Retail Investor Trade and the Pricing of Earnings

Jeremy Michels
[email protected]

University of Pennsylvania

April 2021

Abstract: I examine how retail investor trade is associated with the pricing of earnings by
measuring retail investor trade with the number of Robinhood users holding a firm’s shares. My
findings show that net retail purchasing during the earnings announcement is associated with a
more pronounced market response to earnings surprises for positive earnings surprises but a more
muted response for negative earnings surprises. Further intraday analysis suggests that retail
investors respond to more extreme stock returns following the earnings announcement instead of
the earnings news itself. Finally, in smaller firms, and for both the most positive and negative
earnings surprises, returns drift upward following the earnings announcement when retail trade is
high. Overall, the results show that retail trade is associated with significant changes in how the
market prices earnings information, both in the immediate earnings announcement window and
over longer horizons.

Keywords: Retail investor, individual investor, Robinhood, earnings announcement, post-

earnings-announcement drift.
JEL Codes: G10, G11, G12, G14, G24, G41, G50, M41, O33

I thank seminar participants at The Wharton School for their helpful comments.

Electronic copy available at: https://ssrn.com/abstract=3833565

1. Introduction

Retail investors are becoming increasingly active in their trading behavior, a phenomenon

that has captured the attention of the press, regulators, and the firms whose shares are being traded.1

In several well-publicized instances, surges in retail investor trade have corresponded to wild

fluctuations in firms’ stock prices—fluctuations seemingly divorced from fundamental firm

performance.2 These anecdotes highlight the potential for more engaged retail investors to alter

how markets price accounting information. Understanding the extent to which retail investor trade

alters the association between accounting earnings and price is a first step in understanding how

the recent increase in retail investor trade may impact price efficiency and investor welfare. In this

paper, I take this first step by documenting how the prevalence of retail investors relates to the

association between accounting performance and stock returns. I examine both the market’s initial

reaction to the earnings announcement and how returns evolve in the following weeks.

Beginning as early as Ball and Brown (1968), the accounting literature documents how

the stock market reacts to earnings information. Prior literature documents various properties of

firms and earnings that affect how investors respond to earnings (e.g., Collins and Kothari 1989)

as well as the timeliness of this response (e.g., Bernard and Thomas 1989). Although we have over

50 years of research on how investors price earnings, the properties of this relation can evolve as

market dynamics evolve. As I argue above, the recent swell in retail investor trade has the potential

to upend existing relations.3 Competition among retail-oriented brokerages has pushed trading

commissions to zero, reducing frictions that might otherwise impede retail trade. Technology and

1
See Osipovich (2020), SEC (2021), and Phillips (2021).
2
See McCabe (2021).
3
Barber et al. (2021) confirm this recent increase in retail trade.

Electronic copy available at: https://ssrn.com/abstract=3833565

mobile app-based trading platforms have further increased the ease and accessibility of trading. A

corresponding surge in retail investor trade has the potential to upend existing relations between

accounting earnings and market returns. This recent increase in retail investor activity has parallels

to the increased trading activity associated with the advent of online brokerages in the 1990s (e.g.,

Choi, Laibson, and Metrick 2002). However, there is evidence that traders who transitioned online

during the 90s were relatively experienced and wealthy investors (Barber and Odean 2002). In

contrast, the current surge in retail investing seems driven by inexperienced, first-time investors

(Barber et al. 2021; Tenev 2021), which heightens concerns regarding investor protection and

increases the possibility these trades are disconnected from fundamental firm performance.

Any change in the pricing of earnings associated with the increasing prevalence of retail

traders has potentially far-reaching implications. Concerning investor protection, efficient prices

protect otherwise uninformed investors from paying an unfair price for a given security. Frictions

that mitigate price discovery or introduce excessive volatility can limit this protection. Further,

prices diverging from fundamental value may result in misallocated capital. In response, regulators

are already considering measures designed to discourage frequent and speculative trade.4 On the

one hand, regulators desire open and equal access to financial markets. On the other hand,

regulators have a mission of maintaining orderly and efficient markets, avoiding excessive

volatility that may harm intermediaries, firms, and retail traders themselves.

Motivated by these concerns, I document how the prevalence of retail investor trade relates

to the price-earnings relation, both in the immediate earnings announcement window and over

longer horizons. When retail-investor base increases during the earnings announcement window,

4
See, for example, Osipovich (2021).

Electronic copy available at: https://ssrn.com/abstract=3833565

the immediate market response to earnings surprises is more pronounced. This result is driven by

positive earnings surprises. For negative earnings surprises, increases in retail investor base are

associated with a muted market response.

Further analysis of intraday retail trade and returns suggests that retail investors in my

sample do not respond to earnings news itself but instead react to the stock returns following the

earnings announcement. I find little retail trading activity in the first few trading hours following

the earnings announcement. However, both more positive and more negative stock returns in the

first few trading hours following the announcement of earnings are associated with greater

increases in retail holdings later on the earnings announcement date. These results are consistent

with retail investors reacting to the visibility of more extreme returns driven by earnings events as

opposed to reacting to earnings news itself.

Finally, I show that more extreme earnings surprises, both positive and negative, are

associated with a greater increase in the number of retail traders holding a firm’s shares in the

weeks following the earnings announcement, again consistent with attention-driven trading. When

examining returns over a longer horizon, returns drift upward for both the most positive and most

negative earnings surprises for stocks with the most retail investors or with the largest increase in

retail investors during the earnings announcement. However, these results are concentrated in

smaller firms. For larger firms, any effect of retail trade on share price appears to dissipate quickly.

I also document that the upward drift for firms with increases in retail traders is greater when short-

selling costs are relatively high and also following economic stimulus payments.

Overall, my results suggest greater retail trading activity can disrupt the price-earnings

relation in the immediate earnings announcement window. Further, for small firms, retail trade is

Electronic copy available at: https://ssrn.com/abstract=3833565

associated with a predictable upward drift in returns following both positive and negative earnings

surprises.

These results contribute to the vast literature on the price-earnings relation from the last

half-century. Ball and Brown (1968, 169–70) note that “the information contained in the annual

income number is useful in that if actual income differs from expected income, the market typically

has reacted in the same direction.” This seminal observation provided the foundation for much of

the future research on capital markets in accounting. Researchers often use the strength with which

investors respond to unexpected earnings (the earnings response coefficient, or ERC) to measure

earnings quality. The underlying assumption is that investors respond more strongly to earnings

when it is more useful in their valuation process (see Dechow, Ge, and Schrand 2010, 366 for a

more extended discussion). However, and as noted by Dechow et al., using investors’ response to

earnings as a measure of earnings quality assumes that investors accurately incorporate earnings

innovations into their assessments of firm value. In this paper, I posit that variation in a firm’s

investor base likely affects the degree to which this assumption holds. In other words, earnings

response coefficients may vary not because of differences in the attributes of firms and earnings

but because of differences in the types of investors who consume this information.

Beyond the earnings announcement window, I examine if retail trade around the earnings

announcement is associated with future returns. The idea that the initial market reaction to

unexpected earnings at the earnings announcement may be incomplete is not new. Bernard and

Thomas (1989, 1) note that “Ball and Brown (1968) were the first to note that even after earnings

are announced, estimated cumulative ‘abnormal’ returns continue to drift up for ‘good news’ firms

and down for ‘bad news’ firms.” Bernard and Thomas (1989) examine this phenomenon of post-

earnings-announcement drift (PEAD) in detail, concluding that PEAD results from a delayed

Electronic copy available at: https://ssrn.com/abstract=3833565

response to earnings information. Results on the degree to which a firm’s investor base contributes

to PEAD are mixed. For example, Hirshleifer et al. (2008) find no evidence that individual traders

drive PEAD. However, Ke and Ramalingegowda (2005) show that institutional investors can

exploit PEAD-related mispricing and, in the process, accelerate the speed of price discovery. I

complement this work by demonstrating (1) how retail trade relates to the initial pricing of earnings

in the earnings announcement window and (2) how retail trade associates with later post-earnings-

announcement drift (PEAD). My results showing that both positive and negative earnings surprises

predict positive future returns is more consistent with retail investors responding to the visibility

provided by extreme earnings events as opposed to overreacting or underreacting to earnings

surprises at the earnings announcement, consistent with prior work (Lee 1992; Hirshleifer et al.

2008).

In examining how retail investor trade is associated with the price-earnings relation, I

connect to the larger stream of literature analyzing disparate groups of investors and how these

investors differ in their consumption of accounting information. This literature finds that retail

traders neglect accounting information, even though this information is value-relevant

(Blankespoor et al. 2019). However, this does not necessarily imply that retail trade is therefore

irrelevant with respect to the market’s pricing of earnings. Consistent with prior work, I find little

direct evidence that retail investors react to earnings news. However, I find evidence that retail

investors react to stock returns associated with earnings releases. This coordinated retail trade then

has the potential to further influence stock prices (Barber, Odean, and Zhu 2009; Bushee,

Cedergren, and Michels 2020). Holding risk may prevent more sophisticated investors from

arbitraging mispricing away, particularly in smaller firms. Note that while institutional investors

may hold most of a firm’s shares, inelastic demand from these investors—due, in part, to index

Electronic copy available at: https://ssrn.com/abstract=3833565

investing—increases the likelihood that retail investor demand drives price changes (van der Beck

and Jaunin 2021).

