%20-%20https://sketch.com%20--%3e%3ctitle%3e3ECBBF50-FD08-412A-943D-2CFC5E025A16%3c/title%3e%3cdesc%3eCreated%20with%20sketchtool.%3c/desc%3e%3cg%20id='Stylesheet'%20stroke='none'%20stroke-width='1'%20fill='none'%20fill-rule='evenodd'%3e%3cg%20transform='translate(-926.000000,%20-668.000000)'%20id='BUTTONS'%3e%3cg%20transform='translate(100.000000,%20597.000000)'%3e%3cg%20id='BTN-/-RIGHTS'%20transform='translate(816.000000,%2065.000000)'%3e%3cg%20id='ICON-/-Rights'%20transform='translate(10.000000,%206.000000)'%3e%3cpath%20d='M13.5752841,0%20L1.34943182,0%20C0.59375,0%200,0.586666667%200,1.33333333%20L0,22.6666667%20C0,23.4133333%200.620738636,24%201.34943182,24%20L17.6505682,24%20C18.40625,24%2019,23.3866667%2019,22.6666667%20L19,5.36%20L13.5752841,0%20Z%20M13.5752841,1.89333333%20L17.0838068,5.36%20L13.5752841,5.36%20L13.5752841,1.89333333%20Z%20M1.34943182,22.6666667%20L1.34943182,1.33333333%20L12.1988636,1.33333333%20L12.1988636,5.36%20C12.1988636,6.10666667%2012.7926136,6.69333333%2013.5482955,6.69333333%20L17.6235795,6.69333333%20L17.6505682,22.6666667%20L1.34943182,22.6666667%20Z'%20id='Shape'%20fill='%23595959'%20fill-rule='nonzero'%3e%3c/path%3e%3ccircle%20id='Oval'%20stroke='%23595959'%20stroke-width='1.5'%20fill='%23595959'%20cx='9.5'%20cy='13.5355339'%20r='5.5'%3e%3c/circle%3e%3cpolyline%20id='Path-2'%20stroke='%23FFFFFF'%20stroke-width='1.5'%20transform='translate(9.500000,%2013.035534)%20rotate(45.000000)%20translate(-9.500000,%20-13.035534)%20'%20points='8%2016.0355339%2011%2016.0355339%2011%2010.0355339'%3e%3c/polyline%3e%3c/g%3e%3c/g%3e%3c/g%3e%3c/g%3e%3c/g%3e%3c/svg%3e)
Introduction
This paper uses unique rich survey data to examine the association between religious affiliation and under-five mortality in Mozambique, a country of 23 million people where 142 children out of 1000 new-borns die before reaching the age of five years (WHO, 2011) and about 90% of the population say religion is very important in their lives and most people are affiliated to organized religion (Pew Research Center, 2010). Mozambique is typical of the subcontinent in levels of infant and child mortality. In general, sub-Saharan Africa is the region of the world with the highest under-five mortality rates (WHO, 2011). The United Nations points out that although most causes of child deaths are preventable and treatable, more children die every year in sub-Saharan Africa than anywhere else in the world: of the 8.8 million under-five deaths in the world in 2008, half were in sub-Saharan Africa (United Nations, 2010).
The role of organized religion in Mozambique is also typical of sub-Saharan Africa. Thus, a multi-country survey conducted at the end of the last century showed that in West Africa about 99% of respondents belonged to a religious denomination and 82% attended religious services regularly (Gallup International, 2010). In east and southern Africa the levels of religious involvement are similarly high: approximately 92% reported to belong to a religious congregation in Malawi (Trinitapoli, Reference Trinitapoli2006), about 70% belonged to a Christian denomination in Zambia (Carmody, Reference Carmody2003) and in South Africa about 90% of the black population identified themselves with a Christian church (Garner, Reference Garner2000).
