Grossman (2001) investigated the effect of multiple liabilities of bank share holders on bank failure rates in U.S. states before the Great Depression. Grossman found that double liability did reduce bank failures in periods where bank failures were not abnormally high. However, it did not guarantee bank stability in times of widespread financial distress. Table 2 of Grossman (2001) runs an inverse logit regression of state bank failures (measured by either the failure rate or the asset failure rate) in state \(i\) at time \(t\), on its lagged value, and the failure rate among national banks in that state at time \(t\) as well as a dummy variable for multiple liabilities which takes the value one if the state had double, triple, or unlimited liability for banks for three consecutive years, and zero otherwise.

(a) Replicate the results in Table 2 of Grossman (2001) using the BankFailure stata data set. How are the standard errors and significance affected by robustifying these regressions?

(b) How are the results affected by including state and time dummies. What do you conclude?