The authors have an interesting take on the term 'user fee', suggesting that gas taxes aren't really user fees because (a) they were sometimes used for deficit reduction, (b) they are shared with other surface transportation (transit), and (c) they don't correspond with use. While I don't like either diversion, that doesn't mean that gas taxes aren't user fees, just that Congress can't avoid meddling. Just because gas taxes imperfectly measure use (i.e. it is proportionate to gasoline consumption instead of miles, it is assessed on travel on all facilities, not just highways), doesn't mean it is not highly correlated. It is a surrogate, as are most fees. They are charged only to users of motor vehicles (admittedly only those users who use fuel, but that is approximately all users at this stage of technology). It would be better if user fees (preferably tolls if transactions costs could be reduced, but gas taxes in the interim) covered all costs of operating and maintaining existing streets, roads and highways, so we could depoliticize the issue, and treat it like the public utility it is. It would be better if the charge could vary by location and time of day, it will eventually do so.
It is worth reading his whole post. I will add that "do roads pay for themselves" is the wrong question to ask. Roads are not some cheap date who sticks you with the tab at the end of the night. Everybody gets some value from roads regardless of if they drive. A better question is "do road users pay their way?" Mark Delucchi looked at that question a couple of years ago (David Levinson has done a lot of work on the full cost of transportation modes as well) and Delucchi argues that the per mile cost of driving is about $.20-$.70 per gallon of motor fuel below the full cost, but most of that subsidy is parking, not highways.