The destruction of New York's World Trade Center by religious
fanatics on September 11, 2001 led many commentators to question whether
religion and globalization are compatible. The relationship is
complicated by its theoretical ambiguity: religion can both enhance and
suppress economic activity in general, and its potential network effect
can both create and divert trade. There have been few empirical studies
to shed light on the matter, however. This paper fills this void by
examining the empirical relationship between religion and international
AN INTRODUCTION TO THE AMBIGUOUS ECONOMIC EFFECTS OF RELIGION
Religion is an institution that guides general economic behavior,
and it therefore also affects the important economic activity of
international trade. Religions often promote
"economically-friendly" behavior, such as honesty, diligence,
and the provision of public goods. But, because religions focus on
spiritual issues rather than the "pursuit of happiness," they
may also suppress people's motivation to engage in
welfare-enhancing economic transactions. Religion's overall
influence on trade-enhancing institutions is, therefore, ambiguous.
(Iannaconne, 1998) observes in his survey of the literature on religion
and economic activity that "religion seems to matter, but its
impact is far from uniform." Iannaccone's survey also reveals
how sparse the research on this issue is. How religion influences the
institutions that directly affect international trade has not been
systematically examined by economists.
Religion may also have a network effect that facilitates complex
economic transactions among people in different countries.
Religion's role in creating international trade networks has been
investigated by (Greif, 1989; Grief, 1994; Rauch, 1999; Rauch, 2001;
Rauch & Trindade, 1999). The sharing of religious beliefs can
mitigate problems such as adverse selection, moral hazard, and default.
Therefore, religion can facilitate complex economic transactions among
people in different countries. These network effects of religion are not
necessarily favorable to increased international trade, however.
Networks can divert trade as well as create trade. Furthermore, networks
may hinder the long-run growth of trade by limiting the entry of new
participants and the inclusion of new products. Recent works by (Mokyr,
1990; Holmes & Schmitz, 1995; Parente & Prescott, 2000) showed
that vested interests often obstruct competition and economic change,
suggesting that networks may serve to protect certain participants from
competition from those outside their network.
TESTING THE RELATIONSHIP BETWEEN RELIGION AND TRADE USING THE
These ambiguous theoretical results signal the need for an
empirical study of religion's effects on international trade. In
order to test the institutional and network effects of religion on
trade, an augmented gravity model is applied. The gravity model normally
explains 70 percent or more of the cross section variation in world
trade volumes, and it has proven useful for examining the importance of
potential influences on trade. The model is theoretically attractive
because it can be derived from a number of traditional trade models; see
(Linnemann, 1966; Leamer & Stern, 1970; Anderson, 1979; Deardorff,
The standard gravity equation specifies trade between a pair of
countries to be a negative function of the distance between the
countries and a positive function of their combined national products.
This equation is usually augmented to account for geographic,
ethno-linguistic, and economic conditions. A common form of the gravity
(1) [tot.sub.ij] = [a.sub.0] + [a.sub.1](gd[p.sub.i]gd[p.sub.j]) +
[a.sub.2](po[p.sub.i] po[p.sub.j]) + [a.sub.3][dist.sub.ij] +
[a.sub.4][CONT.sub.ij] + [a.sub.5][LANG.sub.ij] + [a.sub.6][FTA.sub.ij]
in which [tot.sub.ij] is the log of bilateral trade between
countries i and j, gd[p.sub.i]gd[p.sub.j] is the log of GDP for i and j,
[dist.sub.ij] is the log of geographic distance between i and j,
[pop.sub.i] [pop.sub.j] is the log of the product of the populations in
country i and j, and CONT, LANG, and FTA are dummy variables for pairs
of countries that have a contiguous border, a common language, and are
members of the same active free trade area, respectively. For recent
gravity studies see, for example, (Frankel, Stein & Wei, 1995;
McCallum, 1995; Eichengreen & Irwin, 1996; Deardorff, 1998; Frankel
& Romer, 1999; Freund, 2000; Freund & Weinhold, 2000; Frankel
& Rose, 2002).
ADDING RELIGION VARIABLES TO THE GRAVITY EQUATION
For religion to substantially influence a country's
institutions, it must be a dominant religion. Minor religions adhered to
by a few people are unlikely to have much effect on a country's
overall economic institutions and, hence, its aggregate level of
international trade. A dominant religion can be defined as one that is
followed by at least 75 percent of the country's population.
