"The Aggregate Impact of Online Retail", US Census Bureau, Center for Economic Studies Working Paper, 14-23, 2014 [PDF]
To study the impact of online retail on aggregate welfare, I use a spatial model to calculate a new measure of store level retail productivity and each store's equilibrium response to increased competitive pressure from online retailers. The model is estimated on confidential store-level data spanning the universe of US retail stores, detailed local-level demographic data and shortest-route data between locations. From counterfactual exercises mimicking improvements in shipping and increased internet access, I estimate that improvements in online retail increased aggregate welfare from retail activities by 13.4 per cent. Roughly two-thirds of the increase can be attributed to welfare improvements holding fixed market shares, with the remainder due to reallocation. Surprisingly, 8.2 percent of firms actually benefit as they absorb market share from closed stores. Finally, I estimate that the proposed Marketplace Fairness Act would claw back roughly one-third of sales that would otherwise have gone to online retailers between 2007-12.
Product Market Frictions, Selection and Allocative Efficiency (with Cagliarini and Robinson), Journal of Macroeconomics, 33(1), 2011 [PDF]
This paper attempts to reconcile the high estimates of price stickiness from macroeconomic estimates of a New-Keynesian Phillips curve (NKPC) with the lower values obtained from surveys of firms’ pricing behaviour. This microeconomic evidence also suggests that the frequency with which firms adjust their prices varies across sectors. Building on the insights of Carvalho (2006), we present Monte Carlo evidence that suggests that in the presence of this heterogeneity estimates of the NKPC obtained using conventional methods, such as GMM, are likely to considerably overstate the degree of aggregate price stickiness. Furthermore, if roundabout production is a characteristic of the economy the NKPC will falsely suggest that a sizeable fraction of prices are indexed to past inflation. These problems arise because of a type of misspecification and a lack of suitable instruments.
Product Market Frictions, Selection and Allocative Efficiency [PDF]
In models of heterogeneous firms with endogenous exit, selection effects typically arise from increases in input prices. For example, entry of relatively productive firms increases marginal products bidding up input prices and rendering less productivity firms unprofitable leading to their exit. I show that evidence of selection exists in establishment level data, yet crucially, increases in input prices are not the driving mechanism. To reconcile these facts, I present a model where establishments set prices to attract customers, who satisfice in forming relationships with establishments in the presence of search frictions. The extent of these search frictions is a new margin that affects selection. As search becomes less random and more directed, customers are less willing to satisfice, improving allocative efficiency and inducing exit of slower growing firms without corresponding increases in input prices. When search frictions in product markets are increased to match establishment dynamics in Chile, output falls by roughly 14 per cent relative to the model calibrated to the US, reflecting decreased allocative efficiency.
Dispersion in Beliefs and Price Setting [PDF]
The recent literature on price setting has largely focused on the mechanisms that can generate imperfect information across agents and whether they are capable of generating large monetary non-neutralities. In this paper, the dispersion in information across agents is taken from the data and used to discipline a standard menu cost model augmented with information frictions. In the calibrated model, imperfect information has a negligible effect as the selection effect dominates. Real responses to a monetary shock are small and transient relative to most of the literature precisely because the few firms changing prices are those who are furthest from their optimal price, as in Golosov and Lucas (2007). The result still holds even when the level of dispersion is set to that of the maximal levels of dispersion in survey data.