{"paper_id":"2611.26200","title":"A Three-Factor Hull-White Scenario Generator for Macro-Financial Stress Forecasting","authors":["Ahmad Qadi","Thomas Schröder","Francesca Medda"],"paper_date":"2026-11-19","paper_type":"empirical","paper_url":"https://arxiv.org/pdf/2611.26200","page_url":"https://thequant.space/flowcharts/a-three-factor-hull-white-scenario-generator-for-macro-financial-stress/","scores":{"empirical_rigor":8,"math_complexity":8,"quadrant":"Holy Grail"},"proposes":{"status":"available","statement":"The paper builds a transparent, reproducible scenario engine that converts a yield‑curve path into a conditional macro‑financial distribution.","quote":"This paper builds the missing piece: a transparent, reproducible engine that converts a yield-curve path into a conditional macro-financial distribution, and evaluates it to the standard normally applied to forecasting models.","summary":"This paper builds a scenario engine that maps a yield-curve path into a distribution of macro-financial outcomes. A three-factor Hull-White model generates arbitrage-free curves whose factors are simulated under the physical measure, with an estimated and reported market price of risk. A…","strategy":null},"assumptions":{"status":"available","items":[{"assumption":"A three‑factor Gaussian Hull‑White model is used to generate arbitrage‑free curves.","type":"model","stated":"explicit","quote":"The first is a three-factor Gaussian Hull–White model (Hull and White, 1990).","support":"full"},{"assumption":"The Hull‑White factors are simulated under the physical (P) measure, estimated from fifteen years of weekly curve history.","type":"model","stated":"explicit","quote":"Its curves are priced under the risk-neutral measure, so they are arbitrage-free and reproduce the observed discount curve exactly, but the factors that drive them are simulated under the physical measure, estimated from fifteen years of weekly curve history.","support":"full"},{"assumption":"An explicitly estimated market price of risk links the risk‑neutral and physical measures.","type":"model","stated":"explicit","quote":"An explicitly estimated market price of risk links the two. This paper keeps both: curves are priced risk-neutrally, so they are arbitrage-free, and the factors are simulated under the physical measure, with an explicit market price of risk joining the two.","support":"full"},{"assumption":"Macro‑financial variables are modeled in first differences to ensure stationarity.","type":"statistical","stated":"explicit","quote":"The VARX endogenous block contains five weekly vari-ables, all in first differences for stationarity: the S&P 500 weekly log return (mkt_ret); the 6 weekly change in the effective federal funds rate, in decimals (dffr); the weekly change in the Weekly Economic Index of Lewis et al. (2022) (dwei); the weekly change in year-on-year CPI inflation, forward-filled from monthly releases (dinf);","support":"full"},{"assumption":"The splice‑week factor change from the LIBOR–OIS basis adjustment is set to zero in the differenced series.","type":"data","stated":"explicit","quote":"The jump is an artefact of the change of curve, not a market move, and it is kept out of every estimate that uses weekly changes: the splice-week factor change is set to zero in the differenced series, which is equivalent to level-adjusting the SOFR curve to the LIBOR curve at the splice date for that purpose, and the transition across the splice is excluded from the factor-dynamics and volatility estimates.","support":"full"},{"assumption":"The VARX structural identification uses sign restrictions and a relevance filter on the policy‑rate loading.","type":"model","stated":"explicit","quote":"3.5 Structural Identification A contractionary monetary-policy shock is identified by sign restrictions (Uhlig, 2005; Rubio-Ramírez et al., 2010), which do not pin down a single shock but a set of admissible ones, augmented by a relevance filter on the policy-rate loading and by magnitude caps.","support":"full"}]},"datasets":{"status":"available","items":[{"name":"LSEG U.S. dollar zero‑coupon yield‑curve panel","provider":"London Stock Exchange Group","kind":"market-prices","access":"commercial","period":"June 2010–December 2024","universe":"U.S. dollar Treasury zero‑coupon curves","quote":"3 Methodology 3.1 Data The analysis combines macro-financial series from January 2008 with a London Stock Exchange Group (LSEG) zero-coupon yield-curve panel from June 2010. The earlier macro start date supports lagged and rolling variable construction; the VARX sample is governed by zero-curve availability and runs from July 2010 to December 2024 after Friday end-of-week alignment and differencing. Yield-curve data.","support":"full","known_id":"refinitiv"},{"name":"LIBOR swap zero‑coupon curve (up to June 2020) and SOFR OIS curve (after June 2020)","provider":"London Stock Exchange Group","kind":"market-prices","access":"commercial","period":"up to June 2020 / after June 2020","universe":"Swap and OIS zero‑coupon curves","quote":"The term-structure block uses LSEG U.S. dollar zero-coupon curves, a LIBOR swap zero-coupon curve up to June 2020 and a SOFR OIS curve thereafter. The switch takes effect on 19 June 2020, three years before LIBOR cessation.","support":"full"},{"name":"Weekly macro‑financial series (S&P 500 returns, federal funds rate change, Weekly Economic","provider":"FRED and Federal Reserve Bank of Dallas","kind":"macro","access":"public","period":null,"universe":"U.S. aggregate series","quote":"The VARX endogenous block contains five weekly vari-ables, all in first differences for stationarity: the S&P 500 weekly log return (mkt_ret); the 6 weekly change in the effective federal funds rate, in decimals (dffr); the weekly change in the Weekly Economic Index of Lewis et al. (2022) (dwei); the weekly change in year-on-year CPI inflation, forward-filled from monthly releases (dinf); and the weekly change in","support":"full","known_id":"fred"}],"known":[{"id":"refinitiv","name":"Refinitiv / LSEG (Datastream, Eikon, Tick History)","provider":"LSEG","kind":"market-prices","access":"commercial","quote":"3 Methodology 3.1 