China Economic Review 31 (2014) 277–287

Contents lists available at ScienceDirect

China Economic Review

Does religious beliefs affect economic growth? Evidence fromprovincial-level panel data in China

Qunyong WANGa,⁎, Xinyu LIN b

a Institute of Statistics and Econometrics, Nankai University, Room 710, Building of Economics School, No. 94 Weijin Road, Nankai District, Tianjin 300071, PR Chinab Renmin University of China, No. 59 Zhongguancun Street, Haidian District, Beijing 100872, PR China

a r t i c l e i n f o

⁎ Corresponding author. Tel.: +86 13820968313 (MoE-mail addresses: QunyongWang@outlook.com (Q. W

http://dx.doi.org/10.1016/j.chieco.2014.10.0061043-951X/© 2014 Elsevier Inc. All rights reserved.

a b s t r a c t

Article history:Received 5 March 2014Received in revised form 20 October 2014Accepted 20 October 2014Available online 29 October 2014

More and more literature on economic growth and development has increasingly focused onlong-run effects of geographic, historical, and cultural factors on productivity and income percapita. This paper investigates the effect of religious beliefs on economic growth using provincialpanel data from 2001 to 2011 in China. It's verymeaningful to study the role of religion playing ineconomic development since religion has influence on political preference, human capital andwork ethic, especially in current China which is faced with income disparity, environmentalpollution, and official corruption.Our results reveal that, among the different religions, Christianityhas the most significant effect on economic growth. This conclusion is consistent among differentestimators and robust with stability over time. However, no consistent or robust conclusions canbe drawn for other religions. Different estimation methods give different signs or significance.Given the very few studies and limited data resources about China in this field, the paper as atentative study provides a brand new viewpoint.

© 2014 Elsevier Inc. All rights reserved.

JEL classification:O11Z12C01

Keywords:Religious beliefsEconomic growthProvincial panel data

1. Introduction

Religion has played an important role in human civilization and social development for centuries. Smith (1776) andWeber (1930)have laid the theoretic foundation for traditional economics and sociology proposition in the economic growth theory. North (1990)regarded religion, customs, ideology and code of conduct as informal institutions which are receiving more and more recognition inthe process of economic development.

But religion was neglected in development studies for a long time. This is partly due to the perception of religion as irrelevant tomodern societies and a constraint on progress. Recently, many trends, such as the inseparability of religion and politics (especially inIslamic countries), the continuing importance of religion in people's lives and identities, have brought religion back into economicsresearch. As argued by Deneulin and Rakodi (2011), “because religion is fundamental to people as they try to make sense of andgive meaning to their lives, and because it is one of the primary sources of values and morality for the majority of the world's popu-lation, it is important that development studies acknowledges the religious dimension of people's lives; understands the relationshipsbetween religion, societies, and states.” Some literature on economic growth and development has increasingly focused on very long-run effects of geographic, historical, and cultural factors on productivity and income per capita. Some papers have found that religionhas in fact historically played a significant role in the development of political and economic institutions, and in the levels of economicgrowth and development realized. Woodberry (2012) argues that although many aspects of “modernity” have been replicated incountries around the world, religion shaped what spread, where it spread, how it spread, and how it adapted to new contexts.

bile).ang), xinyu.lin@outlook.com (X. Lin).

278 Q. Wang, X. Lin / China Economic Review 31 (2014) 277–287

More andmore studies take religion as one important factor in growthmodels. Some empirical studies confirmed the relationshipbetween religion and economic growth, such as Barro andMcCleary (2003), Grier (1997) andNoland (2005). The null hypothesis thatreligious affiliation is uncorrelatedwith economic performances can frequently be rejected (i.e., religionmatters), but the regressionsdo not yield a robust pattern of coefficients with respect to particular religions (Noland, 2005). There exist some contradictingevidences as to the robustness and sign of the effect. Moreover, it is not yet clear about the way that religion affects the economicgrowth. The debate on the economic effects of religion is far from settled.

Several reasonsmotivate us to focus on the effect of religion on economic growth in China. First, China is amulti-ethnic andmulti-religion country with a long history. According to a 2006 survey by the Pew Global Attitudes Project, 31% of the Chinese publicconsiders religion to be very (12%) or somewhat (19%) important in their lives.1 The proportion of religiously affiliated people isabout 16%. Although the percentage is not high, the sheer number is quite large. They play an important role not only in China butalso in the world. Buddhists represent the largest religious group in China, making up between 11% and 16% of the adult population.Christianity is China's second-largest officially recognized religion with less than 4% of the adult population identified as Christian.There is also a large number of Muslims in China, approximately 1.5% of the total population according to the 2000 census. Fromthe point of economic theory and policy making, it's meaningful to understand the religion's role in the economic development.

Second, China has borrowed many experience and institutional model from developed countries since the introduction of thereform and opening policy. Behind the high economic growth, many problems arise, including official corruption, environmentalpollution,moral decline and income disparity etc. It is evidential that cloning institutionswithout certain intangible aspects of cultureoften proves futile. While education and rule of law might seem straightforward secular policies, the cultural forces that carry theminto a society, including religion, have a lot to do with whether people respect them. Religion affects the way that a society operates.Because religion is an important force that shapes people's values, what they consider worthwhile and valuable, it is integral to social,political, and economic life. This is important with regard to not only how prosperous but also how harmonious that a society canbecome. A personwithout religionwho has nothing to fear and nothing to admire has no bottom line to do anything. The loss of beliefis being regarded as one serious social problem in current China and this explains the above problems at root in China. It's the harmonyof economy instead growth rate that Chinaneedsmost for some time to come. So, it'smeaningful to study the role that religion plays inthe economic growth. If religion is an important factor, it cannot only promote economic growth, but also help economy growhealthily.

Our paper focuses on two questions. First, is there any relationship between religion and economic growth? Second, differentreligious groups have different principles and criteria, and therefore there is a significant difference in the implementation of publicpolicy, resulting in a noticeable difference in economic development. If the relationship between religion and economic growth exists,does it vary among different religions? Synthesizing the contributions:

(1) Given the very few studies in this field about China, our research provides new evidences of the relationship between religionand economic growth. As the biggest transition country in the world, China has attracted much attention because of its highgrowth rate, but few literatures, especially quantitative studies, have given concerns to the relationship between religionand economic growth. Under the background of ignorance of religion in developing studies in the world for a long time,there are some specific reasons for China. First, religion is a somewhat sensitive topic in China. Researchers must take thepolitical effect into account for their relevant studies. Somost researchwas conducted along themainstream. Second, the avail-ability of statistical data about religion is very limited. Except some scattered data here today and there tomorrow, there is nosystematic available official data at all. We use the number of religious institutions in each province as the proxy variable ofreligion. The data come from the China Data Center at the University of Michigan. Our paper provides a tentative studyabout the effect of religion on economic growth.

