A valuable risk measure for high-yield portfolios
Active portfolio managers are increasingly being asked to demonstrate how active they really are, i.e., the degree to which they are taking different positions from the benchmark. Active management essentially takes two forms: unique exposure to systematic factors or idiosyncratic security-specific risk. In a fixed income portfolio, relative performance can be generated in a variety of ways. For example, a portfolio can tilt toward long- maturity securities in anticipation of falling interest rates, tilt toward lower-quality bonds to capture incremental yield, favor certain sectors over others or favor certain securities in lieu of others within a particular sector.
The relative risk of an active portfolio has typically been measured by tracking error — a portfolio’s expected or realized standard deviation from a benchmark return over time. In the case of a high-yield portfolio, the investment process is generally more focused on security selection than on adopting systematic differences relative to the benchmark. Given this emphasis on credit risk, tracking error may not be the most suitable measure for evaluating risk and alpha potential in high-yield portfolios. Measures that hone in on security-specific idiosyncratic risk may be more helpful. Active share serves as one such measure.
While the concept of active share has mainly been applied to equity portfolios to date, it can also provide valuable information on the drivers of risk and performance in credit- oriented fixed income portfolios. In this paper, we make the case for employing active share as a risk measure in the context of managing high-yield strategies.
Tracking error measures the variability of a portfolio’s return around a benchmark index. It captures the width of the distribution around the mean excess portfolio return. For example, a portfolio manager who delivers consistent outperformance of 1% per month will have an annualized mean return of 12% with no tracking error, given the absence in variance in the level of active return. Tracking error is consequently a volatility measure, not a performance measure.
Multifactor models are typically used to identify common factors that “explain” security returns and generate risk forecasts. Within these factor models, tracking error has two components: a systematic component related to common factor risks and exposures, and a specific component that represents idiosyncratic risk unexplained by the common factors.
While tracking error captures both systematic and idiosyncratic risk, for broadly diversified portfolios it is largely driven by systematic differences between a portfolio and a benchmark. High-yield managers are typically more focused on deriving alpha from superior security selection, and consequently they emphasize idiosyncratic risk rather than systematic or market risk. The management of common factor exposure will reduce the systematic component of tracking error. The result is that tracking error frequently understates the degree of “active risk” a manager is taking in a credit-oriented portfolio like high-yield debt. The minimizing of systematic risk, a clear investment objective of many high- yield portfolios, results in muted tracking error and the perception that the manager is not taking sufficient active risk vis-à-vis the benchmark.
Low tracking error can also be a function of the market environment: Risk models are rooted in history. Security factor exposures, factor return volatilities, factor return correlations and security-specific volatilities are based on historical experience, with risk models typically placing more emphasis on recent history. All else being equal, tracking error levels rise and fall with market volatility. There is an inherent lag based on the responsiveness of the models to the recent market environment.
Tracking error may not be the most suitable measure for evaluating risk and alpha potential in high-yield portfolios. Measures that hone in on security-specific idiosyncratic risk may be more helpful.
Tracking error estimates can also be impacted by changes in the correlation structure among risk factors; it is hard to make general statements due to the large number of pairwise relationships and the dependence on the direction of relative factor exposures. Shifts in volatility regimes can cause tracking error estimates to significantly overstate or understate subsequent realized tracking error. This was observed in the aftermath of the 2008–2009 global financial crisis, when tracking errors rose significantly after being at historic lows going into the crisis, following an extended period of low and declining volatility in the period preceding the crisis.
Tracking error signals can also lead to perverse outcomes, particularly in the case of high-yield portfolios subject to upper and lower bounds for tracking error. If tracking errors are low at a point in the credit cycle because of the volatility environment, the measure encourages high-yield managers to overweight potentially mispriced risk. The same is true in periods of stress. If an investor constrains tracking error with an upper bound, the manager’s market exposure is restrained when the market appears most attractive (e.g., in early 2009).
The traditional use of tracking error may very well force decisions we would argue are counter to success in managing high-yield portfolios. Also, given the asymmetry of the asset class, investment decisions based on these perverse signals create performance dispersion — underperformance — that is hard to overcome. In light of the limitations of tracking error as applied to high-yield portfolios, we make the case that active share is a useful complementary or alternative risk measure.
Active share is a measure of how different a portfolio is at the holdings level, based on the percentage of its holdings that differ from the benchmark index. A portfolio’s overall active share is the aggregate of its net long and short positions expressed as a percentage of all of its holdings. For example, if 25% of the portfolio weight is the same as that of the index and the remaining 75% is not, the portfolio has a 75% active share. One can also think of active share as 100% minus the overlap with the benchmark. In contrast to tracking error, active share, by placing equal weight on all active positions regardless of diversification effects, more directly captures the differences between a portfolio and a benchmark at the security level.
