Religious Values , Beliefs and Economic Development

Religious Values, Beliefs and Economic Development Jeffry Jacob† Department of Business and Economics Bethel University and Thomas Osang∗‡ Department of Economics Southern Methodist University February 2010 † Department of Business and Economics, Bethel University, 3900 Bethel Dr., St Paul, MN 55112; Email: [email protected], Phone: 651 635 6715 ∗ Department of Economics, Southern Methodist University, Dallas, TX 75275; E-mail: [email protected], Phone: 214 768 4398 ‡ We would like to thank seminar participants at SMU, Bethel University, Texas Camp Econometrics and the ASREC meeting for their valuable comments and suggestions. Funding for this research project was provided through a Collin Powell Fellowship research grant. 1 ABSTRACT This paper investigates the consequences of religion for economic development. In particular, we examine whether religious attitudes, beliefs and participation contribute to differences in per capita income across countries. Using a large scale international survey on values and religious behavior, we estimate both cross-section and panel data models, controlling for the so called deep determinants of development: Institutions, geography and trade. Our results indicate that religion plays an important role in economic development, but mostly in a non-linear manner. Countries with moderate religious values and behavior tend to have higher income levels than countries on either end of the religious spectrum. JEL: O1, Z12, N1, H1, F1 Keywords: Development, Economics of Religion, Institutions, Openness, Geography 2 1. Introduction According to the first three waves of the World Values Survey (Inglehart, 2000), 83% of the people interviewed believe in God, 75% believe in heaven, 26% attend a place of worship at least weekly and 38% at least once a month. More than 65% of the people questioned find comfort and strength in religion. In contrast, only 7% of the world’s population are not affiliated with any religion or consider themselves as atheists (Barrett et al., 2001). Nevertheless, the distribution of religious beliefs, values, and practice is not uniform across countries. While the people in some countries such as Bangladesh, the Philippines, and Nigeria are predominantly religious both in expressed opinion and religious practice, other countries such as China, Russia, and Denmark display very low levels of religiosity (see Table 1(b) in the Appendix for a ranking of countries by our religiosity index). Thus, given the importance of religion in many people’s lives as well as the observed cross-country heterogeneity in religious beliefs and practice, it is worthwhile to examine whether religious behavior contributes to the existing differences in economic performance across countries. The importance of religion as a determinant of economic development was introduced in the mainstream economic literature almost a century ago. Max Weber (1905) recognizing the far reaching role that religion can play in social transformation, went on to claim that the Protestant Reformation, through its impact on the belief system, was central to the emergence of capitalism. There have been several studies that have challenged the validity of this claim. Tawney (1926) and Samuelsson (1993) argue that the main capitalist institutions which Weber (1905) ascribes to the transformative power of the Protestant Reformation, predated that movement. A weaker version of the 3 Weberian thesis was proposed by Eisenstadt (1968), according to whom it is not a particular theology or belief, but the “transformative potential” of religion that can result in shifts in values and behavior, which in turn can alter outcomes. Iannaccone (1998) provides an excellent survey of this literature on the economics of religion. The direct link between religion and macroeconomic development of a country, though generally acknowledged, has received little attention in terms of empirical research. An exception is the paper by Barro and McCleary (2003) who examine the growth impact of a number of religious variables in a cross-country panel setting. They find a significant positive relationship between belief in hell and economic growth, as well as a significant negative relationship between monthly attendance and growth. Instrumental Variable (IV) estimates are used to account for the potential endogeneity of belief in hell and attendance. Instead of examining the direct link between economic performance and religion, Guiso et al. (2003) examine a variant of the Weberian hypothesis by focusing on how religion shapes people’s attitudes. They find that Jews and Protestants have the greatest aversion to cheating on taxes, while Buddhists and Protestants are least likely to accept bribe. Private ownership is supported most strongly by Catholics, while Hindus and Muslims are the strongest opponents of competition. More recently, Durlauf, Kourtellos and Tan (2006) re-examine the results in Barro and McCleary (2003). In particular, they test the robustness of the earlier results by including a wider set of control variables. To evaluate robustness, they use Bayesian model averaging methods. They confirm as robust the Barro and McCleary finding that monthly church attendance has a significant and negative impact on economic growth, but, in contrast to the Barro and McCleary results, they find that belief in heaven or hell 4 is unrelated to economic growth. On a micro level, Gruber (2005) carries out a subnational study linking religious attendance and economic outcomes in the United States. He finds that higher attendance, instrumented by a geographic measure of religious density, is generally related to better economic outcomes such as higher incomes, higher educational attainment, less reliance on welfare and disability receipt. A common characteristic of the above studies is the use of a set of controls to account for other factors affecting economic performance such as measures of labor, physical and human capital. As Rodrik et al. (2004) point out labor, physical and human capital, while affecting economic development, are in turn determined by deeper and more fundamental factors which fall into three broad categories: Geography, institutions and trade (Acemoglu et al., 2001, Rodrik et al., 2004, Frankel and Romer, 1999, Sachs, 2003, among others). Easterly and Levine (2003) provide a good overview of how each of these three determinants has been treated in the literature with the aim of explaining the vast differences in growth and levels of income amongst countries. Regarding the relative importance of the three deep determinants, Rodrik et al. (2004) report that institutions matter most for economic development once the endogeneity of intuitions and trade has been properly accounted for, leaving a negligible role for geography and trade. Sachs (2003), on the other hand, finds that geography is the most important deep determinant of income and output, while Frankel and Romer (1999) underscore the importance of international trade. Re-examining this issue in a panel data context, Jacob and Osang (2008) find that all three determinants play a significant role in economic development, 5 but that the economic and statistical significance varies with the way in which we measure institutional, trade-based, and geographic determinants. This study contributes to the existing literature on religion and economic development in the following ways. First, our approach can be regarded as a synthesis of two different strands in the development literature: Economics of religion (Barro and McCleary, 2003) on the one hand, and deep determinants of development (Rodrik et al., 2004) on the other. We approach the issue of the impact of religion from the latter strand (Rodrik et. al., 2004). In particular, we extend the Rodrik et al. (2004) approach to study the impact of broad religious values and beliefs on economic development. Second, in contrast to Barro and McCleary (2003) and Durlauf, Kourtellos and Tan (2006), we not only control for the potential endogeneity of the religion variables but also for the endogeneity of important control variables such institutions and trade. Third, we account for the non-linear relationship between most religious variables and economic performance. This is in contrast to the existing literature which has focused almost exclusively on the linear case1. The main findings of the paper are as follows. First and foremost, all aspects of religion considered in the study – attitudes, beliefs and participation as well as an aggregate religion index – appear to matter for a country’s level of economic development, even after adequately controlling for measures of institutional quality, international linkages, and geography. Furthermore, we find that the relationship between religion and income is non-linear (in variables) in most cases. In particular, we find that income levels tend to be the highest for countries with moderate expressions of religiosity 1 An exception is Dincer (2008) who uses a quadratic term in studying the impact of religious and ethnic diversity on corruption. 6 and lowest for those at both ends of the religious spectrum. In terms of robustness, we find that within each of the three aspects of religion some variables are significant and others are not. Finally, cross-section and panel data estimates produce quantitatively and qualitatively similar results. The rest of the paper is organized as follows: Section 2 of the paper describes our empirical methodology, while Section 3 contains a discussion of the dataset. Results are presented in Section 4. Section 5 concludes. 