Corporate Governance Ratings Are they predictors of future performance? Considerable debate has addressed the degree of value in the most widely accepted governance ratings products. The SBA reviews the evidence in two recent studies that weigh in on the verdict. While the predictive results of such products are not bright-line tests, we find evidence that certain ratings may contain information that is useful in obtaining economically significant returns. Rating the Ratings: How Good Are Commercial Governance Ratings? by Robert Daines, Ian Gow and David Larcker In this recent work, the authors recognize the increasingly important role that governance and governancerating firms play in the public markets, and here they seek to provide an outside, independent look into whether the ratings produced by these firms allow investors to predict returns or other performance measures, or negative events such as restatements or litigation, in a significant and economically exploitable fashion. They examine ratings from RiskMetrics Group, GovernanceMetrics International (GMI), The Corporate Library, and Audit Integrity. The Ratings For RiskMetrics, the authors choose the industry Corporate Governance Quotient (CGQ) score as the relevant rating to compare with the other service providers. They do not explain the choice of this score over the CGQ Index score, which scores companies against a much larger group of peers across industries. Although the scores are highly correlated, it would seem that the index CGQ score might have been more applicable for comparison with other ratings as each rating benchmarks firms on a more universal standard of governance practices, rather than just within its own industry. The CGQ score also has several sub-score components used in the analysis, including the audit, board, compensation and takeover defense scores. The overall CGQ score attempts to evaluate the strengths, deficiencies and overall quality of a firm s governance practices and provides a percentile ranking. The GMI score uses a large number of qualitative metrics and ranks firms relative to one another on a scale from 1 to 10. No subscores are provided. The company designed the metric calculation to emphasize governance quality, transparency and compliance with established codes and principles. The Corporate Library provides the Board Effectiveness Rating, which is used as the primary rating by the authors, and a second rating of Best Practices. They also publish four subscores that examine similar aspects of governance when compared to the subscores of RiskMetrics. The Board Effectiveness Rating used by the authors is chosen as it is promoted by The Corporate Library as not merely based on checklists and a set of measured, static factors, but instead relies on the intuition and judgment of The Corporate Library staff, with careful consideration of certain red flag indicators. The Board Effectiveness Rating is issued in letter-grade format A through F. However, they publish a Best Practices rating as well, which is a numerical score based on the rated company s compliance with best practices in checklist-fashion. This score is used in robustness checks along with the subscores. Audit Integrity uses a model based primarily on accounting practices and risk, with governance as a sizable factor but not the main concern. This score is a numerical value between 0 and 100, with an aim of rating the likely quality of the firm s financials, filtered through the lens of governance practices. The authors first finding is that there are low correlations between certain ratings and high amongst others which indicates disagreement in what is most important for good governance 2010 CORPORATE GOVERNANCE REPORT 25
between the rating firms. GMI and CGQ are well correlated and both are correlated to The Corporate Library s Best Practices Rating, which is expected since these ratings have a similar methodology: each is based on a series of criteria that are deemed important or instrumental to establishing good governance practices. The Audit Integrity score is not highly correlated with any of the main ratings (statistically speaking it is correlated with CGQ and The Corporate Library Board Effectiveness, but at low values), which is also expected since it has accounting as a main focus and governance as a secondary risk measure. It is most correlated with The Corporate Library s Audit subscore. Lastly the Corporate Library s Board Effectiveness score is significantly related to Audit Integrity as noted, and to its own subscores and the Corporate Library s Best Practices Rating. Audit Integrity and The Corporate Library s Board Effectiveness ratings differ from the others in terms of area of focus and level of subjectivity, respectively, so low correlations between these ratings and those of GMI and CGQ are not surprising. The authors also note each rating is highly correlated through time. variables known in the financial literature to affect the probability of a restatement as controls, and then perform a duplicate analysis without the use of the control variables. In each case, the authors find GMI and Audit Integrity ratings are predictive of restatements, and the accounting subscore of The Corporate Library as well as its Best Practices rating are statistically related to restatements. The authors attempt to determine the economic relevance of the relationship, focusing on the GMI rating as it showed the greatest predictive ability. In their analysis, the GMI rating correctly identifies more firms that restate than control variables alone; however, the rating results in a larger number of false positives as well. Due to the potentially large consequences of restatements, in both controlled and uncontrolled analyses. Predictive Ability on Negative Outcomes: Class-Action Lawsuits Woodruff-Sawyer provided data showing 3.5 percent of the sample became subject to a class-action lawsuit within the studied period. The analysis is conducted similarly as above where controls such as