Chinese listed firms mirror their industry when adopting AI: average peer adoption drives firms’ own AI uptake more than a single leading adopter, though CIOs and high-market-position firms respond especially strongly to a peer leader.

Following the Herd or the Bellwether: Peer Effects in Firms’ AI Adoption

Siyu Shao, Jianjun Yang, Ling Zhang · Fetched April 27, 2026 · IEEE transactions on engineering management

semantic_scholar correlational medium evidence 7/10 relevance DOI Source

Industry peers' AI adoption levels — both the average peer group and the top peer — are positively associated with a focal firm's AI adoption, with peer-group influence stronger overall and heterogeneous effects by industry digital maturity, CIO presence, and firm market position.

Imitation and competition among peer firms serve as critical drivers of technology adoption and diffusion. Drawing on competitive dynamics and institutional theory, this study investigates the heterogeneous influence of peer firms on a focal firm’s artificial intelligence (AI) adoption by distinguishing between the peer group (the set of industry peers) and the peer leader (the most advanced AI adopter among industry peers). We use a panel dataset of publicly listed Chinese firms from 2012 to 2023 and estimate fixed-effects regression models to test the proposed hypotheses. The results show that the AI adoption levels of the peer group and peer leader positively influence the focal firm’s AI adoption level, with the influence of the peer group being stronger. Industry digital maturity enhances the impact of the peer group while it weakens the effect of the peer leader. The presence of a Chief Information Officer strengthens the influence of the peer group and the peer leader, with the influence of the peer leader being more pronounced. In addition, firms with a high market position tend to imitate the peer leader, whereas firms in middle and low positions are more likely to follow the peer group. Multiple robustness checks confirm the validity of these findings. This study deepens the understanding of AI adoption antecedents and provides a more fine-grained perspective on peer effects, offering actionable insights for policymakers and engineering managers.

Summary

Main Finding

Both the AI-adoption level of an industry’s peer group and that of its peer leader positively influence a focal firm’s AI adoption, but the peer group’s influence is stronger overall. Industry digital maturity, the presence of a Chief Information Officer (CIO), and a firm’s market position systematically shape which peer effect is dominant.

Key Points

Peer group and peer leader effects:
- Higher average AI adoption among industry peers (peer group) → higher AI adoption at the focal firm.
- Higher AI adoption by the most advanced peer (peer leader) → higher AI adoption at the focal firm.
- The peer group effect is larger in magnitude than the peer leader effect on average.
Moderation by industry digital maturity:
- Greater industry digital maturity strengthens the peer group effect.
- Greater industry digital maturity weakens the peer leader effect.
Moderation by internal governance (CIO presence):
- Having a CIO amplifies both peer group and peer leader influences.
- The amplification is stronger for the peer leader effect.
Heterogeneity by firm market position:
- High-market-position firms tend to follow/imitate the peer leader.
- Middle- and low-market-position firms are more likely to follow the peer group.
Robustness:
- Results hold under multiple robustness checks (alternative specifications and checks reported).

Data & Methods

Data: panel of publicly listed Chinese firms, 2012–2023.
Dependent variable: firm-level AI adoption (as measured in the study).
Key independent variables: industry peer group AI adoption (average among peers) and peer leader AI adoption (maximum among peers).
Moderators: industry digital maturity, CIO presence (binary), firm market position (high/middle/low).
Estimation: fixed-effects regression models (firm and/or time fixed effects) to control for unobserved heterogeneity.
Validation: multiple robustness checks (alternative specifications and checks described in the paper).

Implications for AI Economics

Diffusion models should distinguish between peer-group effects and leader-driven effects. Aggregate or symmetric peer specifications can miss important heterogeneity in how adoption spreads.
Market- and institution-level context matter:
- In digitally mature industries, conforming to peer norms and average behavior is a stronger driver of adoption than following frontier leaders.
- In less mature industries, signals from prominent leader firms carry more weight—policy interventions that highlight leader successes can therefore be effective early on.
Firm capabilities and governance (e.g., CIOs) are complementary to external peer influence; internal capabilities enable better uptake of peer information and leader signals. Models of technology adoption should include capability heterogeneity and complementarities.
Status and resource constraints shape imitation strategy:
- Leading firms imitate frontier peers to protect/advance competitive position; lower-status firms rely on broader peer behavior, suggesting limited capacity to interpret or implement leader-specific innovations.
Policy and managerial actions:
- Policymakers: support benchmarks and demonstration projects by leaders in less mature sectors; invest in industry digital infrastructure to accelerate group-driven diffusion where appropriate.
- Managers: appoint or empower CIOs to better translate peer signals into adoption; tailor competitive/imitative strategies to your market position (e.g., high-status firms should monitor leaders closely; smaller firms should track peer averages and practical best practices).
Future research directions: causal identification of mechanisms (learning vs. competition vs. institutional pressure), performance consequences of different imitation strategies, and external validity beyond publicly listed Chinese firms.

