Abstract: We study information acquisition as a motive for trade in the market for municipal bonds, a primary financing source for local governments. The market is decentralized and regional dealers intermediate trades among retail investors. Dealers’ pricing behavior suggests that they learn about demand by trading. We propose and estimate a dynamic model of OTC trading that accounts for dealers’ learning and reveals that they are willing to pay 12% of the intermediation spread for the information acquired through trade. Dealers’ learning incentives benefit investors and issuers, as they strengthen trading in an illiquid market, and interact with policies on post-trade disclosure.
Abstract: In this paper we explore efficiency and optimal policy in decentralized transportation markets that suffer from search frictions, such as taxicabs, trucks and bulk shipping. We illustrate the impact of two externalities: the well-known thin/thick market externalities and what we call pooling externalities. We characterize analytically the conditions for efficiency, show how they translate into efficient pricing rules, as well as derive the optimal taxes for the case where planner is not able to set prices. We use our theoretical results to explore welfare loss and optimal policy in dry bulk shipping. We find that the constrained efficient allocation achieves 6% welfare gains, while the first-best allocation corresponding to the frictionless world, achieves 14% welfare gains. This suggests that policy can achieve substantial gains even if it does not alleviate search frictions, e.g. through a centralizing platform. Finally, simple policies designed to mimic the optimal taxes perform well.
Abstract: In this paper we investigate the importance of fuel oil costs in determining world trade. We use detailed data on ship movements across the globe and transaction-level shipping prices, along with a dynamic model describing the world shipping industry, to measure the elasticity of trade with respect to ship fuel costs. We find that the average estimated elasticity is 0.35, but ranges from 0.1 to about 1.2 depending on the level of the fuel cost. The pass-through of fuel costs to exporters is low, at 0.17. Strikingly, the trade elasticity features a pronounced asymmetry in low vs. high oil prices. As fuel costs decline, the elasticity plateaus. This flattening out of the elasticity is attributed to the equilibrium of the transportation sector and the changes in the relative bargaining positions of ships and exporters in particular. Finally, we use the estimated elasticity to assess the importance of ship design on trade flows: if the large fuel efficiency gains achieved in the 1980s had not been realized, trade would be 12% lower today.
Abstract: In this paper we study the role of the transportation sector in world trade. We build a spatial model that centers on the interaction of the market for (oceanic) transportation services and the market for world trade in goods. The model delivers equilibrium trade flows, as well as equilibrium trade costs (shipping prices). Using detailed data on vessel movements and shipping prices, we document novel facts about shipping patterns; we then flexibly estimate our model. We use this setup to demonstrate that the transportation sector (i) attenuates differences in the comparative advantage across countries; (ii) generates network effects in trade costs; and (iii) dampens the impact of shocks on trade flows. These three mechanisms reveal a new role for geography in international trade that was previously concealed by the frequently-used assumption of exogenous trade costs. Finally, we illustrate how our setup can be used for policy analysis by evaluating the impact of future and existing infrastructure projects (e.g. Northwest Passage, Panama Canal).
Media Coverage: Quartz
Abstract: We provide a guide to estimating matching functions in a spatial context. Several interactions in space take place in a decentralized fashion, such as passengers searching for taxis, ships meeting cargo, exporters meeting importers etc. A convenient modeling device to capture these meetings is the matching function, which has been used extensively in labor market settings. However in the spatial context, data availability is often limited to only one side of the market; for instance it is usually hard to find data on the number of passengers searching for a taxi. We discuss an approach to estimating matching functions that allows the researcher to recover the unobserved side of the market with relatively few assumptions. In addition, our approach obtains the matching function non-parametrically, allowing for significantly more flexibility than is commonly assumed. This additional flexibility can be key when deriving welfare and policy implications.
Joint work with Karam Kang.
Abstract: The availability of a variety of customized products is in general beneficial to consumers. However, these benefits can be eroded in the presence of search frictions, making it difficult to evaluate and compare alternatives. Our paper studies this trade-off in the US municipal bond market, a 4 trillion market on which local government depend to finance investment in infrastructure. When a bond is issued via negotiation between a government and an underwriter, the latter has an incentive to affect search frictions by distorting the choice of bond characteristics. This distortion can be facilitated by government officials’ private interests, which may be aligned with those of the underwriter. We find that state regulations limiting conflicts of interests tend to decrease in the number of uncommon bond characteristics. We then exploit variation in these regulations to show that uncommon bond features increase the underwriter’s market share in bond trades. Motivated by these findings, we build, identify, and estimate a model of bond issuance and trades to quantify the extent of inefficiency in equilibrium and to discuss policy implications.
Abstract: Censoring plays an important role in the estimation of duration models. Most estimation procedures used to estimate such models assume that the censoring times are independent of the duration of interest (perhaps independent conditionally on covariates). Unfortunately, this assumption is fundamentally not testable in the typical setup. In this, paper we point out that many data sets contain additional information on the censoring times that is not assumed in the typical setup. We then use this additional information to propose tests for independence between the censoring times and the duration of interest. We illustrate the ideas using retirement data from the Health and Retirement Study (HRS).