size, beta and the standard deviation of returns are chosen from the applicable literature, and analyses are conducted with and without these variables. Audit Integrity s rating is again shown to be predictive of the outcome, both with and without the inclusion of controlling variables. Other ratings or subscores show significance only in one analysis, either with or without the control variables, but not in both. Again as with the above analysis The problem of board capture is so acute that it simply is not reasonable to construct a system of corporate governance that relies in any meaningful way on boards of directors to improve the corporate performance or prevent corporate deviance. Jonathan Macey, Corporate Governance: Promises Kept, Promises Broken It is further relevant to note that among investors and market participants, the notion of what comprises good governance is hotly debated, so it is not unexpected to see different firms taking different approaches. Predictive Ability on Negative Outcomes: Restatements The authors perform an analysis to determine whether the ratings are useful in predicting negative outcomes such as accounting restatements or class action lawsuits. First, using data from Glass Lewis & Co., another governance researcher, they identify just over six percent of the sample as having one or more accounting restatements over the sample period. They perform the analysis with leverage and several other investors may be more concerned with the presence of false negatives, which makes the analysis more difficult than merely comparing the two errors. A comparison of two portfolios using a buy and hold approach, one excluding firms using the control-predicted approach and a second also excluding firms using the GMI ratings may have been instructive here as to the actual economic feasibility of the ratings. The results were robust to the use of top and bottom deciles alone, as well as using only the firms which are common to all four rating firms, rather than the entire universe of firms covered for each firm. This common firm approach resulted in CGQ becoming predictive on restatements, the significant rating is expected to increase the number of correctly identified firms with a lawsuit over control variables, while increasing the number of falsely-identified firms as well. Once again, a portfolio approach may have been useful in illustrating the effect of following such a rating strategy. The robustness checks performed by the authors support the above findings. Future Operating Performance Prediction The authors also perform an analysis to determine whether governance ratings are useful in predicting future operating performance. They perform the analysis with size as a control variable, and then perform duplicate 26 STATE BOARD OF ADMINISTRATION (SBA)
analyses with and without the use of lagged industry-adjusted ROA as a control. In each case, size and the lagged industry-adjusted ROA variables are highly predictive of future operating performance. The only rating consistently found to have significant ability to predict future operating performance is from Audit Integrity. When used with or without the lagged industry-adjusted return-on-assets (ROA) control, this rating shows mathematical significance at a level which the authors determine likely has economic significance in separate analysis, in which a standard deviation increase in the rating results in nearly a quartile jump in industry ROA performance. The authors note that this relationship is not necessarily causal, as both governance and performance may be jointly influenced by another unconsidered variable. Several other firms ratings or subscore ratings are significant when the lagged industry-adjusted ROA variable is excluded, but the relationship disappears when this control variable is added, with the exception of CGQ board and audit subscores. The loss of significance when the ROA control variable is added implies that the relevant information in the rating that helped determine the future operating performance was captured with the previous period s ROA on an industryadjusted basis. The authors conduct a robustness check of these results using only the firms which are common to all four rating firms, rather than the entire universe of firms covered for each firm. Only Audit Integrity s rating continues to be significantly related to future operating performance, though the significance level falls to within 10 percent. In an analysis on the top and bottom deciles, Audit Integrity s rating remains significant with inclusion of lagged ROA. Firm Value Prediction...there will never be adequate fi xes to executive pay in any industry sector - public or private - as long as there are guaranteed Golden Parachutes or other risk-free contractual provisions that allow executives to bail out of a failing company while both employees and shareholders go down with the ship. Frank Glassner, CEO, Veritas, Executive Compensation Consultants The authors use Tobin s Q as a proxy for firm value, which is a variant of a market to book calculation and known to potentially incorporate other firm aspects such as risk, opportunity set and growth opportunity. The authors conduct an analysis on firm value using Tobin s Q in similar fashion to the above use of ROA for operating performance. The lagged Tobin s Q provides a control which is highly associated with future firm value in each analysis. None of the ratings are shown to predict firm value with any robustness. The authors interpret the results as an indication that the ratings cannot adequately predict future firm value, though model specification could be an underlying cause. Future Stock Returns The authors evaluate the ratings relationship with excess stock returns calculated from the monthly Fama- French returns based on four factors: beta, size, market to book value and price momentum. Since this calculation encompasses the known factors affecting excess returns, no other controlling variables are included. The Corporate