators. The formative measurement model assumes that indicators form or cause the construct. The construct of sexual knowledge level was considered a second-order construct formed by two first-order constructs: reproductive system knowledge level and sexual intercourse knowledge level. The variables of school performance, material situation, and gender were measured in the single-item framework (Mehmetoglu & Venturini, 2021). In path diagrams, formative latent variables tend to be represented by hexagons rather than ovals. Reflective latent variables are usually represented by ovals.

The moderating effect of gender was analyzed through the two-stage path modeling approach, which is suitable for testing models with an interaction between an exogenous latent variable and a dichotomous moderator variable (Mehmetoglu & Venturini, 2021). Additionally, the paper reports the results of the bootstrap t-test (Sanchez, 2013), which was used to evaluate the statistical significance of gender differences. All analyses are replicable, as the modeled data, along with the complete questionnaire, are publicly available online in the Polish Social Data Archive (Bulkowski et al., 2015a). Partial least squares path modeling was performed using the “plspm” R package (Sanchez et al., 2022).

The nearest neighbor method was selected as the imputation method for indicator missing values. The method uses the \(k\) observations most similar to those whose missing values are imputed. The number of considered neighbors, \(k\), is usually between 3 and 7, with 5 often used as the default (Mehmetoglu & Venturini, 2021), which was also applied in the current study. The single imputation was performed using the “VIM” R package (Kowarik & Templ, 2016).

Measures

Adolescent sexual activity was measured using three different items: age of sexual initiation, number of sexual partners, and time since last sexual intercourse. Early sexual initiation was indicated by the age of initiation of sexual intercourse with the categories: has not had sex, 18–19 years of age, 17 years of age, 16 years of age, and 10–15 years of age. Extreme categories of the variable were combined due to low counts. The number of sexual partners in the last 12 months was indicated by a five-category variable with the categories: has not had sex, 1, 2, 3, and 4 or more persons. The short time since last sexual intercourse was represented by a seven-category variable with the categories: has not had sex, more than a year ago, 6–12 months ago, 3–6 months ago, 1–3 months ago, less than a month ago, and less than a week ago. The category “Has not had sex” was created based on a filtering question.

Knowledge of the reproductive system was measured by six eight-point items. Namely, respondents were asked to assign physiological functions to the testicles, phallus, vas deferens, ovaries, uterus, and uterine tubes. Knowledge of sexual intercourse was measured by four three-point items. Respondents were asked to assess statements on the chances of becoming pregnant after the first sexual intercourse, the chances of becoming pregnant post-menstruation, after sexual intercourse in a standing position, and after coitus interruptus. Answers to these questions were recoded to binary variables (correct and incorrect answers). Answers “Do not know” were coded the same as incorrect answers. The knowledge measures were customized to align with the levels of expertise in the country where the study was conducted, as explained by Bulkowski et al. (2015b).

Some previous studies in the area did not include any control variables, which has raised doubts about the spuriosity of the identified relationship between sexual education and sexual activity (Marsiglio & Mott, 1986). Initially, in the current analysis, the set of control variables included the educational levels of parents or guardians, grade point average, and material situation of the household. The latter was removed from the model based on significance testing. The educational level of parents or guardians was indicated by two four-category variables with the categories of elementary or incomplete elementary education, basic vocational education, secondary education, and higher education. Technically, the educational level of parents or guardians was measured with a two-item formative scale. Some variables were measured using single indicators, which is known as single-item measurement (Mehmetoglu & Venturini, 2021). The grade point average was measured by an eight-point indicator, with the lowest category being 2.0 or less and the highest category being 5.0 or more. Poland has a grading scheme ranging from 1 to 6, where 6 is the best, 1 the worst, and 2 the minimum passing mark. The material situation of the household was surveyed using a seven-point question, with the answer options ranging from a situation where a household did not have sufficient money to buy the cheapest food to a situation where a household had sufficient money for all its needs and was saving money for future purposes. In the questionnaire, the question on gender identity had four options; however, only two (man and woman) were selected by respondents. Gender was treated as a moderator in the tested models. Importantly for interpretation, the variable of gender took the value of − 1.00 for girls and 1.00 for boys.

The reflective measurement model of sexual activity met all assessment criteria proposed by Mehmetoglu and Venturini (2021): \(\lambda : 2.39>1.00\), Dillon-Goldstein’s \(\rho :0.92>0.70\), \(AVE:0.82>0.50.\) Formative measurement models require different assessment criteria than those applied to reflective measurement models. The assessment criteria by Mehmetoglu and Venturini (2021) were considered while developing formative measurement models of reproductive system knowledge level, sexual intercourse knowledge level, and educational levels of parents or guardians: \(VIF: 1.14<2.50, 1.07<2.50, 1.07<2.50\).

