Religion as a determinant of fertility rates

Religion as a Determinant in Fertility

Summary and suggestions

The studies reviewed and analyzed take a look at how religion affects birth and fertility in married or paired women throughout the world, Austria, India and Canada in particular. In Canada, minority women make up a large proportion of the population and represent many countries, with recent immigrants being in the majority, and therefore the Canadian study is most relevant to the determinants of the conclusion.

Most studies have found that women's fertility rate is determined by religion and other variants, sometimes equally, such as cultural, socioeconomic, age at time of birth, education, income and urban or rural position. These factors are found to be determinants as indirect causal mechanisms and their importance is borne out in the results of all studies.

Research Questions

Subjects of studies are asked their ages, desired number of children, numbers of live births, marital status, cultural group, racial group and religious denomination, as well as variations of these questions for different studies.

Hypothesis

The hypothesis with which this review and study is undertaken is that religion accounts for, in large part, the number of children born to a woman. Policy makers who wish to reduce the birth rates in their countries may look at the results of such a factor and cooperate with religious leaders in order to effect the desired objectives.

Literature Review

Socioeconomic Factors on Fertility Rate, Compared to Religion

After economic standing, a study by Lutz attempts to account for the effects of culture and religion on fertility rates. This study tries to assess the effects of culture and religion on fertility after accounting for a country's socioeconomic standing. Estimates for covariance models are analyzed for 128 countries between 1950 and 1975. The gross reproduction rate is the dependant variable and independent variables are infant mortality rates, aggregate standing of female education and gross domestic product per person. Categorical variables for religion and cultural region are taken into account as well. Surprisingly, even though socioeconomic standing is higher, European countries have lower fertility rates than Arab countries. Culture and religion have a measurable effect on fertility and in Arab countries it is positive. Catholicism's fertility rate is decreasing, as are other religions, yet Islam maintains a steady fertility rate. As measured by regional variables, the effect of culture on the level of national fertility has been increasing, while socioeconomic effects have decreased.

The socioeconomic indicators of 128 countries are related in Lutz' study by time series during 1975 and are related to gross reproduction rates (GRR) in multivariate and bivariate analysis. These 128 countries represent 97.4% of the world's population and in this study are compared on linear as well as logistic functional bases. China and other countries are grouped geographically and culturally, with variables for these in the equations. Short-term economic determinants are certain aspects of economic and sociological fertility analysis. Changes in the social system and how much norms are forced on the public were considered in the development of the analytical framework, as they effect short-term economic and sociological rates, which in turn affect fertility. Data is obtained from UN publications.

Differences in fertility rates are especially noticeable between less developed countries, as compared to more developed countries, so countries with GRRs above 2.0 and below are noted. "Explanatory values of the models of multivariate analysis are generally very high with the RZs ranging from 0.73-0.97, depending on the weighting used and on the specification of the model. Mutlicollinearity was reduced by transformation and aggregation of variables" explains Lutz.

Religion as a Factor in Fertility Rates Among Hindus and Muslims

In a study by Lutz, the female's age at time of birth was the most important variable in explaining the differences in fertility, yet no causal link in age and birth is assumed. Life expectancy was also a good indicator of the health and quality of life of the mother, which can change birth rates. The educational status of the female was the second most important variable in birth rates. Educational status of females compared to men also shows that social status is relevant for fertility. Other studies bears out the factor of education in lower birth rates. Income effect on fertility appears insignificant; however agricultural determinants were other economic factors included

According to Lutz, religion showed up as significant for differentials in fertility when all the other variables were equal. It appears that in Europe, North America, and China, fertility has a tendency to be lower than would be expected from the level of socioeconomic development when comparing all the countries culturally and geographically, whereas fertility is higher in Islamic and Central American countries. Table 5 of the Appendix shows a study of races, religions and other related factors in birth rates. Longitudinal comparisons show cultural factors have become increasingly important in different fertility rates from 1950 through 1975, while economic development became less important.

