Market design bleg

Some people on Wall Street argue that conventional Treasury bonds are preferred to TIPS, and hence the TIPS spread underestimates actual market inflation expectations. I’m inclined to agree.

The current 5-year TIPS spread is 1.34%, which is based on CPI inflation. That suggests a PCE inflation rate that is barely above 1%, as PCE inflation tends to run at least 0.25% below CPI inflation, and indeed recently the gap has been even larger.

In contrast to this implied 1% PCE inflation forecast, the Fed expects slightly below 2% inflation over 5 years, as does the consensus of private forecasters. I also expect inflation to be only modestly below 2%, which has been the pattern over the past two years.

Of course we could all be wrong, but what if the TIPS spread is biased?
In that case, how can we derive the actual market forecast of inflation?

You could use a CPI futures market. But if this market were tiny relative to the TIPS market (and I believe it is) then arbitrage would probably force CPI futures prices to correspond to TIPS spreads, even if traders thought the actual inflation rate would be higher. (As an analogy, consider the situation if bond market and currency futures market traders disagreed about the future path of exchange rates. The interest parity condition would still hold approximately true due to arbitrage.

Now for my question. How can a less biased inflation indicator be constructed? For instance, what about creating a security that paid $1000 if the 5-year inflation rate ended up being in the 1.4% to 2.6% range, and zero otherwise. My gut instinct is that that outcome is very likely, say over 85% likely. If the market agrees, wouldn’t those contracts sell for at least $500, even if the 5-year TIPS were predicting 1.0% inflation? If so, wouldn’t that prove that inflation expectations are actually well above 1.0%?  I’d certainly pay $500 for that sort of security. Or would arbitragers force any newly created market on inflation expectations to correspond to the implied TIPS prediction, no matter how the new market is designed?

In other words, is there a market design that overcomes the “arbitrage” problem and gets us close to the truth (about people’s expectations), or is this an impossible goal?

Paging Robin Hanson.



20 Responses to “Market design bleg”

  1. Gravatar of Randomize Randomize
    14. October 2019 at 20:17

    If TIPS are undervalued by 65 basis points, I need to stop wasting my money on traditional treasuries!

  2. Gravatar of Matthias Görgens Matthias Görgens
    14. October 2019 at 23:25

    I don’t understand the problem. If the problem in the TIPS spread is about something that’s idiosyncratic to the difference between the two kinds of bonds, arbitrage with the CPI fututes market shouldn’t transfer those idiosyncracies?

    Eg assume holding a TIPS bond gave you a painful electric shock every morning. That would certainly contribute to a bias in the TIPS spread, but you can’t construct an arbitrage to move that spread to the CPI fututes market, or can you?

  3. Gravatar of Martinghoul Martinghoul
    14. October 2019 at 23:46

    The market, on the whole, is quite obviously not risk neutral, and had never been. Secondly, there are significant limits to arbitrage (liquidity, balance sheet availability, etc) that would not allow the market to attempt to price these instruments properly.

    If you think TIPS market is disappointing, you should take a look at the other inflation markets, especially Canadian RRBs.

  4. Gravatar of P Burgos P Burgos
    14. October 2019 at 23:53

    If the arbitrage is specific to bond markets, are there other sorts of financial products that also track inflation (and inflation expectations) pretty closely? Do all of the various commodities have too many idiosyncrasies for any financial product based on their value to serve as gauges of inflation? What about a basket of commodities?

  5. Gravatar of Benjamin Cole Benjamin Cole
    15. October 2019 at 00:00

    If conventional US Treasuries are the largest and most liquid market, then it seems to me they are standard for measuring anticipation of inflation rates.

    US 10-year Treasuries now yield about 1.75%.

    Such a low interest rate suggests either inflation will be very low for the next 10 years, or there is a global capital glut and a flood into safe havens.

    People used to say they were entitled to a return on their savings. This was never technically true, but the idea is that capital must earn a return or it would evaporate.

    However, when there are chronic gluts of capital, then people are entitled to a loss on their savings.

    When the Federal Reserve, or the Hong Kong Monetary Authority, add more capital through QE to already glutted global capital markets…well, then what?

  6. Gravatar of Martinghoul Martinghoul
    15. October 2019 at 00:58

    The nominal treasury market, in spite of being deep and liquid, simply cannot be used as a barometer of inflation or rate expectations. There are all sorts of papers out there talking about the shortage of “safe” assets globally (partly to do with demographics, but there are other considerations). Furthermore, there’s all sorts of evidence that the “risk-free” rate can deviate significantly from any reasonable estimate of “fair value” during periods of risk aversion (basically, skewness and kurtosis of returns matter bigly for “risk-free” rates).

