Popul Res Policy Rev (2011) 30:449–465 DOI 10.1007/s11113-010-9196-8
Does Religion Influence Fertility in Developing Countries Tim B. Heaton
Received: 9 April 2010 / Accepted: 2 November 2010 / Published online: 20 November 2010 Ó Springer Science+Business Media B.V. 2010
Abstract This paper examines religious group differences in fertility in developing nations. Using data from the Demographic and Health Surveys of 30 countries in Asia, Africa and Latin America, this paper documents Muslim/Christian and Catholic/Protestant differences in the number of children under age 5. The paper also considers possible explanations for these differences including level of development, religious mix, social characteristics and proximate determinants of fertility. Muslim fertility is substantially higher than Christian fertility in many countries, but the average difference between Catholics and Protestants is small. Cross-national variation in group differences is at least as large as the average difference. Although level of development, social characteristics and proximate determinants play an important role in religious differences, they do not explain cross-national variation in these differences. Keywords
Fertility Religion Developing countries
Although the relationship between religion and fertility has been the focus of substantial research (Adsera 2006; Frejka and Westoff 2006; Zhang 2008), most of this work focuses on only one, or at best a small number of countries, making generalization difficult. Moreover, more complete data are usually available in more developed contexts so we know less about the relationship in less developed areas where fertility is the highest. In cases like the United States, it is possible to document longer term trends in fertility differentials such as the convergence of Catholic and Protestant fertility (Mosher and Hendershot 1984: Westoff and Jones 1979). There are also some studies documenting cross-national differences in the T. B. Heaton (&) Brigham Young University, Provo, UT, USA e-mail:
[email protected]
123
450
T. B. Heaton
U.S and Europe (Adsera 2006; Frejka and Westoff 2006), and in developing nations (Morgan et al. 2002). Differential fertility has important implications for long term growth of religious groups. Groups with higher fertility increase their share of the total population. Moreover, subgroups within religious communities that have higher fertility will become more prominent within that community. This research compares the fertility of Muslims, Catholics, and Protestants in 30 developing nations, and will examine both individual and nation level characteristics that are related to these differences.
The Role of Religion There is some evidence that Muslims have larger families than other religious groups in developing nations (Jones 2006). Comparison of communities in four Asian countries found that Muslim fertility and desired fertility tends to be higher than in other groups (Morgan et al. 2002). These differences cannot be explained by lower female autonomy or socioeconomic position. In Thailand, Buddhists have experienced a larger fertility transition than Muslims (Knodel et al. 1999). More recently, Muslim fertility has undergone substantial decline in some contexts (Jones 2006). There is less evidence regarding Catholic-Protestant differences in fertility in developing nations. Generalization from this research is only possible when religious influences tend to persist in a variety of different settings. Such an assumption seems more reasonable for some religions than for others. For example, Roman Catholics have a global hierarchy and have articulated a theology of life that may produce some cross-cultural similarity. In contrast, Protestantism is no longer composed of a few mainline groups with origins in Europe. Rather, an increasingly diverse mix of mainline, and emergent local Protestant groups defies simple description. Islam may fall somewhere between these two because there is a unifying text, but substantial local autonomy. Religious group differences in fertility may arise through various mechanisms. Goldscheider’s (1971) hypotheses that differences could be due to social characteristics, particularistic theology or minority group status have guided much of the research on religious differences. The social characteristics hypothesis implies that religious group differences will disappear once relevant social variables are taken into account. To account for Arab Christian/Muslim differences in fertility, Chamie (1981) added the interaction hypothesis that religious group differences may change in response to socio-economic development. Chamie’s model hypothesizes that religious differences will be small at the lowest levels of development, will increase at intermediate levels of development as some religious groups develop faster than others, but will be converge again at high levels of development. The minority group hypothesis postulates that minority religious groups may increase fertility in response to perceived threat of domination from other groups. In contrast to the minority group hypothesis, it may be that religious competition fostered by a diverse mix of religious groups heightens the importance of religious group membership (Stark and Iannaccone 1994; Adsera 2006). This competition model implies that religious differences will be greatest when each
123
Does Religion Influence Fertility in Developing Countries
451
religious group has an equal share of the population. Minority group status, competition among religious groups and socioeconomic composition of religious groups vary widely across national contexts. Indeed, tests for the importance of development and relative size suggest cross-cultural comparison. To adequately test these hypotheses, fertility models need to include social characteristics, level of development and relative size of religious groups using comparative data. Moreover, relationships including development and relative size may be nonlinear. It is also possible that social characteristics do not have the same relationship with fertility across religious groups. It has long been known that mass education plays an important role in fertility decline (Caldwell 1980), but this influence may not be uniform across religious groups. Evidence from Egypt suggests that Muslim education may reinforce traditional familial norms that are, at least indirectly, pronatalist (Faust et al. 1991). In Ghana, Muslims were reluctant to embrace education that was perceived to be secular because it might undermine faith (Heaton et al. 2009). The differential importance of education will be explored by including an interaction term for education and religion. Support for the particularistic theology hypothesis is often inferred if religious group differences remain after relevant control variables are