When It Comes to SMAs, How Do You Measure the Impact of Personalization?
KEY TAKEAWAYS
- Separately managed accounts (SMAs) allow investors to customize their portfolios to meet unique goals and preferences. As portfolios become more personalized, they are likely to deviate more from the starting strategy.
- A forecast of tracking error is often used to set expectations for future differences in performance but has meaningful drawbacks.
- We believe that overlap, along with other holdings-based data lenses, can more reliably inform investor expectations about differences in performance through time.
Many investors want their portfolio to do more than just pursue reliable premiums. They may also want to seek tax efficiency; reflect their environmental, social, and governance (ESG) values; respect any specific security restrictions they might have; or take into account considerations related to their present and future employment situation.
The more personalized the investment solution becomes, however, the further away it gets from the initial investment strategy. What is the best way for investors to evaluate the potential performance differences between a personalized portfolio and the underlying starting strategy?
Tracking error is a common measure of performance deviation between a portfolio and a benchmark. Ex post, or realized, tracking error is defined as the standard deviation of the difference between the portfolio and benchmark returns. This metric can be useful but is not available for a newly constructed strategy. Ex ante, or forecast, tracking error is the expected tracking error derived from a forecasting model and is often used to measure how much the performance of a personalized portfolio is likely to deviate from the performance of the starting strategy. While ex post tracking error may be informative for investors, it is important to exercise caution when interpreting ex ante tracking error. The estimate necessarily relies on many assumptions, typically informed by historical outcomes. As a result, measures like ex ante tracking error can be highly sensitive to even small changes in the estimation choices, such as variable selection and construction, time period, and data frequency.
Another caveat: Tracking error cannot help us distinguish between two types of deviations from a starting strategy: (1) deviations in holdings that can generate short-term differences in realized returns but have no systematic impact on expected performance, and (2) deviations in holdings that can generate persistent differences in expected performance. For both of those reasons, we prefer to focus on alternative ways to evaluate deviations from a starting strategy.
Our new paper “SMAs: Measuring the Impact of Personalization,”1 shows how overlap, along with a few other aggregate portfolio characteristics, can inform investors’ expectations about both short- and long-term performance differences.
Overlap
Overlap can provide clients with a more intuitive and robust measure of proximity to the starting strategy. It measures the proportion of the starting strategy that ends up in the personalized investment solution. The greater the overlap between the two, the lower the expected dispersion in returns. A 99% overlap between the initial strategy and the final portfolio means that $99 out of $100 in the final strategy is invested the same way as in the initial strategy.
Exhibit 1 illustrates the overlap between a hypothetical starting portfolio (represented in yellow) and the investor’s personalized portfolio (blue). Both portfolios hold 10 stocks. Nine are in both portfolios, while stocks D and H are in only one or the other, so D and H contribute 0% to overlap. The shading indicates overlap, which consists of the nine stocks held in both portfolios. These nine make varying contributions due to differences in weighting, ranging from 4% for stock G to 20% for stock I. In total, the overlap of this investor’s portfolio vs. the starting portfolio is 81%. In other words, if one investor puts $100 in the starting strategy and another puts $100 in the final strategy, $81 of their investments will be identical.
For varying levels of overlap across portfolios, how much performance deviation do we observe over time? To study that, we examine the historical overlap and tracking error of three Dimensional US all cap core equity indices vs. the Dimensional US Market Index. The three all cap core equity indices start with the same universe as the Dimensional US Market Index but overweight stocks with higher expected returns (stocks with smaller market capitalizations, lower relative prices, and higher profitability), underweight stocks with lower expected returns, and exclude stocks with extremely low expected returns (small growth low profitability firms and small high investment firms). The three all cap core equity indices differ from one another in the emphasis they place on stocks with higher expected returns. Hence, we can consider these indices as three personalized investment strategies and the market index as the starting strategy. The strategy with the strongest emphasis on higher expected returns (the “Strong” strategy) has the lowest overlap with the starting strategy and the highest tracking error over time. The opposite is true for the strategy with the lightest emphasis on known equity premiums (“Light”), while the third strategy (“Medium”) falls in between. Based on our historical analysis from 1979 through 2020, a steady overlap of about 90% for the Light strategy corresponds to a historical annualized tracking error of about 1%, whereas a steady overlap of about 80% for the Medium strategy is associated with a 2% annualized tracking error. Overlap of about 75% for the Strong strategy corresponds to tracking error of 3% per year.
Additional Data Lenses
Overlap can help inform investor expectations about differences in performance through time, but it should be supplemented by other data lenses.
When building an SMA, differences across top holdings and sector weights can help investors better understand drivers of potential performance deviations vs. the market over shorter time horizons.
Top holdings are particularly informative about short-term performance deviations. For example, in a portfolio that has 95% overlap with a starting strategy, suppose the 5% difference in weights comes from differences of 5 basis points (bps) across 100 small holdings. It is unlikely that the small weight differences across many names will have a meaningful impact on performance deviation in the short term. Some of the small overweights will probably outperform; others will probably underperform. The spread of weight differences across so many names and the small sizes of those weight differences make it likely that the noise in their realized returns will be diversified away and will not have a large impact on short-term relative performance. Now consider an alternative portfolio, also with 95% overlap, but the 5% deviation is driven entirely by one large stock. With the weight deviation concentrated in one name, investors are more likely to experience noticeable day-to-day and month-to-month performance deviations between the starting portfolio and the final portfolio.
Just as differences in top holdings are informative about short-term differences in performance, sector weights across strategies inform investors about the magnitude of potential return deviations from a benchmark or a starting strategy. Differences across top holdings and sector weights can help investors better understand drivers of potential performance deviations vs. the market over shorter time horizons.
What about drivers of long-term differences in expected performance? These differences arise from systematic differences in emphasis on the reliable long-term drivers of expected returns, which are the size, value, and profitability premiums in equities. Hence, a useful starting point is to compare the weighted average market capitalization, relative price, and profitability of the strategies.
While aggregate equity characteristics are helpful tools in setting expectations for long-term systematic performance differences between two strategies, they cannot always paint the full picture. In other words, using aggregate equity characteristics in this way is like evaluating the condition of a car based only on its outward appearance. While the outward appearance is relevant, it does not tell the whole story. You need to look under the hood. For equity portfolios, we can use our relative positioning lens, which breaks down the market into size-value-profitability subsets and shows the relative weights of strategies subset by subset. This data lens provides a granular evaluation of expected return differences among the strategies and can show whether the personalized portfolio has an emphasis on the groups of securities with higher expected returns (small, value, high profitability names) like that of the starting strategy.
As investors embrace personalization in the pursuit of their goals, understanding how their personalized portfolio differs from a starting strategy will help inform their expectations about short- and long-term differences in performance. We believe that a comprehensive evaluation of a personalized equity strategy relative to a starting strategy should include a review of overlap, top holdings and sector weights, aggregate equity characteristics, and relative positioning in order to set proper expectations about short-term and long-term deviations in performance.
Kaitlin Simpson Hendrix is a Researcher and Vice President at Dimensional Fund Advisors