A micro-analytic threshold model to describe the timing of household purchases of consumer durable goods is developed and tested. The model incorporates unobserved heterogeneity via a generalized gamma distribution and accounts for time varying covariates. Further, we employ estimation methods applicable or purchase data observed at periodic intervals of time. The model outperforms other competing models for predicting the timing of purchase of durable goods in terms of fit and predictive ability. In particular, this model outperforms the logit model and the diffusion model. The generalized gamma timing model predicts well the time to purchase the durable good; we show how it can be employed for micro-segmentation of households. Several research directions and applications are described.