Accurate assessments of construction materials stocked in the built environment have received increased attention in the Industrial Ecology literature over the past few years. Many recent models that estimate building material inflows, stock accumulation and end-of-life waste, however, rely on simplistic assumptions about the lifespan of built infrastructure. While several probability distributions have been proposed (normal, Weibull, log-normal, and so on) there is no agreement on which model is best suited for modelling the accumulation of building material stock at urban and national levels. In this study we introduce an analysis of the hazard rate of buildings and discuss alternative distribution functions to model lifespan, testing the fit of five commonly used distributions to real data from the cities of Nagoya (Japan), Wakayama (Japan), and Salford (UK). The results highlight how cities with fast replacement rates are overall best modelled by right-skewed distributions, but single cohort levels express independent behaviours based on their characteristics. We investigate the sensitivity of a top-down stock accumulation model to the choice of different distributions and input parameters uncertainties. The results show that different lifespan distribution functions result in very similar overall stock accumulation at the national level, but have large impacts on calculated demolition waste flows. Differences are more pronounced for cities and the choice of a certain distribution will significantly affect the calculation of the average lifetime. Our results suggest that top-down national material stock accounts have high reliability, and are only weakly affected by the choice of one distribution over another. For cities, it is beneficial to use a distribution based on the characteristics of the buildings analysed, with regard to density and building characteristics. Stock accumulation research would profit from future bottom-up research into building lifespans to validate top-down estimation procedures. Crown Copyright (C) 2017 Published by Elsevier B.V. All rights reserved.