Mozambique also resembles sub-Saharan Africa in Christian denominational diversity. These diverse Christian denominations can be roughly divided into two main groups: the ‘mainstream’ denominations and Pentecostal and similar denominations. The first group encompasses Catholic and mainstream Protestant churches (e.g. Anglican, Baptist, Methodist and Presbyterian). The second group includes imported Pentecostal churches (e.g. the Assemblies of God), and numerous similar but locally initiated churches such as Apostolic and Zionist, whose doctrine and practice are focused on faith-healing. The Catholic and mainstream Protestant churches are historically more connected to institutions of health care and education; they tend to have more socially diverse and better educated members (Addai, Reference Addai2000; Garner, Reference Garner2000; Agadjanian, Reference Agadjanian2001; Takyi & Addai, Reference Takyi and Addai2002; Gyimah et al., Reference Gyimah, Takyi and Addai2006) with better household living conditions (Gyimah et al., Reference Gyimah, Takyi and Addai2006) compared with members of Pentecostal and especially Zionist churches (Schoffeleers, Reference Schoffeleers1991; Agadjanian, Reference Agadjanian2001; Pfeiffer, Reference Pfeiffer2002). Although Apostolic churches share many characteristics with Pentecostals and Zionists, they are characterized by a close-knit and co-operative organization (e.g. Turner, Reference Turner1980; Bourdillon, Reference Bourdillon1983; Jules-Rosette, Reference Jules-Rosette1997), which sets them apart from much more loosely organized Zionists and other Pentecostal churches. Despite socioeconomic differences among Christian denominations, members of sub-Saharan Christian churches as a whole tend to be better educated and wealthier than individuals belonging to Islam or traditional religions, or those not affiliated to a formal religion (Kollehlon, Reference Kollehlon1994; Takyi & Addai, Reference Takyi and Addai2002; Gyimah et al., Reference Gyimah, Takyi and Addai2006; Antai et al., Reference Antai, Ghilagaber, Wedrén and Macassa2009).
There has been little research on the effect of religion on child health and mortality in the region. The few existing studies about religion and child survival in sub-Saharan Africa suggest that the effect of religion on under-five mortality is typically explained by differences in maternal education and differential use of maternal and child health services among members of different religious groups. Gyimah (Reference Gyimah2007) investigated differences in child survival by religious affiliation using the 1998 and 2003 Ghana Demographic and Health Surveys. The bivariate analysis showed that children of Muslim women and of those belonging to traditional religion were at a significantly higher risk of dying compared with children of women belonging to mainstream Christian denominations, but there was no difference in the risk of death of children between mothers affiliated to mainstream Christian denominations and other Christian denominations. However, the association between religious affiliation and child survival observed at the bivariate level disappeared after controlling for socioeconomic factors, especially education.
Another study in Nigeria concluded that the association between belonging to a traditional religion and under-five mortality was explained by religion-related differences in use of maternal and child health services (Antai et al., Reference Antai, Ghilagaber, Wedrén and Macassa2009). In Zimbabwe, Gregson et al. (Reference Gregson, Zhumu, Anderson and Chandiwana1999) observed that members of Zionist and Apostolic churches historically showed higher infant mortality than members of mainstream (‘Mission’) churches, presumably because Zionist and Apostolic churches discouraged the use of both traditional and modern medical services. The authors also noted that the disadvantage of Zionist and Apostolic church members in infant mortality in Zimbabwe has diminished since the early 1990s, probably owing to enforced immunization.
Several studies have looked at the association between religious affiliation and child survival in other parts of the world, notably in Latin America. Using the 2000 Demographic Census of Brazil, Wood et al. (Reference Wood, Williams and Chijiwa2007) examined the association between religious affiliation and child deaths to mothers aged 20–34 years in north-east Brazil. They found a significantly lower death rate among children of women affiliated to traditional Protestant (Baptist and Presbyterian) and Pentecostal denominations when compared with children of Catholic women. This difference was attributed mainly to socioeconomic differences between the religious denominations. According to the study, compared with Catholics, traditional Protestants in north-east Brazil are characterized by higher levels of education and household income, and are more likely to be married and to live in households with piped water. Bivariate results of another study extended to the whole of Brazil found significantly fewer infant deaths among Protestant mothers when compared with Catholic ones; however, after adjusting for socioeconomic and demographic factors, denominational differences in infant deaths became statistically non-significant (Verona et al., Reference Verona, Hummer, Júnior and De Lima2010). The findings from both Brazilian studies illustrate the importance of socioeconomic differences across religious denominations in explaining child deaths.
