The research design largely compensates for the lack of a random sample by having people participate continuously, beginning with when they sign up.

Would I prefer to survey a random sample of vehicle owners on a monthly basis? Sure, if the response rate were high--which it wouldn't be. And even if the response rate were 100 percent, no one would pay the resulting cost of the research--such a survey would cost millions.

What I have is a method that yields surprisingly robust results, with low enough costs that I can provide the results for free to anyone who participates.