Gender Differences in Gender-Role Attitudes: A

Gender Differences in Gender-Role Attitudes: A Comparative Analysis of Taiwan and Coastal China

Su-bao Tu* Pei-shan Liao"

Previous studies have demonstrated that women are more likely than men to support egalitarian gender-role attitudes (Brewster and Padavic 2000; Katsurada and Sugihara 1999; Mason and Lu 1988; Rice and Coates 1995; Wilson and Smith 1995). While considerable attention has been paid to country differences in gender-role attitudes associated with women's—especially manied women's—employment at the individual and societal levels, the industrialization hypothesis used in the previous studies has, however, yielded inconsistent results (Haller and Hoellinger 1994; Alwin, Braun, and Scott 1992; Baxter and Kane 1995; Crompton and Harris 1997; Hsieh and Burgess 1994; Panayotova and Brayfield 1997; Scott and Duncombe 1992). Several issues have remained unexplored. The first is a theoretical explanation for attitudinal gender differences varying by society when considering non-economic social forces, such as social-political ideology, national policy, and other institutional forces. The second is tbe extent to which factors corresponding to gender socialization at the individual, familial, and societal levels structure female and male attitudes differently. The third is national comparison focusing on non-Western populations; previously, attention has focused on Western societies. Furthermore, in order to measure a single concept or attitude, unidimensional scale development and factor analysis have been used to construct gender-role attitudes by assigning interval scores to ordinal response categories. Such research strategies tend to overlook the intertwining of traditional and modem attitudes (Scott and Duncombe 1992), as well as heterogeneity in patterns of responses to attitudinal items due to uncertainty about whether the assigned score is attached to each response category. This practice has compromised some studies using low reliable attitudinal scales (e.g., Cronbach's Alpha less than 0.6) (Mason and Lu 1988). On the other hand. Latent Class Analysis, which identifies distinct patterns within a set of attitudinal items, can avoid that methodological dilerrmia. In order to fill the gap in the literature, this article aims to examine gender difference in gender-role attitudes by comparing two non-Western societies with the same cultural origin: China and Taiwan. More explicitly, we attempt to answer four questions: Is there gender

* address all correspondence to: Su-hao Tu, Center for Survey Research, Research Center for Humanities and Social Sciences, Academia Sinica, 128 Sec. 2 Academia Road, Nangang, Taipei 11529, Taiwan; e-mail: [email protected]. ** Center for Survey Research, Research Center for Humanities and Social Sciences Academia Sinica, Taiwan

620

Joumal of Comparative Family Studies

difference in the diversity in gender-role attitudes, using Latent Class Analysis? Does the gender difference in gender-role attitudes differ by society? Do the determinants of the derived attitude types differ by gender? To what extent does the gender difference in the determinants differ by society? GENDERED SELF-INTEREST: self-actualization vs. economic necessity From the perspective of exchange theory, economic and human resources are important forces related to one's self-interest, and they are associated with a person's gender-role attitudes (Panayotova and Brayfield 1997). Self-interest, motivated by a drive for selfactualization and/or economic necessity, may differ by gender. Women's views of self-interest often relate to lessening the burden of their full-time jobs (family work and paid work) and increasing their own human and economic resources. Men's views of self-interest, in contrast, are likely to support family economic needs only if women's traditional duties (domestic work and childbearing responsibilities) and men's role as primary breadwinner are not disrupted (Baxter and Kane 1995; Scott and Duncombe 1992). Education, employment status, and income are important factors likely to lead women, but not necessarily men, to endorse egalitarian attitudes. While education and employment status were consistently found to contribute to egalitarian attitudes, such positive effects were more significant for women than for men. Gender difference in the effect of employment was especially significant in societies where employment is the overwhelming norm for men (Baxter and Kane 1995; Panayotova and Brayfield 1997; Yi and Kao 1986). The impact of personal income on gender egalitarianism, however, is unclear, with studies providing mixed results, finding it to be positive (Dugger 1991; Wilkie 1993), negative (Cassidy and Warren 1996), and insignificant and indifferent when comparing men's and women's attitudes in Western societies (Baxter and Kane 1995). The inconsistent findings on how income affects gender attitudes leave room for debate over the gendered self-interest hypothesis. PARENTAL ROLE-MODEL Through parental education and employment status, parents serve as role models for children as they develop their beliefs about gender roles (Moen, Erickson, and Dempster-McClain 1997). Those with better-educated parents and higher socioeconomic status (e.g., through the father's occupation) tend to express more egalitarian views and pass them on to children (Kulik 2002). The effect of matemal education on gender-role attitudes was found to be positive, and stronger for men than for women (Powell and Steelman 1982), as well as to interact with daughters' education (Moen, Erickson, and Dempster-McClain 1997; Tallichet and Willits 1986). While scant attention has been paid to gender differences in the effect of parental role models, the notion of a gender schema guiding an individual's perception of gender roles suggests that the effect of the patemal role may be more significant for men, while that of the matemal role may be more significant for women (Smith and Beaujot 1999). FAMILY DEPENDENCE Family is viewed as an essential place for the formation of an individual's gender-role identity.

