Sequence-of-Returns Risk Revisited: Can Time-Series Foundation Models Detect Danger Zones?
DOI:
https://doi.org/10.61190/et356m79Keywords:
sequence-of-returns risk, retirement planning, foundation models, danger-zone detection, GARCH, VIX, Chronos, early-warning systems, withdrawal rates, wealth managementAbstract
Sequence-of-returns risk (SoRR) is the dominant driver of retirement portfolio failure, yet planning practice relies on static probability-of-ruin estimates rather than prospective early-warning signals. This study tests whether a time-series foundation model (TSFM)—Amazon's Chronos—can detect SoRR danger zones, benchmarked against five traditional volatility indicators. Using 395 months of U.S. data (1993–2025) with actual inflation and a real-bond series, we construct danger-zone labels from 15-year, 4% real-withdrawal retirement simulations and evaluate each indicator's detection accuracy and retirement-outcome impact under a defensive-response strategy. As a danger-zone detector, Chronos forecasts are almost perfectly inverted: a naive reading (low forecast as danger) yields an ROC-AUC of 0.022, while reading it contrarily (high forecast as danger) yields 0.978, exceeding every traditional indicator (VIX = 0.654; composite = 0.678). The model's optimism is itself the warning. However, this near-perfect ranking power does not translate into an actionable real-time trigger: an expanding-window contrarian rule fires too rarely to protect portfolios, and both Chronos-based defensive strategies reduce worst-case wealth. Traditional indicators improve point-estimate worst-case (5th-percentile) terminal wealth by 13–51%, but moving-block-bootstrap confidence intervals include zero in every case, so the benefit is not statistically distinguishable from chance. The contribution is a corrected interpretation of TSFM output in financial planning, a rigorously benchmarked comparison, and candid guidance for practitioners on both the promise and the limits of foundation models for retirement risk.
Downloads
References
Ang, A., & Bekaert, G. (2002). International asset allocation with regime shifts. Review of Financial Studies, 15(4), 1137–1187.
Ang, A., & Timmermann, A. (2012). Regime changes and financial markets. Annual Review of Financial Economics, 4(1), 313–337.
Ansari, A. F., Stella, L., Turkmen, C., Zhang, X., Mercado, P., Shen, H., Shchur, O., Rangapuram, S. S., Pineda Arango, S., Kapoor, S., Zschiegner, J., Maddix, D. C., Wang, H., Mahoney, M. W., Torkkola, K., Wilson, A. G., Bohlke-Schneider, M., & Wang, Y. (2024). Chronos: Learning the language of time series. Transactions on Machine Learning Research.
Bekaert, G., & Hoerova, M. (2014). The VIX, the variance premium and stock market volatility. Journal of Econometrics, 183(2), 181–192.
Bengen, W. P. (1994). Determining withdrawal rates using historical data. Journal of Financial Planning, 7(4), 171–180.
Blanchett, D. (2015). Initial conditions and optimal retirement glide paths. Journal of Financial Planning, 28(9), 52–60.
Bluwstein, K., Buckmann, M., Joseph, A., Kapadia, S., & Şimşek, Ö. (2021). Credit growth, the yield curve and financial crisis prediction: Evidence from a machine learning approach (ECB Working Paper Series No. 2614). European Central Bank.
Bollerslev, T. (1986). Generalized autoregressive conditional heteroskedasticity. Journal of Econometrics, 31(3), 307–327.
Bollerslev, T., Chou, R. Y., & Kroner, K. F. (1992). ARCH modeling in finance: A review of the theory and empirical evidence. Journal of Econometrics, 52(1–2), 5–59.
Cooley, P. L., Hubbard, C. M., & Walz, D. T. (1998). Retirement savings: Choosing a withdrawal rate that is sustainable. AAII Journal, 20(2), 16–21.
Das, A., Kong, W., Sen, R., & Zhou, Y. (2024). A decoder-only foundation model for time-series forecasting. Proceedings of the 41st International Conference on Machine Learning (ICML), PMLR 235, 10148–10167.
Engle, R. F. (1982). Autoregressive conditional heteroscedasticity with estimates of the variance of U.K. inflation. Econometrica, 50(4), 987–1008.
