V. Summary and policy implications.
The individual cases reveal that circumstances in each country are quite different. Nevertheless, one can still learn from a comparison of the patterns across countries. For example, as Table 1 demonstrates, while assets may be difficult to measure precisely, women appear to bring far fewer assets to the marriage, both in terms of physical and human capital. (The only exception is in matrilineal Sumatra, where women bring more paddy land to the marriage. However, they have less schooling than their husbands.) While the expenditure share analysis focuses only on physical capital (assets) brought to marriage, there is also much evidence that differences in the human capital of husband and wife--education, age, and experience--have significant effects on intrahousehold allocation (see, for example, Thomas 1994 and Thomas et al. 1997). The importance of human capital on intrahousehold allocation is illustrated in the section on individual educational outcomes.
Table 10 presents the effects of the various measures of bargaining power for all four countries. Each column reports the sign of the significant coefficients for the measures of bargaining power in a specific country. For example, the fourth through sixth rows in the first column indicate that for the share of education expenditures in Bangladesh, the effect of wife's assets are positive and significant, the effect of the husbands assets are insignificant, and the difference between them is positive and significant.
These results show that the unitary model of the household is, on the whole, rejected. This finding is stronger in the Asian countries than in the African ones, in the sense that the unitary model is rejected in more expenditure share equations in the former. While strictly speaking, rejection for any single equation implies the unitary model does not hold, there is a sense in which more rejections for a single country are evidence of the extent to which the unitary framework breaks down. As such, it appears that the model is most strongly rejected in Indonesia and represents a closer approximation to reality in South Africa, in part due to the sample selection issues described above. The rejection of the unitary model has implications for the design of policies designed to transfer resources to households." the identity of the transfer recipient does affect the ultimate outcome of the intervention.
Across countries, the most consistent effect is that relative resources controlled by women tend to increase the shares spent on education (in all countries but Ethiopia). While it is tempting to say that mothers are more altruistic than fathers, this behavior may have a sound economic basis. Given age differences at marriage (women are younger) and gender differences in life expectancy, it is possible that women invest in the education of their children more heavily since they are more likely to rely on them for old age support. In societies where key assets that assure lifetime consumption-smoothing are controlled by men (land, in many cultures; pensions and social security in countries with low female participation in the formal labor market), women may attempt to meet the same long-term needs with other instruments, such as investment in the human capital of healthy and educated children (Guyer 1997: 121).
Having found the most consistent effect at the household level was expenditures on education, we then turned to a more direct examination of educational outcomes. This provides a more relevant test by examining an outcome with which policy makers are directly concerned. In addition, it is a more powerful test since we can distinguish between the effects on boys and girls within households. The evidence described for years of schooling and deviations from cohort means supports the results from the expenditure share regressions (assets brought to marriage, including human capital by husband and wife have differential effects on allocation in the household) but also suggests the mechanisms underlying the earlier results are quite complicated; indeed they differ substantially across the case studies. The expenditure share regressions indicated that more assets in the hands of women had a beneficial impact on budget shares for education, but did not tell us which of the children were benefiting.
For example, in both Bangladesh and South Africa there is evidence that more assets in the hands of women have a positive impact on the educational budget shares. Yet, in Bangladesh, fathers schooling (for the 6-10 year olds) and assets (for the 11-15 year olds) have a negative impact on girls schooling whereas in South Africa it is the opposite, fathers schooling has a positive effect on girls schooling while mothers assets brought to marriage has a negative impact on girls. In South Africa, the pattern may be partly justified using the old-age security hypothesis outlined earlier, but in Bangladesh this is not true and different preferences are more likely the underlying cause. Wealthier Bangladeshi fathers may attach a higher premium to marrying their daughters off earlier, an effect opposite to that of better-educated mothers. Finally, the differences found between parental effects on children of different gender provide further evidence that households in these four countries are not operating within a unitary framework.
While one could rashly recommend unilaterally transferring assets to women, programs designed to transfer assets to women should be designed with caution. First, while the expenditure share evidence suggests more assets in the hands of women leads to higher budget shares for education, the beneficiaries of these gains (boys or girls or both) are different across the countries. These differences appear to be driven by both differences in preferences and underlying economic rationales possibly related to old-age support systems in different countries. An understanding of the latter is an important ingredient into policy making aimed at exploiting these differences. Second, different assets may have different implications for bargaining power if "status" or prestige is attached to a particular asset. For example, in the Sumatran case, paddy land (which is considered a higher form of wealth) and forest land affect expenditure shares in different directions. Indeed, the special meaning or significance attached to ritual transfers such as dowries or brideprice should warn us against asset-transfer interventions that are designed without paying attention to cultural contexts. Lastly, we must also remember the possibility of compensatory (or even retaliatory) action by nonrecipients. The cases of husbands taking control of an irrigation project designed to preserve womens control of rice in the Gambia (Dey Abbas 1997) and Bangladeshi wives borrowing money for their husbands' use from credit programs ostensibly targeted to women (Kabeer 1998) are often mentioned as words of caution to policymakers. Even more disturbing is the possibility of increased domestic violence towards women, should income transfer programs radically alter the distribution of power between husband and wife (Schuler et al. 1997b).
Our results also show that influences on intrahousehold allocation may be operating at different levels, with different implications for policy. For example, in the Ethiopian case, the effect of husband's and wife's assets was dominated by the site-specific characteristics, ethnic and religious differences. This indicates that variations across communities and ethnic groups may be larger than variations in the asset position of men and women within those groups. In this case, legal reforms which affect property rights, or which equalize property rights across groups, might have a larger potential impact on intrahousehold allocation than redistributing resources among men and women within the group. Preliminary results (not reported here) on the distribution of assets upon divorce in Ethiopia, for example, show that having a written marriage contract, which is typical of some ethnic groups but not others, increases the share going to the woman. In the Bangladesh case, on the other hand, despite significant differences across sites, differences in asset positions of men and women within sites are large enough to warrant interventions to increase women's assets relative to men's.
|Printer friendly Cite/link Email Feedback|
|Title Annotation:||Intrahousehold Allocation and Gender Relations: New Empirical Evidence|
|Publication:||Intrahousehold Allocation and Gender Relations-New Empirical Evidence|
|Date:||Oct 1, 1999|
|Previous Article:||IV. The impact of male and female physical and human capital on intrahousehold outcomes.|