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.
What explains the default rate forecast gap?
Today's U.S. high-yield bond market is not as excessively optimistic about the default rate outlook as one would infer from comparing its implied forecast with forecasts produced by quantitatively driven methodologies.
Rather, the current wide disparity can be explained by differences in the universes that the market and the quants are applying to their respective methodologies. The useful byproduct of delving into this issue is a default rate calculation geared to the performance benchmark used by many high-yield managers. We note that no criticism of other default rate calculations and forecasts is intended in this piece; investors benefit from access to a variety of approaches suitable to different applications.
By way of background, a feature of our ongoing analysis gauges the performance outlook for distressed issues by comparing the 12-month U.S. speculative-grade default rate published monthly by Moody's with the high-yield market's implied forecast. This approach was originally described in "How to tell when distressed bonds are attractive" (LCD News, Nov. 28, 2012). In brief, we calculate the market's implied default rate forecast using the distress ratio, defined as the percentage of issues in the ICE BofAML US High Yield Index that are quoted with option-adjusted spreads of 1,000 bps or more. For details on our current methodology see "Estimating the market-implied default rate" (LCD News, Sept. 8, 2015).
Moody's recently published its June 30, 2021 U.S. base case speculative-grade forecast (percentage-of-issuers basis), which came in at 7.43%. Based on the ICE BofA US High Yield Index's June 30, 2020, distress ratio of 17.61%, we derive a 30.16% distressed default rate. (As the distress ratio climbs, a declining percentage of distressed issues default.) Multiplying these two figures yields a 5.31% implied default rate forecast. The gap of 2.12 percentage points between the Moody's and the market-implied forecasts is more than twice the differential of one percentage point that our distressed-debt valuation methodology designates as an extreme.
One plausible explanation of the present gap between the Moody's and market-implied forecasts might be a difference in underlying assumptions. The evidence does not support this explanation, however.
Moody's reports that its forecast assumes a June 30, 2021, U.S. unemployment rate of 7.9%. That is somewhat more optimistic than the view of forecasters surveyed by Bloomberg. The Bloomberg-surveyed experts' median projected unemployment rates are 9.2% in the December 2020 horizon and 7.6% in December 2021. A straight-line interpolation puts their June 2021 forecast at 8.4%, or 1.2 percentage points below the Moody's base case. If Moody's incorporated an economic outlook as pessimistic as the Bloomberg median, its forecast default rate would be higher than it is, and its divergence from the market-implied forecast would be even greater than we currently observe. If we regard the economists' median economic outlook, which is more pessimistic than Moody's, as a proxy for the scenario embedded in high-yield bond prices, we reject the explanation that the market expects a lower default rate than Moody's does because investors are more optimistic about the economy than Moody's is.
There is another, simpler explanation of the seeming gulf between the market's expectation and the output of Moody's forecasting model. The universe on which our version of the market's implied forecast is based is the ICE BofA US High Yield Index. Its composition differs in important ways from the composition of the universe to which the Moody's forecast applies.
To begin with, Moody's U.S. speculative-grade forecast covers issuers that have only rated bonds outstanding, others that have only rated loans outstanding, and some that have both rated bonds and rated loans outstanding. Moody's publishes a bonds-only forecast, but only on a global basis. Consequently, managers whose performance benchmark is the ICE BofA US High Yield Index may achieve a default rate as low as that of the Moody's U.S.-only default rate series, yet experience a higher default rate than their benchmark (see note 1).
We say higher, rather than lower, because Moody's speculative-grade universe is more concentrated in the lowest-rated, i.e., most default-prone, sector.
On Jan. 1, 2019, the most recent date for which figures are available, 44% of speculative-grade issuers within Moody's rated universe resided in the Caa-C category. (U.S.-only figures are not available from the agency.) It is not feasible to generate an exactly comparable figure for the ICE BofA US High Yield Index, as that would require looking up the issuer rating (which may differ from the senior unsecured rating) for all 867 issuers in the index on Jan. 1, 2019, and resolving splits between the rating agencies. A reasonable approximation of the index's proportion of issuers with CCC-C composite ratings, derived by removing duplicate tickers from the index's member listing, is 18%. Measured by issues, the CCC-C component is even lower, at 14%. By the way, the Moody's universe's higher concentration in the bottom rating tier also affects managers who are benchmarked to the Bloomberg Barclays US Corporate High Yield Total Return Index. In turn, the high-yield indexes' ratings mixes differ from that of the S&P Global Ratings' speculative-grade-rated universe.
