The Yield Curve Is Distorted. Yields Must Rise (Part 2)
The US Treasury Note Is ‘Junk’. Mortgage Yields Are Better
The yield curve slope has become an unreliable predictor of the business cycle. The problem runs deeper than the yield curve itself, because the benchmark 10-year ‘risk free’ yield carries most of the distortion. This breakdown tells us less about the real economy, but far more about what should be the ‘true’ level of US Treasury yields. This questions whether the 10-year Treasury note remains the appropriate ‘risk free’ benchmark. Bond term premia are being heavily distorted by both US Fed and Treasury actions. The conclusion is that although a fall into a deep US recession is a less likely outcome of this bias, US 10-year yields are too low by around 110bp.
10-Year Treasury Yields Are ‘Junk’: Mortgage Yields Are Better
Part 1 argued that the US Treasury market – the benchmark for global asset pricing and an oft-used guide to the future path of the American economy – has been heavily distorted by US Treasury and Federal Reserve policies of QE, QT and the deliberate recent skewed supplies of Treasury coupons and bills. We conclude that the underlying US economy is less fragile than the bond market implies and that US 10-year yields should trade some 110bp higher.
This reinforces the case for investing in stocks rather than holding bonds. We continue to favour risk assets, and do not foresee a significant economic recession.
Distortions in yields, from the combined impact of Federal Reserve and US Treasury policies, suggest that the 10-year Treasury may no longer serve as the ‘true’ risk-free benchmark for the financial system. One result is that an unusually depressed 10-year Treasury yield feeds into a larger than warranted inversion in the 10-year less 2-year yield curve, which, in turn, gives a false flag on the economy.
One way to identify the scale of this distortion is to estimate the risk-free rate in an alternative way. We use the US mortgage rate curve, which (given the backstop provided by the GSEs, such Fannie Mae, FNMA) is potentially a quasi-government backed asset. We do this in the following way:
Estimating The ‘Risk Free’ 10-Year Yield From The US Mortgage Curve
We show the estimated risk free rates from the mortgage curve in the chart below (black line) compared to the actual 10-year Treasury yield (orange line). The spread appears to widen during periods when the Fed and the Treasury are actively intervening in the markets to subtly ease conditions, such as the 2008/09 GFC, the COVID emergency and the period following the UK gilt crisis and SVB failure. According to these estimates, the US Treasury 10-year yield should strictly trade around 110bp higher.
The chart below shows the 10-year less 2-year Treasury spread and an ‘adjusted’ yield curve using the implied risk-free 10-year yield from the mortgage curve. Notice how the adjusted yield curve deviates from the actual curve between 1998/2000, 2008/09 and 2020/23. A steeper curve implies easier monetary conditions. Hence, the visibly steeper ‘adjusted’ curve tells that there is greater underlying monetary easing that the conventional yield curve says. We cannot dispute the fact that the US yield curve has flattened, but we can say that the scale of flattening may not yet be sufficient to trigger a recession. What’s more the ‘true’ curve appears to be steepening.
Returning full-circle to the relationship between the yield curve and US business activity, the chart below shows the ISM index charted against the adjusted yield curve (advanced by 12 months). This revised picture visually seems to better capture future turning points in the business cycle and more consistent with the D-star estimates. It suggests that the US economy is ‘bumping along the bottom’ and may just avoid recession. Moreover, the increase in the adjusted yield points to a future mild recovery in business activity later this year. This is an investment environment that risk assets should enjoy
.
Ian ...but even given what you say we first try to adjust a mortgage basket for duration and then convexity. The result is both remarkably stable over time with the 'adjusted' closely tracking the 10-year. Periods of deviation correlate closely with changes in the bill/ coupon mix. We show this. I must admit I was originally sceptical but the results made me rethink. Also closer testing shows that prior to the GFC, mortgages adjusted to Treasuries. Since, the evidence leans the other way. Moreover, using this spread adjustment, the yield curve behaves as it should. No recession signaled, only a flowing and that should be ending now. We shall see.
Hmm, I’ll take the opposite tack here.
One can't compare US Agency MBS securities that are laden with prepayment optionality and negative convexity to a single nominal bullet 10-yr Treasury for a spread comparison. For instance, during the Fed’s QE when most Agency MBS traded at a premium dollar price, the longest option adjusted duration using Sali’s / Citi Yieldbook model (the Sali fixed income folks are impressive!) was 3 years for an Agency MBS 30-yr 2% coupon. 10-year Treasuries usually have a duration of 7-8 years, so any nominal spread comparison is mis-specified due to these duration mismatches when Agency MBS is in a premium price environment.
We usually use OAS analysis that is the excess spread across the entire term structure of Treasuries to arrive at relative value analysis. Agency MBS lower coupons have extended in duration as the market has sold off and now sport a duration in the 7-8 year range (Use ‘TBPF’ screen on Bloomberg). The OAS is roughly 46 bps versus Treasuries, but down from 80 bps last October.
Finally, I’d argue that the Agency MBS market is just as much junk as the US Tsy. Thank you Federal Reserve for buying 35% of the MBS float during QE, but even through QT, they still own 30% plus of the market and really taint the price discovery mechanism for the sector.