Finally, retail investors are not a homogenous group. Fintech innovations are attracting

new investors, increasing variation in the level of trading experience among retail investors.

Contemporaneous work shows that investors on the mobile-centric trading platform Robinhood

engage in more attention-induce trade, consistent with these investors being relatively

inexperienced (Barber et al. 2021). Further, Moss, Naughton, and Wang (2020) find that

Robinhood investors disregard environmental, social, and governance-related disclosures, in

contrast to other work in experimental settings that shows individual investors are sensitive to

these disclosures (Martin and Moser 2016). Thus, prior findings on how retail traders react to

accounting releases may not generalize to the most recent period as the composition and attributes

of retail investors evolve.

I focus on these relatively active and inexperienced retail traders in this paper, measuring

retail trade using the number of Robinhood users holding a particular stock.5 As alluded to above,

Robinhood is a financial services company that provides commission-free trading primarily via a

mobile-app platform. During my sample period, Robinhood published how many of its users held

a particular stock, allowing me to calculate the average number of Robinhood users holding a

firm’s shares going into a firm’s quarterly earnings announcement, as well as how the number of

Robinhood users changed in the three-day window centered on the earnings announcement. In my

analysis, I am not interested in the effects of Robinhood users per se; rather, I use Robinhood data

5
Supporting the assertion that Robinhood investors are relatively inexperienced, Robinhood CEO Vladimir Tenev
stated in Congressional testimony that about half of Robinhood traders self-identify as first-time investors (Tenev
2021).

Electronic copy available at: https://ssrn.com/abstract=3833565

to proxy for the presence and activity of a particular type of retail investor. However, a limitation

of my analysis is that it may not generalize to a broader set of retail investors.

However, these relatively less experienced retail investors are likely of particular interest

to regulators, given investor protection is a central component of the Security and Exchange

Commission’s (SEC) mission, along with maintaining fair, orderly, and efficient markets and

facilitating capital formation. Understanding the association between retail trade and the price-

earnings relation is a first step in understanding more broadly how the increasing participation of

nonprofessional investors affects capital markets and potentially the welfare of these investors.

Understanding how retail investors trade and potentially influence security prices around earnings

announcements is particularly important, given the primacy of these events in explaining stock

returns and the fact that accounting releases are subject to regulatory oversight. Thus, findings in

this setting have the potential to inform a regulatory debate.

The SEC has long restricted ostensibly less sophisticated investors (non-accredited

investors) from participating in less transparent or less liquid markets. Arguably, this is at the

expense of fairness, as these investors are restricted from a set of investment opportunities. The

SEC has recently updated the definition of an accredited investor to allow individuals to qualify

based on professional certifications and credentials, as opposed to only net worth or income. At

the same time, the SEC has had to intervene in public markets to protect investors, and others have

questioned the marketing of a gamified investing experiencing to novice traders (Michaels 2020;

Popper 2020). Less informative or less timely prices also have the potential to harm firms if prices

serve as a feedback mechanism, providing information to managers on the quality of their

investment decisions (e.g., Luo 2005). Such an effect may be particularly harmful to smaller firms

if these firms have the greatest investment opportunities but have less developed information

Electronic copy available at: https://ssrn.com/abstract=3833565

environments overall and thus fewer alternative channels for manager learning. While these

questions are left to future research, I provide initial evidence on how retail trade measured using

Robinhood user holdings is associated with the initial market response to earnings and the

trajectory of returns over the 50 trading days following the earnings announcement. Future

research can help us better understand how this retail trade affects price and price-efficiency over

longer-horizons and the implications of any price distortions on firm and investor welfare.

2. Research Setting and Data

I construct my sample starting with data on the number of Robinhood users holding a

particular stock at a moment in time, as provided by Robintrack.6 Robintrack collected and

archived these data as made available by Robinhood. Robinhood published these data from May

2, 2018, to August 13, 2020. Robintrack typically provides multiple measurements per day of the

number of users who hold a particular stock. I take the number of users from the last reading for a

given day in my primary analysis. For example, my data show that 62,391 users held Hertz (HTZ)

on June 1, 2020. In some analyses, I examine intraday changes in the number of users holding a

stock.

My analysis focuses on quarterly earnings announcements. For each firm’s earnings

announcement, I calculate the average number of Robinhood users pre-announcement (Avg. Users)

as the mean of the number of users holding the firm’s stock each day during the window beginning

65 trading days before the earnings announcement and ending two days before the earnings

announcement. I calculate the change in the number of Robinhood users during a firm’s earnings

announcement (Δ-Users) as the difference between the number of users holding the stock one day

6
https://robintrack.net/

Electronic copy available at: https://ssrn.com/abstract=3833565

following the earnings announcement and the number of users holding the stock one day before

the earnings announcement. In all cases, I adjust the earnings announcement date to the next

trading day if a firm announces earnings after trading hours. In some analyses, I decompose

Δ-Users into three components: (1) the change in users from the trading day before the earnings

announcement to 1:00 pm Eastern Standard Time on the earnings announcement date, (2) the

change in users from 1:00 pm on the earnings announcement date to the end of the earnings

announcement date, and (3) the change in users from the end of the earnings announcement date

to the end of the next trading day. Since the number of users at precisely 1:00 pm on the earnings

announcement date is typically not available, I use the closest reading occurring before 1:00 pm

in this decomposition. Robintrack records user holdings at roughly a one-hour frequency.

I also examine Robinhood user trading activity in a longer window following the earnings

announcement. I define Post-EA Δ-Users as the change in the number of Robinhood users holding

a firm’s stock from the second trading day following the earnings announcement to the 50th trading

day following the earnings announcement, scaled by Avg. Users.

At the earnings announcement, I measure Earnings Surprise as actual quarterly earnings

per share (EPS) less the median analyst forecast of EPS. I scale the difference between actual EPS

and the median forecast by the firm’s share price at the end of the fiscal quarter. I take both actual

EPS and analyst forecasts from IBES.

I measure the market’s response to the earnings surprise with the cumulative three-day

market-adjusted abnormal return centered on the earnings announcement date. I also examine

cumulative abnormal returns (CARs) over the longer horizon of 50 trading days following the

earnings announcement. This research design is detailed in the following section. I use returns data

Electronic copy available at: https://ssrn.com/abstract=3833565

from CRSP. In some analysis, I calculate the intraday return from the market open to 1:00 pm on

the earnings announcement date using price data from TAQ.

I use several control variables to measure factors that likely affect ERCs and potentially

also relate to retail trading activity. Specifically, I control for firm size, growth opportunities, risk,

and leverage (Collins and Kothari 1989). Size is the logarithm of the market value of equity (MVE).

In several analyses, I also include ranks of MVE. MTB is MVE divided by the book value of equity.

Beta is the coefficient from regressing daily firm return less the risk-free rate on market returns

less the risk-free rate over the 252 trading days ending three trading days before the earnings

announcement. Leverage is total debt divided by the book value of equity. Persistence is the

coefficient of EPS regressed on lagged EPS within firm, using up to 10 years of data. I also include

an indicator, Loss, which takes a value of one if EPS is negative, and zero otherwise. Further,

ln(Analysts) is the logarithm of the number of analysts contributing to the earnings forecast on

which Earnings Surprise is based. Volatility is the standard deviation of daily returns over the

period beginning 65 trading days before the earnings announcement and ending two days prior

and Avg. Turnover is the average daily volume divided by shares outstanding during the same

period. I winsorize continuous variables at the 1st and 99th percentiles. Table 1 gives summary

statistics. All variable definitions are given in Appendix A.

Table 2 tabulates key variables used in the analysis. Panel A gives the means of MVE,

Δ-Users, and Avg. Users within each decile of these variables. Both Δ-Users, and Avg. Users are

very skewed, and I use ranks of these variables in later analysis. Panel B of Table 2 shows the

number of observations in a category when the sample is independently divided into terciles of

MVE and Δ-Users. Panel C similarly tabulates terciles of MVE and Avg. Users. These cross-

tabulations show some correlation between size and Robinhood users. However, the off-diagonal

10

Electronic copy available at: https://ssrn.com/abstract=3833565

cells still have a substantial number of observations, consistent with Robinhood users favoring

smaller firms.

3. Analysis and Results

In this section, I describe my analysis and results. First, I describe how investors’

responsiveness to earnings, or ERCs, vary with retail trader presence and activity. Next, I examine

how retail trade at the earnings announcement is associated with future retail trade. Finally, I show

how retail trade at the earnings announcement relates to future returns.

Earnings Response Coefficients.

In my first set of analyses, I examine how the change in the number of Robinhood users

holding a firm’s stock, and the number of users holding a firm’s stock, associates with ERCs. To

do so, I estimate the following regressions:

CARi,t = α + β1 Earnings Surprise + Controls + εi,t (1)

CARi,t = α + β1 Earnings Surprise + β2 Δ-Users + β3 Earnings Surprise* Δ-Users

+ Controls + εi,t (2)

CARi,t = α + β1 Earnings Surprise + β2 Avg. Users + β3 Earnings Surprise*

Avg. Users + Controls + εi,t (3)

I estimate a baseline regression in equation (1), excluding any Robinhood user-related variables.

