Countercyclical capital regime revisited: Tests of robustness

Abstract

Financial entities have been subject to risk-based capital requirements for many years. The experiences of the recent financial crisis led to heightened criticism that these rules are pro-cyclical, meaning they require insufficient capital during economic booms and excessive capital during busts. In response, recent revisions to such rules have included a countercyclical capital buffer, but it is problematic because implementation is discretionary with respect to both timing and amount on the part of regulators. An alternative rules-based countercyclical capital proposal put forth by Smith and Weiher (April, 2012. Countercyclical capital regime: a proposed design and empirical evaluation. Federal housing finance agency, working paper no. 12-2) that embodies a dynamically adjusting stress test for mortgage assets was intentionally structured to address these concerns.This paper tests the robustness of key elements of the Smith and Weiher (SW) countercyclical capital regime. Specifically, we examine each of the key design elements of the stress test, including construction of the trend line, the trough, and the time-path. We also expand on the empirical results presented by SW by showing how the resulting capital requirements might vary when their countercyclical stress test is applied using several different credit models.These tests of robustness support the conclusion that the SW countercyclical capital regime should produce capital requirements sufficient to ensure an entity would remain solvent during severe house price cycles. This conclusion is strongly supported by a test of concept of the countercyclical framework using Fannie Mae's historical book of business. If the countercyclical capital requirement had been in place during the run-up to the recent house price bubble, Fannie Mae would have been sufficiently capitalized to withstand losses it sustained in the subsequent housing crisis. This result is particularly noteworthy given that key components of the SW stress test were designed based upon pre-2002 data.