Sexual Behavior in China:Trends and Comparisons WILLIAM L.PARISH EDWARD O.LAUMANN SANYU A.MOJOLA By ALL ACCOUNTS,the demographic,socio-cultural,and economic landscape of China has changed dramatically over the past several decades.Several studies have examined current sexual behavior,including attention to in- creases in premarital conceptions,divorce rates,extramarital affairs,coital frequency,and relaxed attitudes toward premarital sex(Rindfuss and Morgan 1983;Wang and Yang 1996;Tang and Parish 2000:Farrer 2002;Higgins et al.2002).Social surveys have been limited to the provincial or city level and often based on opportunistic samples (e.g.,Liu et al.1997).The potential threat of a generalized HIV epidemic in China(Detels 2004;Thompson 2005; Beyrer 2003)highlights the need for population-based information on current sexual behavior that could help prevent misguided intervention strategies. For example,worldwide,half of HIV infections are estimated to occur among young people under the age of 24 years(Kiragu 2001).To what extent this rate is a valid estimate also for China is unknown. In 2000,the first national probability survey of adult sexual behavior in China was completed.We use these data to address three questions.First, has there been a revolution in sexual behavior in China?We examine eight trends covering two markers of change:a)the onset of sexual behavior(pu- berty,masturbation,age at first sex and marriage,premarital sex),and b) commodification of sex (use of pornographic materials and commercial sex) and concurrent sex.We present a historical overview of previous studies and analyze findings from the national survey.Second,is China unique in these transformations?We seek to situate findings pointing to a sexual revolution in China within a regional and global context,considering whether China is different or follows patterns similar to those in other parts of the world.Third, what are the implications of these findings for China's risk of a generalized HIV epidemic?As suggestive of possible future disease patterns,we examine POPULATION AND DEVELOPMENT REVIEW 33(41 729-756 (DECEMBER 20071 729 This content downloaded from 128.235.251.160 on Tue,3 Feb 2015 11:39:51 AM All use subject to JSTOR Terms and Conditions
Sexual Behavior in China: Trends and Comparisons William L. Parish Edward O. Laumann Sanyu A. Mojola By all accounts, the demographic, socio-cultural, and economic landscape of China has changed dramatically over the past several decades. Several studies have examined current sexual behavior, including attention to in creases in premarital conceptions, divorce rates, extramarital affairs, coital frequency, and relaxed attitudes toward premarital sex (Rindfuss and Morgan 1983; Wang and Yang 1996; Tang and Parish 2000; Farrer 2002; Higgins et al. 2002). Social surveys have been limited to the provincial or city level and often based on opportunistic samples (e.g., Liu et al. 1997). The potential threat of a generalized HIV epidemic in China (Detels 2004; Thompson 2005; Beyrer 2003) highlights the need for population-based information on current sexual behavior that could help prevent misguided intervention strategies. For example, worldwide, half of HIV infections are estimated to occur among young people under the age of 24 years (Kiragu 2001). To what extent this rate is a valid estimate also for China is unknown. In 2000, the first national probability survey of adult sexual behavior in China was completed. We use these data to address three questions. First, has there been a revolution in sexual behavior in China? We examine eight trends covering two markers of change: a) the onset of sexual behavior (pu berty, masturbation, age at first sex and marriage, premarital sex), and b) commodification of sex (use of pornographic materials and commercial sex) and concurrent sex. We present a historical overview of previous studies and analyze findings from the national survey. Second, is China unique in these transformations? We seek to situate findings pointing to a sexual revolution in China within a regional and global context, considering whether China is different or follows patterns similar to those in other parts of the world. Third, what are the implications of these findings for China's risk of a generalized HIV epidemic? As suggestive of possible future disease patterns, we examine POPULATION AND DEVELOPMENT REVIEW 33(4): 729-756 (DECEMBER 2007) 729 This content downloaded from 128.235.251.160 on Tue, 3 Feb 2015 11:39:51 AM All use subject to JSTOR Terms and Conditions
730 SEXUAL BEHAVIOR IN CHINA results from urine samples for the sexually transmitted infection chlamydia trachomatis-a potential co-factor of HIV (Royce et al.1997;Fleming and Wasserheit 1999;Rottingen et al.2001).Although sexual behavior retains some distinctly Chinese characteristics,our survey data demonstrate that a sexual revolution is occurring in China,reflected by the fact that the sexual behavior of the country's post open-door generation differs markedly from that of earlier generations. Historical overview of sexual behavior in China Throughout the twentieth century,widespread changes in society,in ex- posure to the West,and in government policy helped shape the Chinese family and Chinese sexual behavior.Prior to the 1949 founding of the People's Republic of China,Chinese families were governed in many ways by a Confucianist ideology characterized by filial piety,arranged marriage, and strong pronatalism (Levy 1949).With the establishment of the new socialist government,several changes reshaped the structure and dynamics of Chinese families in sometimes conflicting ways.A well-known example was the raising of women's status through legal reform and encouraging female participation in the labor force and educational system(Tang and Parish 2000)while strictly regulating women's fertility through the one- child policy (Greenhalgh 1990;Scharping 2003).It is within this context of sometimes radical social engineering that biology,culture,and the state became entangled in the varied expressions of sexual behavior. Onset of sexual behavior Declining age at puberty and thus earlier age at sexual maturity are responses to socioeconomic