Analysis of overreaction and underreaction in the American stock market using fuzzy clustering means algorithm

In this paper empirical tests for the analysis of overreaction and underreaction hypothesis in the American stock market are presented. For these tests a new methodology based on the fuzzy clustering means algorithm is proposed. Such methodology is strongly connected with two heuristics of behavioral finance: representativeness and anchoring. The proposed methodology is used to form portfolios through financial ratios of public companies and the results obtained are consistent with the strong influence of overreaction in the American stock market. The analysis is applied for stocks from oil and gas, textile and, steel and iron sectors, with financial indexes ranging from 1999 to 2007.