Secularization and desecularization in our time.

Meisenberg, Gerhard

The role of religion in public life has diminished in Europe and most other advanced societies since the time of the Enlightenment in the 18th century, and personal religiosity appears to be less today than it has been in past centuries. Nevertheless, the secularizing trend is not universal, as demonstrated by the rise of organized religion in the United States since the late 18th century (Finke & Stark, 1993). Predictions by early 20th century intellectuals of the impending demise of religion have been premature. Religious belief persists to the present day even in advanced Western societies.

The two immediate questions about current trends are: (1) is there a (continuing) trend toward lower religiosity in the most advanced societies? (2) Are the less developed countries entering an era of secularization and declining religious belief that is similar to the developments in Europe during the last two centuries?

This study goes beyond the question of whether people are becoming more or less religious over time by examining the correlates and possible causes of the current trends. By specifying the conditions under which religion is likely to thrive or decline, this approach will allow us not only to understand the causes of past and current trends, but also to predict, to some extent, future developments. Several theories will be investigated:

1. Religion, as an arbitrary belief system that is based on faith and emotion rather than reasoned weighting of the evidence, is predicted to decline in countries in which the average intelligence of the population is high, and to rise or remain at a high level in countries in which the average intelligence of the population is low (Lynn et al., 2009; Meisenberg, 2004). Other attributes of advanced societies that might be responsible for religious decline include rising prosperity and consumerism, advances in the educational system, reduced government regulation of the religious economy (Grim & Finke, 2006; Iannaccone et al., 1997), or decreased societal dysfunction (Delamontagne, 2010).

2. Changes in religious culture over time are effected by a mechanism of generational replacement. This hypothesis predicts that in countries in which young people are far less religious than the old, religiosity is likely to decline; and in countries in which young people are more religious than the old, religion is likely to rise.

3. Social learning theories that emphasize the emulation of successful individuals (Mesoudi, 2009) predict that the religious beliefs of the social elite give direction to ongoing trends. Religion is expected to rise in countries in which educated and high-status individuals tend to be more religious than low-status individuals, and to decline in countries in which high-status individuals have below-average religiosity.

4. Religiosity is a familial trait (Myers, 1996; Koenig et al., 2009) that is transmitted by family environment and shared genes. Therefore religion is predicted to rise in countries in which religious people have more children than the less religious, and decline in countries in which religious people have below-average fertility.

5. Because mothers tend to be more involved than fathers in the socialization of children, religion is predicted to rise in those countries in which women are far more religious than men, and decline in those in which women are less religious than men.

Methods

Data about religiosity are from the World Values Survey (WVS) Official Aggregate v.20090901, 2009, available free of charge at www.worldvaluessurvey.org. Data are from 355,298 respondents in 96 countries and territories. The WVS is the largest survey of its kind. Its great advantage is its broad coverage of all "cultural provinces" of the world. Its main disadvantage is the poor representativeness of many country samples, which is especially striking for education. The country-level correlation between the average educational level in the country and the average educational level of respondents in the WVS (definitions are given below) is .492 (N = 94 countries), mainly because of the oversampling of highly educated people in many of the less developed countries.

The following variables were derived from the WVS:

1. Religious belief. This variable was computed from four questions that had been asked in all or nearly all countries: (1) "...would you say you are: A religious person--Not a religious person--A convinced atheist?" (2) "Do you believe in God? Yes/No/Don't know." (3) "How important is God in your life?-10-step scale from 'not at all important' to 'very important.'" (4) "Do you find that you get comfort and strength from religion? Yes/No/Don't know." This measure captures belief in and emotional involvement with, a personal god. This measure can and does produce high scores in polytheistic societies such as India, but yields low scores in societies such as China, which are dominated by agnostic or atheist philosophies such as Confucianism or communism. Individual-level correlations between the four items (Pearson's r) ranged from .579 to .729. In cases of missing data, the score was extrapolated from the available items. A score could be computed for 345,743 respondents. It was transformed to a scale on which the lowest possible score was zero and the highest possible score was 10.

2. Religious trend. This measure was defined as the unstandardized B coefficient for survey year in regression models predicting religious belief with sex, age and survey year for each country.

3. Education. This is a composite variable that was calculated from (a) the age at which formal education ended or (for young respondents) is expected to end; and (b) highest educational degree, from no schooling to university degree. The correlation between these two measures was r = .689. For respondents who were missing one of the two variables, the score was extrapolated from the available measure. A score could be computed for 344,460 respondents. Scores were scaled from zero (no schooling) to 10 (schooling to age 31 + , university degree). Although this measure captures exposure to a secular school system in most instances, it does not measure curricular contents. Nor is it a measure for the cognitive outcomes of schooling, which are described by school achievement or IQ.

The percentage of the population belonging to each of the major religious groups (Catholic, Protestant, Orthodox, Muslim, Hindu, Buddhist/Confucian, pagan and unaffiliated) were from the World Fact Book of the CIA (https://www.cia.gov/library/publications /theworld-factbook/geos/xx.html).

World regions were defined similar to Inglehart et al. (2004). Protestant Europe was defined as the traditionally Protestant countries of northern and central Europe, except Britain. English-speaking countries include the British Isles and English-speaking overseas nations with mainly European-origin population. Catholic Europe & Mediterranean contains the Catholic countries of southern Europe and also Greece, Cyprus and Israel. Middle East refers to the predominantly Muslim countries from Morocco to Pakistan. Africa includes only countries of sub-Saharan Africa. Because of the great cultural and economic differences between different racial groups in Africa (Lynn, 2008), in those countries where the question was asked, only those classifying themselves as "Black" were used in the individual-level analyses. South (+ Southeast) Asia is a heterogeneous group of countries ranging from India to the Philippines. East Asia includes countries with predominantly Confucian culture: China, Japan, Hong Kong, South Korea, Taiwan and Singapore.

In addition to the variables derived from the WVS, the following country-level variables were used:

1. Intelligence was defined as the average of two variables:

(1) Average IQ in the country based on the compilation of Lynn & Vanhanen (2006), with the extensions and amendments reported in Lynn (2010). This "Greenwich IQ" is defined with an average of 100 and standard deviation of 15 for Britain. Measured IQs are available for 136 countries and territories, including 80 in the WVS.

(2) School achievement, based on average scores on standardized school achievement tests. Scores were calculated (in the IQ metric) primarily from the 8th-grade TIMSS assessments in mathematics and science in 1995, 1999, 2003 and 2007, and the PISA assessments of 13-year-olds in 2000, 2003, 2006 and 2009. These data are freely available online. Missing data were extrapolated into this data set from other international scholastic assessment tests as described in Lynn & Meisenberg (2010). Scores are available for 111 countries, including 82 countries in the WVS. The correlation between school achievement and IQ is .907 for the 87 countries having both measures.

The Intelligence score was computed as the average of measured IQ and school achievement, weighted for data quality (see Lynn & Meisenberg, 2010), for countries having both measures, and measured IQ alone or school achievement alone for those having one of the two measures. Of the 96 countries and territories in the World Values Survey, the intelligence score is based on IQ and school achievement for 70 countries, IQ only for 10 countries, and school achievement only for 12 countries. Estimates based on the scores of neighboring countries with similar population, culture and economic development (as described for IQ in Lynn & Vanhanen, 2006) were used for the remaining 4 countries. Scores ranged from 68.5 (Ethiopia) to 107.1 (Singapore).

2. Education is a composite measure for exposure to formal schooling. It was calculated from four datasets: (1) Average years of schooling for adults over the age of 25 from the Barro-Lee dataset: (http://www.cid.harvard.edu/ciddata/Appendix%20Data%20Tables.xl s) for the year 2000, or extrapolated from the latest available data.

(2) School life expectancy for the year 1999 (or extrapolated from earliest available date) from UNESCO at: http://stats.uis.unesco.org/TableViewer/tableView.aspx.

(3) Combined gross enrolment ratio for primary, secondary and tertiary schools in 2002 (United Nations, 2005). (4) Arcsine-transformed averages of the adult literacy rates in 1990 and 2002 (United Nations, 2004). Measures (1) and (2) were averaged, missing data were extrapolated from measure (3), and the remaining missing data were extrapolated from measure (4).

3. lgGDP is the logarithm of gross domestic product adjusted for purchasing power, averaged for the years 1995-2000, from the Penn World Tables (Heston et al., 2009).

4. Corruption is calculated as the average of the corruption score published by Transparency International at; http://www.transparency.org, average of the years 1999-2005, and the corruption score published by the Heritage Foundation at http://www.heritage.org/research/.

