Most factor-based strategies—commonly called Smart Beta—have hundreds of holdings and high overlap with their market benchmark. The far more powerful way to apply factors is to use them first to avoid large chunks of the market and then build more differentiated portfolios of stocks with only the best overall factor profiles. While not as scaleable as smart beta, this alpha-oriented approach has led to much better results for investors.

***

Professionally managed investment strategies have two components: an investing component (seeking alpha) and a business component (seeking assets). Outperformance is one goal, scale is another. Factors—like valuation, momentum, quality, and low volatility—have been largely applied by firms with the business in mind. In the asset management business, two variables matter: fees and assets. Smart Beta has risen to prominence alongside index funds and ETFs, and indexing has significantly reduced fees across the industry. With fees lower across the board, scale becomes a more important consideration for asset managers when deciding what strategies to offer the investing public. When fees fall, assets need to rise. For assets to rise across a business, the strategies offered need to be able to accommodate more invested money.

More assets may be good for the business, but it’s often bad for returns. As the level of assets under management rise, the investable universe of stocks shrinks, trading impact costs rise, and the potential for alpha erodes. In asset management, we find diseconomies of scale—as mutual funds and hedge funds get larger, their performance tends to suffer[i][ii].

Of course, fees matter a great deal, and the nearly free access to broad market index funds is a wonderful thing. But management fees are one thing, and key factors like valuation another. I would rather pay 0.75% for an S&P 500 index fund trading at 12 times normalized earnings than 0.05% for the same market trading at 25 times earnings, which it does as of April 2016. To have a large advantage versus the market—on factors like valuation or shareholder yield—you must build strategies with an emphasis on alpha and the consistency of alpha above all else.

To achieve what we call factor alpha, we believe that investors should use multiple, unique factors to build a more concentrated portfolio of stocks (as few as 50) with the best possible factor profiles. That means not owning wide swaths of the market. Relative to Smart Beta, a focus on factor alpha allows for better returns and significantly better factor advantages. In the rest of this post, I explore the dangers of scale and widespread adoption of any strategy and offer an alternative solution for using factors in the investment process.

Watch Scale Eat Returns

The difference between any portfolio and the market is determined by 1) what stocks you own and 2) how you weight those stocks in a portfolio. To show the impact that these two variables have, we start with the constituents of the S&P 500[iii] and create different portfolios based on a single factor that has worked well historically—valuation[iv]—to demonstrate the effect of moving further and further away from the index. We use this basic example not to recommend this as a strategy but rather to show the effects of both concentration and weighting scheme.

What we tested

I show three versions of this strategy.

• The first sorts all stocks in the S&P 500 on each date by valuation and portfolios of between 50-500 stocks (so the 50 stock version would be the 50 cheapest stocks on that date, and so on). Positions are equally-weighted (e.g. 2% each in the 50 stock portfolio or 1% each in the 100 stock portfolio).
• The second takes the same portfolios with the same stocks (between 50-500 holdings) but weights the positions according to market cap. This method can create very top heavy weightings in the more concentrated portfolios (e.g. IBM at 11.3% of the most recent 50-stock portfolio).
• The third forms the portfolios using a market-cap adjusted valuation factor[v], which multiplies a stock’s weight in the index by its relative valuation. This cap-adjusted value factor rewards companies that are big and cheap and penalizes companies that are small and expensive. Again we use the factor to build portfolios between 50-100 stocks. This is the most scaleable version of the value strategy whose holdings look a lot like major value indexes.

What we found

Here are the results, highlighting two key variables. First, the average forward 1-year excess returns versus the S&P 500 and second the average active share (i.e. the percent of the portfolio that is different from the S&P 500; higher means less overlap with the index).

What we learned

From this simple exercise, we learn the following:

• Concentration and equal weighting lead to portfolios which have better average excess returns and higher active shares.
• The equal weighted portfolios outperform cap-weighted and cap-adjusted value portfolios by an average of 1.8% and 2.0% per year, respectively—a wide margin in the U.S. large cap market.
• More concentrated portfolios have a much better valuation edge: stocks in the portfolio have much cheaper average value percentile scores.

Why this happens

Value (measured by something like a price-to-earnings ratio) is just another way of saying “market outlook.” A low relative valuation for IBM means that the market is pessimistic—relative to other stocks—about IBM’s future. We believe that value works over time because markets becometoo pessimistic about these stocks. Pessimism is good—and the lower the P/E, the more pessimistic the market.

