77629 Development, Modernization, and Childbearing: The Role of Family Sex Composition Deon Filmer, Jed Friedman, and Norbert Schady Does the sex composition of existing children in a family affect fertility behavior? An unusually large data set, covering 64 countries and some 5 million births, is used to show that fertility behavior responds to the presence—or absence—of sons in many regions of the developing world. The response to the absence of sons is particularly large in Central Asia and South Asia. Modernization does not appear to reduce this differential response. For example, in South Asia the fertility response to the absence of sons is larger for women with more education and has been increasing over time. The explanation appears to be that a latent demand for sons is more likely to manifest itself when fertility levels are low. As a result of this differential fertility behavior, girls tend to grow up with signi�cantly more siblings than do boys, with potential implications for their well-being when quantity –quality tradeoffs result in fewer material and emotional resources allocated to children in larger families. JEL codes: J16, J13, O15 A family preference for sons over daughters may manifest itself in various ways. An especially stark dimension is the excess mortality among girls docu- mented in several Asian countries (see, for example, Zeng and others 1993 for China; Muhiri and Preston 1991 for Bangladesh; and Das Gupta 1987 for India). A similar phenomenon has been documented in the Middle East (Yount 2001). Son preference can also manifest itself through lower investments in the human capital of girls. Pande (2003) documents lower nutrition and immuniz- ation rates among girls in India. School enrollment and attainment among girls Deon Filmer (corresponding author) is a lead economist in the Development Economics Research Department at the World Bank; his email address is d�lmer@worldbank.org. Jed Friedman is a senior economist in the Development Economics Research Department at the World Bank; his email address is jfriedman@worldbank.org. Norbert Schady is a senior economist in the Development Economics Research Department at the World Bank; his email address is nschady@worldbank.org. The authors thank Monica Das Gupta, Peter Lanjouw, Cynthia Lloyd, T. Paul Schultz, and three anonymous referees for valuable comments and suggestions, and Ryan Booth and Nicholas Ingwersen for outstanding research assistance. They are grateful for �nancial support from the Hewlett Foundation’s Trust Fund on Fertility, Reproductive Health, and Socioeconomic Outcomes and the Government of Norway. THE WORLD BANK ECONOMIC REVIEW, VOL. 23, NO. 3, pp. 371 –398 doi:10.1093/wber/lhp009 Advance Access Publication October 23, 2009 # The Author 2009. Published by Oxford University Press on behalf of the International Bank for Reconstruction and Development / THE WORLD BANK. All rights reserved. For permissions, please e-mail: journals.permissions@oxfordjournals.org 371 372 THE WORLD BANK ECONOMIC REVIEW lags behind that of boys in many South Asian, Middle Eastern, and North African countries (Filmer 2005).1 This study focuses on one manifestation of a “preference� for sons—a greater propensity for continued childbearing given an all-female rather than an all-male composition of existing children in the family. Such behavior could be the result of taste-based sex discrimination or of economic concerns, such as higher costs of investing in girls than in boys or lower pecuniary returns to investments in girls than in boys. Therefore, while differential fertility-stopping behavior is related to preferences, it is the result of a larger set of factors. There are numerous possible reasons for observing differential fertility-stopping behavior in the developing world. Typically, they derive from conditions found in many traditional rural societies, such as inheritance systems that pass assets to sons, intergenerational insurance systems in which sons care for parents in old age, or production systems with low pecuniary returns to women’s work (and to investments in women’s human capital). General develop- ment processes and modernization, including urbanization, the dissolution of tra- ditional rural communities, and increasing female education and labor force participation, are expected to work against these pressures for differential fertility-stopping behavior in settings where it exists (see, for example, Chung and Das Gupta 2007). This article explores the extent of son-preferred differential fertility-stopping behavior in the developing world; how it varies across countries and regions; whether it is associated with measures of modernization, such as urbanization, women’s education, and wealth; and its potential consequences for household demographic composition and the investment in girls’ human capital. A handful of empirical studies have investigated differential fertility-stopping behavior at various levels of economic development. Hank and Kohler (2000) focus on European countries. Using Fertility and Family Surveys for 17 countries, they �nd substantial heterogeneity across countries, with a tendency toward a mild preference for a mixed-sex composition of children in a family. Their data suggest a preference for girls in the Czech Republic, Lithuania, and Portugal. Andersson and others (2006) use historical data from Denmark, Finland, Norway, and Sweden to show no effect of sex on fertility for second births, a desire for sex balance at third births, and heterogeneity across countries at fourth births (son preference in Finland and daughter preference in the other three countries). For developing countries, most of the literature has focused on individual Asian countries with a prevalence of discrimination against women.2 An 1. See World Bank (2001) for a more general discussion of differences between boys and girls in inputs and outcomes. 2. For example, Park (1983), Arnold (1985), Bairagi (1987), and Larsen, Chung, and Das Gupta (1998) show the strong impact of son preference on future fertility in the Republic of Korea; Arnold, Choe, and Roy (1998), Dre ` ze and Murthi (2001), and Jensen (2007) �nd evidence that son preference affects fertility behavior in India; Haughton and Haughton (1995) show a similar pattern in Vietnam; while Pong (1994) and Leung (1998) document the pattern among ethnic Chinese in Malaysia. One study addresses the issue in Egypt, with a similar �nding of son preference affecting fertility behavior (Yount, Langsten, and Hill 2000). Filmer, Friedman, and Schady 373 important exception to these country-speci�c studies is Arnold (1992, 1997), who considers the impact of sex ratios on subsequent fertility behavior across many developing countries. Arnold (1992) shows that the most typical pattern in the 26 countries he studied is of a preference for at least one son and one daughter. He �nds some weak evidence for son-preferred differential fertility-stopping behavior in North Africa and Sri Lanka. Arnold (1997) ana- lyzes data for 44 countries but focuses largely on the effect of sex ratios on stated fertility preferences and on some fertility behaviors, such as current preg- nancy status and average birth spacing. He �nds regional variation in the extent of an association between sex ratios and the outcomes he analyzes, with the strongest results suggesting son-preferred differential fertility-stopping behavior for the Asian and North African countries. This article uses information on 5 million births by 1.3 million mothers in 64 countries to analyze how the sex mix of children in a family affects fertility decisions in the developing world. The article extends the literature in impor- tant ways. The analysis includes a large number of developing countries from disparate regions. The article documents not only regional patterns in son- preferred differential fertility-stopping behavior, but also within-region differ- ences by location (urban or rural), education (women who have completed primary school and those with less schooling), wealth levels (above and below the median of a composite measure of assets), and over time (different birth cohorts of mothers). The article analyzes the extent to which observed patterns in son-preferred differential fertility-stopping behavior strengthen or weaken as the total number of children decreases. Moreover, �nally, the results are linked to the wider literature on sex composition and resource dissolution in larger families. I. METHODS AND D ATA This section describes the methodology, starting with a model for estimating the impact of the sex balance of children in a family on the probability of sub- sequent births. It then details the data used for the analysis. Estimating the Impact of Sex Balance on Fertility Behavior The basic model estimates: ð1Þ Bwnþ1 ¼ a þ bmn � Mwn þ b fn � Fwn þ uwn for n ! 