The passive ownership share is double what you think it is (2024)
Alex Chinco (URL), Marco Sammon (URL)
Journal of Financial Economics, URL/SSRN
After my post on passive investing (see AgPa #77) and its consequences for active managers, I had a long and very interesting discussion with David Einhorn about the issue. David was very gracious with his time and we ended up agreeing on many things, but also agreed to disagree on a few others. Overall, I think it is fair to say that I am still somewhat more supportive for passive than he is.
So what I am going to do over the following weeks is to challenge my views further. For that purpose, I went back to the episode of the Rational Reminder podcast with Michael Green (URL) from Simplify Asset Management. Mike describes the potential problems of passive investing in great detail and brings a lot of arguments to the table. He also mentions some interesting research papers on the subject. This week’s AGNOSTIC Paper is the starting point and attempts to measure how much of the US equity market is actually passive. Spoiler: it is hard to say precisely, but probably much more than most people think…
Everything that follows is only my summary of the original paper. So unless indicated otherwise, all tables and charts belong to the authors of the paper and I am just quoting them. The authors deserve full credit for creating this material, so please always cite the original source.
Setup and Idea
The idea and goal of the paper is very simple. How much of the US equity market is passive? Passive thereby means that investors do not perform any analyses and simply buy all available stocks in proportion to their market weights (see AgPa #3). Just like the global market portfolio (see AgPa #4), this is quite simple in theory, but a statistical nightmare to quantify in practice.
A common approach in the literature and financial industry is too simply measure the Assets under Management (AuM) of publicly available index funds. This is of course valid and reasonable, but this week‘s authors (and a little common sense) argue that this delivers a lower-bound estimate at best. Index funds and ETFs undoubtedly command a lot of the passive assets, but the market is much larger than that. For instance, we cannot see the AuMs of large institutional investors who manage their passive mandates internally. And what about all the derivatives? Those who invest in the S&P 500 via futures, options or swaps are in a broader definition also passive.1Nerd alert: S&P 500 investors are of course active as they bet on the success of US large caps. I have written several pieces about that (see SA #5, #6, #9), but for the sake of this example, it is fine.
None of that shows up in the AuMs of public ETFs and index funds. The authors therefore argue that the true amount of passive investing is larger than most of us think (hence the title of the paper). More importantly, they also develop a pretty cool methodology to estimate passive AuMs more comprehensively. Instead of fund holdings, they analyze the trading activity around index reconstitution days. In my view, this is in some way brilliant as it is so obvious after thinking about it. You can hide your holdings, but to remain passive you must eventually trade. Of course, you can still try to hide your trades in dark pools or other venues and we will never have perfect data on everything. But overall, the authors’ approach is in my view way more comprehensive than just looking at fund holdings where you know for sure that the data is incomplete.
Data and Methodology
To estimate passive AuMs, the authors first collect constituent changes from the S&P 500, S&P MidCap 400, Russell 1000, Russell 2000, and Nasdaq 100. They also highlight that this is of course just an incomplete subset of indexes, but the best they could realistically do. They mention, for example, that MSCI demanded $240k in licensing fees for daily index weights from 2000 to 2021. CRSP, another index provider, was not willing to offer daily weights at all. In my view, those are quite interesting notes and once again illustrate the market power of the leading index providers (see AgPa #1). The final sample nonetheless covers thousands of index additions and deletions between 2000 and 2021. Stock prices and volume data are from CRSP, a standard and high-quality database for the US market.
Now to the (in my view) very cool methodology. Suppose we have a portfolio worth $1b that aims to track the S&P 500 and a new stock enters the index with a weight of 1%. To remain an index fund, we need to buy 1% x $1b = $10m worth of shares on the reconstitution day. That is the fundamental equation of the paper, but the authors start from the opposite direction. We don‘t know the amount of passive assets (that is what we want to find out), but we have data on trading volume on reconstitution days and index weights. So we can reshuffle the equation and back out the underlying AuMs.
This is the key idea of this week’s paper. The authors use the dollar values of trading volume on reconstitution days to back out estimates for aggregate passive AuMs. They do the calculation for all index changes and average the results across stocks into one final estimate for the total passive AuMs of the respective index.
The key assumption, of course, is that trading volume on reconstitution days really just comes from rebalancing passive investors. This may sound extreme at first glance, but we will see below that it is not so unrealistic. Of course, we can never estimate passive AuMs to the last decimal. But I am quite confident that the methodology of the paper yields reliable estimates and the results seem quite robust. The paper is also published in the Journal of Financial Economics, one of the top publications in finance.
