Stock Price Response to New CEO Earnings News · Stock Price Response to New CEO Earnings News 4 recorded by Audit Analytics between 2005 and 2012. Using panel regressions, we compare

Stock Price Response to New CEO Earnings News

Paul G. Geertsema, David H. Lont, Helen Lu *

(This version: 28 August 2015)

Paul G. Geertsema

Department of Accounting and Finance, the University of Auckland, Private Bag 92019, Auckland

1142, New Zealand

e-mail: [email protected]

David H. Lont

Department of Accountancy and Finance, the University of Otago, Dunedin, Otago 9054, New Zealand

e-mail: [email protected]

* Corresponding author: Helen Lu

Department of Accounting and Finance, the University of Auckland, Private Bag 92019, Auckland

1142, New Zealand

e-mail address: [email protected]

This paper has benefited from presentations at the University of Otago, the University of

Auckland, the 5th Quantitative Accounting Research Network Auckland Conference, the

Accounting and Finance Association of Australia and New Zealand (AFAANZ) Conference

and the New Zealand Finance Colloquium. We gratefully acknowledge helpful comments

from Henk Berkman, Ole-Kristian Hope, Mark Humphery-Jenner, Nick Nguyen, Brett Sceats

and Daniel Moser. We are also thankful for funding from AFAANZ. All remaining errors and

omissions are ours.

mailto:[email protected]




2

Abstract

Stock prices on average rise more on good earnings news announced by new

CEOs compared with established CEOs. This new-CEO attributes effect is more

pronounced for CEOs appointed during challenging situations. By contrast, stock

prices tend to drop less on bad earnings news for new CEOs. This honeymoon effect

is robust to negative accruals. The new-CEO attributes effect is stronger for firms

followed by fewer analysts but the honeymoon effect is stronger for firms followed by

more analysts.

JEL classifications: C23, G14, M40

Keywords: CEO Turnover, Unexpected Earnings, Earnings Response Coefficient,

Meet and Beat Premium, Information Environment, Event Studies


3

1. Introduction

CEO change is an important event in a firm’s life. At the announcement of the

appointment of a new CEO, investors do not know whether he or she is a capable CEO, is the

right fit or has a suitable strategy for a firm, as described in a recent article published in the

Harvard Business Review:2

“there is no positive correlation between how a company’s stock fares upon

the announcement of a new CEO and the share price over that CEO’s

tenure…”.

The truth about the quality of a new CEO and the viability of his or her strategy is gradually

revealed over time. If earnings announcement is an important channel through which this

information is revealed, stock prices should react more strongly to earnings news from new

CEOs compared with that from established CEOs. In addition, while good earnings news can

be seen as proof of the new CEO’s ability and the viability of that CEO’s strategy – and

might affect share prices more strongly than other positive earnings news – it is not

necessarily the case with bad earnings news. CEOs’ “honeymoon” periods, as often referred

to by the financial media, can make bad earnings news less informative. If shareholders

believe that new CEOs need time to organise resources and implement their strategies,

implying bad earnings news might be temporary, share prices could under-react to initial bad

earnings news. This study investigates the asymmetric effect on stock prices from new CEO

earnings news.

Our study uses all firm-quarter observations from the Compustat, CRSP and

Institutional Brokers’ Estimate System (I/B/E/S) merged file and CEO turnover events

2 “When Naming a CEO, Ignore the Market Reaction” by James M. Citrin, Harvard Business Review

January–February 2012.


4

recorded by Audit Analytics between 2005 and 2012. Using panel regressions, we compare

responses in stock prices to good earnings news and bad earnings news for new CEOs and

outgoing CEOs against the benchmark group, or established CEOs.

We test the new-CEO attributes hypothesis (H1a), the honeymoon hypothesis (H2a),

the relation between types of CEO turnover and these two effects (H1b and H2b) and the

impact from the information environment on these two effects (H1c and H2c). The new-CEO

attributes hypothesis (H1a) states that stock prices tend to rise more on good earnings news

from new CEOs because this earnings news resolves some uncertainty regarding the CEO’s

attributes – that is, whether he or she has the ability, is the right fit and has a suitable strategy

for the firm. The honeymoon hypothesis (H2a) proposes that stock prices on average drop

less on bad earnings from new CEOs because investors understand that new CEOs need time

to organise resources and implement changes.

First, we find that stock prices only rise slightly more on average by 0.27% (and

marginally significant at a 10% level) on good earnings news for firms with new CEOs

compared with firms with established CEOs (lending weak support to H1a). While we do not

find strong evidence for the new-CEO attributes hypothesis in the full sample of CEO

turnover events, we find that this effect is more pronounced if the CEO is appointed during a

challenging situation (supporting H1b). In this situation, either the information regarding the

new CEOs’ attributes is very uncertain or the true attributes of the new CEOs are crucial for

the firm. Thus, for a CEO appointed during a challenging situation, good earnings news in the

first year of the CEO’s tenure resolves an important source of uncertainty and leads to higher

increases in stock prices compared with good earnings news from established CEOs. We

choose six alternative proxies for CEO turnover in a challenging situation. We posit that a

higher level of uncertainty regarding the new CEO’s attributes exists if he or she is not

appointed as part of a well-planned routine CEO succession plan or if the CEO is hired


5

externally instead of being groomed and promoted internally (Vancil, 1987). Similarly, some

performance and risk measures – including return on assets (ROA), stock return, stock return

volatility or sales growth prior to the CEO turnover – have been shown to predict forced CEO

turnover events (Fredrickson et al., 1988; Hazarika et al., 2012). We use these measures as a

proxy for forced CEO turnover and propose that the attributes of new CEOs are crucial for

companies that have under-performed their industry peers in these important measures. We

find that if new CEOs are appointed during these challenging situations, stock prices tend to

rise more on their subsequent good earnings news compared with established CEOs by 0.49%

to 1.12% and significant at a 10% or 5% level. This difference is economically significant

because it is sizable compared with the average increase of 1.99% in stock prices in response

to good earnings news in our sample (Appendix III). By contrast, good earnings are not

associated with a higher increase in share prices for CEOs appointed during normal

situations.

In the full sample of CEO turnover events, we find some weak evidence for the

honeymoon effect (lending weak support to H2a). When firms announce bad earnings news,

stock prices drop less on average as a reaction for new CEOs by 0.46% (marginally

significant at a 10% level) compared with established CEOs. The honeymoon effect is not

more pronounced for subsamples of CEOs appointed in a challenging situation (rejecting

(H2b), but it is stronger in a better information environment (supporting H2c).

We hypothesize and find that the information environment influences both the new-

CEO attributes effect and the honeymoon effect (H1c and H2c). We expect that earnings

news resolves more uncertainty regarding the new CEO’s attributes as the asymmetry of

information increases. We proxy for the information environment with the level of sell-side

analyst following of firms because analyst coverage is shown to increase with quality of

accounting information, management presentation and institutional investors’ holdings


6

(Bhushan, 1989; Bushman et al., 2005; Bushman and Smith, 2001; Piotroski and Roulstone,

2004; Lang and Lundholm, 1996). As expected, we find that as the level of analyst following

decreases, the new-CEO attributes effect becomes stronger and more significant – stock

prices rise more on the news of meet-and-beat analyst forecasts for new CEOs compared with

established CEOs, by 0.70% and 1.16% and both significant at a 5% level, respectively, if the

firm has a medium- or low-level of analyst following. By contrast, such a new-CEO attributes

effect does not exist if the new-CEO firm has a high level of analyst following. The

information environment also makes a difference in how investors interpret bad earnings

news from new CEOs. In the face of news of missing analyst forecasts from firms with new

CEOs, analysts seem to play a role in communicating the good prospects of a firm with the

investor community. For firms with new CEOs and that are followed by a large number of

analysts, bad earnings news is associated with a smaller drop in share price compared with

firms with established CEOs by 1% and significant at a 1% level. However, this honeymoon

effect does not exist among firms followed by a small number of analysts. The new CEO

honeymoon effect uncovered in our study is not a result of earnings baths in the first year of

CEO tenure because we control for negative discretionary accruals in all regressions.

Our paper contributes to three areas in the literature: information content of earnings

news, CEO turnover and information environment. With regard to the information content of

earnings news, we take the work by Clayton et al. (2005) a step further and show that, in

general, stock prices only rise slightly more on good earnings news for new CEOs than for

established CEOs. Clayton et al. (2005) find that responsiveness in stock prices to earnings

news is significantly stronger for firms with new CEOs compared with firms with outgoing

CEOs. While they attribute the result to earnings announcements that are more informative

for firms with new CEOs, we believe that firms managed by outgoing CEOs are abnormal

and an inappropriate benchmark for analysing the information content of earnings news from


7

new CEOs. Using a complete dataset of all CEO turnover events in public listed companies in

the U.S., we are able to benchmark earnings news from new CEOs to that from established

CEOs. Our finding suggests that earnings announcements may not resolve a significant

amount of uncertainty regarding a new CEO’s attributes in general. However, following

challenging CEO successions, when it is more uncertain whether new CEOs can perform,

stock prices tend to respond more strongly to their subsequent good earnings news compared

with established CEOs. While Ball and Shivakumar (2008) find that earnings announcements

only account for a small fraction of news impounded in stock prices, we show that in the

context of challenging CEO turnover, earnings announcements are an important channel

through which the information regarding the new CEO’s attributes is revealed.

Our paper also contributes to the literature on CEO turnover. ROAs, stock returns,

stock return volatility and sales growth predict forced CEO turnover (Fredrickson et al.,

1988; Hazarika et al., 2012). Firms that experience non-routine CEO change or appoint a new

CEO from outside tend to perform poorly before and after the CEO turnover (Pourciau, 1993;

Parrino, 1997; Murphy and Zimmerman, 1993). We examine stock price response to earnings

news from firms that experience CEO turnover in these difficult times. We find that for CEOs

appointed in difficult times, meet-and-beat analyst forecasts during the first year of their

tenure are seen as a proof of their quality.

