Information Discovery, Interpretation, and Analysis by Institutional Investors Around Earnings Announcements

Abstract

This study examines how institutional investors allocate trading across the earnings announcement cycle and whether industry trading concentration strengthens that activity. The analysis is motivated by two complementary ideas: public disclosures can increase the value of investors’ prior information, and even sophisticated investors face costly information processing. These perspectives imply that institutional trading need not be concentrated only before disclosure and may be strongest after earnings announcements, when investors combine newly released public information with prior firm- and industry-specific signals. Using daily institutional trading data from Ancerno, we find that institutional net trading is positively related to earnings surprises before, during, and after earnings announcements, with the strongest relation occurring in the post-announcement period. We also document a clear asymmetry: trading is strongly related to positive earnings surprises across all three stages, whereas trading related to negative earnings surprises is concentrated mainly after disclosure. In addition, industry trading concentration strengthens the relation between institutional trading and earnings news across the announcement cycle, especially for positive surprises. These findings provide an integrated view of institutional information processing around a major recurring disclosure event, show that the timing of institutional trading is informative about how earnings news is incorporated into prices, and support the view that industry specialization is linked to stronger earnings-related trading.

1. Introduction

Earnings announcements are among the most important recurring disclosure events in capital markets. Because they are pre-scheduled, widely followed, and highly value-relevant, they create a natural sequence of investor decisions. Before the announcement, investors can gather and refine information about likely earnings news. At the announcement, they must interpret the released signal and assess its implications for firm value. After the announcement, they may continue analyzing whether the news is temporary or persistent and how it affects valuation. Institutional investors are especially important in this setting because their analytical resources, industry expertise, and trading capacity should allow them to play a central role throughout this information-processing cycle. Yet existing evidence provides limited insight into how institutional investors allocate their trading across these stages within a single disclosure setting. This study examines institutional trading before, during, and after quarterly earnings announcements and investigates whether industry trading concentration shapes trading across the full earnings announcement cycle.

This question matters because prior research links institutional ownership to more efficient prices, but provides less direct evidence on the channels through which institutions contribute to price formation. Boehmer and Kelley (2009) show that greater institutional ownership is associated with more informationally efficient prices, and prior work also links institutional presence to weaker return anomalies around earnings news (Bartov et al., 2000; Ali et al., 2004; Ke & Ramalingegowda, 2005).1 If institutional investors improve price efficiency through superior information processing, then the timing of their trading around earnings announcements should reveal when that processing is most valuable and how institutions help move prices toward fundamentals.

Our predictions are motivated by two complementary theoretical ideas. First, public disclosure can increase, rather than eliminate, the value of investors’ prior information. In models of information complementarities, a public signal helps informed investors interpret their existing private or industry-specific information more precisely (Kim & Verrecchia, 1994; Cheynel & Levine, 2020). Under this view, informed trading need not be concentrated before the announcement. Instead, trading can intensify after disclosure, when the earnings release allows investors to complete an information mosaic and refine valuation. Thus, we expect institutional trading to be related to earnings surprises not only before the announcement, but also during and after the announcement.

Second, information processing is costly, even for sophisticated investors. Earnings announcements do not simply reveal whether earnings are above or below expectations; they also require investors to assess persistence, source, and implications for future cash flows. When attention and processing capacity are limited, investors will allocate analytical effort selectively and may not resolve the full implications of the signal immediately (Hirshleifer et al., 2011). This implies that institutional trading should not be evenly distributed across event time. Instead, if the announcement lowers interpretation costs and makes follow-on analysis more productive, then institutions should trade most aggressively after disclosure. Accordingly, we predict that institutional trading will be associated with earnings surprises throughout the earnings announcement cycle, but that the strongest response will occur in the post-announcement period.

These same mechanisms also imply that institutional trading need not respond symmetrically to positive and negative earnings surprises. The asymmetry arises because good and bad earnings news differ not only in sign but also in how easily they can be interpreted and implemented in trades. Favorable surprises are generally easier to act on quickly: institutions can scale long positions immediately, and positive news often maps more directly into upward revisions in expected cash flows or growth opportunities. By contrast, unfavorable surprises are often more costly to exploit. Acting on bad news may require reducing existing positions, rebalancing under mandate constraints, or taking short positions, all of which can be more frictional than buying on good news (Johnson & So, 2018). Negative earnings news is also often more ambiguous. A shortfall may reflect a temporary operational disruption, conservative reporting, transitory macro conditions, or a more persistent deterioration in fundamentals. That ambiguity raises the value of verification and follow-on analysis before investors trade aggressively. Prior research is consistent with this view, showing that investors respond asymmetrically to earnings news when adverse signals are more ambiguous and more costly to trade upon (Williams, 2015). Taken together, this reasoning leads to a clear prediction: institutional trading should respond more strongly and more promptly to positive earnings surprises, whereas trading in response to negative surprises should be weaker around the announcement itself and more concentrated in the post-announcement period.

Prior research provides only a partial view of institutional trading around earnings announcements. Ke and Ramalingegowda (2005), the closest related study, show that changes in transient institutional ownership are associated with earnings surprises, but their quarterly ownership design does not allow a clean separation of pre-announcement discovery, announcement-period interpretation, and post-announcement analysis. Quarterly earnings announcements are well-suited to studying these stages because they create a clear, recurring event timeline, and prices often continue to adjust after disclosure. Using daily institutional trading data and clearly defined event windows, we examine not only whether institutions trade in the direction of earnings news, but also when that trading occurs.

We further examine whether industry trading concentration strengthens institutional engagement around earnings announcements.2 One view is that concentration reflects specialization: investors who focus their trading in particular industries develop deeper expertise, process earnings information more effectively, and trade more aggressively when firms in those industries release important news (Kacperczyk et al., 2005; Choi et al., 2017). An alternative view is that concentration may simply reflect familiarity or overconfidence, in which case it need not improve information processing (Goetzmann & Kumar, 2008; Pool et al., 2015). These competing perspectives lead to a further prediction. If concentration reflects informational advantage, then more concentrated institutions should engage more strongly in pre-announcement discovery, respond more strongly at the announcement, and continue trading more actively afterward as they analyze the implications of the news. If concentration instead reflects non-informational preferences, then this stronger earnings-related trading response should be absent or much weaker.

To examine these issues, we use daily institutional trading data from Ancerno and measure net institutional trading as shares bought minus shares sold, scaled by shares outstanding. We study three periods around quarterly earnings announcements: the two weeks before the announcement, the three-day announcement window, and the two weeks after the announcement.3 We relate trading in each period to earnings surprises measured relative to the final analyst consensus forecast, and we test whether the relation varies with industry trading concentration.

Our evidence shows that institutional net trading is positively related to earnings surprises in all three periods, indicating that institutions participate in information discovery, announcement-period interpretation, and post-announcement analysis. At the same time, trading is not evenly distributed across stages. The strongest relation appears in the post-announcement period, followed by the announcement window, with the weakest relation before the announcement. This pattern is consistent with theories of information complementarities and costly processing, which imply that much of the value of sophisticated trading arises after public disclosure helps investors interpret and act on information. We also find a pronounced asymmetry in the sign of earnings news: institutional trading is more strongly and more promptly related to positive earnings surprises, whereas trading associated with negative surprises is concentrated more heavily after the announcement. Finally, institutions with higher industry trading concentration exhibit stronger earnings-related trading across the announcement cycle, especially for positive surprises, consistent with specialization providing an informational advantage rather than merely reflecting familiarity-based concentration.

