This dissertation is composed of two essays that examine how financial factors affect real economic activities. The first investigates the role of agency problems in financial relationships in an economy with uncertainty shocks. The second examines how much misdirected bank lending accounts for the observed inefficient labor reallocation in the manufacturing sector in Japan's 1990s. A common goal of these essays is to shed light on economic mechanisms that may lower aggregate output and employment in financial crises.The first essay explores a new mechanism through which an increase in uncertainty interacts with agency problems to destroy matches between principals and agents. This is in contrast with inactive waiting predicted by real options theory. To demonstrate the mechanism, I develop a continuous-time dynamic matching model between principals and agents with long-term optimal contracts. In the model, a principal and an agent form a team to run risky projects and adjust their level of risk-taking. Because the agent can divert project payoffs, this agency relationship becomes hard to maintain when higher uncertainty increases the variance of the risky projects and generates more fluctuations in project outcomes. Teams close to termination break up instantaneously when uncertainty increases, causing an immediate reduction in aggregate output since it takes time to set up new teams. In addition, the average level of risk-taking declines among remaining teams because teams with a history of low output take less risk in response to the exogenous increase in project riskiness, in order to reduce the probability of costly separation. This further reduces aggregate output because low-risk projects have low average returns.
In the second essay, my co-authors and I investigate the efficiency of resource reallocation in Japan during the 1990s, a decade of economic recession, by measuring aggregate productivity growth (APG) using a plant-level data set of manufacturers from 1981-2000. We find that resource reallocation contributed negatively to APG, mainly due to inefficient labor reallocation. A possible reason for the inefficient labor reallocation is misdirected (or zombie) bank lending to failing plants. To quantify its impact, we develop a model with plant-level heterogeneity, calibrate it based on the results of plant-level productivity estimation, and conduct a counterfactual exercise, in which there is no zombie lending. We find that, without zombie lending, aggregate productivity would have grown by 1.6% more during the 1990s, mostly due to more efficient labor reallocation.