ESG Signaling on Wall Street in the AI Era

Battle Count: 58

The paper provides portfolio management implications for asset allocation based on ESG score movements (hold, buy, sell, short signals). However, the quantitative trading relevance is limited: (1) the analysis is cross-sectional rather than time-series, (2) the signaling effect is on DTCR (a capital structure variable) rather than direct return prediction, (3) the low R² values suggest limited predictive power for trading signals, (4) no backtesting or trading strategy performance metrics are provided, and (5) the ESG data frequency and availability for real-time trading are not addressed. The paper is more relevant to fundamental portfolio management and credit analysis than to high-frequency or systematic quantitative trading.

The methodology is straightforward: standard OLS cross-sectional regressions with winsorization. Implementation requires: (1) scraping or downloading ESG scores and MVE from Yahoo Finance for S&P 500 firms, (2) computing DTCR from book value of debt and market value of equity, (3) winsorizing variables at 1st and 99th percentiles, (4) running 15 regression specifications (single, double, triple, quadruple variable models with/without MVE control). No specialized software, machine learning frameworks, or complex optimization is needed. Standard statistical packages (R, Python statsmodels, Stata) suffice. The main challenge is data collection and ensuring consistent ESG score definitions.

The methodology is clearly described with explicit regression equations (1)-(8), data source (Yahoo Finance), sample (503 S&P 500 firms as of September 5, 2025), and winsorization procedure. However, no code or replication package is provided. The data is publicly available via Yahoo Finance, but the exact extraction date and methodology for obtaining ESG scores are not fully detailed. Summary statistics and full regression tables are provided, enabling partial replication.