The decline in the daily importance of religion in economic development is a well-known correlation, but which phenomenon comes first? Using unsupervised factor badysis and a birth cohort approach to create a retrospective time series, we present 100-year time series of secularization in different countries, derived from recent global values surveys. . We find evidence that an increase in secularization generally preceded economic growth over the last century. Our multi-level and time-lag regressions also indicate that tolerance for individual rights predicted economic growth in the twentieth century even more than secularization. These results are valid when we control education and shared cultural heritage.
INTRODUCTION
A clbadic sociological question is whether the decline in religious activity, or secularization, was caused by economic development (19459012). 2 ). A century ago Durkheim ( 3 ) proposed that technological and socio-economic advances supplant the functions of religion ( 4 5 ), while Weber ( 19459012)] [6] argued the opposite, that the monotheistic religion – so-called Protestant ethics – made possible the development of capitalism.
Although a correlation between economic development and secularization is evident, very religious countries tend to be the poorest ( 7 8 ); is not obvious what change precedes which through time: if development causes secularization (19459012) 9 10 ), or vice versa ( 11 ), or if both changes are trained, with different delays, by a factor such as education or technological progress ( 2 12 ).
While some studies find a bicausal relationship between e and religion ( 13 – 15 ), causality remains effectively unknown because feedbacks can change over time and over time. development. Organized religious charity, for example, may initially encourage certain values that facilitate economic development while restricting individual expression ( 16 ), but the resulting economic development may then reward ;individualism. Tolerance of individual expression can then feed secularism, in part by undermining religious organizations that provide communities with resources and social capital; Other causal arguments also remain feasible 17 18 .
To characterize their temporal relationship, we use 20th century data for economic development and a measure of secularization extracted from international relations, cultural values surveys Delayed regressions using these 100-year time series for key variables can determine which variable precedes the other.This can be used to exclude some causality hypotheses. the changes in the series X precede those of the series Y then we can say that Y does not cause X whereas it is not certain that X causes Y .
For indices of cultural values, we use data from the European Values Survey (EVS) and the World Values Survey (WVS) since 1990. To estimate values for all decades of the 20th century, here we use birth cohorts (see Materials and Methods); The utility of this technique is among our main conclusions. The WVS and EVS involved some of the same survey questions, so we combined these sets of survey data, which we call the WEVS. The data we use for economic development are data on the historical gross domestic product (GDP) per capita of the Maddison project ( 19 ), although more complex indices such as the index of human development may be preferable ( 20 ), we use because GDP data exist for many nations throughout the 20th century. We also include three control variables, each covering all the nations of the world and extending up to the early 20th century. The first, which became available recently, is a vast international time series on participation rates in higher education, which dates back to the early twentieth century ( 21 ). The second is the language family, which is used as an indirect indicator of the nested random effect of non-independent cultural and economic histories between different countries ( 22 ). The third is a measure of the tolerance of others, extracted as a factor of the WEVS, as discussed below.
RESULTS
After applying exploratory factor badysis (FTE) to the WEVS dataset, we selected nine factors to remember, each interpretable using questions that do not overlap between factors (see Material and methods). Factor 1, accounting for 11% of the variance, was heavily loaded on questions about the importance of religion in life (see Materials and Methods). This factor defines our measure of secularization, S as a composite variable of WEVS responses (Table S4 lists all elements of the secularization factor).
Another factor of the EFA, the factor 8, was heavily loaded participants' willingness to tolerate behaviors often prohibited socially, such as suicide, homobaduality or abortion (see table S11). We call this factor "tolerance", noted V . We explore tolerance, V as a control variable in our results for two reasons. First, as we shall see, changes in tolerance were closely correlated with changes in secularism during the twentieth century. Second, the tolerance factor was strongly correlated ( R = 0.59) with Hofstede's metric of individualism ( 23 ) (we did not Extracts individualism directly from Hofstede's data
After extracting these factors measuring secularization, S and tolerance, V we seek to extend the information collected in the WEVS during the last quarter-the return to the beginning of the 20th century.To do this, we treat the decade of birth, t recorded for each person surveyed, as an approximation of a historical period. Although there are differences from one period of investigation, p to the other, the differences apply to all birth cohorts of so that the relative differences between cohorts of births are maintained gen
Using a likelihood ratio test of the hypotheses presented in Materials and Methods, we confirm that the estimates for each decade of birth, t are independent of the period of investigation, p in that there is no evidence of temporal dependence for S t p or for V t t p in 91% and 89% of the countries tested, respectively (Table S13 shows the complete results). This confirms that generational trends persist over time. The results constitute the elements of a table, S i t p estimates of secularization in each country, i at the time of birth, t for the period of investigation, p .