On the other hand, Rao, in an article called the communalization of population growth, claims it is "dangerous to attribute birth rates to religion. Some propaganda proposes the following myths: that Hindus have only one wife and Muslims many; that Islam forbids family planning and Hinduism does not; that Muslims have a higher birth rate." He supplies statistical evidence to refute these claims, showing the rate of polygynous marriage is 5.80% among Hindus and 5.73% among Muslims, meaning Muslims have a lower incidence. Temporary methods of contraception are allowed in Islam, called "fatwas, but it forbids sterilizations and abortions. In the Muslim countries of Turkey, Egypt, and Indonesia, contraceptives are much used. But Rao claims that when religion and monthly expenditures per capita are compared, that religion is not the significant variable. He shows that socioeconomic factors affect fertility, but Hindus and Muslims have a mix of socioeconomic groups. Indian population forecasts predict that Hindus will eventually outnumber Muslims. Muslim communities have been in decline in the urban and rural areas. He compares the Malabar region of Kerala and Uttar Pradesh, which shows that Malabar has a 40% Muslim population yet has a 15% lower birth rate than Muslims in Uttar Pradesh. Rao concludes "that the evidence does not support the myths; religion is not a primary determining factor."

Bangladesh Fertility Survey (BFS) data from 1975 were used by Chaudhury to test the hypotheses that "the higher the socioeconomic status, the lower the fertility and the narrower the difference in fertility between Muslims and Hindus; and the lower the socioeconomic status, the higher the fertility and the greater the difference in fertility between Muslims and Hindus." He used women married only once and reported to fecund as a core group, including 3914 Muslims and 824 Hindus. To analyze the data, he used multiple classification analysis (MCA). Before adjusting other variables, it was noted that the fertility of Muslims was a little lower than that of Hindus on a variant of 3.89 for Muslims to 3.95 for Hindus. Again, fertility declined with higher socioeconomic status. It increased in proportion to the level of education, being lower with no formal and primary level, and almost converging at the middle and higher educational levels. At lower levels of education (no formal and 1-5 grades), Muslims had.19-.34 children more than Hindus. This difference was reduced to.07 children at the middle level education, i.e., 6-9 grades. At higher levels of education, Muslims had.09 fewer children than Hindus, a significant statistic.

Before adjusting for the effect of other variables, Muslim fertility was higher than that of Hindus at almost age at marriage. This picture reversed when adjustment was made for other variables. At lower ages at marriage, Muslims has.23-.20 more children than Hindus. At the middle age at marriage, Muslims had.08 fewer children than Hindus. Again, the difference was significant statistically from almost no difference between the two groups at the higher age at marriage (20-21 years). Urbanization and education's effect on fertility differed for Hindus and Muslims. Fertility at each level of education was higher for Hindus in urban areas. For Muslims, it was lower in urban areas. While no difference was found in the use of contraception between Muslims and Hindus at higher levels of education, contraceptive use among Hindus was significantly higher than for Muslims at lower levels of education. This suggests that improvement in education will cause fertility to decline, as there is a corresponding increase in the use of contraception, and eventually differences in fertility and use of contraception between Muslims and Hindus will be erased. These findings also indicate that education is the best formula for reducing fertility in poor countries.

In India, Hindu and Muslim differentials in fertility were also studied by Balasubramanian, using census data and 11 surveys. An explanation of fertility differences was offered in that studies show urbanization and modernization trends will have greater effect on the fertility rate than religion. He noted in the general population the proportion of Muslims increased and the proportion of Hindus decreased before and after partition of the country. Afterward (between 1951-71) the proportion of Muslims increased from 9.9%-11.2% while the proportion of Hindus decreased from 84.9%-82.7%. He could not attribute the proportional changes to differences in migration or mortality, but to differences in fertility. Census and survey data show fertility is higher among Muslims than among Hindus, he claims. The total marital fertility rate for Muslim women was 11% higher in urban areas and 20% higher in rural areas than the rate for Hindus in those same areas, according to the 1971 census data. Even when education was controlled, the Muslim rate was higher. The 11 demographic survey findings reveal consistently higher fertility rates for Muslims than Hindus. When education and socioeconomic factors were controlled, studies demonstrated that differences narrowed yet remained.