Religion's network effect depends on whether people in different
countries share the same religion. Therefore, to distinguish between
religion's influence on trade through the institutional channel and
the network channel, three dummy variables are introduced into the
augmented gravity equation: DOM for each pair of countries in which one
trade partner has a dominant religion, DIFDOM when trade partners both
have dominant, but different, religions, and SAMEDOM for country pairs
in which both countries have the same dominant religion. If a dominant
religion's influence on a country's institutions has a general
effect on its ability to engage in international trade, then the DOM and
DIFDOM dummies should be significantly positive, with the latter being
greater in magnitude than DOM. If the sharing of the same dominant
religion has a positive network effect, then SAMEDOM dummy should be
In recognition of potential omitted variable bias, several
institutional and network variables are added to equation (1): a dummy
(LAWij) to capture the network effect of having a common legal structure
using data from (Djankov, La Porta, Lopez-de-Silane & Shleifer,
2002), the bi-lateral average of (Kaufmann, Kraay & Zoido
Lobaton's, 1999) government regulation variable ([burden.sub.ij])
to capture other institutional effects, and two communications channels,
[cyber.sub.i][cyber.sub.j] and [phone.sub.i][phone.sub.j], the log of
the bilateral product of top domain web hosts and telephones per
thousand in countries i and j, respectively [See data sources in the
Appendix]. This leaves the extended gravity model:
(2) [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII].
Most studies estimate the gravity equation (2) in double
logarithmic form so that the estimated coefficients can be interpreted
as elasticities. This technique omits country pairs whose bilateral
trade is zero, or about twenty percent of the observations in this
study. Omitting these observations biases the results. Therefore, the
scaled OLS (SOLS) technique first used by (Eichengreen & Irwin,
1995), in which the dependent variable is expressed as log(1 +
[TRADE.sub.ij]). (Greene, 2003) shows that the transformed variable
approximates a "semi-log Tobit relationship;" for small values
of [TRADE.sub.ij] the logarithm is close to zero, and for large values
of [TRADE.sub.ij] the logarithm of the transformed variable is close to
the logarithm of [TRADE.sub.ij]. This approach yields results that are
similar to those for a Tobit regression, and the double log form is
To check for estimation robustness, equation (2) is also estimated
using nonlinear least squares technique similar to that applied by (Coe
& Hoffmaister, 1999; Coe, Subramanian & Tamarisa, 2002). This
method essentially changes the equation into an exponential form, and by
not actually putting the data in logarithmic form we can include the
observations for which trade is zero.
Column 2 in Table 1 presents the SOLS results for the baseline
gravity model using data on the bilateral trade volumes between 84
countries during the year 1998. As is common for gravity models, nearly
80 percent of the variation in bilateral trade is explained. All signs
are as expected and significant at the 95 percent level. Nonlinear
estimates converted to be compatible with the logarithmic results of the
SOLS regressions are given in Column 5; the similarity between columns 2
and 5 adds robustness to the results.
Columns 3 and 6 in Table 1 present the results for regression
equation (2). All coefficients except DOM are significant and compatible
with the baseline gravity model regression. The coefficient for DOM is
positive but small and not statistically significant; the coefficient
for DIFDOM is significant using SOLS and equal to !0.314, which implies
that, all other things equal, two countries with different dominant
religions trade 31 percent less. Finally, the SAMEDOM coefficient tells
us that when two countries share the same dominant religion, then all
other things equal they trade 38 percent less. These results suggest
that the presence of a dominant religion has no institutional effect, or
possibly a negative institutional effect, on trade. The negative SAMEDOM
coefficient suggests that the network effects related to countries'
sharing of religious institutions cause more trade diversion than trade
THE TRADE IMPACT OF SPECIFIC RELIGIONS
The gravity model can be extended further to distinguish the impact
of specific dominant religions on trade flows. By adding dummies for
specific dominant religions (DOM) and shared dominant religions (SAME),
the regression model becomes:
(3) [tot.sub.ij] = [a.sub.0] + [a.sub.1]([gdp.sub.i][gdp.sub.j]) +
[a.sub.2]([pop.sub.i] [pop.sub.j]) + [a.sub.3][dist.sub.ij] +
[a.sub.4][CONT.sub.ij] + [a.sub.5][LANG.sub.ij] + [a.sub.6][FTA.sub.ij]
+ [a.sub.7][LAW.sub.ij] + [a.sub.8]([burden.sub.ij]) +
[a.sub.10]([phone.sub.i][phone.sub.j]) + [a.sub.11](DOMBuddhist) +
[a.sub.12](DOMCatholic) + [a.sub.13](DOMHindu) + [a.sub.14](DOMJudaism)
+ [a.sub.15](DOMMuslim) + [a.sub.16](DOMProtestant) +
[a.sub.17](SAMEBuddhist) + [a.sub.18](SAMECatholic) +
[a.sub.19](SAMEHindu) + [a.sub.20](SAMEMuslim) +
[a.sub.21](SAMEOrthodox) + [a.sub.22](SAMEProtestant) + [u.sub.ij].
