ECONOMETRIC ESTIMATION OF PROPORTIONAL HAZARD MODELS

When analyzing duration data, covariates are typically assumed to modi fy hazard rates through the use of the proportional hazard model, in w hich the baseline hazard is multiplied by some function of the regress ors and associated parameters, g(.), to determine individual hazard ra tes. The specification for g(.) which is in virtually universal use, g (.) = exp(.), may be overly restrictive. This paper considers one alte rnative-setting g(.) equal to the cumulative density function of the s tandard normal distribution. The relative performance of this function al form for g(.), compared to the exponential transformation, is shown to depend on the parametrization of the baseline hazard rate.