[manyhats]

1. It's important to distinguish between liquidity risk and solvency risk. Liquidity risk is needing to sell an asset that is fundamentally sound and not being able to get a fair price, or any price at all (as happened in 2008/2009 - no/few buyers and many sellers). Solvency risk is the asset goes bad and there's a permanent impairment.

Simplistically:

liquidity risk = broken leg (painful and needs immediate treatment, but recovery over time likely)

solvency risk = horrible cancer

2. CLOs are made up of bank loans, which are generally the most senior obligation of a company, and are the least risky. Bonds, preferred equity and common equity get hit before the loans are impaired. During the period from 1998 through 2016, defaulting loans recovered 66% of their value while bonds recovered 40%. [0]

3. CLOs generally made it through 2008/2009 without any major issues. The structures held up as designed, but prices fell along with other structured products since there were few buyers and many sellers.

If you were able to hold, you did reasonably well. Over the 20 year period from 1994 through 2013, there were 6141 CLO tranches, of which 0.41% defaulted (no AAA or AA defaults) and had a 0.04% loss rate. [1]

4. What has changed since 2008 is that loans are becoming a larger portion of the total value of the business (loan-to-value). This means that if the business goes bad, there's less of a cushion for the loan and recovery rates will be lower.

I don't know how much lower, but some of the junior tranches in a CLO could be impaired in a severe downturn. It's hard to really know, but I _think_ things would need to be worse than 2008 (or the same level of crisis, for a longer period) for AAA tranches in general to be permanently impaired.

5. Accounting plays a role here. If you are required to mark-to-market, then the market clearing price is what the asset is worth, regardless of fundamentals. If there's a panic, you may be a forced seller if you don't have sufficient reserves to get through the disruption and your mark-to-market loss becomes a permanent loss. This applies to any asset class, not just structured products.

6. I don't understand why the article brings up CDS gaming. It's a real issue, and one that is being worked on, but it's not even a rounding error compared to the size of the corporate debt universe. [2]

quote from the FT article:

"Isda’s method of “fixing” CDS has always been akin to “patching” software after hackers exposed weak points. When one window closes, traders simply find a new one."

[0] "JPMorgan Default Monitor 4Q16"

[1] “Twenty Years Strong: A Look Back At U.S. CLO Ratings Performance From 1994 Through 2013”

[2] https://www.ft.com/content/efab718a-40c9-11e9-b896-fe36ec32a...

[o_nate]

These are good points. I would add that generally there's not a problem with having lots of risky debt in the system, as long as it's held by investors who are aware of and can handle the risk. The reason that CDOs led to the financial crisis, was that there was lots of risky debt being packaged as AAA and held by investors (i.e. systemically important banks and insurance companies) who couldn't handle losses on it. So the big questions about CLOs should be: who's buying the AAA debt, is it being mis-rated, and if so, can the holders handle losses on it. In general, I think the rating companies reacted to their vast and obvious failings in 2008 by tightening rating criteria across the board, even in structures such as CLOs which did okay during the crisis. Is it possible they're still being too optimistic? The thing about the rating models is that they're very sensitive to correlation assumptions that are difficult to observe empirically. Also, they tend to have a blind spot about things like fraud. Is it possible that junk-rated corporate defaults could end up being much more highly correlated than the models assume? There are transmission mechanisms that can cause high correlation. For example, an uptick in defaults in one sector could lead to a tightening of credit standards across the board that could lead to rollover risk. (Most of these companies are highly dependent on being able to rollover maturing debt.) Or fraud in underwriting standards could be much more pervasive than assumed. (This one seems unlikely, but who knows.)

[rsln-s]

Another good explainer on why CLOs are not like CDOs: https://www.economist.com/briefing/2019/03/16/should-the-wor...

Roughly speaking, corporate loans are way more transparent and there are fewer of them in a pool, so it's easier to assess the risk.