Another study in Mexico (Valle et al., Reference Valle, Fernández and Potter2009) investigated the effect of religious affiliation of the mother on child mortality among women belonging to indigenous and non-indigenous ethnic groups in Chiapas. For the indigenous group, the study reported significantly lower child mortality among mothers affiliated to the Presbyterian denomination compared with mothers belonging to the Catholic Church, net of demographic and socioeconomic characteristics of the mother. Among non-indigenous women, religious affiliation of the mother was not significantly associated with child mortality in Chiapas. The authors suggested two possible explanations for their results: Presbyterian women are more likely to use health care services than are Catholic women, and there is greater social integration and organization among Presbyterian congregations compared with Catholic ones. The researchers further noted that indigenous health promoters among Presbyterians provided members with health education and helped members with referral to public and private health services. Overall, the literature on the relationship between religious affiliation and child survival appears to suggest that differences in child survival among religious groups are mostly explained by socioeconomic characteristics of those groups and religion-related differences in use of health care services.
The setting
Mozambique, located in southern Africa, won independence from Portugal in 1975. Soon after independence it went through a prolonged civil war (1976–1992) that destroyed much of the socioeconomic infrastructure of the country (Minter, Reference Minter1994; Abrahamsson & Nilsson, Reference Abrahamsson and Nilsson1995). After the signing of a peace agreement and the end of civil war in October 1992, the country embarked in a post-war reconstruction effort, leading to an average economic growth in excess of 7% per year between 1997 and 2010 (UN Mozambique, 2012).
Despite its remarkable economic growth, Mozambique remains one of the poorest countries in the world, with a Gross National Income per capita of US$440 and life expectancy at birth of 50 years in 2010 (World Bank, 2012). It is estimated that the country has more than half of its people living below the national poverty line (Republic of Mozambique, 2010). Although the under-five mortality rate in Mozambique has shown a dramatic decline in the last two decades, from 232 deaths per 1000 live births in 1990 to 142 deaths per 1000 live births in 2009, it remains one of the highest in the world (WHO, 2011). The country's poor health indicators exist in an environment of low literacy, particularly among women: the 2003 Mozambique's Demographic and Health Survey report indicated that 41% of women (aged 15–49 years) and 17% of men (aged 15–64 years) were illiterate among those surveyed (Instituto Nacional de Estatística & Macro International, 2005).
As noted above, most people in Mozambique identify themselves with a religious congregation: among women aged 15–49 years interviewed in the 2003 Demographic and Health Survey, 30.3% were Catholics, 27.2% Protestant or Evangelical and 18.8% Muslims. Women affiliated with Zionist denominations made up about 8.8%, while women who were not affiliated with any religious denomination constituted 14.5% (Instituto Nacional de Estatística & Macro International, 2005). The national religious distribution masks regional differences, with Muslims primarily being concentrated in the north and the coastal strip of central Mozambique, while Zionists are heavily present in the south.
The survey that supplied data for this study was carried out in Chibuto District of the southern province of Gaza (Fig. 1). In 2007 the district had a population estimated at 165,000 inhabitants (Instituto Nacional de Estatística, 2009), and one-third of that population lived in the district capital. Rain-dependent agriculture and remittances from migrants working in the Republic of South Africa and the capital of Mozambique, Maputo, are the basis of the district economy. The district's population is predominantly Christian; the Catholic church, mainstream Protestant churches, Apostolic and numerous small Pentecostal-type churches (mainly Zionist churches and the Assemblies of God) are the prevailing religious organizations (Agadjanian, Reference Agadjanian2005; Agadjanian & Menjívar, Reference Agadjanian and Menjívar2008).
The location of Chibuto District, Gaza Province, Mozambique.
Hypotheses
Following previous studies in sub-Saharan Africa (e.g. Gyimah, Reference Gyimah2007; Antai et al., Reference Antai, Ghilagaber, Wedrén and Macassa2009), it is expected that under-five mortality of children of mothers affiliated to a religious denomination will be lower than mortality of children of non-affiliated mothers, but this difference will be due to other factors such as higher average level of education and higher living standard of affiliated women.
