Recent Posts

Wednesday, December 20, 2017

What Is The 10-Year TIPS Breakeven Telling Us?

Chart: 10-year U.S. Inflation Breakeven

The 10-year inflation breakeven rate for the United States is 1.88% at the time of writing of this article. (Link to primer on breakeven inflation.) The usual way of describing this is to say that the "market is pricing in an average inflation rate of 1.88% over the next 10 years." I believe that the simplest interpretation of this observation is the correct one: on average, markets expect CPI inflation to run below its implicit target over the next decade.

NOTE: This article is a followup to a previous ones on breakevens. I want to write a report on breakeven inflation which will be much more technical than my usual style here. This discussion is probably going to be the most qualitative section, and I am very unsure how to write it. I am throwing ideas at the wall, and see what will stick. I apologise for the fact that some ideas might be repeated, These articles will be condensed into a single section, and the train of thought straightened out at that point.

If You Are an Investor, the Economic Breakeven Is What You Want

The chart above is the spread between the 10-year TIPS and the 10-year nominal (conventional) Treasury (Federal Reserve H.15 fitted curve data; 10-year conventional minus the 10-year TIPS quoted yield.) This measure is usually close to the true economic breakeven: what level of annualised inflation leads to the inflation-linked bond having the same return as the conventional bond? For my purposes here, I am referring to the true economic breakeven, the calculation of which might require some corrections based on data that I do not have access to (such as funding differentials).

If you are an investor with a long investing horizon, you should use the difference between the economic breakeven and your inflation forecast as the input to your investment decisions. Adjusting for an unknown inflation risk premium biases your input in one direction or another; your investing process decision rules should tell you what premium you need to make a position attractive.

However, not all users of breakeven inflation data are investors. Economists (particularly at central banks) want to take market pricing and use them to infer an implied inflation forecast. For a variety of reasons, they wish to remove the effects of any risk premia in yields, to get an unbiased forecast.

We can think of the bias in the breakeven inflation rate as being analogous to a term premium. However, since the breakeven inflation rate is the spread between two bonds with separate term premia. I refer to the bias in the observed breakeven inflation rate as an inflation risk premium -- which could be positive or negative.

Is There An Inflation Risk Premium?

One standard way of inferring an inflation risk premium is to decompose the daily movements of bond yields into risk factors using an affine term structure model. This can be done in many ways, but one possible method is decompose observed breakeven inflation into an unbiased inflation expectations term, and an inflation risk premium.

I am not a fan of such methods. My first objection is straightforward: we do not have enough data to know whether such a premium exists. In conventional bonds, we can look at centuries of money market data to see that a premium exists; in fact, a good portion of finance theory makes an a priori assumption that bonds will outperform cash.

The chart at the top of this article is based on the entire run of 10-year TIPS data in the Fed H.15 Report. Trading started a few years earlier, but even so, we do not have a lot of data. We do not have a lot of 10-year TIPS Notes that matured to see whether they were priced in an inefficient fashion when issued.

(As an aside, the chart compares the 10-year breakeven with the 10-year annualised inflation at the same date. This is allegedly a foolish thing to do: I am comparing the backwards-looking realised inflation rate to the forward-looking 10-year breakeven. Even so, one can see the time series tend to track eachother, outside of the obvious divergence during the Financial Crisis. In an environment of stable inflation around an inflation target, this is perhaps not too naive: investors may be assuming that policymakers have a bias in missing the inflation target, and they use the historical miss to calibrate their estimates of the future miss.)

Looking at the above chart, I could easily explain the 1.88% breakeven in two ways.
  1. The market unbiased forecast is above 2%, but there is an inflation risk premium that drives the observed breakeven down to 1.88% (that is, TIPS are relatively cheap). It would be very hard to argue for a premium of up to 50 basis points (2.38% unbiased forecast).
  2.  There is almost no premium, and the market expects inflation to be below the desired levels of policymakers. (Keeping in mind that CPI is biased above the inflation rate of the PCE deflator.)
Needless to say, that gives us a lot of ground to cover. However, it does rule out some possibilities. For example, it seems implausible that the market has an unbiased forecast of 1% or 3% for inflation. If we wanted to be realistic, we could put error bars around the raw breakeven inflation rate to get the range for unbiased inflation forecasts. We really do not have enough data to calibrate those error bars. If we cannot really measure the effect, why are we attempting to model it?