This study contributes to the literature in three ways. First, it develops an integrated framework for examining institutional trading across pre-announcement information discovery, announcement-period interpretation, and post-announcement analysis within a single disclosure setting. Second, it shows that the timing of institutional trading is itself informative: the strongest earnings-related trading occurs after disclosure, and the response differs systematically between positive and negative earnings surprises. Third, it shows that industry trading concentration is associated with stronger earnings-related trading, supporting the view that specialization enhances institutions’ ability to process and trade on firm-specific information. Taken together, these findings provide new evidence on how institutional investors process earnings news, why that processing is not uniform across event time, and why it differs between favorable and unfavorable earnings shocks.

2. Literature Review and Research Question

Institutional investors are widely viewed as important participants in the incorporation of information into stock prices. Prior research links institutional ownership to greater pricing efficiency and to weaker return anomalies around earnings news, consistent with the view that institutions possess superior resources for acquiring, interpreting, and trading on value-relevant information (Boehmer & Kelley, 2009; Bartov et al., 2000; Ali et al., 2004; Ke & Ramalingegowda, 2005).4 However, that literature says much less about how institutional trading is allocated across the stages of information processing surrounding a major disclosure event. This omission is important because different theories imply different timing patterns in institutional trading. Some theories imply that institutions’ advantage should appear mainly before disclosure, when private information is still valuable. Others imply that institutional trading should remain strong, or even intensify, after disclosure, because public earnings news helps investors interpret prior signals and lowers the cost of further analysis.

Accordingly, the key conceptual gap is not whether institutional investors trade in the direction of earnings news, but when they do so and which economic mechanism best explains that timing. Quarterly earnings announcements provide a useful setting for addressing this gap because they are pre-scheduled, highly salient, and repeated, allowing institutional trading to be studied before, at, and after the same type of information event.

2.1. Competing Views of Institutional Trading Around Earnings Announcements

The first view may be called the information discovery view. Under this view, institutional investors’ main advantage lies in obtaining or inferring firm-specific information before it becomes public. Prior evidence is broadly consistent with this possibility. Institutional trading has been shown to anticipate scheduled corporate events, and in the earnings setting, Hendershott et al. (2015) show that institutional order flow predicts both the sentiment of upcoming earnings announcements and the stock price reaction at disclosure (Campbell et al., 2009; Baker et al., 2010; Hendershott et al., 2015).5 If this mechanism dominates, then institutional trading related to earnings surprises should be concentrated primarily in the pre-announcement window, when private or partially impounded information is most valuable, and should weaken after disclosure once the information becomes public.

A second view emphasizes interpretation and post-announcement analysis rather than discovery alone. Existing evidence shows that institutional investors respond quickly to public disclosures, suggesting that their role extends beyond trading on information acquired before the event. For example, Huang et al. (2020) find that institutions respond rapidly to the tone of corporate news releases, and Lee and Zhu (2022) show that actively managed funds increase trading intensity around earnings announcements in ways associated with faster price adjustment. Related work also suggests that investor trading can continue after disclosure as market participants analyze the implications of new information and trade strategically with one another (Brunnermeier & Pedersen, 2005; Li et al., 2022). The post-earnings announcement-drift literature reinforces this point by showing that stock prices often continue to move in the direction of earnings news after release, implying that the market does not always incorporate the implications of earnings information immediately (Bartov et al., 2000; Ali et al., 2004; Ke & Ramalingegowda, 2005; Keskek & Rees, 2022).

This second view is supported by two complementary theoretical arguments. First, models of information complementarities imply that public disclosure can increase the value of investors’ prior information by helping them interpret existing signals more precisely (Kim & Verrecchia, 1994; Cheynel & Levine, 2020). Second, theories of rational inattention and costly information processing imply that even sophisticated investors face limits in attention and analytical capacity, so they may not fully process the implications of earnings news immediately (Hirshleifer et al., 2009; Blankespoor et al., 2020; Bertomeu et al., 2023). Taken together, these theories imply that institutional investors may trade in all three stages of the earnings announcement cycle, but that trading need not be evenly distributed across them. In particular, if public disclosure helps institutions complete an information mosaic and lowers the cost of follow-on analysis, then earnings-related trading should remain strong, and may even be strongest, in the post-announcement period.

These competing perspectives yield distinct empirical predictions. If the information discovery view dominates, institutional trading should be most strongly related to earnings surprises before disclosure. If the complementarity and costly processing view dominates, institutional trading should be related to earnings surprises throughout the cycle, but should be concentrated more heavily at and especially after the announcement. Existing studies do not cleanly distinguish between these predictions because they typically focus on one stage at a time. This is the central gap in the timing literature that this study aims to fill.

2.2. Industry Trading Concentration: Specialization Versus Behavioral Concentration

A related but distinct question is whether industry trading concentration strengthens institutional engagement around earnings announcements. Here too, the literature offers competing explanations rather than a single interpretation.

Under the specialization view, industry concentration reflects an informational choice. Investors who focus on trading within a narrower set of industries develop deeper expertise, process industry and firm signals more effectively, and are therefore better positioned to trade around major information events. This interpretation is consistent with evidence that more concentrated investors tend to earn superior returns and that industry-focused investors appear to use earnings-related information more effectively (Kacperczyk et al., 2005; Choi et al., 2017). It is also consistent with evidence that industry focus can be associated with deeper expertise, broader information networks, and more efficient use of analytical resources.6 If this view is correct, then institutions with greater industry trading concentration should exhibit stronger earnings-related trading across the announcement cycle, including more active discovery before disclosure, stronger response at disclosure, and more intensive post-announcement analysis.

An alternative explanation is that concentration reflects behavioral rather than informational forces. Concentrated trading may arise from familiarity-based investing, overconfidence, or other non-informational preferences rather than superior information processing (Goetzmann & Kumar, 2008; Pool et al., 2015). Under this view, concentration need not imply better interpretation of earnings news or stronger engagement when firms release major public information. In that case, industry concentration should show little or no systematic association with earnings-related trading across the three stages of the announcement cycle. This alternative is especially relevant because prior evidence suggests that industry information may not always be incorporated into expectations immediately (Keskek & Tse, 2024), making it important to distinguish true specialization from mere concentration in holdings or trading.

Thus, the concentration literature also yields competing predictions. If concentration proxies for specialization, then concentrated institutions should react more strongly to earnings news across stages. If concentration instead proxies for behavioral preference, then stronger stage-specific trading responses should be absent or substantially weaker.

2.3. Research Question and Prediction Development

The literature, therefore, points to two unresolved issues. First, although prior studies show that institutional investors can discover information before disclosure, interpret news at disclosure, and continue trading after disclosure, they do not show how these activities are jointly distributed around the same earnings event. Second, although prior work links industry concentration to superior performance, it remains unclear whether concentration reflects informational specialization or non-informational trading preferences in the earnings announcement setting.

