How should you compare implied volatility across assets?

Mon, Mar 11, 2013 | Jared Woodard

Blog, Volatility

Most traders understand that a “buy low, sell high” approach applies to options markets just as well as it does anywhere else: when options are overvalued, it pays to be a net seller of that premium. But “overvalued” often has nothing to do with the level of implied volatility you might observe on a chart. Options are usually well-bid for a reason, or are cheaply priced for a reason. Finding an asset where volatility is truly under- or overvalued depends on answering the question: relative to what? That’s why we spend so much time looking at implied vol in relation to the trailing standard deviation of returns and to volatility forecasts (on a weekly basis, in fact).

When it comes to ranking the value of the volatilities implied for different assets, things get a little more complicated:

Implied minus realized has an important virtue: the output is an intuitive measure of what might be possibly achieved by a trade. Subtract a percentage point or two for transaction costs/slippage, and assume that future realized vol is equal to current historical vol, and you know what kind of returns to expect from a short delta-hedged straddle. Obviously in this case that latter assumption is nearly always false; figuring out likely future realized vol is the whole trick, and that’s where all of the risk is. And more to the point, a 5 point implied-realized spread in one asset is never automatically less attractive than a 20 point spread in some other asset. Consider the case where options on an equity index are priced at 22% vs. 17% historical vol and some nutty Canadian junior gold miner is priced at 40% vs. 20% historical. I’d argue that the equity index is a better sale, but the simple IV-RV difference won’t reflect that fact in a scan.

Implied as a ratio of realized: ratios don’t care whether option implied vol is 20% or 200%, only what that premium is as a percentage of recent movement in the underlying. In that respect, ratio comparisons are more helpful. You lose all trade-relevant detail (Is a ratio of 1.05 non-exploitable after transaction costs? You can’t tell), but gain the ability to rank different assets more directly. The problem, still, is that the volatility of some assets varies more than that of others, and this simple ratio doesn’t reflect that fact. For example, we publish a ranking of some major assets in the weekly update that looks like this:

1M major asset volatility risk premium (IV/RV). Source: Condor Options

The limitation of a table like this is that it doesn’t do anything about the fact that some assets, like TLT, are almost always near the bottom of the list while emerging markets, commodities, and individual stocks spend a lot of time near the top, but generally just move around a lot. In plainer English, we want trade selection to be cognizant of the fact that the volatility of the risk premium for e.g. EEM or USO is greater than that of TLT. Even if the risk premium in the former group is higher at a given moment, the higher volatility of that risk premium also makes it more difficult to get a reliable handle on likely future realized vol, which again is really the whole ball game.

I have an idea for a trick to solve this problem, and the details are in an update for clients.

Tags: , , , ,

2 Comments For This Post

  1. onlyvix Says:

    Hello Jared! I’m sure you could write a small book about this topic :) I just wanted to add few small points –

    How to compare (in addition to your 2 methods)
    1 – log ratio, or difference of logs = log(implied/realized) will be “better” behaved as log “stretches” small values and compresses large ones. This makes for a somewhat less noisy metric, but ranking is the same as ratio.
    2 – z-score = (implied/realized)/vol of vol – sounds nice in theory but in practice difficult to implement: realized vv is too noisy, implied vv is too noisy for different reasons – too few strikes, wide b-a.

    Why should we compare IVs for different assets?
    That is a critical issue that precedes implementation. If you are comparing similar names for relative value volatility trading e.g. AMZN,GOOG,AAPL,QQQ – you would expect vol of vol to be comparable, and all rankings should be equally effective. If you are comparing different asset classes e.g SLV vs TLT – it cannot be traded from relative value view. Here ranking – by whatever metric – is not applicable, as relative comparison does not as much sense. Your alpha should be viewed as coming from different sources and each asset class may require a separate statistical analysis.

  2. Jared Woodard Says:

    Thanks, D., very thoughtful comments.

1 Trackbacks For This Post

  1. Monday links: force-fed risk - Abnormal Returns | Abnormal Returns Says:

    [...] How should implied volatilies be compared across asset classes?  (Condor Options) [...]

Leave a Reply

Subscription Access

Get Our Free Newsletter

Weekly emails with Options market updates and analysis.


Jared Woodard specializes in trading volatility as an asset class. With over a decade of experience trading options and other volatility products ... Read More


Open All | Close All