taken into account. Theology may also be interpreted and emphasized in different ways in different cultural settings. McQuillan (2004) argues that three conditions must be present for religion to exert an influence on fertility. ‘‘First, the religion in question must articulate behavioral norms that have linkages to fertility outcomes’’ (p. 49). ‘‘Second, a religious group must possess the means to communicate its teachings to its members and to enforce compliance’’ (p. 49). ‘‘Finally, religious groups are more likely to influence the demographic choices of their followers when members feel a strong sense of attachment to the religious community’’ (p. 50). A religious group’s ability to articulate norms, enforce of compliance, and build attachment may also exhibit substantial variation across national settings. Heaton (1989) adds that social ties within religious communities can reinforce pronatalist norms. Lehrer (2004) argues that religious differences in family behaviors exist because religious membership and participation alters the perceived costs and benefits of family behaviors. For example, the costs of contraceptive use would be higher in religions teaching that artificial methods of birth control are immoral. Costs and benefits may be influenced by religious teachings regarding issues such as premarital sex, birth control, polygamy and divorce, but costs and benefits are also influenced by the composition of the communities to which people belong. It is not uncommon for local beliefs about contraception and family behavior to be incorporated into and reinforced by religious communities. Another way to think about the relationship between religion and fertility is via the proximate determinants of fertility. Davis and Blake (1955) provided an analytic framework to understand the processes underlying conception and birth. The three major processes are exposure to intercourse, exposure to conception, and gestation and parturition. Each of these can be further subdivided. Bongaarts (1982) simplified this framework noting that four factors account for much of the global variation in fertility. These four factors are the proportion married, use of contraception, incidence of abortion and breastfeeding. Extending this logic,
123
452
T. B. Heaton
religion will influence fertility if it influences these proximate determinants. This research will focus on two of these factors: marriage and contraception. Religious groups do not generally emphasize teachings that would directly influence breastfeeding. Although many religious groups do have something to say about the acceptability of abortion, accurate abortion data are difficult to obtain. This research will examine the degree to which religious differences in fertility can be explained by age at marriage, divorce and contraceptive use. Contraceptive use plays an important role in mediating the relationship between religion and fertility in developing countries. The most obvious reason for religious group differences in contraception is that members of some groups believe it is not acceptable to use birth control—the particularistic theology hypothesis. Religious congregations can provide a social network facilitating the spread of information about contraception. A Study in Mozambique found that in Catholic and Protestant congregations, socioculturally diverse and inclusive environments were more favorable for the spread and legitimization of modern contraception (Agadjanian 2001). Contraceptive practices vary widely in countries with a Muslim majority (Jones 2006; Hull 2005). In Thailand, the Muslim fertility is higher because Muslims are less likely to use contraception (Knodel et al. 1999). A study of Muslim and Hindu women in Southern India found that 99% of Muslim women did not think their religion permitted contraceptive use compared to 81% of Hindu women (Iyer 2002). Most of the women in both groups did not necessarily agree with their religion’s position on contraception and Muslims were less like to use contraception. The study also found that ‘‘the effect of religion may also be exercised through the local religious community’’ (Iyer 2002, p. 720). In Ghana, contraceptive use is highest among Protestant women, intermediate among Catholics, and lowest among Muslims and traditional groups, but most of these differences are attenuated when controls are added for demographic and socioeconomic characteristics (Addai 1999). Several other studies also find lower contraceptive use among Muslims in a variety of settings (Hogan and Biratu 2004; Akafuah and Sossou 2008; Gaur et al. 2008; Dharmalingam and Morgan 2004; Iyer 2002; Weeks 1988). Less attention has been given to the role of religion in promoting marriage or discouraging divorce. Cross-cultural comparison allows us to consider the degree to which religious group differences persist in different cultural and socioeconomic settings. Claims that the teachings and practices of a religious group motivate fertility in a uniform way imply little cross-cultural variation in religious group differences will be small. For example, a stringent test of the particularistic theology hypothesis would require that the religious group differences in fertility are evident cross-culturally. But some teaching and practices may be flexible and responsive to local conditions. Each of the factors mentioned above could exhibit cross-national variation. Indeed, Chamie’s interaction hypothesis, the minority group status hypothesis and the perspective emphasizing competition in pluralistic settings each directly state that religious differences will be greater in some contexts than in others. Social characteristics of religious groups may vary depending on the historical and geographical context for growth of that religion. Minority status is evident in some national contexts but not others. General tenets and practices of a global religious
123
Does Religion Influence Fertility in Developing Countries
453
community are interpreted in different ways by local congregations. Social ties within congregations may reinforce different norms in different settings. The perceived rewards and costs of having children associated with religious group membership may not be spatially uniform. Finally, the proximate determinants of fertility may show different relationships with religion in different contexts depending on patterns of marriage and contraceptive use. This paper will examine religious group differences in fertility by comparing Muslims with Christians, and then Catholics with Protestants. The paper will also include religious group comparisons in two of the key proximate determinants, namely age at first marriage and use of modern contraception. Analysis is based on 30 less developed countries. The preceding review of the literature suggests the following hypotheses: 1. 2. 3. 4. 5. 6. 7.
