Instead, I use the 1984 NPD data to construct an age-of-auto/income-of-driver profile which allows production of a variable that is a proxy for the ratio of the average income of buyers of leaded gas to the average income of buyers of unleaded gasoline.