Adding to the Moody's/market-implied default rate mismatch is the fact that not every issuer that defaults, thereby entering the numerator of the rating agencies' calculations, appears in the index. For example, some issuers have outstanding-only bonds that fall below the ICE BofA US High Yield Index's $250 million minimum face value requirement. In addition, a corporate bond issuer with a speculative-grade rating may be absent from that index because its only outstanding issues are private placements or convertibles.
Benchmark-matched default rates
The table below details default rates calculated on several bases for the 12 months ending June 30, 2020, all derived from the ICE BofA US High Yield Index. Consistent with standard practice in calculating default rates, an issue must have been in the index on July 1, 2019 to be counted as a default in the period. To illustrate, if a company was downgraded from investment-grade to speculative-grade in December 2019, it is not included in the numerator for the trailing-12-months default rate of June 30, 2020, but will be in the numerator for the full-year 2020 default rate. Our calculation of the default ratio's denominator, the number of issuers in the universe on July 1, 2019, is based on ticker symbols assigned by the bonds' index provider, ICE Indices, LLC. To answer another perennial question, our default counts do include distressed exchanges, following the practice of both Moody's and S&P Global Ratings.
The table's headline number is the percentage-of-issuers default rate, 5.43%. This compares with Moody's 7.32% U.S.-only rate, which includes bond-only, loan-only, and bond-and-loan issuers. The difference of 1.89 percentage points is close to the 2.12-percentage-point difference, noted above, between the Moody's 12-month forecast and the market-implied forecast. This suggests that the forecast gap is largely a function of Moody's inclusion of loan-only issuers, which causes Caa-C issuers to be more heavily represented in the agency's rated universe than in the index to which many high-yield managers are indexed. Note, too, that in this period the index-matched default rate differed immaterially from S&P's U.S. default rate of 5.3%.
Exercises such as decomposing long-run high-yield total return into income and default losses require default rates based on dollar amounts rather than number of issuers. On a face amount basis, we find a default rate of 7.82% for the 12 months ending June 30, 2020. The excess of that rate over the index-matched 5.43% percentage-of-issuers figure reflects the larger average amount outstanding on July 1, 2019 among defaulting issuers, $2.05 billion, than for all issuers, $1.42 billion. Average face amounts for defaulting and all issues were $753.5 billion and $670.5 billion, respectively. In other periods, defaults might happen to be dominated by small, rather than large issuers, for no particular reason that could have been predicted. This unsystematic, high-impact factor, the relative debt outstandings of defaulting issuers, explains why default rate forecasting tends to focus on percentage-of-issuers data.
More important than the face-amount-based default rate for deconstructing total returns is the default rate based on market value. For the 12 months ending June 30, 2020, we calculate a 6.18% market-value-based rate. That is lower than the 7.82% face amount rate because the average beginning-of-period price of defaulting bonds was 74.61, versus 98.22 for all bonds. The deeply discounted July 1 2019 average price of the subsequent defaulters shows that the market generally anticipates defaults well in advance, a point we shall expand upon presently.
On the subject of decomposing index returns, we note that the index provider breaks down the ICE BofA US High Yield Index's return by income and price change. The price change component, however, encompasses both losses on defaulting issues (helpfully, from their prices at the beginning of the selected period, not from par) and the market-weighted net price increase or decrease on non-defaulting issues. Our market-value-based default rate enables analysts to isolate the impact of default losses.
The preceding table also details default rates by industry group, based on the Sector Level 3 classifications employed in the ICE BofA US High Yield Index. Further breaking down default rates by Level 4 classifications, such as Advertising, Cable & Satellite TV, Media Content, Media — Diversified, Printing & Publishing, within the Level 3 Media category, could produce misleadingly distorted numbers due to small sample sizes within the subcategories. Readers will probably not be surprised to see that default rates in Energy and Retail were well above the all-industry 5.43% rate. Contributing to the league-leading 10.53% default rate of Consumer Goods were such well-known names as Dean Foods Co. and Revlon Consumer Products Corp.. Note that more industry groups sustained no defaults (10) than sustained at least one default (8).
Our ratings breakdown is based on the Composite Ratings employed in the ICE BofA US High Yield Index. This approach accords with the practice of managers that calculate their portfolios' ratings mixes on the basis of the average rating of all agencies that rate the issue, or the highest or the lowest of those ratings, than classification by the ratings of one agency. Also in the interest of mirroring managers' practices, we depart from the rating agency practice of grouping defaults by issuer rating (alternatively termed "Corporate Family Rating"). We believe managers and their clients more generally rely on nominal ratings.