Our paper also helps to advance the literature on the information environment. We

show that sell-side analysts play an important role in disseminating information regarding the

new CEO’s quality and the prospect of the firm under the reigns of a new CEO. Research

analyst coverage improves the information environment and valuation for cross-listed firms

(Lang, et al. 2003), increases the dissemination of industry information (Piotroski and

Roulstone, 2004) and reduces insider trading profitability (Frankel and Li, 2004). We show

that research analysts assist the dissemination of information regarding the new CEO’s good


8

attributes. As a result, good earnings news does not have a significantly stronger impact on

stock prices for new CEOs if their firms are covered by a large number of analysts because

their attributes are already known to investors in a transparent informational environment. By

contrast, new-CEO earnings announcements reveal information regarding the new CEO’s

attributes in firms with low- and medium-levels of analyst coverage. We also show that a

good information environment is beneficial in bad times for firms that recently experienced

CEO turnover because stock prices do not tend to drop as much on their bad earnings news

compared with firms with established CEOs.

The remainder of the paper proceeds as follows. In section 2, we describe data and

methodology. Section 3 presents the empirical results, and Section 4 concludes the paper.

2. Data and summary statistics

2.1 Data and sample selection

We identify CEO turnover events using director and officer change filings over the

2005–2012 period, as provided by Audit Analytics. Audit Analytics covers all director and

CEO changes of SEC registrants from January 2005 onwards.3 Panel A in Table 1 describes

the filtering process used to obtain the CEO turnover dataset. Audit Analytics records 12,742

unique CEO turnover events from 2005 to 2012. CEO turnovers due to mergers, acquisitions,

bankruptcies, spin-offs and asset sales were excluded. As many SEC registrants are not

public companies, only 4,227 CEO turnover events can be matched with firm-quarters in the

CRSP and the Compustat Merged file, among which 2,753 firms exist in the I/B/E/S

database. On average, in a given year, 10.7% of firms experience a change in CEO (see Panel

3 An entry is made for the appointment (departure) of each person. When the appointment of a CEO is

matched with the departure of a CEO from the same firm within the previous 90 days, we count the two actions

as one CEO turnover event. When a departure cannot be matched with an appointment, both the departure and

the appointment are accounted for as CEO turnover events. We reviewed the filings related to 20 randomly

sampled unmatched CEO appointments and 20 randomly sampled unmatched CEO departures. Most unmatched

appointments and unmatched departures are genuine CEO turnover events.


9

B of Table 1), implying an average CEO tenure of approximately nine years in the CRSP,

Compustat, I/B/E/S universe of firms from 2005 to 2012.4 Moving to CEO changes by

industry (Panel C of Table 1), firms in the finance, insurance and real estate industry group

have the lowest CEO turnover ratio of 6.8% in the CRSP, Compustat, I/B/E/S space, or

equivalently the longest implied CEO tenure of about 15 years. The highly competitive retail

industry exhibits the highest CEO turnover rate of 14.9%, implying an average CEO tenure of

approximately seven years.

[Insert Table 1 about here.]

The CEO turnover events enable us to create CEO-change dummy variables. We

define new-CEO firm-quarters as firm-quarters where the CEO has been at the helm for no

more than 365 calendar days on the earnings announcement date (referred to as the cut-off

date). Outgoing-CEO firm-quarters are quarters where the CEO leaves the firm within 365

calendar days from the cut-off date. Established-CEO firm-quarters are those quarters that are

neither new-CEO firm-quarters nor outgoing-CEO firm-quarters. Some firm-quarters

between two consecutive CEO changes could be classified as either outgoing-CEO firm-

quarters or new-CEO firm-quarters; our study excludes these ambiguous firm-quarters.

Our study proxies for the expected earnings per share (or, EPS) with the I/B/E/S

consensus forecast (or, the median forecast) and measures the size of earnings news (or,

unexpected earnings) as the deviation in actual earnings per share (or, EPS) from the

expected EPS, scaled by the closing stock price five days before the announcement (as in

DellaVigna and Pollet, 2009). In the context of CEO turnover, using expected EPS from a

time-series model is noisier and less appropriate because a new CEO is often hired to

improve performance (Parrino, 1997). Thus, the change of CEO introduces a break in the

4 As a point of comparison, Bushman et al. (2010) use ExecuComp, which covers S&P 1000 large

companies, and the average CEO tenure of turnover firms is approximately10 years.


10

time-series of earnings, and using earnings during the outgoing CEO period to estimate the

expected earnings for the period managed by new CEOs is inappropriate. Each analyst’s

forecast EPS is adjusted for any share splits and other corporate actions between the forecast

date and the earnings announcement date.

We measure the abnormal stock returns during the three-day window (day-1 to +1)

around the earnings announcement as the accumulated difference between the actual daily

stock return and the normal return from a portfolio matched on size and book-to-market ratio.

We construct size and book-to-market portfolios using the method in Fama and French,

(1992). Specifically, at the end of June in each year, we classify all stocks (with a share code

of 10 and 11, or US common equities) into 25 portfolios by size at the end of June of the

current year and by book-to-market ratio at the end of December of the previous year.

Portfolio returns are equal-weighted returns. As suggested by Daniel et al. (1997) and Barber

and Lyon (1997), this approach generate less biased results than traditional approaches that

use the market model or other factor pricing models to estimate the normal return. 5 The

cumulative abnormal return over the three-day period surrounding the earnings

announcement date6 is the sum of abnormal return over the three event days.

2.2 Descriptive statistics

Table 2 shows that the means of important variables differ significantly between

CEO-change periods and established-CEO periods. During the four quarters prior to CEO

5 Our results are robust to different ways to estimate the abnormal return. In our robustness tests, we

estimate normal return with a market model where we regress a stock’s return on the return to a value-weighted

(or an equal-weighted) portfolio for the window covering 250 trading days up to 10 days before the earnings

announcement date ([-260, -10]). Results are available on request. 6 Following DellaVigna and Pollet (2009), we use the earnings announcement date (rdq) from Compustat

and cross-check using the earnings announcement date (actanndats) from I/B/E/S when possible. Then rdq and

actanndats do not agree, we use the earlier date as the earnings announcement date. If I/B/E/S records that the

earnings are announced after the market closes, the earnings announcement date is moved to the following trading

day. We also eliminate earnings announcements that are potentially contaminated by requiring that the first trading

day after the earnings announcement occur less than seven days after the earnings announcement.


11

turnover and the four quarters after CEO turnover, firms record significantly lower average

cumulative abnormal returns (CARs) in the three days around earnings announcements

(𝐶𝐴𝑅(−1,1)), a higher percentage of earnings misses (BAD) and worse earnings news (or

unexpected earnings, UE) than during firm quarters with established CEOs. For example,

during outgoing- and new-CEO firm-quarters, cumulative abnormal returns average 1.17%

and 0.30% less than those during established-CEO firm-quarters and significant at a 1% level

(column OUT-EST and NEW-EST) – these more negative CARs correspond to worse

earnings news (UE) during these periods with outgoing CEOs and new CEOs. The

percentage of firms that miss the consensus forecast (BAD) is 40% in the four quarters prior

to CEO turnovers and 38% in the four quarters following CEO turnovers, significantly higher

than the 36% during established-CEO firm-quarters. The unexpected earnings are worse for

outgoing CEOs and for new CEOs by −0.13 and −0.08, respectively, and significant at a

1 % level compared with established CEOs.


The change of CEO introduces more uncertainty. Following CEO turnover, firms

tend to deliver more extreme earnings news – positive earnings news (UE×GOOD) averages

0.07 cents per dollar price higher and significant at a 1% level for new CEOs compared with

outgoing CEOs, while negative earnings news (UE×BAD) averages 0.08 cents per dollar

price lower and significant at a 10% level for new CEOs compared with outgoing CEOs.

Earnings also become more volatile and less persistent after the CEO change than before the

CEO change. Earnings volatility (EVOL) increase by 0.03 and earnings persistence (PERS)

further decrease by 0.02, respectively, when moving from outgoing- to new-CEO firm-

quarters.

Our analysis controls for important firm characteristics that affect the relation

between stock returns and earnings news, including accounting loss (LOSS) (Hayn, 1995),


12

earnings volatility (EVOL) (Chambers et al., 2005; Collins and Kothari,1989), growth

opportunities (MB and SIZE) (Kormendi and Lipe, 1987; Skinner and Sloan, 2002), earnings

baths (NEGDA) and downward earnings guidance (GDOWN) (Das et al., 2011). Table 2 also

shows that these important variables are very different during the CEO-change periods,

illustrating the importance for controlling for these variables in all regressions. CEO-change

firms tend to have significantly smaller sizes (SIZE), lower book-to-market ratios (MB),

higher earnings volatility (EVOL), be more likely to be loss-making (LOSS), be more likely

to record earnings-decreasing accruals (NEGDA)7, give more downward guidance (Das et al.,

2011) and have less persistent earnings (PERS).

3. Main results

3.1 The new-CEO attributes effect and honeymoon effect for the full sample

When measuring the response in stock returns to earnings news, we consider both an

intercept effect and the asymmetric slope effects. The intercept effect is captured by an

indicator variable for earnings miss (BAD) to account for the negative average stock returns

in response to the news of missing analyst forecasts (Bartov et al., 2002). Slope effects assess

the response in stock returns to a unit of positive earnings news (GOOD×UE) and that to a

unit of negative earnings news (BAD×UE) separately to account for an asymmetric earnings

response coefficient (or ERC, which measures the change in stock prices in response to each

unit of earnings news) (Hayn, 1995; Basu, 1997). We test the new-CEO attributes hypothesis

(H1a) and the honeymoon hypothesis (H2a) on both the intercept and the slope, by

estimating the following regression model specified in equation (1):

7 We estimate discretionary accruals with the modified Jone’s model in the cross-section by industry

(defined as firms with the same two-digit SIC head) and quarter, as shown in Appendix II.