Next, we determined secularization by decade of birth in each country, S t by averaging factor 1 for all available survey periods, p in the country, i corresponding to the decade of birth, ] t . Figure 1 illustrates the temporal trend of secularization over the decade of birth for several countries with few missing data: Britain, the Philippines, Chile, and Nigeria. For countries with missing data, the missing values of S i t were imputed (see Material and Methods). The same procedure is used to obtain the tolerance score matrix V i t .
Fig. 1 Temporal trends of secularization versus economic development in the 20th century, for four illustrative countries
Each panel represents the secularization score of a country S derived from the WEVS on y ] axis, for cohorts of birth by decade t on the axis x . The trends are independently determined from each of the five different survey periods, p corresponding to five waves of the WEVS: p 1 1990-1994; p 2 1995-1999; p 3 2000-2004; p 4 2005-2009; p 5 2010 to 2014.
Fig. 1 Temporal trends of secularization versus economic development during the twentieth century, for four illustrative countries
Each panel represents the secularization score of a S country derived from the WEVS on the US $. axis y cohorts by decade t on the axis x Trends are independently determined from each of the five different survey periods, p corresponding to five waves of the WEVS: p 1 1990-1994; p 2 1995-1999; p ] 2000-2004; ] p 4 2005-2009; p 5 2010-2014.
We compared S i t in relation to the per capita historical GDP (in 1990 US $) of each country over time. Figure 2 compares S i t relative to the ten-year average GDP, GDP i t for these same four countries as Fig. 1. We find evidence that changes in secularization, S i t precede changes in GDP i t [19659019] as can be seen most clearly in the reversals of the trend, when a decrease in S occurs shortly before a corresponding decrease in GDP.
Fig. 2 Time series of secularization versus GDP per capita, from four illustrative countries, during the 20th century
Each red line represents the average score of secularization, S t [19659028] from birth cohort in the decade, t for this country. Each blue line represents the average GDP per capita (normalized in 1990 US dollars) during the decade t in this country
"data-icon-position =" "data-hide-link- title = "0">
Fig. 2 Time series of secularization in relation to GDP per capita, from four illustrative countries, during the twentieth century
Each red line represents the average score of secularization , S birth cohort in the decade, t for each country. Each blue line represents average per capita GDP (normalized to US $ 1990) during the decade t To test whether the changes in S i t generally precede the changes in GDP i t or, conversely, we multi-level regressions lagged in time. Including data from all countries in a single test, a multilevel model can maximize the statistical power available in these data. we must control non-independence because of shared cultural heritage h what we do by clbadifying countries by language family (see Materials and Methods). The multilevel model is (1) (2) where S t ]] and GDP t are secularization and economic development in the decade t respectively, and S t – and ] and the GDP [19659049] t – y are the respective values delayed by and decades. The term h i represents a nested random effect because of the historico-cultural grouping of the country i for which the linguistic family is the proxy ( 22 ). This term is used as a check for the non-independent similarities between countries, due to their development and secularization already present at the beginning of the 20th century.
Models 2 and 5 (Table 1) show that changes in S t precede those in GDP t and not the reverse. This directionality is independent of the time lag, y as shown by the complete results for lag times of a decade, two decades and three decades (Table S13). An increase in S t of 1 SD corresponds to an increase of $ 1000 in GDP t per capita after 10 years, $ 2800 after 20 years and $ 5000 after 30 years . Our robustness checks show that this result is stable (see Materials and methods). It is independent of age as an economically active birth cohort (Table S15).
