Short on Treasuries

23 Aug 2011 by Jim Fickett (8 May 2015 removed dead link).

On fundamental grounds, Treasuries are significantly overpriced. While several forces could keep yields low, or even take them lower, in the short term, it is quite likely that yields will be higher somewhere in the next few months. An analysis of volatility suggests that holding an inverse ETF bet against Treasuries for a few months will not have excessive tracking error. So I have taken a small position in the Proshares Treasury 20+ year inverse fund.

Treasuries are overpriced, and a correction is likely

There is no guarantee, of course, that Treasuries could not become even more overpriced. If the situation in Europe really goes south, and a large number of investors are looking to get out of euro-denominated investments, that could drive Treasury prices higher, and yields lower, for some time. However so far European politicians have met each market crisis with just enough of a response to calm investors for a while. The world tends not to end, most days, and I think the most likely scenario is that alternations between markets getting out of hand, and politicians pacifying them, will continue. In other words, volatility is likely, and that means opportunities to buy low and sell high.

One other force that could drive yields even lower is all the talk of the US becoming more like Japan. Compared to Japanese Government Bonds paying 1%, the US 10-year Treasury yield of 2.1% looks high to some people. But this ignores inflation. The only way the Japanification argument makes sense is if you believe the US is headed for deflation. I explained some time ago why this is unlikely (Deflation should not be a serious concern for US investors), and the reasons have not changed: the usual aftermath of financial crises is inflation, not deflation and, furthermore, policy in the US (very much unlike Japan) is deeply biased towards inflation.

On fundamental grounds, the 10-year yield is extremely low. It has been hovering around 2.1%. So first, with core inflation running at 1.8%, and with our central bank determined to create inflation, that comes close to guaranteeing a negative real yield. Second, such a yield is almost without historical precedent.

To go short, one must be prepared for volatility

Such extreme conditions tend not to last and, all told, I think it is very likely that sometime in the next few months yields will be higher. So it is quite attractive to take a short position. But the timing is not very clear, and yields could certainly go lower before coming up again.

There is always a cost to going short, and taking a short position makes you a market timer, like it or not. So an important question arises: what is the likely cost of holding a short position through a possibly volatile time?

One easy way to short is to use an inverse ETF. As I've written before (Leveraged and inverse ETFs, 3), multiple positive and negative moves within a longer holding period can be quite damaging for inverse ETFs. Looking at the 10-year rate since the middle of 2008, a reasonable worst case scenario to think about would be this: imagine buying in part way down a deep slide, for example when rates first went below 3% on 26 Nov 2008, and then having to wait until rates moved decisively above 3% several months later, on 28 Apr 2009.

(Data through 22 August from the Fed's H.15 release, and today's read from Bloomberg; click for larger image.)

Before looking at specific historical periods, I'd like to look more generally at the level of volatility in Treasuries, and a rule of thumb I proposed in the earlier post on inverse ETFs: volatility is not too dangerous for a holding period of a month, as long as the daily moves are no more than about 2%.

I have used ProFunds Rising Rates Opportunity 10 (RTPIX) before, which gives daily moves inverse to the price moves of the 10-year bond. How much volatility is there in these price moves?

The following histogram shows the probability density of daily price moves, in percent, over the period Jun 2008 - Aug 2011. The price moves are approximated as (-9) * (daily yield change), because the current duration of the 10-year bond is about 9. This is a rough approximation, but it is good enough for present purposes.

For the 10-year, price moves of more than 2%, in either direction, are very rare.

I have been corresponding with Rich Toscano, of Pacific Capital Associates, about this topic, and he mentioned Proshares Short 20+ Year Treasury (TBF), which gives a daily price change inverse to that of the Barclay's US Treasury 20+ Year Index, which currently has a duration of about 18. For the next histogram, daily price changes over the last three years were approximated as (-18) * (daily yield change), where I used the Fed's 20-year constant maturity yields.

Here you can see that the daily price moves are a little wider, as expected. But even in this case, a move of more than 2%, in either direction, happens only about 10% of the time.

This analysis suggests that for both of these tools, a holding period of days to weeks will result in relatively low tracking error, and is fairly safe. A holding period of months could, theoretically, get more dangerous, depending on the shape of the volatility (a large number of large reversals, for example, is worse than one long trip down and back up).

Coming back to real life, here is the actual performance of RPTIX, the 10-year inverse fund, on the worst-case example above:

RTPIX was down 3% over this five month period, when it “should” have been up very slightly. That is certainly an acceptable tracking error, and would not have much changed the outcome of the trade.

TBF, the 20+ year inverse fund, has a shorter history, that does not cover the above period. But we can look at a somewhat similar case, corresponding to the second main dip in the yield graph above. The 20-year Treasury yield first dropped below 4%, for that dip, on 20 May 2010, and decisively rose above 4% on 7 Dec 2010. Over that period TBF, which “should” have been up slightly, was down 1%. That again is a minor problem, and would not have greatly affected results.

Other considerations

  • Both RTPIX and TBF use part of your investment to buy derivatives, and the remainder is put in a money market fund. So, at least when money market funds are paying an amount discernible without a microscope, the interest earned also helps overall returns.
  • Related also to the mechanics of the fund, if we see a real banking panic (low probability but possible), one needs to worry about counterparty risk on the derivatives.
  • TBF has significantly lower fees than RTPIX.
  • The market is anticipating some sort of hint on further rescues from Bernanke, when he speaks Friday at Jackson Hole. Any hint could influence the market in either direction, of course. My own guess is that the political opposition to more QE remains high enough that he will not be able to go there yet.
  • The last few days show that at least the recent free fall has finished and yields have stabilized.
  • The yield on Japan's 20-year is little under 2%, while the US 20-year is yielding a little over 3%. So a reasonable worst case loss scenario on TBF is about 18% (one point change in yield times duration of 18).

Putting it all together

Fundamentals suggest Treasury yields must rise. The rise could be delayed by a crisis in Europe, some new intervention by the Fed, or more play for the Japanification argument. But I think if investors see the 10-year with a negative real yield, the situation is likely to reverse.

If one accepts that argument, the next question is cost of waiting for the market to come to its senses. A volatility analysis on two inverse ETFs shows that a holding period of a few months is likely to be ok.

The free fall in yields seems to be over for now, making this a reasonable entry point.

TBF has lower fees, so I have taken a position with that fund. Since this does involve market timing, and I consider my main strengths to be in long-term value investing, I've taken quite a small position – 1% of the portfolio.