idea mind. half of your 4-digit GICS code in TSR performance can severely penalizee TSRs of thee compensation plan. lead to ISS So why do some very

Transcription

1 RADFORD REVIEW: Should GICS Codes Be Used In Comparisons? Pay-For-Performance Falling in the bottom half of your 4-digit GICS code in TSR performance can severely penalizee your executive compensation plan. Over the past several months, numerous companies have publicly voiced concerns about Institutional Shareholder Services (ISS) policies on pay-for-performance. The current policy, put into place primarily to evaluate CEO compensation practices for the purpose of making voting recommendations on say-on-pay proposals, uses a fairly simplistic approach focused on shareholder returns. According to ISS, they follow this line of thinking becausee From a shareholder s perspective, performance is predominantly gauged by the company s stock performance over time. 1 While most would agree shareholder returns are generally indicativee of corporate performance, companies are nevertheless concerned that a seemingly one-sizefits-all methodology is being applied towardd the judgment of CEO pay practices. After all, the idea of pay-for-performance is only meaningful and productive when performancee is measuredd effectively with company and industry-specific factors in mind. Revisiting the ISS Pay-For-Performancee Methodology ISSS identifies a company as having sustained underperformance relative to peers by analyzing a company s Total Shareholder Return (TSR)) against the median of the TSRs of thee Russell 3000 companies grouped into its same Global Industry Classification Standard (GICS)) code. If the company s one-year and three-year TSRs are less than the median of the TSRs of these comparator companies, the company will be flagged as underperforming and furtherr analysis will be completed to better understand the various pay elements and whether they create or reinforce shareholder alignment. 1 Ultimately, failing this simple test can lead to ISS recommending a vote against the company s executive compensation plan. According to Standard & Poor s, the publisher of the GICS system, GICS codes are widely- codes, 688 industry codes, and 154 sub-industry codes. For performance measurement purposes, ISS accepted industry classifications consistingg of 10 sectorr codes, 24 industry grouping utilizes the 24 industry grouping codes, commonly referred to as 4-digit GICS codes. So why do some believe the ISS process is an ineffective way of measuring performance? In statements to shareholders, it is clear companies do not believe 4-digit GICS code provide for accurate comparisons because companies includedd in these groupings often represent very different or unrelated lines of business. Comparing the returns of unrelated companies, which respond and react to different market forces, can potentially yield very biased results and lead to false conclusions about relativee performance. Radford is an Aon Hewitt Company 1 From page 12 of ISS U.S. Proxy Voting Guidelines Concise Summary (Digest of Selected Key Guidelines) as of January 3, 2011.

2 To assess the validity of this belief, we analyzed the strength 2 of the relationship between the Russell 3000 companies in each 4-digit GICS code. Specifically, we used revenue comparisons, industry representation, and most importantly, the highly objective and transparent measure we discuss in our next section, stock price correlation. Stock Price Correlation If high taxes were imposed on big oil companies, the stock prices of both Exxon Mobil and Chevron should be affected similarly, while the effect on Ruby Tuesday s stock price should be relatively minimal. Given that ISS methodology is based on stock price returns, there is probably no better measure to assess the ISS process, and relationships between any two companies, than stock price correlation 3, as it measures the alignment between two companies stock price movements. This measure also represents our most objective analysis tool applied in this research, as it is based solely on actual data and not subjective comparisons. In theory, similar companies should be highly correlated in stock price movements as they share parallel attributes and are affected by related market forces. For example, if high taxes were imposed on big oil companies, the stock prices of both Exxon Mobil and Chevron should be affected similarly, while the effect on Ruby Tuesday s stock price should be relatively minimal. While companies in similar fields with comparable lines of business may never have a perfect positive correlation of 1.00, there should be a degree of resemblance in the stock price movements over time. However, when studying specific points in time, it should be noted that anomalies can exist where very different companies can have high stock price correlations (more than 0.5) and very similar companies can have somewhat low stock price correlations (less than 0.4). That said, over a significant period of time, stock price correlations should be indicative of the similarities between companies. To highlight this, the following charts illustrate five relationships between similar peer companies and five relationships between very different companies over a three-year period. Stock Price Correlation Coefficients Between Similar Companies Company 1 Company 2 Correlation 4 Bank of America vs. Wells Fargo Honda vs. Toyota The Home Depot vs. Lowe's Wal-Mart vs. Target Exxon Mobil vs. Chevron Average Median On average, stock price correlation can be a significant objective measure when selecting peers. Obviously, Wells Fargo is a more relevant peer to Bank of America than QUALCOMM, and their stock price correlations agree. Stock Price Correlation Coefficients Between Dissimilar Companies Company 1 Company 2 Correlation 4 Bank of America vs. QUALCOMM Honda vs. M&T Bank The Home Depot vs. ITT Educational Services Wal-Mart vs. PMI Group Exxon Mobil vs. Ruby Tuesday Average Median We have measured the strength of each relationship by comparing stock price movements and specific company attributes, such as market cap and revenue (discussed in a later section), to understand the level of similarity between two companies. 3 A stock price correlation of 1.00 translates to an identical positive relationship in stock price movements, meaning as one company s price moves up, the other company s price moves up by an equal amount. A stock price correlation of translates to an identical negative relationship, meaning as one company s stock price moves up, the other company s price moves down by the same amount. A correlation of 0.00 would mean that no relationship exists between stock price movements. 4 These stock price correlations were calculated using daily returns over a three-year period from January 1, 2008 to December 31, 2010.