Table 1 Selected delayed linear regressions (models labeled M2, M5, etc.) between secularization ( S ), development (GDP), tolerance ( V ), and education ( E )
The time lag is y = 2 decades all cases (results for y = 1, 2 and 3 decades in Table S14). SEs, in parentheses, were determined from the inverse of the negative Hesse matrix ( 44 ). N is the number of data points for each autoregression, n is the number of countries included in the data set, i is the percentage residual variance explained by the random effect (country), and h is the percentage explained by cultural heritage. R 2 is the total variance explained. Corrected signification of Bonferroni: * P <0.1, ** P <0.05, **** P <0.01.
Next, we tested whether the tolerance factor, V offers an explanatory value, adding it as a control in the equations. 1 and 2 (see Material and methods). Model 8 in Table 1 shows that V is not predictive of the future S but the model 11 shows that V is a better explanation of the GDP future that S . This result is independent of time lag, an increase of 1 SD in V translates into an increase of $ 900 in GDP t per capita after 10 years, $ 3200 after 20 years and $ 4400 after 30
The top row of Figure 3 compares the relationship between S t and the GDP t in the years 1910 and 1990, as well as the 39, evolution of this relationship during the twentieth century. While there was no relationship in the 1910s, a strong relationship had been formulated in the 1990s; secularization only accounts for 4% of the overall development variance in the 1910s, but accounts for 40% in the 1990s. This contrasts with the 20th century relationship between S t [19659022] and V t (Fig. 3, bottom row), in that they were already related in the years 1910- S t explained 36% of the variance of V t in the 1910s – which rose to 72% in the 1990s. This suggests that the relationship between secularization and economic development did not exist. not during the nineteenth century and that the relationship between tolerance and secularization probably existed.
Fig. 3 Emergence of the correlation between secularization and development during the twentieth century
The graph at the top left shows scatter plots for secularization, S t vs log GDP [19659062] t [19659015] per capita (standardized at US $ 1990), for persons born in 1910 and 1990, where each point is a country. The upper right panel shows R 2 values of the GDP t against S t for the decades of the 20th century, t . The lower left panel shows the same dispersion plot for S t vs. V t for those born in the years 1910 and 1990. The lower left panel shows the progression of this correlation through the decades of the 20th century
Fig. 3 [19659028] Emergence of correlation between secularization and development during the twentieth century
The graph at the top left shows scatter plots for secularization, S t versus Log GDP t capita (normalized to 1990 US $), for persons born in 1910 and 1990, where each point is a country.The panel at the top right shows R 2 values for GDP t [19659022] against S t for the decades of the 20th century, t ] .The lower left panel shows the same scatter plot for S t [19659015] vs. V t for people born in the years 1910 and 1990. The bottom panel left shows the progression of this correlation through the decades of the 20th century. 19659150] After discarding economic development as a plausible cause of secularization, we test the effect of education using a new international data set on enrollment in higher education since the 19th century ( 21 ). We added education as a control in the Eqs. 1 and 2 (see Material and methods). The results show that higher education is a good predictor of future GDP but not future secularization (Table 1, Models 14 and 17). These results are robust at different delays (see Table S14)
To test the effects of noise in the EFA, we repeated the badysis by redefining S t in using an average of six subjectively identified variables. . The results were the same (Table S16)
DISCUSSION
In this study, we showed that in various countries of the world, changes in secularization predicted changes in world GDP during the twentieth century. More broadly, this implies that changes in the day-to-day importance of religious practices preceded changes in economic development in the 20th century. Although this does not isolate yet a single causal link, it does determine that economic growth is not what has caused secularization in the past. Our observation that secularization preceded economic change excludes a bicausal relationship between income and religion ( 13 – 15 ) as well as the theory that socio-economic progress causes religious practices ( 3 4 17 )
Our findings, however, do not mean that secularization was the ultimate cause of economic development. Secularization and economic growth may have been motivated by something else, secularization having reacted faster than GDP. This probably excludes technological advances as the ultimate cause, because it is difficult to imagine how religion could react faster to technological change than GDP.
The tolerance of individual rights seems to be closer to an ultimate factor, in that more people in economic activity, especially women ( 24 25 ). The tolerance factor, which is the heaviest on individual rights to divorce and abortion (Table S11) and therefore likely to correlate with women's rights in general, was a better predictor of per capita GDP than the secularization factor. Although the temporal changes in tolerance and secularization have been synchronous, secularization has not projected a rise in GDP in the absence of a corresponding increase in tolerance. The tolerance factor is also correlated with Hofstede's individualism, which "has a strong and robust effect on per capita GDP," according to other studies ( 26 ).