Recognizing that religious denominations are diverse (Agadjanian, Reference Agadjanian2005; Mpofu et al., Reference Mpofu, Dune, Hallfors, Mapfumo, Mutepfa and January2011), denominational differences in under-five mortality are expected. Therefore, mortality of children of non-affiliated mothers is compared with mortality of children of mothers affiliated to churches in four denominational groups: Catholic or mainstream Protestant (treated as one group), Apostolic, Zionist and other Pentecostal. It is posited that children of women belonging to Catholic or mainstream Protestant denominations, when compared with children of women not affiliated to an organized religion, will have lower under-five mortality because Catholics and mainstream Protestants are typically better educated and have higher income and better living conditions than members of most other religious groups (Garner, Reference Garner2000; Agadjanian, Reference Agadjanian2001; Takyi & Addai, Reference Takyi and Addai2002; Gyimah et al., Reference Gyimah, Takyi and Addai2006). In addition, Catholic and mainstream Protestant denominations are said to have better access to health services and to encourage the use of these services (Gregson et al., Reference Gregson, Zhumu, Anderson and Chandiwana1999; Agadjanian, Reference Agadjanian2005), which may contribute to child survival.
It is also expected that children born to Apostolic women will have lower under-five mortality when compared with children of non-affiliated women. Religious teachings among Apostolics encourage a co-operative and mutual help life-style (Turner, Reference Turner1980; Bourdillon, Reference Bourdillon1983; Jules-Rosette, Reference Jules-Rosette1997), which is expected to contribute to child survival beyond mother's education and household living conditions. While all Christian churches in general stress the importance of organizational solidarity, the emphasis on mutual social support among Apostolics appears to be most pronounced. Although the effect of mutual help on child survival cannot be directly tested with these data, based on the literature (Jarvis & Northcott, Reference Jarvis and Northcott1987; Taylor & Chatters, Reference Taylor and Chatters1988; Ellison & George, Reference Ellison and George1994; George et al., Reference George, Ellison and Larson2002; Hummer et al., Reference Hummer, Ellison, Rogers, Moulton and Romero2004) it seems reasonable to expect that strong mutual support and help could improve child survival chances among Apostolics.
With respect to children of mothers belonging to Zionist and other Pentecostal denominations, it is expected that their survival advantage over children of non-affiliated mothers will be minor if at all present. Zionist and most Pentecostal denominations are among the poorer religious organizations in southern Africa and do not have the kind of strong social support mechanisms that characterize Apostolics (Gregson et al., Reference Gregson, Zhumu, Anderson and Chandiwana1999; Agadjanian, Reference Agadjanian2005).
Data and Methods
Data
The study used data from a survey conducted in Chibuto District in 2008. The survey was a representative cluster survey of 2019 women aged 18–50 years, both affiliated and not affiliated to a religious denomination. In 82 randomly selected communities (clusters) located in both urban and rural areas of the district, the survey collected information on respondents' complete religious affiliation history since birth until the year of the survey. The survey also collected a variety of information, including information on respondents' demographic and socioeconomic characteristics and complete reproductive histories (births and deaths of all of their children). The survey is therefore unique for sub-Saharan Africa in the amount of details about respondents' religious and reproductive history.
The outcome variable was the hazard of death before age of 5 years. For every child, the risk period begins at the time of birth. The risk period ends when the child dies, completes 5 years or at the time of the survey for children younger than 5 years who were still alive by the survey date. A person-year file in which each child contributes one observation for every year of life before the end of the risk period was built. For children who were still alive by the survey date, the outcome variable was coded 0 in each year of exposure. Children who died between ages 0 and 5 were coded 0 for each year of life before death and coded 1 in the year of death. Each child stops contributing person-years when the risk period ends. This is a typical data restructuring for discrete-time event-history models (Allison, Reference Allison1995). The age of children was operationalized as a set of four dummy variables to capture the variation of mortality risk with age: 0 years, 1 year, 2–3 years, and 4 to exactly 5 years. Children with incomplete birth and death information were excluded from the analysis.