Gender Differences in Gender-Role Attitudes

621

Maniage, childbearing, and large family size strengthen an individual's ties to faniily, especially increasing women's dependence on men, and preventing women, more than men, from developing egalitarian attitudes. Such gender differences in the effect of family dependence are greater in countries where gender inequality is more prevalent (Baxter and Kane 1995; Kulik 2002) and have been found to be associated with marriage and childbearing experience, both of which significantly decrease women's (including married women's) egalitarian attitudes (Alwin, Braun and Scott 1992; Baxter and Kane 1995). According to a study of married respondents in Taiwan, however, in nuclear families, men instead of women tend to have egalitarian attitudes (Huang 1998). The divergent findings on tbe effect of family size deviate from the family-dependence hypothesis and require further clarification. GENDER SOCIALIZATION AT THE SOCIETAL LEVEL: China vs. IWwan As suggested earlier, individual- and familial-level factors contribute to different gender-role attitudes between societies. National culture (including historical experience), social and political ideology, and institutional policies mediate the linkage between social structures and attitudes toward gender roles (Treas and Widmer 2000). Egalitarian attitudes are more likely to be found in capitalist societies where women's employment and educational opportunities, defamilialization, individualism, and social welfare increase as a society becomes industrialized (Alwin, Braun, and Scott 1992). The cbange in gender ideology tends to be more accelerated in socialist societies tban in capitalist societies, in which the evolution of social norms usually changes slowly and takes longer (Wu and Baer 1996). The emergence of capitalism would decrease gender egalitarianism in socialist societies in which gender-role attitudes would more closely resemble those in capitalist societies (Tang and Parish 2000). While the arguments have been challenged by the divergent findings that socialist economic reform does not necessarily reinforce egalitarian attitudes — as seen, for example, in Hungary, the Czech Republic, and China, as opposed to tbe United States, the European countries of Britain and Norway, and Taiwan — the gender gap whereby women have more egalitarian attitudes than men has been consistently found to be smaller in socialist or post-socialist countries than in capitalist countries (Braun, Scott, and Alwin 1994; Crompton and Harris 1997; Hsieb and Burgess 1994; Panayotova and Brayfield 1997). Such findings caution us about the appropriateness of applying the hypotheses proposed to societal differences in gender-role attitudes without emphasizing tbe gender gap in attitudes. Under traditional models, the gender gap in Taiwan would be expected to be larger than that in China, given the two Chinese societies' different political and socio-economic evolution since the 1950s. The gender difference in the determinants of gender-role attitudes inspired by three theoretical perspectives — gendered self-interest, patemal role model, and family dependence — may be differentiated by society based on the following observations. The employmentrate for women in China skyrocketed in the years following the establishment of socialist govemment, especially in large urban areas, but it bas been declining slowly since the post-Mao economic reform of the late 1970s (Thobum and Howell 1995). In comparison, women's employment rate has gradually increased in Taiwan since the economic development of the 1960s, while unmarried women are more active in the labor market than married women (Lee and Sun 1995; Yi and Chien 2002). Women had a higher employment rate in China than