2 where Bwnþ1 is a zero or one outcome variable indicating a birth for woman w with a preexisting number of children n; Mwn is a variable equal to one if woman w had no sons at family size n; Fwn is a variable equal to one if woman w had no daughters at family size n; and the term uwn is a random error. This regression is run separately for each existing family size. 374 THE WORLD BANK ECONOMIC REVIEW The omitted category in the regression is women who have at least one son and one daughter. The coef�cients bmn and bfn can therefore be understood as probabilities of additional childbearing for women who have children of only one sex, relative to those who have children of both sexes. Positive coef�cients are evidence of preferences for a sex mix of children over children of one sex only. A signi�cantly positive difference between the two coef�cients (bmn – bfn . 0) indicates that a woman is more likely to have another birth if she has no sons than if she has no daughters. As in much of the literature (see Key�tz 1968 and Repetto 1972 for early examples), this is referred to as son-preferred differential fertility-stopping behavior. Though sometimes referred to here as “son preference,� the meaning refers exclusively to fertility decisions, as described above, rather than to other possible manifestations of differential behavior toward sons and daughters after birth, as might be evident in differ- ences in mortality, nutritional status, or school enrollment by sex. A negative difference (bmn – bfn , 0) indicates daughter preference in childbearing. Because calculating separate estimates for each pre-existing family size pro- duces a large number of coef�cients for bmn and bfn, for most results the focus is on averages across different family sizes—for individual countries or regions and for speci�c groups (by education, location, wealth, and birth cohort). For this purpose, the means bm and bf are de�ned as follows: X 1 ð2aÞ bg ¼ wgn � bgn for g ¼ m; f n¼2 where wgn is the relative weight for family size n (and the weights sum to one). With independence assumed across parities, the corresponding standard error of bg can also be calculated as follows: sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi X 1 ð2bÞ sg ¼ w2 gn � vbgn for g ¼ m; f n¼2 where vbgn is the square of the estimated standard error of bgn.3 One concern is that including in this analysis women who have not yet com- pleted fertility may bias the results if women who enter childbearing at later ages have different preferences from those who begin childbearing earlier or if birth spacing is partly a function of the sex mix of existing children. To 3. A related alternative approach is to pool all observations at different parities and estimate a model that relates the probability of an additional birth as a function of the share of sons among existing children. Since women appear more than once if they progress beyond three children—for example, a woman with four children would appear twice, once for the transition from two to three children and again from three to four—the model would also include additional controls for the existing family size at each observation. This model can be supplemented with other observable information, such as the location and education of the mother. Analysis of this model serves as a robustness check for the main results and is discussed later. Filmer, Friedman, and Schady 375 overcome this problem, the sample is generally limited to women ages 40–49, on the assumption that these women have completed their lifetime fertility (the data do not include women older than 49). To highlight the largely consistent estimates obtained with the two approaches, results based on the entire sample are occasionally compared with those for women ages 40 –49. An important part of the analysis is the exploration of heterogeneity. In addition to heterogeneity by family size, the article explores differences based on location, education, and wealth. In the case of rural or urban location, the following regression is run: Bwnþ1 ¼ a þ Rw þ bmn � Rw � Mwn þ b fn � Rw � Fwn þ cmn � ð1 À Rw Þ � Mwn ð3Þ þ c fn � ð1 À Rw Þ � Fwn þ uwn forn ! 2 where the Rw is an indicator variable equal to one for women in rural areas; Rw � Mwn and Rw � Fwn equal one for women in rural areas who have had no sons or no daughters; and ð1 À Rw Þ � Mwn and ð1 À Rw Þ � Fwn are equal to one for women in urban areas who have had no sons or no daughters. The aggre- gated coef�cients bm, bf, cm, and cf are reported, along with tests for signi�cant differences between them (based on the formulas in (2a) and (2b)). This arrangement enables testing whether any observed son (or daughter) preference differs in rural and in urban areas by testing whether (bm – bf ) ¼ (cm – cf ), a test of difference-in-differences. A similar logic applies to differences by education levels and wealth. A woman’s reported current residential location de�nes the indicator vari- able used to test for differences between women in urban and rural areas. To test for differences by education, the indicator variable used splits the sample into those who have completed fewer than six years of schooling and those who have completed six or more. (Six years of schooling corresponds to completing primary school in most countries in the sample.4) The analysis by household wealth is based on a composite measure of household durable goods—an approach popularized by Filmer and Pritchett (2001).5 For each country, the indicator variable divides the sample according to whether the household falls above or below the median household wealth scale. To investigate whether son-preferred differential fertility-stopping behavior increases or decreases over time across birth cohorts of women, differential 4. A different approach was also used, calculating the median years of education for women in each country and dividing the sample into those above and those below the median. These results were very similar to those reported here. 5. One drawback with this measure is that it reflects household wealth only at the time of the interview, whereas this study considers the full fertility history of each mother—a history that can stretch back 20 years or more. Thus, the wealth index is not an entirely accurate measure of resources available to mothers at the time of decisions about fertility continuation, although there is a positive correlation between current and previous levels of wealth. Considering these interpretive dif�culties, this article does not stress the results based on wealth. Early applications of this asset index approach include Pollitt and others (1993) and Rivera and others (1995). 376 THE WORLD BANK ECONOMIC REVIEW fertility-stopping behavior is calculated within each country for every one-year birth cohort—for example, women in India born in 1945—and then the corre- sponding regional averages in each year are calculated—for example, for women in South Asia in 1945. A �rst step is to graph these regional averages. As a more formal test of changes in differential fertility-stopping behavior, sep- arate regressions are run on a set of �ve-year birth cohort dummy variables by region, to test for differences in these dummy variables. One concern with these estimates is that any observed changes in differential fertility-stopping behavior across birth cohorts could be driven by changes in the countries that make up the regional averages—some countries have surveys only in earlier years and therefore enter only into calculations of regional averages for early birth cohorts, while other countries have surveys only in later years and enter only into regional calculations for later cohorts. Thus, estimates are also pre- sented that keep �xed the countries in each regional sample and the weight given to each in calculating the regional average. As a �nal step in the analysis, a multivariate framework is applied based on location–education –cohort cells. This is done primarily because, as shown, prevailing fertility rates have a signi�cant effect on estimated differential fertility-stopping behavior and are correlated with other observable factors. The basic regression is then: ðbm À bf Þrht ¼ br Dr þ bh Dh þ bt Dt þ bF Frht þ urht ð4Þ where (bm – bf )rht is the measure of differential fertility-stopping behavior, as before, for a given location–education–birth cohort cell; Dr and Dh are dummy variables for women in rural areas and high-education women; Dt is a measure of a woman’s birth cohort (in practice, birth cohorts in this part of the analysis are aggregated over three