Fertility and Ethnic Diversity Thorsten Janus1 Abstract This paper relates national fertility rates to political factors. I first review evidence that community norms affect fertility, ethnic groups compete politically in many countries, and a group’s relative size affects its ability to compete. This evidence jointly suggests that population growth and fertility may be public goods at the ethnic group level in diverse societies. Next, I present a simple model of ethnic conflict where a high fertility norm within each ethnic group emerges in a unique equilibrium if ethnic diversity is high and/or institutions are weak. In the opposite case, fertility is low. Finally, I test empirically for a link between ethnic diversity and fertility in a cross-national dataset and find a robust positive relationship. Consistent with more intense conflict where institutions are weaker, ethnic diversity’s effect on fertility decreases with income per capita. On the other hand, there is no evidence of a link between fertility and religious diversity.

Keywords: population; fertility; ethnic diversity; social conflict; social norms JEL: D72, D74, J13, O12

1

University of Wyoming. College of Business, Department of Economics and Finance, Dept. 3985, 1000 E.

University Ave., Laramie, WY 82071, United States. Email: [email protected]. Ph: 1-307-766-3384. Fax: 1-307766-5090.

1

1. Introduction A large literature links demography to welfare. While Malthus’ (1798) prediction that population growth would decrease the food supply per capita was not borne out, there is evidence that dense populations put pressure on local natural resources (Dasgupta 1995), high fertility is harmful to women’s health (World Bank 2007, UNICEF 2009) and youth bulges may promote civil conflict (Urdal 2006). High fertility may also sustain a poverty trap (Nelson 1956, Becker and Barro 1989) or decrease income per capita by raising the dependency ratio (Solow 1956, Bloom et al. 2001). On the other hand, large populations may exploit economies of scale (Simon 1981, Krugman 1992) and advance technology because ideas are non-rival (Kremer 1993, Galor and Weil 2000) or scarcity improves the incentive to innovate (Boserup 1965).

Regarding fertility choice, the literature has mostly focused on household variables like parents’ opportunity cost of time, their intrinsic or pecuniary benefits from children (Becker 1960, Rosenzweig and Stark 1997, Schultz 1997), or the distribution of bargaining power in the household (Iyigun and Walsh 2007, Rasul 2008). However, it is important to note that any fertility-related incentives and constraints facing a household may themselves depend on more macro-level factors. In particular, a large literature supports a link between fertility and community-level social norms (Caldwell and Caldwell 1987, Dasgupta 1993, 1995, Watkins 1996, Kohler 2001). Thus, in a widely used textbook Bardhan and Udry (1999, p. 21) explain:

People’s notions of appropriate behaviour concerning the determinants of fertility are strongly influenced by cultural norms. Ideals concerning the age of marriage, birth outside of marriage, birth spacing, breast feeding, and the use of modern methods of birth control are all strongly conditioned by the behaviour of other members of the community

2

Similarly, in recent empirical work Munshi and Myaux (2006) find that contraception decisions in rural communities in Bangladesh depend strongly on contraceptive prevalence in ownreligious group, while there are no cross-religion effects, and Childs et al. (2005) find that recent fertility transitions among Tibetans living in the Tibet Autonomous Region of China and in exile in South Asia are similar in timing and magnitude despite the very different political, economic and social conditions since the 1960s. Among the reasons for sub-Saharan Africa’s historically high fertility rates compared to other regions may be that social norms induce mothers rather than fathers to bear the cost of raising children, kin groups rather than parents bear the cost, or fertility promotes social status (Udjo 1984, Caldwell and Caldwell 1987).

Despite the recognition that social norms can affect fertility and in turn welfare, however, the literature has not paid much attention to the origin of fertility-related norms. Perhaps the simplest answer is to refer to culture and tradition, which have historically been associated more with sociology and anthropology than economics. Alternatively, in order to achieve consistency between norm-based behavior and individual rational choice theory, some papers have depicted fertility norms as one among multiple Nash equilibriums in a social interaction game between community members (Bardhan and Udry 1999 Chapter 3, Manski and Mayshar 2003, Munshi and Myaux 2006). In such games equilibriums with high fertility, for example, can be enforced by the threat of social sanctions on a deviating household (Bardhan and Udry 1999), and there is no guarantee that the outcome will be Pareto-optimal. However, although these multiple equilibrium models explain why agents conform to some particular fertility norm, they do not explain the origin of the norm itself: why is one particular Nash equilibrium chosen and not its alternatives? In fact, against any claim that culture, tradition or multiple equilibriums can explain inefficient fertility practices, one could reasonably object that inefficient behavior should 3

be weeded out by history or renegotiated by the coalition of all players or a subset thereof. If high fertility were instead a uniquely preferred social norm within a community, then its persistence would seem more plausible.

In this paper I follow previous work linking household fertility to social norms. However, rather than study the role of multiple equilibriums in social interactions in leading to high fertility, I focus on the possible role of political factors. In particular, a large body of evidence supports that ethnic politics and ethnic favoritism in public policy is common across countries (Young 1976, Bates 1981, Horowitz 2000, Chandra 2007, Kimenyi and Romero 2008) and many civil conflicts take place along ethnic lines (Fearon and Laitin 2003).2 During the course of such either peaceful or violent ethnic antagonism, a group’s relative size may determine its relative strength. For example, if political leaders are chosen in elections and agents vote for co-ethnic candidates (Young 1976, Bates 1981), then larger groups are more likely to win office. Similarly, large size may be an advantage during military conflict or if peacetime policies are made under the threat of violence (Collier and Hoeffler 2004). In terms of language and culture, larger groups may offer a larger network and tend to tip all of society in their direction.

Combining the evidence from the literature on fertility and social norms and the evidence from the ethnic conflict literature seems to support the following claims: (a) social norms affect fertility, (b) ethnic groups compete politically, and (c) an ethnic group’s relative political strength may depend on its size. In turn, it is possible that (d) up to a point population growth achieved via fertility is a public good at the ethnic group level. In addition, (e) the value of

2

For example, among the 114 civil conflict onsets between 1945 and 1999 recorded in Fearon and Laitin (2003), 75 are also recorded as ethnic conflict onsets.

4

providing this public good should be particularly high in diverse societies with weak institutions to the extent such places (i) ethnic groups fail to internalize each other’s welfare and (ii) constitutional protection of minorities and constraints on the power of government are weak. In these cases, ethnic competition should tend to have a winner-takes-all nature and therefore the benefits of group size should be greater. Consequently, this paper asks whether (a) ethnic conflict can promote high norms and (b) any effect of ethnic conflict on fertility is stronger where the conflict is more intense due to weak institutions.

In order to address these questions, I first present a simple rent-seeking model of ethnic conflict where a high fertility norm within each ethnic group emerges as the unique equilibrium if ethnic diversity is high and/or institutions weak. Otherwise, if ethnic diversity is low or institutions are strong, the only possible norm is low fertility. Second, I test empirically for a link between ethnic diversity and fertility rates in a cross-national dataset. I find that there is a robust positive relationship between the two and that, consistent with more intense conflict where institutions are weaker (Easterly 2001), ethnic diversity’s effect on fertility decreases with income per capita. On the other hand, I find no evidence of a relationship between fertility and religious diversity and briefly discuss some possible explanations. To the extent ethnic diversity increases fertility above a community or country’s welfare-optimum, achieving a fall in fertility may require not just changing local attitudes or economic incentives, but addressing the underlying conflict problem.

