Market Reactions to Quarterly Earnings Surprises: The Impact of

Market Reactions to Quarterly Earnings Surprises: The Impact of Financial Statements’ Data Disclosed with Earnings

Eli Amir London Business School

Sussex Place, Regent’s Park London NW1 4SA United Kingdom Tel: +44 (0)20 7000 7000 (ext. 3182)

[email protected]

And

Joshua Livnat Department of Accounting

Leonard N. Stern School of Business New York University

40 W. 4th St. NY, NY 10012

(212) 998 – 0022 [email protected]

Current Draft: August 2006 The authors gratefully acknowledge Charter Oak Investment Systems Inc. for providing the preliminary and original Compustat quarterly data used in this study. The authors also gratefully acknowledge Standard and Poors’ Compustat for providing the SEC filing dates data. We thank seminar participants at Tel Aviv University for comments and suggestions.


Abstract

This study focuses on market reactions to earnings and other financial statement information that is concurrently disclosed in preliminary earnings press releases. We use a large sample of heterogeneous companies and investigate whether market reactions to earnings surprises vary systematically with the amount of financial statement detail that is included in earnings press releases, after controlling for the timing of announcements and factors related to firms’ information environments. We observe a dramatic increase in the amount of detail disclosed concurrently with quarterly earnings in the years following Regulation Fair Disclosure. We find that level of financial statement detail is positively associated with the magnitude of the earnings surprise (and more so for negative surprises), firm size, volume of trade, likelihood of raising new capital and analysts' coverage. Conversely, it is negatively associated with the timing of press releases. Turning to the market-based analysis, we find that absolute abnormal returns are higher the more detail is disclosed in the preliminary earnings release, and that the level of detail is more strongly associated with unexpected returns when earnings surprises are extreme. We also find that market reactions to earnings surprises decline over time, but increase with the level of financial statement detail in preliminary earnings releases. However, the importance of detail to the interpretation of extreme earnings surprises has declined over time.


1. Introduction

Since Ball and Brown (1968), Beaver (1968) and as reviewed by Lev (1989) and Kothari

(2001), numerous studies have consistently documented a positive and statistically significant

association between excess stock returns and unexpected earnings during the short windows

around preliminary earnings press releases.1 These findings imply that market participants

obtain, process, and react to new earnings information, setting security prices according to the

earnings news. Still, there has been some concern about the role of earnings news in explaining

the total variability of stock prices around preliminary earnings announcements (Lev, 1989), and

the potential deterioration over the years in the association between earnings news and short-

window returns around them (Collins et al., 1997; Francis and Schipper, 1999; Lev and Zarowin,

1999).

Preliminary earnings press releases often convey additional information beyond earnings,

including other financial statements’ data that may be value-relevant for market participants

(Chen et al., 2002). Because preliminary earnings press releases are not mandatory, companies

select the level of detail they wish to disclose concurrently with preliminary earnings. There

exists considerable variation in level of detail across preliminary earnings press releases. Some

of these voluntary additional data are in fact used by market participants to set prices,

incrementally to the effects of the earnings news. For example, Ertimur et al. (2003) provide

evidence that investors react and properly utilize revenue data that are contained in preliminary

1 Most studies used excess stock return and signed unexpected earnings as dependent and independent variables, respectively. Beaver (1968), on the other hand, was the first study to include absolute excess returns and volume of trade as dependent variables and absolute unexpected earnings as independent variables.

2

earnings announcements. Many companies now include detailed income statements, some

include condensed balance sheets (or balance sheet line items), and far fewer include any cash

flow line items in their earnings press releases2.

In a recent study, Francis et al. (2002) randomly selected 30 companies and examined

2,190 earnings press releases over the period 1980-1999. They provide evidence that the amount

of information in press releases (including additional financial statement data) has increased

significantly over time and show that the increased market reaction to earnings press releases

over time is associated with the increase in the amount of information in these disclosures. Lo

and Lys (2001) argue that the disclosure of concurrent information in earnings press releases

may explain the disparity between an increased information content of earnings announcements

and the decrease in the value-relevance of earnings per se; while the increase in concurrent

information causes stronger market reactions around the preliminary earnings press releases over

time. The reaction to the earnings signal alone in fact declines over time.

Until recently it was prohibitively costly to examine the effect of concurrent disclosures on

the market reactions to preliminary earnings press releases using large samples, because data

from press releases had to be manually collected, read and coded. This is why the results in

Francis et al. (2002) are obtained using a small sample of 30 companies randomly selected from

their sample of large and stable 426 companies with complete quarterly financial data over 20

years. In contrast, our study employs the preliminary earnings database available from

Compustat and Charter Oak through WRDS, together with the SEC filing dates supplied to us by

Compustat. These databases allow an investigation of earnings and concurrent information

2 At the time of this draft, IBM preliminary earnings releases contain income statement and balance sheet information, but not cash flow data. See, e.g., IBM’s 2007 Q1’s release on April 17, 2007 at: http://www.ibm.com/investor/1q07/1q07earnings.phtml

3

disclosed during preliminary press releases using a very large sample of heterogeneous

companies with substantially different information environments.

The purpose of our study is to examine (i) whether market reactions to earnings are

affected by other financial statements information concurrently disclosed in preliminary earnings

announcements; (ii) whether the trend of stronger market reactions to earnings information over

time is associated with the increasing level of other financial statement detail in preliminary

earnings announcements; and (iii) whether the information environment plays a role in the level

of detail provided in preliminary earnings releases, and if so, whether market reactions to

preliminary earnings announcements are affected by the level of other financial statement detail

in preliminary earnings announcements after controlling for the differential effects of the

information environment.

As we explore the association between earnings news, level of detail in preliminary

earnings announcements and market reactions, we also control for the timing of the earnings

press release relative to the quarter-end, a factor omitted by Francis et al. (2002), likely because

of their small sample. As Chambers and Penman (1984) and Kross and Schroeder (1984) show,

firms with good (unfavorable) earnings news tend to report earlier (later) than their expected

reporting dates, with market reactions that are in the same direction as the prompt (delayed)

reporting. As we show below, large companies tend to issue preliminary earnings

announcements earlier than small companies, and their announcements usually contain more

financial statement detail than those of small companies. Thus, with the large sample available to

us, we can adequately control for the timing of the earnings release as well as for cross sectional

differences in firms’ information environment and their effects on the association between detail

level and market reactions.

4

The contribution of this study to the literature is along several dimensions. First, it sheds

additional light on the voluntary activity of preliminary earnings announcements, and more

specifically on the level of financial statement detail firms choose to include along earnings in

these announcements. Second, it explains the cross-sectional variation in the amount of financial

statement detail concurrently disclosed with quarterly earnings. Third, it examines whether

market reactions to the preliminary earnings releases are sensitive to the level of detail for a large

sample of firms and for different sub-samples depending on the information environment of

firms. Fourth, it controls for the interaction between the timing of the preliminary earnings

announcements and the level of financial statement detail in measuring market reactions to the

announcements. Finally, it provides additional evidence about the role that financial statement

detail in preliminary earnings announcements plays in explaining the trend of increasing market

reactions to earnings announcements over time.

To facilitate the empirical testing, we design and implement three scoring mechanisms that

measure the amount of financial detail in earnings press releases. The first is based on a simple

tally of the items included in our analysis from the income statement, balance sheet and the

statement of cash flows. The second subjectively assigns weights to each financial statement

item based on the relative importance of the item as depicted by prior studies. The third is based

on “market” weights, which are obtained by regressing the market reaction to earnings on

earnings surprises and an indicator variable that represents whether the financial statement data

item was included in the preliminary earnings announcement. We show that the amount of

financial detail in press releases has increased dramatically following Regulation Fair Disclosure

in late 2000.

5

We find that the level of detail score is positively associated with positive earnings

surprises, but significantly more so with negative earnings surprises. This result suggests that the

likelihood of more disclosure is higher when earnings news is bad, consistent with the argument

that companies increase disclosure to reduce litigation costs.3 We also find that the level of

financial statement detail is positively associated with firm size, earnings volatility, volume of

trade, the likelihood of raising new capital and analysts' coverage, and negatively associated with

the timing of the press release.

Turning to the market-based analysis, we show that, consistent with Francis et al. (2002),

market reactions are stronger the more information is disclosed in preliminary earnings releases.

However, we also show that these additional financial statement details are used more strongly

by market participants when earnings are extreme, i.e., that the absolute unexpected earnings are

more strongly associated with absolute unexpected returns the larger the amount of financial

statement detail concurrently disclosed with earnings. This result generalizes that of Francis et al.

(2002), who apparently due to their small sample size did not investigate this interaction effects.

We also find that while the coefficient on absolute unexpected earnings in explaining absolute

unexpected returns declines over time, the coefficient on the association between detail score and

absolute unexpected returns actually increases over time, even after controlling for variables that

affected by the firm’s information environment. We also find that the coefficient that measures

the interaction between absolute unexpected earnings and level of detail and its association with

absolute abnormal returns has decreased over time. This result explains the disparity between the

decline in the value relevance of earnings and the increasing overall reaction to press releases, as

suggested by Lo and Lys (2001).

3 This argument is consistent with the findings in Skinner (1994) and Kasznik and Lev (1995).

6

Our study continues as follows: Section 2 discusses the research design. Sample selection

and data requirements are discussed in Section 3. The results of our analyses are provided in

section 4, and section 5 provides concluding remarks.

2. Research Design

2.1 Institutional Setting

The majority of companies voluntarily announce their operating results for the preceding

quarter before filing the required 10-Q or 10-K Form with the SEC. The median company

announces preliminary earnings about 27 days after quarter-end, as shown by Easton and

Zmijewski (1993). Most companies actually file their 10-Q or 10-K Form with the SEC on the

last day or two of the allowed period, as is shown by Griffin (2003). Since the preliminary

earnings release is a voluntary activity, each company can choose the level of financial statement

detail it wishes to disclose in the preliminary earnings release. About 95% of companies include

sales along with earnings (Jegadeesh and Livnat, 2006). About 40% of the companies in the

sample of Chen et al. (2002) provide balance sheet data that can be used to estimate current

accruals. As we show below (Table 2), fewer than 10% of our sample observations have line

items from the statement of cash flows.

When a company issues a preliminary earnings report, Compustat extracts as many line

items as it can from the preliminary earnings release and incorporates it in its quarterly database

with an update code of "2". After the company files its 10-Q (or 10-K) Form with the SEC,

Compustat extracts the line items again from the official filing and overwrites the preliminary

figures, denoting them with an update code of "3". The process of extracting data from the

preliminary earnings releases is quite fast, and is usually completed within 48 hours from the

7

press release. The process of extracting the information from the 10-Q and 10-K Forms is much

longer, and in some cases of smaller companies it can take more than a month or two after the

SEC filing for the final data to be included in the Compustat database.