The main predictor in this study was time-varying mother's religious affiliation. First, a distinction was made between women affiliated to an organized religion and non-affiliated women. Second, religious affiliation was categorized as follows: (1) Catholic or mainstream Protestant; (2) Apostolic; (3) Zionist; (4) other Pentecostal (Assembly of God included); and (5) non-affiliated. These categories were derived from respondents' religious history. For each year of observation, religious affiliation of the respondent was lagged by one year. There were only twelve Muslim women in the survey sample, and they were excluded from the analysis.
As controls, socio-demographic variables typically reported to affect child survival were included (Cleland & Sathar, Reference Cleland and Sathar1984; Brockerhoff & Derose, Reference Brockerhoff and Derose1996; Omariba et al., Reference Omariba, Beaujot and Rajulton2007): mother's age, number of previously born children, marital status and the length of preceding birth interval (all time-varying). The length of preceding birth interval was operationalized as follows: less than 2 years and 2 or more years. Mother's education at the time of survey was controlled for as in this setting women typically complete their education before having children. Mother's education has been found to affect the risk of child survival (e.g. Farah & Preston, Reference Farah and Preston1982; Agha, Reference Agha2000; Schellenberg et al., Reference Schellenberg, Nathan, Abdulla, Mukasa, Marchant and Tanner2002). It was coded as follows: no education, 1–4 years and 5 or more years. Mother's place of residence at the time of survey (dichotomized as urban versus rural) was controlled for as this is also expected to influence child survival (Andoh et al., Reference Andoh, Umezaki, Nakamura, Kizuki and Takano2007). To account for the fact that women with sick children might have been selected to joining healing religious organizations, a time-varying variable indicating whether a mother joined a religious denomination due to health problems was controlled for. This variable was lagged by one year. Because information on household material characteristics was available only for the time of survey, these characteristics could not be controlled for. This is acknowledged as a limitation of the analysis.
The statistical model also included a community-level control and a period control. The average female educational level in the community (computed as the aggregate average of respondents' education in a survey cluster) was controlled for. The average female educational level in the community has been found to have an effect on child survival beyond the effect of mother's own education; and it could also be suggestive of the degree of community knowledge about good health behaviour, community environmental hygiene and nutrition (Kravdal, Reference Kravdal2004). Although the community average educational level indicator was based on the information at the time of survey, it was used as a marker of human development differentials of the communities covered in the survey that probably existed over a longer period of time. Finally, a measure of historical period of child's birth was included as an attempt at capturing the influence of broad socioeconomic and political changes in Mozambique in the last 40 years. As has been mentioned, the country went through a period of civil war and post-war reconstruction. It is likely that children born in some years of Mozambican history might have faced an elevated risk of death. Three dummy variables were created to account for the historical period of child's birth: born between 1970 and 1992, born between 1993 and 2000, and born between 2001 and 2008. Selected descriptive statistics for the variables used in the analysis are provided in Table 1.
Selected descriptive statisticsa
The descriptive statistics refer to the year of survey only.
Statistical analysis
To describe the baseline under-five mortality differences between children of non-affiliated women and children of women belonging to each of the denominational categories, five-year probabilities of survival using the Kaplan–Meier (product limit) estimator were estimated. The analysis was done using the LIFETEST procedure in SAS 9.2 (SAS Institute Inc, 2008). The Log-rank test was used to assess the equality of survival functions between the children of non-affiliated women and those of women belonging to each denomination.
The multivariate analysis employed the discrete-time event-history approach. Discrete-time event-history analysis is appropriate because it allows the inclusion in the model of children who are still alive by the time of survey, i.e. censored children (Sear et al., Reference Sear, Steele, McGregor and Mace2002). For analysis, the GLIMMIX procedure in SAS 9.2 was used to fit random-effects discrete-time logistic regression models (SAS Institute Inc, 2008). Models with random effects were employed to take into account that children living in the same village and children born to the same mother may share some unobserved characteristics, which may introduce bias to the results (Das Gupta, Reference Das Gupta1990; Curtis et al., Reference Curtis, Diamond and McDonald1993; Sear et al., Reference Sear, Steele, McGregor and Mace2002; Avogo & Agadjanian, Reference Avogo and Agadjanian2010).
