My devotion to academic economics has been constantly on agent-based economic modeling and simulation. The idea is very much motivated by Herbert Simon, which I called the legacy of Herbert Simon in economics (Chen, 2005, 2007). His interdisciplinary approach to the study of human has deeply inspired my work being interdisciplinary. I started my academic life by following his work on artificial intelligence, and used then-arising computational intelligence to model autonomous agents, a version of bounded-rational agents. After almost two decade’s work, my contribution in this area has gained some degree of international acknowledgement, which provided me opportunities to write overarching articles in the encyclopedia and handbook for young scholars (Chen, 2008a, b).
After years’ work on simulation with software agents, I was attracted recently by experiments with human subjects, including psychology, where we also see the shadow of Herbert Simon. Being the Vice Chair of the IEEE Computational Finance and Economics Technical Committee (IEEE’s Computational Intelligence Society), when I was asked the future research in agent-based computational economics (ACE), my answer is always the integration of human agents and software agents. Chen (2008c) specifies one job of this integration: grounding software agents in studies of human agents, and points out that the heterogeneity of intelligence, personality, and culture, which empirically existing among human agents, has been largely neglected in current agent-based economic modeling. To meet the gap, in Barr et al. (2008), I reiterated that agent-based economics and experimental economics should be integrated with a more coherent framework. In the following, my major work over the past three years will be organized into six categories which can connect the past to the future.
My first work on experimental economics was done almost 12 years ago (Chen, Kuo and Lin, 1996), and since then it has been left aside. My second visit to this area is really motivated by the vision of integrating software agents and human agents as mentioned above. So, since year 2005, under the collaboration with Institute of Physics, Academic Sinica, we started the research on the web-based market experiments, alternatively known as the prediction market. This platform is suitable as a beginning step to see the interaction of human agents with software agents. In this collaboration project, we have published a series of studies on the prediction of future events (Wang, Tseng, Li and Chen, 2006; Wang, et al., 2008; Tseng, Chen, Wang and Li, 2008; Wang, Li, Tai and Chen, 2009; Chen and Wu, 2009).
2 New Elements to ACE
The only available work in agent-based economic modeling which incorporates the idea of modularity is the agent-based models of innovation initiated by Chen and Chie (2004a, b). We proposed a modular economy whose demand side and supply side both have a decomposable structure. While the decomposability of the supply side, i.e., production, has already received intensive attentions in the literature, the demand side is not. Inspired by the study of neurocognitive modularity, Chen and Chie (2004a, b) assumed that the preference of consumers could be decomposable. In this way, the demand side of the modular economy corresponds to a market composed of a set of consumers with modular preference.
Chen and Chie (2007) tested the idea of augmented genetic programming (augmented with automatically defined terminals) in a modular economy. Chen and Chie (2007) considered an economy with two oligopolistic firms. While both of these firms are autonomous, they are designed differently. One firm is designed with simple GP (SGP), whereas the other firm is designed with augmented GP (AGP). These two different designs match the two watchmakers considered by Simon (1965). The modular preferences of consumers define not only the search space of the firms, but also the search space with different hierarchies of the firms. While it is easier to meet consumers’ needs with very low-end products, the resultant profits are negligible. To gain higher profits, firms have to satisfy consumers up to higher hierarchies. However, consumers appear more and more heterogeneous when their preferences are compared at higher and higher hierarchies, which calls for a greater diversity of products. Thus, it shows that the firm using modular design performs better than the firm not using modular design, as Simon predicted.
In experimental economics, it has been an increasing tendency to make intelligence an explicit control variable in experimental economics and to explore its emergent outcomes. However, agent-based computational economics, which is normally claimed as the software counterpart of experimental economics, has paid almost no attention to this development. Chen, Zeng and Yu (2008) is probably the first agent-based model to tackle with this issue. In a context of the agent-based double auction market, they used genetic programming (GP) to model agents’ adaptive behavior.
A more challenging work done by Chen, Zeng and Yu (2008) was to examine the coevolutionary dynamics when competing agents become equally smarter. This brings us closer to the situation discussed in Segal and Hershberger (1999) and Jones (2008). Chen, Zeng and Yu (2008) does show that, even in a competing situation like the double auction game, pairs of smarter agents can figure a way to cooperate so as to create a win-win situation, whereas this collaboration is not shared by pairs of less smarter agents.
AI-ECON Research Center, which is currently under my leadership, is an internationally well-known institute, which is famous for its Taipei Artificial Stock Market (LeBaron, 2006). This research project has constantly generated series of publications since year 2001. Recently ones include Chen and Huang (2005a,b, 2007a,b, 2008), Chen, Liao, and Chou (2008), Chen, Huang and Wang (2009).
