This paper aims to test the microfoundations of consumption models and quantify the macro implications of heterogeneity in consumption behavior. We propose a new empirical method to estimate the sensitivity of consumption to permanent and transitory income shocks for different groups of households. We then apply this method to administrative data from Denmark. The large sample size, along with detailed household balance sheet information, allows us to finely divide the population along relevant dimensions. For example, we find that households who stand to lose from an interest rate hike are significantly more sensitive to income shocks than those who stand to gain. Following a 1 percentage point rate increase, we estimate consumption will decrease by 26 basis points through this interest rate exposure channel alone, making it substantially larger than the intertemporal substitution channel that dominates in representative agent New Keynesian models.