Index Providers: Whales Behind the Scenes of ETFs (2021)
Yu An, Matteo Benetton, Yang Song
Invited for submission to the Review of Financial Studies, URL
The first research paper of AGNOSTIC Papers is somewhat different from what I have done before. It has nothing to do with portfolio backtests or natural language processing, but instead examines a specific area of the asset management industry: Exchange Traded Funds (ETFs) and market indices. Although this is not directly related to investment strategies, the results of this paper brought me to an interesting investment idea that I want to analyze further. Apart from that, I like the paper for its combination of empirical analysis and theoretical modelling.
I am currently thinking about a common structure for the posts of AGNOSTIC Papers. To start with, I divided this one into the following parts:
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
ETFs experienced remarkable growth over the last years1See for example Statista or Blackrock. and made their reputation as one of the most useful financial innovation of all time. Nowadays there are many different types of ETFs, but the vast majority still tracks the performance of an index.2This became widely known as “indexing” or, although not entirely correct, “passive investing”. Popular examples are ETFs tracking the S&P 500- or the MSCI World index. The paper analyzes the role of those indices and, more importantly, the role of the companies providing them (so called index providers).
Let’s go through an example to make it clearer. An index like the S&P 500 is a commercial product that belongs to the intellectual property of an index provider. In this case, this is S&P Global Inc., a publicly traded US company. Such index providers enjoy a very attractive business model because everyone who wants to use their indices must pay them a licensing fee. ETFs are no exception from this. For example, the SPDR S&P 500 ETF pays 3 basis points of its assets as licensing fee each year. Given that the assets of this ETF currently amount to about $395B, S&P Global receives annual licensing fees of $118.5M.3Information as of April 24, 2022 from the ETF’s website. Not bad for just one customer.
In addition to that, some indices are widely recognized standards among finance professionals and the public. MSCI World, S&P 500, Russell 2000, NASDAQ 100 – we all know the names of the few big ones. Again, this is very attractive for index providers because their business model reinforces itself. ETF sponsors, the asset managers that offer ETFs, want to have popular indices for their funds. And the more ETFs track the index, the more popular it becomes. All of this suggests tremendous brand- and market-power of index providers.
The paper analyzes this market power in two steps. First, the authors create a novel dataset and identify several empirical facts about index providers and ETFs. In the second part, they use these facts to formalize a theoretical model of the market. Based on this model, they also suggest how competition among index providers might be improved for the benefit of ETF investors like you and me.
Data and Methodology
The empirical part of the paper is based on a sample of 598 US equity ETFs. The tickers are provided by Morningstar and the sample period ranges from January 2010 to December 2019. In the next step, the authors manually identify the underlying indices. If available, they also include information about the licensing fee agreement from the ETF’s regulatory filings. Finally, they obtain monthly returns, expense ratios, and assets under management (AuM) from the CRSP mutual fund database.
Table 1 of An et al. (2021).
Panel A of the summary statistics provides a well-known, but still interesting insight: the assets of ETFs are heavily concentrated to a very few big funds. The average ETF in this sample had AuM of $2037.29M, however, 75% of all ETFs were smaller than $814.13M. The sample mean is thus heavily distorted and way larger than the median.4For the statisticians out there: since the mean is way larger than the median, the distribution of ETF assets is positively skewed. If you remember the SPDR S&P 500 ETF with almost $400B assets under management, this is no surprise.
Panel B and C show the same data on the ETF sponsor- and index provider-level. Again, the distribution is positively skewed and AuM are very concentrated among the big asset managers. The average ETF sponsor managed $16,939.7M, but 75% of all ETF sponsors had less than $1,154.38M under management. For the index providers, this pattern is surprisingly similar (mean of $15,635M and 75%-quantile of $1,136.98M).
Apart from concentration, the table also shows another sign of market power. The average ETF sponsor only works with 1.68 index providers. This is somewhat strange to interpret, but it basically means that there are close relationships between ETF sponsors and index providers. Like most people choose either Windows or Mac for their computer, most ETF sponsor partner with one or two major index providers and stick with them over time. That makes it hard to compete for smaller players and new entrants.
Overall, I think this is a very cool dataset. Morningstar and CRSP are both leading providers of financial data, so data quality shouldn’t be an issue. In addition to that, all three authors are associated with leading universities and the paper was invited for submission to the Review of Financial Studies, one of the top finance journals. All this makes me confident that the results are based on reliable data and sound empirical methodology.5Matteo Benetton, one of the authors, also presented the paper at a conference in October 2021 and received quite positive feedback.