2. Empirical Methodology Our main objective is to study the relationship between religion and economic development, controlling for the commonly accepted deep determinants of development institutions, trade and geography. We classify religious variables into three categories: Religious attitudes, religious beliefs and religious participation. Attitudes capture people’s perception of religion as a factor in their lives, as exemplified by the question whether they derive comfort and strength from religion. Beliefs measures people’s faith in core religious concepts such as God, Heaven, Hell and Sin. Finally, participation is a measure of the frequency with which people attend religious ceremonies. We use per capita income as our measure of economic development. While other measures of development such as the United Nations’ Human Development Index, inequality measures such as the Gini index, or economic growth have also been used in the literature (Abadie, 2006; Barro and McCleary, 2003), per capita income is by far the most frequently used measure of economic performance and thus makes the results from this study readily comparable to the existing literature exploring the determinants of 7 economic development. In addition to the religion variables, we use one measure of institutional quality (Contract Intensive Money), two measures of global integration (Trade Share and Import Tariff) and one measure of geography (Malaria Ecology). Examining the bivariate scatter plots between per capita income and various religious variables (see Figs. 1(a)-(d)), which demonstrate a non-linear relationship, we use a linear as well as a quadratic term for the religious variables. Thus, our main empirical specification is: InciT = α + β1* Inst i + β 2 * Tradei + β3 * Geog i + γ 1* Religion i + γ 2 * Religion i 2 + ε i (1) where InciT is income per capita in the year 2000, and Insti , Tradei , and Religioni are the time-averaged measures of institutions, trade, and religion, respectively. Geog is a time-invariant measure of geography and ε i is the error term assumed to be normally distributed. Estimation of (1) poses a number of difficulties that need to be addressed. First, institutions, trade and some religion measures are likely to be endogenous due to measurement error, survey bias, and/or reverse causality.2 Consequently, appropriate instruments are needed for these measures.3 Of the various instruments found in the literature for institutions and trade, two stand out due to their desirable properties and widespread use: Settler Mortality as an instrument for institutions (see Acemoglu et al., 2001)4 and Predicted Trade Shares as an instrument for a country’s degree of integration 2 For instance, see Frankel and Romer (1999) We carry out a Hausman (1978) specification test to check the endogeneity of the institutional, trade and religion measures. If found endogenous, we use appropriate instrumental variable techniques to obtain consistent estimates. 4 Acemoglu et al. (2001) argue that settler mortality is a truly exogenous instrument for institutions since it is not correlated with current income. However it determined the colonization strategies, which shaped past institutions. Current institutions were in turn shaped by these past institutions. 3 8 (see Frankel and Romer, 1999). Though these two instruments have been shown to perform well in a number of studies, their first stage diagnostics have not measured up well according to some studies (Dollar and Kraay, 2003). In addition, the data availability of both measures is severely limited in our sample, causing the cross-section sample size to drop to 18 observations or less. Consequently, we use three alternative approaches to construct instruments for the institutions and trade measures. The first approach builds on the idea suggested by several social scientists that a nation’s culture can have an important impact on economic outcomes (Inglehart and Baker, 2000; Landes, 1998). One channel through which culture and values affect economic performance is through institutions. Institutions are defined as the “humanly devised constraints that structure human interactions” (North, 1994). They are the rules of the game which govern how humans interact with each other. Naturally, the enforcement of rules is part of the effectiveness of institutions. The strength of contract enforceability can be gauged by the degree of confidence which citizens have in the establishment. We exploit this relationship by using measures of people’s confidence in the government as instruments for institutional quality. This idea of using underlying cultural values- in this case, confidence in the establishment- as instruments for institutions follows from Grief (1994). Grief (1994) develops a formal model to show the impact of culture on institutions and traces out how cultural differences between two pre-modern trading groups, one from the Muslim and the other from the Latin world, led to widely divergent institutional outcomes.5 5 In a related study Gwin and North (2004) show that a country’s dominant religion is an important determinant of the quality of its institutions. 9 We use a similar approach to find instruments for trade. International trade has been widely credited with increasing competition and improving efficiency. The most noted case of trade-led growth is the East Asian economic miracle. Several studies (for e.g. World Bank, 1993) ascribe the successful adoption of trade promoting policies in these emerging markets to the people’s culture of hard work and openness to exploit the market opportunities. In this spirit, we use data on attitude towards market as instruments for trade. Our second set of instruments is derived by applying the principal components analysis to obtain instrument indices from the set of instruments used in the previous approach6. Specifically, we create an index of the variables capturing the attitude towards the establishment and another one for the square of these variables. An instrument index measuring the attitude towards the market was created in a similar way. Based on the screeplot7, we picked principal components whose eigenvalues were greater than one. In each case, only the first principal component met this criterion. The KaiserMeyer-Olkin (KMO) measure of sampling adequacy, which measures whether a lowdimensional representation of the data is possible, was in the acceptable range for all cases.8 In discussing the appropriate treatment of institutions as exogenous or endogenous, and in the latter case, investigating the choice of the correct instrumental variables, Dollar and Kraay (2003) argue that instruments from within the model perform 6 See Tabellini (2005) and Filmer and Pritchett (2001) for applying principal components analysis in constructing economic indices. 7 A screeplot plots the variances against the number of the principal component. 8 The KMO measure (Kaiser, 1974) was 0.64 and 0.62 for the levels and squares of the variables capturing the attitudes towards the government, respectively. It was 0.52 and 0.51 for the corresponding variables representing attitudes towards the market. 10 better and may be more desirable than external instruments. Thus for our third set of instruments, we follow an approach suggested by Lewbel (1997) to find instruments within the model by using second and third order centered moments of the endogenous institution and trade variables. The religion variables may also be subject to an endogeneity bias. For example, religious attendance may be endogenous as noted in the context of the secularization hypothesis. We use three variables as instruments for the religious variables. The first is a Hirfendahl Index of religious concentration. The idea here is that if a religion is dominant within a country, religious attendance will be strengthened, while a more diffused field of religions may make religious attendance less likely. Gruber (2005) makes a similar argument using religious density as an instrument for religious participation in the United States.9 The other two instruments are a dummy variable for the presence of state religion and an index of government regulation of religion (obtained from Grim and Finke, 2006). For the instruments to be valid, they must meet two conditions: They need to be correlated with the endogenous variables, conditional on the other covariates, and uncorrelated with the dependent variable, conditional on the fitted variables and the exogenous controls10. Regarding the validity of the instruments in the first stage (reduced form regression) we report a battery of diagnostic tests. First, the Cragg-Donald (1993) chi-square statistic tests the null that the matrix of reduced form coefficients has rank=K1 where K=number of regressors, i.e., that the equation is underidentified. This is a test of 9 Note that Barro and McCleary (2003) use religious pluralism as an instrument for attendance. They argue that greater plurality of religions and therefore more competition among them encourages religious participation. Hence they use 1-Hirfendahl Index as instrument. 10 This is equivalent to saying that the instrument should be uncorrelated with the regression error. 11 instrumental relevance and a rejection of the null means that there is at least one available instrument for each endogenous variable (see Hall et al., 1996). Second, we report the Anderson-Rubin underidentification test of the instrumental variables. This test is similar to the Cragg Donald test, but is robust to the presence of weak instruments (see Dufour, 2003, for a discussion). Third, we report Shea’s partial R2 (Shea, 1997). This is a simple diagnostic statistic for determining the strengths of instruments when there are multiple endogenous variables. Weak instruments may cause an identification problem for the partial effects of the endogenous variables in the IV regression framework (Dollar and Kraay, 2003). The final first-stage test we report is the F-stat form of the Cragg Donald (CD) test statistic (see Stock and Yogo, 2002). This statistic tests for the existence of a bias in the IV estimates resulting from weak instruments. For example, the null of a 30% bias in the IV estimator is rejected if the CD F statistic is larger than 4.73 for the case of 2 endogenous variables and 4 instruments (see Stock and Yogo, 2002; Table 1).11 Regarding the second stage restrictions, we carry out Sargan’s test for properly excluding the set of instruments from the second stage regression.12 We also extend our analysis to a panel data setup. Using a panel data approach enlarges the sample size which enables us to include additional control variables such as regional dummies. The estimated panel data model is: Incit = α + β1* Inst it + β 2 * Tradeit + β3 * Geog i + γ 1* Religion it + γ 2 * Religion 2it + ηr + ϑt + ε it where a subscript t indicates a time-varying variable, ηr and ϑt denote region and time dummies, respectively, and ε it is an error term having a normal distribution. 