Although the kind of political economy view of high fertility norms advanced in this paper is distinct from the more common microeconomic approach to fertility, the link between fertility

5

and political economy factors in ethnically divided societies has actually been studied in the broader social science literature (ODI 2008). In order to illustrate, it is worth citing several interesting case studies. For example, Obono (2003, p. 109, see also Udjo 1984 and Kokole 1994) explains the politicized role of fertility in Nigeria as follows:

Nigeria's polity is a federal system whose political center is dominated by three major groups, the Hausa, the Yoruba, and the Igbo. There are innumerable smaller groups - in all, some 374 (Otite 1990: 36; Nigeria 2000: 1) to 389 (Otite 2000: 30) ethnic groups are recognized - many with acute sensitivity to their minority status among the larger ethnic entities and aware that their size counts in terms of claims on state revenues. For most, antinatalism would not be rational if the contradictions endemic to the country's political economy are not addressed.

In East Timor (Franks 1996, p. 164) and Guatemala (Shiffman and del Valle 2006, pp. 55, 66), persecuted groups mistrust government health clinics: The Timorese people do not have confidence in the Indonesian-run health centres and doctors who administer health in their country. Rumors abound in East Timor about forced sterilizations and injections used for no other purpose than to control women’s fertility. Hence, the East Timorese avoid Indonesian clinics and doctors if they can...Their [Indonesia’s] inability to stamp out the people’s resistance to their occupation and the prospect of future generations becoming even more militant ensures that women’s fertility is increasingly integral to any strategy of coercion and control.

[One reason for Guatemala’s high fertility rate compared to Honduras is]...the presence of a large marginalized indigenous population in Guatemala that the state has had difficulty reaching and that is suspicious of modern medical practices…targeted by the army during the country's civil war, many Mayans

6

view the state with suspicion and are reluctant to approach any of its institutions, including its health care facilities.

In Apartheid South Africa (Posel 2005, p. 128, see also Klausen 2002): Inherent in the call for 'white supremacy' was a zealous drive to preserve the 'purity' of the white 'race', by preventing the sexual sullying of the white body. Powerful imagery of the black mass, and the ever imminent threat of its overwhelming (oorstroming) the far smaller numbers of whites, fanned fears of black 'overpopulation' and the imperative of controlling black fertility.

In Yugoslavia and Macedonia (Brunnbauer 2004, pp. 575-76): In 1982, the Federal Assembly of Yugoslavia had passed a resolution on common population policy aimed at establishing “humane and rational” levels of reproduction, and leveling the huge differences in birth rates between the various Yugoslav regions..The Macedonian parliament followed with the Rezolucija za populacionata politika (“Resolution on Population Policy”) in 1987, which propagated the four-member family…The subtext of these resolutions was the fear—especially on the part of the Serbian and Macedonian leaderships—that the Albanian populations in their republics would continue to grow faster than the majority.

In India (Panandiker and Umashankar 1994, pp. 96-97, see also Basu 1997 and Smith 2009): In a newspaper interview in 1989, G. M. Shah, former Muslim Chief Minister of Jammu and Kashmir State, underscored the importance of religion for fertility politics in India, particularly in highly polarized non-Hindu states (Shah, 1989). In Shah's words,

The government has hatched a conspiracy to reduce the Kashmiri Muslim population. Farooq Abdullah [then Chief Minister of the Jammu and Kashmir State] is an instrument of this plot. Our State had an 82 percent Muslim population in 1947; it is now a mere 54 percent as the 1981 census figures reveal. We should reject the government's family planning program. This is aimed at further reducing the Muslim

7

population in Kashmir. Every Kashmiri Muslim should have four wives to produce at least one dozen children. (Shah, 1989: 33)

In the Middle East (Faour 1989, p. 261, see also Winckler 2005 Chapter 4): Like the other Arab Gulf states, Oman is striving to increase its indigenous population to safeguard the stability of the political regime and withstand any possible military incursion by the more populous neighboring countries. Since these states have no wish to retain massive imported labor, they promote and reinforce pronatalist norms …the war between Iran and Iraq has made the Arab Gulf states, including Iraq, even more strongly pronatalist. For example, after the eruption of the Gulf war, Iraq began to use incentives to promote higher fertility, and is now considering the use of disincentives as well to discourage fertility reduction…the presence of large numbers of Palestinians in Jordan, while the Arab-Israeli conflict is not resolved, makes fertility reduction there a highly sensitive issue (National Academy of Sciences, 1974). For Lebanon, demographic problems are intimately related to the distribution and maintenance of political power. The numerical sizes of the major religious sects, as conveyed by the 1932 Census, have determined their shares in the political structure since 1943. All attempts to carry out another census have been futile because there is strong evidence of significant sectarian differences in fertility (Chamie, 1981)...

Further examples of politicized fertility may include that, until recently, leaders among Tibetans in exile in South Asia advocated pronatalism to ensure the survival of the Tibetan people (Childs et al. 2005, Childs and Barkin 2008) and Attane and Courbage (2000) explain that minorities in China use fertility to affirm their group identity in relation to the Han majority. Manski and Mayshar (2003) explain that in Israel in the past half-century parts of the ultra-Orthodox Jewish population have experienced rapid and substantial fertility increase due partly to social norms, while other ethnic-religious groups have had decreasing or moderately increasing fertility. Anson and Meir (1996) find that fertility rates across census areas in Israel are correlated with nationalist support and argue that there is a causal effect (see also Nahmias and Stecklov (2007)

8

on Muslim Palestinians in Israel). Brown and Ferree (2005) study coverage of falling fertility, rising immigration and pronatalist appeals in major newspapers in Great Britain 2000-2, and Kunovich and Deitelbaum (2004) argue that Croatia’s ethnic conflict experience after Yugoslavia’s demise caused a resurgence of traditional values, where women were to serve as ‘mother of the nation’. Shiffman et al. (2002) and Birenbaum-Carmeli (2009) study nationalistic government fertility promotion in Croatia, Serbia, and Israel, and Albanese (2004) links the rise of nationalist regimes in 20th Century Germany, Italy, Russia and Yugoslavia/Croatia to declines in women’s reproductive rights. Rubinstein and Lane (2003) discuss how populations are manipulated in ethnic conflict via mass killings, expulsion, rape and sterilization.

In the rest of the paper, Section 2 presents a simple rent-seeking model of ethnic conflict, where investment and fertility norms within each ethnic group are endogenously determined. A high fertility norm emerges if and only if ethnic diversity is high and/or institutions weak. Section 3 tests for a link between fertility and ethnic diversity in a cross-national dataset. Section 4 concludes the paper. All tables and the proof of the paper’s single proposition are shown in the appendix.

2. A model of Fertility and Ethnic Conflict I assume that there are N ≥ 1 symmetric ethnic groups. Each group i = 1,.., N maximizes its lifetime payoff per capita by choosing population size ni (equal to the labor force) and capital per worker k i . There is a quality-quantity trade-off in fertility (Becker 1960), because on one hand population growth decreases the capital stock per worker, but on the other hand the groups use a labor-intensive technology to appropriate society’s production. A group’s first-period

9

population size, capital stock and income are unity. The discount factor is β ∈ (0,1) , and second period production per capita is k iα , α ∈ (0,1) . On the other hand, group i ’s second period consumption per capita is

1 ni

niη N

N

∑n

∑ n j k αj , where

η j =1

N

∑n kα j

j

is economy-wide output. The

j =1

i

i =1

fraction niη

N

∑ nη i

is the output share that accrues to group i as the result of redistributive

i =1

conflict. The parameter η > 0 measures the decisiveness of the population input to conflict (Tullock 1980, Hirshleifer 1991). Since the returns to conflict should be limited by strong institutions (Easterly 2001) I henceforth interpret η as an inverse measure of institutional quality. Group i solves

max U i = ci1 + βci 2 = (1 − k i ni ) + β k i , ni

niη N

∑n

η

1 ni

N

∑n kα j

j

(1)

j =1

j

j =1

subject to

k i ni ≤ 1 ,

(2)

ki ≥ k > 0 ,

(3)

ni ≥ n > 0 .

