Bloomberg Professional Services
- We find evidence that companies with higher Bloomberg ESG Scores outperformed those with lower ones in terms of total and risk-adjusted returns.
- ESG Scores may function better as return-predicting signals when underpinned by more quantitative information. Accordingly, stewardship teams should engage with their portfolio companies to encourage more quantitative ESG data disclosure.
- The observed outperformance of companies with higher ESG Scores cannot be attributed solely to incidental exposure to traditional risk factors. This suggests the Scores could contain overlooked or underpriced information.
This article was written by Zarvan Khambatta, Head of Sustainable Investments Quantitative Research, Michael Zhang, and Didier Darricau, Senior Sustainable Investments Quantitative Researchers at Bloomberg.
Investors pay attention to non-financial company data, in particular, environmental, social and governance (“ESG”) data and scores, for a variety of reasons. Regardless of their specific motivations, investors are likely to be interested in ESG scores if they can be shown to have a meaningful impact on portfolio returns and risks. Here, we present the findings of our analysis of the effectiveness of using Bloomberg’s ESG Scores (the “Scores”) as a signal that could help differentiate companies’ future stock returns.
PRODUCT MENTIONS
Using the Bloomberg World Large & Mid Cap Index (WORLD Index) and Bloomberg World Large, Mid & Small Cap Index (WLS Index) as examples, we see evidence of companies with higher Bloomberg ESG Scores earning higher returns (as well as higher risk-adjusted returns) than companies with lower ESG Scores.
These excess returns are not entirely accounted for by exposure to known risk factors, such as country, industry or styles (e.g., value, quality), and could point to the Scores containing information that is not captured by traditional factor models.
Nevertheless, the overall evidence is mixed and warrants further analysis. We observe variations in the relationship between ESG Scores and returns across regions, sectors and market capitalizations. While outside the scope of the current analysis, there is also variability in the individual issues that have demonstrated returns prediction or drawdown mitigation potential across different sectors. Each of these elements highlight interesting areas for further research.
Examining performance
We begin by analyzing a universe of companies in a global, broadly diversified equity benchmark index. The Bloomberg World Large & Mid Cap Index (“WORLD Index <Go>” on the Bloomberg Terminal) is a float market-cap-weighted equity benchmark that covers the top 85% of market cap of the measured market. Bloomberg’s ESG Scores cover 96% of the WORLD Index’s constituents and 99% of its market value as of December 2024. In this analysis, we excluded companies in the WORLD Index that were not in the Scores universe.
Each quarter, we sort the companies in the WORLD Index into quintiles based on their Bloomberg ESG Zero-Centered Scores (“ZCS”). ZCS provide a cross-peer group comparable version of the Bloomberg Environmental, Social and Governance (ESG) Score. ZCS range from -10 to 8.5 (higher is better) and represent the difference between a company’s ESG Score and its peer group’s median ESG Score (floored at 1.5). Thus, the median company in a peer group has a ZCS of 0, outperforming companies have ZCS greater than 0 and underperforming companies have ZCS less than 0. Any two companies, from any peer groups, that have the same ZCS could be considered to be out-/under-performing their specific peer group medians by the same margin.
The results in Figures 1a and 1b show that companies in the top ESG quintile generated higher returns and realized higher Sharpe ratios compared to companies in the bottom quintile. This was true for both equal-weighted and market value-weighted quintile portfolios, indicating that the result is not entirely attributable to a “size” effect.
This finding is encouraging, but it should be noted that the differences between the top and bottom quintiles’ Sharpe ratios do not meet the standard of statistical significance. This is because the differences are not extremely large and the history of the back-test is somewhat limited. While random variation cannot be ruled out as the reason for the outperformance we see here, the result and Scores are worthy of further investigation.
Data disclosure and ESG Score effectiveness
We now expand our study universe to an index that also includes small-cap stocks. The Bloomberg World Large, Mid & Small Cap Index (“WLS Index <Go>” on the Bloomberg Terminal) is a float market-cap-weighted equity benchmark that covers the top 99% of market cap of the measured market. Bloomberg’s ESG Scores cover 81% of the WLS Index’s constituents and 96% of its market value as of December 2024. In this analysis, we excluded companies in the WLS Index that were not in the Scores universe.