11 Estimation is done in STATA using ivreg2 (Baum et al., 2003). Since Sargan’s test is valid only in the case of homoscedastic errors, we also report the Pagan and Hall (1983) test for the case of heteroscedasticity of the disturbance term. 12 12 Due to the restricted sample size in our panel specifications, a fixed effect (FE) analysis is not feasible. Thus, we perform pooled OLS and Random Effect (RE) estimations instead. The decision to choose between pooled OLS and RE model is based on the Breuch-Pagan (BP) test for error components. Once, the appropriate model is chosen, we conduct the Hausman (1978) specification test to determine the need for instrumental variable estimation. 3. Dataset Our data spans three decades, 1970s, 1980s, and 1990s. For panel data analysis, we average our time varying macroeconomic variables for each decade to get a maximum of three observations per country. For cross section analysis, we average data over the three decades. Our dependent variable for cross-section study is GDP per capita in 2000, expressed in 1996 international dollars. Information on religion and culture variables is available, however, only for certain discrete time points. Based on the survey time periods, we match these observations to the respective decades considered in our study. Our data on religious attitudes, values and beliefs comes from the first three waves of the World Values Survey (WVS; Inglehart, 2000). The WVS is a large scale international survey aimed at collecting national level data on a wide variety of cultural, religious and political variables. The survey contains socio-cultural information on 59 different countries, which together account for more than 80% of the world population. The three waves of the WVS we use are for the years 1981-84, 1990-93 and 1995-97. We match these time periods to the non-WVS data from the seventies, eighties and nineties respectively. 13 In each wave, an attempt is made to interview a nationally representative sample of at least a 1000 individuals from each country under study. Even though attempt is made to keep the sample large and free from any biases, some under or oversampling might occur. The WVS provides individual weights for each observation to correct for these sampling issues. We conduct our analyses on the weighted sample. Below, we provide a brief discussion of the variables used in our study. 3.1 Measures of Religion As mentioned above, our religion measures cover three different religious dimensions: attitude towards religion, belief in various aspects of religion and religious attendance. These categories are drawn from the WVS (Inglehart, 2000). 3.1 (a) Attitude towards Religion The first category of religious variables measures a country’s attitude towards religion. They are based on two questions: “How important is religion in your life?” (Variable 9) and, “Do you find that you get comfort and strength from religion?” (Variable 191). The first question is measured on a scale of 1-4, with 1 implying “Very Important” and 4 representing the response: “Not at all important”. We recode them so that higher number represents increased importance. The second has two responses: 1=Yes and 2=No. We recode “No” as 0. 3.1 (b) Belief in Religion The next category is an indicator of people’s beliefs in various dimensions of religion. The five questions we look at are, “Do you believe in God, Heaven, Soul, Devil and Sin?” (Variables 182, 188, 185, 186 and 189, respectively). Again, the responses are 14 1=Yes and 2=No. We dichotomize them with 1 representing a positive belief in the respective dimension. 3.1 (c) Religious Attendance Religious Attendance variables are derived from the response to the question: “Apart from weddings, funerals and christenings, about how often do you attend religious services these days?” The responses are: 1= More than once a week, 2= Once a week, 3= Once a month, 4= Only on special holy days, 5= Once a year, 6= Less often and, 7= Never, practically never. From this, we construct three indicator variables: “Attend at least weekly” if response if 1 or 2, “Attend at least monthly” if response is 3 or less and “Attend yearly” if response is 5 or less. 3.1 (d) Index of Religiosity Finally, we aggregate the above three categories of religious variables- attitudes, beliefs, attendance and the fraction of religious population (drawn from Barret et al, 2001) into an Index of Religiosity. This is done by normalizing each of the four religion variables to lie between zero to one and then adding them up to obtain the index. The index thus ranges from lowest possible score zero to a highest possible value of four. Table 1(b) provides a ranking of countries based on our religiosity index. 3.2 Other Explanatory Variables Our measure of institution is contract intensive money (CIM) which was proposed by Clague et al. (1999). It is defined as the ratio of non-currency money to total money. The basic argument for such a measure stems from the fact that in societies where the rules of the game and property and contract rights are well defined, even transactions which heavily rely on outside enforcement can be advantageous. Currency in this setting 15 is used only in small transactions. Agents are increasingly able to invest their money in financial intermediaries and exploit several economic gains. Clague et al. (1999) discuss the various gains from increased use of CIM and augment their use of CIM with case studies. They also show that CIM is a measure of contracting environment and not of financial development, as one might suspect. This measure is thus in line with the definition of institutions as defined above. Moreover, CIM is a more objective measure that is free from some of the biases and measurement errors that affect many survey based measures of institutions. The extent of openness of a country is measured by its share of trade in the GDP as well as import tariffs (obtained from the World Bank, 2003). For geography, we use Malaria Ecology (ME), a recently developed measure of disease environment proposed by Sachs (2003). Sachs has argued that the traditional malaria index (Gallup, Sachs and Melligner, 1998) used in the literature is not a good indicator of the disease environment. The new measure combines temperature, mosquito abundance and vector specificity and, as an ecology-based measure, is predictive of malaria risk. 3.3 Variables used as Instruments Our set of instruments is based on people’s attitude towards the establishment (as instruments for institutions) and markets (as instruments for trade). In particular, we use the following four questions from the WVS as instruments for institutions: “I am going to name a number of organizations. For each one, could you tell me how much confidence you have in them: is it a great deal of confidence, quite a lot of confidence, not very much confidence or none at all?” The organizations we look at are: Government (Variable 142), Parliament (Variable 144), Police (Variable 141) and Armed Forces (Variable 136). The 16 responses are originally coded from 1 to 4, with 1=Great deal of confidence and 4= None at all. However, we rescale them so that a higher number represents a greater degree of confidence. We use three questions from the WVS measuring people’s attitude towards markets as instruments for trade. Specifically, these questions ask the respondent to rank a given statement in the following way: “Now I'd like you to tell me your views on various issues. How would you place your views on this scale? 1 means you agree completely with the statement on the left; 10 means you agree completely with the statement on the right; and if your views fall somewhere in between, you can choose any number in between.” The three statements we look at are: (i) “Competition is good. It stimulates people to work hard and develop new ideas”; (ii) “In the long run, hard work usually brings a better life”; (iii) “Wealth can grow so there's enough for everyone” (Variables 128-130, respectively). We recode the answers to these questions such that higher scores reflect a more positive attitude towards the market. The World Christian Encyclopedia (Barret et al., 2001) is the source of two of the three instrumental variables for religion - the Hirfendahl index of religious concentration and the presence of state religion. The latter is a dummy variable taking the value of one if a state religion is present in two out of the three sample periods. We construct the Hirfendahl index of religious concentration from eight religious preference variables: Percentage of population that is Protestant, Roman Catholic, Orthodox Christian, Jewish, Muslim, Buddhist, Hindu and following a religious faith other than the one listed above 17 (Other Religion)13. The third instrument is the index of government regulation of religion constructed by Grim and Finke (2006). 4. Empirical Results 4.1 Cross-Section Analysis We first estimate the impact of religious attitudes on economic performance within a cross-section context (see Table 2), using measures of institutions, trade and geography as covariates. Given the potential simultaneity between the economic performance variable and measures of institution and trade, we compare and test the OLS estimate (Col. 1 and 6) with various sets of IV estimators (Cols. 2-5 & 7-10). In particular, we use three sets of instruments. The first set of instruments (IV(A)) uses attitudes towards hard work, competition and wealth accumulation and their square terms as instruments for trade; as well as confidence in government, parliament, army and police and their square terms as instruments for institutions. The second set (IV(B)) employs the first principal component of the levels and squares of the variables in set A as instruments for institutions and trade. The third set (IV(C)) uses the second and third order centered moments of each endogenous variable as instruments for that particular variable. We report two IV specifications. The first one includes instruments for 13 The full set of religious groups used by Barrett et al. (2001) is: Roman Catholics, Protestants, Orthodox, Anglicans, Marginal Christians, Independent Christians, Muslims, Hindus, Chinese Folk Religionists, Buddhists, Ethnoreligionists, New Religionists, Sikhs, Jews, Spiritists, Baha’is, Confucianists, Jains, Shintos, Taoists, Zoroastrians, and Other Religionists. We combine Anglicans with Protestants. 