The Topography of Capital: Regime Structure in US Semiconductor Firms 2010--2024
Summary
No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.
Summarise with VIRAT Free to try with your Science Catalogs account.Abstract
We apply a nonparametric drift field estimation framework to the financial state space of 122 US semiconductor firms (SIC 3674) over 2010–2024, using 806 firm-year observations derived from SEC DERA annual filings (658 consecutive annual transitions). From a universe of 7,194 tags surviving numeric validation, a data-driven coverage sieve selects 12 core reporting tags — an income-statement and cash-flow fingerprint of the sector — confirmed identical across coverage thresholds of 70%, 75%, and 80%. These 12 tags generate 66 pairwise log-relative features; principal component analysis reduces the space to six dimensions capturing 79.2% of total variance. Nadaraya-Watson kernel regression estimates the drift field μ̂(x) = E[X_{t+1} − X_t | X_t = x] at the Silverman bandwidth h = 0.5226. Core findings: All six dimensions exhibit mean reversion toward stable marginal attractors (drift-state correlations r ∈ [−0.87, −0.61]), confirmed coordinate-system invariant in raw interpretable ratios without PCA (6/6, r ∈ [−0.95, −0.76]) Six stable marginal equilibrium attractors confirmed by 10,000 bootstrap replicates; attractor locations in two dimensions — receivables cycle (Z2, CI width 0.438) and operating cycle (Z5, CI width 0.507) — are statistically precise and stable across a fourfold bandwidth range Capital structure (Z1, CI width 3.557) exhibits a diffuse flat-band attractor with no stably identified location Three-regime firm classification by equilibrium zone visitation: Peripheral (60.5%), Transit (35.1%), Cycling (4.4%); temporally stable across 2010–2016 vs. 2017–2023 (shifts < 2.5 pp); externally validated by observation span after circularity correction (H = 8.205, p = 0.017) Limitations explicitly stated: Findings apply to the 122 reporting-complete firms (49% of 249 SIC 3674 firms) and cannot be generalized to the full sector. The Cycling regime contains five firms; within-regime inference is limited. Regime proportions are sensitive to PCA dimensionality (k = 6 vs. k = 8 shifts Peripheral by 16.6 pp). All attractors are marginal, not joint multivariate. Robustness: Coordinate-system invariance (6/6 raw ratios mean-reverting without PCA), bandwidth invariance (fourfold range), and threshold invariance (12-tag core identical at T = 0.70, 0.75, 0.80) are all confirmed. Data pipeline: Full SHA-256 hash chain from raw extraction through final classification. Extraction sealed at hash 5da6ee0a…f611d3. All intermediate files verifiable. This paper is the second empirical application in the Business Dynamic Equilibrium research programme, following the published pharmaceutical application (DOI: 10.5281/zenodo.19274393). The drift-field protocol is documented separately (DOI: 10.5281/zenodo.18935867).
Fields of Study & Subjects
Financial Analysis and Corporate Governance
Working Capital and Financial Performance
Corporate Finance and Governance
Publication details
- Journal / source
- Zenodo (CERN European Organization for Nuclear Research)
- Publication date
- 30 March 2026
- Language
- EN
- Document type
- Preprint