ABSTRACT Does the luxury cars are more comfortable and creates more fun to drive? Do people who buy more expensive cars drive more? Does the sunk cost fallacy affect decision makers in repeated situations? We develop a behavioural model of usage of a durable good with mental accounting for sunk costs over some time horizon. The model predicts higher than rational usage that attenuates over time, where the rate of attenuation is faster with higher sunk costs. It nests rational behavior as a special case. We take the model to data on 8,264 cars purchased in Singapore between 2000-2013, when continuing government policy caused the sunk cost associated with buying a new car to vary substantiality.Based on structural estimation, we find robust evidence that sunk cost did affect behavior. The elasticity of driving with respect to the sunk cost is 0.050 (s .e. 0.016). An increase in the sunk cost by S$13,038 (the outcome of government policy between January 2009 and June 2013 ) would be associated with an increase in monthly driving by 90 kilometres or 5.8% in the first four years of ownership. Our results are robust to various checks including alternative explanations in terms of selection, specification of sunk cost, and salience of sunk cost. Similarly, in a dataset of Hong Kong car buyers, we find that driving was influenced by sunk costs.