Results

A path model was built to study the relationships between sexual knowledge level and adolescent sexual activity measured by the age of sexual initiation, number of sexual partners, and time since last sexual intercourse, while controlling for grade point average and educational levels of parents or guardians. The examination of statistical significance led to the exclusion of the control variable of the material situation of the household. The inner (structural) part of the model is visualized in Fig. 1. The level of sexual knowledge was lower in boys (with a mean of − 0.104) than in girls (0.105). Importantly for interpretation, as usually done in the partial least squares framework, the variance of all latent variables was set to one, and all latent means were fixed to zero. The gender difference was comparable in the first-order latent variable of sexual intercourse knowledge level (− 0.093 versus 0.094) and in the first-order latent variable of reproductive system knowledge level (− 0.082 versus 0.083).

Fig. 1
figure 1

Source: Own elaboration

Path diagram of the inner part of the estimated model. Notes: SEX − sexual activity, GPA − grade point average, PAR − educational levels of parents or guardians, GEN − gender (moderator), KNO − sexual knowledge level (second-order construct), REP − reproductive system knowledge level (first-order construct), INT − sexual intercourse knowledge level (first-order construct).

Full size image

Table 1 reports coefficients of the direct paths from the educational level of parents or guardians, grade point average, gender, and sexual knowledge level (second-order latent variable) to adolescent sexual activity measured by the age of sexual initiation, number of sexual partners, and time since last sexual intercourse. The interaction term of gender and sexual knowledge level is also included in this output. Put another way, Table 1 reports the estimate of the moderating effect of gender on the relationship between sexual knowledge level and sexual activity according to the two-stage approach (Mehmetoglu & Venturini, 2021). As shown in Table 1, the gender difference appeared to be highly significant (\(p=0.003\)). Additionally, Table 2 reports gender differences in the path coefficients estimated using the resampling parametric approach (Sanchez, 2013). The difference in the coefficient of the path from the sexual knowledge level (second-order latent variable) to sexual activity appeared to be highly significant (\(p=0.001\)).

Table 1 Coefficients of the direct paths in the inner model with endogenous adolescent sexual activity
Full size table
Table 2 Gender differences in the path coefficients of the inner model with endogenous adolescent sexual activity
Full size table

Less technically, among boys, early and intense sexual activity was associated with a low level of sexual knowledge, also when adjusting for socio-economic controls. Among girls, the relationship between being sexually active and sexual knowledge appeared to be insignificant, also when adjusting for socio-economic controls. The gender difference in the relationship between sexual knowledge level and sexual activity was highly significant according to the two-stage approach (Table 1) and bootstrap t-test (Table 2). In contrast, the control variables of the educational levels of parents or guardians and grade point average appeared to be highly significant covariates among both boys and girls (\(p<0.001\)). More precisely, irrespective of gender, the high educational achievement of respondents and their parents or guardians was protective against early and intense sexual activity.

The outer model shows the relationships between the observed measures and the latent variables of sexual activity, educational levels of parents or guardians, and sexual knowledge level formed by the reproductive system knowledge level and sexual intercourse knowledge level. For example, as reported in Table 3, the first-order latent variable of reproductive system knowledge level and the second-order latent variable of sexual knowledge level were positively loaded by the correct answers to the questions on the physiological functions of the testicles, phallus, vas deferens, ovaries, uterus, and uterine tubes. The first-order latent variable of sexual intercourse knowledge and the second-order latent variable of sexual knowledge level were positively loaded by the correct answers to the questions on the chances of becoming pregnant after the first sexual intercourse, the chances of becoming pregnant post-menstruation, after sexual intercourse in a standing position, and after coitus interruptus.

Table 3 Loadings of the outer model
Full size table

Discussion

The estimated path model indicates that, among boys, a high sexual knowledge level was associated with delayed and decreased sexual activity. The relationship was robust to the inclusion of potential confounders (H1). Among girls, the relationship between sexual knowledge level and sexual activity was insignificant, also with adjustment for control variables. Notably, the gender differences appeared to be highly important and were consistent across the two specifications presented in the “Results” section (H2). The self-selection of boys with low sexual knowledge levels into early and intense sexual activity should be an important area of interest for public health researchers and practitioners. The reduction of gender imbalances is one of the aims of sexual health policy (WHO, 2006). Notably, in the country from which the data analyzed in the article originate, the male-to-female ratio of sexually transmitted infections reached 5:1 (Niedźwiedzka-Stadnik & Zakrzewska, 2021). The most commonly reported emotional state in boys (expectant fathers) upon learning about their partner’s pregnancy was fear and shock (Ratajczak, 2016).