years, to keep the sample sizes reason- able); and Frht is the average number of children born to women in a given location–education –birth cohort cell.6 The resulting sample includes 3,456 observations for 64 countries. Each country-year contributes four observations corresponding to the four location–education groups for women born in that year. In estimating equation (4), observations are weighted by N, the number of women in each cell. By giving greater weight to cells with larger sample sizes, this method more precisely estimates values of differential fertility-stopping behavior. Data Data are from 158 Demographic and Health Surveys (DHS) for the 64 countries listed in the appendix. The data contain the complete retrospective fertility his- tories of 1.3 million women in the 64 countries, as well as socioeconomic 6. Household wealth is not included in this analysis because of the limitations discussed earlier; however, results are largely unchanged when wealth is included. Filmer, Friedman, and Schady 377 information such as educational attainment, ownership of durable goods, and household location.7 For comparisons across developing country regions, countries are assigned to geographic regions following World Bank de�nitions: East Asia and Paci�c, Europe and Central Asia, Latin America and the Caribbean, Middle East and North Africa, South Asia, and Sub-Saharan Africa (see the appendix). Note that the countries observed in the East Asia and Paci�c region include only countries in Southeast Asia and that those in the Europe and Central Asia region include only countries in Central Asia, and hence these regions are referred to here as Southeast Asia and Central Asia. In general, observations in each survey are weighted by their expansion factors, which reflect differences in the probability that households are sampled in the DHS.8 When regional averages are constructed, observations are reweighted so that each country contributes its relative population share to the regional sample; population estimates for 2000 are used.9 A series of robust- ness tests show that the �ndings are largely similar regardless of whether weighted or unweighted regional averages are used. II. EFFECTS OF THE SEX-MIX COMPOSITION OF EXISTING CHILDREN O N F E R T I L I T Y B E H AV I O R This section presents results for the effects of the sex-mix composition of existing children on fertility behavior by region, mothers’ characteristics, mothers’ birth cohort, and implications for gender differences in the number of siblings. Differential Stopping Behavior by Global Region Table 1 presents the results by region. For each region, the 2 þ family size row presents the averages across all family sizes. Although the averages include the results for all family sizes, size-speci�c coef�cients are reported only for family sizes of 2–5 children because the results for higher numbers of chil- dren are very noisy and represent less than 5 percent of the total number of births. 7. Supplemental appendix table S1 presents further descriptive statistics for the study populations including total fertility for women ages 40 and older, the mean son– daughter ratio, the percentage of households without a son, the percentage of households without a daughter, and the ratio of reported “ideal� number of sons to “ideal� number of daughters. 8. When a country has more than one survey, all surveys are pooled and the sampling weights are adjusted so that each survey is equally weighted. For example, surveys were administered in Cambodia in 2000 and 2005. To derive the Cambodia database, data from the two surveys were pooled and the survey weights were adjusted so that each survey contributed half the weighted observations to the analysis. Pooling data across surveys enables increasing the number of observations for each country and therefore increases the precision of the estimates. 9. In other words, if one country has twice the population of another in the same region, it will contribute twice the weighted observations to the analysis. 378 T A B L E 1 . Differential Fertility-stopping Behavior among Women Ages 40 –49 at the Time of the Survey, by Region (Probability of an additional birth as a function of sex-mix composition of existing children) Probability of additional Probability of additional Differential Signi�cance of Mean Mothers’ ideal Region and childbearing after zero childbearing after zero fertility-stopping behavior difference number of ratio of sons to family sizea sons (bm; bmn ) daughters (bf; bfn) (bm – bf; bmn – bfn) (p-value) children daughtersb Latin America and Caribbean 2þ 0.030*** 0.019 0.011 0.541 5.08 0.97 2 0.026*** 0.016 0.009 0.457 3 0.020*** 0.011 0.009 0.211 4 0.041*** 0.048*** 2 0.007 0.724 5 2 0.013** 0.048*** 2 0.061 0.003*** Middle East and North Africa 2þ 0.074*** 0.016** 0.058 0.000*** 6.04 1.13 THE WORLD BANK ECONOMIC REVIEW 2 0.018** 0.014*** 0.004 0.520 3 0.037*** 0.013 0.024 0.033** 4 0.037*** 0.009 0.028 0.065 5 0.056** 0.030* 0.026 0.225 Central Asia 2þ 0.118*** 0.022 0.096 0.000*** 4.14 1.02 2 0.089*** 0.032*** 0.057 0.039** 3 0.122*** 0.011*** 0.110 0.001*** 4 0.166*** 0.060*** 0.106 0.004*** 5 0.168*** 0.032 0.136 0.002*** South Asia 2þ 0.107*** 0.029*** 0.078 0.000*** 4.94 1.37 2 0.054*** 2 0.007** 0.060 0.010*** 3 0.107*** 0.012 0.095 0.062 4 0.137*** 0.020*** 0.116 0.034** 5 0.142*** 0.047*** 0.095 0.010** Southeast Asia 2þ 0.052*** 0.015 0.037 0.040** 4.74 1.01 2 0.035** 0.016*** 0.019 0.354 3 0.031 0.042*** 2 0.011 0.785 4 0.068 0.020** 0.048 0.341 5 0.099** 0.047*** 0.053 0.317 Sub-Saharan Africa 2þ 0.024*** 0.024*** 0.000 0.982 6.63 1.08 2 0.005** 0.002 0.003 0.543 3 0.012 2 0.005 0.017 0.005*** 4 0.021*** 0.010 0.011 0.276 5 0.004 0.010 2 0.006 0.740 **Signi�cant at the 5 percent level; ***signi�cant at the 1 percent level. Note: Table reports the estimated probability of an additional birth as a function of having no boys and no girls. Models are estimated at the region level and include country dummy variables. The sample is limited to women ages 40– 49, who are most likely to have completed their fertility. a. Family size 2 þ estimates are weighted averages for family sizes of two or more children (see text for details). b. As reported by mothers to survey enumerators, who routinely ask mothers for their “ideal� number of children, separately for boys and girls. The ratio is the mean desired number of boys divided by the mean desired number of girls. Source: Authors’ analysis of DHS data shown in the appendix. Filmer, Friedman, and Schady 379 380 THE WORLD BANK ECONOMIC REVIEW F I G U R E 1. Differential Fertility-stopping Behavior by Region and Parity (Five-year Moving Averages) Source: Authors’ analysis of DHS data shown in the appendix. The results show clear evidence that many families in all regions in the devel- oping world prefer a mixed-sex composition of children. All the regional averages of bm and bf are positive, and many are signi�cant: relative to families with both boys and girls, who are the omitted category in the regressions, families with only boys or only girls are more likely to have another birth. In addition, the results shows a son-preferred differential fertility-stopping behavior in many regions in the developing world (see table 1, columns 3 and 4). The largest effects are found for Central Asia, where families are 9.6 percen- tage points more likely to have an additional child if they have had no sons than if they have had no daughters, and South Asia, where the corresponding difference is 7.8 percentage points. Signi�cant, but smaller degrees of son- preferred differential fertility-stopping behavior are apparent in the Middle East and North Africa (5.8 percentage points) and in Southeast Asia (3.7 percentage points). There is no clear evidence of a son-preferred differential in fertility- stopping behavior for either Sub-Saharan Africa or Latin America and the Caribbean.10 Because it is dif�cult to take in all of the coef�cients at a glance, the parity- speci�c results shown in table 1 are summarized in �gure 1. Son-preferred differential fertility-stopping behavior appears to grow with the number of chil- dren in the two regions where it is most pronounced, Central Asia and South Asia. For example, families in South Asia who have already had four or �ve 10. Country-speci�c analyses were also conducted. In the two regions with the clearest evidence of son-preferred differential fertility-stopping behavior (Central Asia and South Asia), these results hold equally for almost all countries in the regions (see supplemental appendix table S2). For the other regions, there is more variability in the country-level results. Filmer, Friedman, and Schady 381 children are approximately 14 