Estimation results for equation (3) are found in columns 4 and 7 in
Table 1. The results indicate that when Catholic, Hindu, Judaism,
Muslim, or Protestant religions are dominant, trade is reduced. The
dominance of the Orthodox religion has an insignificantly positive
institutional effect, and only Buddhism has a significantly positive
institutional effect on trade. The network effects of individual
religions are mostly insignificant. However, when countries share the
Orthodox and Buddhist religions, trade is enhanced. Catholicism has a
negative network effect.
Is religion compatible with globalization? The evidence from the
regressions relating dominant religions and international trade suggests
that dominant religion seems to have negative effect on trade. Focusing
on individual religions reveals that most religions discourage trade. An
exception is Buddhism, whose institutional and network effects both seem
to encourage trade.
Bilateral trade data are from the International Monetary
Fund's Direction of Trade Statistics Yearbook, 2000 (IMF,
Washington, D.C.). Gross Domestic Product in millions of U.S. dollars,
population, telephone lines per 1,000 people, free trade areas, and
number of top domain internet hosts per 1,000 people are from the World
Bank's 2001 World Development Indicators (World Bank, Washington
D.C.). Distance (kilometers between capital cities) is from the U.S.
Geological Survey (ftp://kai.er.usgs.gov/pub/). Common borders, common
languages, and fraction of population claiming adherence to religions
are from the CIA World Factbook 2000,
Anderson, J.E. (1979). A theoretical foundation for the gravity
equation. American Economic Review, 69(1), 106-116.
Coe, D.T. & A.W. Hoffmaister (1999). North-south trade: Is
Africa unusual? Journal of African Economics, 8(2), 228-256.
Coe, D.T., A. Subramanian & N.T. Tamirisa (2002). The missing
globalization puzzle. IMF Working Paper, No. 171.
Deardorff, A. (1998). Determinants of bilateral trade: Does gravity
work in a classical world? in: J.A. Frankel, (Ed.) Regionalization of
the world economy. Chicago, IL: University of Chicago Press, 7-22.
Djankov, S., R. La Porta, F. Lopez-de-Silane & Andrei Shleifer
(2002). Courts: The lex mundi project. NBER Working Paper, No. 8890.
Eichengreen, B. & D.A. Irwin (1995). Trade blocs, currency
blocs and the reorientation of world trade in the 1930s. Journal of
International Economics, 38(1), 1-24.
Eichengreen, B. & D.A. Irwin (1996). The role of history in
bilateral trade flows. NBER Working Paper, No. w5565.
Frankel, J.A. & D. Romer (1999). Does trade cause growth?
American Economic Review, 89(3), 379-399.
Frankel, J.A. & A. Rose (2002). An estimate of the effect of
common currencies on trade and income. Quarterly Journal of Economics,
Frankel, J.A., E. Stein & Shang-Jin Wei (1995). Trading blocs
and the Americas. Journal of Development Economics, 47(1), 61-95.
Freund, C. (2000). Different paths to free trade: The gains from
regionalism. Quarterly Journal of Economics, 115(4), 1317-1341.
Freund, C. & D. Weinhold (2000). On the effects of the internet
on international trade. International Finance Discussion Paper 693,
Board of Governors of the Federal Reserve System.
Greene, W.H. (2003). Econometric analysis (Fifth edition). Upper
Saddle River, NJ: Prentice Hall.
Greif, A. (1989). Reputation and coalitions in medieval trade:
Evidence on the Maghribi traders. Journal of Economic History, 49(4),
Greif, A. (1994). Cultural beliefs and the organization of society:
A historical and theoretical reflection on collectivist and
individualist societies. Journal of Political Economy, 102(5), 912-950.
Holmes, T. & J. Schmitz Jr. (1995). Resistance to new
technology and trade between areas. Quarterly Review, Federal Reserve
Bank of Minneapolis
Iannaccone, L.R. (1998). Introduction to the economics of religion.
Journal of Economic Literature, 36(3), 1465-1495.
Kaufmann, D.,A. Kraay & P. Zoido Lobaton (1999). Aggregating
governance indicators. World Bank Working Paper, No. 2195.
Leamer E.E. & R.M. Stern (1970). Quantitative international
economics. Boston, MA: Allyn and Bacon Publishers.
Linnemann, H. (1966). An econometric study of international trade
flows. Amsterdam: North-Holland.
McCallum, J. (1995). National borders matter: Canada-US regional
trade patterns. American Economic Review, 85(3), 615-623.
Mokyr, J. (1990). The lever of riches. New York, NY: Oxford
University Press. Parente, S.L. & E.C. Prescott (2000). Barriers to
riches. Cambridge, MA: MIT Press.
Rauch, J.E. (1999). Networks versus markets in international trade.