The key issue addressed by this series of studies is the significance of investors’ risk preference in the microstructure dynamics. Simply speaking, we are asking whether investor’s investment performance or wealth depends on his risk preference. The famous riskpreference-irrelevancy theorem, or the Blume-Easely-Sandroni theorem (Sandroni, 2000; Blume and Easley, 2006), states that the investors’ wealth share is determined only by their forecasting accuracy and is independent of their risk preference. Using a different theoretical framework which is more suitable for heterogeneous bounded-rational agents, Chen and Huang (2008) showed that risk preference matters. In particular, we found that the class of constant relative risk aversion is dominant among all classes of risk preferences. The other one, CARA (constant relative risk aversion), which is frequently used in finance, is also a dominated class. In addition, the optimality of the capital asset pricing model (CAPM) is invalidated: with the presence of the CRRA traders, the wealth share of investors who follow CAPM to make portfolio decision eventually shrinks to zero.
Chen, Liao, and Chou (2008) studied the plausibility of sunspot equilibria within the context of agent-based financial models. Sunspot equilibrium concerns the real impacts of purely extrinsic uncertainty. The typical question, when coming to stock markets, is whether the stock price dynamics can be affected by any non-fundamental causes simply because investors believe so. It is not surprising to see that there is mapping between sunspot beliefs and sunspot equilibria, which has already been shown in the literature as an existential proof. However, how agents initially with heterogeneous beliefs can coordinate themselves to converge to such equilibria is largely unknown. In fact, the only few studies using experiments with human subjects all find that it is difficult to coordinate such agents’ beliefs to sunspot equilibria (Marimon, Spear and Sunder, 1993; Duffy and Fisher, 2005).
Most empirically-based ACE models only consider how econometrics can help build or validate the ACE models. Few have explored that the other way around may be equally interesting. Chen, Huang and Wang (2009) presents the first work on the reverse direction, i.e., instead of an econometric foundation of agent-based economics, an agent-based foundation of econometrics. On this regard, ACE can help or present challenges to econometrics.
In policy design, the most difficult issue is to gauge the risk associated with regime changes and policy changes when the surroundings or embeddings are poorly understood. This is particularly true when experimentation of changes is either infeasible or very expensive. One fundamental difficulty pertaining to this challenge is the unpredictable behavior of stake-holders. Unless we can reasonably track their behavioral modes, risk of policy changes can hardly be estimated. As the Chinese proverb goes, an orange becomes a tangerine after crossing the Huai river.
The integration of computational intelligence into agent-based modeling, or the smart agent-based model, has, however, proposed a possible solution to this conundrum. While stake-holders are not infinitely smart or perfectly rational, they are not dumb. Accumulated empirical evidence indicates that they are able to learn and adapt to change. Computational intelligence can help us to make our artificial stake-holders have this feature, and hence the risk simulation based on this bottom-up design may reduce the otherwise greater uncertainty due to inappropriate models of agents.
In Chen and Chie (2008), a concrete demonstration is given to illustrate the idea above. Chen and Chie (2008) shows how an agent-based lottery market can be built and how the potential risk to tax revenue can be gauged after either raising or reducing lottery tax rate. Agent’s decision is based on heuristics (including fuzzy reasoning) and is influenced by his personality trait (aversion to regret). So, the cognitive and psychological parts of agents are both encapsulated in our software agents. Moreover, both parts are evolving, which means agents are learning socially. The essence of the above software modeling is that we don’t fix the agent’s decision rule. Instead, we allow them to learn and to explore the embedding environment without further intervention. Software agents modeled in this way are called autonomous agents. We use fuzzy inference system to model the relation between jackpot size and gamblers’ participation level, and genetic algorithms to evolve the three behavioral aspects of lottery players, including lottery participation decision, conscious selection and aversion to regret.
While my interest in computational intelligence is mainly on its foundational role in building agent-based economic models, I have also conducted several studies which independently apply K means, neural networks, genetic programming, genetic algorithms and decision trees to quantitative economics and finance. Tsao and Chen (2004) justifies the financial applications of genetic algorithms with a statistical foundation. Yu, Chen and Kuo (2004a,b) used the lambda abstraction to enhance the semantics of genetic programming and applied it to financial markets. It is found that better semantics enhance search efficiency. Chen, Kuo and Hsu (2008) gives the most comprehensive examination of the trading performance of genetic programming by using data from eight foreign exchange markets and eight stock markets. The structure of trading strategies discovered by genetic programming has been carefully analyzed in their complexity and heterogeneity.
Barr J. Tassier T, Ussher L, LeBaron B, Chen S.-H., Sunder S (2008), The future of agent based research in economics. Eastern Economic Journal 34: 550-565.
