Recent Posts

Wednesday, February 19, 2020

Inflation Is NOT The Most Significant Factor Determining Bond Prices

One of the pieces of pseudo-science that floats around in popular discussion of bonds is the belief that bond investors are deadly afraid of inflation. In particular, bonds "lose money" every time the Consumer Price Index rises -- which is most months, in most developed countries. As far as I can tell, this is the legacy of some Economics 101 textbook story that has been passed on from "expert" to "expert" over the decades.

The correct answer is that nominal yields largely reflect the expected path of the short-term nominal policy rate, and is thus a reflection of the central bank's "reaction function." (At this point, some people will jump in and start going on about the term premium. However, unless we using an obviously dysfunctional term premium model, the term premium is only a small deviation from the fair value determined by rate expectations.)

The advent of inflation-linked bond markets adds some extra qualifications to the previous statements. (My previous book discusses the inflation-linked markets.) If a bond investor is overweight inflation-linked bonds, they can get a big fat bonus if inflation turns out higher than expected. However, the beliefs about bond investors that I am discussing here were formed in an era when inflation-linked markets did not exist, so economics textbook writers were free to write in whatever campfire ghost story they wished.


The genesis of this article is a question that showed up on the Economics Stack Exchange. The person who posed the question quoted from a primer that came from a less authoritative website. The underlying premise behind the question is that real yields are constant.

Since the arguments I am discussing were pretty shaky, it is difficult to trace where they came from. However, my guess is that the logic worked as follows.
  • Irving Fisher came up with the story that we can decompose nominal bond yields into inflation expectations and a real yield. (I.e., nominal yield = real yield + inflation expectations.) This fit the biases of classical economists.
  • Historically, economists lacked the mathematical tools to deal with yield curves, and so they made up toy models in which a short-term interest rate, and a "bond yield" floated as somewhat independent variables. (Note that this was not purely a "classical economics" view, it goes back to Keynes. Although Keynes was a market participant and was aware of rate expectations, putting an entire yield curve into an economic model was difficult.) They assumed that bond investors were not leveraged, and only cared about the real value of their bond holdings.* As such, they assumed that bond markets cleared based on the real yield.
  • Since it was unclear what would cause the market-clearing real yield to shift, the story went that almost all nominal bond yield movements were due to changes in inflation expectations.
This logic was hard to refute when the only data available was nominal yields. However, the advent of inflation-linked bond markets means that we can dig further.

Inflation Means Bonds Lose Money!

The story that inflation causes bonds -- and only bonds -- to "lose money" is frankly bizarre. 

Imagine that you hold an equity mutual fund, a bond mutual fund, and a money market fund. Each fund has a certain Net Asset Value (NAV). The government suddenly raises the sales tax and boom! -- the price level rises. According to the usual story, you were hit by "losses" on your bond and money market funds, while the equity fund did not lose money. Really?

If we think about this for more than five seconds, we need to look at what happened to the NAV of the funds. Guess what? Since the early 1980s, the United States and other developed markets were in a secular bond bull market. Meanwhile, there was a positive inflation rate most of the time. Bond funds were rising in value, both in nominal and real terms. 

You need to look at actual returns data, and funnily enough, the story is that practically anything can happen, particularly over short holding periods.

U.K. Experience

The first modern inflation-linked bond market was developed in the United Kingdom. At the time, the U.K. had quite poor inflation performance, and in fact, the markets did feature stable real yields and oscillating inflation expectations.
Chart: U.K. 10-year Real (Zero) Yield

The chart above shows the 10-year (zero) real rate from the Bank of England historical yield curve data set. We see that the real yield was in fact stable near 4% in the 1985-1995 period.

However, inflation finally got under control, and the volatility moved to the real yield. The collapse in nominal yields coupled with steady inflation meant that real yields went quite negative. If we look at other developed markets like Canada and the United States, we see an initial period of mis-pricing then the inflation breakeven stabilising around the 2% level. This meant that the big swings in nominal yields resulted in real yield volatility.

In other words, inflation was no longer a source of returns volatility. This is what we would expect to happen if policy results in inflation sticking near target. (Whether the policies that drove this outcome is monetary policy versus other policies is a point of debate between the mainstream and MMTers.)

(Nick Rowe periodically discusses this, and attributes this to Milton Friedman's "thermostat analogy." The idea behind the analogy is that if a home heating system is functioning properly, there will be no correlation between furnace energy consumption and interior temperature -- since the interior temperature is roughly constant. However, this does not imply that the furnace has no effect interior temperatures. Given my control systems background, this point seemed so obvious that I never saw a need to dig into which economists said what about it.)

What Determines Nominal Bond Yields?

The following factors determine nominal bonds yields. (Note: this list is comprehensive, but I wrote it quickly, and may have missed something.)
  • Base fair value is rate expectations, which is an embodiment of the central bank's reaction function. To what extend inflation matters, it is showing up here.
  • Term Premium. There is good evidence that investors want compensation for taking duration risk, although this was not always the case. There are various wacky academic models that allegedly model term premia, but my argument is that those models tend to be GIGO exercises. (The remaining factors might be lumped in with the "term premium," but should be thought of as distinct entities if we are being careful.)
  • Institutional pressures. Regulators can do things to force investors into various parts of the yield curve, distorting markets. (Could be lumped in with a term premium.)
  • Funding cost differentials. If you can fund a bond cheaper than other bonds, its yield should be lower. Regulations can influence funding costs.
  • Taxes. Most of the distortions from taxes have been removed over the years, but we can premiums for discount/premium bonds for tax reasons.
  • Market Squeezes. There are investors with big positions, and dealers test whether they are vulnerable to squeezes.
  • Fat fingers, or rogue algorithms. (Everything else.)
In summary, there are lot of things that can affect bond prices in the real world. To what extent inflation shows up, it is how central banks react to it.

What About Hyperinflation?

I tweeted earlier about this, and the topic of really high inflation "breaking" my arguments was raised. I do not see that as a serious dissent from my view -- the inflation surprise would likely cause a reaction by conventional central bankers. However, based on my experience, almost every time an economist interjects "inflation!" into a discussion of bond yields, they are about to be totally wrong on the duration call. 


* Even if you are concerned about the real value of your portfolio, that does not mean that you are worried about the real yield of bonds. For example, imagine that inflation is 2%, and the bond yield is 1%. Although that is allegedly bad, apply enough leverage to that sucker, and you get a 20% nominal return (if you are lucky...). To be clear, I am not recommending that investors run out to leverage bond portfolios, but want to make it clear that pricing is driven by relative value traders that apply leverage to their positions. As such, they are just interested in positive nominal returns, which are then dialed up.

(c) Brian Romanchuk 2020


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

  2. In the footnote, "apply enough leverage to that sucker"- you are talking about borrowing against a bond that you already own, right? I mean, it seems unreasonable to me to borrow at least 2% to finance a purchase of a bond that pays 1% nominal gain and has a negative real return.

    Well, I guess that my bank borrows the 'money' I keep in my checking account for a whole lot less than the rate of inflation. So maybe it would make a little bit of sense for them to use that to purchase a bond that paid more than they pay me even the bond itself had a negative real return. So long as the money I loaned them had an even more negative real return.

    Maybe you could explain what you mean in the footnote so I don't have to guess? I'm terrible at guessing- which could explain why I leave money in a checking account at negative real rates of return...

    1. If you are a big investor, you borrow at the repo rate (which is what everyone has been going on about). That’s the risk-free short-term funding cost that matters for setting the Treasury yield curve.

      Never something I worried about, but you can get hefty leverage with repo. Pre-crisis you used to be able to get into a position with something like 2% down.


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.