What does the UK experience tell us about cyclicality in banks’ risk appetite?

Matthew Osborne, Alistair Milne & Ana-Maria Fuertes.

Does the risk appetite of banks vary over the cycle? Our recent research paper sheds light on this issue by examining the time-varying correlation between banks’ capital ratios and lending rates which cannot be explained by bank characteristics, such as capital requirements, portfolio risk, size and market share, or macroeconomic factors.  The relationship notably differs between episodes of rapid credit expansion (“good times”), and episodes of crisis with moderate or negative credit growth (“bad times”).  This is difficult to reconcile with traditional theories of bank intermediation, but is consistent with recent theories emphasising cyclical variation in bank leverage and risk appetite.

In our forthcoming research paper, we examine the relationship between lending rates and capital ratios for a panel of 13 large UK banks over the period 1998-2012. Lending rates are proxied by effective rates, that is, the interest received (on an amortised basis) divided by the average stock of loans.  Our “capital ratio” measure is the Tier 1 capital divided by total assets (effectively, a leverage ratio; high leverage, that is, a low ratio of capital to total assets means that, all else equal, a bank has less capital to absorb losses per unit of asset). We estimate a fixed effects panel regression of lending rates on capital ratios (both expressed in levels) and a range of control variables, which include measures of each bank’s portfolio risk, competitive position and size, together with macroeconomic fundamentals.

Our results (summarised in Table 1) show that a positive relationship between capital ratios and lending rates prevails during the pre-crisis “good times” (October 1998 – June 2007). In the “bad times” after the global financial crisis (July 2007 – December 2012), the direction of the relationship reverses, namely, higher bank capital ratios come hand-in-hand with lower lending rates.  This finding echoes those few prior studies which have suggested a negative shift in the relationship between capital ratios and lending rates during periods of financial crisis or economic contraction (Steffen and Wahrenburg 2008; Fischer et al. 2013).  Our paper is, though, the first to find such evidence of cyclical variation for the recent financial crisis.

We find similar evidence of cyclical variation in the relationship between capital ratios and lending rates for sub-sectors.  Secured household lending (residential mortgages) exhibits a positive relationship pre-crisis and a negative one during the crisis.  Unsecured household lending (e.g., credit cards) shows a positive relationship in both pre-crisis and crisis periods, but a significant change in magnitude from strong to weak.  For corporate loans, we find a significant negative association pre-crisis and no association during the crisis.

We ran a number of robustness tests to check this result: (i) formulating the panel model in ‘error correction’ form to capture the short-run dynamics of loan rates, (ii) estimating the panel model using rates on new business lending only, (iii) using data sampled at the quarterly rather than monthly frequency, (iv) including the capital requirement set by the FSA as additional control variable, and (v) allowing not only for bank fixed effects but also time fixed effects so as to ensure that all (observed and unobserved) aggregate common factors influencing bank loan rate decisions are controlled for.

Such instability is difficult to reconcile with well-established theories of financial intermediation, but is consistent with recent theories of bank portfolio decision emphasising cyclical variation in bank leverage and risk appetite.  We briefly review these strands of the literature below.  This view that bank risk-appetite varies cyclically and, in turn, drives credit growth and investment is of course not completely new. It is found in the discussion of ‘animal spirits’ in Keynes’ General Theory and is at the heart of Hyman Minsky’s analysis of speculative investment cycles. As we discuss below, there has been renewed interest in these ideas since the global financial crisis.

Static theories of bank capital structure decision

In traditional theories of bank decision-making prior to the crisis, cyclicality in loan supply and risk exposure enters through time-variation in macro and bank controls.  One example is the theoretical model of Holmstrom and Tirole (1997): bank capital amounts to banks putting sufficient “skin in the game” which increases their incentive to monitor borrower quality; accordingly, higher capital is associated with higher loan volume and lower lending rates.  Other theoretical models suggest that short-term wholesale funding can discipline banks, namely, the possibility of withdrawal of this short-term funding discourages banks from engaging in behaviour that increases the risk of bank failure and, in turn, the anticipated losses on bank debt (Diamond and Rajan 2000). The prediction is that lower capital, resulting from a substitution of short-term wholesale funding for capital, would be associated with higher lending volume and lower lending rates.

Another relevant strand of the theoretical literature incorporates bank capital in broader risk-return decisions (e.g. Hellman et al. 2000; Boyd and De Nicolo 2005) but offers similarly contrasting predictions on the association between bank capital and lending rates. Higher bank risk-taking could either be reflected in greater loan supply (lower lending rates) or a shift in portfolio composition towards higher-risk loans (higher lending rates).

