Risk Update: October 2020 – One big LTCM.


Just as we were putting together our thoughts for this month’s update, we came across the above linked article which is such a good example of the misunderstanding of fragility, our chosen topic for the month. It is a Reuters article, but we might classify it as a “Not the Onion” or “Not the Bee” type article.

The opening line reads: “Taking a break from fighting the coronavirus crisis, the world’s top central bankers will attempt to resolve the existential questions of their profession this week as they tune into the European Central Bank’s annual policy symposium.” We are fairly confident that central bankers are not aiding, in any way whatsoever, in fighting the coronavirus crisis. They could be, however, credited with engaging in a frenetic effort to fight the debt crisis, an extreme fragility of their own making.

This, of course, comes back to our usual argument about the relevance of endogenous risk – where fragility exists, versus exogenous events – the shocks that expose fragility. The fragility comes from the house made of cards. The wind is an inevitability. The above quote is analogous to the builders of card houses fighting the wind crisis. Reading the article, one cannot help but think of Orwell’s “1984” concept of doublethink.

“To know and not to know, to be conscious of complete truthfulness while telling carefully constructed lies, to hold simultaneously two opinions which cancelled out, knowing them to be contradictory and believing in both of them, to use logic against logic, to repudiate morality while laying claim to it, to believe that democracy was impossible and that the Party was the guardian of democracy, to forget whatever it was necessary to forget, then to draw it back into memory again at the moment when it was needed, and then promptly to forget it again, and above all, to apply the same process to the process itself”

Doublethink may very well be the answer to the question that has forever dogged us, as regards central bankers – do they simply not see the problem, or are they in on it? The answer may very well be that they are, knowingly and unknowingly, doublethinkers! (Be careful going too deep down this thought path, it is a mind bender.)

Our point, of course, is that the system is fragile.

We are regularly asked questions, even at times asked to give presentations, along the lines of “can we predict risk events?” The short version of the answer/presentation goes something along the lines of: NO.

We cannot predict when or where lightning might strike and spark a forest fire. We certainly can’t predict if the next major wildfire will be triggered by lightning, or power lines, or campers, or motorcycle backfires, or kids playing with matches…..by definition you can’t predict the unpredictable. What we can do, however, is to take note of fragility. Random unpredictable lightning strikes do not cause uncontrollable forest fires unless they strike an already fragile forest system. Without an accumulation of dry brush and trees, it is pretty hard to get a significantly damaging forest fire. If you want to avoid the risk of property damage due to forest fires, don’t own property in the midst of forests that are clogged with dry brush and trees. If you can’t avoid that, best to buy insurance early, before everybody else realizes the same, and have a well thought out escape plan!

Nassim Taleb defines fragility as something that is harmed by disorder (short optionality/volatility), while coining the term ‘antifragility’ as its opposite, ie something that benefits from disorder (long optionality/volatility). What is the “crisis” that the central bankers are fighting? It is the fragility of the debt driven economic system that, when struck by the spark of a global pandemic, risks erupting into a mass conflagration of destruction.

What does economic system fragility look like?

Figure 1: US Total Federal Debt to GDP

Source: Bloomberg

Even in whatever warped economic education we all endured, we were generally led to understand that when government debt exceeds GDP things aren’t good. Inevitably, with each episode, we are told this time is different. Regardless of arguments as to the right or wrong of the actions that have led us here, the simple fact is that the system is more fragile.

Not surprisingly, the Federal Reserve’s balance sheet looks somewhat similar.

Figure 2: Federal Reserve Balance Sheet Holdings

Source: Bloomberg

Maybe the simplest way to define the fragility of the economy is the lack of loss absorbing capital in the food-chain of leverage. What in our world of derivatives we refer to as ‘uncapitalised tails’. We all got to see this very clearly back in 2008 when the decline in the collateral value of US homes triggered losses that could not be absorbed by the global banking elites, forcing an explicit taxpayer bailout and the implicit stealth tax of inflation being imposed on present and future non-asset holders. Naturally, the above mentioned doublethinkers don’t see the crisis as being the fragility of the system that they promulgate, but rather explain it off as an unpredictable exogenous shock.

