Priced risk in corporate bonds

Abstract: Recent studies document strong empirical support for multifactor models that aim to explain the cross-sectional variation in corporate bond expected excess returns. We revisit these findings and provide evidence that common factor pricing in corporate bonds is exceedingly difficult to establish. Based on portfolio- and bond-level analyses, we demonstrate that previously proposed bond risk factors, with traded liquidity as the only marginal exception, do not have any incremental explanatory power over the corporate bond market factor. Consequently, this implies that the bond CAPM is not dominated by either traded- or nontraded-factor models in pairwise and multiple model comparison tests.

• The paper finds that the bond market factor alone explains nearly all priced risk in corporate bond returns, with multifactor models showing no significant improvement.
• Robust econometric techniques reveal previous multifactor models—like the BBW four-factor model—overstate predictive power due to data and methodological flaws.
• Implications suggest that holding the broad bond market portfolio is optimal for risk-adjusted performance, challenging the value of complex corporate bond multifactor strategies.

Empirical Assessment of Priced Risk in Corporate Bonds

Overview

This paper provides a rigorous empirical reassessment of priced risk in the cross-section of corporate bond expected returns, focusing especially on the validity and incremental explanatory power of multifactor models vis-à-vis the canonical single-factor corporate bond market model (bond CAPM). Motivated by the influential four-factor model of Bai, Bali, and Wen (2019—BBW) and recent literature proposing traded and nontraded risk factors, the study implements extensive cross-sectional and time-series analyses using state-of-the-art identification- and misspecification-robust econometric techniques, a cleansed TRACE/FISD dataset, and a transparent, replicable construction of risk factors.

Data and Factor Construction

The analysis leverages an enhanced version of the TRACE database (July 2002 to December 2016) merged with FISD for bond characteristics. The sample comprises over 860,000 bond-month observations, with rigorous filters applied to remove non-U.S., structured, non-fixed-rate, near-maturity, and lightly traded bonds, resulting in a comprehensive, high-integrity panel for empirical asset pricing.

The BBW four-factor model consists of:

• Bond market factor ($MKTB$): Value-weighted excess returns on all eligible corporate bonds.
• Downside risk factor ($DRF$): 5% rolling Value-at-Risk portfolios, aggregated long–short across ratings.
• Credit risk factor ($CRF$): Spread-based long–short portfolios across VaR, liquidity, and short-term reversal dimensions.
• Liquidity risk factor ($LRF$): Portfolio-level illiquidity proxies (from Bao, Pan, and Wang, 2011), sorted and aggregated similar to $DRF$.

Replication revealed significant construction flaws (lead-lag errors, attenuation of negative return realizations) in the public BBW factor data, artificially inflating the perceived explanatory power of non-market factors. The authors provide corrected risk factors and advocate for stricter adherence to replicable construction protocols.

Traded-Factor Model Performance

Portfolio- and Model-Level Analysis

Analysis across multiple dimensions—including mean-variance efficiency, bias-adjusted Sharpe ratios, and cross-sectional R² (using generalized least squares per Lewellen, Nagel, and Shanken, 2010)—demonstrates that:

• The bond market factor subsumes virtually all the predictive content of alternative traded factors in the cross-section of corporate bond excess returns.
• Incremental Sharpe ratio gains from adding $DRF$, $CRF$, or $LRF$ to $MKTB$ are statistically indistinguishable from zero (e.g., difference in squared Sharpe ratios between BBW four-factor and $MKTB$-only models of 0.001, $DRF$0).
• The only marginally non-redundant factor is $DRF$1 (traded liquidity), which, at a portfolio or individual bond level, exhibits weak but sometimes significant pricing ability.
• Use of robust identification and misspecification adjustment is critical: OLS-based alpha and $DRF$2 statistics can be misleading, inflating the apparent fit especially in the presence of return heterogeneity and model misspecification. GLS-based metrics, which are robust to these issues, reveal very low statistical $DRF$3 (all models, 0.002 to 0.185).

