This commentary is written by Martin Fridson, a high-yield market veteran who is chief investment officer of Lehmann Livian Fridson Advisors LLC as well as a contributing analyst to S&P Global Market Intelligence.
Our Jan. 28, 2015, piece, A new way to debunk the Proportional Spread Thesis, refuted the belief among some practitioners that the high-yield spread-versus-Treasuries (SVT) widens as the underlying Treasury yield increases. We found a -34.91% monthly correlation (R) between the five-year Treasury yield and the option-adjusted spread (OAS) on the benchmark now known as the ICE BofA US High Yield Index, from that index's inception through December 2013. In short, we showed that high Treasury yields were in fact associated with narrow spreads.
This finding contradicts the intuition of many practitioners, who simply feel that when Treasury yields are higher, they need to get a wider spread to compensate for the risk. As we explained, there is no meaningful correlation between the Treasury yield and the default rate, hence no fundamental reason for the risk premium (SVT) to increase when the Treasury rate increases. Indeed, to the extent that high interest rates reflect vigorous demand for capital, a characteristic of a strong economy, default risk ought to be low and spreads should be narrow when Treasury yields are high.
We recently decided it would be worthwhile to delve more deeply into the dynamics of yields and spreads. Experience has taught us that financial relationships can be highly variable over time. To cite a noteworthy example, up until 1958 dividend yields on common stocks exceeded yields on Treasury notes. That is the reverse of what the vast majority of people now active in the financial markets have experienced throughout their investing years. In earlier times, the late financial historian Peter Bernstein explained, investors thought it was natural for common stocks to yield more because they were riskier. They did not think in total return terms, as we do nowadays. The relative rise of yields on bonds from the late 1950s onward may have reflected rising inflation expectations, with common stocks viewed as beneficiaries rather than victims of rising consumer prices. Hedge fund manager Clifford Asness, on the other hand, has explained the change as a function of the two asset classes' comparative volatility.
Design of the study
Lest we fall prey to investment myopia, we decided to examine a longer history of the yield-versus-spread relationship than we studied in our 2015 piece. Specifically, we extended our observation period back to the Sept. 30, 1986, inception of the ICE BofA US High Yield Index and forward to Dec. 31, 2020. Doing so required us to discard OAS, as well as yield calculations more advanced than yield-to-maturity (YTM). Those metrics are available for the index only from December 1996 onward. We fell back on the more primitive method of calculating the spread by subtracting the high-yield index's YTM from the YTM of the ICE BofA US Treasury Index. At the same time, we carried over from the earlier study the exclusion of data for the months October 2008 to March 2009. SVTs for those Global Financial Crisis months are outliers, at +1,663 bps to +1,995 bps.
MORE FRIDSON: What high-yield spreads tell us about the Treasury rate drop
We divided our observation period into three subperiods of roughly equivalent size. The most recent period begins one month after the excluded month of March 2009, which marked the beginning of the Fed’s use of quantitative easing (QE) for the first time since 1951. With only brief interruption, that policy has continued and, with the onset of the COVID-19 pandemic, expanded. We divided the September 1986 to September 2008 interval in half, subject to the odd number of months in that span. Our analysis focused on the subperiods' average yields and spreads, as well as their yield-versus-spread correlations.
Our key findings appear in the table below. The most striking result is a resounding confirmation of our previous finding that higher yields are associated with narrower rather than wider spreads. From one subperiod to the next, the mean Treasury yield declined, while the mean SVT widened.
This finding is partly challengeable on the grounds that the increase in mean SVT from Period 2 to Period 3 is not statistically significant. The table below shows a t-statistic of -0.60 and a p-value of 0.55 for that comparison, far from the values that would indicate significance at confidence levels of 95% or higher. On the other hand, Period 1's mean SVT of +510 bps is different from Period 3's +567 bps with 99% confidence. The difference of means test between Period 1 and Period 2 falls 0.01 short of the canonical 1.96 t-statistic, but we nevertheless maintain that the difference is real in statistical terms. The table does not display the values for the difference between Period 1's +510 bps mean and the mean of +561 bps for the combined Period 2 and Period 3, but the 2.78 t-statistic and 0.01 p-value for that comparison establish significance at the 99% confidence level. In summary, the historical perspective provided by the first table corroborates and strengthens the correlation-based conclusion of our 2015 study.