percentage points more likely to have an additional child if all of their children have been girls rather than boys. This increase in differential fertility-stopping behavior by number of children is perhaps not surprising: the mean number of children is 4.1 in Central Asia and 4.9 in South Asia. Since the average family expects to have a reasonably large number of children, the sex of children in families with fewer children does not matter as much in determining future fertility because parents expect to have more children, regardless of the sex of their children at the time. In families with more children, however, parents are closer to achieving their total desired number of children, and hence the sex-mix composition of children already born becomes an important determinant of future childbearing. Such patterns are less apparent in the Middle East and North Africa, Southeast Asia, and Latin America, in line with either the smaller degree of son-preferred differential fertility-stopping behavior or the absence of such preference in these regions.11 In addition to identifying differences across cohorts in these basic patterns, table 1 is informative about the extent to which the “ideal� balance between the number of boys and girls reported by mothers is a good indication of fertility behavior. This can be seen by comparing columns 3 and 6 of table 1. A clear subjective preference for sons is apparent in South Asia and Middle East and North Africa, as is a clear behavioral preference for sons with regard to the decision to continue child bearing. However, another region that exhibits a sig- ni�cant pattern of son-preferred differential fertility-stopping behavior, Central Asia, reports a subjective preference for a near equality of sons and daughters. In contrast, mothers in Sub-Saharan Africa report a subjective preference for sons, but families do not exhibit son preference in actual fertility behavior.12 In Latin America and the Caribbean, mothers express a slight preference for daughters, 11. Given the preferred parameterization—binary controls for “no sons� and “no daughters�— aggregating results for family sizes of one child with those of family sizes of two or more children would create an inconsistency. With a family size of one child, the model can include only one dummy variable (either “no sons� or “no daughters�). The two models would need to be estimated separately, and the coef�cients on the two variables would merely be transformations of one another. The excluded category in these models would be a family with one son or one daughter. This is unlike the main estimations, where families with children of at least one of each sex serve as the excluded group. The interpretation is therefore slightly different, and so families with only one child are not included in the analysis. A related model was estimated, however, that investigates the probability of an additional birth, controlling for the sex of the �rst child. Supplemental appendix table S3 reports these results, which also show son-preferred differential fertility-stopping behavior in South Asia even for decisions after the �rst child. However, the analysis shows that families in Latin America are signi�cantly more likely to stop child bearing after the �rst birth if that birth is a daughter rather than a son. 12. The lack of observed differential fertility-stopping behavior in Sub-Saharan Africa could be due to several factors, but one important factor is surely the high level of fertility. Completed fertility in Sub-Saharan Africa is by far the highest and the proportion of households with children of only one sex the lowest across all regions. However, supplemental appendix table S1 also suggests that there is wide variation within Sub-Saharan Africa in the ratio of “ideal� number of sons to “ideal� number of daughters. Therefore, to the extent that reported “ideal� ratio reflects latent sex preference in family composition, Sub-Saharan Africa is not a uniformly son-preferring region, unlike, say, South Asia. 382 THE WORLD BANK ECONOMIC REVIEW but actual fertility behavior exhibits no distinct pattern. Clearly, subjectively stated preferences over the sex-mix composition of children more accurately predict actual fertility behavior in some regions than in others.13 Table 2 presents a series of robustness tests to these basic �ndings, focusing on the aggregate effects averaged across all family sizes (number of children). The �rst panel uses the number of women ages 40 –49 as the weight for aggregating across countries within regions rather than the total population of a country. These weights are generated using data on the share of women ages 40–49 and applying these estimates to estimates of the total female popu- lation.14 The stability of the results to this alternative approach to weighting is apparent. The only major difference between this �rst panel and table 1 is that the slight son-preferred differential fertility-stopping behavior found in East Asia is no longer statistically signi�cant. The results are similar if instead of giving greater weight to countries with larger populations, only the expansion factors in the surveys are used (see table 2, second panel). The only difference is that now son-preferred differen- tial fertility-stopping behavior is slightly muted in South Asia—a difference between bm and bf of 4.6 percentage points compared with 7.8 percentage points in table 1. The results are still similar if even these survey weights are disregarded, so that each sample observation in each region is given the same weight (third panel). If anything, these results suggest an even greater degree of son-preferred differential fertility-stopping behavior in Central Asia and South Asia than do the results in table 1. Moreover, �nally, son-preferred differential fertility-stopping behavior continues to be apparent in the three regions where it is most pronounced in table 1—Middle East and North Africa, Central Asia, and South Asia—when all women ages 15 –49 at the time of the survey are included, not just women who are most likely to have completed their fertility (fourth panel).15 Differential Fertility-stopping Behavior by Mothers’ Characteristics This section investigates how the strong son-preferred differential fertility-stopping behavior exhibited in some regions varies across common 13. Supplemental appendix table S4 reports the alternative speci�cation mentioned earlier that pools the parity-speci�c data and estimates differential fertility-stopping behavior as a function of the ratio of sons to total number of children, controlling for family size. Similar to table 1 in this article, this analysis �nds signi�cant son-preferred differential fertility-stopping behavior in the Middle East and North Africa, Central Asia, and South Asia, suggesting that the article’s main �ndings are robust to this alternative measure of differential fertility-stopping behavior. The son-preferred differential fertility-stopping behavior estimates in these three regions actually grow in magnitude when select mothers’ observables such as location, education, and age are also controlled for. These results with covariates are presented in the second panel of Supplemental appendix table S4. 14. Both statistical constructs are from a World Bank database accessed at: http://go.worldbank.org/ N2N84RDV00. 15. Of course, since this panel includes all women, not just those who have completed their fertility, the total number of children is lower in all regions. T A B L E 2 . Differential Fertility-stopping Behavior among Women at the Time of the Survey, with Different Weights, by Region (Probability of an additional birth as a function of sex-mix composition of existing children) Probability of Probability of additional Differential Signi�cance of Mean Mothers’ ideal additional childbearing childbearing after zero fertility-stopping difference number of ratio of sons to Region after zero sons (bm) daughters (bf ) behavior (bm – bf ) (p-value) children daughtersa Women ages 40– 49, population of women ages 40 – 49 adjusted weights Latin America 0.030*** 0.020 0.011 0.545 5.01 0.97 and Caribbean Middle East and 0.076*** 0.016** 0.061 0.000*** 5.99 1.13 North Africa Central Asia 0.120*** 0.023 0.097 0.000*** 4.07 1.02 South Asia 0.109*** 0.028*** 0.081 0.000*** 4.89 1.37 Southeast Asia 0.051*** 0.021 0.030 0.115 4.74 1.01 Sub-Saharan 0.023** 0.024*** 2 0.001 0.925 6.52 1.08 Africa Women ages 40– 49, population-unadjusted weights Latin America 0.018 0.018 0.000 0.984 5.31 0.93 and Caribbean Middle East and 0.072*** 0.016** 0.057 0.000*** 6.46 1.10 North Africa Central Asia 0.133*** 0.049*** 0.084 0.001*** 3.77 1.03 South Asia 0.080*** 0.034*** 0.046 0.001*** 5.45 