Archives of Sexual Behavior

, Volume 41, Issue 4, pp 861–873

Sexual Infidelity in China: Prevalence and Gender-Specific Correlates

  • Na Zhang
  • William L. Parish
  • Yingying Huang
  • Suiming Pan
Original Paper

DOI: 10.1007/s10508-012-9930-x

Cite this article as:
Zhang, N., Parish, W.L., Huang, Y. et al. Arch Sex Behav (2012) 41: 861. doi:10.1007/s10508-012-9930-x

Abstract

The nature of extra-relational sex in societies with rapidly changing sexual mores and widespread commercial sex remains under-explored. The 2006 Sexuality Survey of China provides a national probability survey with data on 3,567 people 18–49 years old who were in a marital (89%) or dating/cohabiting (11%) relationship. In attitudes, extramarital sex was completely unacceptable to 74% of women and 60% of men and either somewhat or completely unacceptable to 95% of women and men. Most (77%) women wanted severe punishment of men’s short-term commercial sex and women’s jealousy was equally elevated by their primary partner’s episodes of commercial and non-commercial sex. Nevertheless, the prevalence of infidelity during the last 12 months was 4.5% (women’s non-commercial sex), 11.0% (men’s non-commercial), and 5.5% (men’s commercial), with each percent matching or exceeding the median for other countries. In multivariate equations for non-commercial infidelity, men’s infidelity was significantly more responsive to sexual dissatisfaction with his primary partner while women’s was more responsive to deficits in love. In commercial sex, men were uninfluenced by primary partner deficits in love, sexual satisfaction or oral sex–pursuing, it would seem, simply a greater variety of sexual partners. In a “trading up” pattern, women partnered with low income men had elevated infidelity. The minority of women reporting early masturbation and premarital sex were just as likely as men with these backgrounds to have elevated infidelity. The Chinese patterns provide ample material for deliberations on gender similarities and differences in extra-relational sex.

Keywords

Infidelity Extramarital sex Extrapair sex Gender differences China Commercial sex 

Introduction

Using a national probability sample from China, this article examined on a range of questions surrounding infidelity (mostly just extramarital sex but also extra partners among people with a stable partner of at least a year’s duration). Much as in many former socialist societies in Europe, is it possible that Chinese abhorrence of extramarital sex has softened (Widmer, Treas, & Newcomb, 1998; Zheng et al., 2011)? Moreover, because it typically involves short-term relationships with little or no emotional commitment, do women find a husband’s/partner’s sex with female sex workers (FSW) more tolerable (Buss, 1994; Zheng, 2009)? Paralleling a rapid rise in divorce, has the prevalence of extramarital sex in China reached levels that rival or exceed those of other countries (Farrer & Sun, 2003; Pan, 2002)? Does extra partnering outside stable relationships involve a mixing by age and education with short-term partners in ways that threaten to accelerate the spread of HIV and other sexually transmitted diseases (Anderson & May 1991; Tucker, Chen, & Peeling, 2010)?

Also, do the conditions that promote infidelity differ between men and women? For example, do women attempt to “trade up” in social and economic status (Atkins, Baucom, & Jacobson, 2001; Buss, 1994; Del Giudice, 2009; Hrdy, 1999; Pillsworth & Haselton, 2006; Schmidt, 2003; Small, 1993)? Is men’s impetus for infidelity motivated more by sexual deficits in their stable sexual relationship while women are motivated more by a search for love and affection (Barash & Lipton, 2001; Blow & Hartnett, 2005; Del Giudice, 2009; Glass & Wright, 1985, 1992; McKeganey, 1994; Monto, 2001; Symons, 1979; Thompson, 1984; Weitzer, 2009)? Or, are men and women more similar than different in their extra-relational quests (Petersen & Hyde, 2010)? The length of our list invites superficiality. Nevertheless, the omnibus nature of the 2006 Chinese Sexuality Survey provides some information on each of these questions.

In several ways, China provides a natural experiment: Along with economic development, sexual mores have changed rapidly (Farrer, 2000; Parish, Mojola, & Laumann, 2007; Zheng et al., 2011). In urban areas, below age 45, urban women were educated almost as much as men and mostly in the labor force with an income about 80% of that of men (NBS, 2010). Could it be that the high levels of education, work, and income change have reduced their tolerance of shortfalls in their current spouse/steady partner and increased their chances of finding additional partners (Tang & Parish, 2000)? By 2006, the number of divorces equaled 20% of all marriages in the same year, which was high for China’s level of economic development (NBS, 2010; Tang & Parish, 2000). Moreover, illegal female commercial sex provided an easy path to infidelity (Hong & Li, 2008; Huang, Henderson, Pan, & Cohen, 2004; Jeffreys, 2004, 2006; Pan, 1999, 2000; Tucker et al., 2010). The extent to which the rise in commercial sex and extramarital sex could be attributed to rich businessmen who travel or to migrants absent family and community controls remained a matter of debate (Chang, 2008; Chen et al., 2009; Gaetano, 2008; Uretsky, 2008; Yang, Latkin, Luan, & Nelson, 2010; Zhang, 2001; Zheng, 2006, 2009).

Method

Participants

The 2006 Sexuality Survey of China was a national probability sample of adults with extensive data on individual backgrounds and sexual behavior. Based on a stratified multi-stage cluster sample, the first stage of the sample began with separate urban and rural strata. Next, the survey used probability proportional to size sampling (more populous locales had a greater chance to be selected) to select 75 urban units (city districts and towns) and 47 rural counties. From each of these primary sampling units, two neighborhoods were selected in each urban unit and one village in each rural unit. Then, at the neighborhood/village level, the survey used probability selection to identify individuals age 18–60 years from the local household register (for both permanent and temporary residents). In addition, before the last step above, in urban neighborhoods lacking a temporary resident register, local officials provided an estimate of the number of unregistered migrants and of locations where unregistered migrants lived. The survey team enumerated adult migrants from these locations and then systematically sampled from the new enumeration in proportion to the estimated number of migrants in the neighborhood.

