A standard real-options model predicts that time-to-build investment decisions could be delayed by uncertainty over future revenue. This paper examines the first-order importance of this mechanism by looking into the micro-data for residential construction during the recent housing boom and bust. We develop a model of sequential irreversible investment with stochastic bottlenecks and estimate the model parameters based on the distribution of time-to-build and new house prices. In the estimated model, the boom period increase in time-to-build is caused by frequent construction bottlenecks, and the decrease in house prices and the increase in uncertainty during the bust generated further delays in construction.