On the Percentage of Market Cap held by Domestic Stock ETFs

Posted on by David Merkel

I don’t have all the resources that I would want in order to do complex analyses.? Give me the database, and the right software, and I can do amazing things.

Even with limited data, and cruddy software, I still have something interesting this evening.? On January 21st, I made measurements of domestic equity ETFs to try to analyze what percentage of domestic equities were held by ETFs.

In order to limit my efforts, I polled the largest 61 domestic stock ETFs, excluding funds that are leveraged or inverse.? (those don’t buy/sell the equities directly, but use derivatives.? Granted, the derivative seller has to hedge, but he very well may cross hedge, messing up the estimates.)? That accounted for 90% of the markets cap of ETFs.? I then took the actual stock holdings of the ETFs and aggregated them, and then compared those holdings to the market capitalizations of the underlying stocks themselves, ending with a percentage of each stock held by the top 90% of ETFs.

I then ran a regression of that variable on several other variables.

SUMMARY OUTPUT
Regression Statistics
Multiple R
0.655372491
R Square
0.429513102
Adjusted R Square
0.428951442
Standard Error
0.013189645
Observations
7,118.000000000
ANOVA
df SS MS F Significance
F
Regression
7.000000000
0.931250612
0.133035802		 764.719743739 –
Residual
7,110.000000000
1.236903529
0.000173967
Total
7,117.000000000
2.168154141
Coefficients Standard
Error 		t Stat P-value Lower 95% Upper 95%
Intercept
0.002247578
0.000264739
8.489772940		 0.000000000 0.001728610 0.002766546
shr insd
(0.000086814)
0.000017370
(4.997942180)		 0.000000593 (0.000120865) (0.000052764)
beta
0.001683951
0.000182311
9.236686447		 0.000000000 0.001326567 0.002041335
shr inst
0.000292763
0.000005154
56.804885114		 – 0.000282660 0.000302866
mktcap
(0.000000003)
0.000000018
(0.151538794)		 0.879555010 (0.000000039) 0.000000033
3m avg volume
(0.000000002)
0.000000002
(1.282295801)		 0.199780706 (0.000000006) 0.000000001
3m realized volatility
0.000076315
0.000005765
13.237614053		 0.000000000 0.000065014 0.000087616
Float/Shs
0.000001184
0.000002810
0.421210862		 0.673613846 (0.000004325) 0.000006693

In short, I learned that ETF holdings of stocks were:

  • Inversely proportional insider holdings
  • Proportional to the stock’s beta, realized volatility, and amount held by institutions, and
  • Seemingly not related to market cap, trading volume or float.

Even the intercept term has some value as it is near the actual average percentage of market cap held by the top 90% of ETFs, which was 2.15%.? Assuming the same proportion applies to the last 10% that would mean that domestic stock ETFs own 2.39% of domestic stocks.? That’s enough to affect pricing at the margin.

Now, that percentage held by the top 90% of domestic ETFs in any common stock was as high as 17.9%, and as low as zero.? In terms of percentage of market capitalization held by the top 90% of domestic ETFs, we hit zero at stock 2912.

Implications

  • Domestic stock ETFs tend to pick more volatile stocks.
  • Domestic stock ETFs tend to pick stocks held by major institutions.
  • Domestic stock ETFs tend to pick stocks less held by insiders.? (They tend to be more boring.)

My summary is that those who create ETFs, even the big successful ones, tend to follow trends.? By their nature, they are extrapolating from what worked in the past, but in the process of doing so, end up overchoosing some names, and in the process add to their volatility.

That’s all for now, I still don’t feel well.

5 thoughts on “On the Percentage of Market Cap held by Domestic Stock ETFs”

  1. Pingback: Sunday links: an unhappy bull market Abnormal Returns

    1. I think it is both, and I hint at that in my article. There is an incentive for managers to choose investments that have done well in the past. Additional capital entering such a trade should raise volatility, as the investments move into the hands of the short-term investors of the ETF.

  3. Hi Mr. Merkel.

    I had a couple of thoughts when reading through this.

    1. You seem to indicate a high level of freedom for the trustees of ETFs. I think that the preponderance of market cap in ETFs represents full [index] replication schema. If that is true, the reality would be that most decision making vis-a-vis the composition of ETFs reflects the volition of the index board, not the ETF sponsors.

    2. You note that “ETFs tend to pick stocks held by major institutions.” ETFs are institutional investors (thus, included in the statistic). With their growing importance as a low cost indexing strategies in the retail space, ETFs make up a grow proportion of this variable. Because of that, I don’t think that the institutional ownership variable makes sense as an explanatory variable. It is kind of circular, like saying my budget of x this month explains my budget of x this month.

    3. What is the degree of multicollinearity of the “3m realized volatility” and beta variables? I don’t suspect that it caused any real problems, but who knows.

    1. 1. Trustees choose the indexes, and choose what constitutes a “creation unit.” Investors choose what indexes they want to use. Trustees aim for indexes that will appeal to investors, which typically involves trend-following.

      2. It is weakly circular; ETFs are only 2.4% of the market, and institutional investors at least 20x larger. The variable makes sense because it is a much larger measure of intermediate-term interest, vs ETF interest, which is short-term.

      3. The two variables are uncorrelated.

      That surprised me, so I did some digging, and found that I made a mistake. The “realized volatility” variable was really a “total returns over the last 3 years” variable.

      That might make more sense, and it indicates that ETFs chase intermediate momentum as well.

Comments are closed.

Theme: Overlay by Kaira.