The coefficient on Earnings Surprise is the ERC. In equations (2) and (3) I modify this regression

to allow the ERC to vary depending on the level of Δ-Users or Avg. Users, respectively. I use

scaled and centered decile ranks of Δ-Users and Avg. Users in these analyses. Thus, these variables

take values from −0.5 to 0.5. I then interact these variables with Earnings Surprise to allow the

ERC to vary with the change in, or level of, Robinhood users. I cluster standard errors by firm in

all regressions. Table 3 gives the results of these regressions. In this table, the RH-Users variable

is Δ-Users in column (2) and Avg. Users in column (3).

11

Electronic copy available at: https://ssrn.com/abstract=3833565

 Table 3 shows that, as expected, Earnings Surprise is strongly related to the CAR in the

three-day window centered on the earnings announcement in baseline regression in column (1).

Column (2) shows how this relation varies with the change in Robinhood users during the three-

day window around the earnings announcement. The positive and significant coefficient on the

interaction of Earnings Surprise*Δ-Users shows that ERCs are more positive when there is a

greater increase in Robinhood users during the earnings announcement. I also estimate a variation

of equation (2), including indicators for each decile of Δ-Users interacted with Earnings Surprise.

Results of this specification are represented in Figure 1, panel A. The figure shows ERCs increase

with Δ-Users, particularly from the sixth decile up. Note that the mean of Δ-Users is positive in

the sixth decile and higher, as shown in Table 2, panel A. This stronger market response to earnings

when the change in Robinhood users is more positive could be consistent with this type of investor

facilitating faster and more complete pricing of earnings information. Alternatively, it could reflect

an overreaction to earnings announcement events. Later analyses attempt to distinguish between

these interpretations.

Column (3) of Table 3 gives the results of estimating equation (3). Results show a

significantly negative coefficient on the interaction of Earnings Surprise with Avg. Users. Thus,

the ERC is lower when the Robinhood user base is larger in the pre-earnings announcement period.

I also estimate equation (3) with indicators for each decile of Avg. Users, instead of the rank

variable. Figure 1, panel B illustrates these results, showing a downward trend in ERCs as

Avg. Users increases. Overall, the results show a higher concentration of relatively unsophisticated

users in a firm’s investor base result in a weaker market response to earnings releases.

In Table 4, I examine several alternative specifications to test the robustness of the results

in Table 3. In particular, I control for firm size in alternative ways. First, in panel A, I scale the

12

Electronic copy available at: https://ssrn.com/abstract=3833565

Δ-Users and Avg. Users variables by MVE before constructing the ranked variables. Second, in

panel B, I interact ranks of Δ-Users or Avg. Users with the ranks of MVE to allow these associations

to vary with firm size. The results indicate the inferences of Table 3 are robust to these alternative

specifications.

Compared to Table 3, panel A of Table 4 shows similar levels of statistical significance on

the coefficient of Earnings Surprise interacted with either Δ-Users or Avg. Users when the ranks

of these variables are based on the user variables scaled by the market value of equity. Similarly,

panel B of Table 4 confirms results are not concentrated in firms of a certain size. Column (2) of

panel B shows the result of more positive ERCs for firm-quarters with higher Δ-Users becomes

more pronounced for larger firms. Column (3) again shows that ERCs are less positive when

Avg. Users is higher. The coefficient on the three-way interaction of Earnings Surprise*Avg.

Users*MVE is not significantly different from zero, proving no evidence this association varies

with firm size. However, ERCs are in general more positive for larger firms, as indicated by the

significantly positive coefficient on Earnings Surprise*MVE.

Again, note that the decile rank variables of Δ-Users, Avg. Users, and MVE are centered,

taking values from -0.5 to 0.5. To provide additional context, the ERC for the second decile of

Δ-Users is 0.27 (t-stat 6.54) when MVE is in the second decile and 0.33 (t-stat 2.11) when MVE is

in the ninth decile. The ERC for the ninth decile of Δ-Users is 0.66 (t-stat 11.35) when MVE is in

the second decile and 1.60 (t-stat 5.98) when MVE is in the ninth decile. Likewise, for Avg. Users,

the ERC for the second decile is 0.71 (t-stat 8.40) when MVE is in the second decile and 1.93

(t-stat 4.07) when MVE is in the ninth decile. When Avg. Users is in the ninth decile, the ERC is

0.38 (t-stat 9.54) for the second decile of MVE and 0.87 (t-stat 5.01) for the ninth decile of MVE.

13

Electronic copy available at: https://ssrn.com/abstract=3833565

Thus, panel B of Table 4 shows the results from Table 3 are not confined to only very large or very

small firms.

Positive Versus Negative Earnings Surprises

Next, I examine how ERCs vary with Δ-Users and Avg. Users depending on the sign of

Earnings Surprise. If retail investors react to earnings news at the earnings announcement, one

might expect more retail buying for more positive earnings announcements and less buying—or

potentially more selling—for more negative earnings announcements. However, if retail investors

respond primarily to the visibility the earnings announcement affords the firm, one would expect

a greater increase in retail investors for more extreme earnings surprises, whether positive or

negative (Lee 1992; Barber and Odean 2008; Bushee, Cedergren, and Michels 2020).

To examine ERCs by the sign of the earnings surprise, I interact key variables with

Pos. Surprise, an indicator taking a value of one when Earnings Surprise is positive and zero

otherwise. Table 1 shows 60% of earnings announcements are above analyst expectations. Table

5, column (1) indicates that the ERC is not significantly different on average for positive surprises

relative to negative surprises. Column (2) shows that ERCs are less positive for higher levels of

Δ-Users when the earnings surprise is negative, as evidenced by the significantly negative

coefficient on the interaction of Earnings Surprise and Δ-Users. The effect of Δ-Users on the ERC

becomes more positive when the earnings surprise is positive, as evidenced by the significantly

positive coefficient on the three-way interaction Earnings Surprise*Δ-Users*Pos. Surprise. Thus,

the prior result in Table 3, column (2) is very much driven by positive earnings surprises. For

negative earnings surprises, the positive association between earnings announcement returns and

Earnings Surprise is attenuated at higher levels of Δ-Users.

14

Electronic copy available at: https://ssrn.com/abstract=3833565

 The results related to Avg. Users in column (3) of Table 5 show no evidence of a differential

effect of Avg. Users on ERCs depending on the sign of the earnings surprise, as the interaction of

Earnings Surprise*Avg. Users*Pos. Surprise is not statistically different from zero. Further, the

coefficient on Earnings Surprise*Avg. Users is also not significantly different from zero. Thus, I

find no evidence ERCs vary with the number of existing Robinhood users going into the earnings

announcement when splitting by the sign of the earnings surprise.

Intraday Retail Investor Trade

The results of Tables 3 and 4 indicate that changes in the number of Robinhood users during

the earnings announcement are associated with significant differences in the relation between

earnings surprises and market returns. The interpretation of this association, however, is less clear.

In this section, I try to discriminate whether retail investors respond to earnings news at the

earnings announcement or the market returns prompted by the earnings news.

To facilitate this analysis, I decompose earnings announcement returns and Δ-Users into

subcomponents using intraday data. First, I calculate the initial market response to the earnings

announcement as the return from the prior day’s closing price to the price at 1:00 pm on the

earnings announcement date. If a firm announces earnings after trading hours, I adjust the earnings

announcement date to the following trading day. For this analysis, I restrict the sample to firms

announcing earnings outside of trading hours, as indicated by IBES. As shown in Panel A of Table

6, this is the vast majority of firms in my sample (98.5%). I label this variable Opening EA Ret.

Thus, this variable represents a firm’s stock return during the first 3.5 trading hours after the firm

announces earnings.

15

Electronic copy available at: https://ssrn.com/abstract=3833565

 Second, I split Δ-Users into three components: the change in users from the end of the prior

day to 1:00 pm on the earnings announcement date, the change in users from 1:00 pm on the

earnings announcement date to the end of the earnings announcement date, and the change in users

from the end of the earnings announcement date to the end of the next trading day. Robintrack data

does not provide minute-by-minute measurements of user holdings, so I take the measurement

closest to, but preceding, 1:00 pm on the earnings announcement date when performing this

decomposition. Several firms are missing either intraday returns or user data, which results in some

attrition of the sample, as documented in panel A of Table 6. About 7% of firms have zero change

in users over the earnings announcement period.

For firms that have non-zero Δ-User, panel B of Table 6 shows the percentage of Δ-Users

that is realized in each of the three windows described in the preceding paragraph. Very little—

only 0.31%—of the change in users over the earnings announcement window occurs by 1:00 pm

on the earnings announcement date. Most of the change in users—almost 70%—occurs between

1:00 pm and the end of the day on the earnings announcement date. The remaining 30% is realized

on the next trading day. This provides initial evidence that Robinhood investors do not react

immediately to the announcement of earnings but instead react to the market returns that follow

the earnings announcement.