Data The analysis combines macro-financial series from January 2008 with a London Stock Exchange Group (LSEG) zero-coupon yield-curve panel from June 2010.","mentions":5,"used_in":2},{"id":"fred","name":"FRED","provider":"Federal Reserve Bank of St. Louis","kind":"macro","access":"public","quote":"The macro-financial series are public and available from the Federal Reserve Economic Data (FRED) database and the Federal Reserve Bank of Dallas.","mentions":2,"used_in":1}],"sample":{"start":null,"end":null,"market":null,"asset_class":null,"frequency":null},"data_resolution":"weekly-monthly"},"transaction_costs":{"status":"not_applicable","modeled":"n/a","headline_net_of_costs":null,"detail":null,"quote":null,"source":"front_matter","rule":"R3","candidates":{"front_matter":"n/a","evidence_card":"n/a","evidence_card_quoted":false},"meaning":"the paper reports no trading or execution result"},"results":{"status":"available","evaluation_type":"forecast-accuracy","findings":[{"finding":"The term‑structure block (G3++) has the lowest out‑of‑sample RMSE in 13 of 20 maturity‑horizon cells, with a tie in a fourteenth cell.","kind":"predictive-accuracy","quote":"G3++ has the lowest RMSE outright in 13 of 20 maturity–horizon cells and ties the random walk in a fourteenth (10Y at h = 26), with 18 the strongest performance at the 1Y, 2Y and 5Y pillars. Diebold–Mariano tests, with HAC bandwidths capped by the number of origins, do not reject equal accuracy against the random walk at the 5% level in any cell.","support":"full"},{"finding":"At weekly frequency, once each variable’s own lags are controlled for, the yield‑curve adds no measurable incremental information about macro‑financial variables, in either the conditional mean or the full predictive distribution.","kind":"predictive-accuracy","quote":"At weekly frequency the curve carries no incremental information about these variables once their own lags are controlled for, so the engine's value lies in conditioning on a specified curve path, which no benchmark can do. Three separate tests, the Mode A comparison, the joint-pipeline decomposition and the CRPS comparison, agree that at weekly frequency, once each variable’s own lags are controlled for, the term","support":"full"},{"finding":"A contractionary monetary‑policy shock is associated with falling equity returns, activity, inflation and the MES proxy on impact.","kind":"statistical","quote":"A contractionary policy shock is associated with falling equity returns, activity, inflation and the MES proxy on impact, and the relevance filter and magnitude caps are what make those responses usable at weekly frequency.","support":"full"},{"finding":"MES and a CoVaR‑inspired covariance proxy move in opposite directions under the same shock, indicating they are not substitutes as systemic‑risk indicators.","kind":"other","quote":"The most informative single result is the comparison between MES and the CoVaR-inspired covariance proxy: the two move in opposite directions under the same shock, MES pushed down through the discount-rate channel while the covariance proxy rises, consistent with the risk-taking channel of Borio and Zhu (2012).","support":"full"},{"finding":"The G3++ forecasts are competitive but not dominant; RMSE improvements are not statistically significant versus a random walk according to Diebold‑Mariano tests.","kind":"predictive-accuracy","quote":"The block is therefore a competitive curve generator rather than a dominant conditional-mean forecaster. That is the outcome its constant price of risk predicts: Duffee (2002) shows that completely affine models should not be expected to beat a random walk in level forecasts, and the pattern here, lower RMSE in most cells without significant rejections, is what that implies.","support":"full"}],"reported_performance":null,"out_of_sample":{"method":"walk-forward","headline_out_of_sample":null}},"independent_evidence":{"status":"none_found","supports":[],"contradicts":[],"qualifies":[],"compared_with":20,"citations":{"cited_by_in_corpus":0,"by_independent_authors":0,"how_cited":{"used_as_baseline":0,"uses_method_or_data":0,"extends":0,"discusses_findings":0,"mentions":0,"unclassified":0},"independent_papers":[],"scope":"Citations between papers in this corpus only; citations from outside it are not counted."}},"implementation":{"status":"available","public":false,"code_url":null,"note":"No repository presented by the authors as their own was found in the paper. Absence here is not proof that no code exists."},"replication":{"status":"none_found","card":null,"note":"No replication of this paper has been run by thequant.space yet."},"extraction":{"text_source":"fulltext","model":"openai/gpt-oss-120b","version":3,"extracted_at":"2026-11-20T11:38:32Z"},"provenance":{"claims":"Everything here is what the authors state in the paper. Nothing was re-run or reproduced by the platform.","quotes":"Assumptions, datasets, cost detail and findings were extracted by a language model. Every quote is verbatim paper text. Where an item has `support`, \"full\" means every figure in it appears in the quoted sentences and \"partial\" means a figure is stated elsewhere in the paper; items with a figure found nowhere in the paper were discarded.","datasets_known":"datasets.known lists well-known vendors and public datasets by a stable id, found by a dictionary over the full text where the sentence says the data is used; it is independent of the model and covers only names in the dictionary.","flags":"transaction_costs.modeled, results.evaluation_type and results.out_of_sample.method are platform classifications (structured-decision models); see https://thequant.space/score-guide/#penalty-flags","comparisons":"supports / contradicts / qualifies are a language model's reading of two papers' quoted findings, restricted to closely related papers that share no author. They are pointers to read, not verified replications, and coverage is partial.","implementation":"code_url is a repository the paper itself presents as its own (rule-scored, ambiguous cases judged by a model), checked for liveness when it was recorded."},"docs":"https://thequant.space/agents/"}