The broad analysis of six global regions shows associations of TFR with each of the five factors explored (Figs. 1, 2, 3, 4, 5, and 7). The similarity of results among and within regions suggests that the relationships (negative or positive) are real and fairly general. Intriguing deviations occur in W and E Europe. Moreover, the factors to which TFR is related are themselves related in interesting ways, especially female education, which is positively correlated with GDP per capita and negatively correlated with religiosity.
TFR is strongly associated with education, contraceptive use, and religiosity (r2 = 0.89, 0.66 and 0.66, respectively). Among regions (Fig. 1), TFR decreased with increasing education for females, supporting earlier studies (e.g. [18, 19, 21, 62]). The number of school years for women increased markedly after 1970 in most regions, but increased less in Africa . The decrease in TFR might also arise indirectly via school year correlations with improved economy, family planning (FP) programs, and media attention to FP, factors which may also lead to smaller families [64,65,66,67].
Below, “among regions” refers to comparisons of regions, and “within region” refers to comparisons of countries within regions.
Western Europe and related countries
Within this region, TFR and education were positively associated, in contrast to all other regions (Fig. 2). This result is consistent with the reversal of TFR decline between 1975 and 2005 in Western countries at high (and increasing) values of the Human Development Index [7, 68, 69]. Also increasing immigration to W Europe may influence TFR (see [70, 71]).
TFR had little or no association with contraceptive prevalence (CPR) or GDP per capita. In contrast to non-European regions, TFR here tended to decline with higher religiosity, partly due to south European countries: among the six countries with highest religiosity, five were in S Europe (Portugal, Italy, Greece, Malta and Cyprus), all with low TFR. But within W European countries, there is evidence that TFR is lowest for religiously unaffiliated or more secular groups [72, 73]. Compared to non-European regions, few European countries have strong religiosity.
E and W Europe had similar average TFR, school years and religiosity, but E Europe had lower CPR and much lower GDP per capita. In contrast to W Europe, TFR in E Europe had no or weak relation to education (Fig. 2). History, post-Soviet economic uncertainty and low GDP per capita may account for higher mean mother’s age at childbirth in E Europe (; see also ). Note that CPR measures modern contraceptives, whereas E European methods include high prevalence of withdrawal, rhythm method, and abortion ). Contraceptive use in E Europe may therefore be higher than in Fig. 1, and the relation TFR versus CPR among regions stronger than shown there.
Within E Europe there was no or weak relation between TFR and GDP per capita (Fig. 5), but note that TFR varies little. Education had high levels in both E and W Europe. Hilevych & Rusterholz  suggested that female labor force participation and contraceptive use favor small families (low TFR) in both E and W Europe. In addition, countries in these two regions may have gone through a ‘second demographic transition’, with a diversity of union and family types and very low TFR (see , and review in ).
Latin America and the Caribbean
Among regions, Latin America and Asia are intermediate in TFR level and religiosity. Latin America had the second lowest GDP per capita and, perhaps surprisingly, the second highest CPR. In many countries, such as Chile, Colombia, Costa Rica and Mexico, family planning activities, policies or programs started and expanded in the 1960’s and 1970’s. Despite resistance from the Vatican, modern contraception became widespread early [30, 39, 80].
Within Latin America TFR declined with more education, but it declined more strongly with increased CPR and GDP per capita (Figs. 4, and 5), suggesting that these factors may be more important than education for TFR in Latin America. School years and GDP per capita were strongly positively associated, suggesting that economic resources sometimes limit education. CPR on the other hand was weakly related to GDP and education, and may partly be limited by other factors – possibly religiosity, through its negative correlation with education. At higher levels of religiosity in Latin America (proportion > 0.8) there is remarkable variation in school years and CPR among countries. At high levels of religiosity, some countries therefore achieve high levels of female education and CPR, in contrast with others at similarly high level of religiosity. This variation deserves further study, see Additional file 1 (part 1).
Arab States had the second highest TFR among the regions, low CPR, and an unusual combination of highest religiosity and second highest GDP per capita among the regions. In some countries, oil resources have led to wealth, but the mean for female school years is low (very low for some countries). Within the region TFR declined strongly with increased education, GDP per capita, and CPR. TFR and religiosity were weakly associated, but note the small variation: almost all countries are highly religious.
The Arab States began implementing FP programs fairly recently, during the 1990’s (; for exceptions, such as Tunisia and Morocco, see ). Effects of FP efforts may come in the future, unless religiosity hinders TFR decline (, and references therein). As in Latin America, at high levels of religiosity (proportion > 0.9) there is large variation in school years, GDP, and CPR among the countries. Arab State social norms, also associated with religion, generally disfavor female empowerment .
This region stands out with much higher TFR and markedly lower CPR than in the other five regions. The level of religiosity is high, similar to Arab States, but GDP per capita is much lower. Within Sub-Saharan Africa, TFR is strikingly negatively correlated with education, GDP and CPR, which all may affect TFR. Two ‘natural experiments’, involving changes in schooling in Nigeria  and Uganda , support the role of education for TFR. School years, GDP and CPR were strongly positively correlated, particularly CPR and school years, suggesting that education favors contraceptive use.
Religious influence may be one contributing reason for high TFR, and for stalling TFR decline in this region. For the eight countries with religiosity above 0.95, females had on average only 1–5 school years. Religiosity was considered an important determinant of fertility in Sub-Saharan Africa by e.g. Caldwell & Caldwell , Akintunde et al.  and Agadjanian & Yabiku . A related and probably strong influence is persistent patriarchal social structure and gender inequality (e.g. ). For Burkina Faso, Mali, Niger and Chad, “One of the key barriers to having desired number of children is sociocultural norms, especially the husband’s role as primary decision-maker and the desire for a large family” .