One common implication of these well-established theories is that, once time-varying macroeconomic and bank factors affecting loan supply and demand have been controlled for, there should be a stable relationship between bank capital and lending rates.  Our findings clearly reject this prediction.

Theories with an explicit role for cyclicality in bank leverage and risk appetite

So what theory applies?  Our findings support recent theoretical proponents of variation in bank leverage and risk appetite (e.g., Geanakoplos 2010; Adrian and Shin 2011; Borio and Zhu 2012). This new literature distinguishes periods of rapid credit expansion – when, for example, the bank and its investors are optimistic about returns or perceive risks as relatively low – and periods of slow credit expansion or contraction – when they may hold opposite views, becoming pessimistic about returns or perceiving risks as being relatively high.

This cyclical variation has been rationalized in distinct ways (Gambacorta and Marquez-Ibañez 2011 provide a review). The ‘leverage cycle’ in Geanakoplos (2010) arises from the interaction of heterogeneity in beliefs and constraints on borrowing. In expansionary periods, optimistic investors are willing to pay high prices for assets – rising prices increase the access of these borrowers to funding which further increases asset prices.

A second rationale hinges on asset price volatility (Brunnermeier and Pedersen 2009). Value-at-risk constraints determine access to leverage and this predicts multiple equilibria with the possibility of periods of low volatility, high asset prices and (by implication) high levels of lending.

A third rationale is behavioural, with reference to potential investor and intermediary irrationality. Periods of low interest rates and rapid growth may lead investors and bankers to underestimate risks. In Borio and Zhu (2012), during periods of low perceived risk and credit expansion (such as the ‘great moderation’ that preceded the global financial crisis), banks see less need to maintain a large buffer of capital; more aggressive banks may operate with lower capital and more portfolio risk. This cyclicality may be reinforced by increased credit demand (Minsky 1986) or by rises in the market value of bank capital (Borio et al. 2001). Another rationale put forward by Gambacorta and Marquez-Ibañez (2011) is that cyclicality in bank lending may be driven by incentive arrangements that focus excessively on short-term performance.

Some final thoughts for policymakers

A couple of caveats should be noted.  Ours is a relatively limited exercise using data on the 13 largest UK banks from 1998 to 2012 and examining only two dimensions of bank decision making. Our research design does not fully control for banks’ determination of their capital ratios alongside loan supply. For these reasons, there are limits to the conclusions that can be drawn from our investigation.  We cannot, for example, predict the impact of higher regulatory capital requirements on lending rates.

First, one should be cautious about studies that draw conclusions about the cost of higher capital requirements from statistical associations between capital ratios and lending rates, such as De Ramon et al 2012 (which acknowledges this caveat) and BIS 2010.  Our findings suggest that there could be cyclicality in such associations, including a switch in direction as well as strength of the association, calling into question whether estimates based on data for a specific time episode can be generalised.  Studies that model the cost of capital using a bottom-up approach, by comparing estimates of the cost of equity and of debt and assuming that banks seek to meet an ROE target, may be more reliable (e.g., Brooke et al. 2015), but might still be improved by allowing for cyclicality in the relationship.

Second, we believe that further research is needed to explore banks’ decision-making during the recent financial crisis, as this has implications for the transmission of financial shocks to the real economy.  In particular, we would encourage further empirical research on the extent to which theories which provide an explicit role for cyclicality in risk-appetite may explain banks’ behaviour, in particular their willingness to accept risk for any given level of bank capitalization, as driven for example by changes in perceptions of risk or beliefs about future returns.

Third, it is necessary to allow for changes in bank risk-appetite in the conduct of monetary, macroprudential and supervisory policy. This will be no easy task given the limited availability of theoretical and empirical models to explain changes in bank behaviour over the credit cycle. There will continue to be an essential role for policy judgement, supported by comparative and historical examination of the banking sector in the UK and in other countries.

Table 1: The relationship between lending rates and the capital ratio (plus a host of control variables)



Matthew Osborne works in the Bank’s Sterling Markets Division, Alistair Milne is from Loughborough University and Ana-Maria Fuertes is from Cass Business School.

If you want to get in touch, please email us at bankunderground@bankofengland.co.uk. You are also welcome to leave a comment below. Comments are moderated and will not appear until they have been approved.

Bank Underground is a blog for Bank of England staff to share views that challenge – or support – prevailing policy orthodoxies. The views expressed here are those of the authors, and are not necessarily those of the Bank of England, or its policy committees.