The statistical model is as follows. Let Yijkt be the event indicator, where Yijkt=1 if child i of woman j and community k dies in year t, and Yijkt=0 if the child survives past year t. The model that is fitted may be described by the following equation:
$$logit(P_{ijkt})=\beta _0+\beta _1X_{ijk}+\beta_2X_{ijkt}+\beta_3Z_k+U_{jk}+U_k$$, where Pijkt=P[Yijkt=1/Xijk, Xijkt, Zk]; β 1, β 2 and β 3 are vectors of coefficients; Xijk is a vector of time-fixed woman-level covariates; Xijkt is a vector of time-dependent woman-level covariates, Zk is a vector of time-fixed community-level covariates; Ujk and Uk are woman- and community-level random effects, respectively.
Results
Table 2 displays comparison of survival probabilities between children of non-affiliated mothers and children of mothers belonging to each of the other religious categories: Catholic or mainstream Protestant, Apostolic, Zionist and other Pentecostal. It is observed that children born to mothers belonging to any religious denomination have higher survival probabilities when compared with children of non-affiliated mothers. However, only the survival probabilities of children of Catholic or mainstream Protestant and Apostolic mothers are significantly different from those of children born to non-affiliated mothers (Table 2, Log-rank test, p<0.01).
Kaplan–Meier child survival probabilities: each denominational category versus no affiliationa
Log-rank test: Catholic or mainline Protestant versus no affiliation, p<0.001; Apostolic versus no affiliation, p=0.007; Zionist versus no affiliation, p=0.170; other Pentecostal versus no affiliation, p=0.165.
Next the multivariate results are presented. Table 3 shows results of the random-effects discrete-time hazard models predicting the effect of mother's religious affiliation on under-five mortality. Models 1 and 2 examine the effect of mother's affiliation to any religious organization on under-five mortality, while Models 3 and 4 examine the effect of mother's affiliation to specific denominational groups. The results are presented as odds ratios, which are obtained by exponentiation of the logistic regression coefficients. The results can also be interpreted as rates because odds are approximately the same as rates when the rates are small or when the events are few compared with the number of periods of exposure to risk.
Odds ratios from random-effects discrete-time hazard models of mother's religious affiliation on under-five mortalitya
Levels of significance: †p<0.1; *p≤0.05; **p≤0.01.
First, whether there is a net difference in under-five mortality between children of non-affiliated mothers and children of mothers affiliated to an organized religion, regardless of the type of denomination was assessed (Models 1 and 2). Model 1 examines the effect of mother's affiliation to any religious organization on the rate of child death and displays the baseline hazard. Model 1 indicates that mother's affiliation to any religious organization is associated with lower under-five mortality. The rate of death of children of women affiliated to an organized religion is 20% lower than that of children of women with no religious affiliation (the reference category). As expected, the baseline hazard indicates that the rate of death of children is significantly higher in the first year of life and decreases with age. When other control variables are added in Model 2, the magnitude of the effect of mother's affiliation to organized religion decreases but the effect remains statistically significant. Net of other factors, the rate of death of children of mothers affiliated to an organized religion is 15% lower. This decrease of the effect of the main predictor is due largely to the effect of maternal education. The findings in Model 2 suggest that affiliation to an organized religion has a beneficial effect on child survival, net of other factors.
Models 3–4 test for the effect of mother's belonging to specific religious denominations on the rate of child death. Model 3 points to a survival advantage of children of women affiliated to any religious denominations, compared with children of women not affiliated to organized religion. However, the coefficients are only statistically significant for the effects of mother's affiliation to the Catholic church or mainstream Protestant denominations and of affiliation to an Apostolic church. Relative to children of non-affiliated women, children born to Catholic or mainstream Protestant mothers and to Apostolic mothers have a rate of death that is lower by 29% and 25%, respectively. In Model 4, when adding other covariates, the pattern of denominational differences in child survival remains similar to that observed in Model 3. The rate of death of children born to Catholic or mainstream Protestant mothers is lower by 24% than that of children born to mothers belonging to the reference group, net of other factors. Compared with children of non-affiliated women, children of Apostolic women display a rate of death that is lower by 22%, controlling for other factors. For both Catholics/mainstream Protestants and Apostolics the effects remain statistically significant (although now only marginally so in the case of Apostolic mothers). The rates of death of children born to Zionist and to other Pentecostal women are not significantly different from that of children of non-affiliated mothers. In sum, the addition of controls, and most importantly, of mother's education, barely changes the effect of religious affiliation on under-five mortality. To explore whether the effects of religious affiliation varies by educational level, interaction between religious affiliation and education was also tested for. The parameter estimates for the interaction terms were not statistically significant (not shown).