This is not to give short shrift to correlation analysis in the present analysis. The first table presents the striking finding that in Period 1 the correlation between the Treasury YTM and the high-yield SVT was positive. That is, the evidence appears to uphold the Proportional Spread Theory.
We hasten to add that the correlation shown for Period 1, 24.23%, is deemed weak by a rule of thumb recently articulated by Hormoz Sohrabi of Tarbiat Modares University. By that standard, correlations of less than 30% in absolute value should be viewed with some skepticism. Period 2's correlation of -34.67% passes muster, however, indicating that investors can be confident that from September 1997 to September 2008 high Treasury rates were genuinely associated with narrow high-yield spreads. These results indicate that although a long historical view confirms that Treasury yields and high-yield SVTs are inversely correlated, that relationship does not always prevail within shorter periods. Period 3's correlation of -22.29% and the full observation period's -16.11% correlation support, but not strongly, the conclusion that high-yield spreads tighten as Treasury yields rise.
Investors should be wary of overinterpreting these results, yet it is intriguing to entertain the possibility that QE — exclusively a Period 3 phenomenon — has weakened the inverse relationship between Treasury YTM and high-yield SVT, which was quite strong in the years preceding that Fed policy change. In difficult economic times, when spreads would otherwise be wide as a function of elevated default risk, QE may hold spreads in check by persuading investors that the Fed is minimizing default risk by keeping capital flowing, no matter what. That constraint on spread-widening, coinciding with downward pressure on Treasury yields, would likely move the Treasury yield-versus-SVT toward — if not necessarily into — positive territory.
Average high-yield spreads were narrower in the old days of high-single-digit Treasury yields than in more recent times, when Treasury rates have averaged less than 5.0%. This finding corroborates the inverse relationship between the underlying government bond yield and the high-yield risk premium documented in our previous research. Within shorter intervals in the long history of the high-yield market, however, this relationship has been highly variable.
High-yield spread moved further from Fair Value in June
Already extremely overvalued by historical standards, the high-yield market edged further away from Fair Value in June. The ICE BofA US High Yield Index's option-adjusted spread, or OAS, moved from -190 bps to -203 bps versus our Fair Value estimate, calculated as described below. That compares with our cutoff of 124.5 bps (one standard error) for deeming the market extremely over- or undervalued. The conclusion would be different if one discarded all pre-2020 data on the premise that the Fed's extraordinarily aggressive intervention in response to the COVID-19 pandemic represents an entirely new chapter in financial history.
To elaborate on the latest changes, our Fair Value estimate for the high-yield OAS dropped to +507 bps in June from +519 bps in May. Economic indicators were mixed, with Capacity Utilization rising to 75.4%, from 75.1%, but with Industrial Production dropping to 0.4% from 0.7%. A rise in the five-year Treasury yield to 0.88% from 0.78% helped drive the Fair Value estimate lower; underlying Treasury rates are inversely correlated with the high-yield spread. The default rate was unchanged at 4.7%. There was no change in the Fed survey of senior loan officers, which is reported only quarterly. As of the most recent update, the number of banks easing credit to large and medium-sized businesses exceeded those tightening credit by 15.1 percentage points. Our dummy variable for quantitative easing remained at 1, indicating that QE remains in force.
On June 30, the ICE BofA US High Yield Index's actual OAS stood at +304 bps, down by 25 bps from +329 bps a month earlier. With the Fair Value estimate down by 12 bps from May, the gap between the actual spread and Fair Value increased by 13 bps, to -203 bps from -190 bps, as noted above. By July 15, however, the actual spread widened to +318 bps, leaving the actual-versus-Fair Value differential 1 basis point less than where it ended May, at -189 bps.
Research assistance by Manuj Parekh and Weiyi Yang.
ICE BofA Index System data is used by permission. Copyright © 2021 ICE Data Services. The use of the above in no way implies that ICE Data Services or any of its affiliates endorses the views or interpretation or the use of such information or acts as any endorsement of Lehmann Livian Fridson Advisors LLC's use of such information. The information is provided "as is" and none of ICE Data Services or any of its affiliates warrants the accuracy or completeness of the information.