As per usual, Deutsche's Aleksandar Kocic put my jumbled and inconclusive suspicions into clear and convincing prose. A brief excerpt:

Volatility declines either when the markets are predictable or when there is no consensus. In a relatively short time we have transitioned from consensus- to dissensus-driven vol selling regime...

... the market has remained in a standby position without ability or desire to take a strong view in any particular direction.

This makes a lot of sense. For a while towards the beginning of the year, the theme of 'complacency' dominated the financial press. The basic conceit was that uncertainty about the future was high, yet 'VIX is so low!' ergo investors must be a mighty complacent lot. At the time, I will admit that I pretty naïvely bought into that notion. But Kocic's interpretation of all the gamma selling is far more convincing in my opinion.

In order for vol to become expensive across the board (or just to revert up from its present cheapness), there must be a convincing consensus on 'why.' Abstract gesticulation about 'uncertainty' or 'geopolitical risk' does not volatility make. Rather, the ever present concept of narrativity is a key input of vol markets.

Just because all the 'fear index' articles bug me in the news, let's also briefly place VIX headlines in their proper context. The index is an aggregation of 30-day ATM impvols on SPX. It therefore represents one specific point (for the sticklers out there, yes, the calls use expiries 23 to 37 days out, so more precisely it represents a small square in the center of the vol surface) on the volatility surface, and fully omits any nuance stemming from changes in the term structure of vol, or the degree of skew. You can imagine weird hypotheticals, like if it were known with certainty that in 90 days something awful would happen, but for whatever reason, SPX can't decline at all in the next 30 days, then VIX could remain dirt cheap in theory, because all of those ATM options which constitute the index wouldn't be in high demand. The bottom line of this brief aside is that VIX is only rightly considered a 'fear gauge' if the fears surround the performance of SPX in the next 30 days. Anything outside of that category is ignored by the VIX.

If you want to read more about the calculation of the index, here is some serious nerdiness.

Vol appears to be at odds with the 'real world' because of a timing mismatch. The 'uncertainties' everyone talks about are very much real, but they are neither immediate nor concrete. The feeling being described when 'uncertainty' (policy, geopolitics, etc) is discussed is more a sense of existential dread about potentialities far out on the horizon than it is an immediate concern about what might happen in the next month. Even if you are gripped by that sense of existential dread, it probably isn't enough for you to pony up for the negative carry associated with a long vol position. There isn't any clear catalyst for why everything should/could blow up in the next 30 days, so looking at 30-day volatilities superficially appears to be at odds with the sense of worry about the state of the world. But in reality it's not: there's just a mismatch between the time horizon of concerns and the time horizon of an index like VIX, or even vol markets in general (because there are much cheaper/cleaner ways to bet on bad stuff happening than buying vol).

To sum up: the uncertainty is long-term, while the vols are short-term. Until long-term worries become immediate problems, vol will continue to languish.

RATES, INFLATION, & OIL

The pricing of government bonds incorporates a plethora of factors, and there are several different paradigms through which one can consider their level and movements. One such way, is the straightforward identity* which states that:

Nominal Yields = Real Yields + Inflation

When analyzing something like 10y UST's, it is often useful to break the object down into components you can analyze individually. It's easier to have intuition about real yields and inflation separately/explicitly than to implicitly consider each when describing nominal yields.

Here's a 10y CMT yield against 5y5y inflation breakevens:

Importantly, don't make the mistake of thinking based on the above that nominal yields = inflation expectations. I rebased the series to 100 to clearly demonstrate the correlation. If I had plotted the values without the rebasing, the white line would be far above the yellow.