development in general and to improved nutrition in particular (Wysack and Frisch 1982;Bongaarts 1980;Bongaarts and Cohen 1998).Despite political,social,and economic upheaval over the past half century (e.g.,the Great Leap Forward in 1958-60,and the Cultural Revolu- tion in 1966-76),Chinese diets improved in quantity,quality,and variety for both urban and rural dwellers(Popkin 1994;Popkin et al.2001).A biological change in the age at puberty and sexual maturity might well have had an impact on the practice of masturbation.Any change in this practice,how- ever,would be striking given that Chinese tradition viewed masturbation as harmful to one's health.Into the 1980s adolescents continued to be warned against masturbation(Ruan 1991;Honig and Hershatter 1988:55,58;Liu et al.1997).These maturation changes might also have contributed to a reduc- tion in age at first sex (Udry and Cliquet 1982). This content downloaded from 128.235.251.160 on Tue,3 Feb 2015 11:39:51 AM All use subject to JSTOR Terms and Conditions
730 Sexual Behavior in China results from urine samples for the sexually transmitted infection chlamydia trachomatis?a potential co-factor of HIV (Royce et al. 1997; Fleming and Wasserheit 1999; Rottingen et al. 2001). Although sexual behavior retains some distinctly Chinese characteristics, our survey data demonstrate that a sexual revolution is occurring in China, reflected by the fact that the sexual behavior of the country's post open-door generation differs markedly from that of earlier generations. Historical overview of sexual behavior in China Throughout the twentieth century, widespread changes in society, in ex posure to the West, and in government policy helped shape the Chinese family and Chinese sexual behavior. Prior to the 1949 founding of the People's Republic of China, Chinese families were governed in many ways by a Confucianist ideology characterized by filial piety, arranged marriage, and strong pronatalism (Levy 1949). With the establishment of the new socialist government, several changes reshaped the structure and dynamics of Chinese families in sometimes conflicting ways. A well-known example was the raising of women's status through legal reform and encouraging female participation in the labor force and educational system (Tang and Parish 2000) while strictly regulating women's fertility through the one child policy (Greenhalgh 1990; Scharping 2003). It is within this context of sometimes radical social engineering that biology, culture, and the state became entangled in the varied expressions of sexual behavior. Onset of sexual behavior Declining age at puberty and thus earlier age at sexual maturity are responses to socioeconomic development in general and to improved nutrition in particular (Wysack and Frisch 1982; Bongaarts 1980; Bongaarts and Cohen 1998). Despite political, social, and economic upheaval over the past half century (e.g., the Great Leap Forward in 1958-60, and the Cultural Revolu tion in 1966-76), Chinese diets improved in quantity, quality, and variety for both urban and rural dwellers (Popkin 1994; Popkin et al. 2001). A biological change in the age at puberty and sexual maturity might well have had an impact on the practice of masturbation. Any change in this practice, how ever, would be striking given that Chinese tradition viewed masturbation as harmful to one's health. Into the 1980s adolescents continued to be warned against masturbation (Ruan 1991; Honig and Hershatter 1988: 55, 58; Liu et al. 1997). These maturation changes might also have contributed to a reduc tion in age at first sex (Udry and Cliquet 1982). This content downloaded from 128.235.251.160 on Tue, 3 Feb 2015 11:39:51 AM All use subject to JSTOR Terms and Conditions
WILLIAM L.PARISH EDWARD O.LAUMANN SANYU A.MOJOLA 731 Particularly in cultures where marriage is nearly universal,it is often,as Carael (1995:76)noted,"the pole around which sexual culture is organized." In China,marriage was the most likely context for first sex.Before 1950,ar- ranged matches were the norm,and a bride and groom typically met for the first time on their wedding day (Wang and Yang 1996).Over the ensuing decades,the Chinese state instituted several legislative measures,including marriage laws in 1950 and 1980 that abolished arranged and polygamous marriages while reinforcing the socialist vision of marriage as a union of equals.State administrative measures helped raise the age at first marriage at various times (Banister 1987;Palmer 1995;Zhang 2002;Yan 2003).Strict control over behavior in schools and work units helped to limit sexual activ- ity outside of marriage (Meijer 1971;Ruan 1991).Starting with a new mar- riage law in 1980,marriage ages of men and women were legally allowed to fall (Banister 1987;Davis and Harrell 1993).This in turn was accompanied by a loosening of work unit and other controls over social behavior.Thus while courting and romantic love matches may have increased the desire for premarital sex in earlier decades,opportunities became more widely available only later (Whyte and Parish 1984;Whyte 1993;Farrer 2002;Yan 2003). Commodification of sex In the 1980s China opened its economy to foreign direct investment and trade and undertook a radical domestic program of economic moderniza- tion (Tang and Parish 2000).One unintended consequence was the arrival in south coastal provinces of vast quantities of pornographic photographs, videos,and other materials originating from the West and Hong Kong (Her- shatter 1997).As a result,while the state maintained a consistent line toward pornography-it remains officially illegal-it can now be easily obtained in urban China.Dissemination of these materials was facilitated by the capacity for VCD production in small workshops and the popular use of VCRs,com- puters,and VCD players.Friendship networks rapidly spread these materials, not only among youth but also among adults.Given the relative scarcity of alternative sources (Gao et al.2001),pornography was a ready source of basic information about sex for Chinese youth (Liu et al.1997).The internet is no doubt gaining increased significance in this regard. The state's approach toward the commercial sex industry varied mark- edly over the twentieth century,from tolerance and implicit acceptance before 1949 to outright prohibition and imprisonment of sex workers in the 1950s (Dikotter 1997;Gil et al.1996;Micollier 2004;Ruan 1991;Pan 2004a, 2004b).Starting in the early 1980s,several factors created increasing demand for and supply of commercial sex.Arguably,the commercial sex industry might find its most favorable soil in patriarchal societies with a sexual double standard:chastity and fidelity expected for women combined with accep- This content downloaded from 128.235.251.160 on Tue,3 Feb 2015 11:39:51 AM All use subject to JSTOR Terms and Conditions
