Research Papers
Market Stabiilty With Machine Learning Agents (with Javier Pereira)
Journal of Economic Dynamics and Control 122 January 2021
- DOI: 10.1016/j.jedc.2020.104032
- Working paper version (pdf), July 2018, revised July 2020.
We consider the effect of adaptive model selection and regularization by agents on price volatility and market stability in a simple agent-based model of a financial market. The agents base their trading behavior on forecasts of future returns, which they update adaptively and asynchronously through a process of model selection, estimation, and prediction. The addition of model selection and regularization methods to the traders' learning algorithm is shown to reduce but not eliminate overfitting and resulting excess volatility. Our results suggest that even a high degree of attention to overfitting on the part of traders who are engaged in data mining is unlikely to entirely eliminate destabilizing speculation. They also accord well with the empirical ``sparse signals" and ``pockets of predictability" findings of Chinco, Clark-Joseph and Ye (2019) and Farmer, Schmidt and Timmermann (2019).
Macroeconomic Gentrification
Manuscript, Hamilton College, February 2019, Revised August 2019.
This paper explores the macroeconomic relationship between product innovation and consumption inequality. We employ a modified version of the agent-based macroeconomic model in Georges (2011, 2018). A rise in rents accruing to salaried workers can shift both production and R&D spending toward products consumed by that class of worker. The increase in income inequality will then have a magnified impact on consumption inequality through induced changes in product quality and availability. We label this process macroeconomic gentrification. We document three channels that contribute to this process at the macroeconomic level: an income channel, a variety channel, and an innovation channel. We further find that the relationship in the model between the distribution of income and long run economic dynamism can be highly nonlinear.
An earlier version of this paper circlutated as “Innovation and Consumption Inequality in an Agent-Based Macroeconomic Model.”
Product Innovation and Macroeconomic Dynamics
In Complex Systems Modeling and Simulation in Economics and Finance, Springer Proceedings in Complexity, 2018. Shu-Heng Chen, Ying-Fang Kao, Ragupathy Venkatachalam and Ye-Rong Du, eds.
- Working paper version, October 2015 (pdf)
Recent evidence points to the importance of product quality and product innovation in explaining firm level dynamics. In this paper we develop an agent-based macroeconomic model in which both growth and business cycle dynamics are grounded in product innovation. We take a hedonic approach to the product space developed in Georges (2011) that is both simple and flexible enough to be suitable for modeling product innovation in the context of a large scale, many agent macroeconomic model.
Risk Preference and Stability Under Learning
Economics Letters, 132, July 2015, 105–108.
In this note, we consider a simple market environment in which traders with finite memory update their forecasting rules at random intervals by OLS. In this context, changes in the perception of market risk can trigger volatility and bubbles. Consequently, higher degrees of risk aversion among traders can have a destabilizing effect on price dynamics. We consider the interaction of this effect with memory, the speed of learning, and the nature of the forecasting rules.
- Working paper version, March 2015 (pdf)
A Hedonic Approach to Product Innovation for Agent-Based Macroeconomic Modeling
Manuscript, Hamilton College, 2009. Revised, September 2011.
This paper develops a hedonic approach to product innovation suitable for large scale agent-based macroeconomic modeling.
Learning Dynamics and Nonlinear Misspecification In an Artificial Financial Market (with John C. Wallace)
Macroeconomic Dynamics, 13:5, 2009, 625-655.
This paper considers the behavior of artificial agents who evolve their forecast rules over time in a simple financial market environment with over-the-counter trading. Asset price dynamics are driven by a combination of nonlinearly misspecified forecast rules and learning how to forecast. We explore the degree to which agents’ forecasts are drawn toward a minimal state variable learning equilibrium as well as a weaker long run consistency condition.
Bounded Memory, Overparameterized Forecast Rules, and Instability
Economics Letters, 98:2, 2008, 129-135.
We consider an environment in which traders with finite memory update their forecast rules at random intervals by OLS. In this context, overparameterization of the forecast rules can destabilize the learning dynamics. This instability tends to be attenuated by greater memory and less frequent updating.
Staggered Updating in an Artificial Financial Market
Journal of Economic Dynamics and Control, 32:9, 2008, 2809-2825.
We consider an environment in which traders search for trading opportunities and update their forecast rules at random intervals. The staggering of this updating process across traders allows differences in opinion to persist over time, generating non-trivial price dynamics.
Learning With Misspecification in an Artificial Currency Market
Journal of Economic Behavior and Organization, 60:1, 2006, Pages 70-84.
Agents evolve their forecast rules over time via a modified genetic algorithm in a simple artificial currency market. These forecast rules can be nonlinearly misspecified. When the misspecification is suppressed, learning tends to be complete. When it is not suppressed, learning can generate persistent exchange rate dynamics.
Learning Dynamics With Nonlinear Misspecification
Manuscript, Hamilton College, July 2002. Revised, October 2002.
