Introduction

Information intermediaries play a critical role in efficient capital allocation by reducing information asymmetries between corporate insiders and outside investors. This research examines how the compensation structure of ESG rating agencies affects the quality of the information they produce. Specifically, it investigates whether intermediaries paid by investors produce different, and potentially higher quality, information than those paid by the subjects of their analysis. While investor-paid models are often viewed as superior due to increased timeliness and frequency of updates, this paper identifies a distinct conflict of interest: the pressure to favour stability over timeliness. Because ESG ratings directly govern index construction, investment mandates, and exclusion screens, rating changes impose concrete portfolio-rebalancing costs on subscribers, creating a demand for rating stability.

A stylised model of investor–pay model incentives

The report develops a theoretical framework to clarify how conflicts of interest arise in an investor-pays ratings model. This “toy model” isolates two distinct and coexisting mechanisms: an incentive for peer alignment and a stability bias. Investors use cross-rater agreement as a heuristic for reliability and treat divergence as a signal of model risk. This creates an incentive for raters to adjust published scores toward peer consensus, sacrificing methodological discipline to enhance perceived informativeness. Additionally, because investors value the continuity of published ratings, any change in a published score can reduce subscriber demand. The interaction of these behavioural channels creates a “traffic light effect” where rating changes are delayed or suppressed.

Institutional background on MSCI ESG Ratings

The empirical analysis focuses on MSCI ESG Ratings, which is characterised as a highly influential commercial rater. MSCI uses a seven-point letter scale from CCC to AAA. The process involves scoring 35 “Key Issues” to obtain a Weighted Average Score (WAS), which is then transformed into an Industry Adjusted Score (IAS) through a benchmarking procedure. This step rescales each firm’s WAS using the cross-sectional distribution of WAS among its industry peers. Crucially, before releasing updated scores, MSCI relies on a committee review process to vet and, if necessary, override model-implied outcomes. This discretionary review is triggered by events such as upgrades to AAA, downgrades to CCC, or changes of two or more notches.

Descriptive evidence of the traffic light effect

The research documents a striking pattern where companies systematically bunch just below or above the thresholds separating adjacent letter ratings. This bunching is difficult to reconcile with a purely mechanical benchmarking procedure and suggests active intervention by committees to limit the extent to which scores cross rating boundaries. Only 24.3% of the 129,702 firm-year observations undergoing a full reassessment experienced any letter rating change. The distribution of IAS over time shows sharp peaks near thresholds, indicating that rating mobility is strongly constrained. This suggests that committee intervention internalises the consequences of rating changes for MSCI’s own ESG index family, where a one-notch move can trigger discrete changes in index eligibility or portfolio weight.

Peer alignment and Refinitiv methodology change

To identify committee adjustments, the researchers reverse-engineer MSCI’s benchmarking methodology to recover methodology-implied scores. They find that the “wedge” between published and implied scores is strongly correlated with divergence from the peer consensus. This relationship is intensified when other raters broadly agree (low dispersion). To sharpen identification, the study exploits a 2020 methodology overhaul by Refinitiv, which was applied retroactively. The results show that MSCI’s adjustments co-move with the legacy ratings that were actually available at the time, rather than the backfilled series that did not exist when decisions were made. This confirms that MSCI conditions its overrides on contemporaneously available peer assessments rather than common fundamentals.

Financial impact

The study examines the financial consequences of this effect by analysing stock returns following downgrade signals. It compares implemented downgrades with those that were suggested by the methodology but suppressed in the published letter grade. Implemented downgrades predict significant negative buy-and-hold returns of approximately 2.8% over the subsequent twelve months. In contrast, suppressed downgrades show no statistically detectable return response at any horizon. This asymmetry suggests that suppressed downgrades contain value-relevant ESG risk information that the market does not fully incorporate because it is hidden from the public. The decision to suppress effectively insulates firms from market consequences, with the cumulative return gap reaching 3 to 4 percentage points in the first quarter following the event.

Conclusions

The paper concludes that the optimality of investor-pays compensation structures is more context-dependent than previously suggested. When ratings are embedded in portfolio rules, subscriber demand for stability can induce a bias that suppresses private, value-relevant information. Published ESG ratings often reflect a compromise between internal methodology and peer assessments. Furthermore, the concentration of adjustments near boundaries implies that ESG index composition is partly shaped by committee discretion. Researchers are cautioned against treating ESG ratings as quasi-random near categorical thresholds in regression-discontinuity designs, as the realised IAS is an endogenous variable rather than a mechanical output.