William L. Parish / Edward O. Laumann / Sanyu A. Mojola 731 Particularly in cultures where marriage is nearly universal, it is often, as Carael (1995: 76) noted, "the pole around which sexual culture is organized." In China, marriage was the most likely context for first sex. Before 1950, ar ranged matches were the norm, and a bride and groom typically met for the first time on their wedding day (Wang and Yang 1996). Over the ensuing decades, the Chinese state instituted several legislative measures, including marriage laws in 1950 and 1980 that abolished arranged and polygamous marriages while reinforcing the socialist vision of marriage as a union of equals. State administrative measures helped raise the age at first marriage at various times (Banister 1987; Palmer 1995; Zhang 2002; Yan 2003). Strict control over behavior in schools and work units helped to limit sexual activ ity outside of marriage (Meijer 1971; Ruan 1991). Starting with a new mar riage law in 1980, marriage ages of men and women were legally allowed to fall (Banister 1987; Davis and Harrell 1993). This in turn was accompanied by a loosening of work unit and other controls over social behavior. Thus while courting and romantic love matches may have increased the desire for premarital sex in earlier decades, opportunities became more widely available only later (Whyte and Parish 1984; Whyte 1993; Farrer 2002; Yan 2003). Commodification of sex In the 1980s China opened its economy to foreign direct investment and trade and undertook a radical domestic program of economic moderniza tion (Tang and Parish 2000). One unintended consequence was the arrival in south coastal provinces of vast quantities of pornographic photographs, videos, and other materials originating from the West and Hong Kong (Her shatter 1997). As a result, while the state maintained a consistent line toward pornography?it remains officially illegal?it can now be easily obtained in urban China. Dissemination of these materials was facilitated by the capacity for VCD production in small workshops and the popular use of VCRs, com puters, and VCD players. Friendship networks rapidly spread these materials, not only among youth but also among adults. Given the relative scarcity of alternative sources (Gao et al. 2001), pornography was a ready source of basic information about sex for Chinese youth (Liu et al. 1997). The internet is no doubt gaining increased significance in this regard. The state's approach toward the commercial sex industry varied mark edly over the twentieth century, from tolerance and implicit acceptance before 1949 to outright prohibition and imprisonment of sex workers in the 1950s (Dikotter 1997; Gil et al. 1996; Micollier 2004; Ruan 1991; Pan 2004a, 2004b). Starting in the early 1980s, several factors created increasing demand for and supply of commercial sex. Arguably, the commercial sex industry might find its most favorable soil in patriarchal societies with a sexual double standard: chastity and fidelity expected for women combined with accep This content downloaded from 128.235.251.160 on Tue, 3 Feb 2015 11:39:51 AM All use subject to JSTOR Terms and Conditions
732 SEXUAL BEHAVIOR IN CHINA tance and tolerance of men's pursuit of multiple sexual partners(Lim 1998). During times of rising incomes for men,this double standard-with many men demanding sex and few women available for nonmarital sex-creates a high demand for a small pool of sexually available women,often commercial sex workers.As in other Asian patriarchies,loss of virginity before marriage to a man other than one's fiance sharply diminishes a woman's standing (Pan 2000;Lim 1998).Nevertheless,the lucrative nature of commercial sex work may present a viable economic option for some women in the face of limited income-earning opportunities (Gil et al.1996).In combination,these factors help fuel a continuous supply of commercial sex workers (Lim 1998; Pan1999,2000,2004a,2004b). Survey of trends in Chinese sexual behavior Despite abundant historical and ethnographic work,as well as smaller pro- vincial studies,a national probability survey of Chinese sexual behavior has until recently been a missing component in the literature on China. Data and methods A national probability survey of sexual behavior,the China Health and Fam- ily Life Survey (CHFLS)collected data from China's adult population aged 20-64 years(with the exclusion of Tibet and Hong Kong).Completed during 1999-2000,the sample for the survey was drawn with a four-stage area- probability sampling process according to standard procedures for complex samples (Levy and Lemeshow 1999).Respondents were selected from 14 strata,48 primary sampling units,and 60 neighborhoods,with probabilities proportional to their respective populations.Large cities and coastal areas were oversampled.For the sake of privacy,respondents were not interviewed at home but instead were asked to come to secure neighborhood facilities for an hour-long interview.In large cities,these facilities were typically private rooms in a neighborhood hotel,while in villages and smaller towns they were rooms in a larger home or in a village meeting place.Before the survey began, an informed consent form approved by the institutional review boards was read to the respondent,and the respondent gave oral and computer-entered consent.'The first half of the interview was administered by the interviewer and the second,more sensitive half,by the respondent,who typically had sole control of the laptop computer containing the questionnaire (CASI: computer-assisted self-administered interviews).Only 13 percent (often older women in the countryside)needed assistance with this latter portion. Of the 5,000 people identified to be in the sample universe,75.3 percent (3,766)finished the interview.There was a slightly lower participation rate (68.5 percent)for urine sample collection.Excluding 337 non-refusals/fail- This content downloaded from 128.235.251.160 on Tue,3 Feb 2015 11:39:51 AM All use subject to JSTOR Terms and Conditions