The initial research design problem that we need to tackle is the identification of those

observations where a preliminary earnings release is issued prior to the SEC filing, as some

companies may issue an earnings press release together or after the SEC filing.4 Compustat

includes the preliminary earnings release date in its quarterly database, and I/B/E/S includes it

for firms covered by analysts. However, the preliminary earnings release date in Compustat may

actually be the SEC filing date subsequent to 1999. Until some time in year 2000, Compustat

used to insert a missing value code if a firm had not issued a preliminary earnings release.

Afterwards, Compustat revised its methodology and began including the SEC filing date for the

earnings report date. Furthermore, to ascertain what market participants obtained in the

preliminary earnings releases rather than the subsequent SEC filings, we require that the

preliminary earnings release date be at least three days prior to the SEC filing date. This requires

us to be able to identify the SEC filing dates for each firm; information that is typically

unavailable in Compustat. However, Compustat provided us with the SEC filing dates for 10-Q

and 10-K Forms for many firms in its database beginning in 1991. Thus, we are able to

determine which companies issued preliminary earnings announcements at least three days prior

to their SEC filings.

To identify the line items that were included in the preliminary earnings release, we use

another database provided to us by Charter Oak, which is also available for purchase through

Compustat in WRDS. Charter Oak has collected the weekly CD-ROMs that were sent to

4 See Stice (2001). After 2002, less than 0.1% of our sample companies issue press release subsequent to the SEC filing.

8

Compustat clients over the years, where some data are designated with an update code of "2" and

others with an update code of "3". Using the update codes of "2", Charter Oak is capable of

constructing a database that includes the preliminary line items reported by companies in their

preliminary earnings press releases. Thus, we can identify which financial statement line items

were included by each company in its preliminary earnings press release. This enables us to use a

very large sample in our study without the need to read and code actual earnings press releases,

as in Francis et al. (2002).

We should, however, point out two potential shortcomings of our research design. Because

the Charter Oak database includes only financial statement data, we limit our study to the effects

of financial statement line items rather than the other potentially useful non-financial and

forward-looking information contained in the preliminary earnings releases, as do Francis et al

(2002). On the other hand, coding such information in preliminary earnings releases may suffer

from subjective assignments and classifications of information.

The second shortcoming of our design relates to conference calls and the potential

disclosure of information to selected parties prior to the issuance of Regulation Fair disclosure in

October 2000 (REG FD). It is commonly assumed that prior to REG FD some companies

disclosed financial statement line items to a select group of analysts and investors in conference

calls and closed meetings. After REG FD, these items are broadcasted to all investors in the open

conference calls. Kohlbeck and Magilke (2002) argue that additional information provided by

conference calls combined with an increased conference call activity is the source of the increase

in information content of earnings. They find that after controlling for conference calls, the

information content of earnings has not increased during 1995-2000, and for small companies

9

has even decreased. This is consistent with additional information beyond earnings being

released in conference calls.

Kimbrough (2005) finds that the initiation of conference calls is associated with a

significant reduction in the serial correlation in analyst forecast errors and post-earnings

announcement drift. He also finds that the reduction in post-earnings announcement drift

surrounding conference call initiation is concentrated in companies with the most significant

drift, particularly small and least heavily traded companies, indicating the potential importance

of controlling for the firm’s information environment.

Bushee et al. (2004) examine the effect of REG FD on managers' decisions regarding the

timing and content of conference calls. Their results suggest that the new rule had a negative

impact on managers' decisions to continue hosting conference calls. Contrary to views expressed

by critics of REG FD, they do not find evidence that the new rule decreased the amount of

information disclosed during the call period. They also find that REG FD increased price

volatility for firms that previously restricted access to their conference calls relative to

companies that prior to the new rule held open conference calls, indicating that additional new

information beyond earnings is likely disclosed in the conference calls.

To the extent that limited-access conference calls and closed meetings are actually the

instrument that is used by investors to learn new value-relevant information that is unavailable in

the preliminary earnings announcements, and to the extent that conference calls are conducted

within the three-day window centered on the preliminary earnings release date, our study will be

biased against finding any effects of the additional disclosures in preliminary earnings releases

beyond earnings on market reactions to earnings surprises prior to 2001. This bias is unlikely to

exist subsequent to REG FD when all investors are supposed to obtain the same material

10

information simultaneously, so we include separate results for the periods prior and after REG

FD. Our study is the first to examine the effects of concurrent disclosures in preliminary press

releases before and after REG FD.

2.2 Detail Scoring Mechanism and its Determinants

To measure the amount of information in the preliminary press release we design and

implement three scoring mechanisms based on the number and relative importance of financial

elements included in each quarterly press release. Initially, we identify 10 income statement

items, 13 balance sheet items and 4 cash flow items to be included in the scoring mechanism.

Second, we assign weights based on three distinct methods – "Equal", "Subjective" and "Market"

Scoring. The Equal Scoring method is the simplest: the score is based on the number of items

included in the quarterly press release. The “Subjective” scoring method is based on assigning

subjective weights of "1", "2" or "3", where "3" denotes higher importance than "2", which, in

turn, is assumed to be more important than "1". We assign the weights based on prior studies.

For example, given the importance of revenues in prior studies (Ertimur et al., 2003) we assign a

weight of "3" to "Sales." We assign a weight of "2" to "Inventory", which can be used to estimate

current accruals, and a weight of "1" to "Depreciation Expense", which typically does not change

much from one quarter to another. Weights for the “Market” scoring methods are based on the

relative coefficients obtained by regressing absolute excess return around earnings

announcements on absolute unexpected earnings, control variables (described below) and an

indicator variable that is equal to "1" if the line item is available in the press release. We explain

and demonstrate the use of the three scoring methods in the Appendix. We report results based

on the Equal Scoring method and conduct sensitivity analysis using the other two methods.

11

The amount of detail in an earnings press release should depend on benefits accruing to the

company from supplying additional information (voluntarily), on disclosure costs, and on capital

market participants' demand for information. Companies may disclose more information in

preliminary earnings press releases to support and interpret good earnings news or to provide

mitigating explanations in cases of bad earnings news. As the propensity to disclose more

information may be different in good and bad news, we examine the association between the

detail score and positive and negative unexpected earnings. Similarly, companies with more

volatile earnings are more likely to increase the amount of detail in earnings press releases to

mitigate the effects of earnings variability on share prices.

Investors' demand for additional financial information may also drive companies to include

more detail in their quarterly earnings press releases. For example, it is expected that larger firms

will disclose additional detail due to their more complex structure of operations than smaller

companies, and also because shares of larger companies are likely to be held by professional

investors with higher demand for more timely financial information. Similarly, the demand for

additional financial information is likely to be higher for firms with larger trading volume, for

firms that seek to raise new capital and for firms that are followed by analysts as these

companies are more visible in the market. Also, more visible companies are more likely to

reduce the cost of capital through supplying more information in a timely manner (Botosan

1997).5

The company also faces direct costs of providing more detail in preliminary earnings

releases. These costs include more timely review of the press release by the company's auditor

before it is disclosed, the firm’s ability to produce other financial statement data with sufficient

5 Regarding trading volume and analysts' coverage, the arguments could be reversed: Companies have higher trading volume and are followed more closely by analysts because they provide more information. As these variables are not the focus of our study, we consider them as control variables in our study and ignore the direction of causality.

12

reliability before finalizing the data for SEC filings, and the potential that some items will have

to be revised by the SEC filing if errors are subsequently discovered (Hollie et al, 2005). A

reasonable proxy for disclosure costs is the timing of earnings announcement. Controlling for the

information itself, companies that release information relatively early are likely to have lower

direct disclosure costs. We therefore include a variable that captures the relative timing of

earnings announcements and expect that, ceteris paribus, early press releases include more

financial detail than late press releases.

We therefore estimate the following model to explain the level of detail score:

SCOREit = α0t + α1tUEit + α2tNEGit x UEit + α3tLOGMKTit + α4tEARNVOLit+ α5tVOLUMEit +

α6tFINAit + α7tANALYSTit + α8tTIMINGit + εit (1)

The variables are defined as follows: LOGMKTit is the logarithm of market value of

equity (in million of dollars) at quarter-end; UEit is the difference between quarterly earnings at

time t and quarterly earnings at time t-4 scaled by market value of equity at quarter-end;

EARNVOLit is the standard deviation of (earnings per share divided by share price), measured

over the last eight quarters, and divided by the mean of that variable during the same period;

VOLUMEit is the quarterly number of shares traded as a percentage of shares outstanding at

quarter end; FINAit is an indicator variable that obtains the value of “1” if the average free cash

flow (net operating cash flow minus capital expenditures) over the prior three years is negative

or if the firm issued stock in the current or subsequent year, and “0” otherwise; ANALYSTit is the

number of analysts that had forecasts of quarterly earnings on the summary IBES file in the

month just before the preliminary earnings announcement; and TIMINGit is an indicator variable

13

that obtains the values of "1", "2" or "3" depending on the number of days between quarter-end

and the preliminary earnings release. Specifically, TIMINGit is "1" if the number of days from

quarter-end to earnings announcement is in the lower quartile of firms for that quarter, "2" if in

the middle two quartiles, and "3" if in the upper quartile (we construct this variable separately for

firms with fiscal quarters 1-3 and quarter 4). We expect positive coefficients on all variables

except TIMING. The coefficient on TIMING is expected to be negative as early press releases are

expected to include more financial detail due to lower disclosure costs.

2.3 The Market Reaction to Preliminary Earnings as a Function of Detail Score

Our market-based tests draw on Francis et al. (2002). Similar to their study, we estimate the

association between the absolute value of excess stock returns in the three-day window centered

on the preliminary earnings announcements and the absolute value of unexpected earnings. The

null hypothesis is that the association between absolute excess returns and absolute unexpected

earnings is not affected by the amount of detail in press releases. The alternative hypothesis is

that the market reaction to earnings increases with the amount of detail in earnings press releases.

Since our sample is large and heterogeneous, we can control for various factors that are

expected to affect the absolute value of excess stock returns. We use the following model:

ABSRETPit = β0t + β1tABSUEit + β2tSCOREit + β3t ABSUEit x SCOREit + β4tTIMINGit +

β5tLOGMKTit + β6tEARNVOLit+ β7tVOLUMEit + β8tFINAit +β9tANALYSTit + νit (2)

The dependent variable – ABSRETPit – is the absolute value of excess stock returns in the

three-day window centered on firm i's preliminary earnings announcement for quarter t, minus

14

the absolute value of excess returns in the three-day window around day -7, where day 0 is the

preliminary earnings announcement date. ABSUEit is the absolute value of quarterly earnings

changes between earnings in quarter t and earnings in quarter t-4, scaled by market value of

equity at quarter-end. All other variables are as defined for Equation (1).

