One Last ETF Simulation: UHN (Heating Oil)

11Oct12

A nice feature of some futures-based ETFs is that, through a bit of simple math, we can simulate how the ETF would have performed historically with remarkable accuracy. Previously I’ve shown this with VXX (1-month VIX) back to 2004, UNG (natural gas) to 1994, and USO (crude oil) to 1985.

One last ETF to add to that list: UHN which tracks heating oil futures. In the graph below I’ve shown my estimate for UHN (red) versus the spot price for heating oil (grey) all the way back to 1980.


[logarithmically-scaled, growth of $10,000]

When the ETF (red) underperforms spot (grey) futures are likely contangoed, and vice-versa.

Note that we don’t see nearly the same drift from spot in UHN as we saw in the other ETFs (meaning the futures term-structure tends to be flatter).

How accurate is the estimate?

The graph below shows the period of time in which the estimate (red) and actual ETF returns (blue) overlap.


[logarithmically-scaled, growth of $10,000]

For a number of reasons, it’s impossible to perfectly simulate futures-based ETF returns, but based on the close fit, I’m confident that we’re well within the ballpark.

In a previous post I showed that (unlike VXX, et al.) blindly following the futures term-structure (going short when contangoed and long when in backwardation) doesn’t work over the long-term with UNG (natural gas) or USO (crude oil).

As you can probably guess based on how closely UHN tracks spot, the same is doubly true here.

Happy Trading,
ms

. . . . .

To stay up to date with what’s happening at the MarketSci Blog, we recommend subscribing to our RSS Feed or Email Feed.



6 Responses to “One Last ETF Simulation: UHN (Heating Oil)”

  1. 1 Andrew Miller

    Do you think that because VIX is mean reverting the biggest driver of returns in VXX is therefore the term structure and one should make investment decisions because of the term structure. Compare this with commodities where the return is from spot price, collateral return and roll yield. Therefore commodities returns are more affected by spot price and collateral return over time?

  2. 3 Jerry

    Hi, love your articles, but I also think you’re forgetting one major component in analyzing the performance of these ETFs – the repo/stock loan borrow rate. How do you incorporate them, if at all?

    • 4 MarketSci

      You’ve lost me Jerry.

      • 5 Jerry

        Hi Michael, to be long or short a stock there is a carrying cost. For example, I know to be short UNG stock can cost anywhere from 0.5%-1% per annum. I was wondering if you took these costs into account when running your simulations?

      • 6 MarketSci

        Carrying cost can have many meanings depending on the asset traded. Normally we wouldn’t use it to refer to an ETF or stock, but okay…

        If you mean the cost of carry for the underlying futures, that has already been accounted for in the price of the futures themselves, and therefore this simulation (based on the ETF’s roll methodology).

        If you mean the yield on the UST investment and/or the ETFs management fee, those have been included/estimated in this simulation.

        Beyond all of these though, the proof is in the pudding. Look at the period of time the simulation and actual ETF returns overlap. If there was a big factor we were missing, you would see drift. You don’t.

        There is of course going to be short-term (mean-reverting) drift because the actual ETF trades at premium/discount. But that’s not something that we can accurately model.

        michael


Leave a Reply

Please log in using one of these methods to post your comment:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s