How to Benchmark Target-Date Funds: A Case Study

Transcription

1 1 How to Benchmark Target-Date Funds: A Case Study Thomas Idzorek, CFA, President, Morningstar Investment Management Division Jeremy Stempien, Director, Investments, Morningstar Investment Management Division, Retirement Solutions We look at a popular family of target-date funds and compare them with the Morningstar Lifetime Asset Allocation Indexes, explaining how to interpret the performance of these important retirement vehicles. By their very nature, target-date funds are complex investments. This complexity makes them especially challenging to benchmark, with each fund containing unique attributes and features that ideally should be addressed individually based on the unique goals and situation of the potential investor. In this article, the assumption is that both a target-date fund family and a target-date benchmark have been selected (rightly or wrongly), and a target-date benchmark must now be used in the most appropriate manner possible. It is important to emphasize the word appropriate, because the use of target-date benchmarks for benchmarking target-date funds is significantly more nuanced than traditional benchmarking, requiring greater skill and interpretation. In all cases, the target-date benchmark should not be unambiguously accepted as good (unless the fund complex licenses a specific target-date index, in which case the glide path and detailed asset class weightings are accepted as given). As such, the performance of a target-date fund relative to a target-date benchmark is open to interpretation. It s important to understand the relative performance the active return relative to the benchmark and from there, determine meaningful insights. Understanding the causes of the relative performance of target-date funds is fundamentally more challenging than it is for traditional benchmarking. Additionally, the scope of the problem goes well beyond the benchmarking of a single fund. Rather, the goal is typically to benchmark all of the five-to-12ish target-date funds that make up a target-date fund family, a process that requires one to analyze and benchmark the 10-to-50 or so sub-funds used within the family in hopes of making an assessment of the performance of the overall fund family. A Case Study This case study focuses primarily on the complications associated with benchmarking the largest retail target-date fund family, the Fidelity Freedom Funds family, relative to the Morningstar Lifetime Moderate Allocation Index family. This analysis focuses on the Fidelity Freedom 2030 fund, but the same process is repeated for each of the other 11 funds that make up the Fidelity Freedom Funds family. Adjusting the Benchmark Figure 1 includes a variety of annualized statistics based on five years of quarterly data for each of the funds that make up the Fidelity Freedom Funds family, the corresponding target-date indexes from the Morningstar Lifetime Moderate Allocation Index family, and the corresponding differences from October 2006 to tember In practice, one would look at these statistics for various periods, such as 12 months, 60 months, 120 months, and since inception.

2 2 Morningstar Indexes Figure 1. Performance Comparison October 2006 tember 2011 The Morningstar Lifetime Moderate Allocation Index SM had a better annualized arithmetic return, geometric return, standard deviation, and Sharpe ratio than the respective Fidelity Freedom Fund over this time period. Arithmetic Geometric Standard Deviation Sharpe Ratio Number of Drawdowns Skewness Kurtosis Fidelity Freedom Morningstar Lifetime Moderate Difference Fidelity Freedom Morningstar Lifetime Moderate Difference Fidelity Freedom Morningstar Lifetime Moderate Difference Fidelity Freedom Morningstar Lifetime Moderate Difference Fidelity Freedom Morningstar Lifetime Moderate Difference Fidelity Freedom Morningstar Lifetime Moderate Difference Fidelity Freedom Morningstar Lifetime Moderate Difference Fidelity Freedom Morningstar Lifetime Moderate Difference Fidelity Freedom Morningstar Lifetime Moderate Difference Fidelity Freedom Morningstar Lifetime Moderate Difference Fidelity Freedom Morningstar Lifetime Moderate Difference Fidelity Freedom Income Morningstar Lifetime Moderate Income Difference Average Difference Source: Morningstar EnCorr.

3 How to Benchmark Target-Date Funds: A Case Study 3 Figure 2. Beta-Adjusted Performance Comparison October 2006 tember 2011 The beta-adjusted performance comparisons of Morningstar Lifetime Moderate Indexes and Fidelity Freedom Funds are closer than shown in Figure 1. This case study examines the Fidelity Freedom 2030 Fund s underperformance of the Morningstar Lifetime Moderate 2030 Index SM by 82 bps. Arithmetic Geometric Standard Deviation Sharpe Ratio Number of Drawdowns Skewness Kurtosis Fidelity Freedom Morningstar Lifetime Moderate Difference Fidelity Freedom Morningstar Lifetime Moderate Difference Fidelity Freedom Morningstar Lifetime Moderate Difference Fidelity Freedom Morningstar Lifetime Moderate Difference Fidelity Freedom Morningstar Lifetime Moderate Difference Fidelity Freedom Morningstar Lifetime Moderate Difference Fidelity Freedom Morningstar Lifetime Moderate Difference Fidelity Freedom Morningstar Lifetime Moderate Difference Fidelity Freedom Morningstar Lifetime Moderate Difference Fidelity Freedom Morningstar Lifetime Moderate Difference Fidelity Freedom Morningstar Lifetime Moderate Difference Fidelity Freedom Income Morningstar Lifetime Moderate Income Difference Average Difference Source: Morningstar EnCorr.

