Forecasting Decomposed Accounting Measures

Jonathan Juhl Severinsen

Student thesis: Master thesis


This project seeks to expand on the current literature, regarding the forecasts decomposed accounting measures. As shown by Fairfield and Yohn (2001), this part of the literature regarding forecasting and accounting, is not a well-covered subject. Therefore, this project has analyzed the effect of decomposing RNOA into PM and ATO and thereafter OPPL, NOA and GP, for the use of forecasting. This is done by creating five different models, that each of which has its own unique role in this project. The first model serves as a benchmark and the subsequent four models, seek to improve upon this. These four models are based on two different schools of thought. The first two of these four models are based on a traditional way of handling multiple linear regressions. The last two models are based on Plenborg, Petersen and Kinserdal (2017) and their way of forecasting value drivers. All five models are evaluated based on several key assumptions, that ensure the models are truly comparable. During the comparison of these five models it has been a focus to have a wide variety of accuracy measures, to thoroughly describe each model. It is then shown, that its possible to improve on the benchmark model, by decomposing RNOA into PM and ATO, and then focusing on forecasting these value drivers separately. However, this was the only model which was deemed an improvement on RNOA, but it was an improvement nonetheless.

EducationsMSc in Finance and Accounting, (Graduate Programme) Final Thesis
Publication date2019
Number of pages96
SupervisorsJeppe Christoffersen