Nonetheless, there are certain limitations. First, the paper uses only US data. It would be interesting to see the market power of index providers in different geographies. But I suspect that international data is very hard to collect, so the authors probably haven’t done that for practical reasons. Second, the sample period from 2010 to 2019 is relatively short and coincides with perhaps the most successful decade of ETFs and indices ever. Maybe this is also due to data availability. But to the best of my knowledge, the authors don’t comment on possible biases from that. Finally, the authors exclude leveraged and synthetic ETFs from the sample. They also don’t comment on this decision. Personally, I would have included them because many of them are also tracking indices.
Important Results and Takeaways
The paper offers a lot of interesting statistics and discussions about econometric methodology. But in my opinion, the key findings are the following empirical facts about index providers and ETFs.
Index providers are an oligopoly
The market of index providers is extremely concentrated. Essentially, there are just five relevant players: S&P Dow Jones Indices, CRSP, FTSE Russell, MSCI, and NASDAQ. Based on AuM, these five companies captured almost 95% of the market as of December 2019. The authors also show that this market concentration is stable over time. The market of ETF sponsors looks very similar, although it is a bit more competitive. As of December 2019, the top five ETF sponsors managed about 94% of ETF assets in the USA.
Table 2 of An et al. (2021).
In addition to that, the authors provide further evidence that one major index provider tends to partner with one major ETF sponsor and vice versa. For example, 97.7% of State Street’s ETF assets were tracking S&P Dow Jones Indices as of December 2019. Again, this figure is stable over time and the general pattern for the four other ETF sponsors is very similar.6See Table 3 on page 12 in the paper. Such long-term relationships are often beneficial for both parties but they also lead to switching costs and hamper competition. In line with this argument, most ETFs never changed their index provider during the 10 year sample period.
ETF investors care about the index
Given that index providers charge substantial licensing fees, the question arises whether their indices are worth it. Clearly not, if you ask the ETF sponsors. Many of them are publicly complaining about high licensing fees and blame investors for their inertia.7See for example these two articles of the Financial Times and Morningstar. The former CEO of Amundi, a large European ETF sponsor, went even further and claimed that licensing fees are not justified by the value index providers add.8Quoted in this article of the Financial Times.
To answer this question more objectively, the authors analyze the relation between AuM of ETFs and the underlying index brands. A fixed effects regression reveals that index providers alone explain about 20% of the variation in ETF assets under management.9See Table 4 on page 14 of the paper. This effect remains statistically significant after controlling for expense ratios, past returns, and the brand of ETF sponsors.
Without the econometric blah blah: ETF investors seem to care about the underlying indices and brands like S&P, MSCI or Russell attract assets. Ironically, however, there is no statistically significant relation between index brands and ETF performance. In fact, indices tracking the same market are often almost identical. For example, the S&P 500 and the Wilshire US Large Cap Index both capture the largest US companies. The nuances are of course different, but ultimately, both indices performed almost identically.10More information about the S&P 500 and the Wilshire US Large Cap Index on their respective websites. The only difference: very few investors know the Wilshire Index and therefore, there are almost no ETFs tracking it.
Remarkably, index providers have thus managed to turn a completely exchangeable product into a sticky brand. Is this something special? No! In fact, many industries work this way. For example, some people are willing to pay a premium for Evian although the water is probably not that unique.
Are ETF sponsors right to complain about licensing fees? That’s hard to answer. On the one hand, a popular index helps them to attract investors which is clearly valuable for them. On the other hand, the market power of index providers can indeed translate into excessive markups. However, we will see that ETF sponsors are not those who end up paying the licensing fees. Instead, most of them are passed through to ETF investors.
Index providers capture 1/3 of ETF fees
To analyze this question in more details, the authors manually collect the licensing agreements for 52 of the 598 ETFs.11It is not mandatory for ETFs to disclose their individual licensing conditions. So the data is very limited and possibly suffers from selection bias. The authors mention these issues themselves. The typical licensing fee consists of two components: X basis points (BP) of AuM plus a fixed fee of $Y per year. For the SPDR S&P 500 ETF, that’s 3BP of AuM plus a fixed fee of $600K per year.
Own illustration based on An et al. (2021). The numbers for the management- and licensing fee are illustrative examples from the SPDR S&P 500 ETF.