3. Sample and Descriptive Statistics

The initial sample contains all companies covered by Compustat with a market value in

excess of $10 million at quarter-end, beginning with the fourth quarter of 1990 and ending in the

third quarter of 2005, or 514,435 firm/quarter observations. We then delete the following

observations: (i) net income before extraordinary items (item 8 on quarterly Compustat) is

missing, (ii) total assets is equal to zero, (iii) total sales for the quarter is below $1 million, and

(iv) foreign-incorporated firms. This yields 334,573 observations.

At this point, we rank all firm/quarters by the preliminary unexpected earnings (preliminary

earnings minus earnings four quarters before, scaled by market value of equity at quarter-end)

and match to CRSP using the WRDS CRSP-Compustat Merged file, leaving 323,107

observations. We delete all observations where the preliminary report date is not at least three

days prior to the SEC filing date, or where the preliminary announcement date is in excess of 120

days after quarter-end, which are either very late announcers or data entry errors, leaving us with

223,272 observations. For the remaining observations, we compute Buy-and-hold excess returns

around preliminary earnings announcements using CRSP and the Fama-French size and book-to-

market 6-group portfolios, ending in a sample with 216,029 observations. This becomes our

sample, although for some tests additional variables will cause elimination of observations with

missing data on these variables.

15

Excess returns are measured in the three-day period (-1,+1) centered on the release of

preliminary earnings (day 0) or day -7. Excess Buy-and-hold returns are calculated as the Buy-

and-hold return from CRSP minus the Buy-and-hold return on the equally-weighted portfolio of

firms with the same size (market value of equity) and book-to-market (B/M) ratio. Daily returns

and cut-off points on the size and B/M portfolios are obtained from Prof. Kenneth French’s data

library, based on classification of the population into six (two size and three B/M) portfolios.

Observations in the top and bottom 0.5% of excess return are deleted from the sample to ensure

that our results are not driven by outlying returns. Portfolio returns are computed each quarter.

Table 1 presents the number of firm/quarter observations per year, median firm size

(market value of equity), percentage of loss-reporting companies and median annual detail score.

Consistent with prior studies, the percentage of loss-reporting companies reached its highest

level in 2001 (34.1%) and decreased afterwards. Median detail score increased over the study

period, reaching 0.69 in 2005, with a dramatic shift in 2000.

(Table 1 about here)

Figure 1 presents mean and median detail score for each quarter during the sample period

(Q4 1990 – Q3 2005). The pattern of detail score has changed considerably during the sample

period. From 1990 to 1996, mean and median detail scores were stable (mean around 0.25 and

median around 0.15). From 1997 to 1999, mean and median scores vary significantly.6 Since

2000, we observe a monotonic increase in scores, primarily because of the disclosure of balance

sheet and cash flow information in press releases. This may be related to Regulation Fair

disclosure (REG FD) in October of 2000, which required firms to disclose material financial

6 Discussions with Compustat executives indicated that there were fewer employees to collect and enter data during the dot.com period, which may have resulted in a smaller number of preliminary line items that entered the Compustat database with the update code of “2”, although those items might in fact have been available in the preliminary earnings releases. Instead, Compustat may have entered the final data with an update code of “3” for these line items. In sensitivity analysis, we repeat the tests without the affected quarters.

16

information to all investors simultaneously. Thus, firms that had previously disclosed balance

sheet and cash flow information to analysts in closed conference calls may have begun including

these additional details in preliminary earnings press releases. Currently, detail scores are close

to 0.70.7

(Figure 1 about here)

Table 2 presents the percentage of companies that include each financial statement item in

their preliminary earnings press releases for selected years. The table demonstrates a significant

increase in reporting detail following REG FD, mostly with respect to balance sheet and cash

flow information. Sales and summary figures such as Income from Continuing Operations, Total

Assets and Total Liabilities have been traditionally the most frequently disclosed items. An

interesting pattern is observed for cash flow items. In 1991, about 6% of companies included

cash flow items in press releases. This percentage decreased during the 1990s (potentially due to

inclusion of many additional smaller firms in the Compustat database), but started to increase

following REG FD. In 2005, about 20% of companies included cash flow items in their press

releases.


In untabulated analysis, we examine whether firms choose the level of financial statement

disclosure and maintain it across adjacent quarters. Out of 183,374 observations (firm-quarters

where adjacent quarter’s data are available), our Score variable (measured between 0 and 1, with

a mean of 0.456 for this sub-sample) has changed by less than 0.01 in 106,391 cases, by less than

0.05 in 140,106 cases and by less than 0.10 in less than 157,663 cases. Thus, firms that choose a

7 We also observe a change in the skewness of the detail score. Prior to 2000, median score was below the mean suggesting that a small subset of observation exhibits high scores pulling the mean upwards. After 2000, median score is generally above the mean suggesting that a relatively small subset of companies exhibits low scores pulling the mean down. This phenomenon demonstrates a fundamental shift in disclosure policy and may be attributed to an external shock such as REG FD.

17

disclosure level tend to maintain it consistently, and seldom make changes in that chosen level of

disclosure.

4. Results

4.1 The Determinants of Detail Score

Table 3 presents median annual detail scores by size levels, timing of earnings

announcements and analysts' coverage. We form three size levels based on market value of

equity at quarter-end, where the group of 'small' and 'large' firms contains the lower and upper

quartiles, respectively, and the 'medium' group contains the middle two quartiles. We also assign

observations into three groups according to earnings announcement timing (25%, 50%, and 25%)

as explained above. Regarding analysts coverage, a firm/quarter observation is considered

covered by analysts if at least one analyst had an earnings forecast for that quarter.

The table shows that detail scores were higher for large firms than for medium firms

between 1990 and 1996. Since 2000, detail scores for medium and large firms are similar and

even slightly larger in some cases for medium size companies. Also, in all quarters, detail scores

for medium size companies are larger than those of small companies, although the difference in

scores decreases after 2001. For example, in 1990 the median score was 0.115 for small firms,

0.148 for medium size companies and 0.269 for large firms. In contrast, in 2005, small firms had

a median score of 0.731, below the score for medium size firms, 0.800, which, in turn, is almost

equal to that of large firms, 0.808.

The timing of earnings announcements (the number of days between quarter-end and the

preliminary earnings announcement date) is also associated with detail score. The table shows

that firms that announce early have a higher score than firms that disclose in the middle 50%,

18

which, in turn, have higher scores than 'late' firms. This behavior is consistent over the entire

sample period. Detail scores also vary significantly by analyst coverage. Companies that are

followed by analysts have higher detail scores than companies that are not followed by analysts.

The difference in scores is attributed in part to differences in firm size as small firms are more

likely not followed by analysts, and firms “neglected” by sophisticated investors may not find it

in their best interest to provide additional detail in preliminary earnings announcements due to

less investors’ pressure or the lower likelihood of accessing the capital markets.


Table 4 provides the results of estimating Equation (1). We estimate this equation

separately for each quarter and report average coefficients and standard errors similar to Fama

and MacBeth (1973). First, we estimate Equation (1) using the entire sample. The coefficient on

unexpected earnings is 0.293 (significant at the 0.01 level, t = 13.02) when unexpected earnings

are positive. This coefficient is 0.026 higher when unexpected earnings are negative and the

difference between the coefficients is significant at the 0.01 level (t = 12.35). This result is

consistent with the argument that the propensity to disclose more information in preliminary

earnings announcements is higher when the earnings news is negative than when it is positive.

Possible explanations for this behavior are the desire to reduce litigation costs by providing

additional explanatory information when earnings changes are negative, or the desire to provide

supplementary information that will help investors react less negatively to the earnings news.

The volatility of earnings over the last eight quarters is also positively associated with

detail score, as reflected by the positive coefficient on EARNVOL (0.007, t = 14.06, significant at

the 0.01 level). This result suggests that companies with more volatile earnings provide more

detailed earnings press releases, as this information is useful for investors in interpreting the

19

earnings changes.

As expected, the coefficient on LOGMKT (firm size) is positive and significant at the 0.01

level, suggesting that, on average, large companies provide more information in earnings press

releases than small companies, likely in response to higher demand for information by

professional investors that tend to be more concentrated in large companies. Furthermore,

companies with higher volume of trade disclose more information in their press releases, as

reflected by the positive and significant coefficient on VOLUME. However, in this case it is

difficult to say whether higher volume translates to more demand for information or that more

information induces higher trading volume.

Companies that are more likely to raise capital in the market have, on average, higher detail

score, as reflected by the positive coefficient on FINA (0.031, t = 8.99). This result supports the

argument that companies that need access to the capital markets respond to a higher demand for

financial information by providing more detailed earnings press releases, attempting to reduce

their cost of capital. Furthermore, as demand for financial information increases with analysts'

coverage, companies that are followed by analysts respond to this higher demand by providing

more detailed earnings press releases. This is reflected in the positive coefficient on ANALYST

(0.003, t = 10.38). Finally, the coefficient on TIMING is negative, as expected, and highly

significant suggesting that, ceteris paribus, early earnings announcements are more detailed than

late announcements due to lower direct disclosure costs.8

We also estimate Equation (1) for three size levels and three levels of timing. Specifically,

we report results for the lower size quartile (denoted as "small"), the middle two quartiles

8 This result may seem counter-intuitive because to have more financial statement detail available for the earnings announcement may require more time after quarter-end. However, as reported in Table 3, larger companies tend to both announce their preliminary earnings earlier than smaller companies, and also tend to have more details in their announcements.

20

("medium") and for the upper size quartile ("large"). Several points are worth noting:

(i) The intercept increases with size, consistent with the results reported in Table 3 that larger

companies provide more information in their earnings press releases.

(ii) The association between positive earnings changes and detail score is stronger in large

companies than in small companies. However, the tendency to disclose more information

when earnings changes are negative is stronger in small companies than in large

companies, although this phenomenon exists in all three size levels. A possible explanation

for this result is that the probability of negative earnings is lower for larger companies and

consequently the propensity to disclose more information when earnings are negative is

diminished.

(iii) The strength of the association between Detail score and earnings volatility decreases with

size. This is expected as earnings volatility decreases with firm size as well.

(iv) The association between analysts' coverage and detail score is strongest for small

companies. This is because the number of "neglected" companies is relatively large in

small companies.

(v) The coefficient on TIMINGit becomes more negative with firm size. This is primarily

because there is much less variation in earnings announcement timing of small companies,

as many of them announce earnings late.