5. Freedom/Democracy is calculated from the scores of political freedom (political rights + civil liberties) from Freedom House at http://www.freedomhouse.org/research/freeworld, averaged over the years 1985-2005, and Vanhanen's democracy index, average 1985-2004, from the Finnish Social Science Data Archive at http://www.fsd.uta.fi/english/data/catalogue/FSD1289/. Democracy and political freedom were highly correlated (r = .847, N = 179 countries).

All statistical evaluations were done with SPSS software.

Results

1. Levels of religious belief

Before analyzing current trends, we need to examine the determinants of country-level and individual differences in religious belief. Table 1 shows that religious belief tends to be highest in sub-Saharan Africa and in the Muslim countries of the Middle East and North Africa, and lowest in East Asia and Protestant Europe. This suggests a negative relationship between religiosity and one or another aspect of cognitive or economic development. This impression is confirmed by the country-level correlations shown in Table 2. Religiosity correlates negatively with all development indicators, with highest correlations for intelligence. Not included in Table 2 are the correlations for the 25 communist and ex-communist countries, which are -.638 for intelligence (p = .001), -.433 for freedom from corruption (p = .032), and less than [+ or -].225 for education, lgGDP, and freedom/democracy.

These results confirm earlier observations of a negative relationship between intelligence and religiosity in comparisons between countries (Lynn et al., 2009; Meisenberg, 2004). In addition to intelligence, corruption seems to be the second most important religiosity correlate when the non-communist and communist countries are considered separately. This supports the observation of Delamontagne (2010) that high religiosity tends to be associated with "societal dysfunction." However, the high correlations among all development indicators make inferences about causal paths tenuous.

2. Current trends

Worldwide, religious belief is rising in 30 countries and declining in 32 countries, out of the 62 for which trend data are available (see appendix). Rising and declining trends are statistically significant at p<.05 in 24 countries each. Religiosity is rising in nearly all of the ex-communist countries of Eastern Europe and the former Soviet Union, and also in the two nominally communist countries China and Vietnam (see appendix and Table 3). Insignificant or declining trends are seen only in ex-communist countries that are geographically and culturally close to the historically non-communist countries of northern and central Europe: Estonia, the Czech Republic, Poland, and perhaps Slovenia. These countries seem to be evolving toward the low level of religiosity typical for Western Europe.

Table 2 includes not only the current level of religious belief, but also the trend in religiosity. It shows that in the worldwide sample (correlations under the diagonal), declining trends are typical for affluent democracies with low corruption. However, when the analysis is limited to countries without communist history (correlations above the diagonal), high intelligence tends to be the best predictor of ongoing religious decline. The discrepancy is caused by the marked revival of religion in the communist and ex-communist countries (Table 3). These countries tend to have relatively high intelligence but mediocre GDP, low freedom and democracy, and high corruption. Among the 18 communist and ex-communist countries with trend data, the rise of religion is favored most by high corruption (r = .651, p = .003) and low freedom/democracy (r = .577, p = .012). Low education (r = .479), low log-transformed GDP (r = .458) and low intelligence (r = .375) are weaker correlates of religious revival (not included in Table 2).

Rising trends are seen also in the two countries of tropical Africa for which trend data are available (Nigeria, South Africa). Trends are inconsistent in Latin America and the Muslim countries of North Africa and the Middle East, and declining trends are typical for non-communist Europe and its overseas offshoots, as well as non-communist East Asia. The trend values in Table 3 were calculated for individual countries, and then averaged for each world region with weighting for the number of respondents and the time span over which the survey was performed.

Regression models predicting the religiosity trend at the country level produced no strong and consistent results for any of the development indicators, possibly because of collinearity between the development indicators. Only communist history had a robust effect. Therefore the hypothesized effects of several other variables were explored in turn. These were the extent to which people become more religious with advancing age (Age effect in Table 4), the extent to which women are more religious than men (Gender effect), the extent to which high education is associated with higher religiosity (Education effect), and the extent to which higher religiosity is associated with higher birth rates (Differential reproduction). Note that the definitions of these variables in Table 4 and the appendix are different from the corresponding variables used in the regression models of Tables 5 to 9.

3. Generational replacement

According to Max Planck (1948, p. 22), "A new scientific truth does not triumph by convincing its opponents and making them see the light, but rather because its opponents eventually die, and a new generation grows up that is familiar with it." If we consider religious belief or atheism as being analogous to new scientific truths, Planck's dictum predicts that longitudinal trends in religiosity are predicted by cross-sectional age differences. In those countries in which young people are far less religious than the old, religion is expected to decline over time for all age groups; and in countries in which young people are more religious than the old, religion is expected to rise.

The age-dependence of religiosity was determined with country-level linear regressions in which religiosity was predicted by sex, age, education and survey year. The unstandardized B coefficient for age, which indexes the extent to which religiosity rises or declines for each year of chronological age, was taken as a measure for the independent effect of age on religious belief. Religiosity increased with age in 86 countries and declined with age in 10 countries. With the exceptions of the Dominican Republic, Israel, Ghana and Zambia, all countries in which young people were more religious than old people were communist or ex-communist countries. Table 4 shows, under Age effect, that in the economically advanced world regions, old people are far more religious than the young. For example, in Protestant Europe each 10 years of chronological age add approximately 0.4 points on the zero-to-10 religiosity scale.

When the age-dependence of religious belief was correlated with the temporal trend, the correlation was found to be -.404 (p = .001, N = 62 countries). As expected, religion is likely to decline over time in those countries in which young people are far less religious than the old. However, because the age-dependence of religiosity is systematically greater in the more advanced world regions (Table 4), this observation needs to be confirmed in regression models that include the development indicators. Table 5 shows regression models in which the religiosity trend is predicted by the age effect and several other measures. In all models, the age-dependence of religiosity is negatively related to the temporal trend. Model 1 shows that the effect is not due to spatial autocorrelation because it is evident even within world regions. The omitted regional variable in Model 1 is Protestant Europe. The other models show weak and inconsistent effects of the development indicators, a tendency for declining religiosity in countries that were highly religious at the start of the trend measurement, and the rise of religion in the communist and ex-communist countries. The adjusted [R.sup.2] values of the models show that a substantial portion of the religiosity trend can be explained with the tested variables.

4. Education effect

Table 2 shows that a low average level of religious belief is predicted most powerfully by high average intelligence in the country. This raises the question of whether high intelligence correlates with low religiosity at the individual level as well. However, the World Values Survey has no measure of intelligence. It only has measures of schooling, which can be used as a proxy for intelligence because the intelligence-education correlation is generally high at the individual level, usually with r>.50 in modern societies (Ceci & Williams, 1997; Meisenberg & Kaul, 2010).

The column labeled Education effect in Table 4 lends partial support to the expectation of a negative relationship between education and religiosity. Tropical Africa is the only world region in which the religiosity-education relationship is mainly positive. However, the effect is very weak in most countries. In all, the relationship between education and religious belief is negative in 70 countries, positive in 25 countries, and zero in one (see appendix). It is most negative in the communist and ex-communist countries, presumably because communist governments used the educational system to discredit religion. Worldwide, the partial correlation between religiosity and education at the individual level (sex, age and country controlled) is -0.053 (N = 328,326 respondents).

The measure for the relationship between education and religiosity was calculated as the unstandardized B coefficient for education in regression models predicting religiosity with sex, age, survey year and education. The correlation of this measure with the religiosity trend was found to be -.513 (N = 62 countries), which is significant at p<.001. This means that in countries in which higher education is associated with higher religiosity, religiosity tends to decline over time. When the analysis is restricted to the non-communist countries, the correlation is reduced to r = -.277 (p = .069, N = 44 countries). This shows that much of this effect is explained by the strongly negative relationship between education and religiosity in the communist and ex-communist countries.

Nevertheless, Table 6 shows that this effect cannot be explained by spatial autocorrelation within predefined world regions (Model 1). It is undiminished when the usual development indicators are controlled (Models 2 and 3), and persists even when the effects of religious denomination are controlled in addition to the development indicators (Model 4). The effect of communism is attenuated in these models because the communist and ex-communist countries tend to have strongly negative relationships between education and religiosity (Table 4).

5. Differential Fertility

Environmental and genetic transmission in the family both imply that children's religiosity is more similar to their parents' than expected by chance. As a consequence, religion is expected to rise in countries in which religious people have more children than atheists, and to decline in countries in which atheists have more children than religious people.