The information–motivation–behavioral–skills model places knowledge within a broader framework of a general theory for the prediction of health-related behavior (Fisher & Fisher, 2000). This research has demonstrated that there is no direct link between knowledge and behavior. It has been argued that knowledge is a necessary but not a sufficient condition for the execution of a given behavior. Instead, knowledge has been proposed to be related to motivational factors and behavioral skills, which, in turn, predict actual behavior (Fishbein & Ajzen, 2011). As explained in the “Introduction” section, boys and girls differ in the motivational factors of sexual activity (Izdebski, 2012; WHO, 2011). These groups of motivations are likely to exhibit differences in susceptibility to the acquisition of sexual knowledge, which is reflected by the revealed moderating effect.

As shown by Kirby (2011), most previous studies have not reported a significant link between participation in sexual education programs and sexual activity. Of those reporting such an association, almost all found this relationship to be negative. This evidence is important in the context of the current analysis because a sexual education teacher was considered to be one of the key sources of information about human sexuality by the surveyed adolescents, as noted in the “Introduction” section. Most noteworthy, the moderating effect of gender on the relationship between sexual education and sexual activity was not considered in the literature review by Kirby (2011). His review focused on the impact of sexual education programs rather than of sexual knowledge in general.

A number of studies have revealed that the feeling of sexual autonomy promotes men’s and women’s sexual function (Haavio-Manilla & Kontula, 1997; Kiefer & Sanchez, 2007; Weinberg et al., 1983). Fine (1988) criticized sexual health policies for suppressing discourse regarding sexual desire, promoting discourse on sexual victimization, and privileging married heterosexuality. Her study was limited to women’s sexuality; however, the idea of incorporating sexual desire into sexual health policies has been extended to men (Lamb, 1997; Lamb et al., 2013). In particular, Lamb (1997) criticized sexual health policies for silencing the sexual desire of men. Considering the high significance of the gender differences in the relationship between sexual knowledge level and sexual activity among adolescents, further studies are needed to explore the silencing effect of sexual knowledge on sexual activity. Notably, in this context, formal sexual education in Poland has been classified as an “abstinence-only” program (type 1 according to WHO, 2010) or as a restrictive model (Jewtuch, 2020; Woźniak, 2015).

Tourangeau and Smith (1996) showed that male respondents are likely to report more opposite-sex partners than their female counterparts, even when omitting any forced intercourse and contacts with prostitutes. They argued that within a closed population, men and women should report the same total number of opposite-sex sexual partners. Overestimation of sexual activity by men or underestimation of sexual activity by women constitutes a limitation of the current analysis; namely, socially undesirable behaviors are misreported in surveys. However, Tourangeau and Smith (1996) revealed that questionnaire self-administration tends to reduce the disparity between men and women in the number of sexual partners reported. As noted in the “Procedure” subsection, in the current study, data were entered anonymously by respondents using computer-assisted self-administered interviews. Over 85% of adolescents completed the questionnaire independently without any interviewer assistance.

The present analysis also reveals that, on average, sexual knowledge level was lower among boys than girls, with the latent variable means of − 0.104 and 0.105, respectively, while the variance of the latent variable was one. For example, as noted in the “Measures” subsection, respondents were asked to assign physiological functions to the testicles, phallus, vas deferens, ovaries, uterus, and uterine tubes. The difference in knowledge of every element of the male reproductive system was not significant between boys and girls at the 0.05 level. However, girls were more aware of elements of the female reproductive system than boys (\(p<0.01\), tests of two independent proportions). Important reasons for the gender gap in sexual knowledge are family and school communication. In the sample of adolescents under analysis, 57% of boys and 66% of girls considered family to be one of the sources of sexual knowledge acquisition. In the independently selected sample of parents of school-age children, 43% of parents of daughters compared to 50% of parents of boys reported that their child was very shy or rather shy when speaking about her or his sexuality. Additionally, boys were less likely to regularly attend sexual education classes than girls (43% versus 52%).

Conclusions

The analysis reveals gender differences in the relationship between sexual knowledge and adolescent sexual activity. The modeling results indicate that, among boys, a low sexual knowledge level was associated with early and intense sexual activity. Among girls, the relationship between sexual knowledge level and sexual activity was insignificant. The self-selection of boys with low sexual knowledge levels into early and intense sexual activity may be of interest to public health, given that adolescent sexual activity can be shaped by sexual health policies (Cowan, 2002). Previous research has demonstrated that the socio-economic characteristics of adolescent fathers differ radically from those of the corresponding population (Ratajczak, 2016). Considering the abovementioned self-selection of boys with low sexual knowledge levels, it should be noted that, additionally, the average level of sexual knowledge was also lower among boys than among girls. The revealed moderating effect was explained by gender differences in the motives of sexual activity of adolescent men and women (Izdebski, 2012; WHO, 2011). Further studies are needed to explore the silencing effect of sexual knowledge on sexual activity.