  7. Gravatar of Benjamin Cole Benjamin Cole
    15. October 2019 at 01:36

    Larry Summers: US economy is just one bad recession away from zero rates or worse


    I guess if some portion of the market expects zero rates down the line, then 1.75% for a 10-year Treasury is a good price….

  8. Gravatar of John Hall John Hall
    15. October 2019 at 02:29

    Interesting post.

    One obvious thing is that if TIPs breakevens are biased one way or another (because of risk premia), then you can measure the bias and adjust for it, assuming it is stable (the liquidity premium isn’t, but the Fed could use special liquidity facilities if needed).

    I know that some people look at inflation expectations embedded in swap curves, but I don’t know if there’s evidence that it’s better or worse than that from breakevens.

    You suggest a security that pays off if inflation is within a range and does not pay off otherwise. This would be valued like a combination of binary options (a cash-or-nothing call at 1.4% and a cash-or-nothing put at 2.6%). I’m not sure this would really get around your issues because there still might be risk premium embedded in the call/put option prices, e.g. the implied volatility of S&P 500 options is skewed because market participants demand more downside protection than upside protection.

    As a side note, I believe I had previously suggested an approach to target NGDP that takes into account the distribution of expectations using options (though I don’t think I had suggested using binary options). This seems consistent with the guardrails approach you have favored recently. The interesting thing to me is what happens when the downside risks to something like NGDP increase, even if the mean doesn’t change, i.e. the tails get fatter. I think you could make a case that the central bank should respond to that. In your binary option approach, then the implied vol should increase when the tails get fatter. So it should take that into account, though I haven’t thought about it thoroughly.

    On a somewhat related basis, here’s a FT article on breakeven inflation and term premia you might find interesting:

  9. Gravatar of Kevin Erdmann Kevin Erdmann
    15. October 2019 at 06:55

    Pre recession tips spreads ranged between 2% and 2 1/2%. That seems about right to me. Do you think expected inflation was biased too low before the recession?

  10. Gravatar of Nick Nick
    15. October 2019 at 07:00

    Scott, there is an inflation swap market. the swaps pay the difference in the index at the start and end, they trade as a zero coupon rate. currently the rate on US 5y cpi swaps is 1.61%, approximately 0.25% above the rate implied by the difference between TIPS and Treasuries.

    The TIPS have a few good reasons for yielding a little more than treasuries, for example they are harder to sell at a time when one needs money quickly, they are also typically harder to pledge as collateral in other transactions.

  11. Gravatar of Tom P Tom P
    15. October 2019 at 07:12

    Basically all bonds, including TIPS, trade at a discount to nominal Treasury bonds because Treasuries provide more liquidity. So while it seems unlikely that you could create a version of TIPS that has the liquidity of the
    (nominal) US Treasury market, you could instead compare TIPS yields to yields on bonds that are less liquid but arguably just as safe; say, corporate debt issued by GSEs. (Although those may be less liquid than TIPS; it’s hard to exactly equalize liquidity.) This liquidity premium is not stable through “big” market moves, but is probably something like 40 bps. The new paper by van Binsbergen, Diamond, and Grotteria seems to accord with my experience in the market.

    The inflation swap market does not suffer from this issue to the same extent, as there isn’t so clearly a liquidity “differential” between the two sides of the market. That is not to say that prices equal expectations in this market, either, though. There could still be a risk-based discount on the side betting for higher inflation because that side would do poorly during a downturn. I’m not sure how big this would be.

    You could also look into the liquidity adjustment for breakeven spreads proposed by Pflueger and Viceira.

  12. Gravatar of Justin Justin
    15. October 2019 at 07:56 should work. Short term 1- and 2-year horizon contracts. Position limits to prevent meaningful arbitrage. Target variable would be core PCE index.

    Failing that you can certainly correct the TIPS spread statistically, or look at 5-year-5-year forwards, which removes the liquidity premium. I’ve always thought the Cleveland Fed’s numbers looked plausible

    I’ll add on that an average inflation miss of 0.6% (core PCE, not CPIU) over ten years is atrocious. 10 years is certainly more than enough time to hit your target, on average, which is basically the same as level targeting. The Fed has decided to assign itself an arbitrary goal for price stability (2% CPCEI) and then run biased policy, deliberately undershooting that target by ~30%, for years. Running an opaque policy that produces a certain amount of Fed-induced market volatility, as expectations oscillate is a feature they’re unlikely to give up. You can guess why.

  13. Gravatar of ssumner ssumner
    15. October 2019 at 08:24

    Kevin, I’d say less so than today, but probably to some extent even back then. Recall that CPI inflation peaked at 5.5% in mid-2008, there was good reason to put some “tail risk” into the prices before 2009.