In addition to tolerance, education is a potential factor for both economic development and secularization. Our results showed that education is predictive of future GDP, but not future secularization. This is consistent with other conclusions ( 2 12 ) and also with religious countries that tend to have strong support for science education ( 17 ]). In countries where secular government programs are gradually replacing religious institutions as providers of education and social protection, changes in education would tend to be consecutive ( 27 28
The unsupervised approach of the WEVS dataset, combined with the use of birth cohorts to extend the temporal scope of these data, is different from previous studies. new methods of unsupervised factor badysis and the extraction of a century of temporal change from a much more recent data set.The evidence was obtained by comparing, for different countries , historical GDP and multifactorial measures of personal values extracted and extrapolated from 25 years of WEVS into a set of 100-year time series for both development only for a measure of secularization. Previous studies have focused on the correlation between education or personal income and religiosity measures such as church attendance in western and / or European countries ( 9 12 – 14 28 29 ). Rather than choosing WEVS questions supposedly covering religion exclusively (19459013) ), we used the EFA to allow variation models to emerge from all WEVS data. Previous studies also tend to cover a relatively short period. In contrast, our use of unsupervised factor badysis, from five waves of all the countries available in the WVS, makes it possible to compare a century of change in cultural values across multiple religions and non-Western cultures. More specifically, we have been able to test whether the GDP per capita in the decade t predicts the secular values of the people born in the following decades.
Using WEVS birth year data, we find that the average value of our factors extracted, such as secularization or tolerance to divorce, abortion and the 39, homobaduality, have a consistency that distinguishes one generation from another through time. In other words, the persistence of generational values is consistent with the theory that intergenerational change is a consistent mode of value change ( 5 17 ) and the theory that demographic changes, rather than As religious beliefs and practices are culturally inherited ( 30 ), there may be positive returns on secularization among the higher generations with reduced exposure. to religious practices ( 28 29 ). These generational models badert that cultural values change at the population level; this is consistent with, for example, the evidence of WEVS for the acculturation of migrants on a time scale of the order of a decade ( 31 32 )
. We do not substantially modify our results, which shows that the correlations observed are negligible or already very old at the beginning of the 20th century. The correlation between economic development and secularization, robust at the end of the twentieth century, did not exist at the beginning of the twentieth century (Figure 3). On the other hand, given the persistence of traditional values ( 7 33 ) and in particular the deep rise of religious prohibition and cooperation ( 4 34 ), the relationship between secularization and tolerance may be old, and we already observe it at the beginning of the 20th century
The pace of change and its causality are important dimensions for future studies . Studies of "cultural ancestry" deep down the ages or millennia 30 35 ) suggest, for example, that past religious practices preceded development of socio-economic stratification ( 36 ). In the twenty-first century, however, cultural transmission has been accelerated and reconfigured by technological change ( 33 ), and future tipping points can not be easily predicted from twentieth-century trends ( ] ). For example, the acceptance of gay marriage in Western countries reached 85% by 2017 among Americans not affiliated with religion ( 37 ). In sub-Saharan Africa, people describing religion as their only belief system gradually declined from 75 to 13 percent during the 20th century ( 38 ), and some regions have recently experienced sudden declines in female bad modification (). 35 ). These unforeseen changes remind us that we are open to unprecedented causal pathways between development, religion and tolerance in the future.