Important Results and Takeaways
Trading data suggest that 1/3 of the US market is passive
The heart of the paper are the estimates for the total passive AuMs and the resulting passive share of the market. The chart below summarizes the headline results and the authors also include holdings-based estimates of passive AuMs from the Investment Company Institute for comparison. The results are in my view very interesting. Rebalancing volume suggests that about 33.5% of the US market was passive as of December 2021. That is twice the amount from the holdings-based estimate and this gap is remarkably stable over time. Remember that those estimates come from only five indexes and are therefore most likely conservative. For example, the authors calculate that the passive share increases to 38.5% by the end of 2021 after adding AuMs from Vanguard funds that track CRSP indexes for which they don‘t have member data.
The authors next provide some more details and robustness around those headline results. In particular, they examine the key assumption that passive investors generate the entire trading volume on reconstitution days. As with any model and simplifying assumption, this is almost certainly not true in practice.
It is well-known (and reasonable) that ETFs and index funds tend to focus their trading on the closing auction. That is because they usually aim to minimize their tracking error versus the index which is calculated from closing prices. So instead of the full-day volume, we could take the other extreme and just look at closing auctions to estimate passive AuMs. But that is probably too restrictive either and risks to exclude some non-index-fund passive investors. So the authors start with a compromise and examine the trading volume from 4pm onwards. The following two charts show the estimates for passive AuMs and the resulting passive share by index and in total.
The dotted gray lines are always below the headline results and suggest less passive AuMs. This is just mechanical and not surprising as full-day volume is always at least as high as the volume from 4pm onwards. If we count less trading towards passive investors, the estimates for their AuMs logically decrease. The estimates are still higher than for the “traditional” holdings-based approach, however (24.6% vs. 16% by end of 2021).
There are also some interesting differences between indexes. Most of the passive ownership in the US market comes from the Russell 1000 and S&P 500 large cap indexes. While this was somehow expectable, the magnitude still surprised me. For example, the passive share for the Russell 2000 and S&P MidCap only stands at 1.1% for each index by the end of 2021. That is a tenth of the large cap indexes. Also remember that the analysis excludes all international indexes which also trade US large caps (e.g. MSCI World, FTSE All-World). The true passive ownership in US large caps is therefore most likely even higher.
For even more robustness and transparency, the authors calculate some more proxies and summarize the results in the following table. As they become increasingly restrictive and count less trading to passive investors, the estimates for the passive ownership share decrease from left to right. In the end, we will never be able to isolate passive trading perfectly. Most of the estimates, however, clearly suggest that passive ownership goes beyond index fund and ETF holdings which I believe is reasonable.
The authors finally present another interesting stress-test. If the methodology works, different index additions and deletions should yield about the same estimate for passive AuMs on any given reconstitution date. I think that’s very intuitive. If index addition A suggests passive AuMs of $5t and index change B gets to $6t, there seems to be a problem.
The authors examine the absolute average difference between the passive ownership estimates from each index change to test this. The results are very promising. For the full sample, the average measurement error across indexes is just 1%-point. So there is some noise and uncertainty around the estimates. Overall, however, index additions and deletions yield quite similar estimates for passive AuMs.
Index changes trigger massive trading volumes
In addition to the passive ownership share, the authors also present some empirical facts about trading volume on reconstitution days. I will go over them rather quickly to not state the obvious for too long. Unsurprisingly, index changes trigger massive trading volumes on reconstitution days. To put things into perspective, the authors normalize daily volumes by Average Daily Volume (ADV) over the previous 22 trading days. The following tables provide summary statistics for the full sample period.
The numbers are by all means impressive. Depending on the index, trading volume ranges from 4.7 to 14.2-times ADV on reconstitution days. I cannot observe a specific pattern for large or small caps and I guess the specific index construction rules have a higher impact than capitalization ranges. Finally, the chart below shows trading volume on reconstitution days in comparison to the previous days. I think the takeaway is quite clear and always the same – index changes trigger massive trading volumes on reconstitution days.
Passive trading affects prices – but (usually) not on reconstitution day
What ultimately matters for investors, of course, is the impact of this intense passive trading on prices. You may intuitively expect that buying (selling) pressure for additions (deletions) will lead to price increases (decreases) of the respective stocks. Unfortunately, it is not so simple. The authors calculate market-cap weighted portfolios of index additions and deletions to test this more thoroughly. The following chart shows average returns of those portfolios in excess of the US market on reconstitution days.
While there are some abnormal returns in the earlier years of the sample, they converged to virtually zero in recent years. So simply buying (selling) index additions (deletions) on reconstitution days doesn’t seem to be a viable strategy. If you think about it, this shouldn’t be surprising. Most index providers announce changes a few days before rebalancing (see below for more). So this is not new information on reconstitution day and thus unlikely to be profitable in a reasonably efficient market. These are bad news for (unsophisticated) investors aiming to profit from index rebalancing. On the other hand, it seems positive for the widely-debated negative side effects of passive investing. Despite their massive trading volumes, passive investors can apparently rebalance without having meaningful impact on prices.
