UC Berkeley

In Chapter 1, I develop a New Keynesian model with inventories and convex costs of labor adjustment. For each of the three empirically observed responses to monetary policy shocks: (1) the slow adjustment of inventories compared to changes in sales; (2) the delayed and gradual response of inflation; and (3) the transitory movement in the aggregate price level in the same direction as the interest rate, also known as the "price puzzle," my model has important implications. First, adjustment costs counteract the financing-cost effect of interest rate changes on inventory holdings, but are still inadequate for the calibrated model to generate countercyclical inventory-to-sales ratios. I find that this financing-cost effect needs to be reduced by 80 percent for the model to predict inventory behavior correctly. Second, firm-specific adjustment costs in production increase the degree of real rigidity for price adjustment, so the response of inflation in the presence of high aggregate marginal costs is still slow-moving and persistent. Finally, the motive of cost smoothing for holding inventories implies that marginal costs should move in the opposite direction as the interest rate, which casts doubt on the use of the cost channel to explain the "price puzzle."

In Chapter 2, I propose a theory of the information channel between home consumption bias and home equity bias. Consumption-revealed information is acquired spontaneously in an investor's daily life and thus is naturally immobile. For this reason, consumption experience more concentrated in home-produced goods endows domestic residents with information advantage in home equities. This channel also helps to explain many empirical facts such as the 90\% correlation between import shares and foreign equity shares.

In Chapter 3, I use individual portfolio data from a China's brokerage firm to test the predictions of Chapter 2. I find that the fraction of local stocks in the brokerage portfolio is 143 percent higher than the fraction of local stocks in the market portfolio. One third of this portfolio locality is explained by business exposure of listed firms, measured by their sales per capita in the brokerage city. The result shows that a rise in sales per capita by \$2.75 leads to a 32 percent increase in the portfolio share relative to the mean. To examine whether business exposure helps investors to gain information other than familiarity, several indicators of business exposure in nonlocal areas are included in a regression. The result suggests that if a nonlocal firm's business is concentrated in other areas, local investors tend to shy away from its stock. However, some coefficients are not statistically significant, so we still cannot reach the conclusion that a firm's sales business has significant amount of information content on stock returns.