[EconLit]
Blume L, Easley D (2006) If youre so smart, why aren’t you rich? Belief selection in complete and incomplete markets. Econometrica 74(4):929-966.
Chen S.-H. (2005) Computational intelligence in economics and finance: Carrying on the legacy of Herbert Simon. Information Sciences 170: 121-131. [SCI]
Chen S.-H. (2007) Computationally intelligent agents in economics and finance. Information Sciences, 177:1153-1168. [SCI]
Chen S.-H. (2008a) Computational intelligence in agent-based computational economics. In Fulcher J, Jain L (eds.), Computational Intelligence: A Compendium, Chapter 13, 517–594, Springer.
Chen S.-H. (2008b) Financial applications: Stock markets. In Wang B (ed.)Wiley Encyclopedia of Computer Science and Engineering, John Wiley & Sons, pp. 481-498. Chen S.-H. (2008c) Software-agent designs in economics: An interdisciplinary framework. IEEE Computational Intelligence Magazine 3(4): 18-22. [SCIE]
Chen S.-H, Chie B.-T (2004a) Functional modularity in the fundamentals of economic theory: Toward an agent-based economic modeling of the evolution of technology. International Journal of Modern Physics 18(17-19):2376-2386.
Chen S.-H, Chie B.-T (2004b) A functional modularity approach to agent-based modeling of the evolution of technology. In Namatame A, Kaizouji T, Aruka Y (eds.) The Complex Networks of Economic Interactions, Lecture Notes in Economics and Mathematical Systems, Springer Vol. 567: 165-178. [SCIE]
Chen S.-H, Chie B.-T (2007) Modularity, product innovation, and consumer satisfaction: An agent-based approach. In: Yin H, Tino P, Corchado E, Byrne W, Yao X (eds.), Intelligent Data Engineering and Automated Learning, Lecture Notes in Computer Science (LNCS 4881), Springer, 1053-1062.
Chen S.-H, Chie B.-T (2008) Lottery markets design, micro-structure, and macro behavior: An ACE approach. Journal of Economic Behavior and Organization 67(2): 463-480. [SSCI]
Chen S.-H, Huang Y.-C (2005a) Risk preference and survival dynamics. In Terano T, Kita H, Kaneda T, Arai K and Deguchi H (eds.), Agent-Based Simulation: From Modeling Methodologies to Real-World Applications, Springer, 2005, pp. 135-143.
Chen S.-H, Huang Y.-C (2005b) On the role of risk preference in survivability. In Wang L, Chen K, and Ong Y. S. (eds.), Natural Computation, Lecture Notes in Computer Sciences 3612, Springer, 2005, pp. 612-621. [SCIE]
Chen S.-H, Huang Y.-C (2007a) The relationship between relative risk aversion and survivability. In Terano T, Kita H, Deguchi H, Kijima K (eds), Agent-Based Approaches in Economic and Social Complex Systems IV, Springer, 2007, pp. 41-48.
Chen S.-H, Huang Y.-C (2007b) Relative risk aversion and wealth dynamics. Information Sciences 177(5):1222-1229. [SCI]
Chen S.-H, Huang Y.-C (2008) Risk preference, forecasting accuracy and survival dynamics: Simulations based on a multi-asset agent-based artificial stock market, Journal of Economic Behavior and Organization 67(3): 702-717. [SSCI]
Chen S.-H, Huang Y.-C, Wang J.-F (2009) Bounded rationality and the elasticity puzzle: An analysis of agent-based computational consumption aapital asset pricing models. In Zambelli S. (ed.) Routledge.
Chen S.-H, Kuo J.-S., Lin C.-C (1996) From the Hayek hypothesis to animal spirits: The phase transition based on competitive experimental markets. In Vaz D, Velupillai K (eds.), Inflation, Institutions and Information: Essays in Honor of Axel Leijonhufvud, Macmillan, 290-318.
Chen S.-H, Kuo T.-W and Hsu K.-M. (2008) Genetic programming and financial trading: How much about ‘What we Know’?” In Zopounidis C, Doumpos M, Pardalos P (eds.), Handbook of Financial Engineering, Chapter 8, Springer.
Chen S.-H, Liao C.-C, Chou P.-J (2008) On the plausibility of sunspot equilibria: Simulations based on agent-based artificial stock markets. Journal of Economic Interaction and Coordination 3(1): 25–41. [EconLit]
Chen S.-H, Zeng R.-J, Yu T (2008) Co-evolving trading strategies to analyze bounded rationality in double auction markets. In: Riolo R (ed.), Genetic programming theory and practice VI, Springer, 195–213.