I do not give investment advice or play the forecasting game. However, if I were working for a fund that was bearish on conventional bonds, I would certainly be looking at index-linked bonds as one way to express that macro view.

Boot-Strapping Issues

Another problem with inflation risk premium modelling is the effect of bootstrapping. If one accepts my assertion that the inflation risk premium in 1-year breakevens is low (as described here), this pins down the range for the 2-year breakeven.

For example, let us assume that the 1-year breakeven has a risk premium of zero. If we look at the 2-year breakeven, even an apparently innocuous 50 basis point risk premium has implausible implications. This is because the 1-year, 1-year forward inflation risk premium would have to be double that premium -- 100 basis points. As an analyst, it would be very easy to structure an attractive trade to take advantage of such a premium; a holding period of one year is easily done at some money managers. As a result, so long as the premiums at the very front end are small, the term structure of the inflation risk premium has to be smooth and back-loaded. The reason why I am skeptical about affine term structure models is that they tended to imply highly implausible trades at the front end of the curve.

Concluding Remarks

The fad amongst researchers to subtracting premia from observed economic breakevens is curious. The limited amount of realised returns data means that we have a hard time validating that inflation expectations are systematically mispriced to a large extent. Meanwhile, the uncertainty around everyone's oil price forecast probably dominates any plausible inflation risk premium estimate.

Appendix: Term Premium Comments

My discussion of the term premium was noted on Twitter. This brought up the article by Matthew C. Klein from last year, on the conventional term premium. I have added this appendix to remind myself to see whether I can reference that discussion when I finally write this up...

(c) Brian Romanchuk 2017


  1. As an investor in TIPs I can tell you another source of "premium." In a truly bad inflationary scenario (sustained 4%+ inflation) I believe it is probable that the government would fiddle with the CPI calculation to lower measured inflation and cpi-linked payments.

    1. That’s the “Boskin Risk.” I started covering TIPS around the year 2000 or so, and TIPS investors substituted his name for various cuss words in discussions. It may be that younger investors are less triggered by his name... (As a Canadian, I didn’t have a dog in that fight.)

      The thing is that TIPS investors have an extremely potent lobby on their side - the oldsters. Their pension payments get whacked by CPI fiddling. My bias is that I could hide behind them, and they would do the dirty political work for me. (Of course, I would probably be firmly in the oldster camp when this might matter.)

  2. Decent article. Actually breakevens are biased low, and have been forever, for reasons I cover here: In a nutshell, since TIPS almost never go on special in the financing markets and nominal Treasuries are almost always special, a long position in breakevens always carries worse than a short position, so it needs to be lower than expectations to compensate for this effect.

    A better measure is inflation swaps; 10yrs are currently at 2.20%. But keep in mind that any fair bet on inflation - such as an inflation swap - should be higher than the market's guess as to ACTUAL inflation since inflation has long tails to the upside; ergo there's an option value to being long BEI as opposed to short.

    1. Thanks. I did note that I do not have access to the finding differential, and that adjustment should be made to the raw breakeven. (I discussed that earlier, and so I did not want to repeat myself too much.)

      My difficulty with inflation swaps is that they are a one-way market; the only natural payers are dealers doing asset swaps on linkers. I was with a Canadian buy side institution, so I am not an expert on the details (there was no market in CAD inflation swaps).My feeling is that “technical factors”can move inflation swaps a lot.

      As for the optionality, I was burying that effect into “expectations”. My fair value for trading purposes is not necessarily the median of my forecast. I am still collecting my thoughts for the report, and I would discuss that effect in another section. However, thanks for bringing it up, so that I hopefully remember to discuss it somewhere. It is obviously a strucural factor that hangs over the market.

  3. This comment has been removed by a blog administrator.

  4. This comment has been removed by a blog administrator.

  5. This comment has been removed by a blog administrator.

  6. This comment has been removed by a blog administrator.


Note: Posts are manually moderated, with a varying delay. Some disappear.

The comment section here is largely dead. My Substack or Twitter are better places to have a conversation.

Given that this is largely a backup way to reach me, I am going to reject posts that annoy me. Please post lengthy essays elsewhere.