A very simple anecdotal example of the food chain of leverage, and how it leads to fragility, might look something like this:

  1. Bank borrows from depositors, paying them a Federal Reserve imposed 0%. The bank is 25x levered, meeting its regulatory requirement of a 4% Leverage Ratio.
  2. Bank lends to CMBS REIT in the form of overnight repos, charging 1%.
  3. The CMBS REIT lends via commercial mortgages at 2%. The REIT pays a dividend of 10% and a Management Fee to the REIT Manager of 2%. Thus, it needs to run at 12x leverage, maintaining circa 8% capital against its assets.
  4. The Property Developer, by circumventing the regulated banking market and tapping the REIT directly, avoids the strict loan-to-value constraints and is able to achieve 10x leverage, holding only 10% capital against his risks.
  5. The Property Developer leases the building to a restaurant at a 3% yield. The restaurant owner has borrowed all of his start-up costs and has no capital.
  6. The diner purchases dinner once a month at the restaurant, using his credit card that he plans to pay off with next month’s wages, he has no capital, and after expenses the restaurant owner has 4% remaining profit.
  7. When the diner gets temporarily laid off due to the pandemic, he stops coming to the restaurant and the restaurant owner can’t pay the lease, the property owner can’t pay the mortgage, the REIT can’t pay the Repo, and the bank can’t pay the depositor. Nobody has sufficient capital to absorb the losses, and the Fed has to step in and buy up all the debt to keep the entire system from going insolvent. Fragility.

Back in our June 2020 Update https://convex-strategies.com/2020/07/17/risk-update-june-2020/ we linked to this BIS report on the growing relevance of leverage in the world of Non Bank Financial Institutions (NBFIs) https://www.bis.org/publ/cgfs65.pdf. In the above example, the CMBS REIT would count as the levered NBFI, not to mention said REIT is likely owned by a bunch of levered investment vehicles (aka hedge funds) as well. We’ve mentioned that we should expect to see more such notes coming out of the likes of the Federal Reserve as they continue to smooth the landing strip for the inevitable touch-down of their increasingly formalized policy of bailing out levered NBFIs. Sure enough, here it is courtesy of Fed Vice Chair of Supervision Randal Quarles:


If we were to summarize, we would paraphrase thus; “we had/have no choice but to bail out leveraged NBFIs as they contribute to fragility in the system”. As we quoted Bill White in our August 2020 Update;

“the policy regime has encouraged the belief that economies face liquidity problems when the underlying problem is really one of insolvency” thus “the path we are on is unsustainable because the underlying problems grow worse over time while the solutions become more constrained”.

Bill is, of course, correct. It is not a liquidity problem, but rather an insolvency problem due to the high levels of leverage/lack of loss absorbing capital.

How is it that these players are allowed to impose risk on the stability of the entire system through their use of leverage? How have we let, what we like to call Agency Bias (ie fiduciaries who participate on the upside but not the downside), imperil the wellbeing of the entire financial universe yet again? Most simply put, by allowing, indeed mandating, that they use a risk methodology that benefits their asymmetric incentive structure to the detriment of end capital owners and society. It is not too much to say that the use of probabilistic, ergodic process, Gaussian based risk methodologies is destroying the world. Bluntly, Value at Risk (VaR) isn’t just wrong, it is dangerous and perhaps even more sinister than that.

As we have said before, the financial world has become one big LTCM. The use of leverage by fiduciary risk takers, fuelled by their ability to misrepresent risk through standardized measures such as VaR and significantly encouraged by the moral hazard expanding practices of policy makers, has built fragility of unprecedented scale. We can use a little VaR Calculator tool to give some perspective on the issue. The below simply shows the respective individual asset class with a mean-variance based distribution of daily returns over an 18 month data series. The light blue distribution is a simple normal distribution over the data period, while the darker blue line is the same but builds in the actual skew of the period. In the box below we show the properties of the distributions and the results of simple VaR and Conditional VaR (cVaR) calculations. We do this over two 18 month data series, one ending on 28th February, 2020, and one ending on 19th March, 2020. The Kurtosis is the measure of how much the Skew Distribution deviates from Normal, or Gaussian, properties, in essence, the excess risk outside of that which was predicted.

