The purpose of this part of my web-site is to acquaint visitors with the broad themes underlying my research, and describe my unpublished work. My research is concentrated in two broad areas.

Understanding Economic Mechanisms

The first area attempts to understand economic mechanisms. So far I have looked at two markets, the market for government securities, and the market for timber in Southern Ontario, with the focus being on the former since the last three years. This area of research has taken me into several fields, notable of which are industrial organization and market microstructure.

Market for Government Securities  

The basic problem that the issuer of government securities (the central bank or the debt management office) has to solve is how to sell a fixed amount of government securities, at a price comparable to the expected resale price of the securities. By looking at selling mechanisms for a wide variety of government securities across several countries, I observe that this basic problem amounts to three decisions that an issuer has to make when issuing government securities. First, whether issuance of government securities should be done through the primary dealer model or not. Primary dealers are intermediaries selected by the issuer. An obligation to participate in the primary issuance, market-making obligation in the secondary market, and reporting obligations are imposed on the primary dealers. In return the issuer awards them advantages. The advantages could be awarded in the primary issuance process itself, like exclusive or preferred access to the primary issuance; alternatively, they could be awarded in other areas in which the central bank interacts with the primary dealers, like payments and settlement system and monetary policy implementation. Second, whether government securities should be sold through auctions or posted prices. Third, if auctions are used, whether discriminatory or uniform price rule should be used.

While there has been extensive theoretical and empirical work on the last issue, there is little work on other aspects of issuance of government securities. I am now writing several papers to understand the various practices being used for issuing sovereign debt. For example, why does the issuer use intermediaries to issue government securities in some countries but not in others? What is the value to the issuer of restricting access to primary issuance to a limited number of wholesale buyers? Why are real return bond auctions always organized as uniform price auctions? What do the wholesale buyers gain from the issuer selling securities through an auction and subsequently buying-back off-the-run securities of similar duration? Why are posted prices used for issuing some government securities while auctions for others? A brief synopsis of the two papers on which I am currently working is given at the end.

A related area that is being explored in the government securities market is the source of demand for government securities, and the extent to which a central bank influences this demand through its role in the payments and settlement system and monetary policy implementation procedures. The ability of the central bank to influence demand follows from the fact that government securities are being used extensively as collateral for the payments and settlement system and monetary policy implementation operations. Finally, this research will attempt to provide a causal link between the debt issuance mechanism and the sources of demand for government securities.

Timber Market in Southern Ontario   

The county of Simcoe conducts up to three sales each year to transfer the harvesting rights to the standing timber on the woodlots that it owns, with each sale involving the auction of multiple woodlots through a first-price, sealed-bid auction for each woodlot. The key design issue for the county is how it should aggregate trees marked for harvesting into woodlots, and whether it should allow resale amongst the bidders subsequent to the auction or not. Currently the county sells both homogenous and heterogeneous woodlots in terms of tree diameter, and does not allow resale. In response to this, I have documented in a paper, that over time there has been tacit coordination of bidding strategies by the three nonfringe sawmills in that they avoid competing with each other on homogenous woodlots, and are able to predict perfectly nonfringe competition on heterogeneous woodlots. Tacit coordination by the nonfringe sawmills is the result of frequent participation, which enables a nonfringe sawmill to learn about the type of woodlots on which other nonfringe sawmills bid.

A question that will be explored in future work is why the sawmills coordinate their bidding strategies. The answer to this question will be based on Gary Becker's work on positive assortive matching in the marriage market. While positive assortive matching of the sawmills with the woodlots type being sold by the county should maximize the joint surplus of the county and the sawmills, the interesting question is what is the consequence of this matching for the division of this surplus between the sawmills and the county. Alternatively, how would the division of the surplus change if the county aggregated trees to form woodlots differently, and/or allowed resale between the winning sawmills subsequent to the auction. For example, while homogenous woodlots with no resale maximize the joint surplus of the county and the nonfringe sawmills due to positive assortive matching, it is unlikely that this maximizes the county's expected revenue as nonfringe sawmills avoid competing with each other on homogenous woodlots.

Objective Prior Inference

The second area of my research is concentrated on objective prior specification in interesting economic models. So far I have looked at the Roy model, and two nonregular models: structural, sealed-bid auctions, and job search model . The motivation for this research comes from the difficulty in specifying subjective priors in the scenario where parameters cannot be easily interpreted as the limiting form of some observables. An example is structural auction models with substantial heterogeneity across auctions and bidders. An alternative in this instance is a ``rule'' for constructing a prior. Bernardo's (1979) reference prior is one such rule-based prior. The basic idea is to obtain a prior that maximizes the expected information from data. In the papers I have worked on, these ``rules'' for determining priors are judged on the basis of how they worked in specific models that I study. This is in the spirit of the approach taken by Laplace and Jeffreys, and continued by several researchers in the statistics literature, including Berger and Bernardo. For example, Berger and Bernardo (1992) document several modifications and refinements made by other researchers to Bernardo's (1979) reference prior idea on the basis of how it performed in numerous applications in statistics. 1