622


in Taiwan' up to 1990; moreover, gender inequality in income was greater in Taiwan than in China^ (Tang and Parish 2000). Women's access to education bas steadily increased in Taiwan. In China, women's access to education was officially encouraged during the Cultural Revolution, but it has been commonly ignored by parents during the post-Mao regime, especially in mral areas (Rai 1995; Tang and Parish 2000). Taiwan has had a steady increase in the proportion of nuclear families-' since the initiation of family-planning programs on the island in 1965 (Lee and Sun 1995). In contrast, since the late 1970s, a strictly implemented one-child policy in China has resulted in a high percentage of nuclear families there," with a higher proportion in urban than in rural areas (Rai 1995; Tang and Parish 2000; Yi 2002; Zhang 2001). Furthermore, socialist ideology has liberated women by mobilizing tbeir labor-force participation as routine social practice in China for several decades. In contrast to top-down political forces in China, feminist ideas in Taiwan have gradually increased with tbe emergence of democracy and individualism, along with increasing contacts witb tbe Westem world since the 1970s (Hsieb and Burgess 1994). In contrast with Taiwan, the economic, educational, family, and socialist policies in China were more thoroughly implemented in urban than mral areas, and changed radically during different periods. Gender-role attitudes in the two social milieus can thus be differentiated significantly by urban and cohort experiences to a greater extent in China than in Taiwan. China's smaller gender gap in employment opportunity and income, its larger gap in access to education, greater disparities in mral-urban policy implementation, and greater difference of cohort experiences when compared witb Taiwan imply that personal education, birth cohort, and urban experience appear tofiguremore importantly among the factors stmcturing gender-role attitudes in China than in Taiwan. It also reflects that employment, income, parental role-models, and family-related variables appear more significant in Taiwan than in China. The argument by Tang and Parish (2000) that the emergence of a capitalist economy and the abmpt easing of one-child regulations would revive traditional faniily arrangements and undermine the maintenance of socialist legacies in China, however, leaves room for debate over tbe difference in tbe explanation for gender-role attitudes between Taiwan and China.

' The employment rate for women was 91 percent in Shanghai, according to 1990 census data (Abbott et al. 1995), and 44 percent in Taiwan, according to manpower statistics in 1990 (Directorate-General of Budget, Accounting and Statistics 2002). ^ The coefficients translate into females making 89 percent as much as males in China. In Taiwan, the coefficient translated into females making only 67 percent as much as men (Tang and Parish 2000:286). ' The average number of persons per nuclear family, defmed as the family including parents and children only, was 4.7, according to Weinstein et al. 1990 (quoted in Lee and Sun 1995), while the figure for nuclear families was 63.6 percent up to 1990, according to the 1990 population and housing census (DirectorateGeneral of Budget, Accounting and Statistics, Executive Yuan, 2002). " The percentage of nuclear families was 71.4 percent in China in 1995, with around a 4 percent difference between rural and urban areas. The average number of persons per family was 3.23 in urban areas and 4.48 in rural areas in 1995, according to the National Bureau of Statistics of China, http://www.stats.gov.cn/ndsj/ information/iml.html".


623

DATAANDMETHOD

Two sets of survey data, using multistage probability sampling, were collected from Taiwan in 1996 and coastal areas of China in 1997.'A total of 2,801 and 2,907 completed face-to-face interviews from Taiwan and coastal China, respectively, were included in the final analysis for the purposes of this particular study. As the dependent variable, the respondents were asked whether they agreed with four statements reflecting the ideal of women helping the family and supporting husbands' careers, as well as the segregation and combination of gender roles suggested by Mason and Lu (1988) and Scott and Duncombe (1992). These statements were: (1) husbands should work outside the home, and wives should maintain the home; (2) husbands should share housework when their wives have a job; (3) women naturally take much better care of families than men do; and (4) a woman achieves through her husband's career achievement. The responses to those items were strongly disagree, disagree, agree, strongly agree, and no opinion. The response categories were further reduced by combining "strongly agree" with "agree," "strongly disagree" with "disagree," and then by combining "no opinion," which comprise a small proportion, with the majority. The independent variables were conceptualized and measured as follows. Concerning the importance of economic and human resources in the self-interest and parental-role models, respondent's occupation was listed as the current or latest job. Respondent's father's occupation was recorded as the job held wben tbe respondent was age 15. Six categories, based on employment status and level of professional and technological skill, were defined: (1) administrative staff and managers, (2) professionals, (3) technological workers, (4) manual workers, (5) fanners, and (6) the unemployed and otbers (e.g., students and homemakers). Dummy variables were created with the category of manual (unskilled) workers as the reference category. Personal annual income is measured by the total of incomes per year in U.S. dollars, while education (for self, father, and mother) was measured by years of schooling. Among the family-dependence variables, a nuclear family was defined as a family consisting of parents and children living together. A composite of marital and childbearing status was defined by creating four categories to account for the combinations of respondents' marital status and the number of children: (1) never-married single,* (2) married or living with a partner with no or one child, (3) married with two children, and (4) married with more than two ' More explicitly, Taiwanese aged 20-65 were randomly selected using three-stage stratified sampling.Ten stata were defmed based on urbanization, community development, and regional differences of townships and districts, the Primary Sampling Unit (PSU). Proportion to Population Size Sampling (PPS) was used in the first two stages, the selection of PSU's in each Stratum and Secondary Sampling Units (SSU, Li, the administrative unit under township) in each PSU, respectively. Household registered residents in each SSU were systematically sampled. Five-stage stratified sampling was employed in the random selection of Chinese aged 20-65 and residing in 18 coastal cities. Two strata were defined according to the population size in the 18 cities, using 2 millions as the cutting point. PPS was used in the first four stages, the selection of cities in each stratum, districts in each city, blocks or townships in each district, and community committee(Administration district, Li), respectively. Households in each Li were systematically sampled, while one household member was randomly selected. ' The marital status in this study was focused on whether the respondents have married. "Never-married single" doesn't include those who divorced. None of the never-married respondents in the two data sets reported having a child.

