Excessive risk-taking in the financial sector contributed to the Global Financial Crisis of 2007-2009, leading many policymakers to consider tightening existing regulations or imposing new ones altogether. One such proposal is to tighten bank capital requirements, which have been a cornerstone of international financial regulation since the 1980s. At least part of capital requirements’ attractiveness to regulators is their simplicity: they promote safety in the financial sector by imposing an upper bound on banks’ risk-weighted leverage. The additional capital serves two primary purposes. First, it buffers against losses on loans and other investments that might lead a bank into financial distress or even failure. Second, it mitigates the extent to which banks can “shift risk” to taxpayers, profiting from gains but not bearing the full brunt of losses.
Simply put then, if the banking system was short of capital during the last financial crisis, why not make sure that it has enough before the next one hits?[i]
One cause for concern is that raising capital requirements may impose real costs. For one, banks who are less able to finance loans with cheap debt may ultimately reduce their credit supply, raising borrowing costs for firms and households. Second, borrowers may simply turn to other parts of the financial sector that are not subject to the same regulations, shuffling risk around the system rather than eliminating it (see, e.g., Buchak et al. (2020). If banks are somehow “better” at lending than other institutions,[ii] this substitution may also increase risk outside the financial sector as well as inside it. While these assumptions are hotly debated between practitioners, regulators, and researchers, we cannot assess the merits of changing capital requirements without examining these costs seriously.
Furthermore, both concerns have become increasingly relevant over time. To illustrate, let’s focus on business debt as I do in my recent paper. Since the implementation of capital requirements in the 1980s, total debt has increased five-fold, while the share of debt coming from bank loans has decreased from 35% to a low of 10% after the 2007-2009 recession.
What would happen if we raised capital requirements?
My paper quantifies these potential costs of raising capital requirements while accounting for several key features of the modern financial system. The benefits are well known: avoiding financial crises, bank failures, and all the associated pain of financial distress. The costs, by contrast, are more opaque. How much would banks cut lending in the face of tighter capital requirements? Even if they cut lending a lot, would it matter? Given the prevalence of alternative financial institutions such as corporate bond markets, “shadow banks,” and others, this latter question is especially important.
To answer these questions, I construct a novel quantitative economic model with both bank and non-bank lending which captures key institutional features of the modern U.S. banking and corporate borrowing landscape. I find that raising the capital requirement to 26% from the current minimum of 8% would effectively eliminate bank failures while yielding miniscule costs in terms of foregone investment or increased volatility. These results hold in both the long run and over the transition from the current regime to the stricter one.
Why do I find that the potentially large costs of increasing capital requirements are in fact so small? A straightforward principle guides my quantitative analysis and leads to these findings. Three links connect capital requirements and aggregate macroeconomic outcomes:
- capital requirements constrain banks’ capital structures (deposits, leverage)…
- which shape banks’ pricing and lending decisions…
- leading to changes in firms’ capital structures (amount and source of funds) and investment patterns.
To understand the total impact of the regulation, we need to understand and quantify the forces shaping each link. In principle, each link may either dampen or amplify the impact of the capital requirement. It turns out that while the first link is very active, each successive link dissipates the effect.
Link 1: the direct impact of capital requirements on bank capital structure is huge! I find that bank leverage declines almost one-for-one with the capital requirement. Given that banks overwhelmingly prefer to finance with debt—the economywide rates of earnings retention and equity issuance in the banking sector are quite low on a quarterly basis—the capital requirement is a large driver of their leverage ratio. Thus, as the capital requirement increases from 8% to 26%, banks’ debt to assets ratio drops from 92% to 74%. Accordingly, failures and distress become much less likely. Of course, though, the point of the paper is that this direct effect is far from the end of the story. Therefore, we turn to the indirect effects captured in links two and three.
Link 2: large changes in bank capital structure do NOT translate to large changes in bank lending. Even while sharply reducing leverage, banks may still change their composition of financing so that their balance sheets do not shrink one-for-one with leverage. They can do this either by retaining earnings or issuing new equity. Despite the latter being costly and unpopular—indeed, banks only issue equity at a quarterly rate of 0.09% in the model, consistent with the data—the former is not. In both the model and the data, banks earn considerable profits on a per unit basis, with net interest margins that are remarkably stable between 3% and 4%. Therefore, banks reduce the size of their balance sheets only slightly. The standard balance sheet identity then implies an identically small drop in bank lending.
