{"paper_id":"2401.05080","title":"Markowitz Portfolio Construction at Seventy","authors":["Stephen Boyd","Kasper Johansson","Ronald Kahn","Philipp Schiele","Thomas Schmelzer"],"paper_date":"2024-01-10","paper_type":"empirical","paper_url":"https://arxiv.org/pdf/2401.05080v1","page_url":"https://thequant.space/flowcharts/markowitz-portfolio-construction-at-seventy/","scores":{"empirical_rigor":6.5,"math_complexity":7,"quadrant":"Holy Grail"},"proposes":{"status":"available","statement":"The paper proposes an extension of the classic Markowitz portfolio construction called Markowitz++ that adds practical constraints and handles uncertainty in return forecasts.","quote":"When we need to distinguish the extension of Markowitz’s portfolio construction that we recommend from the basic Markowitz method, we refer to it as Markowitz++.","summary":"More than seventy years ago Harry Markowitz formulated portfolio construction as an optimization problem that trades off expected return and risk, defined as the standard deviation of the portfolio returns. Since then the method has been extended to include many practical constraints and objective…","strategy":null},"assumptions":{"status":"available","items":[{"assumption":"Asset returns are modeled as Gaussian for the utility analysis.","type":"model","stated":"explicit","quote":"Specifically if we model the returns as Gaussian, and use the exponential utility U(R) = 1 −exp(−γR), then the expected utility is the risk-adjusted return, up to an additive constant"},{"assumption":"Synthetic return predictions are used to simulate a proprietary mean prediction method.","type":"data","stated":"explicit","quote":"we use synthetic return predictions to simulate a proprietary mean prediction method"},{"assumption":"Bid‑ask spread transaction costs are linear in trade size and proportional to half the spread.","type":"execution","stated":"explicit","quote":"(κspread)T|z|, where κspread ∈Rn is the vector of one‑half the asset bid‑ask spreads"},{"assumption":"Short‑selling incurs an additional 5% annualized spread over the effective federal funds rate.","type":"execution","stated":"explicit","quote":"When shorting an asset we add a 5% annualized spread over the effective federal funds rate to approximate the shorting cost in our simulation"},{"assumption":"Trade limits are enforced as bounds on the trade vector relative to forecasted trading volumes.","type":"execution","stated":"explicit","quote":"Trade limits impose lower and upper bounds on trades, as zmin ≤z ≤zmax"},{"assumption":"Cash weight is constrained between –5% and 100% to guarantee feasibility.","type":"other","stated":"explicit","quote":"We limit the cash weight to lie between −5% and 100% (which guarantees feasibility)"}]},"datasets":{"status":"available","items":[{"name":"S&P 100 daily adjusted close price data","provider":null,"kind":"market-prices","access":"public","period":"2000-01-04 to 2023-09-22","universe":"74 assets (S&P 100 constituents after filtering)","quote":"We use daily adjusted close price data from 2000-01-04 to 2023-09-22. … We"},{"name":"Bid‑ask spread data","provider":null,"kind":"market-prices","access":"public","period":null,"universe":null,"quote":"In addition to the price data, we use bid‑ask spread data to estimate the trading costs"},{"name":"Effective federal funds rate","provider":null,"kind":"macro","access":"public","period":null,"universe":null,"quote":"the effective federal funds rate [Fed23] for short term borrowing and lending"}],"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":"Bid‑ask spread cost = (κspread)T|z|; short‑selling cost = 5% annualized spread over federal funds rate; forecast spread = average of last 5 days; κshort =7.5% annualized, κborrow = federal funds rate.","quote":"For forecasting, we set κshort to 7.5% annualized, and κborrow to the effective federal funds rate.","source":"front_matter","rule":"R3","candidates":{"front_matter":"yes","evidence_card":"yes","evidence_card_quoted":true},"card_detail":"Bid‑ask spread cost = (κspread)T|z|; short‑selling cost = 5% annualized spread over federal funds rate; forecast spread = average of last 5 days; κshort =7.5% annualized, κborrow = federal funds rate.","card_quote":"For forecasting, we set κshort to 7.5% annualized, and κborrow to the effective federal funds rate.","meaning":"costs are part of the reported headline results"},"results":{"status":"available","evaluation_type":"trading-pnl","findings":[{"finding":"Basic Markowitz yields poor performance: 3.7% return, 14.5% volatility, Sharpe 0.19, leverage 9.3, drawdown 78.9%.","kind":"performance","quote":"Basic Markowitz 3.7% 14.5% 0.19 1145.2 9.3 78.9%"},{"finding":"Tuned Markowitz++ further improves Sharpe to 4.65 with 41.8% return and lower turnover.","kind":"performance","quote":"Tuned Markowitz++ 41.8% 8.8% 4.65 38.6 1.6 6.4%"},{"finding":"Adding simple weight limits (10% long, –5% short) raises Sharpe to 1.69 compared to basic Markowitz.","kind":"performance","quote":"Weight‑limited 20.2% 11.5% 1.69 638.4 5.1 30.0%"},{"finding":"Parameter tuning converges after 45 back‑tests to specific γ values, increasing Sharpe while keeping leverage, turnover, and volatility reasonable.","kind":"performance","quote":"This tuning process converged after 45 back‑tests to the parameter values γrisk = 4 × 10−2, γhold = 0.64, γtrade = 0.64, γlev = 5 × 10−4, γturn = 1.6 × 10−3."}],"reported_performance":null,"out_of_sample":{"method":"unclear","headline_out_of_sample":null}},"independent_evidence":{"status":"available","supports":[],"contradicts":[],"qualifies":[],"compared_with":16,"citations":{"cited_by_in_corpus":10,"by_independent_authors":7,"how_cited":{"used_as_baseline":0,"uses_method_or_data":1,"extends":0,"discusses_findings":0,"mentions":1,"unclassified":8},"independent_papers":[{"paper":{"paper_id":"2407.17645","title":"Hopfield Networks for Asset Allocation","paper_date":"2024-07-24","authors":["Carlo Nicolini","Monisha Gopalan","Jacopo Staiano","Bruno Lepri"],"empirical_rigor":6,"page_url":"https://thequant.space/flowcharts/hopfield-networks-for-asset-allocation/"},"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":true,"code_url":"https://github.com/cvxgrp/markowitz-reference","note":null},"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":"2026-11-14T18:16:37Z"},"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/"}