6 Funcionamiento 59
6.4 Tablas de función
6.4.3 Menú C, Selección
somewhat higher than the small separate effects. In addition it should be noted that the effects of the individual value variables does vary across estimation techniques (Appendix 6.1). The signs are all according to expectations, but for age at birth of the first child and a partner’s education only the MCMC model shows statistical significant results. For traditional household structures and age difference all models do. Despite these somewhat less robust results, it is safe to conclude that overall micro-level values do have the expected impact on women’s employment.
3.14 To the extent that other household members hold more traditional values, the smaller the probability that a woman in ✓ that household is gainfully non-agriculturally employed.
3.15 To the extent that more traditional values are internalised, the smaller the probability that a woman is gainfullynon
✓ agriculturally employed.
Box 6.6
3.16 To the extent the institutionalisation of conservative Islam into the state increases, the smaller the probability ✓ that a woman in that country is gainfully non-agriculturally employed.
from entering the educational system and that consequently their likelihood of employment decreases. It should be noted however, that the effect found for traditionalism at the district level is not consistently found across models using different estimation methods.71
Based on the analyses in this section, it can be concluded that values and norms seem to have a direct and an indirect effect on women’s employment. If we include the possible indirect effects, both hypotheses (3.12 and 3.13) seem to be corroborated. The support for Hypotheses 3.13, however, rests solely on the indirect effect through education.
At the micro-level, four variables have been used to measure the impact of the (internalised) values of women and their household members: (1) the age of the woman at the birth of her first child, (2) the age difference with the partner, (3) the partner’s education, and (4) the presence of traditional household structure. The more traditional the values, the lower the expected employment propensity. The results are summarised in Box 6.6.
Each of these variables appears to be related to women’s employment, but all decrease in strength considerably after controlling for other micro-level factors. Moreover, the age of the woman when she gave birth to her first child shows a small negative effect (p<0.01) where a positive was expected.72 Overall, it still seems that the cluster as a whole shows the expected
results. The decrease in effects after controlling for factors such as education and having children suggests indirect effects as well. Over the full range, increasing age difference between spouses decreases the odds on employment by about 50%. Rises in the educational level of the husband shows a steady increase in odds on women’s employment, and the odds are 12% higher for women with tertiary-educated partners compared to women with uneducated partners. Lastly, women living in household with the least traditional structure (nuclear, not polygynous) have 28% higher odds of employment than women in the most traditional (extended, polygynous) structures, 1.4% of all women included here.
While each of these effects is limited in strength, it should be taken into account that they all work simultaneously, meaning that the effect of the micro-level values cluster as a whole is larger.73 With regard to Hypotheses 3.14 and 3.15, the four value variables I have used do
not neatly distinguish between the two hypotheses; as they overlap it is difficult exactly to determine whether the behaviours used as proxies tap into a woman’s or, for instance, her partner’s values. Only a partner’s education seems rather safe to declare an indicator of the partner’s values. The other variables tap into both the value system of the household in which a woman was raised, her own values, and the values of the partner. Yet, since three out of four variables show statistically significant effects in the expected direction, I do think the evidence is solid enough to consider both the hypotheses supported.
In the analyses, the institutionalisation of conservative Islam was included to test the idea that institutionalised values also influence women’s employment (Box 6.7). In all models and the bivariate analyses, this country-level variable is the only one that is consistently and strongly related to women’s employment. The more conservative Islam is institutionalised in a country, the lower the probability that a woman in that country is employed. Most countries are coded 0 (13) or 2 (5) on this variable ranging from 0 to 6. Women in a country with a score of 2 have odds of employment 58% lower than women in a state in which Islam is not institutionalised at all. Overall, in these analyses, many different value variables simultaneously, but also differently, affect women’s employment. This strongly supports the view that ‘values’ is not just one influencing factor, while different values are often treated as one in cultural explanations. Moreover, values can be internalised by women or by their direct environment. They can take the form of societal norms, and they can become institutionalised in the state. Distinguishing between these helps to think about what and how changes in value patterns might or might not influence women or not and how. Moreover, a change in the values upheld in society might change women’s position but not directly lead to changes in the institutionalised values. Through institutionalisation, ‘old’ values can have unexpected long-lasting effects. On the other hand, institutionalised values might be more easy to change than societal norms or internalised values. Also, results suggest that values can be grounded in different sources, as is evident when surveying the difference between Islam-based values to values more generally related to traditional gender roles.
3.17 To the extent that traditional gendered family roles are more institutionalized in government policies, the smaller the
?/? ✓probability that a woman in that country is gainfully non-agriculturally employed.
3.18 To the extent that the size of the social safety net in a country increases, the smaller the probability that a woman in that
?/? ✓country is gainfully non-agriculturally employed.
Box 6.8
3.19 To the extent that the level of democracy in a country is higher, the greater the probability that a woman in that country
?/? ✓ isgainfully non-agriculturally employed.
