Religion and international trade: does the sharing of a religious culture facilitate the formation of trade networks?

American Journal of Economics and Sociology, The, Oct, 2007 by Joshua J. Lewer, Hendrik van den Berg

IV

Adding Religion to the Gravity Equation

To QUANTIFY THE CULTURAL INFLUENCE OF RELIGION, we construct variables that quantify the sharing of the beliefs, rituals, and rules of eight major world religions across pairs of countries. The eight religions are Buddhism, Roman Catholicism, Confucianism, Hinduism, Judaism, Islam, Orthodox Catholicism, and Protestantism. We use data on these religions' prominence in each country provided in the CIA Factbook. These data are far from perfect, given our purpose in this study. For example, the term "Orthodox Catholicism" refers to the Eastern Orthodox Churches, but some of the Orthodox churches are in communion with the See of Rome, and thus closely related to the Roman Catholic Church. And, in the case of Protestantism, the CIA data group together a very diverse set of religious organizations, traditions, and cultures. Perhaps the biggest weakness of the CIA data is that they do not measure the intensity of a religion's cultural influence. However, there is no real alternative, as Barro and McCleary (2002, 2003) also concluded. The CIA data set does cover a very large sample of countries, and it at least separates the sample into eight different religions.

We introduce the religion variables into the gravity model in two alternative forms. First, we calculated average pairwise religion network variables, which are simply the average percentages of the populations in a pair of countries that follow each of the eight major religions. For example, the average pairwise Buddhism variable for countries i and j would be ([Buddhist.sub.i] [Buddhist.sub.j]/2, where the variable [Buddhist.sub.i] is the percentage of people who are Buddhist in country i. Alternatively, we calculated multiplicative pairwise religion network variables, which in the case of Buddhism are the product of the shares of Buddhist in the populations of a pair of countries, or ([Buddhist.sub.i] * [Buddist.sub.j]). The multiplicative pairwise variable has an interesting property in that it represents the chance that two people from different countries who meet randomly share the same religious beliefs. Average pairwise variables are generally larger than their corresponding multiplicative pairwise variables, which implies that the average pairwise religious network variable overstates the possibility of two people of the same religion meeting randomly. The average pairwise variable may be the more accurate measure of the potential for trade networks, however, because people who share religious beliefs are likely to seek each other out, especially in the case where foreign trade partners are being sought. To enhance the robustness of our results, we run alternative regressions using each of the pairwise measures.

In order to accurately capture the network effects of religion on international trade, it is necessary to isolate religion's network effects from religion's general institutional effects. It is also necessary to isolate religion's network effects from other potential sources of network effects, such as improved communications and common languages. The standard form of the gravity model given by Equation (1) already controls for some institutional and network effects, such as the effects of a common language or legal heritage. In order to ensure further against omitted variable bias, we add several other institutional and network variables to Equation (1). A dummy ([law.sub.ij]) captures the network effect of sharing a common legal tradition. The product of top domain web hosts in countries i and j ([cyber.sub.i][cyber.sub.j]) captures communications network effects. We also add a government regulation variable ([burden.sub.ij]), an index of how well intellectual property rights are protected ([intel.sub.ij]), and the weighted average of tariff levels ([tariff.sub.ij]) for each pair of countries to capture other institutional characteristics.