Time-Dependent and Time-Invariant Covariates within a Proportional Hazards Model : A Financial Distress Application

Research in the area of financial distress often uses proportional hazards models to determine the influence of covariates on the duration of time that precedes financial distress. A critical issue in the use of a proportional hazards model is the use of time-invariant and time-dependent covariates. Time-invariant covariates remain fixed while time-dependent covariates change during the estimation of the model. Although the choice of covariates might substantially affect the estimation of the proportional hazards model, existing literature does not consider the potential effect of this choice on model estimation. Using a sample of financially distressed and non-financially distressed firms to estimate proportional hazards models with time-invariant and time-dependent covariates, this paper suggests the choice of time dependence substantially influences model estimation and that covariate selection should be given more serious consideration in financial distress research