Banner for 2025 Financial Markets Conference

May 19, 2025

Research Session 1: Bank Capital and the Growth of Private Credit

In this first research spotlight, Harvard University's David Scharfstein explored his paper "Bank Capital and the Growth of Private Credit." The session, moderated by W. Scott Frame of the Bank Policy Institute, examined whether regulatory arbitrage can explain the growth of private credit.

Transcript

David Scharfstein: Thank you. Thank you very much for the opportunity to present my work. Thanks to the Atlanta Fed for hosting this conference, and for including me—and this is joint work with Sergey Chernenko and Robert Ialenti.

You all know that private credit has grown dramatically over the last decade, and what we're interested in is trying to understand what explains the growth of private credit, and what are the potential risks to financial stability. So, there are a few potential explanations.

First of all, before I go further, there are a lot of definitions of private credit. The largest share of private credit—and again, depending on how you define it—is direct lending. That's lending largely to private equity-sponsored businesses. And so, there's been obviously the increase in demand for financing of private equity transactions and the growth of private equity-backed firms.

What we're trying to really understand is why has that demand been met by non-bank financial institutions (NBFIs) rather than by banks, and one potential explanation is what is commonly referred to as the "shadow banking narrative," which is that more stringent bank capital requirements after the Great Financial Crisis (GFC) caused lending to migrate to more highly leveraged nonbanks, which are less subject to capital requirements—and they're more highly leveraged. We're going to present evidence that I think throws cold water on that perspective.

A second explanation is, capital requirements make it relatively more attractive for banks to lend to lenders than lend to risky middle market, private equity-sponsored firms, and we'll go through some calculations that suggest why that is the case. We'll talk about bank regulation and supervision leading to higher operating costs for banks relative to private credit funds, and then we'll explore a fourth explanation, which is private credit investors underestimate risk, and the fees are too high relative to the risk that they're assuming.

There's some evidence suggesting that alpha in private credit is statistically indistinguishable from zero, but with wide confidence intervals such that there could be sizable positive—or sizable negative—excess returns. So, we'll think a little bit about if that can explain some of what's going on in the growth of private credit.

So, we're going to look at private credit through the lens of business development companies (BDCs), which are actively managed investment companies that are regulated under the Investment Company Act of 1940—which means that they have to disclose their financial statements and their portfolio holdings, so it gives us a very nice lens into what private credit funds are holding. It comprises about 20 percent of private credit, but it also tells us about private credit funds more broadly—in large part because there is typically co-investment between BDCs and affiliated funds that are sponsored by the same alternative asset manager.

In addition, there's evidence suggesting that BDCs and US private credit funds have similar leverage, so it gives us some lens into the broader space. BDCs have to hold at least 70 percent of their assets in eligible investments, which are private companies in the US—or those that are public but have market caps under $250 million, and the SEC has leverage restrictions, which since 2008 say that assets over debt, or asset coverage ratios (ACRs), have to be at least 150 percent—or flipping it around, debt-to-asset ratio under 67 percent.

So, these are pass-through entities. There's no entity level taxation, but dividends are taxed as ordinary income and they have to distribute at least 90 percent of their income. There are a lot of numbers here, but we're looking at BDCs as of 2023 Q2. Things have changed since then, to some extent, in terms of credit spreads tightening, and we're looking at more recent data in ongoing research.

Key things to take away from this are that the average debt-to-asset ratio is about 50 percent—so, under the 67 percent cap—that roughly half is bank debt and half are unsecured bonds. We're looking here largely at the traded BDCs, and so that I think leads to a little bit more use of unsecured bonds. But there's also securitized debt, so there are collateralized loan obligations (CLOs) that the equity in the CLOs is owned by the BDC, and they're funding with CLOs as well. So, the composition of the assets of the BDCs is loans. Large, senior secured loans are the main asset being held by BDCs, but there is also equity, and then there's also equity in CLOs.

We don't have direct data on the portfolio firms. We know the terms of the financing, we know the name of the company and what industry they're in—which we use in some of our analysis, but we don't have the underlying capital structure of those firms. But there have been surveys of this, and the typical BDC portfolio company has a debt-to-EBITDA ratio of 6.4, and has the equivalent to a B- credit rating.