For the first example we use the Bloomberg Barclays US High Yield Index. One could think of this as a proxy for the credit risk on a bank’s balance sheet, or the portfolio for a credit based investment strategy. In the pre-pandemic “crisis” period it shows a moderate Skew and Kurtosis, with the respective levels of VaR and cVaR. When we slide the data series to capture the three weeks of the “crisis”, you can see the relatively sharp increases in all of the risk measures, despite just adding circa three new weeks to the 18 month data series. Visually, it jumps out just how much the new Skew Distribution diverges from a Normal Distribution and starts to show its left-hand fat tail dynamic (hint, it obviously isn’t normally distributed). VaR and even cVaR were not meaningfully relevant predictors of risk.

Figure 3: US Corp High Yield Index 18mth daily returns data to 28 Feb 2020

Source: Convex Strategies, Bloomberg

Figure 4: US Corp High Yield Index 18mth daily returns data to 19 March 2020

Source: Convex Strategies, Bloomberg

What really stands out is how the moderate increase in Skew leads to a quite dramatic increase in Kurtosis. It is indeed the case that you can say Kurtosis is convex to Skew. Rest assured, we are aware that the Skewed Distribution merely indicates how much off the Normal Distribution HAS been, not even remotely how far off it COULD be.

We can do the same exercise with the S&P Index as the underlying asset. Again, as you would expect, we see that the historical based VaR/cVar were pretty poor predictors of potential risk. The earlier data set, for the equity risk, has a higher standard deviation (in VaR speak that equates to “higher risk”) and slightly less Skew than the High Yield Credit Index, thus smaller Kurtosis (less excess risk, per our simple definition). When we add the “crisis” period to the data set, we see the similar expansion in Skew and the markedly greater expansion in Kurtosis, though not nearly as much as in the High Yield Credit Index, which raises the obvious question as to which one is truly riskier. The equity risk, per this notably flawed comparison, at least has some up tail to go with its obviously fat down tail. Gaussian, including our Skew version, is a bad way to describe equity risk, just not as bad as it is to describe credit risk, which is much more concave, ie. limited upside with accelerating downside.

Figure 5: SPX Index 18mth daily returns data to 28 Feb 2020

Source: Convex Strategies, Bloomberg

Figure 6: SPX Index 18mth daily returns data to 19 March 2020

Source: Convex Strategies, Bloomberg

An interesting one to look at, at least in our opinion, is the Bloomberg Barclays US Treasury Index. As you would expect, in the pre-crisis data set, it shows relatively small amounts of VaR-type risk, ie standard deviation, nor potential excess risk of Skew and Kurtosis. Still, for something that is commonly utilized as a “risk mitigant” in investment portfolios or treated as a risk-free asset on bank balance sheets, that adjustment in Kurtosis after the three week “crisis” period is quite telling.

Figure 7: US Treasury Index 18mth daily returns data to 28 Feb 2020

Source: Convex Strategies, Bloomberg

Figure 8: US Treasury Index 18mth daily returns data to 19 March 2020

Source: Convex Strategies, Bloomberg

Kurtosis, our simplified measure of excess risk, on US Treasuries expanded from 0.8 to 10.3, or 12.875x. That is a significantly greater relative expansion of Kurtosis than either the equity index, 4.727x, or the credit index, 4.33x. Imagine from a bank perspective, a supposedly zero Risk Weighted Asset that they hold no capital against just incurred a 12.875x jump in left-tail Kurtosis, even worse for the traditional investment portfolio that is treating fixed income as a risk mitigant and de facto giving themselves capital relief from holding (likely levered) positions in it against their equity/growth holdings. Of course, there may be some possibility that market participants, allowed to treat said assets as risk free or risk reducing, may have levered any and all available positive carry in said assets, leaving repo/funding counterparties in a bit of a panic when the amount of actual risk disclosed itself. We suspect this has something to do with Vice Chair Quarles’ comment in the linked paper above that;

“the pressure on longer-dated government bonds was sufficient to impair pricing for some of these bonds in this normally deep and liquid market, an outcome that we would not normally expect.”

