This dissertation consists of three essays on topics in macroeconomics. In the first chapter, I construct a macroeconomic model with a heterogeneous banking sector and an interbank lending market. Banks differ in their ability to transform deposits from households into loans to firms. Bank size differences emerge endogenously in the model, and in steady state, the induced bank size distribution matches two stylized facts in the data: bigger banks borrow more on the interbank lending market than smaller banks, and bigger banks are more leveraged than smaller banks. I use the model to evaluate the impact of increasing concentration in US banking on the severity of potential downturns. I find that if the banking sector in 2007 was only as concentrated as it was in 1992, GDP during the Great Recession would have declined by 40% less it did, and would have recovered twice as fast. In the second chapter, my co-author and I investigate the impact of firm capacity constraints on aggregate production and productivity when the economy is driven by aggregate and idiosyncratic demand shocks. We are motivated by three observed regularities in US GDP: business cycles are asymmetric, in that large absolute changes in output are more likely to be negative than positive; capacity and capital utilization are procyclical, and increase the procyclicality of measured productivity; the dispersion of firm productivity increases in recessions. We devise a model of demand shocks and endogenous capacity constraints that is qualitatively consistent with these observations. We then calibrate the model to aggregate utilization data using standard Bayesian techniques. Quantitatively, we find that the calibrated model also exhibits significant asymmetry in output, on the order of the regularities observed in GDP. The third chapter explores the role of distance in equilibrium selection. I consider a model economy with multiple steady state equilibria where a high productivity and a low productivity technology are available for use in production. The high productivity technology requires a fixed set up cost for production. Sectors are linked by localized production complementarities. I consider selection under a learning rule in which agents imitate their most successful neighbor. As distance between neighbors decreases, the possible profits from industrialization increase, and the likelihood that the learning rule process converges to a steady state matching the H equilibrium increases. The result suggests that, in the presence of localized technology spillovers, there may be important gains to economic growth from infrastructure development. / text