So, one of the things that we do is—a starting point is—would BDCs be considered well capitalized, according to bank capital frameworks? This is trying to address the question of: Is risky credit migrating to more leveraged entities? Basically, at a first pass you would answer "no." On a non-risk adjusted basis, or non-risk weighted basis, the leverage of BDCs is, there's 50 percent capital backing the assets of the BDC.

But the typical BDC has assets that are riskier than a typical bank, and so what we do is use the standardized approach to assess the capital of these BDCs—and we're looking at loans getting a 100 percent risk weight. There's equity in private companies, which would get a 400 percent risk weight; CLO equity, a 1,250 percent risk weight; and BDCs are also providing loan commitments—some of which may be undrawn—which then get a 50 percent risk weight.

BDCs, importantly, use fair value accounting, whereas banks use historical cost accounting. So, we have to do some work to try to do an apples-to-apples comparison, and so we adjust the fair value accounting to an estimate of the amortized cost and how that would affect the measured equity. And then we have a calculation of the allowance for loan and lease losses, to, again, try to put it in an apples-to-apples comparison, and the mean, risk-weighted capital ratio would be 36.4 percent.

So again, the typical asset of a BDC is going to be riskier than a bank, and this leads to a capital ratio of 36.4 percent, given they have 50 percent equity-over-asset ratio. And the other thing you could do is you could say the standardized approach is limited. It's giving a 100 percent risk weight to a set of loans to risky middle market firms that would do poorly in a stress scenario, and so what we then do is look at BDC capital using the stress testing approach.

And so, we look at the portfolio, estimate portfolio loss rates based on industry composition, whether it's a loan equity, equity in a CLO; we'll calculate what the pre-provision net revenue is over time, what the net loss rate is, and then an ending of period stressed capital ratio. And when you do all of that, what you end up with is a mean stress capital ratio of 30 percent, with the interquartile range of basically between 20-40 percent.

So again, under the stress testing approach—and we're using the Fed's 2023 stress testing methodology—what you see is that the BDCs do quite well. So, the idea that, again, that capital—that this activity—is moving outside of the banking sector to riskier entities, or entities that are funded in a more risky fashion, doesn't seem to be right.

Okay. So the next thing we do is try to understand, though, a bit more about why banks are not interested in making loans on balance sheet—or not making loans on balance sheet, and so what we do is a simple exercise where we compare middle market lending on a bank's balance sheet to lending to private credit funds.

BDCs receive a fair amount of their funding from banks in the form of credit facilities. Middle market lending, on average in 2023 Q2, was paying SOFR plus 600 basis points; those spreads have tightened. We estimate an expected loss of 160 basis points, given the average credit rating—or assumed credit rating—of a portfolio company.

Under the stress testing methodology, a generous capital allocation would be something like 20 percent capital. The funding costs we assume are roughly SOFR plus 55 basis points, so we're thinking about the wholesale funding costs of a bank. And then we estimate operating expenses of about 1.4 percent of assets, based on the operating expenses that alternative asset managers report in their 10-Ks and 10-Qs. So that's an estimate, and we get an estimate of return on equity (ROE) of 14 percent for that activity.

And then we contrast that to making an over-collateralized loan to a special purpose vehicle (SPV)—so this is the way the loans to BDCs and private credit funds are structured by banks, is that they're typically over-collateralized loans to an SPV, with an advance rate of somewhere between 60-75 percent for senior secured loans. This qualifies for Social Security Fairness Act (SSFA) treatment as a securitization, and for reasonable parameters will typically get a 20 percent risk weight. That's contrasting with, on average, a much higher risk weight for lending to middle market firms.

The average loan to a BDC or private credit fund, is about SOFR plus 230 basis points—with, I would argue, a de minimis expected loss, given the overcollateralization and low operating expenses, which we estimate at 0.2 percent of assets. So, banks have a choice of making a loan where they can earn SOFR plus 230 basis points with a 20 percent risk weight and de minimis expected loss rates, versus making a loan at SOFR plus 600 basis points with expected loss of 160 basis points and considerably higher operating expenses, given the need to identify, amongst thousands of potential borrowers, who wants funding versus a more narrow set of private credit funds that would receive financing.

