Worst Case Scenario, or Inconsistent Scenario? A Reply to Frank Partnoy
On Thursday I wrote a very critical reply to Professor Frank Partnoy’s feature article in the Atlantic entitled “Will the Banks Collapse”. My article was subsequently linked to by Matt Levine in his Bloomberg newsletter, in Marketwatch and at the American Prospect and went somewhat viral. As of this writing, the article has over 24,000 unique views which is quite a lot for a substack post. Yesterday evening, Frank Partnoy wrote out a long reply on twitter. I’m thrilled to be a part of an extended back and forth about a financial and macroeconomic issue with long form writing on the internet. This is especially thrilling because I was an avid consumer of macroeconomics and financial crisis debates which sprung up in the econoblogosphere a decade ago. Participating in such a back and forth feels like a throwback to a classic era. I greatly appreciate Professor Partnoy’s and am glad that my critical analysis provoked such a substantive response.
First, there is the question of “facts”. Professor Partnoy says:
I’ll also note upfront that, although you disagree with some of my points, you don’t dispute any of the facts I report. I have been writing regularly for @TheAtlantic over the years, and their level of rigor in checking facts is phenomenal.
Partnoy is correct that my piece doesn’t dispute any facts he stated in his article. Instead, I focused on his framing and presentation of those facts. However, as I state below, I disagree with the interpretation and presentation of these facts so fundamentally that I still think they fall within the realm of fact checking. I do not think that popular press pieces should present Collateralized Loan Obligations as similar to Collateralized Debt Obligations when the unique and dangerous feature of Collateralized Debt Obligations is that they were securitizations of already structured products and not securitizations of underlying loans directly. Partnoy states in his feature article that:
The reforms were well intentioned, but, as we’ll see, they haven’t kept the banks from falling back into old, bad habits. After the housing crisis, subprime CDOs naturally fell out of favor. Demand shifted to a similar—and similarly risky—instrument, one that even has a similar name: the CLO, or collateralized loan obligation. A CLO walks and talks like a CDO, but in place of loans made to home buyers are loans made to businesses—specifically, troubled businesses
It is simply not the case that these are similar instruments and it is certainly not the case that they are “similarly risky”. A CDO made up of lower tier parts of a Mortgage Backed Security which will completely implode if default rates rise to the low double digits is just not comparable in risk to a CLO. Mortgage Backed Securities are the proper analogue to Collateralized Debt Obligations and if Partnoy’s piece were edited to reflect this, a large degree of its rhetorical force would be weakened and the premise of the article may not even hold up to scrutiny.
Next, Partnoy addresses the issue of Credit Default Swaps:
We also agree that “some of the most irresponsible gambles from the last crisis—the speculative derivatives and credit-default swaps you may remember reading about in 2008—are less common today ….” That’s what I wrote, it’s a common view, and another of your main points.
I do not believe we agree on this point, at least based on the article. I didn’t claim that CDOs, synthetic CDOs and credit default swaps on those CDOs were “some of the most irresponsible gambles”. I claimed that they are what defined the financial crisis of 2008 and are what made it a global financial crisis. Without these elements, CLOs just aren’t capable of matching the 2008 crisis to any degree. This is a point that’s widely accepted in the financial press. Yves Smith’s early Great Financial Crisis book Econned, established this in greater and more precise detail than any other writer up until now. However, this point is widely accepted. It is even stated as a fact in the movie based on MIchael Lewis’s book, The Big Short. Now, it's still possible for Collateralized Loan Obligations to see very large losses even if they are not similarly risky to Collateralized Debt Obligations because of truly extraordinary default rates. But if this happens it will still be a very different crisis from 2008 because the central feature of that crisis was trillions of dollars of losses that emerged from a much smaller quantity of underlying individual loans with low double digit default rates. A crisis with 50% or more default rates is qualitatively different in fundamental ways and such a crisis, as we’ll discuss more below, would be widespread across loans and not concentrated in structured financial products this time.
Professor Partnoy than comments on the Federal Resrve’s liquidity interventions
We agree about the Fed’s willingness to provide liquidity. You don’t address the hotly disputed question of statutory authority for Fed lending programs. Many experts question 13(3) authority for the TALF term sheets, based on Dodd-Frank. (I have a thread here on that.)
It is not clear to me who Partnoy is referencing when he refers to “many experts” and in general it is difficult to assess claims that rely on unspecified experts. One possibility is that it's a reference to Lev Menand’s excellent paper “Unappropriated Dollars: The Fed’s Ad Hoc Lending Facilities and the Rules That Govern Them”. However, Lev doesn’t argue that TALF doesn’t currently have statutory authority, as he explicitly says on twitter:
Completely agree w/ @DanAwrey on 13(3). The Fed's new facilities plainly comply w/ the CARES Act, which is more recent and specific. My complaint is with Congress for not doing a better job of clarifying the scope of the Fed's authority post-CARES Act / amending 13(3) directly.