To complete the analysis in Table 3, we estimate Equation (1) separately for three levels of

announcement timing. Consistent with our prior findings, late earnings announcements include

less detail than early announcements, as reflected by the lower intercept for the 'late' group

relative to the 'early' group. The association between earnings changes and detail score is

consistent across different levels of announcement timing. In particular, detail score is higher in

21

negative earnings changes than in positive earnings changes in all three timing levels, and the

difference in detail score between positive and negative earnings changes is significant at the

0.01 level in all cases. All other variables exhibit similar behavior and are not sensitive to

different timing levels.9


4.2 Detail Score and the Market Reactions to Earnings Announcements

Before we turn to analyzing the effect of detail score on absolute excess stock returns, we

estimate the earnings response coefficients using the following model:

RETPit =γ0t + γ1tUEit + ηit (3)

RETPit is excess Buy-and-hold stock return in the three days centered on the preliminary

earnings announcements of firm i in quarter t; and UEit is the change in quarterly earnings from

quarter t-4 to quarter t scaled by market value of equity at quarter-end. The model is estimated

cross-sectionally each quarter for 60 quarters. The regression coefficients as well as standard

errors are obtained using the Fama and MacBeth (1973) approach. The results, which are

reported in Table 5, are consistent with prior studies. The earnings response coefficient is

positive and statistically significant, 0.022 (t = 10.72), and the average quarterly R2 is very low

(0.003). We also estimate the model for the three size groups (small, medium and large

companies), and obtain results consistent with prior studies; the earnings response coefficients

decrease with size. Thus, we feel that our sample is not different from samples used in previous

9 We regressed each of the quarterly coefficients in Equation (1) on a time counter. The results (not tabulated) indicate that most coefficients are quite stable over time. An exception is the coefficient on unexpected earnings, which increases steadily and significantly over time, suggesting that firms increasingly tend to provide additional detail when earnings surprises are larger. We also repeated the analysis with Subjective and Market Scoring obtaining similar results.

22

market reactions studies so the influence of detail on the association between market reactions

and earnings surprises is not due to a unique sample peculiarities.


Table 6 presents the results of estimating Equation (2). We also present results for two

distinct time periods – Pre REG FD (40 quarters from Q4 1990 to Q3 2000) and Post REG FD

(20 quarters from Q4 2000 to Q3 2005).

Consider first the results for all quarters. The coefficient on absolute unexpected earnings

(ABSUEit) is positive, as expected from prior studies, and significant at the 0.05 level (0.004, t =

2.17) for the entire sample period. This coefficient is larger in the Post REG FD period than in

the Pre REG FD period, where the latter is insignificantly different from zero. The coefficient on

SCOREit is positive and significant (0.003, t = 3.63), in line with Francis et al (2000), suggesting

that absolute excess stock returns around preliminary earnings announcements increase with the

amount of financial statement information disclosed concurrently with earnings. Moreover, the

coefficient on the interaction variable between detail score and absolute unexpected earnings

(ABSUEit x SCOREit) is also positive and significant at the 0.01 level (0.030, t = 4.48),

suggesting that additional financial statement detail is used more strongly by investors when

earnings surprises are extreme.

The coefficient on the detail score is significant and larger in the Post REG FD period than

in the Pre REG FD period, where the latter is insignificantly different from zero. In contrast, the

interaction between score and absolute unexpected earnings is significantly different from zero

for the Pre REG FD period, but insignificantly different from zero in the Post REG FD period.

These results are new to the literature and provide further contribution beyond Francis et al.

(2002), using a much larger and heterogeneous sample and controlling for the information and

23

regulatory environment of companies as well as the timing of the preliminary earnings

announcements.

Turning to the control variables in the model, the coefficient on TIMING is insignificantly

different from zero, once we control for other dimensions such as size and analyst following. As

expected, the coefficient on firm size (LOGMKTit) is negative (-0.003, t = -15.33) suggesting that

absolute excess returns around earnings announcements decrease with firm size, likely because

larger firms have more informative environments, so earnings affect prices less strongly. Also,

the positive coefficient on earnings volatility (0.001, t = 9.09) suggests that firms with more

volatile earnings are likely to have stronger market reactions when their earnings are announced.

Similarly, the coefficient on VOLUMEit is positive, as expected (0.010, t = 10.88), suggesting

that firms with higher share turnover are likely to also have stronger market reactions around

earnings announcements, likely because they are followed more closely by market participants.

Finally, the coefficients on FINAit and ANALYSTit are both positive, although the first is

insignificantly different from zero, suggesting that companies that are more likely to raise new

capital and companies that are followed by analysts are paid greater attention by market

participants when their earnings are announced, hence their larger absolute excess returns around

earnings press releases. These coefficients are not materially different before and after REG FD.


Since the timing of the preliminary earnings announcement is correlated with the level of

detail in the press announcements and firms’ size, we estimate Equation (2) separately for three

timing groups (early, middle, and late) and three size levels (small, medium, and large). The

results are reported in Table 7. Referring to the timing groups, the regression intercepts increase

with timing suggesting that excess stock returns are more volatile as earnings are released

24

relatively late, although this may also be related to the smaller size of late firms. Also, the

coefficients on absolute unexpected earnings (ABSUEit) are positive, but significant at the 0.05

level or better just for the early group. The coefficient on SCOREit is significantly different from

zero only for the early and middle group. Lastly, the coefficient on the interaction variable

ABSUEit x SCOREit is significant only for the 'middle' group – the 50% of the observations that

are neither early nor late. These results suggest that detail score provides more explanatory

power for extreme earnings when the timing of the press release is not extreme. In early or late

announcements, the timing of information release likely reveals additional information to the

market as well as information transfer between companies in the same industry, reducing the

explanatory power of SCOREit.

Turning to the size levels, several results are worth noting:

(i) The regression intercepts indicate that earnings news are more informative for smaller

companies with weaker information environments. Also, the coefficient on absolute

unexpected earnings is larger for small companies than for large companies, indicating the

larger role earnings news play for smaller firms.

(ii) The coefficient on SCOREit is significant only for medium size companies, suggesting that

the richer information environment of larger firms makes the additional level of detail in

the preliminary earnings announcement less relevant for investors, who may have obtained

that information from other sources.

(iii) The coefficient on ABSUEit x SCOREit is significant only for small companies. This result

suggests that absolute stock returns depend much less on the release of concurrent

information with earnings for large firms with richer information environments.

25

Overall, the results in Table 7 suggest that both timing of press releases and firm size are

major factors in explaining the association between absolute unexpected earnings and absolute

excess stock returns. The Table identifies sub-samples in which the detail score is a significant

contributor to the explanatory power of the association between earnings news and absolute

excess returns incrementally to timing and size. In particular, detail score is incrementally

significant in companies that are neither late in releasing preliminary earnings nor large. Recall

that large companies typically announce early and for those that provide more detail in their

preliminary announcements we expect a stronger market reaction. Small companies that report

late should have weaker market reactions.


As Francis et al. (2002) and Lo and Lys (2001) point out, there have been significant

changes over time in the association between absolute excess returns and absolute unexpected

earnings. These findings motivate us to examine the time series behavior of the coefficients

obtained from Equation (2). We estimate the following models using the 60 coefficients from the

quarterly cross-sectional regressions:

β1t = δ0 + δ1COUNTERt +λt (4a)

β2t = δ0 + δ1COUNTERt +λt (4b)

β3t = δ0 + δ1COUNTERt +λt (4c)

The dependent variables in Equation (4) are the regression coefficients obtained from Equation

(2) and COUNTER is a time variable from 1 (the first quarter in our data, Q4 1990) to 60 (the

26

last quarter of our data, Q3 2005). Recall that the cross-sectional regressions control for the other

variables (earnings volatility, analyst following, etc.

The results, which are reported in Table 8, show that the association between absolute

excess returns and absolute unexpected earnings has changed significantly over time, and also

that this association has changed significantly over time according to the level of financial

statement detail disclosed in the preliminary earnings announcements. Specifically, the

coefficient on unexpected earnings, ABSUE, has decreased during the sample period (-0.006, t =

-2.85) suggesting that the contribution of absolute unexpected earnings alone to the explanation

of absolute returns has deteriorated over time. In contrast, the role of information provided

concurrently with earnings has increased over time, as reflected by the increase in the coefficient

on SCORE (0.005, t = 2.81). However, the interaction between concurrent information and

earnings news has also decreased over time (-0.028, t = -2.21), suggesting that the additional

effects of financial statement detail on stock returns is for extreme earnings is decreasing over

time. Overall, the results in Table 8 highlight the increasing importance of concurrent

information in press releases over time and its effects on market reactions. These results also

support the argument in Lo and Lys (2001) that the value relevance of earnings has decreased

over time, but that the quantity and relevance of concurrent information in preliminary earnings

announcements have increased over time.


Our final analysis, provided in Table 9, focuses on matched samples rather than on linear

regressions. We identify companies with detail score in the upper quartile of the distribution and

match them with similar companies whose detail scores are in the bottom quartile during the

27

same quarter. We then compute average excess returns and average absolute excess returns

around the preliminary earnings announcements for the high and low detail samples.

In the first matching, we identify companies with high detail score and companies with low

detail score that are similar in size (where the absolute size difference is capped at 10% of the

high detail score company). The average excess return is larger for companies with high detail

score than for companies with low detail score (at the 0.05 level). Furthermore, the average

absolute excess returns are larger for companies with high detail score by 0.88% (t = 22.69). This

result is consistent with our prior analysis and with Francis et al. (2002). The second match is

performed by firm size and the decile of unexpected earnings. The difference in average excess

returns is close to zero, but the difference in average absolute excess returns increases to 1.12%

(t = 29.14). The third matching adds the number of analysts as another factor while the fourth

matching takes into account all of the above and the timing of press releases. The results remain

consistent – no difference in average excess returns but significant differences in absolute excess

returns. The results in Table 9 clearly show that increasing the amount of financial statement

detail disclosed concurrently with earnings is associated with stronger market reactions around

the earnings press release dates.


4.3 Sensitivity Analysis

All our analyses are performed using the subjective and market weighting to determine the

detail score. The results are insensitive to the alternative scoring method.

We estimated Equation (2) separately for negative and positive unexpected earnings. The

results about the effects of detail on the association between absolute excess returns and absolute

28

unexpected earnings are stronger for negative unexpected earnings changes than for positive

ones, indicating that market participants incorporate more of the non-earnings information when

earnings decline.

To assess the effect of survivorship on our analysis and to facilitate better comparison with

the results obtained by Francis et al. (2002), we estimated Equation (2) using companies with at

least 55 quarters of data. We find that the coefficients on detail score (SCOREit) and the

interaction between detail score and absolute unexpected earnings (ABSUEit x SCOREit) are

larger and statistically more significant than those reported in Table 6.