To examine this possibility directly, I regress decile ranks of the change in users from

1:00 pm on the earnings announcement date to the end of the earnings announcement date on

Earnings Surprise and Opening EA Ret. Table 7, panel A gives the results. Column (1) shows that

Earnings Surprise is significantly and positively associated with the change in users during the

second half of the earnings announcement day. In contrast, column (2) shows that Opening EA

Ret. is significantly and negatively associated with this change in users. Thus, more positive returns

16

Electronic copy available at: https://ssrn.com/abstract=3833565

in the first few trading hours following the announcement of earnings are associated with a lower

change in users in the intraday window that follows. Column (3) of panel A shows these

associations persist when both Earnings Surprise and Opening EA Ret. are included in the

regression.

Next, in Table 7, panel B, I allow these associations to vary by the sign of either Earnings

Surprise or Opening EA Ret. To do so, I estimate models where I fully interact all variables with

an indicator for either Earnings Surprise or Opening EA Ret. being greater than zero. The first two

columns of panel B give the results based on the sign of Earnings Surprise, and the last two

columns give the results based on the sign of Opening EA Ret. The results based on the sign of

Earnings Surprise show that Earnings Surprise is only significantly associated with the change in

users from 1:00 pm to the end of the earnings announcement date when the Earnings Surprise is

positive. This could be consistent with retail investors buying in response to positive earnings news

but not selling in response to negative earnings news. Selling for retail traders may be restricted

since they likely do not already hold many of the subject firms in their portfolios and are unlikely

to sell short. However, Opening EA Ret. is negatively associated with the change in users during

the second half of the earnings announcement date for both positive and negative earnings

surprises.

Interacting all covariates with indicators for the sign of Opening EA Ret. in the last two

columns of panel B reveals that Opening EA Ret. is negatively associated with the change in users

from 1:00 pm to the end of the earnings announcement date when this return is negative but is

positively associated with this change in users when the return is positive. Thus, more extreme

initial returns, either negative or positive, are associated with greater increases in users. Again, this

increase does not occur concurrently with the opening returns on the earnings announcement date

17

Electronic copy available at: https://ssrn.com/abstract=3833565

as very little user activity is observed in the first few trading hours following the earnings

announcement (see again Table 6, panel B). Overall, these results appear most consistent with

retail investors reacting to returns on the earnings announcement date instead of the earnings news

itself. Consistent with prior work, these traders appear suspectable to attention-induced

purchasing, buying in response to both more positive and more negative returns (e.g., Barber and

Odean 2008).

Post-Earnings Announcement Retail Investor Trade

The results of the previous sections show that retail trading activity is associated with

significant differences in the initial market response to earnings. Further, retail activity appears to

be driven by returns in the first few hours of trading following the earnings announcement instead

of the earnings release itself. In the following section, I document how this retail activity associates

with the evolution of prices over a longer horizon following the earnings announcement. Before

doing so, however, I offer evidence in this section on how retail trade evolves following the

earnings announcement. Specifically, I document how Earnings Surprise associates with the

percentage change in retail holdings over the next 50 trading days, Post-EA Δ Users.

The first column of Table 8 shows no evidence of an association between Earnings

Surprise and the change in the number of Robinhood users holding a firm’s shares over the 50

trading days following the earnings announcement. When examining this relation by the sign of

Earnings Surprise, however, the second column reveals a negative association between Earnings

Surprise and the growth in the number of users for negative earnings surprises but a more positive

(and significantly positive overall, p-value=0.01) association for positive earnings surprises. Thus,

more extreme earnings events, regardless of sign, are associated with an increase in Robinhood

traders, consistent with these traders engaging in attention-induced trade.

18

Electronic copy available at: https://ssrn.com/abstract=3833565

 Post-Earnings-Announcement Drift

In my next set of analyses, I examine how firms’ stock returns evolve in the period

following the earnings announcement depending on the level of Robinhood user engagement. In

this analysis, I sort firm-quarters into portfolios based on terciles of Δ-Users or Avg. Users, MVE,

and Earnings Surprise. Each tercile is formed independently. Within each portfolio, I then track

how returns evolve, beginning two trading days following the earnings announcement and ending

50 trading days after the earnings announcement.

I summarize these results in Figure 2 for Δ-Users and in Figure 3 for Avg. Users. Both

figures show a distinct upward drift in returns in the terciles where Δ-Users or Avg. Users is largest

and where MVE is smallest. This upward trajectory in returns exists for the most positive and the

most negative portfolios of Earnings Surprise. Thus, unlike traditional post-earnings-

announcement drift, where returns drift in the direction of the earnings surprise (e.g., Bernard and

Thomas 1989), returns for small firms drift upwards following both positive and negative earnings

surprises when the increase in the number of retail investors at the earnings announcement is

greatest.

I quantify the amount of drift in each portfolio by calculating the area between the return

lines in Figures 2 and 3 and the ending CAR on the 50th trading day following the earnings

announcement. Similar to Blankespoor, deHaan, and Zhu’s (2018) adjusted intraperiod timeliness

measure, I also adjust for return overreactions and reversals, although such overreactions are

limited in my portfolios. Unlike Blankespoor et al. (2018), I do not scale by the final cumulative

abnormal return since I wish to capture both the amount and the speed of price revision following

the earnings announcement. I label this area the “PEAD-Area.” Note these areas will be larger

when returns drift upward or downward to a greater degree and when they do so more slowly.

19

Electronic copy available at: https://ssrn.com/abstract=3833565

These areas will be smaller when there is little price revision or when the price revision happens

shortly after the earnings announcement.

As an illustrative example, consider Figure 2 where the Δ-Users Tercile=3 and the MVE

Tercile=1 (the graph in the upper right). For the most negative Earnings Surprise tercile portfolio,

the PEAD-Area is roughly equal to the area between the solid line and the horizontal line where

CAR equals 10. Table 9 reports that this area is equal to 205.264. Table 9 reports all the PEAD-

Areas related to portfolios formed on Δ-Users. Table 10 reports the PEAD-Areas for portfolios

formed on Avg. Users.

To test the statistical significance of these results, I take an approach conceptually similar

to Bushman, Smith, and Wittenberg‐Moerman (2010). Specifically, I create a null distribution for

the PEAD-Area of each portfolio. To do so, I randomize the assignment of Δ-Users and MVE

tercile pairs (or Avg. User and MVE pairs) among firms and recalculate the PEAD-Areas. I repeat

this process 10,000 times to construct a distribution of PEAD-Areas under the null hypothesis that

Δ-Users and MVE (or Avg. Users and MVE) do not matter for the path of returns post-earnings

announcement. Note that I do not randomly reassign Earnings Surprise terciles. Rather, I

randomize within each Earnings Surprise tercile.

Figure 4 presents the results of this process for the most positive Earnings Surprise tercile

and the most positive Δ-Users tercile. The results for the smallest MVE tercile are given in Panel

A. Panel B gives the results for the largest MVE tercile. The histogram in each panel gives the

distribution of the randomized PEAD-Areas. The solid line represents a kernel density estimate

for the density of the null distribution. Panel A shows that in the smallest MVE tercile the observed

PEAD-Area is far above what would be expected under the null of Δ-Users and MVE being

unrelated to the path of post-earnings-announcement returns (p-value=0.000). Panel B, however,

20

Electronic copy available at: https://ssrn.com/abstract=3833565

shows the observed PEAD-Area is not significantly outside what would be expected based on the

null distribution when examining the largest MVE tercile (p-value=0.5764). Figure 5 provides a

similar illustration, but for the most negative terciles of Earnings Surprise. Again, in the smallest

MVE tercile, the observed PEAD-Area is significantly greater than that expected under the null

hypothesis. And again, the PEAD-Area in the largest MVE tercile is not significantly outside the

null distribution.

The right panel of Table 9 gives the results of testing of the PEAD-Areas against the null

distributions for all portfolios related to Δ-Users. In the right panel, the first number in each cell

gives the difference between the PEAD-Area and the mean of the null distribution to which the

portfolio relates. P-values based on the null distribution are given in parentheses. Table 10 presents

similar results for portfolios related to Avg. Users.

These tables confirm what is apparent from visual inspection of Figures 2 and 3: the upward

drift in the highest terciles of Δ-Users or Avg. Users is significant for the smallest firms but not for

the largest firms. This again suggests that more extreme earnings surprises may prompt attention-

induced purchasing among retail trading, which can translate into elevated prices for smaller firms.

Post-Earnings-Announcement Drift: Additional Subsample Analysis

Finally, I examine the relation between retail investor trade at the earnings announcement

and future returns in several subsamples. First, I examine the degree to which the results are

concentrated in firms whose shares are costly to borrow (i.e., costly to sell short). Second, I

examine whether the results vary during quarters where retail investors received an economic

impact payment (i.e., a stimulus check) from the government, potentially relaxing a budget

constraint that would otherwise limit retail trade.

