Essays on the Interaction of Demand Shocks with Geography

Norris, Jordan James

doi:https://doi.org/10.21985/n2-7bm5-1d77

Work

Essays on the Interaction of Demand Shocks with Geography

Public

Download PDF

Download All Files (.zip)

The world is an increasingly interdependent place. Regions of economic activity are deeply connected with each other through the movement of goods, labor and ideas. This has deep consequence for both the economist and policymaker: in the evaluation of economic activity, an exogenous shock in any given region does not only affect outcomes in that region, but, in principle, in all regions. Correctly accounting for these spillover effects is necessary for both valid identification of local effects and measurement of the aggregate impact of a shock. Furthermore, beyond measurement, such spatial aspects generate rich mechanisms that are absent in non-spatial (single-location) models. In my thesis, I push the frontier of this in the context of regional demand shocks. This all comes down to a single equation, that is recurrent throughout my thesis Y_i = sum(j=1,...,N) beta_ij X_j + eps_i, for i = 1,...,N where i indexes location, N is the number of locations, Y is the outcome of interest in the location, X is a demand shock in the location, eps is the error term, and the matrix beta_ij describes the spatial landscape of interactions: the effect of an exogenous demand shock in location j on the outcome in location i. It is beta which is the object of interest and my thesis is devoted to learning about this. In chapter 1, Fiscal Multipliers in Integrated Local Labor Markets, I allow for spatial heterogeneity and interdependence in the analysis of fiscal multipliers. That is, unlike the literature, not constraining beta to be proportional to the identity matrix. I conduct analysis at the state-level in the US, the Y variable is GDP, and the X variable is US Federal Government defense procurement. I parameterize beta to be a function of the network of interstate trade flows, along with unobservable elasticities which I estimate, and investigate how national fiscal multipliers depend on the internal geography of trade. In the literature, there is wide recognition that fiscal multipliers are greater in times of recession than in normal times; that is, the presence of a state-dependent fiscal multiplier. My findings suggest a meaningful analogy of geography-dependent fiscal multipliers, with the expectation that the fiscal multiplier is greater when the spending is concentrated geographically, in regions that are smaller and more-closed to trade. The intuition is that the multiplication mechanism underlying fiscal multipliers exhibits convexity; therefore concentrating the spending geographically achieves greater returns by exploiting this. In chapter 2, Disproportionate Gains: A Home Market Effect in an Almost Arbitrary Geography, I revisit a canonical mechanism in the International Trade literature dating back to the 1980s: The Home Market Effect. This theory models a mechanism for trade deriving from demand: in the presence of increasing returns to scale in production and transportation costs, countries with greater demand for an industry become a net-exporter in that industry. However, to this day, the theory only gives sharp predictions in two-location models. By making one additional assumption relative to the canon --- the matrix of interregional iceberg trade costs is positive semi-definite --- I'm able to prove that the home market effect remains on average in an arbitrary number of locations. Intuitively, the result remains only on average, and not bilaterally for all countries, because when there are more than two countries, some countries are a big market for the industry, relative to other countries. A shift in demand from these large markets can indirectly harm the country receiving the demand because it at the same time reduces the potential market size. This attenuates the gains from increasing demand in a location. In chapter 3, Random Coefficients: Identification with Spatial Spillovers, I dispense with structural assumptions and consider how far one can go in recovering own-treatment effects in the presence of spillovers. The challenge is a parameter problem: there are N observations, yet N squared coefficients: the effect on the outcome in i from a shock in j for all ij pairs. When the interest is only on own-treatment effects, I show that under a statistical assumption --- ij spillovers are uncorrelated with the product of i and j shocks --- I am able to get identification. This is appealing because increasing geographic disaggregation in a research design increases the sample size, N. Yet, the unfortunate inevitability is that this also increases the severity of spillovers as the distance between observations decreases (transitioning from states to counties to blocks, for example). However, I show that nonetheless we are still able to recover the object of interest --- the average own-treatment effect --- despite these spillovers.

Creator