{"paper_id":"2502.20978","title":"Using quantile time series and historical simulation to forecast financial risk multiple steps ahead","authors":["Richard Gerlach","Antonio Naimoli","Giuseppe Storti"],"paper_date":"2025-02-28","paper_type":"empirical","paper_url":"https://arxiv.org/pdf/2502.20978v2","page_url":"https://thequant.space/flowcharts/using-quantile-time-series-and-historical-simulation-to-fore/","scores":{"empirical_rigor":7,"math_complexity":7.5,"quadrant":"Holy Grail"},"proposes":{"status":"available","statement":"The paper proposes a quantile‑based, semi‑parametric historical simulation method (QFHS) for multi‑step‑ahead VaR and ES forecasting.","quote":"A method for quantile-based, semi-parametric historical simulation estimation of multiple step ahead Value-at-Risk (VaR) and Expected Shortfall (ES) models is developed.","summary":"A method for quantile-based, semi-parametric historical simulation estimation of multiple step ahead Value-at-Risk (VaR) and Expected Shortfall (ES) models is developed. It uses the quantile loss function, analogous to how the quasi-likelihood is employed by standard historical simulation methods.…","strategy":null},"assumptions":{"status":"available","items":[{"assumption":"The relationship between VaR and ES in the conditional return distribution does not change over time.","type":"model","stated":"explicit","quote":"the proposed method implicitly assumes, and is applicable to, any data or model where the relationship between VaR and ES in the conditional return distribution does not change over time"},{"assumption":"The conditional return distribution is constant (or its VaR‑ES relationship is time‑invariant).","type":"model","stated":"explicit","quote":"which includes the case where the conditional return distribution is constant"},{"assumption":"GARCH(1,1) errors are i.i.d. with a strictly increasing continuous CDF, zero mean and unit variance.","type":"statistical","stated":"explicit","quote":"where the CDF of ϵt (i.e. of D(µ, σ)), is assumed to be strictly increasing and continuous on the real line, with zero mean and unit variance."},{"assumption":"Quantile loss function replaces Gaussian QL in the filtered historical simulation.","type":"model","stated":"explicit","quote":"where the Gaussian QL function is replaced by the quantile loss and resampling is based on residuals scaled by the series of conditional quantiles rather than variances"},{"assumption":"CAViaR model estimation can use a quantile level (αest) different from the target risk level (α0).","type":"model","stated":"explicit","quote":"the QFHS brings additional flexibility to the standard CAViaR modelling framework by allowing αest, the quantile level used to estimate the CAViaR model, to be decoupled from the target risk level α0"}]},"datasets":{"status":"available","items":[{"name":"Daily open, high, low and adjusted close of SP500, FTSE100, ASX200, MSFT, GEN, A","provider":null,"kind":"market-prices","access":"public","period":"01/2000 to 09/2023","universe":"Three major stock indices and three US listed stocks","quote":"We consider daily data on the open, high, low and adjusted close of three stock market indices, SP500, FTSE100, ASX200 and three stocks listed on the New York Stock Exchange, Microsoft (MSFT), Gen Digital (GEN) and Agilent Technologies (A). Data is observed from 01/2000 to 09/2023."},{"name":"Synthetic GARCH(1,1) simulations with varying ν and ξ","provider":null,"kind":"synthetic","access":"unknown","period":null,"universe":null,"quote":"Data is simulated from the specified GARCH(1,1) model, employing each paired combination of values for ν, ξ, for N = 500 replications, each of size T = 3000 observations."}],"known":[],"sample":{"start":null,"end":null,"market":null,"asset_class":null,"frequency":null},"data_resolution":"daily"},"transaction_costs":{"status":"not_applicable","modeled":"n/a","headline_net_of_costs":null,"detail":null,"quote":null,"source":"front_matter","rule":"R1","candidates":{"front_matter":"n/a","evidence_card":"unclear","evidence_card_quoted":false},"meaning":"the paper reports no trading or execution result"},"results":{"status":"available","evaluation_type":"forecast-accuracy","findings":[{"finding":"QFHS outperforms filtered HS (FHS) for high kurtosis and skewed data, especially when αest is in [0.05, 0.15].","kind":"performance","quote":"Finally, in risk forecasting applications to both real and simulated data, we find that QFHS‑based predictors generated by choosing αest in [0.05, 0.15] tend to outperform their FHS‑based counterparts for high kurtosis and skewed data"},{"finding":"The most accurate QFHS approach typically uses an estimation quantile αest greater than the target risk level α0.","kind":"predictive-accuracy","quote":"the most accurate QFHS approach, with typically smallest RMSEs, has αest > α0, highlighting an advantage of our flexible approach"},{"finding":"For one‑step‑ahead VaR forecasts, estimating CAViaR at αest ∈ {0.05,0.1,0.15} yields lower RMSE than using αest = α0.","kind":"predictive-accuracy","quote":"it is more accurate to estimate the CAViaR model at αest =∈ {0.05, 0.1, 0.15} and take the sample 100α0 percentile of a bootstrap sample of the quantile standardised returns"},{"finding":"QFHS methods compare less favorably to GARCH‑FHS when the DGP has Gaussian errors, but improve as kurtosis increases.","kind":"performance","quote":"Figures 1 and 2 and Table 1 highlight that QFHS methods don’t compare favourably with GARCH-FHS (GHS) when the DGP has Gaussian errors"}],"reported_performance":null,"out_of_sample":{"method":"holdout","headline_out_of_sample":null}},"independent_evidence":{"status":"none_found","supports":[],"contradicts":[],"qualifies":[],"compared_with":32,"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":1,"extracted_at":"2026-11-14T17:44:06Z"},"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/"}