13

𝐶𝐴𝑅(−1,1) = 𝛽0 + 𝛽1 × 𝐺𝑂𝑂𝐷 × 𝑂𝑈𝑇 + 𝛽2 × 𝐺𝑂𝑂𝐷 × 𝑁𝐸𝑊

+𝛽3 × 𝐵𝐴𝐷 + 𝛽4 × 𝐵𝐴𝐷 × 𝑂𝑈𝑇 + 𝛽5 × 𝐵𝐴𝐷 × 𝑁𝐸𝑊

+𝛽6 × 𝐺𝑂𝑂𝐷 × 𝑈𝐸 + 𝛽7 × 𝐺𝑂𝑂𝐷 × 𝑈𝐸 × 𝑂𝑈𝑇 + 𝛽8 × 𝐺𝑂𝑂𝐷 × 𝑈𝐸 × 𝑁𝐸𝑊

+𝛽9 × 𝐵𝐴𝐷 × 𝑈𝐸 + 𝛽10 × 𝐵𝐴𝐷 × 𝑈𝐸 × 𝑂𝑈𝑇 + 𝛽11 × 𝐵𝐴𝐷 × 𝑈𝐸 × 𝑁𝐸𝑊

+𝜽 × 𝑵𝑶𝑵𝑳𝑰𝑵𝑬𝑨𝑹 + 𝝁 × 𝑪𝑶𝑵𝑻𝑹𝑶𝑳 + 𝝉 × 𝒀𝒆𝒂𝒓𝒒𝒖𝒂𝒓𝒕𝒆𝒓 + 𝜎 (1)

The new-CEO attributes hypothesis predicts a larger increase in the stock price on the news

of meet-and-beat for new CEOs (H1a: 𝛽2 > 0) or a stronger response to each unit of good

news for new CEOs (H1a: 𝛽8 > 0). The honeymoon hypothesis suggests a smaller drop in

the stock price on the news of missing analyst forecasts for new CEOs (H2a:𝛽5 > 0) or a

weaker response to each unit of bad news (H2a:𝛽11 < 0). All regression analysis in this

paper controls for the non-linear reactions to earnings news (𝑵𝑶𝑵𝑳𝑰𝑵𝑬𝑨𝑹) (Chambers et

al., 2005; Das and Lev, 1994) by including an interaction term between UE and |UE|. We also

control for all well-known factors that affect stock returns’ sensitivity to earnings news. The

vector of control variables (𝑪𝑶𝑵𝑻𝑹𝑶𝑳) includes SIZE, MB, EVOL, LOSS, GDOWN,

NEGDA, as well as interaction terms between each of these control variables and BAD

because these variables affect stock prices differently depending on whether the earnings

news is good or bad (Skinner and Sloan, 2002). Still, some firm characteristics and the time

effect are not captured in our model; thus all our regressions include a firm-fixed effect and

year-quarter dummies. 𝑡 −tests for significance in all regressions are clustered by year-

quarter because many firms announce earnings in a quarter, and returns in these announcing

firms are cross-correlated (Petersen, 2009). Appendix I contains a description of all variables.


Table 3 reports the correlation coefficient among the three-day cumulative abnormal

returns around the earnings announcement date (CAR(-1,1) and important explanatory

variables in our analysis. The explanatory variables are not highly correlated, with correlation


14

coefficients ranging between −0.48 (between UE(t) and BAD(t) to 0.26 (between SIZE(t-1)

and MB(t-1)). These low correlation coefficients suggest that multicollinearity is not a

concern. These correlation coefficients also have expected signs. For example, the correlation

coefficient between SIZE(t-1) and EVOL(t-1) is −0.25, suggesting that the earnings of small

companies tend to be more volatile.


Column (1) in Table 4 reports the main regression results using equation (1). We

find weak evidence for the new-CEO attribute hypothesis. The estimate for 𝛽2 (or the

coefficient on GOOD×NEW) is 0.27 and marginally significant at 10%, suggesting that

stock prices only rise slightly more on the news of meet-and-beat analyst forecasts from new

CEOs compared with established CEOs (lending weak support to H1a). By contrast, stock

prices do not tend to react differently to each unit of good earnings news because the

estimated for 𝛽8 (GOOD×UE×NEW) is insignificant. Estimates for coefficients on control

variables all have expected signs. Adjusted R-squareds in regression (1) is 14.7%, similar to

those in comparable studies (for example, Table 4 of Chambers et al., 2005).

At first sight, this finding is surprising because it appears to contradict Clayton et al.

(2005), who find a significant and strong average reaction in stock prices to earnings

announcements for new CEOs more than for their predecessors. While they attribute this

result to new CEOs and interpret it as an evidence that new CEOs’ earnings announcements

contain information regarding their ability and strategy, it is also possible that investors

interpret outgoing CEOs’ earnings announcements differently. We find that this is indeed the

case. To explain why our findings differ from Clayton et al. (2005), we follow their approach

and regress CARs on unexpected earnings using the following regression:


15

𝐶𝐴𝑅(−1,1)

= 𝛼0 + 𝛼1 × 𝐺𝑂𝑂𝐷 × 𝑈𝐸 + 𝛼2 × 𝐺𝑂𝑂𝐷 × 𝑈𝐸 × 𝑂𝑈𝑇 + 𝛼3 × 𝐺𝑂𝑂𝐷 × 𝑈𝐸 × 𝑁𝐸𝑊

+𝛼4 × 𝐵𝐴𝐷 × 𝑈𝐸 + 𝛼5 × 𝐵𝐴𝐷 × 𝑈𝐸 × 𝑂𝑈𝑇 + 𝛼6 × 𝐵𝐴𝐷 × 𝑈𝐸 × 𝑁𝐸𝑊


Column (2) in Table 4 presents the regression results. Clayton et al. (2005) study CEO-

change firms only and compare ERCs for new CEOs with those for outgoing CEOs. By

contrast, benefiting from the complete dataset of CEO turnover events in the U.S. from Audit

Analytics, we study all firm-quarters and use firms with established CEOs as the benchmark

group. Like Clayton et al. (2005), we find that the ERC to good earnings news is stronger for

new CEOs compared with outgoing CEOs by 1.20 and significant at a 5% level

(GOOD×UE×NEW−GOOD×UE×OUT). However, this difference is almost entirely due to

the outgoing-CEO effect – for outgoing CEOs, the ERC on good earnings news is weaker by

1.20 and significant at a 5% level compared with established CEOs (GOOD×UE×OUT),

while the ERC for new CEOs is not significantly different from the benchmark group. The

estimate for 𝛼3 (or the coefficient on GOOD×UE×NEW) is −0.01 and insignificant,

suggesting that stock prices do not tend to rise more on good earnings news from new CEOs

compared with established CEOs (rejecting H1a). Both Clayton et al. (2005) and our study

find that good earnings news from new CEOs contains more information than that from

outgoing CEOs. While Clayton et al. (2005) attribute the difference to the possible rich

information in earnings news from new CEOs, we show that the difference should be

attributed to the lack of information content in the earnings news from outgoing CEOs.

We also find weak evidence for the honeymoon hypothesis (H2a). In column (1) of

Table 4, the estimate for 𝛽5 (or the coefficient on BAD×NEW) is 0.46 and marginally

significant at 10%, suggesting that stock prices tend to drop a little less on the news of

missing analyst forecasts from new CEOs compared with established CEOs (lending weak

support to the honeymoon hypothesis, H2a). By contrast, stock prices do not tend to react


16

differently to each unit of bad earnings news because the estimate for 𝛽11 (BAD×UE×NEW)

is insignificant. New CEOs tend to record earnings decreasing accruals (Pourciau, 1993;

Wilson and Wang , 2010; Choi et al., 2014), and therefore missing analyst forecasts may be

a result of new CEOs giving earnings a “bath”. If so, it is possible that the new CEO

honeymoon effect (H2a) is simply an artefact of muted stock price responses to missing

analyst forecasts resulting from earnings baths. We show that the honeymoon effect

uncovered in this paper is robust to the new-CEO earnings bath. To address this concern, all

our regressions include a control variable for negative discretionary accruals (NEGDA) and

an interaction term between earnings miss and negative discretionary accruals

(BAD×NEGDA). If the firm records a negative discretionary accrual, the three-day abnormal

returns surrounding earnings announcement dates are 0.95% higher on average.8 This

suggests that investors regard negative discretionary accruals as less value-relevant and stock

prices on average rise more on the news of meet-and-beat if the firm has recorded negative

discretionary accruals.

3.2 CEOs appointed in challenging times

CEO successions can be non-eventful or challenging. We hypothesize that a smooth

CEO succession introduces little uncertainty, but that a challenging CEO turnover brings in

new sources of uncertainty. Thus, we expect that, following a challenging CEO turnover but

not a normal CEO turnover, stock prices respond more strongly to good earnings news for

new CEOs compared with established CEOs. This hypothesis is the new-CEO attributes

hypothesis conditioned on types of CEO turnover (H1b). Following these challenging CEO

turnover events, investors in these companies may anticipate bad earnings news and stock

prices drop less on subsequent bad earnings news compared with news from firms with

8 This result is not tabulated.


17

established CEOs. Thus, we hypothesize that the honeymoon effect is stronger for new CEOs

who take the helm in a challenging situation than for those appointed in a normal situation.

This is the honeymoon hypothesis conditioned on types of CEO turnover (H2b).

