As any parent will confirm, teenagers know one thing. Unfortunately, it is “every”-thing. The rest of the population, by contrast, recognizes and appreciates the complexity of the world in which we inhabit and all the uncertainty that this necessarily generates. One of the key lessons life has to teach us is how to deal with the risks arising from uncertainty, and as often happens – albeit not always – what is true in life is also true in finance.
Borrowing the lexicon of Donald Rumsfeld there are two types of risk that we encounter in our daily lives: known-unknowns and unknown-unknowns[1]. Of these two we can dismiss the second category. Not because they are not important, they are – often extremely so – but because, by definition, we have absolutely no way of quantifying or dealing with such risks on an ex ante basis[2]. Such events just hit us and we have no alternative but to deal with the consequences. This unavoidable and irreducible risk may not always result in positive outcomes but it unquestionably adds to the rich tapestry of life and who would wish to inhabit a world of certainty (no, seriously – think about it!)
However, in terms of known-unknowns we can do something; indeed, this is the bread and butter of risk managers around the world. Two recent examples of known-unknowns are the UK EU referendum (outcome: Brexit), and the upcoming US presidential elections (outcome: fingers crossed!). We may not know in advance the outcome, but we do know that these events would occur and when.
We have documented the use of crowd-sourced sentiment indicators to help investors navigate Brexit in several earlier posts (including last week) so we won’t dwell on that here[3].
In terms of the US elections our sentiment-based US Political Risk Indicator (PRI)[4] has been helpful in detecting shifts in the political mood during the campaign before the opinion polls. Throughout August our crowd-sourced measure of public anger towards the US government was elevated. Given one clear differentiating factor between the two candidates is that Clinton is viewed as being firmly “establishment” whereas Trump is firmly “anti-establishment” – this negativity towards the government flagged a loss of support for the former Secretary of State; a trend borne out in the polls as Clinton’s lead diminished from over six percentage points at the beginning of the month to barely 2% by month-end.
However, during September we observed a sharp drop in government anger, a move that presaged her bounce back in the polls from almost even at the time of the first debate to a more comfortable six-point lead (as we suggested it would in the following tweet – see exhibit below). Suffice to say that crowd-sourced sentiment indicators provide a useful additional prism by which to assess such event risks.
Exhibit 1. Sentiment-Based Political Risk Indicator – US
Source: www.amareos.com
That said, event risk is just one type of known-unknown. In finance, the risk most investors focus on is market risk, which is defined as,
“the possibility that an investment loses money as a result of adverse changes in market prices[5].”
The standard method investors employ to mitigate this type of risk is to calculate an asset’s fundamental value – let’s call it the “model” for ease of use – and to then compare this to the current market price. By only buying assets that are considered undervalued (ie the market price is below the model price) and selling assets which are overvalued (ie the market price is above the model price), the aim is to create a favourable skew to expected returns and thereby lower market risk. This all sounds very good, but this approach has one, very serious, flaw.
For investment signals to be generated using this approach the market price and the model valuation must divergence periodically[6] in order for investors to benefit from the subsequent convergence. If this were not the case, and the model valuation always equalled the market price, we would inhabit the world of efficient markets, where the “market is always right” and the incentive for active investing would disappear. (Thankfully, this world is only a reality in certain dark corners of academia where the light of experience has yet to shine.)
Rather obviously in such periods of divergence, “fundamentals” are of limited use to investors in assessing market risk because the information pertinent to the asset in question is already incorporated into the model valuation. After all, consider the extreme situation of a bubble where fundamentals become utterly irrelevant as market forces are driven by other factors.
One “fix”, or “hack” in more modern parlance, to such a problem is to compare divergences between the market price and the model valuation and make assumptions about market risk based on historical precedent. Indeed, this is the approach many investors follow out of necessity. But, in doing so, one opens oneself up to yet another risk, namely that the underlying model is deficient in some sense either because it is incomplete or incorrectly specified.
As many investors found out to their peril during the Great Financial Crisis when events that were not supposed to happen happened and correlations that were not supposed to occur occurred, the P&L consequences arising from model risk can be extremely hazardous.
This is where the use of crowd-sourced sentiment indicators is, in our view, extremely valuable. As we documented in a previous blog[7], the basic hypothesis as to how emotions and crowd psychology impact the market is as follows:
“excessive optimism (pessimism) leads to periods of market overvaluation (undervaluation) and that this overvaluation (undervaluation) results in low (high) future returns. In other words, sentiment is positively correlated with contemporaneous price but negatively correlated with long-run future returns.”
Hence, monitoring and quantifying the prevailing mood of the crowd is ideally suited to assessing situations (and gauging market risk) when asset prices and fundamentals are diverging.