Muslim fertility is higher that Christian fertility Catholic fertility is higher than Protestant fertility Religious differences depend on the concentration of the religious group and the relationship may be nonlinear Religious differences depend on the level of development and the relationship may be nonlinear Religious differences are explained by social characteristics including education and urban residence The effects of education on fertility vary by religious group Religious differences are explained by proximate determinants of fertility including age at first union, use of modern contraception and marital disruption.
Comparison across countries also raises the question of how much variability exists across countries. In order to consider cross-national differences, this analysis uses multi-level modeling which explicitly calculates the variability in fertility within and between countries. This analysis implies two additional hypotheses: 8. 9.
Religious group differences in fertility will vary cross-nationally Inclusion of level of development, religious mix, social characteristics and proximate determinants in the model will account for the cross-national component of variability in fertility.
Data and Methods This study utilizes the Demographic and Health Surveys (www.measuredhs.com). The survey includes basic household information and a detailed interview with women in the childbearing ages. The research reported here focuses on the survey of women in reproductive ages because they are most able to provide active information about fertility. In some countries, the sampling frame included all women of reproductive age but in others, only women who have been in a union (legal marriage or consensual union) were included. To maintain comparability, analysis reported here is restricted to women who have been in a union. A vast majority of births in these countries occur to women who have been in a union. These surveys
123
454
T. B. Heaton
are designed to provide national level data on critical aspects of fertility and reproductive health. The core questionnaire and research methodology is comparable across countries, but countries have the option of excluding particular questions and including additional modules on specific topics. While these data offer the advantage of comparability, there are several limitations. Several countries opted not to include the question about religious group membership so they cannot be included here. Analysis is also constrained by the categories of religious groups included. Different countries used different coding schemes. Only a small group of countries had sufficient information on Hindus and Buddhists so these groups are not included. Coding schemes do not distinguish among various types of Muslims. On the other hand, Christians may or may not be divided into Catholic or Protestant and several different subcategories of Protestants are reported in different countries. To preserve anonymity and reduce sampling error, results are not reported for religious groups with fewer than 100 respondents. Our measure of fertility is the number of children born to respondents in the last 5 years. Limiting births to the last 5 years creates a measure that is current (analysis was replicated for total children ever born and major results are similar to those reported here). Age of the woman and age squared are included as a control variable because fertility increases in late adolescence, peaks in the mid-twenties and declines in later years. The level of development is measured by the human development index as calculated by the United Nations (hdr.undp.org). The relative size of religious groups is measured by aggregating the number in each group to the national level and dividing by the total number of respondents. Intermediate variables include age at first union, marital disruption (intact = 0, divorced or separated = 1) and ever having used modern contraception (no = 0, yes = 1). Social characteristics include education and rural/urban residence (rural = 0, urban = 1). Education is coded on a 6 point scale ranging from 0 for those with no education to 5 for those with post-secondary education. Religious group contrasts are complicated because each country has a different religious mix and includes different categories in the data. This makes it impossible to create a universal code that fits all countries. To deal with this complexity, this paper codes two basic contrasts—Muslims compared to Christians and Roman Catholics compared to Protestants. Several countries report more than one Protestant group. These groups are combined for the analysis reported here. Separate models are estimated for these two comparisons. Multi-level poisson regression is used to examine cross-national variation in the relationship between fertility and individual and country level characteristics. Poisson regression is appropriate for count data such as number of children born in the last 5 years. The basic form of this model is: LogðEðFertij jXij ÞÞ ¼ aj þ RbXij þ eij : a ¼ c0 þ RcZj þ fj where Fert is the fertility of the ith woman in country j, Xs are individual level variables and Zs are country level variables. The model postulates that individual level variables influence fertility and that country level variables affect country level
123
Does Religion Influence Fertility in Developing Countries
455
fertility. The model also specifies error at the individual level (e) and at the country level (f). The main questions of interest are: (1) how large is the b value for religion; (2) how much this value is altered when other variables are added; (3) do other variables have the hypothesized influence on religion; (4) how much cross-country variability is there in the b for religion; and (5) is this variability reduced by adding other variables to the model? The Stata procedure xtmepoisson is used to estimate these models.
