On the Predictability of Trade and Growth

Abstract

This thesis is composed of three chapters, which are tied together by their focus on discussing factors that affect our ability to predict changes in international trade and economic growth. The first chapter theoretically and quantitatively evaluates the hypothesis that, due to the existence of large firms (granularity), idiosyncratic shocks to individual firms can lead to significant variation in the growth of countries. I embed granularity, through finiteness in the set of firms, in a general equilibrium environment featuring monopolistic competition, growth, and international trade. Firm productivities grow according to idiosyncratic productivity shocks, which obey a Gibrat's law proportional growth process, and are the only source of growth in the model. I derive an approximate analytic mapping for the standard deviation of GDP growth in this framework, which is non-zero due to granularity. This mapping depends on only a few key parameters, which I estimate for a wide-range of countries using firm-level micro data. My results indicate that idiosyncratic shocks to firms can play a significant role in generating both short-run macroeconomic fluctuations and variation in longer-run growth trends, particularly for countries that engage heavily in international trade. Empirically, I show that the model does well in matching relative differences in GDP volatility across OECD countries. The second chapter discusses the granular hypothesis and the importance of Pareto tails in generating aggregate uncertainty and instability due violations of the Central Limit Theorem. I argue that the importance of Pareto tails has been significantly overstated and that significant aggregate uncertainty can arise even in the cases where the Central Limit Theorem holds. I revisit the debate on the distribution of firm sizes and show that, when appropriate statistical methods are used, there is significant variation across countries in whether the distribution of firm sizes follows a Pareto distribution or a lognormal distribution. Despite this variation, I show that these differences are largely irrelevant in determining how much aggregate uncertainty we can expect to arise due to granularity, indicating that the presence of Pareto tails is largely irrelevant and that the pathways through which microeconomic heterogeneity can lead to aggregate uncertainty and instability are more general than previously thought. The third chapter, joint work with Timothy J. Kehoe and Kim J. Ruhl, develops a methodology for predicting the impact of trade liberalization on exports by industry (3-digit ISIC) based on the pre-liberalization distribution of exports by product (5-digit SITC). We evaluate the ability of our methodology to account for the industry-level variation in export growth by using our model to predict � the growth in industry trade from the North American Free Trade Agreement (NAFTA). We show that our method performs significantly better than the applied general equilibrium models originally used for the policy evaluation of NAFTA. We find that the most important products in our analysis are not the ones with zero pre-liberalization trade, but those with positive, yet small amounts of pre-liberalization trade.