We empirically examine the distributional consequences of income-based versus place-based recycling of carbon tax revenues when automobile demand varies substantially over geographic space. Using a large household dataset from Japan, we estimate a discrete-continuous choice model that parsimoniously accounts for the geographic distribution of incomes, public transit, and portfolio preferences. The model outperforms a naive random-coefficient model in explaining the observed spatial distribution of automobile demand, and allows us to estimate the price and income elasticities that vary by income and public transit density. The estimated model is used to simulate the distributional impacts of income-based versus place-based revenue recycling on carbon emissions and consumer welfare. Our results show the following: first, the improvement in consumer welfare from rebates substantially outweighs the increase in negative externalities from the rebound in carbon emissions; second, place-based recycling outperforms income-based recycling in mitigating welfare losses for low-income and rural households, which face the greatest welfare losses from the carbon tax.