The Institutional Determinants of Bilateral Trade Patterns

• comm_religion: captures whether a pair of countries has the same dominant religion.
This indicator was constructed by determining the religion with the maximum share
of people in the total population.

• comm_VA, comm_PS, comm_GE, comm_RQ, comm_RL, comm_CC: capture
whether a pair of countries has a similar quality of governance, as measured by the
corresponding institutional proxies.

Appendix B. Various Specifications for the Basic Gravity Model

In this appendix, we describe some alternative choices of specifying the basic gravity
model of bilateral trade that we have considered.

In specifications 4 to 6 of Table 2, we investigate alternative specifications of the basic
set-up, replacing either the GDP variables by country-specific dummies, the log of distance
by distance itself, or both.

Specifications 4 and 6 have been estimated with a full set of dummies for home and
foreign countries, and without constant term. The dummy variables represent all country-
specific factors that might be relevant for their propensity to trade, either in the role as
exporter or as importing country. In this way, any influence of country-specific factors,
including GDP, on the estimated regression coefficients for the bilateral variables is
controlled for. The resulting parameter estimates for bilateral variables generally become
more precise and do not suffer from omitted variable biases

Following the introduction of country-specific dummies in specification 4, the effect
sizes for the bilateral dummy variables rise, and they become statistically more significant.
For example, the effect of language commonality becomes more pronounced and
statistically significant at the 1%-level. Also, the impact on the effect size for common
border is substantial. However, country dummies lower the coefficient (positively defined)
for log distance and lower its statistical significance. Apparently, omitted variable biases
pronounced the effect size for log distance and improved its statistical significance, in
contrast to their effect on the bilateral dummies. One possible explanation may be that
geographical distance is positively correlated with the presence of a developed and a less

Bilateral Trade Flows and Institutions

<<   21   22   23   24   25   >>

More intriguing information