plvch
plvch · Research note

The ETF fee paradox.

US investors have never had cheaper access to the market: the average ETF dollar now pays 0.13% a year. The industry, meanwhile, has moved its launch effort almost entirely to active products, with the median new fund charging 0.69%. The money is starting to follow.

§ 01
The shape of the market

The United States now has roughly as many ETFs as listed companies, and the two things most often written about them are both true: fund fees keep falling, and new funds keep getting more expensive. To be fair, a 0.60–0.70% median fee on recent launches is still progress relative to many older cohorts of active mutual funds. But sheer launch volume does not make the menu better, and it certainly does not make choosing from it easier for the typical investor. The contradiction sits in the word "average". This note rebuilds the full fee distribution from the primary source — every fee an ETF discloses in its prospectus and every census form it files with the SEC, from 2010 to 2026.

US '40 Act ETFs
3,954
SEC N-CEN census, trailing four quarters to 2026Q1.
Assets
$13.0T
Fiscal-year average net assets, as reported to the SEC.
What the average dollar pays
0.13%
Asset-weighted net expense ratio. The median fund charges 0.50%.
Industry fee revenue
$16.4B/yr
Implied: each fund's assets × its net fee, summed.

Sort the funds into five fee buckets — 3,820 of the 3,954 report both a fee and assets — and count three things for each: funds, assets, and the fee revenue those assets produce.

Figure 01
Funds, assets and fee revenue by expense-ratio bucket
Source · SEC N-CEN + Risk/Return XBRL
share of fund count share of assets share of fee revenue
8%
73%
25%
< 0.10%
10%
10%
12%
0.10–0.20%
28%
12%
30%
0.20–0.50%
48%
5%
28%
0.50–1.00%
7%
0.4%
4%
≥ 1.00%
3,820 of 3,954 US '40 Act ETFs in the fiscal-2025 census (the rest lack a joined fee or assets). Net expense ratio per fund from prospectus Risk/Return XBRL filings; assets are fiscal-year average net assets from N-CEN. Revenue = assets × fee per fund, summed by bucket.

Funds charging under 0.10% are 8% of the shelf and hold 73% of the money. This is the cheap core built by Vanguard, BlackRock and State Street, where the price war happened. Funds above 0.50% are the reverse case: more than half the shelf, about a twentieth of the money. Revenue evens the picture out. The 0.50–1.00% bucket alone earns $4.6B a year, roughly as much as the entire $9.5T core.

Investors in aggregate pay very little, because the money concentrates in a few dozen cheap funds. The industry still makes three quarters of its fee income on the rest of the shelf.

§ 02
Fifteen years of fees

The price of the average fund never really fell. The price of the average dollar did.

"ETF fees keep falling" is true of the asset-weighted average, the fee paid by the typical dollar. The fee charged by the typical fund has sat near half a percent for fifteen years and has drifted up since 2020. The two lines below come from the same prospectus filings; only the weighting differs.

Figure 02
Median fund fee vs asset-weighted fee · 2011–2025
Source · SEC Risk/Return XBRL + N-CEN
median fund (what the typical ETF charges) asset-weighted (what the typical dollar pays)
0.0% 0.2% 0.4% 0.6% 2011 2013 2015 2017 2019 2021 2023 2025 median fund · 0.50% average dollar · 0.12% fund-level assets reported from 2019 0.38 points apart
Net expense ratio per fund per year (net of waivers where a waiver exists, otherwise gross), from prospectus Risk/Return XBRL. Asset weighting uses fiscal-year average net assets from N-CEN, available at fund level from 2019 (2018 is omitted as partial). The fee-year-2025 panel value is 0.12%; the fiscal-2025 census snapshot used elsewhere in this note puts it at 0.13%. Pre-2019 fee history covers funds that survived to file an N-CEN census.

Trackers with longer asset records, Morningstar and ICI among them, show the asset-weighted line falling through the 2010s before settling near current levels. The median never joined that decline. Fees compressed where investors moved their money; the rest of the shelf never repriced. And the shelf kept growing: 3,954 funds in the fiscal-2025 census against about 2,000 in 2019, with 250–670 organic launches a year against roughly 100–200 closures.

§ 03
The launch conveyor

Most new funds are active, and priced accordingly.

The census marks each fund's first filing, which gives a reliable launch ledger from 2020. Launches nearly tripled between 2020 and 2025. Their composition changed more: the share that are actively managed (no index, a manager's discretion, often an options overlay) rose from 44% to 85%.

Figure 03
New ETF launches per year, active vs index
Source · SEC N-CEN, first filings
active index-tracking
250
2020
341
2021
433
2022
390
2023
539
2024
670
2025
First-ever N-CEN filing per fund series, by calendar year of filing. "Active" = the fund does not report tracking an index. Counts exclude mutual-fund conversions and pre-existing funds that filed the form late (anything arriving with more than $500M of first-year assets). 2025 may still gain a few late filers; the partial 2026 cohort is excluded.