Balasubramanian claims that one of three hypotheses is generally used to explain the differences. The first hypothesis attributes the fertility differences to background or socioeconomic characteristics of the two populations, but this explanation can not be supported when socioeconomic controls are introduced. The second hypothesis states that minority status is the cause of high fertility. But this hypothesis is easily refuted. In predominantly Muslim countries, Hindu minorities have lower fertilityrates. The third hypothesis claims fertility differences are due to religious beliefs concerning reproduction. However, both Islam and Hinduism are pronatalist religions, even though the two religions differ in regard to their beliefs concerning marriage, reproductive behavior, and fertility control. These differences have differing impacts on the variables influencing fertility. The intermediate variables include age at marriage, marriage stability, and contraceptive use.

In contrast to Hindu beliefs, Islamic beliefs support polygamy, allow for easy divorce in case of infertility, and allow widows to remarry. Yet Hindu women are more likely than Muslim women to return to their family for a period of time following the birth of their first and second children, and this reduces exposure to the risk of pregnancy for Hindu women. Muslim women are less receptive to family planning. A national survey indicated that 17% of Hindu women compared to 13% of Muslim women ever used contraception. Hindu women not only use family planning, but sterilization rates are higher among them.

Rural and Urban Effects on Fertility

Balasubramanian's goal was to demonstrate how, while positive, modern changes are occurring in India, religious impacts on intermediate variables will decrease, and religious differentials in fertility will decline.

A study bearing on this subject by Matin was done to determine whether in Bangladesh a wife's educational level significantly affects her desire for more children, or whether Bangladesh has not yet reached the threshold of socioeconomic development where education has any effect on the number of additional children desired. The analysis was carried out in a rural-urban context and also at different stages in the woman's life cycle, as defined by parity at time of interview. Two to three live births were considered early stage, and four or more were defined as later stage. The analysis was carried out on 2791 women selected from the Bangladesh Fertility Survey of 1975-76 who at the time of interview met the criteria of being married, fecund, parity of two or more, and having a husband present in the household. Matin explained:

The method used to perform the analysis was the Maximum Likelihood estimation method for the truncated dependent variable developed by J. Tobin. The method is a hybrid of probit analysis and classical regression and was chosen because the dependent variable, number of additional children desired, had a clustering of values at the lower limit of O. Separate regressions were run by life cycle stage (parity) and rural-urban background. The effect of variables such as the education of the husband, age, age at first marriage, work status of wife, number of living male children, and infant mortality were held constant in order to find the impact of education on desired fertility (Matin 41).

The resulting study found the number of additional children desired by rural wives was greater than that desired by urban wives at all stages of the life cycle. The mean number of additional children desired by a rural wife at the early stage of the life cycle was.57 and in the later stage.15. Urban women desired.29 and.12. Neither educational levels of wife and husband nor the age of the wife at first marriage were significant. The age of the woman was not significant in urban areas but was negatively significant in rural areas. The percent of dead children was positively significant in both urban and rural areas for early stage women but for later stage women only in rural areas. In rural areas also the number of living male children was positively significant. As a result of the study, the author believes Bangladesh is not at the developmental level where education affects desire for additional children and desire for fertility is more influenced by income, infant mortality, and preference for male children.

Fertility decisions and desires in Bangladesh

Sirageldin reported on two models that were developed to examine fertility behavior in Bangladesh. The first model dealt with the total number of ever-born children to a couple; the second examines sequential decisions that characterize the desire for an additional child. The "Chicago-Columbia" or "New Home Economics" approach was used, but to the usual economic variables were added sociological and demographic variables; and fertility was examined in relation to the prices of child services consumed as well as a valuation of the mother's time. The data for the study were drawn from a sample of 3088 currently married women respondents to the 1968/69 Impact Survey (an extended KAP survey). Whether the families were Hindu or Muslim was a factor in this study. Examination of this study is important as a control study and to determine if worldwide, other factors affecting fertility are similar to the effect that religion makes.

The model used four interior variables: total live births, number of dead children, current income, and female labor force participation. These were examined in terms of 14 exterior variables, including property ownership, age, literacy, awareness of family planning, rural vs. urban, type of family, size of family, and schooling. The model was built on four equations with parameters "estimated by two-stage least squares technic and then subjected to multivariate analysis" (Sirageldin 329-350).

