Saleem Bahaj, Jonathan Bridges, Cian O’Neill & Frederic Malherbe.
It’s not just what you do; it’s when you do it – many decisions in life have “state contingent” costs and benefits. The payoffs from haymaking depend crucially upon the weather. Putting fodder away for a rainy day can be quick, cheap and prudent when skies are blue. But results may take a soggy and unproductive turn, if poorly timed. The financial climate is similarly important when assessing the costs and benefits of macroprudential policy changes. We argue that it is best to build the countercyclical capital buffer when the macroeconomic sun is shining. We find strong empirical evidence to support our claim.
Within the macroprudential realm, state contingent considerations come to the fore when setting the countercyclical capital buffer (CCyB). The CCyB forms a key part of the new framework of capital requirements for UK banks and the Financial Policy Committee (FPC) has published a high-level strategy for varying it, based on five core principles. Within that, the primary focus is to vary the CCyB – and hence banking sector resilience – to match the evolving risk of bank losses on UK exposures. At any given point, this should ensure that the banking system can withstand stress without restricting essential services, such as the supply of credit, to the real economy.
The FPC’s strategy also acknowledges that – while not the primary objective – changing the CCyB may in itself affect credit growth. Clearly, any direct impact on lending influences the costs and benefits of making macroprudential hay at any given time. It is therefore important for policymakers to know whether the scale of the lending response to a change in capital requirements is likely to vary according to the prevailing financial climate. We seek to shed light on this question in our paper What determines how banks respond to changes in capital requirements?
Two sequential questions summarise whether a change in bank capital requirements will affect credit provision. First, under what conditions does a change in regulatory requirements affect actual bank capital ratios at all? Second, if actual capital ratios do respond, to what extent does the bank adjust through the denominator (ie by a change in risk weighted assets – and within that lending to the real economy), rather than the numerator (by a change in capital resources):
*Within total risk weighted assets, we focus on loans to the private non-financial sector (households and private non-financial corporations).
From a theoretical perspective, an impact of changing capital requirements on bank lending requires a violation of the conditions for the Modigliani and Miller theorem. Otherwise the composition of bank liabilities should not affect decisions over the asset side of their balance sheet. In our paper, we demonstrate that frictions associated with legacy assets in the banking sector and government guarantees provide reasons for such violations. Moreover, we show that the perceived health of legacy assets and the perceived prospects for new lending are likely to impact the responsiveness of bank lending to regulatory change. In other words, the financial climate can matter for the transmission of macroprudential policy.
Assessing the likely impact on lending of a change in the CCyB is ultimately an empirical question. The CCyB is a new tool as part of a new macroprudential regime, so the effect of changing it is necessarily uncertain, until experience accrues over time. Nevertheless, we want answers now. We therefore glean empirical insight from individual bank responses to changing capital requirements in the previous, microprudential regime. Relationships from past regimes should be extrapolated with caution. But even so, our results point to striking state contingency in the response of bank lending to changes in capital requirements.
Our empirical analysis uses the panel dataset of Bridges et al (2014). This marries individual bank lending data from 1989-2007 with confidential, bank-specific changes to capital requirements. Over our sample, there were 50 changes to capital requirements, with an average size of about ½ pp. These were roughly equally split between increases and decreases (Chart 1).
Chart 1: Break down of capital requirement changes through time, by sign and size
Notes: The bar chart indicates the number of occasions policymakers increased (blue bar) and decreased (red bar) a capital requirement on a bank in our sample in each year. Dashed line indicates the change in the average UK banking system capital requirement weighted by bank lending stocks to the non-financial private sector (household and private non-financial corporations). Moves of less than a basis point in size are set to zero.
We ask this dataset two questions.
First, on average, did regulatory changes matter for bank lending? Chart 2 suggests yes, though not emphatically. It plots the unconditional responses of key variables to a 25bp capital requirement increase, constructed using panel local projections (Jordà 2005). The top left chart in the panel shows the path of the capital requirement itself – the initial increase in our experiment is 25bps, which subsequently decays somewhat. The top right panel shows that an increase in regulatory requirements did lead to an increase in actual bank capital ratios – regulation was binding. The adjustment was not immediate, as banks held voluntary buffers above the regulatory minimum, enabling a smooth transition to the higher requirement. But within a year, adjust they did.
Chart 2: Unconditional responses of key variables to a 25bp capital requirement increase
Notes: These impulses are constructed from a panel local projection using data from 18 UK banks covering 1989-2007. The shock variable is a 25bp change in the bank’s capital requirements. Confidence intervalspresented are two Driscoll and Kraay (1998) standard errors. Linear model estimated.
