ETHIOPIAN DEVELOPMENT RESEARCH INSTITUTE<br />URBANIZATION AND FERTILITY RATES IN ETHIOPIA <br />Fanaye Tadesse and Derek Headey<br />IFPRI ESSP-II<br />Ethiopian Economic Association Conference<br />July 21, 2011<br />Addis Ababa<br />1<br />
1. Introduction<br />Ethiopia has a long history of Malthusian population dynamics (Pankhurst 1985). <br />The population-dense highlands face shrinking farm sizes, deforestation and soil degradation<br />All these problems are related to high fertility rates<br />Also widely accepted that reducing fertility can produce a demographic dividend via reduced age dependency <br />Hence Ethiopian government t has long sought to reduce fertility<br />Good news is that fertility rates are falling<br />
1. Introduction – cont’d<br />But Ethiopia still has the largest rural-urban fertility differential in the world: 6 children in rural areas; 2.4 children in urban areas. <br />So why is the rural-urban fertility gap so large in Ethiopia?<br />Previous work does not satisfactorily explore this question.<br />Some papers look at proximate determinants of fertility in the Bongaarts framework or explain fertility for certain proportion of the population<br />A large World Bank study used DHS and other datasets to look at fertility rates, including by rural and urban areas<br />But not systematic tests for rural-urban differences<br />
Specific objectives of the paper<br />1) What causes fertility in Ethiopia?<br />2) How do these causes differ by rural and urban areas? And<br />3) How will different socio-economic development scenarios influence Ethiopia’s future fertility trends?<br />
2. Economic theories of fertility<br />Economic theories tend to emphasize demand-side determinants of fertility rates, following Gary Becker’s seminal work. <br />Emphasis is on choice, rather than biological factors<br />Children possess both consumption good characteristics and investment good characteristics<br />Quality and quantity tradeoffs, which posits a likely substitution from quantity to quality as family income increases (Becker and Lewis 1973). <br />Opportunity costs matter – children come with both explicit costs and implicit costs – e.g. women’s time use<br />
Child mortality rates influence expectations; risk averse parents may have more children than actually desired<br />Women’s education tends to reduce fertility, but the level of education that affect fertility is still not clear. <br />Age at marriage also matters, but as a proximate determinant it is closely connected to other factors, such as female education, poverty, location, etc<br />Urbanization seems to affect fertility (Kuznets 1974), sometimes even when other controls are introduced<br />But not clear why – many rural-urban differences in poverty, education, infrastructure, and unobservables<br />2. Economic theories of fertility<br />
3. Data and Estimation<br /> The data used for this study is EDHS (Ethiopian Demographic and Health Survey) of 2005. <br />The ERHS is a nationally representative survey of 14,070 women between the ages of 15 and 49 and 6,033 men with ages between 15 and 59. <br />Topics include family planning, fertility, child mortality, child health, nutrition and knowledge of HIV/AIDS. <br />Geographic Information Systems (GIS) estimates of travel times to nearest facilities were merged with the DHS data<br />GIS data adds info on isolation on, and may also be important since rural-urban divide can be arbitrary<br />
3. Data and Estimation<br />Dependent variables are number of children born and desired number of children<br />Former is more like revealed reference, latter is stated preference<br />Desired number has two phrasing depending on age:<br /> “If you could go back to the time you did not have any children and could choose exactly the number of children to have in your whole life, how many would that be?”<br /> “If you could choose exactly the number of children to have in your whole life, how many would that be?”<br />
3. Data and Estimation<br />Some potential problems . . . <br />DHS does not have information on consumption or income variables, but measures a wealth index constructed from the information on asset holdings<br />Possible endogeneity of Child Mortality <br />Rather than including the Child mortality of the household, we took the average of the child mortality in the cluster.<br />So it is a locally formed expectation rather than actual for the household.<br />
<ul><li> where the are the parameters to be estimated,
the expression ( ) are interactions of the explanatory variables with the urban dummy.
The interaction variable obviously tests whether the effects of the explanatory variables on fertility differ by location.
We also separately estimate urban and rural equations and conduct Chow tests to check for parameter differences among these groups.
We use Probit regressions since these are count variables</li></li></ul><li>4. Results – basic descriptive statistics<br />Differences in dependent variables<br />
Explanatory variables: All differences significant at 5% level<br />
4. Results – basic descriptive statistics<br />In terms of some more proximate determinants, contraceptive use among women is just 18%: 37% urban, 15% rural<br />Reasons for not using contraceptives vary across rural and urban areas<br />Rural women are more ignorant of contraceptives, desire more children, and face slightly more opposition from husbands<br />Age at marriage differences are also significant at about 1.2 years (15.8 versus 17.0 in rural areas).<br />In both areas legal age at marriage is below legal age (18 yrs)<br />
4. Results – regression results<br />We begin with national regressions that interact explanatory variables with an urban dummy<br />Most of the interaction terms are significant, suggesting not only differences in levels, but differences in impacts too <br />The urban dummy becomes insignificant once interaction terms are introduced.<br />Wealth index not very significant in either rural or urban areas<br />Much larger effect of female education (only secondary and above) and work status<br />Child mortality has big effects, especially in urban areas<br />
Regression Results (Poisson marginal effects)number of children born <br />
4. Results – regression results<br />For desired number of children, most results are quite similar, but rural-urban differences appear to be less significant<br />Some evidence that village level contraceptive knowledge, access to radio matters, suggesting a role for fertility policies<br />
5. Conclusions<br />Rural-urban fertility gap is explained by both difference in levels of explanatory variables and differences in impacts. <br />Wealth, by itself, does not seem to matter much<br />Most policy-relevant findings related to female secondary education, age at marriage, and raising awareness of family planning goals and technologies<br />Female secondary education likely to have high returns because in addition to reducing fertility, it can increase incomes and improve nutrition outcomes<br />Currently female education is so low in rural areas (3.2%) that there is huge scope for expansion <br />
5. Conclusion<br />In terms of future research we plan to more formally decompose rural-urban differences into level effects, parameter effects and unexplained effects (e.g. Oaxaca decomposition).<br />We may also update with forthcoming 2010 DHS<br />We can explore regional effects more as there are fertility differences across regions, even within rural and urban areas<br />