Now, notice the trend in the price-to-earnings ratios (figure to the right) for the different portfolios today (the equally weighted versions). If market pessimism (low P/E) signals an opportunity, then the opportunity clearly grows as the portfolio gets more and more different from the S&P 500.

It helps to see what these simple value portfolios would hold today. Here are the top ten holdings for each of the 50-stock versions of our value strategy, along with each portfolio’s weighted average market cap[vi]. You can see, as you move left to right, that the top ten look more and more like the overall market, because the market caps get bigger and bigger.

Size vs. Edge

If you are a value investor, or a quality investor, or a yield investor, a key question is: how big is your edge versus the overall market? If you believe in price-to-book, your goal should be to achieve significant portfolio discounts versus the market on that measure.

But the most notable difference across the above portfolios is their size. When market cap is used as a variable when building a portfolio, it obscures any other edge that exists. Exxon Mobile has a price-to-book of 1.95x but is a huge company so has a weight of 3.6% in the Russell 1000 Value index. Seadrill—a maligned energy stock—has a price to book of 0.18x. So it is much “cheaper” than Exxon, but it’s only a $1.6B company, so its weight in the index is 0.01%, which may as well be zero.

If Exxon went up 40%, it would push the overall index up 1.44%. If Seadrill went up 40%, it would push the index up 0.004%. Seadrill would have to go up 14,400% to have the same impact as Exxon going up 40%.

For the Russell 1000 Value, Exxon is the far more important stock. But if you cared more about value than size, then the weights would be very different. Exxon is the biggest stock, but it’s only in the 50th percentile when sorted by price-to-book instead of by market cap. Seadrill is in the cheapest percentile.

From the perspective of an active investor, this is very odd because cheapness should matter much more than market cap when deciding a stock’s weight in a portfolio.

Different Means More Potential

If you weight stocks based on market cap—like the Russell 1000 Value does—then your portfolio will always have a lot of overlap with the market (a low active share). With a lot of overlap, there is only so much alpha you can earn.

Active share—our preferred measure of how different a portfolio is from its benchmark—is not a predictor of future performance, but it is a good indicator of any strategy’spotential alpha. The chart to the right shows why. On every date through history (1962-2015), we bestow ourselves with perfect foresight so that we can build portfolios that will achieve the highest possible 1-year forward excess return at each level of active share between 0% and 90%. We use only stocks in the S&P 500 itself and allow a maximum position size of 5% in the portfolios, to avoid piling into just a few of the best-performing stocks. This chart shows the maximum possible excess return (average of all historical periods) vs. the S&P 500 by active share level. This potential cuts both ways, so we also show the worst case scenarios by active share level. At any given level of active share, the “potential” excess return skews more to the positive (dotted-line).

More active portfolios have more potential for excess than less active portfolios[vii]. No one has perfect foresight, so nobody achieves alpha like this with any consistency. But these “best case scenarios” show the power of being different.

Factors like value have been a good way of putting yourself on the positive side of this potential curve above. Indeed, we saw the same thing with our different value portfolios: the more concentrated portfolios (higher active share), the better the average results. What is amazing is that in the early 1980’s, well over half of mutual funds had an active share above 80%, but as of 2009, only 20% or so of funds were this active. Large, scaled up institutions now control a majority of the market. In 1950, between 7-8% of the market was managed by large institutions. In 2010, that number was 67%[viii].

Costs Matter

So far we’ve discussed paper returns only. We saw that—gross of costs—returns to the S&P 500 value strategy get worse as the number of stocks grows. Now let’s look at the same issue net of trading costs. Simple trading commissions are a real cost, but our focus here is on market impact costs, which matter more for big asset managers. Once you start getting too big (trading billions of dollars), your trading moves the price of the names you are buying and selling—you pay a higher price when buying and get a lower price when selling than if you were an individual trading a $100,000 account. Market impact is a cost that doesn’t get enough attention, because end investors can’t see it, and asset managers don’t report it. The table below shows how annualized market impact costs grow with assets (these are estimates, skilled traders can beat these estimates by a little or a lot).

To show the cost drag, we expand our universe to the Russell 3000, which includes small- and mid-cap stocks where market impact is even more important. We again build portfolios based on valuation that have between 50 and 3000 stocks, and assets under management between $50MM and $50B. We rebalance these portfolios on a rolling annual basis, so the holding period is at least one year for each position. It is important to note that the more concentrated equal-weight portfolios have more exposure to smaller-cap stocks, so the impact numbers are higher than if we performed the same analysis on the S&P 500 universe.

