S&P 500 firms that publicly pursue more and broader AI initiatives enjoy higher market valuations and improved operational efficiency; board power and involvement alter these benefits—concentrated power boosts valuation gains but can erode operational efficiency, while active boards amplify returns to diversified AI strategies.
Competing With Artificial Intelligence: Board Governance And Competitive Ai Actions
Using NLP on S&P 500 press releases (2010–2022), the paper finds firms that pursue more and more diverse AI-related competitive actions exhibit higher market valuations and better operational efficiency, with board governance shaping these associations.
Artificial intelligence (AI), marked by autonomy, learning, and algorithmic opacity, is increasingly central to firm strategy, yet firms differ widely in their ability to realize sustained performance gains.We conceptualize AI adoption as a set of AI competitive actions, including AI-related R&D, acquisitions, partnerships, product launches, and signaling activities.Drawing on the attention-based view (ABV), we examine how these actions influence firm performance and how board governance shapes their effectiveness.Using Natural Language Processing (NLP) to identify AI competitive actions from press releases of S&P 500 firms (2010)(2011)(2012)(2013)(2014)(2015)(2016)(2017)(2018)(2019)(2020)(2021)(2022), we find that firms engaging in more AI actions and a broader portfolio of actions achieve higher market valuation and operational efficiency.Governance conditions these effects: power disparity strengthens market valuation but weakens operational efficiency, while board involvement amplifies the benefits of diversified AI strategies.These findings highlight the importance of governance in shaping strategic attention toward AI initiatives.
Summary
Main Finding
Firms that undertake more AI-related competitive actions and a broader portfolio of such actions achieve higher market valuation (Tobin’s Q) and better operational efficiency (gross profit margin). Board governance conditions these benefits: concentrated power (power disparity) amplifies valuation effects but undermines operational efficiency, whereas active board involvement strengthens both valuation and operational gains from diversified AI strategies.
Key Points
- Conceptual framing: AI adoption is treated as a set of observable competitive actions (not a single homogeneous capability). AI’s distinct features — autonomy, learning, opacity — prompt firms to experiment across multiple initiatives.
- AI competitive action types (identified from press releases): AI R&D, AI acquisitions, AI partnerships, AI marketing/exposure, AI product launches (six categories in total).
- Scope measure: number of distinct AI action categories a firm undertakes in a year; used as a measure of strategy diversity.
- Main hypotheses supported:
- H1: More AI competitive actions → better firm performance.
- H2: Greater diversity (Scope) of AI actions → better performance.
- H3: Power disparity moderates effects (strengthens valuation, weakens operational efficiency).
- H4: Board involvement strengthens AI actions → performance links.
- Theoretical anchor: Attention-Based View (ABV) — boards shape managerial attention and resource allocation to complex, uncertain AI initiatives.
Data & Methods
- Sample: S&P 500 firms, 2010–2022.
- Data sources: press releases from Factiva (for AI actions via NLP), BoardEx (governance measures), Compustat (financials).
- NLP/detection: dictionary-based keyword search plus classification to identify AI-related press releases and classify them into six action types.
- Governance measures:
- Power disparity: concentration of decision authority among top executives.
- Board involvement: extent of board engagement in strategic decision-making.
- Outcome measures:
- Market valuation: Tobin’s Q.
- Operational efficiency: gross profit margin.
- Identification and estimation:
- Propensity score matching (PSM) to mitigate selection bias in AI action adoption.
- Pooled panel regressions with independent variables lagged one year; industry and year fixed effects included.
- Tests of interaction terms between AI action measures and governance variables to assess moderation.
Implications for AI Economics
- Heterogeneity of AI strategy matters: Economics analyses should distinguish types and portfolios of AI initiatives rather than treating AI as a single binary capability.
- Governance as a second-order determinant of AI returns: Board structure and engagement materially shape how firms convert AI experimentation into market and operational value. Models of firm productivity gains from AI should incorporate governance interactions.
- Signaling versus coordination trade-off: Concentrated authority can improve investor signaling (raising valuation) but may harm operational coordination needed to realize efficiency gains — an important nuance for interpreting market-based measures of AI returns.
- Measurement advances: Using observable competitive actions (press releases, acquisitions, product launches) combined with NLP offers a viable proxy for firm-level AI activity; economists can refine such measures for large-sample empirical work.
- Policy and managerial relevance: Firms may boost AI returns by diversifying AI initiatives and increasing meaningful board engagement. Policymakers and investors should consider governance when assessing firm-level prospective gains from AI investment.
- Directions for future research:
- Stronger causal identification (e.g., instrumental variables, difference-in-differences around exogenous shocks).
- Deeper NLP to capture intensity, technical sophistication, or domain of AI actions.
- Heterogeneity analyses across industries, firm size, and AI application types.
- Longer-term dynamics (productivity J-curve, persistence of valuation effects).
Assessment
Paper Typecorrelational
Evidence Strengthlow — The paper documents robust associations between NLP-coded AI actions and firm outcomes over a large panel, but it lacks a clear source of exogenous variation to rule out reverse causality, omitted variable bias, or signaling effects (e.g., firms with good prospects both talk more about AI and perform better). Thus claims about AI causing performance gains are not strongly supported.
Methods Rigormedium — Strengths include a large longitudinal sample (S&P 500 firms, 2010–2022) and use of NLP to systematically identify AI-related actions plus governance interaction analysis; weaknesses include reliance on press-release text (measurement error and signaling), observational identification, and likely limited robustness to unobserved confounders.
SamplePress releases from S&P 500 firms from 2010–2022 coded via NLP for AI-related R&D, acquisitions, partnerships, product launches, and signaling; firm outcomes include market valuation metrics and operational efficiency from financial statements; board governance measures come from corporate governance/board datasets for the same firms and years.
Themesadoption governance
IdentificationAssociational analysis: NLP is used to code AI-related competitive actions from S&P 500 press releases (2010–2022) and those measures are regressed on firm outcomes (market valuation, operational efficiency) with controls and governance interaction terms; no clear exogenous shock, instrument, or difference-in-differences strategy is described, so causal claims rely on observational variation and model controls (likely including fixed effects).
GeneralizabilityOnly large, publicly listed U.S. firms (S&P 500) — results may not generalize to smaller firms or non-U.S. contexts, Press releases capture public signaling and may not reflect actual AI adoption or internal implementation, NLP classification errors and changing language about AI over time may bias measures, Heterogeneous effects across industries may limit cross-sector generalization, Findings for market valuation may reflect investor sentiment rather than real productivity gains
Claims (6)
| Claim | Direction | Confidence | Outcome | Details |
|---|---|---|---|---|
| Firms engaging in more AI actions achieve higher market valuation. Firm Revenue | positive | high | market valuation |
0.3
|
| Firms with a broader portfolio (greater diversity) of AI competitive actions achieve higher market valuation. Firm Revenue | positive | high | market valuation |
0.3
|
| Firms engaging in more AI actions and a broader portfolio of actions achieve higher operational efficiency. Organizational Efficiency | positive | high | operational efficiency |
0.3
|
| Board power disparity strengthens the positive relationship between AI competitive actions and market valuation. Firm Revenue | positive | high | market valuation (conditional on board power disparity) |
0.3
|
| Board power disparity weakens the positive relationship between AI competitive actions and operational efficiency. Organizational Efficiency | negative | high | operational efficiency (conditional on board power disparity) |
0.3
|
| Board involvement amplifies the benefits of diversified AI strategies. Adoption Rate | positive | high | beneficial effects of diversified AI strategies (noted for market valuation and operational efficiency) |
0.3
|