A Simulation Experimental Study Of The Traders Adaptive Learning Behavior Effect On The Stock Market

The abnormal phenomena which are contradictory to the efficient market theoryare troubling for the standard financial theories.So,the new different theories andmethods are appearing.Agent-based computational finance views the finance marketas a complex adaptive system.By building the artificial finance market and using thestatistic theories and methods,it explores the reasons that cause the abnormalphenomena from the micro level.This dissertation uses the methods of theAgent-based computational finance.The stock market is viewed as complex adaptivesystem consisting of many boundedly rational and dynamic heterogeneous traders.Inorder to survive in the stock market,traders should learn from each other as newinformation becoming available and adapt their behavior accordingly over time.It isthe interacting of the adaptive traders causing the complexity of stock market and theabnormal phenomena of the market.The idea of bottom_up approach provides a neworiginal method for the study of stock market.Meanwhile,the conclusions based onthis study have the theoretical and realistic significance.This dissertation has seven chapters,the main contents as follows.Firstly,wedivide the learning mechanism of traders into individual learning and social learning.Based on the divide,we construct the four artificial stock markets:trades only haveindividual learning;traders only have individual learning;traders have both learningmechanism;some traders only have individual learning and the others only haveindividual learning.By using the statistical methods to analyze these four simulativeexperimental data,we can know how the different learning mechanism of tradersaffects the stock market clearly.At last,we introduce the price limit institute into theartificial stock market and make it has the properties of the Chinese stock market.Ascomparing the simulation results with the Shenzhen stock exchange constituent index,we investigate the problems of probability and validity for using the Agent-basedcomputational finance methods to study the Chinese stock market.In this dissertation,the main conclusions are the following:(1) The trader'slearning ability can make it more rational and then improves the efficient of the stock market.But the effect of learning is limited,even with the more powerful learningability,traders are not able to know everything.The stock market is not completelyefficient.(2) The social learning mechanism is an important reason that causes thevolatility of stock price.(3) The speed of learning can affect the stock marketsignificantly no matter what kind of learning ability traders have.(4) The price limitinstitute can decrease the degree of volatility of stock price,but at the same time,italso affects the trader's predication of the future and then prevents the stock pricefrom converging to the theoretical price.(5) The simulation results that come from theartificial stock market which has the properties of the Chinese stock market exhibitthe stylized facts such as excess volatility and clustered volatility.It proves that usingthe Agent-based computational finance methods to study the Chinese stock market isfeasible.The major innovation of this dissertation includes:(1) Using a new method,the Agent-based computational finances,study thestock market.Build the relative artificial stock markets at Matlab 7.1 and accomplishall programs independently.(2) Investigate the trader's different learning mechanism effect on the stockmarket in a uniform artificial stock market structure.Especially dividing the tradersinto individual learning trader and social learning trader,study on how the behaviorand the constitution of these two types of the traders affect the stock market.(3) Investigate the effect of price limit institute to the stock price in the artificialstock market and the problems of probability and validity for using the Agent-basedcomputational finance methods to study the Chinese stock market.