{"paper_id":"2310.00553","title":"Robust Asset-Liability Management","authors":["Tjeerd de Vries","Alexis Akira Toda"],"paper_date":"2023-10-01","paper_type":"empirical","paper_url":"https://arxiv.org/pdf/2310.00553v3","page_url":"https://thequant.space/flowcharts/robust-asset-liability-management/","scores":{"empirical_rigor":6,"math_complexity":8,"quadrant":"Holy Grail"},"proposes":{"status":"available","statement":"The paper proposes a bond portfolio solution based on ambiguity‑averse preferences that generalizes classical immunization and can be computed via a simple generalized least squares problem.","quote":"We address this question by proposing a bond portfolio solution based on ambiguity-averse preferences, which generalizes classical immunization and accommodates arbitrary liability structures, portfolio constraints, and interest rate perturbations.","summary":"How should financial institutions hedge their balance sheets against interest rate risk when managing long-term assets and liabilities? We address this question by proposing a bond portfolio solution based on ambiguity-averse preferences, which generalizes classical immunization and accommodates…","strategy":null},"assumptions":{"status":"available","items":[{"assumption":"Bonds and liability pay out on finitely many dates (discrete payouts).","type":"model","stated":"explicit","quote":"Assumption 1 (Discrete payouts). The bonds and liability pay out on finitely many dates, whose union is denoted by {tn}N n=1 ⊂(0, T]."},{"assumption":"Admissible portfolios are non‑empty, closed and satisfy value matching.","type":"model","stated":"explicit","quote":"Assumption 2 (Portfolio constraint). The set of admissible portfolios Z ⊂RJ is nonempty and closed. Furthermore, all z ∈Z satisfy value matching: P(x) = J X j=1 zjPj(x)."},{"assumption":"Cumulative discount rates are continuous functions on [0,T] with x(0)=0.","type":"model","stated":"explicit","quote":"Let the yields be in C[0, T], the vector space of continuous functions on [0, T] endowed with the supremum norm"},{"assumption":"Basis functions hi are polynomials of degree i with hi(0)=0 (Assumption 3).","type":"model","stated":"explicit","quote":"For each i, it is natural to choose hi such that hi is a polynomial of degree i with hi(0) = 0, for Assumption 3 then holds (Lemma D.2)."},{"assumption":"Perturbations h represent realized changes in the cumulative discount rate over holding periods of 1–100 trading days, with formation dates spanning 1996–2023.","type":"data","stated":"explicit","quote":"h to be the realized change in the cumulative discount rate from date s to date s + d, for holding periods of d = 1, . . . , 100 trading days and formation dates s spanning 1996–2023."},{"assumption":"Rebalancing trades incur proportional transaction costs equal to half the quoted bid‑ask spread, ranging from 0.5 bps to 3 bps.","type":"execution","stated":"explicit","quote":"The proportional cost is half the quoted bid-ask spread, rising from half a basis point of price at the short end to three basis points at twenty years..."}]},"datasets":{"status":"available","items":[{"name":"empirical yield curve data","provider":null,"kind":"market-prices","access":"unknown","period":"1996–2023","universe":null,"quote":"h to be the realized change in the cumulative discount rate from date s to date s + d, for holding periods of d = 1, . . . , 100 trading days and formation dates s spanning 1996–2023."},{"name":"simulated yield curve paths","provider":null,"kind":"synthetic","access":"unknown","period":null,"universe":null,"quote":"Figure 5: Distribution of absolute return error on simulated paths."},{"name":"discount curve database","provider":null,"kind":"market-prices","access":"unknown","period":null,"universe":null,"quote":"We use their basis functions and discount curve database to check the robustness of our results."}],"known":[],"sample":{"start":null,"end":null,"market":null,"asset_class":null,"frequency":null},"data_resolution":"daily"},"transaction_costs":{"status":"available","modeled":"yes","headline_net_of_costs":null,"detail":"Proportional cost equal to half the bid‑ask spread, from 0.5 bps (short end) to 3 bps (20‑year), applied to each rebalancing trade.","quote":"The proportional cost is half the quoted bid-ask spread, rising from half a basis point of price at the short end to three basis points at twenty years...","source":"front_matter","rule":"R3","candidates":{"front_matter":"yes","evidence_card":"yes","evidence_card_quoted":true},"card_detail":"Proportional cost equal to half the bid‑ask spread, from 0.5 bps (short end) to 3 bps (20‑year), applied to each rebalancing trade.","card_quote":"The proportional cost is half the quoted bid-ask spread, rising from half a basis point of price at the short end to three basis points at twenty years...","meaning":"costs are part of the reported headline results"},"results":{"status":"available","evaluation_type":"pricing-calibration","findings":[{"finding":"RI(2) outperforms KRD by roughly fifteenfold for pension liabilities and ninefold for medium liabilities in 95th percentile underfunding ratio.","kind":"performance","quote":"Under the leverage cap (Panel B), RI(2) is feasible for four of the five liabilities and outperforms KRD in each, by roughly fifteenfold for pension (0.15% against 2.27%) and ninefold for medium."},{"finding":"RI(2) attains the lowest mean error and mean‑squared error for both liabilities, with the lowest gross leverage and turnover among competitors.","kind":"performance","quote":"RI(2) attains the lowest mean error and mean squared error for both liabilities, at the lowest gross leverage and turnover among the competitive"},{"finding":"Unconstrained RI portfolios require median gross leverage of only 1.1 (RI(0)), 2.3 (RI(1)), and 2.6 (RI(2)), whereas exact HD needs leverage >43 and is infeasible.","kind":"performance","quote":"Exact HD demands a gross leverage of 43 for fullHorizon and above 200 for longRun, so it is infeasible in every panel and omitted, whereas the un‑constrained RI portfolios require a median gross leverage of only 1.1 (RI(0)), 2.3 (RI(1)), and 2.6 (RI(2))."},{"finding":"KRD’s mean terminal error reaches 13–16% with a 99th percentile growing linearly to 30% over ten years, reflecting a persistent duration gap.","kind":"performance","quote":"KRD fares much worse: its mean terminal error reaches 13–16%, with a 99th percentile that grows linearly to 30% over the ten years."},{"finding":"Transaction costs amount to at most a few basis points per year and do not affect the ranking of methods.","kind":"performance","quote":"Transaction costs do not affect any of these rankings: they amount to at most a few basis points per year (Table 2), and because costs are linear in the spread and RI(2) has the lowest turnover…"}],"reported_performance":null,"out_of_sample":{"method":"holdout","headline_out_of_sample":null}},"independent_evidence":{"status":"none_found","supports":[],"contradicts":[],"qualifies":[],"compared_with":30,"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":"2026-11-14T18:21:37Z"},"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/"}