4 4 Morningstar Indexes Over this five-year period, in almost all cases, the Morningstar Lifetime Moderate Allocation Index SM had a better annualized arithmetic return, geometric return, standard deviation, and Sharpe ratio than the respective Fidelity Freedom Fund. While this suggests poor relative performance for the Fidelity Freedom Funds over this particular time period, before reaching this conclusion one must dig deeper. The final row of Figure 1 shows that, on average, the Fidelity Freedom Fund family underperformed the respective Morningstar Lifetime Moderate Allocation Index family by 181 basis points per year during this five-year analysis. Looking at the respective glide paths demonstrates where this difference comes from. This analysis revealed the varying level of equity exposure between the Fidelity Freedom funds and the Morningstar Lifetime Moderate Indexes. Almost all the Fidelity Freedom funds have less equity exposure than the corresponding target-date index. Different equity exposure levels would typically lead one to expect different risk and return characteristics between the target-date fund and its corresponding index. It is widely recognized that the returns of a benchmark should be riskadjusted, or more specifically, beta-adjusted so that the fund and benchmark in question have similar overall risk exposures (i.e., level the playing field). To do this, a rolling 12-month beta was calculated for each of the Fidelity Freedom Funds relative to their corresponding Morningstar Lifetime Moderate Allocation Index, and then a separate beta-adjusted return series was created using these rolling betas for each of the Morningstar Lifetime Moderate Allocation Indexes. In Figure 2, the Fidelity Freedom funds were compared to each, relative to a beta-adjusted version of the Morningstar Lifetime Moderate Allocation Index. In comparison to Figure 1, the beta-adjusted performance comparison in Figure 2 is much closer. For example, the average annual arithmetic return differential shrunk from 181 basis points of underperformance per year for the Fidelity Freedom Funds to 100 basis points of underperformance per year. Note the highlighted negative 82 basis points of underperformance of the Fidelity Freedom 2030 fund. From a target-date benchmarking decomposition standpoint, Figure 2 provides us with a starting point: a return differential or active return between the fund in question and the corresponding betaadjusted return of the benchmark that must be decomposed. More specifically, the quantity that must be decomposed and understood is the following: (Formula A) Fund Return 2 Beta-Adjusted Benchmark Return 5 Active Return The active return is then decomposed into four parts asset allocation, sub-fund alphas, glide path fee, and other using Formula B. Figure 3. Morningstar Lifetime Moderate 2030 Fund Active Return Distribution, October 2006 tember 2011 Active return attribution analysis for the Fidelity Freedom 2030 Fund leads back to the negative 82 bps annual return differential identified in Figure Morningstar Lifetime Moderate 2030 (Beta-Adjusted) Fidelity Freedom Active Return Morningstar Lifetime Moderate 2030 (Beta-Adjusted) Fidelity Freedom 2030 Effective Asset Allocation Return (a) Asset Allocation (b) Weighted Average Sub-Fund Alpha Fidelity Freedom 2030 Overall Fund Fee Weighted Average Sub-Fund Fees (c) Glide Path Fee (d) Other Source: Morningstar EnCorr.

