Please Note: The associated research paper that underpins this data is currently under a Revise & Resubmit (R&R). The data might be subject to change over the next few months. Any changes will be explicitly documented in the Changelog on GitHub. If you have suggestions for the data / spot any issues, please reach out to alexander.dickerson1@unsw.edu.au.
🚀 TRACE Data Processing Pipeline: trace-data-pipeline
🔗Stage 0: Transaction-Level Cleaning (public beta)
⚡Process intraday TRACE transaction data to a daily panel, whilst eliminating data errors at the transaction-level.
📄 Download detailed (>400-page) Enhanced TRACE data reports on the Data tab.
🔗Stage 1: Daily Bond Panel (public beta)
⚡Augment the cleaned daily data from Stage 0 into a comprehensive daily pricing panel including clean prices, accrued interest, yields, credit spreads, duration and convexity.
📊 Download the daily bond panel from the Data tab.
💻 Code: trace-data-pipeline GitHub repository.
🔗 Contact alexander.dickerson1@unsw.edu.au for comments/suggestions.
341 Factors with Alternative Data from “Factor Investing with Delays”: Time-series of value-weighted factors formed with excess bond returns using the OSBAP database and error corrected TRACE data. Factors are very closely aligned regardless of data source. Code available on GitHub includes a script to evaluate the similarity of the 341 factors across the 3 databases. Across all data sources, most factor means are statistically indistinguishable from one another.
The “Co-Pricing Factor Zoo” factors: Includes the traded bond, stock and nontraded factors from the paper. Download the zoo with bond factors formed with excess returns here. The identical file with bond (duration-adjusted) factors can be accessed here. See Table A.1 of the Internet Appendix of the paper for detailed construction notes.
Bond Level Machine Learning Panel Data with 341 Predictors: Includes bond-level panel data with 341 bond and stock predictors spanning 2002-07 to 2022-12. Used in the new working paper, “Factor Investing with Delays”, by Dickerson, Nozawa and Robotti (2024). Time-series of the 341 out-of-sample factors can be downloaded here. See the dedicated “Machine Learning Data” tab for more information.
Duration-Matched Treasury Returns for the WRDS (TRACE) panel from the forthcoming RFS paper, “Duration-Based Valuation of Corporate Bonds” by van Binsbergen, Nozawa and Schwert (2024). Contributed by Yoshio Nozawa. Over the past 40-years, the vast majority of the total returns to investing in corporate bonds can be attributed to gains on long-term Treasury securities (duration) rather than compensation for credit and liquidity risk.
Priced risk in corporate bonds
This website provides the correctly constructed Bai, Bali & Wen (2019, BBW) four factors, and reproducible code and data used in “Priced risk in corporate bonds“, by Dickerson, Mueller & Robotti (2023). If you use the factors, code or data, please cite our paper:
@article{Dickerson-Mueller-Robotti-2023,
title={Priced Risk in Corporate Bonds},
author = {Alexander Dickerson and Philippe Mueller and Cesare Robotti},
journal={Journal of Financial Economics},
volume={150},
pages={article 103707},
year={2023},
publisher={Elsevier}}
A key finding in our paper, is that the bias adjusted sample squared Sharpe ratio of the bond market portfolio (MKTB) is, in fact, greater than holding all four of the BBW bond factors. Economically, and from an investment perspective, it is optimal to simply hold the market portfolio.