(2) It is a challenge to infer the causality from religion to economic growth. One important reason is the endogeneity of religion.Different with Barro and McCleary (2003) and many other studies, we use the subnational panel data in our analysis, so it iseasy to control for country-specific cultural, economic, political, and institutional influences. Unlike many cross-nationalresults, some economic phenomena that vary across countries such as differences in inflation rates, trade policies, or judicialpractices can be ignored. We use Hausman–Taylor estimator, Amemiya–McCurdy random effect estimator and fixed effectvector decomposition estimator to solve the endogeneity problem because of the time invariance of the religion variable.This avoids the arbitrariness of choosing instrument variables to some extent.

The structure of the paper is as follows. Section 2 reviews the current literature on this topic, and in Section 3 we make aneconometric analysis using the provincial panel data to estimate the effect of religion on economic growth. Robustness check ofmodel stability is given in Sections 4 and 5 concludes.

2. Literature review

The empirical literature on economic growth and development has moved from the study of proximate determinants to the anal-ysis of ever deeper, more fundamental factors, rooted in long-term history. Spolaore and Wacziarg (2013) survey the new literatureand provided a framework to discuss different channels through which inter-generationally transmitted characteristics may impacteconomic development, biologically and culturally. The evidence suggests that economic development is affected by traits that

1 http://www.pewforum.org/2008/05/01/religion-in-china-on-the-eve-of-the-2008-beijing-olympics/.

279Q. Wang, X. Lin / China Economic Review 31 (2014) 277–287

have been transmitted across generations over the very long run. An important issue is whether historically transmitted traits haveaffected development through their direct impact on productivity, or have operated indirectly as barriers to the diffusion of produc-tivity enhancing innovations across populations. In this section, we focus on the literature about the role played by religion in theeconomic development.

2.1. The mixed evidence about the relationship between religion and growth

Most empirical studies found a positive relation between religion and economic growth. Grier (1997) uses data from Spanish andFrench (predominantly Catholic) and British (Protestant) colonies and found that the difference in religion is one of the factors thataffect the colony growth. They also found that the growth rate of Protestantism is positively and significantly correlated with realGDP growth. Based on theWVS (World Value Survey) data of 59 countries, Barro andMcCleary (2003) find that the economic growthis positively correlated with the degree of religion and negatively correlated to church participation. This means that the relationshipbetween religion and economic growth depends on the individual subjective attitude whether has the religion, rather than whetherto join the church. Guiso, Sapienza, and Zingales (2003), using data of 66 countries from WVS, show that active religious system isconducive to economic growth and the increase of per capita income. Noland (2005) finds that religious affiliation is correlatedwith economic performance, but the regressions do not yield a robust pattern of coefficients with respect to particular religions.Differentwith someother studies, his evidence does not support the notion that Islam is inimical to growth. This is basically consistentwith Pryor (2007) who finds no special Muslim economic system using cluster analysis. The presence of Islam has relatively littleinfluence on most economic or social performance indicators. In the analysis of Muslim economies, religion does not appear to be auseful explanatory variable. Alon and Chase (2005) find a positive correlation between the religious freedom and GDP per capitausing a cross section analysis of 54 countries. In their opinion, it's a country's long run interest to expand not only economic freedombut also religious freedom.

There are a few papers with contradicting evidences. For example, Durlauf, Kourtellos, and Tan (2012) use the BMA (BayesianModel Averaging) method to study the robustness of the conclusion of Barro and McCleary (2003). They find that the results arenot robust to changes in the baseline model specifications. With updated dataset, Eum (2011) finds that the significant correlationbetween economic growth and religious activities or beliefs may not hold constant for different time periods. Neither religiousfragmentation nor religious polarization was statistically significant with updated dataset.

2.2. Influence mechanism of religion on economic growth

Althoughmost studies confirmed the relationship between religion and economic growth, it's not clear how religious beliefs affecteconomic growth. Summarizing the literature, we think that it works through but not limited to the following ways. First, religionaffects political preference, the quality of governance and economic policies. Alesina and Giuliano (2011) find that Protestantismtends to decrease and Catholicism tends to increase preferences for redistribution in their analysis with data from the General SocialSurvey and theWorld Values Survey. Luttmer and Singhal (2011) establish a causal effect of culture on preferences for redistributionby analyzing the attitudes of immigrantswithin Europe from different countries of origin. Platteau (2008) argues that the influence ofreligion increaseswhen the state falls into crisis, owing to its impotence or excessive absolutism. Butwhen political and religious func-tions appear to bemerged, religion is the handmaiden rather than themaster of politics. D’Antonio, Tuch, and Baker (2013) describestheways that how religion interacts with partisan affiliation to influence the polarized voting behavior of members of American Con-gress over the last several decades. Wald and Calhoun-Brown (2014) establish that religion plays a fascinating and crucial part inAmerica's political process and culture at large, and that religion continues to be a major part of the American cultural and politicalmilieu while explaining that it must interact with many other factors to influence political outcomes in the United States. Bastenand Betz (2013) exploit a historical quasi-experiment inWestern Switzerland and found that Reformed Protestantism reduces refer-enda voting formore leisure by 14, redistribution by 5, and government intervention by 7 percentage points. These preferences trans-late into higher per capita income as well as greater income inequality.

