In addition to the three hypotheses considered above, other explanations fall into the general class of "Type I" error. For example, one could argue that investors in 1990 had rational expectations about the expected costs and benefits of takeover defenses, where the expected costs are more severe agency problems and the expected benefits are higher takeover premiums. Then, when the hostile takeover market largely evaporated in the early 1990s?perhaps because of macroeconomic conditions unrelated to takeover defenses-Dictatorship firms were left with the costs but none of the benefits of their defenses. Over the subsequent decade, the expected takeover premiums eroded as investors gradually learned about the weak takeover market. Simple calculations suggest that this explanation cannot be that important. Suppose that in 1990 the expected takeover probability for Dictatorship firms was 30 percent, and the expected take
over premium conditional on takeover was also 30 percent. Further suppose that both of these numbers were zero for Democracy firms. Then, the unconditional expected takeover premium for Dictatorship firms would have been only 9 percent, which is approximately the relative underperformance of these firms for only a single year. In sum, we find some evidence in support of Hypothesis I and no evidence in support of Hypothesis II. For Hypothesis III we find that industry classification can explain somewhere between
one-sixth and one-third of the benchmark abnormal returns, but we do not find any other observable characteristic that explains the remaining abnormal return. The sub index regressions, which might be helpful in distinguishing between Hypotheses I and III,
are not powerful enough for strong inference. We conclude that the remaining performance differences, which are economically large, were either directly caused by governance provisions (Hypothesis I), or were related to unobservable or difficult-to-measure characteristics correlated with governance provisions (Hypothesis III).