John Rust - US grants

This award is for continued research on the development and application of computational methods for estimating microeconomic decision models. Under the previous grant the investigator has made significant contributions to the development of computational methods and has produced several useful applications. He has also produced several computer programs which have proved valuable to other researchers working on related issues. The project will pursue a number of different, but related topics. Among these are: (1) Empirical applications and software implementation of estimation algorithms for dynamic choice and equilibrium models; (2) Extending the algorithms to handle continuous choice variables; (3) The development of algorithms for estimating general equilibrium models. This research is important because it will contribute to a deeper understanding of microeconomic decision process, as well as pointing toward new directions for estimating general equilibrium models.

This project conducts a series of human double auction market experiments to generate a matching data set for a tournament of "robot traders" who compete in an artificial double auction market. The double auction market is a simplified version of the AURORA electronic double auction market recently developed by the Chicago Board of Trade. Precise comparisons will be made of the behavior of human and robot traders. Existing experimental data on human double auction markets are not directly comparable to the computerized market because the latter divides time into discrete bid-offer and buy-sell steps, and adopts a new institutional design, Chicago Rules, that allows only holders of the highest outstanding bid or lowest outstanding offer to execute transactions. The project uses the experiments to study the impact of Chicago rules, the effects of variations in period length in a market with discrete trading intervals, and the consequences of restrictions on the messages that can be transmitted and received by traders. The results of this research will improve experimental methods and our understanding of widely used market institutions. By matching human experimental data with the results from the tournament of computerized robot traders, the project will be able to provide strong tests of the ability of "artificially intelligent" agents to mimic the behavior of human traders in Double-Auction markets. If computer simulations could complement human experiments, this would dramatically increase the number and type of experiments that investigators could use to study the effects of institutional changes. The results of the experiments should also provide new insights into the performance of double auctions. This is important because the double auction is the type of trading institution used by the Chicago Board of Trade, the New York Stock Exchange, and many other securities and commodities exchanges throughout the world.

This research develops and empirically tests dynamic programming models of inventory investment and price determination by durable commodity intermediaries. These firms acquire bulk quantities of durable commodities such as steel, lumber, or coal from producers and subsequently sell their inventories to customers at a markup. The investigators have established a working relationship with one such intermediary, a U.S. steel supplier, which has provided them with detailed daily data on purchases, sales, and inventory holdings for over 2,000 products since June, 1997 with monthly updates on an on-going basis. A major component of the research is micro-oriented, focusing on developing computational and econometric methods to determine whether the firm's behavior can be accurately approximated as an optimal trading strategy from the solution to a dynamic programming problem. Since the data provide identities of each of the intermediary's customers and the prices and quantities they purchased, the investigators can study whether dynamic models of optimal price discrimination can explain significant variations in prices charged to different customers. The project develops new econometric procedures for correcting the problem of dynamic selectivity bias resulting from endogenous sampling of the prices at which the intermediary acquires new inventories. The research also draws macro-level implications by aggregating the micro-level models to develop a market-level (disequilibrium) model of price determination. The investigators analyze the impact of macro shocks and cyclical fluctuations in interest rates on the behavior of these markets to provide new insights into the role of inventories as propagating or stabilizing mechanisms in business cycles. Finally, this research program extends the state of the art on numerical methods for continuous dynamic programming problems such as the inventory problem by comparing discrete approximation methods such as uniform discretization, quadrature-based methods, and random and low-discrepancy multi-grid methods to continuous approximation methods based on polynomial and neural network approximations to the value function and decision rule. The investigators also evaluate Euler equation methods (e.g. parameterized expectations), and methods from the AI/learning literature such as "Q-learning" and neuro-dynamic programming.

This research will develop theoretical and computational approaches for characterizing efficient social insurance institutions. We define a social insurance institution as a compulsory government-run program that provides a set of state-contingent taxes (premiums or contributions) and transfers (benefits) that cover a well-defined set of risks. An efficient social insurance institution is one that provides a given level welfare to the individuals in the system at minimal cost. Our research will attempt to provide characterizations of efficient social insurance institutions that cover the following risks: 1) longevity, via mandatory pensions or "old age insurance", 2) disability, via disability insurance, 3) unemployment, via unemployment insurance, and 4) health care costs, via medical insurance. Another objective of social insurance is lifetime redistribution of income and/or wealth. This can be viewed as insurance for a fifth class of risks, namely insurance for individuals who have certain fixed characteristics or "types" that may lead to permanently lower lifetime employment, earnings, wealth, and welfare.
Social insurance programs are large and pervasive in developed economies. In the U.S., spending on Social Security (Old Age, Disability, and Unemployment Insurance), Medicare, Medicaid amounted to 48.3\% of total Federal spending and 9.8\% of GDP in 2000 (U.S. Congressional Budget Office). All forecasts indicate that unless benefits are reduced, social insurance spending will grow rapidly over the coming decades as the baby boomers age and start to retire. Although there have been proposals to shift the costs of social insurance from the government to the private sector via various "privatization" schemes, a variety of moral hazard and adverse selection problems hinder the operation of private insurance markets. We take market incompleteness as the principal rationale for mandatory government provision of insurance, and as the basic point of departure for our analysis. We assume that the government can compel universal participation, but we also assume that it faces the same informational constraints as private insurance institutions would face if they existed. We deal with these fundamental informational asymmetries via two very different but related strategies: 1) a "dynamic mechanism design" (DMD) approach where we search for an efficient policy over an infinite-dimensional space of all possible policies that satisfy certain participation and incentive constraints, and 2) a "parametric mechanism design" (PMD) approach where we search for an "approximately efficient" social insurance institution in a finite-dimensional subspace of social insurance institutions and where incentive constraints are ignored.

