The maximum likelihood approach to model estimation (MLE) gives several advantages over ordinary least squares (OLS) regression when applied to our research question. MLE gives us the flexibility to estimate pooled models with fine control over which coefficients are identified by the entire data set versus which coefficients are event specific. For example, we estimate models that allow autocorrelation to be different in different markets, and we also estimate models where autocorrelation is treated as a pooled estimate. Another important advantage of the MLE approach is the flexible error structure. Although OLS applied to an event study would assume that the variance of the error term is fixed (the same before and the same after the event), MLE is free from this restriction and allows us to estimate models with different error variances before and after the event. Empirically, we find this feature is important.