we decided to employ a U.S. large-cap growth ETF with slightly fewer stock holdings and a longer time period. 32 The founder-CEO index has a relatively high cor- relation with the VUG (0.92). Later in this article, we will show performance analytics and attribution relative to this benchmark. We considered using more popular indexes, such as S&P 500 Growth or Russell 1000 Growth, for compar- ison purposes. However, the founder-CEO index only has a correlation of 0.84 with the S&P 500 Growth Index and a correlation of 0.88 with the Russell 1000 Growth Index. Furthermore, due to the complex data intensity of our factor analysis utilized on Bloomberg, we chose a U.S. large-cap growth index with fewer constituents than the S&P 500 Growth and Russell 1000 Growth to allow a more compre- hensive factor analysis (shown later). Given all of the consid- erations involved, we believe the VUG is an appropriate fit for our analysis. 33 As we will discuss in the performance analytics sec- tion, the relative performance of the founder-CEO index versus benchmarks varies considerably across sectors. Much of the relative performance in generating alpha derives from a few sectors, including IT, Health Care, and Staples. Interestingly, one of the worst-performing sectors for the founder-CEO index comes from Real Estate (which has a relative overweight) and Industrials (which has a relative underweight). 34 As our results in the “Performance Attribution” sec- tion show later in the article, the IT sector for founder CEOs contributes significant alpha generation to the portfolio per- formance due to stock selectivity but loses some alpha due to low allocation to the IT sector (relative to the benchmark). 35 The 30-stock portfolio in the founder-CEO index is subject to wide variations in composition with a one- or two-stock movement during rebalancing (e.g., dropping two health care stocks and adding two IT companies). In con- trast, a 500-stock index would be less likely to experience strong shifts in sector weights. We realize that it might be preferable to use a founder-CEO index with many holdings, although the nature of such an objective, while keeping true to a high level of data integrity, becomes impractical. We source founder CEOs from many data sources: Bloomberg, Factset, CapitalIQ, ExecuComp, and company websites. Data are often difficult to find and inconsistent across sources. Many opportunities exist for errors of omission. We identify approximately 500 companies per quarter that trade on a major U.S. exchange and have a founder currently present in the company. From the approximately 500 companies, approximately 150 include a founder CEO. Each quarter, the approximately 150 founder-CEOs are rebalanced and ranked by market cap. The top 30 are selected and placed in the portfolio with an adjusted market cap weight. Although we track approximately 150 publicly traded founder-CEO
companies, approximately 30 constituents fit the U.S. large- cap category. As we drift below the top 30 U.S. large-cap founder-CEO constituents, the market capitalizations begin to fall sharply, thus negating the benefits of a well-defined benchmark. Given the wide variation of returns among U.S. large, mid, and small cap, we believe it is best to limit the number of constituents (while meeting the minimum threshold of diversification) and use a U.S. large-cap growth benchmark. After determining the 30 stock selections per quarter, we apply a smoothing factor to the index to reduce turnover (which approximates less than 10% per quarter). The founder-CEO index that we employ is consistent with the founder-CEO index on Bloomberg (ticker: CEOF) devel- oped by EntrepreneurShares butt calculated and distributed by Thompson-Reuters. The computations applied in this analysis employ a bottom-up dividend reinvestment approach (which assumes that dividends are reinvested back into issuing company), compared to the top-down dividend reinvestment approach (in which dividends are reinvested back into index). We employ this approach (with disclosure) to be consistent with the excess return computations shown in the “Perfor- mance Attribution” and “Factor Analysis” sections (which assume a bottom-up dividend reinvestment approach in the computational algorithm). Over an extended period of time, the compounded nature of the reinvestment methodology can generate potentially wide variations of returns. 36 Despite IT and Health Care being shown as (signifi- cantly) underweight during the time period of the study, as we will see in the “Performance Attribution” section, much of the performance contribution comes from three key sec- tors: IT, Health Care, and Consumer Discretionary. 37 This time period corresponds with a complete market cycle, including the 2007–2009 stock market recession. The time period also corresponds with an actual performance track record implemented (along with other variables) in an entrepreneur model along with a founder-CEO index (pub- lished and disseminated by Thomson Reuters). As noted in a prior section, many companies included in our study have been delisted due to acquisitions, corporate actions, and bank- ruptcies. These older records, in particular, have been very time consuming to gather, although they help ensure accu- racy, completeness of data, and elimination of survivorship bias. Data have been computed on the eVestment database with U.S. large-cap growth as a benchmark universe. 38 We perform the analysis for the December 2006 through December 2015 time periods. The one-year period (and year to date) correspond with calendar year 2015; the two other bar charts correspond with the three-year and five- year periods (dating from December 31, 2015), respectively. 39 eVestment is a database widely used by the financial industry’s top consultants and includes one of the most com- prehensive sets of returns for investment professionals around
L EADERSHIP M ATTERS : C RAFTING A S MART B ETA P ORTFOLIO WITH A F OUNDER -CEO T WIST
W INTER 2017
Made with FlippingBook - Online magazine maker