Chen S.-H, Wu W.-S (2009) Price errors from thin markets and their corrections: Studies based on Taiwan’s political futures markets. Advances in Econometrics, forthcoming. [SSCI]
Delli Gatti D. Gaffeo E. Gallegati M. Giulioni G. Kirman A. Palestrini A. Russo A. (2007) Complex dynamics and empirical evidence. Information Sciences, 177(5): 1204-1221.
Devetag G, Warglien M (2003) Games and phone numbers: Do short-term memory bounds affect strategic behavior? Journal of Economic Psychology 24:189-202. Devetag G, Warglien M (2008) Playing the wrong game: An experimental analysis of relational complexity and strategic misrepresentation. Games and Economic
Behavior 62:364-382.
Duffy J, Fisher E (2005) Sunspots in the laboratory. American Economic Review 95(3):510– 529
Gallegati M. Palestrini D. Delli Gatti D. Scalas E (2006) Aggregation of heterogeneous interacting agents: the variant representative agent framework. 1(1):5-19.
Gode D, Sunder S (1993) Allocative efficiency of markets with zero intelligence traders: Market as a partial substitute for individual rationality. Journal of Political Economy 101:119137
Hall R. E. (1988) Intertemporal substitution in consumption. Journal of Political Economy 96:339-357.
Jones G (2008) Are smarter groups more cooperative? Evidence from prisoners dilemma experiments, 1959-2003. Journal of Economic Behavior and Organization, forthcoming.
Jones G, Schneider W (2006) Intelligence, human capital, and economic growth: A Bayesian averaging of classical estimates (BACE) approach. Journal of Economic Growth 11(1):71-93.
LeBaron B (2001) Evolution and time horizons in an agent-based stock market. Macroeconomic Dynamics 5:225–254.
LeBaron B (2006) Agent-based computational finance. In: Tesfatsion L. Kenneth J (eds.) Handbook of Computational Economics, 2-24: 1187-1233, Elsevier.
Marimon R, Spear S, Sunder S (1993) Exceptionally driven market volatility: An experimental study. Journal of Economic Theory 61(1):74-103.
Neely C. Roy A. Whiteman C. (2001) Risk aversion versus intertemporal substitution: A case study of identification failure in the intertemporal consumption capital asset pricing model. Journal of Business and Economic Statistics, 19: 395-403.
Ohtsubo Y, Rapoport A (2006) Depth of reasoning in strategic form games. The Journal of Socio-Economics 35:31-47.
Ram R (2007) IQ and economic growth: Further augmentation of Mankiw-Romer-Weil model. Economics Letters 94(1):7-11.
Sandroni A (2000). Do markets favor agents able to make accurate predictions? Econometrica 68:1303-1341.
Segal N, Hershberger S (1999) Cooperation and competition between twins: Findings from a prisoners dilemma game. Evolution and Human Behavior, 20:29-51.
Simon H (1965) The architecture of complexity. General Systems 10: 63-76
Stoker T (1993) Empirical approaches to the problem of aggregation over individuals. Journal of Economic Literature, 31:1827-1874.
Tsao C.-Y., Chen S.-H. (2004) Statistical analysis of genetic algorithms in discovering technical trading strategies. Advances in Econometrics 17:1-43.
Tseng J.-J., Chen, S.-H.,Wang S.-C., Li S.-P. (2008) Scale-free network emerged in the markets: Human traders versus zero-intelligence traders. In Terano T, Kita H, Takahashi S, Deguchi H (eds.), Agent-Based Approaches in Economic and Social Complex Systems V, Springer, 245-254.
Wang S.-C, Tseng J-J., Li S.-P., Chen S.-H. (2009) Prediction of bird flu A(H5N1) outbreaks in Taiwan by online auction: Experimental results. New Mathematics and Natural Computation 2:271-280.
Wang S.-C., Tseng J.-J., Tai C.-C., Lai K.-H., Wu W.-S., Chen S.-H., and Li S.-P. (2008) Network topology of an experimental futures exchange. European Physics Journal B 62:105 111.
Wang S.-C, Li S.-P., Tai C.-C., Chen S.-H. (2009) Statistical properties of an experimental political futures markets. Quantitative Finance, Vol. 9, No. 1, 2009, 9-16
[SSCI].
Weede E, Kampf S (2002) The impact of intelligence and institutional improvements on economic growth. Kyklos 55(3):361-380.
Yu T, Chen S.-H., Kuo T.-W. (2004a) Discovering financial technical trading rules using genetic programming with lambda abstraction. In O’Reilly U, Yu T, Riolo R, and Worzel B (eds.), Genetic Programming Theory and Practice II, Springer, 11-30.
Yu T, Chen S.-H., Kuo T.-W. (2004b) A genetic programming approach to model international short-term capital flow. Advances in Econometrics 17: 45-70.