Authored by Peter Tchir via Brean Capital,
I continue to be focused on tail risk. While I think valuations may be stretched, my bigger concern is that the real risk lies in market structure – more specifically to volatility and correlation strategies that quickly propagate through the financial markets with the aid of algorithmic trading.
I will rehash and add to that argument, but I think it is worth spending some time re-examining prior ‘events’. I want to take a closer look at the premises that drove many of those prior episodes, looking for similarities.
CDS in 2006/2007
There were two main products that contributed to the financial crisis that came from the credit derivative side. One was more responsible for the rapid tightening of credit spreads and the other was more of a reason why they blew out so quickly.
For both, the fatal flaw was that Market Risk dwarfs Default Risk and they both hit ‘trigger’ events.
CPDO or Constant Proportion Debt Maturity
This was a product that became the rage in 2006 (as I remember it).
The product was packaged in structured notes and Special Purpose Vehicles (“SPVs”). They were rated AAA and paid a coupon of LIBOR + 150 for example when generic BBB+ credit spreads were around 50. This was an incredibly tantalizing proposition for many investors – AAA rated with more coupon – how could you resist?
Here is roughly what a $100 million investment would look like
- The SPV takes in money which is used as collateral to sell CDX IG CDS (a 125 name index that had an average rating of BBB+)
- You would initially sell 12 times the notional – i.e, a $100 million investment into the AAA rated SPV would generate selling $1.2 billion of the index (which was BBB+)
- If the spread on the index was about 0.5% per annum, the CDS selling would generate 6% per annum or 3% semi-annually
- At this point the SPV has two ‘assets’, the initial funds, invested in a high-quality LIBOR asset plus the CDS generating 6%. So the SPV is earning LIBOR+6% and only paying out LIBOR+1.50% - generating a buffer – this buffer played a key part in getting the AAA rating
Before going any further let’s just spend a moment thinking about the financial alchemy that LEVERAGES a BBB+ asset more than 10 times and calls it AAA. Investors could plunk down some money on a AAA asset and get a better coupon than was possible on a BBB+ asset – that was the driver. Every dollar of this investment generated a large multiple of selling in the CDS market – driving spreads tighter.
Now at the end of 6 months the index would ‘roll’ to the new index (CDX indices rolled semiannually). This is where the trade gets even stranger.
At the end of 6 months the SPV would buy back all of the old CDS index and sell some amount of the new ‘on the run’ index o There was an assumption that the new index would be at a higher spread than the old index because CDS curves were steep (i.e. 5 year CDS was typically more than 4.5 year CDS). The CPDO model gave some value to this roll effect.
If CDS spreads widened during the 6 month period, the SPV would sell MORE CDS at the roll. Once again, not a typo – if say the index widened from 50 to 60, the SPV would buy back $1.2 billion of the old index and sell $1.3 billion of the new index. The assumption was that the new index would be around 65 because of the roll – giving you even more carry – allowing reserves to build up more quickly to cover the loss.
The SPV locks in losses on the old index but the excess income offsets some of that loss (the buffer) and for the next period you get more income as the initial spread is wider and you have sold more index (the ultimate in doubling down)
This worked extremely well so long as there was mean reversion.
I think I forgot to mention that if the SPV ever dropped below a certain valuation, it had to terminate. It was an entity that had a limited amount of capital backing a much greater exposure of CDS. Since the counterparties selling the CDS to the SPV couldn’t get more money than what the SPV had – at some point they would terminate their trades with SPV – locking in losses.
The rating agencies made so many mistakes in these deals it is hard to know where to start, but mean reversion, steep curves and a focus on historical defaults rather than ratings migration all worked to sink these trades in an epic way.
- Credit spreads flattened – credit risk became a greater concern so the curve wasn’t as steep for the index – meaning the roll strategy was less interesting
- Credit spread widening in the index was driven by names that were downgraded below BBB- and therefore couldn’t be in the new index. So the new index was actually at a lower spread than the old index – because the names that came out were such a big driver of the spread. That meant it was harder to recoup losses. (the rating agencies had been right about the low likelihood of default in a 6 month period, but they totally missed the impact of downgrades locking in losses that could not be recovered)
- Each successive index had its own batch of names that would tumble – reinforcing the problem
- Credit spreads blew out far faster than modeled and did not mean revert
CPDO’s ‘triggered’ at losses of around 80% to 90% - probably some of the fastest losses ever on a AAA product. That caused massive buying of CDS protection – the exact opposite of what the trade had promised for the market.
