The art of pension plan management consists of skillful plan design, asset allocation, investment selection, implementation and risk management. Most plans use multiple investment managers to attain a proper degree of diversification, and selecting those managers takes a lot of time and typically requires expert, external advice.
The process does not end with selection, however: Managers must be monitored on performance, risk, process and style in order to confirm that they are performing their role in the overall portfolio construction.
The methods employed to carry out that monitoring are well established. But given the importance of getting portfolio allocation right in a low-growth, low-return world, it’s worth examining whether there are new ways of thinking about and assessing whether and how a manager adds value beyond a passive benchmark.
Traditionally, pension plan managers and consultants have used tracking error (TE) to monitor the active risk their appointed asset managers take. Tracking error is the standard deviation of periodic excess returns above or below a benchmark index, usually measured either daily or monthly (see Figure 1). The theory is that if a given manager usually posts a tracking error within a certain range – say 300 to 500 basis points – a deviation above or below this range may be a sign of change in the manager’s risk posture and potentially even its portfolio management style. (To be sure, there are a number of ways to measure how a manager stacks up to a benchmark, including the portfolio’s risk factor exposures. The analysis in this article focuses on historical tracking error as well as a complementary measure.)
Plans pay attention to both high and low range exceptions – too high a degree of tracking error may signal that a manager is moving toward more concentrated positions, whereas too low a degree may mean the manager is shifting to become a “closet indexer.”
While it is true that some investment managers explicitly manage to a tracking error, more commonly managers think about active positions in terms of market value deviations from their benchmark, or sometimes even disregard the benchmark entirely in their portfolio construction, a so-called benchmark-agnostic style of management. One could argue that if managers actually managed to a target tracking error consistently, a median estimate of the historical tracking error of all equity managers should be relatively stable over time. Using data from the eVestment manager database, we did just such a calculation in Figure 2.
Figure 2 appears to suggest the median manager decided to take substantially more tracking error during 2001-2003 and 2008-2009 than in the other years in the sample. And yet we believe this would likely be the wrong conclusion: More likely, tracking error of the median manager is influenced by the degree of volatility in the markets.
The missing piece in this calculation is a measure called cross-sectional volatility. Simply put, cross-sectional volatility measures the dispersion of stock returns in a given period. It is typically calculated as the standard deviation of returns for all stocks in the relevant investment universe during one day or month. If cross-sectional volatility were zero, and the market index were up 5%, this would indicate that every stock moved in parallel the same 5%. As a corollary, in a zero cross-sectional volatility environment, stock picking would add no value because all stocks would have the same performance.
Cross-sectional volatility is highly correlated with overall market volatility. If we use VIX, the index commonly followed in the market for S&P 500 Index stock options, as the proxy for a forward-looking estimate of volatility, we can see that cross-sectional volatility of the S&P 500 and overall market volatility are coincident indicators of risk (see Figure 3).
A manager holding the same active positions vis-à-vis the benchmark index throughout the period we have shown here would see its tracking error fluctuate in relation to the changes in cross-sectional volatility. In other words – and we believe this is critical – while using tracking error as a measure of the manager’s active risk is valid, such a metric does not tell you whether a change in tracking error was caused by a change in the manager’s investment style or simply by changes in market volatility.
As Figure 4 illustrates, over time the relationship between tracking error and cross-sectional volatility seems relatively stable for the (theoretical) median manager.
Cutting out the market noise: active share
In 2006, Yale researchers Antti Petajisto and Martijn Cremers published an article on a new measure of active management called active share. Active share is based on the actual market value deviations from a manager’s benchmark and therefore is independent of the prevailing market volatility as measured by the benchmark index. Focusing on portfolio holdings and weightings rather than returns, active share can simply be defined as the percentage of a portfolio that is different than the benchmark. Active share measures range from 0% (entirely passive, with positions identical to the benchmark index) to 100% (highly active, with positions entirely different from the benchmark index).
