Skew and Trend following
In this publish I speak a widely known stylised truth of the investment enterprise: "Trend following is a positively skewed strategy".
Spoiler alert: sure it is (sort of), however it's plenty greater complex (and thrilling!) than you may think.
A short primer on effective skew
So what truely is high quality skew? Essentially it is an asset, or trading approach, whose returns have the subsequent profile:
- A high proportion of relatively poor returns
- The losing returns are smaller in magnitude than the winning returns
Or... In case you prefer a pretty photograph:
Or, in case you select maths, then the skew is the 1/3 second of the statistical distribution, and positive skew approach there's extra skew relative to a Gaussian everyday distribution which has 0 skew.
It's typically felt that superb skew is a great element, and people are commonly willing to pay a premium for owning assets with tremendous skew (and vice versa for negative skew) [where by 'a premium', I mean the assets have a higher risk adjusted return than you would expect when risk is measured purely by the second moment - standard deviation].
A coherent explanation of this comes from behavioural finance, and specifically prospect principle. A cognitive bias results in people overweighting the chances of low chance effects. They get fixated on the small risk of a massive advantage that advantageous skew gives. Equally with poor skew human beings get scared of the small danger of large losses which can be threatened by using poor skew.
(It may also be well worth readingthis paper, written via a gaggle of people I used to paintings with, and a few different human beings I haven't worked with).
In fact, we are able to test to look if we get paid for skew. If I have a look at the skew during the last three months of each day returns, and notice how properly that predicts the subsequent 3 months annualised Sharpe Ratio, then I find that with terrible skew the common SR is 0.33. With superb skew it's -0.016. The difference is statistically tremendous; if I do a regression the p-cost is 0.01.
(At this factor you is probably thinking 'ah-ha! I can use skew as a predictor in a trading strategy. I can be wealthy!' .This isn't an authentic concept! See as an example,this paper.)
Intuitively, why should fashion following produce definitely skewed returns?
Trend following is efficaciously like shopping for a synthetic straddle* (a mixture of long positioned and make contact with alternatives). This is a well known and fairly vintage end result (see the seminal Fung and Hsieh 2001). Intuitively this makes experience, seeing that each techniques will do nicely if volatility rises, and do badly if prices continue to be pinned. It's similarly widely recognized that any lengthy volatility approach, like shopping for straddles, have to produce undoubtedly skewed returns: a number of small bad returns when expenses don't circulate and we hand over our premium, plus a smaller number of big fine returns whilst prices circulate sufficient for one leg of our straddle to be inside the money.
* actually it is a look again straddle, however the difference isn't always crucial right here.
So, advantageous skew is sincerely one of the reasons why people like to allocate to fashion following techniques, the others being:
- Linear diversification; low correlation with traditional asset classes
- Non linear diversification; good performance in tail events like 2008 (if you're from a fixed income background like yours truly, you can also think of this as 'positive convexity')
- They sometimes even make money!
However fashion following additionally has it's issues. People don't like the long drawdowns that fashion following kind techniques produce, but those are an inevitable result of fine skew (for a given threat adjusted return the size of the average drawdown might be better than for definitely skewed assets). For instance, suppose you are trading a method with a Sharpe Ratio of 0.5 and an annual risk goal of 25%. With zero skew a terrible drawdown (one this is carried out 10% of the time) will be 9.Three% in significance. With tremendous skew that would upward push to eleven%, and might be simply three.7% with negative skew (skews of 1 and -2 respectively).
If fashion following generates wonderful returns (and there's no clean proof it has stopped doing so) then human beings should be extra scared of those ugly drawdowns than they may be of the blessings I've indexed above. But (spoiler alert!) there is probably something else happening.
The proof
Economists and quant finance 'specialists' regularly faux to be scientists (many of them have real Phds in actual clinical topics). So, allow's pretend to be scientists and in reality take a look at to peer if the proof supports our expectations.
I'm going to use three varieties of trend following buying and selling rule: a 2,eight day EWMAC; all of the way as much as a sixty four,256 day EWMAC (Exponentially weighted transferring average crossover). Finally the outcomes will be calculated over the 40ish futures contracts in my dataset. The entire element is being completed under the auspices ofpysystemtrade, and you could discover the standard ugly code right here.
* definitely 2,8 is sincerely a chunk costly to alternate, but costs do not affect the calculation of skew on account that they simply shift the distribution of returns to the left a piece.
For reasons that will become obvious I'm going to degree skew over unique time durations: day by day, weekly, monthly, and annual returns.
Let's start with the each day returns
 |
| Skew by using buying and selling rule, daily returns |
Let me give an explanation for these plots. The y-axis is the measurement of skew, and the x-axis is the quick parameter price in the shifting common pair (2,four,8,... 64). Each dot represents the skew dimension for a unmarried tool, and for a single five yr length. This gives an illustration of the uncertainty in our skew estimate (sure...I cannot stop banging on about uncertainty).
Here are the median values for every rule:
2_8 = -0.04, 4_16 = -zero.07, 8_32 = -0.Fifty one, 16_64 = -zero.73, 32_128 = -zero.Ninety four, 64_256 = -zero.82
So... WTF?! Negative skew throughout the board, with significantly poor values for the slower crossovers. Something weird occurring here.
Let's take a look at the alternative time intervals out:
 |
| Skew by trading rule, weekly returns |
 |
| Skew by trading rule, monthly returns |
 |
| Skew by trading rule, annual returns |
Interesting. It seems like for bigger time durations the estimate of skew does certainly turn out to be high quality. We can see this if we plot the median values for each rule, with the aid of time period:
 |
| Skew of a fashion following policies profits, measured at extraordinary time horizons, from left to right: day by day, weekly, monthly, annual |
The results run from (at the left) every day, to (at the right) annual. Generally, skew gets greater positive the slower the time period we use. The exception to this are the very quickest trading guidelines, which have a 'candy spot' for skew on the monthly time period.
