Home Business & Finance
Multivariate analysis of differences in male-female views on gender equality
The WVS6 asked respondents (both male and female) to mark, on a scale of 1 to 10, how essential it was that “in a democracy women had the same rights as men” - hereafter, referred to as “gender equality” (GE) - with 1 representing “GE is not essential” and 10 representing “GE is essential”. For the purposes of the econometric estimation reported in this section, these responses were recoded as “GE is not essential” (original coding: 1-6; new coding: 1), “GE is not necessarily essential” (original coding: 7-9; new coding 2), and “GE is essential” (original coding 10; new coding: 3). Of the 73,281 respondents to this question, 26.7% viewed GE as “not essential”; 29.9% thought it was “not necessarily essential”; and 43.4% regarded GE as “essential”. Not surprisingly, these proportions varied by gender: 40.1% of men, in contrast to 46.4% of women, regarded GE as “essential” while, at the other end of the scale, the proportions of men and women who regarded GE as “not essential” were, respectively, 29.4% and 24.2%.
These are the raw figures. The interesting question is the shape of these male-female differences after non-gender factors (inter alia religion, region, age, social class, income) had been taken into account. For example, Islam, by offering fewer rights to women than it does to men, is predicated on the belief that men and women are not equal. So are male and female Muslims similarly minded in regarding GE as “not essential”, or is there a difference of opinion between Muslim men and women regarding the importance they attach to GE? And do Muslim men and women attach lower importance to GE than their counterparts from other religions?
In order to answer these questions a multinomial model was estimated in which the dependent variable took the value 1 for respondents who viewed GE as “not essential”; 2, for respondents who thought GE was “not necessarily essential”; and 3, for respondents who regarded GE as “essential”. In a multinomial logit model with / mutually exclusive possible outcomes, indexed, / = 1,...,/, for each individual i, indexed i = 1,..., N, the dependent variable Y, is defined as taking the value j for individual i у = j , if that outcome / occurs for individual i. If outcome / is taken as the base outcome, the multinomial logit represents, for each individual (i= 1,, N), the logarithm of the odds ratio of outcome j (j = 1,. . ., J - 1) to the base outcome, / as a linear function of К determining variables (indexed, k = 1, . . . , K), with Xjk representing the value of variable k for individual i:
where /r = Pr(Y) = /), ^рц = 1 and (3jk are the coefficients associated ;=i
with /th outcome for the kth determining variable, with by definition, ,8jk = 0 (k = 1,..., K). The assumption is that these coefficients do not vary across the individuals in the sample.
In order to test the hypothesis that men and women, on average, held differing views on GE, the model allowed all the determining variables to interact with the gender of the person concerned: these interaction effects allowed the estimated coefficients to be different for men and women with appropriate statistical tests allowing the statistical significance of these differences to be assessed.26 The multinomial interaction model for GE was estimated using all the determining variables shown in Tables 6.12 and 6.13 but, since the focus of interest in this section is religion and region, the results in Table 6.14, displayed in terms of the predicted probabilities of the two extreme attitudes towards GE, namely, “GE is essential” and “GE is not essential”, pertain only to these two variables.
The first point of interest about the results in Table 6.14 was that the predicted probability of regarding GE as essential was, for every religion and for every region, significantly lower for men than for women while the predicted probability of regarding GE as not essential was, except for the West where the difference was insignificantly small, significantly higher for men than for women.
The second point of interest was that the predicted probability of regarding GE as essential was lowest, and the predicted probability of regarding
Table 6.14 Inter-gender differences in the predicted probabilities of agreeing that “Men and women have equal rights in a democracy
Source: Own calculations, World Values Survey 6. Note: Estimated on data for 72,471persons.