Following the management and finance literature, we choose six alternative proxies

for a challenging executive turnover, including non-routine CEO turnover, external

succession, low ROAs, subpar stock return, excessive stock return volatility and slow sales

growth. CEO successions can be a non-eventful and smooth “relay” process in which the

CEO title is passed to an selected heir when the incumbent retires according to plan (Vancil,

1987). Following Pourciau (1993), we refer to this type of CEO turnover as routine CEO

succession. Vancil (1987) also lauds many cases of inside CEO successions (or internal

successions) as successful processes in which internal candidates are groomed and tested. We

conjuncture that both routine CEO turnover and internal successions tend to be smooth CEO

succession processes, so that the change of CEO does not bring in a significant amount of

uncertainty. By contrast, both non-routine CEO successions and external successions

introduce a new source of uncertainty, and stock prices respond more strongly to good

earnings news after these two types of CEO successions. We use information in Audit

Analytics to classify a CEO turnover as routine/non-routine and internal/external. Appendix

IV describes the criteria for and procedure in classifying CEO turnover events into different

types. In addition to the information contained in 8-K filings from Audit Analytics, we also

consider four performance variables during the 12-month period (or, rolilng four-quarter

period) leading up to the CEO departure as proxies for whether the departing CEO is forced

out due to poor performance (Fredrickson et al., 1988; Hazarika et al., 2012). During the 12-

month period prior to the CEO turnover, if the firm’s ROA falls in the lowest one-third

among its industry peers (or companies sharing the same two-digit SIC head), we assume that

the CEO turnover takes place in a challenging situation. Otherwise, the CEO turnover is a


18

normal turnover. Similarly, during this 12-month period leading up to the CEO departure, if

the firm’s stock returns, stock return volaility and sales growth fall in the worst one-third

within an industry, the CEO turnover is classified as challenging.

To test whether the new-CEO attributes effect and the honeymoon effect are more

pronounced for new CEOs appointed during a challenging situation (H1b and H2b), all terms

that contain OUT (or NEW) in equation (1) are further interacted with a CHAL and a NORM

separately in equation (3):

𝐶𝐴𝑅(−1,1) = 𝛾0 + 𝛾1 × 𝐺𝑂𝑂𝐷 × 𝑂𝑈𝑇 × 𝑁𝑂𝑅𝑀 + 𝛾2 × 𝐺𝑂𝑂𝐷 × 𝑂𝑈𝑇 × 𝐶𝐻𝐴𝐿

+𝛾3 × 𝐺𝑂𝑂𝐷 × 𝑁𝐸𝑊 × 𝑁𝑂𝑅𝑀 + 𝛾4 × 𝐺𝑂𝑂𝐷 × 𝑁𝐸𝑊 × 𝐶𝐻𝐴𝐿

+𝛾5 × 𝐵𝐴𝐷

+𝛾6 × 𝐵𝐴𝐷 × 𝑂𝑈𝑇 × 𝑁𝑂𝑅𝑀 + 𝛾7 × 𝐵𝐴𝐷 × 𝑂𝑈𝑇 × 𝐶𝐻𝐴𝐿

+𝛾8 × 𝐵𝐴𝐷 × 𝑁𝐸𝑊 × 𝑁𝑂𝑅𝑀 + 𝛾9 × 𝐵𝐴𝐷 × 𝑁𝐸𝑊 × 𝐶𝐻𝐴𝐿

+𝛾10 × 𝐺𝑂𝑂𝐷 × 𝑈𝐸

+𝛾11 × 𝐺𝑂𝑂𝐷 × 𝑈𝐸 × 𝑂𝑈𝑇 × 𝑁𝑂𝑅𝑀 + 𝛾12 × 𝐺𝑂𝑂𝐷 × 𝑈𝐸 × 𝑂𝑈𝑇 × 𝐶𝐻𝐴𝐿

+𝛾13 × 𝐺𝑂𝑂𝐷 × 𝑈𝐸 × 𝑁𝐸𝑊 × 𝑁𝑂𝑅𝑀 + 𝛾14 × 𝐺𝑂𝑂𝐷 × 𝑈𝐸 × 𝑁𝐸𝑊 × 𝐶𝐻𝐴𝐿

+𝛾15 × 𝐵𝐴𝐷 × 𝑈𝐸

+𝛾16 × 𝐵𝐴𝐷 × 𝑈𝐸 × 𝑂𝑈𝑇 × 𝑁𝑂𝑅𝑀 + 𝛾17 × 𝐵𝐴𝐷 × 𝑈𝐸 × 𝑂𝑈𝑇 × 𝐶𝐻𝐴𝐿

+𝛾18 × 𝐵𝐴𝐷 × 𝑈𝐸 × 𝑁𝐸𝑊 × 𝑁𝑂𝑅𝑀 + 𝛾19 × 𝐵𝐴𝐷 × 𝑈𝐸 × 𝑁𝐸𝑊 × 𝐶𝐻𝐴𝐿

+𝜽 × 𝑵𝑶𝑵𝑳𝑰𝑵𝑬𝑨𝑹 + 𝝆 × 𝑪𝑶𝑵𝑻𝑹𝑶𝑳 + 𝝉 × 𝒀𝒆𝒂𝒓𝒒𝒖𝒂𝒓𝒕𝒆𝒓 + 𝜎 (3)

The new-CEO attributes hypothesis conditioned on turnover type (H1b) suggests that,

following challenging CEO turnovers, stock prices tend to rise more on the news of meet-

and- beat (𝛾4 on 𝐺𝑂𝑂𝐷 × 𝑁𝐸𝑊 × 𝐶𝐻𝐴𝐿) or the ERC to good earnings news is stronger (𝛾14

on 𝐺𝑂𝑂𝐷 × 𝑈𝐸 × 𝑁𝐸𝑊 × 𝐶𝐻𝐴𝐿) than during periods with established CEOs. The

honeymoon hypothesis conditioned on turnover type (H2b) suggests that, following

challenging CEO turnonvers, on average stock prices drop less on news of missing analyst

forecasts (𝛾9 on 𝐵𝐴𝐷 × 𝑁𝐸𝑊 × 𝐶𝐻𝐴𝐿) or react more weakly to each unit of bad earnings

news (𝛾19 on 𝐵𝐴𝐷 × 𝑈𝐸 × 𝑁𝐸𝑊 × 𝐶𝐻𝐴𝐿) than during periods with established CEOs.


19

Table 5 reports the regression results.

[insert Table 5 about here]

Results in Table 5 show that stock prices tend to rise more on good earnings news

for new CEOs appointed during challenging times, but not for those appointed during normal

times (supporting H1b). Loadings on 𝐺𝑂𝑂𝐷 × 𝑁𝐸𝑊 × 𝐶𝐻𝐴𝐿 (𝛾4) are positive for all six

alternative proxies for challenging turnovers, with four coefficients significant at a 5% level

and one at a 10% level. On average, stock prices tend to rise more on good earnings news

following these challenging CEO turnovers by 0.49% to 1.12%. The difference in CARs is

not only statistically significant, but also economically significant because they are sizable

compared with the average CAR of 1.99% to all good earnings news in our full sample (see

CAR(-1,1)×GOOD in Appendix III). By contrast, we do not see evidence for the honeymoon

effect conditioned on types of turnover (thus rejecting H2b). Investors do not appear to

consider the nature of CEO turnover or be lenient towards bad earnings news if the firm has

to replace its CEO in times of great uncertainty.

3.1 Information environment

In a highly asymmetric informational environment, earnings news can resolve a

significant amount of uncertainty regarding the new CEO’s attributes. Thus, we hypothesize

that the new-CEO attributes effect should increase in the asymmetry of the firm’s

informational environment (H1c). By contrast, if a firm that has just replaced its CEO and

missed an analyst forecast, analysts can help investors focus on the long-term prospects of the

firm. Thus, we expect the honeymoon effect to be stronger among firms with more analyst

coverage (H2c). Sell-side analysts act as important information intermediaries in the capital

markets. The existing literature shows that a high level of analyst following is associated with

a better information environment with less information asymmetry (Bushman and Smith,


20

2001; Lang et al. 2003; Bushman et al., 2005; Piotroski and Roulstone, 2004). Thus, we use

the number of analysts as a proxy for information environment and investigate whether the

information environment has an impact on both the new-CEO attributes effect and the

honeymoon effect. In addition to the explanatory variables in equation (1), we include the

number of analysts (ANALYST), the interaction term between good earnings news from new

CEOs and number of analysts and the interaction term between bad earnings news from new

CEOs to test H1c and H2c, using equation (4):

𝐶𝐴𝑅(−1,1) = 𝜃0 + 𝜃1 × 𝐴𝑁𝐴𝐿𝑌𝑆𝑇 + 𝜃2 × 𝐺𝑂𝑂𝐷 × 𝑂𝑈𝑇

+𝜃3 × 𝐺𝑂𝑂𝐷 × 𝑁𝐸𝑊 + 𝜃4 × 𝐺𝑂𝑂𝐷 × 𝑁𝐸𝑊 × 𝐴𝑁𝐴𝐿𝑌𝑆𝑇

+𝜃5 × 𝐵𝐴𝐷 + 𝜃6 × 𝐵𝐴𝐷 × 𝑂𝑈𝑇

+𝜃7 × 𝐵𝐴𝐷 × 𝑁𝐸𝑊 + 𝜃8 × 𝐵𝐴𝐷 × 𝑁𝐸𝑊 × 𝐴𝑁𝐴𝐿𝑌𝑆𝑇

+𝜃9 × 𝐺𝑂𝑂𝐷 × 𝑈𝐸 + 𝜃10 × 𝐺𝑂𝑂𝐷 × 𝑈𝐸 × 𝑂𝑈𝑇 + 𝜃11 × 𝐺𝑂𝑂𝐷 × 𝑈𝐸 × 𝑁𝐸𝑊

+𝜃12 × 𝐵𝐴𝐷 × 𝑈𝐸 + 𝜃13 × 𝐵𝐴𝐷 × 𝑈𝐸 × 𝑂𝑈𝑇 + 𝜃14 × 𝐵𝐴𝐷 × 𝑈𝐸 × 𝑁𝐸𝑊

+𝝎 × 𝑵𝑶𝑵𝑳𝑰𝑵𝑬𝑨𝑹 + 𝝉 × 𝒀𝒆𝒂𝒓𝒒𝒖𝒂𝒓𝒕𝒆𝒓 + 𝝆 × 𝑪𝑶𝑵𝑻𝑹𝑶𝑳 + 𝜎 (4)

