Essays in macroeconomics and production networks

Chapter 1 provides a theory of financial frictions as a mechanism for uncertainty shocks to drive aggregate total factor productivity (TFP) fluctuations, and shows how input-output linkages amplify the impact of financial frictions on aggregate economy. Financial frictions distort the allocation of capital and induce capital wedges. These wedges differ across sectors due to different sectoral technology nature and financing demands. Sectors that have more financing needs are disproportionately vulnerable to financial frictions. This variation in distortion across sectors generates inefficiency in capital allocation and reduces aggregate measured TFP. Intermediate input shares amplify the effect of this inefficiency and generate larger TFP fluctuations.;Chapter 2 estimates and calibrates the theoretical model in Chapter 1 into US data at 14 sectors, and quantifies the amplification magnitude from input-output linkages. Financial frictions can drive aggregate TFP fluctuations and play a crucial role when uncertainty shocks hit the economy. Adding input-output linkages can further amplify the effects of both TFP and uncertainty shocks. In particular, aggregate output drops an additional 84% under TFP shocks and an additional 40% under uncertainty shocks with input-output linkages. The amplification from input-output linkages is the key for capital misallocation to contributing significantly in generating TFP fluctuations. Furthermore, there are dramatic differences in sectors' sensitivities to a tightening of borrowing constraint. Compared to other sectors, an increase in the dispersion of the return to capital in the Finance sector has the largest impact on aggregate output.;Chapter 3 settles the debate between Eeckhout (2004, 2009) and Levy (2009) and offers a simple but neglected explanation for the heavy tail observed in the city size distribution. Using the same data set and statistics reported in Eeckhout (2004), I show that U.S. city sizes are not lognormally distributed and that the upper tail indeed follows a power law. I then show that the aggregate city size distribution is a mixture of lognormal distributions, and that this mixture generates the heavy tail. Finally, I relax the independence assumption and show that city sizes are positively dependent in the data, and that the power law distribution is robust under this positive dependence.