1.41 Southeast Asia 0.055*** 0.017 0.038 0.048** 4.84 0.99 Sub-Saharan 0.032*** 0.017** 0.015 0.165 6.62 1.04 Africa Women ages 40– 49, no weights Latin America 0.031*** 0.031*** 0.000 0.977 5.17 0.92 and Caribbean Middle East and 0.075*** 0.013*** 0.061 0.000*** 5.82 1.15 Filmer, Friedman, and Schady North Africa (Continued ) 383 384 TABLE 2. Continued Probability of Probability of additional Differential Signi�cance of Mean Mothers’ ideal additional childbearing childbearing after zero fertility-stopping difference number of ratio of sons to Region after zero sons (bm) daughters (bf ) behavior (bm – bf ) (p-value) children daughtersa Central Asia 0.150*** 0.017 0.133 0.000*** 3.77 1.05 South Asia 0.119*** 0.025*** 0.094 0.000*** 4.67 1.34 Southeast Asia 0.044*** 0.020*** 0.024 0.020** 4.95 0.99 Sub-Saharan 0.025*** 0.019*** 0.006 0.482 6.73 1.06 Africa Full sample of women, population-adjusted weights Latin America 0.042*** 0.026*** 0.016 0.134 5.08 0.95 and Caribbean THE WORLD BANK ECONOMIC REVIEW Middle East and 0.063*** 0.020*** 0.043 0.000*** 6.04 1.12 North Africa Central Asia 0.124*** 0.037*** 0.087 0.000*** 4.14 1.03 South Asia 0.102*** 0.013*** 0.089 0.000*** 4.94 1.35 Southeast Asia 0.046*** 0.023*** 0.023 0.100 4.74 1.01 Sub-Saharan 0.018*** 0.021*** 2 0.003 0.609 6.63 1.09 Africa **Signi�cant at the 5 percent level; ***signi�cant at the 1 percent level. Note: Table reports the estimated probability of an additional birth as a function of having no boys and no girls. Models are estimated at the region level and include country dummy variables. Estimates are for families with three or more children (see text for details). a. As reported by mothers to survey enumerators, who routinely ask mothers for their “ideal� number of children, separately for boys and girls. The ratio is the mean desired number of boys divided by the mean desired number of girls. Source: Authors’ analysis of DHS data shown in the appendix. Filmer, Friedman, and Schady 385 measures of “modernization�—rural–urban location, education, and wealth. Although results are reported for all regions, the discussion focuses on Central Asia and South Asia, where the aggregate results show the greatest son- preferred differential fertility-stopping behavior. The patterns are somewhat different in the two regions. In both South Asia and Central Asia, there is son-preferred differential fertility-stopping behavior in both urban and rural regions, among more and less educated women, and among both households with more and those with less wealth (table 3, columns 3 and 7). However, the difference-in-difference results suggest that in South Asia son-preferred differential fertility-stopping behavior is higher in urban than in rural areas (although not signi�cantly so), among women with more education levels than those with less, and in households with more wealth than in those with less. Some of the differences are quite large: For example, women with six or more years of schooling are 19 percentage points more likely to have an additional child if they do not have boys than if they do not have girls (column 3), while women with less than six years of schooling are only 7 percentage points more likely to do so (column 7).16 In Central Asia, the picture is more mixed: Son-preferred differential fertility-stopping behavior is also higher in urban than in rural areas, but higher among women with low levels of education than among those who have completed at least primary school. Further, there is no signi�cant difference among households in Central Asia at different wealth levels. Many express the belief that as societies and economies develop, the tra- ditional social practices that may enforce or perpetuate a preference for sons weaken. This could happen, for example, if women gain greater autonomy and control a greater share of the household’s economic resources (see, for example, the discussions in Haddad, Hoddinot, and Alderman 1997). Under this assumption, greater son-preferred differential fertility-stopping behavior might be expected in rural than in urban areas, among women with less edu- cation, and among poorer women. The results here do not support that, however, either overall or for regions in which son preference is most pro- nounced (see table 3). This is consistent with earlier �ndings of greater male preference in Indian households with more educated household heads (Behrman 1988). Differential Fertility-stopping Behavior over Time To examine changes across birth cohorts, differential fertility-stopping behav- ior is calculated for each regional cohort cell, as described above. The results 16. Women who are educated or live in urban areas potentially have greater access to technologies that allow them to select the sex of a child. This might affect a small number of the women in the sample (those in the latest cohorts in some countries). However, the effect on estimated differential fertility-stopping behavior is not clear since differential fertility-stopping behavior is by de�nition a behavior conditional on the existing sex mix of children, regardless of whether that mix arose through natural means or with the assistance of sex-selective technology. 386 T A B L E 3 . Differential Fertility-stopping Behavior by Select Mother or Household Characteristics for Women Ages 40–49, by Region (Probability of an additional birth as a function of sex-mix composition of existing children) Probability Probability Probability of additional Probability of additional of additional childbearing Mean of additional childbearing Mean childbearing after zero Differential number childbearing after zero Differential number after zero daughters fertility-stopping of after zero daughters fertility-stopping of Difference-in-difference Region sons (bm) (bf ) behavior (bm – bf ) children sons (bm) (bf ) behavior (bm –bf ) children (column 3 –column 7) Urban Rural Difference Latin America and Caribbean 0.041*** 0.049*** 2 0.009 4.46 0.044** 2 0.011 0.055 6.05 2 0.064 Middle East and North Africa 0.048*** 0.009 0.039*** 5.08 0.076*** 0.019 0.057*** 6.94 2 0.018 Central Asia 0.125*** 0.033** 0.091*** 3.55 0.098*** 0.036** 0.063*** 5.07 0.028 South Asia 0.137*** 0.032*** 0.105*** 4.27 0.098*** 0.026*** 0.072*** 5.22 0.033 Southeast Asia 0.077*** 0.023** 0.054*** 4.29 0.042** 0.013 0.029 4.94 0.025 Sub-Saharan Africa 0.041*** 0.030** 0.012 5.55 0.019** 0.023** 2 0.004 7.05 0.016 Six or more years of schooling Less than six years of schooling Difference Latin America and Caribbean 2 0.003 0.063*** 2 0.066*** 3.46 0.031*** 0.006 0.025 5.91 2 0.090** THE WORLD BANK ECONOMIC REVIEW Middle East and North Africa 0.109*** 0.044*** 0.064*** 3.78 0.074*** 0.011 0.062*** 6.57 0.002 Central Asia 0.107*** 0.046*** 0.061*** 3.64 0.136*** 2 0.001 0.137*** 4.65 2 0.076** South Asia 0.198*** 0.004 0.193*** 3.32 0.094*** 0.029*** 0.066*** 5.35 0.128** Southeast Asia 0.062*** 0.020 0.042** 4.20 0.049*** 0.023 0.026 5.19 0.017 Sub-Saharan Africa 0.047*** 2 0.007 0.054** 5.10 0.019 0.027*** 2 0.008 7.05 0.062** Above-median-wealth households a Below-median-wealth households a Difference Latin America and Caribbean 0.020 0.043** 2 0.023 3.55 0.056*** 0.053*** 0.003 5.07 2 0.026 Middle East and North Africa 0.042*** 0.037*** 0.005 5.17 0.040** 0.008 0.032 6.55 2 0.027 Central Asia 0.119*** 0.028 0.091*** 3.66 0.116*** 0.027 0.089*** 4.67 0.002 South Asia 0.144*** 0.028*** 0.116*** 4.43 0.086*** 0.026** 0.060*** 5.54 0.056** Southeast Asia 0.079*** 0.036*** 0.043 4.23 0.042** 2 0.003 0.045** 4.98 2 0.002 Sub-Saharan Africa 0.033*** 0.008 0.025 6.31 0.026** 0.019 0.007 6.62 0.018 **Signi�cant at the 5 percent level; ***signi�cant at the 1 percent level. Note: Table reports the estimated probability of an additional birth as a function of having no boys and no girls. Models are estimated at the region level and include country dummy variables. Estimates are for families with two or more children (see text for details). a. The analysis by household wealth is based on a composite measure of household durable goods, with households categorized as above or below the median of a composite measure of assets. Source: Authors’ analysis of DHS data shown in the appendix. Filmer, Friedman, and Schady 387 F I G U R E 2. Differential Fertility-stopping Behavior by Region and Mother’s Year of Birth (Five-year Moving Averages) Source: Authors’ analysis of DHS data shown in the appendix. are summarized in �gure 2, which shows the �ve-year moving average of differential fertility-stopping behavior by region. In most regions, there is no systematic pattern. In South Asia, however, son-preferred differential fertility- stopping behavior increases across birth cohorts and is almost 15 percentage points higher for