624


children. The group of never-married singles was used as the reference category. Birth cohort and urban experience were used as socialization variables implying life and residential experiences in different societies. Using 1949 (when China officially became a communist society) and 1965 (the beginning of the Cultural Revolution) as cutoff points, cohort experience was defmed by three birth cohorts according to those bom (1) in 1949 or before, (2) between 1950 and 1965, and (3) in 1966 or after, while tbe cohort bom in or after 1966 was the reference group. Urban experience was measured by whether a respondent was currently residing in a city or had lived in one before the age of 15. Latent Class Analysis (LCA) was employed to identify major response pattems in genderrole attitudes for both genders in the Taiwanese and the Chinese data,' respectively. The results of the LCA were used as the dependent variable. Logistic regression or multinomial logistic regression, where applied, was used to examine the effects of the independent variables on the types of gender-role attitudes. In addition to tbe main effect of the variables defmed earlier, the interaction effect of self-education with mother education was included in the analytical models. RESULTS Sample Description Sample characteristics are reported in Table 1. For Taiwanese, the most frequently reported occupation among men was manual worker, followed by professional worker, while the most commonly reported occupation among women was technological worker, followed by manual worker. For those from coastal China, farmer was tbe most frequently cited occupation for both women and men, followed by manual worker, wbile about one-third of the women in both data sets were unemployed, homemakers, or others. An income gap between women and men is noticeable in the two societies. The average of personal income for Taiwanese women was less than half of tbat for their male counterparts. In coastal China, the average of women's income was about two-thirds that of men. More than half of the respondents from coastal China had a junior high degree. Tbe Taiwanese respondents who bad a primary-school, senior high school degree, and a college degree particularly for men respectively are all more than a quarter. In both societies, more men than women had a high school (junior or senior high) degree or above, while more women than men had education less than primary school. There is not much country difference in tbe gender gap when concerning the five education levels. Similar proportions of botb genders in Taiwan and coastal China reported low patemal educational attainment, with the latter showing a higher percentage of respondents whose fathers bad junior high school degrees. Mothers' educational attainment was even lower, with more than 90 percent of Taiwanese women's and men's mothers having completed no more than primary school. Comparatively, mothers' educational attainment was a bit higher Hereafter, "the Chinese data" or "the Chinese respondents" stands for the respondents from coastal China. hina

625

Gender Differences in Gender-Role Attitudes Table 1: Sample Charactenstics Coastal China

'Mwan

\^nables

%

% 1420 92 311 198 577 122 120

N 1279 15 187 336

1. (1) (2) (3) (4) (5) (6)

PsrsMialoocifalim Adnrinislralion Profesaaialwater Tedmologicalwater Manualwater Fma UnarployedandotheB

2 3. (1) (2) (3) (4) (5)

Pasraialincane (1,OOOUS$) Pereaialeducatiai Nofamaleducalion Primaiysdioolorless Junia-highsdxwl SoiiOThigfischool Collegpa-beyond

19(17.4)^ 1230 83(10.0) 1435 92 25 17 30.6 371 25.9 282 162 197 26.6 380 265 263 377 17.4

1175 1366 125 418 221 363 239

4. (1) (2) (3) (4) (5)