Second, religion affects education, health and human capital accumulation. Becker andWoessmann (2009) argue that religionwasimportant for economic success in the sense thatwithout intention it resulted in an uneven accumulation of human capital. Protestanteconomies prospered because instruction in reading the Bible generated the human capital crucial to economic prosperity. Usingcounty-level data from late-nineteenth-century Prussia, they found that Protestantism indeed led to higher economic prosperity,but also to better education. Some evidences about the relationship between education and religion have been found, includingAmerica (Chiswick, 1983; Steen, 2005), Canada (Tomes, 1985), British (Brown & Taylor, 2007; Sawkins, Seaman, & Williams, 1997)and Australia (Kortt & Dollery, 2012) etc. For example, Chiswick (1983) find that Jewish men had 16% higher earnings and a 20%higher rate of return on investment in schooling. Tomes (1985), using data from the 1971 Canadian Census, estimates the relationshipbetween earnings and religious affiliation for white males aged between 25 and 64 years. He reports that Jews receive a 46% higherrate to schooling (compared to theweighted average of returns to schooling) and that Protestants received a 9% higher rate of return.Steen (2005) uses the data from National Longitudinal Survey Youth (NLYS) and observes that both men and women aged between35 and 43 years whowere raised as Catholics benefited from an earnings premium than their Protestant counterparts. Using individ-ual level data from the British National Child Development Study (NCDS), Brown and Taylor (2007) support a positive associationbetween education and church attendance, and they find that current participation in religious activities is positively associatedwith past religious behavior. Kortt and Dollery (2012) suggest that Catholic men benefit from a wage premium of 6.7% using thedata from the Household Income Labor Dynamics Survey of Australia. Using national data from the Child Development Supplement

280 Q. Wang, X. Lin / China Economic Review 31 (2014) 277–287

to the Panel Study of Income Dynamics, Chiswick and Mirtcheva (2013) reveal a generally positive and statistically significant asso-ciation between religion and health, especially for the psychological health of children ages 12–15.

Third, religion is beneficial to decreasing transaction cost through “work ethnic”. Weber (1930) argues that the rise of capitalismwas facilitated by “Ascetic Protestantism.”North (1990) declares that the interrelated phenomena of “ideas, organized ideologies, andeven religious zealotry playmajor roles in shaping societies and economies.” Thus, “in the absence of any unified political structure orformal rules, religious precepts usually have imposed standards of conduct on the players”. Work ethic works tominimize the costs ofeconomic transactions, a large part of any economic system.

Other possible channels were also discussed. For example, Dolansky and Alon (2008) find that religion not only affects the eco-nomic development, political structure, legal environment and social behaviors, but also has an enormous effect on internationaltrade environment in Japan. Moreover, national income and the religious diversity can significantly influence the investment plan.Research shows that religions have influences on medical care (Ellison, 1991) and crime (Evans, Cullen, Dunaway, & Burton, 1995).Gruber (2005) analyzes the religious activities of community and the social and economic status of the believers. He concludes thatthe religious market intensity has a positive correlation with religious activities participation, education and income.

On a more general level, culture does matter for political and economic outcomes (Basten & Betz, 2013). Religion and culture aretypically mixed inextricably. The economics profession is only beginning to come to grips with the complexities that may eventuallyhave to be confronted (McCloskey, 2010).

3. Empirical analysis

The complexity and variety of every religion make it almost impossible to reach any conclusion on purely theoretical grounds.Econometric models help us to find the evidence in the real world in spite that every model has some shortcomings. As Box said inhis book “Empirical model-building and response surfaces”, “Essentially, all models are wrong, but some are useful”. In this section,we make an econometric analysis using balanced provincial panel data of China.

3.1. Model and data

We propose an extensive Cobb–Douglas model as follows:

Y ¼ A exp α1 ln religionð Þ þ γ1educ þ γ2beliefð Þ Kβ1Lβ2Qβ3 ; ð1Þ

where Y is the economic output. A is the efficiency coefficient to capture something like technical progress. In order to avoid the biascaused by omitting variables, based on theoretical and empirical studies, we add the control variables as many as possible to isolatethe effect of religion from other confounding effects, including K (physical capital), L (labor), Q (knowledge capital), educ (humancapital), and belief (social capital). Religion is the religious beliefs which we concern most. exp() means the exponent function andln() means nature logarithm function.

The corresponding panel econometric model of function (1) is:

ln Yitð Þ ¼ ln Að Þ þ α1 ln religionð Þit þ γ1 educit þ γ2 belief it

þ β1 ln Kitð Þ þ β2 ln Litð Þ þ β3 ln Qitð Þ þ μ i þ vit :ð2Þ

We explain each variable in detail. Religion is usually excluded from developing economics. Apart from the reasons we discussedabove, another reason is that it's hard to quantify. Some studies use monthly attendance of religious activities, dummy variables ofbelief in hell or heaven, religious composition of societies, or the population distribution according to each religion in individualnations. All these data is not available at provincial level for China. There is still no systematic and continuous microsurvey dataabout religion in China. So, considering the limited availability of religious data,we use the number of religious institutions as religiousvariable. A religious institution is an institution that is not-for-profit and that is established for religious purposes. It consists of church,mosque and temple etc. Its objects and activities reflect its character as a body instituted for the promotion of some religious object,and the beliefs and practices of the members must constitute a religion. In China, most religious activities are organized by the reli-gious institutions, somore religious institutionsmeanwider and deeper effect of religion on social life. Since the existence of religiousinstitutions is objective, this statistical indicator reflects the religious belief more accurately than the number of followers. The lattermay be underestimated by the official statistics because people could hide their real beliefs out of consideration of promotion or beingafraid of prejudice. The data come from the China Data Center at the University of Michigan (CMR, China Data Center).

Y is regional GDP (unit: one hundred million RMB). K is the fixed asset investment (unit: one hundred million RMB). L is theemployment for each city according to the employment of the three divisions of industry overview (unit: ten thousand). We selectpatent grants as the observed value for Q, the output for knowledge. The existing statistical data of social capital is scarce, andsome commonly used proxy indicators, such as area density of nongovernmental organization, the proportion of unpaid blooddonation, have no data on province level. So we use the estimation of social capital from Zhang and Ke (2002). Unfortunately, theydid not continue their estimation, so we assume that the social capital be a constant variable. Descriptive statistics are listed Table 1.

The last year for obtainable religious data is 2004, while the period of other variables is from 2001 to 2011. Because the number ofreligious institutions rarely changed after 2000, we take the religious variable as constant. This is reasonable from the point of taking

Table 1Statistical description of variables.