We repeated the market reaction tests with an additional dummy variable indicating

whether the firm reported a loss during the quarter or not. The results on the detail score

coefficients were qualitatively the same. We also repeated the analysis without the first two

quarters of 1999, when Compustat may have been unable to record all the preliminary

information in a timely manner due to staff shortages. The market reactions results are a bit

stronger than those reported in Table 6 with respect to the effects of the detail score, and we

obtain the same conclusions as those in Table 8 about the evolution of the coefficients on

earnings, detail score and the interaction between the two over time.

5. Summary and Conclusions

This study investigates the effects of additional financial statement disclosure in

preliminary earnings announcements on the association between absolute unexpected earnings

and absolute excess returns in the three-day window centered on the preliminary earnings

announcements. It shows that market reactions to unexpected earnings increases with the amount

29

of financial statement detail concurrently disclosed with earnings, after controlling for variables

that affect the firm's information environment.

The study is based on a very large sample of firms and uses a unique database to study the

effects of additional financial statement disclosure in preliminary earnings announcements. This

extensive sample allows us to investigate the determinants of the level of detail in preliminary

earnings announcements, their evolution over time, and control for the effects of the information

environment on the relationship between earnings surprises, excess returns and the level of

additional financial statement detail in preliminary earnings announcements.

We find that the level of financial statement detail in preliminary earnings releases has

increased over time and in particular after the enactment of Regulation Fair Disclosure in 2000.

Typically, there is a positive relationship between the level of additional financial statement

detail in preliminary announcements and size, investors’ interest in the company, the company’s

earning volatility and firms’ likelihood to access the capital markets. There is an inverse

relationship between the announcement timing and the level of detail in the announcement;

larger firms tend to announce early and to have more financial statement detail in their

preliminary earnings announcements than smaller firms.

We also find that the association between market reactions and unexpected earnings is

affected positively and significantly by the level of additional financial statement detail in

preliminary earnings announcements. These effects are stronger for smaller firms and for firms

that announce early. Extending Francis et al. (2000), we find significant explanation for the

interaction between the level of financial statement detail and earnings surprises as it relates to

market reactions; this indicates that more extreme earnings surprises which are accompanied by

greater detail tend to be more strongly associated with greater market reactions. This implies that

30

investors likely seek and utilize other than earnings information when earnings surprises are

extreme, probably attempting to better assess the persistence of the earnings surprises.

The results of this study have implications for academics, managers and investors. They

highlight the determinants and effects of additional (to earnings) voluntary disclosures, and how

they are utilized by market participants. The results can be used by management to assess the

likely effects of a decision to increase or decrease the level of additional financial statement

detail in preliminary earnings announcements. Finally, investors should pay close attention to

firms that decide to increase or decrease the level of financial statement detail in their

preliminary earnings announcements. Such decisions are likely driven by the firms’ intention to

attract more active and sophisticated investors in cases of increasing the level of detail, or to “fly

under the radar screen” in cases of decreasing the level of detail. These actions have implications

beyond the additional information that is provided or withheld.

31

References Ball, R. and P. Brown. 1968. “An Empirical Evaluation of Accounting Income Numbers.”

Journal of Accounting Research, vol. 6, pp. 159–177. Beaver, W. H. 1968. "The information Content of Annual Earnings Announcements." Journal of

Accounting Research, vol. 6 (Supplement), pp. 67-92. Botosan, C. 1997. "Disclosure Level and the Cost of Equity Capital." The Accounting Review,

vol. 72, no. 3 (July), pp. Bushee, B.J., D.A. Matsumoto, and G..S. Miller. 2004. "Managerial and Investor Responses to

Disclosure Regulation: The Case of Reg FD and Conference Calls." The Accounting Review, vol. 79, no. 3 (July), pp. 617-643.

Chambers, A., and S. Penman. 1984. "Timeliness of Reporting and the Stock Price Reaction to

Earnings Announcements." Journal of Accounting Research, vol. 22, pp. 21-47. Chen, S., M. L. DeFond and C. W. Park, 2002. "Voluntary Disclosure of Balance Sheet

Information in Quarterly Earnings Announcements." Journal of Accounting and Economics, vol. 33, pp. 229-251.

Collins, D, E. Maydew, and I. Weiss. 1997. "Changes in the Value-Relevance of Earnings and

Book Values in the Past Forty Years." Journal of Accounting and Economics, vol. 24 (December), pp. 39-67.

Easton, P., and M. Zmijewski. 1989. "Cross Sectional Variation in the Stock Market Response to

Accounting Earnings Announcements." Journal of Accounting and Economics, vol. 11 (July), pp. 117-141

Easton, P., and M. Zmijewski. 1993. "SEC Form 10K/10Q Reports and Annual Reports to

Shareholders: Reporting Lags and Squared Market Model Prediction Errors." Journal of Accounting Research, vol. 31, no. 1, pp. 113-139.

Ertimur, Y., J. Livnat and M. Martikainen. 2003. “Differential Market Responses to Revenue and

Expense Surprises." Review of Accounting Studies, vol. 8, nos. 2-3. Fama, E. F., and J.D. MacBeth. 1973. “Risk, Return, and Equilibrium: Empirical Tests.” Journal

of Political Economy, vol. 81, pp. 607-636. Francis, J., K. Schipper. 2002. "Have Financial Statements Lost Their Relevance? Journal of

Accounting Research, vol. 37 (Autumn), pp. 319-352. Francis, J., K. Schipper, and L. Vincent. 2002. "Expanded Disclosures and the Increased

Usefulness of Earnings Announcements." The Accounting Review, Vol. 77 (3), pp. 515-546.

32

Griffin, P. 2003. "Got Information? Investor Response to Form 10-K and Form 10-Q EDGAR

Filings." Review of Accounting Studies, vol. 8, pp. 433-460. Hollie, D., J. Livnat and B. Segal. 2005. “Oops, Our Earnings Were Indeed Preliminary”: Market

Reactions to Companies that Subsequently File Different Earnings with the SEC,” Journal of Portfolio Management, vol. 31, pp. 94-104.

Jegadeesh, N., and J. Livnat. 2006. "Revenue Surprises and Stock Returns." Journal of

Accounting and Economics, Forthcoming. Kasznik, R., and B. Lev. 1995. "To warn or not to warn: Management disclosures in the face of

an earnings surprise." The Accounting Review, vol. 70, pp. 113-134. Kimbrough, M.D. 2005. "The Effect of Conference Calls on Analyst and Market Underreaction

to Earnings Announcements. The Accounting Review, vol. 80, no. 1, pp. 189-219. Kohlbeck, M.J. and M.J. Magilke. 2002. "The Impact of Concurrent Conference Calls on the

Information Content of Earnings Announcements" (February). Working paper, University of Wisconsin (available at http://ssrn.com/abstract=302230).

Kothari, S.P. 2001. “Capital Markets Research in Accounting.” Journal of Accounting and

Economics, vol. 31, pp. 105-231. Kross, W. and D. Schroeder. 1984. “An Empirical Investigation of the Effects of Quarterly

Earnings Announcements on Stock Returns.” Journal of Accounting Research, vol. 22, pp. 153-176.

Lev, B. 1989. "On the Usefulness of Earnings and Earnings Research: Lessons and Directions

from Two Decades of Empirical Research." Journal of Accounting Research, vol. 27 (Supplement), pp. 153-201.

Lev, B., and P. Zarowin. 1999. "The Boundaries of Financial Reporting and How to Extend

Them." Journal of Accounting Research, vol. 37 (Autumn), pp. 353-385. Lo, K., and T. Lys. 2001. "Bridging the Gap between Value Relevance and Information

Content." Working paper, Northwestern University. Skinner, D.J. 1994. “Why Firms Voluntarily Disclose Bad News.” Journal of Accounting

Research, vol. 32 (Spring), pp. 38-60. Stice, E. 1991."The Market Reaction to 10-K and 10-Q Filing and to Subsequent The Wall Street

Journal Earnings Announcements." The Accounting Review, vol. 66, pp. 42-55.

33

Wilson, G. P. 1987. “The Incremental Information Content of the Accrual and Funds Components of Earnings After Controlling for Earnings.” The Accounting Review, vol. 62, pp. 293-322.

34

Table 1 Sample and Descriptive Statistics*

Year N Median

Size %

Loss Mean Equal

SCORE 1990 2,197 126.3 18.5 0.3111991 9,910 146.1 18.3 0.3031992 10,803 148.2 17.0 0.2911993 11,664 165.2 18.2 0.2961994 14,079 144.1 14.9 0.2821995 15,247 165.3 15.8 0.2921996 16,241 194.7 18.0 0.3371997 18,036 217.1 19.5 0.3841998 17,840 215.2 23.3 0.4521999 17,117 220.9 24.0 0.3792000 16,766 239.1 28.9 0.5242001 16,020 249.5 34.1 0.5792002 13,871 272.2 29.0 0.6052003 13,166 293.5 25.1 0.6432004 13,643 433.0 20.1 0.6532005 9,429 476.9 19.8 0.690All 216,029 219.5 22.0 0.444

*Notes: 1. N – The number of firm/quarter observations in each year during the sample period (Q4 1990

– Q3 2005). The sample includes all quarter/year observations that issued preliminary earnings reports and for which return data is available. We use only preliminary earnings reports that were issued 3 or more days prior to SEC filing.

2. Size - Market value of equity in millions of dollars. 3. %Loss – the percentage of companies that reported negative earnings during the quarter. 4. SCORE - A variable between zero and one that captures the amount of financial statement

detail in earnings press releases. The computation of detail scores is demonstrated in the Appendix.

35

Figure 1

Detail Score over time (Q4 1990 – Q3 2005)

Detail over Time

00.20.40.60.8

119

904

1991

419

924

1993

419

944

1995

419

964

1997

419

984

1999

420

004

2001

420

024

2003

420

044

Quarter

010002000300040005000

Mean Median N

The Figure presents mean Score, median Score and the number of observations (N) per each quarter. Score is a variable between zero and one that captures the amount of financial detail in earnings press releases. The computation of detail scores is demonstrated in the Appendix.