Results Table 1 reports the contrasts for each country where the requisite religious categories are present and cell sizes meet the 100 person rule. These data indicate substantial religious group variation in fertility, social characteristics and proximate determinants both within and between countries. Average age at first marriage varies from the mid-teens to the early twenties, modern contraceptive use is almost nonexistent in some cases and near universal in others, and fertility ranges from .2 to over 1.0 children. The differences in fertility are graphed in Figs. 1 and 2. Figure 1 shows than Muslim fertility is generally higher than Christian fertility, but that there is substantial cross-national variation. For example in Kazakhstan the difference is about .3 children, and in Indonesia Muslims have lower fertility than Christians. There is not a clear geographic pattern in these differences. Asian and African countries are represented toward the top, middle and bottom of the graph. Figure 2 shows a somewhat different distribution for the Catholic/Protestant distribution. The top half of the graph looks like the mirror image of the bottom half. On average, the difference does not appear to be large, but cross-national variation is substantial. African and Latin American countries can be found at each end of the spectrum. Table 2 reports statistical analysis for the differences between Christians and Muslims. Numbers under model 1 show differences between Christians and Muslims with age of the mother and age squared as the only control. Results indicate that Christians fertility is about 8% lower than Muslim fertility (1 - e-.085 = .081), a substantial difference for national populations. The standard deviation for the Christian/Muslim difference is larger than the coefficient. In other words, the variability in the Christian/Muslim difference, as measured by the standard deviation of the coefficient across countries, is as larger than the average difference. These results are consistent with hypothesis #1 and hypothesis #8. Models 2 and 3 add national levels of development based on the Human Development Index (HDI). Each model includes HDI as a variable and results indicate that fertility is substantially lower in more developed countries. Model 2 examines the interaction hypothesis that religious differences will be greater at intermediate levels of development. Three dummy variables were created indicating low development (HDI \ .4), moderate development (.4 \ HDI \ .6), and high development (HDI [ .6) and each of these dummy variables is multiplied by the variable Christian. If the interaction hypothesis is accurate then the interaction terms will be small at low and high levels of development, but large at intermediate levels
123
456
T. B. Heaton
Table 1 Characteristics of religious groups in demographic and health surveys Country
Religion
(n)
Age at % Used Children % Education Ever first modern born Urban divorced marriage contraception \5 years
Indonesia, 2007
Muslim
(26185) 19.3
79.2
.56
41.3
2.54
6.4
62.4
.67
30.5
2.68
5.6
20.6
61.4
.63
23.0
2.37
6.1
(12234) 17.2
52.0
.70
52.8
1.37
5.9
20.0
45.7
.70
42.6
2.28
8.6
(1850)
21.2
77.9
.50
48.4
4.09
13.4
Christian
(1104)
20.3
86.2
.21
80.6
4.08
23.0
Catholic
(7350)
21.0
57.8
.74
52.8
3.38
6.4
Protestant (1190)
20.8
65.0
.75
47.4
3.31
4.0
Muslim
19.0
19.8
.99
33.9
2.30
5.1
Protestant (1309)
19.7
90.4
.46
85.4
2.36
13.0
Protestant (3598)
20.6
Catholic
(1406)
Muslim Christian
(7186)
Kazakhstan, 1999
Muslim
Philippines, 2003
India, 2006
Brazil, 1996
(525)
Catholic
(6756)
19.9
89.3
.56
78.3
2.37
13.8
Guatemala, 1995
Catholic
(5298)
18.0
28.0
1.06
30.4
1.02
11.1
Protestant (2409)
17.8
29.4
1.09
30.6
0.98
8.6
Haiti, 2006
Catholic
(3876)
18.9
57.2
.79
45.0
1.25
13.8
Protestant (3061)
20.0
53.8
.82
46.2
1.49
12.6
Muslim
(3287)
17.4
14.6
1.18
41.5
0.42
3.3
Catholic
Benin, 2006
Burkina Faso, 2003
Cote d’Ivoire, 2005
Ghana, 2003
Guinea, 2005 Mali, 2006 Nigeria, 2003
Senegal, 2005 Togo, 1998
123
(3755)
19.1
30.9
1.01
50.9
0.51
8.6
Protestant (3393)
18.4
25.6
1.11
35.7
0.49
5.3
Catholic
18.1
24.1
1.01
22.2
0.49
7.7
Protestant (510)
18.0
27.7
1.02
25.1
0.51
6.7
Muslim
(5705)
17.2
22.9
1.05
22.3
0.24
4.0
Catholic
(973)
21.0
–
.49
33.4
1.26
14.7 16.6
(2241)
Protestant (983)
20.4
–
.58
43.9
1.34
Muslim
(2717)
20.5
–
.56
49.1
0.43
6.5
Catholic
(625)
18.7
48.8
.80
27.5
1.67
11.7
Protestant (2277)
18.9
51.2
.85
43.6
1.95
14.9
Muslim
(801)
18.7
33.8
.99
34.2
0.65
4.6
Muslim
(5880)
16.2
15.7
.94
26.6
0.28
4.1
Christian
(533)
16.7
25.1
.87
24.6
0.62
12.0
Muslim
(11791) 16.6
19.2
1.10
34.6
0.39
4.1
Christian
(404)
17.3
26.7
1.02