Pairwise Model Comparison

Robust pairwise and multiple-model comparison tests (Barillas et al., 2020) consistently fail to find statistically significant outperformance of any multifactor or alternative traded-factor model relative to the bond CAPM. The findings robustly contradict earlier BBW results, as well as those of models using intermediary capital or default/term factors.

Nontraded Factor Models

Numerous nontraded risk factors have been proposed as drivers of corporate bond premium variation, including macroeconomic uncertainty (Jurado-Ludvigson-Ng uncertainty, Bali et al., 2021), aggregate liquidity (Amihud 2002; Pástor-Stambaugh 2003), volatility risk (VIX changes), intermediary capital (He, Kelly, Manela, 2017), and long-run consumption risk (Elkamhi, Jo, Nozawa, 2023).

This study constructs factor-mimicking portfolios for each and rigorously projects pricing performance versus the bond CAPM:

• None of the nontraded factor-mimicking portfolios command significant risk premia after controlling for $DRF$4.
• Alphas, bias-adjusted Sharpe ratios, and nested/non-nested model comparison tests indicate no incremental economic or statistical value from adding any nontraded factor (except transiently for the macro uncertainty mimicking portfolio, which is not robust to correct measurement of estimation error in weights).
• Both portfolio-level and bond-level Fama–MacBeth (1973) cross-sectional regressions corroborate the absence of robust pricing power for nontraded factors.

Bond-Level Analysis

Fama–MacBeth two-pass regressions at the individual bond level, using post-ranking betas to address attenuation bias, confirm:

• Only $DRF$5 and, to a lesser extent, $DRF$6 generate significant risk premia proxies; all other traded and nontraded factors are subsumed.
• Analysis is robust to alternative test portfolios (credit spread, size, industry, maturity), different databases (TRACE, WRDS, ICE), and extended time samples.

Methodological Contributions and Statistical Recommendations

• The study identifies that lead-lag errors, ex-post factor selection, and improper use of OLS-based $DRF$7 and alphas have contaminated prior empirical findings in corporate bond asset pricing.
• It emphasizes robust estimation and inference—GLS, misspecification adjustments, and correct bootstrapping for mimicking portfolios—as essential best practice.
• The findings imply that spurious factor identification is likely when factors are selected based purely on empirical fit; theory-driven selection and strong economic rationale are necessary for credible multifactor model construction.

Implications and Future Directions

Theoretical Implications

The results suggest the absence of robust, priced common risk factors in corporate bond returns beyond aggregate bond market risk and, possibly, traded liquidity. This stands in contrast to the equity literature, where the multifactor paradigm retains partial empirical support.

Practical Implications

From a practical asset allocation viewpoint, holding the value-weighted bond market—without additional factor exposure—delivers maximum attainable risk-adjusted performance, net of implementation costs and trading frictions. The findings considerably diminish the empirical motivation for complex corporate bond multifactor portfolios, especially when considering the substantial trading costs in OTC bond markets.

Future Research

The failure to identify reliable priced factors suggests the need to explore alternative model specifications, such as:

• Frequency-domain/frequency-dependent risk (as in Bandi et al., 2021; Neuhierl and Varneskov, 2021), which may uncover priced structure missed by canonical time-domain factors.
• Transaction cost–adjusted model performance metrics, adapting recent advances in equity factor selection under costs (Detzel, Novy-Marx, and Velikov, 2023).
• Theoretical investigations into bond market microstructure effects and liquidity supply, reflecting the OTC market environment.

Conclusion

This comprehensive reassessment finds little robust evidence of incremental priced risk factors in the cross-section of corporate bond expected returns, with the exception of marginal liquidity effects. The bond CAPM, using the aggregate market portfolio, is not statistically or economically dominated by popular traded or nontraded alternative specifications, undermining the practical relevance of multifactor risk models for corporate bonds. The findings call for greater methodological rigor in empirical corporate bond pricing research, robust data handling, and a reevaluation of theory in light of persistent model misspecification and empirical redundancy of popular factors.