Within an issuer's capital structure there can be two or more ratings, reflecting different seniority levels. Our methodology, we believe, accords with the way practitioners address that matter. To illustrate our approach, suppose an issuer has one senior issue with a B3 Composite Rating and another with a CCC2 Composite Rating. We count that issuer in the numerators of both the ratio for all issuers with at least one B issue outstanding and the ratio for all issuers with at least one CCC issue outstanding.
In the 12 months ending June 30, 2020, no issue that began the period with a Composite Rating of BB3 or higher defaulted. Like the market, the rating agencies anticipated defaults to the extent that at the issue level, all defaults began the period at B1 or less. (Note that an issue rated in the B category may be a secured bond of a company with an issuer rating — which is more indicative of the probability of default — in the CCC category.) Further dispelling popular canards about ratings accuracy, default rates increased with each step down in major rating tier, from 0.00% on BB to 6.04% on B to 17.32% (not shown in the chart) on the CCC-C tier as a whole. The rate escalated sharply from CCC, at 14.02%, to 70.00% on CC. Statistical noise arising from a very small sample size accounts for the C category's default rate not fully upholding in this particular period the monotonic increase by rating category that consistently validates agency ratings in historical analyses.
The final statistic shown in the table is the percentage of issuers that defaulted in the 12 months ending June 30, 2020 that had at least one issue priced at a distressed level on July 1, 2019. We define distress according to the criterion for inclusion of issues in the ICE BofA US Distressed High Yield Index, i.e., an option-adjusted spread (OAS) of +1,000 bps or greater. At 67.4%, this ratio is far below the comparable 93.1% figure for full-year 2019, as shown in the table below. Historically, we have found that the market anticipates about 95% of defaults by assigned distressed spreads. That is the foundation of our above-described use of the distress ratio in calculating the market-implied default rate forecast.
The exceptionally low ratio in the latest 12 months underscores the suddenness of the switch from a vibrant economic expansion to a pandemic-induced recession. To cite one dramatic example of an abrupt change in actual and perceived default risk, the July 1, 2019 OAS range on issues of subsequent defaulter Hertz Global Holdings Inc. was +212 bps to +469 bps. One Intelsat bond's OAS was just +203 bps. At the time, the ICE BofA BB US High Yield Index's OAS was +235 bps. In light of the nearly overnight change in economic conditions, it is little wonder that the market's June 30, 2019 implied 12-month default rate forecast for the period ending June 30, 2020 undershot the actual outcome by a margin of 2.58% to 5.43%.
Also noteworthy in comparing the two tables above are the respective headline, percentage-of-issuer default rates, 3.34% in 2019 and 5.43% in the 12 months ending June 30, 2020. The 2019 rate is 1.01 percentage point below Moody's U.S. speculative-grade rate for the period, again upholding the thesis that the agency's inclusion of loan-only issuers in its denominator contributes to the disparities between its default rate forecasts and the market's implied forecasts. A separate calculation discloses that the first-half 2020 rate annualizes to 7.90%. As in the 12 months ending June 30, 2020, S&P Global's U.S. default for full-year 2019, at 3.1%, was very close to the index-matched rate (3.34%).
We have presented evidence that the present, wide disparity between Moody's U.S. default rate forecast and the market's implied forecast is attributable in substantial measure to widely divergent ratings mixes between the rating agency's U.S. speculative-grade universe and the ICE BofA US High Yield Index, a widely used performance benchmark. S&P Global's U.S. default rate has closely approximated the index-matched rate in recent periods. Details of the index-matched default rate analysis should be helpful to end investors in evaluating their high-yield managers.
Our findings also show that at the beginning of 2019 the market recognized the vast majority of the year's defaulters as being highly vulnerable to default. This was less so for the 12 months ending June 30, 2020, underscoring the suddenness of the decline in economic conditions as a result of the COVID-19 pandemic. In both periods, Composite Ratings exhibited foresight, with no defaults occurring on issuers rated BB at the outset and with default rates on issuers rated CCC or lower far exceeding default rates on issuers rated B, according to our methodology.
High-yield spread remains extremely tight for prevailing conditions
As shown in the chart below, the option-adjusted spread, or OAS, on the ICE BofA US High Yield Index remains extremely tight, given the factors that ordinarily drive it. The Federal Reserve's extraordinary support for credit markets is keeping the high-yield risk premium at a substantially lower level than we would normally expect to see during a recession. Some of this disparity may be reduced next month if our credit availability measure, which is reported only once a quarter, shows a substantial liberalizing of bank lending standards in response to the Federal Reserve's intervention and some signs of improvement in economic conditions in recent months.