The model for demand for additional children added five exterior variables including sex of children, desire for children, and perceived need for education of children. The model was researched using standard probit analysis. Interpretation of the two models showed that 1) Income was positively related to completed family size but has no effect on desire for additional children;

2) female education, female employment, and cost of fertility control had no effect in either model;

3) Age at marriage had a positive effect on completed family size but none on desire for additional children;

4) Urban women had more live births, but rural women were more likely to want additional children;

5) Sex preference for boys is intense in Bangladesh.

The study concludes that:

1) Economic well-being effects fertility;

2) the more adequate couples consider their income, the more likely they are to want more children;

3) Female education and employment have no effect on either completed family size or desire for more children;

4) There are no clear effects of family planning programs on either; and 5) desire for more children decreases as the number of children, particularly sons, increases.

Effect of Religion through Fertility Norms and Possibility of Dissolution

This paper develops two hypotheses concerning husbands' and wives' religious affiliations on fertility. The hypotheses are based on two central ideas. First, religions differ in their fertility norms and corresponding tradeoffs between the quality and quantity of children; differences in religious beliefs between husband and wife may thus lead to conflict regarding fertility decisions and possible resolution through bargaining. Second, if the spouses were not religiously compatible, there was a higher probability of marital dissolution and thereby the possibility of having children. Data from the 1987-1988 National Survey of Families and Households conducted in the U.S. suggest that both of these effects help explain the observed links between religious composition of unions and fertility.

The relationship between religion and fertility: Evidence from Austria

Data from the Austrian Family and Fertility Survey were used to examine for the first time the relationship between religion and fertility in first unions in Austria. Although Austria is a Catholic country, results from a Poisson hurdle model show that both women's denominational affiliation and religiosity affect the number of children born. Simply whether unions have religious composition does not result in clear evidence. Furthermore religion has is no effect of on the timing of births. There is a puzzle to be made from the results, however: Females and unions of other than Catholic or no religious affiliation have a higher transition rate to third birth.

The Religious Denomination Factor in Fertility in Canada

The data in the sample revealed substantial differences between visible minority groups in the area of religious beliefs (Table 1). For example, most Chinese and Japanese women recorded that they had no religion. Islam was the religion most often recorded by Arab women, and more than half of South Asian women recorded they were Sikhs or Hindus. Could the low fertility of the former and the high fertility of the latter be related to those differences in religion, as suggested by the particularized ideology hypothesis?

The Census recorded total fertility rates of major religious denominations in 2000- 2001, indicating that religious groups do differ in fertility. Unfortunately, unlike the question on visible minority groups, the religion question was not asked in the 1996 Census. The most fertile women were Muslims and Hindus, at 2.41 and 2.00 children per woman respectively. In contrast, Buddhists (1.34), Orthodox Christians (1.35) and women with no religion (1.41) had the lowest fertility rates, at 1.36 children per woman on average. While the two extremes, Sikhs, Jews and "other Christians" exhibited fairly high fertility, Protestants and Catholics were close to the national total fertility rate of 1.57. This category mostly includes persons who reported they were "Christian," without providing further detail.

It should be noted that the results may be connected with the frequency of religious practice; based on data from the Ethnic Diversity Survey, which presented percentages of people who attend religious services at least once a month for each religious group. The denominations with the largest proportions of followers who attend services regularly, the Sikh, Hindu and other Christian denominations, are among the groups with the highest fertility, while Buddhists, Orthodox Christians and Catholics have the lowest proportions of followers who practice their faith regularly and, surprisingly, the lowest fertility rates. Therefore, the possibility that religion exercises its effect on fertility at least partly through religious practice cannot be ruled out -- provided, of course, the effect actually exists and is not masking the effects of underlying variables. Muslim women are a notable exception, as their fertility is high while their religious attendance is not much higher than average; it would be interesting to explore this phenomenon in subsequent research. In any case, it seems at first glance possible to establish a link between the fertility of some visible minority groups and that of religious groups. For example, the Arab and South Asian groups, the most fertile visible minority groups, include a large proportion of women who belong to the religious groups with the highest fertility rates (Muslim, Hindu and Sikh). Women with no religion showed fertility lower than the average for the total population made up major proportion of the Japanese and Chinese groups, who were the least fertile visible minority groups in 2000-2001.