732 Sexual Behavior in China tance and tolerance of men's pursuit of multiple sexual partners (Lim 1998). During times of rising incomes for men, this double standard?with many men demanding sex and few women available for nonmarital sex?creates a high demand for a small pool of sexually available women, often commercial sex workers. As in other Asian patriarchies, loss of virginity before marriage to a man other than one's fiance sharply diminishes a woman's standing (Pan 2000; Lim 1998). Nevertheless, the lucrative nature of commercial sex work may present a viable economic option for some women in the face of limited income-earning opportunities (Gil et al. 1996). In combination, these factors help fuel a continuous supply of commercial sex workers (Lim 1998; Pan 1999, 2000, 2004a, 2004b). Survey of trends in Chinese sexual behavior Despite abundant historical and ethnographic work, as well as smaller pro vincial studies, a national probability survey of Chinese sexual behavior has until recently been a missing component in the literature on China. Data and methods A national probability survey of sexual behavior, the China Health and Fam ily Life Survey (CHFLS) collected data from China's adult population aged 20-64 years (with the exclusion of Tibet and Hong Kong). Completed during 1999-2000, the sample for the survey was drawn with a four-stage area probability sampling process according to standard procedures for complex samples (Levy and Lemeshow 1999). Respondents were selected from 14 strata, 48 primary sampling units, and 60 neighborhoods, with probabilities proportional to their respective populations. Large cities and coastal areas were oversampled. For the sake of privacy, respondents were not interviewed at home but instead were asked to come to secure neighborhood facilities for an hour-long interview. In large cities, these facilities were typically private rooms in a neighborhood hotel, while in villages and smaller towns they were rooms in a larger home or in a village meeting place. Before the survey began, an informed consent form approved by the institutional review boards was read to the respondent, and the respondent gave oral and computer-entered consent.1 The first half of the interview was administered by the interviewer and the second, more sensitive half, by the respondent, who typically had sole control of the laptop computer containing the questionnaire (CASI: computer-assisted self-administered interviews). Only 13 percent (often older women in the countryside) needed assistance with this latter portion. Of the 5,000 people identified to be in the sample universe, 75.3 percent (3,766) finished the interview. There was a slightly lower participation rate (68.5 percent) for urine sample collection. Excluding 337 non-refusals/fail This content downloaded from 128.235.251.160 on Tue, 3 Feb 2015 11:39:51 AM All use subject to JSTOR Terms and Conditions
WILLIAM L.PARISH EDWARD O.LAUMANN SANYU A.MOJOLA 733 ures(moved away,never at home,data handling errors),the completion rate was 80.8 percent overall and 73.5 percent for the urine sample collection.Of the total sample,17.9 percent (897)refused to be interviewed (before,at the start,or during the interview),and 6.8 percent(340)either refused to pro- vide a urine sample or there were sample handling problems.These response rates compare favorably with those of other large surveys on sexual behavior, notably the US National Health and Social Life Survey 1992 response rate of 79 percent(Laumann et al.1994),the British National Survey 2000 rate of 65.4 percent(Fenton et al.2001a),and the French ACSF rate of 76.5 percent (ACSF 1992;Spira et al.1994). When weighted according to sampling fractions,the data set of suc- cessful interviews had too few individuals in their 20s and too few in their 50s and between 60 and 65,as judged by national census figures and annual population surveys conducted by the government.Thus in terms of biases in response rates,both youth(who one might assume were less traditional) and older individuals (who one might assume were more traditional)were undersampled.We also further adjusted the analysis weights to make the age distribution approximate what is found in national-level data sources. Since those who failed to provide a urine sample were only marginally dif- ferent from those who did,judged by tests of significance,we made no fur- ther adjustments among those who provided urine.After adjusting weights for both sampling fractions and age distributions,we found that percentage distributions for urban residence,age,and education closely paralleled data available in the national census and other national-level statistics,implying that the adjusted sample was representative of the total adult working-age population. The sensitive nature of surveys on sexual behavior poses a challenge in terms of gathering high-quality data (Catania et al.1990;Fenton et al. 2001b;Cleland et al.2004;Curtis and Sutherland 2004).Even with a rela- tively high response rate,misreporting on specific items can still exist.In any cross-sectional survey,differences between cohorts can merely reflect differences in the freedom of different age groups to report particular kinds of behavior,as opposed to capturing real behavioral shifts over time.Neverthe- less,our findings are supported by the close correspondence between many of the trends discerned in the survey and historical and current trends cited in the overview above.Methodologically,a combination of both face-to-face interviews and computer-assisted self-administered interviews-the latter for the more sensitive half of the survey-has no doubt helped to minimize self-presentation bias (Catania et al.1990). To assess potential validity of the responses,two strategies were used in pre-testing.First,50 husband-wife pairs were given shortened versions of the questionnaire.The kappa statistics for reports on their shared sexual behavior averaged a modest 0.35.Nevertheless,this degree of agreement is in the same This content downloaded from 128.235.251.160 on Tue,3 Feb 2015 11:39:51 AM All use subject to JSTOR Terms and Conditions
