Prognostisering av räntabilitet på eget kapital - En jämförelsestudie av tre regressionsmodellers prognosförmåga applicerat på svenska data

Sammanfattning: A multivariate cross-sectional model is used in this thesis to proxy for expected earnings and to estimate return on equity for 214 companies over the period 2009-2013, using Swedish data. The model, which has never been tested on Swedish data before, is first confirmed to function as a forecasting model for expected return. Furthermore, the model is evaluated through a comparison with two univariate models based on the assumption that return on equity follows a mean reversion process. Forecast accuracy is calculated as the difference of estimated returns and actual returns. The results show that the univariate models' forecasts are superior to the multivariate model's.