On June 30, the ICE BofA US High Yield Index's OAS stood at +644 bps, down marginally from +654 bps one month earlier. Over the same span, our Fair Value estimate declined to +1,141 bps from +1,264 bps. That reduction reflected a rise in capacity utilization to 68.6% from 64.8% and an improvement in industrial production to 5.4% from 1.4%. The speculative-grade default rate, a backward-looking indicator with little impact on the spread, edged up to 5.3% from 4.6%. The five-year Treasury yield, which is inversely correlated with the spread, inched down to 0.29% from 0.31%. As alluded to above, the highly influential credit availability measure, the percentage of banks tightening lending standards minus the percentage loosening them, remained at 41.6%. We also kept the quantitative easing dummy variable at 1, meaning that it remains in place, regardless whether or not the term is currently being employed by the Federal Reserve. The net impact was that the gap between the actual spread and our Fair Value estimate moved to –496.8 bps, from –610.5 bps in May. That narrowing of the valuation gap was partly reversed by subsequent market action, however, as the ICE BofA US High Yield Index's OAS tightened to +574 bps on July 17, from the abovementioned +644 bps at the end of June.
These conclusions are drawn from the updated methodology presented in "Fair Value update and methodology review" (LCD News, Jan. 24, 2018). In brief, we find that 80% of the historical variance in the ICE BofAML US High Yield Index's OAS is explained by six variables:
- Credit availability, derived from the Federal Reserve's quarterly survey of senior loan officers;
- Capacity utilization;
- Industrial production;
- Current speculative-grade default rate;
- Five-year Treasury yield; and
- A dummy variable for the period covered by quantitative easing.
Each month we calculate a Fair Value spread based on the levels of these six variables. The extent of high-yield overvaluation or undervaluation is determined by the difference between the actual OAS and the Fair Value number ("estimated"). We define an extreme valuation as a divergence of one standard deviation (124.5 bps) or more from Fair Value. The monthly difference between the actual and estimated OAS is tracked in the chart above.
Covenant quality was better, not worse, in Q2 than Q1
Moody's and FridsonVision agreed on the direction of change in covenant quality between May and June 2020. Quality improved and furthermore, the two organizations' series agreed on the magnitude of the improvement. On a scale of 1 (strongest) to 5 (weakest), covenants strengthened from 4.51 to 4.50 by Moody's methodology. The FridsonVision score also improved by 0.01, from 4.46 to 4.45.
It was an entirely different story for the quarterly comparison. This series is more meaningful than the monthly one, which is more affected to statistical noise due to smaller new-issue counts. (There were 92 new issues in Q1'20 and 153 in Q2'20, while the monthly counts during the period ranged from 4 to 58.) Moody's reported a deterioration in covenant quality from 4.48 in the first quarter to 4.52 in the second quarter. FridsonVision, by contrast, reported an improvement, from 4.49 to 4.44.
The minor deterioration indicated by Moody's methodology was, in reality, the result of a change in the new-issue ratings mix from the first to the second quarter. In the earlier period, the Ba category accounted for 30.4% of issues. The Ba share nearly doubled to 56.2% in the latter period. That change created the appearance of a worsening in quality because in every month of the first half of 2020, Ba bonds had a weaker score than B and Caa bonds.
FridsonVision's calculations, which filter out the impact of quarter-to-quarter variations in ratings mix, captured the actual strengthening in covenant quality between the first and second quarters of 2020. That improvement was most apparent in the contrast between the range of monthly Ba scores in the first quarter, 4.75–5.00, and the second quarter, 4.64 to 4.67. (Remember, lower numbers indicate better quality.)
By way of background, "Covenant quality decline reexamined" (LCD News, Oct. 1, 2013) explains how we modify the Moody's CQ Index to remove noise arising from month-to-month changes in the calendar's ratings mix. On average, covenants are stronger on Caa bonds than on Bs, and stronger on Bs than on Ba bonds. Therefore, for example, if issuance shifts downward in ratings mix in a given month, without covenant quality changing within any of the rating categories, the Moody's CQ Index will show a spurious improvement. We eliminate such false signals by holding the ratings mix constant at an average calculated over a historical observation period. (To be fair, Moody's acknowledges the impact of ratings mix in its monthly commentary on its CQ Index.)
Research assistance by Lu Jiang and Zhiyuan Mei.
ICE BofA Index System data is used by permission. Copyright © 2020 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.
1. End investors who want to include default rates in their evaluation of high-yield managers should tabulate not only defaults on issues held by the managers, but also defaults that occurred shortly after the managers sold the bonds at deep discounts to face value.