{"paper_id":"2609.35359","title":"From Cointegration to Out-of-Sample Failure: A Pairs-Trading Case Study on PEP-KO","authors":["Davide Graziano"],"paper_date":"2026-09-28","paper_type":"empirical","paper_url":"https://arxiv.org/pdf/2609.35359","page_url":"https://thequant.space/flowcharts/from-cointegration-to-out-of-sample-failure-a-pairs-trading-case-study-on-pep-ko/","scores":{"empirical_rigor":8,"math_complexity":6.5,"quadrant":"Holy Grail"},"proposes":{"status":"available","statement":"The paper proposes a cointegration‑based pairs‑trading strategy for PepsiCo and Coca‑Cola and evaluates its statistical robustness and economic exploitability.","quote":"This paper examines whether a cointegration-based pairs trading strategy between PepsiCo and The Coca-Cola Company is statistically robust and economically exploitable.","summary":"This paper examines whether a cointegration-based pairs trading strategy between PepsiCo and The Coca-Cola Company is statistically robust and economically exploitable. We first test for cointegration and estimate the spread's mean-reversion dynamics over 2013-2018, then hold these statistical…","strategy":"cointegration-based pairs trading PEP‑KO"},"assumptions":{"status":"available","items":[{"assumption":"Log‑price series follow a first‑order autoregressive (AR(1)) process.","type":"model","stated":"explicit","quote":"We model the log-price process as a first-order autoregressive process, AR(1), Xt = α + ρXt−1 + εt, εt i.i.d. ∼N(0, σ2 ε), where we assume Cov(X0, εt) = 0 for all t."},{"assumption":"Prices are adjusted for dividends and stock splits.","type":"data","stated":"explicit","quote":"Throughout, prices are retrieved as adjusted daily closing prices, i.e. corrected for dividends and stock splits."},{"assumption":"Transaction cost is a fixed fraction of traded notional applied on entry and exit.","type":"execution","stated":"explicit","quote":"A transaction cost c, expressed as a fraction of traded notional, is applied once on entry and once on exit of each leg."},{"assumption":"Risk budget, maximum exposure, and volatility‑estimation window are fixed ex‑ante during optimisation.","type":"model","stated":"explicit","quote":"The policy parameters other than θ are fixed ex ante, with a risk budget w = 0.5%, a maximum notional exposure Nmax = 25%, a transaction cost c = 0.20%, and a volatility‑estimation window L = 64 trading days."},{"assumption":"Unit‑root and cointegration tests assume the standard asymptotic critical values (adjusted for estimated cointegrating relationship).","type":"statistical","stated":"explicit","quote":"We use the adjusted critical values to account for the fact that the cointegrating relationship is estimated from the data."},{"assumption":"The optimal policy is selected by maximising the annualised net Sharpe ratio over a predefined grid of threshold policies.","type":"model","stated":"explicit","quote":"The optimal policy θ⋆is selected by maximising the annualised net Sharpe ratio"}]},"datasets":{"status":"available","items":[{"name":"Adjusted daily closing prices of PEP and KO","provider":null,"kind":"market-prices","access":"unknown","period":"2013–2018","universe":"PEP and KO","quote":"The statistical analysis is based on the prices of PEP and KO over the period 2013–2018."},{"name":"Training window for policy optimisation","provider":null,"kind":"market-prices","access":"unknown","period":"2018–2023","universe":"PEP and KO","quote":"Optimising J over the threshold grid, using only the 2018–2023 training window, selects θ⋆= (z⋆ entry, z⋆ exit) = (2.15, 0.50)."},{"name":"Out‑of‑sample test period","provider":null,"kind":"market-prices","access":"unknown","period":"2023–present","universe":"PEP and KO","quote":"We evaluate θ⋆= (2.15, 0.50) on the 2023–present period as an OOS test."}],"known":[],"sample":{"start":"2018","end":"2023","market":"PepsiCo and Coca‑Cola stocks","asset_class":"equities","frequency":"daily"},"data_resolution":"daily"},"transaction_costs":{"status":"available","modeled":"yes","headline_net_of_costs":true,"detail":"c = 0.20% per leg, applied on entry and exit of each trade.","quote":"a transaction cost c, expressed as a fraction of traded notional, is applied once on entry and once on exit of each leg","source":"front_matter","rule":"R3","candidates":{"front_matter":"yes","evidence_card":"yes","evidence_card_quoted":true},"card_detail":"c = 0.20% per leg, applied on entry and exit of each trade.","card_quote":"a transaction cost c, expressed as a fraction of traded notional, is applied once on entry and once on exit of each leg","meaning":"costs are part of the reported headline results"},"results":{"status":"available","evaluation_type":"trading-pnl","findings":[{"finding":"In‑sample optimisation yields 9 round‑trip trades, annualised net Sharpe 0.67 and net P&L $16,941 on $100,000 capital.","kind":"performance","quote":"generating 9 round‑trip trades, an annualised net Sharpe ratio of J(θ⋆) = 0.67, and a net P&L of $16,941 on the initial capital of $100,000."},{"finding":"Excluding the COVID‑19 window reduces Sharpe to 0.28, net P&L to $5,759 and trade count to 8.","kind":"performance","quote":"trade count falls from 9 to 8, the annualised Sharpe ratio from 0.67 to 0.28, and net P&L from $16,941 to $5,759."},{"finding":"Adjusted Sharpe Ratio is 0.281 and Deflated Sharpe Ratio is 0.423, indicating limited statistical evidence of skill.","kind":"statistical","quote":"The DSR is 0.423, implying that the observed Sharpe ratio is not statistically distinguishable from the maximum Sharpe ratio that could arise by chance when searching over 901 zero‑skill policies."},{"finding":"Out‑of‑sample test produces a single losing trade with net P&L ≈ –$28,000 and Sharpe –1.19.","kind":"negative","quote":"the reported Sharpe ratio of −1.19 is not a meaningful estimate of OOS performance."},{"finding":"Post‑hoc Engle‑Granger test on OOS data fails to reject no cointegration (p = 0.120) and hedge ratio reverses sign.","kind":"statistical","quote":"A post‑hoc Engle–Granger test no longer rejects the null of no cointegration (p = 0.120). Moreover, the rolling OLS and Kalman filter adaptive estimation of the hedge ratio fail to"}],"reported_performance":{"strategy":"cointegration-based pairs trading PEP‑KO","sharpe":0.67,"sharpe_benchmark":null,"annual_return_pct":null,"total_return_pct":16.94,"max_drawdown_pct":null},"out_of_sample":{"method":"holdout","headline_out_of_sample":true}},"independent_evidence":{"status":"none_found","supports":[],"contradicts":[],"qualifies":[],"compared_with":25,"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":"2026-11-14T17:04:26Z"},"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/"}