Here, the results between equally-weighted and market value-weighted quintile portfolios were not consistent. Results are presented in Figures 2a and 2b. The market value-weighted portfolios had similar results to those in Figure 1b—the top quintile outperformed the bottom quintile. This is unsurprising since the smallest companies do not have much influence in market value-weighted portfolios. However, the equal-weighted quintile portfolios indicated that the Scores were not capable of accurately predicting companies’ future returns.
We hypothesize that a possible reason for the observed discrepancy between the equal-weighted results in the WORLD and WLS indices is the addition of several thousand smaller stocks in the WLS Index and the known relationship between company size and disclosure of ESG data.
The effectiveness of the Scores as investment signals may depend on the amount of quantitative data feeding into them. Since smaller companies have, in aggregate, tended to disclose much less information than their larger counterparts, it is conceivable that the Scores of companies with low disclosure do not have the same predictive power with respect to future returns as Scores of companies with high levels of disclosure.
To explore whether the quantity of disclosed data influences the Scores’ ability to differentiate future stock returns, we partition the companies in the WLS Index into four tiers of ESG data disclosure levels. Within each of these four tiers, we form quintile portfolios based on companies’ ZCS, as before. The set criteria for levels of disclosure and the resulting counts of securities in each tier are presented in Figure 3.
Evaluating performance across the four disclosure tiers, we see evidence of a positive link between ESG Scores and stock returns at High and Average levels of quantitative data disclosure. However, when quantitative disclosure is Low or non-existent for the Environmental and Social pillars, we do not see such a pattern. These results, shown in Figures 4a and 4b, support the hypothesis that the value of ESG Scores in predicting returns is influenced by the amount of quantitative data disclosure supporting the Scores.
Repeating the analysis from Figures 1 and 2, but this time filtering only for those companies with High or Average levels of quantitative data disclosure shows improvement in returns prediction in both the WORLD Index and WLS Index universes. The results, shown in Figures 5 and 6, reinforce our belief that the ESG Scores are more effective when they are informed by more data disclosure.
When filtered for data disclosure that is High or Average, companies in the top quintile outperformed companies in the bottom quintile in both the WORLD and WLS indices. To further investigate whether this outperformance can be attributed to information contained in the ESG Scores, and not simply to incidental exposure to other risk factors, we conducted a return attribution analysis using Bloomberg’s PORT MAC3 equity risk model.
To do this, we form equal-weighted long-short portfolios for both the WORLD Index and WLS Index universes. These portfolios are long the companies in the top quintile and short the companies in the bottom quintile of Figures 5a and 6a, respectively. As shown in Figures 7a and 7b, a substantial portion of the long-short portfolios’ returns is not “explained” by exposures to factors such as Industry, Country, Currency or Equity Style (e.g., value, quality).
This unexplained share is termed the “Selection Effect”. In our back-test, the Selection Effect accounted for 38.3% out of 64.7% of cumulative long-short portfolio returns in the WORLD (High and Average disclosure) universe and 9.3% out of 23.7% of cumulative total returns within the WLS disclosure-filtered universe.
Note that this return attribution is based on monthly down-sampled risk exposures from Bloomberg’s PORT MAC3 equity risk model, that are produced at a daily frequency. As a result, the attribution results shown here are approximations and may not exactly match analyses performed in PORT <GO>.
Conclusion
We see evidence of outperformance resulting from the use of the Bloomberg ESG Scores to construct portfolios. This effect is more pronounced when the Scores are informed by higher levels of quantitative data disclosure. Thus, investors who believe they can benefit from these signals should encourage their portfolio companies to increase quantitative data disclosure.
Investors, whether explicitly focused on ESG criteria or not, may benefit from studying and incorporating these signals into their investment processes. Furthermore, the performance differentials are not fully explained by traditional risk factors, suggesting that the Scores may contain under-utilized information.
Bloomberg Terminal subscribers can access Bloomberg ESG Scores at ESG SCORE <GO>. The Scores’ methodologies can be found at BESG <GO> on the Bloomberg ESG Scores tab. For more information click here or request a demo.