18 institutions and trade variables only (IV(A), IV(B) and IV(C)), while the second one includes instruments for the religion variables as well (IV-R(A), IV-R(B) and IV-R(C))14. The main finding from Table 2 is that the two variables measuring religious attitudes - importance of religion (Cols. 1-7) and religion as a source of comfort and strength (Cols. 8-14) - exhibit a statistically significant impact on per capita income. The positive sign of the level estimate and the negative sign for the quadratic term imply that income levels initially rise, then taper off, and eventually decline as religious attitudes gain strength.15 The Hausman specification test indicates that only IV(A) estimates are preferred to OLS at the 5% level (col. 2 and 9), but not IV(B) and IV(C). Furthermore, when comparing IV(A) to IV-R(A), we get mixed results. For the importance of religion variable, we cannot reject the IV(A) null (col. 2), while for the comfort in religion variable, the IV(A) specification is rejected in favor of IV-R(A). Thus, the two preferred specifications in Table 2 are those in columns 2 and 10. (Table 2 about here) Regarding the strength of the instruments used in the IV regressions, we find that, for the specification in col. 2, all tests indicate the validity of the instrument set with the exception of the relatively low value of the CF-F statistic of 0.8. For the specification in col. 10, the CD Underid. Test now rejects the instruments in the first stage, while the CD 14 The sample sizes reported in Tables 2 through 5, ranging from 22-34, are substantially smaller than the 95-97 wave of the World Value Survey (59 countries). The decline in the sample size is due to data limitations with regard to the CIM variable as well as certain instrumental variables (IV(A) and IV(B)). Replacing the CIM measure with a more widely available measure of institutional quality (Rule of Law , see Kauffman et al, 2003) increases the sample size only for the OLS and IV(C) specifications and yields quantitatively and qualitatively similar results to those reported in Table 2. These results are available from the authors upon request. 15 Note that linear and quadratic terms are also jointly significant in all specifications, as reported in the table. 19 – F statistic is even lower (0.08). All other tests, however, underscore the relevance of the instrument set including Sargan’s overidentification test. Note that the Pagan-Hall test indicates that – across all specifications – the null of homoscedastic error terms cannot be rejected. With one exception (import tariffs in Col. 13), all covariate estimates have the expected signs, that is positive for CIM and negative for import tariffs and malaria ecology. Furthermore, the majority of CIM estimates are statistically significant at the 10% level. However, only four import tariffs and malaria ecology estimates are significant. The imprecision in the estimation of these two deep determinants is most likely the result of the small sample size in most specifications in Table 2(a) and not a reflection of the dominance of institutional measures over trade and geography variables. In all subsequent tables we use CIM and Import Tariff as our primary measures of institutions and trade, respectively. We always present the OLS results as a benchmark, but report only those IV estimates which are preferred to the OLS estimates according to the Hausman test. For each IV estimator, we report the same first and second stage diagnostic tests as in Table 2. (Table 3 about here) Table 3 examines the role of religious beliefs in economic development. In particular, we use five different forms of religious beliefs: In God (Cols. 1-5), Heaven (Cols.6-11), Soul (Cols. 12-17), Devil (Col. 18. 19) and Sin (Col. 20, 21). The strongest results from Table 3 are with respect to Belief in God (Cols. 1-5) which is statistically significant in both linear and second order term in all cases. Using alternate trade measures does not affect these results. Belief in Heaven (Cols 6-11) also shows a non- 20 linear pattern but is statistically significant in fewer cases. When combined with Import Tariff (Cols 6 & 7), only the second order term is significant in the OLS regression, while both terms are statistically significant in the IV (A) estimator (Col 7). Replacing Import Tariff with Trade Share (Cols. 8-11), we find that OLS is rejected against all IV estimates. Interestingly, neither OLS nor IV estimates yield significant results, except the second order term in IV(A) (Col. 9). For Belief in Soul (Cols. 12-17), only IV(A) and IV(C) yield statistically significant results for both, linear and the second order term while only the latter is significant in Cols. 15 and 16. The remaining religious belief variables - Belief in Devil and Belief in Sin- are never statistically significant. While the last two religious belief variables can be thought of as representing the deterring aspect of religion, the first three –Belief in God, Heaven and Soul- reflect, at least to some degree, the redeeming side of religion which may explain their stronger explanatory power. Except for Shea’s Partial R2 for IV(A) (Col. 2) and the CD F-statistic for all IV estimators, the first and second stage diagnostic statistics confirm the relevance of the chosen instruments. Regarding the other covariates, CIM has a positive and statistically significant impact on economic development in 15 out of 21 cases. With one exception, the Import Tariff estimates are negative and statistically significant, while Trade Share is statistically significant (and has the right sign) in four out of eleven cases (Cols. 9, 10, 15 & 16). As expected, Malaria Ecology exerts a negative impact on development but is significant in only half of all cases. (Table 4 about here) 21 In Table 4, we look at the third category of religious variables: Religious Participation or Attendance. Attendance has been linked to the extent of religiosity in several previous studies. One concern with using attendance as a determinant of economic development has become known as the secularization hypothesis: In the early stages of development, places of religious worship tend to serve as important venues for networking and social capital formation. As a result, greater attendance can have a positive impact on income levels. However, as a country develops, formal institutions start maturing causing the demand for places of worship as facilitators of social capital to decline. As a result, attendance will decline. This potential feedback from income to attendance may bias the attendance estimates. In our analysis, we control for this potential endogeneity by using the instruments for religion discussed earlier. We consider three levels of religious participation: Attend religious worship at least once a week, at least once a month, and at least once a year. Weekly attendance exerts a linear negative impact on development (Cols. 1 and 3). Since estimates for monthly attendance are statistically insignificant across most specifications, we only report the benchmark OLS results (Col. 3). For both these attendance variables, the linear terms are negative and the quadratic terms are positive. In contrast, Yearly attendance demonstrates the nonlinear relationship found in the previous tables – a positive linear and a negative second order term (Cols. 5-9). All second order terms are statistically significant at the 10% level. The linear terms are significant too, but most of them at a lower level (e.g. 15% level). While CIM is statistically significant in only one case (Col. 6), its point estimates are similar to the previous results (see Tables 2 and 3). Import Tariff and Trade Share 22 have the expected signs (except for Cols. 2 and 10), but only Import Tariff are statistically significant. Malaria Ecology has the expected negative sign in all cases and is statistically significant at the 10% level in all but one. In terms of the validity of the instruments, the AR joint significance test, which is robust to the presence of weak instruments, rejects the null of underidentification of the first stage equation in Cols. 6, 8 and 9. The Shea’s partial R2 values are also reasonably high in most cases, indicating that collinearity between instruments is not a problem. Except for Col. 6, Sargan’s overidentification test indicates that the instruments are correctly excluded from the second stage regressions. Next we combine the variables measuring religious attitudes, beliefs, participation and population size into an index of religiosity (Table 5).16 The results confirm our previous findings. The religiosity index has a non linear impact on economic development. Per capita income levels increase with the index of religiosity, taper off and then eventually experience a decline as the index increases. Furthermore, the linear and square terms of the index variable are significant in four out of six specifications. The magnitude of the impact remains robust to alternative choice of trade policy variables. As before, CIM, Malaria Ecology and Import Tariff are mostly statistically significant and display the expected signs. 4.2 Panel Data Results As discussed in section 2, we also investigate the impact of religious attitudes, beliefs, attendance in the context of a panel data model. As in the previous section, we report IV estimates where appropriate.. 16 For a ranking of countries by the religiosity index, see Table 1(c). 