So the alchemy of leveraging BBB and getting AAA, which seemed likely to provide a tidal wave of selling of CDS protection, unwound horribly and rapidly.
In some ways I apologize for dragging you through this product, but I think we will see how many of the themes apply to today’s market.
LSS or Leveraged Super Senior
Here at least we can ‘blame Canada’ as it was Canadian Conduits that were the largest participants in this trade.
The conduits could issue Asset Backed Commercial Paper (ABCP) and buy assets with the proceeds. The commercial paper was, by definition, short term in nature. The conduits could employ varying amounts of leverage depending on the quality of asset they purchased.
They too were left searching for the Holy Grail in a tight credit spread environment. They really needed to buy high quality assets. I think AAA CLO paper at the time might have been paying around LIBOR + 25 or something.
Along came CDS to the rescue. The synthetic CDO business was cranking on all cylinders. There were buyers of ‘first loss’ because the expected returns were so high. The rated tranches all looked cheap relative to anything available in the cash market, but this left an overhang of ‘super senior’ paper. Let’s call this the 15-100 tranche of the index. Basically if you referenced the index, this ‘standardized’ tranche had not credit losses until losses totally 15% of the index had accumulated. That is a MASSIVE amount of credit losses for a corporate bond index to have. It had never happened before and still hasn’t happened (CDX IG 8 or 9 has had the highest losses and I don’t think either of them made it to 3% in realized credit losses of a 10 year period). So no problem?
Actually, a big problem.
The super senior paid maybe 0.08%. Expected loss was 0.00000000000000% so the 0.08% was cheap but not many people are in the business of selling risk at that price. But now if we just take the super senior and leverage it say 20 times.
The risk of losses due to credit is still 0.00000000000000000% but now you have 1.6% of income to play with. Once you finish paying for structuring, rating and a bit for the house, you now can have an asset with low credit risk that pays 1% - blowing away anything else available in the market.
With 20 times leverage you could absorb a 5% loss on the super senior tranche before getting wiped out. Of course, the swap providers want out before they are at significant risk, so maybe the unwind trigger was struck at -60% or a 3% loss on the super senior tranche.
For the super senior tranche to lose 3% in value, spreads would have to move from about 8 bps to 60 bps – that seemed hardly possible when the raw index was trading at 50 bps. How would it be possible for the super senior tranche to trade so wide?
Ahhh, we forgot about rolling the ABCP. The conduit was funding longer term risk with shorter term assets. As credit spreads started widening, it didn’t take a genius to start getting nervous about lending to conduits to had extremely large notional exposures to CDS (there is always a point in credit where you can talk until you are blue in the face about actual risk, where people can’t get over notional exposure).
Once this weakness was exposed – the ‘lottery’ ticket of shorting super senior was exposed and quickly became a vicious cycle.
There are so many similar examples, like non-recourse total return swaps on leveraged loans, that it could take up the entire morning, but there was a whole series of products that took ‘safe’ assets and leveraged them, that initially provided the impetus for further tightening, but eventually became problematic in the unwind. In this example, not even the ‘mezz’ tranche ever got touched with credit default losses, let alone the senior tranches – but it didn’t matter.
ABX
I won’t spend much time on this, you can watch the movie, but there are a few fatal flaws that should be highlighted as a reminder.
The first flaw was in the residential mortgage backed security rating themselves. The assumption that home prices on a national basis would never decline was proven wrong. You can then argue over whether things link NINJA (No Income No Job No Assets) and sloppy portfolio construction contributed (they did) but the flaws around the resi mortgage backed market could have been contained.
It was the ABS of ABS or CDO^2 (CDO squared) that was the real problem.
- It allowed BBB assets, often the hardest to sell, to be repackaged into tranches, where the bulk of it was AAA again, making it easier to sell BBB risk and allowing the market to grow rapidly
- Problems in the original methodology were compounded by using ‘normal’ recovery rates for BBB tranches – when in reality – as many argued – they were so thin with so little subordination that you should awesome that if there was a loss at a BBB tranche – the entire tranche would be wiped out (they modelled ‘expected’ loss rather than thinking about the conditional probability that if there is a loss it would be the entire tranche)
- Then, the assumption that each BBB mortgage tranche was more or less independent of the others also turned out to be wrong – BBB tranches were under pressure universally – (they incorrectly modelled the correlation).
