For banks, fair value adjustments do infl uence dividend policy

Most researchers who investigate the interplay between fair value accounting (FVA) and the fi nancial crisis look at the time period during the crisis. This paper investigates a potential role for FVA prior to the crisis: If FVA led to increased accounting profi ts with the recognition of transitory gains through profi t and loss during the boom, and if those increased profi ts provided the rationale for increased dividends, then bank capital became riskier prior to the crisis, and this would have made the system more prone to failure. A study by Goncharov and Van Triest (2011) found no empirical support for an increase in dividends in response to unrealised positive fair value adjustments to income. In contrast, when the setting is limited to only South African banks, this paper fi nds that South African banks did pay dividends from unrealised transitory gains. This fi nding is based on a combination of three strands of evidence: a panel regression of the annual dividends declared by the large South African universal banks that showed that those banks probably ignored the unrealised nature of FVA profi ts when dividends were determined; monthly data from the total South African bank system in a co-integrated regression that showed that unrealised fair value profi ts from the banking book raised the average level of bank profi ts materially; and simple descriptive statistics on distributions that showed that South African banks distributed a greater proportion of profi ts during the critical period of 2004 to 2008 when unrealised fair value profi ts from the banking book raised the level of bank profi ts. The fi nding that South African banks did pay dividends from unrealised transitory gains was also confirmed by bank representatives and the post-fi nancial crisis disclosure of one of the South African banks.Key words: fair value accounting, dividend policy, earnings persistence, bank