Participants were interviewed not at home but in a nearby meeting place (e.g., hotel, school, neighborhood office) with family members absent. The questionnaire was in a laptop computer, with a same-sex interviewer controlling the computer initially followed by the participant controlling the computer during the last, more sensitive part of the interview. Among the participants 18–49 years of age used below, all but 3% of urban and 9% of rural participants entered data in the computer with little or no help, which should have produced more forthright responses about sensitive behavior (Whisman & Snyder, 2007). The study procedures were approved by a Renmin University human subjects committee and, prior to beginning the interview, participants gave informed consent.

In the initial sample of 10,203 individuals, 26% could not be found at the address where they were registered—leaving 7,553 individuals who had the potential to be interviewed. Excluding refusals, temporary absences, computer glitches, and faked interviews, 5,404 interviews were completed and available for analysis.1 Or, compared to the 7,553 individuals who could potentially be interviewed, the response rate was 71.5%.

After completion, results from the survey were compared with the age, gender, and urban/rural residence statistics in census results to produce post-stratification population weights. Compared to the census, the final data set required particularly large weights for two groups: (1) The young ages 18–29 required larger weights in both urban and rural areas and particularly for those aged 18–24 in the countryside. This was consistent with the young being difficult to locate simply because they were more migrant—both from village to city and within cities (Landry & Shen, 2005). (2) Older people age 50 and above required larger weights, particularly women. This was consistent with older people refusing to be interviewed. Because of the over-sampling of urban areas and lower SEs, the later analysis of precursors used the urban portion of the sample. Using Stata 11.2, the percentages and odds-ratios were adjusted by post-stratification weights, with SEs adjusted by Huber-White corrections for sample clusters (Skinner, Holt, & Smith, 1989).

Procedure

Five analytical choices were of note: (1) To compare the magnitude of simple descriptive differences (in percentages/means) between men and women, we presented Cohen’s (1977, 1988) d statistic. This standardized mean difference score was calculated as the mean score for men minus the mean score for women divided by the pooled within-gender SD. (2) To evaluate whether men and women (or men in commercial and non-commercial sex) responded differently to a potential correlate (e.g., income), we used a Wald test for the significance of differences in odds ratios (ORs) across equations (Allison, 1999). (3) To evaluate whether candidate correlates improved multivariate models, we compared Akaike Information Criterion (AIC) for goodness-of-fit across models (Burnham & Anderson, 2002). (4) To deal with possible endogeneity problems (e.g., feedback effects from infidelity to sexual dissatisfaction), we used instrumental variable probit equations (Ameniya, 1978). In these equations, what we used was not the observed values of “weak love,” “sexual dissatisfaction” or “relaxed surveillance,” but instead the predicted values of each, based on exogenous variables that produced independent variables “purged” of possible feedback effects from infidelity. (5) Most of the analysis that follows was limited to participants age 49 years and below, both because most infidelity occurs before this age and because many comparable studies from other countries use this upper age limit.

Measures

The precursors for infidelity were grouped by sexual history, opportunity, and relationship quality. Sexual History was the sum of the following three dichotomous precursors, each of which was coded as 0 or 1 before summing. The resulting summary scale ranged from 0 (no risks) to 3 (3 risks): (1) “Early sexual contract” was sexual contact before age 14. (2) “Premarital sex” required special coding for the never-married (single, cohabiting). Because all the never-married were sexually active (with a steady sexual partner over the last 12 months), we coded them as having premarital sex. (3) “Masturbation” was based on reported frequency of masturbation during the past year, with values dichotomized at monthly or more often versus less than monthly (or none).

Opportunity (surveillance relaxed) included two items. The first was a summary scale for three dichotomous items: (1) “Travel” out of town without one’s spouse/steady partner for a week or more last year. (2) “Lived apart from spouse/steady partner” for part or all of the last year. The positive values for this item included the 2% of urban participants who were dating but not living together. It included the 11% of married people who were separated from their spouse by out-of-town work for most or part of the year. (3) “Social life” was for spouse/partner absent activities outside of regular work hours, including entertaining, dining parties, calling on friends, and group travel, with activities more than monthly during the year counted as one and less frequent activities counted as zero.

Migrant was the second way in which the relaxation of standard social controls (by family and community) was indexed. Rural-to-urban migrants were counted as migrant if they remained without an official urban household status. Three-fifths of these migrants had been in the city less than 5 years, close to three-fourths were below age 40; four-fifths were in sales/service, self-employment, and manual occupations; one-fourth were either never married or separated from their spouse much or most of the previous year.

Opportunity (resources/work) was the sum of two dichotomous items: (1) “Income” was dichotomized at 1,000 yuan (about US$ 125) per month. This is not as modest as it might seem, because most of the urban participants in this sample lived not in major metropolitan areas but instead in much smaller towns and small cities where average incomes were lower. Manager/owner occupation (of enterprise with ≥8 employees) was included because preliminary analysis showed this to be the group most prone to both non-commercial and commercial infidelity. For descriptive purposes, we will also show the prevalence of infidelity among other occupational groups, including the currently non-employed.

Relationship quality was indexed by two items, both of which remained separate in the analysis: “Love weak” or, literally, “How intense (deep) are your feelings for your spouse?” was on a 4-point scale: “very deep,” “somewhat deep,” “not so deep,” and “feelings already broken apart.” In initial descriptive results, the love scale was dichotomized between the first two and last two responses. “Sexual dissatisfaction was an index produced by summing five questions about responses to sex with one’s spouse/steady partner, including frequency of orgasm, physical satisfaction, emotional satisfaction, thrill, and shame (α = .60). For the minority (2.5%) who had no sex with the steady partner in the last year, this index was assigned the median value. To make the index more comparable to other items in the analysis, the index was divided by a constant to produce a continuous scale from 0 to 2. In initial descriptive results, the scale was dichotomized at the unweighted median.

Results

Descriptive Patterns

Our results are in two parts: first, for descriptive patterns and, then, for precursors of infidelity. This first part on descriptive patterns examined three questions: people’s attitudes towards infidelity, the level of infidelity in China compared with other countries, and the nature of infidelity partners (including mixing by age and education and whether partnering was with strangers or acquaintances).

