{"paper_id":"2312.17375","title":"Causal Discovery in Financial Markets: A Framework for Nonstationary Time-Series Data","authors":["Agathe Sadeghi","Achintya Gopal","Mohammad Fesanghary"],"paper_date":"2023-12-28","paper_type":"methodological","paper_url":"https://arxiv.org/pdf/2312.17375v3","page_url":"https://thequant.space/flowcharts/causal-discovery-in-financial-markets-a-framework-for-nonst/","scores":{"empirical_rigor":6,"math_complexity":8,"quadrant":"Holy Grail"},"proposes":{"status":"available","statement":"The paper proposes a new causal structure learning algorithm called CDNOTS for nonstationary time series that handles lagged dependencies.","quote":"We introduce a refined version (CDNOTS) which is designed specifically to account for lagged dependencies in time series data.","summary":"This paper introduces a new causal structure learning method for nonstationary time series data, a common data type found in fields such as finance, economics, healthcare, and environmental science. Our work builds upon the constraint-based causal discovery from nonstationary data algorithm…","strategy":null},"assumptions":{"status":"available","items":[{"assumption":"Objective assumptions such as stationarity and linearity are statistically testable and should be aligned with the data.","type":"statistical","stated":"explicit","quote":"These are assumptions which are statistically testable such as stationarity and linearity."},{"assumption":"Algorithmic assumptions include choices like significance threshold and type of conditional independence test, which cannot be directly evaluated.","type":"model","stated":"explicit","quote":"These assumptions refer to decisions in the causal discovery algorithm which can neither be directly evaluated nor justified purely by domain expertise."},{"assumption":"Subjective assumptions such as pseudo‑sufficiency, consistency, faithfulness, and randomness are not statistically testable and rely on domain expertise.","type":"other","stated":"explicit","quote":"These are assumptions which cannot be statistically tested and require domain expertise to evaluate. Pseudo-sufficiency, consistency, faithfulness, and randomness (assumptions made by CDNOTS) fall in this category."}]},"datasets":{"status":"available","items":[{"name":"Normalized Log Price‑to‑Book ratio (N. LPTB) and stock returns (Ret) for financial compani","provider":"Bloomberg","kind":"fundamentals","access":"commercial","period":null,"universe":"financial securities","quote":"Results of CDNOTS KCIT run on financials data for log Price‑to‑Book ratio (LPTB) and stock returns (Ret). Normalized LPTB (N. LPTB) refers to normalizing the LPTB by the mean and standard deviation of the LPTB of financial companies in that quarter.","known_id":"bloomberg"},{"name":"Synthetic data experiments","provider":null,"kind":"synthetic","access":"unknown","period":null,"universe":null,"quote":"Our synthetic data experiments indicate the effectiveness of ParCorr when data is limited, as observed in our country‑level experiments."},{"name":"Economic data case study (KCIT and RCoT)","provider":null,"kind":"macro","access":"unknown","period":null,"universe":null,"quote":"Appendix C. Economic Data Case Study with KCIT and RCoT"}],"known":[],"sample":{"start":null,"end":null,"market":null,"asset_class":null,"frequency":null},"data_resolution":"weekly-monthly"},"transaction_costs":{"status":"available","modeled":"no","headline_net_of_costs":null,"detail":null,"quote":"In Figure 15b, we see this simple approach outperforms a market‑cap weighted strategy","source":"front_matter","rule":"R3","candidates":{"front_matter":"no","evidence_card":"no","evidence_card_quoted":true},"card_detail":null,"card_quote":"In Figure 15b, we see this simple approach outperforms a market‑cap weighted strategy","meaning":"a trading or execution result is reported without costs"},"results":{"status":"available","evaluation_type":"trading-pnl","findings":[{"finding":"CDNOTS identifies statistically significant lagged relationships and nonstationarity in the financial data.","kind":"statistical","quote":"we see statistically significant lagged relationships and statistically significant nonstationarity, implying we would not be able to trust the results of an algorithm that cannot handle these types of properties."},{"finding":"A simple strategy based on normalized LPTB yields higher Sharpe ratios (0.76‑0.81) than a market‑cap weighted portfolio (Sharpe 0.63).","kind":"performance","quote":"Threshold: 0.0, Sharpe: 0.76 ... Threshold: 0.25, Sharpe: 0.81"},{"finding":"Linearity test for normalized LPTB and returns rejects linearity with p‑value 0.0.","kind":"statistical","quote":"... find the p‑value to be 0.0 (i.e., we can reject the null hypothesis that the relationship is linear)."}],"reported_performance":null,"out_of_sample":{"method":"none","headline_out_of_sample":null}},"independent_evidence":{"status":"none_found","supports":[],"contradicts":[],"qualifies":[],"compared_with":27,"citations":{"cited_by_in_corpus":3,"by_independent_authors":3,"how_cited":{"used_as_baseline":0,"uses_method_or_data":0,"extends":0,"discusses_findings":0,"mentions":0,"unclassified":3},"independent_papers":[],"scope":"Citations between papers in this corpus only; citations from outside it are not counted."}},"implementation":{"status":"available","public":true,"code_url":"https://github.com/bloomberg/causal-ts","note":null},"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":"2026-11-14T16:34:53Z"},"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/"}