Falher'seducaliai Nofamaleducaliai Primaiyschoolorless JuniOTbi^sdiool Saviorhi^sdKX)l CoBegporbq^raid

31.9 428 10.0 9.4 5.9

1385 442 592 139 130 82

1257 391 566 105 121 74

5. (1) (2) (3) (4) (5) (6)

Falha-'soccuMlion Administralicn Piofessionalwater Technologicalwater Manualwater Rmer Unmployedandothas

7.0 8.4 14.6 265 423 12

1346 94 113 197 356 570 16

6.5 9.0 15.0 283 39.9 13

6. (1) (2) (3) (4) (5)

Mote-'seducatiai Noftxmaleducalirai PriniaiysdKX)lOTless JuniorhigJisdMol Saiiorhi^sdKX)l Collegea-beyond

567 345 5.4 2.4 1.0

1401 795 484 75 34 13

17.8

1436 256

7 Nudearfemily (l)Yes

6.5 21.9 13.9 40.6 8.6 8.5

12 14.6 263 192 3.8 34.9

Female

Male

Female

Male

245 49 447

%

N

%

N

9.9 7.6 107 277 285 15.6

i4«y 147 113 159 413 425 232

2.8 7.6 147 187 21.9 313

41 110 215 273 363 457

1.4(23) 1403 0.9(0.9) 1338 1477 1445 32 48 10.0 144 13.8 204 18.9 273 59.4 878 502 726 83 122 10.7 155 153 225 102 147 353 35.9 207 2.9 52

1331 470 478 275 39 69

16

10.1 6.3 13.9 262 429 0.6

1323 133 83 184 347 568 8

11.4 7.1 14.8 26.8 392 0.7

1267 145 90 188 339 4% 9

56.0 345 5.5 2.9 1.1

1316 737 454 73 38 14

57.0 26.1 13.1 2.0 1.8

1424 812 372 186 28 26

59.4 24.9 11.6 2.8 13

1356 806 338 157 38 17

137

1366 187

115

1491 1066

652

1461 952

31.1 45.0 8.4 9.6 5.9

1276 83 115 191 362

509

313 39.6 20.6 1.8 6.7

1387 434 549 286 25 93

626


Table 1: (Continued) Sample Characteristics Iknvan Male \yables 8. (1) (2) (3) (4)

Marital-diildbearingstatus Never-mamedsin^ Manied^Mrtna-O-ldnld Mamed/t)ai1ner2d]i]drai Manied4Mrtnei>2ciiikiren

9. Agp 10. (1) (2) (3)

Birthcdiot Bcminl949orbefae Bom 1950-1965 Bominl966OTafta-

11. Lftbanexpedenoe (1) Yes

Coastal China

Female

Male

Female

%

N'

%

N

%

N

%

15.7 15.4 29.8 39.1

1436 225 221 428 562

9.0 128 343 43.9

1366 123 175 468 600

128 553 18.6 133

1491 191 825 277 198

53 555 207 \%5

40.4(9.0) 25.8 54.4 19.8

1461 78 811 302

m

1436 40.1(9.1)

1366 41.9(11.1) 1491 425(11.2) 1461

1436 370 781 285

23.4 55.6 21.0

1366 320 759 287

31.6 45.9 225

1491 472 684 335

33.9 453 20.8

1461 495 662 304

513

1366 701

67.8

1491 1011

73.0

1461 1066

136 48.0

N

1. N=vali(l cases. 2. The number outside of the parentheses is the mean; the one in the parentheses is the standard deviation.

in coastal China, where more than one-tenth of w o m e n and men reported their mothers having a junior high school degree. In addition, for both genders in the t w o societies, around 4 0 percent of the respondents' fathers were farmers, followed by manual workers. A s for family structure, both w o m e n and men in Taiwan shared similar likelihood of either living in a nuclear family or being married/living with a partner. Similar pattems for both genders were found in coastal China, with a much higher percentage of respondents in China than in Taiwan living in a nuclear family. Furthermore, more than half of women and men from coastal China were married or living with a partner with no or one child, while a much higher percentage of Taiwanese of both genders reported being m a n i e d or living with a partner and with m o r e than t w o children. T h e average age for w o m e n and men in both societies was about the same, with a higher proportion of those b o m between 1950 and 1965. M o r e than half of Taiwanese w o m e n in this study reported having lived in urban areas, while fewer than half of Taiwanese m e n did. Compared with the Taiwanese respondents, more than half of w o m e n and m e n from coastal China reported having lived in urban areas, while the percentage was higher for females. Latent Classes of Gender-Role Attitudes As shown in Table 2, the proportion of affirmative responses to tbe four statements ranged roughly from 52 percent to 97 percent for both genders in Taiwan. Agreement with items one, three, and four and disagreement with item two indicated a more traditional orientation. Except foritem one, with which about half of the respondents agreed, most of the Taiwanese