Box 6.9
theoretical factor expec-
tation re- sults operationalised variable expec- tation re- sults accepted
H3.1: Presence male breadwinner - ✓ Partner present - ✓
Working male is head of household - ✓
H3.2: Presence children - ✓ presence of children + ✓
H3.4: Male labour supply - ✓ Non-employed men in district - ✓
H3.6: Urbanisation + ✓ Living in a city + ✓
Degree of urbanisation in district + ?✓
H3.8: Service sector + ✓ Number of service sector jobs in district + ✓
H3.10: Human capital + ✓ Education + ✓
H3.11: Socio-economic class network + ✓ Occupation(al status) partner + ✓
H3.12: Norms public presence women (pro) + ✓ Relative presence women in education and labour
+ ✓
H3.14 & H3.15: Traditional values household members &
traditional values woman
-
-
✓ ✓
Partner’s education
Age difference with partner (partner- women)1
Traditional family structure1
Age at birth first child1
+ - + - ✓ ✓ ✓ x
H3.16: Institutionalisation conservative Islam - ✓ Codification of Islam in constitution - ✓ accepted, but also counteracting effects found
H3.3: Economic development + ✓ Country: GDP/c + x
District: Wealth +/- ✓ Possibly in-
cludes a nega- tive indirect effect
Cautiously accepted
H3.5: Light manufacturing + ?✓ Proportion skilled labour + ?✓
H3.13: Norms traditional care roles (pro) - ?✓ Prevalence of traditional households - ?✓
H3.17: Traditional gendered family role policies - ?/? ✓ Gender unequal policies and laws - ?/? ✓
H3.18: Social safety net - ?/? ✓ Family allowance - ?/? ✓
not accepted
H3.7: Foreign direct investment + x Ln (FDI three years average (% GDP)) + ? x
H3.9: Public sector size + x Government expenditures + ? x
H3.19: Democracy + x Freedom House civil liberties and politi- cal rights
+ ? x
Notes: (1) While these value proxies are related to either the degree of traditionalism of the household member or of the women, the variables are expected to capture both.
6.7 tHe roLe of tHe governMent: SHapIng opportunItIeS and needS
The last group of variables has drawn attention to how government policies reflect values and shape opportunities and needs. Reinforcing dominant gender roles was expected to influence employment negatively as were policies decreasing economic need (Box 6.8); on the other hand, giving women room to influence policies was expected to increase employment (Box 6.9).
The bivariate models (Table 6.1) indicate that more traditional family policies are related to lower employment, as well as an increase in family allowances. The latter is only significant at the 5% level, but given the limited number of countries (28), this still seems to support the idea that family allowances discourage women from entering the labour market. However, for both variables the results do not hold when other variables are included in the model. Different groups of variables can account for the difference found between models. Consequently, it is unclear whether the bivariate results are spurious, top-down indirect, or absent because of technical reasons.
Spuriousness could firstly be derived from disappearing effects after the inclusion of economic and labour market structure variables and societal norms. In this case, it could also be argued that the policies work through district-level norm variables. However, the policy variables focus on very specific policies, whereas the norm variables focus on general norms. As the measurement of the policy variables comes closer to the direct actions of women than to the norms, it is unlikely that these particular policies have a large influence on norms, without having a direct effect on women’s employment. Secondly, the traditional family policies might lead to a lower educational participation of women, and consequently to lower employment rates. The family allowance policies might shape households’ fertility decisions and have an indirect effect that way. However, the effect of policies on fertility can be expected to be fairly limited (see Olmsted, 2003: 87–90) and thus even smaller with regard to women’s employment. Thirdly, the relatively weak operationalisation might mean that a more encompassing variable such as state Islamisation more effectively taps into the overall policy climate. In short, policies might have indirect effects, but the expectations are not strongly supported (see Box 6.8).
Lastly, democracy was expected positively to influence women’s employment, mainly through policy-making. The raw association with women’s employment (Table 6.1) shows that women in the most democratic countries in the dataset have 37% higher odds of employment than women in the least democratic country. However, including the state institutionalisation of conservative Islam takes away that effect for my sample of 28 countries, in which the two variables overlap substantially.74
Overall, the results for the political variables certainly indicate that family policies and benefits, as well as state structures, all relate to women’s employment, and that they relate to women’s employment in the expected direction. However, the results from these analyses are too ambiguous to draw strong conclusions about the influence of state structures and policies measured by the variables in this thesis. More attention for these kinds of factors, using better operationalisations and more cases, could shed more light on the possible influences of policies and polities.
6.8 ConCLuSIonS: overaLL reSuLtS
6.8.1 SUMMARY
Summarising, of the nineteen hypotheses formulated in Chapter 3, eleven are rather
unambiguously accepted in this chapter (see Box 6.10). The hypothesis on economic development is supported as well, but aside from the hypothesised positive effect there seems to be a negative effect coming from economic development. On four other hypotheses, the results do suggest an influence, but the results are less robust. Lastly, three hypotheses are provisionally rejected. For these, I found no to hardly any proof that they have a general effect on women’s employment in the 28 countries. This is not to say, that these factors (FDI, public sector size, and democracy) cannot have an effect in a certain context;75 however, based on the data used here
77 For each variable the full range on this variables was