Library s Board Effectiveness score and Audit Integrity s rating are both found to be positively associated with future excess returns, though the authors do not explore the economic feasibility of the magnitude of this relationship. They find an unexpected negative relationship between excess returns and the CGQ board score, and suggest that possibly the higher board scores are somehow associated with a lower cost of capital, though this is contrary to the intended purpose of the score. The positive relationship of Audit Integrity s rating with future excess returns is robust to the common firm sample and to top and bottom decile analyses, with results that seem attributable to its identification of highly negative alpha in firms in the lowest decile. Voting Analyses Risk Metrics is one of few companies that not only issues governance ratings, but also provides recommendations for proxy voting. It is the largest provider of such recommendations, which is a primary line of business for the firm. The authors evaluate the relationship between the CGQ rating issued by Risk Metrics and voting recommendations on management proposals of various types, and also examine whether the CGQ is related to the outcome of the final shareholder vote, which could be assumed if the score encompassed information that is useful to shareowners in determining their votes. To control for the additional factors shareholders consider when casting a vote, the authors include the prior year s excess returns, and for compensation voting items, proposal dilution, burn rate and overhang are included as well. The authors find a weak but statistically significant relationship between CGQ and voting recommendations for all recommendations, and specifically director elections and compensation 2010 CORPORATE GOVERNANCE REPORT 27
Many factors including CEO influence over director nominations, the complexity of pay plans and packages, the reliance on compensation consultants whose incentives may discourage objectivity and social influences, such as group bias toward collegiality undermine boards ability and willingness to bargain at arm s length over executive compensation. Compensation Accomplices Mutual Funds and the Overpaid American CEO The Corporate Library/AFSCME/Shareholder Education Network, April 2009 items, with or without the presence of controlling variables. Their analysis suggests very large changes in CGQ would be required on average to have a significant effect on the probability of Risk Metrics making a certain recommendation for any proposal type, due to the weakness of the relationship. Their analysis of CGQ score and the final outcome of the vote suggest that CGQ score is not related to the final vote outcome, though a strong relationship exists between the Risk Metrics recommendation and the outcome. Summary The authors conclude that the rating services have little or no predictive ability and suppose that any predictive benefit of the ratings would be economically small, though they point to Audit Integrity as a potential exception. It is suggested that although governance attributes may be of importance, the rating services may not display enough reliability in their ranking methodologies. They ponder why investors use these services and come to these potential reasons: they are misled by claims made by the ratings providers; they buy the ratings merely for access to the underlying data; they buy ratings as proof they are diligent in performing their fiduciary duties; or the model specified in this research may simply not capture the true impact of the ratings. They close by seeming to implore the firms to comply with the virtue of transparency by disclosing the right model to be used if this last possibility is the case. An Investor s Perspective It is somewhat difficult to reconcile the repeated findings of significant relationships between ratings and future outcomes, particularly the ability to predict negative events, and the authors conclusion of no predictive ability amongst the ratings firms. However, in consideration of the authors final suppositions on the popularity of the ratings, it fits within our experience to use the data provided by the rating firms partly as suggested above: in order to evaluate the underlying data. Large investors such as the SBA tend to have a great deal of assets invested passively. The result of such an investment strategy is that certain firms will be held in the portfolio that have, in our opinion, poor governance policies and may take liberties in areas like board composition or executive compensation that seem detrimental to the long-term shareholder. We use ratings to identify such companies and seek to proactively encourage the adoption of more shareownerfriendly practices. These ratings are helpful considering that large numbers of companies are held within large public investors portfolios, and data tracking of this large number of firms is very time-intensive. We believe that better governance practices will allow shareowners to protect themselves in the long run. It is about preservation of value as well as appreciation. However, it does appear that there is value in these ratings if used in more pro-active stock selection, as evidenced by this paper s findings, suggesting that investors might make better returns given certain criteria. The assessment of excess returns, or alpha, in this analysis is done with the Fama-French model, which seems to be more popular in academic analyses than in practice. In future research it would be interesting to note if the choice of calculation for alpha leads to considerable difference in results. In the analysis on voting recommendations and CGQ scores, some of the results can be better understood by considering Risk Metric s use of rule-based recommendations for many proposals. This is favored by subscribing shareowners both for its predictability and clarity. The bar for good governance is constantly sought to be raised by investors, and they often support basic tenets such as improving shareholder rights, independent boards, and fair compensation plans universally and without consideration of what the firm may already be doing or their present performance. Risk Metrics has to satisfy customers and therefore tends to provide proxy recommendation research that is generally satisfying to the ideals and assessments of its customers. If it did not represent the values of subscribers in its recommendations, there would not be demand for its service. It is our opinion that the Risk Metrics recommendation is a reflection of the sentiments of subscribing shareowners, and it is not swaying any material amount of votes in corporate ballots. In fact, it is the shareowners that are in fact swaying the 28 STATE BOARD OF ADMINISTRATION (SBA)