3 Analysis of 4-Digit GICS Codes using Stock Price Correlation Due to the fact numerous companies claim peer firms analyzed in ISS methodology provide inaccurate comparisons, we assessed the relationship between ISS-selected comparator firms in our study using pair-wise correlation coefficients 5 for the Russell 3000 companies comprising each of the 24 4-digit GICS codes. Specifically, we analyzed daily stock prices over a threeyear period from January 1, 2008 to December 31, The following table illustrates the median and average pair-wise correlations of companies in each GICS code. For more detail on this portion of the study, please see Appendix A. In our study, we analyzed three-year stock price correlations of the Russell 3000 companies from each 4-digit GICS code. In the majority of GICS codes, there is a very low alignment, on average, in stock price movements of the companies comprising these codes. Correlation Coefficients for Russell 3000 Companies Comprising Each 4-Digit GICS Code GICS Code Industry Group 6 Number of Companies Analyzed Median Correlations Average 1010 Energy Materials Capital Goods Commercial & Professional Services Transportation Automobiles & Components Consumer Durables & Apparel Consumer Services Media Retailing Food & Staples Retailing Food Beverage & Tobacco Household & Personal Products Health Care Equipment & Services Pharmaceuticals, Biotechnology & Life Sciences Banks Diversified Financials Insurance Real Estate Software & Services Technology Hardware & Equipment Semiconductors & Semiconductor Equipment Telecommunication Services Utilities Average Median As you can see, the average and median correlation coefficients across all of the GICS codes are approximately 0.40, which does not indicate a considerably high level of stock price correlation between the companies analyzed. It should be noted though that a few GICS codes (particularly 1010, 4030, 4040, and 5510) yield fairly high correlation coefficients (above 0.50), which implies a stronger alignment of stock price movements of the companies comprising these codes. Yet, over half of the GICS codes yield correlations less than 0.40 with three codes (3020, 3030, and 3520) coming in at less than 0.30, indicating a very low overall alignment in stock price movements. Additionally, about one third of the codes fell between 0.40 and 0.50, implying an average, or even inconclusive, level of alignment. 5 Represents the correlation between each possible pair of companies in each GICS code. 6 Given that some companies do not have three consecutive years of data, we were not able to analyze stock price correlations for every Russell 3000 company in each 4-digit GICS code. Overall though, we were able to obtain the necessary information for over 98% of these companies, indicating very statistically significant results.

