A search for all equity ETF's available to U.S. investors in Bloomberg leads to a list of 969 candidates, a surprisingly large number of options for a relatively new investment vehicle. Given that most focus on large capitalization stocks here in the U.S. (not all, but most), this means that there has to be overlap in the underlying stock holdings...in some cases a lot of overlap.

The unfortunate result is that investors are lured into believing that 1) their exists choice within ETF categories, and 2) building portfolios of ETF's represent a diversified set of exposures. This is generally not the case.

A distinction without a difference

When I wrote about my concerns with Low Volatility ETF's here, here, and here, I noticed that the turnover in those ETF's was very low, which suggested that the billions of dollars flowing into the low volatility theme were being channeled into a relatively concentrated group of stocks. It turns out, dividend ETF's are more popular by a factor of five, and there tends to be a lot of overlap in the underlying holdings.

As you can tell from the chart below, dividend ETF's are by far the largest "Smart Beta" ETF category, having amassed $126 billion in assets as of the third quarter of 2016.

The challenge with dividends as an investment factor, as opposed to Value, Momentum, or Quality, is that only so many companies pay dividends. As buybacks have become more popular in the U.S., the share of stocks paying high dividend yields has declined. The chart below breaks down dividend payers listed on the S&P 500 by their level of yield. While the share of companies paying small yields of less than 1% has decreased, the share of companies paying greater than 4% has also declined. Historically, about 12% of S&P 500 stocks offer a yield of 4% or greater. As of the third quarter, the tally is half that at 6%.

So, as a generation of Baby Boomers seeks alternative solutions for income generation in the face of low fixed income yields, the dividend opportunity set is also shrinking. Even still, with average dividend yields in excess of most fixed income, you can bet dividend ETF's will continue to be recipients of investment dollars.

Circling back on the point of overlap, let's take for example, four of the most popular dividend ETF's by assets:

• Vanguard Dividend Appreciation ETF (VIG)
• iShares Select Dividend (DVY)
• Vanguard High Dividend Yield ETF (VYM)
• SPDR S&P Dividend ETF (SDY)

After pulling the holdings as of 1Q 2016, I used some fancy programming pyrotechnics to generate a map of their holdings. The graphic below illustrates the overlap between these ETF's. The large circles represent an ETF ticker. Each of the spokes is an individual stock holding and is scaled to its weight across the ETF's. The spokes connect each stock to the ETFs in which it is held. For the sake of simplification, only underlying holdings with a greater than 3% weight receive a label.

A few observations. The largest names tend to be those that are held in common between all of the ETF's--those stocks in the middle of the map. The stocks that are unique to an ETF tend to be the smallest weights and line the periphery. It's surprising how similar the ETF's are; only 20% of the total weight across the 4 ETF's is unique. 80% of the assets have at least one holding in common with the other three ETF's. 30% have two holdings in common. And 9% of the assets are held in common across all four ETF's.

Self-perpetuating momentum

Here's why this matters. I mentioned earlier that $126 billion resides in dividend ETF's. If the stocks occupying the largest weights are the ones that are most commonly held across dividend ETF's, then the massive flows we have seen over the last few years have disproportionately gone into these stocks.

The characteristics of the most highly-weighted names across these popular ETF's betray exactly this point. Flows have pushed up valuations of the stocks holding the greatest weights, and those with the greatest weights have substantially better momentum (momentum being the average of 12 month trailing total return).

The table below takes all of the stocks held in the four popular dividend ETF's and treats them as one super portfolio. It then segregates the super portfolio into stocks that occupy weights greater than and less than 1%. I then calculate average characteristics on the two groups of stocks.

The <1% group is cheaper by 37%, 17%, 28%, and 32% relative to sales, earnings, cash flows and book value than the >1% group, and their dividend yield is 10% higher. Given the choice between these two groups of stocks, any competent investor would choose the <1% group! Unfortunately, the <1% group represents a disproportionately small amount of the ETF's holdings, just 19%. The more expensive and lower-yielding group represents a whopping 81% share.

Why?

In a word...scale. ETF's prefer stocks with large market capitalizations because they are designed to accommodate large amounts of assets that are not limited by capacity constraints. Some of the highest yields are offered by lower capitalization stocks that are beyond the reach of the Vanguards, iShares and SPDR's of the world. This creates opportunities for smart investors willing to go against the grain.