Market efficiency theory and the earnings announcement premium at the Oslo Stock Exchange

Abstract

Lamont and Frazzini (2007) document that a trading strategy consisting of buying every stock expected to announce within the coming month and selling short every stock not expected to announce the coming month generates a large and statistically significant earnings announcement premium in the U.S. stock market between 1972 and 2004. Lamont and Frazzini (2007) claim that the main explanation for the earnings announcement premium is uninformed or irrational demand by individual investors, coupled with imperfect arbitrage by sophisticated investors. Their results are not in accordance with weak-form market efficiency in the U.S. stock market in the sense that historical information can be used to predict future stock prices. This thesis will test if related trading strategies based on predicted quarterly earnings announcement dates generates an earnings announcement premium at the Oslo Stock Exchange in the period between 1999 and 2007. Contrasting with the results of Lamont and Frazzini (2007) the results presented in this thesis, that are not statistically significant, show that various versions of the trading strategy based on predicted earnings announcement dates seem to generate negative monthly average excess returns. Further, a L/S portfolio trading strategy based on actual announcement dates does not generate average monthly returns statistically significantly larger than zero. This indicates that improved methods for predicting earnings announcement dates would not assist in forming L/S portfolios generating positive excess returns over the sample period. Consequently, it seems there was no earnings announcement premium at the Oslo Stock Exchange in the sample period between 1999 and 2007. The results presented in this thesis can therefore not reject market efficiency at the Oslo Stock Exchange. The main reasons for the presented results, which are differing from the results of Lamont and Frazzini (2007), are the following: Firstly, there is a possibility that the dataset of earnings announcement dates utilised in this analysis is not representative for the sample period regarding the real coverage of earnings announcement dates. Moreover, there is a possibility that the patterns found by Lamont and Frazzini (2007) are random, and caused by for example data-mining, and that in reality there is no earnings announcement premium.