(4)

10

(2) ensures feasibility and (3) and (4) bound the future capital stock and population from below. (3) is plausible if part of the initial capital stock cannot be ‘eaten’ and (4) if today’s agents can ensure their own or their descendants’ survival regardless of what is optimal for the group. To focus, I avoid the less interesting corner solutions where the feasibility constraint (2) is binding

  N  by assuming (A1) n ≥ 0.5 and (A2): k ≤   ( N − 1) β (η − 1) 

1/ α

. To ensure a unique solution I also

assume α ≠ ( N − 1)(η − 1) .

Proposition 1 In the symmetric equilibrium

{

}

(i) If α > ( N − 1)(η − 1) then n = n and k = min k , ( βα / N n)1 /(1−α ) .  β ( N − 1)(η − 1)  (ii) α < ( N − 1)(η − 1) then k = k and n = min n,  1−α Nk  

(iii) The fertility rate tends to rise when ethnic diversity is greater and institutions are weaker.

Proof See the appendix.

To explain the intuition behind Proposition 1, it is useful to write the first-order conditions for investment and population growth (a6)-(a7) in the appendix as follows:

βα k α −1 / N + λ k = n

k = β (η − 1)

N −1 kα + λn N n

(6)

(7)

11

Condition (6) equates the marginal investment return with the marginal cost. Due to imperfect property rights, group i only receives a share 1 / N of the social investment return. Consequently, investment is inefficiently low given fertility if N > 1 . On the right hand side, a rise in (one plus) fertility n raises the marginal cost of equipping each future group member with a unit of capital. Condition (7) equates the marginal cost of fertility, which is the cost of equipping future members with capital, with the marginal benefit of fertility. This marginal benefit consists of (a) the discounted value of greater ability to appropriate society’s future output via the contest success function. One can show (see the appendix) that this gain is βη

( N − 1) k α . At the same N n

time, however, for a fixed amount of wealth population growth dilutes income per capita, which leads to a loss β

1 ( N − 1)k α . The net benefit is the right hand side of (7). N n

From (6) and (7) the returns to capital and fertility when no constraints are binding ( λ k = λ n = 0 ) are MBk / MCk = βαk α −1 / Nn and MBn / MCn = β ( N − 1)(η − 1)k α −1 / Nn . Therefore, the return to capital (production) exceeds the return to fertility (appropriation) for all values of k and n if and only if α > ( N − 1)(η − 1) . When this is true, then all ethnic groups prefer low fertility and high investment. On the other hand, when α < ( N − 1)(η − 1) then they prefer low investment and high fertility. A rise in N or η - a rise in ethnic diversity or a fall in the quality of institutions – make the necessary and sufficient conditions for the high fertility equilibrium, α < ( N − 1)(η − 1) and

12

β ( N − 1)(η − 1) Nk

by n =

1−α

> n , more likely to hold. Further, fertility within that equilibrium, which is given

β ( N − 1)(η − 1) Nk

1−α

, will rise.

3. Evidence Linking Fertility to Ethnic Diversity This section studies the relationship between fertility and ethnic diversity empirically. Before proceeding I note that there actually are not many macro-econometric studies of the determinants of fertility and none study the link between ethnic diversity and fertility on which I focus. Barro (1991) finds a negative cross-country relationship between income and the net fertility rate, where the net fertility rate is defined as the total fertility rate – the number of children that would be born to a woman if she were to live to the end of her childbearing years and bear children in accordance with prevailing age-specific fertility rates - times one minus the under-age 5 mortality rate. Barro and Lee (1993) find an inverted u-shaped cross-country relationship between income and total fertility after controlling for infant mortality. Ahituv (2001) find a negative relationship between total fertility and human and physical capital in a country panel. Eckstein et al. (1999) and Herzer et al. (2010) find a negative long-run relationship between income and net fertility rates in time-series and panel data, respectively.

Although a panel dataset would be ideal to study ethnic diversity and fertility empirically, unfortunately it is difficult to find time-varying data on ethnic diversity. Consequently, I use the cross-national data on ethnic and religious diversity provided by Alesina et al. (2003) and estimate the ethnic diversity-fertility link on a cross-section of countries. As the dependent

13

variable I use a country’s average net fertility rate between 2002 and 2007 computed from the World Bank’s World Development Indicators (WDI). Since the latest ethnic diversity observation in Alesina et al. is from 2001, the fact that net fertility is averaged over 2002-2007 should limit the risk of reverse causation from fertility to the ethnic diversity measure. I estimate the model with OLS and robust standard errors.

In order to control for other determinants of macro-level fertility, first, using WDI I compute each country’s GNI per capita in thousands of 2005 US PPP dollars. The PPP correction presumably gives a better measure of living standards and income- and opportunity cost factors affecting the household fertility decision than uncorrected income measures. To test the hypothesis that ethnic diversity raises fertility more where institutions are weak, I also add an interaction term between ethnic diversity and GNI per capita. Second, I control for institutional quality using a measure of executive constraints from the PolityIV dataset. For example, the quality of national institutions may affect households’ labor market opportunities or their demand for informal social networks, which could in turn affect desired fertility. This measure ranges from 1 to 7 with a score of 1 indicating the least constrained executive. Third, I add the WDI primary completion rate (World Bank 2007) as well as the net female enrollment rate in secondary education from GenderStats to measure the flow, if not necessarily the stock, of education.3 I further include the ratio of females to males enrolled in secondary schools from the World Bank GenderStats dataset to measure gender equality. Fourth, I include the rural population share from FAOStat because the demand for household labor may be greater in rural

3

However, Ahituv (2001) finds that current enrollment predicts fertility better than a country’s average education

level. He conjectures that the reason is large measurement errors for the latter variable.

14

areas. Fifth, I control for the percent of roads that a paved and the percent of the population with access to an improved water source (both from WDI) since infrastructure capital could be a substitute or complement to household labor. Finally, I include region dummies with Western developed countries as the omitted region. Including region dummies may be important because, for example, sub-Saharan Africa has historically had high ethnic diversity as well as fertility rates (Caldwell and Caldwell 1987). Thus, omitting region effects could create a spurious fertility-ethnic diversity correlation. All the control variables are measured in 2001 or 2000 depending on data availability. Table 1 shows the summary statistics.

Main Results The results in Table 2 support that ethnic diversity is positively related to fertility. In addition, although the interaction term between diversity and income is only significant in 5 of the 8 regressions, it is consistently negative. This at least partly supports that the relationship between ethnic diversity and fertility is more positive in poor countries. In terms of the quantitative effect, the coefficients from Model 8 imply that a rise from the sample minimum to the maximum ethnic diversity level (0 to 0.93) at the sample mean GNI per capita ($US 9,958) is associated with a moderate rise in net fertility of 0.6 children or 0.5 standard deviations. For a poor country with a per capita income of $1000, however, the rise in fertility would be 1.27 or just over one standard deviation. One standard deviation (0.26) rise in ethnic diversity implies a rise in net fertility of 0.14-0.29 children (0.11-0.24 standard deviations) as income moves from $1000 to the sample mean.