Chinese Attitudes

The first question concerned people’s attitudes towards infidelity? For the population 49 years old and younger, three survey items suggested that infidelity met with disapproval among most people: (1) In the year 2000 version of the current survey, urban and rural participants 49 years old and below were asked whether it was acceptable for people to have an outside partner after they were married (n = 3,111). The response “completely unacceptable” was 86% among women and 60% among men (p < .05). The more general response combining “completely” and “somewhat” unacceptable was 95%, with only four percentage points separating men and women. Married and rural responses were statistically no different from others in the sample. (2) In the 2006 survey, views about commercial sex were also conservative. Specifically, a survey item noted that the current government policy was that, if arrested as a client of a sex worker, a man could be fined up to 5,000 yuan (several months’ salary for the average urban worker), detained up to 15 days, and his family informed of his behavior. When asked their opinion of this policy, 77% of women and 66% of men reported that this level of punishment was either “just right” or “too lenient.” (3) When men reported on their wife/steady partner’s jealousy, that jealousy increased more for commercial than for non-commercial infidelity. Specifically, in an ordinal logistic equation that included his age, early sexual history, travel, autonomous social life, and income, men reported increased spouse/steady partner jealousy (on a 4-point scale) if during the last 12 months he had either a non-commercial partner (OR = 1.63, p < .05) or commercial partner (OR = 3.68, p < .01). The two ORs were statistically different from one another at p < .05. Adding additional controls for age, education, and prior sexual history left this pattern unchanged.

Comparative Statistics

Our second question was whether Chinese infidelity levels were high compared to other societies. Table 1 shows the Chinese prevalence by type. Extramarital prevalence was for 89.4% of all participants who were in a marital relationship. The infidelity prevalence included people in cohabiting and dating relationships of at least a year’s duration. For non-commercial infidelity, urban exceeded rural by only about one percentage point for both women and men. For men’s commercial sex, the urban–rural difference rose to a full six percentage points and this, in turn, was reflected in the urban–rural gap for any infidelity.
Table 1

Sexual infidelity during the last 12 months by location and type (%)

 

Women

Men (by type of infidelity)

Anya

Non-commercial

Commer-cialb

Extramarital sex

 Total

4.2

13.6

10.4

5.3

  95% CI

(3.1–5.8)

(11.2–16.4)

(8.3–13.0)

(4.2–6.8)

Infidelity (in any sexual relationship)c

 Total

4.5

14.3

11.0

5.5

  95% CI

(3.3–6.2)

(12.0–17.0)

(8.9–13.5)

(4.4–6.9)

 Urban

5.2

17.5

11.6

9.1

  95% CI

(4.0–6.7)

(15.3–20.0)

(10.0–13.8)

(7.4–11.1)

 Rural

4.0

11.8

10.5

2.7

  95% CI

(2.2–7.2)

(8.2–16.6)

(7.2–15.1)

(1.5–4.9)

Observations

 Extramarital sex

1,747

1,551

1,551

1,547

 Infidelity

  Total

1,846

1,721

1,721

1,716

  Urban

1,398

1,312

1,312

1,310

  Rural

448

409

409

406

Note. For Chinese population aged 18–49, with relationship that lasted ≥12 months. Source (for all tables): 2006 Sexuality Survey of China

aBecause some men reported both types, the non-commercial and commercial percentages sum to more than the “any” percentage

bFor sex with female sex worker (xiaojie), excluding sex in exchange for gifts and favors—which for all relationships would have added 2.0% to the “commercial sex” total and 1.2% to the “any infidelity” total

cFor both men and women, the infidelity denominator included 5.3% cohabiting and 5.3% dating for a year or more

For comparative purposes, we focused on the prevalence of extramarital sex. The most common age range for studies done in other societies was 15–49. None of the Chinese participants below age 20 were married. This makes it possible to compare the narrower Chinese age range of 18–49 with the typical age range of 15–49 (Fig. 1). For men, Fig. 1 shows the prevalence for non-commercial sex, on the assumption that studies in other countries often missed commercial sex. However, adding commercial sex for a total Chinese male prevalence of 13.6% led to a similar conclusion. That conclusion was that Chinese males were close to the median of 13.2% for men in this collection of studies. Many countries from the Caribbean and sub-Saharan Africa were higher, South American countries were at about the same level, and much of the developed West was at the same or lower level. For women, the Chinese prevalence of 4.2% was above the median of 0.8% and, indeed, above all but three other countries.
Fig. 1

Extramarital sex during the last 12 months by gender. Note. Data for total married population (urban and rural) in 36 countries. Typically for population aged 15–49, though European ages were more variable. * indicates country with name omitted for readability. For all countries, the median for women was 0.8% and for men 13.2%. Source: 2006 Sexuality Survey of China (Druckerman 2007, pp. 61–63)

Situating China’s level of commercial sex was more difficult. The easily available statistics were available neither for men who were married nor for men in other types of steady sexual relationships. Instead, the statistics were for all men age 15–49, including the sexually inactive. Adapting that standard, and assuming that all Chinese men below age 18 had no commercial sex in the past year, then, Chinese men’s prevalence of commercial sex in the last 12 months was 4.2% (95% CI 3.5–5.2). This exceeded the median of 2.4% for 55 countries, which had a prevalence ranging from 1.1% at the 25th percentile to 4.6% at the 75th percentile (Caraël, Slaymaker, Lyerla, & Sarkar, 2006; DHS, 2010; Pan, Parish, & Huang, 2011).

One additional pattern of note was how sub-Saharan African and Caribbean prevalence differed between non-commercial and commercial sex. In non-commercial extramarital sex, the number of countries from sub-Saharan Africa and the Caribbean with a prevalence exceeding that of China was 20 out of 22 (91%) (Fig. 1). In contrast, in the same comparison for prevalence of commercial sex, the number of countries from sub-Saharan Africa was only 6 out of 28 (21%). For all other countries outside sub-Saharan Africa and the Caribbean, the percentage exceeding China was in a narrower range of 21% (non-commercial, n = 14) to 30% (commercial, n = 27).

