{"paper_id":"2503.01099","title":"A Dynamic Model of Private Asset Allocation","authors":["Hui Chen","Giovanni Gambarotta","Simon Scheidegger","Yu Xu"],"paper_date":"2025-03-03","paper_type":"empirical","paper_url":"https://arxiv.org/pdf/2503.01099v1","page_url":"https://thequant.space/flowcharts/a-dynamic-model-of-private-asset-allocation/","scores":{"empirical_rigor":7,"math_complexity":8.5,"quadrant":"Holy Grail"},"proposes":{"status":"available","statement":"The paper proposes a state-of-the-art dynamic model of private asset allocation that uses Deep Kernel Gaussian Processes to compute optimal investment policies.","quote":"We develop a state-of-the-art dynamic model of private asset allocation. It captures (1) the illiquidity inherent in private asset investing, (2) the lag between initial capital commitments, subsequent capital calls, and eventual distributions, (3) time-varying market and business cycle conditions, (4) serial correlati","summary":"We build a state-of-the-art dynamic model of private asset allocation that considers five key features of private asset markets: (1) the illiquid nature of private assets, (2) timing lags between capital commitments, capital calls, and eventual distributions, (3) time-varying business cycle…","strategy":null},"assumptions":{"status":"available","items":[{"assumption":"Risk-free bonds have zero risk weight, public stocks and private equity have positive risk weights with PE weight not lower than stocks.","type":"model","stated":"explicit","quote":"That is, risk-free bonds carry a zero risk weight as they are risk free while public stocks and PE carry positive risk weights. The risk weight for PE is no lower than that of public stocks."},{"assumption":"A quadratic cost is incurred when the LP’s risk weight exceeds a threshold θ=1.","type":"model","stated":"explicit","quote":"That is, no costs are assessed if the LP’s risk weight (4) falls below the threshold θ; quadratic costs are assessed above the threshold θ. Without loss of generality, we can take the threshold to be θ “ 1."},{"assumption":"The macroeconomic environment follows a two‑state Markov chain with recession (s=1) and expansion (s=2) states.","type":"model","stated":"explicit","quote":"We take state s “ 1 to be a recessionary state and state s “ 2 to be an expansionary state. We set the transition probabilities for the macroeconomic state to p12 “ 0.25 and p21 “ 0.05 based on the duration of NBER recessions in the post‑WWII sample."},{"assumption":"PE call and distribution rates are regime‑dependent and estimated from LMI data.","type":"data","stated":"explicit","quote":"We estimate regression (25) separately for recessionary (s “ 1) and expansionary (s “ 2) regimes to obtain"},{"assumption":"The liquidation discount for PE stakes is based on secondary‑market transaction cost estimates.","type":"model","stated":"explicit","quote":"We calibrate the discount from liquidating PE investments αpsq based on estimates of transaction costs in the secondary market for selling PE stakes from Nadauld,"},{"assumption":"Public‑stock transaction costs are negligible relative to private‑equity costs.","type":"execution","stated":"explicit","quote":"Since transaction costs for public stocks are negligible in comparison, this approach offers a cost‑effective means of managing risk over the business cycle."}]},"datasets":{"status":"available","items":[{"name":"Liberty Mutual Investments PE holdings","provider":"Liberty Mutual Investments","kind":"alternative","access":"proprietary","period":"1996Q4‑2023Q4","universe":"Buyout funds (when number of buyout investees first reaches 10)","quote":"We use quarterly data provided by Liberty Mutual Investments (LMI) on its PE holdings to calibrate various PE‑related parameters."},{"name":"Kenneth French’s risk‑free rate and aggregate stock returns series","provider":"Kenneth French","kind":"macro","access":"public","period":null,"universe":null,"quote":"We calibrate the returns processes as follows. We take the series for the risk‑free rate and aggregate stock returns from Kenneth French’s website."},{"name":"NBER recession dates","provider":"National Bureau of Economic Research","kind":"macro","access":"public","period":null,"universe":null,"quote":"In what follows, we use NBER recession dates when estimating parameter values that depend on the macroeconomic state."}],"known":[{"id":"french-library","name":"Kenneth French Data Library","provider":"Kenneth R. French","kind":"market-prices","access":"public","quote":"We take the series for the risk-free rate and aggregate stock returns from Kenneth French’s website.","mentions":1,"used_in":1}],"sample":{"start":null,"end":null,"market":null,"asset_class":null,"frequency":null},"data_resolution":"weekly-monthly"},"transaction_costs":{"status":"available","modeled":"yes","headline_net_of_costs":null,"detail":"PE liquidation discount α calibrated to 66% purchase price in recessions and 90% in expansions; risk‑budget violation costs quadratic above threshold.","quote":"We calibrate the discount from liquidating PE investments αpsq based on estimates of transaction costs","source":"evidence_card","rule":"R2","candidates":{"front_matter":"unclear","evidence_card":"yes","evidence_card_quoted":true},"card_detail":"PE liquidation discount α calibrated to 66% purchase price in recessions and 90% in expansions; risk‑budget violation costs quadratic above threshold.","card_quote":"We calibrate the discount from liquidating PE investments αpsq based on estimates of transaction costs","meaning":"costs are part of the reported headline results"},"results":{"status":"available","evaluation_type":"unclear","findings":[{"finding":"Under the optimal policy the cumulative default rate over a ten‑year horizon is only 0.1%, versus 13.6% for a naive policy that ignores business cycles.","kind":"performance","quote":"Notably, under the optimal policy, the cumulative default rate is only 0.1% over the ten‑year investment horizon."},{"finding":"Higher risk‑charge scenarios reduce portfolio risk but also lower average annualized returns and certainty‑equivalent wealth.","kind":"performance","quote":"... The average annualized return decreases to 7.08% under the second scenario (down from 8.41% for the baseline)"},{"finding":"The optimal allocation keeps the intended private‑equity transition profile and adjusts public‑stock exposure to manage business‑cycle risk.","kind":"theoretical","quote":"To first order, the optimal allocation maintains the intended transition profile for the PE portfolio, as if short‑term business cycle fluctuations did not exist."},{"finding":"Deep Kernel Gaussian Processes enable solving the high‑dimensional dynamic model that would be infeasible with traditional methods.","kind":"theoretical","quote":"We instead use Machine Learning methods—Deep Kernel Gaussian Process dynamic programming—to make the problem computationally feasible."},{"finding":"The model’s optimal policies can differ dramatically from those derived using static mean‑variance heuristics.","kind":"performance","quote":"obtained under our fully dynamic framework can differ dramatically from those derived using heuristic methods."}],"reported_performance":null,"out_of_sample":{"method":"none","headline_out_of_sample":null}},"independent_evidence":{"status":"none_found","supports":[],"contradicts":[],"qualifies":[],"compared_with":44,"citations":{"cited_by_in_corpus":1,"by_independent_authors":1,"how_cited":{"used_as_baseline":0,"uses_method_or_data":0,"extends":0,"discusses_findings":0,"mentions":0,"unclassified":1},"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":"2026-11-14T17:43:52Z"},"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/"}