In principle, banks could raise loan rates in order to offset any losses stemming from the restriction on cheap debt financing, delivering an adverse effect to total lending even while keeping the banking sector relatively unchanged. However, the presence of the growing “non-bank” financial sector precludes this response. Since banks must compete for marginal loans with non-banks, they face a highly elastic demand curve. Therefore, raising interest rates even a small amount leads to a large reduction in loan volume, which is not profitable. The data confirms this: for the medium- to low-risk borrowers (the bulk of credit volume in the U.S.), the gap between the interest rates on bank loans and corporate bonds is quite small. Therefore, neither loan rates nor loan volumes change very much in response to heightened capital requirements.
Link 3: firms’ capital structures and investment patterns change very little. Two things can still go wrong on the “real side” of the economy, despite the small effects on the “financial side.” First, even a slight decrease (increase) in bank lending (loan rates) may lead to a large reduction in bank borrowing by firms. Second, the overall riskiness of lending in the economy may change in an adverse way.
Neither of these effects is large. As described above, the disciplining presence of the non-bank financial sector essentially kills any large adjustment in loan rates. Bank loan volume decreases only slightly, and much of this decrease is absorbed by the non-bank sector. In terms of overall riskiness of lending, there is simply very little evidence in the data that banks have a sizeable comparative advantage in mitigating the riskiness of loans relative to other classes of lenders. For example, charge-off rates on corporate bonds and bank loans of similar risk classes are quite similar, controlling for seniority. Combining this with the small change in loan volumes leads to a negligible effect on overall risk.
To summarize, because banks can readily retain earnings to increase their capital base in response to increased capital requirements, and because bank loan terms are not especially sensitive to changes in bank capital structure, the net effect of the change from regulation is small. In the long run, the bank share of total debt actually increases by 3.6 percentage points as banks become more price competitive so they can retain more earnings. This is accompanied by small increases in total output and investment and small decreases in loan risk—a win-win situation.
All of these effects hold in the long run, but there is still one potential cost that has not yet been mentioned. Specifically, what if the transition from today’s weaker capital requirement regime to the more stringent one is itself costly? Sure, in the long run banks can retain enough earnings to make the costs small, but what about in the short run? I show that this concern is also minor: unless there is a large string of bad aggregate shocks over the transition path, banks can readily meet the new requirement with minimal costs to output and investment if given a phase-in period of five years.
Policy implications: where do we go from here?
Ultimately, my paper contributes to a long line of literature debating the costs and benefits of capital requirements. What I bring uniquely to this classical discussion is a quantitatively rigorous model which includes many of the potential margins along which capital requirements are costly. In particular, I show how the presence of non-bank lending disciplines the banking sector such that large responses in terms of reduced loan volume are effectively ruled out.
Where do these insights leave policymakers? First, it should leave them more comfortable raising capital requirements on traditional commercial banks: the industry can bear it. Consistent with the cutting insights of Admati et al. (2013), I find that many of the concerns attached to strengthening capital requirements are over-stated (such as the cost of non-deposit financing) or actually work the other way (competition from the non-bank sector).
Second, it should encourage them to shift their focus to the non-traditional parts of the financial sector, whose regulatory framework is much more complex and relatively weak. One area of inquiry my paper leaves open is the financial instability induced by non-bank financial institutions themselves, and this is likely to be one of the most pressing concerns for regulators in the near future.
Kyle Dempsey is an Assistant Professor at The Ohio State University. This post is adapted from his paper, “Macroprudential Capital Requirements with Non-Bank Finance,” available on the ECB’s website.
[i] This is, of couse, an over-simplification of the origins of the financial crisis. For a more detailed look, see, for example, the excellent accounts of Brunnermeier (2009) and Gorton and Metrick (2012).
[ii] There are any possible reasons for this, all of which have received extensive attention in the financial economics literature. Banks might be better at screening borrowers if they have private information gleaned through long term lending relationships (see, for example, Petersen and Rajan (1994)). Or, banks may be able to “monitor” borrowers to impose good behavior (see, for example, the seminal papers of Diamond (1984) and Holmstrom and Tirole (1997)).