William L. Parish / Edward O. Laumann / Sanyu A. Mojola 733 ures (moved away, never at home, data handling errors), the completion rate was 80.8 percent overall and 73.5 percent for the urine sample collection. Of the total sample, 17.9 percent (897) refused to be interviewed (before, at the start, or during the interview), and 6.8 percent (340) either refused to pro vide a urine sample or there were sample handling problems. These response rates compare favorably with those of other large surveys on sexual behavior, notably the US National Health and Social Life Survey 1992 response rate of 79 percent (Laumann et al. 1994), the British National Survey 2000 rate of 65.4 percent (Fenton et al. 2001a), and the French ACSF rate of 76.5 percent (ACSF 1992; Spira et al. 1994). When weighted according to sampling fractions, the data set of suc cessful interviews had too few individuals in their 20s and too few in their 50s and between 60 and 65, as judged by national census figures and annual population surveys conducted by the government. Thus in terms of biases in response rates, both youth (who one might assume were less traditional) and older individuals (who one might assume were more traditional) were undersampled. We also further adjusted the analysis weights to make the age distribution approximate what is found in national-level data sources. Since those who failed to provide a urine sample were only marginally dif ferent from those who did, judged by tests of significance, we made no fur ther adjustments among those who provided urine. After adjusting weights for both sampling fractions and age distributions, we found that percentage distributions for urban residence, age, and education closely paralleled data available in the national census and other national-level statistics, implying that the adjusted sample was representative of the total adult working-age population. The sensitive nature of surveys on sexual behavior poses a challenge in terms of gathering high-quality data (Catania et al. 1990; Fenton et al. 2001b; Cleland et al. 2004; Curtis and Sutherland 2004). Even with a rela tively high response rate, misreporting on specific items can still exist. In any cross-sectional survey, differences between cohorts can merely reflect differences in the freedom of different age groups to report particular kinds of behavior, as opposed to capturing real behavioral shifts over time. Neverthe less, our findings are supported by the close correspondence between many of the trends discerned in the survey and historical and current trends cited in the overview above. Methodologically, a combination of both face-to-face interviews and computer-assisted self-administered interviews?the latter for the more sensitive half of the survey?has no doubt helped to minimize self-presentation bias (Catania et al. 1990). To assess potential validity of the responses, two strategies were used in pre-testing. First, 50 husband-wife pairs were given shortened versions of the questionnaire. The kappa statistics for reports on their shared sexual behavior averaged a modest 0.35. Nevertheless, this degree of agreement is in the same This content downloaded from 128.235.251.160 on Tue, 3 Feb 2015 11:39:51 AM All use subject to JSTOR Terms and Conditions
734 SEXUAL BEHAVIOR IN CHINA range as agreement about other aspects of family life such as spousal violence and relative social status of parents (average =0.27),suggesting that it was not the sensitivity of sexual behavior alone that caused reporting problems. Another 50 respondents had repeat interviews after a gap of two months.In the repeat interviews,the 21 items about sexual behavior had average kappa statistic values of 0.75 across interviews.In the final survey,the biomarker data-the urine sample collected to test for chlamydia trachomatis-helped us to assess and confirm trends in reported sexual behavior. We analyzed both urban and rural trends.We give estimates for urban areas in more detail,both because the cutting edge of social change is in urban areas and because of a purposive oversampling of cities that produced more robust estimates for urban areas.In the urban portion of the sample, 44 percent of respondents grew up in rural areas and migrated to cities,often decades ago.Although these persons differed in many ways from those who grew up in cities,we count them here simply as "urban,"leaving the explora- tion of their differences from other urbanites for a later analysis. Data were analyzed using STATA 8.0/9.2(Stata Corp)svy methods with results adjusted for sample strata,primary sampling units,and popula- tion weights.For Figures 1-3,event history analyses were used to calculate cohort survival curves (the cumulative percent experiencing each event), while for Figure 4,and Tables 1-4,weighted cross-tabulations were used. Tests for trend analyses were also performed using Cox and logistic regres- sion adjusted for sample design.We present our results by cohort-grouping respondents into four or five cohorts based on the year a respondent turned age 20(Figures 1-4,Tables 1-4)or the year the respondent turned age 15 for the initial analysis of puberty. Results To address the first question of whether there has been a revolution in sexual behavior,we investigate transformations in a)the onset of sexual behavior, including trends in age at puberty,masturbation,first sex,first marriage,and premarital sex;and b)commodification of sex (including trends in use of pornographic materials and commercial sex)and concurrent sex. Onset of sexual behavior in urban areas Puberty.In the survey data,the average age at puberty declined.Among urban women,age at menarche declined from a high of age 15.9 years for those aged 15 in 1950 to a low of age 13.6 in 1995,a reduction of more than two years.A similar two-year decline in average age at puberty was observed This content downloaded from 128.235.251.160 on Tue,3 Feb 2015 11:39:51 AM All use subject to JSTOR Terms and Conditions