23 In Table 6 we study the impact of religious attitudes (Comfort in Religion, Cols. 1-4) and religious beliefs (Belief in God, Cols. 5-7). In Cols. 1 and 2, we report the pooled OLS and RE results, respectively. Comfort in Religion has a statistically significant and non-linear impact on income in the pooled OLS case. The BP test, however, indicates the presence of error components and thus the appropriateness of the RE model. In that specification, the linear term of the Comfort in Religion variable is no longer significant. In Cols. 3 and 4, we add region and time dummies to the OLS and RE model, respectively. Compared to the specifications without dummies, the parameter estimates for the Comfort in Religion variable are no longer significant, but sign and magnitude of the two terms are similar. The two time dummies and most of the region dummies are statistically significant and have the expected signs. For the Belief in God variable, we report the benchmark pooled OLS case (Col. 5), pooled OLS with region and time dummies (Col. 6) as well as pooled IV with region and time dummies (Col. 7). In all three cases Belief in God has a statistically significant and non-linear impact. The Hausman specification test indicates again the appropriateness of the IV approach (Col. 7) over the corresponding non-IV model (Col. 6). Regarding instrument quality, all tests indicate the validity of the instruments in the reduced form regression as well as their proper exclusion from the second stage regression. As in the cross-section models before, most control variables have the expected sign and are generally statistically significant. The exceptions are the Trade Share estimates, two of which not only have the wrong sign but are statistically significant at the 5% level (Cols 6 & 7). 24 Table 7 investigates the impact of religious attendance. The results for Weekly Attendance are not statistically significant in the panel framework when both linear and non-linear terms are included. Thus, we restrict our attention to just the linear form (Cols 1-4). Weekly Attendance exhibits a negative sign and is highly significant in all four specifications. This result points to a clear trade-off between intense religious participation and economic performance: The higher the percentage of people attending religious ceremonies at least once a week, the lower a country’s per capita income. In contrast, the Yearly Attendance variable demonstrates the familiar non-linear pattern of positive linear and negative quadratic term and is statistically significant near the 10% level in most cases (Cols. 7-9). These results suggest that countries with moderate religious participation rates (i.e. a moderate number of people attending religious ceremonies at least once a year) have higher income levels than countries with both, extremely low (i.e., most people never attend religious ceremonies) or extremely high participation rates (i.e. most people participate in religious ceremonies at least once a year). All control variables perform fairly well in Table 7. CIM has the correct sign and is statistically significant at the 5% level in two third of the cases. The Trade Share variable is positive and statistically significant around the 10% level in three out of four cases. The Import Tariff covariate is also statistically significant in three specifications but has the expected sign in all five cases. Malaria Ecology always has the expected sign and is statistically significant at the 5% level. As in Table 6, the diagnostic tests confirm the appropriateness of the instruments in the IV(C) specification. 25 5. Summary and Conclusions We find that religion matters for the economic performance of countries, even after controlling for the influence of such important determinants as public institutions, international linkages, and geography. In contrast to previous findings in the literature, however, our evidence points to a predominantly non-linear relationship between measures of religiosity and per capita income. Countries with moderate values for their religious indicators tend to enjoy higher levels of income than those with extremely high or low values. There is a plausible explanation for this non-linearity. When comparing low- to mid-level religious countries, the latter may benefit from incentives and behavioral modifications triggered or provided by religious beliefs and practice. In the context of countries with Christian faith, this could be called the “Protestant Ethics” effect. When comparing mid- to high-level religious countries, the latter may experience income losses due to reduced labor productivity as a result of the extensive involvement of their citizens in religious practice or due to barriers to scientific research, gender equality, and educational attainments, among others, justified by restrictive religious beliefs and attitudes. In addition to our main ‘non-linearity’ finding, several more specific results emerge concerning the impact of religion on economic performance. The two religious attitude variables that we investigate - importance of religion and religion as a source of comfort and strength – both have a strongly significant (non linear) effect on per capita income levels. Among the religious beliefs variables those representing the redeeming aspect of religion – Belief in God, Heaven and Soul - have a more pronounced statistically significant impact on income than those capturing the punishing aspects 26 (Belief in Devil and Sin). Like some of the previous empirical studies, we find a negative relationship between weekly religious participation and income levels. Participation in religious ceremonies at least once a year, however, demonstrates the same non-linear relationship with income as the religious attitudes and beliefs variables. We also find that the three important controls – institutions, trade and geography - exhibit strong linkages with income both statistically and economically. While endogeneity of the explanatory variables (including some of the religion variables) is an ongoing issue, we find that our instrumental variables for institutions, trade, and religion perform well in most cases. The robustness of our results is supported in two ways. First and foremost, we use several religion variables for each of the three categories that we examine: Two for religious attitudes, five for religious beliefs and three for attendance. In addition, we construct an aggregate index of religiosity based on the three categories. With the exception of two of the attendance measures all measures point to the same inverted Ushape relationship between religion and income, albeit by various degrees of significance. Second, we find that most of our cross-section findings carry over to the panel data framework where we can control for region and time effects. There are several possible extensions of our work. As noted in Section 2, economic development has several dimensions besides per capita income. It will be interesting to explore the linkage between religion and some other development measures such as educational attainment, health, income inequality and attitude towards violence. Following Gruber (2005), it would be interesting to examine the role of religion and economic outcomes at the individual level, with individual data coming from several countries as in Guiso et al. (2003). 27 References Abadie, Alberto (2006), "Poverty, Political Freedom, and the Roots of Terrorism," American Economic Review, 96(2), 50-56. Acemoglu, Daron, Simon Johnson and James A. Robinson (2001), “The Colonial Origins of Comparative Development: An Empirical Investigation,” American Economic Review, 91(5), 1369-1401. Barrett, David B., George T. Kurian, and Todd M. Johnson (2001), World Christian Encyclopedia (2nd edition), Oxford: Oxford University Press. Barro, Robert J., Rachel M. McCleary (2003), “Religion and Economic Growth Across Countries,” American Sociological Review, 68 (5), 760-781 Baum, Christopher F., Mark E. Schaffer, and Steven Stillman (2003), “Instrumental variables and GMM: Estimation and testing,” Stata Journal, 3(1), 1-31. Central Intelligence Agency (2001), The World Factbook, 2001, Washington, DC: U.S. Government Printing Office. Clague, Christopher, Stephen Knack and Mancur Olson (1999), “Contract-Intensive Money: Contract Enforcement, Property Rights, and Economic Performance,” Journal of Economic Growth, 4(2), 185-211 Cragg, John G. and Stephen G. Donald (1993), “Testing Identifiability and Specification in Instrumental Variable Models,” Econometric Theory, 9, 222 – 240. Dincer, Oguzhan C.(2008), “Ethnic and religious diversity and corruption,” Economic Letters, 99, 98-102. Dollar, David and Aart Kraay (2003), "Institutions, trade, and growth : revisiting the evidence," Policy Research Working Paper Series 3004, The World Bank. Dufour, Jean. M. (2003), “Identification, Weak Instruments, and Statistical Inference in Econometrics,” CIRANO Working Paper. Durlauf, Steven N., Kourtellos, Andros and Tan, Chih Ming (2006), "Is God in the Details? A Reexamination of the Role of Religion in Economic Growth" (September 27, 2006). Available at SSRN: http://ssrn.com/abstract=933989 Easterly, William and Levine, Ross (2003), “Tropics, Germs, and Crops: How Endowments Influence Economic Development,” Journal of Monetary Economics, 50(1): 3-39 Eisenstadt, Shmuel N. (1968), The Protestant Ethic and Modernization: A Comparative View. New York: Basic Books. Filmer, Deon and Lant H. Pritchett (2001), “Estimating Wealth Effects without Expenditure Data--Or Tears: An Application to Educational Enrollments in States of India,” Demography, 38 (1), 115-132. Frankel, Jeffrey A. and David Romer (1999), “Does Trade Cause Growth?,” American Economic Review, 89(3), 379-99 28 Gallup, John Luke, Jeffrey D Sachs and Andrew D Mellinger (1998), “Geography and Economic Development,” Harvard Institute of Economic Research Working Papers, Harvard - Institute of Economic Research. Greif, Avner (1994), “Cultural beliefs and the organization of society: A historical and theoretical reflection on collectivist and individualist societies,” Journal of Political Economy, 102(5), 929-950 Grim, Brian J. and Roger Finke (2006), "International Religion Indexes: Government Regulation, Government Favoritism, and Social Regulation of Religion," Interdisciplinary Journal of Research on Religion, 2(Article 1), 1-40 