Differential fertility was defined as the partial correlation between religiosity and number of children in country-level regression models predicting number of children with religiosity, age and survey year, calculated separately for 4 demographic groups: young females aged 15-40, young males aged 17-45, old females aged 41-80, and old males aged 46-80. The score is the unweighted average of these four partial correlations. The consistently positive signs in the Differential reproduction column of Table 4 show that throughout the world religious people tend to have more children than atheists. The appendix shows that only 14 out of 94 countries have a negative religiosity-fertility relationship. The correlation between the religiosity trend and the religiosity-fertility partial correlation is .106 (N = 62 countries), which is not significant. However, Table 7 shows that significant positive relationships emerge consistently in regression models that control for confounding variables. Although the [beta] coefficients show that the differential fertility effect is of only moderate size, the t values show it to be rather consistent.

6. Gender effect

The effect of differential fertility on the religious trend shows that religion is transmitted in families. Assuming that much of the differential fertility effect is mediated by the family environment rather than shared genes, we have to ask whether mothers or fathers are more important for the religious upbringing of their children. Table 4 shows that throughout the world, women are more religious than men. In countries with high average religiosity, both men and women are highly religious and the gender difference is small. Gender differences are larger in low-religiosity countries, and are largest in advanced nations with Christian tradition. Gender differences are smallest in the Muslim countries of North Africa and the Middle East. This region has high average religiosity, and it is quite possible that Islam is intrinsically more attractive for men whereas Christianity is more attractive for women. The appendix shows that Mali and Saudi Arabia are the only countries in the World Values Survey in which men are (marginally) more religious than women.

For the regression models, the extent to which women's religiosity exceeds men's was measured as the unstandardized B coefficient for female gender in regressions predicting religiosity with gender, age and survey year. The correlation of this measure with the temporal trend for religiosity was r = .286 (p = .024, N = 62 countries). This suggests that in countries in which women are far more religious than men, religion is more likely to rise; and where women have low religiosity relative to men's, religion is more likely to decline. Table 8 shows some regression models. The magnitude of the gender difference (women more religious than men) predicts a rising trend for religiosity independent of the other predictors. This trend is conventionally significant (p<.05) in 3 of the 4 models.

7. Complex models

Table 9 presents models that contain the religiosity-gender relationship, religiosity-education relationship, and/or religiosity-fertility relationship in addition to other predictors. The religiosity-age relationship is not included because causal effects on the religiosity trend are also likely to change the relationship between religiosity and age.

Significant effects of these variables were obtained in models controlling for the development indicators. Whereas the effect of gender difference is of moderate magnitude and rather consistent, the education effect is suppressed in models that also contain differential fertility, as seen in the comparison between models 2 and 3. Conversely, the education effect appears to reduce the effect of differential fertility, as shown by a comparison of the differential fertility effect in models 2 and 4. These results confirm that religiosity tends to rise in countries in which religious people have less education and higher fertility than atheists, and in which women are far more religious than men.

Discussion

The present study points to several conditions that appear to strengthen or weaken religious belief in our time. With reasonable projections for future trends in these conditions, we can to some extent predict rising and declining trends for religion in different world regions.

1. Communist rule

The recent rise of religion in the communist and ex-communist countries is best understood as a rebound phenomenon that restores a "normal" level of religious belief in these countries after the end of state-imposed atheism. The considerable magnitude of this trend implies that it can only be a temporary phenomenon. It remains to be seen whether the rising trend continues for some time into the future, or whether it has ended already in recent years.

2. Development indicators

In today's world, religiosity is far lower in the more advanced societies than in the less developed countries (see Tables 1 and 2 and the appendix). This state of affairs is usually attributed to a secularization process that has been evident in the European countries since the times of the Enlightenment and the Industrial Revolution. The precise reasons for this centuries-long trend are not known. However, the secularization hypothesis predicts that the inverse relationship of religiosity with development indicators is strongest for those indicators that most closely represent the aspects of modernity that are responsible for the religious decline.

Table 2 shows that current religiosity is somewhat more closely related to intelligence than to the other development indicators. This confirms the results of earlier work showing that both high religiosity and some other "anti-modern" cultural traits are related to low intelligence and absence of communist rule (Meisenberg, 2004). Assuming that today's intelligence differences between countries have, to some extent, been in existence throughout the last two centuries, this suggests that rising intelligence was a major cause of declining religiosity. In its simplest form, the hypothesis states that religious belief systems are irrational, and that therefore intelligent people are likely to reject religious world views in favor of scientific ones (Lynn et al., 2009). Thus it is perhaps no coincidence that East Asia is not only the world region with the highest intelligence (approximately 6 points higher than Europe), but also the one with the lowest religiosity (see Table 1).

Rising intelligence during the 20th century is a possible cause of religious decline because intelligence has been rising massively in those advanced societies for which informative data are available. This observation is known as the Flynn effect (Flynn, 1987). It has been variously attributed to more prolonged and intensive schooling (Blair et al., 2005), better nutrition (Lynn, 1990), reduced inbreeding (Mingroni, 2007, but see Woodley, 2011), and cultural amplifier effects (Dickens & Flynn, 2001). Low corruption appears to be the second most important correlate of low religiosity, after high intelligence. Therefore it is possible that declines in corruption or related aspects of societal dysfunction may have contributed to the weakening of religion in today's advanced societies, as is also suggested by the observations of Delamontagne (2010) in the United States.

Turning from the current level of religiosity to the recent trend in religiosity, we find that with the exception of the communist and ex-communist countries, religiosity has recently been declining in a majority of the more advanced societies (see Table 3 and appendix). This observation shows that secularization in the sense of declining religiosity has not yet ended in these societies. It is possible that the secularizing trend in the advanced Western societies will be reversed by the immigration of people from more religious countries, especially those with Muslim tradition. This trend is not (or not yet) evident in the World Values Survey data. In Protestant Europe, for example, the World Values Survey interviewed 22494 Protestants and 6381 Catholics, but only 205 Muslims (0.7% of the total).

For the non-communist countries, Table 2 shows weak negative associations of the religiosity trend with all development indicators. In other words, religiosity tends to be stable in less developed countries but to decline in the more advanced societies. The religiosity gap between the less developed countries and the more advanced countries got bigger in recent years. Table 3 confirms this conclusion by showing that religiosity is declining in Europe and its English-speaking offshoots as well as in non-communist East Asia, but remains virtually unchanged in Latin America and the Middle East. There may be a rising trend in parts of sub-Saharan Africa.

In the regression models of Tables 5 to 9 we frequently encounter development indicators with negative signs. Intelligence is the development indicator that is most often associated with religious decline. Also low corruption is associated with declining religion in many of the models. Education, freedom/democracy and log-transformed GDP rarely or never have independent effects. Therefore it appears that rising intelligence, and possibly declining corruption, have not only acted over the last one or two centuries to depress the level of religiosity in the most advanced world regions. They still have this effect today.

Other work has pointed to an association between high religiosity and "societal dysfunction" (Delamontagne, 2010; Paul, 2009). The association of low corruption with low religiosity (Table 2) and a declining religiosity trend (Table 2, Tables 5-9) suggest that high corruption or some other aspect of societal dysfunction does indeed support the persistence of religion in our time. Alternatively, strong religious worldviews may be a cause of corruption and other forms of societal dysfunction.

The recent declines of religiosity in a majority of the most advanced societies are not easily compatible with the views of Stark & Iannaccone (1994, p. 230), who concluded that any apparent religious declines were caused by structural features of the dominant churches and that "... the concept of secularization be dropped for lack of cases to which it could apply."

Recent religious declines in the most advanced societies also refute, for this group of countries at least, Peter Berger's thesis that the world today has entered a phase of "desecularization" (Berger, 1999). The question rather is: For how long is the decline of religion in the most advanced societies likely to continue? High intelligence seems to be inimical to religion, but combining all available evidence (Beaujean & Osterlind, 2008; Cotton et al., 2005; Shayer & Ginsburg, 2009; Sundet et al., 2004; Teasdale & Owen, 2008), it appears that in the most advanced countries, intelligence is no longer rising in cohorts born after about 1980. Intelligence may have started declining in some countries. Such declines have been reported from Norway (Sundet et al., 2004), Denmark (Teasdale & Owen, 2008) and Britain (Shayer & Ginsburg, 2009), although intelligence may still be rising slightly in the United States (Beaujean & Osterlind, 2008).

Possibly, the development of human intelligence is reaching its biological limits in the most advanced societies. If so, the advanced societies can expect fairly constant--or stagnating--levels of intelligence during the 21st century, although intelligence will eventually decline due to differential fertility favoring the less intelligent. For the United States at the end of the 20th century, the rate of the genetic decline has been estimated at up to 1.2 IQ points per generation (Meisenberg & Kaul, 2010), although this rate is likely to be smaller in ethnically homogeneous countries.