734 Sexual Behavior in China range as agreement about other aspects of family life such as spousal violence and relative social status of parents (average = 0.27), suggesting that it was not the sensitivity of sexual behavior alone that caused reporting problems. Another 50 respondents had repeat interviews after a gap of two months. In the repeat interviews, the 21 items about sexual behavior had average kappa statistic values of 0.75 across interviews. In the final survey, the biomarker data?the urine sample collected to test for chlamydia trachomatis?helped us to assess and confirm trends in reported sexual behavior. We analyzed both urban and rural trends. We give estimates for urban areas in more detail, both because the cutting edge of social change is in urban areas and because of a purposive oversampling of cities that produced more robust estimates for urban areas. In the urban portion of the sample, 44 percent of respondents grew up in rural areas and migrated to cities, often decades ago. Although these persons differed in many ways from those who grew up in cities, we count them here simply as "urban," leaving the explora tion of their differences from other urbanites for a later analysis. Data were analyzed using STATA 8.0/9.2 (Stata Corp) svy methods with results adjusted for sample strata, primary sampling units, and popula tion weights. For Figures 1-3, event history analyses were used to calculate cohort survival curves (the cumulative percent experiencing each event), while for Figure 4, and Tables 1-4, weighted cross-tabulations were used. Tests for trend analyses were also performed using Cox and logistic regres sion adjusted for sample design. We present our results by cohort?grouping respondents into four or five cohorts based on the year a respondent turned age 20 (Figures 1-4, Tables 1-4) or the year the respondent turned age 15 for the initial analysis of puberty. Results To address the first question of whether there has been a revolution in sexual behavior, we investigate transformations in a) the onset of sexual behavior, including trends in age at puberty, masturbation, first sex, first marriage, and premarital sex; and b) commodification of sex (including trends in use of pornographic materials and commercial sex) and concurrent sex. Onset of sexual behavior in urban areas Puberty. In the survey data, the average age at puberty declined. Among urban women, age at menarche declined from a high of age 15.9 years for those aged 15 in 1950 to a low of age 13.6 in 1995, a reduction of more than two years. A similar two-year decline in average age at puberty was observed This content downloaded from 128.235.251.160 on Tue, 3 Feb 2015 11:39:51 AM All use subject to JSTOR Terms and Conditions
WILLIAM L.PARISH EDWARD O.LAUMANN SANYU A.MOJOLA 735 for urban men-from age 17.5 among men aged 15 in 1950 to age 15.5 by 1995.These trends were consistent with proxies for improved nutrition over the same period:self-reported height increased by 4.5 cm(1.8 inches)among men and by 6.3 cm (2.5 inches)among women in urban areas from 1950 to 1995.The decline in age at puberty implies that both men and women were biologically ready for sexual activity at younger ages in more recent years. Masturbation.Self-reports of age at first masturbation suggest that over time the practice became increasingly common in urban areas.In the earliest cohorts who reached age 20 in the 1950s and 1960s,only about 39 percent of men and 5 percent of women had masturbated by age 30.In contrast,by the 1990s cohort,over 70 percent of men and 30 percent of women had mastur- bated by the time they reached age 30-a six-fold increase for women and a 30 percentage point increase for men.Consistent with more general trends in Chinese society,most of the change in onset of masturbation occurred in the final two cohorts,who reached age 20 in the 1980s and 1990s.(Results appear in Figure 1,where a trend line further to the left implies an earlier and more rapidly accelerating start in the percent of respondents reporting age of onset of masturbation. First sex.In urban areas,median age at first sex (intercourse)moved first toward older and then toward younger ages(see Figure 2).For men,the period of most delayed (farthest to the right)age at first sex was in the 1970s: at the median(the 50th percentile),first sex took place at about age 25.This was a time when economic conditions and administrative policies to delay marriage led both to higher marriage ages and to a later onset of sex,which often did not occur until marriage.There was a sharp shift toward earlier ages at first sex from the 1980s cohort on-indicated by a move to the left in the trend lines of Figure 2.What is most distinctive about men's patterns in the 1990s was a sharp shift to the left at around age 20.Median age for men's first sex in this cohort fell to about age 22.5,a decline of more than two years. Among women,the earliest ages at first sex were found in the 1950s and 1960s,when age at marriage and the first sex that went with marriage remained low.By the 1980s,women's median age at first sex had risen from a low of 21.5 years to almost 24 years,but it fell to about 22.5 a decade later in the 1990s-as in the case of men,a decline of almost two years.It is noteworthy that no more than one-fourth of either men or women had sex before age 20.Chinese urban teenagers,in other words,remained mostly virgins even in the 1990s. First marriage.Among urbanites,trends in first marriage displayed most clearly the effects of tight state regulation of marriage age.For both women and men,the two oldest cohorts had both the earliest (1955-69)and most delayed(1970-79)median ages at first marriage.This is consistent with his- This content downloaded from 128.235.251.160 on Tue,3 Feb 2015 11:39:51 AM All use subject to JSTOR Terms and Conditions