Characteristics of Infidelity Partners

Our third question was whether infidelity involved a mixing by age and education with short-term partners in ways that might accelerate the spread of HIV and other sexually transmitted diseases. In an analysis of non-commercial partners from the last 3 years, most partners were met at work, at school, in the neighborhood or through introductions by friends and family (Table 2). Though slightly more so for men, about a third were met at entertainment venues (e.g., dance halls), over the web or other impersonal settings. Most partners were known for some time before sex. That knowledge was more than a year for 40% of women’s partners and 25% for men’s partners. Only 3% of either men’s or women’s sexual partners involved no prior contact. No more than 10% involved acquaintance of only a day or two. Two-thirds of relationships lasted for less than a year, and less than 10% involved only a single sex episode (“one night stand”).
Table 2

Characteristics of infidelity partners (non-commercial)

Characteristics

Women

Men

da

Relationship casual?

 Where met partner

  introduced by friend, etc.

43%

19%**

−0.56

  at school

6%

9%

0.11

  at work place

20%

34%*

0.30

  entertainment venue, web

31%

38%

0.15

 Time known before 1st sex

  no prior contact

3%

3%**

0.31

  1–2 days

6%

10%

 

  <1 month

23%

36%

 

  <1 year

25%

27%

 

  ≥1 year

43%

25%

 

 Duration

  sex only once

7%

9%

0.07

  <1 month

12%

23%

 

  1 month–1 year

44%

36%

 

  >1 year

36%

32%

 

Respondent–partner similar?

 Education gap, male–female (mean)

0.2

0.1

 

 Age gap, male–female age (median)

2

3

 

  Female partner age ≤ 20

5%

 

  Male ≥ 10 years older than partner

19%

 

  Male ≥ 15 years older than partner

8%

 

Sex with partner

 began when respondent age (median)

26

28

 

 always used condom

39%

33%

−0.13

 not wanted the 1st time

31%

5%*

0.84

 felt some shame

55%

37%*

0.39

 emotionally satisfying

77%

91%*

0.44

Maximum observations

 partners (the unit of analysis)

96

238

 

 participants (who reported partners)

85

160

 

Note. Data for urbanites who had a spouse/steady partner for 3 years or more and who were <52 years-old when interviewed. Partners restricted to those in the last 3 years. Up to two partners per respondent. Of all partners, 30% were continuing (had sex in the last 2 weeks). Results unweighted

p < .05; ** p < .01

aCohen’s d

Most partners were not that different from the participant. Consistent with many partners being met at work, at school or by introduction by friends, most partners had about the same level of education and about the same age as the participant. The 2–3 year age gap, with men being slightly older than female partners, was about the same as found for marital partners in the data set. Among the men’s partners, only 5% were below age 20, and most of those were no younger than 18 or 19. That said, there was a substantial minority of men (18%) who reported that their partner was 10 years younger than themselves or even 15 years younger (8%).

Precursors of Infidelity

Our analysis of precursors of infidelity proceeds by reporting in succession bivariate patterns, multivariate patterns, and then tests for the robustness of the observed patterns.

Bivariate Patterns

Two pathways might have led to men reporting more infidelity than women: (1) First, men might have had more exposure to precursors to infidelity (e.g., premarital sex, masturbation, travel, independent social life, high income) (panel A of Table 3). Cohen’s d provides a measure of the size of the difference in men’s and women exposure, with positive values indicating more male exposure and negative values indicating more female exposure. Conventionally, d values below 0.20 are seen as very small to nil and values around 0.50 are seen as moderate. By that convention, women were more exposed to sexual dissatisfaction, four exposures had little or no relation to gender (early sexual contact, living apart, migrant, weak love), and all the remaining exposures were greater for men (often at the moderate 0.50 level). (2) Second, men might have been more responsive to precursors, particularly to precursors where men were highly represented (panel B of Table 3). Both pathways were represented in Table 3. Three examples illustrate the possible patterns: (a) Men were both more likely to have high income (panel A) and, though high income had no influence on women’s behavior, high income sharply increased men’s infidelity (panel B). (b) Though women’s infidelity was greater when they were either a migrant or partnered with a low income man (panel B), these were two precursors in which they were underrepresented (panel A). (c) Though over-represented among those reporting sexual dissatisfaction (panel A), women’s sexual dissatisfaction failed to translate into more infidelity (panel B).
Table 3

Precursors and prevalence of sexual infidelity during the last 12 months

Precursors

A. Distribution of precursors

B. Infidelity prevalence (%) by precursor

Percentages

Non-commercial sex

Commercial sex

Women

Men

d

Women

Men

Men

Sexual history

 Early sexual contact (<age 14)

  Yes

5.2

5.4

0.01

14.0**

23.8 **

30.0***

  No

   

4.7

10.9

7.9

 Premarital sex (before 1st marriage)

  Yes

20.0

41.3***

0.48

9.8***

16.9***

13.8***

  No

   

4.1

7.9

5.9

 Masturbation (during year)

  ≥monthly

7.8

18.6***

0.32

12.1***

16.6***

15.6***

  Less

   

4.6

10.4

7.6

Opportunity (surveillance relaxed)

 Travel ≥ week in year (w/out regular partner)

  Yes

13.6

33.2***

0.48

8.2+

12.8

14.4***

  No

   

4.7

9.7

6.5

 Lived apart from regular partner part/all of year

      

  Yes

14.2

19.3**

0.14

6.6

19.7***

13.5*

  No

   

5.0

9.7

8.1

 Social life (w/out regular partner)

  More

43.5

56.5***

0.59

6.5

15.2***

12.6***

  ≤monthly

   

4.7

6.9

4.7

 Migrant (rural-to-urban w/out urban registration)

  Yes

12.5

15.3*

0.08

9.5**

12.3

9.7

  No

   

4.6

11.5

9.0

Opportunity (resources/work)

 Income high (1000+ yuan/month)

  Yes

36.8

59.2***

0.46

4.9

13.3*

12.5***

  No

   

5.4

9.1

4.3

 Occupation

      

  Administrator (government)

1.0

2.1***

 