36

Table 2 Reporting Frequency by Financial Statement Item and Year*

1991 1993 1995 1997 1999 2001 2003 2005Income statement (10 items): Sales 0.955 0.958 0.881 0.896 0.903 0.900 0.880 0.894Cost of goods sold 0.148 0.162 0.285 0.551 0.528 0.796 0.820 0.858SG&A expenses 0.285 0.293 0.303 0.427 0.401 0.606 0.653 0.687Interest expense 0.275 0.282 0.235 0.319 0.307 0.435 0.464 0.491Special items 0.277 0.324 0.385 0.542 0.533 0.752 0.772 0.805Depreciation expenses 0.127 0.130 0.114 0.128 0.181 0.220 0.258 0.328Non-operating income 0.327 0.327 0.325 0.462 0.428 0.654 0.681 0.713Income tax expense 0.605 0.589 0.494 0.635 0.656 0.885 0.893 0.912Net income from continuing operations 0.990 0.990 0.992 0.982 0.964 0.970 0.946 0.947Extraordinary items 0.079 0.099 0.041 0.052 0.057 0.075 0.109 0.114 Balance sheet (13 items) Cash (and equivalents) 0.240 0.229 0.227 0.337 0.327 0.600 0.683 0.735Accounts receivable 0.191 0.182 0.200 0.299 0.318 0.570 0.658 0.692Inventory 0.193 0.183 0.193 0.289 0.290 0.525 0.608 0.657Current assets 0.233 0.218 0.198 0.308 0.279 0.490 0.539 0.542Property Plant and Equipment 0.278 0.248 0.227 0.327 0.310 0.557 0.627 0.659Total assets 0.320 0.299 0.279 0.421 0.430 0.728 0.798 0.833Short-term debt 0.194 0.175 0.171 0.243 0.249 0.446 0.535 0.575Accounts payable 0.166 0.152 0.172 0.255 0.293 0.523 0.606 0.655Current liabilities 0.235 0.220 0.200 0.311 0.280 0.491 0.542 0.544Long-term debt 0.280 0.255 0.232 0.339 0.328 0.566 0.656 0.714Other liabilities 0.228 0.212 0.208 0.304 0.291 0.533 0.619 0.670Total liabilities 0.296 0.287 0.274 0.415 0.421 0.713 0.785 0.826Stockholders equity, total 0.248 0.273 0.247 0.372 0.338 0.688 0.760 0.815 Cash flow (4 items) Net operating cash flow 0.066 0.043 0.028 0.034 0.013 0.073 0.170 0.222Capital expenditures 0.062 0.041 0.025 0.031 0.013 0.067 0.158 0.207Total investing cash flows 0.064 0.042 0.026 0.032 0.013 0.071 0.166 0.218Total financing cash flows 0.064 0.042 0.027 0.032 0.013 0.071 0.165 0.218 *Note: The table presents percentage of companies that include each financial statement item in their preliminary earnings press releases. Frequencies are reported for odd years.

37

Table 3 Median Annual Detail Score (Equal Scoring) by size levels and analysts coverage*

Detail Score by Firm Size Detail Score by Timing of

Earnings Announcements Detail Score by

Analyst CoverageYear Small Medium Large Early Middle Late Yes No 1990 0.115 0.148 0.269 0.269 0.167 0.115 0.115 0.200 1991 0.115 0.148 0.269 0.269 0.154 0.115 0.115 0.200 1992 0.115 0.136 0.250 0.240 0.154 0.115 0.115 0.190 1993 0.115 0.154 0.269 0.250 0.154 0.115 0.115 0.208 1994 0.100 0.120 0.200 0.154 0.120 0.115 0.100 0.154 1995 0.105 0.120 0.154 0.136 0.115 0.115 0.105 0.148 1996 0.115 0.250 0.280 0.250 0.263 0.115 0.115 0.280 1997 0.231 0.318 0.308 0.346 0.308 0.192 0.154 0.333 1998 0.333 0.455 0.385 0.560 0.450 0.308 0.308 0.476 1999 0.120 0.308 0.381 0.412 0.308 0.143 0.125 0.333 2000 0.391 0.615 0.632 0.680 0.600 0.417 0.368 0.654 2001 0.578 0.692 0.654 0.722 0.692 0.500 0.476 0.720 2002 0.615 0.731 0.714 0.731 0.731 0.538 0.538 0.731 2003 0.654 0.769 0.769 0.789 0.762 0.565 0.615 0.769 2004 0.680 0.769 0.800 0.800 0.769 0.579 0.615 0.789 2005 0.731 0.800 0.808 0.808 0.800 0.714 0.667 0.800

*Notes: 1. The Table presents median detail scores for each year during the sample period (Q4 1990 –

Q3 2005) for three size levels (small, medium, large), three levels of earnings announcement timing (early, middle, late) and by analyst coverage (Yes, No).

2. The sample includes all quarter/year observations that issued preliminary earnings reports

and for which return data is available. We use only preliminary earnings reports that were issued 3 or more days prior to SEC filing.

3. Detail Score - A variable between zero and one that captures the amount of financial

statement detail in earnings press releases. The computation of detail scores is demonstrated in the Appendix. The Table presents statistics using the Equal Scoring method.

4. Size levels - Formed every quarter based on market value of equity. The small and large

levels include the lower and upper quartiles, respectively, and the medium level includes the middle two quartiles.

5. Timing levels - Measured every quarter based on the number of days from quarter-end to

earnings announcement. Early and late levels include the lower and upper quartiles, respectively, and the middle level includes the middle two quartiles.

6. Analyst Coverage - A firm/quarter observation is considered as covered by analysts if at least

one analyst is covering the firm during that quarter.

38

Table 4 Determinants of Detail Score for the Total Sample, by Size Levels and by Timing Levels

Total

SampleSize Levels

Small Medium Large Timing Levels

Early Medium Late Variable Sign Mean

Coeff. Mean Coeff.

Mean Coeff.

Mean Coeff.

Mean Coeff.

Mean Coeff.

Mean Coeff.

INTERCEPT + 0.363 0.323 0.433 0.461 0.353 0.337 0.253t-value 13.67 17.42 19.35 26.11 13.19 10.98 10.84 UE + 0.293 0.222 0.319 0.401 0.327 0.288 0.274t-value 13.02 9.15 13.30 23.06 14.06 12.82 10.48 UE x NEG ? 0.026 0.031 0.025 0.014 0.024 0.026 0.029t-value 12.35 9.29 11.24 3.44 7.02 8.21 11.70 LOGMKT + 0.010 -- -- -- 0.011 0.010 0.009t-value 5.63 -- -- -- 5.82 4.02 6.54 EARNVOL + 0.007 0.010 0.008 0.003 0.007 0.007 0.007t-value 14.06 12.16 11.64 3.83 5.98 9.74 9.96 VOLUME + 0.056 0.062 0.036 0.044 0.050 0.062 0.045t-value 11.06 7.90 6.28 6.09 6.15 11.89 8.22 FINA + 0.031 0.040 0.013 0.032 0.039 0.024 0.032t-value 8.99 7.73 3.96 5.89 5.57 5.79 7.06 ANALYST + 0.003 0.035 0.010 0.002 0.003 0.002 0.007t-value 10.38 20.31 19.22 5.30 6.05 6.70 15.62 TIMING - -0.036 -0.028 -0.035 -0.041 -- -- --t-value -23.41 -14.09 -20.67 -21.69 -- -- -- Average R2 0.141 0.140 0.142 0.010 0.079 0.089 0.163Average observations 3,264 811 1,595 858 829 1,645 790

Notes: 1. The Table presents mean coefficients and standard errors based on 60 quarterly regressions

(Fama and MacBeth, 1973 regressions) for the following model:

SCOREit = α0t + α1tUEit + α2tNEGit x UEit + α3tLOGMKTit + α4tEARNVOLit+ α5tVOLUMEit + α6tFINAit + α7tANALYSTit + α8tTIMINGit + εit (1)

2. Variable definitions:

a. SCORE – A variable between zero and one that captures the amount of detail in a

39

preliminary earnings press release. See details in the Appendix. b. UE – Unexpected Earnings, measured as quarterly earnings in quarter t minus quarterly

earnings in quarter t-4, scaled by market value of equity at quarter end. c. NEG – An indicator variable obtaining the value of "1" if the change in quarterly

earnings is negative, and "0" otherwise. d. LOGMKT – Logarithm of market value of equity (in million of dollars) at quarter-end. e. EARNVOL – The standard deviation of (earnings per share over share price) over the last

8 quarters divided by the mean of (earnings per share over share price). f. VOLUME -- Quarterly number of shares traded as a percentage of shares outstanding at

quarter end. g. FINA – An indicator variable that obtains the value of “1” if average free cash flow over

the prior 3 years is negative (financing is required) or if the firm issued stock in the current or subsequent year (financing actually used), and “0” otherwise.

h. ANALYST – The number of analysts following the company that quarter. i. TIMING – Indicator variable. Initially, we measure the number of days between quarter-

end and the preliminary earnings announcement date. We assign each observation into one of three groups according to the lag from quarter-end: lowest 25% into early group, middle 50% into the middle group and largest 25% into the late group. We conduct this analysis separately for quarters 1-3 and quarter 4.

3. Coefficients and standard errors in bold are significant at the 0.01 level (2-tailed test).

40

Table 5 The Association between signed unexpected earnings and unexpected return

Fama-MacBeth Regressions using 60 quarters from Q1/90 to Q1/05

Entire Sample

Size levels Small medium large

Variable Sign Coeff. t-stat.

Coeff.t-stat.

Coeff. t-stat.

Coeff.t-stat.

INTERCEPT ? 0.004 0.009 0.002 0.003t-statistic 10.56 11.74 5.15 5.97 UE + 0.022 0.037 0.015 0.009t-statistic 10.72 10.71 5.95 4.36 Average R2 0.003 0.008 0.002 0.001Average number of observations 3,601 900 1,801 900

Notes: 1. The Table presents results for the following model: RETPit = β0t + β1tUEit + νit (3); where

RETPit is stock return in the three days centered on the preliminary earnings announcement of firm i in quarter t; and UEit is the change in quarterly earnings from quarter t-4 to quarter t scaled by market value of equity at the end of quarter t.

2. The model is estimated separately for 60 quarters. The regression coefficients and standard

errors are obtained using the Fama and MacBeth (1973) approach. 3. Results are provided for the entire sample and also for three size levels (small, medium and

large), where size is measured as market value of equity. Small and large portfolios contain the lower and upper quartiles of the sample, respectively, and the medium size portfolio contains the middle two quartiles.

4. Coefficients and standard errors in bold faces are significant at the 0.01 level (2-tailed test).

41

Table 6 Explaining the Absolute Value of Market Reaction to Preliminary Earnings

Announcements Using Three Methods of Detail Scoring (Fama-MacBeth Regressions)

Equal Scoring

Pre Post All REG FD REG FD Quarters

Variable Sign Coeff. t-stat.

Coeff. t-stat.

Coeff. t-stat.