21

Electronic copy available at: https://ssrn.com/abstract=3833565

 Figure 6 gives the results when splitting the sample by firms that are relatively costly to

short. I use the same approach as Beneish et al. (2015), classifying stocks as hard-to-borrow, or

“special,” when the daily cost of borrowing score (DCBS) from Markit Securities is greater than

two. Thus, to bifurcate the sample, I use the indicator variable, Special, which takes the value of

one when a firm’s shares are relatively costly to borrow, and zero otherwise. The graphs in the

first row of Figure 6 show results when I split the sample by Special. The graphs show little post-

earnings-announcement drift for the more easily shorted firms (Special=0). For the firms that are

costly to short, there is an upward drift in all Earnings Surprise terciles (p-value<0.01).

The second row of Figure 6 shows the results when further restricting the sample to firms

with the greatest Δ-Users and the smallest MVE tercile. These charts show some evidence of an

upward drift for both more easily borrowed and more difficult to borrow stocks. However, the

upward drift is markedly larger in magnitude for the more costly to borrow stocks. The statistical

significance of the drift is also greater in when Special=1 (p-value<0.01) than when Special=0 (p-

value<0.05). In sum, Figure 6 indicates that short-selling constraints contribute to the observed

upward drift in price for firms more heavily purchased by retail investors, consistent with these

constraints limiting the ability of more sophisticated investors to correct mispricing.

Finally, Figure 7 compares results in the period following disbursement of Economic

Impact Payments (i.e., “stimulus checks”) made pursuant to the Coronavirus Aid, Relief, and

Economic Security Act (CARES Act), signed into law on March 27, 2020. The first of these checks

were received in mid-April of 2020. Thus, I divide the sample around the Stim Check period,

spanning April 13, 2020, to May 13, 2020. Fortuitously, this window overlaps with the earnings

announcement for the first quarter of 2020 for many firms.

22

Electronic copy available at: https://ssrn.com/abstract=3833565

 The graphs in the first row of Figure 7 show the post-earnings announcement returns for

firms outside the Stim Check period, comparing across the smallest and largest MVE terciles for

firms in the highest tercile of Δ-Users. Similar to the results shown in Figure 2, returns drift upward

for the smallest firms (p-value<0.01) but not for the largest firms.

The second row of Figure 7 shows the results of a similar analysis, except I restrict the

sample to within the Stim Check period. In the smallest firms, the magnitude of the drift is

amplified greatly. Overall, the results appear consistent with the Economic Impact Payments

prompting additional retail trade, perhaps by relaxing the budget constraint of retail investors. This

additional trade appears to exert upward pressure on smaller firms’ share prices.

4. Conclusion

Reduced trading costs and more accessible trading platforms are encouraging an increasing

number of retail traders to participate actively in capital markets. This paper provides initial

evidence on how this growing prevalence of retail trade associates with the pricing of accounting

earnings. To do so, I leverage data on the holdings of retail investors on the trading platform

Robinhood. I test how the number and changes in the number of Robinhood users holding a firm’s

stock during the earnings announcement is associated with earnings surprises and stock returns. I

analyze positive and negative earnings surprises separately, as well as intraday trading and returns,

to better understand whether retail investors respond to earnings news or stock returns driven by

the underlying earnings news. Finally, I examine retail trade and stock returns in the weeks that

follow the earnings announcement.

My results show activity by Robinhood traders is associated with significant changes in

the price-earnings relation. More retail traders in a firm’s investor base pre-earnings announcement

23

Electronic copy available at: https://ssrn.com/abstract=3833565

is associated with a weaker market response to earnings surprises. Net purchasing by these retail

investors during the earnings announcement window is associated with a stronger market response

to earnings surprises. However, this second result varies depending on the sign of the earnings

surprise. For negative earnings surprises, a greater increase in Robinhood users is associated with

a weaker market response to earnings. For positive earnings surprises, a greater increase in

Robinhood users is associated with a stronger market response.

Intraday analysis suggests that these retail investors respond to the initial market response

to earnings news as opposed to the earnings release itself. Little retail trade is observed in the first

trading hours following the earnings announcement. Instead, most of the retail trade appears to

occur in the second half of the trading day. This retail trade is significantly associated with returns

realized earlier in the trading day. More extreme returns, both positive and negative, are associated

with greater increases in retail holdings, consistent with attention-induced trade.

Following the earnings announcement, more extreme earnings surprises, both positive and

negative, are associated with increased retail purchasing. Moreover, when net retail purchasing is

greatest, more extreme earnings surprises—again, both positive and negative—are associated with

an upward drift in future returns for smaller firms. Overall, the results are most consistent with

retail traders responding to the visibility provided by more extreme earnings surprises instead of

the information content of earnings. Thus, despite the evidence that retail traders neglect the

information content of earnings, trade by these retail investors is still associated with significant

changes in the price-earnings relation due to how these investors respond to earnings-related price

movements.

A limitation of this study is that I only focus on a particular subset of retail investors,

namely those trading on Robinhood. Data constraints also limit my analysis to a specific time

24

Electronic copy available at: https://ssrn.com/abstract=3833565

period. My results may not generalize to other populations of retail investors or other time

windows. Further, platform-specific features of Robinhood may contribute to the results I observe.

Investor behavior driven by features of the investing platform meant to increase user engagement

may not generalize to other settings. Nevertheless, the behavior of the retail investors in my sample

may be of particular interest to regulators given the apparent inexperience of these investors. This

provides an opportunity to study the behavior of investors who perhaps stand to benefit most from

investor protections. My results suggest these investors respond to the visibility of earnings news

and the associated market returns instead of the underlying information on fundamental firm

performance, as both positive and negative events induce purchasing. This purchasing, coordinated

by the earnings announcement, appears to exert upward pressure on stock prices, particularly

among small firms and firms that are costly to short. The effect is amplified when retail traders

receive an infusion of cash in the form of an economic stimulus payment. Future work can help us

better understand if and when prices revert and whether this results in a wealth transfer away from

retail investors. Further, future work might illuminate whether these apparent price distortions

influence firm decisions and capital allocation.

25

Electronic copy available at: https://ssrn.com/abstract=3833565

Appendix A: Variable Definitions

CAR .......................... The cumulative three-day market-adjusted return centered on the earnings
announcement date.
Opening EA Ret........ The return from the prior day’s closing price to 1:00 pm on the earnings
announcement date.
Earnings Surprise .... Actual EPS less the median analyst forecast of EPS scaled by price.
Δ-Users .................... The difference between the number of Robinhood users holding a stock one
day following the earnings announcement and the number of users holding
a stock one day before the earnings announcement.
Avg. Users ................ The mean number of Robinhood users holding a stock each day during the
window beginning 65 trading days before the earnings announcement and
ending two days before the earnings announcement.
MVE ......................... Market value of equity.
Size ........................... Logarithm of MVE.
MTB.......................... MVE divided by the book value of equity.
Beta .......................... The coefficient from regressing daily firm return less the risk-free rate on
market returns less the risk-free rate over the 252 trading days ending three
trading days before the earnings announcement.
Leverage................... Total debt divided by the book value of equity.
Persistence ............... The coefficient of EPS regressed on lagged EPS within firm, using up to 10
years of data.
Loss .......................... An indicator variable taking a value of one if EPS is negative, and zero
otherwise.
Analysts .................... The number of analysts contributing to the media forecast on which
Earnings Surprise is based.
Volatility................... The standard deviation of daily returns over the period beginning 65
trading days before the earnings announcement and ending two days before
the earnings announcement.
Avg. Turnover .......... Average daily volume divided by shares outstanding during the period
beginning 65 trading days before the earnings announcement and ending
two days before the earnings announcement.
Pos. Surprise ............ An indicator variable taking a value of one if Earnings Surprise is positive,
and zero otherwise.
Post-EA Δ-Users ...... The change in the number of Robinhood users holding a firm’s stock from
the second trading day following the earnings announcement to the 50th
trading day following the earnings announcement, scaled by Avg. Users.
Special ...................... An indicator variable taking a value of one if a firm’s stock is costly to
borrow and zero otherwise (see Beneish, Lee, and Nichols 2015).