voting policies and recommendations of Risk Metrics. Lastly, aspects of good governance can often be mimicked by poorermanaged firms in an attempt to signal quality. This would have the greatest disguising impact on rating firms that use more objective, list-constructed methodologies. It is instructive that an accounting quality research firm with consideration of governance practices came out as a clear winner in this research. Paring the salient aspects of governance with accounting may allow a better identification of firms that may be faking good practices in both, as poorly managed and overseen companies may be apt to do, which increases the risk of being identified as such. GMI Ratings and Corporate Performance: 2003 to 2008 By Robert Watson and Kevin Spellman A stated goal of GovernanceMetrics International is to provide governance ratings for companies to shareowners for use in identifying and avoiding high-risk investments. The underlying premise is that good governance and transparency will lead to superior returns and lower risk, partly accomplished by a lower cost of capital. This research paper examined the relationship between GMI s governance ratings and performance outcomes in US firms for the period beginning in 2003, when GMI began rating firms, through 2008. The initial firms rated by GMI were biased toward a larger size, since the S&P 500 was the first index to be rated. These early rated firms had statistically significant lower growth in sales, lower total shareholder returns and higher earnings than firms added in later years. However, the market to book ratio between early and later rated firms were similar. While the average GMI rating score did not change significantly over time, the use of red flag indicators to highlight specific practices or occurrences at companies that raise the risk of longer-term underperformance did increase. Also, the distribution of firms among the various industry categories tended to shift as more firms were added; the addition of smaller and mid-sized companies added more financial and information technology firms, while decreasing the percentage of materials, consumer staples and utilities sector firms, which is important as the performance measures vary by year and by industry. To compensate, the analysis of performance was performed using total returns and returns adjusted for size and sector. The ratings also have some underlying relationships with company variables that were controlled. The authors conducted an analysis of the ratings which revealed they were slightly positively correlated with past and future returns and profitability and had a slight negative correlation with past and future sales growth. The largest correlations were between size and the rating. Using data covering all of the ratings in all years, the initial year, and only those firms that were continuously rated over the years, the authors showed consistent and similar correlations between the ratings and the variables above for all three groups. To determine if information in the GMI ratings not captured by the controlling variables was useful in predicting future performance, the authors analyzed the GMI ratings and subsequent twoyear performance, along with the controlling variables. The regression results showed a significant positive relationship between GMI rating and future performance for each data set, indicating that the ratings have explanatory power of future performance in addition to that of the control variables. As a robustness check, the authors used the residuals from the analysis where the GMI ratings were shown to be related to size, industry and prior measures of returns, sales growth and profitability. The residuals contain any aspects of the ratings which are not predictable by these variables and therefore represent a source of new and different information. Again, the analysis of the residuals and future performance suggest that the GMI ratings have additional information over the control variables helpful to predicting future performance. This suggests the ratings have value to investors in addition to the variables that are commonly used in predicting future performance. The authors used the ratings to further segregate the firms into three portfolio simulation groups of high, medium and low scores. Beginning from 2003, the portfolios were rebalanced annually. In addition, the analysis was performed on the initial 2003 rating sample and evaluated in a buy-and-hold approach. A final portfolio approach simply avoided investing in any companies with a red flag marker. Lower rated GMI firms were outperformed by medium and high rated firms on a sector-relative and absolute basis across several portfolio simulations. The authors do note the potential for mean reversion as the market acts to fully incorporate the elements of good or bad governance into returns over time, and the potential for early skewing of ratings coverage to larger cap firms to distort returns. Overall, the results bolster the evidence that GMI ratings may contain information that is additive to that of commonly assessed performancepredicting variables and suggests that such information may be used to obtain economically significant returns. 2010 CORPORATE GOVERNANCE REPORT 29