4 Given that a majority of GICS code-based groupings yield low stock price correlations, there may be some truth behind claims that ISS methodology does not employ similar enough peer companies, which can skew TSR comparisons. Still, as was mentioned above, this is not a universal concern. A few groupings have high alignments of stock price movements, so they may in fact serve as relevant comparisons. To further evaluate the legitimacy of Russell 3000 companies in each GICS code as relevant comparators, we also analyzed each company by other, more subjective, metrics to assess differences in size and structure. Alignment of Other Key Metrics Among 4-Digit GICS Codes Stock price correlation is certainly not the only metric used when creating a peer group, so it will not be the only metric considered in this analysis. Other metrics such as industry, market capitalization and revenue are commonly considered to ensure companies are comparing themselves against other companies of similar size and structure. To further assess claims that 4-digit GICS codes are not suitable comparator groups, we also studied some of these specific measures, which are not necessarily related to shareholder return or stock price correlation. In particular, we analyzed the number of employees, total market capitalization, profit margin, revenue, sector and industry representation, and fiscal year end for each GICS code group. We found an enormous amount of variability inside each code, which implies that some of these companies may not provide for relevant comparisons on any level. In GICS code 3520, companies are compared to other companies with more than 10,000 times their revenue. To exemplify these findings, we will specifically discuss GICS code 3520, which represents companies in the Pharmaceuticals, Biotechnology & Life Sciences Group. We chose this code because of the recent claim made by Amgen Inc. to shareholders that ISS methodology fails to provide an accurate comparison of Amgen s financial results, citing that many of the comparator companies are engaged in entirely different and unrelated businesses. In the table below, we compare some of these peer selection measures for Amgen against Russell 3000 companies comprising GICS code EE Headcount Market Cap Profit Margin Revenue Amgen Inc. 17,000 $52,330,000, % $15,170,000,000 Russell 3000 Companies Comprising GICS Code 3520 Min 5 $26,690, % $1,230, th Percentile 82 $222,095, % $27,420,000 Average 3,991 $5,631,192, % $2,172,731,636 Median 220 $512,130, % $107,470, th Percentile 740 $2,120,000, % $448,260,000 Max 116,500 $178,270,000, % $67,810,000,000 Sample Size Four different sectors are present in the 4-digit GICS code representing Pharmaceuticals, Biotechnology and Life Sciences. Studying the table above, it is easy to see that this GICS code covers immensely different companies. In particular, Amgen is significantly above the 75 th percentile but nowhere near the maximum of each metric. According to the data, this particular code includes three companies with less than ten employees and three companies with at least 100,000 employees. Additionally there are six companies with less than two million dollars in revenue and five companies with more than 20 billion dollars in revenue. Few companies, if any, would consider such discrepancies acceptable when developing their own peer groups for compensation benchmarking purposes. Nor would most shareholders. As was mentioned above, we also analyzed Amgen s sector and industry 7 (Healthcare and Biotechnology) vs. the sector and industry makeup of the other Russell 3000 companies comprising this GICS code. While the majority of companies in this code represent the Biotechnology industry, it is less than half, at 44%. And among the remaining 56%, we find companies from the Business Software and Services industry to the Specialty Chemicals industry. Overall, four broad sectors are actually present in the 3520 GICS code, including Basic Materials, Healthcare, Services, and Technology. 7 Represents the stock classification system derived by Hemscott Inc., a Morningstar company, and used by data providers such as Yahoo! Finance. This classification system is widely regarded as the standard grouping method for the marketplace.

