This dissertation is composed of three chapters. In the first two chapters, I study the welfare and distributional consequences of government policies in economies with credit constraints and heterogeneous agents. Given the complexity to compute such models, the third chapter reviews the basics of parallel computing in macroeconomics. In the first chapter, I quantitatively assess the welfare maximizing policy during the Great Recession, when the government had access to two policy instruments: a) offer bailouts to banks, and b) subsidize the mortgage refinancing of households. The implementation of these instruments involves a trade-off, shaped by a dead-weight loss caused by foreclosures and an information friction on house prices. The main finding is that a subsidy-only policy would have generated welfare gains of up to 0.4% in consumption equivalent terms when compared to the HAMP and TARP benchmark. In the second chapter, I study the general equilibrium effects of government-supplied student loans in the educational markets of developing countries. Our main finding suggests that subsidized student loans can lead to a widening gap in the quality of education offered by top-10 versus top-50 educational institutions. Given the complexity to compute models with heterogeneous agents, the third chapter describes the basics of parallel computing in economics, reviews widely-used implementation routines and compares performance gains using as a test bed a standard life-cycle problem.