However, religion is a cultural phenomenon that is transmitted through social learning in families and the wider society. Cultural transmission implies inertia. It implies that the anti-religious effect of high intelligence is not fully realized immediately. Intelligence rather determines the set point to which religiosity will gravitate in the course of several generations. Even if intelligence remains unchanged at its current level in advanced societies, religion may continue its decline for some more generations until this set point has been reached.

Also corruption and other forms of societal dysfunction are not likely to change dramatically in these countries during the 21st century, although there is a possibility of rising dysfunction in the wake of rising income inequality. Until recently, at least, countries with high average intelligence used to have low income inequality (Meisenberg, 2007, 2008a), probably because an ample supply of skilled workers reduces the skill premium in the labor market. However, income inequality has been rising in most industrialized countries in the recent past (Atkinson, 2004). It is predicted to rise even more if economic development continues to raise the demand for cognitively demanding work while the supply of high intelligence is no longer rising. If rising income inequality brought about by stagnating or declining intelligence leads to societal dysfunction, conditions for religion may again improve.

What is predictable, however, is an impending secularization trend in modernizing societies that experience rising intelligence and declining corruption and disorder. Therefore we can expect declining religiosity in Latin America and the Middle East in the near future.

But why does religion still remain strong in the less developed countries of Latin America and the Muslim Middle East although many of these countries are advancing economically, and intelligence seems to be rising in at least some of them (Batterjee, 2011; Colom et al., 2006; Khaleefa et al., 2009; Meisenberg et al., 2005)? An answer is suggested by Figure 1, which shows the relationship between country-level intelligence and the current level of religiosity for the noncommunist countries. We see that religiosity declines with rising intelligence only at average IQs higher than about 85. A plot predicting religiosity with log-transformed GDP (not shown) looks similar, but with somewhat lower [R.sup.2] of .586. Non-linear fits with corruption and freedom/democracy have [R.sup.2] values of .563 and .544, respectively.

Scientific and non-religious ideological worldviews appear to become serious competitors of traditional religion only when a fairly high level of intelligence--roughly, an average IQ of 80 to 90-has been reached. Most countries in Latin America and the Middle East are reaching this level only now. A secularization trend starting within the next few decades is therefore predictable for these countries. Table 3 and the appendix show that the decline of religion may be starting already in some of the non-communist countries of South Asia and Southeast Asia.

Outside the pale of abandoned communism, tropical Africa is the only world region that experiences a substantial rise in religiosity today (Table 3). Table 4 shows that tropical Africa is also the only world region in which the relationship between education and religiosity is positive rather than negative. The latter may well be due to a positive relationship between religiosity and intelligence. In an Afro-Caribbean population, high verbal intelligence but not high education was found to be an independent predictor of high religiosity (Meisenberg et al., 2006). Although unusual in today's world, the mainly positive relationship of religiosity with intelligence and education in African and African-descended populations conforms to anthropological observations that religious belief and religious ritual both rise with rising cultural complexity in small-scale societies (Zern, 1984).

[FIGURE 1 OMITTED]

Combining this observation with evidence on secularization in modern societies, we can postulate an inverted U-shaped relationship between cultural/cognitive complexity and religion--or at least the more-or-less dogmatic and more-or-less monotheistic religions that prevail in our time. Well-formed religious beliefs, it seems, tend to be most important at intermediate stages of cultural evolution and average population IQs of 70 to 90. Much of Sub-Saharan Africa today still seems to be evolving from a pristine state of vague animism, magic beliefs and practices or relative agnosticism to a more complex condition dominated by intrusive dogmatic religions. This is comparable to the religious evolution of Europe from pre-Roman times to the time of the Crusades.

3. Generational replacement

Table 5 shows that, other things being equal, a strongly positive age-religiosity relationship (old people far more religious than the young) predicts a negative (declining) trend of religiosity over time. This observation needs little comment. It confirms the received wisdom that young people are more apt than the old to endorse new cultural trends whereas the old are more likely to persist in ideas and beliefs that they acquired in their youth. Assuming that today's young people will likewise persist in the beliefs that they acquired in their youth, we can predict that countries with a strongly positive age-religiosity relationship will experience declining religiosity in the next few decades.

4. Emulation of educated individuals

There are two reasons to expect that the beliefs of the educated set the trend for the future. First, social learning is thought to be guided by model-based biases such that beliefs, attitudes, skills and behaviors are learned preferentially from successful individuals. In the words of Alex Mesoudi (2009, p. 939), "...people might preferentially copy models who are particularly successful or prestigious or who are older or richer than themselves....The reason for this is fairly intuitive: Copying a successful person's behavior is likely to result in the adoption of adaptive behavior because successful people become successful because of their adaptive behavior." A second reason to expect such a relationship is that highly educated people control the mass media and the educational system. People are quite gullible (e.g., DellaVigna & Kaplan, 2007), and through their control of the media and the schools, the educated should be able to impose their worldviews on everyone else.

This prediction is thoroughly refuted by the results presented in Tables 6 and 9, which show the opposite relationship. Religion tends to rise in countries in which the less educated are more religious. One possible explanation is that in modern societies, high education is not a trait that most people value for its own sake. People are known to emulate media celebrities, as shown, for example, in a study by Cheng et al. (2007), but not necessarily those with high education or great knowledge or wisdom.

Another reason is suggested by the results shown in Table 9. Here we see that a positive religion-education relationship no longer predicts declining religiosity when the model contains a measure of differential fertility by religiosity. When differential fertility is omitted, the effect of the education-religiosity relationship on the religiosity trend is restored. The likely explanation is that the negative education effect is mediated in large part by differential fertility. The education-fertility relationship is substantially negative in nearly all countries (Meisenberg, 2008b). In countries in which high education is associated with high religiosity, religion is likely to decline because religion is transmitted in the family and the educated have fewer children than the less educated.

We can ask the question of whether religion is the only cultural trait for which the less educated sections of the population are the trendsetters. This question has not been investigated (or even asked) to any extent. However, we must contemplate the possibility that other cultural traits, for example popular music, reading habits and sexual mores, have been trending towards the preferences of the less educated sections of the population in the recent past, and that such trends have been driven to some extent by differential reproduction favoring the less educated.

5. Differential fertility

As seen in the appendix, the relationship between religiosity and fertility is positive in most countries. The likely reason for this is quasi-Darwinian selection. Religious traditions that encourage high fertility are more likely to survive and spread than those that reduce the reproduction of their adherents.

The importance of differential reproduction for the fortunes of religious denominations in the United States during the 20th century has been recognized by earlier workers (Hout et al, 2001). The present results augment this insight by showing that not only religious denomination, but also the strength of religious belief is affected by differential reproduction.

Differential fertility by religiosity is a fairly consistent predictor of the religiosity trend, reaching statistical significance in most models of Tables 7 and 9. The effect of differential fertility is expected because everything else being equal, children's religious beliefs tend to resemble those of their parents (Koenig et al., 2009; Myers, 1996), and a positive relationship between religiosity and fertility increases the proportion of children who grow up in religious homes. What is remarkable is that even the small differences in differential fertility between countries are sufficient to produce significant results when religiosity trends are compared in different countries. Most likely, the effect is strong enough to be visible because parental effects are augmented by peer influence.

Differential fertility needs to be considered in any analysis of religious trends. To give only a few examples of countries with large samples in the World Values Survey, Turkey has a robust positive religiosity-fertility partial correlation of +.174, and religiosity is increasing by .20 points per decade on the zero-to-10 scale. The United States have a religiosity-fertility partial correlation of +.116 and a slightly declining religiosity trend of -.14 points/decade. For Britain and Germany the religiosity-fertility partial correlations are only +.024 and +.017, respectively, and religiosity is declining more steeply at .26 points/decade in Britain and .38 points/decade in Germany. Because Britain, Germany and the United States are very similar on all development indicators, differential fertility appears to be a major reason for the continued strength of religion in the United States but not the European countries.

6. Gender difference

Tables 8 and 9 show that religion tends to rise in countries in which women are far more religious than men. This result is expected if mothers are more important than fathers as transmitters of religious beliefs in a majority of countries. Studies in the United States (e.g., Dudley & Dudley, 1986; Koenig et al., 2009) and Australia (Eaves et al., 1990) did indeed find that children's religiosity resembles their mother's more than their father's. The generally higher female than male religiosity can therefore contribute to the survival of religion in modern societies.