Pricing follows the mix. The median fund launched in 2011 charged 0.53%; the median 2025 launch charges 0.69%, about five times what the average invested dollar pays. The decomposition matters: index launches have cost a fairly flat 0.40–0.50% throughout, while active launches drifted from about 0.55% to 0.70%. Most of the rise in the blended line comes from the mix flipping to 85% active.

Figure 04
Median fee of newly launched ETFs, by vintage
Source · SEC Risk/Return XBRL
all launches active launches index launches
0.0% 0.2% 0.4% 0.6% 0.8% 2011 2013 2015 2017 2019 2021 2023 2025 active launches index launches all launches · 0.69%
First fee observation per fund, grouped by the year of that observation — a close proxy for launch year, since prospectus fees are filed at registration. Net of waivers where a waiver exists. Early active cohorts are small (active ETFs were rare before 2016), hence the noise in that line. Pre-2018 vintages include only funds that survived into the N-CEN era.

For the investor this is, so far, a story about the shelf rather than the portfolio: index funds still hold fourteen of every fifteen dollars. But the active side is compounding. Active ETFs held $44B in mid-2019 and hold $872B now, a twentyfold rise over a period in which index assets grew two and a half times. Each of those dollars also earns the sponsor more: at current fees, active funds collect about a quarter of industry revenue on 6.7% of assets.

Figure 05
Active ETFs: share of funds vs share of assets · 2019–2026
Source · SEC N-CEN
active share of fund count active share of assets
0% 15% 30% 45% 2020 2021 2022 2023 2024 2025 2026 52% of funds 6.7% of assets · $872B
Trailing-four-quarter census windows, 2019Q2–2026Q1. Active = the fund does not report tracking an index. Assets are fiscal-year average net assets; the active total includes conversions.

Half the shelf, a fifteenth of the money. The natural question is who is doing the launching.

§ 04
The giants' new shelf

Eight of the ten largest sponsors launch above their own book.

It would be tidy if the expensive shelf belonged to boutiques while the giants tended the core. The ledger says otherwise. For each of the ten largest sponsors, the chart compares the asset-weighted fee of the funds it ran before 2020 with the median fee of what it has launched since. Mutual funds converted into ETFs, a large flow at Dimensional, JPMorgan and Fidelity, are excluded along with other transfers of existing assets; what remains is new product.

Figure 06
Incumbent book fee vs median fee of 2020–25 launches
Source · SEC N-CEN + Risk/Return XBRL
fee of the pre-2020 book → median fee of new launches ← launches priced below book
Vanguard $7.0T · 11 launches
0.04 0.10 book → new launches
2.4×
BlackRock · iShares $3.0T · 155 launches
0.15 0.20
1.3×
State Street · SPDR $741B · 35 launches
0.12 0.15
1.3×
Schwab $412B · 9 launches
0.07 0.05
0.7×
Invesco $314B · 58 launches
0.28 0.22
0.8×
Dimensional $192B · 25 launches
0.20 0.24
1.2×
JPMorgan $186B · 43 launches
0.19 0.35
1.8×
First Trust $150B · 156 launches
0.63 0.85
1.3×
VanEck $91B · 28 launches
0.42 0.49
1.2×
Fidelity $90B · 40 launches
0.19 0.46
2.4×
Whole market average dollar → median launch
0.13 0.60
4.6×
0% 0.30% 0.60% 0.90%
Top 10 sponsors by current AUM, in %; the right-hand column is the launch fee as a multiple of the book fee. Book = asset-weighted net fee today of the sponsor's funds first filed before 2020, conversions and other transfers included; launch fee = median first-observed net fee of organic 2020–25 launches (across the market, 78 first filings arriving with over $500M of assets are treated as transfers, not launches). Bottom row: the market-wide asset-weighted fee against the market-wide median launch.

Eight of the ten price new product above their existing book. The exceptions are the two retail price-fighters, Schwab and Invesco. The step is smallest where the cheap-core franchise is largest: Vanguard's eleven launches in six years come in at a median 0.10%. It is widest where the sponsor is building an active franchise: JPMorgan launches at 0.35% and Fidelity at 0.46%, each more than twice its book.

