Essays on identification in macroeconomics

Abstract

This thesis studies new methods to estimate the effects of interventions at the macroeconomic level using regional variation. Chapter 1 studies the relationship between elasticities at the regional level and elasticities at the macroeconomic level. To this end, I analyze regional versions of canonical macroeconomic models, along with different policies of interest, such as aggregate government spending. At the regional level, I show that there are two relevant elasticities for policy analyses, the micro-local elasticities and the micro-global elasticities. The former measure how a region reacts to a regional policy, the latter measure how a region reacts to an aggregate policy. I then show that the macro elasticity of interest is a function of the micro-global elasticities exclusively. Finally, I show that if we fix a policy and an outcome variable, the mapping from the micro-global elasticities to the macro elasticity is the same across models.

Chapter 2 proposes a new approach (Regional Structural VAR (RSVAR)) to estimate macroeconomic elasticities using regional data that avoids the problem of model-specific estimates. I first define a class of models, which includes the most widely used models for policy analysis, that gives equilibrium regional equations that contain the micro-local and micro-global elasticities. I then specify different sets of identification assumptions, along with estimators of the macroeconomic elasticities, and show that the estimators proposed are consistent. The crucial assumption underlying all of these results is that regions are heterogeneous in the sense that they react differently to the same shocks. Chapter 3 provides a new identification strategy to estimate the fiscal multiplier in the US. Using state-level data for the period 1971-2008, I apply a RSVAR approach to recover the national fiscal multiplier.

The instrument employed at the state level is the official declarations of natural disasters by the federal government. My results suggest a very precisely estimated fiscal multiplier of around 1, depending on the specification used. Thus, it is not possible to rule out the possibility that government spending crowds out/in private spending. However, the range of the estimates obtained (0.7 - 1.2) suggests that either effect should be small.