627

respondents, more commonly men, reported traditional gender-role attitudes. When compared with Taiwanese, higher proportions of women and men from coastal China accepted a traditional division of marital roles (item one) but were less traditional toward women's familial and career roles (items three and four). Except for item four about women's career role, gender difference in which women are more modem than men toward items one to three was found to be similar between Taiwan and coastal China. Table 2: Distribution of Responses to Gender-Role Attitudes Male Items

Female

Agree

Disagree

Agree

Disagree

%(N)

%(N)

%(N)

%(N)

1. Husbands should work outside the home. and wives should maintain the home. 2. Husbands should share housework when their wives have a job. 3. Women naturally take much better care of families than men do. 4. A woman achieves through her husband's career achievements.

56.0 (803) 97.1 (1,394) 82.0 (1,177) 91.8 (1317)

44.0 (632) 2.9 (41) 18.0 (258) 8.2 (118)

523 (715) 973 (1,329) 78.4 (1,071) 90.8 (1,240)

47.7 (651) 27 (37) 21.6 (295) 92 (126)

n . Coastal China 1. Husbands should work outside the home. and wives should maintain the home. 2 Husbands should share housework when their wives have a job. 3. Women naturally take much better care of families than men do. 4. A woman achieves through her husband's career achievements.

72.8 (1,061) 96.7 (1,408) 80.5 (1,173) 89.6 (1,306)

27.2 (397) 33 (48) 19.5 (284) 10.4 (151)

69.5 (1,009) 97.5 (1,415) 74.4 (1,080) 90.7 (1,316)

305 (442) 25 (36) 25.6 (371) 93 (135)

L Taiwan

The cross-classification of the four dichotomous indicators of gender-role attitudes resulted in 16 (2"*) response pattems for each Latent Class Analysis (LCA) among women and men in the Taiwanese and Chinese data, using the MLLS A program adapted by Eliason (1988).* Table 3 reports the goodness-of-fit statistics for four latent class models. A model with two latent classes fit both gender groups in the two datasets well, while a restriction was imposed in the Chinese male group. A two-class model was thus used to represent the latent class association among the four items of gender-role attitudes. For each gender group, a variable with two latent classes was used as the dependent variable in the multivariate analysis. * Since originally categories of the response to the items used for the present study were ordinal, zero-order correlations of the items, from strongly agree to strongly disagree, were examined prior to Latent Class Analysis. As shown in the appendix, similar patterns of inter-item correlations, from moderate to low correlation, between women and men were found in each society.

628


Table 3: GkKMlness of Fit of Latent Class Models

Number of latent classes I

L-square

d.f.

P

205.94 1.71

11 7

.000 >.1O

189.07 9.61

11 7

.000 >.1O

280.13 11.41 5.69

11 6 1

.000 .10>p>.05 .1O

Hwwwi A. Male:

1 2 B. Female:

1 2 IL Coastal China A. Male:

1 2* 3 B. Female:

1 2

*: This model is selected because of its small value of index of dissimilarity (.01) and high percent of correctly allocated responses (93.06 percent).

Similarities among male and female pattems of gender-role attitudes were further examined. We compared the V? fit of the model in which equality restrictions were imposed on conditional probabilities across genders with the \} fit of the model where the conditional probabilities werefi-eeto vary with gender. Similar results between Taiwan and coastal China indicated that it is statistically acceptable to impose the same estimated conditional probabilities upon most of the response categories of the four items (results not shown) across genders. By the same token, the similarities of latent class probabilities were examined. The tests rejected the assumption that the latent class probabilities did not vary with gender (results not shown). Finally, the similarity of the estimated probabilities of responses in each class across the two datasets for either gender was examined with equality restrictions imposed. The equality of some items across the two datasets was acceptable, as indicated in italic in Table 4. The estimated conditional probabilities and latent class probabilities are presented in Table 4. The probability of Taiwanese agreeing with all four items for both genders in Class I was high, from .73 to .97. In other words, the Taiwanese respondents in Class I tended to be traditional toward gender roles. The estimated probabilities that Taiwanese men and women classified in Class I, named as "traditional type," were .73 and .66, respectively. In contrast, both women and men in Class II had low probabilities of agreeing with item one but high probabilities of agreeing with items two and four, while about half of them agreed with item three. The respondents classified in Class II, with probabilities of .27 and .34 for men and women, respectively, indicated a mixed type of gender-role attitude. This type was defined as "mixed pro-husband's-career type," because the respondents in this type were almost equally supportive of women's inborn family role (item three) but traditional toward a woman's