While the original paper discussed the use of active share for identifying managers that had the potential to outperform, active share can be used for monitoring the existing manager roster as well. In conjunction with tracking error, active share can help provide a more accurate picture on whether managers are conforming to their style and taking enough active risk to warrant keeping them as part of the overall portfolio construction.
Interpreting active share
The basics of interpreting an active share number are simple enough. Below 60% active share, managers are considered closet indexers. From 60% to 80%, managers are defined as moderately active, and above 80% managers are seen as highly active (see Figure 5).
In evaluating your manager lineup, 60%, then, is the first key hurdle – the minimum active share required to be considered “active.” In our view, it rarely makes sense to hire managers with low active share, as it is difficult for these managers to outperform net of fees. We believe investors are better off hiring truly active managers or outright passive strategies rather than investing with these closet indexers.
Beyond this basic categorization, however, some judgment is required to determine the adequacy of active risk: Is a manager’s 65% active share sufficient? How about 75%? Or 85%? An acceptable level of active share depends on several factors. One guiding principle may be to compare a manager’s active share against its own historical record and that of similar managers within the sub-asset class.
For example, a small cap manager benchmarked to the Russell 2000 and a large cap manager benchmarked to the S&P 500 may have very different active shares due to structural differences in benchmarks and the way those managers approach portfolio construction. Based on the number of index holdings, the percentage in the top 10 and the weighting of the largest holding, it is clear the Russell 2000 is a less concentrated benchmark than the S&P 500.
The differences in these benchmarks directly affect a manager’s ability to take active risk. Consider a situation in which a portfolio manager has positive research opinions on large benchmark positions. For example, let’s assume a small cap manager has high conviction in the benchmark’s largest holding, Ocwen Financial, and initiates a 4% portfolio weight. Because Ocwen is 0.31% of the Russell 2000 (as of 31 Dec. 2012 – see Table 1), this position alone represents a 3.69% active weight and therefore contributes 1.85% to the portfolio’s active share. By contrast, a large cap manager’s 4% portfolio weight in the benchmark’s largest holding, Apple, results in just a 0.03% contribution to active share. To achieve the same active weight as the small cap manager, the large cap manager would need a 7.63% weighting in Apple, a relatively large single-stock position. Since the small cap manager takes less portfolio “capacity” to achieve the same active weight, it is ultimately easier to achieve a higher active share without incurring potential concentration risk. As a result, managers in different equity asset classes and with different benchmarks should be held to different active share standards.
Another key variable in interpreting active share is the plan’s overall allocation between active and passive strategies. In such an analysis it is helpful to look at the weighted average active share of the portfolio’s active managers. If the plan is primarily allocated to active strategies, then an average active share in the “moderately active” range may be appropriate. However, if the plan has a meaningful allocation to passive – as is often the case in a classic “core-and-satellite framework” – then the plan should demand higher active risk from its active managers. This example reinforces the judgment element of evaluating portfolio structures, as the degree of active risk may depend on a plan’s risk tolerance and risk budget.
Putting it all together: active share in the monitoring process
Given the importance of active share in understanding risk and the potential for outperformance, we believe it should be monitored over time (see Table 2). While tracking error is driven in part by market volatility, active share is a more direct expression of a manager’s willingness to take benchmark risk. As this is largely a function of a manager’s philosophy, active share should be relatively stable over time.
Changes in active share that constitute a clear trend could indicate a change in a manager’s investment discipline. So setting a range of expectations for active share for each manager may help identify circumstances where additional evaluation of a manager’s investment strategy may be warranted. We believe that an annual review of quarterly active share numbers should be sufficient to identify these potential changes.
Again, managing a portfolio of managers and investment strategies means understanding risk and the potential for return generation as well as unintended biases. While tracking error has been held as a key measure for active risk, it may include elements that reflect market conditions rather than managers’ actual decisions on risk. We believe active share, as perhaps the best additional objective measure of active risk, should be a critical part of this equity toolkit.