Active share has garnered considerable attention in recent years following the publication of seminal academic research conducted by Cremers and Petajisto.1 While subsequent studies have questioned some of their findings, particularly as they relate to predicting manager skill and performance, active share is increasingly viewed as a useful alternative or complement to tracking error.
The two gauges emphasize different aspects of active management: Active share serves as a reasonable proxy for security selection differentiation versus the benchmark, whereas tracking error is a proxy for systematic factor risk, though it does also take account of security selection differences.2 For example, a diversified high-yield credit manager can have a high active share because of large active positions in individual issues, in tandem with a low tracking error derived from minimal exposure to factor risks such as duration, sectors and quality.
While active share has not been commonly applied to fixed income portfolios, this metric can serve as a complement to tracking error, as it has with equity portfolios. This is particularly the case for segments of the market, such as high yield, where security selection is often a key driver of active returns. Making use of active share in the context of fixed income portfolios introduces some subtle complexities not encountered in equities, such as whether to measure exposure at the issue or issuer level.
Active share serves as a reasonable proxy for security selection differentiation versus the benchmark, whereas tracking error is a proxy for systematic factor risk, though it does also take account of security selection differences.
Exhibit 1 highlights some of the nuances of applying active share to a fixed income portfolio. The benchmark in this example consists of four bonds (issues) from three companies (issuers). The portfolio is invested in each of these three issuers, but not each of the bonds, and in one off- benchmark issuer/issue, with weights that differ from the benchmark weights.
One can make the following observations based on this example:
Active share at the issue level will be higher than at the issuer level. An overweight to one issue relative to another from the same issuer will contribute to the issue-level active share calculation, but not to the issuer-level active share calculation. For instance, in comparing Security Three from issuer Energy Co. C with Security Four from the same issuer, there is only a 25% overlap at the issue level, but a 35% overlap at the issuer level.
Heavily concentrated benchmarks make achieving high active share more difficult. Unless a portfolio manager is willing to meaningfully underinvest in a heavily weighted issuer (in the example portfolio, issuer Technology Co. B represents 40% of the benchmark), the manager will use up some potential active share by holding some exposure of the heavily weighted benchmark issuer for risk management purposes. One could infer from this simple example that the portfolio manager is unwilling to be more than 25% under- or overweight any issuer and thus uses up 15% of the potential active share with the 25% underweight position. If the portfolio manager wanted to be overweight this issuer, he or she would use up 40% of the potential active share from the overlap created.
More diversified benchmarks allow for higher active share. For this reason, we typically see higher active share in small-cap equity portfolios than in large-cap, all else being equal. Favoring smaller (and perhaps less liquid) issuers/issues will tend to lead to greater active share. In the portfolio in Exhibit 1, going overweight issuer Financial Co. A requires only a 10% overlap with the benchmark, whereas going overweight issuer Technology Co. B would require a 40% overlap — 30% less active share, all else being equal. In general, active share needs to be interpreted in the context of the type of portfolio and the concentration of the investment universe. Moreover, limited liquidity in some benchmark securities can also exacerbate concentration and mitigate active share.
Portfolio positions that are off-benchmark are pure active share. In the example portfolio, the 15% weight in issuer Consumer Staples Co. D represents an off-benchmark position that contributes its weight to the active share of the portfolio. Too many off-benchmark positions, however, can be indicative of a benchmark-agnostic investment process or an inappropriately specified benchmark.
If we take tracking error as a reasonable proxy for systematic factor exposure, and active share for security selection differentiation versus the benchmark, we can assemble a matrix showing different kinds of fixed income managers based on their level of active share and tracking error, as in Exhibit 2.3
A diversified high-yield credit manager (upper left quadrant) can be active despite the low tracking error when there is low exposure relative to the benchmark in terms of duration, foreign exchange, swap spread, volatility and sectors, for example. Conversely, a macro manager focused on systematic macroeconomic factors can generate a large tracking error even without large deviations from the index holdings (lower right quadrant). A credit/macro manager combines the two approaches by taking positions in individual credits as well as in systematic factors. A “closet indexer” scores low on both active dimensions while often claiming to be an active manager. A pure index fund has almost zero tracking error and active share.
