Trading on Parsimonious Prediction Models: Simply Genius or Genuinely Too Simplistic? - A 'fundamental analysis'-based test of market efficiency in the U.S.

Sammanfattning: In the mature field of fundamental analysis (FA) and market efficiency tests, the study of Skogsvik & Skogsvik (2010) stands out in that it generates significant abnormal returns using a parsimonious ROE-based investment strategy in the Swedish market. To test the robustness of their results across countries and time periods, the aim of this thesis is twofold: First, it is investigated whether a simple FA-based trading strategy can generate similar excess returns on a large U.S. manufacturing sample. Second, market efficiency and its time-series behavior in the U.S. market are assessed. This paper replicates Skogsvik & Skogsvik's (2010) investment criteria by employing a prediction model of medium-term ROE changes and an indicator variable revealing mispriced stocks based on residual income valuation. Stock positions are taken in a contrarian fashion and held for 36 months over the period 1979-2014. Despite a strong prediction accuracy of 68% for future medium-term ROE changes, no evidence of significant abnormal hedge returns is found. Additional tests suggest that mitigating critical model weaknesses could improve the results, but also impose non-trivial cost of complexity. Moreover, the theoretical abnormal return potential is found to diminish over time in the U.S. market. While this supports the notion of increasing efficiency, preliminary evidence of 'decoupling' between prices and fundamentals in more recent years might point towards 'crazy' rather than 'efficient' prices.