The sins here were relying on historical relationships and expecting them to continue into the future and ignoring the fact that the CDO squared model, which made it easier to sell BBB tranches, may have changed the underwriting standards.
Long Term Capital Management and Russia
LTCM and Russia hit at almost the same time. There are just a few things that I want to highlight about this period of weakness.
It almost started a month earlier. One of the LTCM models was that the high yield market constantly mispriced yield to call bonds. Since the bonds were high yield, investment grade buyers wouldn’t buy them even if they were almost certainly going to be called. High yield buyers didn’t really like to own them as they were a drag compared to the rest of the portfolio (this is when high yield still actually had high yields). My understanding is that LTCM owned almost the entire issue of Host Marriot Travel Plaza bonds and TelMex bonds. Those were the last two large re-financings that were done in the high yield and EM markets before the crisis hit (basically less than a month before). We will never know what would have happened to LTCM or the markets had the Russia crisis hit before those bonds got called away. So in this case their model ‘arrogance’ for lack of a better word paid off – but as we all felt later that summer, when it failed, it failed badly (as a side note, I’m not sure I like the current steady stream of money into short term high yield bond funds – SHYG and SJNK but that is for later).
Ultimately LTCM blew up, rather spectacularly when things like swap spreads failed to ‘revert to the mean’. LTCM was rather proud that their models often had them acting as ‘liquidity’ providers (the near certain callable high yield fits that definition more or less too).
So when things ‘broke’ ranges they stepped in and provided liquidity and were rewarded when prices mean reverted (maybe it was just the original buy the dip (BTD) capital). It did tend to be on slightly more complex products that offered high leverage (like swap spreads).
When the ranges broke and then broke again approaching ‘unbelievable’ levels, at least according to what the models had predicted, then the margin calls and unwinds started.
At the same time, we were hit with the Russia Crisis. That was another ‘crisis’ where in hindsight the market had plenty of warnings, but failed to react. It reminds me of the so-called ‘Lehman Moment’ which everyone remembers so clearly as being a trigger, yet it doesn’t show up in prices that week at all.
While different than the prior examples, it is another example of models breaking, though it also introduced the world to the Greenspan Put.
VIX
I have written about VIX a lot lately – there seem to be only three possible outcomes
First, VIX continues to retrace lower while VIX futures continue to remain steep providing the ‘free money’ trade everyone is now looking at. The total return and sharp ratio for this trade have been off the charts good. So maybe we revert to that and everyone can sell vol – so that it we truly have the tail wagging the dog.
The second option is that VIX reprices in such a way that the ‘risk/reward’ is more balanced. That the view that if you just sell the VIX and roll it (or own one of the short VIX funds) you have the best long term investment ever invented isn’t so obvious. This mostly comes to curve shape.
The third option is that short VIX causes trigger events in the ETFS and ETNs that track it, causing 100% losses, with such a spike that it distorts markets – putting an end to the direct VIX speculation – at least for a little while
Option 1, the ‘free money’ option doesn’t seem likely – the markets are efficient at pricing it out.
Option 2 makes more sense, but I cannot discount option 3 at all. If anything, option 3 reminds me a lot of CPDO – that selling VIX has become a ‘yield’ or ‘income’ product and that the risk of getting stopped out isn’t being priced in. (it is funny, that virtually every article that talks about shorting stocks takes the time to mention that your losses can be infinite, but that is never associated with selling VIX futures (I’ve still never seen an infinite stock price – though Volkswagen in 2008 sure felt like it).
Before I move beyond VIX, I should highlight that SVXY and XIV both took in money on Friday as investors sold more volatility – so much for fear.
Low- Or Min-Vol Funds
There are two types of funds in this category – those that buy low volatility stocks (the $7.6 billion SPLV is an example) or those that try and create portfolios of stocks with low volatility (the $13.4 billion USMV is an example).
Low and Min Vol Fund Flows
Both funds have experienced inflows since March, with SPLV having an incredibly strong day on Friday in terms of inflows. Both funds were down about 1.3% over the past two days – outperforming the S&P 500 and USMV hit is 52-week high on Wednesday – so current performance isn’t an issue.
I have one problem with this strategy in general – I believe the investors are relatively weak hands.
The appeal of low volatility funds leads me to believe, that at best, these are funds being targeted by investors very reluctant to be in the stock market, but feeling an obligation to be in it – for now. That is the ‘weak’ hand view. The worst case is these are investors who are taking these ‘low vol’ funds and levering them up to outperform the broad market – there are some indications of that, though less than we saw in the fall of 2016 when the low vol strategies actually had high volatility and led the market down.
