{"paper_id":"2305.17285","title":"Critical density for network reconstruction","authors":["Andrea Gabrielli","Valentina Macchiati","Diego Garlaschelli"],"paper_date":"2023-05-26","paper_type":"empirical","paper_url":"https://arxiv.org/pdf/2305.17285v1","page_url":"https://thequant.space/flowcharts/critical-density-for-network-reconstruction/","scores":{"empirical_rigor":3,"math_complexity":7.5,"quadrant":"Lab Rats"},"proposes":{"status":"available","statement":"The paper introduces the concept of reconstructability for financial network reconstruction and analyzes the critical link density threshold that separates reconstructable from unreconstructable regimes.","quote":"Here we consider one of the most successful network reconstruction methods, producing random graphs with desired link density and where the observed constraints (related to the market size of each node) are replicated as averages over the graph ensemble, but not in individual realizations. In this work we introduce the problem of reconstructability of the network under ensemble methods.","summary":"The structure of many financial networks is protected by privacy and has to be inferred from aggregate observables. Here we consider one of the most successful network reconstruction methods, producing random graphs with desired link density and where the observed constraints (related to the market…","strategy":null},"assumptions":{"status":"available","items":[{"assumption":"The problem is framed within undirected networks, interpreting each link as the bilateral exposure.","type":"model","stated":"explicit","quote":"While these relationships are clearly directional, for simplicity in this paper we will frame the problem within the context of undirected networks, where we interpret each link as representing the bilateral exposure (the sum of the exposure in both directions) between two nodes.","support":"full"},{"assumption":"Outgoing and incoming link weight sums (total interbank assets and liabilities) are observable and used as node strength constraints.","type":"data","stated":"explicit","quote":"Individual links are unobservable, while the sum of the outgoing link weights and the sum of the incoming link weights for each node are both observable, as they represent the total interbank assets and the total interbank liabilities that can be derived from the balance sheet of the corresponding institution.","support":"full"},{"assumption":"Sparse network regime is defined by a link density scaling d(z) ≃ kN with finite k > 1.","type":"model","stated":"explicit","quote":"This implies that a choice for z could be z = zsparse where zsparse is such that d(zsparse) ≃k N , k > 1 (4) with k finite. Clearly, the value zsparse realizing the above condition depends on the entire strength sequence, and in general on the strength distribution when N →∞.","support":"full"},{"assumption":"Empirical strengths are generated by i.i.d. sampling from a fixed probability density f(s) that does not depend on N.","type":"statistical","stated":"explicit","quote":"To calculate mini=1,N{s∗ i } as a function of N, we assume that the N values {s∗ i }N i=1 of the empirical strength are realized via sampling N times (in an i.i.d. manner) from some probability density function f(s). The same results can be derived more rigorously by analysing the large N convergence of the probability distribution of the random variable WSL (whose possible realized values will be denoted as wSL).","support":"full"},{"assumption":"Link weights are deterministic functions of the observable margins when a link is realized.","type":"model","stated":"explicit","quote":"In the simplest specifications of the model [7,15], the link weights are deterministic functions of the observable margins: if the link is realized, the random variable wij is assigned a value s∗ i s∗ j 2W ∗pij(z), otherwise it is given a zero value.","support":"full"},{"assumption":"The strength distribution f(s) is assumed to be independent of network size N (sparse networks with 2W* = N).","type":"statistical","stated":"explicit","quote":"The same results can be derived more rigorously by analysing the large N convergence of the probability distribution of the random variable WSL (whose possible realized values will be denoted as wSL).","support":"full"}]},"datasets":{"status":"available","items":[{"name":"Bankscope dataset","provider":"Bureau Van Dijk Bankscope","kind":"fundamentals","access":"commercial","period":"2006-2013","universe":"119 anonymized European banks","quote":"Our dataset consists of a subset of N = 119 anonymized European banks that were publicly traded between 2006 and 2013 [21]. For each of the N banks and each year t in the data, we have access Critical density for network reconstruction 17 10 1 10 3 10 5 10 7 Total Assets 2007 (mil USD) 10 1 10 0 10 1 10 2 10 3 10 4 10 5 10 6 10 7 Interbank Assets 2007 (mil USD) 10 1 10 3 10 5 10 7 Total Assets 2007 (mil USD) 10 1 10","support":"full"}],"known":[],"sample":{"start":null,"end":null,"market":null,"asset_class":null,"frequency":null},"data_resolution":"weekly-monthly"},"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":"none","findings":[{"finding":"There exists a minimum critical link density below which the reconstruction method enters an unreconstructability phase, where constraints are reproduced on average but not in typical individual realizations.","kind":"theoretical","quote":"We show that there is a minimum critical link density below which the method exhibits an `unreconstructability' phase where at least one of the constraints, while still reproduced on average, is far from its expected value in typical individual realizations.","support":"full"},{"finding":"Dense networks are always reconstructable, whereas sparse homogeneous networks are always unreconstructable.","kind":"theoretical","quote":"We also find that, while dense networks are always reconstructable, sparse networks are unreconstructable if their structure is homogeneous, while they can display a crossover to reconstructability if they have an appropriate core-periphery or heterogeneous structure.","support":"full"},{"finding":"Introducing heterogeneity via a core‑periphery structure can restore reconstructability in sparse networks.","kind":"theoretical","quote":"3.2 Core-periphery structure We now consider a less trivial setting where we introduce some heterogeneity in the network in the form of a core-periphery structure, where nodes in the core have a larger strength and nodes in the periphery have a smaller strength. The presence of a core-periphery structure has been documented in real financial networks [17,18,19,20].","support":"full"},{"finding":"The scaling of the critical density differs from the thresholds for the emergence of a giant component or a unique component in the graph.","kind":"theoretical","quote":"We establish the scaling of the critical density for various theoretical and empirical distributions of interbank assets and liabilities, showing that the threshold differs from the critical densities for the onset of the giant component and of the unique component in the graph.","support":"full"},{"finding":"Empirical analysis on the Bankscope data confirms the presence of two regimes (reconstructable and unreconstructable) depending on network size, strength distribution details, and the decay parameter β.","kind":"other","quote":"Realistic situations typically fall into one of those two cases. The next Section provides an empirical example that confirms this picture.","support":"full"}],"reported_performance":null,"out_of_sample":{"method":"none","headline_out_of_sample":null}},"independent_evidence":{"status":"available","supports":[],"contradicts":[],"qualifies":[],"compared_with":15,"citations":{"cited_by_in_corpus":3,"by_independent_authors":1,"how_cited":{"used_as_baseline":0,"uses_method_or_data":1,"extends":0,"discusses_findings":0,"mentions":0,"unclassified":2},"independent_papers":[{"paper":{"paper_id":"2506.21426","title":"Evolution and determinants of firm-level systemic risk in local production networks","paper_date":"2025-06-26","authors":["Anna Mancini","Balázs Lengyel","Riccardo Di Clemente","Giulio Cimini"],"empirical_rigor":8.5,"page_url":"https://thequant.space/flowcharts/evolution-and-determinants-of-firm-level-systemic-risk-in-local-production/"},"how":"uses_method_or_data","quote":null}],"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":2,"extracted_at":"2026-11-14T20:02:45Z"},"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/"}