Symbol Variable Description Mean Std Min Max

Y GDP (billion RMB) Gross National Product 8514.74 8721.34 139.16 53,210.3K Physical capital Fixed investment 4430.251 4632.28 85.77 26,749.7L Labor input (thousands) Total employed population 2260.04 1569.01 125 6485.6Q Knowledge capital Patent licensing 10,309.23 21,770.95 7 199,814educ Human capital Average schooling (years) 8.08 1.15 3.74 11.55belief Social capital Trust index 38.13871 51.85 2.7 218.9religion Religious belief Number of religious institutions 23.33 42.75 0.07 238.39Buddhism Buddhism Number of Buddhism inst. 362.91 496.71 0 2429Taoism Taoism Number of Taoism inst. 83.68 186.81 0 971Islam Islam Number of Islam inst. 377.03 863.71 0 3179Christianity Christianity Number of Christianity inst. 390.6774 567.14 0 2442Others Other religions Number of other religions 1118.516 3622.18 0 20,663

Note: std in the fifth column denotes standard deviation. Data sources include (1) Data of GDP, physical capital, labor input, knowledge capital and human capital areextracted from the China Statistical Yearbook from2001 to 2012. (2) Data of social capital is from Zhang and Ke (2002). (3) Religious data are extracted from China DataCenter at the University of Michigan.

281Q. Wang, X. Lin / China Economic Review 31 (2014) 277–287

religion as a part of culture. It's only changing very slowly, and we can view it as constant in short run. Table 2 lists the standarddeviation (std) and coefficient of variation (cv) of the religious variables from 2000 to 2004, both the number on the whole andthe number for each religion (Buddhism, Taoism, Islam, and Christianity). The nationwide coefficient of variation is only 0.07. Thatmeans the annual variation degree is very small relative to the mean. So it is feasible for us to take religious variables as constant,i.e., the values from 2005 to 2011 are fixed at its levels of 2004.

Fig. 1 is the scatter plot of ln(GDP) and religion. Two basic facts can be read from Fig. 1. First, there is an obvious relationshipbetween the two variables although we cannot infer the causal relationship without controlling other variables. Second, Tibet,Qinghai, Xinjiang, Ningxia, Gansu, Fujian, Zhejiang, the religious relationships between GDP and the number of religious institutionsare clearly distinct from the other provinces. Tibet, Qinghai, Ningxia, Gansu are both major ethnic enclaves and large religious settle-ments inwestern China. In Zhejiang and Fujian, two provinces in eastern China, the religious development has a long history. Accordingto information released by the State Administration for Religious Affairs in P.R.C., Buddhism and Taoism were transmitted to Zhejiangprovince more than 1800 years ago, and it is 1400 years ago that Islam was introduced into China. Catholicism and Christianity wereintroduced into China about 400 years and 150 years ago respectively. Among the “top ten fairy grottoes”, three are located in Zhejiang.Hangzhou Phoenix Temple of Islam is one of the four most ancient temples in Southeast China.While Fujian province does not harborany significant religions, local folk beliefs predominant landscape is highly developed, including animals, Rain God, various folk beliefssuch as ancestor worship. Fujian is the only pilot province of folk beliefs place registration system.

Since ancient times, there has been a variety of religious coexistence and common development in Xinjiang. In addition to Islam,Buddhism, Christianity, Lamaism (Tibetan Buddhism), other religions, such as Roman Catholic, Eastern Orthodox, and Shamanismalso play an important part. There exist a considerable number of places of worship in Xinjiang. This is a reason why Xinjiangseems to be an outlier in Fig. 1 (Table 3).

Taking into account the differences of religious development in the area, we add a dummy area variable into the model. The firstgroup includes Zhejiang, Fujian, Tibet, Gansu, Qinghai, Ningxia and Xinjiang which are labeled on Fig. 1. The base group (dummyvariables with 0) consists of the other 24 provinces. The proportions of the number of religious institutions for each religion areshown in Table 2. As a whole, the total proportion of Buddhism, Christianity, Taoism and Islam accounts nearly 50%.

3.2. Model estimation

Endogeneity and time invariance are two obvious features of our model. First, the religious belief may be endogenous because itcould depend directly on economic growth which means the endogeneity may come from the bi-directional causality. In a sense,religion and economics have long been intertwined. This was implied in the secularization hypothesis of Weber which says thateconomic development causes individuals to become less religious, as is usually measured by church attendance and religion.Endogeneity may come from another source, i.e., omitting important variables which are correlated with both economic growthand religion. To address the issues of the endogeneity problem, Barro and McCleary (2003) use a dummy variable which indicates

Table 2Statistics of the religious institutions from 2000 to 2004.

Buddhism Christianity Taoism Islam Others Total

Mean 11.35 12.35 2.63 2.39 16.09 45.49Std 1.021 1.698 0.499 0.363 4.586 5.936cv 0.0899 0.1389 0.189 0.151 0.2851 0.0722

Note: std and cv in the first column denote standard deviation and coefficient of variation respectively. Data source is China Data Center of the University of Michigan:http://chinadataonline.org/. The definitions of variables are in Table 1.

Fujian

Gansu

NingxiaQinghai

Tibet

Xinjiang

Zhejiang4

68

1012

log(

GD

P)

0 50 100 150 200 250

the number of religious institutions

Fig. 1. Scatter plot of ln(GDP) and the number of religious institutions in China. Note: Data sources include China Statistical Yearbook from 2001 to 2012, China DataCenter at the University of Michigan. The definitions of variables are in Table 1.

282 Q. Wang, X. Lin / China Economic Review 31 (2014) 277–287

the presence of a State Religion in 1970 and a measure of religious plurality as the instrumental variables of religiosity. In this paper,the use of sub-national panel data is helpful to decrease this risk because it is easy to control country-specific effects, such as historicalculture, political institution, monetary policy and other institutional influences. Unlike many cross-national results, some economicphenomena that vary across countries such as differences in inflation rates, trade policies, or judicial practices can be ignored byusing sub-national data.Moreover, all unobserved individual variables can be relatively easily dealtwith in panel data, such as culture,geography, and initial income.

Second, there are two time invariant variables, religious belief and social capital. In a fixed-effect model, these variables areswept away by the within transformation. The random effect estimator is inconsistent if the individual effect is correlated with theindependent variables. Hausman and Taylor (1981, hereafter HT) provid an efficient instrumental variable estimation of the modelwhen the individual effects are correlated with some of the time invariant variables and some of the time varying variables. Valid in-struments are given by the other time invariant and time varying variables in the equation. Amemiya and McCurdy (1986, hereafterAM) propose an instrumental estimator under stricter conditions for this model. However, the small sample properties of HT and AMestimator are not satisfactory. Plümper and Troeger (2007) propose a fixed effect vector decomposition approach (FEVD). The proce-dure consists of three stages: the first stage runs a fixed-effects model without time-invariant variables, the second stage decomposesthe unit effects vector into a part explained by the time-invariant variables and an error term, and the third stage re-estimates thefirststage by pooled-OLS including the time invariant variables plus the error term of stage 2. Their Monte Carlo simulations revealed thatthis method works better than pooled OLS, random effect and HT estimator.