On the other hand, it does not seem possible to generalize this relationship to all visible minority groups. Some seemingly dichotic statistics resulted. Latin American and Filipino women have above-average fertility. Most of them are Catholic, yet this denomination's fertility is slightly below the national average. About half of all Southeast Asian women report that they are followers of the least fertile religion (Buddhism), yet the fertility of Southeast Asian women is higher than the national average. West Asian women, for whom Islam is the major religion, have a fertility rate similar to that of white women.

Results of the Multivariate Analysis

The foregoing findings suggest that religious composition may account for part of the resulting figures on fertility in some visible minority groups. But visible minority groups also differ from one another, and from the rest of the population in their composition with respect to other fertility-related demographic and socio-economic characteristics (Table 1, Appendix).

Some other surprising results were: In South Asian and Arab groups, about 60% of women aged 15 to 49 were married in 2001, compared with approximately 44% in the white group and less than 30% in the Black group. Table 1 also shows that about 40% of Arab and West Asian women were members of a household below the low-income cut-off, compared with roughly 14% of Japanese and white women.

A second point of interest is the fact that though recent immigrants made up more than half the population of Filipino, Arab, Korean and West Asian women, a quarter of all Japanese women, and only 3% of all white women. Additionally, many visible minority women tended to attend school full time, to have a university degree, and to live in a census metropolitan area, though in these cases too, significant variations between visible minority groups was evident.

In another analysis, the results of a logistic regression model estimating the fertility of visible minority groups based on a combination of religious denomination and various fertility-related demographic and socio-economic factors was made. The model by progressively adding the control variables to a base model that included only the visible minority groups, age groups and number of children one year earlier (parity). The model's dependent variable is the probability that a woman aged 15 to 49 is living with at least one Canadian-born child under the age of 1. That variable is interpreted as a measure of fertility.

When there was control for variables such as religion, marital status, income, period of immigration and generation of immigrants, a number of visible minority groups continued to exhibit fertility differences from other minority women and from the white women. All other things being equal, Black, Arab, Filipino and Latin American women are significantly more fertile than white women (the reference group), and had a 65%, 25%, 19% and 14% greater chance respectively of having given birth to at least one child in the year preceding the 2001 Census. Chinese, Korean and West Asian women are less fertile than white women, as their odds of having given birth in 2000-2001 are 18%, 24% and 34% lower respectively. Aboriginal women are 56% more fertile than the reference group, even though Indian reserves. Three territories (where fertility is higher) were excluded from the sample. (See Tables 6 and 7) Controlling for demographic and socio-economic composition alters the fertility differentials revealed by the total fertility rates. The various models in the study showed that introduction of control variables generated three distinct effects: in Japanese, Southeast Asian and South Asian women it erased fertility differences with the reference group; in Arab women it diminishd the differences without erasing them completely; and in the case of West Asian and Black women it accentuated fertility differences, making them the least and most fertile groups respectively in 2000-2001.

Discussion

How fertility gaps between visible minority groups found in the descriptive analysis be accounted for? The characteristics hypothesis, which assumes that the fertility differences are due to the effects of underlying socio-economic differences in comparable groups, seems to work for three groups: Japanese, South Asian and Southeast Asian. The low fertility of the Japanese, and the higher fertility of the South Asian and Southeast Asian groups disappeared completely when parity, religion, marital status and income variables are introduced as controls. It also explained part of the high fertility rate of Arab women: the gap between them and the reference group narrowed but did not disappear as variables are phased in to the model. This means that if demographic and socio-economic characteristics of Japanese, South Asian, Southeast Asian and Arab women are identical to those of white women, we could expect fertility of Japanese, South Asian and Southeast Asian groups would not be significantly different from the fertility of the white group and that Arab women would be less fertile than they currently are, though they would still be more fertile than white women.

The results in table 5 might suggest that the characteristics hypothesis also works to some extent for Filipino and Latin American women, whose fertility advantage shrinks when their socio-economic composition is taken into account. However, in both cases, the odds ratio in the full model is not significantly different from the odds ratio in the model that controls only for the effects of age and parity. It may be appropriate in essence, to say that socio-economic characteristics of Filipino and Latin American women do not explain, even marginally, the fact that they are more fertile than white women, while in European countries, this is not true.