15

Regarding the control variables, somewhat surprisingly income per capita turns out to be weakly or even positively related to net fertility. However, there are three reasons why this result may be less surprising than it first appears. First, the coefficient on income is only positive and significant after controlling for female education, which may be one of the primary channels through which development decreases fertility. Before that the sign on income is negative (and barely significant in Model 1).4 Second, adding the coefficients on income and ethnic diversitytimes-income in Models 7-8 shows that the income-fertility relationship is actually still negative in diverse societies (roughly, when the ethnic diversity measure exceeds 0.70, which is the sample mean plus one standard deviation). Third, many studies in demography find either no effect or only a weak effect of socioeconomic factors on fertility, and generally demographers lack a unified theoretical framework to explain national fertility rates (Coale and Watkins 1986, Cleland and Wilson 1987, Hirschman 1994). Among the other controls, as expected institutional quality (executive constraints), the primary completion rate, the female-male secondary school enrollment ratio and the female secondary enrollment rate are all negatively related to fertility although the former three are insignificant after controlling for female secondary enrollment. The reason I drop the primary completion rate after Model 5 is that it is highly correlated (0.89) with female secondary enrollment. Once the other controls are added, there is no evidence that the rural population share or the infrastructure variables matter.5 4

There is also evidence that fertility declines may diffuse across countries (NRC Committee on Population 2010).

This could make the relationship between income and fertility more difficult to detect. 5

The following changes also do not affect the results for ethnic diversity: using the percent of GDP produced in

agriculture (from WDI) instead of the rural population share, using the female-male ratio in primary enrollment or tertiary education (GenderStats) instead of the ratio in secondary enrollment, or including life expectancy or female life expectancy at birth (GenderStats).

16

Additional Results Table 3 shows results from using alternative controls on the right hand side. For brevity the table omits the region fixed effects and the constant term. Model 9 controls for the female labor force participation rate from GenderStats as a measure of the female opportunity cost of time or gender equality. Model 10 includes the percent of parliamentary seats held by females, also from GenderStats, as either another gender equality measure or a control for the extent to which the identities of legislators affect government policies relating to fertility (Pande ). Model 11 includes the percent females in the population (again from GenderStats): smaller female populations might be a symptom of male dominance or face increased social pressure to bear children to sustain family dynasties. Model 12 includes internet use (from WDI) as a proxy for information, communication and exposure to Western values. Model 13 controls for the ratio of a country’s plurality ethnic group to the second largest group (from Fearon and Laitin 2003): although the model in the previous section assumed equally sized groups, in practice some ethnic groups are larger than others in the same country, and one theory in demography states that minorities use fertility to enhance their security or social mobility (Goldscheider and Uhlenberg 1969, Anson and Meir 1996, Agadjanian 1999). Model 14 includes the ratio of military expenditure to GDP (from WDI) in case nationalistic governments promote fertility (Courbage 1999, Shiffman et al. 2002, Birenbaum-Carmeli 2009). In all cases the main results from before are unchanged.

Next, in models 15 and 16 I control for the literacy rate of females aged more than 15 and 15-24, respectively (both from GenderStats). These regressions drop the female secondary enrollment rate since it is quite highly correlated (0.61) with the first literacy measure and almost perfectly

17

correlated (0.96) with the second measure, as well as to increase the number of observations. In both regressions, ethnic diversity and its interaction with income are insignificant, but with only 12 and 11 degrees of freedom in the two regressions, it is difficult to reject the null hypothesis of no effect. Moreover, the coefficients on ethnic diversity and the interaction term have the same sign as before, they are still quite large and in model 16 they are almost significant (p=.104 and p=.149, respectively). The last result is hardly surprising since female literacy for 15-24 year olds may effectively be a proxy for female secondary enrollment. Measuring the right hand side variables in 2000 instead of 2001 gives a few more degrees of freedom and significant coefficients on ethnic diversity 1.66 (p=0.03) and 1.34 (p=0.07) in Models 15-16 with the interaction term still insignificant.6 Overall, though, the sample sizes are probably too small to allow inference.

In Tables 4 and 5 I try in two ways to deal with the well known risk of omitted variables problems in cross-country regressions. First, Table 4 relates ethnic diversity not to fertility but to the future change in fertility: however the world’s average fertility rate has changed between 2002 and 2007, if all else constant the change was more positive in ethnically diverse countries then this supports a positive effect of ethnic diversity on fertility. The dependent variable in Table 4 is a country’s average fertility rate in 2006-7 minus its average fertility rate in 2002-5. I use these averages rather than simply the difference between 2007 and 2002 to maximize data availability and choose the cutoff in 2005 for the same reason.7 To control either for (i) the fact that time-invariant omitted variables could lead to high ethnic diversity as well as fertility in some countries, or (ii) the possible presence of a ‘convergence effect’ by which all countries 6

Scatterplots of fertility against ethnic diversity for the relevant observations show no evidence that outliers drive the results either when the right hand side is measured in 2001 or when it is measured in 2000. 7 All other possible cutoffs yield similar results but fewer observations.

18

converge asymptotically to, say, replacement fertility or the fertility rate of developed countries, I also include a control for initial fertility. Again to increase data availability all right hand side variables are measured in 2000, but otherwise the controls are the same as in the main regressions in Table 2. The results show that more ethnically diverse countries have indeed had increasing fertility rates compared to less diverse countries: the coefficient on ethnic diversity is positive and significant in all but one of the regressions. On the other hand the interaction term between diversity and income, while negative like before, is rarely significant. As an alternative approach, in Table 5 I rerun the levels regressions from Table 2 except I now include initial fertility on the right hand side and (again to raise the number of observations) measure all the right hand side variables in 2000. As in Table 4, controlling for initial fertility should help to capture any time-invariant omitted variables driving both diversity and fertility. The results show that ethnic diversity is again positively and significantly related to future fertility, although the interaction term is again mostly insignificant.

Finally, in Table 6, I briefly examine the link between fertility and religious rather than ethnic diversity: in principle, group conflict within a country could be organized along religious rather than ethnic lines (McQuillan 2004).8 However, when I rerun the regressions in Table 2 and replace the ethnic diversity measures with religious diversity (also from Alesina et al. and all measured in 2001), I find no evidence of a relationship with fertility. One reason for the contrast with ethnic diversity may be that ethnic groups compared to religious groups find it easier to enforce high fertility norms because the members interact more frequently. Ethnic group leaders compared to religious leaders may also be better placed in the political system to affect policy determinants of fertility, such as family planning or fertility-related education. 8

Moreover,

Like Table 3, Table 4 omits the results for the region effects and the constant term for brevity.

19

religious competition may be more a question of religious doctrines competing against each other rather than religious groups competing within a country. For example, Catholicism, Islam and Judaism are sometimes described as promoting fertility, but if such competition exists it may primarily be global or regional. It should then be relatively independent of religious diversity in any one country.

4. Conclusion This paper has studied the link between national fertility rates and political conflict between ethnic groups. I find that in a cross-national sample the data supports a positive relationship between ethnic diversity and fertility and this relationship is stronger for poor countries, which the political conflict may be more intense due to weak institutions. These findings are consistent with not just economic, but also political factors affecting fertility rates. I have hypothesized that the optimal fertility rate at the ethnic group level, which is determined by political factors, may be enforced among households via social norms. However, I do not test for this norm enforcement empirically. To the extent high fertility is harmful for economic development but sustained by political factors, addressing the problem may require addressing the underlying political conflict.

20

References Agadjanian, Victor 1999. Post-Soviet Demographic Paradoxes: Ethnic Differences in Marriage and Fertility in Kazakhstan. Sociological Forum 13, 3, 425-46 Ahituv, A., 2001. Be Fruitful or Multiply: On the interplay between Fertility and Economic Development, Journal of Population Economics 14, 1, 51-71. Albanese, Patrizia 2004. Abortion & Reproductive Rights under Nationalist Regimes in Twentieth Century Europe. Women’s Health and Urban Life 3, 1, 8-33 Alesina, A., A Devleeschauwer, W Easterly, S y. Kurlat and R Wacziarg 2003. Fractionalization. Journal of Economic Growth 8, 2, 155-94 Anson, Jon and Avinoam Meir 1996. Religiosity, Nationalism and Fertility in Israel. European Journal of Population 12, 1, 1-25 Attane, I., & Courbage, Y. (2000). Transitional Stages and Identity Boundaries: The case of Ethnic Minorities in China. Population and Environment, 21, 257–280. Bardhan, Pranab and Christopher Udry 1999. Development Microeconomics. New York: Oxford University Press. Barro Robert J 1991. Economic Growth in a Cross Section of Countries. Quarterly Journal of Economics 106, 2, 407-443 Barro Robert J and Jae Woo Lee 1993. Losers and Winners in Economic Growth. NBER Working Paper No. 4341 Basu, Alaka Malwade 1997. The ‘Politicization’ of Fertility to Achieve Non-Demographic Objectives. Population Studies 51, 1, 5-18 Bates, Robert H 1981. Markets and States in Tropical Africa: The Political Basis of Agricultural Policies. Berkeley: University of California Press.