28.0
0.66
3.5
Catholic
(653)
18.6
33.2
.96
40.4
1.95
9.7
Protestant (1626)
19.0
42.4
.98
38.9
2.30
8.8
Muslim
(3136)
15.6
12.3
1.14
35.9
0.69
5.0
Muslim
(10525) 17.4
22.6
1.00
37.6
0.45
6.6
Christian
(420)
18.7
43.1
.77
56.4
1.15
11.0
Muslim
(1056)
18.1
19.9
1.02
49.1
0.40
7.7
Catholic
(1058)
18.8
36.3
.89
40.5
1.19
10.5
Protestant (1109)
18.5
34.0
.93
37.2
1.06
11.8
Does Religion Influence Fertility in Developing Countries
457
Table 1 continued Country
Religion
Cameroon, 2004 Catholic
(n)
Age at % Used Children % Education Ever first modern born Urban divorced marriage contraception \5 years
(2975)
18.3
56.6
.90
51.5
2.01
13.1
Protestant (2695)
18.1
53.4
.90
44.5
1.94
12.7
Muslim
(1540)
15.2
16.1
1.07
43.3
0.60
8.1
Catholic
(1634)
16.9
14.5
.91
41.0
0.78
14.8
Protestant (2559)
17.0
10.5
.93
37.0
0.68
14.0
Muslim
(459)
15.8
9.8
.95
47.3
0.36
13.5
Catholic
(797)
16.6
11.9
1.15
36.0
0.80
10.2
Protestant (617)
16.5
17.5
1.16
43.0
1.13
12.3
Muslim
(3434)
15.5
5.5
1.09
49.4
0.26
13.2
Muslim
(15702) 19.3
76.4
.72
40.9
2.12
7.0
Christian
(820)
20.1
77.8
.61
54.4
2.46
5.3
Christian
(6408)
16.1
31.9
.89
28.3
.77
18.7
Muslim
(3590)
16.6
16.2
1.07
15.7
.27
10.5
Catholic
(1267)
19.0
57.1
.91
29.0
1.85
14.5
Protestant (3461)
19.2
62.8
.96
30.2
2.14
14.3
Muslim
(819)
17.6
20.4
1.14
39.9
.62
17.1
Madagascar, 2004
Catholic
(2323)
19.3
49.1
.82
68.3
2.13
19.5
Protestant (2315)
19.5
52.4
.80
66.0
2.26
15.8
Malawi, 2004
Catholic
(2075)
17.4
56.9
1.10
12.1
1.19
14.1
Protestant (5979)
17.4
54.5
1.08
13.2
1.13
14.1
Muslim
(1619)
17.1
40.3
1.13
10.2
.73
16.1
Catholic
(2630)
17.2
55.6
.95
37.1
.89
16.7
Protestant (4279)
17.7
48.6
.96
44.0
.79
20.6
Muslim
(1543)
16.3
53.0
.97
32.5
.59
13.0
Catholic
(849)
21.3
72.7
.79
41.2
1.85
15.6
22.4
77.6
.74
50.9
2.22
16.1
CAR, 1995
Chad, 2004
Egypt, 2005 Ethiopia, 2005 Kenya, 2003
Mozambique, 2003
Namibia, 2000
Protestant (2383) Rwanda, 2005
Uganda, 2007
Zambia, 2002
(3128)
20.3
25.0
1.15
17.5
1.09
22.2
Protestant (3548)
Catholic
19.8
21.2
1.23
20.2
.96
21.5
Muslim
(152)
19.3
46.7
1.20
58.6
1.39
27.6
Catholic
(2980)
17.7
34.1
1.26
10.4
1.02
17.1
Protestant (2641)
17.4
45.8
1.25
14.8
1.34
18.3
Muslim
(727)
17.0
52.8
1.24
27.8
1.44
14.9
Catholic
(1309)
17.5
49.6
1.11
29.3
1.59
19.3
Protestant (4351)
17.5
51.0
1.12
31.0
1.55
18.4
19.0
88.0
.66
39.3
2.43
19.7
18.7
84.5
.78
32.8
2.26
21.0
Zimbabwe, 2006 Catholic
(618)
Protestant (4960)
of development. Instead, the coefficients get progressively larger at higher levels of development. In other words, the Christian/Muslim difference is greater the higher the development. These data do not provide an adequate test of the theory since the
123
458
T. B. Heaton
Kazakhstan Philippines Senegal Kenya Ethopia Cameroon Nigeria Ghana Benin Togo Egypt Mali Guinea Burkina Faso CAR Coted' lvoire Rwanda India Madagascar Uganda Chad Indonesia
-0.20
-0.10
0.00
0.10
0.20
0.30
Fig. 1 Muslim/Christian difference in children born in the last 5 years
most developed countries are not included.1 Nevertheless, results suggest that the difference continues to increase even at relatively high levels of development. Model 3 simplifies the equation by adding a simple interaction between Christian and HDI. The coefficient for this term in consistent with the observation that the Christian/Muslim fertility gap increases at higher levels of development. Variation in the constant term is smaller when HDI is included, indicating that some of the variability in fertility across countries is attributable to level of development. On the other hand, variation in the coefficient for Christian does not change much when development is added, indicating that most of the national differences in the fertility gap are not explained by level of development. Model 4 is designed to test the minority group status hypothesis and the religious competition hypothesis. One interaction term is created by multiplying the national percent Christian by the variable Christian. If the minority group status hypothesis holds then this coefficient will be positive because the Christian/Muslim difference will be less negative when Muslims constitute a larger share of the population. A second interaction term is created by multiplying the percent Christian squared by 1
In a model that did not include an estimate of the variance in the coefficient for Christian, results were consistent with the interaction hypothesis. There are relatively few countries at the higher end of development so that it is difficult to estimate a variance component for Christian and the nonlinear relationship between development and the Christian/Muslim difference in fertility simultaneously.
