Financial news flow and investor crowd behavior: Testing an overreaction hypothesis

Research has demonstrated that stocks with prior period returns that under performed (i.e., losers) will have superior returns in subsequent periods compared to stocks that in prior periods were the out performers (i.e., winners). The results have become the basis for the development of an overreaction hypothesis and provided evidence of a violation of the Efficient Markets Hypothesis (EMH). However, other studies have found that the difference in returns can be attributed to characteristics of the firms, not to past performance. In this dissertation, a test for overreaction is conducted at the level of the S&P 500 futures, thus removing the question of firm characteristics (and related disputes about risk) from the results. Furthermore, instead of identifying losers vs. winners based on past price performance, this dissertation constructs a bear and bull news intensity data series and tests S&P 500 futures returns against prior period news levels. Evidence is presented showing that bearish prior period news levels produce superior, risk-adjusted subsequent period returns (i.e., a bad news portfolio) compared with returns following prior period bullish news (i.e., good news portfolio). Linear regression tests are then conducted using a ratio of prior period bear to bull news intensity as the independent variable. The tests produced the correct signs and significant t-stats on the news variable, thus providing additional support for an investor overreaction hypothesis.