We use the HT, AM and FEVD method to estimate model (2). In comparison with Barro and McCleary (2003), one importantadvantage of our model is that we avoid the arbitrariness of choosing instruments. The result is presented in Table 4. The result ofrandom effect GLS estimation is also listed in comparison. Note that we also add the trend into the model as a control variable.

The Hausman statistics show that the endogeneity is very significant, so the GLS estimation estimates are inconsistent and biased.AmongHT, AMand FEVD, the result of FEVD estimation is very similar to that of HT.Most coefficients of AM are also similar to those ofHT except ln(religion) which is themost concerned variable in our model. All variables have the right sign as expected. Ln(religion) issignificantly positive at 0.01 level in all four methods which implies a significant role of religion in promoting economic growth. Thisresult is consistentwithmost empirical studies in the literature.Wedescribed severalways of how religion affects growth in Section 2.In our opinion, the most important and practical way religion affects economic growth in China is through “work ethic” proposed byWeber. Work ethic is a value based on hard work and diligence. On a larger scale, religious denominations affect economics by

Table 3Proportion of religious institutions for each religion.

Buddhism Christianity Taoism Islam Others

Group 1 10.95% 6.82% 3.63% 20.51% 58.09%Group 2 24.86% 36.81% 3.50% 7.38% 27.45%Overall 15.56% 16.75% 3.59% 16.16% 47.95%

Note: Group 1 consists of the provinces in the base category (dum = 0), i.e., Zhejiang, Fujian, Tibet, Gansu, Qinghai, Ningxia and Xinjiang. Group 2 consists of the otherprovinces (dum = 1). Data source is China Data Center at the University of Michigan. The definitions of variables are in Table 1.

Table 4Estimation result of model (2).

ln(Y) GLS HT AM FEVD

ln(religion) 0.1060 ⁎⁎⁎ 0.7430 ⁎⁎⁎ 0.3610 ⁎⁎⁎ 0.7802 ⁎⁎⁎

ln(K) 0.3230 ⁎⁎⁎ 0.2160 ⁎⁎⁎ 0.2320 ⁎⁎⁎ 0.2126 ⁎⁎⁎

ln(L) 0.4080 ⁎⁎⁎ 0.0463 0.1260 ⁎⁎ 0.0379ln(Q) 0.0489 ⁎⁎⁎ 0.0235 ⁎ 0.0251 ⁎⁎ 0.0216 ⁎

belief (index) 0.0047 ⁎⁎⁎ 0.0087 ⁎⁎⁎ 0.0071 ⁎⁎⁎ 0.0089 ⁎⁎⁎

educ (years) 0.0865 ⁎⁎⁎ 0.0651 ⁎⁎⁎ 0.0636 ⁎⁎⁎ 0.0631 ⁎⁎⁎

trend 0.0630 ⁎⁎⁎ 0.1020 ⁎⁎⁎ 0.0968 ⁎⁎⁎ 0.1040 ⁎⁎⁎

dum −0.3410 ⁎⁎⁎ −1.9540 ⁎⁎⁎ −1.2030 ⁎⁎⁎ −2.0334 ⁎⁎⁎

constant 1.1490 ⁎⁎⁎ 3.4450 ⁎⁎⁎ 3.5960 ⁎⁎⁎ 3.4770 ⁎⁎⁎

N 341 341 341 341Hausman χ2(8) = 245.82⁎⁎⁎ 609.79 ⁎⁎⁎ 211.77 ⁎⁎⁎

Note: trend is a linear time trendvariable. dum is dummyvariablewith 0 denoting Zhejiang, Fujian, Tibet, Gansu, Qinghai, Ningxia andXinjiang, and 1 denoting the otherprovinces. *, **, *** represent the parameter estimator is significant at 0.1, 0.05, 0.01 level. Data sources include China Statistical Yearbook from 2001 to 2012, China DataCenter at the University of Michigan and Zhang and Ke (2002). The definitions of variables are in Table 1.

283Q. Wang, X. Lin / China Economic Review 31 (2014) 277–287

creating bonds of trust and shared commitment among small groups, reducing corruption and increasing respect for law. The loss ofprofessional ethics is becoming an obstacle for economic development in China currently. This aggregates the speculation of earningquick buckmentality. One direct evidence is the LDI (Lottery Development Index)which is defined as the proportion of lottery sales inGDP. As a developing country, the LDI of China reached 0.37% at 2005, whichwas almost on par with that of America. There are morethan 10 million lottery citizens in China. The personal lottery consumption is extraordinary high relative to their disposable income.Lottery consumption consists of 6.09%–7.62% of disposable income per capita. According to a survey conducted by the Institute ofPsychology of Chinese Academy of Science in 2007, 4.4% lottery citizens spend more than 20% of their income on lottery. Whilemost religionspromote thrifty and diligentwork, these traits help people keep down-to-earthworking spirit and thepractical learningattitude. These qualities are also necessary for lending and trade and helpful to decrease transaction cost and promote economicgrowth.

As to the political way through which religion affects growth, we do not think that the way works much in China. The politicaleffect of religion is very limited in China under the leadership of Communist Party. At least it does no work in the way describedabove. But the central government has been arranging some preferential fiscal policy to support the development of the minoritynationality area, such as transfer payment in minority areas besides the general and special transfer payment, and some specialfunds. The growth rate of transfer payment for minority areas is 15.6% from 1978 to 2008. China has established some specialfunds since 1950s, including education allowances in 1951, subsidy for minority areas in 1955, and development funds for minorityarea development in 1992.

As argued by Brock and Durlauf (2001), it's indeed a challenge to infer the causality from religion to economic growth. The diffi-culty lies in the inherent open-endedness of growth theory which means a particular theory is relevant does not logically precludeother theories from also being relevant. So the statistical relationship between growth and religion, especially the causal relationship,needs to account for the plethora of growth determinants that exist in the empirical literature. In spite of the data management andmodel estimation methods, we are well aware of the difficulty in interpreting the observed correlations as causal effects. FollowingGuiso et al. (2003), we prefer to interpret our results as more precisely estimated partial correlations.