Column (1) in Table 6 presents the regression results. Because H1c states that the

new-CEO attributes effect becomes stronger if the firm managed by a new CEO is followed

by fewer analysts, we expect and find the estimate on 𝐺𝑂𝑂𝐷 × 𝑁𝐸𝑊 × 𝐴𝑁𝐴𝐿𝑌𝑆𝑇 (𝜃4) to be

negative and significant. If the firm managed by a new CEO does not have any analyst

following, stock prices on average rise more on its news of meet-and-beat, by 0.81% and

significant at a 5% level, compared with established CEOs. If the firm is followed by one

more analyst, this difference in positive stock returns decreases by 0.07% and is significant at

a 5% level. We expect and find the estimate on 𝐵𝐴𝐷 × 𝑁𝐸𝑊 × 𝐴𝑁𝐴𝐿𝑌𝑆𝑇 (𝜃8) to be positive

and significant because H2c states that the honeymoon effect becomes stronger if the firm

managed by a new CEO is followed by more analysts. If a firm managed by a new CEO is

covered by one more analyst, the stock price tends to drop less on its news of missing analyst

forecasts by 0.09% and the difference is significant at a 10% level.


21

[insert Table 6 about here]

The number of analysts following a firm does not vary dramatically among firms –

50% of firms are followed by one to four analysts and only 10% of firms are followed by

more than 10 analysts. Thus, in order to make sure our results are not driven by extreme

observations, we use three levels of analyst coverage instead of the number of analysts

providing coverage. Specifically, at the end of each calendar quarter, we sort firms on the

number of analysts into three terciles. We run a regression as specified in equation (5), which

is an extension of equation (1). In equation (5), we interact new-CEO meet-and-beat

(𝐺𝑂𝑂𝐷 × 𝑁𝐸𝑊 ) and new-CEO miss (𝐵𝐴𝐷 × 𝑁𝐸𝑊) with each of the three indicator

variables for the level of analyst coverage (LOW, MED and HIGH) to test the information

environment hypotheses (H1c and H2c):

𝐶𝐴𝑅(−1,1)

= 𝛿0 + 𝛿1 × 𝑀𝐸𝐷 + 𝛿2 × 𝐻𝐼𝐺𝐻 + 𝛿3 × 𝐺𝑂𝑂𝐷 × 𝑂𝑈𝑇

+𝛿4 × 𝐺𝑂𝑂𝐷 × 𝑁𝐸𝑊 × 𝐿𝑂𝑊 + 𝛿5 × 𝐺𝑂𝑂𝐷 × 𝑁𝐸𝑊 × 𝑀𝐸𝐷 + 𝛿6 × 𝐺𝑂𝑂𝐷 × 𝑁𝐸𝑊 × 𝐻𝐼𝐺𝐻

+𝛿7 × 𝐵𝐴𝐷 + 𝛿8 × 𝐵𝐴𝐷 × 𝑂𝑈𝑇

+𝛿9 × 𝐵𝐴𝐷 × 𝑁𝐸𝑊 × 𝐿𝑂𝑊 + 𝛿10 × 𝐵𝐴𝐷 × 𝑁𝐸𝑊 × 𝑀𝐸𝐷 + 𝛿11 × 𝐵𝐴𝐷 × 𝑁𝐸𝑊 × 𝐻𝐼𝐺𝐻

+𝛿12 × 𝐺𝑂𝑂𝐷 × 𝑈𝐸 + 𝛿13 × 𝐺𝑂𝑂𝐷 × 𝑈𝐸 × 𝑂𝑈𝑇 + 𝛿14 × 𝐺𝑂𝑂𝐷 × 𝑈𝐸 × 𝑁𝐸𝑊

+𝛿15 × 𝐵𝐴𝐷 × 𝑈𝐸 + 𝛿16 × 𝐵𝐴𝐷 × 𝑈𝐸 × 𝑂𝑈𝑇 + 𝛿17 × 𝐵𝐴𝐷 × 𝑈𝐸 × 𝑁𝐸𝑊


Column (2) in Table 6 presents the regression results for equation (5). We find support for

both the information environment hypothesis for new-CEO attributes (H1c) and for

honeymoon effect (H2c). If firms that recently experienced CEO turnover have a high level

of analysts following, stock prices do not react more strongly to good earnings news for new

CEOs as compared with established CEOs because the estimate for 𝛿6 on 𝐺𝑂𝑂𝐷 × 𝑁𝐸𝑊 ×

𝐻𝐼𝐺𝐻 is not significantly different from zero. As the level of analyst following decreases, the

new-CEO attributes effect turns significant – stock prices rise more on good earnings news

for new CEOs compared with established CEOs, by 0.78% and 1.13%, respectively, both


22

significant at a 5% level (see slope estimates for GOOD×NEW×MED and

GOOD×NEW×LOW). For firms with new CEOs and followed by a large number of analysts,

news of missing forecasts is associated with a smaller drop in share price compared with

firms with established CEOs by 0.95% and significant at a 5% level (𝛿10 on 𝐵𝐴𝐷 × 𝑁𝐸𝑊 ×

𝐻𝐼𝐺𝐻). This result suggests that the stock market is lenient towards bad earnings news from

these CEO-change firms with a large number of analysts following, perhaps because analysts

play an important role in helping investors focus on the long-term prospect of the firm. This

“honeymoon” effect does not exist among firms followed by a small number of analysts

because the slope estimates on 𝐵𝐴𝐷 × 𝑁𝐸𝑊 × 𝐿𝑂𝑊 and 𝐵𝐴𝐷 × 𝑁𝐸𝑊 × 𝑀𝐸𝐷 are

insignificant and small (−0.20 and 0.39).

4. Conclusion

The first few earnings announcements following high-profile CEO successions tend

to attract a great deal of attention from the financial media. It is possible that earnings news

reveals important information regarding the new CEO’s quality, as suggested by Clayton et

al. (2005). Clayton et al. (2005) find that response in stock price to earnings news is stronger

for new CEOs compared with outgoing CEOs and attribute this result to the richer

information content in new CEOs’ earnings news. We show that the period preceding CEO

turnover is an inappropriate benchmark group because it is an abnormal period – stock prices

tend to rise less on good earnings news and drop more on bad earnings news for outgoing

CEOs compared with established CEOs. We find that the difference in ERCs between periods

with new CEOs and those with outgoing CEOs is driven by a weaker response in stock prices

to earnings news from outgoing CEOs, not a stronger response to that from new CEOs. When

analysing the information content of earnings news during the period following CEO

turnover, firms with established CEOs are the appropriate benchmark group. Our study uses

this benchmark group and finds an asymmetric response in stock prices to earnings news


23

from new CEOs. Stock prices tend to rise more on the news of meet-and-beat from firms with

new CEOs, and this new-CEO attributes effect is more pronounced if the CEO is appointed

during a challenging situation. Interestingly, as the information environment becomes more

asymmetric (or as the firm’s analyst following decreases), stock prices rise more on the news

of meet-and-beat from new CEOs compared with established CEOs. This finding suggests

that analyst coverage acts as a substitute for earnings announcements in spreading out good

news regarding the new CEO’s quality across the investor community. However, they do not

spread out the bad earnings news during CEO successions. Instead, firms with more analyst

coverage seem to benefit from their transparent information environment if they miss analyst

forecasts during a CEO succession. If the CEO-change firm has a large number of analysts

following, its stock price tends to drop less on the news of missing analyst forecasts

compared with firms with established CEOs.


24

References

Ball, R., Shivakumar, L., 2008. How much new information is there in earnings? Journal of

Accounting Research 46 (5), 975-1016.

Barber, B. M., Lyon, J.D., 1997. Detecting long-run abnormal stock returns: The empirical

power and specification of test statistics. Journal of Financial Economics 43 (3), 341-

372.

Bartov, E., Givoly, D., Hayn, C., 2002. The rewards to meeting or beating earnings

expectations. Journal of Aand Economics 33 (2), 173-204.

Basu, S., 1997. The conservatism principle and the asymmetric timeliness of earnings'

surprises. Journal of Accounting and Economics 24 (1), 3-37.

Bhushan, R. 1989. Firm characteristics and analyst following. Journal of Accounting and

Economics 11 (2), 255-274.

Bushman, R.M., Piotroski, J.D., Smith, A.J., 2005. Insider trading restrictions and analysts'

incentives to follow firms. The Journal of Finance 60 (1), 35-66.

Bushman, R.M., Smith, A.J., 2001. Financial accounting information and corporate

governance. Journal of Accounting and Economics 32 (1), 237-333.

Chambers, D.J., Freeman, R.N., Koch, A.S., 2005. The effect of risk on price responses to

unexpected earnings. Journal of Accounting, Auditing & Finance 20 (4), 461-482.

Choi, J.-S., Kwak,Y.-M., Choe, C., 2014. Earnings Management Surrounding CEO Turnover:

Evidence from Korea. Abacus 50 (1), 25-55.