629

achieving through her husband's career achievement (item four). At the same time they endorsed gender equality concerning the division of family labor (items one and two). Table 4: Latent Class Probabilities of Gender-Role Attitudes'* Ikhvan Male Latent Class Class probability

Coastal China Female

Male

Female

I

n

I

n

I

n

I

n

.73

.27

.66

.34

.75

.25

.71

.29

Response probability 1. Husbands should work outside the home, and wives should maintain the home. Agree .73 .11 .73 .11 .89 .23 .89 Disagree .27 .89 .27 .89 .11 .77 .11 2. Husbands should share housework when their wives have a job. Agree .97 .98 .97 .98 .97 .95 .99 Disagree .03 .02 .03 .02 .03 .05 .01 3. Women naturally take much better care of famihes than men do. Agree .95 .46 .95 .46 .95 .36 .95 Disagree .05 54 .05 .54 .05 .66 .05 4. A woman achieves through her husband's career achievements. Agree .95 .81 .95 .81 .93 81 .93 Disagree .05 .19 .05 .19 .07 .07 19

.23 .77 .95 .05 .25 .75 .81 .19

*: The equal numbers in italic are results of equality restriction across data sets for either gender.

Latent Class Probabilities indicated that gender difference in gender-role attitudes in coastal China was found similar to that in Taiwan (Table 4). Class I, in which both women and men are classified, could be defined as "traditional type." Differing somewhat from thefindingsin the Taiwanese data, those from coastal China classified in Class II had high probabilities of agreeing with items two and four but low probabilities of agreeing with items one and three. Class n for the Chinese respondents was thus defined as "modem pro-husband's-career type," because those in this type were traditional toward a woman's achieving through her husband's career achievements but modem toward the remaining items. Similar pattems, in which most of women and men in the two Chinese societies across the Taiwan Strait were classified as "traditional type," echo what Parish and Tang (2002) suggested: In socialist societies the emergence of capitalism leads gender-role attitudes to resemble those in capitalist societies. In addition to cultural similarities and the gradual closeness of social systems, the reasons behind the results may come from similar sample characteristics, such as respondents' age, parental education, and father's occupation. Despite the similarity, diversity existed across societies and genders. First, the respondents from coastal China were more likely than the Taiwanese respondents to be in the traditional type. Second, the gender gap was slightly larger in Taiwan than in coastal China, in terms of a higher probability of traditional attitudes among men than among women, and a lower probability of nontraditional attitudes among men than among women. The findings imply