The relationship of religious trends with female religiosity suggests environmental transmission within families. We nevertheless have to ask whether genetic effects are involved as well. Behavior genetic studies have shown that the religiosity of children and adolescents depends on shared family environment rather than shared genes (Abrahamson et al., 2002; Boomsma et al., 1999; Eaves et al., 2008; Koenig et al., 2009). However, beyond age 20 the effect of genes becomes significant (Kendler & Myers, 2009; Koenig et al., 2008), and measures of religiosity in adults can have heritabilities as high as 40% or 50% (Bouchard et al., 1999; Bouchard et al., 2004; Bradshaw & Ellison, 2008; Waller et al., 1990). Therefore we can predict some directional selection on "pious genes" in countries with substantial fertility differentials. The genetic effect is expected to be negligible in a single generation, but it can potentially be cumulative across generations if differential fertility is maintained for a long time. It can contribute to the survival of religion on time scales of centuries to millennia.

Differential fertility and female religiosity may have been important in past historical epochs as well. Rodney Stark (1996) observed that women played important roles in the early Christian communities of the Roman Empire, and speculated that Christians had higher birth rates than pagans because of their rejection of contraception, abortion and infanticide. We will never know whether Stark's analysis of early Christianity is on target, but similar mechanisms seem to be at work today.

7. Limitations of the study

The most obvious limitations of the study are weaknesses in the data, especially for the religiosity trend. Although 15 of the 62 countries with trend data have time series of 24 to 26 years, for 19 countries the time series spans only 5 to 9 years. Longer time series of at least 20 years for all countries would be desirable, but are not available at this time. Not only the trend data, but also the measures for age effect, education effect, gender difference and differential fertility are based on moderate sample sizes and therefore are not expected to be highly accurate.

The statistical significance values need to be interpreted with caution because of possible spatial and cultural autocorrelation, also known as "Galton's problem." The problem is that neighboring countries, and countries with similar culture and history, tend to be similar on many dimensions (Eff, 2004). Therefore the data points are not strictly independent, leading to the possibility of inflated statistical significance values and false positives. However, Tables 5-8 show that the effects of the focal variables are not abolished when world regions are controlled. This means that the effects are not due to the chance association of a predictor variable with religiosity in one or a few world regions.

Another limitation is that we have good data (or any data) only for the recent past. We do not know for how long today's differences in country-level intelligence, differential fertility by religion, and other measures have been in existence. Therefore attempts at reconstructing the causal influences on the secularizing trends of the last two centuries are speculative even if the causal influences on religiosity have not changed over time. In reality, the determinants of religious trends are not necessarily the same across generations. For example, it is possible that rising intelligence and education have been most important for religious decline in the 19th and early 20th centuries, whereas prosperity, social disorder, and/or differential fertility are more important today.

The results must not be interpreted as showing that the religiosity of the world population is stagnating or declining (see last row in Table 3). Fertility rates are higher in high-religiosity countries than in low-religiosity countries. The country-level correlation (Pearson's r) between total fertility rate (1975-2004 average) and religiosity is .657 (N = 94 countries). Therefore the proportion of the world population living in high-religiosity countries is rising rapidly. The average religiosity of the world population is therefore rising today, and it will almost certainly keep rising through most or all of the 21st century, even if the average religiosity within most countries declines. We can expect that religion will remain strong in those countries in which religious people have substantially higher fertility rates than atheists. Finally, because secularization appears to be driven by high levels of intelligence and social harmony, religion is predicted to re-emerge as a major force during the collapse of modern civilization--whenever that may be.

Appendix

Religiosity: Average religiosity on a scale from zero (atheist) to 10 (maximally religious).

Trend: Change in religiosity per decade.

Age effect: Extent to which religiosity rises for every decade of chronological age.

Education effect: Extent to which the religiosity of the most educated persons (university degree, schooling to age 31 +) differs from the religiosity of unschooled persons.

Gender effect: Extent to which female religiosity exceeds male religiosity.

Differential reproduction: Percent difference in number of children between the most religious individuals (religiosity score of 10) relative to the nonreligious (religiosity score of 0). Data points that are based on a total sample size of less than 1400 or for which at least one of the demographic categories (e.g., young males, old females...) has a sample size of less than 300 are in parentheses. These should be considered less accurate than data derived from larger samples.

Country Religiosity Trend Age
 effect

Albania 7.65 0.27
Algeria 8.67 0.14
Andorra 6.05 0.52
Argentina 8.24 0.41 0.18
Armenia 7.65 -0.13
Australia 6.84 -0.42 0.33
Austria 7.58 0.28 0.23
Azerbaijan 8.91 0.05
Bangladesh 9.45 0.39 0.05
Belarus 6.42 1.33 0.28
Belgium 6.70 -0.37 0.36
Bosnia 7.77 -0.08
Brazil 9.25 0.13 0.09
Bulgaria 6.06 1.07 0.19
Burkina Faso 9.19 0.01
Canada 7.84 -0.16 0.33
Chile 8.52 -0.35 0.19
China 4.00 0.99 -0.03
Colombia 9.22 -0.05 0.09
Croatia 7.72 0.23
Cyprus 8.13 0.25
Czech Rep. 5.07 -0.46 0.51
Denmark 5.90 0.05 0.50
Dominican R. 8.74 -0.04
Egypt 9.64 0.03 0.04
El Salvador 9.01 0.09
Estonia 5.44 -0.25 0.26
Ethiopia 9.04 0.17
Finland 6.85 0.15 0.38
France 5.70 -0.19 0.38
Georgia 8.91 0.79 -0.14
Germany 5.63 -0.38 0.40
Ghana 9.55 -0.02
Greece 8.06 0.22
Guatemala 9.05 0.03
Hong Kong 5.26 0.17
Hungary 5.97 0.72 0.53
Iceland 7.55 -0.02 0.42
India 8.46 -0.17 0.08
Indonesia 9.38 -0.34 0.06
Iran 9.35 -0.24 0.05
Iraq 9.22 0.07
Ireland 8.42 -0.07 0.29
Israel 8.25 -0.10
Italy 8.05 0.25 0.22
Japan 5.46 -0.12 0.26
Jordan 9.56 0.13 0.06
S. Korea 5.69 -0.25 0.25
Kyrgyzstan 8.31 -0.00
Latvia 6.92 0.71 0.15
Lithuania 7.47 0.63 0.36
Luxembourg 6.54 0.36
Macedonia 7.84 -0.05
Malaysia 8.67 0.03
Mali 9.42 0.01
Malta 9.18 -0.30 0.15
Mexico 8.63 -0.00 0.14
Moldova 8.48 0.46 0.10
Morocco 9.67 -0.50 0.06
Netherlands 6.28 -0.24 0.37
New Zealand 6.49 -0.48 0.28
Nigeria 9.48 0.10 0.02
Norway 5.98 -0.33 0.47
Pakistan 9.54 0.03
Peru 9.05 -0.17 0.15
Philippines 9.23 -0.30 0.03
Poland 8.86 0.05 0.13
Portugal 8.17 0.39 0.29
Puerto Rico 9.22 -0.16 0.11
Romania 8.80 0.87 0.15
Russia 6.52 1.03 0.19
Rwanda 9.48 0.01
Saudi Arabia 9.27 0.04
Serbia & M 7.00 1.76 0.10
Singapore 8.64 0.14
Slovakia 7.34 0.48 0.38
Slovenia 6.50 0.22 0.27
S. Africa 8.94 0.13 0.07
Spain 7.15 -0.42 0.43
Sweden 5.19 -0.15 0.40
Switzerland 7.24 -0.38 0.33
Taiwan 6.60 -0.88 0.24
Tanzania 9.46 0.02
Thailand 7.31 0.11
Trinidad &T. 9.33 0.07
Turkey 8.99 0.20 0.13
Uganda 9.31 0.01
Ukraine 7.29 0.98 0.07
United K. 6.38 -0.26 0.43
N. Ireland 7.86 -0.36 0.37
USA 8.61 -0.14 0.16
Uruguay 7.26 0.20 0.26
Venezuela 9.11 0.06
Vietnam 5.12 1.30 0.05
Zambia 9.23 -0.03
Zimbabwe 9.31 0.00