0.0

4.0

3.2

  Professional/technical

5.5

6.6

 

3.7

6.0

15.8**

  Manager/owner (enterprise)

1.0

3.8

 

7.3

29.2**

33.6***

  Clerical

24.2

17.7

 

5.5

10.4

9.2

  Self-employed (peddlers …)

15.5

20.7

 

7.2

11.8

9.1

  Sales/service

19.7

14.0

 

5.8+

13.9

10.3

  Manual (reference)

12.0

21.1

 

1.8

9.3

4.9

  Other

3.6

6.0

 

3.7

15.3

6.3

  Not working (& students)

17.5

7.9

 

5.7

9.3

3.8+

 Low income partner (<300 yuan/month)

  Yes

8.2

15.4***

0.23

13.3***

9.2

7.5

  No

   

4.5

12.0

9.5

Relationship quality

 Love weak

  Yes

9.0

6.9+

−0.08

14.0***

30.6***

17.5**

  No

   

4.3

10.2

8.5

 Sexual dissatisfaction

  High

59.1

32.3***

0.50

5.6

17.2***

10.9+

  Low

   

4.6

8.6

8.0

Other conditions

 Relation

  Dating/cohabiting

5.8

10.8*

0.14

10.0*

15.4+

14.8*

  Remarried

2.6

3.0

0.18

2.6

13.2

8.3

  In 1st marriage (reference)

91.6

86.3

 

5.0

11.1

12.7

Age

      

 18–29

21.4

20.1

−0.03

7.6

15.4*

8.9

 30–39

43.3

44.3

0.02

5.1

11.2

11.7**

 40–49 (reference)

35.2

35.6

 

5.2

10.0

6.0

Note. Urban data for 1398 women and 1310 men aged 18–49 years, with relationship that lasted ≥12 months

+p < .10; * p < .05; ** p < .01; *** p < .001 across columns (panel A) or rows (panel B)

Multivariate Patterns

The multivariate analysis in Table 4 used summary measures from the previous table. Besides examining the significance of ORs within each equation, the ORs in panel A were compared across equations so as to detect significant differences in men’s and women’s precursors to infidelity. When two ORs were statistically different from one another at p < .05, a superscript column number indicated the column that was statistically different. The results for “sum of high income and manager” in panel A provide an example. In the set of three ORs, the male non-commercial OR of 1.57 was larger than the female OR of 0.88 (at p < .05). Similarly, in this same set of results, the men’s commercial OR of 3.57 was statistically greater than the both the women’s OR and the men’s non-commercial OR (at p < .05), resulting in both a superscript 1 and a superscript 2 indicating statistically greater than both.
Table 4

Infidelity by precursors (multivariate equations)

Precursors

Correlate range

A. Ordinary logistic equations (odds ratios/z-statistics)

B. Instrumental variable equations (probit coefficients/z-statistics)

Non-commercial Sex

Commercial sex

Non-commercial sex

Commercial sex

Women

Men

Men

Women

Men

Men

(1)

(2)

(3)

(4)

(5)

(6)

Sexual history

 Sum of early sexual contact, premarital sex, & masturbation

0–3

2.33**

1.72**

2.21**

0.36**

0.14*

0.39**

 

(3.94)

(4.63)

(5.90)

(3.75)

(2.07)

(5.39)

Opportunity (surveillance relaxed):

 Sum of travel, lived apart, & social life

0–3

1.14

1.43*

1.62**

0.08

0.25+

0.39*

 

(0.80)

(2.52)

(3.59)

(1.01)

(1.72)

(2.51)

 Migrant

0–1

2.39**

0.961

0.941

0.43**

−0.02

0.08

 

(2.88)

(−0.20)

(−0.20)

(2.59)

(−0.17)

(0.53)

Opportunity (resources/work)

 Sum of high income & manager

0–2

0.88

1.57*1

3.06**1,2

−0.06

0.20*

0.52**

 

(−0.44)

(2.46)

(4.68)

(−0.48)

(2.06)

(4.96)

 Low income primary partner

0–1

3.40**

0.871

1.071

0.54**

−0.12

0.07

 

(3.12)

(−0.53)

(0.21)

(3.10)

(−0.79)

(0.46)

Relationship (w/primary partner)

 Love weak

0–3

1.74*

1.21

0.911

0.89**

na

−0.03

 

(2.52)

(1.24)

(−0.49)

(3.14)

 

(−0.34)

 Sexual dissatisfaction

0–2

0.89

2.47**1

1.322

na

1.68**

0.13

 

(−0.40)

(4.27)

(1.04)

 

(4.09)

(1.02)

Other conditions

 Relation (ref.—in 1st marriage)

  Dating/cohabiting

0–1

0.51

0.49*

0.97

−0.23

−0.38+

0.00

 

(−1.39)

(−2.12)

(−0.07)

(−0.84)

(−1.91)

(0.01)

  Remarried

0–1

0.48

1.19

1.59

−0.36

0.17

0.19

 

(−0.68)

(0.25)

(1.04)

(−0.78)

(0.56)

(0.58)

 Age (reference—40–49)

  18–29

0–1

1.03

1.34

0.60

0.19

0.48*

−0.31

 

(0.07)

(1.11)

(−1.26)

(0.98)

(2.34)

(−1.58)

  30–39

0–1

0.69

0.92

1.37

−0.09

0.11

0.15

 

(−1.45)

(−0.34)

(1.09)

(−0.60)

(0.82)

(1.14)

Observations

 

1,398

1,310

1,310

1,396

1,309

1,309

Pseudo R2

 

0.10

0.09

0.15

   

Wald test of exogeneity (p)

    

0.02

0.003

0.40

Note. Numbers in bold are large compared to other odds ratios in the same row (panel A). Numbers in underlined italics indicate instrumental variables (panel B). Constants not shown “na” indicates “not applicable.” In comparisons of odds ratios across columns, p < .05 is indicated by a superscript number for the column being compared. Within each column, significance of an individual odds ratio indicated by + p < .10; * p < .05; ** p < .01

Across the equations in Table 4 (panel A), several results emerged. First, one precursor (sex history) was shared, both across gender and across commercial and non-commercial sex (ORs = 2.33, 1.72, 2.21 and no superscript numbers for significant differences across columns). Second, there were significant gender differences in the role of other precursors. For women, the significant precursors were migration, low income partner, and absence of love. These were not important for men. Instead, whether in commercial or non-commercial sex, men were more prone to infidelity when they had more opportunity, either because of relaxed surveillance (travel, living apart, gender-specific social life, ORs = 1.43, 1.62) or more resources and work opportunities (income, managerial work, ORs = 1.57, 3.06). Among men, the role of sexual dissatisfaction differed by type of infidelity. While sexual dissatisfaction increased the chances of having a non-commercial partner (OR = 2.47), sexual dissatisfaction was irrelevant to having a commercial partner.

