Results from the NPTS
The National Personal Transportation Survey (NPTS) for 1995 has information on usage of vehicles and enough information to disaggregate by age, model year, and make. One can see the relationship between age and annual miles. The change in odometer readings is suspect especially because of its volatility and its negative values in some years. The median curve implies a certain amount of bunching around round numbers (multiples of 1000). However, both the mean and the median curve show a strong downward trend in miles per year as a function of age. The small deviation between the two curves suggests that any outlier problem in the mean curve is not critical.
If we disaggregate by brand, there is significant variation in curves across brands. All of the curves are downward sloping except for those with trancated panels. Thus there is promise to use some brand specific mileage/aging parameters as measures of age effects on benefits.
Vehicle ownership per household varies with family size but in somewhat surprising ways. Vehicle ownership graphs show that a) in households with at least one driver, the modal number of vehicles is 2 and b) in small households, a significant proportion of households own more vehicles than have drivers. The distribution of miles by drivers in the household increases from households with one driver to households with three but then is pretty stable. Also, the density of miles per vehicle, ranked by vehicle, disaggregated by number of vehicles in the household, shows that household mileage is shared among the first two vehicles and, to a lesser extent among the third vehicle. But other vehicles are used only marginally.
We can model the effect of car characteristics on mileage and estimate the parameters of the model.