{"paper_id":"2401.00188","title":"Enhancing CVaR portfolio optimisation performance with GAM factor models","authors":["Davide Lauria","W. Brent Lindquist","Svetlozar T. Rachev"],"paper_date":"2023-12-30","paper_type":"empirical","paper_url":"https://arxiv.org/pdf/2401.00188v1","page_url":"https://thequant.space/flowcharts/enhancing-cvar-portfolio-optimisation-performance-with-gam-f/","scores":{"empirical_rigor":8,"math_complexity":7.5,"quadrant":"Holy Grail"},"proposes":{"status":"available","statement":"The paper proposes a discrete‑time econometric model that combines autoregressive filters with factor regressions to predict stock returns for portfolio optimisation.","quote":"We propose a discrete-time econometric model that combines autoregressive filters with factor regressions to predict stock returns for portfolio optimisation purposes.","summary":"We propose a discrete-time econometric model that combines autoregressive filters with factor regressions to predict stock returns for portfolio optimisation purposes. In particular, we test both robust linear regressions and general additive models on two different investment universes composed of…","strategy":null},"assumptions":{"status":"available","items":[{"assumption":"The studied markets (DJIA and S&P 500) are highly efficient.","type":"market","stated":"explicit","quote":"This choice is motivated by the fact that these markets are highly efficient, and positive results indicate that this method should be extended to more inefficient markets to obtain improved performances."},{"assumption":"Returns are modelled with an ARMA‑GARCH time‑series model and the standardized innovations are assumed to follow an NIG distribution that is in the domain of attraction of the normal distribution, allowing Gaussian assum","type":"model","stated":"explicit","quote":"Here, the standardised innovation for asset hi,t is assumed to follow an NIG distribution, which is in the domain of attraction of the normal distribution, allowing the estimation to be obtained under Gaussian assumptions."},{"assumption":"No short‑selling is allowed in the optimisation.","type":"execution","stated":"explicit","quote":"we also impose a portfolio turnover constraint of 5% and rule out short-selling positions in order to limit additional complexities"},{"assumption":"A flat transaction cost of two basis points per trade is applied.","type":"execution","stated":"explicit","quote":"transaction in order to take into account re‑balancing fees that may heavily"},{"assumption":"A rolling estimation window of 765 days is used to fit the models.","type":"data","stated":"explicit","quote":"For each trading day, we estimate the model for a rolling window of Tw = 765 days, generate S = 10000 one-step-ahead scenarios for asset returns"},{"assumption":"The factor set consists of 20 Bloomberg‑derived momentum, fundamental and technical indicators.","type":"data","stated":"explicit","quote":"We describe here the 20 factors used in the regressions by dividing them into three main categories: momentum, fundamental, and technical indicators."}]},"datasets":{"status":"available","items":[{"name":"Dow Jones Industrial Average constituents","provider":"Dow Jones","kind":"market-prices","access":"public","period":null,"universe":"29 assets (DJ30)","quote":"two different investment universes composed of the Dow Jones Industrial Average and the Standard & Poor’s 500 indexes"},{"name":"Standard & Poor’s 500 constituents (sub‑space)","provider":"S&P","kind":"market-prices","access":"public","period":null,"universe":"29 assets (S&P500 sub‑space)","quote":"two different investment universes composed of the Dow Jones Industrial Average and the Standard & Poor’s 500 indexes"}],"known":[{"id":"bloomberg","name":"Bloomberg","provider":"Bloomberg L.P.","kind":"market-prices","access":"commercial","quote":"In this work, we used the Bloomberg-calculated or company-reported EPS, which follows the Bloomberg’s adjusted EPS consensus estimate.","mentions":23,"used_in":1}],"sample":{"start":null,"end":null,"market":null,"asset_class":null,"frequency":null},"data_resolution":"daily"},"transaction_costs":{"status":"available","modeled":"yes","headline_net_of_costs":null,"detail":"A flat cost of two basis points per transaction is applied to all trades.","quote":"transaction in order to take into account re‑balancing fees that may heavily","source":"front_matter","rule":"R3","candidates":{"front_matter":"yes","evidence_card":"yes","evidence_card_quoted":true},"card_detail":"A flat cost of two basis points per transaction is applied to all trades.","card_quote":"transaction in order to take into account re‑balancing fees that may heavily","meaning":"costs are part of the reported headline results"},"results":{"status":"available","evaluation_type":"trading-pnl","findings":[{"finding":"The GAM factor model shows higher explanatory power than the robust linear regression, with R² shifted to the right, lower MAE and lower BIC.","kind":"statistical","quote":"It is clear that not only the location but also the mass of the distribution of R2 of the GAM is shifted toward the right with respect to the RLR case, which suggests that the GAM factor model has a greater explanatory power. The distributions of the MAE and the BIC information criterion confirm that the GAM regression"},{"finding":"GAM regressions yield a substantially higher proportion of significant regressions across all 20 factors, especially the 30‑day RSI momentum indicator.","kind":"statistical","quote":"regressions for all the factors, and the momentum indicator RSI with a 30-day interval achieved the highest score."},{"finding":"Out‑of‑sample portfolio performance improves when the GAM factor model is used, as shown by higher portfolio values over the six‑year horizon.","kind":"performance","quote":"A comparison of the results of the optimal strategies obtained with the three different models also shows an increase in the out-of-sample performance when the GAM regression is used."},{"finding":"Summary statistics indicate higher total and annualised out‑of‑sample returns, lower average turnover and reduced maximum drawdowns for the GAM‑based optimal strategy.","kind":"performance","quote":"Table A.4 contains summary statistics such as the total and annualised out-of-sample returns, the average turnover (AvgT), and the maximum drawdowns (MD) for each optimal strategy."},{"finding":"The GAM factor model explains a larger share of variance in the innovations compared to the ARMA‑GARCH model, as shown by PCA analysis.","kind":"statistical","quote":"The amount of variance explained by the time series model and the GAM factor model, respectively, can be computed by mean of Principal Components Analysis (PCA)."}],"reported_performance":null,"out_of_sample":{"method":"walk-forward","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":"2026-11-14T18:17:13Z"},"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/"}