A new Fed paper explains how to construct a real-time distress index, using the case of the corporate bond market. The index is based on metrics that describe the functioning of primary and secondary markets and, unlike other distress measures, does not rely on prices and volatility alone. Thus, it includes issuance volumes and issuer characteristics on the primary side and trading volumes and liquidity on the secondary market side. Making use of a broad range of data on market functioning reduces the risk of mistaking a decline in asset values for actual market distress. Distress in a market that is critical for funding the economy and the financial system has predictive power for future economic dynamics and can be a valuable trading signal in its own right. It can be used for more advanced trend following and for detecting price distortions.

The below are quotes from:
Boyarchenko, Nina, Richard Crump, Anna Kovner, and Or Shachar (2021), “Measuring Corporate Bond Market Dislocations”, Federal Reserve Bank of New York Staff Report No. 957 January 2021.

The post ties up with this site’s summary on price distortions.

The idea

“We use granular data on both primary market issuance and secondary market trading to construct a broad set of measures of corporate bond market conditions…We propose a new measure of market functioning, the U.S. corporate bond market distress index, to quantify joint dislocations in the primary and secondary corporate bond markets.”

“[We] apply index methods to create a comprehensive measure of market functioning. We adapt the methodology of the Composite Indicator of Systemic Stress (Hollo et al., 2012), but apply it to the systematic distress of a market rather than of the economy. The basic intuition is to use insights from portfolio theory to optimally combine measures of different facets of dislocation into a single index.”

“Several principles guide the index.

  • First, while information on prices and price volatility is included, changing prices in either the primary or the secondary market are not by themselves a sufficient statistic to measure market disruptions: price changes are consistent with functioning markets when risk and risk tolerance change.
  • Second, market liquidity – both in the primary market, capturing the ability of issuers to issue new debt, and in the secondary market, capturing the ability of market participants on both sides of the market to transact – plays a key role in the index.
  • Third, the standardized metrics take into account the real-time historical properties of market conditions, so that the index can be backtested.”

“The broad range of indicators that underlie the corporate bond market distress index, spanning both primary and secondary market activity, in both price and quantity terms, reduce the risk that the index increases without a corresponding episode of market stress.”

The data

“A key feature of the corporate bond market distress index is that it combines both primary market and secondary market measures to offer a full picture of corporate bond markets. We calculate the index weekly beginning in 2005, using data that would be available in real time. Primary markets measures come from Mergent and include issuance data on volume and pricing as well as issuer characteristics. For secondary markets, we exploit the rich secondary market trading data available for corporate bonds through the Trade Reporting and Compliance Engine (TRACE) and include measures that reflect both the central tendencies and other aspects of distributions of volume, liquidity, non-traded bonds, spreads and default-adjusted spreads.”

“We use corporate bond transactions data from a regulatory version of TRACE, which contain price, uncapped trade size, and buyer and seller identities as well as other trade terms…Transactions are required to be reported in real-time, with 15 minutes delay, with occasional cancelled or corrected trades.”

“Our final sample thus has 34,074,792 unique bond-trade observations in the secondary market, corresponding to 31,018 unique CUSIPs, or 2,711 unique issuers. In the primary market, we have 58,381 unique issues, corresponding to 1,945 unique issuers.”

The calculation

We standardize each metric using its empirical cumulative distribution function, allowing us to combine variables with different units without assuming a particular parametric transformation. We group metrics into sub-categories by the type of information they capture and, for each sub-category, construct the category-specific sub-index as the equal-weighted average of the standardized constituent series. In this way, we do not overweight sub-categories for which we have more measures, as would be the case in an index computed as an equal-weighted average or the first principal component of all measures. Finally, we combine the sub-indices using time-varying correlation weights, corresponding to an optimal portfolio allocation interpretation of the index.”

“We construct five sets of weekly metrics of secondary market functioning, capturing secondary volume, liquidity, duration-matched spreads, default-adjusted spreads and conditions for non-traded bonds.

  • Volume: We use four metrics of trading volume in the secondary market: dealer-to-customer volume as a fraction of gross trading volume (“intermediated volume”), average dealer-to-customer trade size, ratio of customer buy volume to customer sell volume (which we dub “customer buy-sell pressure ratio”), and turnover. Intermediated volume captures how easily customer volume can be absorbed by dealers in the market, with a lower intermediated volume indicating that the same dealer-to-customer volume generates a greater dealer-to-dealer volume. Turnover measures the fraction of amount outstanding that trades every day…Turnover tends to be high and intermediated volume, average trade size and customer buy-sell pressure ratio all tend to be low during periods of market stress.
  • Liquidity: We construct four standard metrics of market liquidity for corporate bonds: effective bid-ask spread, Thompson and Waller (1987) spread, price impact, and imputed round-trip cost…The four spreads co-move tightly together, rising during periods of market distress.
  • Duration-matched spreads: To capture information about the pricing of the corporate bond market relative to Treasuries, we compute duration-matched spreads…at the bond-level, and construct time series of average spreads, spread volatility (time series standard deviation), and interquartile range of spreads (cross-sectional standard deviation). To keep the index interpretable as a real-time index of market conditions, we compute the average spread and spread volatility from an ARCH-in-mean model…Though all three metrics increase during periods of broad market distress…spread volatility tends to normalize much more quickly and does not increase as much during less significant periods of disruptions.
  • Default-adjusted spreads: Duration-matched spreads capture the pricing of corporate bonds relative to similar duration Treasuries but reflect both expected default rates and default risk premia. To isolate the latter, we construct default-adjusted spreads at the bond-level, and construct time series of average spreads, spreads volatility (time series standard deviation), and interquartile range of spreads (cross-sectional standard deviation). To keep the index interpretable as a real-time index of market conditions, we estimate the predictive regression for the default-adjusted spread on an expanding window basis…As with the duration-matched spreads, all three metrics increase during periods of broad market distress, with spreads volatility normalizing much quicker than the other two measures.
  • Non-traded bonds: While TRACE provides a wealth of information on market conditions for bonds that are actually traded on the secondary market, TRACE does not capture information about market conditions for bonds that are not regularly traded. Instead, we use price quotes from ICE – BAML for bonds included in ICE- BAML U.S. corporate bond indices to construct average quoted duration-matched and default-adjusted spreads. The difference between these quoted average spread series and their traded counterparts – the quoted-traded spread – thus captures the relative conditions for non-traded bonds.”

