Premium
Forecasting Earnings Using Geographical Segment Data: Some UK Evidence
Author(s) -
Roberts Clare B.
Publication year - 1989
Publication title -
journal of international financial management and accounting
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.818
H-Index - 37
eISSN - 1467-646X
pISSN - 0954-1314
DOI - 10.1111/j.1467-646x.1989.tb00007.x
Subject(s) - earnings , econometrics , random walk , economics , margin (machine learning) , inflation (cosmology) , statistics , finance , computer science , mathematics , physics , machine learning , theoretical physics
This study examines the question of whether or not the geographical segment data disclosed by UK companies can be used to generate forecasts of earnings that outperform forecasts based upon past consolidated data. One year ahead forecasts of attributable earnings or net income before extraordinary items are generated for both geographical sales data combined with a consolidated attributable earnings to sales margin and segmental earnings data. The forecasts are based upon forecasts of changes in the GNP of individual countries, both with and without the addition of forecasted inflation rates. It is found that models based upon both geographical segment sales and segment earnings outperform the random walk and random walk plus drift consolidated models for the years 1981 to 1983. The difference in the sizes of the errors generated by the segment data based models and the consolidated data based models are significant in the majority of cases especially when the errors are truncated at 100%. However, there is no additional advantage in terms of forecast accuracy in using segment earnings data rather than segment sales data.