5 How to Benchmark Target-Date Funds: A Case Study 5 (Formula B) Active Return 5 Asset Allocation a) Return of the Effective Asset Allocation Fund 2 (Beta3Return of the Benchmark) Sub-Fund Alphas b) 1 Weighted Average Pure (Net of Fee) Alphas of Sub-Funds Glide Path Fee c) 2 (Overall Fund Fee 2 Weighted Average Fee of Sub-Funds) Other d) 1 Other The goal now is to estimate the inputs for the decomposition formula and then decompose the active returns relative to the benchmark period by period. This return attribution/decomposition framework was applied to the Fidelity Freedom 2030 fund in order to better understand the annual arithmetic return difference of negative 82 basis points between Fidelity Freedom 2030 fund and the Morningstar Moderate 2030 (betaadjusted) series. Moving forward, quarterly data (corresponding to the quarterly reported holdings) was used, in an attempt to decompose the slightly more than negative 20 basis points of average quarterly underperformance. Like most target-date funds, the Fidelity Freedom Funds are funds of funds. When looking at the evolving allocations to the underlying sub-funds that have made up the Fidelity Freedom 2030 fund during the past five years, an allocation at each point can be identified. The fewest number of sub-funds used during the analysis period was 21 and the most was 26. To determine the effective asset allocation, each of the individual sub-funds was analyzed to establish their respective individual effective asset allocations and, eventually, the pure alphas for each sub-fund. As part a of Formula B demonstrates, the active return from asset allocation is calculated by comparing the weighted average performance of the effective asset class allocations of the Fidelity Freedom 2030 fund relative to the performance of the beta-adjusted benchmark. When looking at the evolving estimated effective asset allocation of the Fidelity Freedom 2030 fund based on holdingsbased style analysis and combining it with the quarterly returns associated with the appropriate asset class index, one is able to determine the period-by-period return of the fund s effective asset allocation. On average, over a five-year period, the effective asset allocations of the Fidelity Freedom 2030 fund added approximately 5 basis points of excess return per quarter. Sub-Fund Alphas Turning to the weighted-average alpha of the sub-funds, one should focus on part b of Formula B. Here the goal is to determine the part of active return attributed to sub-fund alphas. More specifically, weighted-average, period-by-period alpha is estimated for each of the sub-funds for the given target-date fund. In this working example, the details for the sub-funds that make up the Fidelity Freedom 2030 fund are shown. Earlier the term pure alpha was used to emphasize that these are not alphas relative to a single-factor benchmark, such as the S&P 500, but alphas relative to a custom, evolving multi-asset class benchmark. A returns-based style analysis is completed for each sub-fund using a rolling 36-month window to create the Average

6 6 Morningstar Indexes evolving custom multi-asset class benchmark. The period-byperiod alpha of each fund is then estimated relative to its own custom benchmark. Since, the entire Fidelity Freedom Funds family has used 40 underlying funds, each of which must be analyzed individually. Coupling the quarter-by-quarter alphas with the weight of each sub-fund in the Fidelity Freedom 2030 fund, the quarter-byquarter weighted average alpha can be calculated. This will provide the average quarterly alpha for each of the sub-funds, as well as the average for the weighted average quarterly alphas, which is negative 25 basis points. In contrast to the effective asset allocation decisions that added approximately 6 basis points, the sub-funds of Fidelity Freedom 2030 hurt performance by approximately 25 basis points per quarter. This analysis helps identify the quality of the underlying sub-funds, as each fund will be analyzed individually. Glide Path Fee The sub-fund alpha analysis was completed net-of-fees so that it could be determined whether or not the individual sub-funds were adding or detracting alpha after accounting for their individual fees. The underlying sub-fund fees were already accounted for. The weighted average quarterly sub-fund fees during this time period were 19 basis points. For the purchaser of a target-date fund, the most transparent number is the all-in fund fee. To avoid potentially doublecounting, in part c of Formula B, focus first on the all-in Fidelity Freedom 2030 fee (75 basis points per year, or per quarter), and then subtract from the all-in fee the weighted average fee of the underlying sub-funds that were already accounted for. One might think of this remaining fund fee as the glide path fee. For some fund families this glide path fee is extremely low or potentially negative, as the fund companies view the target-date fund family as a tool for driving assets to the sub-funds rather than an additional revenue source. The final part of Formula B, part d, represents other. It is a plug that allows the decomposition in each quarter to hold perfectly. It should be thought of as an error term resulting from estimation errors associated with parts a through c of Formula B. For the Fidelity Freedom 2030 fund, the average value for other is nearly zero, indicating that over time the errors are cancelling each other out. Putting the Pieces Together Figure 3 summarizes the active return attribution analysis for the Fidelity Freedom 2030 fund. Working with the return attribution framework developed in Formulas A and B, Figure 3 contains the following information: 3 Active Return 3 Asset Allocation 3 Weighted Average Sub-Fund Alphas 3 Weighted Average Sub-Fund Fees 3 Other The final column of Figure 3 contains the averages for the various quarterly series, respectively. Annualizing these quarterly active returns leads back to the negative 82-basis-point annual arithmetic return differential identified earlier in Figure 2. Conclusion The importance of target-date funds to America s retirement system is difficult to overstate. Yet, to date, the industry has lacked definitive guidance on how to benchmark target-date funds and target-date fund families. Frustratingly, benchmarkrelative comparisons will not tell you if the glide path itself is good. With that in mind, Morningstar proposes a comprehensive framework for decomposing the active return of a fund relative to a benchmark into four components: asset allocation, sub-fund alphas, glide path fee, and other. The process is both data- and labor-intensive, though necessary given the importance of target-date funds. K In the case of Fidelity Freedom 2030, the weighted average quarterly fee of the sub-funds comes extremely close to the basis point all-in fund fee per quarter, indicating that there is virtually no additional glide path fee, and the all-in fund fee charged by the Fidelity Freedom 2030 fund is simply the weighted average fee of the underlying sub-funds.