Country Education Gender Differential
 effect effect reproduction

Albania -1.55 0.90 21.0
Algeria -0.05 0.02 (39.2)
Andorra -2.67 1.14 (38.4)
Argentina -1.04 0.77 46.8
Armenia -0.74 1.08 20.7
Australia -0.44 0.89 24.4
Austria -0.83 0.87 50.2
Azerbaijan -0.56 0.23 53.6
Bangladesh -0.26 0.11 (58.0)
Belarus -1.51 1.37 17.7
Belgium -0.58 0.80 41.5
Bosnia -2.20 0.50 16.1
Brazil -0.41 0.26 17.8
Bulgaria -2.52 1.18 18.4
Burkina Faso 0.33 0.17 (22.4)
Canada -1.13 0.83 40.5
Chile -0.76 0.71 20.8
China -1.48 0.42 15.6
Colombia -0.24 0.32 45.8
Croatia -1.86 0.60 41.7
Cyprus -0.23 0.82 (-2.6)
Czech Rep. -0.15 0.75 10.6
Denmark -0.46 0.95 22.8
Dominican R. 0.17 0.24
Egypt 0.08 0.10 1.0
El Salvador 0.17 0.22 (-17.4)
Estonia -0.74 1.05 -9.3
Ethiopia -0.23 0.24 (14.9)
Finland 0.32 1.23 19.1
France -0.72 0.87 18.9
Georgia 0.06 0.44 -8.9
Germany -0.84 0.84 2.4
Ghana 0.24 0.04 (7.1)
Greece -3.05 0.79 (27.8)
Guatemala -0.25 0.24 (22.4)
Hong Kong 1.46 0.77 (0.6)
Hungary -2.92 1.25 20.6
Iceland -1.66 1.09 31.3
India -0.57 0.42 21.1
Indonesia 0.48 0.13 33.2
Iran -0.36 0.12 63.8
Iraq 0.01 0.12 28.5
Ireland -0.35 0.58 29.9
Israel -2.81 0.57
Italy -1.13 0.96 52.3
Japan -0.84 0.74 3.4
Jordan -0.07 0.27 14.8
S. Korea 2.06 1.13 6.0
Kyrgyzstan -2.31 0.31 (42.0)
Latvia -1.30 1.46 -7.1
Lithuania -1.29 1.10 13.7
Luxembourg -2.26 0.68 (54.1)
Macedonia -3.18 0.23 37.2
Malaysia 1.06 0.24 (10.1)
Mali 0.32 -0.02 (5.4)
Malta -0.58 0.38 28.9
Mexico -0.82 0.51 13.9
Moldova -2.16 0.76 36.9
Morocco -0.05 0.04 -6.5
Netherlands -0.85 0.65 33.8
New Zealand 0.28 0.83 29.6
Nigeria 0.00 0.07 -24.2
Norway -0.52 1.16 18.4
Pakistan -0.13 0.05 -27.0
Peru -0.27 0.45 9.7
Philippines 0.20 0.16 -5.9
Poland -2.09 0.49 46.2
Portugal -2.69 1.07 56.4
Puerto Rico -0.12 0.28 29.5
Romania -0.93 0.63 45.2
Russia -1.18 1.51 10.4
Rwanda -0.17 0.03 (-15.5)
Saudi Arabia 0.06 -0.02 (-79.6)
Serbia & M -3.16 0.71 16.5
Singapore -1.12 0.32 (41.7)
Slovakia -2.83 0.86 21.6
Slovenia -4.23 0.74 33.3
S. Africa 0.54 0.60 36.7
Spain -0.76 1.21 27.5
Sweden 0.41 0.97 1.6
Switzerland -0.35 0.79 33.8
Taiwan -1.08 0.69 6.4
Tanzania 0.31 0.11 (53.0)
Thailand 0.10 0.01 18.3
Trinidad &T. -0.16 0.20 (-30.1)
Turkey -2.15 0.36 80.2
Uganda 0.32 0.18 (38.4)
Ukraine -1.33 1.24 16.9
United K. 0.43 0.95 7.2
N. Ireland -0.37 0.84 19.5
USA -0.35 0.53 45.0
Uruguay -1.99 1.23 26.2
Venezuela 0.08 0.28 18.8
Vietnam -1.20 0.82 17.4
Zambia 0.69 0.14 (-3.9)
Zimbabwe 0.93 0.35 (-9.9)

References

Abrahamson, Amy C., L.A. Baker & A. Caspi (2002) Rebellious teens? Genetic and environmental influences on the social attitudes of adolescents. Journal of Personality and Social Psychology 83: 1392-1408.

Atkinson, Anthony B. (2004) Increased income inequality in OECD countries and the redistributive impact of the government budget. In: A. Cornia (ed): Inequality Growth and Poverty in an Era of Liberalization and Globalization, pp. 221-248. Oxford: Oxford Univ. Press.

Beaujean, A. Alexander & S.J. Osterlind (2008) Using item response theory to assess the Flynn effect in the National Longitudinal Study of Youth 79 children and young adults data. Intelligence 36: 455-463.

Berger, Peter L. (1999) The Desecularization of the World: Resurgent Religion and World Politics. Grand Rapids, MI: W.B. Eerdmans Publishing.

Blair, Clancy, D. Gamson, S. Thorne & D. Baker (2005) Rising mean IQ: cognitive demand of mathematics education for young children, population exposure to formal schooling, and the neurobiology of the prefrontal cortex. Intelligence 33: 93-106.

Boomsma, Dorrett I., E.J.C. de Geus, G.C.M. van Baal & J.R. Koopmans (1999) A religious upbringing reduces the influence of genetic factors on disinhibition: evidence for interaction between genotype and environment on personality. Twin Research 2: 115-125.

Bouchard, Thomas J., M. McGue, D. Lykken & A. Tellegen (1999) Intrinsic and extrinsic religiousness: genetic and environmental influences and personality correlates. Twin Research 2: 88-98.

Bouchard, Thomas J., N.L. Segal, A. Tellegen, M. McGue, M. Keyes & R. Krueger (2004) Genetic influence on social attitudes: another challenge to psychology from behavior genetics. In: Lisabeth F. DiLalla (ed): Behavior Genetics Principles. Perspectives in Development, Personality, and Psychopathology, pp. 89-104. Washington DC: American Psychological Association.

Bradshaw, Matt & C.G. Ellison (2008) Do genetic factors influence religious life? Findings from a behavior genetic analysis of twin siblings. Journal for the Scientific Study of Religion 47: 529-544.

Ceci, Stephen J. & W.M. Williams (1997) Schooling, intelligence, and income. American Psychologist 52: 10511058.

Cheng, Andrew T.A., K. Hawton, C.T.C. Lee & T.H.H. Chen (2007) The influence of media reporting of the suicide of a celebrity on suicide rates: a population-based study. International Journal of Epidemiology 36: 1229-1234.

Colom, Roberto, C.E. Florez-Mendoza & F.J. Abad (2006) Generational changes on the draw-a-man test: a comparison of Brazilian urban and rural children tested in 1930, 2002 and 2004. Journal of Biosocial Science 35: 33-39.

Cotton, Sue M., P.M. Kiely, D.P. Crewther, B. Thomson, R. Laycock & S.G. Crewther (2005) A normative and reliability study for the Raven's Coloured Progressive Matrices for primary school aged children from Victoria, Australia. Personality and Individual Differences 39: 647-659.

DellaVigna, Stefano & E. Kaplan (2007) The Fox News effect: media bias and voting. Quarterly Journal of Economics 122: 1187-1234.

Delamontagne, R. Georges (2010) High religiosity and societal dysfunction in the United States during the first decade of the twenty-first century. Evolutionary Psychology 8: 617-657

Dickens, William T. & J.R. Flynn (2001) Heritability estimates versus large environmental effects: the IQ paradox resolved. Psychological Review 108: 346-369.

Dudley, Roger L. & M.G. Dudley (1986) Transmission of religious values from parents to adolescents. Review of Religious Research 28: 3-15.

Eaves, Lindon J., N.G. Martin & A.C. Heath (1990) Religious affiliation in twins and their parents: testing a model of cultural inheritance. Behavior Genetics 20: 1-22.

Eaves, Lindon J., P.K. Hatemi, E.C. Prom-Womley & L. Murrelle (2008) Social and genetic influences on adolescent religious attitudes and practices. Social Forces 86: 1621-1647.

Eff, E.A. (2004) Spatial and cultural autocorrelation in international datasets. Department of Economics and Finance Working Paper Series, February 2004, retrieved 21 February 2006 from: www.mtsu.edu/~berc/working/spatial%20autocorrelation%20z.pdf

Finke, Roger & R. Stark (1993) The Churching of America 1776-1990. Winners and Losers in Our Religious Economy. Newark (NJ): Rutgers Univ. Press.

Flynn, James R. (1987) Massive IQ gains in 14 nations: what IQ tests really measure. Psychological Bulletin 101: 171-191.