Guiso, Luigi, Paola Sapienza, and Luigi Zingales (2003), “People’s Opium? Religion and Economic Attitudes,” Journal of Monetary Economics, 50, 225-282. Gruber, Jonathan (2005), "Religious Market Structure, Religious Participation, and Outcomes: Is Religion Good for You?," Advances in Economic Analysis & Policy, Berkeley Electronic Press, 5(1), 1454-1454 Gwin, Carl R. and North, Charles M (2004), “Religion’s Role in the Rule of Law”, Working Paper. Hall, Alastair R., Glenn D. Rudebusch, and David W. Wilcox (1996), “Judging Instrument Relevance in Instrumental Variables Estimation,” International Economic Review, 37(2), 283-298. Hausman, Jerry (1978), “Specification Tests in Econometrics,” Econometrica, 46(6), 1251-1271. Iannaccone, Laurence.R. (1998), “Introduction to the Economics of Religion,” Journal of Economic Literature, 36 (3), 1465–1496. Inglehart, Ronald and Wayne E. Baker (2000), “Modernization, Cultural Change and the persistence of Traditional Values,” American Sociological Review, 65, 19-51 Inglehart, Ronald (2000), “World values surveys and European values surveys, 19811984, 1990-1993 and 1995-1997 [Computer File]”, Ann Arbor-Michigan, Institute for Social Research, ICPSR version. Jacob, Jeffry and Thomas Osang (2008), "Institutions, Geography and Trade: A Panel Data Study," RePEC working paper. Kaiser, H. F. (1974), “An Index of Factor Simplicity,” Psychometrika, 39, 31-36 Kaufmann, Daniel, Aart Kraay and Massimo Mastruzzi (2003), “Governance Matters III: Governance Indicators for 1996–2002”. www.worldbank.org/wbi/governance/govdata2002/ Landes, David S., (1998), The Wealth and Poverty of Nations, Little, Brown, and Company, London. Lewbel, Arthur (1997), “Constructing Instruments For Regressions With Measurement Error When No Additional Data Are Available, With An Application To Patents And R&D,” Econometrica, 65(5), 1201-1213. 29 North, Douglass C. (1994), “Institutions Matter,” Working Paper Archive at WUSTL. Paper # 9411004 Pagan, Adrian. R. and Anthony D. Hall (1983), “Diagnostic Tests As Residual Analysis,” Econometric Reviews, 2(2), 159-218. Rodrik, Dani, Arvind Subramanian and Francesco Trebbi, (2004). “Institutions Rule: The Primacy of Institutions over Geography and Integration in Economic Development,” Journal of Economic Growth, 9(2):131-165 Sachs, Jeffrey D (2003). “Institutions Don't Rule: Direct Effects of Geography on Per Capita Income”, National Bureau of Economic Research, Inc, NBER Working Papers: 9490. Data on Malaria Ecology obtained from: http://www.earth.columbia.edu/about/director/malaria/index.html Samuelsson, Kurt (1993), Religion and Economic Action: The Protestant Ethic, the Rise of Capitalism, and the Abuses of Scholarship, Toronto: U. Toronto Press. Shea, John. (1997), “Instrument Relevance In Multivariate Linear Models: A Simple Measure,” Review of Economics and Statistics, 49(2), 348-352. Stock, James H. and Motohiro Yogo (2002), "Testing for Weak Instruments in Linear IV Regression," NBER Technical Working Papers 0284. Tabellini Guido (2005), “Culture and institutions: economic development in the regions of Europe,” Working Paper. Tawney, Richard H (1926), Religion and the Rise of Capitalism, New York: Harper and Row. Weber, Max ([1905]1958), The Protestant Ethic and the Spirit of Capitalism, translated by Talcott Parsons.New York: Free Press. World Bank (1993), The East Asian Miracle: Economic Growth and Public Policy, New York: Oxford University Press. World Bank (2003), World Development Indicators CD ROM, 2003, Washington DC, USA. 30 Table 1(a): Summary Statistics I: Distribution of Religious Attitudes and Beliefs Attitudes Beliefs in.. (% of "yes" answer) Country Argentina Armenia Australia Austria Azerbaijan Bangladesh Belarus Belgium Brazil Bulgaria Canada Chile China Colombia Czech Republic Denmark Dominican Republic Estonia Finland France Georgia Germany Ghana Hungary Iceland India Ireland Italy Japan Korea, Rep. Latvia Lithuania Macedonia, FYR Mexico Moldova Religion Comfort in Important Religion (% (On a scale of of "yes" 1-4) answer) 2.90 65.99% 2.82 69.84% 2.52 48.99% 2.70 60.76% 3.10 87.65% 3.80 98.78% 2.39 56.55% 2.33 50.64% 3.44 87.67% 2.18 46.07% 2.79 64.75% 3.18 81.06% 1.51 3.43% 3.36 90.70% 1.77 2.10 23.70% 3.33 80.99% 1.92 84.46% 2.37 51.17% 2.27 37.55% 3.18 82.65% 2.26 51.19% 3.81 96.67% 2.52 46.49% 2.65 75.73% 3.32 87.57% 3.28 79.43% 2.87 67.58% 1.93 40.02% 2.61 2.11 62.55% 2.46 68.30% 2.88 57.52% 3.11 82.85% 2.88 92.30% Attend church at least… God 91.96% 85.64% 82.38% 86.69% 97.79% 98.62% 60.21% 76.68% 98.86% 53.72% 91.21% 96.50% Heaven 65.05% 41.21% 63.74% 47.10% 61.27% 97.87% 28.70% 36.19% 79.30% 21.59% 73.87% 78.04% Soul 79.01% 64.88% 79.08% 72.49% 69.40% 97.99% 59.21% 61.71% 83.04% 41.36% 85.58% 81.70% Devil 46.43% 40.28% 44.63% 23.05% 47.29% 95.91% 28.82% 20.84% 50.26% 17.12% 43.98% 54.94% Sin 68.46% 65.39% 71.38% 66.37% 75.31% 87.71% 62.88% 48.86% 85.50% 40.81% 75.11% 87.11% 99.10% 81.68% 87.12% 40.65% 85.45% 57.60% 92.74% 51.77% 78.27% 64.58% 93.24% 78.51% 98.95% 61.33% 83.55% 95.74% 96.66% 89.53% 61.45% 60.35% 65.93% 86.61% 83.84% 95.25% 90.59% 16.01% 81.14% 20.58% 57.96% 30.72% 62.43% 39.19% 92.39% 23.61% 58.17% 58.83% 87.02% 49.99% 38.39% 53.38% 20.49% 70.14% 45.71% 75.11% 53.43% 42.96% 88.72% 63.49% 77.19% 53.30% 75.33% 77.86% 89.01% 29.91% 87.42% 77.90% 86.63% 74.80% 73.11% 72.67% 75.80% 85.96% 79.14% 77.01% 72.32% 9.89% 76.18% 26.00% 38.73% 19.44% 55.60% 19.81% 73.91% 14.17% 17.27% 39.33% 55.33% 37.54% 19.12% 46.88% 25.76% 59.30% 38.01% 51.95% 44.92% 24.27% 83.46% 56.57% 69.20% 43.79% 83.17% 61.56% 86.17% 42.55% 69.82% 73.77% 87.33% 70.40% 29.11% 57.50% 60.14% 88.65% 83.52% 78.80% 83.90% Weekly 28.85% 7.49% 22.27% 25.50% 5.92% 63.75% 4.40% 26.27% 34.70% 6.34% 29.66% 26.06% 0.35% 45.81% 6.36% 2.22% 44.25% 3.63% 3.98% 10.89% 9.60% 17.92% 80.22% 16.33% 2.41% 50.15% 80.18% 35.94% 2.93% 18.31% 3.99% 15.90% 10.95% 48.02% 10.82% Monthly 50.76% 29.50% 32.44% 44.12% 13.98% 76.47% 10.16% 33.58% 51.90% 12.50% 43.27% 45.20% 0.88% 66.57% 10.79% 9.04% 55.26% 8.63% 11.80% 17.32% 26.48% 32.21% 82.42% 25.42% 10.16% 64.51% 86.68% 50.78% 12.68% 39.64% 12.26% 31.47% 17.69% 67.61% 22.65% Yearly 77.91% 74.65% 68.79% 67.90% 70.99% 90.99% 43.30% 48.61% 68.92% 44.86% 66.85% 63.64% 2.85% 80.43% 35.71% 40.83% 73.35% 50.39% 57.17% 38.15% 74.42% 57.14% 84.62% 57.44% 54.81% 90.53% 92.95% 77.13% 71.33% 65.90% 56.16% 77.34% 71.16% 82.04% 75.51% 31 Table 1(a), Contd.: Summary Statistics I: Distribution of Religious Attitudes and Beliefs. Attitudes Beliefs in.. (% of "yes" answer) Religion Comfort in Important Religion (% (On a scale of of "yes" 1-4) answer) Country God Heaven Soul Devil 2.34 Netherlands 45.48% 66.56% 39.71% 71.72% 18.73% 3.87 Nigeria 97.99% 98.62% 96.93% 88.72% 65.10% 2.32 Norway 40.88% 69.58% 47.24% 57.40% 27.23% 3.75 Pakistan 97.36% 3.37 Peru 91.77% 97.99% 85.85% 89.72% 68.81% 3.76 Philippines 92.99% 99.75% 98.23% 96.51% 92.15% 3.33 Poland 85.35% 96.60% 80.13% 85.84% 50.47% 2.56 Portugal 67.30% 85.66% 55.73% 66.20% 27.60% 3.62 88.77% 99.31% 91.99% 94.93% 78.83% Puerto Rico 3.10 Romania 75.68% 93.71% 57.49% 76.07% 42.26% 2.22 46.30% 55.63% 25.61% 59.63% 27.27% Russian Federation 2.31 Slovak Republic 2.38 Slovenia 54.20% 63.52% 32.83% 58.58% 21.81% 3.54 90.06% 98.39% 89.00% 85.38% 55.84% South Africa 2.62 60.88% 89.58% 56.79% 73.42% 37.36% Spain 2.08 Sweden 30.18% 53.37% 32.62% 59.17% 14.25% 2.47 Switzerland 54.91% 83.10% 44.28% 85.11% 29.47% 2.52 Taiwan 67.90% 76.92% 59.40% 76.70% 66.89% 3.52 88.61% 97.56% 88.28% 88.88% 76.69% Turkey 2.55 Ukraine 88.55% 76.53% 44.24% 67.62% 47.25% 2.42 47.88% 81.15% 61.38% 69.92% 33.11% United Kingdom 3.30 United States 81.04% 96.55% 87.28% 92.84% 71.91% 2.53 54.97% 86.65% 50.85% 60.96% 27.16% Uruguay 3.41 Venezuela, RB 88.68% 99.08% 88.44% 90.19% 58.17% Attend church at least… Sin 48.10% 71.06% 49.20% Weekly 21.48% 83.09% 5.21% Monthly 34.56% 85.63% 13.34% Yearly 54.28% 87.81% 48.76% 94.11% 99.83% 90.75% 68.22% 96.52% 76.66% 59.37% 42.88% 70.00% 60.69% 33.29% 51.77% 18.64% 2.06% 33.33% 22.50% 46.99% 32.51% 4.60% 18.11% 11.19% 38.26% 10.12% 14.07% 42.94% 13.25% 30.92% 64.07% 89.92% 79.34% 41.16% 65.06% 30.64% 7.03% 40.22% 34.24% 63.30% 43.97% 11.63% 33.52% 23.74% 41.23% 17.60% 24.08% 57.59% 23.27% 49.33% 80.85% 97.33% 93.01% 53.09% 81.54% 85.64% 30.07% 61.72% 63.84% 73.57% 61.69% 38.46% 59.17% 56.05% 65.71% 56.12% 43.72% 74.56% 33.79% 77.92% 53.28% 71.16% 62.97% 34.37% 60.42% 43.88% 91.70% 72.78% 72.80% 89.76% 52.53% 93.51% 32 Table 1 (b): Summary Statistics II: Index of Religiosity Countries (Highest to Lowest Index of Index Value) Attendance Bangladesh 0.760 Philippines 0.774 Pakistan 0.760 Nigeria 0.836 Ghana 0.751 Puerto Rico 0.636 Ireland 0.781 Peru 0.619 Poland 0.711 Colombia 0.616 Venezuela, RB 0.543 South Africa 0.592 Dominican Republic 0.585 United States 0.566 Turkey 0.487 India 0.691 Brazil 0.535 Mexico 0.630 Chile 0.465 Romania 0.520 Argentina 0.534 Italy 0.544 Canada 0.466 Azerbaijan 0.387 Spain 0.460 Lithuania 0.472 Georgia 0.442 Portugal 0.438 Macedonia, FYR 0.422 Korea, Rep. 0.422 Austria 0.478 Iceland 0.287 Moldova 0.429 Australia 0.433 Taiwan 0.301 Switzerland 0.400 Finland 0.336 Ukraine 0.325 Slovak Republic 0.454 Slovenia 0.424 Armenia 0.422 United Kingdom 0.293 Germany 0.388 Belgium 0.361 Norway 0.275 Uruguay 0.246 Netherlands 0.369 Fraction of Religious population 1.000 1.000 1.000 1.000 1.000 1.000 0.984 1.000 0.946 1.000 1.000 0.980 1.000 0.984 0.985 1.000 1.000 1.000 1.000 1.000 1.000 0.995 0.949 0.808 0.985 0.801 0.692 0.970 0.927 1.000 0.926 0.990 0.655 0.916 0.971 0.956 0.974 0.771 0.907 0.921 0.653 0.890 0.853 0.926 1.000 0.843 0.856 Index of Religious Attitude 0.970 0.935 0.955 0.973 0.959 0.896 0.807 0.880 0.843 0.873 0.869 0.893 0.821 0.818 0.883 0.853 0.868 0.803 0.803 0.766 0.693 0.696 0.672 0.825 0.632 0.649 0.810 0.656 0.648 0.654 0.641 0.709 0.821 0.560 0.654 0.583 0.552 0.761 0.578 0.568 0.702 0.541 0.539 0.545 0.494 0.591 0.520 Index of Religious Belief 0.956 0.973 0.956 0.841 0.881 0.923 0.826 0.873 0.808 0.788 0.859 0.800 0.844 0.877 0.886 0.691 0.794 0.756 0.797 0.692 0.702 0.645 0.739 0.702 0.640 0.781 0.740 0.607 0.660 0.582 0.591 0.632 0.690 0.682 0.648 0.605 0.643 0.617 0.455 0.460 0.595 0.637 0.554 0.489 0.501 0.556 