26

Electronic copy available at: https://ssrn.com/abstract=3833565

References
Ball, Ray, and Philip Brown. 1968. “An Empirical Evaluation of Accounting Income
Numbers.” Journal of Accounting Research 6 (2): 159–78. https://doi.org/10.2307/2490232.
Barber, Brad M., Xing Huang, Terrance Odean, and Christopher Schwarz. 2021.
“Attention Induced Trading and Returns: Evidence from Robinhood Users.” SSRN Scholarly
Paper ID 3715077. Rochester, NY: Social Science Research Network.
https://doi.org/10.2139/ssrn.3715077.
Barber, Brad M., and Terrance Odean. 2002. “Online Investors: Do the Slow Die First?”
The Review of Financial Studies 15 (2): 455–88. https://doi.org/10.1093/rfs/15.2.455.
———. 2008. “All That Glitters: The Effect of Attention and News on the Buying
Behavior of Individual and Institutional Investors.” The Review of Financial Studies 21 (2): 785–
818. https://doi.org/10.1093/rfs/hhm079.
Barber, Brad M., Terrance Odean, and Ning Zhu. 2009. “Do Retail Trades Move
Markets?” Review of Financial Studies 22 (1): 151–86. https://doi.org/10.1093/rfs/hhn035.
Beck, Philippe van der, and Coralie Jaunin. 2021. “The Equity Market Implications of the
Retail Investment Boom.” SSRN Electronic Journal. https://doi.org/10.2139/ssrn.3776421.
Beneish, M. D., C. M. C. Lee, and D. C. Nichols. 2015. “In Short Supply: Short-Sellers
and Stock Returns.” Journal of Accounting and Economics 60 (2): 33–57.
https://doi.org/10.1016/j.jacceco.2015.08.001.
Bernard, Victor L., and Jacob K. Thomas. 1989. “Post-Earnings-Announcement Drift:
Delayed Price Response or Risk Premium?” Journal of Accounting Research 27: 1–36.
https://doi.org/10.2307/2491062.
Blankespoor, Elizabeth, Ed Dehaan, John Wertz, and Christina Zhu. 2019. “Why Do
Individual Investors Disregard Accounting Information? The Roles of Information Awareness
and Acquisition Costs.” Journal of Accounting Research 57 (1): 53–84.
https://doi.org/10.1111/1475-679X.12248.
Blankespoor, Elizabeth, Ed deHaan, and Christina Zhu. 2018. “Capital Market Effects of
Media Synthesis and Dissemination: Evidence from Robo-Journalism.” Review of Accounting
Studies 23 (1): 1–36. https://doi.org/10.1007/s11142-017-9422-2.
Bushee, Brian, Matthew Cedergren, and Jeremy Michels. 2020. “Does the Media Help or
Hurt Retail Investors during the IPO Quiet Period?” Journal of Accounting and Economics 69
(1): 101261. https://doi.org/10.1016/j.jacceco.2019.101261.
Bushman, Robert M., Abbie J. Smith, and Regina Wittenberg‐Moerman. 2010. “Price
Discovery and Dissemination of Private Information by Loan Syndicate Participants.” Journal of
Accounting Research 48 (5): 921–72. https://doi.org/10.1111/j.1475-679X.2010.00384.x.

27

Electronic copy available at: https://ssrn.com/abstract=3833565

 Choi, James J., David Laibson, and Andrew Metrick. 2002. “How Does the Internet
Affect Trading? Evidence from Investor Behavior in 401(k) Plans.” Journal of Financial
Economics 64 (3): 397–421. https://doi.org/10.1016/S0304-405X(02)00130-7.
Collins, Daniel W., and S. P. Kothari. 1989. “An Analysis of Intertemporal and Cross-
Sectional Determinants of Earnings Response Coefficients.” Journal of Accounting and
Economics, SPECIAL ISSUE, 11 (2): 143–81. https://doi.org/10.1016/0165-4101(89)90004-9.
Dechow, Patricia, Weili Ge, and Catherine Schrand. 2010. “Understanding Earnings
Quality: A Review of the Proxies, Their Determinants and Their Consequences.” Journal of
Accounting and Economics 50 (2–3): 344–401. https://doi.org/10.1016/j.jacceco.2010.09.001.
Hirshleifer, David A., James N. Myers, Linda A. Myers, and Siew Hong Teoh. 2008. “Do
Individual Investors Cause Post-Earnings Announcement Drift? Direct Evidence from Personal
Trades.” The Accounting Review 83 (6): 1521–50. https://doi.org/10.2308/accr.2008.83.6.1521.
Ke, Bin, and Santhosh Ramalingegowda. 2005. “Do Institutional Investors Exploit the
Post-Earnings Announcement Drift?” Journal of Accounting and Economics 39 (1): 25–53.
https://doi.org/10.1016/j.jacceco.2004.02.002.
Lee, Charles M. C. 1992. “Earnings News and Small Traders.” Journal of Accounting
and Economics 15 (2): 265–302. https://doi.org/10.1016/0165-4101(92)90021-S.
Luo, Yuanzhi. 2005. “Do Insiders Learn from Outsiders? Evidence from Mergers and
Acquisitions.” The Journal of Finance 60 (4): 1951–82. https://doi.org/10.1111/j.1540-
6261.2005.00784.x.
Martin, Patrick R., and Donald V. Moser. 2016. “Managers’ Green Investment
Disclosures and Investors’ Reaction.” Journal of Accounting and Economics 61 (1): 239–54.
https://doi.org/10.1016/j.jacceco.2015.08.004.
McCabe, Caitlin. 2021. “GameStop Shares Swing Sharply in Volatile Trading Session.”
Wall Street Journal, March 10, 2021, sec. Markets. https://www.wsj.com/articles/gamestop-
shares-swing-sharply-in-volatile-trading-session-11615408279.
Michaels, Dave. 2020. “Bankrupt Hertz Suspends $500 Million Stock Sale as SEC Poses
Questions on Deal.” Wall Street Journal, June 17, 2020, sec. Markets.
https://www.wsj.com/articles/bankrupt-hertz-suspends-500-million-stock-sale-as-sec-poses-
questions-on-deal-11592424158.
Moss, Austin, James P. Naughton, and Clare Wang. 2020. “The Irrelevance of ESG
Disclosure to Retail Investors: Evidence from Robinhood.” SSRN Scholarly Paper ID 3604847.
Rochester, NY: Social Science Research Network. https://doi.org/10.2139/ssrn.3604847.
Osipovich, Alexander. 2020. “Individual-Investor Boom Reshapes U.S. Stock Market.”
Wall Street Journal, August 31, 2020, sec. Markets. https://www.wsj.com/articles/individual-
investor-boom-reshapes-u-s-stock-market-11598866200.

28

Electronic copy available at: https://ssrn.com/abstract=3833565

 ———. 2021. “GameStop Frenzy Emboldens Supporters of Stock-Trading Tax.” Wall
Street Journal, March 7, 2021, sec. Markets. https://www.wsj.com/articles/gamestop-frenzy-
emboldens-supporters-of-stock-trading-tax-11615132800.
Phillips, Matt. 2021. “Hungry for Investors, Some Companies Woo the Little Guy.” The
New York Times, April 13, 2021, sec. Business.
https://www.nytimes.com/2021/04/13/business/stock-market-investors.html.
Popper, Nathaniel. 2020. “Robinhood Has Lured Young Traders, Sometimes With
Devastating Results.” The New York Times, July 8, 2020, sec. Technology.
https://www.nytimes.com/2020/07/08/technology/robinhood-risky-trading.html.
SEC. 2021. “Statement of Acting Chair Lee and Commissioners Peirce, Roisman, and
Crenshaw Regarding Recent Market Volatility.” January 29, 2021.
https://www.sec.gov/news/public-statement/joint-statement-market-volatility-2021-01-29.
Tenev, Vladimir. 2021. Testimony of Vladimir Tenev, Robinhood Markets, Inc.
https://docs.house.gov/meetings/BA/BA00/20210218/111207/HHRG-117-BA00-Wstate-
TenevV-20210218.pdf.

29

Electronic copy available at: https://ssrn.com/abstract=3833565

Table 1: Summary Statistics
Count Mean Std. Dev. p25 Median p75
CAR 22,906 -0.00 0.10 -0.05 0.00 0.05
Earnings Surprise 22,906 -0.00 0.03 -0.00 0.00 0.00
Δ-Users 22,906 44.74 236.71 -6.00 0.00 11.00
Avg. Users 22,906 2,448.96 8,331.71 109.80 353.32 1,212.18
MVE 22,906 8.51 24.32 0.33 1.33 4.88
Size 22,906 20.98 2.01 19.60 21.01 22.31
MTB 22,906 3.56 7.54 1.16 2.15 4.46
Beta 22,906 1.08 0.45 0.80 1.07 1.35
Leverage 22,906 0.82 2.74 0.12 0.53 1.19
Persistence 22,906 0.28 0.33 0.04 0.25 0.52
Loss 22,906 0.38 0.48 0.00 0.00 1.00
Analysts 22,906 8.11 6.60 3.00 6.00 11.00
Volatility 22,906 0.03 0.02 0.02 0.03 0.04
Avg. Turnover 22,906 0.01 0.01 0.00 0.01 0.01
Pos. Surprise 22,906 0.60 0.49 0.00 1.00 1.00
Post-EA Δ-Users 22,795 0.23 0.78 -0.05 0.03 0.20
Special 20,221 0.11 0.31 0.00 0.00 0.00

All variable definitions are given in Appendix A

30

Electronic copy available at: https://ssrn.com/abstract=3833565

Table 2: Tabulations of Key Variables
Panel A: Means of key variables by decile
Decile: MVE Δ-Users Avg. Users
1 0.04 -114.49 24.37
2 0.15 -16.24 60.44
3 0.32 -6.02 111.16
4 0.58 -2.42 182.66
5 1.01 -0.45 285.15
6 1.68 1.43 444.88
7 2.77 4.22 722.08
8 4.90 12.19 1,231.62
9 11.16 45.66 2,585.63
10 61.52 527.38 18,752.33
Total 8.51 44.74 2,448.96