Using these techniques we will not only be able to characterize the form of efficient social insurance institutions, we will also be able to quantify the degree of inefficiency in current social insurance institutions. We expect to be able to characterize optimal social security and disability insurance programs as part of a comprehensive, integrated analysis of social insurance in the U.S. and other developed economies. Given the large share of GDP devoted to social insurance, the potential cost savings from discovering more efficient social insurance programs provides a strong, practical rationale for this research.

In environmental and natural resource economics, the management of renewable natural resources is of considerable interest to economists because the prospect of stewarding a resource forever has practical importance and real-world relevance. Moreover, because past failures in managing certain renewable natural resources properly have lead to the extinction of several species, developing practical strategies for managing remaining renewable natural resources is both timely and useful. This project investigates in detail the management of one renewable natural resource, timber, in the province of British Columbia, Canada.

For a Timber Supply Area (TSA) in British Columbia the investigators have obtained unique access to extremely detailed site-level data, which are used by timber-supply managers in the British Columbia Ministry of Forests when making harvesting decisions. In particular, they have access to elaborate linked biological and geographical data in the form of a Geographical Information System (GIS) at the grid level which, in this case, is a hectare or 100 meters square. Thus, for every hectare in the Fraser TSA, which is located near Vancouver, British Columbia and contains several hundreds of thousands hectares, officials at the Ministry of Forests provide the investigators with a wide variety of biological, engineering, and geographical information relevant to harvesting timber. In addition, they also have access to the harvesting strategies proposed and, in some cases, the decisions implemented by the Minister of Forests, so they can compare our estimated decisions with actual ones.

In developing practical harvesting solutions for timber, they apply the method of stochastic dynamic programming. The goal of the project is to make the following contributions: First, this project takes geography seriously, both in the planar sense and in the three-dimensional sense. Second, it takes site-specific heterogeneity seriously both on the cost side in terms of felling and transportation and on the growth and yield side in terms of heterogeneous stands of timber. Third, it models initial conditions. In particular, it does not take as the starting point a steady-state allocation, or even an optimal allocation. Instead, it takes the existing uneven-aged timber stand as given and derives the optimal policy function, the optimal timber-harvesting profile, in terms of this age distribution. Fourth, this project uses best-practice biological methods to model the dynamics of uneven-aged forest growth and yield. Fifth, in the past economists and foresters have typically demonstrated their methods by making extremely simple assumptions concerning the stochastic nature of timber prices and volumes so that closed-form examples could be solved. This research implements recent developments in computational methods so that practitioners can solve numerically for the optimal policy function in realistic biological, economic, and spatial environments.

Proposal No: 0635806
Institution: University of Maryland, College Park
NSF Program: ECONOMICS
Principal Investigator: Rust, John
Title: Collaborative Research on Models of Bargaining and Price Determination of Residential Real
Estate, with and without Real Estate Agents.

Residential real estate accounts for a large share of wealth and GDP in modern economies. In addition, the majority of households own their homes and the sale and/or the purchase of a home is often the largest financial transaction a household engages in. Surprisingly, there are few models available to analyze the housing transaction process. This research will study the transaction process in the housing market. It will develop computationally tractable models of the behavior of buyers, sellers, and intermediaries in the housing market, and estimate the models using a high frequency micro data on individual housing transactions, which include how list prices are revised over time, information on each visit by buyers, and outcomes of bargaining for a large sample of homes over a long period of time the PIs will collect in both the US and the UK. The model has 3 agents---seller, buyers, and real estate agents who interact in several bargaining rounds. Real estate agents are modeled as having access to a technology and data (the MLS) that can increase the arrival rate of buyers as well as match potential buyers to sellers.

This research extends, applies, and empirically implements theories of dynamic decision making and bargaining under incomplete information to the housing market, in order to describe the operation and efficiency of these large markets. This is the first attempt to apply dynamic decision making to the housing market, a significant contribution to the existing literature. Understanding the transaction process in the housing market is important in itself. In addition, there are potential public policy benefits resulting from better analytical models of the residential real estate market. For example, the U.S. Department of Justice is currently investigating the U.S. National Association of Realtors to determine whether it has created unfair barriers to entry, particularly in restricting access to the MLS, in order to maintain large real estate commissions. The result of this research could be very useful in such litigation.

The analysis will be extended to include other real estate intermediaries and an endogenous choice of whether to sell via a real estate agency, or to sell by owner. This will allow researchers industrial organizational issues connected with real estate agents, including endogenous determination of real estate contracts and commissions. Using data from both the US and the UK allows the PIs to shed light on institutions, laws, and customs affect the relative efficiency of different forms of organization of the housing market. The results from this research should also help households and real estate agents understand the trade-offs at play when formulating home selling or buying strategies.