The question then is how one might calibrate a reasonable level of active share in a high-yield portfolio. Petajisto suggests that active share of less than about 60% in a large-cap US equity portfolio is closet indexing.4 There is general acknowledgment, however, that the appropriate active share threshold is a function of the benchmark and portfolio concentration, as well as the degree to which the investment process and risk oversight allow a manager to be underweight large benchmark names. The more concentrated the benchmark, the harder it is to achieve a high active share; the more concentrated the portfolio, the easier it is to achieve a high active share; and the more constraints on the portfolio construction process relative to the benchmark, the harder it is to achieve a high active share.
In general, one might expect active share in high-yield portfolios to be lower than equities because of the asymmetric nature of fixed income returns. The negative skew of fixed income returns is a function of the asset class: In the best-case scenario, an investor’s principal is returned at maturity, along with the promised yield on the bond; in the worst case, the issuer defaults and the investor receives only a percentage of the principal due, based on the agreed “haircut,” or is not repaid at all. This return asymmetry argues for broader diversification, which creates more overlap and hence less active share.
The credit focus of high-yield portfolios and, hence, the larger degree of idiosyncratic security-specific risk compared with most other fixed income strategies, suggests that active share may be a particularly relevant risk measure for these portfolios.
If we consider the case of high yield relative to large-cap US equities, what then is an appropriate threshold for active share in a high-yield portfolio? As a way to approach this question, we examined a number of portfolio scenarios based on the Barclays US High Yield 2% Issuer Capped Index, which held 1,055 issuer names on 28 July 2015.
In the table in Exhibit 3, an active share range is shown for each of nine scenarios. These were calculated based on portfolios with 150, 200 and 250 issuers invested in three ways, on the proportion of names held among the largest 250 benchmark names (70%, 80% or 90%) and the proportion of the portfolio’s names randomly distributed among the smallest 805 benchmark names — the balance of the 1,055 issuer names and the top 250 issuer names (30%, 20% or 10%).
The top 250 benchmark names account for 70% of the weight of the overall benchmark, suggesting a reasonable amount of concentration among the top index names. The 250 largest names were chosen as the focus for these portfolios largely because liquidity constraints mean that as the size of a portfolio grows, the more it will be forced to own these larger names. See further details in the Methodology below.
For instance, as the table shows, a portfolio of 250 issuers, with 80% of those issuers held in the high-yield benchmark among the largest 250 benchmark names and 20% held in the smallest 805 benchmark names, would generate an active share in the region of 58% to 63%. The matrix indicates that active share can be increased along these two dimensions — the number of issuers and the proportion held in the largest 250 benchmark names — by reducing the number of issuers held in the portfolio and increasing the proportion of the portfolio held in the smaller names.
These scenarios show that in typical portfolio parameter ranges, one generally adds 3–5 percentage points of active share by increasing the proportion of the portfolio’s names held in smaller benchmark names by an additional 10 percentage points, and one adds 4–8 percentage points of active share by decreasing the number of issuers in the portfolio by 50.5
It should be noted that the levels of active share referenced in the high-yield portfolio scenarios would not necessarily apply in other segments of the fixed income market. For instance, a TIPS (Treasury Inflation-Protected Securities) or government/ mortgage portfolio would likely have a much lower active share than the high-yield examples above, in part due to the homogeneity of issuers in these portfolios.
Like equities, active share can provide valuable information on the drivers of risk and return in credit-oriented fixed income portfolios. This approach reveals that a diversified high-yield credit manager can be active despite a low tracking error.
Active managers are increasingly being called upon to demonstrate the degree of active exposure to systematic factors or idiosyncratic security-specific risk in a portfolio. While tracking error has been the traditional measure of relative risk of an active portfolio, it may not be the most appropriate measure in the case of high-yield managers focused on security selection. In these instances, active share may serve as a helpful complement — or even an alternative — to tracking error.
Like equities, active share can provide valuable information on the drivers of risk and return in credit-oriented fixed income portfolios. This approach reveals that a diversified high-yield credit manager can be active despite a low tracking error. This is consistent with a high-yield manager’s investment process, which frequently entails minimizing systematic risk while seeking to maximize returns from the security selection process.
1 Cremers, K. J., and A. Petajisto. “How Active Is Your Fund Manager? A New Measure That Predicts Performance.” Review of Financial Studies, 2009, vol. 22, no. 9: 3329–3365.
3 The methodology for this graphic borrows from a similar identification of four basic types of equity managers by Cremers and Petajisto in their 2009 paper (cited above).
4 Petajisto, A., “Active Share and Mutual Fund Performance.” Financial Analysts Journal, 2013, vol. 69, no. 4: 73–93.
5 This relationship is nonlinear, particularly at extremes.