Abstract
In this thesis the impact of Fair Value Accounting (FVA) and Historical Cost Accounting (HCA) on the earnings volatility of U.S. bank holding companies is investigated. The effect of accounting regime on reported financial data per se and in connection with the market situation is a long-discussed issue in the discipline of accountancy, and re-fuelled by the current financial crisis. The thesis has two research questions, both of which relate to the effects from the implicit assumption of homoscedasticity (of earnings volatility) made in some cross-sectional work when panel data are available. The finding that most earnings series are heteroscedastic, under both accounting regimes FVA and HCA, questions the inferences made from cross-sectional analyses. This result aligns with Pesaran and Smith (1995) and Willet and Falta (2010): panel data should be examined with time series approaches at the firm-level first. Taking then parameter averages, if one wishes, would produce more meaningful sample information. Data were acquired through the Federal Reserve Bank of Chicago which enabled a total sample of 529 banks across five different time periods, the longest series being from 1986 to 2010. Tests for the equality of variances were used to analyse the volatility of variance over time for an understanding of the earnings volatility behaviour. The results and conclusion from this work are that i) data quality matters, particularly in relation with the different interpretations of ‘earnings’, ii) earnings volatility (measured through income measures generated from a generic balance sheet) under HCA and FVA are, by majority of the sample, heteroscedastic, and iii) the average variance of fair value income measures increases more than of historical cost income measures. The overall findings support previous research which argued that FVA increases the volatility in earnings.