Why So Many Inflation Market Haters All of a Sudden?
The inflation market offers such wonderful opportunities for profit since so few people understand the dynamics of inflation, much less of the inflation market.
One of the things which continually fascinates me is how the inflation trade has become sort of a “risk on” kind of trade, in that when the market is pricing in better growth expectations, reflected in rising equity prices, inflation expectations move with the same rhythm.[1] The chart below (source: Bloomberg) shows the 10-year inflation breakeven rate versus the S&P 500 index. Note how closely they ebb and flow together, at least until the latest swoon in inflation expectations.
Your knee-jerk reaction might be that this is a spurious correlation caused by the fact that (a) bond yields tend to rise when growth expectations rise and (b) when bond yields rise, the components of bond yields – including both real rates and expected inflation – tend to rise. But look at the chart below (source: Bloomberg), covering the same period but this time charting stocks versus real yields.
If anything, real yields ought to be more correlated to movements in equities than inflation expectations, since presumably real economic growth is directly related to the real growth rates embedded in equity prices. But to my eyes at least, the correlation between real interest rates and equity prices, for which there is a plausible causal explanation, doesn’t look nearly as good as the relationship between stocks and inflation expectations.
It doesn’t make any sense that long-term inflation expectations should be so closely related to equity prices. I know I have mentioned before – in fact, regular readers are probably sick of me pointing it out – that there is no causal relationship between growth and inflation. Really, it is worse than that: one really needs to torture the data to find any connection at all, causal or not. The chart below (source: Bloomberg) shows quarterly real GDP in yellow against core inflation in white.
I’ve pointed out before that the big recession in ’08-09 saw almost no deceleration in core inflation (and none at all if you remove housing from core inflation), but it’s hard to find a connection anywhere. The next chart (source: Bloomberg) makes the point a different way, simply plotting core inflation (y-axis) against real GDP (x-axis) quarterly back to 1980.
In case anyone out there is protesting that there should be a lagged relationship between growth and inflation, I am happy to report that I was able to get an r-squared as higher as 0.257 with a lag of 12 quarters. Unfortunately, the lag goes the wrong way: high inflation precedes high growth, not vice-versa. And I don’t know anyone who proffers a reasonable explanation of a causality running in that direction. Lags the other way fail to produce any r-squared over 0.1.
So, how to explain the fact that since the end of August we have seen 10-year real rates rise 29 basis points while 10-year breakevens were falling 12bps (producing a net rise in nominal rates of only 17bps)? The explanation is simple: the market is wrong to treat breakevens as a “risk on/risk off” sort of trade. Breakevens have cheapened far too much. I don’t know if that means that TIPS yields have risen too much, or nominal yields have risen too little (I rather expect the former), but the difference is too narrow. The short end of the US curve (1 year inflation swaps are at 1.44%) implies either that core inflation will decline markedly from its already-depressed level well below median inflation, or that energy prices will decline sharply and much further than implied by gasoline futures.
I think that one of the reasons US inflation has been under pressure is that it is currently at a very large spread versus European inflation, and earlier this month it was at the highest level in at least a decade (see chart, source Enduring Investments).
This may look like an appealing short, perhaps, but based on our internal analysis and some historical relationships we track we actually believe the spread is too low by about 50bps. And think about it this way: if Europe really is in the process of inheriting Japan’s lost decade, then 10-year expectations for the US ought to be much, much higher than in Europe. I don’t really think Europe will end up there, because the ECB seems to be trying to do the right thing, but it is not unreasonable to think that there should be a hefty premium to US inflation over Europe.
[1] Of course, the correlation of levels won’t be very good because stocks have an upward bias over time while inflation expectations do not. To run the correlation, you’d have to de-trend stocks but I’m trying to make a visceral point here rather than a quantitative one.
Leave a Reply to Michael AshtonCancel reply
“In case anyone out there is protesting that there should be a lagged relationship between growth and inflation, I am happy to report that I was able to get an r-squared as higher as 0.257 with a lag of 12 quarters. Unfortunately, the lag goes the wrong way: high inflation precedes high growth, not vice-versa. And I don’t know anyone who proffers a reasonable explanation of a causality running in that direction. Lags the other way fail to produce any r-squared over 0.1.”
