Understanding The Lack Of Relationship Between Supply And Bond Yields

Academics and strategists have spent decades trying to prove that increased deficits raise bond yields, and publishing a flood of papers that allegedly prove this link. However, nobody competent takes any of those papers seriously. This article outlines why these papers are largely doomed to failure, and are mainly exercises in how to figure out academics tried to lie using statistics.

If the Proof Existed, You Would Hear About It

If a robust econometric study linking fiscal deficits to bond yields, you would hear about it. The reality is that you do not. Instead, you hear people discussing obscure papers where academics allegedly proved something or other. In every case, the textual claims in the abstract fall apart once you look at the actual mathematical methodology.

(An alternative version is to come up with an effect that historically moved the 10-year yield by 30 basis points or less. Given that a good ISM report used to move the 10-year by that amount in seconds in the gold old days, such an effect is statistically indistinguishable from zero, and thus not safe to extrapolate.)

Econometric Problem is Extremely Difficult

There are major difficulties in trying to link deficits to bond yields. There are a number of factors that explain this.

• If we go backwards in time, bond yield data from before 1980 or even 1990 are not comparable to the current environment. Currencies were pegged, and/or interest rate regulations existed. For example, the rate hikes deployed by the Bank of England to defend the pound during the ERM debacle are not a typical feature of floating currency sovereign interest rates.
• Even when we have floating currencies, the number of "independent" observations is not particularly large. Once the ERM fiasco settled down, the developed countries largely had synchronised interest rate cycles. The largest outlier -- Japan -- blows up any theory that links debt/deficits to bond yields, so our intrepid academics either ignore Japan, or add epicycles to their theories to explain it away. The other set of outliers were the "flat curve" countries, mainly the U.K. and Australia. Those countries feature hyperactive housing markets, where the financing tends to be floating rate. As such, there was sensitivity to the policy rate, and the curves remained flat. 
• In countries featuring a cycle and steep curves (U.S., Canada), the yield curve and deficits are cyclic. The level of rates drops during recessions -- exactly when deficits are the largest. Since that is the "wrong" answer, academics start to pretend that they can compensate for the effect of the cycle. 

To what extent "supply and demand" show up in yield curves, it is typically in the form of squeezes in long duration bonds. The U.K. curve was a relative value horror show since the mid-1990s, courtesy of pension regulations. Meanwhile, the market is closely held by big players, and they have a tendency to squeeze foreign relative value investors who move into their turf. (The exact same thing happens in Canada.) The other example is the ability of central banks to squeeze the long end during massive QE expansions of their balance sheets. If the central bank buys up all the bonds, there is not much of market left to seek out fair value.

However, in both cases, the issue is that long-term yields are too low, which is not the preconceived answer that academics and strategists want to see. They want higher yields.

Institutions Matter

Why don't we see bond yields rocketing to the moon on supply? Institutions.

Every developed country has the equivalent of the U.K.'s Debt Management Office. They launch long, boring discussions with both the buy side and the sell side to see what maturity structure investors are interested in. They only issue long-dated debt if there is a proven demand for it. If they did something stupid and dropped billions of 50-year debt on the market that could not be absorbed, those bureaucrats would be looking for new employment. So we do not see that happening.

Instead, short-term maturities are used to deal with the swings in the fiscal balance. Very few investors disagree with the premise that short-dated yields conform fairly closely to rate expectations -- as otherwise, they would be raking in profits off of mis-priced bonds.

Concluding Remarks

If researchers spend decades looking for something and do not find it, the best course of action is to ask why this is so. In the case of the link between deficits and bond yields, learning even a tiny amount about bond issuance practices offers the explanation.

Technical Appendix: You Want Theory? I'll Give You Theory

The first thing to note is that the classical story about interest rate determination -- it is used to match financing ("loanable funds") -- is wildly incorrect. Within a currency, flows are circular. Financing is always available for someone -- but perhaps not you. Credit spreads tell us which borrowers are having a hard time dipping into the flows. Where circular flows appear to break down in cross-currency; cross-currency basis swaps is the wholesale market that link different currency blocs. (This is why they blow out during crises.)

The markets that matter are instead the following.

• The market for DV01 ("duration"). As can be verified by your friendly risk management professional, every single fixed income instrument has a DV01. Meanwhile, all responsibly managed institutions have DV01 risk limits. The market for DV01 is balanced by taking investors taking directional ("duration") risk, and by all issuers who decide what maturity point to borrow at. Hard-working relative value quants scour the swap curve, looking for any dislocations (which helps eliminate "preferred habitats").
• The market for high quality cash flows. Liability-matching investors need cash flows that will reliably be there, far in the future. A really large notional in a 1-month receiver swaption might have plenty of DV01, but no long-dated cash flows, so this factor is not entirely fungible with DV01. The central government is the best supplier of these cash flows. (Mortgages amortise, and corporate managements are too unstable to be trusted to provide long-term cash flows.)
• There is a demand for inflation-linked cash flows. As discussed in Section 4.6 of Breakeven Inflation Analysis, the central government is almost the sole supplier of these cash flows. (Other issuers often just swap out the inflation risk, with the dealers using central government linkers as the inflation supply hedge.)

The central government may not have a total monopoly of long-dated duration supply, but it is close. A monopolist can either choose to fix the quantity or price. Debt management offices follow a hybrid strategy: supplying a fixed quantity that is constantly adjusted so that prices remain at a "reasonable" level. If yields are "too high" or "too low," the monopolist just mis-judged the market. (Since trading long-dated debt is rarely particularly easy, one can argue that the debt management offices historically did a good job -- but past performance may not be predictive.)

At the short end, the multiplicity of sources of duration, and the ability of investors to take large positions means that pricing gravitates to something near fair value -- which is rate expectations plus a small term premium.

In summary, even if one has the hypothetical capacity to predict the yield of bonds extremely accurately (a very lucrative tool) and one determines that bond yields are "too high," one is not measuring an effect of the fiscal deficit or the stock of debt, one just found that the debt management office messed up (they issued too many long-dated bonds relative to 2- to 5-years). Other than the limiting case where the stock of debt is too small to meet the needs of liability-matching investors, such forecast errors can be manifested at any level of the fiscal deficit.

(c) Brian Romanchuk 2020