5 In sum, the variability of these measures among companies in the 3520 GICS code is significant. While all of the companies have businesses related to Pharmaceuticals, Biotechnology and Life Sciences, they have very different sizes, structures, and business models. Unsurprisingly, when we return to correlation coefficients for this particular code (median of ), it appears these companies also do not possess much alignment in stock price movements. All of this evidence tends to support Amgen s public argument. These findings also hold true across the other 23 4-digit GICS codes. We find codes where companies are compared to organizations with more than 50,000 times their number of employees and almost 70,000 times their revenue. In one code, we find as many as seven different sectors and 51 different industries. This amount of variability cannot reasonably be expected to yield relevant comparisons between all companies inside each GICS code. For a summary of the metrics analyzed for all 24 GICS codes, please see Appendix B. Analysis Beyond 4-Digit GICS Codes So is there a comparator group that provides a more representative evaluation than the 4-digit GICS code? Entities comprising certain 6-digit and 8- digit GICS codes yielded similar results to their parent 4-digit GICS codes. According to ISS guidelines, if a company fails its test (i.e. finishes in the bottom half of its 4-digit GICS code for TSR), further analysis is required. While ISS does not explicitly discuss the methodology behind this secondary test, ISS has stated that it involves comparing performance for all companies in each respective 6-digit GICS code with revenue within a factor of two of the focus company. Radford also assessed the strength of the relationships between the companies comprising certain 6-digit and even 8-digit GICS codes. Specifically, we looked at the 6-digit and 8-digit codes within GICS code groups 3520 and 4530, respectively, and found little difference from the results shared above. In fact, only code , representing Life Sciences Tools and Services, showed any significant improvement in the median stock price correlation, going from to , which is still not indicative of a high correlation of stock price movements. The other analyzed codes yielded median correlations almost identical to the median correlation of each respective 4-digit GICS code. To further assess ISS secondary methodology, we also randomly selected a few focus companies comprising each of the 6-digit and 8-digit codes evaluated above, and studied the stock price correlations between the focus company and the companies with revenue (from the most recent fiscal year) within a factor of two. When applying this methodology, we saw stock price correlations climb approximately 40% from the average and median correlations of the entire code; however, given the original level of correlation in the overall codes, this was not a very significant increase. For example, GICS code yielded a median stock price correlation of When studying the stock price correlation between Regeneron Pharmaceuticals and the 11 other companies in GICS code with revenue within a factor of two of Regeneron, the median stock price correlation increases to Again, this is still not indicative of a high correlation in stock price movements. Conclusion To receive information on your specific GICS code, please Dan Kapinos at dkapinos@radford.com. Based on this research, we do not believe it is reasonable to judge a company s CEO compensation practices based primarily on a review of corporate performance against median one-year and three-year TSRs for Russell digit GICS code groupings. While Radford certainly agrees with the intention behind ISS methodology to hold companies accountable for performance, this only works when performance is measured fairly. While some companies comprising the same 4-digit GICS code may in fact be worthwhile comparators, far too often, the current methodology ends up comparing entities in very different sectors with vastly different sizes. In sum, based on the analysis presented above, Radford believes GICS codes, no matter how narrow, do not, in and of themselves, provide for suitable comparator groups for most companies. Additionally, it is not even clear that separating comparators by revenue within GICS codes will work either. Rather, we believe an

6 entirely different methodology should be explored in assessing the connection between executive compensation and corporate performance. Otherwise, companies will continue to be compared to radically different firms and be penalized for it. It appears unlikely ISS will significantly modify its approach in the near term. According to Mary Hughes, of the BNA Daily Tax Report, ISS Executive Director Patrick McGurn recently indicated during a webcast that a matrix of different ways of examining pay for performance is under consideration. However, on the matter of using GICS codes for determining appropriate peer groups, McGurn said that they would not be moving away from that methodology. As such, we expect public sparring between companies and ISS to only grow louder and more contentious in future proxy seasons. While pay-for-performance is the future of executive compensation, we must first learn to measure performance fairly and accurately. Going forward, companies should be prepared to invest in developing their own explanations (similar to Amgen) for how they evaluate and view pay-for-performance. And unless current conditions change, this will likely need to include a discussion of the potential pitfalls of using broad 4-digit GICS codes to develop comparator groups when making performance evaluations. Contact Us For assistance in developing your say-on-pay proposals and for more information on ISS pay-for-performance tests, please contact your local Radford Consultant at consulting@radford.com. If you d like to see how your company stacks up against the companies comprising your 4-, 6-, or 8-digit GICS code, please contact Dan Kapinos at +1 (215) or dkapinos@radford.com. For more information, please contact us at: Terry Adamson Senior Vice President +1 (215) tadamson@radford.com Jon W. Burg, FSA Vice President +1 (415) jburg@radford.com Dan Kapinos, CEP Consultant +1 (215) dkapinos@radford.com Press Contact Alex Cwirko-Godycki +1 (415) acwirko@radford.com Locations Austin, Boston, Chicago, Hong Kong, London, New York, Philadelphia, San Francisco, San Jose About Radford For more than 35 years, Radford has provided compensation market intelligence to the technology and life sciences industries. Global survey databases, which include more than 4.2 million incumbents, offer current, reliable data to nearly 2,000 clients. Leveraging Radford survey data, our thought-leading global Radford Consulting team creates tailored solutions for the toughest global business and compensation challenges facing companies at all stages of development. In addition to our consulting team, we also offer equity valuation assistance via Radford Valuation Services, and leading-edge market analyses and survey services with Radford Analytic Services. Radford's suite of surveys includes the Global Technology, Life Sciences, and Sales Surveys, as well as the US Benefits Survey. For more information on Radford, please visit About Aon Hewitt Aon Hewitt is the global leader in human capital consulting and outsourcing solutions. The company partners with organizations to solve their most complex benefits, talent and related financial challenges, and improve business performance. Aon Hewitt designs, implements, communicates and administers a wide range of human capital, retirement, investment management, health care, compensation and talent management strategies. With more than 29,000 professionals in 90 countries, Aon Hewitt makes the world a better place to work for clients and their employees. For more information on Aon Hewitt, please visit