Assessment

Paper Typecorrelational Evidence Strengthmedium — Findings are supported by panel fixed-effects models and robustness checks, which reduce some confounding from time-invariant heterogeneity, but the study lacks a clear exogenous source of variation (e.g., IV, natural experiment, or staggered adoption design) to fully rule out simultaneity, reflection problems, or common industry shocks driving both peers and focal firms. Methods Rigormedium — Appropriate use of longitudinal firm-level data and fixed effects strengthens internal controls; heterogeneity and robustness analyses add credibility; however, identification remains observational without stronger causal tools (instruments, quasi-experimental variation, or explicit strategies addressing peer-effect reflection/selection), and measurement details of the AI-adoption variable are not described in the summary. SamplePanel of publicly listed Chinese firms observed annually from 2012 to 2023; firm-year observations with a firm-level AI adoption measure (paper constructs an AI adoption index and defines peer group as industry peers and peer leader as the most advanced adopter within the industry); covariates include firm characteristics, industry digital maturity, and an indicator for presence of a Chief Information Officer; sample restricted to listed firms. Themesadoption org_design innovation IdentificationFirm-level panel two-way fixed-effects regression using within-firm variation over 2012–2023, industry-level peer definitions (peer group average and single peer leader) as key regressors, controls for observable firm and industry covariates, heterogeneity analyses (industry digital maturity, CIO presence, market position) and multiple robustness checks; no exogenous shock, instrumental variable, or natural experiment reported in the summary. GeneralizabilityRestricted to publicly listed Chinese firms — may not generalize to private firms or SMEs, China-specific institutional, market, and regulatory context may limit applicability to other countries, Industry-level peer definitions may miss firm-specific network ties or geographic spillovers, Findings conditional on the 2012–2023 period; diffusion dynamics could differ in other time windows, Dependent on how AI adoption is measured (disclosure/patent/investment proxies), which may not capture all adoption modalities

Claims (8)

Claim	Direction	Confidence	Outcome	Details
The AI adoption levels of the peer group positively influence the focal firm’s AI adoption level. Adoption Rate	positive	high	focal firm AI adoption level	0.3
The AI adoption level of the peer leader (the most advanced AI adopter among industry peers) positively influences the focal firm’s AI adoption level. Adoption Rate	positive	high	focal firm AI adoption level	0.3
The influence of the peer group on a focal firm’s AI adoption is stronger than the influence of the peer leader. Adoption Rate	positive	high	focal firm AI adoption level (relative effect sizes of peer group vs. peer leader)	0.3
Industry digital maturity enhances (strengthens) the impact of the peer group on a focal firm’s AI adoption. Adoption Rate	positive	high	focal firm AI adoption level (moderated by industry digital maturity for peer group effect)	0.3
Industry digital maturity weakens the effect of the peer leader on a focal firm’s AI adoption. Adoption Rate	negative	high	focal firm AI adoption level (moderated by industry digital maturity for peer leader effect)	0.3
The presence of a Chief Information Officer (CIO) strengthens the influence of both the peer group and the peer leader on a focal firm’s AI adoption, with the influence of the peer leader being more pronounced when a CIO is present. Adoption Rate	positive	high	focal firm AI adoption level (moderated by presence of CIO for peer group and peer leader effects)	0.3
Firms with a high market position tend to imitate the peer leader, whereas firms in middle and low market positions are more likely to follow the peer group. Adoption Rate	mixed	high	focal firm AI adoption level (differential peer influence by firm market position)	0.3
The main findings are robust to multiple robustness checks. Adoption Rate	null_result	high	robustness of reported peer effect findings	0.3