The effect of bank capital on lending is a critical determinant of the linkage between financial conditions and real activity, and has received especial attention in the recent financial crisis. We use panel-regression techniques-following Bernanke and Lown (1991) and Wilcox (1993, 1994)-to study the lending of large bank holding companies (BHCs) and find small effects of capital on lending. We then consider the effect of capital ratios on lending using a variant of Lown and Morgan's (2006) VAR

more »

... model, and again find modest effects of bank capital ratio changes on lending. These results are in marked contrast to estimates obtained using simple empirical relations between aggregate commercial-bank assets and leverage growth, which have recently been very influential in shaping forecasters' and policymakers' views regarding the effects of bank capital on loan growth. Our estimated models are then used to understand recent developments in bank lending and, in particular, to consider the role of TARP-related capital injections in affecting these developments. * for helpful conversations. The views expressed here are our own and do not necessarily reflect the views of the Board of Governors or the staff of the Federal Reserve System. 1 If leverage is procyclical, the reduction in bank assets associated with a $1 reduction in capital is even larger. 1 developed state-level equations linking bank loan growth to bank capital ratios and employment, along with bank-level equations for a single state (New Jersey). 2 This paper draws on this earlier literature, as well as more recent approaches, to examine how bank capital affects bank lending in the U.S. In all cases, we find relatively modest effects of bank capital on lending and more important roles for factors such as economic activity and increased perception of risk by banks. Using balance-sheet data for large bank holding companies (BHCs), we consider variants of two of the approaches developed in the earlier literature: specifically, the Wilcox (1993, 1994) and Bernanke and Lown (1990) models. For our purposes, the main difference between these two approaches is that the Bernanke-Lown approach considers the effect of actual BHC capital-to-asset ratios on BHC loan growth while the Hancock-Wilcox approach looks at the effect of deviations of BHC capital levels relative to an estimated target. Both approaches have merits and shortcomings-which we will discuss at length-but both also agree in finding relatively small effects of BHC capital on loan growth. A potential concern surrounding both sets of results is survivor bias: Our panel only includes BHCs that were still in operation at the end of our sample period (2008:Q3). We address this issue by also considering a more aggregate approach; namely, a modified variant of one of Lown and Morgan's (2006) vector autoregression (VAR) models. This allows us to investigate the dynamic and general-equilibrium effects of an exogenous change in bank capital ratios. Again, however, we find relatively modest effects of bank capital-ratio changes on loan growth even using this very different approach. Our estimates of the effect of bank capital on lending are considerably smaller than the effects suggested in recent statements by U.S. Treasury officials. For example, on the likely effects on lending of capital injections and anticipated capital raising following the stress tests, such statements suggest that a $1 capital injection generates between $8 to $12 of lending capacity. 3 These magnitudes seem more consistent with the view described earlier that banks actively manage their assets to maintain constant bank capital ratios. This view has been quite prominent of late, and is apparently based on a scatterplot for aggregate leverage and commercial-bank asset growth reported by Adrian and Shin (2007) . 4 This scatterplot for commercial banks is reproduced in the lower-left 2 Another important paper in this literature is Peek and Rosengren (1995) , which used New England data to fit a model of bank deposits. Because it focuses on liabilities, this methodology is less suitable for studying the effect of bank capital on loan growth, however. 3 Our estimates are one-fifteenth to one-quarter as large. See Treasury Department Press Release (TG-95) "Treasury Secretary Timothy Geithner, Opening Remarks as Prepared for Delivery to the Congressional Oversight Panel," April 21, 2009, for the quoted lending estimates. 4 Adrian and Shin (2007) do not focus on the effect of bank capital ratios on loan growth. Rather, their paper presents a pictorial representation of this relation as part of a set of motivating facts before moving on to consider the sources of procyclical leverage among securities brokers and dealers.