Figure 07
What 2020–25 launches contribute today: assets vs fee revenue
Source · SEC N-CEN + Risk/Return XBRL
Sponsor Launches '20–'25 Median launch fee Share of sponsor's AUM Share of sponsor's revenue
Vanguard110.10%0.1%0.2%
BlackRock · iShares1550.20%1.3%2.0%
State Street · SPDR350.15%0.9%1.4%
Schwab90.05%3.4%1.8%
Invesco580.22%3.6%2.4%
Dimensional250.24%25.1%31.2%
JPMorgan430.35%52.1%65.4%
First Trust1560.85%29.1%34.9%
VanEck280.49%2.3%2.4%
Fidelity400.46%11.6%23.1%
All '40 Act ETFs2,6230.60%4.3%15.4%
Current fiscal-2025 assets and implied fee revenue of funds organically launched 2020–2025, as a share of each sponsor's totals. Conversions and transfers count as incumbent book; 2026 first-filers are excluded on both sides. Launch fees are taken within two years of the first census filing; coverage 97.4% of 2,623 launches.

Funds launched in 2020–25 hold 4% of industry assets and already earn 15% of industry fees. At JPMorgan they are most of the business: JEPI and JEPQ, the 0.35% options-income pair launched in 2020 and 2022, hold about $58B between them, and the post-2020 range produces two thirds of the firm's ETF fee income. Fidelity's launches produce 23% of its fee income on 12% of its assets; First Trust earns a third of its fees on launches priced at a median 0.85%. Even at Vanguard and BlackRock, where new funds barely register, the revenue share runs ahead of the asset share.

None of this required repricing the old funds. The old funds stay cheap, and the new revenue is built beside them.

§ 05
If you hold ETFs

What this means if you hold ETFs.

01

The 0.13% average is something you opt into.

It exists because most money sits in a few dozen funds priced under 0.10%, and that price is open to anyone. The median fund still charges 0.50%; nothing about the trend moves your money to the cheap side on its own.

02

New funds are priced like active management, because most of them are.

The median 2025 launch charges 0.69%, and 85% of launches are active. A brand known for cheap funds tells you little about its new ones: Fidelity's pre-2020 book averages 0.19%, its launches since come in at 0.46%.

03

The discount sits in the old funds.

The funds the price war repriced are the older, larger ones. When offered a new fund, the useful comparison is the incumbent in the same family with similar exposure, which is usually several times cheaper.

Methods & caveats

Everything above is computed from raw SEC structured datasets — no commercial fund databases. Universe and assets: Form N-CEN census filings, 2018Q3–2026Q1 (fund counts use trailing-four-quarter windows; assets are fiscal-year average net assets). Fees: prospectus Risk/Return XBRL summaries, 2010Q4–2026Q1; a fund's fee is the net expense ratio where a fee waiver exists, otherwise the gross ratio, taken as the median across share-class values of the latest filing. Launches: a fund's first-ever N-CEN filing (reliable from 2020). Sponsor brands group the SEC's legal trusts (e.g. five iShares trusts → BlackRock).

  • The universe is '40 Act ETFs. ETFs organized as unit investment trusts (SPY, QQQ, DIA — roughly $1.1T, all cheap and passive) and '33 Act trusts (crypto, gold) file no fund-level census and are excluded; including them would push the asset-weighted fee and the concentration further in the directions shown.
  • Vanguard series blend ETF and mutual-fund share classes; Vanguard's AUM is overstated and its figures blend classes. Its asset-weighted fee here (0.04%) sits slightly below the published ETF-class average (~0.05–0.07%).
  • "Organic" launches exclude 78 first filings that arrived with more than $500M of first-fiscal-year assets: mutual-fund conversions (Dimensional's range, several JPMorgan and Fidelity funds), funds adopted from other issuers, and pre-existing funds whose trusts filed their first census late (the form dates from 2019). The rule will also catch a rare organic launch that grew that fast in its first year.
  • Asset-weighted fees here (0.13%) run a couple of basis points below Morningstar's published 0.15–0.16%, explained by the two caveats above. Counts, assets, and launch mix reconcile with published industry totals.
  • Implied revenue = assets × net fee; it ignores securities-lending income, which subsidizes some cheap funds.

Data and pipeline: SEC DERA structured data sets (N-CEN, N-PORT, Risk/Return). Analysis code reproducible from the raw quarterly files.

Disclaimer

This is independent research commentary, written for general information. It is not investment advice, an offer, or a recommendation to buy, sell or hold any security, and it takes no account of anyone's individual circumstances. A fund's fee is one attribute among many; nothing here implies that any fund named is a good or bad investment.

All figures are the author's calculations from public SEC filings, may contain errors, and describe a moment in time — fees, assets and product lineups change. Sponsor, fund and index names (Vanguard, BlackRock, iShares, State Street, SPDR, Schwab, Invesco, Dimensional, JPMorgan, First Trust, VanEck, Fidelity, JEPI, JEPQ and others) are trademarks of their respective owners and are used for identification only; nothing here implies affiliation with, or endorsement by, any of these firms. If you find an error in the data or the method, the underlying filings are public and the author would rather hear about it than not.