William L. Parish / Edward O. Laumann / Sanyu A. Mojola 735 for urban men?from age 17.5 among men aged 15 in 1950 to age 15.5 by 1995. These trends were consistent with proxies for improved nutrition over the same period: self-reported height increased by 4.5 cm (1.8 inches) among men and by 6.3 cm (2.5 inches) among women in urban areas from 1950 to 1995. The decline in age at puberty implies that both men and women were biologically ready for sexual activity at younger ages in more recent years. Masturbation. Self-reports of age at first masturbation suggest that over time the practice became increasingly common in urban areas. In the earliest cohorts who reached age 20 in the 1950s and 1960s, only about 39 percent of men and 5 percent of women had masturbated by age 30. In contrast, by the 1990s cohort, over 70 percent of men and 30 percent of women had mastur bated by the time they reached age 30?a six-fold increase for women and a 30 percentage point increase for men. Consistent with more general trends in Chinese society, most of the change in onset of masturbation occurred in the final two cohorts, who reached age 20 in the 1980s and 1990s. (Results appear in Figure 1, where a trend line further to the left implies an earlier and more rapidly accelerating start in the percent of respondents reporting age of onset of masturbation.) First sex. In urban areas, median age at first sex (intercourse) moved first toward older and then toward younger ages (see Figure 2). For men, the period of most delayed (farthest to the right) age at first sex was in the 1970s: at the median (the 50th percentile), first sex took place at about age 25. This was a time when economic conditions and administrative policies to delay marriage led both to higher marriage ages and to a later onset of sex, which often did not occur until marriage. There was a sharp shift toward earlier ages at first sex from the 1980s cohort on?indicated by a move to the left in the trend lines of Figure 2. What is most distinctive about men's patterns in the 1990s was a sharp shift to the left at around age 20. Median age for men's first sex in this cohort fell to about age 22.5, a decline of more than two years. Among women, the earliest ages at first sex were found in the 1950s and 1960s, when age at marriage and the first sex that went with marriage remained low. By the 1980s, women's median age at first sex had risen from a low of 21.5 years to almost 24 years, but it fell to about 22.5 a decade later in the 1990s?as in the case of men, a decline of almost two years. It is noteworthy that no more than one-fourth of either men or women had sex before age 20. Chinese urban teenagers, in other words, remained mostly virgins even in the 1990s. First marriage. Among urbanites, trends in first marriage displayed most clearly the effects of tight state regulation of marriage age. For both women and men, the two oldest cohorts had both the earliest (1955-69) and most delayed (1970-79) median ages at first marriage. This is consistent with his This content downloaded from 128.235.251.160 on Tue, 3 Feb 2015 11:39:51 AM All use subject to JSTOR Terms and Conditions
736 SEXUAL BEHAVIOR IN CHINA FIGURE I Trends in age at first masturbation among four cohorts of men and women in urban China Men 80 Cohort Significance" ◆1990-2000 ps0.01 ■-1980-89 ps0.01 ±1970-79 40 ◆1955-69 20 10 20 25 30 Age Women 80 Cohort Significance" ◆1990-2000ps0.01 ■-1980-89 ps0.01 +1970-79 ps0.10 40 ◆1955-69 20 10 15 20 25 30 Age NOTES:Cumulative percent of cohort (defined by year a respondent reached age 20)who had first masturbated at each age.Trend lines censored at current age,and for clarity of presentation all results truncated at age 30 because of a plateau in the data after that age. "Significance for whether more recent cohorts differed statistically from the oldest cohort in an event history analysis. SOURCE:China Health and Family Life Survey,1999-2000 torical findings that marriage ages rose sharply in the 1970s when govern- ment pressure to raise marriage ages was most intense and then fell slightly thereafter (e.g.,Banister 1987;Xenos and Gultiano 1992;Wang and Yang 1996;Wu 2003).The data suggest that the effect of government policy was greater on women:while we note little variation in age at first marriage for men(from 24.2 years to 25.2 years)during this period,we note much greater variation among women (from 21.5 years to 24.5 years).Comparing across the results for age at marriage(Figure 3)and age at first sex(Figure 2),the exceptions to parallel trends are most apparent for the 1990s.In that decade, This content downloaded from 128.235.251.160 on Tue,3 Feb 2015 11:39:51 AM All use subject to JSTOR Terms and Conditions
736 Sexual Behavior in China FIGURE 1 Trends in age at first masturbation among four cohorts of men and women in urban China Men 301-1 Cohort Significance3 ^1990-2000 p<;0.01 g 60!.^-*^*-^r^ -m-1980-89 p<s0.01 fc J M-^*""^""^ -A-1970-79 ? 40:.Jjr**.^^-^^^.m.m.m.m.ii _^_l955_69 15 10 20 30 25 Age Women 801-1 Cohort Significance3 +-1990-2000 ps;0.01 - 60. -B-1980-89 psO.01 ? -^-1970-79 psO.10 | 40. -#-1955-69 15 10 20 30 25 Age NOTES: Cumulative percent of cohort (defined by year a respondent reached age 20) who had first masturbated at each age. Trend lines censored at current age, and for clarity of presentation all results truncated at age 30 because of a plateau in the data after that age. Significance for whether more recent cohorts differed statistically from the oldest cohort in an event history analysis. SOURCE: China Health and Family Life Survey, 1999-2000 torical findings that marriage ages rose sharply in the 1970s when govern ment pressure to raise marriage ages was most intense and then fell slightly thereafter (e.g., Banister 1987; Xenos and Gultiano 1992; Wang and Yang 1996; Wu 2003). The data suggest that the effect of government policy was greater on women: while we note little variation in age at first marriage for men (from 24.2 years to 25.2 years) during this period, we note much greater variation among women (from 21.5 years to 24.5 years). Comparing across the results for age at marriage (Figure 3) and age at first sex (Figure 2), the exceptions to parallel trends are most apparent for the 1990s. In that decade, This content downloaded from 128.235.251.160 on Tue, 3 Feb 2015 11:39:51 AM All use subject to JSTOR Terms and Conditions
