Sunteți pe pagina 1din 4

The Dynamics of Quote Prices in an Articial Financial Market with Learning Effects.

Andrea Consiglio Valerio Lacagnina Annalisa Russino February 1, 2006

1 Abstract
In this paper we study the evolution of bid and ask prices in an electronic nancial market populated by portfolio traders who optimally choose their allocation strategy on the basis of their views about market conditions. Recently, a growing literature has investigated the consequences of learning about the returns process1 . There has been an increasing interest in analyzing what are the implications of relaxing the assumption that agents hold correct expectations. In particular, it has been asked the fundamental question of understanding if typical asset-pricing anomalies (like returns predictability, and excess volatility) can be generated by a learning process about the underlying economy. In this paper we focus on the process by which information is incorporated into prices, examining the relationships among the dynamics of price changes, and the time variation of liquidity and of trading activity. We design an order book market system. Agents enter the market sequentially, and they trade to adjust their portfolio according to their optimal target
Corresponding author: University of Palermo, Dept. of Statistics and Mathematics Silvio Vianelli, Palermo, IT. Email: consiglio@unipa.it University of Palermo, Dept. of Statistics and Mathematics Silvio Vianelli, Palermo, IT. Email: ricopa@unipa.it University of Palermo, Dept. of Statistics and Mathematics Silvio Vianelli, Palermo, IT. Email: arussino@dssm.unipa.it 1 See Barberis (2000); Bossaerts (1999); Brennan and Y. (2001); Guidolin and Timmermann (2005); Lewellen and Shanken (2002); Y. (2001). See Hommes (2005), for a recent survey on heterogeneous agent models.

allocations. At each instant in time, the probability of entering the market is an increasing function of the distance of the agents current portfolio from the optimal allocation of wealth. Once an agent is selected to enter the market, the agents ability to make his desired trades effective depends on the state of the order book. We x the order type submission strategy: all agents submit a market order for the quantity available at the best quote and, if the quantity they want to trade is higher, they place, for the residual quantity, a limit order at a price such that their order will be rst in the queue of orders written in the book2 . We create heterogeneity assuming that investors have imperfect information about the joint distribution of returns. In particular, we allow agents to hold arbitrary priors about the univariate marginal distribution of returns, and we make agents update those distribution using past realized market prices. We concentrate our attention on analyzing the impact of a learning process about the marginal distributions of returns assuming that agents have a constant common view of the assets association structure. They correctly apply a copula function to generate the joint distribution of returns to be used to determine the optimal portfolio allocations. We assign to all agents the same investment horizon, but we create asynchronous updating implicitly assuming that different agents (or groups of agents) entered the market at different moments in time. Finally, we assume that investors are myopic in the sense that, at the beginning of the investment horizon they choose their portfolios as if there will be no further trading. At the end of the investment horizon agents use the observed market prices to update the joint distribution of returns and choose their new optimal portfolio3 . In this way we simplify the optimization problem making investors demand functions dependent on time only through the learning process. In Consiglio and Russino (2005) we used the same setting and we assigned to the investors a prospect-type utility function (Kahneman and Tversky, 1979). We have shown that, under learning, the automated auction system generates irregular price series characterized by sharp increases and decreases (looking like bubbles and crashes), but that the jumps in the price series are not related to sudden changes in the optimal portfolio weights. We analyzed the unconditional distribution of price changes and we provided evidence supporting the hypothesis that
2 In an automated nancial market, trading occurs through an electronic order book without involving nancial intermediaries. In this market setting, agents have two types of choices to make: rst, they have to decide the sign and the size of the order; second, they have to choose what type of order to submit (limit or market order). In this paper we concentrate our attention on the interactions between the endogenous order ow, driven by the evolution of the optimal target portfolio allocations, and the price dynamics, imposing exogenously a common order type submission strategy to all agents. See Consiglio et al. (2005) for an analysis of the effect of allowing the agents to choose the type of order to submit on the basis of the information about market conditions revealed by the state of the book, in a setting with exogenously assigned target allocations. 3 See Consiglio and Russino (2005) for the details of the model that we implement.

the parameters characterizing the learning process affect signicantly the evolution of market liquidity, and that the variability of market liquidity determines the observed bubble-like phenomena. This paper is an extension of Consiglio and Russino (2005) in two directions. First, we analyze what is the role played by the assumed portfolio optimization model in affecting the market dynamics. That is, maintaining constant all the parameters governing the learning process, we compare two settings where we change the utility function assigned to the agents. In the rst setting, we use, as in Consiglio and Russino (2005), a prospect-type utility function. In particular, we assume that each investor has an initial level of wealth and a target growth rate to reach within his investment horizon. The investor must determine an asset allocation strategy so that the portfolio growth rate will be sufcient to reach the target. We model the utility function in terms of deviations, measured at regular intervals, from a specied target, and we assume that investors are more sensitive to downside movements. In the second setting, we assign to the investors a standard mean-variance reduced utility function and we study the differences that emerge. Second, we study how the time-series behavior of price changes, market liquidity, and trading activity is affected by the learning process. We use the highfrequency data relative to the order ow and the structure of the book to analyze the short-run impact of those variables on price changes. Following Engle and Patton (2004), we estimate a VAR model for the bid and ask price changes and we include as regressors variables representing the characteristics of the orders arriving in the market and the structure of the book.

References
Barberis, N.: 2000, Investing for the Long Run when Returns are Predictable, Journal of Finance 55, 225264. Bossaerts, P.: 1999, LearningInduced Securities Price Volatility, Working Paper, California Institute of Technology. Brennan, M. and Y., X.: 2001, Stock Price Volatility and Equity Premium, Journal of Monetary Economics 47, 249283. Consiglio, A., Lacagnina, V. and Russino, A.: 2005, A simulation analysis of the microstructure of an order driven nancial market with multiple securities and portfolio choices, Quantitative Finance 5(1), 7187.

Consiglio, A. and Russino, A.: 2005, How Does Learning Affect Market Liquidity? A Simulation Analysis of a Double-Auction Financial Market with Portfolio Traders, http://ssrn.com/abstract=876415. Engle, R. F. and Patton, J.: 2004, Impacts of Trades in an Error-Correction Model of Quote Prices, Journal of Financial Markets 7, 125. Guidolin, M. and Timmermann, A.: 2005, Properties of Asset Prices Under Alternative Learning Schemes, Working Paper 009, Federal Reserve Bank of St. Louis. Hommes, C. H.: 2005, Heterogeneous Agent Models in Economics and Finance, in K. L. Judd and L. Tesfatsion (eds), Handbook of Computational Economics, Vol. 2, Elsevier Science. Kahneman, K. and Tversky, A.: 1979, Prospec Theory: an Analisys of Decision under Risk, Econometrica 47(2), 263291. Lewellen, J. and Shanken, J.: 2002, Learning, Asset-Pricing Tests, and Market Efciency, The Journal of Finance LVII, 11131145. Y., X.: 2001, Learning about Predictability: The Effect of Parameter Uncertainty on Dynamic Asset Allocation, The Journal of Finance 56, 205246.

S-ar putea să vă placă și