Grim, Brian J. & R. Finke (2006) International religion indexes: government regulation, government favoritism, and social regulation of religion. Interdisciplinary Journal of Research on Religion 2: article 1 (online).

Heston, Alan, R. Summers & B. Aten (2009) Penn World Table version 6.3. Center for International Comparisons of Production, Income and Prices at the University of Pennsylvania, accessed Dec 15, 2010, at: http://pwt.econ.upenn.edu/php_site/pwt63/pwt63_retrieve.php.

Hout, Michael, A. Greeley & M.J. Wilde (2001) The demographic imperative in religious change in the United States. American Journal of Sociology 107: 468-500.

Iannaccone, Laurence R., R. Finke & R. Stark (1997) Deregulating religion: the economics of church and state. Economic Inquiry 35: 350-364.

Inglehart, Ronald, M. Basanez, J. Diez-Medrano, L. Halman & R. Luijkx (2004) Human Beliefs and Values. A Cross-Cultural Sourcebook Based on the 1999-2002 Values Surveys. Mexico: Siglo XXI Editores.

Kendler, Kenneth S. & J. Myers (2009) A developmental twin study of church attendance and alcohol and nicotine consumption: a model for analyzing the changing impact of genes and environment. American Journal of Psychiatry 166: 11501155.

Khaleefa, Omar, A. Sulman & R. Lynn (2009) An increase of intelligence in Sudan, 1987-2007. Journal of Biosocial Science 41: 279-283.

Koenig, Laura B., M. McGue & W.G. Iacono (2008) Stability and change in religiousness during emerging adulthood. Developmental Psychology 44: 532-543.

Koenig, Laura B., M. McGue & W.G. Iacono (2009) Rearing environmental influences on religiousness: an investigation of adolescent adoptees. Personality and Individual Differences 47: 652656.

Lynn, Richard (1990) The role of nutrition in secular increases in intelligence, Personality and Individual Differences 11: 273-285.

Lynn, Richard (2008) The Global Bell Curve. Augusta, GA: Washington Summit. Lynn, Richard

(2010) National IQs updated for 41 nations. Mankind Quarterly 50: 275-296. Lynn, Richard, J. Harvey & H. Nyborg

(2009) Average intelligence predicts atheism rates across 137 nations. Intelligence 37: 11-15.

Lynn, R., & G. Meisenberg (2010) National IQs calculated and validated for 108 nations. Intelligence 38: 353-360.

Lynn, Richard & T. Vanhanen (2006) IQ & Global Inequality. Augusta, GA: Washington Summit.

Meisenberg, Gerhard (2004) Talent, character, and the dimensions of national culture. Mankind Quarterly 45: 123-168.

Meisenberg, Gerhard (2007) Does multiculturalism promote income inequality? Mankind Quarterly 47: 3-39.

Meisenberg, Gerhard (2008a) How does racial diversity raise income inequality? Journal of Social, Political and Economic Studies 33: 3-26.

Meisenberg, Gerhard (2008b) How universal is the negative correlation between education and fertility? Journal of Social Political and Economic Studies 33: 205-227.

Meisenberg, Gerhard, E. Lawless, E. Lambert & A. Newton (2005) The Flynn effect in the Caribbean: generational change of cognitive test performance in Dominica. Mankind Quarterly 46: 29-69.

Meisenberg, Gerhard, E. Lawless, E. Lambert & A. Newton (2006) The social ecology of intelligence on a Caribbean island. Mankind Quarterly 46: 395-433.

Meisenberg, Gerhard & A. Kaul (2010) Effects of sex, race, ethnicity and marital status on the relationship between intelligence and fertility. Mankind Quarterly 50: 151-187.

Mesoudi, Alex (2009) How cultural evolutionary theory can inform social psychology and vice versa. Psychological Review 116: 929-952.

Mingroni, Michael A. (2007) Resolving the IQ paradox: heterosis as a cause of the Flynn effect and other trends. Psychological Review 114: 806-829.

Myers, Scott M. (1996) An interactive model of religiosity inheritance: the importance of family context. American Sociological Review 61: 858-866.

Paul, G.S. (2009) The chronic dependence of popular religiosity upon dysfunctional psychosociological conditions. Evolutionary Psychology 7: 398-441.

Planck, Max (1948) Wissenschaftliche Selbstbiographie. Leipzig: Barth Verlag.

Shayer, Michael & D. Ginsburg (2009) Thirty years on--a large anti-Flynn effect? (II): 13- and 14-year-olds. Piagetian tests of formal operations norms 1976-2006/7. British Journal of Educational Psychology 79: 409-418.

Stark, Rodney (1996) The Rise of Christianity. Princeton (NJ): Princeton University Press.

Stark, Rodney & L.R. Iannaccone (1994) A supply-side reinterpretation of the "secularization" of Europe. Journal for the Scientific Study of Religion 33: 230-252.

Sundet, Jon Martin, D.G. Barlaug & T.M. Torjussen (2004) The end of the Flynn effect? A study of secular trends in mean intelligence test scores of Norwegian conscripts during half a century. Intelligence 32: 349-362.

Teasdale, Thomas W. & D.R. Owen (2008) Secular declines in cognitive test scores: a reversal of the Flynn effect. Intelligence 36: 121-126.

United Nations (2004) Human Development Report 2004. United Nations, retrieved 15 Oct 2008 at: http://hdr.undp.org/en/media/HDR04_complete.pdf

United Nations (2005) Human Development Report 2005. United Nations, retrieved 10 Oct 2008 at; http://hdr.undp.org/en/media/HDR05_complete.pdf

Waller, Niels G., B.A. Kojetin, T.J. Bouchard, D.T. Lykken & A. Tellegen (1990) Genetic and environmental influences on religious interests, attitudes and values: a study of twins reared apart and together. Psychological Science 1: 138-142.

Zern, David S. (1984) Religiousness related to cultural complexity and pressures to obey cultural norms. Genetic Psychology Monographs 110: 207-227.

Gerhard Meisenberg *

Ross University Medical School, Dominica (Eastern Caribbean)

* Address for correspondence: Gmeisenberg@rossmed.edu.dm

Table 1 Levels of religious belief (Belief) and other measures, by
world regions. Religious belief and education are on zero-to-10 scales,
and intelligence on the 100/15 IQ scale. GDP, per-capita GDP adjusted
for purchasing power. N values indicate the number of respondents with
information about religious belief in the World Values Survey.

Region N countries Belief N

Prot. Europe 8 6.14 38835
Cath. Eur. / Med. 11 7.27 40888
English-speaking 6 7.60 32583
Ex-communist 23 7.17 83857
Latin America 13 8.78 43481
Middle East 9 9.33 36124
South(east) Asia 7 8.36 22157
East Asia 6 5.57 17334
Africa 11 9.25 23509
World 94 7.69 345743

Region Education Intelligence GDP

Prot. Europe 8.85 99.0 24706
Cath. Eur. / Med. 7.32 95.8 21796
English-speaking 9.07 98.3 23134
Ex-communist 6.79 92.8 6186
Latin America 5.98 85.1 8090
Middle East 4.67 84.6 4950
South(east) Asia 5.07 87.2 4221
East Asia 7.19 105.3 19391
Africa 3.10 71.6 1596
World 6.74 90.1 10975

Table 2 Correlations of country-level religiosity and religiosity
trend (Rel. trend) with several predictors. Correlations below the
diagonal are for all countries, and correlations above the diagonal
are for countries without communist history only. Sample sizes:
N = 92 (all countries) and 67 (non-communist countries), except
correlations with trend, which are for 61 and 43 countries,
respectively. Statistical significance of the trend correlations:
*, p<.05; **, p<.01; ***, p<.001. For all other comparisons,
correlations higher than .215 (92 countries) or .255 (61
countries) are significant at p<.05.

 Religiosity R. Intell. Edu
 trend

Religiosity 1 .298 -.760 -.723
Rel. trend -.130 1 -.318 * -.259
Intelligence -.736 -.063 1 .849
Education -.592 -.221 .799 1
lgGDP -.497 -.423 ** .781 .857
Freedom/ -.414 -.451 *** .599 .755
Dem.
No -.464 -.499 *** .630 .723
corruption

 lgGDP Fr./Dem No
 Corr.

Religiosity -.698 -.726 -.742
Rel. trend -.237 -.067 -.253
Intelligence .891 .764 .796
Education .906 .827 .826
lgGDP 1 .819 .824
Freedom/ .817 1 .779
Dem.
No .813 .785 1
corruption

Table 3 Change in religious belief per decade (Trend). Religious
belief is measured on the same zero-to-10 scale as in Table 1.
Numbers in parentheses are for countries without communist history
only.

