Further Robustness Checks Sample Clauses

Further Robustness Checks. Before concluding, we subject the main results reported in Models I to V in Table 7 to a battery of additional robustness checks. First, we revisit our a priori choices regarding our distance measures. We rerun all models employing traveling times in levels (rather than logs). Results are mostly unaffected, although (not surprisingly) the coefficients entail somewhat smaller effects. In Model III, the coefficient on Distance to Lender equals -1.1***, implying a decrease in the annual outlays for the median borrower of BEF 33 per extra minute of travel time to the lender. Again, we cannot reject the equality restriction in Model V. We also replace the ln(1+Distance to Closest Competitors) (i.e., the 25-percentile measure) by the (possibly more noisy) ln(1+Distance to the actual Closest Competitor) in all specifications discussed so far. We report the coefficients for the representative Models II, III, and V in Appendix B with the label “actual”, but do not tabulate standard errors to conserve space. As can be seen from the table, coefficients are unaffected both in sign and magnitude, though (not unexpectedly) significance levels are somewhat higher. We also replace traveling time by physical distance in all specifications and again report the three models in Appendix B, labeled “physical”. Remember that (based on our calculations) physical distance may well be less indicative as a measure of transportation costs for entrepreneurs. Nevertheless, while the coefficients decrease somewhat in size (for longer trips physical distance in kilometers exceeds traveling time in minutes), signs and significance levels are broadly unaffected. We remain concerned that technological developments and/or the location of competitors determine the choice of lender, partly driving our results rather than spatial price discrimination. Hence, we add the Starting Year of the Relationship (assuming technology progresses linearly through time) and the (national) Number of Branches to all models. Even though ln(1+Duration of Relationship) and Starting Year are by construction almost interchangeable and highly correlated (-0.91), both coefficients are significant. In Model V, the coefficients are 55.4*** and 4.5***, respectively. The coefficient on the Number of Branches is insignificant. More importantly, however, the results for the distance coefficients of interest in all discussed models remain virtually unaltered. Next, we also split the sample by contracting year. We find just...