WILLIAM L.PARISH EDWARD O.LAUMANN SANYU A.MOJOLA 737 FIGURE 2 Trends in age at first sex(intercourse)among four cohorts of men and women in urban China 100 Men Cohort Significance" ◆-1990-2000 ps0.01 80 量1980-89 ps0.01 ±1970-79 60 ●1955-69 40 20 15 20 25 30 35 Age 100 Women Cohort Significance" ◆1990-2000 80 量-1980-89 60 女1970-79 ps0.01 aAne nun 。1955-69 40 20 20 25 30 35 Age NOTES:Cumulative percent of cohort(defined by year a respondent reached age 20)who had first sex (intercourse)at each age.Trend lines censored at current age,and for clarity of presentation all results truncated at age 35 because of a plateau in the data after that age. Significance for whether more recent cohorts differed statistically from the oldest cohort in an event history analysis. SOURCE:China Health and Family Life Survey.1999-2000 men's age at first sex began to move ahead of age at first marriage-implying increased premarital sex.The same is true for women,although the pattern was much more muted. Premarital sex.In urban areas,fewer than 15 percent of either men or women who reached age 20 before 1964 had experienced premarital sex (Figure 4).Subsequently,however,the prevalence of premarital sex began to increase-primarily with her fiance among women and with a greater variety of partners among men.For all types of premarital sex (see Notes to Figure 4),the peak in more recent years was 30 percent for women and 40 percent This content downloaded from 128.235.251.160 on Tue,3 Feb 2015 11:39:51 AM All use subject to JSTOR Terms and Conditions
William L. Parish / Edward O. Laumann / Sanyu A. Mojola 737 FIGURE 2 Trends in age at first sex (intercourse) among four cohorts of men and women in urban China 1001- III! Cohort Significance3 yj^C^^^ "^ ~^~ 1990-2000 psO.01 w 80 ...^..........j^c... _m_ 1980_89 p^o.01 20 15 25 35 30 Age 100|Women-, ?rfr 1 1 1 f Cohort Significance3 jfe^*^ " -^~ 1990-2000 ^ MT 80 _>__ Y 9g0_89 ? 6Q .Sji.... f -*- 1970-79 p<;0.01 / I // J - - 1955-69 20 15 25 35 30 Age NOTES: Cumulative percent of cohort (defined by year a respondent reached age 20) who had first sex (intercourse) at each age. Trend lines censored at current age, and for clarity of presentation all results truncated at age 35 because of a plateau in the data after that age. Significance for whether more recent cohorts differed statistically from the oldest cohort in an event history analysis. SOURCE: China Health and Family Life Survey, 1999-2000 men's age at first sex began to move ahead of age at first marriage?implying increased premarital sex. The same is true for women, although the pattern was much more muted. Premarital sex. In urban areas, fewer than 15 percent of either men or women who reached age 20 before 1964 had experienced premarital sex (Figure 4). Subsequently, however, the prevalence of premarital sex began to increase?primarily with her fiance among women and with a greater variety of partners among men. For all types of premarital sex (see Notes to Figure 4), the peak in more recent years was 30 percent for women and 40 percent This content downloaded from 128.235.251.160 on Tue, 3 Feb 2015 11:39:51 AM All use subject to JSTOR Terms and Conditions
738 SEXUAL BEHAVIOR IN CHINA FIGURE 3 Trends in age at first marriage among four cohorts of men and women in urban China Men 100 Cohort Significance? ◆1990-2000 ps0.01 80 1980-89 ★1970-79 60 ◆1955-69 40 20 19 25 27 29 31 33 35 Age Women 100 Cohort Significance" ◆1990-2000 ps0.01 80 1980-89 ps0.05 ★1970-79 ps0.01 60 Ane nun) ◆1955-69 40 20 17 19 21 23 25 27 29 31 33 35 Age NOTES:Cumulative percent of cohort (defined by year a respondent reached age 20)who had married at each age.Trend lines censored at current age,and for clarity of presentation all results truncated at age 35 because of a plateau in the data after that age. "Significance for whether more recent cohorts differed statistically from the oldest cohort in an event history analysis SOURCE:China Health and Family Life Survey,1999-2000 or more among men.And,for the very youngest cohort,which was only age 20-24 at the time of the survey,the percentage having premarital sex was, of course,understated because respondents had not yet reached the median ages at marriage for men and women. In Figure 4,types of premarital sexual relationship range from those that were commercial to those that led to marriage.In the middle were non- commercial relationships of varying length that did not lead to marriage. Also,among the youngest cohorts were some never-married respondents This content downloaded from 128.235.251.160 on Tue,3 Feb 2015 11:39:51 AM All use subject to JSTOR Terms and Conditions
738 Sexual Behavior in China FIGURE 3 Trends in age at first marriage among four cohorts of men and women in urban China 1 oo U r^?- Cohort Significance3 ^J^J^t---* " ' -#-1990-2000 psO.01 ^ 80.^*^p^. -B-l 980-89 | 60|.Jffi.I ^-1970-79 | J/% -*_1955-69 ol??==^=irf?-1-1-1-1-1-1-1 17 19 21 23 25 27 29 31 33 35 Age Women _ , ^ _. .? I oo_ -^?^?^?g Cohort Significance3 Jj^iMr** T -+-1990-2000 p<;0.01 w 80.*0fc. _?_1980-89 p<;0.05 w j/Jil..... -^-1970-79 psO.01 1 / d/ / -,-1955-69 17 19 21 23 25 27 29 31 33 35 Age NOTES: Cumulative percent of cohort (denned by year a respondent reached age 20) who had married at each age. Trend lines censored at current age, and for clarity of presentation all results truncated at age 35 because of a plateau in the data after that age. Significance for whether more recent cohorts differed statistically from the oldest cohort in an event history analysis. SOURCE: China Health and Family Life Survey, 1999-2000 or more among men. And, for the very youngest cohort, which was only age 20-24 at the time of the survey, the percentage having premarital sex was, of course, understated because respondents had not yet reached the median ages at marriage for men and women. In Figure 4, types of premarital sexual relationship range from those that were commercial to those that led to marriage. In the middle were non commercial relationships of varying length that did not lead to marriage. Also, among the youngest cohorts were some never-married respondents This content downloaded from 128.235.251.160 on Tue, 3 Feb 2015 11:39:51 AM All use subject to JSTOR Terms and Conditions