Ultimately, the market power of index providers affects ETF investors because they pay most of the licensing fees in form of higher ETF expense ratios. Strikingly, the fraction of licensing fees continuously increased from 31.4% in 2010 to 35.7% in 2019.12See Table 7 of the paper on page 18. Index providers are therefore the hidden champions and capture more than 1/3 of the ETF industry’s revenues.
Index providers are extremely profitable
In the theoretical part of the paper, the authors formalize a model of the ETF industry. I skip the mathematics because I don’t think they are interesting for a non-academic audience (and because I don’t understand most of them myself). The results are very interesting because the model estimates marginal costs of ETF sponsors and index providers which are not directly observable from the data.
Table 9 of An et al. (2021).
Let’s start with Panel B. The average ETF13Note that the authors restrict the sample to the 20 largest ETFs for this estimation. Those are representative for the industry as they capture most of the assets. charged 9.3BP of management fees in 2019. But the marginal costs, i.e. the cost for managing an additional dollar, were just 5.4BP. This gives ETF sponsors a healthy markup of 42% (so called Lerner-Index). Although they like to portray themselves as the victims of greedy index providers, ETF sponsors also have market power over their investors.
Index providers, however, are indeed on a different level. The average licensing fee in 2019 was 4.4BP with marginal costs of only 1.6BP. This translates into a very high markup of 63.4%. This is in line with estimates that the profit margin of the leading index providers was around 65% in 2019.14See this article of the Financial Times. Put into perspective, this number is extraordinary. Apple, one of the most profitable companies that ever existed, had an operating profit margin of “just” 30% in 2021. The same number for MSCI is 54% and that includes some businesses that are probably less profitable than index provision.15I got the operating profit margins from the companies’ annual reports.
Conclusions and Further Ideas
The paper highlights an area of the asset management industry that is indeed “behind the scenes”. To be honest, I look at many of these indices every day and I have never consciously thought about the tremendous underlying business model. I really like the structure and process of the paper. The authors first collect robust empirical facts and subsequently use them to formalize a model that helps to derive unobservable insights. In my opinion, that’s exactly the way how science improves our industry.
What are the implications of those four key results? The authors highlight two points. First, ETFs should be legally required to report their licensing agreements. Mandatory disclosures enhance transparency for both investors and competitors and thereby make the market more efficient. Second, the authors address competition in the index market. According to their estimates, eliminating the market power of index providers could reduce ETF expense ratios by up to 30%. They admit, however, that this is a theoretical best case and hardly realistic. They are also silent about specific policies to approach this issue.
Given that my motives are less noble than those of scientists, my personal conclusion is somewhat different: I consider the oligopoly of index providers a very interesting investment idea. Four of the top index providers are publicly traded companies, so we can easily invest in them. Buying into the oligopoly and holding the companies for the long term could be a profitable investment.
Of course, the success of this strategy depends on the purchase price. Investors are not stupid (at least some of them) and recognized the great business model of leading index providers. Therefore, the shares of MSCI and the other firms performed very well over the last years and are not cheap by common valuation ratios. Nonetheless, the paper got me interested in this industry and I want to analyze this further. So stay tuned for a follow-up!
- AgPa #66: Machine-Learned Manager Selection (2/4)
- AgPa #65: Machine-Learned Manager Selection (1/4)
- AgPa #64: Fund Manager Multitasking
- AgPa #63: Fire the Winners and Hire the Losers
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 (*). Please also read the Disclaimer.
|1||See for example Statista or Blackrock.|
|2||This became widely known as “indexing” or, although not entirely correct, “passive investing”.|
|3||Information as of April 24, 2022 from the ETF’s website.|
|4||For the statisticians out there: since the mean is way larger than the median, the distribution of ETF assets is positively skewed.|
|5||Matteo Benetton, one of the authors, also presented the paper at a conference in October 2021 and received quite positive feedback.|
|6||See Table 3 on page 12 in the paper.|
|7||See for example these two articles of the Financial Times and Morningstar.|
|8||Quoted in this article of the Financial Times.|
|9||See Table 4 on page 14 of the paper.|
|10||More information about the S&P 500 and the Wilshire US Large Cap Index on their respective websites.|
|11||It is not mandatory for ETFs to disclose their individual licensing conditions. So the data is very limited and possibly suffers from selection bias. The authors mention these issues themselves.|
|12||See Table 7 of the paper on page 18.|
|13||Note that the authors restrict the sample to the 20 largest ETFs for this estimation. Those are representative for the industry as they capture most of the assets.|
|14||See this article of the Financial Times.|
|15||I got the operating profit margins from the companies’ annual reports.|