the latest birth cohorts than for the earliest ones. The other region with a high degree of son preference, Central Asia, shows an initial increase in son-preferred differential fertility-stopping behavior, followed by a decrease, although the absolute levels remain high throughout. To test whether these changes across birth cohorts are signi�cant, differen- tial fertility-stopping behavior is �rst regressed on a linear cohort trend, separ- ately by region. Each observation is weighted by the number of women in that cohort-year cell, which gives greater weight to the more precisely calculated cell averages. The coef�cient on the cohort trend in this regression for South Asia is highly signi�cant (0.007, with a standard error of 0.002), which suggests that son-preferred differential fertility-stopping behavior has been increasing by about 0.7 percentage points with each successive cohort. The corresponding coef�cient for Southeast Asia is also signi�cant (0.005, with a standard error of 0.002). None of the other coef�cients is close to standard levels of signi�cance. There are two potential problems with �gure 2 and the corresponding regression analysis. The �rst is that a linear cohort trend may not do justice to the data; this is particularly apparent for Central Asia, with its inverted U-shaped pattern. To address this concern, differential fertility-stopping behav- ior is regressed on �ve-year birth cohort dummy variables, again separately by region. The results—the regression analog of the pattern observed in �gure 2— again show the clearest pattern for South Asia, where son-preferred differential 388 THE WORLD BANK ECONOMIC REVIEW fertility-stopping behavior rises monotonically across �ve-year birth cohorts (table 4). The increase is 10-fold, from 0.017 for the cohort born in 1941–45, to 0.170 for the cohort born in 1961–65. The second, more dif�cult problem is that the regional averages for different birth cohorts may be driven by different countries, depending on the years in which they conducted the DHS. For example, the data from Sri Lanka, where the only DHS was carried out in 1987, enters the average for South Asia for the early birth cohorts but not for the later ones, while the data for Nepal, where DHS were carried out in 1996, 2001, and 2006, enters the regional averages for the later birth cohorts, but not the earlier ones. To address this concern, the sample was limited to countries with a DHS both in 1995 or earlier and in 2000 or later. This greatly reduces the number of countries, from 65 to 27. However, cohort-speci�c measures of son-preferred differential fertility- stopping behavior can be calculated for these countries for women born in every year between 1945 and 1960, and thus regional averages can be calcu- lated that keep the weights �xed for each country across birth cohorts. (The sample is limited to women ages 40 and older, as before.) When both the sample of countries and the weight of each country in the regional average are kept �xed, son-preferred differential fertility-stopping behavior still increases across birth cohorts in South Asia, although the pattern is less dramatic and the difference across cohorts is no longer signi�cant (see table 4, bottom panel). In other regions, the patterns are less clear and are gen- erally not signi�cant. What is clear is that there is no decline in son-preferred differential fertility-stopping behavior in any region where it exists for yet another standard measure of modernization—the passage of time. A S I M P L E M U L T I VA R I A T E F R A M E W O R K The sociodemographic characteristics explored in table 3—mother’s education, urban location, and household wealth—are likely correlated with each other. Thus, it is possible that the association between son-preferred differential fertility-stopping behavior and each of these characteristics is really driven by one main social indicator. Furthermore, prevailing fertility levels may have an effect on differential fertility-stopping behavior since in a high-fertility environ- ment fewer families face differential stopping decisions because of the greater likelihood of mixed-sex composition at larger family sizes. This section thus uses the aggregated location–education–cohort cell data described earlier to estimate the multivariate framework given by equation (4). In bivariate regressions, urban residence and higher educational attainment are both associated with higher differential fertility-stopping behavior, although not signi�cantly so (table 5, columns 1 and 2). These results are con- sistent with those in table 3. In addition, however, there is a signi�cant negative association between the average number of children and differential fertility-stopping behavior (column 3)—the point estimate implies that Filmer, Friedman, and Schady 389 T A B L E 4 . Differential Fertility-stopping Behavior Regressed on Region Interacted with Five-year Cohorts of Mother Birth Year, for Women Ages 40 –49, by Region F-testb Mothers’ birth Region-cohort All interactions First and last Region year cohort interactiona equal equal All countries for cohorts 1941– 65 Latin America and 1941– 45 2 0.004 0.784 0.904 Caribbean 1946– 50 0.013 1951– 55 2 0.009 1956– 60 0.025 1961– 65 0.001 Middle East and 1941– 45 0.062 0.851 0.733 North Africa 1946– 50 0.055 1951– 55 0.031 1956– 60 0.010 1961– 64 0.040 Central Asia 1946– 50 0.017 0.412 0.403 1951– 55 0.085** 1956– 60 0.141*** 1961– 65 0.094 South Asia 1941– 45 0.017 0.001*** 0.000*** 1946– 50 0.067*** 1951– 55 0.078*** 1956– 60 0.120*** 1961– 65 0.170*** Southeast Asia 1941– 45 0.024 0.027** 0.874 1946– 50 0.002 1951– 55 0.013 1956– 60 0.108*** 1961– 63 0.033 Sub-Saharan Africa 1941– 45 2 0.001 0.025** 0.895 1946– 50 0.000 1951– 55 0.034 1956– 60 2 0.047*** 1961– 65 2 0.006 Countries with differential fertility-stopping behavior for cohorts 1946– 60 c Latin America and 1946– 50 0.020 0.410 0.491 Caribbean 1951– 55 2 0.020 1956– 60 0.000 Middle East and 1946– 50 0.050 0.593 0.311 North Africa 1951– 55 0.024 1956– 60 0.010 Central Asia 1946– 50 0.084 0.710 0.456 1951– 55 0.147*** (Continued ) 390 THE WORLD BANK ECONOMIC REVIEW TABLE 4. Continued F-testb Mothers’ birth Region-cohort All interactions First and last Region year cohort interactiona equal equal 1956– 60 0.148*** South Asia 1946– 50 0.093*** 0.219 0.275 1951– 55 0.080*** 1956– 60 0.120*** Southeast Asia 1946– 50 0.007 0.124 0.615 1951– 55 2 0.038 1956– 60 0.024 Sub-Saharan Africa 1946– 50 0.018 0.042** 0.037** 1951– 55 0.016 1956– 60 2 0.035** **Signi�cant at the 5 percent level; ***signi�cant at the 1 percent level. a. The results in this column are the coef�cients of the interaction terms. b. The F-tests are region speci�c. The results are the p-values for the F-tests. Data are weighted by sample size. c. Countries include Bangladesh, Bolivia, Burkina Faso, Cameroon, Colombia, Co ˆ te d’Ivoire, Dominican Republic, Egypt, Ghana, Haiti, India, Indonesia, Kenya, Madagascar, Malawi, Mali, Morocco, Namibia, Niger, Nigeria, Peru, Philippines, Rwanda, Senegal, Tanzania, Turkey, Uganda, Zambia, and Zimbabwe. Source: Authors’ analysis of DHS data shown in the appendix. a decrease in average family size of one child more than offsets a switch from rural to urban location and almost offsets a switch from low to high schooling levels. The key results include the measures of location, education, and the mean number of children for each country, year, location, and education cell (see table 5, columns 4 and 5). Once the average number of children is included in the model, the association between son-preferred differential fertility-stopping behavior and urban residence and between differential fertility-stopping behav- ior and education becomes negative (column 4). This reverses the bivariate �ndings and suggests that the higher son-preferred differential fertility-stopping behavior in urban areas and among more educated mothers can be “explained� by differences in overall fertility levels.17 Including global dummy variables for each birth year, as a way of flexibly controlling for any secular changes, barely affects the results for these three indicators (column 5). In sum, the cell-level results suggest that the number of children women expect to have over their lifetimes is an important determinant of son-preferred differential fertility-stopping behavior. When fertility levels are high, the 17. This �nding is in character with Das Gupta and Mari Bhat (1997), who argue that fertility decline may lead to an intensi�cation of discrimination against girls if the total number of children that couples desire falls more rapidly than the total number of desired sons. Filmer, Friedman, and Schady 391 T A B L E 5 . Multivariate