3.3. Estimation by religion

There are considerable variations in different religious doctrine that encourage or limit people's specific behaviors. Therefore, theeffect of religion may vary with different religions. We further studied the influence of different religious sects on economic growth.We use five variables for the number of institutions of Buddhism, Taoism, Islam, Christian and other religions. We don't treat theMarxism as a religion or belief in our paper although Marxism is still the dominant ideology in current China.

So model (2) changes into

ln Yitð Þ ¼ ln Að Þ þ α1 ln Buddhismð Þit þ α2 ln Taoismð Þit þ α3 ln Christð Þitþ α4 ln Islamð Þit þ α5 ln Othersð Þit þ γ1 educit þ γ2 belief it

þ β1 ln Kitð Þ þ β2 ln Litð Þ þ β3 ln Qitð Þ þ μ i þ vit:

ð3Þ

All religion variables take thenatural logarithm.Weuse themethod of GLS, HT, AMand FEVD to estimatemodel (3). The results arelisted in Table 5.

We can get the following conclusion based on the result of model estimation.

(1) To summarize, the most robust conclusion we can make is about Christianity. The results of all four models are consistent.Among the religions, the coefficient of Christianity is the most consistent. All models give the same sign and have significanceat 1% level. GLS, HT, AM and FEVD estimation results show that the variable of ln(Christ) has a significant positive effect oneconomic growth. Christianity has a long cultural history of more than 2000 years. In spite of the short history of Christianity

Table 5Estimation result of model (3).

GLS HT AM FEVD

ln(Buddhism) −0.0993 ⁎⁎⁎ −0.0964 −0.1450 −0.0964ln(Taoism) 0.0608 ⁎⁎ −0.0060 0.1210 −0.0060ln(Christ) 0.1040 ⁎⁎⁎ 0.4600 ⁎⁎⁎ 0.1610 ⁎⁎ 0.4603 ⁎⁎⁎

ln(Islam) −0.0420 ⁎⁎⁎ −0.1120 −0.0509 −0.1121 ⁎⁎

ln(Others) 0.0795 ⁎⁎⁎ 0.0232 0.1330 ⁎ 0.0232ln(K) 0.2890 ⁎⁎⁎ 0.2130 ⁎⁎⁎ 0.2260 ⁎⁎⁎ 0.2130 ⁎⁎⁎

ln(L) 0.3080 ⁎⁎⁎ 0.0379 0.1020 ⁎ 0.0379ln(Q) 0.0388 ⁎⁎⁎ 0.0216 ⁎ 0.0239 ⁎ 0.0216 ⁎

belief (index) 0.0040 ⁎⁎⁎ 0.0051 0.0047 ⁎⁎⁎ 0.0051 ⁎⁎⁎

educ (years) 0.0718 ⁎⁎⁎ 0.0631 ⁎⁎⁎ 0.0623 ⁎⁎⁎ 0.0631 ⁎⁎⁎

trend 0.0762 ⁎⁎⁎ 0.1040 ⁎⁎⁎ 0.0990 ⁎⁎⁎ 0.1040 ⁎⁎⁎

dum −0.2760 ⁎⁎ 0.00 −0.6260 ⁎⁎ 0.00Constant 2.0710 ⁎⁎⁎ 3.6110 ⁎⁎⁎ 3.7130 ⁎⁎⁎ 3.6110 ⁎⁎⁎

N 341 341 341 341

Note: trend is a linear time trendvariable. dum is dummyvariablewith 0 denoting Zhejiang, Fujian, Tibet, Gansu, Qinghai, Ningxia andXinjiang, and 1 denoting the otherprovinces. Data sources include China Statistical Yearbook from 2001 to 2012, China Data Center at the University ofMichigan and Zhang and Ke (2002). The definitionsof variables are in Table 1.⁎ Represent the parameter estimator is significant at 0.1 level.⁎⁎ Represent the parameter estimator is significant at 0.05 level.⁎⁎⁎ Represent the parameter estimator is significant at 0.01 level.

284 Q. Wang, X. Lin / China Economic Review 31 (2014) 277–287

in China, it has a vital impact on religion in China. Christian religious group has a high proportion in the population of Chinesereligion. In 2004, Christian institutions constitute 16.75%, not including the underground parts.In Christian view, economic core principle is tomaintain economic freedomon the premise of private property rights. Economicfreedom is conducive to creating values, providing a better life for human society. Christianity maintains that economicdevelopment should contribute to the all-round development of human beings, which is consistent with the people-orienteddevelopment concept. Christianity advocates that “love your neighbor as yourself”. They believe that it is people's right toenjoy economic freedom, and that requires carryingout the relevant obligations. Christians support normal investmentbehaviorrather than speculation. The former is beneficial to create values,while the latter is characterized by disrupting themarket order.As another example, Christians hold the opinion that the development of economy and the institutions should be people-oriented. The development of human nature should be paid attention to when we focus on the economic development. There-fore, humanities education is beneficial to the accumulation of human capital and the improvement of population quality.In essence, the doctrine of Christianity has a great degree of consistency with the ideas of market economy. Creations of moneyand God are regarded as the core goal in economic and religious areas. People pursue to accumulate wealth and glory of God.Christian helps increase the level of economicmarketization, and the development ofmarket economy aggravated the commer-cialization and marketization of Christianity. The equality and fraternity that are advocated by Christianity correspond to thepublicity and impartiality in the market economy. The pursuit of truth corresponds to the rule of equivalent exchange in themarket economy. Improving the quality of product, attitude of service and service projects conforms to the thought of“serving people” in Christian. The cultivation of the consciousness in market economy is the embodiment of the affirmationand praise in Christian. The spread and development of Christianity in some ethnic minority areas in China have obvious effecton the local economic development. It not only can promote the degree of social civilization, raise the level of culture andeducation, but also can regulate the behavior of people.To sum up, Christianity is developing with economic growth. It is conducive to social harmony and stability. People's physical,psychological and ethical is increasing with the help of Christianity. Christianity is beneficial to the accumulation of humancapital, to promote the further development of economy.

