How precisely do published regressions measure the reallocation of green investment?
Citable version that has not completed AIRR screening. Not peer reviewed. Screening record
After an increase in environmental political risk, how much of the estimated fall in green investment is offset by other investment? The main asset-scaled regressions in Gustafson, Lel and Qin imply a point ratio of 28/37. Three control coefficients violate a necessary accounting restriction by more than rounding can explain. Marginal standard errors alone give a conservative asymptotic 95% interval of [0.020, 2.956], rounded outwards. Green capital expenditure subsequently rebounds. Its two horizon coefficients sum to 0.45 per thousand of assets, but either sign of the population sum is compatible with the marginal evidence.
Preview PDF on this page v1
Uses your browser's PDF viewer. Open the PDF directly if a preview is unavailable.
Cite this paper
Lluis Eriksson (2026). How precisely do published regressions measure the reallocation of green investment?. AIRR.SCIENCE, ARR-2026-1HEWDY1GFM84PS5P, v1. https://airr.science/papers/ARR-2026-1HEWDY1GFM84PS5P/versions/v1/ Research preprint; ARR-2026-1HEWDY1GFM84PS5P v1. Not peer reviewed by AIRR. Working paper; not admitted to the AIRR accepted collection.
Use this version link to cite the manuscript you read. A Working paper label means the version has not completed AIRR screening. AIRR screening is not journal peer review. Import BibTeX, RIS or CSL JSON into your reference manager to apply a citation style.
Version history
The paper identifier remains stable. Each version has its own immutable release, timestamp and version identifier.
- v1 · source snapshot available · viewing
Verification record
- Frontier-model screening
- Not assessed
- Source integrity
- Pass
- Bibliographic integrity
- Partial
- Reproducibility
- Partial
- Lean 4
- Not applicable
Recorded under ARR-VERIFY-1.0. AIRR verification and screening are not peer review.
AI assistance statement
OpenAI Codex participated in analysis, writing and the first review of the exact PDF. Its shared-context assessment is 3.60/10. Report SHA-256 e496a2e920a752aec07d1868ee8182b74f1dcd1711a3c53eb2166025384802c7. The author administers AIRR. Independent peer review has not been documented.
Frontier-model screening
Status: not_assessed. Any listed reports correspond to this exact version under ARR-SCREEN-1.0; no absent assessment is represented as a pass.
Model assessments and review context
- Distinct models
- 1
- Range
- 3.60–3.60
- Method
- bounded capability-weighted mean · exact version only
Limited evidence · 0/1 benchmark matches. This is an evidence description, not a probability of correctness.
Only one distinct model; One provider; shared blind spots possible; Fresh blinded review context not documented for every model; Mathematical benchmark not applied to this subject; Adverse reports or unresolved objections require editorial disposition
OpenAI · gpt-6.2-astra · involved in manuscript · review context: shared3.60
Recommendation: Minor Revision · Material objections: 0
Working-paper assessment of the exact three-page PDF. The baseline ratio and two-horizon CAPEX contrast answer distinct questions. Three control discrepancies exceed 0.015. In the dynamic table the two component discrepancies are -0.20 and +0.06. Allowing unknown covariance admits full baseline offset and either sign for the sum of CAPEX coefficients. I participated in analysis and drafting in the same chat. This is an involved-model assessment, without independent replication of the underlying economic data.
- Correctness confidence
- The signed discrepancy and three-number rounding bound agree with the primary tables. The marginal-error test inverts a bound for each candidate ratio and admits r=1. Two separate regressions at distinct horizons support the stated limited interpretation.
- Rigor
- The paper quantifies worst-case covariance and rounding separately. The normal approximation and common-operator assumption are explicit. No cluster-level sampling data were used to test finite-sample coverage.
- Novelty
- The regression consistency diagnostic and CAPEX-specific implications are a distinct application to a named source. Fieller inversion and the worst-case variance bound have explicit antecedents.
- Significance
- Readers gain a reproducible reason to distinguish the reported direction of reallocation from the share recovered. The analysis addresses one empirical paper and does not change a policy estimate using microdata.
- Reproducibility
- Two builds matched byte for byte and all three final pages were read. Numerical results from the pinned scripts matched the archived results in the PDF build. Table extraction depends on publisher layout and the original underlying firm-level regressions were not re-estimated.
Strengths
- Explicit falsifiable restriction on all common-design regression coefficients; the headline regressor alone passes it.
- Two different normalizations and the timing of controls are central to the interpretation and correctly distinguished.
Weaknesses
- A single-paper audit does not establish how often reporting differences of this size occur or which data transformation caused them.
- A future revision would benefit from reporting the numerical cross-horizon correlation threshold and a cluster-count sensitivity beside the asymptotic interval.
Potential errors
- The term full compensation could be read as a causal project-level quantity if the abstract is read in isolation. Sections 2 and 4 limit it to a coefficient comparison.
- Worst-case covariance sets admit singular joint limits. The actual panel structure may restrict this set and tighten inference; the reported intervals are conservative.
Strong novelty candidates
- The linked baseline and two-horizon additivity failures and the reduced-form inference obtained when covariance cannot be reconstructed from the total.
Unresolved material objections
None reported.
arr:assessment:0e7aa79b-07da-4d5e-9642-aa5c1c6a1c83 · prompt ARR-ASSESS-1.0 · response SHA-256 27a3e7c5fccb48e2f7453a8ccb00ba258cbb536c403702ff3011f85df0ff7b19 · canonical PDF SHA-256 d06f8b38dac08a92a883b7e123634dc43e0dc2c6f33dd22c29f79d1dacf09542
A model assessment is not peer review or a correctness certificate. AIRR preserves disagreement, exact-version provenance and later reassessments.
Editorial disclosure
Working-paper deposit under exact-hash permission. Collection screening and independent peer review remain outstanding.