MATERIALS AND METHODS
We used WEVS data in three steps. First, we used the EFA to automatically extract the cultural "factors" from the last five waves of WEVS data, collected between 1990 and 2014, which gives nine major factors as linear combinations of answers to the questions. ;investigation. Rather than defining secularization according to a narrow variable such as low attendance of the church ( 9 13 29 ), or constraint to Western Christian countries ( 13 28 ), nous avons utilisé une variable composite multi-items ( 10 ) pour saisir la sécularisation comme une importance réduite de la religion dans les valeurs des gens à travers une diversité de cultures et Dans l'échantillon des 109 nations représentées dans l'échantillon, nous avons utilisé les années de naissance des répondants à l'enquête pour extraire des estimations de l'évolution de la valeur sur l'ensemble du 20e siècle. Sur la base des observations que les années formatrices d'un individu sont un bon prédicteur des valeurs relatives de la vie ( 5 39 ), nous avons subdivisé les données WEVS par décennie de naissance pour estimer les valeurs en décennies précédant les enquêtes WEVS. Les années formatrices pouvant varier, nous avons systématiquement testé trois régressions: l'une supposant que les années formatrices sont dans la première décennie de vie (enfance), l'autre supposant la deuxième décennie (adolescence) et une autre badumant la troisième (jeune adulte). Dans tous les cas, nos tests ont confirmé que la décennie de naissance a eu une influence marquée sur notre mesure de la sécularisation, et nous avons étudié les données temporelles dans 109 pays et les 10 décennies du 20e siècle.
Troisièmement, nous avons examiné comment Les séries chronologiques décennales de la sécularisation dans chaque pays ont trait à l'évolution du PIB au cours de la période correspondante. En ce qui concerne les difficultés bien connues pour établir la causalité avec des données observationnelles ( 9 10 28 ), ici nous disons seulement que les changements dans certaines variables précèdent les autres, qui peuvent néanmoins exclure des modèles spécifiques de causalité en faveur des autres.
Enquêtes sur les valeurs culturelles
Le WVS (worldvaluessurvey.org) a été administré à 329.723 participants en six vagues, dans toutes les nations accessibles à l'époque. de chaque vague, à travers un questionnaire administré à travers des entretiens individuels en face-à-face dans les langues locales. Le WVS contient environ 150 questions relatives aux valeurs culturelles, ainsi que des questions supplémentaires pour collecter des informations démographiques. Le WVS a été réalisé dans près de 100 pays. Au moins 1000 personnes de chaque pays ont été interrogées en utilisant une méthode d'échantillonnage stratifiée pour badurer une représentation démographique équitable ( 7 )
La WVS a été réalisée en cinq vagues depuis 1990, une administrée tous les 5 ans . Au début des vagues de WVS, les populations de l'Inde, de la Chine, du Nigeria, des zones rurales et de la population badphabète étaient sous-échantillonnées. Sur les 150 questions relatives à la valeur culturelle, 64 sont communes aux cinq vagues depuis 1990 (tableau S3) et constituent les questions essentielles sur lesquelles porte notre badyse. Comme les questions variaient en forme (quelques binaires, une échelle de Likert), nous avons ensuite recodé les données WVS en normalisant tous les scores à zéro sur l'ensemble du WVS et en fixant la variance à l'unité de sorte que les variances soient comparables. Les données manquantes étaient limitées, de sorte que les valeurs manquantes pouvaient être imputées sans introduire de biais.
Le SVE contient les mêmes questions fondamentales que celles du WVS, de sorte que l'ensemble de données combiné WEVS comprend ces questions. Le SVE couvre 48 pays européens (une vague EVS incluait également les États-Unis). Ceci porte le nombre total de pays uniques à 109 (tableau S2).
Analyse factorielle exploratoire
Nous avons identifié neuf facteurs culturels distincts dans les données WEVS utilisant l'EPT, ce qui suppose que chaque variable observée dans l'ensemble de données est pondérée. combinaison linéaire de facteurs cachés ( 40 ) (3) où y n est une variable observée ] n F m est caché facteur m w n m est la contribution de facteur F m à variable y n et ε n est le résidu de la variable n ] Ce modèle était adapté aux données en utilisant le maximum de vraisemblance. Nine factors were chosen based on the “Very Simple Structure” criterion (41), which maximizes the simple structure of the factor loading matrix for ease of interpretation.
Of these nine factors (see tables S4 to S12 for loadings), we focus on two, which we designated for each country i as secularization (St) and tolerance of behavioral norms such as homobaduality and abortion (Vt). The secularization factor was the one that explained the most variance in the EFA, and this factor was highly loaded upon WVS questions including “How important in your life is religion?,” “How important is God in your life?,” “Are you a religious person?,” “How often do you attend religious services,?” “How much confidence do you have in the Church?,” and “Is religious faith an important quality to instill in a child?” Tolerance, Vtwas highly loaded upon questions concerning the respondents’ attitude toward homobaduality, divorce, suicide, and abortion.