Panel B: Observations by terciles of MVE and Δ-Users

Terciles of Δ-Users
Terciles of MVE 1 2 3 Total
1 2,410 2,880 2,289 7,579
2 2,597 2,748 2,268 7,613
3 3,217 1,691 2,806 7,714
Total 8,224 7,319 7,363 22,906

Panel C: Observations by terciles of MVE and Avg. Users

Terciles of Avg. Users
Terciles of MVE 1 2 3 Total
1 2,660 2,514 2,405 7,579
2 3,356 2,475 1,782 7,613
3 1,614 2,625 3,475 7,714
Total 7,630 7,614 7,662 22,906

31

Electronic copy available at: https://ssrn.com/abstract=3833565

Table 3: ERCs by Change in Users and Average Users Deciles
Dependent Variable: CAR
(1) (2) (3)
RH-Users variable (scaled rank) = Δ-Users Avg. Users
Earnings Surprise 0.473*** 0.481*** 0.543***
(13.99) (14.25) (12.73)
RH-Users -0.052*** -0.003
(-19.30) (-1.11)
Earnings Surprise*RH-Users 0.522*** -0.425***
(5.34) (-3.41)
Size -0.001** -0.001 -0.001*
(-2.13) (-1.48) (-1.82)
MTB 0.000 0.000 0.000
(0.99) (0.16) (1.14)
Beta 0.005*** 0.005*** 0.005***
(2.80) (2.96) (2.78)
Leverage 0.000 0.000 0.000
(0.72) (1.43) (0.61)
Persistence -0.001 -0.002 -0.001
(-0.35) (-1.06) (-0.42)
Loss -0.025*** -0.024*** -0.024***
(-13.58) (-13.44) (-12.41)
Ln(Analysts) 0.005*** 0.005*** 0.005***
(3.48) (3.52) (3.63)
Volatility 0.369*** 0.417*** 0.376***
(7.89) (8.91) (7.90)
Avg. Turnover -0.530*** -0.577*** -0.527***
(-6.23) (-6.64) (-6.02)
Constant 0.017 0.007 0.013
(1.37) (0.53) (0.97)
Adj. R-Sq. 0.037 0.068 0.038
N 22,906 22,906 22,906
* p<0.10, ** p<0.05, *** p<0.01 (for two-tailed tests, t-stats in parentheses)
The dependent variable is the three-day market-adjusted abnormal return centered on the earnings
announcement date. Earnings Surprise is EPS less the median analyst forecast, scaled by price.
Δ-Users and Avg. Users are in centered and scaled decile ranks, taking values from -0.5 to 0.5.
Full variable definitions are given in Appendix A.

32

Electronic copy available at: https://ssrn.com/abstract=3833565

Table 4: Alternative Specifications
Panel A: Scaling by Firm Size
(1) (2) (3)
RH-Users variable (scaled rank) = Δ-Users/MVE Avg. Users/MVE
Earnings Surprise 0.473*** 0.481*** 0.910***
(13.99) (14.38) (10.74)
RH-Users -0.047*** 0.004
(-16.19) (1.15)
Earnings Surprise*RH-Users 0.379*** -1.122***
(5.17) (-5.92)
Controls Yes Yes Yes
Adj. R-Sq. 0.037 0.062 0.041
N 22,906 22,906 22,906

Panel B: Interacting Firm Size

(1) (2) (3)
RH-Users variable (scaled rank) = Δ-Users Avg. Users
Earnings Surprise 0.810*** 0.716*** 0.971***
(9.76) (9.26) (7.69)
RH-Users -0.051*** -0.005*
(-17.28) (-1.67)
Earnings Surprise*RH-Users 1.074*** -0.893**
(4.78) (-2.48)
Earnings Surprise*RH-Users*MVE (scaled rank) 1.460*** -1.221
(2.72) (-1.41)
MVE (scaled rank) -0.028*** -0.037*** -0.028***
(-3.10) (-4.06) (-2.72)
Earnings Surprise*MVE (scaled rank) 0.869*** 0.641*** 1.103***
(4.53) (3.51) (3.73)
RH-Users*MVE (scaled rank) -0.042*** 0.008
(-5.21) (1.03)
Controls Yes Yes Yes
Adj. R-Sq. 0.039 0.072 0.041
N 22,906 22,906 22,906
* p<0.10, ** p<0.05, *** p<0.01 (for two-tailed tests, t-stats in parentheses)
The dependent variable is the three-day market-adjusted abnormal return centered on the earnings
announcement date. Earnings Surprise is EPS less the median analyst forecast, scaled by price.
Δ-Users, Avg. Users, and MVE related variables are in centered and scaled decile ranks, taking
values from -0.5 to 0.5. Full variable definitions are given in Appendix A.

33

Electronic copy available at: https://ssrn.com/abstract=3833565

Table 5: Positive Versus Negative Earnings Surprises
Dependent Variable: CAR
(1) (2) (3)
RH-Users variable (scaled rank) = Δ-Users Avg. Users
Earnings Surprise 0.251*** 0.250*** 0.275***
(5.38) (5.36) (4.86)
Pos. Surprise 0.041*** 0.041*** 0.040***
(26.55) (26.90) (25.47)
Earnings Surprise*Pos. Surprise -0.105 -0.098 -0.063
(-1.30) (-1.22) (-0.68)
Earnings Surprise*RH-Users -0.280** -0.086
(-2.13) (-0.53)
Earnings Surprise*RH-Users*Pos. Surprise 1.165*** -0.157
(5.32) (-0.59)
RH-Users*Pos. Surprise 0.069*** -0.015***
(12.22) (-3.24)
RH-Users -0.101*** 0.006
(-22.26) (1.49)
Controls Yes Yes Yes
Adj. R-Sq 0.068 0.109 0.068
N 22,906 22,906 22,906
* p<0.10, ** p<0.05, *** p<0.01 (for two-tailed tests, t-stats in parentheses)
The dependent variable is the three-day market-adjusted abnormal return centered on the earnings
announcement date. Earnings Surprise is EPS less the median analyst forecast, scaled by price.
Pos. Surprise is an indicator for Earnings Surprise being greater than zero. Δ-Users and Avg. Users
are both in centered and scaled decile ranks, taking values from -0.5 to 0.5. Full variable definitions
are given in Appendix A.

34

Electronic copy available at: https://ssrn.com/abstract=3833565

Table 6: Intraday Retail Investor Trade
Panel A: Intraday Sample
Observations % of Full Sample
Full Sample 22,906 100.00%
Less: Announce Outside of Trading Hours 341 1.49%
Less: Missing Intraday Trade Data 345 1.51%
Less: Missing Intraday Return Data 91 0.40%
Intraday Sample (for Table 7) 22,129 96.61%
Less: Δ-Users = 0 1,585 6.92%
Sample for Panel B (below) 20,544 89.69%

Panel B: Percentage Δ-Users Realized in Different Time-Windows

End of prior day to 1pm on EA Date to End of EA date to
Window: 1pm on EA Date end of EA Date end of next day
Percent: 0.31% 69.92% 29.77%

35

Electronic copy available at: https://ssrn.com/abstract=3833565

Table 7: Change in Users, Earnings Surprise, and Prior Intraday Returns
Panel A: Change in Users: From 1pm to End of EA Date
(1) (2) (3)
Earnings Surprise 3.376*** 5.770***
(4.36) (7.19)
Opening EA Ret. -5.388*** -5.672***
(-17.24) (-17.98)
Size 0.060*** 0.054** 0.058***
(2.82) (2.57) (2.74)
MTB -0.020*** -0.019*** -0.019***
(-4.88) (-4.70) (-4.94)
Beta 0.027 0.067 0.055
(0.48) (1.21) (1.00)
Leverage 0.043*** 0.041*** 0.043***
(4.08) (3.96) (4.19)
Persistence -0.201*** -0.206*** -0.207***
(-2.81) (-2.93) (-2.94)
Loss 0.046 -0.116** -0.076
(0.87) (-2.19) (-1.44)
Ln(Analysts) 0.041 0.062 0.054
(1.01) (1.55) (1.36)
Volatility 6.816*** 8.199*** 8.577***
(4.79) (5.77) (6.01)
Avg. Turnover 2.162 -0.223 -0.224
(0.82) (-0.08) (-0.08)
Constant 3.849*** 3.920*** 3.855***
(9.23) (9.47) (9.33)
Adj. R-Sq. 0.007 0.030 0.033
N 22,129 22,129 22,129
* p<0.10, ** p<0.05, *** p<0.01 (for two-tailed tests, t-stats in parentheses)
The dependent variable in all models is the change in the number of users between 1:00 pm and
the end of day on the earnings announcement in decile ranks. Earnings Surprise is EPS less the
median analyst forecast, scaled by price. Opening EA Ret. is the return from the prior day’s
closing price to 1:00 pm on the earnings announcement date. Full variable definitions are given
in Appendix A.