The bottom charts in the panel show the extent to which banks adjusted their capital ratios via their stock of loans to the real economy (the denominator) versus their regulatory capital resources (the numerator). There is evidence that both were affected. The bottom left panel suggests that the level of loans typically fell by about 0.6% after 1 ½ – 2 years in response to a 25bp increase in capital requirements. But this effect is not statistically significant. In contrast, the response of the level of regulatory capital is both larger (2.3% after 7 quarters) and statistically significant. In short, we find that on average banks’ adjustment mostly occurs on the capital margin.
Our second question is the more interesting one: is bank behaviour dependent on the prevailing state of the financial cycle? Local projection methodology is advantageous here in that it can easily be adapted to construct non-linear impulse responses conditional on other variables (Jordà 2013, Jordà 2015). In particular, we incorporate an interaction term for the aggregate rate of credit growth in the lead up to the change in capital requirement. We can then construct an impulse response similar to Chart 2, but conditional on buoyant and subdued aggregate credit conditions respectively. In other words, we can investigate whether banks respond differently to regulatory change in ‘good’ versus ‘bad’ times.
Chart 3 illustrates our key finding: the financial climate at the time of a policy change potentially matters a lot for policy transmission:
- When aggregate credit growth was relatively strong, banks reduced lending by a statistically insignificant amount in response to an increase in capital requirements – the blue line.
- But when aggregate credit growth was already subdued, the impact of regulatory change on lending to the real economy was statistically significant and much bigger, cumulating to around 4% over 2 years – the red line.
Table 1 spells this distinction out. In response to a 25bp increase in capital requirements, actual capital ratios increase by around 17bps in both strong and subdued aggregate lending environments. But the nature of adjustment differs markedly. In strong conditions, almost all of the increase in the actual capital ratio is met by increased capital resources. That is consistent with it being relatively easy for banks to issue capital or retain earnings in “good times.” In stark contrast, when lending conditions were already subdued, all of the increase in the capital ratio is achieved through a reduction in lending. In fact, in this case the lending response more than accounts for the total increase in the capital ratio – capital resources actually fall a little as the bank de-levers. That is consistent with it being much harder and more expensive for banks to raise capital in bad times.
Chart 3: Response of cumulative real economy lending to a 25bp capital requirement increase conditional on financial conditions
Notes: These impulses are constructed from a panel local projection using lending data (to households and private non-financial corporations) from 18 UK banks covering 1989-2007. The shock variable is a 25bp change in the bank’s capital requirements. We interact the shock with the state of the economy to calculate differentiated impulses. Blue line is aggregate credit growth 1 S.D. above and red line is aggregate credit growth 1 S.D. below the sample average. Confidence intervals presented are two Driscoll and Kraay (1998) standard errors.
Table 1: Approximate contributions to the increase in actual capital ratios after a 25bp capital requirement increase
| Aggregate credit growth above sample mean | Aggregate credit growth below sample mean | |
| Memo: Capital requirement | +25bp | +25bp |
| Response: Actual capital ratio | ~ +16bp | ~ +17bp |
| Of which: | ||
| Capital up | ~ +14bp | ~ –2bp |
| Lending down | ~ +2bp | ~ +19.bp |
Notes: Numbers based on point estimates of the response to a 25bp change in capital requirements at the four quarter horizon. A first order log approximation is used, assuming banks have an initial capital ratio of 12% and that total risk weighted assets grow at the same rate as lending to the private non-financial sector. These approximations mean that the sum of the “of which” components may not exactly equal the overall capital ratio response.
Our findings – to the extent that they can be translated across regimes – have implications for macroprudential policy:
- First, they support the idea of adjusting capital requirements in a countercyclical fashion. The red line in Chart 3 highlights that having to raise capital requirements when credit is already relatively scarce is likely to be particularly costly, in terms of real economy lending. Leaving it until a financial downpour before scrambling to put aside hay for a rainy day means that you have left it too late.
- Second, in the face of policy lags and inherent uncertainty about the future state of the financial cycle, it is perhaps advisable to adjust the CCyB in a fairly opportunistic fashion. If conditions are sufficiently strong that the lending impact is likely to be relatively benign, then it may be worth raising the CCyB because the costs are low. That is consistent with the FPC’s intention to set the CCyB above zero before the level of risk becomes elevated.
Overall, despite our best efforts to improve our machinery and forecasting prowess in both the agricultural and financial sectors, we would still do well to heed John Heywood’s words of 1546:
Whan the sunne shinth make hay, whiche is to say,
Take time whan time comth, lest time steale away
Saleem Bahaj, Jonathan Bridges, Cian O’Neill work in the Bank’s Macroprudential Strategy and Support Division and Frederic Malherbe is an Assistant Professor of Economics at London 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.
Comments will only appear once approved by a moderator, and are only published where a full name is supplied.
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.