As in many cases, reality is somewhat more complicated. The authors explain that many passive investors (ETFs, index funds, internal indexers, …) arrange their trades before the reconstitution day. How those that work? ETFs, for example, approach so called rebalancing facilitators (banks, hedge funds) some months before reconstitution day with their desire to trade those giant volumes on that particular day. The authors mention, for example, that the Russell rebalancing on the third Friday in June is typically arranged between March and May. Nobody, of course, knows future rebalancing volumes with certainty. But rebalancing facilitators do the best they can to predict it and so prepare for the large volumes on reconstitution days.
There are two important consequences for prices. First, if rebalancing facilitators predict the index changes and rebalancing volumes accurately, there will be (almost) no price impact on reconstitution days. Only unexpected volumes could lead to buying or selling pressure and thus price movements. The lack of abnormal returns of additions and deletions on reconstitution days suggests that this procedure works rather well.
Second, reconstitution days are not the right date to analyze the price impact of passive investors. If we expect in April that a stock will be added to an index in June, traders will buy it already in April to provide the liquidity on reconstitution day. The price impact from passive investors thus happens when their counterparties prepare to trade with them, not when they actually trade. The authors illustrate this with demand elasticities. The idea behind this basic economic concept is quite intuitive. An elasticity of -3 tells us that demand falls by 3% if prices increase by 1%. The higher (lower) this number in absolute terms, the more elastic (inelastic) the demand.
In the chart above, the authors estimate elasticities for the demand of passive investors over the 60 trading days (3 months) before reconstitution days. Intuitively, their demand should be very inelastic. They just track their underlying index and buy the stocks no matter what’s happening. This is not what we find on reconstitution days. The average elasticity on those days stands at -39.2 which suggests a very elastic demand. As mentioned above, this is because of the pre-arranged trades and we need to look at the preceding days to get the real picture.
The white lines in the charts above mark the announcement days of index changes. Those are the days at which rebalancing facilitators know with (almost) certainty what’s going to happen on reconstitution day. It is therefore reasonable to assume that most pre-arranged trades are done by this date and (most of) the price impact from passive investors has happened. The data and the charts support this. Elasticities increase after those announcement days, but are fairly low before. For example, the elasticity for the Russell 1000 is only -1.2 at announcement day and even lower in the 3 months before. This suggests that passive investors, as their critiques like to mention, are indeed mostly price-insensitive and bring inelasticity to the market.
Conclusions and Further Ideas
This post became way longer than I originally thought because I read the paper multiple times. I really like it when researchers translate simple theoretical concepts into practice and measure something that was unknown before. This paper is clearly from that category. The methodology is intuitive and doesn’t require understanding of advanced statistical wizardry. Yet it is still very effective and does the job. So I am a fan and I hope that the authors, or someone else, will find the data and budget to do a follow-up with international indexes to get an even broader understanding of passive investing.
I also like the paper for its focus. The authors measure how much of the US equity market is passive, but except for the elasticities, they are mostly silent about possible consequences. In my view, this is the right way to go. What to make from the higher-than-expected passive AuMs is a (quite complicated) question for itself. Proponents of passive investing may argue that fears of too much passive are overdone as the market still works quite well. At the same time, opponents can point to low demand elasticities and rightly mention that most people do not care about the potential side-effects of passive investing.
Another point I thought about when reading the paper is how dynamic and adaptable financial markets actually are. We have a system that allows to trade trillions of dollars in a single afternoon without wild price swings or other problems. In my view, this is simply fascinating and shows how complex the underlying market structures actually are. It also shows that apparently simple and intuitive ideas like “High volume must lead to price movements” not necessarily work (also see StanStu #1). Passive investing is simple in terms of investment analysis, but it is very sophisticated in terms of infrastructure and implementation.
As I mentioned earlier, my motivation for reading all those papers is primarily selfish. I (currently) believe that passive investing is for most people the only required investing strategy and a very hard-to-beat benchmark. So it may not be surprising that some of my own money also sits in passive ETFs. Given this position, I think it is very important to understand the arguments of the critics in order to determine if the investment is actually at risk. As Charlie Munger (2016) said, we should “[…] really try and destroy our previous ideas”. Despite all its advantages, this also applies to passive investing.
- AgPa #83: How Much of the US Market is Passive?
- AgPa #82: Equity Risk Premiums and Interest Rates (2/2)
- AgPa #81: Equity Risk Premiums and Interest Rates (1/2)
- AgPa #80: Forget Factors and Keep it Simple?
This content is for educational and informational purposes only and no substitute for professional or financial advice. The use of any information on this website is solely on your own risk and I do not take responsibility or liability for any damages that may occur. The views expressed on this website are solely my own and do not necessarily reflect the views of any organisation I am associated with. Income- or benefit-generating links are marked with a star (*). All content that is not my intellectual property is marked as such. If you own the intellectual property displayed on this website and do not agree with my use of it, please send me an e-mail and I will remedy the situation immediately. Please also read the Disclaimer.