123
Does Religion Influence Fertility in Developing Countries
459
Zimbabwe Rwanda Philippines Benin Haiti Guatemala Togo Coted' lvoire CAR lndonesia Burkina Faso Zambia Mozambique Cameroon Egypt Chad Kenya Uganda Malawi Madagascar Ghana Nigeria Namibia Brazil
-0.20
-0.10
0.00
0.10
0.20
0.30
Fig. 2 Protestant/Catholic difference in children born in the last 5 years
the dummy variable Christian. The competition hypothesis is supported if this term has a negative coefficient. A positive linear term and a negative squared term indicate and inverted u-shaped curve because the Christian/Muslim gap increases to a point and then begins to decrease. Coefficients are just the opposite of this prediction and are not statistically significant. In other words, these data provide little support for either the minority group or religious competition hypotheses. Models 5 and 6 include two of the key social characteristics that influence fertility, namely education and urban/rural residence. Consistent with prior research, fertility declines as the mother’s education increases and fertility is lower in urban areas. Of particular interest for this paper, the social characteristics account for a fraction of the Christian/Muslim difference. The coefficient for Christian in this model is two-thirds the size of the coefficient in Model 1. Education and urban residence account for some of the cross-national variation in fertility as well as for cross-national variation in the Christian/Muslim difference in fertility. Model 6 adds an interaction between religion and education. The coefficient for this interaction is negative, indicating that education plays a greater role in reducing Christian fertility than in reducing Muslim fertility. This interaction also accounts for a small portion of the cross-national variation in the Christian/Muslim difference in fertility. Model 7 adds measures of three of proximate determinants. Christian women tend to marry about one year later than Muslim women. There are also large
123
123
-.139*
Christian*high
.188*
27148*
Sd (Christian)
Wald (v2)
.233*
.150*
Sd (constant)
27165*
.144* 27167*
.141*
27149*
.148*
.195* 30044*
.125*
.209* 30278*
.114*
30256*
.142*
.227*
-.559*
Ever divorced
Random effects
-.001
Contraception
-.065*
-.005*
.292*
- 3.581*
7
-.078*
-.059*
-.023*
-.165*
.020
-.005*
.299*
-3.712*
6
Age at 1st union
.233*
-.051*
Education*Chris
-.167*
Education
.410
-.056*
-.005*
.299*
-3.675*
5
Urban
-.398
-.032
-.005*
.294*
-3.728*
4
%Christian*Chris
-.438
-1.089*
.137
-.005*
.294*
-3.173*
3
%Christian2*Chris
.187*
-.091*
Christian*mod
Christian*HDI
-1.096*
-.005*
-.016
Christian
Christian*low
-.005*
-.085*
Age2
.294*
-3.169*
2
HDI
-3.726*
.294*
Constant
Age
Effects of
1
Table 2 Multilevel poisson regression predicting children born in last 5 years: Christians compared to Muslims
32697*
.106*
.192*
-.514*
.067*
.041*
-.062*
-.029*
-.157*
.074
.010
.034
-.974*
-.005*
.291*
-3.151*
8
460 T. B. Heaton
Does Religion Influence Fertility in Developing Countries
461
Christian-Muslim differences in contraceptive use in several countries. In a few cases, Christians are 5 to six times more likely to have ever used modern contraception. The average across all countries is 2.26, indicating that Christians are more than twice as likely to have used contraception. On average the percentage divorced is about 1 third higher among Christians than among Muslims. We would expect Christians to have lower fertility because they tend to marry later, use contraception more and divorce more than Muslims. As expected, later marriage and marital disruption are associated with lower fertility. Women who have used modern contraceptives have higher fertility than women who have not. This is because women often delay contraceptive use until they have had children. Inclusion of these proximate determinants does not explain cross-national variation in fertility or in the Christian/Muslim difference in fertility. The final model includes all variables that were statistically significant in each of the prior models. As with prior models, national level of development, education, urban residence and proximate determinants each influence fertility. In this model, the coefficient for Christian is positive at each level of development, but is not statistically significant. Note that this model includes interactions with religion and development, and religion and education. The results imply that among the least educated women, Christian fertility is not lower than Muslim fertility. As the education increases, however, the religious gap increases. Although these variables help explain the average Catholic/Muslim difference, they only account for about one-third of cross-national variation in this difference. Table 3 considers Catholic/Protestant differences in fertility. The average difference in model 1 is essentially zero and the standard deviation is this difference across countries is also small. Subsequent models parallel those in Table 2. The difference between Christian and Muslim fertility is not closely related to level of development (models 2 and 3) or to the percentage of the population that is Protestant. While fertility is lower among more educated women and in urban areas (model 5), inclusion of these variables does not influence the magnitude of the Protestant/Catholic difference. Proximate determinants also influence fertility (model 7). After each of the statistically significant variables are included in the model, the coefficient for Protestant remains small and statistically not significant. In short, the overall difference in Catholic/Protestant fertility is small, regardless on the controls variables included in the model.