Finke, M., Pfau, W. D., & Blanchett, D. (2013). The 4 percent rule is not safe in a low-yield world. Journal of Financial Planning, 26(6), 46–55.
Fouliard, J., Howell, M., & Rey, H. (2021). Answering the Queen: Machine learning and financial crises (BIS Working Papers No. 926). Bank for International Settlements.
Goel, A., Pasricha, P., & Kanniainen, J. (2024). Time-series foundation AI model for Value-at-Risk forecasting. arXiv preprint arXiv:2410.11773.
Goswami, M., Szafer, K., Choudhry, A., Cai, Y., Li, S., & Dubrawski, A. (2024). MOMENT: A family of open time-series foundation models. Proceedings of the 41st International Conference on Machine Learning (ICML), PMLR 235, 16115–16152.
Gu, S., Kelly, B., & Xiu, D. (2020). Empirical asset pricing via machine learning. Review of Financial Studies, 33(5), 2223–2273.
Guidolin, M., & Timmermann, A. (2007). Asset allocation under multivariate regime switching. Journal of Economic Dynamics and Control, 31(11), 3503–3544.
Guyton, J. T., & Klinger, W. J. (2006). Decision rules and maximum initial withdrawal rates. Journal of Financial Planning, 19(3), 48–58.
Hamilton, J. D. (1989). A new approach to the economic analysis of nonstationary time series and the business cycle. Econometrica, 57(2), 357–384.
Kitces, M. E., & Pfau, W. D. (2015). Retirement risk, rising equity glide paths, and valuation-based asset allocation. Journal of Financial Planning, 28(3), 38–48.
Milevsky, M. A. (2006). The calculus of retirement income: Financial models for pension annuities and life insurance. Cambridge University Press.
Milevsky, M. A. (2016). It's time to retire ruin (probabilities). Financial Analysts Journal, 72(2), 8–12.
Milevsky, M. A., & Robinson, C. (2005). A sustainable spending rate without simulation. Financial Analysts Journal, 61(6), 89–100.
Pfau, W. D. (2015). The lifetime sequence of returns: A retirement planning conundrum (SSRN Working Paper No. 2544637). https://ssrn.com/abstract=2544637
Pfau, W. D. (2019). The four approaches to managing retirement income risk (SSRN Working Paper No. 3501673). https://doi.org/10.2139/ssrn.3501673
Pfau, W. D., & Kitces, M. E. (2014). Reducing retirement risk with a rising equity glide path. Journal of Financial Planning, 27(1), 38–45.
Rasul, K., Ashok, A., Williams, A. R., Ghonia, H., Bhagwatkar, R., Khorasani, A., Darvishi Bayazi, M. J., Adamopoulos, G., Riachi, R., Hassen, N., Biloš, M., Garg, S., Schneider, A., Chapados, N., Drouin, A., Zantedeschi, V., Nevmyvaka, Y., & Rish, I. (2023). Lag-Llama: Towards foundation models for probabilistic time series forecasting. arXiv preprint arXiv:2310.08278.
Reimann, C. (2024). Predicting financial crises: An evaluation of machine learning algorithms and model explainability for early warning systems. Review of Evolutionary Political Economy, 5(1), 51–83.
Samitas, A., Kampouris, E., & Kenourgios, D. (2020). Machine learning as an early warning system to predict financial crisis. International Review of Financial Analysis, 71, 101507.
Whaley, R. E. (2000). The investor fear gauge. Journal of Portfolio Management, 26(3), 12–17.
Downloads
Published
Issue
Section
License
Copyright (c) 2026 Harry Vadalkar
This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License.
Author(s) retain the copyright and full publishing rights without restriction.
Author(s) grant the Journal right of first publication with the work simultaneously licensed under a Creative Commons Attribution-NonCommercial 4.0 International License that allows reusers to distribute, remix, adapt, and build upon the material in any medium or format, for noncommercial purposes only. Reusers must acknowledge the work's authorship and initial publication in this Journal.
Noncommercial means not primarily intended for or directed towards commercial advantage or monetary compensation.
In addition, FSR grants to the UGA Libraries a worldwide, non-exclusive license to all content published by the Journal, including metadata, that is necessary to publish, transmit, and index the Journal and to preserve its content over time.