Extending this analysis further led us to develop the notion of crowd fail, where a strong sentiment skew implied either by high levels of sentiment (excessive optimism) or low levels of sentiment (extreme pessimism), invalidates one of the prerequisites for the collective wisdom of the crowd to outperform the vast majority of individuals[8].
To demonstrate this effect, consider the exhibit below which plots the level of crowd-sourced sentiment towards the S&P500 versus the 2-year ahead price change in the index.
Exhibit 2. S&P500 Sentiment vs. 2-year Forward Returns (2010-)
Source: www.amareos.com
We find negative correlation between the level of sentiment and future equity price returns. During periods of high sentiment, the forward change in the level of the S&P500 over the next two years is lower than the period average and significantly lower than when sentiment was low[9]. This result is entirely consistent with the aforementioned hypothesis
Given that sentiment for the S&P500 presently stands -0.36 this leads us to conclude that the medium-to-longer term outlook for US equities remains favourable. Certainly, sentiment readings are nowhere near levels that preceded the Great Recession slump (1.7) or the 2011 (2.4) or 2015 (2.2) significant equity market corrections.
For long-term investors with deep-pockets this is all well and good, but the vast majority of financial market participants cannot afford to make an investment and then head off to the beach and forget about it for a couple of years (if only!).
So, to complement our sentiment indicators, Amareos also produces daily measures of Fear and Stress[10] (available on all 6,000 underlying assets covered). These emotions tend to be associated with bouts of financial turmoil as demonstrated in the exhibit below which plots crowd-sourced Fear towards the S&P500 and the VIX index (this equity volatility index is often, inappropriately we would argue, referred to as the equity market’s Fear gauge[11]).
Exhibit 3. Crowd-Sourced S&P500 Fear Sentiment vs. VIX
Source: www.amareos.com
As can be readily seen, spikes in the crowd-sourced US equity Fear indicator typically correspond to spikes in the VIX index[12]. However, more recently, what we have noticed is that while both series continue to oscillate in tandem, Fear has remained somewhat elevated unlike the VIX, which stands towards the lower-end of its range (even after this week’s blip higher). This indicates that the crowd is more concerned about the near-term outlook for US stocks than is readily apparent from market-price based measures.
Given the two rather obvious known-unknowns on the near-horizon namely the US Presidential election campaign and the next couple of FOMC meetings[13], the recent rise in US equity Fear leads us to be much more cautious about the near-term outlook for Wall Street[14] even though the medium-to-longer term outlook remains constructive. This is something risk managers would be advised to take into account when assessing the current financial environment.
[1] We have excluded unknown-knowns, which according to the philosopher Slavoj Zizek are “things that we know, but which we are unaware of knowing”. This is both to avoid confusion and also because it is not especially pertinent to finance.
[2] We can’t know what we don’t know, at best we can guess.
[3] See: https://amareos.com/blog/fear-and-loathing-in-london/
[4] We discussed the Political Risk Indicator in an earlier note see: https://amareos.com/blog/political-risk/. For those interested in getting our daily sentiment-flavoured amuse bouche you can follow us @Amareos_info.
[5] Market risk is also known as systematic risk. Unlike unsystematic risk, it cannot be reduced diversification, but it can be hedged.
[6]In the case of a quantitative model chosen to minimize residuals (the bit unexplained by the model) the extent of this divergence is, by construction, as great as the convergence.
[7] See: https://amareos.com/blog/emotions-and-markets/
[8]See: https://amareos.com/blog/outsmarting-the-crowd/
[9] In the six-year period in question the two-year return in the S&P500 was hardly ever negative but this in no way detracts from the point being made.
[10] This daily data is available to view either on the Amareos web portal or on our App available on the Eikon Studio.
[11] More correctly the VIX is a measure of investor demand for equity options, which although highly correlated with fear can also reflect non-emotional driven demand and supply.
[12] Sharp-eyed readers will note that the spikes in the Fear index are notably higher in more recent years. We judge this to be a reflection of the underlying market trend. In 2010 having been battered by the Great Recession and significant losses investor fear would naturally tend to be more modest than in later years when stock prices had risen significantly and there was more to lose both financially and emoitonally.
[13] Just like last year the Fed has persistently signalled these are windows of opportunity to hike the Fed funds rate another 25bp.
[14] In addition, as some commentators have noted, given the positive correlation between equities and bond prices, any weakness in either could force a bout of deleveraging by risk-parity funds. As this note is already fairly lengthy we would direct readers interested in this topic to a Black Swan Economic Consultants (BSEC) note published last September discussing the challenges facing this type of fund – see: https://www.linkedin.com/pulse/great-asset-allocation-problem-ryan-shea