Region Trend N countries N respondents

Prot. Europe -.191 8 37,530
Cath. Eur. -.166 7 35,316
 / Mediterranean
English-speaking -.224 7 32,583
Ex-communist .639 16 69,044
Latin America .031 8 37,407
Middle East .005 5 25,585
South/ Southeast -.020 (-.128) 5 (4) 19,422 (16936)
 Asia
East Asia .038 (-.266) 4 (3) 14,571 (10663)
Africa .122 2 11,818
World .039 (-.126) 62 (44) 283,276 (207,838)

Table 4 Effects of age, gender, and education on religiosity
in 96 countries and territories, and differential fertility by
religion. Shown are the effects of each decade of age, female
gender, and an increase in the educational level from no schooling
to university graduate. Differential reproduction is shown as the
percent difference in number of children of the most religious
(religiosity score of 10) compared to staunch atheists
(religiosity score of zero). See also explanation of the
variables in the appendix.

World region Age Gender Education
 effect effect effect

Prot. Europe .406 .917 -0.490
Cath. Europe .332 .925 -1.090
English-speaking .307 .759 -0.355
Excommunist .188 .870 -1.693
Latin America .135 .473 -0.539
Middle East .072 .114 -0.549
South Asia .063 .285 -0.238
East Asia .165 .654 -0.395
Africa .034 .259 0.321
World .172 .644 -0.776

World region Diff. N (countries)
 Reprod.

Prot. Europe 15.8 8
Cath. Europe 36.3 12
English-speaking 30.0 7
Excommunist 20.7 23
Latin America 20.9 13
Middle East 28.0 9
South Asia 23.8 7
East Asia 10.7 6
Africa 10.8 11
World 22.7 96

Table 5 Prediction of the religiosity trend with the relationship
between religiosity and age. Age effect is the extent to which
religiosity rises or declines with advancing age in each country
(see text). Standardized [beta] coefficient and t statistic are
shown.

 Model 1 Model 2

 t t

Religiosity -.519 3.69
Intelligence -.368 2.33
Education .001 0.01
lgGDP .170 0.75
No corruption -.146 0.75
Freedom/Democr. .177 0.99
Communism .500 3.55
Age effect -.473 3.37 -.458 3.17
Catholic Europe -.079 0.60
English-speaking -.150 1.17
Ex-communist .361 2.09
Latin America -.178 1.16
Middle East -.253 1.70
South Asia -.116 0.79
East Asia -.128 1.01
Africa -.092 0.76
% Orthodox
% Buddhist
% Unaffiliated
N 62 61
Adj. R2 .418 .610

 Model 3 Model 4

 t t

Religiosity -.498 3.66 -.652 5.00
Intelligence -.341 2.56
Education
lgGDP
No corruption -.153 1.26
Freedom/Democr. .208 1.50
Communism .531 4.80 .248 1.86
Age effect -.460 3.45 -.455 3.68
Catholic Europe
English-speaking
Ex-communist
Latin America
Middle East
South Asia
East Asia
Africa
% Orthodox .120 1.14
% Buddhist -.296 3.66
% Unaffiliated -.215 2.06

N 61 61
Adj. R2 .626 .700

Table 6 Prediction of the religiosity trend with the
relationship between religiosity and education.
Education effect is the extent to which religiosity
differs between people with different exposure to
formal education (see text). Standardized [beta]
coefficient and t statistic are shown.

 Model 1 Model 2

 t t

Religiosity -.387 2.86
Intelligence -.323 2.01
Education .191 0.91
lgGDP -.028 0.12
No corruption -.306 1.54
Freedom/Democr. -.096 0.52
Communism .246 1.37
Education effect -.307 2.44 -.295 2.60
Catholic Europe .015 0.11
English-speaking -.044 0.33
Ex-communist .507 3.04
Latin America .083 0.61
Middle East .043 0.35
South Asia .192 1.54
East Asia .057 0.47
Africa .133 1.17
% Orthodox
% Buddhist

N 62 61
Adj. R2 .363 .588

 Model 3 Model 4

 t t

Religiosity -.378 2.90 -.388 4.54
Intelligence -.306 2.16
Education
lgGDP
No corruption -.231 1.71 -.323 2.40
Freedom/Democr. -.148 1.15
Communism .347 2.49
Education effect -.248 2.53 -.266 3.31
Catholic Europe
English-speaking
Ex-communist
Latin America
Middle East
South Asia
East Asia
Africa
% Orthodox .296 3.52
% Buddhist -.217 2.76

N 61 61
Adj. R2 .604 .676

Table 7 Prediction of the religiosity trend with the
religiosity-fertility partial correlation (Differential
fertility). Standardized [beta] coefficient and t
statistic are shown.

 Model 1 Model 2

 t t

Religiosity -.401 3.08
Intelligence -.384 2.45
Education .118 0.63
lgGDP .008 0.04
No corruption -.272 1.43
Freedom/Democr. -.122 0.69
Communist history .445 3.15
Differential fertility .257 2.23 .307 3.36
Catholic Europe .005 0.04
English speaking -.081 0.61
Excommunist .712 4.60
Latin America .139 1.02
Middle East .077 0.61
South Asia .213 1.67
East Asia .090 0.73
Africa .145 1.26
% Orthodox

N 62 61
adj. R2 .352 .618

 Model 3 Model 4

 t t

Religiosity -.400 3.17 -.429 3.59
Intelligence -.373 2.69 -.335 2.54
Education
lgGDP
No corruption -.250 1.92 -.258 2.10
Freedom/Democr.
Communist history .496 4.14 .316 2.42
Differential fertility .283 3.38 .271 3.42
Catholic Europe
English speaking
Excommunist
Latin America
Middle East
South Asia
East Asia
Africa
% Orthodox .261 2.77

N 61 61
adj. R2 .634 .673

Table 8 Prediction of the religiosity trend with the
relationship between religiosity and the extent to
which women's religiosity exceeds men's (Gender
difference, for explanation see text). Standardized
[beta] coefficient and t statistic are shown.

 Model 1 Model 2

 t t

Religiosity -.197 1.30
Intelligence -.283 1.74
Education -.141 0.73
lgGDP .052 0.22
No corruption -.200 1.00
Freedom/Democr. -.060 0.33
Communist history .446 2.98
Gender difference .286 1.92 .336 2.39
Catholic Europe .086 0.64
English speaking -.020 0.15
Ex-communist .669 4.30
Latin America .206 1.40
Middle East .209 1.36
South Asia .307 2.12
East Asia .086 0.69
Africa .178 1.45
% Orthodox

N 62 61

 Model 3 Model 4

 t t

Religiosity -.231 1.65 -.225 1.64
Intelligence -.305 2.13 -.248 1.81
Education -.296 2.41
lgGDP
No corruption -.300 2.11
Freedom/Democr.
Communist history .422 3.23 .367 3.13
Gender difference .286 2.32 .295 2.33
Catholic Europe
English speaking
Ex-communist
Latin America
Middle East
South Asia
East Asia
Africa
% Orthodox .279 2.74

N 61 61

Table 9 Prediction of the religiosity trend with
religiosity-education B coefficient (Education effect),
religiosity-fertility partial correlation (Differential
fertility), and the extent to which women's religiosity
exceeds men's (Gender difference). Standardized [beta]
coefficient and t statistic are shown. Education and
log-transformed GDP are omitted because they were
ineffective in all models.

 Model 1 Model 2

 t t

Religiosity -.256 1.89
Intelligence -.350 2.57
No corruption -.221 1.44
Freedom/Democr. -.204 1.38
Communist history .331 2.33
Education effect -.210 1.37 -.089 0.80
Differential fertility .122 0.87 .242 2.41
Gender difference .250 1.73 .302 2.45
Catholic Europe .019 0.14
English speaking -.019 0.15
Excommunist .575 3.23
Latin America .195 1.36
Middle East .203 1.37
South Asia .320 2.29
East Asia .110 0.90
Africa .220 1.84

N 62 61
adj. R2 .387 .659

 Model 3 Model 4

 t t

Religiosity -.272 2.00 -.252 1.87
Intelligence -.326 2.37 -.357 2.63
No corruption -.316 2.31 -.223 1.46
Freedom/Democr. -.199 1.35 -.199 1.35
Communist history .288 2.09 .386 3.12
Education effect -.226 2.36
Differential fertility .287 3.46
Gender difference .255 2.15 .305 2.49
Catholic Europe
English speaking
Excommunist
Latin America
Middle East
South Asia
East Asia
Africa

N 61 61
adj. R2 .628 .661