66
reasons of companies’ elimination, which defines the bankruptcy and acquisition.
“Permanent Numbers” (PERMNO) are used to merge different data sets. Since the
missing values of covariates are relatively small, we simply deleted companies with
missing covariates.
In the WRDS data set, companies are separated as “active companies” and “in-
active companies”. In this research, following the definitions in Chapter 3 on events
and censoring, the active companies were cases of independent censoring, and inactive
companies were further categorized into events or dependent censoring.
In total, 173 companies were included in the analysis, among which 21 companies
went bankrupt, 68 are active companies, i.e. the independent censoring, and 84 were
acquired by other companies and lost their original names. The percentages are
12.14%, 39.31% and 48.55%, respectively.
As mentioned, the time horizon we considered for this research is from 1994 to
2005, which covers 6 years before the passage of the GLB Act and 6 years after.
Under the proposed method, the observed time points are sorted in ascending
order. Note that at the end of 2005, there are 68 active companies. Since their
observed time points are tied, we added a random number of days from 1 to 30 days
to each tied company to facilitate the analysis. This approach is reasonable since
all independent censoring will contribute to the denominator of the partial likelihood
function. The order of them will affect neither the parameter estimates, nor the results
of this research. In addition, there are only a few companies that went bankrupt on