    Nick, I thought the two markets were closer. Why can’t that difference be arbitraged away?

    Tom, Good point. How do GSE bond yields compare to equivalent Treasuries?

  14. Gravatar of Michael Rulle Michael Rulle
    15. October 2019 at 08:33

    While Tips always sounded great, and I certainly thought they made great sense, the fact is, for all practical purposes the market has rejected them. There is likely something about the payout structure which creates cognitive dissonance. They were treated like risky assets in 2007-2009. Early in my career, we created an instrument which was effectively a bond that would convert automatically into equity in ten years——but at a tremendous discount to the implied forward price. It was a pure risk-less arbitrage. We even had Fisher Black “certify” its favorable pricing features. It was a complete bust.

    Markets like simple things—-or at least things which they believe are simple. Your idea, while fascinating is too geek like (Even as I have always been a sucker for great geek ideas) and its liquidity would likely be non existent. Bridgewater’s Ray Dalit, whose all weather idea is perhaps brilliant, always knew the weak link in his Risk Parity idea was how to trade “inflation”. Commodities are not really an asset in the traditional sense, and with the exception of a stretch in the 70’s, earn zero in real terms. So they pushed strongly for TIPS. For 40 years Risk Parity with just stocks and bonds has far out performed All Weather. You will need a pretty large inflation period to earn that back.

    I have no solution for this. T-bills capture expected inflation, and Tips were meant to capture “unexpected” inflation—where the real risk lies. My best guess is buy and hold some amount of commodities and Tips, but at a low percent of portfolio.

    It will be hard to find an instrument which definitively captures forward inflation expectations.

  15. Gravatar of Brian Donohue Brian Donohue
    15. October 2019 at 08:39

    There are at least 3 issues with TIPS compared to regular Treasuries:

    1. Liquidity. Greater liquidity among Treasuries means they enjoy a lower yield than TIPS.

    2. Sharpe ratio. TIPS provide inflation protection that nominal Treasuries, so nominal Treasuries have a higher expected yield similar to how (riskier) corporate bonds have a higher yield than Treasuries.

    3. TIPS aren’t “marked down” in the event of deflation, so, as the risk of deflation increases, TIPS yields are lower than Treasuries.

    Of course, there’s yet another issue of how well expected inflation is (or is not) realized.

    So yeah, it’s not a perfect match, but it’s a huge amount of information compared to a pre-TIPS world, and, generally speaking, CHANGES in TIPS spreads are indicative of changes in the stance of monetary policy.

    Let’s understand the tools we have rather than pining for what’s around the next corner but may never get here.

  16. Gravatar of Tom P Tom P
    15. October 2019 at 10:00


    AAA corporates in the 3-5 year maturity range are trading about 34 bps above Treasuries; in the 7-10 year range, the number is 37 bps.

    US Agency debt is trading at +28 bps in the 3-5 year range and +50 bps in the 7-10 year range.

    I think this is fairly consistent with the idea that the liquidity premium is around 40 bps in “normal” times.

  17. Gravatar of John Hall John Hall
    15. October 2019 at 10:49

    Also Scott, the TIPs have an implied put option. If there is deflation by maturity, then you still get back your original principal.


  18. Gravatar of stoneybatter stoneybatter
    15. October 2019 at 14:30

    There is a market for inflation caps and floors, which are what you are describing. Their pricing is hard to find and I’m sure they have liquidity issues as well, but here is a relevant summary of them:

  19. Gravatar of Nick Nick
    16. October 2019 at 03:53

    Scott “Nick, I thought the two markets were closer. Why can’t that difference be arbitraged away?”

    In short, it can and to some extent it is, but returns are not amazing for a few reasons. Most people who would arbitrage it away would be leveraged players. in principle one will have to carry the arbitrage to maturity of the 2 bonds, you will have to borrow the treasury and lend the tips, that spread probably averages something in the region of 15-20bps. then you have regulatory capital you need to deploy, this would be of order 60bps (you dont lose it you just have to set it aside for 5 years). so to make $1million you need to deploy $3m in a margin account for 5 years, a return of circa 7%, whilst very attractive for something that is a high certainty you do run regulatory risks and mark to market risks. Repo is an obvious risk as you may have observed recently.

    I should note that one can do better than this when these trades are part of a larger portfolio.

  20. Gravatar of Nick Nick
    16. October 2019 at 03:56

    sorry calculation was wrong. youd need circa $6m in a margin account for 5 years to return $1m, so a return of circa 3.5%.

Leave a Reply