630


the inadequacy of Parish and Tang's (2002) closeness hypothesis to explain gender gap in attitudes between socialist and capitalist societies. In particular, conceming sample characteristics in this study, the gender difference in respondents' occupation, personal income, marital-childbearing experience, and family size may also provide some explanation. Determinants of Gender-Role Attitudes Binary logistic regression was used to identify the factors differentiating the pattems of gender-role attitudes. "Mixed pro-husband's-career type" and "modem pro-husband's-career type" derived from Latent Class Analysis were assigned as the reference categories for the Taiwanese and the Chinese data, respectively. In order to answer research questions three and four, we first examined gender difference in the determinants of attitude pattems using the combination of the Taiwanese and the Chinese data and then compared gender difference in the combined dataset. Conceming the two Chinese societies as a whole, all but one factor significantly associated with the pattems of gender-role attitudes — respondent education, a common factor for both genders — were found different by gender (Table 5). Education consistently infiuenced the likelihood of having non-traditional attitudes for both men and women. Society difference was found significant only among women. When controlling for other factors, Taiwanese women were more likely than coastal Chinese women to hold traditional attitudes.' In the female model, those with lower personal income and younger age were more likely to hold non-traditional attitudes. It was found only in the male model that men with fathers who worked as administrative staff were more likely than men with manual-worker fathers to have traditional attitudes. Those having lived in urban areas tended to hold modem attitudes. The results support the gendered self-interest perspective only in terms of more significant effects of personal income and education in the female model than in the male model. Similarly, the parental role-model perspective was supported only in terms of the more significant effect of father occupation in the male model than in the female model. Model chi-square indicated that the specified model is significantly different fiom the constant model (at .001 significant level), meaning that the factors in the model are important for ' This finding seems to be a bit deviant from that suggested by Latent Class Analysis, the probability for Taiwanese women to be classified in traditional type was lower than that for the Chinese women. The possible explanation may be that the reference category was, in fact, differently defined for the two societies. Strictly speaking, "Mixed pro-husband's-career type" found in the Taiwanese data was less modem that "Modem pro-husband's-career type" found in the Chinese data. It is, therefore, not perfectly adequate to make a precise conclusion based on the minor and diverse fmdings. '" There are several similar measures, usually called Pseudo R^, representing explained variance for limited dependent variables models like logistic regression. Different from R^ in OLS regression, Pseudo R^ is calculated based on the change of model chi-square to indicate the extent to which the inclusion of independent variables in the model reduces the goodness-of-fit. The most original Pseudo R^ (usually named as R ' J is model chi-square divided by (model chi-square + -2 Log likelihood). The derived Pseudo R^ does not usually have a maximum of 1, perfect prediction, because it mostly relies on the percentage of cases in the largest category of the dependent variable. Adjustments were made by different scholars. Among the adjusted Pseudo R^'s, this study used Nagelkerke R^. Given that Pseudo R^ can stand for the percentage of "variation" in the dependent variable explained by the logistic regression model, as suggested by Pampel (20(X)), the pseudo measure could be treated as only a rough guide instead of as a precise measure of goodness-of-fit (Menard 1995; Pampel 2000; Veall and Zimmermann 1994).


631

Table 5: Logistic Regression on Gender-Role Attitudes: Total Sample Male model'

P

Variable

Exp(p)

2.83***

Intercept Country (coastal China)" Taiwan

-.14

Female model

P

Exp(p)

1.09** .87

.36*

1.43

Economic and human resources Current occupation (Manual workers)" Administration Technological workers Farmers Professional workers Unemployed and others Personal income (High)" Low^ Medium Respondent's education

-.07 -.01 .09 -.26 -.26

.93 .99 1.10 .77 .77

-.21 -.14 -.08 -.23 .22

.81 .87 .93 .80 1.25

.33 -.23 -.08**

1.39 .80 .93

.44** .33* -.09***

1.56 1.39 .92

Role-model variables Father's education Father's occupation (Manual workers)" Administration Technological workers Fanners Professional workers Unemployed and others Mother's education

-.01

.99

-.01

.99

.49* .20 .13 -.01 .30 .04

1.64 1.23 1.14 .99 1.35

.01 .08 .20 -.02 -.03 .01

1.01 1.08 1.22 .98 .97 1.01

l.m

Family structural variables Nuclear family (No)' Yes Marital-childbearing status (Never-married singles)" Manied with no or one child Married with two children Married with more than two children

-.05

.96

.19

1.21

-.33

.72 .69 .89

-.09 .15 .15

.91 1.17 1.17

-.37t -.11

Cohort and urban experience Birth cohort (Bom in 1966 or after)" Bom in 1949 or before Bom 1950-1965 Urban experience (No)" Yes

.09 .18

1.10 1.20

.38* .27*

1.47 1.31

-.29*

.75

-.04

.97

632


Table 5: (Continued) Logistic Regression on Gender-Role Attitudes: Total Sample Male model'

P

Variable

Exp(p)

Female model

P

Exp(p)

Interaction effect Mother education*R's education Valid N Model chi-square (d.f.=24) Nagelkerke R-square

-.00

.00

2280 86.79*** .06

2,165 180.12*** .12

1: Reference category of the dependent variable is "mixed pro-husband's-career type" and "modem prohusband's-career type" for the Taiwanese and the Chinese data, respectively. 2: Incomes with different scale in the two societies were classified into three categories using 33.33 percent and 66.67 percent as cutting points. The incomes for the cutting points (1000 U.S.$) were 6.19 and 15.66 for Taiwan and .51 and 1.17 for coastal China. *: Reference category of the dummy variable. t: p