Abstract
Can traders in a financial market learn whether to be informed and which information to use in their demand for risky assets? We describe in this paper an agent-based model where heterogeneous traders seek short-term profits and differ in their choices to use or discard some signals. In the model, a vector of fresh news/signals is available at every period and some (but not all) the signals affect the stochastic payoff of the stock.
Under an evolutionary dynamics favouring higher myopic returns we find that, in equilibrium, traders mostly end up in either discarding all signals or being (perfectly) informed using all the relevant signals (paying the related costs). Moreover, the rate of use of information strongly depends on the “complexity” of the market: an excessively large abundance of signals to be screened or a high volatility of the market, result in large shares of passive agents who overestimate the market’s risk; conversely, low market complexity is associated with a more intense use of information and aggressiveness of informed traders.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
Notes
- 1.
We discard “technical” information, mostly derived from time-series and historical data. Many traders may use such “information” but the model has little to say in this respect as no past observation is used, see [4] for an evergreen examination of technical trading.
- 2.
Admittedly, the agents in our model learn in a very basic way, as they have no memory or expectations and update their behaviour based on a single random match. A discussion of more sophisticated reinforcement learning approaches (with an extended bibliography) is in [3], where a form of collective intelligence is built to maximise returns. In contrast, we assume agents are selfish and myopic.
- 3.
It is useful to add to the description of what our agents do, a list of things they do not do: they do not explicitly maximise any utility function, they do not have memory, they do not search in a set of possible alternative strategies or, if they do so, they may need several periods in which they compare the outcomes with a single strategy, they do not try to anticipate the equilibrium price based on the shares of strategies in the population, they do not save or accumulate wealth strategically. In a nutshell, they keep a strategy till they stumble on concrete evidence that someone else makes higher returns and occasionally flips some bits.
- 4.
The “overall” number of set bits is (# of set bit)/NM and the number of “relevant” bits is (# of set bits/SM).
- 5.
Another interpretation leads to overconfidence on the part of informed traders: in principle, once all relevant component of \(\theta _t\) are used, there is no intrinsic noise other than \(\epsilon \) and super-rational agents should set \(v_i=v_\epsilon \). The model shows that there is evolutionary pressure to adjust downwards the individual risk assessment, or, that it pays off for the informed to be overconfident at equilibrium.
- 6.
The reason why this happens is not entirely clear and this phenomenon is known as the “equity premium puzzle”, see [5].
References
Benartzi, S., Thaler, R.: Myopic Loss aversion and the equity premium puzzle. Q. J. Econ. 110(1), 73–92 (1995)
Grossman, S.J., Stiglitz, J.E.: On the impossibility of informationally efficient markets. Am. Econ. Rev. 70(3), 393–408 (1980)
Huang, Y., Zhou, C., Cui, K., Lu, X.: A multi-agent reinforcement learning framework for optimizing financial trading strategies based on timesnet. Expert Syst. Appl. 237, 121502 (2024)
Lo, A.W., Mamaysky, H., Wang, J.: Foundations of technical analysis: computational algorithms, statistical inference, and empirical implementation. J. Finan. 55(4), 1705–1765 (2000)
Mehra, R., Prescott, E.C.: The equity premium: a puzzle. J. Monet. Econ. 15(2), 145–161 (1985)
Newton, J.: Evolutionary game theory: a renaissance. Games 9(2), 31 (2018). https://doi.org/10.3390/g9020031 ,https://www.mdpi.com/2073-4336/9/2/31
Robson, A.J., Orr, H.A.: Evolved attitudes to risk and the demand for equity. Proc. Natl. Acad. Sci. 118(26), e2015569118 (2021)
Sandholm, W.H.: Population Games And Evolutionary Dynamics. MIT Press, Cambridge (2010). Economic learning and social evolution
Schredelseker, K.: Pascal’s wager and information. J. Forecast. 33(6), 455–470 (2014)
Smith, J.M.: Evolution and the Theory of Games. Cambridge University Press, Cambridge (1982). https://doi.org/10.1017/CBO9780511806292
Acknowledgements
We thank the audiences at WEHIA 2022 in Catania, City University London and AAU Klagenfurt for their comments and remarks. Luca Gerotto and Marco Tolotti provided many insightful comments and ideas on related previous work.
Author information
Authors and Affiliations
Corresponding author
Editor information
Editors and Affiliations
Rights and permissions
Copyright information
© 2024 The Author(s), under exclusive license to Springer Nature Switzerland AG
About this paper
Cite this paper
Pellizzari, P. (2024). Learning Whether to be Informed in an Agent-Based Evolutionary Market Model. In: Villani, M., Cagnoni, S., Serra, R. (eds) Artificial Life and Evolutionary Computation. WIVACE 2023. Communications in Computer and Information Science, vol 1977. Springer, Cham. https://doi.org/10.1007/978-3-031-57430-6_25
Download citation
DOI: https://doi.org/10.1007/978-3-031-57430-6_25
Published:
Publisher Name: Springer, Cham
Print ISBN: 978-3-031-57429-0
Online ISBN: 978-3-031-57430-6
eBook Packages: Computer ScienceComputer Science (R0)