This paper considers the behavior of artificial agents who evolve their forecast rules over time in a simple market environment. The forecast rules can be nonlinearly misspecified and thus induce interesting dynamics in the single state variable. Learning fails to eliminate this misspecification under the social learning algorithm considered here.
Adjustment Costs, Learning, and Indeterminacy
Journal of Economic Dynamics and Control 28:1 October 2003, Pages 101-116
I compare the adjustment cost and learning approaches to resolving indeterminacy of perfect foresight equilibrium in a continuous time model with local indeterminacy. Rapid learning tends to select different perfect foresight equilibria of the original model than does the inclusion of vanishing adjustment costs. This difference has implications for the dynamic response of the economy to unanticipated shocks.
An Efficiency Wage Model With Persistent Cycles
Economics Bulletin, 5:3, 2002, pp. 1--6.
This note develops an efficiency wage model which displays persistent cycles under perfect foresight. Limit cycles arise from the dependence of current labor supply on both recent labor market conditions and the expected rate of job creation.
The Credit Channel of Monetary Policy Transmission: Evidence from Stock Returns in the 1990s (With Elizabeth J. Warner)
Economic Inquiry, 39:1, January 2001.
This paper offers a novel test of the credit view of the monetary policy transmission mechanism using stock market returns. We identify Fed policy shocks using newspaper accounts and track daily stock prices immediately following the shocks. If the credit channel is important, then firms that are dependent on bank credit and internal funds should receive a relatively greater benefit (loss) from a Fed easing (tightening) than firms with access to non-bank credit at favorable terms. We identify ten policy shocks during the expansion of 1993-94 and the "credit crunch" period of the 1990-91 recession and find little evidence supportive of an operative credit channel.
Profit-Shares, Bargaining, and Unemployment
Economic Inquiry, 36:2, April 1998.
This paper considers the effect of profit sharing on unemployment in a simple model with an employment externality, overhead costs, and free entry. When wages are determined by Nash bargains and employment is set by firms, equilibrium unemployment is unambiguously reduced by mandated profit sharing but may be either mitigated or exacerbated by a bargained profit share.
Debt and Taxes in a Marxian Growth Model (with Thomas R. Michl)
Review of Radical Political Economics, 28:3, September 1996.
We analyze government debt in the context of a growth model that is Marxian in the sense that its dynamics depend on a capitalist class which accumulates wealth in the form of capital and government debt.
Adjustment Costs and Indeterminacy in Perfect Foresight Models
Journal of Economic Dynamics and Control; 19:1-2, Jan.-Feb. 1995, pages 39-50.
This paper suggests a method for resolving indeterminacy of equilibrium in perfect foresight models and applies it to several existing models. Quadratic adjustment costs are introduced into models with multiple equilibria, and the limiting behavior of the resulting solutions is studied as the adjustment costs go to zero. The method is shown to be useful in selecting among feasible equilibria in examples in which there is local indeterminacy.
Unemployment Persistence under Profit Sharing
Economics Letters; 45:3, July 1994, pages 329-34.
In a simple dynamic model with an employment externality generated by union bargaining, an increase in a mandated rate of profit sharing reduces not only the steady-state unemployment rate but also the persistence of unemployment out of steady state.
A Dynamic Macroeconomic Model of the Labor-Managed Economy
Journal of Comparative Economics; 18:1, February 1994, pages 46-63.
This paper presents a macroeconomic model of a labor-managed economy under decentralized employment setting. In the model, individual labor-managed firms maximize rents. As a result, the behavior of aggregate employment hinges on the relationship between the employment strategies of individual labor-managed firms and opportunities for reemployment in other labor-managed firms. The author shows that, within the framework developed here, the allocation of labor in a labor-managed economy is identical to that in a competitive capital-managed economy and may be Pareto superior to that in a comparable imperfectly competitive capital-managed economy in which there is decentralized bargaining. These results are based on the unity of decision making and residual claimancy in the labor-managed firm but are fundamentally macroeconomic in nature. (c) 1994 Academic Press, Inc.
Emphasizing Controversy in the Economics Curriculum (with Fred Moseley and Christopher Gunn)
Journal of Economic Education 22:3, Summer 1991.
"The Status and Prospects of the Economics Major" (Siegfried et al. 1991) is a welcome and timely reexamination of the undergraduate economics curriculum. We enthusiastically agree with a number of points in the article, such as the suggestion to examine the underlying social values implicit in different economic theories, the emphasis on empirical tests of theories, the importance of active learning, the emphasis on contextual inquiry, and the related proposal for a breadth requirement. However, we disagree with the report on one important and far-reaching point: it completely ignores the controversial nature of economics. We argue that the economics curriculum should emphasize the important controversies within economics and should challenge students to learn to critically evaluate the different sides of these ongoing debates.