The results from Eqs. 1 and 2 in the columns of Table 4 provided another opportunity to examine the role of exposure to potential precursors (e.g., high income) and relative responsiveness to those precursors (e.g., the large ORs for high income/managerial position for men). For women and men, we recorded the mean exposure values for each precursor (mean sexual history, mean migrant, etc.). In turn, these exposure values were inserted into Eqs. 1 and 2 to produce a set of four estimates of average non-commercial infidelity, ranging from a low of 3.7% for women in the women’s equation (Eq. 1) to a high of 9.4% for men in the men’s equation (Eq. 2), for a total gap of 5.6 percentage points. Then, there were two intermediate estimates when we substituted the mean exposure values for the opposite gender into each equation. From this, we calculated that at mean exposure levels the effect of exposure (substituting means for the opposite sex) accounted for 29% of the total women-men’s infidelity gap while the higher male responsiveness accounted for 71% of the total gap.

Robustness

Endogeneity

We address five sets of questions about the robustness of our findings. The first was about possible endogeneity (feedback) effects. That is, it could be that lessened love, sexual dissatisfaction, and frequent travel, socializing, and living apart were at least as much a consequence as a cause of infidelity. For weak love, sexual dissatisfaction, and relaxed surveillance, we tried to deal with possible endogeneity by using instrumental variable probit equations (Table 4, panel B).2 We could not directly compare the ORs from the logistic equations in panel A of Table 4 with the probit coefficients panel B. We could, however, compare the z-statistics (and significance levels) in each panel. In these comparisons, weak love and sexual dissatisfaction maintained their significance across panels. The relaxed surveillance sum for travel, living apart, and an active social life were weakened. Nevertheless, for men’s commercial sex, the new results remained marginally significant at p < .10 and the results for men’s commercial sex remained significant at p < .05. Or, in short, the probit results in panel B suggest that we were not seriously misled by ordinary logistic results in panel A.

Several other precursors were not subjected to instrumental variable analysis. For the sum of income and managerial position, we judged that both occurred before the infidelity. Similarly, we judged that “sexual history” was relatively immune to feedback from infidelity. For most participants, premarital sex occurred before finding a new infidelity partner. If there was a feedback from infidelity to masturbation, we assumed that masturbation would have declined, not increased. Our interpretation was that masturbation was a marker of pre-existing heightened interest in sex. As an additional check for possible infidelity feedback effects on masturbation, we recalculated the summary measure for history substituting “masturbation before age 17” in place of “current masturbation.” The results in Table 4, panel A remained much as before, with the OR (and z statistic) for sexual history in each column now being 2.07 (3.46), 1.36 (2.80), and 2.08 (5.11). Though marginally weaker, these results remained large and robust.

Omitted Variables

The second question is whether we overlooked any variable that would have changed the patterns in Table 4. None of the patterns were changed by adding candidate variables one at a time, i.e., the significant ORs in panel A of Table 4 remained significant at p < .05. Moreover, most of the added variables were insignificant on their own (e.g., ever experienced same-sex intercourse, resident in major city, male smoking, Communist Party membership). Some additional items were highly significant on their own (e.g., how frequently one thinks of sex, sociosexuality [permissive sex attitudes]), but these seemed so subject to feedback effects from infidelity that we did not pursue this analysis.

Oral Sex

Oral sex with spouse/steady partner was an omitted variable of some note. Contrary to a literature suggesting that men seek commercial sex in order to enjoy more exotic sexual practices (e.g., McKeganey, 1994; Monto, 2001), men who had commercial sex reported no shortage of sexual practices at home. Oral sex provided an example. Instead of reducing commercial sex, regular receipt of oral sex from one’s regular partner increased reports of commercial sex, measured both by statistical significance and reduction in AIC values (OR = 1.54, p < .05, AIC∆ = 4.14). Of course, increased oral sex with one’s regular partner could be subject to feedback effects from commercial sex. However, an attempt to deal with possible feedback effects through an instrumental variable analysis (as in Table 4, panel B) failed to alter the conclusion that oral sex increased reports of commercial sex, suggesting, perhaps, that oral sex with one’s regular partner provided yet another index of sexual history and/or participant’s sexual drive.

Migrant Women

Why migrant women (and not migrant men) were prone to infidelity (OR = 2.39) remained puzzling (Table 4). Compared to other women, migrant women were only marginally more likely to be single or in cohabiting relationships or to be younger—and, besides, those conditions were already included as controls in the multivariate equations. Could it have been that they were more vulnerable from being overrepresented in sales/service and self-employment as peddlers and the like? Possibly, but when these occupations were added to the logistic equation in Table 4, the role of migrant status persisted (OR = 2.35, z = 2.69). Is it because the migrant women were sex workers? They reported only a modest number of extra partners during the year, which would be inconsistent with most FSWs. Thus, the pattern for migrant women seemed robust, and explanations of the pattern must be sought elsewhere.

Other Possible Issues

We might have influenced the results by several other analysis choices: One might have been to include items in the equations that were highly related to one another. The items most highly related were “weak love” and “sexual dissatisfaction” (r ≈ 0.40). Inserting these items one-at-a-time into Eqs. 1–3 produced results paralleling those in Table 4, panel A, i.e., sexual dissatisfaction remained insignificant for women and both love and sexual dissatisfaction remained non-significant for commercial sex among men. For non-commercial sex among men, the results were more complex. Taken one-at-a-time, both love (OR = 1.53, z = 2.96, p = .003) and sexual dissatisfaction (OR = 1.68, z = 5.14, p < .001) were significant. Thus, it was only when competing for influence in the full equation of Table 4 that sexual dissatisfaction was distinctly more important than weak love (p < .02).

