The Journal of Fixed Income

Can the Returns of Real Treasuries (TIPS) Be Predicted?

Riccardo Rebonato — 2024-03-01

We analyze the predictability of excess returns in US Treasury Inflation Linked Securities (TIPS). We find the unconditional carry strategy to be profitable, with a Sharpe ratio more attractive than what found for nominal bonds. We study return-predicting factors, and we find that the most universally accepted factor for nominal returns, the slope, has no predictive power at all. We argue that this raises questions about the validity of common explanations of why the slope predicts nominal returns well. However, we also find that both realized inflation and a variation of the Cochrane-Piazzesi factors predict well, and give rise to strategies with high Sharpe ratios. We offer an explanation of why these factors are effective.

How do Alternatives to LIBOR Measure Up?

Faten Sabry — 2024-03-01

The USD LIBOR panel has ceased as of June 30, 2023 and market participants have been transitioning to the Secured Overnight Financing Rate (SOFR) as the alternative benchmark. In this article, we examine the relation between SOFR and LIBOR, as well as analyze various additional benchmark rates that were considered by regulators, academics, and industry experts. We conduct statistical analysis to evaluate how well the adjusted benchmark rates have tracked 1-month LIBOR using historical data. First, we use the mean absolute error to quantify the distance between 1-month LIBOR and each benchmark rate, after adjusting for term and spread. Next, we employ a time-series analysis to assess the degree to which each benchmark co-moved with 1-month LIBOR. We find that while benchmark rates, including SOFR, have generally tracked 1-month LIBOR rates well in the long run, the relation weakens in times of market dislocation, such as during the 2007-2009 Global Financial Crisis and the 2020 COVID-19 pandemic.

Synthetic Credit Ratings and the Inefficiency of Agency Ratings

Doron Nissim — 2024-03-01

This study develops and evaluates a model that generates synthetic credit ratings using accounting and market-based information. The model performs well in explaining agency ratings, suggesting that fitted values for unrated companies are likely to be reasonably precise. Moreover, the synthetic ratings explain cross sectional differences in CDS spreads, even after controlling for contemporaneous agency ratings. Compared to synthetic ratings, agency ratings explain a greater proportion of the variation in CDS spreads, but their differential informativeness is relatively small and has declined substantially over the last decade. This decline is possibly due to post-crisis SEC regulation that limits rating agencies’ ability to obtain confidential information from rated companies. Consistent with the finding that agency ratings do not fully impound the information in synthetic ratings, the difference between synthetic and agency ratings predicts changes in agency ratings in subsequent months, especially for small companies. There is no evidence of substantial improvement over the last four decades in the timeliness of agency ratings with respect to the information in synthetic ratings. Investors in large companies appear to process the synthetic rating information in a timely fashion, as the difference between synthetic and agency ratings does not predict changes in CDS spreads or in the stock prices of these companies. For small companies, however, there is some predictability.

The Stock-Bond Multiscale Correlation

Sofiane Aboura — 2024-03-01

Using a wavelet local multiple correlation methodology that blends multiscale and multivariate approaches, this article investigates the relationships between U.S. stocks and bonds. A spillover analysis completes this methodology. The S&P 500 stock index and U.S. Treasury notes with maturities ranging from 1 to 30 years make up the data. From 1985 to 2022, it is sampled on a daily basis, and from 1990 to 2022, it is sampled on a monthly basis to take inflation and volatility into consideration.

The results suggest that the correlation sign for long-term scales is usually negative between bonds and inflation or stocks and volatility, and positive between stocks and inflation or bonds and volatility. Moreover, volatility often has more influence than inflation. The two scales where a recent correlation shift, from negative to positive, occurred have been the monthly and yearly scales. The major conclusions demonstrate that the stock-bond correlation is prone to time-scale effects in addition to being time-varying at all maturities.

Simulating long-horizon returns on government bonds

pavol Povala — 2024-03-01

We outline a macro-finance model for simulating long-horizon returns on government bonds. We show how the determinants of return distributions change across different investment horizons, and discuss differences in prospective returns on long- and short-duration bonds. Using a realistic calibration, we show that it is unlikely that long-duration bond returns will match the experience in the last few decades. Assuming that macro trends are flat on average, the return distributions of short- and long-duration bonds are comparable, despite long-duration bonds earning a term premium. For long-duration bonds to generate returns closer to historical experience, long-term growth prospects would likely need to deteriorate from today's levels.

Geographic Disaggregation of House Price Stress Paths: Implications for Single-Family Credit Risk Measurement

Alex Bogin — 2024-03-01

We explore the impact of geographic disaggregation of house price stress paths on single-family credit risk measurement. Specifically, we focus on the value added of moving from national, to state-level, to core-based statistical area (CBSA)-level house price paths on estimates of mortgage credit related stress losses. To ensure the robustness of our results, we estimate losses across two different loan portfolios and three credit models. We find that CBSA-level paths provide additional insight on localized credit risk and can be reliably constructed using quarterly house price indices. Further, the variation in results across credit models suggests an implicit confidence interval around any one stress loss estimate. Accounting for this uncertainty through a model risk add-on could potentially offer a more conservative view of portfolio credit risk.