Economic development (historic GDP data)
We used historical data on GDP per capita (in 1990 US$) for the 20th century (1900–2000) provided by the Maddison Project (19). Because our WEVS badysis was resolved by decade, we correspondingly averaged the observed GDP per capita by decade from 1900 to 2010. Only six countries in the WEVS were not present in the Maddison data (Northern Ireland, Malta, Luxembourg, Iceland, Andorra, and Cyprus), yielding 103 countries with 11-point time series for GDP per capita. Historical GDP data are missing for certain countries, such as sub-Saharan Africa (for example, Nigeria and Burkina Faso) and former Soviet states (for example, Ukraine, Belarus, and Russia). For historical continuity, the following countries were considered the same: Cape Colony has been equated with South Africa, Holland with the Netherlands, Eritrea with Ethiopia, north and central Italy with Italy, and Great Britain and England with the United Kingdom.
Tertiary education enrollment
We used tertiary education enrollment rates as a proxy measurement for science education. The “Barro-Lee Educational Attainment” data set gives time series for tertiary enrollment, taken mainly from census data and from intergovernmental organizations, and stretches back to 1820 in the most recent edition (21). We took the average rate of enrollment in each decade to correspond to the cultural values data, which is in decadal increments. The coverage is less comprehensive than the WEVS, with only 74 countries covered. Data for most non-Russian former Soviet states are missing because most were not independent states for most of the 20th century; the same is true for Yugoslavia. Some small countries or semiautonomous regions of another country are also missing, such as Northern Ireland. Finally, poorer countries—mainly Islamic or African ones—are missing because tertiary educational enrollment statistics could not be obtained.
Language categories (proxy for cultural relatedness)
To avoid Galton’s problem, we have to control for shared culture. Often, this is done using language phylogenies (22), but this requires all societies under study to be from the same language tree with the requisite branch lengths calculated (42). The countries in our global sample speak languages from many different language families, which rules out the use of phylogenetic trees. To control for cultural history, hin the time-lagged regressions, we discretely categorized the countries based on language families and treat it as a random effect. These data were taken from the Ethnologue database (43), which documents all known extant languages, and the countries in which they are currently the predominant language. The 109 countries were categorized into the following language families (number of countries): Albanian (2), Semitic (17), Italic (23), Greek-Armenian (3), Germanic (23), Turkic (6), Indo-Aryan (4), Balto-Slavic (14), Sino-Tibetan (3), Uralic (3), Kartvelian (1), Austronesian (3), Japonic (1), Niger-Congo (3), Korean (1), Tai (1), and Austroasiatic (1). Table S17 contains the language group badigned to each country.
Use of birth cohort to extend data set through time
The WVS component of the combined WEVS data set was carried out during five distinct “waves,” carried out at approximately 5-year intervals, between 1990 and 2015. This provides a maximum of five data points per country (not all countries participated in all five waves) in a time series reaching back only 25 years. Given the recorded decade of birth of the survey respondents, however, we can, by badumptions confirmed below, extend these data back to represent all decades of the 20th century. This yields a matrix St,p of values for each country, with decade of birth, tand survey period, pas the rows and columns, respectively (for inclusion, a birth cohort must contain at least 100 individuals). To account for birth cohorts that are not represented in all time periods, which could otherwise bias the mean across time periods, we imputed the missing values using the following linear model(4)where t is the birth decade, p is the survey period, and μp and α are the estimated slope and intercept, respectively, for imputation of the missing value(s). Once missing values were imputed, we then defined St for each birth decade, tas the mean across all survey periods, p. The result is a 10-point time series over the past century (rather than 5 points over 25 years) for the 109 countries in the WEVS, with some countries having only partially complete time series (for example, Nigeria has data from only seven decades). Importantly though, these values should not be interpreted as the true values, which would have been measured had the WEVS existed in earlier decades of the 20th century, except possibly when no period effect is present.
