Robinhood's 0.04% wealth share had the biggest per-dollar price impact

A 0.04% wealth share, a 7.66% small-stock price effect

Robinhood's per-dollar price impact exceeds every other investor group's in this paper, including hedge funds'. Its accounts held about 0.04% of US equity wealth from 2018 to 2020. Van der Beck, Cohen and Jaunin estimate that removing that wealth and distributing it among other investors would have changed prices in the smallest fifth of US stocks by 7.66%. Across the market, the value-weighted price change is only 0.25%. The paper is clear about the small aggregate effect. The striking claim concerns price movement per dollar invested, and that is the claim worth examining.

How the demand estimates get their force

The authors use the Koijen-Yogo demand system. It expresses each investor's stock weight relative to an outside asset as a log-linear function of log market equity and five characteristics: book equity, investment, profitability, market beta and dividends over book. Latent demand is the stock-level residual left after those characteristics. The coefficient on own market equity, γ, determines price elasticity, approximately 1 minus γ for a diversified holder. Because prices are endogenous, the authors instrument γ with a hypothetical market cap: the value implied if other investors assigned equal weights to stocks they had held for at least eight quarters. Market clearing then gives prices under counterfactual wealth or demand.

The Robinhood holdings measure takes some construction. Robintrack scraped account counts for each stock from March 2018 until the API closed in August 2020. Following Welch, the authors treat a stock's fraction of all account-positions as its representative portfolio weight, then scale it using equity AUM in Robinhood's S-1 and 10-Q. More than 3,500 institutions come from 13F filings. Households receive the shares left over. The resulting universe contains 2,539 stocks.

The estimated tilts are enormous. Robinhood's book-equity coefficient is -19.42, beside 3.01 for households and 0.55 for hedge funds. Its beta coefficient is 28.03, versus 6.64 for hedge funds. Meanwhile, the investors absorbing a change in Robinhood demand respond little to price: investment advisors hold 48.06% of wealth and have elasticity 0.10; banks are at 0.04 and pension funds at 0.03.

Robinhood itself has the highest estimated elasticity, 0.79. Its per-dollar effect is still largest because its holdings depart sharply from the asset-weighted average, while the holders taking the other side are inelastic. Dividing total repricing by wealth share yields marginal impact of 6.74 for Robinhood, 4.51 for hedge funds and 2.74 for households.

The large-cap 9.72 needs a careful reading

The lead over other groups can be narrow. Insurance companies reach 6.61, only 0.13 behind Robinhood. The authors' strongest answer comes from alternative holdings proxies: a price-weighted measure and two mappings from holder counts to portfolio weights, fitted on 13F data. In the baseline redistribution counterfactual, those proxies put Robinhood between 7.28 and 8.59. Across all counterfactuals, the range is 5.96 to 8.77. Under withdrawal, however, the baseline estimate falls to 5.96, below insurers' 6.61. The paper does not estimate insurers' withdrawal result.

The elasticity claim is more exposed to the choice of proxy. Price-weighted holdings correlate just 0.32 with the baseline measure. They produce γ of 0.73 and elasticity 0.34, slightly below hedge funds' 0.35. Robinhood leads on elasticity under three of the four proxies (0.79, 0.78, 0.61), with price weighting the exception (0.34).

The size split gives me still more pause. Marginal impact climbs from 3.44 in the bottom quintile to 9.72 in the top. In that top quintile, Robinhood has a 0.02% wealth share and the estimated repricing is 0.20%. The authors attribute the pattern to inelastic passive owners of large caps. Perhaps, but a ratio built on a 0.02% denominator is sensitive to mistakes in AUM or portfolio weights. They also acknowledge that the baseline could overweight small stocks if wealthier accounts hold portfolios closer to market weights. Pre-December-2019 AUM is filled in with a polynomial based on account counts. And the appendix puts December 2019 AUM at $11.7B, where the main text says $13.5B.

I would treat 9.72 as model output.

The small-cap estimate has more underneath it: 2.23% of wealth corresponds to 7.66% repricing. It is less sensitive to a tiny denominator, although the authors' warning about equal account weights overstating small-stock holdings matters more for this result than for 9.72.

What is behind the 13% variance share?

The abstract credits Robinhood trades with explaining over 13% of small-cap return variation during the 2020 lockdown. The calculation sequentially decomposes Q2 2020 cross-sectional return variance. Supply changes switch characteristics and shares outstanding; the institutional, household and Robinhood demand changes each switch a group's AUM and latent demand to their next-quarter values. Robinhood goes last. The authors argue that concavity therefore gives it the least room to contribute. Its share exceeds 7% across all stocks and is negligible in the top quintile. Robinhood equity holdings rose from $13.5B to $53B in 2020, so that 13% includes AUM growth as well as shifts in demand.

This is a one-quarter attribution to returns occurring at the same time. We did not find a return-predictability test or a tradable portfolio in the paper. The nearest evidence is a regression of latent demand: lagged absolute returns carry a coefficient of 1.270 (t of 3.54), and lagged share volume a coefficient of 0.194 (t of 19.21). Within R² is 0.251 across 9,810 stock-quarters, compared with 0.004 or less for most other groups. The prose rounds the t-statistics to 5 and 20, although Table 5 gives 3.54 for absolute returns. Share volume enters in raw shares rather than turnover, leaving room for size and popularity to influence that coefficient.

Individual counterfactuals show the V-shape around past returns. Ford's Q1 return was -48.06%, against -56.82% without Robinhood: its estimated effect cushioned the fall. Carnival returned 24.68% in Q2, against -5.41% without Robinhood, an amplification of 30.09 points. The ten stocks were selected ex post from the largest size decile for the biggest absolute moves, subject to at least 0.1% Robinhood ownership. That selection helps reconcile a 30.09-point Carnival effect with the negligible top-quintile variance share.

There is also an internal check on coefficient stability. The paper describes the pre-2020 estimates as largely unchanged; γ (0.22) and book equity (-18.77) support that description. Beta falls from 28.03 to 20.97, though, and dividends to book moves from -0.08 to 4.15.

Why we could not reproduce the demand estimates

We could not run this analysis. Robintrack account counts, needed for Robinhood portfolio weights, ended in August 2020. Estimating group elasticities also requires investor-level 13F holdings and AUM. Prices and volume by themselves cannot distinguish a shift in demand from movement along an inelastic demand curve.

The paper nevertheless makes a credible case for 2020 small caps. A few tens of billions of attention-driven money reached stocks where Robinhood owned 2.23% of market cap and moved prices. Robinhood reported $160B in platform assets by mid-2025; the authors caution that this may include assets custodied elsewhere. I would put much more weight on the per-dollar ranking if a post-2020 holdings re-estimate found the small-cap variance share rising with that AUM.