The cost estimates reported in the table[ix] are based on 5 years of simulated trading in the actual value portfolios between 2010 and 2015. These are based on actual market conditions, not hypotheticals. We’ve highlighted the point at which impact (annualized) crosses 1%. For cap-weighted portfolios, you reach $30 billion in the most concentrated portfolio before crossing 1% impact costs.  But in the equal weighted portfolios, you reach the 1% threshold much more quickly in the more concentrated portfolios. We’ve already seen that equal weighting and concentration have delivered better results. This table proves that the more concentrated value portfolios cannot accommodate the kind of scale that large asset managers are after. If you are seeking alpha, you’d equal weight and you’d be willing to have fewer names in the portfolio. If you were seeking assets, you’d do what the industry has done: build broader smart beta indexes that focus on the large cap market or weight based on market cap.

Back to Smart Beta

We’ve seen that scale and excess return are mortal enemies. As Buffett said in his 1994 letter to shareholders, in which he warned of lower future growth rates for Berkshire Hathaway, “a fat wallet…is the enemy of superior investment results.” Price-to-book was arguably the first smart beta factor and has likely suffered from its own popularity as a measure of value. Hundreds of billions are invested based on price-to-book, but we’ve watched it deteriorate since 1993, when Fama and French first held it out as the defining value factor:

Science fiction master William Gibson wrote in his book Pattern Recognition that“commodification will soon follow identification.” In another passage, also fromPattern Recognition, Gibson is talking about clothes, but he could be talking about Smart Beta ETFs:

This stuff is simulacra of simulacra of simulacra. A diluted tincture of Ralph Lauren, who had himself diluted the glory days of Brooks Brothers, who themselves had stepped on the product of Jermyn Street and Savile Row, flavoring their ready-to-wear with liberal lashings of polo knit and regimental stripes. But Tommy surely is the null point, the black hole. There must be some Tommy Hilfiger event horizon, beyond which it is impossible to be more derivative, more removed from the source, more devoid of soul.

Smart Beta is the commodification of the most common historically proven factors. By definition, a commodity must be widely available. In asset management, that means it must be able to accommodate lots of invested money. We haven’t seen many active strategies with hundreds of billions of dollars behind them consistently beat a simple market index. Even Warren Buffett has been slowed—though not stopped–by scale.

Factor investing has huge potential benefits. Factor investing strategies tend to be cheaper than traditional active management. Properly managed, factor-investing strategies are also very disciplined. But, if a given strategy can accommodate $100B in assets, you may want to look elsewhere. Always avoid saturated strategies. For most of history, the factors behind Smart Beta strategies weren’t big targets. Now they are. Beware of popularity, beware scalability, and beware newly accepted “measures” of a strategy or idea. Too often, popular measures become targets and then lose their meaning and their edge.

Factor Alpha

The philosophical roots of the factor-alpha approach are notably different than those of Smart Beta.

First, what you don’t own matters. If Apple or Microsoft don’t look attractive, we believe you should own none of either in your portfolio. We start with a weight of zero in every stock, not with the market weight. Stocks are guilty until proven innocent. This naturally leads to higher active share and a portfolio with a greater overall potential for alpha.

Second, alpha comes from the relative advantage a portfolio has versus the market measured across key factors. Greater spreads—like bigger discounts or higher shareholder yields[x]—have led to better excess returns through time. Portfolios should focus on just the stocks with the best factor profiles. To achieve these big factor advantages, portfolios should be more concentrated than has become typical for smart beta strategies.

Sample the most popular Smart Beta ETFs and you’ll find the opposite: high overlap with the S&P 500 or Russell 1000. USMV, the popular “low volatility” ETF has an active share of 46% to the S&P 500. That means it has more in common with the S&P 500 than does an equal-weighted version of the S&P 500 with an active share around 50%. Strategies designed for factor alpha often have active shares higher than 80%. They are still well diversified, but more diversification is not always better.  In the case of factor investing, diversification often means diluted factor exposures. If factors work best at their extremes, then diversification means moving away from your edge and towards the market return.

Often, a picture tells the story better than words can. Below, we show the unmistakable difference between popular smart beta approaches vs. the factor alpha approach. We recreate the spirit of a Morningstar style box, but instead of using market cap and value vs. growth as the dimensions on the chart, we instead use the factors that we’ve found to be most predictive of future excess return: shareholder yield and quality (where quality is a combination of valuation, earnings growth, earnings quality, and financial strength).

