{"paper_id":"2305.03224","title":"Carbon Price Forecasting with Quantile Regression and Feature Selection","authors":["Tianqi Pang","Kehui Tan","Chenyou Fan"],"paper_date":"2023-05-05","paper_type":"empirical","paper_url":"https://arxiv.org/pdf/2305.03224v1","page_url":"https://thequant.space/flowcharts/carbon-price-forecasting-with-quantile-regression-and-featur/","scores":{"empirical_rigor":6.5,"math_complexity":7.5,"quadrant":"Holy Grail"},"proposes":{"status":"available","statement":"The paper proposes two new methods, Sparse Quantile Group Lasso (L‑SQG) and Adaptive Sparse Quantile Group Lasso (L‑ASQG), to improve carbon price forecasting.","quote":"We named these two methods as Sparse Quantile Group Lasso (L-SQG) and Adaptive Sparse Quantile Group Lasso (L-ASQG), respectively.","summary":"Carbon futures has recently emerged as a novel financial asset in the trading markets such as the European Union and China. Monitoring the trend of the carbon price has become critical for both national policy-making as well as industrial manufacturing planning. However, various geopolitical…","strategy":null},"assumptions":{"status":"available","items":[{"assumption":"The EU carbon price time series is stationary.","type":"statistical","stated":"explicit","quote":"the ADF test value for the EU data is -13.013, the p-value is about 0.00, and the test value at the 5% significance level is -2.882 which is greater than the ADF test value, indicating that the time series is stable."},{"assumption":"The carbon price data does not follow a normal distribution.","type":"statistical","stated":"explicit","quote":"This indicates that the data does not meet the assumption of normality. However, we can still explicitly predict their quantile values with quantile regression."},{"assumption":"Sparse and group-level penalties induce sparsity in the selected factors.","type":"model","stated":"explicit","quote":"For L-SQG, we added sparsity at both per-parameter level with L1-penalty and per-group level with L2-penalty with the parametercontroling the relative importance of each level."},{"assumption":"Adaptive weights improve model flexibility by assigning different penalty weights.","type":"model","stated":"explicit","quote":"For L-ASQG, we further added adaptivity parameters to improve the flexibility of the model by assigning different penalty weights."}]},"datasets":{"status":"available","items":[{"name":"EU ETS carbon futures returns","provider":null,"kind":"market-prices","access":"unknown","period":null,"universe":null,"quote":"For EU carbon price data, we utilize the carbon futures returns in the EU ETS such a…"},{"name":"Guangzhou ETS carbon prices","provider":null,"kind":"market-prices","access":"unknown","period":null,"universe":null,"quote":"Table 1.Statistical tests for the Guangzhou ETS carbon prices"}],"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":"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":"L‑ASQG achieves the lowest error metrics on the EU ETS dataset (MSE 0.0194, MAE 0.1241, RMSE 0.1393, MAPE 2.6600).","kind":"performance","quote":"our proposed method L-ASQG has the lowest MSE (0.0194), MAE (0.1241), RMSE (0.1393) and MAPE (2.6600) compared to other Lasso-based models on the European ETS carbon price dataset."},{"finding":"L‑ASQG also attains the lowest error metrics on the Guangzhou ETS dataset (MSE 0.0362, MAE 0.1495, RMSE 0.1903, MAPE 3.0894).","kind":"performance","quote":"our proposed L-ASQG model also has the lowest MSE (0.0362), MAE (0.1495), RMSE (0.1903) and MAPE (3.0894)."},{"finding":"L‑SQG does not improve performance over the original L‑QG on the datasets.","kind":"negative","quote":"the results of the L-SQG method proposed by us are basically not better than L-QG, indicating that only adjusting the sparsity for per-parameter and per-group levels cannot improve the results obviously unless the data has larger dimensions."},{"finding":"Quantile regression (qr_Lasso) outperforms linear regression (lm_Lasso) on both EU and Chinese carbon price datasets.","kind":"predictive-accuracy","quote":"For both the lm_Lasso model and the qr_Lasso model, it is clear that qr_Lasso performs significantly better than lm_Lasso on both datasets, indicating that for high-dimensional time series models, quantile regression is significantly better than using only linear regression (least squares)."},{"finding":"The L‑ASQG model performs best under the median quantile (τ=0.5) with the lowest evaluation indicators.","kind":"performance","quote":"First, we observe that the model has the best prediction performance when τ= 0.5, and the values of all evaluation indicators are the lowest."}],"reported_performance":null,"out_of_sample":{"method":"holdout","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":1,"extracted_at":"2026-11-14T18:27:09Z"},"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/"}