And when you take into account the cost, the loss rate, the capital requirements, we estimate an ROE of 33 percent. So, that's an argument that actually, banks are happy lending to private credit funds, and that what we have is an ecosystem that's evolved where BDCs, private credit funds, are lending to middle market firms bearing risk in less leveraged capital structures, and raising funding from banks.

And I would argue that what's happening here is that...what is the bank's edge? The bank's edge, its competitive advantage, is in raising cheap financing, and therefore it would want to use that cheap financing to maximum extent with relatively little capital and relatively little risk, and therefore lending to the private credit fund is optimal in that framework.

There's a second question, which is not just why do you prefer to lend to banks, but when we see some banks engage in private credit or in lending to sponsored middle market firms, it actually takes—and there's been a number of announcements about partnerships between banks and alternative asset managers, private credit funds—it takes the form of lending off balance sheet, not on a bank balance sheet. What we observe is, when we talk about banks getting into private credit, it's not lending to middle market firms on their balance sheet, it's lending off balance sheet, through asset management arms or through private credit funds.

So, that's a bit of a puzzle, given the advantage that banks have in funding; they can use more leverage at lower cost to make these loans. On the other hand, if you do it through an asset manager, there is an advantage because there's no double taxation—these are flow-through entities, a lot of the holders are tax exempt—and there's lower regulatory and supervisory compliance costs. So, what we do is an analysis comparing those two strategies of lending to middle market firms, and our estimates suggest that as long as the regulatory and supervisory compliance costs are more than 50-100 basis points, banks are going to prefer sponsoring private credit funds rather than lending on balance sheet.

And what we have in mind with the supervisory and regulatory compliance costs, there's the actual cost and bureaucratic cost of complying with regulations and supervision, but we also have in mind limitations on the kinds of assets and kinds of activities that banks can undertake, on balance sheet versus doing it off balance sheet. Alternatively, you could argue that if asset management fees are 50-100 basis points too high—that is, there is actually negative alpha—banks will prefer sponsoring private credit funds rather than lending on balance sheets. So, there are two alternative explanations that could explain these facts.

Okay. I'm going to come back to that. All right. So, that's an analysis that, to us, helps explain why we've seen a growth in private credit, rather than lending by banks on balance sheet.

Now the next part of the talk, I want to talk a bit about potential financial stability concerns of private credit. Here I've just listed six possibilities; one is that banks incur losses on loans to BDC. We think that those risks are remote, given the very high overcollateralization of these loans. A second possibility is that banks incur losses on loans they also make to BDC portfolio companies. A third, which we're going to analyze in detail, is deleveraging—that is, that violation of regulatory leverage limits—that is, asset coverage minimums—and financial covenants forces BDCs to reduce lending and liquidate assets.

A fourth possibility is difficulty rolling over debt, forcing BDCs to reduce lending and liquidate assets. We just note that only 11 percent of total debt matures within two years, during our sample period.

A fifth possibility is redemptions by equity investors. About a third of the publicly traded BDCs are closed-end funds, so there really isn't any redemption issue. But a third—about $100 billion—are in perpetual BDCs, which do offer quarterly 5 percent liquidity redemption possibilities, and that could put pressure on selling assets or reduced lending.

And then finally, there could be a spike in portfolio company defaults and distress that leads to negative macroeconomic spillovers.

Okay. So, our focus is on this deleveraging piece. The basic point here is that the Investment Company Act of 1940 requires BDCs to maintain 200 percent asset coverage ratio. Again, in 2018 that was reduced to 150 percent, and most BDCs have moved to that level. They can't incur additional debt or pay dividends if this asset coverage ratio is violated—and if you can't pay dividends, it jeopardizes the regulated investment company status for corporate tax purposes.

Bank loans also have financial covenants, and some of those might even be more restrictive; so, what we do is apply a dynamic stress testing methodology to measure asset coverage ratios under stress. And so, our analysis has three key ingredients: first is a macro scenario, which builds on the Fed's severely adverse scenario. A second is portfolio valuation, given the Fed's severely adverse scenario. So, the idea here is that as you get into a stress scenario, credit spreads widen, defaults increase—that reduces the fair value of the assets, and then forces the BDC towards bumping into that asset coverage constraint.