INTERCEPT ? 0.024 0.022 0.029 t-statistic 14.61 8.48 16.81 ABSUE + 0.003 0.005 0.004 t-statistic 1.25 3.27 2.17 SCORE + 0.001 0.007 0.003 t-statistic 0.66 6.20 3.63 ABSUE x SCORE + 0.036 0.018 0.030 t-statistic 4.63 1.87 4.88 TIMING + -0.000 0.001 0.000 t-statistic -0.61 1.11 0.29 LOGMKT - -0.003 -0.003 -0.003 t-statistic -11.51 -10.90 -15.33 EARNVOL + 0.001 0.002 0.001 t-statistic 5.60 12.64 9.09 VOLUME + 0.008 0.013 0.010 t-statistic 7.16 11.08 10.88 FINA + 0.000 0.001 0.001 t-statistic 0.70 0.44 0.80 ANALYST + 0.001 0.001 0.001 t-statistic 14.65 7.86 15.72 Average R2 0.012 0.023 0.016 Average observations 3,200 3,339 3,246 Quarters 40 20 60

42

Notes: 1. Equation (2) is estimated separately for each quarter and regression coefficients and standard

errors are obtained using the Fama and MacBeth (1973) approach. The Table presents mean coefficients and corresponding t-statistics.. The model is:

ABSRETPit = β0t + β1tABSUEit + β2tSCOREit + β3t ABSUEit x SCOREit + β4tTIMINGit + β5tLOGMKTit + β6tEARNVOLit+ β7tVOLUMEit + β8tFINAit +β9tANALYSTit + νit (2)

2. Variable definitions:

a. ABSRETP – The dependent variable is absolute value of excess stock returns in the 3-day window centered on the preliminary earnings announcements date (day 0) minus the absolute value of excess returns in the 3-day window around day -7.

b. ABSUE – Absolute value of quarterly earnings changes, scaled by market value of equity at quarter-end. The change in quarterly earnings is the difference between earnings in quarter t and earnings in quarter t-4.

c. SCORE – A variable that captures the amount of detail in a preliminary earnings press release. We report results for three scoring methods (Equal, Subjective and Market) as demonstrated in the Appendix.

d. TIMING – Indicator variable. Initially, we measure the number of days between quarter-end and the preliminary earnings announcement date. We assign each observation into one of three groups according to lag from quarter-end: lowest 25% into the early group, middle 50% into the middle group and largest 25% into the late group. We conduct this analysis separately for quarters 1-3 and quarter 4.

e. LOGMKT – Logarithm of market value of equity (in million of dollars) at quarter-end. f. EARNVOL – The standard deviation of {earnings per share over share price} during the

prior 8 quarters divided by the mean of {earnings per share over share price} during the same period.

g. VOLUME -- Quarterly number of shares traded as a percentage of shares outstanding at quarter end.

h. FINA – An indicator variable that obtains the value of “1” if the average free cash flow over the prior 3 years is negative (financing is required) or if the firm issued stock in the current or subsequent year (financing actually used), and “0” otherwise.

i. ANALYST – The number of analysts following the company for that quarter. 3. Coefficients and standard errors in bold are significant at the 0.01 level (2-tailed test).

43

Table 7 Explaining the Absolute Value of Market Reaction to Preliminary Earnings

Announcements Using Report Detail and by Timing Groups - Fama-MacBeth Regressions Equal Scoring Method

Timing Groups

Early Middle Late Size Levels

Small Medium Large Variable Sign Coeff.

t-stat. Coeff. t-stat.

Coeff.t-stat.

Coeff.t-stat.

Coeff. t-stat.

Coeff.t-stat.

INTERCEPT ? 0.021 0.024 0.025 0.038 0.018 0.021t-statistic 9.21 14.56 10.86 8.96 7.45 8.66 ABSUE + 0.008 0.003 0.004 0.004 0.004 0.002t-statistic 2.16 0.84 1.70 1.10 1.62 1.08 SCORE + 0.003 0.002 0.002 0.002 0.004 0.001t-statistic 2.95 2.44 1.18 0.79 3.88 1.30 ABSUE x SCORE + 0.027 0.048 0.14 0.046 0.016 0.018t-statistic 0.96 4.20 0.94 3.00 1.87 0.93 TIMING + -- -- -- -0.001 0.000 0.001t-statistic -- -- -- -1.54 0.64 2.27 LOGMKT - -0.003 -0.003 -0.003 -0.007 -0.002 -0.002t-statistic -8.22 -13.17 -7.83 -5.98 -6.10 -7.95 EARNVOL + 0.002 0.001 0.001 0.002 0.001 0.001t-statistic 7.04 5.86 4.14 6.83 5.54 3.44 VOLUME + 0.011 0.010 0.008 0.004 0.011 0.011t-statistic 6.94 9.15 3.96 1.91 10.55 8.54 FINA + -0.000 0.001 0.001 0.001 0.000 0.001t-statistic -0.30 0.83 0.62 0.42 0.15 1.90 ANALYST + 0.001 0.001 0.001 0.001 0.001 0.001t-statistic 7.05 10.51 6.74 1.68 9.76 8.50 Average R2 0.036 0.020 0.020 0.022 0.022 0.030Average observations 819 1,632 778 799 1,582 846

Notes: 1. Equation (2) is estimated separately for 60 quarters and regression coefficients and standard

errors are obtained using the Fama and MacBeth (1973) approach. The Table presents mean

44

coefficients and corresponding t-statistics. The dependent variable is absolute excess stock returns around preliminary earnings press releases minus the absolute excess return seven days prior to the preliminary earnings release date. Estimation is done separately for three timing groups and three size levels. The model is:

ABSRETPit = β0t + β1tABSUEit + β2tSCOREit + β3t ABSUEit x SCOREit + β4tTIMINGit + β5tLOGMKTit + β6tEARNVOLit+ β7tVOLUMEit + β8tFINAit +β9tANALYSTit + νit (2)

2. Timing Groups - Initially, we measure the number of days between quarter-end and the

preliminary earnings announcement date. We assign each observation into one of three Timing groups according to lag from quarter-end: lowest 25% into early group, middle 50% into the middle group and largest 25% into the late group. We conduct this analysis separately for quarters 1-3 and quarter 4.

3. Size levels - Size is measured as market value of equity. Small and large portfolios contain

the lower and upper quartiles of the sample, respectively, and the medium size portfolio contains the middle two quartiles.

4. See Table 6 for variable definitions. 5. Coefficients and standard errors in bold faces are significant at the 0.05 level (2-tailed test).

45

Table 8

Trends in the Association between Absolute Unexpected Return and Absolute Unexpected Earnings, Detail Score and the Interaction between Absolute Unexpected Earnings and

Detail Score*

Equal Score Dependent Variable ABSUE

β1t SCORE

β2t ABSUE x SCORE

β3t INTERCEPT x 100 1.192 -0.242 7.204 t-statistic 5.55 -2.15 4.51 TIME COUNTER x 100 -0.006 0.005 -0.028 t-statistic -2.85 2.81 -2.21 R2 0.04 0.06 0.04

*Notes: 1. The Table presents results for three linear regressions. The dependent variable in each

regression consists of 60 quarterly coefficients obtained from the following regression: ABSRETPit = β0t + β1tABSUEit + β2tSCOREit + β3t ABSUEit x SCOREit + β4tLOGMKTit +

β5tEARNVOLit+ β6tVOLUMEit + β7tFINAit + β8tANALYSTit + νit (2) 2. See Table 6 for variable definitions. 3. Coefficients and t-statistics in bold indicate significance at the 0.05 level.

46

Table 9 Differences in Absolute Excess Returns between Companies with High Detail Score and Companies with Low Detail Score*

High Detail Score Low Detail Score Differences in t-statistics Matching by

Excess Return

AbsoluteExcess Return

Excess Return

Absolute Excess Return

Excess Return


Excess Return


Firm Size 0.0041 0.0568 0.0029 0.0480 0.0011 0.0088 2.34 22.69 Firm Size and Earnings Surprise Decile 0.0042 0.0577 0.0050 0.0464 -0.0008 0.0112 -1.33 29.14 Firm Size, Number of Analysts and 0.0045 0.0595 0.0045 0.0518 0.0000 0.0077 0.04 12.80 Earnings Surprise Decile Firm Size, Number of Analysts, 0.0044 0.0611 0.0053 0.0519 -0.0009 0.0092 -0.84 11.50 Earnings Surprise Decile and Timing Notes: 1. The Table presents excess stock returns and absolute value of excess stock returns for a sample of companies with high detail

score (score in the upper quartile) and a matched sample of companies with low detail score (score in the lower quartile). The table also presents the differences in excess returns and absolute value of excess returns and their significance levels.

2. Matching techniques: (i) Matching by Size (size difference is capped at 10%); (ii) Matching by Size and Earnings Surprise decile; (iii) Matching by Size, Earnings Surprise decile and the number of analysts following the firm, (iv) matching by Size, Earnings Surprise decile, number of analysts following the firm and the announcement timing (early, medium, late).

3. Excess returns are measured in the 3-day period (-1,+1) around the release of preliminary earnings. Excess Buy-and-hold returns are calculated as the Buy-and-hold return from CRSP minus the Buy-and-hold return on the portfolio of firms with the same size (market value of equity) and book-to-market (B/M) ratio. Daily returns and cut-off points on the size and B/M portfolios are obtained from Prof. Kenneth French’s data library, based on classification of the population into six (two size and three B/M) portfolios. Observations in the top and bottom 0.5% of excess return are deleted from the sample to ensure that our results are not driven by outlying returns. Portfolio returns are computed each quarter. The table presents average quarterly returns and t-statistics based on a Fama and MacBeth (1973) approach.

4. Standardized unexpected earnings are measured as: )( 4

4

−

−

−−−

=tt

tt

XXVARXXSUE δ , where Xt is earnings for quarter t, δ is the average of

Xt- Xt-4 over the prior eight quarters, and the variance of Xt- Xt-4 is also estimated over the prior eight quarters.

47

Appendix Scoring Methods

To measure disclosure detail, we initially identify 10 income statement items, 13 balance

sheet items and 4 cash flow items that form the basis for our scoring mechanism. We then assign

weights to each item based on the following methods - Equal Scoring, Subjective Scoring and

Market Scoring. For Equal Scoring, we assign equal weights of "1" to each item found on the

quarterly press release, but only if the item also appears in the subsequent SEC filing. If the item

does not appear in the subsequent SEC filing, we do not assign the item any weight, and ignore it

in our scoring. Under Subjective Scoring, we assign each item a relative weight of “1”, “2”, or

“3”, where a weight of “3” indicates high relative importance. For example, the item “Sales” is

assigned a weight of “3” whereas the item “Income Tax Expense” is assigned a weight of “1”.