These gaps motivate the central question of the study: how do institutional investors allocate trading activity across pre-announcement discovery, announcement-period interpretation, and post-announcement analysis, and does industry trading concentration strengthen that engagement? The theoretical discussion above implies clear, testable predictions. If institutional investors’ advantage is primarily pre-disclosure discovery, earnings-related trading should be concentrated before the announcement. If public disclosure complements prior information and lowers processing costs, earnings-related trading should appear in all three stages and be strongest after the announcement. Likewise, if industry concentration reflects specialization, more concentrated institutions should exhibit stronger earnings-related trading across stages; if it reflects familiarity or overconfidence, such a pattern should not emerge.

3. Sample Selection and Data

3.1. Institutional Investors

To examine institutional investors’ trading behavior during the information discovery, immediate response, and information analysis periods, we use daily institutional trading data from Ancerno Ltd. (formerly Abel Noser Corporation), New York, NY, USA.7 Our sample period begins in 2002 and ends in 2010, which is the last full year in which Ancerno provided fund-level identifiers, referred to as client codes, before restricting access to protect client confidentiality.8 The period from 2002 to 2010 captures a substantial share of institutional trading activity. For example, Hu et al. (2018) estimate that Ancerno data represent approximately 12 percent of total trading volume reported in the Center for Research in Security Prices (CRSP).

The Ancerno database does not explicitly reveal the identities of institutional investors. Instead, it uses unique codes (“clientcode”) to track each investor throughout the sample period, allowing us to aggregate trades by institutional investors. Additionally, Ancerno categorizes the institutional investor by type (“clienttype”) as either a mutual or pension fund. The database also provides firm-level identifiers (CUSIP and TICKER symbols), along with specific details like the execution date (“tradedate”), the volume of shares traded (“volume”), and the execution price (“price”). Importantly, Ancerno also reports the direction of trade (side = 1 for buys and side = −1 for sells).

Our main sample consists of firms that announce quarterly earnings during the period from 2002 to 2010 and exhibit sufficient institutional trading activity in the Ancerno database. Consistent with prior research (Bhattacharya et al., 2018; Geiger et al., 2022), we exclude thinly traded firms to ensure reliable measurement of institutional trading behavior. Specifically, firms are required to be traded on more than five days and by more than ten distinct Ancerno funds within the one-month window surrounding the earnings announcement. This window includes the two weeks before the announcement, the three-day announcement window, and the two weeks following the announcement. These restrictions ensure that our analyses focus on firms with meaningful institutional participation across all information processing stages.

3.2. Financial Data

We obtain quarterly financial statement data from Compustat and ensure that firms have the necessary data to calculate the market-to-book ratio (MTB) at the end of each quarter, along with their price and assets (AT). Earnings surprise (ESURP) is computed using data from I/B/E/S. Additionally, we require firms to have stock return data available from CRSP to calculate their stock price momentum (MOMENTUM) and eliminate firms with a price of less than $1. Applying these data requirements and sample screens yields a final sample of 107,917 quarterly earnings announcements during the period from 2002 to 2010.

4. Variable Measurement and Descriptive Statistics

4.1. Net-Buy Trading Metric

We compute institutional net-buying activity during each window (i.e., two-week pre- and post-announcement periods, and 3-day announcement period) for each quarterly earnings announcement. We start by calculating the difference between the total number of shares purchased by institutional investors and the number of shares sold, then dividing this by the total number of shares outstanding. This serves as our measure of net institutional investor buying for the period, which is computed separately for the pre- and post-earnings announcement periods. Our pre-earnings trading period (NET_BUY [−15, −2]) is the two weeks beginning 15 days before the earnings announcement date. Our post-earnings trading period (NET_BUY [+2, +15]) is the two weeks beginning 2 days after the earnings announcement date, and finally, our event period (NET_BUY [−1, +1]) is the three days around the earnings announcement date. In untabulated analyses, we confirm that our findings remain robust when using both shorter periods (e.g., one week) and longer periods (e.g., one month) for information discovery and analysis phases. Following the prior literature (e.g., Geiger et al., 2022), we calculate net institutional trading for the three event periods (i.e., pre- and post-earnings and during earnings) as follows:

$$\sum _{\begin{array}{c}p\in Event\\ Period\end{array}}NET\_BUY=\sum _{\begin{array}{c}p \in Event\\ Period\end{array}}({\displaystyle \frac{{BUY}_{ip}-{SELL}_{ip}}{{SHO}_i}})/T$$

(1)

where BUY_ip (or SELL_ip) represents the total number of shares purchased (or sold) by Ancerno investors in firm i during period p, SHO_i (in millions) refers to the total number of shares outstanding (CSHOQ) for firm i at the end of the quarter, and T is the number of days in the respective event periods. We then divide $(\frac{{BUY}_{ip}-{SELL}_{ip}}{{SHO}_i})$ by T to find the average daily net buying in each of our three periods. Therefore, ∑NET_BUY represents institutional investors’ daily average percentage net buying activity in firm i during our event periods.

4.2. Industry Trading Concentration

Following Geiger et al. (2022), we construct the industry trading concentration metric for each investment fund annually based on the percentage of the fund’s total trading activity within each two-digit industry SIC (e.g., 20 is the two-digit code for Food & Kindred Products). The trading concentration metric, CONCMET_f,i, is determined by taking the ratio of a fund’s dollar trading activity (i.e., buy plus sell) within industry i, scaled by the fund’s total dollar trading activity for that year.9 Specifically,

$${CONCMET}_{f,i}={\displaystyle \frac{\sum _{f,i}{(BUY}_{f,i}+{SELL}_{f,i})}{\sum _f({BUY}_f+{SELL}_f)}}$$

(2)

where BUY_f,i (or SELL_f,i) represents the total dollar value of shares purchased (or sold) by the Ancerno fund f in industry i during the year, and BUY_f (or SELL_f) represents the total dollar value of shares purchased (or sold) by Ancerno fund f over the entire year.10 For example, $\sum _{f,i}{(BUY}_{f,i}+{SELL}_{f,i})$ denotes fund f’s total dollar trading activity in industry i for the year.11

Second, we rank institutional investors annually based on their trading concentration metric, CONCMET_f,i, and categorize them into three groups according to their level of industry concentration. Finally, our scaled trading concentration metric, CONS, takes a value of 0 for the lowest tercile, 0.5 for the middle tercile, and 1 for the highest concentrated trading funds.

4.3. Earnings Surprise

We measure earnings surprise, ESURP, as the actual earnings minus the last mean analyst consensus forecasts before the earnings announcements as provided by the I/B/E/S summary file, scaled by the absolute value of the actual. We replace the absolute value of the actual with 0.05 if it is less than 0.05 to prevent division by zero or to mitigate the impact of a small denominator effect.12

4.4. Descriptive Statistics

We report the descriptive statistics for our sample of quarterly earnings announcements (N = 107,917) in Panel A of

Table 1. Descriptive statistics and correlations.

