We focus on a discussion of Goyal and Welch (2006) and Campbell and Thompson (2008) about the predictability of stock returns. While the former suggest that no regression model exists that can forecast the equity premium more accurately than its historical average, the latter claim that there are, in fact, such regressions if one restricts the model appropriately. The goal of this thesis is to add our input to this discussion by applying the heterogeneity and autocorrelation robust test approach introduced in Ibragimov and M ̈uller (2010) and understanding its mathematical background. To this end, we first present the results of Bakirov and Sz ́ekely (2005), which provide the conservativeness of the two-sided t-test for significance levels α ≤ 0.05 under certain departures from i.i.d. normality. We eventually apply this method of Ibragimov and M ̈uller (2010) in two ways: First, we assess the significance of the proposed regressions’ slope coefficients and compare the results to inferences from alternative methods. Second, we examine whether a regression’s forecast is significantly more accurate than the historical average forecast. Our results show that even a critical investor can find a regression model that predicts the equity premium more accurately than the historical average based on data up to the year 2005.     «

We focus on a discussion of Goyal and Welch (2006) and Campbell and Thompson (2008) about the predictability of stock returns. While the former suggest that no regression model exists that can forecast the equity premium more accurately than its historical average, the latter claim that there are, in fact, such regressions if one restricts the model appropriately. The goal of this thesis is to add our input to this discussion by applying the heterogeneity and autocorrelation robust test app...     »