Many academic researchers, policy makers, and other practitioners have concluded that fair value accounting can lead to suboptimal real decisions by firms, particularly financial institutions, and result in negative consequences for the financial system. This conclusion is sustained by the belief that fair value accounting was a major factor contributing to the 2008-2009 financial crisis by causing financial institutions to recognize excessive losses, which in turn caused excessive sales of assets and repayment of debt, thereby leading to procyclical accounting leverage. Leverage is procyclical when it decreases during economic downturns and increases during economic upturns. In our paper, Does Fair Value Accounting Contribute to Procyclical Leverage?, which was recently made publicly available on SSRN, we examine whether there exists any link between fair value accounting and procyclical accounting leverage.
To address this question, we develop a model of commercial bank actions taken in response to economic gains and losses on their assets throughout the economic cycle to meet regulatory leverage requirements. We focus on commercial banks because of the central role they play in the financial system and the allegation that their actions in response to fair value losses contributed to the financial crisis. Our model and empirical tests based on the model establish that procyclical accounting leverage for commercial banks only arises because of differences between regulatory and accounting leverage, and not because of fair value accounting.
Studying whether commercial banks exhibit procyclical accounting leverage and, if they do, whether bank regulation or fair value accounting is its source is important to helping policy makers determine how best to minimize the effects of exogenous shocks to financial asset prices on the macro economy. Although many claim that fair value accounting caused banks to take actions that resulted in procyclical decreases in bank leverage during the financial crisis, there is no direct evidence that this is the case. Moreover, we are unaware of any research regarding the role of bank regulation as a cause of these actions.
Our model of bank behavior in response to changes in asset prices assumes the objective of a bank is to maximize return on equity. To achieve this objective, a bank uses debt financing whenever possible to acquire risky assets, thereby maximizing accounting leverage. However, bank regulatory leverage constraints limit both the amount of assets a bank can purchase using debt financing and the riskiness of the assets in its investment portfolio. In particular, a bank’s regulatory leverage—the ratio of risk-weighted assets (where weights are set by a regulator) to regulatory capital—cannot exceed an amount set by the regulator. Thus, our model assumes that banks maximize accounting leverage subject to a regulatory leverage constraint. The model shows that procyclical accounting leverage results only when the average regulatory risk weight of assets purchased (sold) in response to increases (decreases) in asset values is less than the average asset risk weight of the assets in its investment portfolio prior to the purchase (sale). That is, absent differences in regulatory risk weights across assets, accounting leverage cannot be procyclical. Thus, procyclical accounting leverage is attributable to bank regulatory requirements and not fair value accounting. Regulatory risk weights likely change countercyclically, i.e., risk weights increase (decrease) during economic downturns (upturns). The model shows that counter-cyclical regulatory risk weights only serve to exacerbate any accounting leverage pro cyclicality.
We empirically test predictions of the model using quarterly financial statement and regulatory data for a sample of US commercial banks from 2001 to 2010. Following prior research, we begin by estimating the relation between change in accounting leverage and change in assets and find that the relation is significantly positive, which indicates that accounting leverage is procyclical. However, consistent with the predictions of our model, this procyclical relation evaporates when change in each bank’s weighted average regulatory risk weight is included in the estimating equation. We next estimate the relation between change in accounting leverage and change in assets disaggregated into comprehensive income, other changes in equity, and change in debt.
We find that comprehensive income and other changes in equity are significantly negatively related to change in accounting leverage, and change in debt is significantly positively related. These are the expected relations between change in leverage and change in debt and equity. When we estimate this relation disaggregating comprehensive income into net income, fair value components of other comprehensive income, and the remaining components of other comprehensive income, all three components of comprehensive income are significantly negatively related to change in accounting leverage. Thus, we find no evidence that fair value accounting—whether reflected in net income or other comprehensive income—is a source of pro cyclical accounting leverage.
Because of the asymmetry in accounting for gains and losses and that the concerns about fair value accounting and procyclicality arose during the economic downturn that followed the financial crisis, we estimate the relations described above separately for quarters of up and down economic markets. Inferences based on these separate estimations are the same as those for the full sample, except that there is no evidence of procyclical accounting leverage in down markets. To test more directly whether there is a link between fair value accounting and procyclical accounting leverage, we estimate the relation between change in accounting leverage and change in assets arising from fair value comprehensive income, other changes in assets, and their interaction. We find no evidence of a relation between the interaction and change in accounting leverage.
The full paper is available for download here.