Panel A: Descriptive Statistics
Variable	Mean	Median		Min	P25		P75		Max
ESURP	3.01	3.85		−232.00	−5.45		17.39		144.19
POS_ESURP	14.56	3.85		0.00	0.00		17.39		144.19
NEG_ESURP	−11.55	0		−232.00	−5.45		0.00		0.00
MOMENTUM	0.11	0.07		−2.79	−0.12		0.30		8.34
MVE ($Billions)	5.21	0.88		0.00	0.32		2.87		505.71
AT ($Millions)	11,039.19	1116.43		2.79	327.87		3954.90		3,293,755.00
MTB	2.68	2.05		0.24	1.32		3.34		9.35
PRICE	49.84	20.73		1.00	10.78		34.60		141,600.00
SHO (Millions)	168.31	47.5125		1.44	24.15		111.08		29,049.60
NET_BUY [−15, −2]	−5.64	2.69		−1445.10	−235.26		229.03		1669.45
NET_BUY [−1, +1)	−45.65	0		−3370.67	−412.91		386.13		3443.79
NET_BUY [+2, +15]	31.64	11.7185659		−1627.05	−233.51		298.03		1737.59
Panel B: Correlations
	[1]	[2]	[3]	[4]	[5]	[6]	[7]	[8]		[9]
ESURP	1	0.09821	0.0096	−0.00489	0.04575	0.00055	0.01379	0.01276		0.05324
[1]		<0.0001	0.0017	0.108	<0.0001	0.8569	<0.0001	<0.0001		<0.0001
MOMENTUM		1	−0.01591	−0.02632	0.24131	−0.00037	0.04902	0.01697		0.06046
[2]			<0.0001	<0.0001	<0.0001	0.9026	<0.0001	<0.0001		<0.0001
MVE			1	0.47764	0.10574	0.12708	−0.01317	−0.00343		−0.01955
[3]				<0.0001	<0.0001	<0.0001	<0.0001	0.2616		<0.0001
SIZE				1	−0.04775	0.04935	−0.00318	0.00029		−0.00687
[4]					<0.0001	<0.0001	0.2964	0.923		0.0241
MTB					1	−0.00521	0.03162	0.02		0.05016
[5]						0.0869	<0.0001	<0.0001		<0.0001
PRICE						1	−0.00091	0.00405		0.00098
[6]							0.7641	0.1835		0.7464
NET_BUY [−15, −2]							1	0.29416		0.20986
[7]								<0.0001		<0.0001
NET_BUY [−1, +1)								1		0.31035
[8]										<0.0001
NET_BUY [+2, +15]										1
[9]

Panel A of Table 1 reports descriptive statistics relating to our population of 107,917 quarterly earnings announcements between 1 January 2002 and 31 December 2010. All variables are defined in Appendix A. See Appendix A for detailed variable definitions; Panel B of Table 1 presents Pearson correlations relating to our population of 107,917 quarterly earnings announcements between 1 January 2002 and 31 December 2010. All variables are defined in Appendix A. See Appendix A for detailed variable definitions.

. We find that average net buying during discovery period, NET_BUY [−15, −2], interpretation period, NET_BUY [−1, +1], and analysis period, NET_BUY [+2, +15], are −5.64, −45.65, and 31.64, respectively, suggesting that institutional investors are net sellers in discovery and interpretation periods and net buyers in the analysis period during our sample period. Specifically, the average percentage daily net selling during the interpretation period is 0.0046 percent.13 In terms of economic magnitude, the net selling activity by Ancerno investors during the three days surrounding the announcement date, [−1, +1], is approximately $77,280,000 ($5.6 billion mean MVE × 0.0046 × 3 days). That is, the funds in our study, on average, net sold $77,280,000 equity securities per firm during the 3 days around earnings announcements.

The mean (median) ESURP is 3.01% (3.85%), consistent with prior research suggesting that firms, on average, tend to report positive surprise at earnings announcements. MVE (in $billions) is the market value of equity and is computed as price times number of shares outstanding at the end of the quarter, with a mean value of $5.2 billion. MTB is market-to-book calculated as MVE divided by the book value of common stock (CEQQ) as of the most recent fiscal quarter. AT (in $millions) is the total assets at the end of the quarter, and PRICE is the end-of-quarter stock price. MOMENTUM is the cumulative value-weighted excess returns computed from the days −253 to −31 relative to earnings announcement day (i.e., CAR [−253, −31]). Finally, all the continuous variables are winsorized at the 5% level to mitigate the effect of outliers.

In Panel B of Table 1, we report the correlation coefficients and find that net buy in each of the discovery, interpretation, and analysis periods is positively correlated with ESURP, providing preliminary support for the institutions’ ability in discovering, interpreting, and analyzing earnings announcement information. A notable distinction that emerges when examining the correlations between net buying activity and ESURP is that the correlation coefficient is four times greater during the analysis period compared to either the discovery or interpretation periods.

5. Empirical Models and Results

5.1. Earnings Surprises and Institutional Trading

We first examine the institutional investors’ ability to discover, interpret, and analyze the earnings information as reflected in their trades by estimating the following model:

NET_BUY [t₁, t₂] = β₀ + β₁ESURP + β₂MTB + β₃SIZE + β₄MOMENTUM + ε

(3)

where NET_BUY [t₁, t₂] is average daily net buying activity between the days t₁ and t₂ relative to the earnings announcement day. The other variables are defined as in the previous section.

In

Table 2. Institutional investor trading and overall earnings surprise.

	Dependent Variable: Net Trading Activity
	NET_BUY [−15, −2]	NET_BUY [−1, +1)	NET_BUY [+2, +15]
ESURP	0.102 ***	0.300 ***	0.687 ***
	(0.008)	(0.000)	(0.000)
MTB	5.501 ***	13.50 ***	12.91 ***
	(0.000)	(0.000)	(0.000)
AT	−0.0287	−0.0275	−0.0510 **
	(0.144)	(0.443)	(0.015)
MOMENTUM	50.11 ***	12.90	62.54 ***
	(0.000)	(0.199)	(0.000)
Constant	−48.38	−23.13	24.38
	(0.396)	(0.764)	(0.424)
N	107,718	107,718	107,718
Adj-R2	0.006	0.002	0.008

Table 2 presents coefficient estimates from model (3). NET_BUY [t₁, t₂] = β₀ + β₁ESURP + β₂MTB + β₃AT + β₄MOMENTUM + ε. All variables are defined in Appendix A. See Appendix A for detailed variable definitions; numbers in parentheses are p-values calculated using standard errors per White (1980), and standard errors are clustered at the firm and industry level. *** and ** denote statistical significance at the 1%, 5%, and 10% levels, respectively.

