An accurate estimate of vehicle miles traveled (VMT) is an essential input into a variety of emissions, climate change, energy and infrastructure-planning models. We examine different measures of annual miles driven from the 2017 National Household Travel Survey (NHTS), which includes a self-reported and an imputed measure of annual VMT for a nationally representative sample of U.S. households. We compute two additional estimates from the odometer reading, and the trip diary information contained in the NHTS, respectively. We compare these four measures of annual VMT, seek to identify low/high mileage segments of the fleet, and examine the quality of the data and their internal validity. The price elasticity of VMT (−0.275 to −0.20) is comparable across these four measures. Quantile regressions indicate that the households with higher driving intensity are less responsive to the fuel cost per mile. As cars get older, the difference between the four VMT measures tends to increase; this implies large discrepancies in the CO₂ emissions that would be saved, for example, when older cars are removed from the fleet, due to natural turnover or when “cash for clunkers” programs are implemented.