Smart beta might not be all that smart.
Academic researchers have scoured market history looking for attributes in stocks that give them the ability to outperform. In their quest, they have found hundreds of investment factors. Yet only a handful of them are used within smart beta ETFs—most notably those that take into account a company’s size, value, quality/profitability, momentum and volatility.
Michael Markov, the founder and CEO of Markov Processes International (MPI), a Summit, N.J.-based global investment research firm, says financial advisors should be mindful of the fact that only a handful of factors have been effective, and that not every smart beta product will perform well in every market.
In his criticisms, he joins Rob Arnott, the chairman and CEO of Research Affiliates, who has in the past said that a blindfolded monkey throwing darts at a board of stocks could replicate many of the benefits of smart beta. His argument is that if you choose a random subset of stocks from a capitalization-weighted benchmark index, you will invariably tend to tilt toward smaller capitalizations and more value-style stocks than the benchmark index.
And according to an analysis by Markov’s firm, many smart beta stock portfolios make only small deviations from the benchmarks, and randomness indeed explains much of the difference in the returns generated by different factor exposures.
“In reality, these products are not any different from managed products being offered 20 to 30 years ago,” Markov says, “except they’re accessing these market anomalies that investors believe will produce returns via an automated process.”
He says that, in the end, investors should question how much different smart beta products really are from the cap-weighted benchmarks.
His firm’s analysis places smart beta ETFs within a style box on a four-quadrant grid, with axes defined by a product’s correlations with market capitalization-weighted or equal-weight indexes, and by the product’s tilt toward value or growth styles. Equal-weight and market-cap weight versions of the Russell 1000 Value and Growth indexes define the corners of the grid.
According to the grid, an index more similar to market-cap weighting has more concentrated stock positions, while an index resembling equal weighting is dispersed.
Rather than studying the holdings, Markov used William Sharpe’s method of tracking a portfolio’s style exposures by comparing its returns to those of benchmark indexes.
Markov placed seven smart beta ETFs on the grid. Four of them bore some resemblance in composition to a benchmark, cap-weighted growth index at the time of the analysis—indicating they were concentrated in particular areas of the market, or performing similar to their benchmarks even though they’re advertised as fully differentiated “smart” products. These included the iShares Edge MSCI USA Momentum Factor ETF (MTUM), which tracks a momentum-weighted index; the PowerShares S&P 500 Low Volatility ETF (SPLV), which tracks a low volatility weighted index; the iShares Edge MSCI USA Quality Factor ETF (QUAL), a quality-weighted ETF; and the Guggenheim S&P 500 Pure Growth ETF (RPG), which tends toward indexes with more growth-style stocks.
Among the other three funds on the grid, the one most resembling a cap-weighted value index was the PowerShares FTSE RAFI US 1000 ETF (PRF), which uses a Research Affiliates fundamentally weighted index.
Two funds did stand out from the benchmarks, however. The PowerShares S&P 500 High Beta ETF (SPHB), which tracks an index weighting toward beta-sensitive stocks, and the Guggenheim S&P 500 Pure Value ETF (RPV), which tracks an index that gives more weight to value-style stocks, both by and large resembled an equal-weight value index. At the time of the analysis, they were highly differentiated from their related market-cap benchmark indexes and were not concentrated in a particular area of the market.
A portfolio’s resemblance to an equal weight index, in Markov’s analysis, is due to the randomness of its behavior. A market cap index, meanwhile, at least takes into consideration the size of its constituents.
“An equal-weight index is an index with no information whatsoever on any stock,” says Markov. “It’s a ‘monkey portfolio’ which provides no purposeful exposure to any factors, but by virtue of its allocation and rebalancing does capture the risk premium related to smaller companies.
“Market-cap portfolios, on the other hand, provide exposure to market beta, which is itself considered a factor by academic researchers.”
Tom Goodwin, the senior research director at FTSE Russell, says it’s not exactly fair to call an equal-weight index “dumb” because even though equal weighting doesn’t measure a factor exposure, it is used by advisors to passively access both the size and value factor premiums.
“It feels like [Markov] is trying to say that as an index trends closer towards equal weighting, it becomes less smart,” says Goodwin. “To me, the only thing he’s really measuring is concentration. There’s no way to create an index less concentrated than equal weight, while cap-weighted indexes tend to be very top heavy. We know that most factor indexes have less concentration, but to varying degrees. It depends on the underlying strategies.”
Science And Statistics
Keep in mind that equal weighting the S&P 500 does not change the fact that it’s an index whose components are drawn from the largest companies listed on U.S. indexes. Even a weighting scenario that favors the smaller components in the S&P 500 fails to create the exposure to small- and mid-cap companies that you’d achieve by equal-weighting the larger Russell 1000.
Markov says we can’t infer anything about the effectiveness of a smart beta ETF just because its holdings resemble a market cap or equal-weight benchmark at a specific point in time. And he isn’t questioning how effective these ETFs are. It may be mere coincidence that some ETFs resemble benchmarks in the style box analysis. He notes that his analysis represents a snapshot of ETF construction, and that different factor ETFs may reallocate through the markets differently over time, with their indexes at any point capable of being represented as a blend of equal and market-cap weightings that tilt toward growth or value.
An ETF’s plot point on the style boxes describes the blend of the four benchmarks that approximate its performance at the time of the analysis. So the performance of an ETF plotted dead in the middle of the style box could be approximated by allocating to each of the benchmarks equally. Thus, the performance of an ETF plotted three-quarters of the distance to an equal-weight value benchmark from a market-cap value benchmark could be approximated by allocating 75% to equal-weight value and 25% to market-cap value. On Markov’s style box, the performance of the fundamentally weighted PowerShares FTSE RAFI US 1000 ETF roughly corresponds to the performance of an 80%/20% blend of the market cap-weighted and equal-weighted Russell 1000 Value indices.