This necessitated their intervention to address wrongly behaving markets. The official premise being that markets behaved wrongly relative to their models of risk, as opposed to an alternative premise that their models of risk behaved wrongly relative to markets. Funny old game.

This simple mechanism helps clarify why it is low risk (high leverage) things like AAA Super Senior Tranche Sub-Prime CDOs or Eurozone Sovereign Credits, and their respective regulatory treatment, that tend to necessitate bank bailouts and central bank interventions, as opposed to explicitly risky credit activities. You could say the mismeasurement of risk is more dangerous than the riskiness of risk. As the old Mark Twain saying goes: “It ain’t what you don’t know that gets you into trouble. It’s what you know for sure that just ain’t so.”

You could stretch this to a conclusion that the “system” would be safer if banks took actual risk that paid some actual return, eg High Yield Credit, capitalized that risk appropriately and then managed positively convex, negatively correlating exposure, eg CBOE Long Vol Index type stuff. Instead they seemingly prefer faux riskless strategies with virtually nil capital protecting against the mis-measured future losses. It is possible to see this as a less systemically risky financial system! On a return and risk basis this simply looks like our Barbell vs Balanced Portfolio view, replacing SPX with the High Yield Credit Index. Not surprisingly, this results in something that has less risk, more return, and better true capitalization.

Figure 9: Credit Barbell vs Credit Balanced

Source: Convex Strategies, Bloomberg

Or in scattergram view:

Figure 10: Credit Barbell vs Credit Balanced Jan 2005-Sept 2020

Source: Convex Strategies, Bloomberg

Figure 11: Credit Barbell vs Credit Balanced Jan 2020-Sept 2020

Source: Convex Strategies, Bloomberg

Common sense might inform us that the US Treasury Index’s historical risk parameters could be impacted by the level of yields at a given point in time. If we expand our pre-crisis data series all the way back to the highs of yields in 1981, making it a 460month series, we see significantly positive/righthand Skew and Kurtosis.

Figure 12: US Treasury Index 18mth daily returns data to 28Feb 2020

Source: Convex Strategies, Bloomberg

With this view, the riskless cum risk mitigating dynamics make a lot of sense. Imagine the pain felt when, as rates approach zero, the Skew and Kurtosis flip to the other side. Imagine it even more if you are running something like Risk Parity or any other such vol optimizing strategy. Be aware, for this little missive we are only talking about the “wrongness” of mean-variance Gaussian type models on the individual asset class level, ie variance. At the portfolio level there are two key variables that are presumed constant in their use as a measurement of risk, variance (volatility) and covariance (correlation). Likely we will come back to the covariance issue in future notes but leave it to say Kurtosis can also be convex to correlation, and likely will be when and where it hurts the most (reflexivity is yet another issue for further discussion!).

This brings us back to where we started, the system is fragile. It is impossible to predict when/where/how a spark may occur, but the good news is you don’t need to! It is possible to a) recognize and to some extent even measure fragility, eg think reverse testing stress analysis, thus giving the chance to avoid it, and b) there are ways to add positive convexity to a portfolio. We say it over and over again: avoid things that don’t participate on the upside but have correlated asymmetric downside and add truly protective convex protection against the downside, thus allowing you to take more upside participating risk. Arithmetic Average annual Returns and Value at Risk models are harbingers of Agency Bias strategies, run for the benefit of the fiduciary. Geometric Compounding and Convexity are beacons for the benefits of end capital owners.

Read our Disclaimer by clicking here