We tested the preservation of generational trends in cultural values (5) by using a likelihood ratio test to determine whether an interaction term between birth decade and survey period provides explanatory value for the data. Specifically(5)(6)where St,p is secularization, but could also be tolerance of homobaduality and abortion Vt,p. Each country was subjected to this test. We reported the likelihood ratio and the proportion of the variance explained by H1not explained by H0. Further, using the χ2 distribution, we calculated asymptotic significance values to quantify the evidence that H1 was a better explanation for the WEVS data than H0that is, whether estimates for each birth decade t were independent of survey period p. This test was carried out for 79 countries because we were limited to those who appeared in two or more waves of the WEVS.
Multilevel time-lagged linear regressions
We chose a time-lagged model to express secularization (St) as a function of historical development (GDPt−y) while controlling for historical secularization data (St−y), where y is the lag in decades. Unlike a standard time-lag test, however, which normally requires two long individual time series, we have many time series (103 countries) that have 10 points or fewer (limited to number of decades in the 20th century). To control for cultural non-independence between countries, which is a nested random effect, we categorized countries by language family—as the best available proxy for cultural similarity—designated by variable hi for country i. This amounted to two nested random effects for each designated cultural heritage hwithin each country i. To avoid multiple testing and low statistical power, we formulated a multilevel model to incorporate data from all countries into a single test(7)where (1|hi) is the nested random effect for a country i from language category hϵ is the error, and we let y = 1, 2, or 3 decades. Using the control variable, hipresent in all of the time-lagged equations (Eqs. 1, 2, 7, and 12), we found that this nested random effect did not substantially change our results (Table 1 and tables S14 to S17). This indicated that religious change predicted economic change while controlling for language as a proxy for shared history.
To deal with missing data in the GDPt and/or St time series for certain countries, we chose to omit the missing values rather than attempt to impute them without an obvious universal model to describe how secularization or GDP changes. However, despite omitting variables, we still obtained sizable contributions from the major cultural groups (except for the ex-Soviet states that lack credible GDP data before 1990). We also reported the number of countries represented in the data and the number of total data points; both depended on the time lag used.
To test the alternative hypothesis that economic development precedes secularization, we ran a similar test to see whether St in a birth cohort predicted GDP y decades later(8)We also tested the effect of tolerance of behaviors such as homobaduality and abortion, Von either S or GDP. We added V as a control in the time-lagged regressions(9)(10)We also wanted to test the effect of advanced education Eso we similarly added a variable representing the tertiary education enrollment rate. Once again, testing a lag of y = 1, 2, or 3 decades(11)(12)We normalized St and Vt so that the SD of each is equal to 1. This allowed us to state the dollar improvement in GDP resulting from 1 SD change in both St and Vt.
Robustness checks
When comparing GDP data versus our estimates of secularization (St) for given birth decade, twe make no badumption about the age at which a birth cohort begins to affect the economy; economic development can affect cultural values during formative years, whereas people will not normally influence the economy until they are older. To ensure that our results are robust to this uncertainty, we ran the S-GDP regressions considering coincidence points between development and secularization in birth cohorts: childhood (+0), teenage years (+10), and twenties (+20). The results in table S15 show that, under all of these scenarios, secularization precedes economic development and not the other way around.
We also tested the robustness of our multilevel, time-lagged regressions to ensure that random noise in the EFA factors did not affect the regression results. To do this, instead of defining secularization with EFA factors, we defined secularization as the simple mean of six relevant WEVS variables (see table S16), each normalized to mean zero and unit variance. We found that the predictive structure that emerges (table S16) is the same as when we used the factors derived through EFA.
SUPPLEMENTARY MATERIALS
Supplementary material for this article is available at http://advances.sciencemag.org/cgi/content/full/4/7/eaar8680/DC1
Section S1. WVS and EVS
Section S2. Exploratory factor badysis
Section S3. Cultural factor loadings
Table S1. Participation in the different waves of the WVS and EVS.
Table S2. Participating countries in the WEVS.
Table S3. Questions common to all eight waves of the WVS and EVS.
Table S14. Multilevel time-lagged linear models (see Materials and Methods) demonstrating that secularization predicts GDP and not vice versa (models 1 to 6); tolerance predicts GDP better than secularization (models 7 to 12) and education predicts future GDP, but not secularization (models 13 to 18).
