Historical in-sample results
In addition to analyzing risk-adjusted returns (in that the analysis is examining returns after controlling for specific risk factors), we want to highlight identifying characteristics of our sample funds by showing the distribution of the sample across the seven current Morningstar subcategories (excluding funds that may have moved categories over time) [Figure 2]. Funds that were liquidated or merged are represented in the category that they occupied at the time of the liquidation/merger.
It should be no surprise that more than one third of the funds are in the Allocation — 50% to 70% Equity category because most investors have a risk tolerance that translates into holding a portfolio that has, on average,a 60% equity allocation. More broadly, two thirds of the sample fall into three categories: the 30% to 50%, 50% to 70%, and 70% to 85% Equity allocation categories. Many funds in these categories are default options in retirement plans and are formally called qualified default investment alternatives (QDIAs). A far smaller portion of the funds fall into extremely conservative (i.e., 15% to 30% Equity) or very aggressive (85%+ Equity) categories.
Additionally, Morningstar's enhanced return data for U.S. categories can produce a separate return stream to represent the category average performance. The enhanced return data makes adjustments for new funds, for fund liquidations and mergers, and for funds that may have changed categories. These adjustments help to correct for survivorship bias and provide a better approximation of the true historical performance of the category.
The 30% to 50%, 50% to 70%, and 70% to 85% Equity allocation categories again stand out in the overall sample in terms of age (in months) and asset size (in millions of dollars) [Figure 3]. The World Allocation category, though newer, also sports substantial average net assets. A closer examination of average fees across the seven categories shows that many are tightly dispersed between 80 and 87 bps in the stated expense ratio. It also confirms that investors are typically paying higher expenses for funds with higher equity allocations that are also often perceived to be more active (i.e., Tactical Allocation and World Allocation), and paying less for funds that have more fixed income exposure (i.e., Allocation — 15% to 30% Equity).
Regression results for category averages identify alpha
In addition to applying our return-generating model to test the full sample, we also test each category. A new alpha term is calculated for every month for the five-year rolling return windows from December 1990, or whenever a category had sufficient return history, through December 2016 [Figure 4]. This creates a time series of up to 312 highly overlapping regression parameters.
At first glance, these results confirm several widely held beliefs. First, the category averages have a sensitivity to the Market factor that is consistent with the stated equity allocation and, in most cases, falls comfortably into the stated equity range. For example, the Allocation — 50% to 70% Equity category has an average Market factor weight of 0.64. Next, it appears that most managers also maintain exposure to small-cap stocks and have a bias for value over growth, as demonstrated by positive average values for both the SMB and HML factors. It also appears that on average, the return generating model we use has a high degree of explanatory power across the seven category averages. This is supported by the high regression r-square values in Figure 4.
Most importantly, the average monthly net-of-fee alpha is positive for the majority, but not all, of the category averages. Four of the seven categories have rather large positive average monthly alpha values, while three (Allocation — 15% to 30% Equity, Tactical Allocation, and Allocation — 30% to 50% Equity) have small, insignificant, and even slight negative average monthly alpha. Despite the latter three categories, it is fair to say that the average monthly alpha for the overall multi-asset fund sample can translate into a mean annualized net-of-fee performance differential. The 11 bps of monthly alpha in the Allocation — 70% to 85% Equity category, for example, can translate into approximately 1.36% of annualized net-of-fee, spendable alpha.
Alpha over time for sample of multi-asset managers
The average alpha return for each of the 1990–2016 rolling periods is plotted as a time series along with dashed lines indicating the average alpha over the entire time series [Figure 5]. Both sample groups — the trimmed sample and the 10% tails sample — are shown.
These calculations produce several observations:
- The average monthly alpha is +9 bps, which translates into approximately 1.13% in annualized net-of-fee alpha.
- Results showing the median estimate yield similar results, and suggest that the average returns do not appear to be skewed by any large positive or negative outliers.
