Financial market practitioners are, if nothing else, pragmatic in their approach to analysis. Finding something that works is great. And when something stops working it tends to be discarded. After a twitter convo last night with some smart folks I’m going to outline something that I have tried and discarded in the realm of seeking an answer to the question as to how the long dated real yields across the world might be determined by macroeconomic variables.
Some years ago when I was a full-time bond geek I asked my then CIO (the wonderful Michael Hughes of BZW Equity-Gilt Study fame) how he thought about assigning relative value in the inflation-linked bond market. (Please bear in mind that this preceded the ‘popular’ preoccupation with discerning the true r-star by some years.)
Simple, he told me. Think of the real yield as the bid for international capital: there should be an inverse relationship with current account balances. I regressed 10yr real yields with current account deficits as a percentage of GDP, and there was an okay fit. Markets are forward-looking so I took consensus c/a balances (proxied by IMF forecasts). A better fit.
Here’s the chart* with and without the UK using today’s IMF forecasts:
It’s not the worst chart in the world at first glance (although @Barnejek might disagree). It gave me a speck of hope that there might be some useful stuff here that could improve on King & Low 2014.
King & Low sought to provide a ‘world interest rate’ series that can be used by all sorts of people. And it has since been heavily cited. They basically take an unweighted average of available ten year inflation-linked bond yields, and also a GDP-weighted average of the same instruments (finding them pretty indistinguishable). As far as papers by former G7 central bank governors go it is unextraordinary. But I like the idea of a global real yield and for play for having a stab at it.
As a practitioner, developments within one market appear to drive others, although the ‘aggressor’ market appears to change. (So, for example, it could be US Treasuries setting the tone for the world’s bond markets until the ECB decides to institute QQE and starts driving the car. And then the BoJ rocks up and declares Yield Curve Control. That sort of thing.) In financegeek language, developed market bond yields appear to be cointegrated. This is uncontroversial.
Finding a world interest rate that recognises adjustments for the international balance of payments (which looks from these charts like they might be a *thing*) could be interesting. Rather than looking at the simple average of the historically small number of issued bond yields and calling that the world interest rate, you could instead plot the course of the intercept of the regression line on the chart over time: a line not distorted by a country’s idiosyncratic balance of payment positions (or at least aggregating and controlling these idiosyncratic distortions).
When I tried this approach I came up with an answer that was not radically different from King & Low. The green line in the chart below is not far from King & Low’s blue line (which I extended by updating their dataset) and is modestly below (taking account of the degree to which King & Low’s issuers typically run current account deficits, and that they hadn’t made any credit spread adjustments to their data). Not that different. But better?
However, we now reach the disappointing part of the story. How stable is the relationship shown above in the nice scatter/bubble charts? Not very. As the next two charts show, the r-square of the relationship goes to pot 2008-2014 (although here I am using actual c/a balances rather than forecast ones). And the slope of the line actually flips from negative to positive in 2008 (admittedly, some really reallly crazy stuff was going on in international inflation-linked markets then – stuff that I still talk to people about today, so maybe that’s forgivable).
I haven’t GDP-, M3-, or duration-equivalent-bond-market-size-weighted the dots to find this r-square, so maybe things improve with a bit more work.
Bottom line: The cutting room floor is littered with ideas that don’t quite make the cut. I really like this way of thinking about international real yields, but am disturbed by the episodic nature of the relationship actually ‘working’. Pragmatism leads me to look in other directions if I want to link international real yields to international macroeconomic variables.
If anyone has had more success in this approach please do drop a comment in below.
* I take the average IMF c/a balance for 2017-2021 as a percentage of GDP and compare here to 10yr real yields on inflation linked bonds as at 28th November 2017. I have adjusted the UK for the RPI-CPI ‘wedge’, and the Eurozone real yields by the difference in their 5yr CDS and Germany’s 5yr CDS so that it doesn’t become a credit quality chart. The bubble sizes are common currency GDP; I was wondering whether they should be common currency M3 or bond market size in 10yr duration equivalents. Not sure.