And then we need to make some assumptions—behavioral assumptions—about how BDCs stay in compliance with these ACR minimums. And so, we do a simulation of what would happen, and what you see is that the asset coverage ratio declines initially, and in order to stay in compliance with the asset coverage ratio, to stay away from that 150 percent, there can be deleveraging that takes two forms. One is that as loans are repaid rather than reinvesting those and lending to other companies, the BDC uses it to pay down debt. And then a second possibility is as you bump into that constraint, that you're actually forced to sell assets.

And in this baseline scenario, we find that there could be aggregate asset sales of about 5 percent, and free cash flows equal to 4 percent of assets used to repay debt. So, the BDC on average is 9.5 percent smaller than it otherwise would be. Clearly, there are lots of different assumptions one can make, lots of different scenarios; and, we're doing that.

So, this assumes that BDCs maintain their debt structure; we're not modeling the ability to refinance maturing debt, we're not modeling redemptions from perpetual non-traded BDCs, we're not accounting for lines of credit and undrawn commitments that BDCs extend to portfolio firms. That is, there could be drawdowns by portfolio firms of these undrawn commitments, that then forced the BDCs themselves to draw down on their credit facilities; that's another thing that we want to be looking at.

I would say we're making some very strong assumptions about how fair value responds to an increase in credit spreads. We're using some pretty aggressive mark to market, based on what happens in the broadly syndicated loan market—and if you look at how BDCs mark their portfolios, I think they market a little less aggressively than what happens in the broadly syndicated loan market.

And then, of course, we're only modeling one scenario. There could be more rapid defaults, and a stagflation scenario could be more challenging to portfolio companies.

Let me just conclude with some concluding thoughts, which is that the growth of private credit I don't think is easily explained by a standard regulatory arbitrage story, that "shadow banking narrative" I mentioned. I think private credit is fueled by access to bank funding, and we argue that it's more attractive for banks to lend to private credit funds given the favorable capital treatment spreads and lower origination costs.

Compliance, including leveraged loan guidance, could be part of the explanation. But more fundamentally, what I see is that the growth of private credit (and the growth of bank lending to these private credit funds, and to non-bank financial institutions more generally), suggests to me that banks don't particularly—at least, in certain sectors—likely do not have an edge in originating risky loans, given high supervisory compliance costs, given a lack of focus. But banks do have an edge in raising low-cost funding, and thus have incentives to make safe loans to private credit funds and other NBFIs that have an edge in originating risky loans.

The view from a financial stability standpoint is that I think the risk is limited to banks, but the broader risk to the economy from deleveraging is worth further analysis. Thank you.

W. Scott Frame: Thank you, David. I definitely recommend this paper; I learned an awful lot reading it. I think you clearly articulate the risks in your analysis, and it's also a very accessible paper. You don't have to be a PhD to follow along with it, so I definitely recommend that.

While we're waiting here for questions to come in, I just wanted to ask a couple of follow-up questions about the analysis. I was struck by the fact that you estimate the profitability of banks lending to a private credit fund—in this case, a BDC—was on the order of 30 percent, which is extraordinary. So, I guess a natural question for me was: What do you think this might reflect?

Scharfstein: So, I think there are a couple of possibilities. One is that there are risks that we're not modeling. One in particular is that these are not just bank term loans, but they are credit facilities with the option for BDCs to draw down on those facilities, particularly during times of stress when liquidity is at a premium. And so, there may be a sense in which, if you properly risk-adjusted, that that 30 percent looks a little less ... on a risk-adjusted basis, is lower.

A second possibility is that banks and large banks are in a privileged position in terms of being able to offer this liquidity, and so there may be limited competition that leads to those spreads. I think it would be interesting to see how those numbers evolve over time. We've certainly seen increased competition in private credit, leading to very compressed spreads in loans to portfolio companies, and it would be interesting to see if we observe that on the loans by banks to private credit funds.

Frame: Now, your financial stability analysis focused on credit losses arising during an adverse economic environment—I think you used the 2023 Fed stress test scenarios—and then, the mechanics of that was to explore how this might result in BDCs deleveraging in the face of contractual and statutory leverage constraints, and the paper describes the risk to bank lenders to BDCs as being remote, and we touched on that briefly. I was wondering if you could flesh out a little bit more the reasoning for that perspective.