Studies also show that relationships between fertility of visible minority groups and the fertility of major religious denominations are less evident than the descriptive data seemed to suggest. The higher fertility of Arab and South Asian women appears to correspond with the higher fertility of Muslim, Hindu and Sikh women, while below-average fertility in Japanese and Chinese women seemed to be related to the below-average fertility of women with no religion.

When we control for visible minority group, religion and other socio-economic variables at the same time, Arab women still remain more fertile than white women, while Chinese women remain less fertile, suggesting that their respective fertility may not be due to their religious beliefs. In the same way, the fertility of South Asian women is not explained by the high proportion of Sikhs or Hindus in the group, since the fertility of these two religious groups is no different from the fertility of the Catholic group when controlled for marital status. In fact, it may be possible that the high proportion of married people in the Sikh and Hindu groups is due to religious beliefs or traditions. If so, religion would be a major factor, with the effect mediated by marital status (See Tables 2 and 3).

In addition, it was found that the fertility of Japanese women was not significantly different from that of white women. Since religion was not considered in that model, this suggests that their low fertility may not be due to religious composition. In fact, five religious denominations continued to present fertility differences with Catholics: Jewish, Muslim and other Christian women, who shared a higher fertility, while Orthodox Christians and women without religious affiliations had a lower fertility than Catholics.

Summary and Suggestions model has been found to suggest that religion plays a part in the fertility rates of women in Canada and India, along with socioeconomic factors and other variables. Data from the tables in the Index demonstrate the scattered effect that is evident from the studies done. The trend in the tables, however, demonstrates a strong affinity to race and culture. Yet part of culture is the religion endemic to that culture, and therefore it may be concluded that religion affects fertility rates, as much as other variables do.

Suggestions for further action are as follows:

1. Matin and others recommend that measures be taken to improve socioeconomic levels and provide employment opportunities for women.

2. It is also suggested that additional analysis be done to identify the factors important in determining fertility decisions in urban areas and that separate policies be adopted for rural and urban areas.

3. Further research may also be done by examining the desired number of children by individuals' religion. This may help to understand whether the reasoning of pronatalist religious ideologies will hold or not hold.

APPENDIX

Looking at both individuals' denominational affiliation and their religious belief in more detail, Table 1 indicates that individuals with religion on average have some 0.5 children more than individuals without religious affiliation: Catholics, Protestants and individuals with other religion have about 1.7 children whereas individuals with no religion have about 1.2 children. Furthermore, the distribution of the differences between observed and expected frequencies shows no clear pattern for Protestants and females with other religion. However, contrasting Catholic women and women without denominational affiliation, there is a negative difference between observed and expected frequencies for the zero-birth outcomes and positive differences for the non-negative birth outcomes. This indicates that Catholic women are less likely than expected to have zero births and more likely than expected to have positive and particularly higher order birth outcomes. A reversed picture shows for women without denominational affiliation: There are large positive differences between observed and expected frequencies for both the zero-births and the one-child outcomes and large negative differences for the two and more children outcomes. This therefore indicates that women with no religion are more likely than expected to either have zero births or one child only and that they are less likely than expected to have positive and higher order birth outcomes. The Chi2-test statistic of 134.48 suggests rejecting the assumption of independence of religious affiliation and the number of births.

Differentiating for individuals' religious belief, the findings presented in the lower part of Table 1 also suggest for a relationship between religiosity and birth outcomes. On average, there is monotonic decline in the number of births by religious belief: On the one side of the spectrum, individuals with a strong religious belief have almost 2.1 children, while on the other side of the spectrum, the average number children decreases to some 1.2 children for women who do not have a religious belief. As for the differences between observed and expected number of frequencies of births, Table 1 furthermore indicates that strong religious believers are less likely than expected to have zero births or one child only, while they are more likely than expected to particularly have higher order birth outcomes, i.e. three and more children. While the findings is not clear cut for individuals with some religious belief, non-believers and strong nonbelievers are more likely than expected to either have zero births or one child only and that they are less likely than expected to have two or more children. Complementary to the finding for religious affiliation, the Chi2-test statistic of 43.51 also suggests rejecting the assumption of independence of religious belief and the number of births" (Heineck 9).