UC Berkeley

This dissertation investigates the role of trade and trade frictions in shaping the internal structure of economies over time. The first chapter investigates how trade costs in generating the spatial distribution of wages and employment across regions, a classic question in economic geography. It make several contributions to the extensive theoretical and empirical literature on this question. First, building on the recent literature I show that for a wide class of economic geography models the positive implications of changes in trade costs are entirely captured by two reduced form elasticities: the elasticities of wages and employment with respect to market access. Second, I develop a novel instrumental variable approach to consistently estimating these elasticities from changes in observed wages and employment using exogenous changes in the incomes of each location's trading partners. I implement this approach using data on U.S. MSAs between 1990 and 2007 and find that wages and employment are quite sensitive to differences in market access due to trade costs. Counterfactual simulations indicate that eliminating trade costs would result in large shifts in employment from the Northeast towards the South and West and a flattening of the city size distribution. More modest reductions in trade costs result in qualitatively similar outcomes that remain quantitatively large.

The second chapter investigates how trade in intermediate inputs across industries varies with the level of development, and how this variation is related to the cross-country variation in productivity. We know that specialization is a powerful source of productivity gains, but how production networks at the industry level are related to aggregate productivity in the data is an open question. This chapter constructs a database of input-output tables covering a broad spectrum of countries and times, develop a theoretical framework to derive an econometric specification, and document a strong and robust relationship between the strength of industry linkages and aggregate productivity. We then calibrate a multisector neoclassical model and use alternative identification assumptions to extract an industry-level measure of distortions in intermediate input choices. We compute the aggregate losses from these distortions for each country in our sample and find that they are quantitatively consistent with the relationship between industry linkages and aggregate productivity in the data. Our estimates imply that the TFP gains from eliminating these distortions are modest but significant, averaging roughly 10\% for middle and low income countries.

The third chapter brings these two themes together to explore how trade costs across industries and space shape the spatial distribution of industries. The motivation and specific context is the decline of the U.S. manufacturing belt over the post-war period and the spread of industrial production to the South and West. To study the causes of this geographic dispersion of industry, this chapter first develops a multi-industry model with many locations, local external economies and input-output relationships across industries. The second contribution is to develop an estimation strategy for the parameters, including the strength of local Marshallian externalities and the size of trade costs, that does not rely on the availability of comprehensive internal trade data. I then apply this strategy to data on U.S. industry location across cities between 1970 and 1995. I find that trade costs have declined substantially over this time period, and that local external economies are on average quite strong at the industry level. These findings together suggest that only modest productivity convergence together with the decline in trade costs are sufficient to explain the decline of the manufacturing belt.