Scharfstein: Sure. So, these credit facilities that banks are offering to BDCs are over-collateralized—meaning that, every $70 or $65 that they're lending is backed by $100 of collateral. It's in a diversified pool of loans across industries. There are, for sure, industry concentrations; you would need a fair amount of correlation in defaults, with lower recovery rates, to get into a default scenario on those bank loans.

I think if you look at the attachment points for loss in a loan from these credit facilities, relative to, say, a AAA tranche in a CLO of broadly syndicated loans: it's similar, and I think we know that AAA tranches and CLOs have I don't think ever been pierced. So, I think there's a reasonable amount of confidence about this, but time will tell.

Frame: I'm going to turn to the audience questions, and I noticed a couple here that I think touch on the supervisory and regulatory costs and helping us flesh that out a little bit more. We have one question from Margaret Tahyar: "Could you please expand on how the leveraged lending guidelines are part of the explanation here?"

And I know Sergey, in an earlier paper, had done some thinking about this.

Scharfstein: Yes. So, Sergey, my co-author's professor at Purdue had done some work looking at this and found that there was that leverage guidance around loans with a debt-to-EBITDA ratio in excess of six, and found that where you hit that threshold, the lending went down dramatically on the part of banks. My understanding—and I'm not a regulator, supervisor or lawyer—is that this was guidance around risk management practices associated with leveraged lending.

I think what we're saying, without getting into the details of the specifics of how that played out, our estimates are saying if compliance costs with making those risky loans are on the order of 100 basis points, then you can explain the movement; you can explain why banks would rather do this through a private credit vehicle rather than on the balance sheet. And I think that those compliance costs could be a combination of actual cost of compliance and the work that it takes to comply with the regulations on risky credit, but it could also be limitations on the types of loans so that there is, let's say, a loan that has a debt-to-EBITDA ratio in excess of six, which is the median loan here, and you think twice about that and it may be a profitable opportunity for a lender, but in the bank framework, it doesn't really work.

Frame: Beth Hammack asked a question that was related to that, which I think you've answered now. She also, I think, was pushing a little bit on the assumption, and she asked: "How sensitive is the ROE calculation to the expense ratio, and how is the bank expense ratio on secured loans so much lower than to risky companies?"

Scharfstein: I think there are two things. One is, there's the origination of the loan—so, it's identifying who that borrower is. Private credit funds, alternative asset managers, have built up platforms to—and relationships with private equity sponsors to—identify lending opportunities. There's a due diligence process, there's an underwriting process to making those loans, and they're generally fairly small loans.

And if you contrast that to a relationship with a much smaller number of alternative asset managers that a bank would have to have, the scale of those loans is much larger. The risk is much lower, so the underwriting process is less costly. I think there's a pretty significant difference in the operating expenses of those two activities.

Frame: Okay. Constance Hunter had an interesting question here; she makes the point that these private credit funds, or this intermediation activity has been around for a long time, and what do we think is causing the significant growth here just in the last three or four years?

Scharfstein: Right. So, first of all, it is true that private credit—non-bank credit—has been around for a long time. I do think it's worth noting that it existed pre-GFC, but in a very different form. So, you had GE Capital, CIT, Heller Financial—these were essentially private credit, but they were done in structures that were much more leveraged with much more short-term funding (and the outcome, of course, was not a pretty one).

Private credit has been through a fairly benign economic environment; returns have been strong, in a lower interest rate environment, and have looked good. It is difficult to measure the risk of private credit. There are some analyses that look at risks in net asset value in reported values of private credit, and so those analyses tend to, I think, understate the correlation risk with the market, underestimate its beta with the market, because they kind of smooth values.

So, I worry a little bit that there is an underestimation of risk, with high returns relative to benchmarks that the industry uses that suggests that; and so, it is possible that there's more flow into private credit than is warranted by the actual risk-adjusted returns. But it's hard to know because it is hard to estimate risk in these funds, and to know whether or not there is excess returns.

Frame: We have a question from Sai Srinivasan. You touched on this a little bit at the outset, but he asked: "Given that BDCs account for only about 20 percent of the private credit ecosystem, are there any concerns or limitations around extrapolating the analysis from BDCs to the broader private credit system?"

