Here, we use the same long-term dummy specification as in Table 6, which presents the results for a long-term dummy that takes the value of 1 in the fourth year following the opening of the POH. We also estimated µ for time = ln(t) and µ for time = t. In all cases, the results are robust to the results
presented in Tables 8 and 9. As the specification for time = 1 for Year 4 has the highest overall likelihood value, we use these as the presented results. As a further robustness check, we estimated Model 1 (no lagged dependent variable) for each of the 21 markets; the results are remarkably robust to the Model 2 estimates, although with lower likelihood values.