{"paper_id":"2308.02491","title":"Mapping Global Value Chains at the Product Level","authors":["Lea Karbevska","César A. Hidalgo"],"paper_date":"2023-06-12","paper_type":"empirical","paper_url":"https://arxiv.org/pdf/2308.02491v1","page_url":"https://thequant.space/flowcharts/mapping-global-value-chains-at-the-product-level/","scores":{"empirical_rigor":6,"math_complexity":4,"quadrant":"Street Traders"},"proposes":{"status":"available","statement":"The paper proposes the “Backward & Forward” method that combines machine‑learning and trade‑theory to infer product‑level value‑chain relationships from fine‑grained international trade data.","quote":"Here, we introduce a method based on machine learning and trade theory to infer product-level value chain relationships from fine-grained international trade data.","summary":"Value chain data is crucial to navigate economic disruptions, such as those caused by the COVID-19 pandemic and the war in Ukraine. Yet, despite its importance, publicly available value chain datasets, such as the World Input-Output Database'', Inter-Country Input-Output Tables’’, EXIOBASE'' or the…","strategy":null},"assumptions":{"status":"available","items":[{"assumption":"Imports are allocated to sectors in the same proportion as local inputs (proportional allocation).","type":"model","stated":"explicit","quote":"That is, they assume, for instance, that if 20% of the steel produced in a country is used for the production of machinery, then 20% of the steel imported from any country will also be used for the production of machinery."},{"assumption":"A location with RCA > 1 in a product is specialized in that product.","type":"data","stated":"explicit","quote":"RCA larger than 1 in a product, we say that the location is specialized in that product since it exports more than what it is expected for a location of the same size and for a product with the same global market."},{"assumption":"Specialization patterns across geographies can be used as features to recover global value‑chain information.","type":"model","stated":"implied","quote":"Our hypothesis is that by exploiting specialization patterns across multiple geographies we can generate features that when fed in a machine learning model can recover information about global value chains."},{"assumption":"The fine‑grained trade data (2017‑2020) is sufficiently representative of product‑level flows.","type":"data","stated":"implied","quote":"We leverage fine-grained international data compiled by the Observatory of Economic Complexity"},{"assumption":"The OECD Inter‑Country Input‑Output table provides a valid benchmark for fine‑tuning the product‑level dataset.","type":"data","stated":"implied","quote":"construct a product-level dataset fine-tuned on the OECD Inter‑Country Input‑Output table"}]},"datasets":{"status":"available","items":[{"name":"Observatory of Economic Complexity trade data","provider":"Observatory of Economic Complexity (oec.world)","kind":"macro","access":"public","period":"2017-2020","universe":"1,226 HS4 and 5,890 HS6 products across 9 countries (351 regions)","quote":"spanning from the year 2017 to 2020. This is data on exports and imports at the regional level for 1,226 HS4 level and 5,890 HS6 level unique products."},{"name":"OECD Inter‑Country Input‑Output table","provider":"OECD","kind":"macro","access":"public","period":null,"universe":null,"quote":"construct a product-level dataset fine-tuned on the OECD Inter‑Country Input‑Output table"}],"known":[],"sample":{"start":null,"end":null,"market":null,"asset_class":null,"frequency":null},"data_resolution":"unclear"},"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":"The Backward & Forward method outperforms a random baseline, identifying at least one correct input for more than 40% of the data—over twice the random baseline performance.","kind":"performance","quote":"the “Backward & Forward” method outperforms the random baseline, identifying at least one correct input for more than 40% of the data, which is more than twice as much as the capabilities of a random baseline model."},{"finding":"The method correctly identifies three inputs for complex products such as cars, integrated circuits, computers, and telephones.","kind":"performance","quote":"We also demonstrate that the method identifies correctly three inputs for complex products such as cars, integrated circuits, computers, and telephones."}],"reported_performance":null,"out_of_sample":{"method":"none","headline_out_of_sample":null}},"independent_evidence":{"status":"none_found","supports":[],"contradicts":[],"qualifies":[],"compared_with":8,"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":"2026-11-14T18:25:53Z"},"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/"}