Constructing Superior Equity Portfolios by Cambridge Associates
- A common perception among investors that employ active equity management is that the “donut” structure—an all-active manager structure often composed of four or more high conviction managers—is more aggressive, more expensive, and riskier than the “coresatellite” structure—which blends active and passive management by adding a large passive core component to the donut—because of the donut structure’s heavier reliance on concentrated, high tracking error, high fee managers. In contrast, the large passive element of the core-satellite structure is presumed to reduce these risks. The research we present in this report calls into question these perceptions.
- Over the 17-year period examined in this report, the donut structure has offered higher returns net of fees with similar levels of risk to the core-satellite structure. Though the core-satellite structure had a tighter range of return outcomes, performance for the donut structure was superior across all four performance quartiles. The median outcome for the donut structure was 72 bps higher, while the upper range (1st percentile) of outcomes outperformed by 140 bps. Notably, even the worst (99th percentile) outcomes of the donut structure were slightly superior, beating those of the core-satellite structure by 23 bps.
- Any historical analysis of returns—especially one encompassing 17 years—is subject to survivorship bias. In addition,because the donut structure relies more heavily on active management, it is more likely to benefit from that bias than the core-satellite structure. We conducted a detailed analysis of the impact of that bias and found that survivorship bias accounted for 15 bps of the median donut portfolio’s relative performance—meaning that 57 bps of the 72 bps outperformance is attributable solely to manager structure.
- Does the higher return of the donut structure come with an unpalatable level of risk? Our analysis indicates the answer is no. While the average volatility of the core-satellite portfolios is indeed slightly lower (16.8 versus 17.1, consistent across all four performance quartiles), the higher return of the donut structure across quartiles makes for a higher risk-adjusted return.
- The 17-year period of our analysis contains several unique market environments: the tech bubble, the tech bust, the easy-credit bull market from 2002 to 2007, the 2008 liquidity crisis, and the stimulus-driven rally since 2009. In each of these five periods, the donut structure had higher median performance and better performance in the top quartile than the core-satellite structure. However, the 99th percentile portfolios of the donut structure underperformed those of the core-satellite structure in each of the five periods. The underperformance of the donut structure’s 99th percentile portfolios was more pronounced over two of these periods: the speculative environment of the tech bubble between 1996 and March 2000 and the liquidity crisis of 2008–09.
- For most rolling three-year periods since 1996, the median and upper percentiles of the donut structure materially outperform those of the core-satellite structure, while the performance of the lower percentiles is generally similar. However, there are extended periods where the rolling three-year performance of the worst-performing (90th percentile) donut portfolios substantially underperformed the worst-performing core-satellite portfolios.
- For investors that adopt a manager structure populated by concentrated, high tracking error managers, whether in a donut or core-satellite structure, perhaps the greatest risk they face is the natural tendency to focus on short-term results. Given the cyclicality of manager returns, the temptation to fire a manager based on poor recent performance—and to replace it with one showing strong recent performance—can lead to disappointing results. Success for either structure depends on patience and a long-term view. A five-year time frame—encompassing a more complete market cycle—is likely a better yardstick of success or failure than shorter evaluation periods.
- While rebalancing is not critical to the success of the donut structure, it seems to have a more positive impact for the donut structure than for the core-satellite structure. Further, because rebalancing has a small positive impact over time—particularly in reducing the magnitude of negative outcomes—investors may find it worthwhile to put in place target allocations for individual managers to establish a framework (and impose discipline) for annual rebalancing.
- Investors with core-satellite portfolios might benefit from revising their rationale for implementing with such a structure. If it is to protect portfolio downside or mitigate volatility, investors may not be getting the result they expect. Recognizing the behavioral challenge that might exist in moving to a donut structure, we would encourage investors to consider how their equity returns may be augmented by abandoning a traditional core-satellite structure and allowing a well-selected portfolio of high conviction equity managers to fuel the return engine in their equity portfolios.
Constructing Superior Equity Portfolios
For investors that choose to employ active equity management, a great deal of time and effort is spent trying to identify managers that can beat their benchmarks. But how you structure your equity manager portfolio may be just as important as the managers you choose in outperforming the market. Manager structure takes many forms, ranging from completely passive to completely active. This report takes an in-depth look at two common equity manager structures to assess whether investors may fare better with one structure or the other. The first is the “donut” structure, typically composed of four or more high conviction managers diversified by style, capitalization, or strategy. The second is the “core-satellite” structure, which essentially blends active and passive management by pairing the donut structure with a large passive “core.”
A common perception among investors is that the donut structure is more aggressive, more expensive, and riskier than the core-satellite structure because of its heavier reliance on concentrated, high tracking error, high fee managers. In contrast, the large passive element of the core-satellite structure is presumed to reduce these risks. However, our research calls into question these perceptions: over the 17-year period examined in this report, the donut structure has offered higher returns net of fees with similar levels of risk to the core-satellite structure. Our analysis suggests that, at a minimum, investors should reassess whether a core-satellite structure is as likely to help them earn their payout as a donut structure.
To weigh the tradeoffs between the donut and core-satellite structures, we developed a model to construct portfolios at random using return data from a subset of managers in the Cambridge Associates LLC Investment Manager Database. In selecting managers for this analysis, we wanted to achieve three things:
1. Sufficiently long time horizon. Recognizing that there are extended periods in which certain factors fare better or worse—or where active management as a whole tends to struggle relative to passive management—extending the analysis as far back as possible while maintaining sufficiently robust data proved critical in creating an even playing field for both structures. The period between January 1997 and December 2013 offered the most robust data set while encompassing multiple market cycles. Excluding any managers that reported performance for only part of that 17-year period, we began with 713 equity products with complete performance records.
2. Diversification by style and capitalization. To neutralize the impact of factor exposures, each portfolio created by our model contains only one manager from each of the all-cap, large-cap value, large-cap growth, small-cap value, and small-cap growth categories. If an investment strategy did not fit into one of these buckets (e.g., single sector strategies, enhanced index strategies, REITs), it fell out of consideration. After isolating managers into each of these five categories, we further whittled our universe down to 409 equity products.
3. High Active Share.1 Both the donut and core-satellite structures attempt to add value by emphasizing concentrated, high conviction managers. To select those managers from our universe of 409, we used the findings of an earlier Cambridge Associates report as a guide.2 That report found that the combination of high active share, high concentration, and modest tracking error had the greatest power to predict superior returns.3 However, as there is such a strong correlation between active share and concentration4 we simply used active share as a proxy for both criteria. Moreover, high active share in isolation is a strong predictor of superior return, while the same cannot be said of concentration, as illustrated in Table 1. Selecting for only the top quartile by active share in each of the style and capitalization categories described above, we were left with a final universe of 108 strategies diversified by style and capitalization, filtered for active share, and with complete performance records.
See full PDF below.