Conclusions This paper reports religious group differences in fertility in developing nations. These comparisons are important because fertility is high in many of these countries and religious diversity is increasing. Generalization about the relationship between religion and fertility requires cross-cultural data and several of the hypothesis regarding the nature of religious differences cannot be adequately tested without comparative data. Several of the hypotheses guiding this research receive at least partial support. Muslim fertility is higher than Christian fertility in most of the countries considered, but the magnitude of the difference varies considerably. The
123
123
-.014
Protestant*high
14393*
.020
Sd (Protestant)
Wald (v2)
.206*
Sd (constant)
14402*
.020
.183* 14402*
.020
.205* 14398*
.018
.168* 16782*
.000
16782*
.000
.168*
15974*
.016
.191*
18034*
.017
.172*
-.412*
-.437*
Random effects
Ever divorced
.041*
-.140*
.012
-.092*
-.178*
-.345
.005
-.005*
.261*
-2.911*
8
-.050*
.002
-.005*
.262*
-3.071*
7
Contraception
.001
-.088*
-.189*
.005
-.005*
.266*
-3.160*
6
Age at 1st union
Education*Prot
-.189* -.088*
Urban
.005
-.005*
.266*
-3.160*
5
Education
.303 -.238
-.082
-.005*
.256*
-3.225**
4
%Protestant2*Prot
-.081
-.738*
.045
-.005*
.256*
-2.838*
3
%Protestant*Prot
.183*
.011
Protestant*mod
Protestant*HDI
-.748*
-.005*
-.007
Protestant
HDI
-.005*
.003
Age2
.256*
-2.833*
2
Protestant*low
-3.223*
.256*
Constant
Age
Effects of
1
Table 3 Multilevel poisson models predicting children born in the last 5 years: Protestants compared to Catholics
462 T. B. Heaton
Does Religion Influence Fertility in Developing Countries
463
Catholic/Protestant difference is much smaller. Presence of religious group differences supports the conclusion that religion matters, but models that include interaction terms and estimate cross-national variation indicate that the nature of the relationship is complex. The Muslim/Christian difference grows wider at higher levels of development and at higher levels of educational achievement. These patterns suggest that changes in attitudes and behaviors that influence childbearing and that accompany socioeconomic progress are most likely to occur in Christian as compared to Muslim groups. Perhaps Islam’s emphasis on familial roles of women insulates Muslims from social changes associated with development. A more complete test of this hypothesis would require longitudinal data and more complete measures of variables that account for religious group difference. Level of development appears to be unrelated to the Catholic/Protestant difference in fertility. Social characteristics and proximate determinants do account for some of the religious difference in fertility between Muslims and Catholics. Religion is associated with level of education, type of residence, marriage timing, contraceptive use and divorce. These factors could arise from a variety of sources including historical circumstances accompanying the spread of religion, the type of people attracted to different religious groups, and teaching and institution building of the religious groups. These factors must be taken into account to completely understand the complexities of fertility differences. The data provide little support for the importance of relative group size in explaining fertility. It would appear that minority group status or religious group competition related to group size are not directly involved in decisions regarding family size. One of the most important conclusions from this analysis is that religious group differences in fertility vary dramatically across countries. In the case of the Christian/Muslim comparison, cross-national variability is as large as the average difference, and in the Catholic/Protestant comparison, cross-national variability also larger than the average difference. Efforts to account for this variability reported here are rather disappointing because most of the variability remains unexplained. Perhaps the grouping of religious groups is far to crude to capture distinctions within Islam and Protestantism. The mix of Protestant groups varies widely across countries such that including all Protestants in one category provides only the most basic comparison. DHS surveys do not include a measure of religious participation so it is not possible to distinguish between active participants and members who do not participate. Finally, historical conditions surrounding the initiation and spread of religion contain nuances that may difficult to incorporate in quantitative measures. Whether inclusion of a more complete set of controls could account for crossnational variation remains to be seen. Still, documentation of variability underscores the difficulty in making broad generalization about the nature of religious influence. These results have important implications for long-term changes in the relative size and composition of religious groups. Coefficients from Model 1 of Table 2 indicate that Muslim fertility is about 9% higher than Christian fertility. Given the caveats that this sample of countries is not necessarily representative of all developing countries and that conversion also plays and important role in the size of religious groups, results imply that Islam will experience higher growth than Christianity, but this will vary
123
464
T. B. Heaton
substantially across countries. In contrast, Roman Catholics and Protestants will maintain nearly equal shares of population due to the small overall difference in fertility. Another implication is that growth will be somewhat higher among less educated Muslims relative to other groups. This difference will add to the challenge of improving educational opportunities for Muslim populations. Several deficiencies limit the conclusions that can be drawn from this study. Although the sample includes countries from a wide array of developing countries, the sample is not random and is far from complete. The most developed countries are not included in the database. Several countries in DHS elected not to include the variable on religion. Hopefully, the demonstrated importance of religion will provide justification for the inclusion of religion in future DHS surveys. Measures of religious group are rather crude, and vary across countries. Moreover, the data do not include any measures of religious participation or strength of belief. Research in the United States indicates that belief and participation matter more that group affiliation. Despite these limitations, results do confirm broad differences and wide variability in religious group differences in fertility. They also suggest that hypotheses regarding the role of development, social characteristics and proximate determinants are important factors underlying religious group differences.