21

Becker, Gary S 1960. An Economic Analysis of Fertility. In: Demographic and Economic Change in Developed Countries. Princeton: National Bureau of Economic Research, 209-31. Becker, G S and Robert J Barro 1989. Fertility Choice in a Model of Economic Growth. Econometrica 57, 2, 481-501. Birenbaum-Carmeli, Daphna 2009. The Politics of ‘The Natural Family’ in Israel: State Policy and Kinship Ideologies. Social Science and Medicine 69, 7, 1018-24 Bloom, David E, David Canning and Jaypee Sevilla 2001. Economic Growth and the Demographic Transition. NBER Working Paper 8685 Boserup, Ester 1965. The Conditions of Agricultural Growth: The Economics of Agrarian Change under Population Pressure. Chicago: Aldine Brown, Jessica Autumn and Myra Marx Ferree 2005. Close Your Eyes and Think of England. Gender and Society 19, 1, 5-24 Brunnbauer, Ulf 2004. Fertility, Families and Ethnic Conflict: Macedonians and Albanians in the Republic of Macedonia, 1944-2002. Nationalities Papers 32, 3, 565-98 Caldwell, John C. and Pat Caldwell 1987. The Cultural Context of High Fertility in sub-Saharan Africa. Population and Development Review 13, 3, 409-437 Chandra, Kanchan 2007. Why Ethnic Parties Succeed: Patronage and Ethnic Head Counts in India. New York: Cambridge University Press Childs, Geoff, Melvyn C Goldstein, Ben Jiao and Cynthia M Beall 2005. Tibetan Fertility Transitions in China and South Asia. Population and Development Review 31, 2, 337-49 Childs, Geoff and Gareth Barkin 2008. Reproducing Identity: Using Images to Promote Pronatalism and Sexual Endogamy among Tibetan Exiles in South Asia. Visual Anthropology Review 22, 2, 34-52

22

Cleland, John and Chris Wilson 1987. Demand Theories of the Fertility Decline: An Iconoclastic View. Population Studies 41, 1, 5-30 Coale, Ansley J and Susan Cotts Watkins (eds) 1986. The Decline of Fertility in Europe. Princeton: Princeton University Press Collier, Paul and Ankle Hoeffler 2004. Greed and Grievance in Civil War. Oxford Economic Papers 56, 4, 563-95 Courbage, Yussef 1999. Economic and Political Issues of Fertility Transition in the Arab World—Answers and Open Questions. Population and Environment 20, 4, 353-80 Dasgupta, Partha 1993. An Inquiry into Well-Being and Destitution. Oxford: Clarendon Press. Dasgupta, Partha 1995. The Population Problem: Theory and Evidence. Journal of Economic Literature 33, 4, 1879-1902. Easterly, William and Ross Levine 1997. Africa’s Growth Tragedy: Policies and Ethnic Divisions. Quarterly Journal of Economics 112, 4, 1203-50. Easterly, William 2001. Can Institutions Resolve Ethnic Conflict? Economic Development and Cultural Change 49, 4, 687-706. Eckstein, Z. Mira, P., and Wolpin, K., 1999, A Quantitative Analysis of Swedish Fertility Dynamics. Review of Economic Dynamics 2, 1, 137-165. Faour, Muhammad 1989. Fertility Policy and Family Planning in the Arab Countries. Studies in Family Planning 20, 5, 254-63 Fearon, James D. and David Laitin 2003. Ethnicity, Insurgency, and Civil War. American Political Science Review 97, 1, 75-90. Franks, Emma 1996. Women and Resistance in East Timor: “The Centre, as They Say, Knows Itself by the Margins”. Women’s Studies International Forum 19, 1-2, 155-68

23

Galor Oded and David N. Weil 2000. Population, Technology and Growth: From Malthusian Stagnation to the Demographic Transition and Beyond. American Economic Review 90, 806-828. Goldscheider, Calvin and Peter Uhlenberg 1969. Minority Group Status and Fertility. American Journal of Sociology 74, 4, 361-72 Herzer, Dierk, Holger Strulik and Sebastian Vollmer 2010. The Long-run Determinants of Fertility: One Century of Demographic Change 1900-1999. Harvard Initiative for Global Health, Program on the Global Demography of Aging Working Paper No. 63. Hirschman, Charles 1994. Why Fertility Changes. Annual Review of Sociology 20, 203-33 Hirshleifer, Jack 1991. The Paradox of Power. Economics and Politics 3, 177-200 Horowitz, Donald 2000. Ethnic Groups in Conflict. 2nd Ed. Berkeley: University of California Press. Iyigun, Murat and Randall P Walsh 2007. Endogenous Gender Power, Household Labor Supply and the Demographic Transition. Journal of Development Economics 82, 1, 138-55 Kimenyi, Mwangi S and Roxana Guiterrez Romero. Identity, Grievances and Economic Determinants of Voting in the 2007 Kenyan Elections. University of Connecticut Department of Economics Working Paper No. 38. Klausen, Susanne 2002. The Uncertain Future of White Supremacy and the Politics of Fertility in South Africa 1930-1939. Presented at the University of Natal African Studies Seminar Kohler, Hans-Peter 2001. Fertility and Social Interaction: An Economic Perspective. Oxford: Oxford University Press Kokole, Omari H 1994. The Politics of Fertility in Africa. Population and Development Review 20, Supplement: The New Politics of Population: Conflict and Consensus in Family Planning, 73-88.

24

Kremer, Michael 1993. Population Growth and Technological Change: One Million BC to 1990. Quarterly Journal of Economics 108, 3, 681-716. Krugman, Paul 1992. Geography and Trade. Leuven University, MIT Press Kunovich, Robert M and Catherine Deitelbaum 2004. Ethnic Conflict, Group Polarization and Gender Attitudes in Croatia. Journal of Marriage and Family 66, 5, 1089-1107 Malthus, Thomas Robert 1798/1999. An Essay on the Principle of Population. Oxford University Press Manski, Charles F and Joram Mayshar 2003. Private Incentives and Social Interactions: Fertility Puzzles in Israel. Journal of the European Economic Association 1, 1, 181-211 McQuillan, Kevin 2004. When Does Religion Influence Fertility? Population and Development Review 30, 1, 25-56. Munshi, Kevin and Jacques Myaux (2006). Social Norms and the Fertility Transition. Journal of Development Economics 80, 1, 1-38 Nahmias, Petra and Guy Stecklov 2007. The Dynamics of Fertility amongst Palestinians in Israel from 1980 to 2000. European Journal of Population 23, 1, 71-99 National Research Council Committee on Population 2010: The Role of Diffusion Processes in Fertility Change in Developing Countries: Report of a Workshop. By Holly Reed, Rona Briere and John B. Casterline. Washington, DC: National Academy Press Nelson, R 1956. A Theory of the Low-Level Equilibrium Trap in Underdeveloped Economies. American Economic Review 46, 5, 894-908. Obono, Oka 2003. Cultural Diversity and Population Policy in Nigeria. Population and Development Review 29, 1, 103-11