The thread I found from Partnoy which references TALF points to Dodd Frank’s direction to remove references to credit rating agency ratings in existing regulations. This is an interesting and important point, however I don’t see any evidence that directing agencies to remove administrative rule-making references to credit ratings amounts to a statutory ban on regulatory agencies discretionarily relying on credit ratings in their lending or purchasing decisions. Even if it did, that would not eliminate these facilities and may in fact make them more broadly accessible and equitable.
Most fundamentally, none of this alters the fact that liquidity freezing up for a single bank because solvency concerns emerge is not something that will happen today like it did in 2007-2008 because the Federal Reserve’s interventions have been fundamentally different and far faster. This part of Partnoy’s “worst case scenario” just is not credible given the Federal Reserve’s current operating framework.
Finally, Partnoy moves on to our disagreements:
4 – We disagree about whether “the losses from CLOs, combined with losses from other troubled assets like those commercial-mortgage-backed securities, will lead to serious deficiencies in capital.” I wrote that, and many experts worry about cumulative losses; you are skeptical.
I focused my analysis on Collateralized Loan Obligations because they are the central focus of Partnoy’s piece. If Partnoy’s piece had focused attention on “other troubled assets”, I'd have been more sympathetic to his conclusions, though would question whether poor underwriting was the central issue with these “troubled assets”. However, the focus of his analysis was not those other assets. There is no evidence in the piece that bank holdings of CLOs are the most concerning parts of their asset holdings. Which brings us to the question of how to represent the size of CLO holdings:
– We disagree about whether leverage loan/CLO size is significant for the big banks. You say the total is “just” 60% of Tier 1 capital. I think 60% of Tier 1 is a big deal. I wouldn’t say “just.” Also, experts and regulators compare based on capital, not assets, given leverage.
Here, Partnoy misquotes my reply. My use of “just” in that paragraph was meant to convey just how tiny CLOs are in terms of the balance sheets of the major banks. It’s worth quoting that section at length:
Returning to the OCC report, we find out that Wells Fargo has more than 1.7 trillion dollars in assets. 29.7 billion is about 1.7% of Wells Fargo assets. CLOs are similarly minor for the other large banks. Their investments are also restricted to the top tranches, which as we explained earlier really do have a very substantial cushion of losses. Even in the extreme scenario where every single CLO experiences 100% losses, the exposure of the Global Systemically Important Banks is minor. Even Partnoy acknowledges that their total holdings of leveraged corporate loans overall amount to just 60% of tier 1 capital. High yield corporate debt is certainly a financial risk for the banking system but that story is far more boring and conventional than Professor Partnoy wants to present it as.
As I said, leveraged loans as a whole are significant but Collateralized Loan Obligations are not. Nowhere does my piece claim that leveraged corporate loans are not a financial stability concern for the banking system. That said, it is important that even complete losses on CLOs and leveraged loans wouldn’t wipe out Global Systemically Important Bank’s tier 1 capital. Next, we have the question of whether total assets are useful for comparisons.
The more leveraged a financial institution is, the less the default rate has to rise in order to wipe out a bank’s equity. However, the default rate on a specific set of assets is only relevant to that overall question of capital adequacy if those assets are a significant percentage of total assets. The nominal value of a specific set of assets is only usefully comparable to equity if you expect almost total loss on those assets and only those assets. Professor Partnoy’s article clearly makes the case that this extreme scenario is relevant. However, a major overarching point of my piece is that that extreme scenario is only plausible when it is implied that top tranche CLOs are like subprime top tranche CDOs. I’m claiming that this is an apples to oranges comparison. Without this critical analogy, that extreme scenario which involves unique losses for CLOs is not relevant and thus contextualizing CLO holdings relative to total assets in the banking system is a valid and important form of contextualization.
The next major issue that Professor Partnoy brings up is the question of “default correlation”
We disagree about the worst-case. The key is default correlation, and experts have a range of views. I link to the leading academic article showing correlation assumption flaws. People should simply look at leveraged borrower names to make judgments about correlation.
The linked article that Partnoy references does indeed provide a convincing case that rating agencies are underestimated default correlations, which is the likelihood that a specific portfolio of loans will experience default events at the same time. However, the article also argues that market prices have reflected investor belief that default correlations are higher than rating agency models suggest which means higher default correlations were already “priced in”. The article’s central premise is that many of these securities don’t deserve triple AAA ratings. This argument is a long way from the apocalyptic vision that Partnoy presents. Regardless of their rating, the top tranche of Collateralized Loan Obligations takes losses only once the lower tranches have absorbed high levels of losses. There will have to be truly gigantic levels of defaults before AAA CLO holders see significant losses.