Among regions, Asia resembled Latin America in TFR, GDP per capita and religiosity, though with lower average CPR (Fig. 1). Within Asia, lower TFR was associated with longer female education and higher GDP, and especially with higher CPR. As in Latin America, several countries with TFR below replacement level had CPR values above 70% (Thailand, South Korea and Hong Kong). FP programs have been important historically in these and other Asian countries . In central Asia, however, Pakistan, Tajikistan and Afghanistan had TFR above 3.5 and low levels of CPR. An interesting exception in central Asia is Azerbaijan, with the lowest CPR (Fig. 4) but with TFR at 2.1. Many female school years (10.6), low religiosity (proportion 0.5), use of traditional contraception  and economic conditions  may together explain this exception.
The Asian countries show a rather strong positive correlation between education and GDP, and an even stronger negative association between education and religiosity.
Role of different factors
To help clarify factors of likely importance for TFR in different global regions, we studied five potential major agents that could be quantified. Social norms are also important  but often difficult to quantify. For example, large desired family size characterizes Sub-Saharan Africa. Korotayev et al.  related this norm to polygyny, high status of polygynous men, extended families, and child fosterage within kinships. The latter two aspects enable females to carry out traditional hoe agriculture without reducing the number of children, contributing to high TFR. And in modern urban Africa, abolition of postpartum sex taboos reduces birth intervals and may contribute to high TFR when large desired family size persists [49, 90, 91].
To limit the number of factors and relationships we did not analyze infant and child mortality, gender roles and female labor force participation rates, which may all play a role [9, 25, 92,93,94,95,96]. These factors seem likely to bear some relation to female education, contraceptive use and GDP per capita. Family planning programs include contraception and education directly related to fertility, and was analyzed in four regions. Lower TFR was associated with stronger FP programs in Asia, Arab States and Latin America, but only weakly so in Sub-Saharan Africa. In a study of 40 countries 2003–2010, TFR levels “were lowest in the presence of both good social settings and strong programs”, but Sub-Saharan Africa was the least successful region (, based on data from track20.org). Yet, in 2014, the mean values for program strength were similar in all four regions in our study. However, FP programs in Asia and Latin America started earlier, and many of them are considered successful (, and references therein). Duration, change in social norms, institutional support and international funding are important for success of FP programs [27, 40, 46].
Lower TFR was associated with higher FP program strength in three regions. For Sub-Saharan Africa, Arab States and Asia, FP programs were under-represented in low and high TFR countries, compared to our full sample of countries. Incentives for starting FP programs may be lower in countries with relatively low TFR. And such programs might be difficult to start in poor, high-TFR countries with strong religion, corruption or conflicts. Nevertheless, the results in Fig. 7 suggest that FP programs recently have been effective also within relatively narrow TFR ranges in Asia and Arab states, but not in Sub-Saharan Africa.
Among regions, the TFR versus GDP per capita relationship was the weakest of the four (Fig. 1). Without Sub-Saharan Africa, the slope of the regression would be near zero. But within four regions, TFR’s negative relation to GDP per capita was strong or relatively strong (Fig. 5). So why is TFR not associated with GDP per capita in E and W Europe, in line with economic hypotheses, and despite equally large variation in GDP per capita as in Latin America? And why are school years, potentially improving child ‘quality’, not negatively associated with TFR in E and W Europe? The relation is even reversed, TFR increasing with school years in W Europe.
Evidence for a quantity-quality trade-off, between increased family size and investment in child quality, is mixed (, and references therein). In India, trade-off was strongest in rural areas . In this study, TFR declined with increasing GDP per capita especially in the three poorest regions (Sub-Saharan Africa, Asia, Latin America). Is there a self-reinforcing loop, where increased wealth motivates higher child quality and other changes that reduce TFR, the reduction feeding back positively on economic development and wealth? According to Canning & Schultz , TFR declines can boost income per capita through reduced youth dependency rates, and may have positive long-term economic effects (see also [27,28,29]).
This study is, as far as we know, the first to relate TFR to religiosity together with other major factors in global regions and many countries. Both among the regions (Fig. 1) and within two of them (Asia and Sub-Saharan Africa, Fig. 3), TFR increased with degree of religiosity. Moreover, stronger religiosity is associated with lower education, CPR and GDP per capita in at least five regions. Among Arab States, effects of the large differences in wealth seem to override effects of the small differences in strength of religiosity.
We quantified religiosity from Gallup surveys, but did not distinguish between religions as regard TFR. There are probably differences [37, 86, 99], but using the same basic measure greatly simplifies regional and global analyses. In a study in the US, religiosity measured as here was more useful than religious affiliation and showed “a substantially positive effect on fertility”, without any gender difference . Most earlier studies analyzed religious affiliation and TFR. Global TFR 2010–2015 was substantially lower for religiously non-affiliated (1.7) than for affiliated (2.6) .
Why is fertility associated with religiosity? Beside declarations from the Vatican and other religious leaders [30, 31], possible reasons are belief in supernatural influence on things we desire, such as “good crops, protection, health and fertility” [33, 101], and fatalistic views about fertility, such as children “are up to God” [46, 89]. Human sociality and norms, history, type of religion and other conditions influence TFR-religion relationships [86, 99, 102]. Religiosity probably contributes to maintaining high TFR in Sub-Saharan Africa, Arab States and parts of Asia and Latin America, in part by suppressing factors that reduce TFR. Yet FP programs have been successful even in strongly religious countries, as shown by encouraging results in Iran , Tunisia , and Rwanda .