A second analytic choice was our emphasis on the urban part of the data set. To examine the consequences of this choice, we repeated the analysis in Table 4 (panel A), this time joining the rural with the urban data sets. With only one exception, the significant ORs paralleled those in Table 4, panel A. The exception was that in the combined data set while “weak love” was significant in bivariate results it was non-significant in the multivariate equation (OR = 1.23, z = 1.04). Also, while the added precursor of “urban residence” had no effect on non-commercial infidelity it did increase commercial infidelity (OR = 2.34, p = .01).

Third, what would have occurred if we examined only the currently married (extramarital sex) rather than all couples in steady relationships (infidelity)? To find out, we reran Table 4 (panel A) with only the currently married. The results paralleled those in Table 4, that is, while specific ORs varied slightly, significance levels remained much as before. Finally, what would happen if we combined men’s commercial and non-commercial infidelity into a single “infidelity” measure? Doing so produced an equation with significance levels quite similar to those for men’s non-commercial infidelity, that is, compared to the commercial equation “sexual dissatisfaction” was a significant influence on infidelity of any type.

Discussion

Chinese Attitudes

Our data on descriptive patterns examined three questions. The first descriptive question was whether Chinese popular disapproval of extramarital sex had softened. The answer was no. In our national probability sample of people 49 years of age and below, 95% of Chinese men and women disapproved of extra-marital sex. When compared to the combined “always/almost always wrong” responses of other countries, China’s 95% disapproval was close to disapproval levels in the U.S. (94%) and much sterner than in Russia (62%) or in most other formerly socialist states of Europe (Widme et al., 1998; cf. Zheng et al., 2011). Thus, unlike formerly socialist European societies, Chinese attitudes towards extramarital sex remained highly disapproving.

Moreover, that disapproval included men’s visits to FSW. Most women wanted harsh punishments of men who visited FSW. When men reported on jealousy levels from their wife/steady partner, those levels were even greater when he reported a FSW rather than a non-commercial sex partner in the last year. The negative reactions were inconsistent with the frequent supposition that women are less concerned about their partner’s short-term sexual liaisons that lack emotional involvement (e.g., Buss, 1994, 2000). The reactions were consistent, however, with ethnographic reports on how Chinese clients of FSW are anxious about their spouse possibly discovering their FSW visits (Zheng, 2009) and with women’s political support of anti-FSW legislation in places such as Taiwan with a similar set of cultural practices (Huang et al., 2004). Nevertheless, one might conjecture that these observations are but “examples that prove the rule,” both that jealous reactions are deeply ingrained and that women are particularly fearful of resources flowing to FSW.

Comparative Statistics

Our second descriptive question was whether Chinese levels of extramarital sex had begun to rival or exceed levels in other societies. The answer was yes, with levels for both non-commercial and commercial sex meeting or exceeding the median levels for other societies. The substantial levels for China were consistent with other reports, including ethnographic accounts about increasing levels of extramarital sex (Farrer & Sun, 2003), reports of spousal jealousy and violence being fueled by suspicions about partner behavior (Wang, Parish, Laumann, & Luo, 2009), and more private investigators being hired to report on spousal behavior (Jeffreys, 2010). In a world of one-child families and potentially bitter child custody fights, these issues can only be intensified (Alford & Shen, 2004; Fong, 2004).

Mixing and Public Health

Our third descriptive question was whether extra partnering outside stable relationships involved a mixing by age and education with short-term partners in ways that threatened to accelerate the spread of HIV and other sexually transmitted diseases. Sub-Saharan Africa (SSAf) is often mentioned as a place with frequent sexual mixing by age and other characteristics, with older men gaining sex with young women below age 20 through the exchange of small gifts and favors (Caldwell, 2006; Gregson, 2002; Kelly et al., 2003; Luke, 2003, 2005; Wamoyi, Fenwick, Urassa, Zaba, & Stones, 2011). Some argue that the origins of Sub-Saharan African patterns in contrast to much of Asia stem from traditional patterns of land ownership that led to the tight control of women’s sexuality in Asia and less so in Africa. The modern consequence, the argument goes, has been a greater availability of young women for casual sex in sub-Saharan Africa and less availability in China and some other parts of Asia (Caldwell, 2006, Goody, 1976). With little availability of ordinary young women for casual sex, Asian men turned to commercial sex with dedicated sex workers.

This literature is consistent with the sharp distinction between how Chinese non-commercial and commercial sexual behavior compare with the rest of the world. In comparison to SSAf and the Caribbean (where some of the SSAf practices were repeated), prevalence of non-commercial sex in China was both less common (Fig. 1) and less likely to involve short-term partners who were much younger, below age 20, or of different background than the participant (Table 2). For commercial sex in contrast, Chinese prevalence matched or exceeded the levels of both SSAf and much of the rest of the world, typically involving an older, high income man with a younger, often rural-origin woman (Zheng, 2009). Thus, it was only with the smaller pool of FSW that many of the SSAf patterns were repeated.

Precursors and Gender Differences

The final part of our study examined whether the conditions promoting infidelity differed greatly for men and women. The answer depended both on differential exposure to precursors by gender and differential responsiveness to precursors. For example, men were likely both to have high income (exposure) and then to use that income for commercial sex (responsiveness). The overall pattern was that men were more likely both to have more exposure (see the larger positive d values in Table 3) and to be responsive when exposed (see the larger odds-ratios in Table 4, panel A). This pattern was summarized by observing that in predicted values produced by equations for non-commercial sex, the total women-men’s infidelity gap was 29% attributable to differential exposure and 71% attributable to differential responsiveness. This strong role of “responsiveness,” in turn, invites reflection on whether future generations of women, freed even more from old societal constraints and in much greater demand because of increasingly male-skewed sex ratios, might become much more responsive (Goodkind, 2011; Petersen & Hyde, 2010) or whether more innate sources will in many ways continue to restrain their behavior (e.g., Lippa, 2009).