The puzzle
Does it make feel that tremendous skew simplest appears at certain frequencies of size, with a greater infrequent size required for slower buying and selling techniques? Yes, it does. Think approximately a fairly sluggish fashion following rule. Maybe it modifications it is positions each few months. When it isn't always converting it's positions, then it is skew of every day returns might be dictated by way of the skew of the underlying property.
So if it's trend following say equities (negative skew), then half the time you'd expect to see negative skew of (when it's long), and half the time (when it's short) you'd see positive skew. Overall your skew will be zero (and this result should hold for positive skew assets as well).
However if you begin looking at annual returns, you are much more likely to see the generally high quality skew of fashion following. The point at which the skew turns into considerably positive will rely upon the speed of the trend following rule. With the faster policies we see positive skew with weekly and month-to-month returns; with the slower regulations it is not till we get to annual returns that the superb skew reveals itself.
(This isn't an original locating. See this, written with the aid of someone else I used to work with)
But... that doesn't explain one thing. Why is the skew strongly negative at the shorter time frames? It should be zero, or close to it.
The only explanation is that trend following strategies like to be long negatively skewed assets, and short positively skewed assets.
This is kind of exciting (nicely I think it is!). Perhaps the effective returns of trend following (a 'definitely skewed' buying and selling approach) aren't that sudden at all, if it simply loads on to negatively skewed property. Perhaps trend following is just a way of gathering the poor skew top rate.
And... Questioning some extra... It type of makes feel. If negative skew assets earn a top class within the market, then on common they will move up extra often than they pass down. And belongings which go up extra frequently than they go down, will have a tendency to showcase more bullish developments. And assets which showcase greater bullish tendencies, properly they may be bought with the aid of fashion following techniques.
This is all assuming that negative skew assets are negative before we buy them, and remain so . I will check this in a second.
What is the conditional relationship between skew and trend following
Let's do the following exercise. We'll find out the median skew, conditional on a trading rule being long or short, for a given trading rule. I'm going to measure the skew over a period of a month, using daily returns.
First, let's look at the skew of a given instrument in the month after a trading rule has taken it's position. Remember its this skew that matters in determining what the skew of the returns of a trading rule will be (at least for the slower rules, which will 'inherit' the skew of the underlying asset).
 | |
|
If a trading rule is long, then within the month following the forecast being made the skew is bad. If the rule of thumb is brief, then the skew is toward 0, or maybe tremendous. The impact is more substantive for slower guidelines (faster regulations could have changed their function at some stage in the subsequent month anyway, perhaps more than one instances).
This is a affirmation of our in advance instinct that slower fashion following guidelines are probable to have terrible skewed returns, because while they are long the underlying asset is negatively skewed; and when they may be quick the underlying asset is undoubtedly skewed (giving the approach the other: more negatively skewed returns).
Now we look at the skew in the monthbeforethe trading rule decides what position it is taking:
 | |
|
Now this is exciting. The slowest shifting common does what we might assume; it has a tendency to be brief while skew has been effective (or as a minimum much less negative), and is going lengthy when skew has currently been poor. This confirms the principle that we end up loading up on bad skew as trend fans due to the fact negatively skewed property are more likely to have wonderful waft (as a reward for that awful skew).
But for all the other trading rules we get the opposite effect*! For them the story is very weird: if skew has recently been negative they go short. But (from the previous graph), they then end up being short assets which subsequently have positive skew (which gives the trading strategy negatively skewed returns). The skew flips sign.
* in truth the penultimately slow trading rule is sort of flat.
If skew has been positive, the rule is going lengthy, and then the underlying asset has bad skew (which once more gives the buying and selling strategy negatively skewed returns).
What goes on right here? One viable rationalization is that this; for unstable belongings strongly poor skew generally appears after a sharp sell off. After this sort of promote off maximum trading rules will move quick. But skew, like volatility on the proper time horizon, is a mean reverting parameter. The buying and selling rule starts offevolved with the skew the 'proper' manner spherical for producing positive skew (it goes short recent negative skew, and long recent effective skew) however then the sign of skew flips, and it finally ends up with exactly the incorrect role!
The slowest transferring average isn't always affected by this; alternatively it is much more likely to choose up the secular wonderful drift from negatively skewed belongings.
Summary
Trend following rules do indeed have the positive skew you'd expect... but only at the right time horizon. For slower trend following rules you don't see them appear until you are using annual returns. At shorter time horizons they have persistently negative skew.
An asset that is negatively skewed at one time horizon, and undoubtedly skewed at every other is... Bizarre. Should we want to own it? I guess it relies upon on our very own 'funding horizon'. If you only examine annual returns, you will love trend following! If you have a look at more common returns... You will be less impressed. Given the lengthy drawdowns of fashion following strategies, you'll be exceptional off searching at your portfolio each two decades or so :-)
For the slowest fashion following rule I use it looks as if this takes place because negatively skewed property have a return top rate, which results in positive glide. So slow trend following guidelines will have a secular lengthy bias to negatively skewed assets.
For other trend following rules this rationalization is inaccurate. Instead, they tend to short property whose skew has currently long past terrible, and vice versa. It appears probable this is because of sharp selloffs in unstable belongings creating each terrible skew and bearish latest developments. However skew is suggest reverting; so the other regulations turn out to be being brief property which sooner or later have high quality skew, and vice versa.
This also approach that in case you're making plans to apply poor skew as a buying and selling sign in aggregate with fashion following, it is going to be a extremely good diversifier! Except for the slowest moving average crossover, the momentum rule will typically do the other to a skew trading rule: it will brief bad skewed assets, and cross lengthy definitely skewed assets.