7 Appendix A: Stock Price Correlation Detail of 4-Digit GICS Codes Correlations GICS Industry Group # of Companies Analyzed 1 Min 2 25th Percentile Median Average 75th Percentile Max 1010 Energy Materials Capital Goods Commercial & Professional Services Transportation Automobiles & Components Consumer Durables & Apparel Consumer Services Media Retailing Food & Staples Retailing Food Beverage & Tobacco Household & Personal Products Health Care Equipment & Services Pharmaceuticals, Biotechnology & Life Sciences Banks Diversified Financials Insurance Real Estate Software & Services Technology Hardware & Equipment Semiconductors & Semiconductor Equipment Telecommunication Services Utilities Average Median Footnotes: 1 Represents the number of Russell 3000 companies in each code with at least three years of stock price data that were analyzed in our study. 2 Please note all correlations have been rounded to four digits, so a correlation of does not imply no correlation, but an extremely low correlation.

8 Appendix B: Summary of Analysis of Non-Shareholder-Return Metrics for 4-Digit GICS Codes Max/Min Factor 1 Number of 2 GICS Industry Group EE Headcount Market Cap Revenue Sectors Industries Fiscal Years 1010 Energy 51350x 6265x 68520x Materials 3278x 346x 1869x Capital Goods 9500x 3778x 36370x Commercial & Professional Services 3659x 183x 231x Transportation 810x 1362x 1521x Automobiles & Components 2912x 323x 1147x Consumer Durables & Apparel 3722x 600x 655x Consumer Services 1578x 3905x 351x Media 906x 933x 736x Retailing 3311x 1154x 1601x Food & Staples Retailing 1587x 798x 1011x Food Beverage & Tobacco 25375x 1579x 1916x Household & Personal Products 2309x 1317x 3904x Health Care Equipment & Services 4211x 1045x 10595x Pharmaceuticals, Biotechnology & Life Sciences 23300x 6679x 55130x Banks 5036x 5531x 63983x Diversified Financials 31556x 2752x 18144x Insurance 5647x 410x 589x Real Estate 7320x 256x 1353x Software & Services 34236x 2435x 3834x Technology Hardware & Equipment 3707x 5096x 26827x Semiconductors & Semiconductor Equipment 1535x 1620x 2803x Telecommunication Services 6530x 3006x 1829x Utilities 3000x 326x 18275x Average 9849x 2154x 13466x Median 3715x 1340x 1893x Footnotes: 1 Represents the multiplication factor between the minimum and maximum of each metric to illustrate the spread between companies in each code. 2 Represents the number of each metric represented in each code. On average, almost 4 sectors, over 15 industries, and about 13 fiscal year ends are present among these codes.