Paper: arXiv 2611.20911
Authors: Alejandro Rodriguez Dominguez
Abstract
We analyze the transport of portfolio exposures between moving subspaces under a varying positive definite risk metric. The reference evolution is the induced connection obtained from the ambient connection of minimum norm in fixed asset coordinates. An explicit expansion of risk weighted Procrustes identifies a leading local error that couples metric deformation with covariant motion normal to the subspace. A rank two construction with an exact continuous solution shows that this term can reduce global accuracy to first order even when the metric commutes with its derivative. Geometric mean alignment and a symmetric Procrustes construction cancel the term and yield reversible updates with second order global accuracy. We establish uniform error bounds on nonuniform partitions, characterize the cases in which the Procrustes defect vanishes, and distinguish discretization error from errors in estimated geometry. We also characterize the connection induced by principal square root whitening and its relation to the selected transport. Manufactured examples confirm the predicted defect and convergence rates. A historical geometry stress test then estimates the defect from preceding observations and finds that it predicts the subsequent separation between risk weighted Procrustes and both second order updates, while the two second order schemes remain much closer to each other. Portfolio comparisons separately evaluate target persistence, turnover, and economic implementation.
Complexity vs Empirical Score
- Math Complexity: 9.0/10
- Empirical Rigor: 7.0/10
- Quadrant: Holy Grail — high math complexity, high empirical rigor
Why this score: This paper presents a highly mathematical and theoretical framework for portfolio subspace transport under varying risk metrics, demonstrating significant mathematical complexity. It also includes empirical validation through manufactured examples and historical stress tests, indicating a good level of empirical rigor. The proposed second-order schemes offer a novel improvement over existing methods.
Research Flowchart
flowchart TD
A["Research Goal: Analyze transport of portfolio exposures between moving subspaces under varying risk metric"] --> B{"Key Methodology: Second order transport schemes, Procrustes analysis, Geometric mean alignment"};
B --> C["Data/Inputs: Varying positive definite risk metric, Asset coordinates, Historical geometry observations"];
C --> D{"Computational Processes: Risk weighted Procrustes expansion, Rank two construction, Uniform error bounds, Principal square root whitening"};
D --> E["Key Findings: Leading local error couples metric deformation with covariant motion; Geometric mean alignment & symmetric Procrustes yield second order accuracy; Predicted defect and convergence rates confirmed; Historical stress test validates defect prediction"];
E --> F["Outcomes: Improved schemes for portfolio updates (target persistence, turnover, economic implementation)"];
Evidence Card
What the authors claim, on what data, and who has checked it. Read from the paper's own text by a language model: every quote is verbatim, nothing was re-run. How to read this card
Proposes
The paper proposes a geometric‑mean alignment and a symmetric Procrustes transport scheme that cancel the leading local defect of risk‑weighted Procrustes, yielding reversible endpoint updates with second‑order global accuracy on non‑uniform partitions.
Source sentence
Geometric mean alignment and a symmetric Procrustes construction cancel the term and yield reversible updates with second order global accuracy. It has local error O(h3) and global error O(h2 max) for the selected connection. Proof.
Assumes 3
- Smooth paths of fixed rank and uniformly positive risk metrics are assumed for the convergence analysis. model
Source sentence
The convergence analysis assumes smooth paths of fixed rank and uniformly positive metrics, with spectral separation required when an eigenspace perturbation bound is used.
- Spectral separation is required when an eigenspace perturbation bound is used. statistical
Source sentence
The convergence analysis assumes smooth paths of fixed rank and uniformly positive metrics, with spectral separation required when an eigenspace perturbation bound is used.
- The risk metric is positive definite and may commute with its derivative in certain configurations. model
Source sentence
A rank two construction with an exact continuous solution shows that this term can reduce global accuracy to first order even when the metric commutes with its derivative. The analysis also characterizes the constant metric, fixed subspace, and other configurations in which the defect vanishes and risk weighted Procrustes retains second order accuracy.
Data
equities · daily
- 48 industry portfolios market-prices public
Source sentence
We therefore add a historical geometry diagnostic using the 48 industry panel and the same Ledoit–Wolf metric and rank five representation used in the portfolio experiment. The first panel contains 48 industry portfolios and the second contains 100 portfolios formed at the intersections of ten size groups and ten book to market groups.
- 100 size‑value portfolios market-prices public
Source sentence
The first panel contains 48 industry portfolios and the second contains 100 portfolios formed at the intersections of ten size groups and ten book to market groups. The candidate sample is 2 January 1990 through 30 April 2026.
- Manufactured examples synthetic unknown
Source sentence
Manufactured examples confirm the predicted defect and convergence rates. Noncommuting metric path and convergence order The second manufactured problem uses six assets and a rank three representation.
Costs
Modeled
Proportional transaction costs are applied; turnover is reported annualized in the primary results table.
Source sentence
Signals, constraints, costs, and persistence settings are common to all methods, as specified in Section C. Table 4: Primary results after proportional costs.