Another example relates to the balance sheet: The items “Current Assets” and “Current

Liabilities” are assigned weights of “3” because they facilitate the estimation of operating cash

flows, while the items “Accounts Payable” and “Account Receivable” are assigned a weight of

“2” because they facilitate estimation of operating cash flows with greater accuracy and facilitate

estimation of current accruals. Under the Market Scoring mechanism we initially estimate the

following regression model:

ABSRETPit = β0t + β1tITEMjit + β2tTIMINGit + β3tLOGMKTit + β4tEARNVOLit+ β5tVOLUMEit +

β6tFINAit +β7tANALYSTit + νit (A1)

Where ITEMjit is an indicator variable that obtains the value of "1" if item j (27 items)

appears in company i's press release for quarter t, and "0" otherwise. All the other variables are

48

as defined above. We obtain average coefficients and corresponding standard errors using the

Fama and MacBeth, 1973 approach for 60 quarters.

We compute market weights for each item as:∑

jj

j

ββ

. Table A1 lists the financial elements

and weights for each scoring mechanism.

(Table A1 about here)

To obtain a score for each quarterly observation, we initially examine the 10-Q filing of the

company in order to determine the maximum score possible. For instance, the item “inventory”

is included in computing the maximum score only if the item was separately disclosed in the 10-

Q filing. So technically, it is possible for each firm/quarter to have a different maximum score.

Only then, we compute the actual score for the preliminary press release and divide it by the

maximum score. We demonstrate the scoring mechanisms for the quarterly press release of AAR

Corp. The actual score is 0.500, 0.509 and 0.770 using the Equal, Subjective and Market scoring

mechanisms, respectively.

49

Table A1

Weights Equal

Weights Subjective

Weights Market

Income statement (10 items): Sales 1 3 12 Cost of goods sold 1 2 2 SG&A expenses 1 2 5 Interest expense 1 1 1 Special items 1 2 4 Depreciation expenses 1 1 1 Non-operating income 1 2 3 Income tax expense 1 1 5 Net income from continuing operations 1 3 28 Extraordinary items 1 1 0 10 18 61 Balance sheet (13 items) Cash (and equivalents) 1 1 4 Accounts receivable 1 2 2 Inventory 1 2 3 Current assets 1 3 6 Property Plant and Equipment 1 1 4 Total assets 1 3 2 Short-term debt 1 2 2 Accounts payable 1 2 1 Current liabilities 1 3 6 Long-term debt 1 2 2 Other liabilities 1 1 2 Total liabilities 1 2 2 Stockholders equity, total 1 3 3 13 27 39 Cash flow (4 items) Net operating cash flow 1 3 0 Capital expenditures 1 3 0 Total investing cash flows 1 2 0 Total financing cash flows 1 3 0 4 11 0 Maximum Score 27 56 100

50

An example of an application of the scoring mechanism to a press release AAR REPORTS FISCAL YEAR 2006 FIRST QUARTER RESULTS • 22% sales growth; 109% income growth • 25% commercial sales growth; 16% defense services sales growth • 45% growth in Asia sales WOOD DALE, ILLINOIS (September 21, 2005) — AAR (NYSE: AIR) today reported net sales from continuing operations of $199.6 million for the first quarter of fiscal 2006, an increase of 22% compared to the prior year. Income from continuing operations more than doubled to $5.3 million or $0.15 per diluted share, from $2.5 million or $0.08 per diluted share in the year ago period, which included a $1.0 million pre-tax gain on extinguishment of debt. The sales and earnings growth for the quarter were driven by increased sales in the Aviation Supply Chain, Maintenance, Repair & Overhaul and Structures & Systems segments. Within the Aviation Supply Chain segment, sales increased 25% reflecting strong demand from our commercial and defense customers as AAR creates and provides cost-effective solutions for managing their supply chains. MRO segment sales increased 78% reflecting the commencement of operations at the Indianapolis Maintenance Center as well as increased sales at the Oklahoma maintenance facility. Structures and Systems sales increased 14% as the Company experienced stronger volumes across all businesses within this segment. Sales in the Aircraft Sales and Leasing segment were lower primarily due to joint venture accounting treatment, which excludes joint venture revenues from consolidated net sales. “This is a great start to fiscal 2006,” said David P. Storch, President and CEO of AAR. “We are seeing the benefits of the actions we have taken over the past few years to strengthen and grow the Company. Increasingly, our airline and defense customers are turning to AAR to meet their maintenance and engineering, supply chain and deployment needs.” Higher sales and operational efficiencies drove an improvement in gross profit margin from 16.2% to 17.4% year over year. Although selling, general and administrative costs increased as the Company made investments and prepared for growth, they declined as a total percentage of sales from 12.2% to 12.0%. Net interest expense was $0.5 million lower for the quarter due to lower average borrowings. The Company also made progress in its goal to improve asset performance, increasing working capital turnover from 3.0x to 3.6x. Investments in inventory primarily to support supply chain programs resulted in an operating cash outflow of $22 million for the quarter. We expect the returns from these investments to favorably impact results in future periods. Subsequent to the Company’s quarter-end, two valued customers, Delta Air Lines and Northwest Airlines, filed for bankruptcy. AAR was proactive in minimizing its exposure to these events, and the financial impact was minimal and was provided for in the first quarter. The Company continues to support both airlines by providing cost-effective solutions for their maintenance and supply requirements. Storch added, “We made progress on many fronts. Our results were significantly stronger than last year, and we won several new programs during the quarter, including the Airbus A400M cargo system program, a contract to provide pallets to the U.S. Air Force and the United Kingdom’s Royal Air Force E-3D AWACS and MESA Air Group supply chain programs, all of which should have a positive impact on future periods.” Storch continued, “We are also very pleased with the progress made at our recently-opened Indianapolis Maintenance Center, as the business unit produced profitable results in the quarter.” AAR is a leading provider of diverse products and value-added services to the worldwide aviation/aerospace industry. Headquartered in Wood Dale, Illinois, with locations around the world, AAR serves commercial and government aircraft fleet operators and independent service customers by providing Aviation Supply Chain services; Maintenance; Repair and Overhaul services; Structures and Systems manufacturing and Aircraft Sales and Leasing. Further information can be found at www.aarcorp.com.

51

AAR CORP. and Subsidiaries Three Months Ended August 31,

2005 2004 (Unaudited)

Consolidated Statements of Operations (In thousands except per share data)

Sales $ 199,588 $ 163,773 Cost and Expenses:

Cost of sales 164,906 137,248 Selling, general and administrative 23,901 20,040 Equity in earnings of joint ventures 205 ---

Operating income 10,986 6,485 Gain on extinguishment of debt --- 995 Interest expense 4,122 4,463 Interest income 459 283 Income from continuing operations before income taxes 7,323 3,300 Income tax expense 2,065 787 Income from continuing operations 5,258 2,513 Discontinued Operations:

Operating loss, net of tax --- (227) Net income $ 5,258 $ 2,286

Share Data: Earnings per share - Basic:

Earnings from continuing operations $ 0.16 $ 0.08 Loss from discontinued operations --- (0.01) Earnings per share – Basic $ 0.16 $ 0.07

Earnings per share – Diluted:

Earnings from continuing operations $ 0.15 $ 0.08 Loss from discontinued operations --- (0.01) Earnings per share – Diluted $ 0.15 $ 0.07

Average shares outstanding – Basic 32,961 32,243 Average shares outstanding – Diluted 37,040 36,198

August 31 August 31 2005 2004 (Unaudited) (Derived from audited financial statements)

Consolidated Balance Sheet Highlights Cash and cash equivalents $ 24,411 $ 50,338 Current assets 462,526 474,542 Current maturities of recourse LTD 437 2,123 Current liabilities (excluding debt accounts) 164,118 156,280 Net property, plant and equipment 72,565 71,474 Total assets 757,826 732,230 Recourse long-term debt 201,288 199,919 Total recourse debt 201,725 202,042 Total non-recourse debt 28,612 28,862 Stockholders’ equity 319,902 314,744 Book value per share $ 9.69 $ 9.66 Shares outstanding 33,004 32,586

52

Three Months Ended August 31,

2005 2004 Sales By Business Segment (In thousands - unaudited) Aviation Supply Chain $ 107,111 $ 85,846 Maintenance, Repair & Overhaul 37,972 21,281 Structures and Systems 51,360 44,948 Aircraft Sales and Leasing 3,145 11,698

$ 199,588 $ 163,773

Diluted Earnings Per Share Calculation (In thousands except per share data, Unaudited) Net income as reported $ 5,258 $ 2,286 Add: After-tax interest on convertible debt 306 313 Net income for diluted EPS calculation $ 5,564 $ 2,599 Basic shares outstanding 32,961 32,243 Additional shares due to: Assumed exercise of stock options 475 351 Assumed conversion of convertible debt 3,604 3,604 Diluted shares outstanding 37,040 36,198 Diluted earnings per share $ 0.15 $ 0.07

53

Scoring Sheet

Equal Scoring Subjective Scoring Market Scoring AIR Quarter ended 8/2005 Filing Prelim. Actual Max Actual Score Actual Max Income statement (10 items): Sales 199.588 199.588 1 1 3 3 12 12 Cost of goods sold 158.815 164.906 1 1 2 2 2 2 SG&A 23.901 23.901 1 1 2 2 5 5 Interest expense 4.122 4.122 1 1 1 1 1 1 Special items 0.000 0.000 1 1 2 2 4 4 Depreciation expenses (in C/F statement) 6.091 NA 0 1 0 1 0 1 Non-operating income 0.664 0.664 1 1 2 2 3 3 Income tax expense 2.065 2.065 1 1 1 1 5 5 Net income from continuing operations 5.258 5.258 1 1 3 3 28 28 Extraordinary items 0.000 0.000 0 0 0 0 0 0 Balance sheet (13 items) Cash (and equivalents) 24.411 24.411 1 1 1 1 4 4 Accounts receivable 119.771 N/A 0 1 0 2 0 2 Inventory 277.232 N/A 0 1 0 2 0 3 Current assets 462.526 462.526 1 1 3 3 6 6 Property Plant and Equipment 175.471 N/A 0 1 0 1 0 4 Total assets 757.826 757.826 1 1 3 3 2 2 Short-term debt 2.280 N/A 0 1 0 2 0 2 Accounts payable 94.426 N/A 0 1 0 2 0 1 Current liabilities 166.398 N/A 0 1 0 3 0 6 Long-term debt 228.057 N/A 0 1 0 2 0 2 Other liabilities 22.061 N/A 0 1 0 1 0 2 Total liabilities 437.924 437.924 1 1 2 2 2 2 Stockholders equity, total 319.902 319.902 1 1 3 3 3 3 Cash flow (4 items) Net operating cash flow -21.698 N/A 0 1 0 3 0 0 Capital expenditures 4.695 N/A 0 1 0 3 0 0 Total investing cash flows -4.089 N/A 0 1 0 2 0 0 Total financing cash flows -0.128 N/A 0 1 0 3 0 0 13 26 28 55 77 100 Score 0.500 0.509 0.770