Empirical Part I – Testing for Predictability

Abstract

Chapter 5 presents tests of the above-stated hypothesis, which is derived from the research question formulated in the introduction. Besides the motivation outlined along the formulation of the first research question, there are additional factors which favour tests of predictability based on past returns. As shown in the literature review, there is an extensive number of potential forecasting variables and some of them seem to have promising forecasting power. However, each of them has its drawbacks. A major constraint for Swiss companies is data availability. There are almost no such long-time series available, as there are for US data, which makes it difficult to test certain variables since no sufficient sample can be gathered. This is especially true for the company level or for certain market segments. However, with respect to the effects of the real economy on return predictability, a market segment perspective seems important. Another challenge is the data frequency. Some variables such as accounting numbers are updated only one to four times a year. Additionally, historical practices contribute to a low data frequency. For instance, it is common for Swiss companies to pay dividends on an annual basis rather than each quarter. Moreover, some time series of the above-mentioned variables are difficult to handle econometrically. Testing the Random Walk Hypothesis might overcome some of the stated problems.