References Addai, I. (1999). Does religion matter in contraceptive use among Ghanaian women? Review of Religious Research, 40, 259–277. Adsera, A. (2006). Religion and changes in family-size norms in developed countries. Review of Religious Research, 47, 271–286. Agadjanian, V. (2001). Religion, social milieu, and the contraceptive revolution. Population Studies, 55, 135–148. Akafuah, R. A., & Marie-Antoinette, S. (2008). Attitudes toward and use of knowledge about family planning among Ghanaian men. International Journal of Men’s Health, 7, 109–121. Bongaarts, J. (1982). The fertility-inhibiting effects of the intermediate fertility variables. Studies in Family Planning, 13, 179–189. Caldwell, J. C. (1980). Mass education as a determinant of the timing of fertility decline. Population and Development Review, 6, 225–255. Chamie, J. (1981). Religion and fertility: Arab christian-muslim differentials. Cambridge: Cambridge University Press. Davis, K., & Blake, J. (1955). Social structure and fertility: An analytic framework. Economic Development and Cultural Change, 4, 211–235. Dharmalingam, A., & Philip Morgan, S. (2004). Pervasive Muslim-Hindu fertility differences in India. Demography, 41, 529–545. Faust, K., Bach, R., Gadalla, S., Khattab, H., & Gulick, J. (1991). Mass education, Islamic revival, and the population problem in Egypt. Journal of Comparative Family Studies, 22, 329–341. Frejka, T., & Charles, F. W. (2006). Religion, religiousness and fertility in the U.S. and in Europe. Max Planck Institute for Demographic Research Working Paper WP 2006-013. Gaur, D. R., Goel, M. K., & Goel, M. (2008). Contraceptive practices and related factors among females in predominantly rural Muslim area of North India. Internet Journal of World Health & Societal Politics, 5, 3. Goldscheider, C. (1971). Population modernization and social structure. Boston: Little-Brown. Heaton, T. B. (1989). Religious influences on Mormon fertility. Review of Religious Research, 30, 401–412.
123
Does Religion Influence Fertility in Developing Countries
465
Heaton, T. B., James, S., & Oheneba-Sakyi, Y. (2009). Religion and socioeconomic attainment in Ghana. Review of Religious Research, 51, 71–86. Hogan, D. P., & Biratu, B. (2004). Social identity and community effects on contraceptive use and intentions in Southern Ethiopia. Studies in Family Planning, 35, 79–90. Hull, T. H. (2005). People, population and policy in Indonesia. Jakarta and Singapore: Equinox Publishing. Iyer, S. (2002). Religion and the decision to use contraception in India. Journal for the Scientific Study of Religion, 41, 711–722. Jones, G. W. (2006). A demographic perspective on the Muslim World. Journal of Population Research, 23, 243–265. Knodel, J., Fray, R. S., Sricatcharin, P., & Peracca, S. (1999). Religion and reproduction: Muslims in Buddhist Thailand. Population Studies, 53, 149–164. Lehrer, E. L. (2004). Religion as a determinant of economic and demographic behavior in the United States. Population and Development Review, 30, 707–726. McQuillan, K. (2004). When does religion influence fertility? Population and Development Review, 30, 25–56. Morgan, S. P., Stash, S., Mason, K. O., & Smith, H. L. (2002). Muslim and non-Muslim differences in female autonomy and fertility: Evidence from four Asian countries. Population and Development Review, 28, 515–538. Mosher, W. D., & Hendershot, G. E. (1984). Religion and fertility: A replication. Demography, 21, 185–191. Stark, R., & Iannaccone, L. R. (1994). A supply-side reinterpretation of the ‘Secularization’ of Europe. Journal for the Scientific Study of Religion, 33, 76–88. Weeks, J. R. (1988). The demography of Islamic Nations. Population Bulletin. No. 43. Washington DC: Population Reference Bureau. Westoff, C. F., & Jones, E. F. (1979). The end of Catholic fertility. Demography, 16, 209–217. Zhang, L. (2008). Religious affiliation, religiosity, and male and female fertility. Demographic Research, 18, 234–257.
123