I think the issue here is because you are looking at GDP yoy growth and Core Inflation yoy as equivalents — they are not. GDP yoy is first moment of a nominal amount, and Core CPI is first moment of a price construct. To make them comparable, you have to raise both of the variables to their second moments. Therefore the second moment of CPI should be compared with the second moment of GDP. The second moment state defines their variance. Try it and you will see a very good match when GDP is lagged 2 quarters, which makes sense — definitely much much better than an r-squared of 0.1.
Also, I think you are doing a very good job at explaining the phenomenon of inflation to the investing public. This is giving credit where credit is really due.
Anonymous fan
Thank you! Always good to have fans!
If I understand you, you mean to take say the standard deviation of changes in GDP versus the standard deviation of changes in CPI. When I do that (and just for grins taking the standard deviation of the last 8 quarters, which isn’t a true standard deviation because the y/y overlap but good enough for our purposes), I do indeed get a higher r-squared of about 0.37 from 1970-present. Interestingly, that doesn’t improve if I merely do 1985-present, which surprised me.
If in fact that’s what you mean, then it’s interesting but something short of Keynesian claims isn’t it? It means that growth volatility leads to price volatility with some lag. That makes PERFECT sense to me! But it doesn’t say that growth leads to increases in price.
Am I understanding your point correctly? I must admit I’ve never thought about the second moment of GDP.
I really love comments like this because it gets me thinking about new things. That’s one big reason I write this article. Someone on Twitter yesterday also suggested that headline inflation leads core inflation ex-housing…which seems weird, but when I looked at it seemed quantitatively interesting. I can’t puzzle out the causality, but it gives me something to think about. So thanks for raising the level of discourse!!
I often learn something from the broader themes of your articles without understanding the details. What are “breakevens” and “inflation swaps?” Maybe you could you write an article called “Retail Inflation Investing: 101.” Or recommend readings if you don’t want to write the article. In any case, thanks for the articles. I like that you think for yourself.
Thanks John, and thanks for the article suggestion. I might do that!
A “breakeven” describes a position of being long TIPS and short nominal Treasuries. Since Treasury yields are approximately equal to real yields (represented by TIPS) and a premium for inflation expectations, doing this trade exposes you to only the changes in inflation expectations. So if TIPS sell off, but nominal bonds sell off more, you make money as inflation expectations rise.
Inflation swaps do the same thing but in a single package. In a conventional zero-coupon inflation swap, one party pays a fixed rate of interest, compounded and paid at the end of the swap, and the other party pays the total increase in the price index (price index at end / price index at beginning – 1). So buying the swap means you are expecting inflation to be higher than current expectations embedded in the swap, or you are expecting inflation expectations to rise, or both.
Actually the dichotomy in the inflation-GDP relationship started in 1987, when the Fed allowed interests to be paid on deposits — that completely screwed the concept of money demand (you had then, and still have, CDs that are paying higher than the policy rate of same maturity). That impacted and changed the relationship of money supply (all flavors) to a lot of variables (growth, and yes, inflation too) and skewed money velocity to the downside. Even the US Dollar was impacted as well, as you can imagine. So when I look at growth, inflation and the USD, there is pre-1987 and post-1987. Actually for the USD, there is also a special period between 1971 — when the currency’s link to gold was severed– until 1987 — that is the most problematical part of the US Dollar’s history.
Using moments is an engineering solution a lot of people may not be familiar with. There is a quick and dirty way of showing the real covariance of inflation and GDP, by going into the second derivative of both variables — which is not exactly as elegant as the moment-type of solution, but will be sufficient to prove the point. Use actual change of the percentage change to obtain the second derivative. You may be surprised by what you get.
Extending this concept further: most investors believe that the ebb and flow of GDP is the seminal factor that governs the changes in the USD. Use the same simple trick to see the real USD-GDP covariance — it shows that the causality really goes to other way. Which makes full sense — the USD impacts net exports which is a direct component of GDP (that is fact the Fed is saying today, complaining about the strong US Dollar). Go all the way using the same simple trick and put all three together — GDP, CPI and USD. This will give you the temporal order which governs all three. Enjoy . .
Anonymous fan
Thanks. I typically make the inflation break at 1993-94, model-wise…the parameterizations seem to allow most conventional models of inflation to fit well on both sides but with different weightings. (This is one reason economists developed the idea of “anchored expectations” although we cannot measure accurately either consumers’ expectations nor their anchoring.) I am always very excited when I find a model that bridges that period without a dummy variable.