** Significant at 5% level; * significant at 10% level.
GE as not essential was highest, among Hindu and Muslim men (Table 6.14: 24.8% and 34.1%, for essential and 36.8% and 33.8% for not essential, respectively). This result was mirrored for women. The predicted probability of regarding GE as essential was also lowest, and the predicted probability of regarding GE as not essential was also highest, among Hindu and Muslim women (Table 6.14: 29.1% and 41.7%, for essential and 31.4% and 27.2% for not essential, respectively).
In terms of the regions, the predicted probability of men and women regarding GE as essential was lowest for African countries (Table 6.14: 26.7% and 33.1% for men and women, respectively) and highest for Western countries (Table 6.14: 57.5% and 61.5% for men and women, respectively). Conversely, the predicted probability of men and women regarding GE as not essential was highest for African countries (Table 6.14: 37.2% and 31.7% for men and women, respectively) and lowest for Western countries (Table 6.14: 15.5% and 15.3% for men and women, respectively).
On the basis of these predicted probabilities one can define gender assertion as the ratio of the predicted probabilities of women and men regarding GE as essential where this ratio is represented by a, a > 1: a = 1 means that there is no gender assertion since the predicted probabilities of women and men regarding GE as essential are the same; the greater the value of a the greater the divergence between these two probabilities and, therefore, the greater the degree of gender assertion. So, for example, Table 6.15 shows that the overall degree of gender assertion is a = (0.460/0.405) x 100 = 114 or, in other words, 14% higher than the “no assertion” base value, a = 100.
In a similar vein, one can define gender deference as the predicted probabilities of men and women regarding GE as non-essential where this ratio is represented by 6, 6 > 1: 6 = 1 means that gender deference is at its highest since the predicted probabilities of women and men in regarding GE as non- essential are the same; the greater the value of S, the greater the divergence between these two probabilities and, therefore, the lower the degree of gender deference. So, for example, Table 6.15 shows that the overall degree of gender deference is <5 = (0.291/0.244) x 100 = 119 or, in other words, 19% lower than the maximum deference value, 6 = 100.
Combining the two concepts of gender assertion and gender deference, one can define gender tension (GT) as the disjoint between women and men in their respective views of whether GE is essential or non- essential. Following from the above discussion, GT may be measured as a weighted average of a (gender assertion) and 6 (gender deference): GT = a x wn + 8xwt where wa and ws are the weights assigned to, respectively, assertion and deference, wn + wg = 1. If a = 6 = 1, GT = 1, and there will be no gender tension. On the other hand, GT will be larger, the greater the values of о (high degree of gender assertion) and 8 (low degree of gender deference). Without loss of generality, in computing the degree of GT, one could place equal weight on assertion and deference by setting
Table 6.1 S Estimates of gender assertion, deference, and tension by religion and
Source: Own calculations, World Values Survey 6.
Note: Index of Gender Assertion = (Probability of women regarding gender equality as essential/ Probability of men regarding gender equality as essential) x 1100; Index of Gender Deference = (Probability of men regarding gender equality as non-essential/Probability of women regarding gender equality as non-essential) x 1100; Index of Gender Tension = (Index of Assertion + Index of Deference)/2.
wa =wb = 1/2. Then the overall degree of GT is (114 + 119)/2 = 116.5, that is, 16.5% above its “no tension” value.
Table 6.15 shows that, at 123.2 - or 23% above the baseline no tension value of 100 - gender tension was highest, both in respect of assertion (122.2) and of deference (124.2), among Muslims and lowest among Catholics (111.3), non-Catholic Christians (114.7), and Buddhists (109.1). It is important to emphasise that the gender tension referred to is latent, as opposed to overt, tension. It does not mean that Muslim men and women are more likely to be at war with each other than men and women from other religions. What it does mean, however, is that Muslims displayed the greatest dissonance between male and female attitudes towards women’s rights and that Christians (Catholic and non-Catholic) and Buddhists displayed the smallest. Mirroring this finding, the highest level of gender tension - at 134.2 or 34% above the baseline no tension value of 100 - was in Islamic countries, and the lowest level was in Western countries (104.4), Latin America (106.6), and Asia (107.7).