- Comparatively, the combined 10 percent tails of the sample had consistently more volatile results, with an average monthly alpha of -1 bps (-8 bps of annualized alpha).
In addition to considering month-to-month and long-term average alpha, it is also relevant to the active management debate to know how often alpha has been positive as a percentage of all periods. For the denominator, the sample had 2,483 funds and 312 monthly observations (12 months x 26 years); taking into account funds with history that does not span the entire sample period, the result was 337,487 individual monthly alpha observations. For the numerator, there were 194,767 alpha scores greater than zero. In total, multi-asset funds have produced positive alpha in approximately 58% of the five-year periods covered in this study.
This track record of generating positive alpha began to deteriorate in 2009. The rolling average 60-month alpha fell into negative territory in 2012 and continued to decline until very recently. The onset of this downturn in risk-adjusted performance coincided with the market bottom in 2009 and has persisted through the years of extraordinary monetary policies that have challenged active managers, as described in the introduction. It is worth noting that, during this time period, a well-documented shift in the correlation structure across asset classes occurred, which raised cross-asset correlations and made it more difficult for active managers to add performance through well-timed allocation shifts [Figure 7].
Plotting the rolling 5-year average pairwise correlation across six asset classes (inverted) helps to illustrate this relationship: (1) U.S. investment-grade bonds (Bloomberg Barclays U.S. Aggregate Bond Index), (2) international equities (MSCI EAFE Index), (3) high yield (Bloomberg Barclays High Yield Index), (4) U.S. large-cap stocks (Russell 1000 Index), (5) U.S. small-cap stocks (Russell 2000 Index), and (6) commodities (GSCI) versus the average multi-asset fund alpha from our sample. The 12-month lag illustrates that it takes some time for the change in correlation structure to flow through to the alpha-generating ability of the multi-asset managers.
Visually, these two series have a correlation of -0.54 to one another, with the average multi-asset fund alpha having a -0.63 beta (-11.2 T-stat) to the average 6-asset-class pairwise correlation. Furthermore, plotting a 60-month moving average of the correlation variable shows that it tracks the alpha variable even more closely. These two series have a correlation of -0.74 with one another, and the alpha term has a beta of -1.08 (-17.6 T-Stat) to the correlation moving average variable.
This illustration sheds light on a few issues. First, it confirms a relationship between the average cross-asset correlation and the alpha that multi-asset managers have produced. Second, there appears to be evidence that correlations have rolled over from their peak, and this has coincided with the average multi-asset fund alpha bouncing off its lows.
The positive monetary impact for investors
The major question for all investors, whether individuals or institutions, is the impact of active management on performance in dollar terms. The enhanced return data for each of the seven category averages can be used to calculate the results of a hypothetical investment in each category over the sample period, which can then be compared with a style benchmark using the same return generating model from the equation in Figure 1. While there are multiple ways to calculate a style benchmark, in this case, the style benchmark return in each month is calculated as the sum of the return of each factor (i.e., Market, SMB, HML, and BOND) multiplied by the average weight for the category. This style benchmark effectively represents an attempt to mimic the returns of a specific fund using the risk factors from the equation in Figure 1. In theory, it can serve as the portfolio an investor may hold if he/she were attempting to replicate the returns of the referenced category average.
Given that the biggest category by number of funds (35% of the sample fund universe) is the Allocation — 50% to 70% Equity category, we chose it as the proxy for the average multi-asset fund category return for comparison with the style benchmark. As shown earlier [Figure 4] the Allocation — 50% to 70% Equity category had an average monthly net-of-fee alpha of 6 bps. Clearly, over long periods of time, even an incremental increase in return can produce a meaningful impact. A $10,000 investment starting at the beginning of 1986 in the average fund would have grown to just over $96,000, while the same investment in the style benchmark would have grown to just under $79,000, a $17,000 difference. That $17,000 difference is true net-of-fee "spendable alpha" for the investor, thanks to the impact of compounding returns.