I have been running into cross-market yield comparisons in the news flow in recent weeks. For example, the raw U.S. Treasury/German bund yield spread often comes up in valuation discussions.The simple rule of thumb is that one should never make such cross-currency yield comparisons; they only matter if the currency value is being pegged. Since I do not have a handy source for euro-denominated bond yields, I will use the Canada-U.S. comparison.
The chart at the top of this article shows the Canadian dollar/U.S. dollar exchange rate since 1995. It moves around, but there are no currency crises in sight.
We can now turn to the hobby horse of economics textbooks: uncovered interest rate parity. According to economists, investors allegedly expect the currency to move in a way to respect interest rate differentials. Let us now look at how useful that theory is for the CAD-USD exchange rate. The chart above shows the 3-month annualised change in the exchange rate, versus the 3-month Treasury bill spread -- which is an annualised interest rate (so we are comparing apples-to-apples). You do not need a doctorate in econometric analysis to see that the Treasury bill spread bears no resemblance whatsoever to realised currency movements. Since there are very few instances where the actual exchange rate change is even within a few hundred basis points of the yield spread, no one in the real world cares about uncovered interest rate parity. This means that the parables told in economics textbooks about exchange rate determination fall apart.
Outside the world of hedge funds, fixed income investors have limited risk budgets. For active investment grade bond managers, an outperformance target of 20 basis points a year is fairly typical, and the risk budget is set accordingly. There is no way to absorb foreign exchange volatility into a fixed income risk budget. (In many cases it would not even be allowed. The "separation of church and state" says that forex managers cannot take duration risk, and bond managers cannot take forex risk.) The only way to trade foreign bonds is to hedge the currency risk. A hedged foreign currency bond position can be thought of as:
- the foreign currency bond;
- a foreign currency loan to finance the bond;
- a local currency deposit to act as collateral.
There is a number of ways of implementing this (bond futures, swaps, a bond and a currency swap or currency forward), but they all end up being economically equivalent to this structure. This means that your operations in the foreign currency look identical to a highly-leveraged investor who operated in that currency. All you care about is the following: will the bond outperform the short-term financing cost of the position? (That explains why the path of short rates matter.) So, you could care less about the absolute yield level of the foreign bonds, rather how they compare to the local currency term structure.
However, there is one group of large investors who are exceptions to my discussion above: foreign currency reserve managers. This is about the only class of investors who act like the investors in economics textbooks, buying and selling foreign currency short-term debt. Although these investors are important, they are not enough to dictate the level of the yield curve.
The only reason to expect nominal yields across different currencies to track each other is the tendency for central banks to herd. One could reasonably argue that the Bank of Canada does not entirely set its policy rate in a vacuum; rather it does look at the level of its policy rate versus the Fed. That said, we have seen large divergences (on the order of hundreds of basis points) in the past, and these divergences have been quite persistent. (The Aussie-Japanese spread historically captured the imagination of many strategists.) As a result, a certain amount of correlation of yields is reasonable to expect (and observed in the data). This means that we can quite often use hedged foreign currency bonds as a proxy for local currency duration. (The strength of the correlation varies greatly; U.S. Treasurys are a good proxy for Canadian dollar duration; bunds, not so much.) However, these relatively persistent policy rate divergences can support observed bond yield spreads, and there is no a priori reason to believe that the fair value of the spread should be zero.
Appendix: Disclaimer Regarding Long-Term ForwardsI would note that I am not entirely happy with my article title; it is too emphatic. However, a title such as "Comments On Currency Volatility Versus Fixed Income Volatility" is not a recipe for an article to get read. The issue revolves long-term spreads. For at least some currency pairs, one can have confidence regarding long-term reversion in currency levels. As a result, a wide spread on 30-year bonds does suggest an attractive long-term trade (although the holding period is probably too long for most institutional investors).
If I look at CAD-USD, there are good psychological reasons to expect that the currencies will not move "too far" away from the 1:1 exchange rate. The countries have similar inflation targets, and there is a lot of nominal cross-border price comparisons. A 2:1 exchange rate (either way) is probably the limit at which Canadians at least would probably take action to restore parity (outside of some hypothetical currency crisis which might allow a temporary deviation). So a wide bond yield spread on 30-year bonds could imply what appears to be an implausible 30-year forward exchange rate, and so suggest a long-term cross-currency bond trade. (I have written in the past how I bought U.S. Treasurys on this theory earlier, noting that I am a Canadian.)
There are other currency pairs that probably have this property: AUD-NZD, euro versus European currencies (EUR-GBP, EUR-CHF). I leave the plausibility of these arguments as an exercise to the reader.
The key to the matter is that such trades are in fact a currency trade, and would be verboten for most institutional fixed income investors. Nevertheless, one could argue that the yield spreads do matter in this case, contra the title of this article.
(c) Brian Romanchuk 2018