Further, it’s important to grasp that to the extent to which the top tranches of CLOs are threatened on the scale that Partnoy speculates about, no other portfolio of loans would be safe. As I said in my original reply:
This structure may encourage originating lower quality loans and investors may overpay as a result of tranches being mis-rated by credit rating agencies. However, it can’t be said that this structure, in and of itself, magnifies losses for investors in the senior tranche. In fact, it works as intended- other investors bear the bulk of the risks.
In fact, it’s likely that collateralized loan obligations made up of the top portion of a portfolio of loans will do the best of any of the Bank’s corporate loans. When a bank originates and holds a loan, it immediately takes losses if the borrower defaults. In contrast, it takes very substantial levels of defaults before holders of the top portion, or “senior tranche” of a collateralized loan obligation realize any losses. For any given portfolio of loans, given the same overall default rate and the same loan modifications, senior tranche holders will experience less losses than outright holders of a portfolio of loans. This structure may encourage originating lower quality loans and investors may overpay as a result of tranches being mis-rated by credit rating agencies. However, it can’t be said that this structure, in and of itself, magnifies losses for investors in the senior tranche. In fact, it works as intended- other investors bear the bulk of the risks.
If all these different businesses are failing at the same time, it's because of a lack of aggregate nominal income because of the Coronavirus Depression and the inadequate response from government policymakers. The entire portfolio of business loans will be under threat, not simply securitized products. The issue of “default correlations” in the Coronavirus Depression is very distinct from the 2017 concerns about default correlations. It is impossible to “price in” extraordinary events like this as they are uncertain, not risky. Attempting to center the problem of today’s crisis in poor underwriting by lenders is a category error.
My problem with the article is not the assessment of worst case scenarios. My problem is that the article bends over backwards to uniquely put worst case scenarios on Collateralized Loan Obligations relative to all other bank assets which I think is unwarranted and the motivation seems to be to connect this crisis with the fraudulent practices of the 2008 crisis which is unsupportable. I think the possibility of mass corporate- and household- defaults is an important worst case scenario to assess and the consequences are indeed potentially devastating. The issue is that that worst case scenario is not grounded in poor underwriting but the complete collapse of nominal incomes across the economy as a result of the shutdown. No amount of underwriting can protect from such catastrophic collapses in spending and incomes. In fact, a major justification for sustained government deficit spending in recessions is precisely that the private sector cannot insure against such events without stockpiles of government safe assets which can only be generated by big fiscal deficits. As I say in my original reply:
Losses from this crisis may lead to “serious deficiencies in capital”, but if they do it will not be because of fancy structured products but the failure of good old-fashioned loans because of a good old-fashioned depression
In 2008, the macroeconomic issues flowed out from the practices of individual financial institutions and the rigged system that they set up. Today, it is macroeconomic dynamics which are leading to risks and uncertainties for individual financial institutions. Professor Partnoy’s piece fundamentally reverses the causality in its heroic attempts to find a culprit.
Finally, we reach Partnoy’s final points:
We disagree about likely leveraged loan loss recoveries. Many experts say recovery assumptions are much too high, particularly given the crisis (and could approach zero). Reasonable people disagree - it's a prediction. We'll see as defaults increase … I gave my worst-case scenario to several experts and asked them to put a probability on it. A two-SD worst-case event would suggest 5%, which is about what I think (and is std for VAR). But the expert consensus was near 20%. What % would you put on it? Maybe we agree.
Partnoy’s article suggests that covenant lite leveraged loans will experience 80-90% losses- as opposed to 30-35% losses on defaulted loans as a result of less loan covenants. There is simply no evidence for this assertion and he does not proffer any. My argument that this is a “maximalist” assumption does not rely on current industry recovery assumptions just as my argument about the top tranches of CLOs doesn’t rely on defending their triple A ratings. This is another area where he is attempting to ground today’s macroeconomic issues in poor loan underwriting in a way that is unwarranted. If default events are leading to total losses on such a widespread basis, there is likely a much more systematic macroeconomic cause that is underlying these issues and not a microeconomic question such as loan covenants.
This leaves us with the question of the “likelihood” of his “worst case scenario”. It’s hard to know what to make of this number without context. Which experts were asked? What precisely was the worst-case scenario described to them? All I can say is that the threats to bank balance sheets are macroeconomic and thus apply across the assets in their balance sheets and are not concentrated in structured financial products this time around. Leveraged corporate loans may see default rates rise before other bank assets but AAA CLOs will be the least impacted. However, it's much more likely that small and medium sized business loans will experience, as a category of loans, problems first. Most importantly, the lack of liquidity issues for the banking system this time around and the forbearance to borrowers that regulators are already encouraging means this will be a slow moving, extended problem rather than a fast moving crisis. I would be interested in where the experts Partnoy spoke to would point to as the source of a sudden and sharp banking collapse.
I’d like to thank Professor Partnoy for his extended and detailed reply and look forward to more discussions of this issue which is certainly not going away.