Finds 5
- The symmetric Procrustes update has local error O(h^3) and global error O(h^2) for the selected connection, confirming second‑order accuracy on non‑uniform partitions. theoretical
Source sentence
It has local error O(h3) and global error O(h2 max) for the selected connection. Proof.
- The historical geometry diagnostic predicts the subsequent separation between risk‑weighted Procrustes and both second‑order updates, while the geometric‑mean and symmetric constructions remain much closer to each other. predictive-accuracy
Source sentence
A historical geometry stress test then estimates the defect from preceding observations and finds that it predicts the subsequent separation between risk weighted Procrustes and both second order updates, while the two second order schemes remain much closer to each other. The historical geometry diagnostic links the theoretical defect to estimated market geometry without using portfolio performance: the ex ante
- Geometric‑mean transport reduces annualized turnover by 48.06% in the industry panel and 42.89% in the 100‑portfolio panel relative to independent local refitting. performance
Source sentence
TO local TO geo. ∆CE, bp [95%] 48 industries 8.458% 8.364% 2.997 1.557 −9.37 [−95.93, 81.48] 100 size–value 6.781% 5.609% 4.654 2.658 −117.20 [−279.61, −1.64] At the primary persistence weight, geometric mean transport reduces annualized turnover by 48.06% in the industry panel and 42.89% in the 100 portfolio panel relative to independent local refitting, with paired intervals excluding zero in both cases.
- Annual certainty‑equivalent differences relative to local refitting are –9.37 bp (industry) and –117.20 bp (100‑portfolio), with the industry interval spanning zero and the 100‑portfolio interval below zero. performance figure from elsewhere in the paper
Source sentence
The corresponding annual certainty equivalent differences are −9.37 and −117.20 basis points.
- Both metric‑based second‑order updates (geometric‑mean and symmetric Procrustes) have smaller discrepancies from the composed reference than ambient projection, and remain substantially closer to each other than to risk‑weighted Procrustes. theoretical
Source sentence
The two second order rules remain substantially closer to each other than either is to risk weighted Procrustes. 4.6. Both metric based updates have smaller discrepancies from the composed reference than ambient projection, with the geometric mean and risk weighted Procrustes results closer to each other.
Checked by others
No other paper in the archive has been found to support or contradict these findings (compared with 17 related papers). Coverage is partial, so this is neither replication nor refutation.
Code
No repository by the authors found in the paper.
Extracted 2026-11-17 · same card as JSON · A language model chose these items: check the source sentence before relying on one.
Cited References (32)
- Dynamic Causal Portfolio Choice: Hedging the Rotation of the Common-Driver Manifold (2026) — link
- The parallel-transported (quasi)-diabatic basis. (2022) — link
- Approximating the parallel transport of an induced connection (2022) — link
- A Grassmann manifold handbook: basic geometry and computational aspects (2020) — link
- Positive Definite Matrices (2019) — link
- Is Affine-Invariance Well Defined on SPD Matrices? A Principled Continuum of Metrics (2019) — link
- Quantum mechanics in an evolving Hilbert space (2016) — link
- Dynamic Trading with Predictable Returns and Transaction Costs (2013) — link
and 24 more
- Large Covariance Estimation by Thresholding Principal Orthogonal Complements (2011) — link
- A Generalized Approach to Portfolio Optimization: Improving Performance by Constraining Portfolio Norms (2009) — link
- Functions of matrices - theory and computation (2008) — link
- Sparse and stable Markowitz portfolios (2007) — link
- Geometric Means in a Novel Vector Space Structure on Symmetric Positive-Definite Matrices (2007) — link
- Procrustes Problems (2005) — link
- A Differential Geometric Approach to the Geometric Mean of Symmetric Positive-Definite Matrices (2005) — link
- A well-conditioned estimator for large-dimensional covariance matrices (2004) — link
- Risk Reduction in Large Portfolios: Why Imposing the Wrong Constraints Helps (2003) — link
- The Geometry of Algorithms with Orthogonality Constraints (1998) — link
- Common risk factors in the returns on stocks and bonds (1993) — link
- Computing the polar decomposition with applications (1986) — link
- Numerical methods for computing angles between linear subspaces (1971) — link
- Portfolio Selection (1971) — link
- The Rotation of Eigenvectors by a Perturbation. III (1970) — link
- A generalized solution of the orthogonal procrustes problem (1966) — link
- Data Library: U.S. research returns, 2026, https://mba.tuck.dartmouth.edu/ pages/faculty/ken.french/data_library.html , consulted November 14, 2026 (2026)
- Optimization algorithms on matrix manifolds (2009)
- Geometric Numerical Integration , 2nd ed. (2006)
- A Riemannian Framework for Tensor Computing (2005) — link
- Matrix Perturbation Theory (1991) — link
- Optimum Consumption and Portfolio Rules in a Continuous-Time Model* (1975) — link
- Transport under a Varying Risk Metric
- Second-order steps
0 of 32 cited works are in the corpus; resolved via arXiv source bundles, Semantic Scholar and OpenAlex. Corpus-wide: 2596 linked papers.