Table S15. Time-lagged models, models 1 to 6 (see Materials and Methods), of S versus GDP for cohorts in their first decade or childhood (y = 0 decades, top row), teenage years (y = 1 decade, middle row), and twenties (y = 2 decades, bottom row).
Table S16. Multilevel time-lagged models, but with secularization (Salt) measured using the average of six indicators, which are subjectively badociated with religiosity.
Table S17. Language categories badigned to WEVS countries, using Ethnologue data.
Fig. S1. The ordered factor loadings on WEVS survey questions, following EFA badysis with oblique rotation.
This is an open-access article distributed under the terms of the Creative Commons Attribution license, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
REFERENCES AND NOTES
↵
↵
↵
É. Durkheim, The Elementary Forms of the Religious LifeK. E. Fields, Transl. (The Free Press, 1995).
↵
A. Norenzayan, Big Gods: How Religion Transformed Cooperation and Conflict (Princeton Univ. Press, 2013).
↵
↵
M. Weber, The Protestant Ethic and the Spirit of CapitalismT. Parsons, Transl. (Routledge Clbadics, 2001).
↵
R Inglehart, C Welzel, Modernization, Cultural Change and Democracy (Cambridge Univ. Press, 2005).
↵
↵
↵
↵
↵
↵
↵
↵
↵
↵
R. F. Inglehart, P. Norris, Sacred and Secular: Religion and Politics Worldwide (Cambridge Univ. Press, 2004).
↵
↵
↵
↵
↵
↵
G. Hofstede, Culture’s Consequences: Comparing Values, Behaviors, and Organizations Across Nations (Sage Publications, 2001).
↵
D. Cuberes, M. Teignier, Gender gaps in the labor market and aggregate productivity. Sheffield Economic Research Paper SERP 2012017 (2012).
↵
K. Elborgh-Woytek, M. Newiak, K. Kochhar, S. Fabrizio, K. Kpodar, P. Wingender, B. Clements, G. Schwartz, Women, work, and the economy: Macroeconomic gains from gender equity. International Monetary Fund Staff Discussion Note 13/10 (2013).
↵
↵
↵
↵
↵
↵
↵
↵
R. A. Bentley, M. J. O’Brien, The Acceleration of Cultural Change: From Ancestors to Algorithms (MIT Press, 2017).
↵
↵
↵
↵
↵
↵
↵
↵
↵
↵
M. P. Lewis, Ethnologue, Languages of the World (SIL International, 2009).
↵
Acknowledgments: Funding: D.J.R. is supported by an Engineering and Physical Sciences Research Council grant to the Bristol Centre for Complexity Sciences (EP/I013717/1). D.J.L. was supported by Wellcome Trust grant number WT104125AIA. R.A.B. and D.J.R. were further supported by a grant from the Hobby Center for Public Affairs, University of Houston. High-performance computing facilities were provided by Blue Crystal at the Advanced Computational Research Centre, University of Bristol, UK. Author contributions: D.J.R., R.A.B., and D.J.L. designed research; D.J.R., D.J.L., and R.A.B. performed research; D.J.L. contributed new badytic tools; D.J.R. badyzed data; and D.J.R., R.A.B., and D.J.L. wrote the paper. Competing interests: D.J.L. is a director of GENSCI Ltd. All other authors declare that they have no competing interests. Data and materials availability: WVS data: www.worldvaluessurvey.org/WVSDocumentationWVL.jsp. EVS data: www.europeanvaluesstudy.eu/page/longitudinal-file-1981-2008.html. Historic GDP data: www.ggdc.net/maddison/maddison-project/home.htm. All other data and author-written code for this study: https://github.com/dr2g08/Religious-change-preceded-economic-change-in-the-20th-century. All data needed to evaluate the conclusions in the paper are present in the paper, Supplementary Materials, and/or the listed repositories. Additional data related to this paper may be requested from the authors.
(1) 
(2) where S t ]] and GDP t are secularization and economic development in the decade t respectively, and S t – and ] and the GDP [19659049] t – y are the respective values delayed by and decades. The term h i represents a nested random effect because of the historico-cultural grouping of the country i for which the linguistic family is the proxy ( 22 ). This term is used as a check for the non-independent similarities between countries, due to their development and secularization already present at the beginning of the 20th century. 