“We construct two sets of weekly metrics of primary market functioning:

  • Issuance: We construct four metrics of primary market issuance: year-over-year growth rate of dollar amount issued, year-over-year growth rate in the number of bonds issued, and issuance relative to maturing within the next year, in both dollar and number of bonds terms. Considering issuance on a year-over-year growth rate basis allows us to account for both the overall positive time trend in bond issuance as well as seasonality in the timing of corporate bond issuance, while issuance relative to maturing within the next year captures the ability of companies to satisfy their re-financing needs…Intuitively, while the growth rate of dollar amount issued captures the volume of debt issued in the market, the growth rate in the number of bonds issued proxies for the number of issuers able to access the corporate bond market.
  • Primary-secondary spread: Finally, we construct metrics of the spread between prices of bonds traded in the secondary market and the prices of bonds issued in the primary market. As with the secondary market pricing, we construct two measures: duration-matched and default-adjusted spreads…While the primary-secondary spread is positive and relatively small during ‘normal’ periods, the spread becomes negative and large during periods of distress. That is, while during normal times primary market pricing reflects a positive spread to prevailing secondary market prices and issuers are freely able to access the market, market access during downturns is restricted to better-performing issuers, and the average price in the primary market is above the average price in the secondary market. The primary-secondary duration-matched spread is more volatile than the primary-secondary default-adjusted spread.”

“We begin by standardizing each individual metric using the empirical cumulative distribution function of the metric. The appeal of this transformation is that it allows us to combine variables with different ‘natural’ units by imposing a common support without assuming a particular parametric transformation, as would, for example, be the case with a z-score transformation…We use the expanding sample transformation in our construction of the index as it corresponds more closely to the objective of monitoring market conditions in real time and allowing a true test of the approach with historical data.”

“For each category, we construct the category-specific sub-index as the equal-weighted average of the standardized constituent series. [The figure below] plots the time series of all 7 sub-indices. Although each individual sub-index is quite noisy…the combined index is not.”

“The final step in the construction of the corporate bond market distress index is to combine the sub-indices using time-varying correlation weights, corresponding to an “optimal” portfolio allocation interpretation of the index.”

The findings

“Although at times the Corporate bond Market Distress Index gives similar signals to frequently used measures of financial market stress, such as financial conditions indexes, and measures of market risk aversion, such as the VIX, the information…is distinct and captures conditions specific to the corporate bond markets.”

“Increases in the secondary-market-related sub-indices tend to somewhat lead increases in the primary market- related sub-indices, consistent with the conventional wisdom that trading-activity based measures react more quickly to changing economic conditions…We observe that during periods of broad market distress, conditions across both the primary and secondary markets deteriorate, amplifying the individual contribution of each market to the corporate bond market distress index. During normal times, however, the secondary-primary market amplification spiral does not arise.”

“We construct the index separately for investment-grade and high-yield bonds…We find the investment-grade index to be a better predictor of future activity.”

The benefits for macro trading

“The corporate bond market distress index broadens market distress measurement away from just identifying periods of high credit spreads or periods of increased illiquidity in secondary markets alone. Together with the real-time nature of the index, this makes the corporate bond market distress index a valuable summary metric of market distress and functioning.”

Corporate bond market functioning, over and above the information contained in credit spreads alone, has predictive information about future real outcomes…The corporate bond market in the U.S. is a major source of funding for U.S. businesses, representing more than half of the total debt outstanding of non-financial corporations. Distress in the corporate bond market is thus likely to have meaningful consequences for economic outcomes more broadly.”

“We document that the corporate bond market distress index is an economically and statistically significant predictor of cumulative one-year-ahead economic activity as measured by a variety of indicators, even after controlling for standard predictors, such as the term spread. We further show that, in predictive regressions where we include both the corporate bond market distress index and measures of credit spreads, the corporate bond market distress index remains economically and statistically significant, which is generally not the case for credit spreads. This may reflect the fact that the measure’s incorporation of primary market measures adds a dimension of access to credit…There is information in the level and type of activity in both the primary and secondary corporate credit markets, beyond the pricing of credit.”