(2) No consistent conclusion can be drawn for other religions. The coefficients of Buddhism in all four models are negative, butsome are insignificant. The same case applies for ln(Islam) (with negative signs) and ln(Others) (with positive signs). Thecoefficients of Taoism have different signs and significance in the four models.The goal of Buddhism is transcendentalismbut not secular interests.What Buddhism focuses on is not goods but freedom fromsuffering. The physical production serves for the soul purification. In Tibetan Buddhism, for example, it advocates tolerance andsubmission. In addition, Buddhist economics argues that themost rational method is tomeet the needs of the local productionusing the local resources. It is considered as highly uneconomical to import from far distance. These doctrines are contrary tothe market economic concept and leading to serious lag of marketization and trade. Our conclusion is not consistent with Tu,Bulte, and Tan (2011) who used household survey in rural Tibet and found that there exists a positive but inverted U-shapedrelation between the intensity of religious beliefs and income.As to Islam, the evidences are mixed as reviewed in Section 2. The most widely known Muslim economic doctrine is theprohibition against interest, or usury. However, Pryor (2007) explained in detail that prohibition against interest is obviouslynot absolute. Almost all Muslim governments pay an interest rate on their government bonds andmany pay interest on postalsavings aswell. Over three quarters of the total lending of all Muslim banks take the formof various fixed-rate-of-return instru-ments, such as the mark-up loan. As Pryor (2007) said, “In many nominally Muslim countries, religious injunctions are not

285Q. Wang, X. Lin / China Economic Review 31 (2014) 277–287

incorporated into economic institutions or governmental policies. In addition, some doctrines are obeyed only in form (but notin content) or are simply disregarded.” Some studies claim that historically the subservient role of women inMuslim nations isa source of economic drag. To isolate half of the talent of a population from productive activity is not helpful to productivity. Inaddition to that, leavingwomen uneducated makes things even worse because of the role that mothers play in the intellectualdevelopment of children.

4. A robustness check

As described in the literature review, the conclusion in Barro andMcCleary (2003)was found not robust tomodel specification andupdated dataset. The shortage of robustness arises both frommodelmisspecification and structural change. The estimatorwill changeif we use different subsets of variables. The structural changes can be caused by external shocks or internal reform. This is a usual casefor most developing countries, or transition countries. As to China, the reform policy causes a gradual institutional change whichbecomes an important source of model instability. Moreover, our sample covers the period of financial crisis which began at theend of 2007. So it's interesting to check if our conclusion is stable over time. We check the robustness of the models (2) and (3)using rolling regression method. We first estimate the model using sample from 2001 to 2006, i.e. 6 observations. Then were-estimate the model by shifting the sample forward. So we estimate the model for sample 2001–2006, 2002–2007, etc. We collectthe coefficients of religious variables. Fig. 2 plots the recursive coefficient and its 95% confidence interval of ln(Religion) in model (2).The x-axis is the last year of each sample. The coefficient is always significant. The maximum value appeared in 2008, which wascaused by the 4 trillion investment of the central government to cope with the financial crisis. The estimator changes from 0.28 forsample 2001–2006 to 0.67 for sample 2003–2008.

Fig. 3 plots the recursive coefficients of ln(Buddhism), ln(Taoism), ln(Christ), ln(Islam), and ln(Others) in model (3). On the whole,the fourmodels give consistent results. The same variable in the fourmodel has the same sign and same significance in general. So, wejust make the plot of HT estimators in order to keep the graph clear. The coefficients for each variable remain constant relatively. Thebiggest changes happen at 2008. This may be resulted from the financial crisis which happened at the end of 2007. Among the fivegroups, Christianity is themost prominentwith coefficient greater than zerowhichdenotes a positive relationship between Christianityand growth. On the contrary, the coefficients of Buddhism, Taoism, and Islam fluctuate around zero. There is no significant and stableconclusion we can make for them.

In sum, the models have a reasonable stability over time which verifies the robustness of our conclusions.

5. Conclusions

Religion, a complex objective existence, is a kind of informal institutional factor and a part of the social capital and cultural capital.No matter how we categorize it, it has a significant impact on the subjective and objective world. Religious factors help to solve theproblems of model misspecification and endogenous problem. As Deneulin and Rakodi (2011) say, “religion needs to be broughtback in to development research so that our understanding of challenging development issues can be improved.”

Thepaper is a tentative research about the case of Chinawith very limited data resources. According to the extended Cobb–Douglasproduction function,we estimate a panelmodelwith time invariant variables. The results of GLS, HausmanTaylor, AmemiyaMaCurdy,

0.2

.4.6

.81

coef

ficie

nt

2006 2007 2008 2009 2010 2011year

Fig. 2. Plot of recursive coefficient of religious variable inmodel (2). Note: The coefficient on y-axis is the elasticity of religion onGDP. The x-axis denotes the last year ofthe sample in recursive regression. Data sources include China Statistical Yearbook from2001 to 2012, ChinaData Center at theUniversity ofMichigan and Zhang andKe(2002). The definitions of variables are in Table 1.

-.4-.2

0.2

.4.6

coef

ficie

nt

2006 2007 2008 2009 2010 2011year

BudhismChristianityTaoismIslamOthers

Fig. 3. Plot of recursive coefficients of religious variables inmodel (3). Note: The coefficient on y-axis is the elasticity of each religion on GDP. The x-axis denotes the lastyear of the sample in recursive regression.Data sources include China Statistical Yearbook from2001 to 2012, ChinaData Center at theUniversity ofMichigan and Zhangand Ke (2002). The definitions of variables are in Table 1.

286 Q. Wang, X. Lin / China Economic Review 31 (2014) 277–287

and fixed effect vector decomposition estimation show that religion has a significant positive influence on economic growth overall.This is consistent with most empirical studies. This means that religious beliefs and religious institutions can become a driving forcein determining national economic results. Among the different religions, Christianity has the most significant effect on economicgrowth. This conclusion is consistent among different estimators and robust with stability over time. However, no consistent or robustconclusions can be drawn for other religions. Different estimation methods give different signs or significance.

From the point ofmodel specification, it is necessary to introduce religion into themodelswhenwe do the research and analysis ofinfluence factors of economic growth. From the point of policy making, as McCleary said, the lesson of the research results isn't thatgovernments should boost religion, but that they should recognize it has some value, and avoid regulating it too heavily. The biggerapplication of research into religion isn't to foster religious imperialism but to build a better-informed economics, and in the long run,better policy. Therewon't bemanna fromheaven. But theremight less poverty over time since it seems to take a long time for religiousshifts to catalyze economies. And since religion is a very long-term historical factor of current development, the policy should focus onovercoming long term constraints.

