Acknowledgements References

Notes

1. Theoretically, this income effect is relevant also in settings where children are net economic contributors in the long run, because costs at a young age must be covered (in the absence of well-functioning capital markets or fostering arrangements). However, a positive effect is very rarely found in empirical studies of any society. While a higher income may stimulate the demand for children, given quality requirements, the latter may also be sensitive to income.
2. There may also be an effect of enrollment itself, because it signals educational goals. For example, women in secondary school may prefer not to have a child yet, because a birth would make it much more difficult to complete the education, which not least has implications for life-time income.
3. However, education will not necessarily contribute to improve women's status. For example, it has been shown in some Asian countries that the uneducated actually may have more freedom of movement than the better-educated (e.g. [Balk 1994]).
4. Guilkey and Jayne [Guilkey and Jayne 1997] reported small differences between men's and women's fertility desires in Zimbabwe. A similar conclusion has been reached in other African studies, while some researchers (e.g. [Bankole and Sing 1998]) have documented substantial differences.
5. Boy preferences reflect women's generally subordinate position in society at present and an implicit expectation that this will be the case also for the next generation. While daughters' involvement in domestic work may be a great advantage for the often over-burdened mothers, they are likely to earn a lower income than sons, even if they receive the same education. This is because of the unacceptability of certain kinds of work for women, their weaker rights to land and inheritance, their more limited access to credit, and other factors. Besides, they may primarily have obligations to parents-in-law rather than their natal kin, depending on the family system. The motive for having a son may, in fact, be extra strong for women, because they are often much younger than their husbands and can expect a long period of widowhood. However, there are likely to be individual variations in boy preferences. Women who themselves have got an education, and who have a relatively strong position compared to husband and in-laws, may expect to back up their daughters so strongly that they would be just as rewarding in the long run as the sons.
6. Spelling this out more explicitly, the first better-educated women in a society where the most remunerative forms of economic activity have traditionally been considered inappropriate for women may perhaps be granted the right to take decisions in child care and some other daily activities, but not be allowed to make use of their human capital in formal employment.
7. The latter is dominant. In Zimbabwe, as in many other African countries, the mean duration of amenorrhoea is only a few months shorter than the mean duration of insusceptibility [Kirk and Pillet 1998].
8. The probability pij of, say, wanting another child for individual i in district j is given by
    log (pij / (1-pij)) = 1 xij +2 yj +3xij yj + uj ,
where 1 is the effect of a vector of individual characteristics xij , 2 is the effect of district characteristics yj , and 3 is an interaction effect (i.e. describing how the effects of individual characteristics change across district characteristics). The unobserved factors are represented by the random term uj , which is assumed to be normally distributed with mean 0.
9. Predictions show that the 15% difference in first-birth rates corresponds to an approximately 9-month difference in median age at first birth between these two educational groups. As a very simplistic argument (in lack of appropriate simulation tools), this delayed motherhood means that women with a complete primary education are exposed to the higher-order birth rates for a 9-month shorter period, with an impact on cumulated fertility that may be about 1/8 child, given higher-order birth rates that correspond to a median duration of about 6 years between births. More importantly, the 15% lower higher-order birth rates among the women with a primary education contributes another 1/2 child. A similar calculation is done for secondary education and referred to below.
10. Assume that the proportions in the five educational categories in the bottom-10 districts are b1, b2, b3, b4 and b5 , that the corresponding proportions in the top 10-districts are t1, t2, t3, t4 and t5 , and that fertility at these educational levels are 0, a2, a3, a4 and a5 relative to that at the lowest educational level. Then the individual-level impact is
    I = a2 (t2-b2) + a3 (t3-b3) + a4 (t4-b4) + a5 (t5 - b5)
11. This holds more generally. Assume, for example, that the expected number of births within a certain age interval for a woman i in district j (j=1,2) is given by fij = f0 + uij + pj, where uij is a dichotomous educational variable with values 0 and 1, and pj is the proportion of women in this age group in district j with uij=1. It is not difficult to show that, under usual assumptions about independence, the average expected number of births to all women with uij=1 (pooled together across both districts) would be + c higher than the corresponding average for all women with uij=0, where 0< c <1. This difference corresponds to the effect of individual education in a model where aggregate education is left out. Thus, if the proportion pj is increased from p1 to p2, the expected change in average fertility according to the true model is ( + ) (p2 - p1), whereas that according to a purely individual model would be ( + c) (p2 - p1).
12. Further investigation revealed no clear clues about the reasons for this. According to models where also age, parity, duration since last previous birth, religion and the urban/rural character of the current place of residence were included (not shown), a high average length of education in the district significantly enhances a woman's reported participation in decisions about work. It also weakens her boy preferences, increases her wealth (as proxied by various consumer items), reduces the probability that she reports to be non-working, reduces the probability that she lives in a polygynous marriage, and increases her general level of knowledge (as indicated by her knowledge about the protective effect of breastfeeding and about proper behaviour when a child has diarrhea). Most of this suggests relatively weak fertility desires. On the other hand, there was no impact on other potentially important determinants, such as death of older children, the use of foster parents for older children, participation in decisions about purchases, and (if she reported to work) whether she earns cash and can keep this herself. Besides, her wages (if she worked for cash) were actually lowest in districts with a generally high educational level, while her own education increased wages very sharply. When all these variables were included in the model for fertility desires, the estimated effect of aggregate education was not changed. This means that they are either relatively poor indicators of women's status, knowledge, wealth, work activity, wage potential and family situation, or that other dimensions are generally much more important. (Inclusion of these variables also explained only about 1/3 of the individual effects of a secondary education longer than 2 years).

 

Acknowledgements References

A Search for Aggregate-Level Effects of Education on Fertility, Using Data from Zimbabwe
Øystein Kravdal
© 2000 Max-Planck-Gesellschaft ISSN 1435-9871
http://www.demographic-research.org/Volumes/Vol3/3