Correlates of Differential Fertility-stopping Behavior Regression Variable (1) (2) (3) (4) (5) Urban 0.014 2 0.023** 2 0.021** (0.010) (0.010) (0.010) Six or more years of schooling 0.027 2 0.026*** 2 0.022*** (0.020) (0.009) (0.009) Mean number of children 2 0.021* 2 0.029** 2 0.027** (0.011) (0.013) (0.012) Birth year dummy variables No No No No Yes Number of observations 3,456 3,456 3,456 3,456 3,456 R-squared 0.00 0.01 0.04 0.05 0.06 *Signi�cant at the 10 percent level; **signi�cant at the 5 percent level; ***signi�cant at the 1 percent level. Note: Numbers in parentheses are robust standard errors. Each observation is a country, urban – rural, high– low education, year of birth cell. Data are weighted by sample size and country population in 2000. Source: Authors’ analysis of DHS data shown in the appendix. absence of boys in earlier births is not an important driver of childbearing decisions—at all but the largest family size, most couples expect to have more children, no matter what the sex-mix composition of earlier births. However, as family size decreases, a higher fraction of couples �nd themselves having to choose whether to have an additional child at a point when they are already close to their expected family size and all their children are of the same sex. At this point, the sex-mix composition of their children—in particular, whether there is at least one boy—appears to play an important role in their decision. Sex Differences in Number of Siblings If families are more likely to have an additional child when they have no sons than when they have no daughters, girls may grow up in households with more siblings than do boys. Of course, the number of siblings that boys or girls have will also be determined by mortality—which may vary with family size and by a child’s sex. The mean number of siblings for girls and boys ages 0 –15 years is higher for girls than for boys in regions where there is son-preferred differential fertility-stopping behavior (table 6). For example, in South Asia girls have about 0.13 more siblings than boys, on average; in Central Asia, the compar- able number is 0.10. In contrast, in Sub-Saharan Africa, boys and girls have the same number of siblings on average. Moreover, if girls are discriminated against relative to boys after birth in regions where there is son-preferred differ- ential fertility-stopping behavior, like South Asia and Central Asia, and 392 THE WORLD BANK ECONOMIC REVIEW T A B L E 6 . Mean Number of Siblings of Children ages 0–15 Children of women ages 40 and older All children Sons– Sons – Region Sons Daughters daughters Sons Daughters daughters Latin America and 4.99 5.06 2 0.07*** 3.08 3.14 2 0.06*** Caribbean Middle East and North 5.27 5.29 2 0.02 3.67 3.73 2 0.06*** Africa Central Asia 4.27 4.37 2 0.10** 2.63 2.77 2 0.14*** South Asia 4.59 4.72 2 0.13*** 2.81 2.96 2 0.15*** Southeast Asia 4.46 4.52 2 0.07*** 2.82 2.86 2 0.04*** Sub-Saharan Africa 5.49 5.49 0.01 3.55 3.56 2 0.01** **Signi�cant at the 5 percent level; ***signi�cant at the 1 percent level. Source: Authors’ analysis of DHS data shown in the appendix. therefore suffer excess mortality,18 these results would generally underestimate the differences in sibship size by sex that result from son-preferred differential fertility-stopping behavior. An extensive literature documents associations between larger family size and poorer outcomes for children in developed and developing countries (see, for example, Behrman and Wolfe 1986; Horton 1986; Conley and Glauber 2006, and the references therein). Having more siblings dilutes household and parental resources and may result in quantity–quality tradeoffs. Estimating the causal effect of the number of siblings on child outcomes is dif�cult, however, because of the likelihood of omitted family characteristics that may bias results. Nevertheless, insofar as some of the association between the number of children and poor outcomes is causal, it suggests that son preference, as mani- fested in sex-speci�c differential fertility-stopping behavior, may have adverse implications on the outcomes for girls, who will tend to grow up in larger families. Moreover, the differences in family size by children’s sex are largest in regions where girls are more likely to suffer discrimination in other ways, in particular in South Asia (see table 6). III. CONCLUSION This article has investigated the fertility response to the sex-mix composition of children in a family using data from 158 DHS carried out in 64 countries. Sex composition of earlier births is a signi�cant determinant of subsequent fertility in many developing countries. Fertility behavior is consistent with son prefer- ence in many regions of the developing world, with the clearest patterns appar- ent in South Asia and Central Asia. Speci�cally, the absence of sons increases 18. On India, see, for example, Das Gupta (1987), Behrman and Deolalikar (1990), and Rose (1999). Filmer, Friedman, and Schady 393 the probability of an additional birth by signi�cantly more than the absence of daughters. This phenomenon is referred to as son-preferred differential fertility-stopping behavior. Exploration of heterogeneity shows that widely used measures of “moderniz- ation,� including urbanization, higher education levels, and household wealth, are associated with an increase in son-preference, as captured in differential fertility-stopping behavior. The presumption that this manifestation of son pre- ference will dissipate over time is also not supported by the data. The results from regressions using a simple multivariate framework suggest that this may be a result of reductions in family size with increased modernization. While it is possible that greater urbanization, female education, and household wealth all reduce a latent son preference, the reductions in fertility that accompany modernization also make it more likely that a latent son preference can be detected in behavior. For this reason, social policies that aim to limit fertility may, as an unintended consequence, bring son-preferred differential fertility-stopping behavior to the fore. Finally, one implication of son-preferred differential fertility-stopping behav- ior is that girls tend to have more siblings than boys. This is an important �nding in itself, as it likely has consequences for the development of boys and girls in infancy, childhood, and adolescence. Moreover, insofar as there are quantity–quality tradeoffs that result in fewer material and emotional resources allocated to children in larger families, son preference in fertility decisions can have important indirect implications for investments and for the well-being of girls relative to boys. S U P P L E M E N TA RY MAT E R I A L Supplemental appendix to this article is available at http://wber.oxfordjournals. org/. APPENDIX: SAMPLE COUNTRIES, SURVEYS, AND NUMBER OF MOTHERS AND BIRTHS Number of mothers Number of Country Region Year of survey observed births observed Armenia Central Asiaa 2000, 2005 8,648 21,583 Bangladesh South Asia 1993– 94, 1996– 97, 36,169 127,486 1999– 2000, 2004 Benin Sub-Saharan 1996, 2001, 2006 22,688 95,989 Africa (Continued ) 394 THE WORLD BANK ECONOMIC REVIEW Continued Number of mothers Number of Country Region Year of survey observed births observed Bolivia Latin America and 1989, 1993– 94, 31,431 121,101 Caribbean 1998, 2003– 04 Brazil Latin America and 1986, 1991– 92, 12,050 37,871 Caribbean 1996 Burkina Faso Sub-Saharan 1992– 93, 1998– 99, 19,168 84,320 Africa 2003 Burundi Sub-Saharan 1987 2,777 11,886 Africa Cambodia Southeast Asiab 2000, 2005 20,721 81,447 Cameroon Sub-Saharan 1991, 1998, 2004 14,243 56,254 Africa Central African Sub-Saharan 1994– 95 4,388 16,936 Republic Africa Chad Sub-Saharan 1996– 97, 2004 10,508 47,187 Africa Colombia Latin America and 1986, 1990, 1995, 50,573 141,967 Caribbean 2000, 2005 Comoros Sub-Saharan 1996 1,695 7,913 Africa Congo, Rep. of Sub-Saharan 2005 5,152 16,687 Africa ˆ te d’Ivoire Co Sub-Saharan 1994, 1998– 99, 11,895 45,803 Africa 2005 Dominican Latin America and 1986, 1991, 1996, 33,677 113,636 Republic Caribbean 1999, 2002 Ecuador Latin America and 1987 3,117 11,835 Caribbean Egypt Middle East and 1988, 1992– 93, 70,394 276,509 North Africa 1995– 96, 2000, 2003, 2005 Ethiopia Sub-Saharan 2000, 2005 19,482 84,055 Africa Gabon Sub-Saharan 2000– 2001 4,499 16,878 Africa Ghana Sub-Saharan 1988, 1993– 94, 14,449 55,788 Africa 1998– 99, 2003 Guatemala Latin America and 1987, 1995, 1998– 16,804 72,032 Caribbean 99 Guinea Sub-Saharan 1999, 2005 11,672 50,058 Africa Haiti Latin America and 1994– 95, 2000, 16,294 63,814 Caribbean 2005 Honduras Latin America and 2005 13,991 50,093 Caribbean India South Asia 1992– 93, 1998– 244,831 800,833 