As Basten and Betz (2013) illustrate, the growing interest in investigating the issues raised byWeber is still far been limited to thestatistical testing of certain implications ofWeber's arguments as formulated by economists. The full economic impact of a culture andreligion is not likely to be captured by examining the statistical significance of certain correlations posited among variables taken to beidentifying characteristics of a culture and the economy.

Acknowledgement

We are grateful to the anonymous referees and Professor Jian Yang of the University of Colorado (Denver) for their valuable andconstructive comments, which substantially improve our paper. Liting Su of the University of Kansas helped the revision of this paper.Qunyong Wang acknowledges the financial support from NFSC (Project No. 71101075).

References

Alesina, A., & Giuliano, P. (2011). Family ties and political participation. Journal of the European Economic Association, 9(5), 817–839.Alon, I., & Chase, G. (2005). Religious freedom and economic prosperity. Cato Journal, 25, 399.Amemiya, T., & MaCurdy, T.E. (1986). Instrumental-variable estimation of an error-components model. Econometrica, 869–880.Barro, R.J., & McCleary, R. (2003). Religion and economic growth across countries. American Journal of Sociology, 68, 760–781.Basten, C., & Betz, F. (2013). Beyond work ethic: religion, individual, and political preferences. American Economic Journal: Economic Policy, 5(3), 67–91.Becker, Sascha O., & Woessmann, Ludger (2009). Was Weber wrong? A human capital theory of protestant economic history. Quarterly Journal of Economics, 124(2),

531–596.Brock, W., & Durlauf, S. (2001). Growth empirics and reality. World Bank Economic Review, 15(2), 229–272.Brown, S., & Taylor, K. (2007). Religion and education: evidence from the National Child Development Study. Journal of Economic Behavior & Organization, 63(3),

439–460.Chiswick, B.R. (1983). The earnings and human capital of American Jews. Journal of Human Resources, 18, 313–336.Chiswick, B. R., & Mirtcheva, D. M. (2013). Religion and child health: Religious affiliation, importance, and attendance and health status among American youth[J].

Journal of family and economic issues, 34(1), 120–140.D’Antonio, W. V., Tuch, S. A., & Baker, J. R. (2013). Religion, Politics, and Polarization. Rowman & Littlefield.

287Q. Wang, X. Lin / China Economic Review 31 (2014) 277–287

Deneulin, S., & Rakodi, C. (2011). Revisiting religion: development studies thirty years on. World Development, 39(1), 45–54.Dolansky, E., & Alon, I. (2008). Religious freedom, religious diversity, and Japanese foreign direct investment. Research in International Business and Finance, 22(1),

29–39.Durlauf, S.N., Kourtellos, A., & Tan, C.M. (2012). Is God in the details? A reexamination of the role of religion in economic growth. Journal of Applied Econometrics, 27(7),

1059–1075.Ellison, Christopher G. (1991). Religious involvement and subjective well-being. Journal of Health and Social Behavior, 32, 80–99.Eum, W. (2011). Religion and economic development— a study on religious variables influencing GDP growth over countries. (Doctoral dissertation). Berkeley: University

of California.Evans, T.D., Cullen, F.T., Dunaway, R.G., & Burton, V.S. (1995). Religion and crime reexamined: the impact of religion, secular controls and social ecology on adult crim-

inality. Criminology, 33(2), 195–224.Grier, R. (1997). The effect of religion on economic development: a cross national study of 63 former colonies. Kyklos, 50(1), 47–62.Gruber, J.H. (2005). Religious market structure, religious participation, and outcomes: is religion good for you? Advances in Economic Analysis & Policy, 5(1).Guiso, L., Sapienza, P., & Zingales, L. (2003). People's opium? Religion and economic attitudes. Journal of Monetary Economics, 50(1), 225–282.Hausman, J.A., & Taylor, W.E. (1981). Panel data and unobservable individual effects. Econometrica, 49, 1377–1398.Kortt, M.A., & Dollery, B. (2012). Religion and the rate of return to human capital: evidence from Australia. Applied Economics Letters, 19(10), 943–946.Luttmer, Erzo F.P., & Singhal, Monica (2011). Culture, context, and the taste for redistribution. American Economic Journal: Economic Policy, American Economic

Association, 3(1), 57–79.McCloskey, D. (2010). Bourgeois dignity: why economics can't explain the modern world. Chicago: University of Chicago Press.Noland, M. (2005). Religion and economic performance. World Development, 33(8), 1215–1232.North, D.C. (1990a). Institutions, institutional change and economic performance. Cambridge university press.Platteau, J.P. (2008). Religion, politics, and development: lessons from the lands of Islam. Journal of Economic Behavior & Organization, 68(2), 329–351.Plümper, T., & Troeger, V.E. (2007). Efficient estimation of time-invariant and rarely changing variables in finite sample panel analyses with unit fixed effects. Political

Analysis, 15(2), 124–139.Pryor, F.L. (2007). The economic impact of Islam on developing countries. World Development, 35(11), 1815–1835.Sawkins, J.W., Seaman, P.T., & Williams, H.C.S. (1997). Church attendance in Great Britain: an ordered logit approach. Applied Economics, 29, 125–134.Smith, A. (1776). An inquiry into the nature and causes of the wealth of nations. Indianapolis: Liberty Classics.Spolaore, E., & Wacziarg, R. (2013). How deep are the roots of economic development? Journal of Economic Literature, 51(2), 325–369.Steen, T. P. (2005). Is there an earnings premium for Catholic women? Faith and Economics, Spring: Evidence from the NLS Youth Cohort 21–39.Tomes, N. (1985). Religion and the earnings function. The American Economic Review, 245–250.Tu, Q., Bulte, E., & Tan, S. (2011). Religiosity and economic performance: micro-econometric evidence from Tibetan area. China Economic Review, 22(1), 55–63.Wald, K.D., & Calhoun-Brown, A. (2014). Religion and politics in the United States. Rowman & Littlefield.Weber, Max (1930). The protestant ethic and the spirit of capitalism. London: Allen and Unwin (Orig. pub. 1904).Woodberry, R. (2012). The missionary roots of liberal democracy. American Political Science Review, 106(2), 244–274.Zhang, W., & Ke, R. (2002). Trust in China: a cross regional analysis. Economic Research Journal, 10, 59–70.