25

Overseas Development Institute 2008. Donors and the Political Dimensions of Health Care Reform: the Cases of Tanzania and Uganda. By Kent Buse and David Booth with Grace Murindwa, Aziza Mwisongo and Andrew Harmer. ODI Working Paper 7. Panandiker, V. A. Pai and P. K. Umashankar 1994. Fertility Control and Politics in India. Population and Development Review 20, Supplement: The New Politics of Population: Conflict and Consensus in Family Planning, 89-104. Posel, Deborah 2005. Sex, Death and the Fate of the Nation: Reflections on the Politicization of Sexuality in Post-Apartheid South Africa. Africa 75, 2, 125-53 Rasul, Imran 2008. Household Bargaining over Fertility: Theory and Evidence from Malaysia. Journal of Development Economics 86, 2, 215-41 Rosenzweig, Mark R and Oded Stark 1997. Handbook of Population and Family Economics Volume 1A, 241-420. Amsterdam: Elsevier. Rubinstein, Robert A and Sandra D Lane 2003. Population, Identity and Political Violence. Social Justice: Anthropology, Peace and Human Rights 3, 3-4, 139-52 Schultz T. Paul 1997. Demand for Children in Low Income Countries. In: Handbook of Population and Family Economics Volume 1A, 350-430. Amsterdam: Elsevier Shiffman, Jeremy and Ana Lucia Garces del Valle 2006. Political History and Disparities in Safe Motherhood between Guatemala and Honduras. Population and Development Review 32, 1, 5380 Shiffman, Jeremy, Marina Skrabalo and Jelena Subotic 2002. Reproductive Rights and the State in Serbia and Croatia. Social Science & Medicine 54, 4, 625-642 Simon, Julian 1981. The Ultimate Resource. Princeton: Princeton University Press

26

Smith, Sara Hollingsworth 2009. A geopolitics of intimacy and anxiety: Religion, Territory, and Fertility in Jammu and Kashmir. PhD dissertation, University of Arizona. Solow, Robert 1956. A Contribution to the Theory of Economic Growth. Quarterly Journal of Economics 70, 1, 65-94 Tullock, Gordon 1980. Efficient Rent-Seeking. In: JM Buchanan, RD Tollison and G Tullock (Eds) Toward a Theory of the Rent-Seeking Society. College Station, Texas A&M University Press, 97-12 Udjo, Eric O 1984. Obstacles to Successful Fertility Control in Nigeria. Social Science and Medicine 19, 11, 1167-71 Ujah, IAO, OA Aisien, JT Mutihir, DJ Vanderjagt, RH Glew and VE Uguru 2005. Factors Contributing to Maternal Mortality in North-Central Nigeria: A Seventeen-year Review. African Journal of Reproductive Health 9, 3, 27-40 UNICEF 2009. The State of the World’s Children. New York: United Nations Children’s Fund Urdal, Henrik 2006. A Clash of Generations? Youth Bulges and Political Violence. International Studies Quarterly 50, 3, 607-629 Watkins, Susan Cotts 1996. Social Interactions and Contemporary Fertility Transitions. Population and Development Review 22, 4, 639-82 Winckler, Onn 2005. Arab Political Demography. Volume 1: Population Growth and Natalist Policies. Sussex Academic Press World Bank 2007. Population Issues in the 21st Century: The Role of the World Bank. World Bank Health, Nutrition and Population Discussion Paper 40583. Young, Crawford 1976. The Politics of Cultural Pluralism. Madison: University of Wisconsin Press

27

Appendix: Tables and Proof of Proposition 1 Table 1: Summary Statistics Variable Net fert. 2002-8 Ethnic diversity Rel. diversity

GNI/CAP (PPP) Exec constr. Prim cmpl. rate Rural pop share F/M sec enr F sec enr net %Paved roads Improved water

Observations Mean 182 183 201 172 146 136 197 142 98 131 176

S.d. 2.76 0.44 0.44 9.96 4.78 84.25 0.45 95.41 59.37 49.34 81.75

Min 1.22 0.26 0.23 11.59 2.11 25.32 0.24 18.52 29.35 33.72 19.34

Max 1.15 0 0 0.22 1 18.83 0 28.45 2.86 0.8 21

5.94 0.93 0.86 52.83 7 130.7 0.91 138.61 99.97 100 100

Notes: Ethnic and religious diversity are from Alesina et al. (2003). Net fertility - the number of children that would be born to a woman if she were to live to the end of her childbearing years and bear children in accordance with prevailing age-specific fertility rates times one minus the under-age 5 mortality rate - along with GNI/Cap (in $US 2005 PPP), the primary completion rate, the percent of roads that are paved, and access to improved water sources are from World Development Indicators (WDI) and author’s calculations. Executive constraints are from Polity IV and the rural population share from FAOStat. The ratio of girls to boys in secondary schooling and the female secondary enrollment rate are from World Bank GenderStats.

28

Table 2: Ethnic Diversity and Fertility ethnic GNI/Cap ethnic*GNI/Cap eeurop lamerica asia ssafrica nafrme

(1) 0.893** (0.336) -0.020+ (0.011) -0.026 (0.017) -0.784** (0.210) 0.065 (0.239) 0.144 (0.280) 1.380** (0.315) 0.358 (0.252)

exconst

(2) 1.005** (0.316) -0.014 (0.010) -0.038* (0.017) -0.861** (0.208) 0.113 (0.236) -0.162 (0.269) 1.111** (0.325) 0.103 (0.254) -0.115** (0.035)

prim cmpl

(3) 1.066** (0.378) -0.004 (0.016) -0.036 (0.024) -0.779* (0.310) -0.032 (0.359) -0.241 (0.422) 0.669 (0.461) -0.028 (0.347) -0.062+ (0.037) -0.015** (0.005)

F/M Sec Enr

(4) 1.151** (0.432) 0.006 (0.019) -0.045 (0.029) -0.654+ (0.340) 0.168 (0.396) -0.209 (0.458) 0.704 (0.489) 0.001 (0.378) -0.095** (0.034) -0.009 (0.006) -0.009* (0.004)

F Sec Enr Net

(5) 1.502** (0.520) 0.010 (0.025) -0.055 (0.038) -0.640 (0.428) -0.256 (0.475) -0.524 (0.580) 0.214 (0.564) 0.162 (0.440) -0.076 (0.046) 0.005 (0.008) -0.007 (0.006) -0.018* (0.008)

rural pop

(6) 1.746** (0.483) 0.028 (0.022) -0.075* (0.032) -0.338 (0.317) 0.099 (0.441) -0.146 (0.450) 0.496 (0.476) 0.474 (0.413) -0.074 (0.047)

(7) 1.020+ (0.583) 0.054** (0.017) -0.078* (0.029) 0.285 (0.291) 0.630 (0.459) 0.520 (0.488) 1.059 (0.649) 0.419 (0.477) -0.125 (0.075)

(8) 1.457** (0.415) 0.054** (0.018) -0.082** (0.024) 0.241 (0.302) 0.411 (0.419) 0.026 (0.481) 0.943* (0.454) 0.486 (0.407) -0.057 (0.050)

-0.004 (0.005) -0.014+ (0.007) 0.714 (0.724)

-0.003 (0.006) -0.020* (0.008) -0.151 (0.553) -0.002 (0.003)

0.002 (0.005) -0.019* (0.008) 0.229 (0.696)

%roads paved imprvd water Constant Observations R-sqrd

2.271** (0.300) 152 0.75

2.911** (0.342) 135 0.79

3.894** (0.646) 98 0.83

4.162** (0.725) 81 0.84

3.914** (0.775) 60 0.86

2.944** (0.956) 68 0.88

3.480** (0.789) 48 0.89

-0.013 (0.009) 3.434** (0.960) 67 0.89

Notes: Robust standard errors in parentheses. + significant at 10%; * significant at 5%; ** significant at 1%. Ethnic diversity is from Alesina et al. (2003). Net fertility - the number of children that would be born to a woman if she were to live to the end of her childbearing years and bear children in accordance with prevailing age-specific fertility rates times one minus the under-age 5 mortality rate - along with GNI/Cap (in thousands of $US 2005 PPP), the primary completion rate, the percent of roads that are paved, and access to improved water sources are from World Development Indicators (WDI) and author’s calculations. Executive constraints are from Polity IV and the rural population share from FAOStat. The ratio of girls to boys in secondary schooling and the female secondary enrollment rate are from World Bank GenderStats.