Finally, because information on household characteristics is available only for the time of the survey, which cannot be extrapolated to the distant past, this same set of models was run restricting the period of observation to 5 years preceding the survey and including controls for household material conditions (assuming that these conditions did not change much during that period). The results of these additional tests were essentially the same as those presented here (results are not shown but can be provided upon request).
Discussion and Conclusion
In this study, the hypothesis that children of women affiliated to an organized religion, irrespective of the religious denomination would have a survival advantage over children of non-affiliated women was first tested. The findings show that affiliation to any religious organization significantly decreased the hazard of under-five mortality net of other factors. Given that religious organizations in sub-Saharan Africa are diverse in their composition and characteristics, whether the effect of affiliation to an organized religion on under-five mortality rate could vary by the type of mother's denomination was also assessed. The analysis by religious denomination yielded support for the hypothesis that affiliation to Catholic or mainstream Protestant churches and affiliation to an Apostolic denomination would offer a child survival advantage relative to non-affiliation. Despite the similarity of the results for the two denominational categories, the mechanisms through which their child survival advantage is achieved are probably very different. Previous studies in sub-Saharan Africa have observed that Catholic or mainstream Protestant churches tend to have members with better education (Agadjanian, Reference Agadjanian2001; Takyi & Addai, Reference Takyi and Addai2002; Gyimah et al., Reference Gyimah, Takyi and Addai2006). However, in the analysis, the advantage of Catholics and mainstream Protestants over non-affiliated women persisted even after controlling for educational differences. Instead, the results are interpreted in light of the earlier observations that the Catholic and mainstream Protestant churches are better connected to modern medical services, both because they have members who work in the health sector and because they place less faith in miracle healing than do other denominations (e.g. Gregson et al., Reference Gregson, Zhumu, Anderson and Chandiwana1999; Agadjanian, Reference Agadjanian2001, Reference Agadjanian2005). These connections cannot be documented directly with data and this interpretation therefore remains tentative. In comparison, the favourable effect on child survival of maternal affiliation to an Apostolic church may stem, it is argued, from the high level of social cohesiveness and mutual support among members of Apostolic churches.
The analysis also demonstrated that the survival of children of women belonging to Zionist and other Pentecostal denominations, i.e. denominations that generally lack the health sector connections of Catholics and mainstream Protestants and the social support networks of Apostolics, was not significantly different from that of children whose mothers were not affiliated to a religion. It is important to note, however, that while the survival chances of children of Catholics/mainstream Protestants and Apostolics were significantly higher than those of children of non-affiliated women, the differences among religious denominations were not statistically significant after controlling for other factors (results not shown). The findings on child survival advantage related to mother's affiliation to specific religious groups are similar to those of other studies (e.g. Antai et al., Reference Antai, Ghilagaber, Wedrén and Macassa2009; Valle et al., Reference Valle, Fernández and Potter2009).
As organized religion continues to wield enormous influence in everyday life in sub-Saharan Africa (Carmody, Reference Carmody2003; Gyimah et al., Reference Gyimah, Takyi and Addai2006; Trinitapoli, Reference Trinitapoli2006; Gallup International, 2010) and child mortality in the region remains the highest in the world (WHO, 2011), efforts to improve child survival in the region need to consider the role of religion. The present case study of Mozambique illustrates the need for a better understanding of the role of religious groups' organizational characteristics and of their position vis-à-vis the health sector. Future research and policy should further explore and engage the implications of these and other dimensions of religious life for child health and well-being.
Acknowledgments
The data used in this study were collected with the support of the Eunice Kennedy Shriver National Institute of Child Health & Human Development (grant #R01 HD050175). Insightful comments from anonymous reviewers are also gratefully acknowledged.