Journal of International Economics, 48(1), 7-35.
Rauch, J.E. (2001). Business and social networks. Journal of
Economic Literature, 39(4), 1177-1203.
Rauch, J.E. & V. Trindade (1999). Ethnic Chinese networks in
international trade. Review of Economics and Statistics, 48(1), 7-35.
Joshua J. Lewer, West Texas A&M University
Table 1: The Gravity Model and Dominant Religions
Equation Equation Equation
(1) SOLS (2) SOLS (3) SOLS
Constant -6.695 -4.883 -4.714
(-24.62) ** (-14.99) ** (-14.10) **
[gdp.sub.i][gdp.sub.j] 0.687 0.666 0.668
(75.78) ** (30.87) ** (29.34) **
[pop.sub.i] [pop.sub.j] -0.072 -0.101 -0.111
(-5.81) ** (-4.09) ** (-4.26) **
[dist.sub.ij] -0.594 -0.689 -0.696
(-24.61) ** (-28.02) ** (-27.73) **
CON[T.sub.ij] 0.759 0.59 0.575
(6.83) ** (5.49) ** (5.32) **
LAN[G.sub.ij] 0.521 0.243 0.305
(7.72) ** (3.24) ** (4.02) **
FT[A.sub.ij] 0.475 0.293 0.282
(8.47) ** (5.29) ** (5.03) **
DO[M.sub.ij] (-5.26) **
LA[W.sub.ij] 0.365 0.354
(7.96) ** (7.61) **
[burden.sub.ij] 0.282 0.366
(4.34) ** (5.54) **
[cyber.sub.i]- 0.081 0.087
[cyber.sub.j] (8.28) ** (8.23) **
[phone.sub.i] -0.188 -0.215
[phone.sub.j] (-7.59) ** (-8.23) **
Buddhist (3.47) **
Catholic (-3.04) **
Hindu (-3.17) **
Judaism (-1.86) *
Muslim (-1.72) *
Protestant (-2.56) **
Buddhist (2.11) **
Catholic (-3.33) **
Orthodox (2.62) **
[R.sup.2] 0.783 0.800 0.800
Equation (1) Equation (2) Equation (3)
Nonlinear Nonlinear Nonlinear
Constant -7.674 -6.777 -6.703
(-18.17)** (-10.39)** (-10.56)**
[gdp.sub.i][gdp.sub.j] 0.851 0.861 0.895
(43.09) ** (13.19) ** (11.94) **
[pop.sub.i] [pop.sub.j] -0.061 -0.137 -0.178
(-1.43) (-1.49) (-1.82) *
[dist.sub.ij] -0.526 -0.596 -0.501
(-21.06) ** (-19.79) ** (-13.76) **
CON[T.sub.ij] 1.207 1.042 1.069
(6.69) ** (5.46) ** (5.48) **
LAN[G.sub.ij] 0.457 0.348 0.420
(3.92) ** (2.50) ** (2.83) **
FT[A.sub.ij] 0.473 0.362 0.344
(4.17) ** (3.17) ** (2.48) **
DO[M.sub.ij] (-2.75) **
LA[W.sub.ij] 0.467 0.451
(6.13) ** (5.08) **
[burden.sub.ij] 0.347 0.592
-1.32 (2.11) **
[cyber.sub.i]- 0.105 0.107
[cyber.sub.j] (3.14) ** (2.28) **
[phone.sub.i] -0.259 -0.328
[phone.sub.j] (-3.55) ** (-2.95) **
Buddhist (2.39) **
Catholic (-2.57) **
Hindu (-2.56) **
Protestant (-2.14) **
Orthodox (2.34) **
[R.sup.2] 0.816 0.823 0.834
** indicates significant at the 95% level, and * at the 90% level.
With 84 countries, there are 3486 data points (=84*(83/2)). Dominant
Buddist countries are Japan and Thailand, dominant Catholic countries
are Argentina, Austria, Belgium, Bolivia, Brazil, Chile, Colombia,
Costa Rica, Dominican Republic, Ecuador, El Salvador, France,
Guatemala, Honduras, Ireland, Italy, Mexico, Nicaragua, Panama,
Paraguay, Peru, Philippines, Poland, Portugal, Spain, and Venezuela,
dominant Hindu countries are India and Nepal, the dominant Judaic
country is Israel, dominant Muslim countries are Algeria, Bangladesh,
Indonesia, Iran, Mauritania, Saudi Arabia, Tunisia, and Turkey,
dominant Orthodox countries are Belarus, Georgia, Greece, Moldova, and
Ukraine, dominant Protestant countries are Denmark, Estonia, Finland,
Norway, Sweden, and the United Kingdom..
Notes: Figures in parentheses are heteroskedasticity-consistent