Herd Behavior Experimental Testing and its Effect on Stock Pricing Using Agent-Based Simulation and STGP Algorithm

The aim of this study is to investigate herding behavior bias and its impact on pricing mechanism using STGP algorithm and agent based simulation. For this purpose, 2000 virtual agents that categorized in two groups include a small subset of the total population called "Best agents" and a main group called other "Virtual agents" (the rest of the population after deducting the best agents), receive 2500 real historical quotes (the opening and closing price, the lowest and highest prices, volume and value of transactions) of the financial instruments and would trade virtual market. Virtual agent transactions create a price range in the form of endogenous ones that are used to test hypotheses. Software training is for 2500 days of transactions starting from October 2003. The exam period is June 2013 to December 2018. The results indicate less herding behavior in the groups of Virtual agents, due to the high genetic diversity, the existence of diverse trading strategies and greater coherence due to mechanisms Competitive and evolutionary pricing of an entire market, but herding behavior does not cause for mispricing of assets, bubble and crash in the long run and prices return to their intrinsic value.