Modeling cargo clearance duration using unshared frailty models

Abstract

Frailty models are survival models, which take into account the heterogeneity in the data by including random effects in the model fit. A frailty is a multiplicative latent effect on the hazard function, which is estimated together with the other model parameters. This frailty effect is assumed to have unit mean and finite variance, and is estimated from the data. Hence frailty models provide an alternative approach to parametric survivals models when the latter fail to account for most of the variability in the observed data. The data for this study is sourced from the Kenyan Revenue Authority. Cargo clearance duration at Kenyan borders is the time variable, while the explanatory variables include the weight of the cargo, the year of shipment, the continent of the cargo destination and the trade custom regime. An unshared frailty survival model is fitted to analyze the data by assuming a gamma frailty distribution and an exponential, Gompertz or Weibull survival distribution.