Scharfstein: So, the large asset managers will sponsor BDCs and private credit funds that don't meet the BDC structures, and my discussions with folks in the industry is that they almost do a pro rata allocation of loans across those different funds, and so I think it gives a pretty good measure of what is held in the portfolios of non-BDC private credit funds. There are surveys of the leverage of private credit funds, and if anything they're lower than they are for BDCs; less reliance, I think, on bonds, but I think it generally holds up.

Frame: Andreas Lehnert asks—and this speaks to your financial stability analysis—he asks: "If BDCs really have negative alpha, is there a broader financial stability concern that investors might pull back from the sector as a whole at some point that could create a credit crunch for private credit borrowers?"

Scharfstein: Yes, I think that is possible. And first of all, I don't know what alpha is, if it's positive, negative, 0—again, it's hard to measure that. I think one of the things we see in CLOs in the broadly syndicated loan market is that they're generally fairly durable; they do okay through stress, but formation of new CLOs basically dries up during stress periods. So, I would anticipate something similar in the private credit space as well, that if we go through a period of very significant stress, that there would be less formation of private credit funds, and then that you'd have the kind of deleveraging going on for existing funds, of the sort that I've modeled.

Frame: So, Ian Harnett asked, following up on this: "Are there any additional risks that you might consider as part of a stagflation-type scenario?" I know we've been discussing this, and whether you have any thoughts about the nature or scale of these risks that might be unique in a stagflationary scenario.

Scharfstein: Right. So, that's actually something we're working on now—in particular, what the effect of tariffs would be. Obviously, as was mentioned earlier this morning, there's a tremendous amount of uncertainty; a large share of the BDC portfolios are in software, business services, health care—the kinds of firms that private equity firms invest in. If I had to guess, just without looking at the data, I would say that those industries would be less directly affected; but of course, across the portfolio there would be firms that would be more directly affected, as well as just generally if we were in a recession scenario.

The stress testing analysis that we did, and the deleveraging analysis we did, assumes that the Fed steps in and lowers rates; that relieves a lot of pressure on firms, which are borrowing on a floating rate basis as a spread to SOFR. If the Fed doesn't step in and lower rates and we are in a stagflation situation, then obviously that creates more stress on the portfolio companies. And separately, the BDCs tend to benefit from a rate increase, everything else being equal, because they're making floating rate loans and they have some fixed-rate liabilities, so their income tends to be asset sensitive—it goes up when rates go up.

But I would think that the bigger effect would be stress on the portfolio companies. As I said, that's something that we're looking into right now.

Frame: We had a question from Roy Henricksson, asking: "In the data that you used for BDCs, and you discussed (briefly) the availability of some information about the specific firms that are creditors to these funds; can you see through to any of the information about covenants in those agreements, or do you just see the names, and maybe their industry code, or something like that in the data?"

Scharfstein: We see names, the interest rates that they're borrowing at, their seniority in the capital structure—but we don't see specifically the covenants.

Frame: And it's I think a natural question in this environment, because I think one of the appeals, as I understand it, of private credit is you can negotiate some of these covenants.

Scharfstein: And that is one of the reasons that private equity sponsors—portfolio companies—go to private credit, is you're dealing with one or a few lenders as opposed to the broadly syndicated loan market—which is also for larger firms, but where it is much harder to renegotiate. And so, there is concern at times that private credit funds are kicking the can down the road and maybe a little too lenient. On the other hand, it does allow for more flexibility in a stress scenario.

Frame: One question I had, as you talked about this stress scenario, is that my understanding is that BDCs allow for some limited, periodic redemptions. Did you have any thoughts about, in that case, applying that to your financial stability analysis to get a sense of: would that amplify the channel that you were discussing?

Scharfstein: It's relevant for perpetual BDCs, which is about a third of the sector. And yes, if you can do a quarterly redemption, I think the board has the option to not allow that—to gate it. But if they allow it, and there's a redemption of 5 percent, and you're bumping up against the asset coverage ratio, that could lead to asset sale or less lending. So, it would increase the effect, to some extent. I think we're making fairly strong assumptions already about what the fair value reduction would be. But yes, I think it would, for a limited set of BDCs, have an impact.

Frame: Okay. Well, thank you very much, David. Please, join me in thanking David Sharfstein from HBS. We're going to adjourn right now, and I believe the next session will start at 10:50. So, go ahead and refill your coffee cups, and come back in by 10:50.