29

Table 3: Additional Results ethnic GNI/Cap ethnic*GNI/Cap exconst2 rural pop F Sec Enr Net F lbr frc prtcp

(9) 1.707** (0.381) 0.023 (0.020) -0.070* (0.028) -0.079+ (0.043) 0.635 (0.672) -0.016** (0.005) 0.007 (0.006)

%F in parl

(10) 1.893** (0.400) 0.046* (0.018) -0.083** (0.029) -0.066 (0.053) 1.342* (0.511) -0.016** (0.005)

(11) 1.389** (0.375) 0.048* (0.021) -0.079** (0.022) -0.079+ (0.046) 0.390 (0.604) -0.022** (0.005)

(12) 1.762** (0.407) 0.026 (0.022) -0.076* (0.029) -0.080+ (0.047) 0.697 (0.683) -0.016** (0.005)

(13) 1.777** (0.465) 0.027 (0.022) -0.079* (0.030) -0.050 (0.061) 0.566 (0.718) -0.017** (0.006)

(14) 2.135** (0.347) 0.051** (0.017) -0.095** (0.026) -0.107* (0.044) 1.512** (0.431) -0.013** (0.005)

(15) 0.802 (0.735) 0.023 (0.033) -0.060 (0.071) -0.179+ (0.084) 1.005 (0.576)

0.000 (0.007)

% F Pop

-0.011 (0.043)

internet

0.003 (0.005)

plu/Sec

-0.005 (0.004)

mili/gdp

0.009 (0.011)

ltrcy F 15+

-0.018* (0.007)

ltrcy F 15-24 observations R-sqrd

(16) 1.241 (0.700) 0.026 (0.032) -0.085 (0.055) -0.224** (0.066) 1.012 (0.657)

70 0.88

64 0.90

74 0.89

70 0.88

60 0.88

66 0.91

24 0.91

-0.026** (0.008) 23 0.94

Notes: Robust standard errors in parentheses. + significant at 10%; * significant at 5%; ** significant at 1%. Apart from the variable sources explained below Table 2, the female labor force participation rate, percent of parliamentary seats held by females, percent females in the population, and literacy rates for females aged more than 15 and 15-24 are from World Bank GenderStats. Internet use and military expenditures are from WDI. The ratio of a country’s plurality ethnic group to the second largest group is from Fearon and Laitin (2003).

30

Table 4: Change in Fertility from 2002-05 to 2006-07 netfert 2000 ethnic GNI/Cap ethnic*GNI/CAP eeurop lamerica asia ssafrica nafrme

(17) -0.023* (0.009) 0.071** (0.027) 0.003+ (0.001) -0.005* (0.002) 0.009 (0.030) -0.047 (0.030) -0.069* (0.032) -0.064+ (0.038) -0.048 (0.031)

exconst

(18) -0.022* (0.010) 0.065* (0.030) 0.001 (0.001) -0.003 (0.002) 0.009 (0.030) -0.057+ (0.029) -0.076* (0.032) -0.061 (0.040) -0.037 (0.031) 0.008* (0.004)

prim cmpl

(19) -0.025 (0.016) 0.097* (0.043) 0.002 (0.002) -0.005+ (0.003) 0.014 (0.042) -0.061 (0.040) -0.082 (0.052) -0.070 (0.046) -0.033 (0.041) 0.008 (0.005) -0.000 (0.000)

F/M Sec Enr

(20) -0.016 (0.015) 0.112+ (0.064) 0.003 (0.003) -0.007 (0.005) 0.032 (0.046) -0.056 (0.048) -0.069 (0.063) -0.063 (0.052) -0.013 (0.039) 0.005 (0.007) -0.000 (0.000) 0.001 (0.001)

F Sec Enr Net

(21) -0.012 (0.019) 0.092 (0.066) 0.001 (0.003) -0.004 (0.005) -0.025 (0.044) -0.072 (0.043) -0.123+ (0.064) -0.079 (0.051) -0.037 (0.040) 0.005 (0.009) -0.001+ (0.001) 0.001 (0.001) 0.001 (0.001)

rural pop

(22) -0.008 (0.017) 0.101* (0.043) 0.002 (0.002) -0.005 (0.003) -0.004 (0.031) -0.069+ (0.038) -0.098* (0.042) -0.064 (0.051) -0.018 (0.029) 0.008 (0.007)

(23) 0.000 (0.018) 0.107* (0.043) 0.003 (0.003) -0.006 (0.004) -0.018 (0.050) -0.100+ (0.052) -0.113* (0.051) -0.122+ (0.065) -0.012 (0.038) 0.006 (0.006)

(24) -0.010 (0.015) 0.109* (0.041) 0.001 (0.002) -0.004 (0.003) -0.023 (0.033) -0.075+ (0.039) -0.113** (0.040) -0.073 (0.049) -0.019 (0.026) 0.006 (0.008)

0.000 (0.001) 0.000 (0.001) -0.009 (0.042)

0.001 (0.001) 0.000 (0.001) 0.046 (0.042) -0.000 (0.000)

0.001 (0.001) 0.001 (0.001) -0.044 (0.058)

%roads paved imprvd water constant observations R-sqrd

0.031 (0.042) 145 0.49

-0.002 (0.050) 128 0.54

0.009 (0.085) 93 0.56

-0.070 (0.088) 77 0.55

-0.025 (0.084) 57 0.59

-0.099 (0.073) 68 0.61

-0.128 (0.100) 46 0.72

-0.001 (0.001) -0.007 (0.076) 63 0.62

Notes: Robust standard errors in parentheses. + significant at 10%; * significant at 5%; ** significant at 1%. For variable sources, see Table 1.

31

Table 5: Controls for Initial Fertility netfert 2000 ethnic GNI/Cap ethnic*GNI/CAP eeurop lamerica asia ssafrica nafrme

(25) 0.892** (0.027) 0.227* (0.089) 0.004 (0.003) -0.011* (0.005) 0.014 (0.065) -0.144* (0.067) -0.170* (0.071) -0.040 (0.085) -0.146* (0.073)

exconst

(26) 0.884** (0.031) 0.229* (0.093) 0.001 (0.003) -0.007 (0.005) -0.019 (0.058) -0.200** (0.060) -0.234** (0.068) -0.042 (0.084) -0.153* (0.067) 0.014 (0.012)

prim cmpl

(27) 0.881** (0.047) 0.154 (0.145) -0.003 (0.006) -0.003 (0.009) -0.107 (0.119) -0.295* (0.118) -0.473** (0.144) -0.147 (0.117) -0.236+ (0.122) 0.015 (0.016) -0.001 (0.001)

F/M Sec Enr

(28) 0.902** (0.042) 0.208 (0.170) 0.003 (0.005) -0.011 (0.009) -0.031 (0.086) -0.252* (0.101) -0.402** (0.133) -0.085 (0.103) -0.149+ (0.077) 0.004 (0.019) -0.001 (0.002) 0.002 (0.002)

