This paper investigates the impact of urban form on households' fuel consumption and car emissions in France. We analyze more particularly three features of cities commonly referred to as the `three D's' (Cervero and Kockelman, 1997): Density, Design and an innovative measure of Diversity. Individual data allow us to circumvent selection issues, as some households may live in a location consonant to their socioeconomic characteristics or travel predispositions, while instrumental variables help control for other endogeneity issues. The results suggest that, by choosing to live at the fringe of a metropolitan area instead of its city-center, our mean-sample household would bear an extra-consumption of approximatively six fuel tanks per year. More generally, doubling residential Density would result in an annual saving of approximatively two tanks per household, a gain that would be much larger if compaction were coupled with better Design (stronger jobs centralization, improved rail-routes or buses transiting to job centers and reduced pressure for road construction), and more Diversity (continuous morphology of the built-up environment). Another important finding is that the relationship between metropolitan population and car emissions is not linear but bell-shaped in France, contrary to the US, which suggests that small cities do compensate lack of Density by either a better Design or more Diversity.