Economic theory suggests that the environmental externalities created by driving cars can be corrected by imposing taxes. In this paper we focus on one possible such tax, namely an annual registration fee that grows with the CO2 emissions rate of a vehicle, of the kind adopted in recent years in countries with important car markets, including Germany. We ask three research questions. First, how effective is this type of tax at redirecting new cars towards models with lower emissions rates? Second, are these effects stable over time? Third, when new information is issued about the true CO2 emissions rates of cars, how does the car market process that information? We use monthly new car sales figures at a very disaggregated level from Germany from Jan 2011 to March 2019, a period during which the tax, which is relatively modest, was revised three times. With the first two revisions the threshold above which a car's emissions are penalized was tightened; with the third, it is the car's emissions rate that changes, due to the adoption of a new testing procedure, the WLTP. We find that, overall, the elasticity of new car sales with respect to the tax is low (-0.06). Focusing on specific periods around the time of the tax revisions results in higher estimated tax elasticities (-0.12 to -0.31), and efforts to match cars with emissions rates just below the threshold with similar vehicles just above the thresholds even stronger elasticities (-0.122 to-0.563, depending on the period). When we separate the tax amount into the previous tax plus the shock due to the switch to the WLTP test procedure, we find that the market is insensitive to the tax shock, but processes fully the corresponding change in the fuel economy of the vehicle. Much literature has focused on whether consumers discount very heavily or neglect the future costs of owning a car, compared to the price they pay for it. We find that even within future costs, some appear to be more salient than others.