The goal is to show where the different portfolios plot on the yield and quality continuum. A portfolio in the lower left would have the strongest relative readings on both quality and yield. The central dot in each circle represents the average current shareholder yield and quality readings for the portfolio, and the surrounding circle encompasses 75% of the portfolio’s weight. The trend is clear. As you move from the broad Russell 1000 (in grey), to the Russell 1000 Value (in green), to the biggest “fundamental index” smart beta approach (in orange), what you see is a tilt towards factors, but with very broad exposure. The factor alpha approach is entirely different, by design: a much better and tighter exposure to the key factors.

We can get even more granular a visualization: by plotting the position (good to bad on quality, vertical axis, and shareholder yield, horizontal axis) and portfolio weight (size of the circle) of every stock in a variety of popular ETF smart beta strategies: S&P 500 (SPY), Russell 1000 Value (IWD), Fundamental Index (PRF), and Minimum Volatility (USMV). Compare these with the final chart: the factor-alpha approach, and you can see the clear difference in portfolio construction and position in the shareholder yield and quality themes.

There are some outliers in the factor-alpha approach: these are positions whose factors profiles are deteriorating. The one of the far right, for example, is XL group, which had a strong shareholder yield until it issued a big chunk of shares for an acquisition last year. As the strategy rebalances, these positions will be sold down in favor of stocks in the lower left corner.

Wonky calculation note: these all use different starting universes, so percentiles are calculated on the broadest set of investable U.S. stocks, about 2,500 or so. This is why you see a slight skew towards higher shareholder yield everywhere in these charts: these are all large cap-ish strategies which have higher dividend and buyback yields than smaller cap companies which are included when calculating the percentages but often not shown on the chart.

Into the Future

Indexes and Smart Beta factors are affected and changed by asset flows into strategies which target those indexes. Hundreds of billions of dollars flowed into low price-to-book strategies, and price-to-book has suffered as a result[xi]. Fund flows affect everything.

Mark Twain said, “I was seldom able to see an opportunity until it had ceased to be one.” We often become aware of market strategies only after they’ve been identified, commodified, and scaled away. Smart Beta factors are a commodity. There is an ETF for everything, from value, to obesity, to put writing. When making an investment, consider the motivations of the manager or sponsor company—are they oriented toward gathering assets or earning alpha over time? If you believe in value, is a market-cap weighted value portfolio–where, for example, Exxon is your top holding despite being in the 50th percentile by price-to-book– really the best expression of that factor? Factor alpha has won for investors in the past.

Nothing is perfect. Other quants will disagree, citing the fact that fewer positions likely means higher tracking error, lower information ratios, and so on. I believe that as asset managers continue to put out hundreds of smart beta strategies, that a more differentiated approach will continue to win in the future. My suggestion is: if you believe in factors enough that you are willing to move away from the simple, cheap, market portfolio, don’t do it via tilt. Do it in a way that gains true exposure to factors. A neat parting idea: consider blending a 5-basis point S&P 500 or Russell 1000 fund with a truly deep value portfolio (or some other factor). The combination might get you similar overall factor exposures as a smart beta option, but at a cheaper price.

Always always always, ask yourself: is this strategy about alpha, or about assets?

[i] Does Fund Size Erode Mutual Fund Performance? The Role of Liquidity and Organization By JOSEPH CHEN, HARRISON HONG, MING HUANG, AND JEFFREY D. KUBIK*

[ii] How AUM Growth Inhibits Performance By Andrea Gentilini

[iii] Prior to 1990, we use the top 500 stocks by market cap to represent the S&P 500 rather than its actual constituents which are not available.

[iv] Value defined as sales/price, earnings/price, ebitda/ev, free cash flow/ev, and shareholder yield, weighted equally

[v] valuation percentile * weight in the S&P 500, think of it like a “contribution to total cheapness”

[vi] Sorted by value ranking for equal weighted and cap-adjusted value portfolios

[vii] The potential works both ways: more active portfolios also have more potential for underperformance.

[viii] https://www.sec.gov/News/Speech/Detail/Speech/1365171515808

[ix] ITG TCI analysis

[x] Shareholder yield = dividend yield + net buyback yield (percent of shares outstanding repurchased, net of any issuance, over the past 1-year period.

[xi] There are, or course, issues beyond flows that affect factor performance