Clayton, M.C., Hartzell, J.C., Rosenberg, J., 2005. The Impact of CEO turnover on equity

volatility. The Journal of Business 78 (5), 1779-1808.

Collins, D.W., Kothari, S.,1989. An analysis of intertemporal and cross-sectional

determinants of earnings response coefficients. Journal of Accounting and Economics

11 (2), 143-181.

Daniel, K., Grinblatt, M., Titman, S., Wermers, R., 1997. Measuring mutual fund

performance with characteristic‐based benchmarks. The Journal of Finance 52 (3),

1035-1058.

Das, S., Kim, K., Patro, S., 2011. An analysis of managerial use and market consequences of

earnings management and expectation management. The Accounting Review 86 (6),

1935-1967.

Das, S., Lev, B., 1994. Nonlinearity in the Returns‐Earnings Relation: Tests of alternative

specifications and explanations. Contemporary Accounting Research 11 (1), 353-379.

DellaVigna, S., Pollet, J.M., 2009. Investor inattention and Friday earnings announcements.

The Journal of Finance 64 (2), 709-749.


25

Fama, E.F., French, K.R., 1992. The cross‐section of expected stock returns. The Journal of

Finance 47 (2), 427-465.

Frankel, R., Li, X., 2004. Characteristics of a firm's information environment and the

information asymmetry between insiders and outsiders. Journal of Accounting and

Economics 37 (2), 229-259.

Fredrickson, J.W., Hambrick, D.C., Baumrin, S., 1988. A model of CEO dismissal. Academy

of Management Review 13 (2), 255-270.

Hayn, C. 1995. The information content of losses. Journal of Accounting and Economics 20

(2), 125-153.

Hazarika, S., Karpoff, J.M., Nahata, R., 2012. Internal corporate governance, CEO turnover,

and earnings management. Journal of financial Economics 104 (1), 44-69.

Kormendi, R., Lipe, R., 1987. Earnings innovations, earnings persistence, and stock returns.

Journal of Business 60 (3), 323-345.

Lang, M.H., Lins, K.V., Miller, D.P., 2003. ADRs, analysts, and accuracy: Does cross listing

in the United States improve a firm's information environment and increase market

value? Journal of Accounting Research 41 (2), 317-345.

Lang, M.H., Lundholm, R.J., 1996. Corporate disclosure policy and analyst behavior.

Accounting Review 71 (4), 467-492.

Murphy, K.J., Zimmerman, J.L., 1993. Financial performance surrounding CEO turnover.

Journal of Accounting and Economics 16 (1), 273-315.

Parrino, R. 1997. CEO turnover and outside succession a cross-sectional analysis. Journal of

Financial Economics 46 (2), 165-197.

Petersen, M.A. 2009. Estimating standard errors in finance panel data sets: Comparing

approaches. The Review of Financial Studies 22 (1), 435-480.

Piotroski, J.D., Roulstone, D.T., 2004. The influence of analysts, institutional investors, and

insiders on the incorporation of market, industry, and firm-specific information into

stock prices. The Accounting Review 79 (4), 1119-1151.

Pourciau, S. 1993. Earnings management and nonroutine executive changes. Journal of

Accounting and Economics 16 (1-3), 317-336.

Skinner, D.J., Sloan, R.G., 2002. Earnings surprises, growth expectations, and stock returns

or don't let an earnings torpedo sink your portfolio. Review of Accounting Studies 7

(2-3), 289-312.

Vancil, R.F., 1987. Passing the baton: Managing the process of CEO succession: Harvard

Business School Press.

Wilson, M., and Wang, L. W., 2010. Earnings management following chief executive officer

changes: the effect of contemporaneous chairperson and chief financial officer

appointments. Accounting & Finance 50 (2), 447-480.


26

Table 1 CEO turnover events

A. CEO turnover events from Audit Analytics, exclusions and merge with CRSP/Compustat and I/B/E/S

(1) CEO turnover events 12,742

Less:

(2)

Mergers/acquisition, bankruptcies,

asset sales/spin-offs 1,162

11,580

(3) Merged with CRSP/Compustat 4,227

(4) Merged with IBES 2,753

B. CEO turnover events by year

Year

(quarter)

CEO turnover events

(CRSP/Compustat/IBES)

as

% of firms

2005 365 10.9%

2006 387 11.2%

2007 380 10.9%

2008 430 13.0%

2009 322 10.2%

2010 276 8.8%

2011 307 10.3%

2012 286 10.1%

Q1 795

Q2 686

Q3 609

Q4 552

Total (average)

2,753 10.7%

C. CEO turnover events by industry

Industry

SIC

header

CEO turnover events

(CRSP/Compustat/IBES)

as % of

unique firms

Agriculture, Forestry, Fishing 01-09 4 7.5%

Mining 10-14 108 8.6%

Construction 15-17 33 10.2%

Manufacturing 20-39 1,199 10.6%

Transportation 40-43 29 10.7%

Public utilities 44-49 188 8.3%

Wholesale trade 50-51 66 10.2%

Retail trade 52-59 271 14.9%

Finance, insurance, real estate 60-69 344 6.8%

Services 70-89 511 10.7%

Public administration 91-99 NA NA

Total (average) 2,753


27

Table 2 Differences in means for key variables

This table presents the means of important variables for new-CEO firm-quarters, outgoing-CEO firm-quarters and established-CEO firm-quarters and their differences in means.

Asterisks ***, ** and * next to a coefficient estimate indicate significance levels of 1%, 5% and 10%, respectively. Standard errors are calculated assuming unequal variance

in two samples. Appendix I includes definitions of variables.

Established OUT NEW Differences in means

N Mean N Mean N Mean

OUT-

EST

NEW-

EST

NEW-

OUT

Cumulative abnormal return

CAR(-1,1) 96,707 0.05 9,208 -1.13 9,229 -0.26 -1.17*** -0.30*** 0.87***

CAR(-1,1)×GOOD 40,237 2.03 3,641 1.32 3,668 2.19 -0.71*** 0.16 0.88***

CAR(-1,1)×BAD 21,954 -3.25 2,384 -4.72 2,144 -3.48 -1.47*** -0.23 1.24***

Unexpected earnings (Expected earnings is the median IBES forecast)

UE(t) 66,197 0.00 6,406 -0.13 6,083 -0.08 -0.13*** -0.08*** 0.05*

UE(t)×GOOD 42,738 0.37 3,835 0.39 3,808 0.46 0.02** 0.09*** 0.07***

UE(t)×BAD 23,459 -0.67 2,571 -0.90 2,275 -0.98 -0.23*** -0.31*** -0.08*

BAD(t) 67,461 0.36 6,582 0.40 6,298 0.38 0.05*** 0.02*** -0.03***

Firm characteristics

MB(t-1) 96,460 5.51 9,129 5.55 9,250 5.29 0.04 -0.23*** -0.27***

SIZE(t-1) 99,326 5.49 9,495 5.44 9,725 5.33 -0.05 -0.16*** -0.12***

EVOL(t) 93,251 0.04 8,760 0.06 9,070 0.08 0.01*** 0.04*** 0.03***

LOSS(t) 104,854 0.28 10,012 0.40 9,897 0.42 0.12*** 0.14*** 0.02***

NEGDA(t) 67,077 0.46 7,266 0.47 7,114 0.49 0.01 0.03*** 0.02**

GDOWN(t) 67,654 0.50 6,605 0.55 6,313 0.55 0.05*** 0.05*** 0.01

PERS(t-1) 96,031 0.15 9,110 0.16 9,452 0.14 0.01 -0.01 -0.02**


28

Table 3 Correlation matrix

This table contains pairwise Pearson correlation coefficients among important variables. Appendix I includes definitions of variables. Asterisks ***, ** and * next to a coefficient

estimate indicate significance levels of 1%, 5% and 10%, respectively.

CAR(-1,1) UE(t) BAD(t) SIZE(t-1) MB(t-1) EVOL(t-1) LOSS(t) NEGDA(t) PERS(t-1)

UE(t) 0.22***

BAD(t) -0.30*** -0.48***

SIZE(t-1) 0.02*** 0.07*** -0.12***

MB(t-1) 0.00 0.06*** -0.08*** 0.26***

EVOL(t-1) -0.01*** -0.08*** 0.05*** -0.25*** -0.24***

LOSS(t) -0.14*** -0.24*** 0.23*** -0.30*** -0.03*** 0.26***

NEGDA(t) 0.01*** -0.07*** 0.04*** 0.00 -0.01** 0.02*** 0.10***

PERS(t-1) 0.00 0.00 0.00 0.00 0.02*** -0.04*** -0.01*** 0.01**

GDOWN(t) -0.02*** -0.04*** 0.07*** -0.04*** -0.08*** 0.00 0.08*** 0.02*** 0.01***


29

Table 4 Stock price response to new-CEO earnings news

This table presents estimation results for the new-CEO attributes effect and the honeymoon effect for all CEO

turnover events. Column (1) presents regression results using the following equation:

𝐶𝐴𝑅(−1,1) = 𝛽0 + 𝛽1 × 𝐺𝑂𝑂𝐷 × 𝑂𝑈𝑇 + 𝛽2 × 𝐺𝑂𝑂𝐷 × 𝑁𝐸𝑊

+𝛽3 × 𝐵𝐴𝐷 + 𝛽4 × 𝐵𝐴𝐷 × 𝑂𝑈𝑇 + 𝛽5 × 𝐵𝐴𝐷 × 𝑁𝐸𝑊

+𝛽6 × 𝐺𝑂𝑂𝐷 × 𝑈𝐸 + 𝛽7 × 𝐺𝑂𝑂𝐷 × 𝑈𝐸 × 𝑂𝑈𝑇 + 𝛽8 × 𝐺𝑂𝑂𝐷 × 𝑈𝐸 × 𝑁𝐸𝑊

+𝛽9 × 𝐵𝐴𝐷 × 𝑈𝐸 + 𝛽10 × 𝐵𝐴𝐷 × 𝑈𝐸 × 𝑂𝑈𝑇 + 𝛽11 × 𝐵𝐴𝐷 × 𝑈𝐸 × 𝑁𝐸𝑊


Column (2) presents results from a regression similar the approach used in Clayton, Hartzell and Rosenberg (2005),

as below:

𝐶𝐴𝑅(−1,1) = 𝛼0 + 𝛼1 × 𝐺𝑂𝑂𝐷 × 𝑈𝐸 + 𝛼2 × 𝐺𝑂𝑂𝐷 × 𝑈𝐸 × 𝑂𝑈𝑇 + 𝛼3 × 𝐺𝑂𝑂𝐷 × 𝑈𝐸 × 𝑁𝐸𝑊

+𝛼4 × 𝐵𝐴𝐷 × 𝑈𝐸 + 𝛼5 × 𝐵𝐴𝐷 × 𝑈𝐸 × 𝑂𝑈𝑇 + 𝛼6 × 𝐵𝐴𝐷 × 𝑈𝐸 × 𝑁𝐸𝑊


Estimation results for control variables, non-linear terms and year-quarter dummies are omitted. Regressions include

firm fixed effects. Standard errors are clustered by year-quarter. Asterisks ***, ** and * next to a coefficient estimate

indicate significance levels of 1%, 5% and 10%, respectively. Appendix I includes a detailed description of variables.

(1) (2)

CAR(-1,1) CAR(-1,1)

GOOD×OUT (β1) 0.05

GOOD×NEW (H1a: β2>0) 0.27*

BAD (β3) -3.97***

BAD×OUT(β4) -0.86***

BAD×NEW (H2a: (β5>0) 0.46*

GOOD×UE(β6) 8.53*** α1 9.79***

GOOD×UE×OUT(β7) -1.32 α2 -1.20**

GOOD×UE×NEW (H1a: β8>0) 0.42 α3 -0.01

BAD×UE (β9) 1.71*** α4 1.86***

BAD×UE×OUT (β10) -0.65 α5 0.59*

BAD×UE×NEW (H2a: β11<0) -0.06 α6 -0.50**

Const. (β0) 4.11*** α0 -0.54***

Adj R-sqr(%) 14.7 9.9

N 42,155 74,028

F-test for equality in coefficients

GOOD×NEW -GOOD×OUT 0.23

BAD×NEW -BAD×OUT 1.31***

GOOD×UE×NEW-GOOD×UE×OUT 1.74* 1.20**

BAD×UE×NEW-BAD×UE×OUT 0.59 -1.09**


30

Table 5 CEO turnover in challenging situations and stock price response to new-CEO earnings news

This table reports the CEO honeymoon effect following turnovers in challenging situations. We use six alternative proxies for a challenging turnover situation (refer to Appendix

I for more details). Regressions use an equation similar to equation (1) (used for column (1) in Table 3) and the difference is that in this table CEO turnovers are grouped into

the CHAL type and the NORM type, using the following equation:

𝐶𝐴𝑅(−1,1) = 𝛾0 + 𝛾1 × 𝐺𝑂𝑂𝐷 × 𝑂𝑈𝑇 × 𝑁𝑂𝑅𝑀 + 𝛾2 × 𝐺𝑂𝑂𝐷 × 𝑂𝑈𝑇 × 𝐶𝐻𝐴𝐿 + 𝛾3 × 𝐺𝑂𝑂𝐷 × 𝑁𝐸𝑊 × 𝑁𝑂𝑅𝑀 + 𝛾4 × 𝐺𝑂𝑂𝐷 × 𝑁𝐸𝑊 × 𝐶𝐻𝐴𝐿

+𝛾5 × 𝐵𝐴𝐷 + 𝛾6 × 𝐵𝐴𝐷 × 𝑂𝑈𝑇 × 𝑁𝑂𝑅𝑀 + 𝛾7 × 𝐵𝐴𝐷 × 𝑂𝑈𝑇 × 𝐶𝐻𝐴𝐿 + 𝛾8 × 𝐵𝐴𝐷 × 𝑁𝐸𝑊 × 𝑁𝑂𝑅𝑀 + 𝛾9 × 𝐵𝐴𝐷 × 𝑁𝐸𝑊 × 𝐶𝐻𝐴𝐿

+𝛾10 × 𝐺𝑂𝑂𝐷 × 𝑈𝐸 + 𝛾11 × 𝐺𝑂𝑂𝐷 × 𝑈𝐸 × 𝑂𝑈𝑇 × 𝑁𝑂𝑅𝑀 + 𝛾12 × 𝐺𝑂𝑂𝐷 × 𝑈𝐸 × 𝑂𝑈𝑇 × 𝐶𝐻𝐴𝐿

+𝛾13 × 𝐺𝑂𝑂𝐷 × 𝑈𝐸 × 𝑁𝐸𝑊 × 𝑁𝑂𝑅𝑀 + 𝛾14 × 𝐺𝑂𝑂𝐷 × 𝑈𝐸 × 𝑁𝐸𝑊 × 𝐶𝐻𝐴𝐿

+𝛾15 × 𝐵𝐴𝐷 × 𝑈𝐸 + 𝛾16 × 𝐵𝐴𝐷 × 𝑈𝐸 × 𝑂𝑈𝑇 × 𝑁𝑂𝑅𝑀 + 𝛾17 × 𝐵𝐴𝐷 × 𝑈𝐸 × 𝑂𝑈𝑇 × 𝐶𝐻𝐴𝐿

+𝛾18 × 𝐵𝐴𝐷 × 𝑈𝐸 × 𝑁𝐸𝑊 × 𝑁𝑂𝑅𝑀 + 𝛾19 × 𝐵𝐴𝐷 × 𝑈𝐸 × 𝑁𝐸𝑊 × 𝐶𝐻𝐴𝐿

+𝜽 × 𝑵𝑶𝑵𝑳𝑰𝑵𝑬𝑨𝑹 + 𝝆 × 𝑪𝑶𝑵𝑻𝑹𝑶𝑳 + 𝝉 × 𝒀𝒆𝒂𝒓𝒒𝒖𝒂𝒓𝒕𝒆𝒓 + 𝜎 (3)

Estimates for the slope coefficients on the nonlinear terms (𝑵𝑶𝑵𝑳𝑰𝑵𝑬𝑨𝑹), control variables (𝑪𝑶𝑵𝑻𝑹𝑶𝑳) and year-quarter dummy variables (𝒀𝒆𝒂𝒓𝒒𝒖𝒂𝒓𝒕𝒆𝒓)

are omitted. Appendix I includes a detailed description of variables. Regressions include firm fixed effects. Standard errors are clustered by year-quarter.

Asterisks ***, ** and * next to a coefficient estimate indicate significance levels of 1%, 5% and 10%, respectively.


31

Non-RT External Low ROA Low Ret High Vol Low SalesGrwth

CAR(-1,1) CAR(-1,1) CAR(-1,1) CAR(-1,1) CAR(-1,1) CAR(-1,1)

GOOD×OUT×NORM (γ1) 0.19 0.19 -0.13 0.25 0.04 -0.04

GOOD×OUT×CHAL (γ2) -0.31 -0.27 0.06 -0.60 -0.10 0.05

GOOD×NEW×NORM (γ3) -0.23 0.23 -0.09 0.01 -0.08 0.31

GOOD×NEW×CHAL (H1b: γ4>0) 0.69** 0.49* 0.96** 0.73** 1.12** 0.22

BAD (γ5) -3.97*** -3.97*** -3.96*** -3.98*** -3.97*** -3.96***

BAD×OUT×NORM (γ6) -0.82** -0.71** -0.44 0.32 -0.18 -0.85*

BAD×OUT×CHAL (γ7) -1.24** -1.50** -1.46** -2.22*** -2.13*** -0.78**

BAD×NEW×NORM (γ8) 0.14 0.47 0.40 0.27 0.74** 0.69

BAD×NEW×CHAL (H2b: γ9>0) 0.91* 0.27 0.51 0.76* -0.21 0.32

GOOD×UE (γ10) 8.51*** 8.53*** 8.53*** 8.53*** 8.52*** 8.53***

GOOD×UE×OUT×NORM (γ11) -0.54 -1.75 0.65 0.69 -0.77 0.53

GOOD×UE×OUT×CHAL (γ12) -1.61 -0.36 -3.07** -3.15*** -1.66 -2.60*

GOOD×UE×NEW×NORM (γ13) 1.47 1.07 2.23** 1.38 1.89* 0.42

GOOD×UE×NEW×CHAL (H1b: γ14>0) -1.33 -0.41 -1.44 -0.67 -1.57 0.58

BAD×UE (γ15) 1.72*** 1.71*** 1.73*** 1.70*** 1.75*** 1.71***

BAD×UE×OUT×NORM (γ16) -1.35** 0.07 0.18 -0.68 -0.11 0.79

BAD×UE×OUT×CHAL (γ17) -0.16 -1.42* -1.13 -1.10* -1.38* -1.35*

BAD×UE×NEW×NORM (γ18) -1.02 0.41 -0.39 -0.24 -0.23 0.78

BAD×UE×NEW×CHAL (H2b: γ19<0) 1.08 -0.70 0.22 0.37 -0.16 -0.51

Const (γ0) 4.11*** 4.09*** 4.11*** 3.99*** 4.13*** 4.10***

Adj R-sqr(%) 14.8 14.7 14.7 14.8 14.7 14.7

N 42,155 42,155 42,155 42,155 42,155 42,155


32

Table 6 Analyst coverage and stock price response to new-CEO earnings news

This table presents results from regressions where we condition good earnings news from new CEOs on the

level of analyst coverage.