F Sec Enr Net

(29) 0.942** (0.049) 0.293+ (0.170) -0.007 (0.007) -0.004 (0.010) -0.238* (0.107) -0.321** (0.106) -0.471** (0.138) -0.180 (0.109) -0.238* (0.097) -0.002 (0.026) -0.006** (0.002) 0.000 (0.002) 0.009** (0.003)

rural pop

(30) 0.951** (0.048) 0.384** (0.140) 0.000 (0.006) -0.011 (0.008) -0.118 (0.078) -0.269** (0.099) -0.339** (0.111) -0.103 (0.124) -0.124 (0.081) 0.007 (0.024)

(31) 0.980** (0.037) 0.401** (0.138) -0.002 (0.005) -0.008 (0.007) -0.168+ (0.091) -0.390** (0.115) -0.371** (0.102) -0.318* (0.119) -0.156* (0.075) 0.010 (0.018)

(32) 0.951** (0.049) 0.395** (0.134) -0.002 (0.006) -0.009 (0.008) -0.155+ (0.087) -0.292** (0.103) -0.365** (0.113) -0.123 (0.132) -0.112 (0.079) 0.007 (0.026)

-0.000 (0.001) 0.004+ (0.002) 0.094 (0.125)

-0.001 (0.002) 0.005* (0.002) 0.191 (0.157) -0.001 (0.001)

-0.000 (0.001) 0.006+ (0.003) 0.080 (0.141)

%roads paved imprvd water constant observations R-sqrd

0.068 (0.107) 145 0.98

0.065 (0.141) 128 0.98

0.253 (0.285) 93 0.98

0.044 (0.240) 77 0.99

0.102 (0.260) 57 0.99

-0.362 (0.228) 68 0.99

-0.373 (0.290) 46 0.99

-0.002 (0.002) -0.237 (0.234) 63 0.99

Notes: Robust standard errors in parentheses. + significant at 10%; * significant at 5%; ** significant at 1%. For variable sources, see Table 1.

32

Table 6: Fertility and Religious Diversity religion GNI/Cap religion*GNI/Cap

(33) 0.065 (0.363) -0.032* (0.012) -0.010 (0.018)

Exconst

(34) -0.142 (0.379) -0.038** (0.014) 0.006 (0.020) -0.118** (0.037)

prim cmpl

(35) 0.028 (0.445) -0.029+ (0.015) -0.001 (0.020) -0.061 (0.037) -0.018** (0.005)

F/M Sec Enr

(36) 0.277 (0.438) -0.012 (0.014) -0.018 (0.021) -0.094* (0.036) -0.008 (0.006) -0.017** (0.005)

F Sec Enr Net

(37) 0.596 (0.434) -0.002 (0.013) -0.026 (0.020) -0.089+ (0.051) 0.007 (0.009) -0.013* (0.006) -0.019* (0.008)

rural pop

(38) 0.617 (0.420) -0.005 (0.011) -0.018 (0.022) -0.086 (0.052)

(39) 0.275 (0.537) 0.018 (0.018) -0.025 (0.030) -0.109 (0.083)

(40) 0.314 (0.531) 0.007 (0.015) -0.014 (0.025) -0.064 (0.060)

-0.011+ (0.006) -0.014+ (0.008) 0.213 (0.667)

-0.014 (0.009) -0.016 (0.010) -0.088 (0.630) -0.004 (0.004)

-0.007 (0.007) -0.018* (0.009) 0.182 (0.681)

%roads paved imprvd water Observations R-sqrd

153 0.72

136 0.76

99 0.80

82 0.83

60 0.85

68 0.85

49 0.87

-0.010 (0.010) 68 0.86

Notes: Robust standard errors in parentheses. + significant at 10%; * significant at 5%; ** significant at 1%. Apart from the variable sources explained below Table 2, religious diversity is from Alesina et al. (2003).

33

Proof of Proposition 1 Denoting the multipliers for (2)-(4) in the main paper by λ ki ni , λ k1 and λ ni , group i' s Lagrange problem is

L = ci1 + βci 2 = (1 − k i ni ) + β

nηi N

∑n

η

1 ni

N

∑n kα + λ j

j =1

j

k i ni

(1 − k i ni ) + λki k i + λ ni (ni − n) .

(a1)

j

j =1

The first order conditions for k i and ni are

− ni + β

nηi N

∑n

η

αk iα −1 − ni λk n + λk = 0 i i

i

(a2)

j

j =1

 N  ηnη −1  nη  j   i ∑ j ≠i   1  − ki + β 2 ni   N η   ∑ n j    j =1 

 α  − n k ∑ j j  N niη i≠ j α  − k i λ ki ni + λ ni = 0 njk j + N ∑ (ni ) 2  η j =1 nj ∑  j =1 

(a3)

and the complementary slackness conditions. Using symmetry, dropping the group identifying subscript, and simplifying implies that in the symmetric equilibrium, where I assume and later verify that (2) is non-binding ( λ ki ni = 0 ),

βα k α −1 / N + λ k = n

(a4)

34

N −1 kα k=β (η − 1) + λ n . N n

(a5)

These can be usefully written

βα Nn

k α −1 +

λk n

=1

β (η − 1)( N − 1) Nn

k α −1 +

(a6)

λn k

=1

(a7)

The rest of the proof proceeds in four steps. Step 1 shows that either (3), (4), or both are binding. Step 2 shows that α > ( N − 1)(η − 1) implies part (i) of the proposition. Step 3 shows that

α < ( N − 1)(η − 1) implies part (ii). Step 4 shows that part (iii) follows from (i) and (ii).

Step 1. Suppose that neither (3) nor (4) bind ( λ k = λ n = 0 ). Then (a6)-(a7) can only hold if α = ( N − 1)(η − 1) , but α ≠ ( N − 1)(η − 1) . Thus either (3), (4), or both are binding.

Step 2. If α > ( N − 1)(η − 1) , then from comparing (a6)-(a7) either (a) λ n > λ k = 0 or (b)

λ k , λ n > 0 . In Case (a) we have n = n and from (a6) k = (βα / N n)1 /(1−α ) . Condition (2) can be ignored if k n = ( βα / N n)

1 /(1−α )

 βα n =  α  Nn

   

1 /(1−α )

≤ 1 or N n α ≥ βα , which is true by (A1) since

N ≥ 2 , n > 0.5 and β , α < 1 . In case (b), again n = n and now also k = k . This case applies if

35

and only if, using (a6),

βα Nn

k α −1 < 1

or k > (βα / N n)1 /(1−α ) . This establishes part (i) of the

proposition.

Step 3. If α < ( N − 1)(η − 1) , then from comparing (a6)-(a7) either (a) λ k > λ n = 0 or (b)

λ k , λ n > 0 . In Case (a) we have k = k and from (a7) n =

ignored if n k = again k = k

N N − 1 β (η − 1) α , which is true by (A2). In Case (b) k ≤ 1 or k ≤ 1−α ( N − 1) β (η − 1) N k

and now also n = n . This case applies

β (η − 1)( N − 1) Nn

N − 1 β (η − 1) . Condition (2) can be 1−α N k

k

α −1

 β ( N − 1)(η − 1)   < 1 or k >  Nn  

if and only if, using (a7),

1 /(1−α )

. This establishes part (ii) of the

proposition.

Step 4. The only possible fertility rate above the minimum n is n = shown in parts (i) and (ii), this can only happen if

 β ( N − 1)(η − 1)  k <   Nn  

N − 1 β (η − 1) > n . As 1−α N k

α < ( N − 1)(η − 1)

and

1 /(1−α )

. Both conditions are more likely to hold when ethnic diversity N is

high and institutions are weaker and therefore η is higher. Moreover, conditional on these necessary conditions holding n =

N − 1 β (η − 1) , which is increasing in N and η .  1−α N k

36