0.490 Religiosity Index 3.685 3.682 3.671 3.651 3.591 3.455 3.397 3.372 3.307 3.277 3.272 3.265 3.250 3.244 3.241 3.236 3.197 3.189 3.064 2.978 2.929 2.880 2.827 2.722 2.717 2.703 2.684 2.672 2.657 2.657 2.636 2.619 2.595 2.592 2.574 2.545 2.505 2.474 2.394 2.373 2.372 2.361 2.333 2.320 2.271 2.237 2.234 33 Table 1 (b), Contd.: Summary Statistics II: Index of Religiosity Countries (Highest to Lowest Index of Index Value) Attendance Hungary 0.376 Japan 0.373 France 0.242 Sweden 0.228 Bulgaria 0.275 Latvia 0.309 Belarus 0.247 Estonia 0.257 Denmark 0.221 Czech Republic 0.207 Russian Federation 0.174 China 0.021 Fraction of Religious population 0.890 0.879 0.968 1.000 0.899 0.630 0.660 0.576 0.954 0.685 0.609 0.467 Index of Religious Attitude 0.548 0.441 0.472 0.410 0.503 0.577 0.581 0.662 0.381 0.442 0.509 0.206 Index of Religious Belief 0.343 0.442 0.424 0.388 0.349 0.496 0.480 0.437 0.301 0.455 0.455 0.147 Religiosity Index 2.156 2.135 2.105 2.026 2.026 2.011 1.967 1.932 1.856 1.789 1.747 0.842 34 Table 1 (c ): Summary Statistics III: All Variables Variable Obeservations Mean Dependent Variable GDP per capita in 2000 (in 1996 intern'l $) Religion Important (on a scale of 1-4) % find Comfort and strength in Religion % Believe in God % Believe in Heaven % Believe in Soul % Believe in Devil % Believe in Sin % Attend at least Weekly % Attend at least Monthly % Attend at least Yearly Religiosity Index (on a scale of 0-4) a Contract Intensive Money (CIM) Rule of Law (Average for the 1990s) a Import Tarrifs a Trade Share Malaria Ecology Instruments for Institutions b a Confidence in Goverenment a Confidence in Army a Confidence in Parliament a Confidence in Police Instruments for Trade c a Attitude Towards Competetion a Attitude Towards Hard Work a Attitude Towards Wealth Accumulation Instruments for Religion a Presence of State Religion Index of Government Regulation of Religion † Hirfendahl Index of Religious Preference Std. Dev. Min Max 130 8454.19 a Religion Variables 59 2.79 56 68.20% 55 83.12% 55 58.63% 55 74.88% 55 42.94% 55 69.22% 58 26.04% 58 37.61% 58 64.37% 59 2.69 Other Explanatory Variables 8529.20 481.84 44008.48 0.58 0.22 0.15 0.23 0.14 0.21 0.18 0.22 0.24 0.18 0.59 1.51 3.43% 51.77% 16.01% 29.91% 9.89% 24.27% 0.35% 0.88% 2.85% 0.84 3.87 98.78% 99.75% 98.23% 97.99% 95.91% 99.83% 83.09% 89.92% 97.33% 3.69 107 76.34 171 0.01 145 9.40 164 72.64 160 3.71 Variables used as Instruments 13.95 0.97 14.08 43.95 6.52 32.99 -1.83 0.03 0.18 0.00 94.52 2.21 160.65 242.92 31.55 45 59 58 59 2.33 2.65 2.36 2.50 0.32 0.39 0.34 0.37 1.81 1.95 1.76 1.85 3.30 3.76 3.44 3.20 58 58 58 7.49 6.59 6.45 0.53 0.81 0.77 6.27 4.24 2.64 8.98 8.63 7.78 59 0.36 0.48 57 2.55 2.72 173 0.58 0.24 a: Variables are time varying. Cross section averages are reported in the table. b: These variables are drawn from the WVS. Individual responses raged from 1-4 c: These variables are drawn from the WVS. Individual responses raged from 1-10 †: Data from World Christian Encylopedia, which has religious preference data for 173 countries 0 0 0.09 1 9.2 0.98 35 Table 2: Impact of Religious Attitudes - Cross Section Results 1 2 3 Ln CIM 3.070 5.079 4.152 (2.68)* (2.80)*** (2.96)*** Ln Import Tariff -0.084 -0.340 -0.283 (1.66) (2.29)** (2.44)** Malaria Ecology -0.015 -0.016 -0.021 (0.74) (0.62) (1.02) Religion Important 4.280 4.734 4.259 (2.43)* (2.89)*** (2.26)** Religion Important Sq -0.887 -0.926 -0.863 (2.92)** (3.31)*** (2.75)*** Observations 35 22 21 R-squared 0.86 0.91 0.94 Joint Test of Religion Var. 0.0000 0.0000 0.0000 H1 Endogeneity Test 0.0087 H2 Endogeneity Test 1.0000 CD Underid. Test 0.0000 0.2011 AR Underid. Test 0.0000 0.0000 Shea's Partial R2: Instn 2 Shea's Partial R : Trade 7 1.267 (0.69) -0.011 (0.15) 0.008 (0.30) 8.999 (1.68)* -1.775 (2.00)** 34 0.83 0.0002 0.0906 0.3432 0.0558 0.4351 0.0000 0.0000 0.0120 0.3097 0.5133 0.0001 0.40 0.56 0.53 0.41 0.75 0.78 0.25 0.57 0.62 0.49 0.46 2 OLS 6 2.929 (1.97)** -0.014 (0.22) -0.015 (0.71) 4.047 (2.43)** -0.874 (3.05)*** 35 0.85 0.0000 0.1438 0.78 Shea's Partial R : Relgn Shea's Partial R : Relgn CD F stat Sargan Overid. Test PH Heteroscedasticity test Reported Modela 5 3.750 (1.56) -0.241 (1.23) 0.012 (0.29) 9.446 (2.01)** -1.776 (2.23)** 21 0.88 0.0000 0.79 2 2 4 2.530 (0.98) -0.303 (1.17) -0.041 (1.32) 4.027 (2.23)** -0.820 (2.76)*** 22 0.91 0.0001 0.7225 0.15 0.11 0.81 0.6101 0.9474 0.48 0.20 0.1277 0.5286 1.03 0.1421 0.9792 0.16 0.31 0.8421 0.9944 6.57 0.0180 0.8686 0.12 0.34 0.2461 0.8057 IV (A) IV-R (A) IV ( B) IV-R ( B) IV (C) IV-R (C) a: IV(A): Uses attitudes towards hard work, competition and wealth accumulation and their square terms as instruments for trade and confidence in government, parliament, army and police and their square terms for institutions. IV(B): Uses first principal components of (i) levels and (ii) squares of the variables in IV(A) as instruments for institutions and trade. IV(C): Uses the second and third order centered moments of the endogenous variables as instruments. In IV-R, we use as intruments not only instrument for trade and institution variables but also for the religion variables. Specifically, we use the Hirfendahl index of religious preference, an index of government regulation of religion, and the presence of a state religion as instruments in sets A, B and C. Notes: Dependent variable is Ln GDP per capita in 2000 (in 1996 international $). All explanatory variables are defined in Section 3. Absolute value of z statistics in parentheses ; +/*/**: significant at 10% / 5% /1%, respectively; p values of test statistics in italics. Explanation of the various diagnostic and specification tests reported: Joint test of religion var.: The Wald test of the null that the two variables are not jointly significant. H1 (Hausman, 1978) Endogeneity Test: Specification test of OLS null vs. IV alternative; H2 (Hausman, 1978) Endogeneity Test: Specification test of IV null vs. IV-R alternative. CD (Cragg and Donald, 1993) Underid. Test: CD chi square test of the null that the matrix of reduced form coefficients has rank=K-1 where K=number of regressors. AR ( Anderson-Rubin) Underid. Test (Hall et al., 1996): Similar to CD Underid. Test but robust to the presence of weak instruments. We report the chi-square version of the test. Shea's Partial R2 : Shea's (1997) "partial R-squared" is a diagnostic statistic for determining the strengths of instruments when there are multiple endogenous variables. CD (Cragg and Donald, 1993) F stat: Test for weak instruments (Stock and Yogo, 2002). For example, the null of a 30% bias in the IV estimator is rejected if the CD F stat. is larger than 4.73 for the case of 2 endogenous variables and 4 instruments. Sargan Overidentification Test: Tests the validity of the null of the exclusion restrictions in the second stage regression; PH (Pagan and Hall, 1983) Heteroscedasticity test: Under the null the errors are homoskedastic. 36 Table 2 (contd.): Impact of Religious Attitudes - Cross Section Results 8 9 10 11 Ln CIM 2.864 3.439 1.432 1.792 (2.47)* (2.04)** (0.86) (0.71) Ln Import Tariff -0.059 -0.417 -0.509 -0.329 (1.07) (2.61)*** (2.67)*** (1.10) Malaria Ecology -0.028 -0.038 -0.038 -0.053 (1.46) (1.79)* (1.88)* (1.99)** Comfort in Religion 7.280 9.775 15.345 9.027 (2.95)** (3.39)*** (2.70)*** (2.65)*** Comfort in Religion Sq -7.427 -8.537 -12.668 -8.272 (3.83)** (4.05)*** (3.21)*** (3.62)*** Observations 34 21 20 21 R-squared 0.87 0.92 0.93 0.92 Joint test of Religion 0.0000 0.0000 0.0000 0.0000 H1 Endogeneity Test 0.0141 0.6840 H2 Endogeneity Test 0.0000 CD Underid. Test 0.0001 0.8923 0.1796 AR Underid. Test 0.0000 0.0000 0.3835 Shea's Partial R2: Instn 2 Shea's Partial R : Trade 0.0845 0.6172 0.0000 0.0000 0.0407 0.8102 0.7467 0.0002 0.38 0.56 0.53 0.41 0.66 0.46 0.20 0.26 0.63 0.42 0.24 2 OLS 14 1.062 (0.63) -0.003 (0.040) -0.025 (1.28) 12.332 (1.55) -11.891 (2.15)** 33 0.86 0.0001 0.74 Shea's Partial R : Relgn Shea's Partial R : Relgn CD F stat Sargan Overid. Test PH Heteroscedasticity test Reported Modela 13 2.450 (1.65)* 0.013 (0.20) -0.030 (1.49) 7.240 (3.12)*** -7.769 (4.21)*** 34 0.86 0.0000 0.1648 0.79 2 2 12 1.629 (0.90) -0.377 (1.57) -0.039 (1.84)* 13.901 (1.96)* -11.847 (2.46)** 20 0.94 0.0000 0.14 0.08 0.41 0.4475 0.9540 0.26 0.08 0.6459 0.9415 0.76 0.1425 0.9005 0.16 0.21 0.6839 0.8655 6.66 0.0445 0.9623 0.11 0.20 0.1540 0.9542 IV (A) IV-R (A) IV ( B) IV-R ( B) IV (C) IV-R (C) a: Same as Table 2 Notes: Same as Table 2 37 Table 3: Impact of Religious Beliefs - Cross Section Results 1 2 3 Ln CIM 3.224 4.426 3.883 (1.96)+ (1.96)* (2.26)* Ln Import Tariff -0.088 -0.383 (1.38) (2.16)* Ln Trade Share 0.058 (0.30) Malaria Ecology -0.056 -0.066 -0.049 (2.18)* (2.53)* (1.84)+ Belief in God 19.892 17.908 22.119 (2.26)* (2.24)* (2.41)* Belief in God Sq -13.933 -12.319 -15.774 (2.47)* (2.39)* (2.71)* Belief in Heaven 4 8.131 (3.94)** 5 -1.216 (0.44) 0.168 (0.76) -0.044 (1.52) 19.774 (2.05)* -14.266 (2.32)* -0.213 (0.80) -0.099 (2.80)** 25.699 (2.62)** -18.355 (2.94)** Belief in Heaven Sq R-squared Joint test of Religion H1 Endogeneity Test CD Underid. Test AR Underid. Test Shea's Partial R2: Institution Shea's Partial R2: Trade CD F stat Sargan Overid. Test PH Heteroscedasticity test Reported Modelb 34 0.79 0.0098 OLS 22 0.0228 0.0584 0.0000 0.0000 0.00 0.00 0.86 0.2279 0.7977 IV (A) 35 0.76 0.0003 22 0.80 0.0001 0.0363 0.0000 0.0000 0.86 0.77 0.76 0.3701 0.8627 35 0.69 0.0000 0.0161 0.0000 0.6483 0.43 0.58 4.89 0.5029 0.4521 OLS IV (A) IV (C) 6 4.101 (2.41)* -0.138 (2.03)+ 7 4.366 (1.98)* -0.548 (3.42)** 8 5.522 (2.99)** 9 10.536 (4.57)** 10 9.126 (2.08)* 11 -0.955 (0.30) 0.231 (1.14) -0.027 (0.85) 0.438 (1.80)+ -0.002 (0.07) 0.799 (1.79)+ -0.023 (0.43) 0.001 0.00 -0.103 (2.27)* -0.037 (1.28) -0.043 (1.52) 3.42 (1.29) -3.763 (1.82)+ 34 0.77 0.0255 7.924 (2.27)* -6.625 (2.60)** 22 0.87 0.0058 0.0250 0.0000 0.0000 0.82 0.81 1.11 0.3615 0.7630 1.53 (0.56) -2.716 (1.23) 35 0.72 0.0028 6.018 (1.28) -6.139 (1.74)+ 22 0.76 0.0004 0.0073 0.0000 0.0000 0.86 0.78 0.78 0.6397 0.9175 6.769 (1.16) -6.685 (1.53) 22 0.68 0.0024 0.0351 0.0836 0.1051 0.31 0.30 1.06 0.4798 0.8166 -0.279 (0.09) -1.356 (0.55) 35 0.60 0.0015 0.0076 0.0000 0.8943 0.39 0.61 4.36 0.6423 0.6067 OLS IV (A) OLS IV (A) IV ( B) IV (C) a: Same as Table 2 Notes: Same as Table 2 38 Table 3 (contd.): Impact of Religious Beliefs - Cross Section Results 12 13 14 Ln CIM 4.217 2.424 5.904 (2.29)* (0.98) (3.01)** Ln Import Tariff -0.153 -0.709 (2.20)* (3.97)** Ln Trade Share 0.266 (1.24) Malaria Ecology -0.055 -0.085 -0.042 (1.85)+ (3.10)** (1.33) Belief in Soul 9.254 24.572 8.81 (1.28) (2.62)** (1.13) Belief in Soul Sq -7.219 -15.941 -7.67 (1.46) (2.69)** (1.44) Belief in Devil 15 10.877 (4.16)** 16 9.604 (2.01)* 17 -2.046 (0.55) 0.511 (1.77)+ -0.036 (0.96) 21.741 (1.51) -15.796 (1.70)+ 0.989 (1.96)* -0.059 (1.09) 28.873 (1.58) -20.273 (1.72)+ -0.019 (0.06) -0.134 (2.68)** 3.139 (0.34) -3.875 (0.62) Belief in Devil Sq 18 3.539 (2.09)* -0.144 (2.27)* 19 3.363 (1.51) -0.553 (3.36)*** 20 3.497 (1.93)+ -0.144 (2.15)* 21 3.692 (1.56) -0.528 (2.99)*** -0.056 (2.10)* -0.082 (3.18)*** -0.071 (2.51)* -0.090 (3.47)*** 0.836 (0.45) -2.096 (1.23) 2.521 (1.32) 2.933 (-1.86)* 2.214 (0.72) -2.576 (1.08) 34 0.74 0.1739 3.668 (1.24) -3.316 (1.50) 22 0.84 0.1780 0.0646 0.0000 0.0000 0.82 0.77 0.94 0.2369 0.5755 OLS IV (A) Belief in Sin Belief in Sin Sq Observations R-squared Joint Test of Religion Var. H1 Endogeneity Test CD Underid. Test AR Underid. Test Shea's Partial R2: Institution Shea's Partial R2: Trade CD F stat Sargan Overid. Test PH Heteroscedasticity test Reported Modela 34 0.73 0.1862 22 0.86 0.0245 0.0297 0.0000 0.0000 0.72 0.68 0.57 0.3507 0.4300 35 0.69 0.0166 22 0.68 0.0156 0.0076 0.0000 0.0000 0.84 0.76 0.69 0.6665 0.9642 22 0.57 0.0382 0.0125 0.0437 0.0000 0.35 0.34 1.29 0.5339 0.7870 35 0.51 0.0336 0.0030 0.0002 0.7712 0.37 0.60 3.88 0.5687 0.6970 34 0.77 0.0240 22 0.86 0.0319 0.0371 0.0000 0.0000 0.79 0.77 0.86 0.3132 0.4149 OLS IV (A) OLS IV (A) IV ( B) IV (C) OLS IV (A) a: Same as Table 2 Notes: Same as Table 2 39 Table 4: Impact of Religious Participation - Cross Section Results 1 2 3 4 Ln CIM 1.885 3.193 -6.022 2.177 (0.99) (1.42) (1.40) (1.14) Ln Import Tariff -0.168 -0.157 (2.89)** (2.60)* Ln Trade Share 0.168 -0.119 (0.71) (0.34) Malaria Ecology -0.089 -0.075 -0.171 -0.069 (2.68)* (1.90)+ (2.96)*** (2.32)* Attend Weekly -3.264 -3.189 -5.818 (2.19)* (1.70) (2.40)** Attend Weekly Sq 3.013 2.416 4.712 (1.54) (0.98) (1.56) Attend Monthly -1.573 (0.94) Attend Monthly Sq 0.449 (0.25) Attend Yearly 6 1.586 (0.56) -0.055 (0.71) 7 3.006 (1.52) 8 5.935 (2.43)* 9 8.437 (1.43) 10 -4.531 (1.26) 0.212 (1.01) -0.068 (2.36)* 0.169 (0.72) -0.058 (1.90)+ 0.767 (1.32) -0.057 (1.23) -0.099 (0.32) -0.14 (3.31)** 8.805 (1.99)* -8.307 (2.46)* 22 0.0029 0.0868 0.0000 0.0000 0.85 0.86 1.09 0.2058 0.8202 10.635 (1.85)* -9.094 (2.07)** 22 0.63 0.0958 0.0911 0.2517 0.0064 0.23 0.22 0.65 0.1646 0.9589 7.724 (1.54) -8.251 (2.13)* 35 0.57 0.0003 0.0042 0.0001 0.6215 0.38 0.58 4.09 0.9253 0.6266 IV (A) IV (B ) IV (C) -0.069 (2.53)* -0.08 (2.25)* 6.12 (1.56) -6.169 (2.02)* 34 0.73 0.0135 0.0975 0.0003 0.0808 0.37 0.60 3.61 0.0292 0.1494 6.466 (1.46) -6.581 (1.95)+ 35 0.71 0.0045 IV (C) OLS 34 0.75 0.0622 35 0.66 0.0636 35 0.46 0.0064 0.0036 0.0002 0.5060 0.36 0.60 3.70 0.9727 0.7381 34 0.74 0.1093 6.325 (1.56) -5.83 (1.87)+ 34 0.75 0.0639 OLS OLS IV (C) OLS OLS Attend Yearly Sq Observations R-squared Joint Test of Religion Var. H1 Endogeneity Test CD Underid. Test AR Underid. Test Shea's Partial R2: Institution Shea's Partial R2: Trade CD F stat Sargan Overid. Test PH Heteroscedasticity test Reported Modela a: Same as Table 2 Notes: Same as Table 2 5 2.429 (1.33) -0.155 (2.45)* 40 Table 5: Impact of Religiosity Index - Cross Section Results 1 2 3 Ln CIM 3.355 2.988 3.656 (3.38)** (2.40)** (3.54)** Ln Import Tariff -0.096 -0.379 (1.85)+ (2.52)** Ln Trade Share 0.191 (1.31) Malaria Ecology -0.029 -0.048 -0.033 (1.51) (2.50)** (1.60) Religiosity Index 3.860 5.774 3.119 (2.06)* (3.26)*** (1.66) Religiosity Index Sq -0.800 -1.104 -0.702 (2.42)* (3.56)*** (2.08)* Observations 36 22 36 R-squared 0.83 0.92 0.83 Joint Test of Religion Var. 0.0037 0.0001 0.0003 H1 Endogeneity Test 0.0155 CD Underid. Test 0.0000 AR Underid. Test 0.0000 4 5.582 (4.70)*** 5 10.671 (1.30) 6 0.369 (0.21) 0.308 (1.84)* -0.042 (1.72)* 4.866 (2.14)** -0.986 (2.45)** 22 0.86 0.0002 0.0020 0.0000 0.0000 0.953 (1.23) -0.032 (0.59) 4.914 (1.16) -0.910 (1.21) 22 0.51 0.4775 0.0131 0.7506 0.3207 0.045 (0.22) -0.056 (2.37)** 3.840 (1.91)* -0.884 (2.42)** 36 0.77 0.0000 0.0150 0.0002 0.9279 2 0.79 0.92 0.07 0.38 2 0.68 0.60 0.5294 0.6278 IV (A) 0.82 1.03 0.8676 0.8829 IV (A) 0.13 0.19 0.9951 0.9978 IV (B ) 0.58 3.81 0.6543 0.7325 IV (C) Shea's Partial R : Institution Shea's Partial R : Trade CD F stat Sargan Overid. Test PH Heteroscedasticity test a Reported Model a: Same as Table 2 Notes: Same as Table 2 OLS OLS 41 Table 6: Impact of Religious Attitudes and Beliefs - Panel Data Results 1 2 3 Ln CIM 3.576 1.678 1.265 (4.64)** (2.29)* (2.16)* Ln Import Tariff -0.039 -0.053 -0.012 (1.22) (2.18)* (0.57) Ln Trade Share Malaria Ecology Comfort in Religion Comfort in Religion Sq -0.022 (1.44) 4.532 (2.84)** -5.023 (3.87)** -0.049 (2.93)** 0.796 (0.51) -2.113 (1.69)+ -0.065 (5.01)** 0.625 (0.56) -1.436 (1.50) 4 -0.164 (0.33) -0.008 (0.45) -0.088 (5.10)** -0.989 (1.03) -0.060 (0.074) Belief in God Belief in God Sq East Asia and Pacific East Europe and Central Asia Latin America North America South Asia Sub-Saharan Africa Middle East and North Africa 1980 1990 Observations Number of countries R-squared Joint Test of Religion Var. BP error components test H Test of Pooled (P) OLS vs. P IV CD Underid. Test AR Underid. Test 61 34 0.80 0.0000 61 34 . 0.0000 0.0000 -0.206 (1.64) -0.614 (2.98)** -0.500 (3.92)** 0.526 (4.07)** -1.491 (7.35)** -0.262 (1.42) -0.299 (1.20) 0.070 (0.80) 0.224 (2.41)* 61 34 0.94 0.0003 -0.392 (2.00)* -0.767 (2.33)* -0.734 (4.28)** 0.508 (2.19)* -1.800 (7.21)** -0.334 (1.02) -0.339 (1.08) 0.153 (3.74)** 0.324 (6.55)** 61 34 . 0.0013 0.0448 5 4.544 (4.37)** 6 1.521 (2.07)* 7 1.917 (2.55)** 0.089 (0.80) -0.035 (1.70)+ -0.176 (2.06)* -0.069 (4.21)** -0.251 (2.83)*** -0.062 (3.94)*** 11.642 (2.33)* -8.662 (2.69)** 7.347 (2.33)* -5.569 (2.65)* -0.525 (3.86)** -0.813 (3.71)** -0.669 (4.53)** 0.408 (2.51)* -1.945 (8.64)** -0.431 (2.15)* -0.448 (1.59) 0.127 (1.35) 0.333 (2.92)** 63 35 0.91 0.0011 7.654 (2.74)*** -5.759 (3.09)*** -0.588 (4.65)*** -0.801 (4.11)*** -0.698 (5.15)*** 0.337 (2.28)** -1.993 (9.50)*** -0.470 (2.63)*** -0.457 (1.83)* 0.117 (1.38) 0.342 (3.22)*** 63 35 0.91 0.0000 63 35 0.67 0.0000 0.0589 0.0000 0.0000 2 0.75 2 0.73 30.15 0.1867 0.0630 IV (C) Shea's Partial R : Institution Shea's Partial R : Trade CD F stat Sargan Overid. Test PH Heteroscedasticity test a Reported Model OLS RE OLS RE OLS OLS a: Same as Table 2 Notes: Same as Table 2 42 Table 7: Impact of Religious Participation - Panel Data Results 1 2 3 a Ln CIM 3.762 1.933 (3.29)** (2.60)** a Ln Trade Share 0.177 (1.57) a Ln Import Tariff Malaria Ecology Attend Weekly 0.372 (3.69)** 4 5 6 7 0.681 (0.95) 0.887 (1.38) 3.655 (3.45)** 1.593 (2.10)* -0.009 (0.01) -0.151 (1.89)+ -0.087 (0.91) -0.104 -0.029 0.023 (2.95)** (1.13) (0.74) -0.042 -0.059 -0.069 -0.064 -0.051 -0.06 -0.082 (2.26)* (2.70)** (4.69)** (3.62)** (2.75)** (3.54)** (4.13)** -1.298 -1.334 -0.874 -0.97 (3.65)** (3.30)** (4.08)** (3.71)** Attend Yearly 2.905 (1.15) -3.017 (1.57) Attend Yearly Sq East Asia and Pacific East Europe and Central Asia Latin America North America South Asia Sub-Saharan Africa Middle East and North Africa 1980 1990 Observations Number of countries R-squared Joint Test of Religion Var. BP error components test H Test of Pooled (P) OLS vs. P IV CD Underid. Test AR Underid. Test 68 35 0.66 68 35 -0.52 (4.45)** -1.106 (6.52)** -0.892 (7.21)** 0.309 (2.07)* -2.07 (9.80)** -0.582 (2.74)** -0.413 (1.50) 0.164 (1.84)+ 0.343 (3.26)** 68 35 0.92 0.0000 -0.559 (3.37)** -1.122 (5.26)** -0.925 (6.08)** 0.338 (1.55) -1.966 (7.62)** -0.58 (1.98)* -0.403 (1.35) 0.179 (3.30)** 0.369 (4.84)** 68 35 65 34 0.67 0.0163 2.731 (1.59) -2.524 (1.91)+ -0.237 (1.81)+ -1.096 (4.76)** -0.762 (5.42)** 0.31 (1.98)+ -1.74 (7.58)** -0.718 (3.07)** -0.491 (1.59) 0.099 (0.98) 0.222 (2.04)* 65 34 0.89 0.0340 3.156 (1.91)+ -3.015 (2.37)* -0.297 (2.35)* -1.234 (5.54)** -0.942 (6.38)** 0.294 (1.95)+ -1.976 (8.47)** -0.699 (3.09)** -0.651 (2.19)* 0.217 (2.04)* 0.315 (2.85)** 65 34 0.87 0.0096 0.0001 0.0079 0.0000 0.0000 2 0.38 2 0.61 7.22 0.2552 0.0818 IV (C) Shea's Partial R : Institution Shea's Partial R : Trade CD F stat Sargan Overid. Test PH Heteroscedasticity test a Reported Model OLS RE OLS RE OLS OLS a: Same as Table 2 Notes: Same as Table 2 43 Figures 1(a) – (d) AZE ARM USA DNK NOR ISL CHE CAN IRL AUS JPN FIN SWEGBRBEL NLD AUT FRA ITA ESP PRT KOR SVN CZE SVK ARG URY EST HUN CHL POL MEX BLR RUS LVA ZAFLTU BRA TUR VEN BGR COL DOM MKD GEO PER UKR ROM PHL AZEARM IND MDA BGD GHA 10 IRL 9 POL PHL CHN 8 8 9 10 USA NOR CAN DNK CHENLD AUSAUT ISL JPN FIN BEL SWE FRAGBR ITA ESP KORSVN PRT CZE ARG SVK HUN CHL URY EST MEX BLR RUS LVA ZAF LTU BRATUR VEN BGR COL DOM MKD GEO UKR PER ROM CHN IND MDA BGD 7 7 GHA NGA 6 6 NGA 0 .2 .4 .6 % attending church atleast weekly Log GDP per capita, 2000 .8 0 .2 .4 .6 % attending church atleast yearly Log GDP per capita, 2000 JPN FRANLD NOR BELFIN HUN BLR 9 RUS AUT PRT KOR SVN EST CHE AUS ISL GBR CAN ITA ESP CZE URY LVA CHL POL MEX ZAF BRA TUR VEN COL DOM GEO PER ROM UKR 8 MKD ARM IND PHL AZE Fitted values USA IRL NOR CHE CAN DNK AUS ISL JPN SWE NLD BEL FIN AUT ITA FRA GBR ESP KOR SVN PRT SVK ARG LTU BGR USA IRL EST 9 DNK LVA RUS HUN URY BLR LTU BGR UKR ARG MKD CHN CHL POL MEX BRAZAF VENTUR COL GEO DOM PER ROM PHL 8 SWE 1 Figure 1(b): Relationship Between % of Population Attending Religious Ceremonies At Least One a Year and Per Capita Income. 10 Figure 1(a): Relationship Between % of Population Attending Religious Ceremonies At Least One a Week and Per Capita Income. 10 .8 Fitted values ARM AZE IND MDA BGD GHA 7 7 MDA PAK BGD GHA NGA 6 NGA 6 .5 .6 .7 .8 % believing in God Log GDP per capita, 2000 .9 1 Fitted values Figure 1(c): Relationship Between % of Population Believing in God and Per Capita Income. 1.5 2 2.5 3 3.5 Importance of Religion on a Scale of 1-4 Log GDP per capita, 2000 4 Fitted values Figure 1(d): Relationship Between Importance of Religion (scale: 14) and Per Capita Income. 44