36

Electronic copy available at: https://ssrn.com/abstract=3833565

Table 7: Change in Users, Earnings Surprise, and Prior Intraday Returns (continued)
Panel B: By sign of Earnings Surprise or Initial Intraday Returns
Earnings Surprise Opening EA Ret.
<= 0 >0 <= 0 >0
Earnings Surprise 0.322 6.638*** 3.831*** 6.866***
(0.27) (3.84) (3.88) (6.02)
Opening EA Ret. -10.999*** -2.403*** -23.082*** 14.674***
(-23.15) (-5.64) (-56.25) (28.05)
Size 0.136*** 0.048* 0.330*** 0.024
(4.77) (1.72) (13.22) (0.91)
MTB -0.026*** -0.012** -0.024*** -0.024***
(-4.19) (-2.38) (-5.12) (-4.40)
Beta 0.101 -0.050 -0.077 -0.452***
(1.29) (-0.68) (-1.12) (-6.26)
Leverage 0.049*** 0.030** 0.048*** 0.045***
(3.25) (2.17) (3.89) (3.18)
Persistence -0.096 -0.235*** -0.074 -0.133
(-1.01) (-2.66) (-0.89) (-1.53)
Loss -0.071 -0.025 -0.074 -0.122*
(-0.99) (-0.35) (-1.19) (-1.80)
Ln(Analysts) 0.080 0.033 0.107** -0.173***
(1.51) (0.64) (2.25) (-3.40)
Volatility 9.170*** 9.184*** 6.190*** 10.395***
(4.61) (4.58) (3.48) (5.66)
Avg. Turnover -11.515*** 3.622 -10.818*** -8.637**
(-3.00) (1.05) (-3.26) (-2.18)
Constant 1.919*** 4.129*** -2.604*** 4.210***
(3.45) (7.47) (-5.25) (8.06)
Adj. R-Sq. 0.051 0.197
N 22,129 22,129
* p<0.10, ** p<0.05, *** p<0.01 (for two-tailed tests, t-stats in parentheses)
In the first two columns, all variables are interacted with indicators for Earnings Surprise being
either less than or equal to, or greater than, zero. In the last two columns, all variables are interacted
with indicators for Opening EA Ret. being less than or equal to, or greater than, zero. The
dependent variable in both models is the change in the number of users between 1:00 pm and the
end of day on the earnings announcement in decile ranks. Full variable definitions are given in
Appendix A.

37

Electronic copy available at: https://ssrn.com/abstract=3833565

Table 8: Post-Earnings Announcement Retail Investor Trade
Dep. Var.: Post-EA Δ-Users Post-EA Δ-Users
Earnings Surprise -0.329 -1.324***
(-1.25) (-3.66)
Pos. Surprise -0.027**
(-2.31)
Pos. Surprise*Earnings Surprise 3.189***
(4.40)
CAR -0.270*** -0.250***
(-4.05) (-3.70)
Δ-Users 0.121*** 0.119***
(7.09) (6.97)
Size -0.032*** -0.026***
(-6.94) (-5.55)
MTB -0.003*** -0.003***
(-4.02) (-3.27)
Beta 0.015 0.018
(1.00) (1.16)
Leverage 0.007** 0.006**
(2.50) (2.01)
Ln(Analysts) 0.049*** 0.051***
(5.07) (5.29)
Volatility 3.709*** 3.216***
(9.23) (7.71)
Avg. Turnover -3.827*** -4.220***
(-6.54) (-7.12)
Constant 0.728*** 0.614***
(7.95) (6.62)
Adj. R-Sq. 0.023 0.025
N 22,795 22,795
* p<0.10, ** p<0.05, *** p<0.01 (for two-tailed tests, t-stats in parentheses)
The dependent variable, Post-EA Δ-Users, is the change in the number of Robinhood users holding
a firm’s stock from the second trading day following the earnings announcement to the 50th trading
day following the earnings announcement, scaled by Avg. Users. Earnings Surprise is EPS less
the median analyst forecast, scaled by price. Pos. Surprise is an indicator for Earnings Surprise
being greater than zero. CAR is the three-day cumulative market-adjusted return centered on the
earnings announcement. Δ-Users is the scaled decile rank of the change in users over the earnings
announcement window, taking values from -0.5 to 0.5. Full variable definitions are given in
Appendix A.

38

Electronic copy available at: https://ssrn.com/abstract=3833565

Table 9: PEAD-Areas by Δ-Users, MVE, and Earnings Surprise Terciles
PEAD-Areas Test of PEAD-Areas
Δ-Users Tercile Δ-Users Tercile
MVE EarnSurp 1 2 3 1 2 3
Tercile Tercile:
46.009 15.536 205.264 18.070 -11.635 175.638
1
(0.1758) (0.6938) (0.0000)
104.539 46.160 206.911 76.325 23.132 175.111
1 2
(0.0096) (0.0760) (0.0010)
136.717 17.210 233.416 84.947 -34.163 181.780
3
(0.0043) (0.2219) (0.0000)
22.347 69.403 38.667 -8.713 38.646 7.646
1
(0.6170) (0.0434) (0.3538)
28.052 69.610 29.731 9.375 51.795 7.079
2 2
(0.1738) (0.0019) (0.2503)
40.771 47.219 39.451 -11.637 -4.597 -12.822
3
(0.8667) (0.7630) (0.8310)
16.298 15.738 16.850 -17.583 -24.670 -16.096
1
(0.6506) (0.5157) (0.6300)
3 8.434 23.100 9.729 -5.883 5.159 -5.837
2
(0.4296) (0.2873) (0.5739)
12.247 62.565 30.827 -40.807 1.910 -22.362
3
(0.2238) (0.4476) (0.5764)
p-values in parentheses.
Compare values to charts in Figure 2. In the left panel, numbers reflect the PEAD-Area associated
with each portfolio. In the right panel, the first number in each cell is the difference between the
PEAD-Area and the mean of the null distribution to which the portfolio relates. P-values are in
parentheses.

39

Electronic copy available at: https://ssrn.com/abstract=3833565

Table 10: PEAD-Areas by Avg-Users, MVE, and Earnings Surprise Terciles
PEAD-Areas Test of PEAD-Areas
Avg-Users Tercile Avg-Users Tercile
MVE EarnSurp 1 2 3 1 2 3
Tercile Tercile:
30.982 48.566 212.525 3.216 20.553 183.900
1
(0.4736) (0.1468) (0.0000)
26.268 126.435 283.070 3.198 98.242 250.568
1 2
(0.3635) (0.0030) (0.0000)
31.783 144.528 246.227 -19.921 93.062 194.712
3
(0.6393) (0.0031) (0.0000)
98.552 22.112 66.948 70.007 -9.439 33.299
1
(0.0061) (0.6176) (0.0849)
80.210 24.017 150.310 64.015 3.960 124.322
2 2
(0.0000) (0.3413) (0.0010)
56.486 38.751 68.742 4.671 -13.168 15.598
3
(0.6718) (0.7768) (0.3630)
56.430 14.047 24.107 15.900 -20.037 -8.641
1
(0.2571) (0.5423) (0.6150)
3 56.957 13.879 27.706 38.676 -1.638 13.518
2
(0.0064) (0.8169) (0.0421)
62.592 38.270 41.754 .546 -16.787 -10.344
3
(0.4494) (0.5365) (0.7783)
p-values in parentheses.
Compare values to charts in Figure 3. In the left panel, numbers reflect the PEAD-Area associated
with each portfolio. In the right panel, the first number in each cell is the difference between the
PEAD-Area and the mean of the null distribution to which the portfolio relates. P-values are in
parentheses.

40

Electronic copy available at: https://ssrn.com/abstract=3833565

Figure 1: ERCs by Change in Users and Average Users

Note: Whiskers represent 95% confidence intervals.

41

Electronic copy available at: https://ssrn.com/abstract=3833565

 Figure 2: By Δ-Users and MVE: Cumulative Abnormal Returns (CAR) Over 50 Trading Days Following EA
Electronic copy available at: https://ssrn.com/abstract=3833565

42
 Figure 3: By Avg. Users and MVE: Cumulative Abnormal Returns (CAR) Over 50 Trading Days Following EA
Electronic copy available at: https://ssrn.com/abstract=3833565

43
Figure 4: Null Distributions and Hypothesis Testing for Most Positive Earnings Surprise and
Most Positive Δ-Users Tercile
Panel A: Smallest MVE Tercile

Actual PEAD-Area = 233.416

(from Table 9)
p-value = 0.0000

Panel B: Largest MVE Tercile

Actual PEAD-Area = 30.827

(from Table 9)
p-value = 0.5764

44

Electronic copy available at: https://ssrn.com/abstract=3833565

Figure 5: Null Distributions and Hypothesis Testing for Most Negative Earnings Surprise
and Most Positive Δ-Users Tercile
Panel A: Smallest MVE Tercile

Actual PEAD-Area = 205.264

(from Table 9)
p-value = 0.0000

Panel B: Largest MVE Tercile

Actual PEAD-Area = 16.850

(from Table 9)
p-value = 0.6300

45

Electronic copy available at: https://ssrn.com/abstract=3833565

 Figure 6: Cost of Short Selling and CAR Over 50 Trading Days Following EA
Electronic copy available at: https://ssrn.com/abstract=3833565

46
 47
Figure 7: Stimulus Check Period Versus Rest of Sample
Electronic copy available at: https://ssrn.com/abstract=3833565