, we report the results from the estimation of model (3). We find the coefficients on ESURP in the discovery, interpretation, and analysis periods are 0.102, 0.300, and 0.687, respectively, and are statistically significant within each respective period. Our findings suggest that institutional investors as a group trade in the direction of earnings surprise in each period, providing evidence for their ability to discover, interpret, and analyze earnings information. We then compare the coefficients on ESURP in each period to assess the relative importance of each role for institutional investors’ trading activity. Specifically, we find that the average net daily buying associated with ESURP in the information analysis period is about 2.3 times (0.687 versus 0.300) larger than that in the interpretation period, and about 6.7 times (0.713 versus 0.092) larger than that in the discovery period. Thus, our results reveal that approximately 63% of average daily net trading associated with earnings surprise is concentrated during the information analysis period, followed by 27% during the interpretation period, and only around 9% during the discovery period.14 Our findings suggest that institutional investors primarily trade on earnings information based on their efforts towards information analysis.

It is important to mention that our results rely on average daily net buying rather than cumulative total net buying within each period. We adopted this approach to facilitate meaningful comparisons, considering that the interpretation period spans only 3 days, whereas the discovery and analysis periods extend over 14 days each. Thus, we also calculate the cumulative total net buying associated with ESURP within each period by multiplying the ESURP coefficient by the number of days in each period. Using the cumulative total net buying, the coefficient on ESURP becomes 1.428, 0.900, and 9.618 in the discovery, interpretation, and analysis periods, respectively. Interestingly, despite the relatively small average daily net buying in the discovery period, the cumulative total net buying over the entire discovery period surpasses that occurring within the shorter 3-day interpretation period. Additionally, our findings based on cumulative total net buying reveal that over 80% of trading related to ESURP is concentrated in the information analysis period. This highlights the significance of the analysis phase in terms of trading activity related to earnings surprises. These findings also complement prior studies documenting that institutional investors play an important role in incorporating earnings information into stock prices (Bartov et al., 2000; Ke & Ramalingegowda, 2005). Our results extend this literature by showing how institutional trading is distributed across the stages of information discovery, interpretation, and post-announcement analysis around earnings announcements.

5.2. Positive and Negative Earnings Surprises

Next, we investigate whether institutional investors’ engagement across the discovery, response, and analysis periods varies based on the direction of news. We estimate the following model:

NET_BUY [t₁, t₂] = β₀ + β₁POS_ESURP + β₂NEG_ESURP + β₃MTB + β₄SIZE + β₅MOMENTUM + ε

(4)

where POS_ESURP (NEG_ESURP) is the positive (negative) earnings surprise equal to ESURP if the surprise is greater than (less than) zero, and equal to zero otherwise. The other variables are defined as in the previous section.

In

Table 3. Institutional investor trading and positive and negative earnings surprises.

	Dependent Variable: Net Trading Activity
	NET_BUY [−15, −2]	NET_BUY [−1, +1)	NET_BUY [+2, +15]
POS_ESURP	0.341 ***	0.763 ***	1.170 ***
	(0.000)	(0.000)	(0.000)
NEG_ESURP	−0.0702	−0.0323	0.340 ***
	(0.217)	(0.791)	(0.000)
MTB	5.913 ***	14.30 ***	13.75 ***
	(0.000)	(0.000)	(0.000)
AT	−0.0278	−0.0258	−0.0492 **
	(0.151)	(0.474)	(0.016)
MOMENTUM	50.18 ***	13.02	62.66 ***
	(0.000)	(0.194)	(0.000)
Constant	−55.16	−36.24	10.68
	(0.324)	(0.640)	(0.727)
N	107,718	107,718	107,718
Adj-R2	0.007	0.002	0.009

Table 3 presents coefficient estimates from model (4). NET_BUY [t₁, t₂] = β₀ + β₁POS_ESURP + β₂NEG_ESURP + β₃MTB + β₄AT + β₅MOMENTUM + ε. All variables are defined in Appendix A. See Appendix A for detailed variable definitions; numbers in parentheses are p-values calculated using standard errors per White (1980), and standard errors are clustered at the firm and industry level. *** and ** denote statistical significance at the 1%, 5%, and 10% levels, respectively.

, we present the results from the estimation of model (4). Our results reveal a noticeable difference in the magnitudes of coefficients for POS_ESURP across the different periods. Specifically, the coefficients for POS_ESURP are 0.341 during the discovery period, 0.763 during the interpretation period, and 1.170 during the analysis period. Importantly, all of these coefficients are statistically significant, providing support for institutional investors’ ability to discover, interpret, and analyze earnings information for firms reporting positive earnings surprise. Moreover, our results indicate that the impact of POS_ESURP becomes increasingly pronounced as we progress from the discovery period to the analysis period. When comparing the average daily trading activity in each period, our findings reveal that institutional investors’ trading related to POS_ESURP during the information analysis period is approximately 1.5 times greater than that during the interpretation period and approximately 3.4 times greater than that during the discovery period. This underscores the significantly heightened trading activity during the information analysis phase in relation to positive earnings surprises. When we use the cumulative total net buying rather than the average daily during each period, the coefficient on POS_ESURP becomes 4.774, 2.289, and 16.38 in the discovery, interpretation, and analysis periods, respectively. Our findings based on cumulative net buying suggest that, over the course of approximately a month centered around earnings announcements, around 52% of POS_ESURP-related trading takes place in the information analysis period. In contrast, about 33% and 15% of this trading activity occur in the discovery and interpretation periods, respectively. This distribution highlights the information analysis phase as the dominant period for trading related to positive earnings surprises.

Regarding NEG_ESURP, we find that the coefficients are −0.07, −0.032, and 0.340 in the discovery, interpretation, and analysis periods, respectively. Interestingly, the NEG_ESURP coefficient is negative, but statistically insignificant during the discovery period, indicating no evidence that institutional investors systematically discover negative earnings surprises prior to the announcement. Similarly, we find no evidence of information interpretation as the NEG_ESURP coefficient is insignificant in this period. We find, however, significant evidence for information analysis following the announcement of negative earnings surprises. Thus, our results suggest that institutional investors’ contribution to improving price efficiency in relation to negative earnings surprises is mainly attributable to their ability to analyze information. Furthermore, when comparing the coefficients for NEG_ESURP to those for POS_ESURP in each respective period, our findings underscore that institutional investors show a tendency to concentrate on companies expected to report positive news during the discovery phase. Additionally, they tend to interpret and analyze positive news more promptly compared to negative news.

5.3. Earnings Surprises and Industry Trading Concentration

In this section, we examine whether institutional investors with higher levels of industry trading concentration differ in their use of earnings information. Two competing perspectives offer different predictions regarding industry trading concentration. The information advantage view suggests that investors concentrate their trading in particular industries to develop deeper expertise and acquire superior firm- and industry-specific information, which should enhance their ability to process and trade on earnings news. In contrast, behavioral explanations attribute portfolio concentration to factors such as familiarity or overconfidence, which do not necessarily imply superior information processing. Under this view, higher industry concentration would not be expected to improve investors’ ability to discover, interpret, or analyze earnings information. If industry specialization reflects informational advantages, institutional investors with higher industry trading concentration should exhibit stronger trading responses to earnings information across the discovery, interpretation, and analysis periods surrounding earnings announcements. To investigate this argument, we classify investors into groups based on their degree of industry trading concentration and examine their net trading activities around earnings announcements for each group. We then estimate the following regression model:

NET_BUY [t₁, t₂] = β₀ + β₁CONS + β₂ESURP + β₃CONS*ESURP + β₄MTB + β₅SIZE + β₆MOMENTUM + ε

(5)

where CONS is our measure of industry trading concentration, taking the value of 0 for low industry trading concentration funds and 1 for funds with moderate or high industry trading concentration.15 All other variables are as defined previously.

