67
the same day. Their observed dates were randomly added or reduced by one day. As
a result, there are no ties.
4.3 How covariates are determined
We selected covariates related to bankruptcy and acquisition using a stepwise selection
algorithm assuming independent censoring. We set a р-value of 0.30 as the threshold
for both variable entry and stay. That choice was based on the consideration of the
potential change in р-values for covariates after accounting for dependent censoring.
Outputs are shown in Table 4.1 for bankruptcy; and in Table 4.2 for acquisition
(being acquired). A total of five covariates are selected for the model of bankruptcy
and of acquisition. Two of them are significant for both bankruptcy and acquisition.
They are size and liability. Age is a significant predictor in the model for bankruptcy;
Growth and Profit are significant predictors for acquisition. Profit, growth and size
were also used in Fama and French (2001) when studying disappearing dividends. Age
was also used in Gretchen’s Ph.D. dissertation (2004) regarding dividend initiation
policy. Size was also used in Marlett et al. (2003) when studying the impact of the
1999 GLB Act. Intuitively, growth, profitability, size, age and liability are reasonable
measures of a company’s financial status. Therefore, they can be related to companies’
survival.