2000, 2005– 06 (Continued ) Filmer, Friedman, and Schady 395 Continued Number of mothers Number of Country Region Year of survey observed births observed Indonesia Southeast Asiab 1987, 1991, 1994, 111,864 370,441 1997, 2002– 03 Kazakhstan Central Asiaa 1995, 1999 6,013 14,972 Kenya Sub-Saharan 1988– 89, 1993, 22,504 94,497 Africa 1998, 2003 Kyrgyzstan Central Asiaa 1997 2,776 8,781 Lesotho Sub-Saharan 2004 4,832 14,708 Africa Liberia Sub-Saharan 1986 4,231 17,264 Africa Madagascar Sub-Saharan 1992, 1997, 2003– 15,447 61,383 Africa 04 Malawi Sub-Saharan 1992, 2000, 2004 23,353 92,634 Africa Mali Sub-Saharan 1987, 1995–96, 21,004 98,580 Africa 2001 Mexico Latin America and 1987 5,776 22,676 Caribbean Morocco Middle East and 1987, 1992, 2003– 18,970 80,669 North Africa 04 Mozambique Sub-Saharan 1997, 2003 16,530 63,195 Africa Namibia Sub-Saharan 1992, 2000 8,490 28,318 Africa Nepal South Asia 1996, 2001, 2006 23,042 84,505 Nicaragua Latin America and 1997– 98, 2001 18,971 70,977 Caribbean Nigeria Sub-Saharan 1990, 1999, 2003 17,209 74,438 Africa Niger Sub-Saharan 1992, 1998, 2006 18,194 87,107 Africa Pakistan South Asia 1990– 91 5,905 27,369 Paraguay Latin America and 1990 3,970 153,46 Caribbean Peru Latin America and 1986, 1991–92, 60,700 217,275 Caribbean 1996, 2000, 2004 Philippines Southeast Asiab 1993, 1998, 2003 26,609 98,932 Rwanda Sub-Saharan 1992, 2000, 2005 17,876 771,14 Africa Senegal Sub-Saharan 1986, 1992–93, 23,525 102,547 Africa 1997, 2005 South Africa Sub-Saharan 1998 8,223 22,934 Africa Sri Lanka South Asia 1987 5,388 17,701 Sudan Sub-Saharan 1989– 90 5,277 25,805 Africa Tanzania Sub-Saharan 1991– 92, 1996, 23,504 96,542 Africa 1999, 2004 (Continued ) 396 THE WORLD BANK ECONOMIC REVIEW Continued Number of mothers Number of Country Region Year of survey observed births observed Thailand Southeast Asiab 1987 6,025 17,803 Togo Sub-Saharan 1988, 1998 8,825 37,051 Africa Trinidad and Latin America and 1987 2,440 7,837 Tobago Caribbean Tunisia Middle East and 1988 3,856 16,463 North Africa Turkey Central Asiaa 1993, 1998, 2003 18,861 59,996 Uganda Sub-Saharan 1988– 89, 1995, 20,946 92,326 Africa 2000– 2001, 2006 Uzbekistan Central Asiab 1996 3,018 96,50 Vietnam Southeast Asiab 1997, 2002 10,742 29,900 Yemen Middle East and 1991– 92 5,378 29,803 North Africa Zambia Sub-Saharan 1992, 1996– 97, 17,013 70,726 Africa 2001– 02 Zimbabwe Sub-Saharan 1988– 89, 1994, 17,881 62,855 Africa 1999, 2005– 06 64 countries 6 regions 158 surveys 1,336,484 4,931,081 a. None of the countries observed in this region is in the part of the region traditionally referred to as Eastern Europe, and so this region is referred to in the analysis as Central Asia only. b. None of the countries observed in this region is in the part of the region traditionally referred to as the Paci�c or in the Northeastern region of Asia, and so this region is referred to in the analysis as Southeast Asia only. REFERENCES Andersson, G., H. Karsten, M. Rønson, and A. Vikat. 2006. “Gendering Family Composition: Sex Preferences for Children and Childbearing Behavior in the Nordic Countries.� Demography 42(2):255–67. Arnold, F. 1985. “Measuring the Effect of Sex Preference on Fertility: The Case of Korea.� Demography 22(2):280 –88. ———. 1992. “Sex Preference and Its Demographic and Health Implications.� International Family Planning Perspectives 18(3):93–101. ———. 1997. Gender Preferences for Children. DHS Comparative Studies 23. Calverton, Md.: Macro International. Arnold, F., M. K. Choe, and T. K. Roy. 1998. “Son Preference, the Family-building Process, and Child Mortality in India.� Population Studies 52(3):301–15. Bairagi, R. 1987. “A Comment on Fred Arnold’s “Measuring the Effect of Sex Preference on Fertility.� Demography 24(1):137 –38. Behrman, J. R. 1988. “Intrahousehold Allocation of Nutrients in India: Are Boys Favored? Do Parents Exhibit Inequality Aversion?� Oxford Economic Papers 40(1):32–54. Behrman, J. R., and A. B. Deolalikar. 1990. “The Intrahousehold Demand for Nutrients in Rural South India: Individual Estimates, Fixed Effects, and Permanent Income.� The Journal of Human Resources 25(4):665– 96. Filmer, Friedman, and Schady 397 Behrman, J. R., and B. L. Wolfe. 1986. “Child Quantity and Quality in a Developing Country: Family Background, Endogenous Tastes, and Biological Supply Factors.� Economic Development and Cultural Change 34(4):703 – 20. Chung, W., and M. Das Gupta. 2007. “The Decline of Son Preference in South Korea: The Roles of Development and Public Policy.� Population and Development Review 33(4):757–83. Conley, D., and R. Glauber. 2006. “Parental Educational Investment and Children’s Academic Risk: Estimates of the Impact of Sibship Size and Birth Order from Exogenous Variation in Fertility.� Journal of Human Resources 41(4):722–37. Das Gupta, M. 1987. “Selective Discrimination against Female Children in Rural Punjab, India.� Population and Development Review 13(1):77– 100. Das Gupta, M., and P. N. Mari Bhat. 1997. “Fertility Decline and Increased Manifestation of Sex Bias in India.� Population Studies 51(4):307–15. ` ze, J., and M. Murthi. 2001. “Fertility, Education, and Development: Evidence from India.� Dre Population and Development Review 27(1):33– 63. Filmer, D. 2005. “Gender and Wealth Disparities in Schooling: Evidence from 44 Countries.� International Journal of Education Research 43(6):351– 69. Filmer, D., and L. Pritchett. 2001. “Estimating Wealth Effects without Income or Expenditure Data—or Tears: Educational Enrollment in India.� Demography 38(1):115–32. Haddad, L., J. Hoddinot, and H. Alderman. 1997. Intrahousehold Resource Allocation in Developing Countries: Models, Methods, and Policy. Baltimore: Johns Hopkins University Press. Hank, K., and H.-P. Kohler. 2000. “Gender Preferences for Children in Europe: Empirical Results from 17 FFS Countries.� Demographic Research 2 (Article1). www.demographic-research.org/Volumes/ Vol2/1/2-1.pdf. Haughton, J., and D. Haughton. 1995. “Son Preference in Vietnam.� Studies in Family Planning 26(6):325– 37. Horton, S. 1986. “Child Nutrition and Family Size in the Philippines.� Journal of Development Economics 23(1):161– 76. Jensen, R. 2007. “Equal Treatment, Unequal Outcomes? Generating Sex Inequality through Fertility Behavior.� Harvard University, John F. Kennedy School of Government, Cambridge, Mass. http:// www.watsoninstitute.org/pub_detail.cfm?id=811. Key�tz, N. 1968. Introduction to the Mathematics of Population. Reading, Mass.: Addison-Wesley. Larsen, U., W. Chung, and M. Das Gupta. 1998 “Fertility and Son Preference in Korea� Population Studies 52(3):317 –25. Leung, S. 1998. “On Tests for Sex Preferences.� Journal of Population Economics 1(2):95–114. Muhiri, P. K., and S. H. Preston. 1991. “Effects of Family Composition on Mortality Differentials by Sex among Children in Matlab, Bangladesh.� Population and Development Review 17(3):415– 34. Pande, R. 2003. “Selective Gender Differences in Childhood Nutrition and Immunization in Rural India: The Role of Siblings.� Demography 40(3):395– 418. Park, C. B. 1983. “Preference for Sons, Family Size, and Sex Ratio: An Empirical Study in Korea.� Demography 20(3):333– 52. Pollit, E., K. S. Gorman, P. Engell, R. Martorell, and J. A. Rivera. 1993. “Early Supplementary Feeding and Cognition: Effects over Two Decades.� Monographs of the Society for Research in Child Development, Serial No. 235, 58(7):1– 99. Pong, S. 1994. “Sex Preference and Fertility in Peninsular Malaysia.� Studies in Family Planning 25(3):137– 48. Repetto, R. 1972. “Son Preference and Fertility Behavior in Developing Countries.� Studies in Family Planning 3(4):70–76. Rose, E. 1999. “Consumption Smoothing and Excess Female Mortality in Rural India.� Review of Economics and Statistics 81(1): 41– 49. 398 THE WORLD BANK ECONOMIC REVIEW Rivera, J. A., R. Martorell, M. T. Ruel, J. P. Habicht, and J. D. Haas. 1995. “Nutritional Supplementation during Pre-school Years Influences Body Size and Composition of Guatemalan Adolescents.� Journal of Nutrition 25(4S):1068S –77S. World Bank. 2001. Engendering Development through Gender Equality in Rights, Resources, and Voice. Washington, D.C.: World Bank and New York: Oxford University Press. Yount, K. M. 2001. “Excess Mortality of Girls in the Middle East in the 1970s and 1980s: Patterns, Correlates, and Gaps in Research.� Population Studies 55(3):291–308. Yount, K. M., R. Langsten, and K. Hill. 2000. “The Effect of Gender Preference on Contraceptive Use and Fertility in Rural Egypt.� Studies in Family Planning 31(4):290–300. Zeng Yi, T. P., G. Baochang, X. Yi, L. Bohua, and L. Yongpiing. 1993. “Causes and Implications of the Recent Increase in the Reported Sex Ratio at Birth in China.� Population and Development Review 19(2):283 – 302.