This research program studies the interplay between household- and firm heterogeneity on the one hand and macroeconomic outcomes on the other hand. It does so by combining macroeconomic theory, recent developments in causal machine learning methods and novel use of micro-simulation models. The overarching purpose of the program is to address two questions: First, what type of household- and firm heterogeneity is important to understand macroeconomic outcomes? Here we take a micro-data driven machine-learning approach to identify crucial dimensions of heterogeneity. Key here is that we do not let our search for important heterogeneity be limited only to predictions from necessarily stripped-down theoretical models. Instead, our methods let high-quality Scandinavian micro data speak freely. Second, how can we design efficient economic policies that leverage these heterogeneities? The results can be used to design fiscal policies that stimulates the economy at a minimized fiscal cost. In the light of recent large-scale rescue packages, the societal gains of even small efficiency improvements in this direction would be enormous.