A Dash of Vol

Jake (@Econompic) mentioned this when chatting about the Steady Vol strategy.

I started poking around with the data and found a similar result.

Very odd.

This means, the previous month's realised volatility will tell us more about next month's volatility than today's Vix level does.

To me, at least, this is not very intuitive at all (yet).

A while ago I read a slew of literature about the efficacy of the Vix as a forecast tool, some were pro some against but seeing it roundly trounced by last month's realised vol throws the cat amongst the pigeons!

In any case, if steady vol is something to aspire to, achieving a steadier vol should lead to an improved performance, right?

That's exactly what we see with the Realised Steady Vol strategy. 0.6 vs 0.58 Sharpe, over 25 years.

Probably not large enough a difference to be significant however.

Another point I realised after reading Jake's post was if the previous month's realised returns have a high R2 against the Vix, why not use them to weight a holding in the Vix?

Usually the Vix is in Contango, which means you'll lose money by holding it.

But if your holding is weighted along the lines of

1/Realised Vol

You turn a profit.

The Sharpe is nothing to write home about.

But that's not the point!

You have now begun to treat volatility as an asset class.

So instead of being like every other long-only-S&P-index-holder, you allocate a little (10% on average) to the Vix along the lines of the above, the Sharpe over the last 25 years goes from 0.4 to 0.75.


Diversification is super-important.

The only free lunch in finance.

Especially when you can tap into something as special and orthogonal as volatility.

Note, that usually I'd recommend benchmarking this portfolio against and equally weighted strategy. In this case 50/50 Vix and S&P 500 or 50/-50 just leaves us with almost zero profit, which reinforces the idiosyncratic nature of volatility and why it's nice to find a way to make it investable.

Thanks to Jake for his write-up and sparking this tangent.

Code for the Lazy Backtest IDE is here.