The overall pattern above hid four more fine-grained patterns: First, in multivariate results, non-commercial infidelity was driven for women more by reported absence of love whereas for men the influence of love was superseded by reported sexual dissatisfaction with his wife/steady partner. This male–female pattern was consistent with some literature (Barash & Lipton, 2001; Blow & Hartnett, 2005; Del Giudice, 2009; Glass & Wright, 1977, 1985, 1992; Symons, 1979; Thompson, 1984; Weitzer, 2009).

Second, in multivariate results, men’s commercial infidelity was unrelated to the quality of love or sex with his primary partner. Instead, men’s commercial sex seemed driven by a simple desire for a greater quantity of sexual partners. This conclusion was consistent with the observation that the absence of specific sexual acts with his primary partner (e.g. oral sex) failed to increase commercial infidelity. The Chinese pattern was similar to Western patterns for men seeking indoor (e.g., escort, call girl) sex workers (e.g., Bernstein, 2007; Sanders, 2008a, b), though not for men seeking outdoor (e.g., street walker) sex workers (Lever & Dolnick, 2000; Levitt & Dubner, 2009; McKeganey, 1994; Monto, 2000, 2001; Weitzer, 2009). The origin of this indoor/outdoor disjunction begs explanation, though we can note that through the 2000s the majority of sex work in Chinese cities was “indoors” (Huang et al., 2004). The tendency for men in China to seek a greater variety of partners irrespective of sexual satisfaction at home will be seized on by different schools of thoughts–some emphasizing evolutionary origins (Barash & Lipton, 2001; Del Giudice, 2009; Schmidt, 2003; Symons, 1979) and others noting older cultural traditions of male bonding through commercial sex in several parts of East and Southeast Asia (Allison, 1994; FHI, 2006; Phinney, 2008; Uretsky, 2008; Zhang, 2001; Zheng, 2006, 2009).

Third, multivariate results were at least partially consistent with a literature suggesting that in extra-relational sex women attempt to “trade up” in social and economic status (Atkins et al., 2001; Buss, 1994; Del Giudice, 2009; Hrdy, 1999; Pillsworth & Haselton, 2006; Schmidt, 2003; Small, 1993). Chinese women with a low income husband/steady partner were more likely to report extra partners during the year. Conversely, high income men were more likely to report non-commercial infidelity, assisted perhaps by their ability to dip further down in the status hierarchy of women seeking higher status men.

Fourth, the responsiveness to sexual history was quite similar across gender (Table 4). Though fewer women reported a sexual history that included early sex, premarital sex, and masturbation, among the minority of women with this type of history infidelity was just as likely as among men. This pattern invites more exploration of biological and other sources of early differentiation in sexual tendencies (Del Giudice, 2009; Lippa, 2009; Udry, 2000).

Limitations of our study include the possibility of differential recall and candidness–by country, gender, income, and other characteristics (Brewer, Roberts, Muth, & Potterat, 2008). For example, in China, responses about infidelity were typically entered confidentially into a computer. In other countries, though the participant was interviewed in the absence of third parties, responses had to be given orally to the interviewer (e.g., ORC, 2006). This methodological difference might have diminished infidelity reports elsewhere as compared with China (Treas & Giesen, 2000; Whisman & Snyder, 2007). Sampling remained a challenge in China, particularly for a young, highly mobile population that might differ in sexual behavior (Landry & Shen, 2005).

In sum, China has entered a new age in which continuing public disapproval of infidelity is combined with levels of actual infidelity that match levels in much of the rest of the world. Though non-commercial infidelity appears not to involve rapid turnover of partners with age and status mixing that would speed the spread of sexually transmitted diseases, men’s use of commercial sex does. Differences in patterns of infidelity by gender are in part consistent and in part inconsistent with the existing infidelity literature, thereby inviting further study.

Footnotes
1

Based on comparisons to elapsed seconds per interview item in a year 2000 version of this survey, interviews were judged as “fake” when seconds/item dropped below 7 and these seconds were more than two SDs below the predicted elapsed times in a median regression using age, education, urban residence, and regular web use. This criterion eliminated 10% of completed interviews (with eliminated interviews concentrated in September when student interviewers were rushing to return to school). An additional set of analyses eliminated all observations by the five most problematic interviewers, even when elapsed seconds for a specific observation were in a non-problematic range. This set of analyses produced results that were similar to those reported in this article.

 
2

By variable, the exogenous predictors were as follows: Love: partner had extra sexual partner during relationship, was ever hit by the participant, and education; participant’s health; Sexual dissatisfaction: partner had extra sexual partner during relationship, education, height, and travel; participant’s history of drunkenness and health; Opportunity (surveillance relaxed): partner’s frequent social activities and travel; participant’s education, history of drunkenness, and health. For each of the exogenous equations, the F value ranged from 10.1 to 36.3.

 

Acknowledgments

For suggestions on earlier drafts, we thank Ye Luo, William Jankowiak, Edward O. Laumann, and Martin K. Whyte. The University of California Berkeley Demography Center and its Director, Michael Hout, provided a supportive home for the analysis. Financial support for data collection and analysis included Important National Science & Technology Specific Projects of Chinese Government (2008ZX10102), The Ford Foundation, Beijing (1065-0331, 1070-0226), and NICHD (HD056670-01, Gail Henderson, PI).

Copyright information

© Springer Science+Business Media, LLC 2012

Authors and Affiliations

  • Na Zhang
    • 1
  • William L. Parish
    • 2
    • 4
  • Yingying Huang
    • 1
    • 3
  • Suiming Pan
    • 3
  1. 1.Sociology DepartmentRenmin University of ChinaBeijingChina
  2. 2.Sociology DepartmentUniversity of ChicagoChicagoUSA
  3. 3.Institute of Sexuality & GenderRenmin University of ChinaBeijingChina
  4. 4.BerkeleyUSA

Personalised recommendations