Table 4. Institutional investor trading and industry concentration.

	Dependent Variable: Net Trading Activity
	NET_BUY [−15, −2]	NET_BUY [−1, +1)	NET_BUY [+2, +15]
ESURP	−0.00127	0.00553	0.0468 ***
	(0.740)	(0.425)	(0.000)
CONS	0.659 *	−5.525 ***	5.135 ***
	(0.080)	(0.000)	(0.000)
ESURP × CONS	0.0258 ***	0.0639 ***	0.0865 ***
	(0.001)	(0.000)	(0.000)
MTB	1.279 ***	3.155 ***	2.404 ***
	(0.000)	(0.000)	(0.000)
AT	−0.00649 *	−0.00973	−0.00846 **
	(0.080)	(0.128)	(0.040)
MOMENTUM	4.729 ***	−3.047 **	6.841 ***
	(0.000)	(0.022)	(0.000)
Constant	−11.15	1.649	1.595
	(0.111)	(0.879)	(0.702)
N	646,308	646,308	646,308
Adj-R2	0.001	0.001	0.002

Table 4 presents coefficient estimates from model (5). NET_BUY [t₁, t₂] = β₀ + β₁CONS + β₂ESURP + β₃CONS*ESURP + β₄MTB + β₅SIZE + β₆MOMENTUM + ε. All variables are defined in Appendix A. See Appendix A for detailed variable definitions; numbers in parentheses are p-values calculated using standard errors per White (1980), and standard errors are clustered at the firm and industry level. ***, **, and * denote statistical significance at the 1%, 5%, and 10% levels, respectively.

reports the results from the estimation of model (5), which examines whether industry trading concentration moderates the relation between institutional trading and earnings surprises. We find that the coefficients on ESURP in the discovery, interpretation, and analysis periods are −0.001, 0.005, and 0.047, respectively, and are statistically significant only in the analysis period. This pattern indicates that, excluding industry specialization, institutional investors primarily trade on earnings information after the announcement rather than before or immediately at disclosure. In contrast, the coefficients on the interaction term, CONS*ESURP, are 0.026, 0.064, and 0.086 in the discovery, interpretation, and analysis periods, respectively, and are statistically significant in each period. Our results suggest that institutional investors’ ability in information discovery prior to earnings announcements, interpretation at the announcement, and post-announcement information analysis significantly increases with their industry concentration. Specifically, we find that industry-concentrated institutions demonstrate the ability to discover and interpret earnings information, whereas there is no evidence of discovery or interpretation among low industry concentration institutions. Our results also reveal that institutional investors with higher levels of industry trading concentration analyze earnings news more promptly.

Next, we examine whether the effect of industry trading concentration is different for positive versus negative earnings surprises by estimating the following model:

NET_BUY [t₁, t₂] = β₀ + β₁CONS + β₂POS_ESURP + β₃NEG_ESURP + β₄CONS*POS_ESURP + β₅CONS*NEG_ESURP + β₆MTB + β₇SIZE + β₈MOMENTUM + ε

(6)

We present the results in

Table 5. Institutional investor trading, industry concentration, and earnings surprise sign.

	Dependent Variable: Net Trading Activity
	NET_BUY [−15, −2]	NET_BUY [−1, +1)	NET_BUY [+2, +15]
POS_ESURP	0.0258 ***	0.0265 *	0.0864 ***
	(0.001)	(0.050)	(0.000)
NEG_ESURP	−0.0210 ***	−0.0105	0.0176 ***
	(0.000)	(0.341)	(0.007)
POS_ESURP × CONS	0.0492 ***	0.148 ***	0.150 ***
	(0.000)	(0.000)	(0.000)
NEG_ESURP × CONS	0.00937	0.00506	0.0415 ***
	(0.392)	(0.825)	(0.001)
CONS	0.129	−7.422 ***	3.685 ***
	(0.782)	(0.000)	(0.000)
MTB	1.353 ***	3.287 ***	2.545 ***
	(0.000)	(0.000)	(0.000)
AT	−0.00633 *	−0.00944	−0.00816 **
	(0.084)	(0.139)	(0.041)
MOMENTUM	4.740 ***	−3.027 **	6.863 ***
	(0.000)	(0.023)	(0.000)
Constant	−12.01 *	0.736	0.228
	(0.079)	(0.946)	(0.955)
N	646,308	646,308	646,308
Adj-R2	0.001	0.001	0.003

Table 5 presents coefficient estimates from model (6). NET_BUY [t₁, t₂] = β₀ + β₁CONS + β₂POS_ESURP + β₃NEG_ESURP + β₄CONS*POS_ESURP + β₅CONS*NEG_ESURP + β₆MTB + β₇SIZE + β₈MOMENTUM + ε. All variables are defined in Appendix A. See Appendix A for detailed variable definitions. Numbers in parentheses are p-values calculated using standard errors per White (1980), and standard errors are clustered at the firm and industry level. ***, **, and * denote statistical significance at the 1%, 5%, and 10% levels, respectively.

. We find that the coefficients on POS_ESURP in the discovery, interpretation, and analysis periods are 0.0258, 0.0265, and 0.0864, respectively, and are statistically significant in the discovery and analysis periods, but marginally significant in the interpretation period. In contrast, the coefficients on the interaction term, CONS*POS_ESURP, are 0.049, 0.148, and 0.150 in the discovery, interpretation, and analysis periods, respectively, and are statistically significant in each period. Thus, we find evidence that institutional investors’ ability to discover, interpret, and analyze positive earnings news is positively related to their industry concentration.

Interestingly, our analysis reveals that during the discovery period, low industry concentration institutions do not exhibit trading behavior consistent with the anticipation of negative earnings news, as indicated by the highly significant coefficient of −0.021 for NEG_ESURP. Surprisingly, these institutions tend to be net buyers rather than net sellers during this phase, potentially hindering price efficiency leading up to the announcement of negative earnings surprises. Crucially, the combined coefficients for NEG_ESURP and CONS*NEG_ESURP, which represent high industry concentration institutions in the discovery period, amount to −0.012 and are statistically insignificant (p-value = 0.205). This suggests that industry concentration does not materially enhance these institutions’ ability to discover negative earnings news prior to the announcement.

Turning to the interpretation period, our results show that the coefficient for NEG_ESURP is insignificant for low-concentration institutions, implying that they do not notably contribute to price efficiency through the interpretation of negative earnings news. Likewise, the combined coefficient for NEG_ESURP and CONS*NEG_ESURP in this period totals 0.005 and is statistically insignificant, indicating that industry concentration does not significantly impact the interpretation ability of these institutions. However, our findings provide evidence that industry concentration positively affects institutions’ capacity to analyze negative earnings information in the right direction, as indicated by the significant positive coefficient of 0.0415 on the interaction term, CONS*NEG_ESURP, in the analysis period. This finding suggests that industry specialization contributes to the processing of negative earnings information primarily through enhanced post-announcement information analysis rather than through pre-announcement discovery or immediate interpretation.

