competitive advantage variable is not affected by excluding these two variables, indicating that our main results were not biased due to endogeneity.
We also explore the sensitivities to the specific sample of countries. We report two tests: including all source and host countries, even when they did not have any in- or outward investments in banking, except for off-shore financial centers; and including all source countries, but studying only developing countries as host countries. The first sample is much larger, almost 15,500 observations compared to the 8,838 observations in most other regressions. The second sample of developing host countries only is somewhat smaller, 7,076 observations. The regression results are reported in rows 10 and 11, with the first regression results for all countries and the second for the sample of developing countries only.
When we consider all countries, row 10, we find the institutional competitive advantage variable to be again statistically significant at the 1 percent level with the same negative sign and a somewhat larger coefficient. The result for the sample of only developing host countries (row 11) shows that the institutional competitive advantage variable remains significant at the one percent level, with the semi-elasticity slightly lower than the base regression.
We next explore whether results are similar when, instead of using a simple average to calculate the denominator of the institutional competitive advantage, we use a source GDP-weighted average to account for the fact that some source countries are larger, have more banks that can invest abroad, and thus are potentially more important competitors than other countries are. The result (row 12) shows that this measure of institutional competitive advantage is similar in sign and significance, but displays a lower semi-elasticity than the base regression. It suggests that large source countries have other reasons than institutional competitive advantage that make them able and willing to invest aboard.
Lastly, we explore the sensitivity of our results to the specific variable used to capture institutional quality, the KKM variable. One constraint we face is the time dimension: few other institutional environment indicators are available for the same, long time period as our banking entry data are. Also the large number of countries in our sample excludes the use of some institutional indicators. At the same time, we need