I won’t harp on the low vol indices and how they select stocks (for now). On the other hand, the min vol strategies do account for correlation to create an index that “in the aggregate, have lower volatility characteristics relative to the broader” equity market.
The risk that the model doesn’t work seems high to me (it is working well now, as it hit 52 week high and had low volatility – but the realized volatility on the fund is picking up – which could make holders nervous. The potential to disappoint investors by having higher volatility when the money has come looking for an oasis of calm in the equity market – seems high.
When I look at these funds, the direction of flows on weakness seems to be most likely out of stocks altogether. If this was the last refuge for bears struggling to maintain an equity allocation – then on fear they are shifting out of equities as there is no allegedly safer bastion than funds with names like low vol. If, even worse, investors have leveraged this up, they are getting stopped out.
On strength, there can be outflows from these funds into more aggressive sounding funds, but on weakness, I think flows are biased to coming out of the market.
I think this combination of funds targeting low volatility with weak investors (if I’m right) is another potential issue and yet another link between stocks, volatility and correlation.
Correlation
Correlation within the stock market is increasing – which will drive volatility higher if it continues.
The CBOE S&P 500 Implied Correlation Index
By this measure, correlation within the stock market is increasing. Whether it continues is anyone’s guess – but this will impact some risk management decisions.
Risk Parity - Not A Problem Yet
The 800 pound gorilla in the room in terms of cross asset vol targeting is the world of Risk Parity. This continues to be a popular strategy – ranging from sophisticated model driven ones, to the less formal “Risk Parity Lite” or “Homebrew Risk Parity” which can be as simple as QQQ and TLT for the home player to the ‘spoos and blues’ coined by Zervos for non-Risk Parity funds attempting a simple replication.
Salient Risk Parity Index 20 Day Realized Vol
The Salient Risk Parity index targets 10% volatility with an equal weight across 4 risk assets.
While the fund will be driven by ‘expected’ volatility rather than its own realized volatility – I think it is worthwhile seeing how volatile it has been.
I am seeing a similar uptick in actual risk parity funds that I track, but all are still below their normal levels.
My understanding is that many funds cap leverage or employ floors on volatility and correlation so that they don’t get maximum exposure at the wrong time.
One problem with any VAR (value at risk model) is that it can cause you to have excessively large positions when volatility is tame and correlations are helping – so many employ caps and floors to limit the risk of being caught in a vol/correlation regime change shift.
In a nutshell, I think the combination of not being fully levered and reasonable returns during the past two weeks means we aren’t yet seeing any meaningful de-risking from Risk Parity strategies.
However, if volatilities increase and correlations increase (a positive correlation between equities and sovereign debt being the most dangerous) then we might finally see some renewed pressure from these funds to reduce positions across asset classes.
I can’t help from thinking what could cause bond yields to go higher and stocks to fall – which would be the trigger to get some selling from these funds. It is hard to think of anything obvious, though idea that we could see some sort of fight around the debt ceiling and some offhand remarks (or tweets) that spook both bondholders and equity holders doesn’t seem out of the realm of possibility.
Find Tail Risk Hedges
I think a vol spike will be the trigger to any movement – so I continue to think trades that benefit from rising volatility will outperform outright direction trades.
Having said that, I think a vol spike would mean that we face a potential 5% to 10% move in the S&P 500. That would translate into 2% to 6% in high yield (though I think credit will outperform) and as per my note yesterday – think BKLN is an intriguing short idea – because it could move close to 1:1 with high yield on the way down, but won’t move 1:1 if markets rally. With IG CDX roll coming up, I think that would mean a 5 to 15 bp move in IG spreads (again, this isn’t a credit issue so the move will be muted). LQD might be a better short (buy puts) than CDX as you get a lot of duration and the overhang in the bond market seems real and LQD finally had small outflows. The only issue with puts on LQD is that turns it into a yield trade rather than a spread trade – which I prefer. If you choose to short LQD, happy to help with some long treasury positions to keep it a spread trade.
Here I am at the end of 9 pages and I haven’t said anything about fundamentals. It is because I don’t have that much to say – I am more worried that financial engineering has run amuck – yet again – and an ‘unlikely’ event is more likely than know.
For full disclosure, I am preparing to buy SVXY or XIV in my personal account – but not today and not until I see how this all plays out – or at least until I can stop thinking about CPDO and LSS the minute I hear about VIX and vol targeting strategies.