5.4. Additional Results

5.4.1. Mutual Versus Pension Funds

Institutional investors differ in their investment horizons, trading incentives, and portfolio turnover. In supplementary analyses, we extended our investigation by categorizing institutional investors into two distinct groups: mutual funds and pension funds, as identified by the client type (clienttype) identifier provided by Ancerno Ltd. Mutual funds typically face stronger short-term performance pressures and exhibit higher trading frequency, whereas pension funds tend to have longer investment horizons and lower trading turnover. These differences may influence how actively investors respond to earnings-related information. Motivated by prior research indicating that mutual funds and pension funds exhibit contrasting characteristics, including investment horizon and portfolio turnover, we aim to explore potential variations in how these two groups discover, interpret, and analyze earnings information to inform their trading decisions. To explore this possibility, we conducted all analyses separately for mutual funds and pension funds to gain deeper insights into their respective behaviors in relation to earnings information.

In

Table 6. Mutual and pension fund trading.

	Dependent Variable: Net Trading Activity (NET_BUY [t1, t2])
	Mutual Funds			Pension Funds
	[−15, −2]	[−1, +1)	[+2, +15]	[−15, −2]	[−1, +1)	[+2, +15]
POS_ESURP	0.128 **	0.559 ***	0.510 ***	0.0515 ***	0.119 ***	0.155 ***
	(0.020)	(0.000)	(0.000)	(0.000)	(0.001)	(0.000)
NEG_ESURP	−0.0186	0.0584	0.106 ***	−0.000511	0.0268 *	0.0287 ***
	(0.526)	(0.401)	(0.004)	(0.936)	(0.089)	(0.000)
MTB	20.35 ***	50.68 ***	23.89 ***	5.450 ***	14.46 ***	6.441 ***
	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)
AT	−0.0282	−0.0573	−0.0425	−0.0103	−0.0206	−0.0157 *
	(0.275)	(0.324)	(0.137)	(0.147)	(0.204)	(0.056)
MOMENTUM	32.50 ***	−42.74 ***	53.94 ***	10.91 ***	−10.69 **	14.20 ***
	(0.000)	(0.008)	(0.000)	(0.000)	(0.040)	(0.000)
Constant	−91.80	−104.5	23.65	−15.40	−109.3 ***	−54.71 ***
	(0.134)	(0.336)	(0.487)	(0.460)	(0.002)	(0.003)
N	107,993	107,993	107,993	107,993	107,993	107,993
Adj-R2	0.014	0.013	0.015	0.011	0.013	0.013

Table 6 presents coefficient estimates for mutual and pension funds from model (4). NET_BUY [t₁, t₂] = β₀ + β₁POS_ESURP + β₂NEG_ESURP + β₃MTB + β₄AT + β₅MOMENTUM + ε. All variables are defined in Appendix A. See Appendix A for detailed variable definitions. Numbers in parentheses are p-values calculated using standard errors per White (1980), and standard errors are clustered at the firm and industry level. ***, **, and * denote statistical significance at the 1%, 5%, and 10% levels, respectively.

, we present the results from estimating model (4) separately for each institution type. We observe significantly positive coefficients for POS_ESURP across all periods, both for mutual funds (see columns 1–3) and pension funds (see columns 4–6), indicating that both mutual funds and pension funds are actively engaged in information discovery, interpretation, and analysis of positive earnings information. Furthermore, similar to the trends observed in our full sample results, we note that the magnitude of the coefficient on POS_ESURP mostly tends to increase as we transition from the discovery period to the analysis period for both mutual funds and pension funds. This result highlights the information analysis phase as the dominant period for trading related to positive earnings surprises by both types of funds. However, a notable distinction emerges in the trading patterns of these two fund types. Mutual funds exhibit trading activity related to positive earnings information that is approximately three to four times greater than that of pension funds across all periods. For example, in the interpretation period, the POS_ESURP coefficient is 0.559 in column 2 compared to 0.119 in column 5 of Table 6. It is important to highlight that this discrepancy cannot be attributed to differences in fund sizes, as total assets under management by pension funds exceed those by mutual funds (e.g., Del Guercio & Tkac, 2002), suggesting instead that differences in trading behavior and investment objectives drive the observed variation.

Turning to negative earnings surprises, we find no evidence that either mutual funds or pension funds systematically discover negative earnings news prior to the announcement, as reflected by the statistically insignificant coefficients on NEG_ESURP during the discovery period. However, following the announcement of negative earnings surprises, both types of funds exhibit significant evidence of engaging in information analysis, as evidenced by the significant positive coefficients on NEG_ESURP during the analysis period. Intriguingly, our results suggest that pension funds do not significantly interpret earnings news negatively during the interpretation period, as indicated by the insignificant coefficient on NEG_ESURP. In contrast, there is some indication that mutual funds respond to negative news during the interpretation period, as demonstrated by the marginally significant positive coefficient on NEG_ESURP. Consequently, our findings imply that both types of funds primarily contribute to enhancing price efficiency concerning negative earnings surprises through their ability to analyze information.

We further investigate whether the effect of industry concentration on information discovery, interpretation, and analysis differs based on the type of institution. To explore this, we separately estimate model (5) for each institution type and present the findings in

Table 7. Mutual and pension fund trading and industry concentration.

	Dependent Variable: Net Trading Activity (NET_BUY [t1, t2])
	Mutual Funds			Pension Funds
	[−15, −2]	[−1, +1)	[+2, +15]	[−15, −2]	[−1, +1)	[+2, +15]
POS_SURP	0.0391 ***	0.0460 *	0.150 ***	0.0126 ***	0.00702	0.0228 ***
	(0.006)	(0.076)	(0.000)	(0.000)	(0.111)	(0.000)
NEG_SURP	−0.0346 ***	−0.0282	0.0316 **	−0.00741 ***	0.00727 **	0.00361
	(0.002)	(0.183)	(0.012)	(0.001)	(0.035)	(0.122)
POS_SURP × CONS	0.0816 ***	0.232 ***	0.217 ***	0.0169 ***	0.0634 ***	0.0835 ***
	(0.002)	(0.000)	(0.000)	(0.005)	(0.000)	(0.000)
NEG_SURP × CONS	0.0171	0.00641	0.0601 **	0.00160	0.00372	0.0229 ***
	(0.416)	(0.884)	(0.017)	(0.701)	(0.629)	(0.000)
CONS	0.225	−10.80 ***	7.384 ***	0.0318	−4.044 ***	−0.0147
	(0.793)	(0.000)	(0.000)	(0.888)	(0.000)	(0.955)
MTB	2.163 ***	5.395 ***	4.182 ***	0.542 ***	1.179 ***	0.908 ***
	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)
AT	−0.00972	−0.0141	−0.0121 *	−0.00294 **	−0.00476 **	−0.00423 ***
	(0.109)	(0.209)	(0.085)	(0.044)	(0.036)	(0.006)
MOMENTUM	7.003 ***	−5.374 **	10.76 ***	2.478 ***	−0.679	2.961 ***
	(0.000)	(0.021)	(0.000)	(0.000)	(0.304)	(0.000)
Constant	−22.90 **	11.97	7.080	−1.113	−10.49 **	−6.623 **
	(0.042)	(0.556)	(0.302)	(0.734)	(0.048)	(0.043)
N	323,154	323,154	323,154	323,154	323,154	323,154
Adj-R2	0.002	0.001	0.004	0.002	0.001	0.004

Table 7 presents coefficient estimates from model (6). NET_BUY [t₁, t₂] = β₀ + β₁CONS + β₂POS_ESURP + β₃NEG_ESURP + β₄CONS*POS_ESURP + β₅CONS*NEG_ESURP + β₆MTB + β₇SIZE + β₈MOMENTUM + ε. All variables are defined in Appendix A. See Appendix A for detailed variable definitions. Numbers in parentheses are p-values calculated using standard errors per White (1980), and standard errors are clustered at the firm and industry level. ***, **, and * denote statistical significance at the 1%, 5%, and 10% levels, respectively.