As you may have noticed from past experience, inflation happens to matter quite a bit for bonds! The above chart demonstrates that graphically. Below, you can see that just as inflation expectations matter a great deal for nominal yields, changes in the price of crude oil matter quite a bit for changes in realized and expected inflation:

This provides a useful starting point (upon which I will build in future pieces) in analyzing the relationship between energy and rates markets. There is a good reason that interest rate strategists and traders keep a close eye on OPEC, inventory builds, and production statistics. An interesting question to consider going forward is how structural shifts in energy markets will impact the relationship between crude and rates. If renewables continue to gain traction, or even hit a sort of inflection point in adoption, that creates a very bearish oil environment. Superficially, one might be inclined to say: more solar/wind/hydro/etc = less demand for crude = lower oil price = lower inflation expectations = lower nominals yields = you should extend duration. But the crucially weak link in this logic is the very relationship, heretofore steady, between crude and inflation expectations. The relationship has held thus far because the cost of energy (usually obtained initially through burning fossil fuels like crude and its distillates) is a big part of CPI, and more importantly, is a key cost input in many other components of the CPI basket, thus carrying a disproportionate impact on the level of consumer prices. This linkage is vulnerable to change because if the price of oil declines due to alternatives becoming more attractive (one can envisage a scenario where environmental regs create huge Pigouvian incentives against burning crude), then the weight of crude (both explicit and implied through other goods) in the CPI basket will fall too, and the change will be muted.

It's true that humans, on net, don't change. But to say that 'this time can never be different,' while often useful, is too blunt an instrument in this case. Technology has, and does change, and that change creates a new environment in which the unchanged humans operate. It could very well be the case that soon, this time will be different, and the rates-crude linkage will break down. Not that that is likely anytime in the foreseeable future, of course. It's just a potentiality to keep an eye out for if ever rates and crude decouple inexplicably, alongside a perhaps less easily noticed change in the global energy regime.

USEFUL HEURISTIC: 'THE BAND OF AGNOSTICISM'

I came across a useful heuristic this morning as I was reading a fairly old (2002) piece on exchange-rate determination from Michael Rosenberg. In one section, he describes the dividing line between when volume is constituted of short-term versus long-term traders. To simplify, the traders with the short-term time horizons trade primarily off of momentum, whereas the FX 'value investors' work off of value estimates (sell when dear, buy when cheap). It is a useful framework: as lovely as theoretical models of fair value (FV) are, actual prices are a product of transactions, and transactions are a product of people (or computers!) coming together and agreeing to engage on specified terms. Therefore, understanding who constitutes the market (and sometimes who constitutes which side of the trade: there are situations where the bid is predominantly one kind of participant and the ask a very different kind) is extremely useful in forecasting what matters: the price at which transactions can be done.

But as I further considered the framework, I realized it is even more intuitive and useful for explaining market dynamics than how it is used in the report. For now, I will just point out what an instinctive/straightforward explainer it is of when 'value investors' (in any asset class) do or don't step in.

For a value-based trader to enter the market, the observed price must fall outside the band of agnosticism. But while the FV model may rely on a static relationship between inputs, the width of the band above and below the FV number fluctuates based on subjective conviction. High conviction = narrow band of agnosticism. Low conviction = wide band of agnosticism.

Thus, during times of volatility or uncertainty, as conviction weakens, the band of agnosticism widens. This is how uncertainty translates into realized volatility. If uncertainty were low, and value investors had high conviction in their FV estimates, their band of agnosticism would be sufficiently narrow that, say, a 150bps fall in price would be enough to induce them to buy (remember: the value-based traders enter only when the price falls outside their bands). This lowers volatility. One can imagine a theoretical scenario in which the band is at, say, +/- 1bps (theoretical maximum of certainty), so that a 0.01% change in either direction results in buying or selling back to FV. This would result in realized vol of 0%.

Whereas if market conditions are otherwise identical, but uncertainty abounds, the band of agnosticism will widen, such that even, say, a 300bps fall isn't enough to induce value-based traders to buy. In this scenario: high uncertainty = lack of conviction = more volatility.

Maybe this is all obvious or even tautological to the reader, but I have thus far found this straightforward notion to be a useful workhorse of analysis.

FOOTNOTES

* For more pedantic readers, you probably recognize that this is an approximation of (1+i)=(1+r)(1+pi). The approximation is a very safe one to make, however, especially in low yield, low inflation regime at present. But I didn't want to break out annoying LaTeX syntax or Taylor Expansions in the body of the post, for purposes of concision. For more mathy details, see here.