. For mutual funds, our analysis reveals that industry concentration enhances their engagement to discover, interpret, and analyze positive earnings news information. However, it primarily improves their ability to analyze negative earnings news and has no discernible impact on the discovery and interpretation of negative earnings news.

Interestingly, our results for pension funds align closely with those for mutual funds, displaying a similar pattern for the effects of industry concentration on information discovery, interpretation, and analysis roles concerning both positive and negative earnings surprises. Hence, the results in Table 7 indicate that although pension funds trade less intensively overall, the specialization effect documented earlier remains qualitatively similar across both institutional investor types. This finding suggests that the documented relationship between industry specialization and institutional information processing is not driven solely by a particular class of institutional investors.

Despite these similarities, a notable difference emerges in the intensity of trading responses across institution types. High-industry-concentration mutual funds consistently exhibit significantly stronger trading activity linked to positive earnings information, consistently surpassing that observed among high-concentration pension funds across all periods, amounting to approximately three to four times greater activity. We also observe comparable differences between the two groups in the analysis period following negative earnings announcements, while differences during the discovery and interpretation periods are less pronounced.

Taken together, these findings highlight meaningful heterogeneity in how mutual funds and pension funds deploy industry specialization in their trading strategies, with mutual funds consistently adopting a more aggressive trading stance around earnings announcements. This finding is consistent with mutual funds being more active traders than pension funds (e.g., Del Guercio & Tkac, 2002).

5.4.2. Using Different Windows for Information Discovery and Analysis

In untabulated robustness analyses, we extended our investigation of information discovery and analysis by institutions by considering both shorter and longer time windows. Specifically, we examined shorter time windows ranging from 2 to 5 days and longer ones spanning from 2 to 30 days. Intriguingly, our main findings and conclusions remained consistent and unchanged when we employed these alternative windows. This robustness in results across different time frames suggests that the patterns we have identified regarding information discovery and analysis are robust and not dependent on the specific time periods chosen for analysis. This lends further credibility to the reliability of our research findings in this context.

6. Conclusions

This study examines how institutional investors trade around quarterly earnings announcements and whether industry trading concentration is associated with stronger institutional engagement across the earnings announcement cycle. Using daily institutional trading data from Ancerno, combined with CRSP, Compustat, and I/B/E/S data for the 2002–2010 period, we distinguish among three stages of earnings-related information processing: information discovery before the announcement, interpretation at disclosure, and post-announcement information analysis.

The results show that institutional investors trade in ways consistent with engagement in all three stages. Institutional net trading is positively associated with earnings surprises before the announcement, during the announcement window, and after the announcement. At the same time, institutional trading is not evenly distributed across these stages. The largest share of average daily institutional trading occurs after the earnings announcement, followed by the immediate announcement window, with the smallest share occurring before the announcement. These findings suggest that a substantial portion of institutional trading around earnings announcements occurs after information becomes public, consistent with institutions continuing to analyze and act on disclosed earnings news.

The results also reveal important asymmetries based on the direction of earnings news. Institutional trading is strongly related to positive earnings surprises across the discovery, interpretation, and post-announcement analysis periods. In contrast, for negative earnings surprises, institutional trading is not significantly related to earnings news during the pre-announcement or immediate response periods, but becomes significantly negative in the post-announcement period. These patterns indicate that institutional trading around earnings announcements differs systematically between favorable and unfavorable earnings news and suggest that institutions respond more strongly and more promptly to positive surprises than to negative surprises.

The study further shows that industry trading concentration is positively associated with institutional engagement across the earnings announcement cycle. Institutional investors with more concentrated industry trading exhibit stronger trading in the direction of earnings news before the announcement, at disclosure, and after the announcement. These effects are strongest for positive earnings surprises and substantially weaker for negative surprises. Supplementary analyses also indicate that mutual funds trade more actively around earnings announcements than pension funds, although both investor types exhibit similar patterns in how trading is allocated across the three stages.

These findings contribute to the literature in several ways. First, the study provides a more integrated view of institutional information processing around a major recurring disclosure event by jointly examining trading consistent with information discovery, interpretation, and post-announcement analysis within the same setting. Prior research typically examines these activities separately. Second, the evidence clarifies when institutional investors are most actively engaged in earnings-related trading. Rather than being concentrated primarily before or at the announcement, institutional trading is strongest after the announcement, consistent with theories emphasizing the value of public information in sharpening prior beliefs and the role of costly information processing in shaping the timing of trading activity. Third, the results contribute to the literature on institutional specialization by showing that industry trading concentration is associated with stronger institutional engagement around earnings announcements, consistent with the view that industry focus is linked to informational advantage rather than reflecting only behavioral concentration.

More broadly, the results have implications for understanding price formation around earnings announcements. They suggest that institutional investors do not simply react mechanically to disclosed earnings news, but participate throughout the broader information-processing cycle surrounding earnings announcements. The concentration of trading in the post-announcement period is particularly important because it suggests that a meaningful portion of earnings-related trading occurs after public disclosure, when investors continue to interpret the implications of earnings news for firm value. In this sense, the findings provide new evidence on how sophisticated investors contribute to the incorporation of earnings information into prices.

This study is subject to several limitations. First, the Ancerno data capture a large and widely used sample of institutional trading, but they do not represent the full universe of institutional investors. Second, the sample period ends in 2010, reflecting the availability of consistent fund identifiers in the Ancerno database. Third, although the analysis is designed around distinct periods before, during, and after earnings announcements, trading in these windows should be interpreted as consistent with information discovery, interpretation, and analysis rather than as direct proof of the underlying mechanisms. Finally, the study documents systematic patterns in institutional trading around earnings announcements, but it does not identify the specific information sets or decision processes used by individual investors.

These limitations point to several opportunities for future research. Future studies could examine whether similar trading patterns arise around other important corporate disclosures, such as management forecasts, 8-K filings, or regulatory events. Additional work could also examine whether the documented asymmetry between positive and negative earnings surprises extends to other settings or varies across market conditions. Finally, future research could explore alternative measures of institutional specialization and investigate whether the relation between specialization and earnings-related trading differs across other classes of institutional investors.