Column (1) presents results where we interact good earnings news from new CEOs and bad earnings news from

new CEOs with number of analysts. The regression uses the following equation:

𝐶𝐴𝑅(−1,1) = 𝜃0 + 𝜃1 × 𝐴𝑁𝐴𝐿𝑌𝑆𝑇 + 𝜃2 × 𝐺𝑂𝑂𝐷 × 𝑂𝑈𝑇

+𝜃3 × 𝐺𝑂𝑂𝐷 × 𝑁𝐸𝑊 + 𝜃4 × 𝐺𝑂𝑂𝐷 × 𝑁𝐸𝑊 × 𝐴𝑁𝐴𝐿𝑌𝑆𝑇

+𝜃5 × 𝐵𝐴𝐷 + 𝜃6 × 𝐵𝐴𝐷 × 𝑂𝑈𝑇

+𝜃7 × 𝐵𝐴𝐷 × 𝑁𝐸𝑊 + 𝜃8 × 𝐵𝐴𝐷 × 𝑁𝐸𝑊 × 𝐴𝑁𝐴𝐿𝑌𝑆𝑇

+𝜃9 × 𝐺𝑂𝑂𝐷 × 𝑈𝐸 + 𝜃10 × 𝐺𝑂𝑂𝐷 × 𝑈𝐸 × 𝑂𝑈𝑇 + 𝜃11 × 𝐺𝑂𝑂𝐷 × 𝑈𝐸 × 𝑁𝐸𝑊

+𝜃12 × 𝐵𝐴𝐷 × 𝑈𝐸 + 𝜃13 × 𝐵𝐴𝐷 × 𝑈𝐸 × 𝑂𝑈𝑇 + 𝜃14 × 𝐵𝐴𝐷 × 𝑈𝐸 × 𝑁𝐸𝑊


Column (2) presents results where we classify a firm as a high-analyst (middle- or low-) coverage firm if it is

ranked among the top 1/3 (middle 1/3 or bottom 1/3) of all firms in terms of number of analyst following in a

quarter. The regression uses the following equation:

𝐶𝐴𝑅(−1,1) = 𝛿0 + 𝛿1 × 𝑀𝐸𝐷 + 𝛿2 × 𝐻𝐼𝐺𝐻 + 𝛿3 × 𝐺𝑂𝑂𝐷 × 𝑂𝑈𝑇

+𝛿4 × 𝐺𝑂𝑂𝐷 × 𝑁𝐸𝑊 × 𝐿𝑂𝑊 + 𝛿5 × 𝐺𝑂𝑂𝐷 × 𝑁𝐸𝑊 × 𝑀𝐸𝐷 + 𝛿6 × 𝐺𝑂𝑂𝐷 × 𝑁𝐸𝑊 × 𝐻𝐼𝐺𝐻

+𝛿7 × 𝐵𝐴𝐷 + 𝛿8 × 𝐵𝐴𝐷 × 𝑂𝑈𝑇

+𝛿9 × 𝐵𝐴𝐷 × 𝑁𝐸𝑊 × 𝐿𝑂𝑊 + 𝛿10 × 𝐵𝐴𝐷 × 𝑁𝐸𝑊 × 𝑀𝐸𝐷 + 𝛿11 × 𝐵𝐴𝐷 × 𝑁𝐸𝑊 × 𝐻𝐼𝐺𝐻

+𝛿12 × 𝐺𝑂𝑂𝐷 × 𝑈𝐸 + 𝛿13 × 𝐺𝑂𝑂𝐷 × 𝑈𝐸 × 𝑂𝑈𝑇 + 𝛿14 × 𝐺𝑂𝑂𝐷 × 𝑈𝐸 × 𝑁𝐸𝑊

+𝛿15 × 𝐵𝐴𝐷 × 𝑈𝐸 + 𝛿16 × 𝐵𝐴𝐷 × 𝑈𝐸 × 𝑂𝑈𝑇 + 𝛿17 × 𝐵𝐴𝐷 × 𝑈𝐸 × 𝑁𝐸𝑊


Estimates for the slope coefficients on the nonlinear terms (𝑵𝑶𝑵𝑳𝑰𝑵𝑬𝑨𝑹), control variables (𝑪𝑶𝑵𝑻𝑹𝑶𝑳) and

year-quarter dummy variables (𝒀𝒆𝒂𝒓𝒒𝒖𝒂𝒓𝒕𝒆𝒓) are omitted. Regressions include firm fixed effects. Standard

errors are clustered by year quarter. Asterisks ***, ** and * next to a coefficient estimate indicate significance

levels of 1%, 5% and 10%, respectively. Appendix I includes a detailed description of variables.


33

(1) (2)

CAR(-1,1) CAR(-1,1)

MED -0.11

ANALYST 0.00 HIGH 0.08

GOOD×OUT 0.05 GOOD×OUT 0.04

GOOD×NEW 0.81** GOOD×NEW×LOW (H1c: δ4>0) 1.13**

GOOD×NEW×ANALYST (H1c: θ4 <0) -0.07** GOOD×NEW×MED 0.78**

GOOD×NEW×HIGH (H1c: δ6=0) -0.28

BAD -3.96*** BAD -3.96***

BAD×OUT -0.86*** BAD×OUT -0.86***

BAD×NEW -0.09 BAD×NEW×LOW (H2c: δ9=0) -0.20

BAD×NEW×ANALYST (H2c: θ8>0) 0.09* BAD×NEW×MED 0.36

BAD×NEW×HIGH (H2c: δ11>0) 0.95**

GOOD×UE 8.54*** GOOD×UE 8.56***

GOOD×UE×OUT -1.31 GOOD×UE×OUT -1.30

GOOD×UE×NEW 0.14 GOOD×UE×NEW -0.07

BAD×UE 1.71*** BAD×UE 1.71***

BAD×UE×OUT -0.65 BAD×UE×OUT -0.65

BAD×UE×NEW -0.19 BAD×UE×NEW -0.24

Const 4.10*** Const 4.12***

Adj R-sqr 14.7 Adj R-sqr 14.7

N 42,155 N 42,155


34

Appendix I Variable definitions

Variable Definition (Compustat code)

A Total assets at the end of each quarter (atq).

Accruals Total accruals, equal to income before extraordinary items minus CFO.

AR Receivables at the end of each quarter (rectq).

BAD A dummy variable that takes the value of 1 if UE is negative and 0

otherwise.

Book value of equity The book value of equity a firm is the book value of stockholders’ equity

(seqq), plus balance sheet deferred taxes (txditcq) and minus the book

value of preferred stock. Depending on availability, we use the

redemption (pstkrvq), liquidation (pstkl), or par value (pstkq) (in that

order) to estimate the book value of preferred stock. If stockholders’

equity is unavailable, we measure stockholders’ equity as the book value

of common equity (ceqq) plus the par value of preferred stock (pstkq),

or the book value of assets (atq) minus total liabilities (ltq) (in that

order). (See Davis et al., 2000)

CAR(-1,1) Cumulative abnormal returns during the three day period around

earnings announcement date, computed as the raw buy-and-hold return

less the return to the portfolio matched on size and book-to-market

portfolio, as in Fama and French (1992). At the end of June in each year,

we classify all CSRP stocks (with share code of 10 and 11, or, US

common equities) in to 25 portfolios by size at the end of June of the

current year and by book-to-market ratio at the end of December of the

previous year. Portfolio returns are equal-weighted returns.

CFO To obtain quarterly cash flow from operations in the second, third and

fourth fiscal quarter, we subtract accumulated year-to-date cash flow

from operations ended in ended in the previous quarter from that ended

in this quarter (oancfy) and that; quarterly cash flow from operations in

the first fiscal quarter equals to the year-to-date operating cash flow.

DA Discretionary accruals estimated cross-sectionally from the modified

Jones Model Dechow et al. (1995). A larger DA indicates a higher level

of discretionary accruals or upward earnings management. Appendix II

includes a description of the estimation method.

EVOL Volatility in earnings (niq) over quarter 𝑡 − 3 to quarter 𝑡 scaled by the

market capitalisation (cshoq×prccq) of the firm at the end of quarter 𝑡.

GDOWN A dummy variable that takes the value of 1 if the median of analyst

forecasts decreases during the period between the last earnings

announcement and the current earnings announcement, and 0 otherwise.

GOOD A dummy variable that takes the value of 1 if UE is non-negative and 0

otherwise.

HIGH A dummy variable that takes the value of 1 if a firm has a high level of

analyst following in that quarter. At the end of each calendar quarter,

firms are sorted into terciles (HIGH, MED and LOW) based on the

number of active analyst following. An analyst needs to issue at least

one valid forecast before the quarterly earnings announcement. Refer to

“UE” for the definition of “valid”.

LOSS Takes the value of 1 if the quarterly net profit (niq) is below zero and 1

otherwise.

LOW A dummy variable that takes the value of 1 if a firm has a low level of

analyst following in that quarter.

MB The market-to-book decile number of a firm. At the end of each calendar

quarter, all firms are sorted into 10 portfolios based on their market-to-

books ratios. Market value is the market value of equity (prcc ×cshoq).

Book is the book value of equity.

MED A dummy variable that takes the value of 1 if a firm has a medium level

of analyst following in that quarter.


35

BAD A dummy variable that takes the value of 1 if UE is negative and 0

otherwise.

NEGDA NEGDA takes the value of 1 if the discretionary accrual during a quarter

is negative and 0 otherwise. Discretionary accruals estimated from the

modified Jones Model Dechow et al. (1995) for each industry (firms

with the same two-digit SIC head) and calendar year-quarter. See

Appendix II for the estimation model.

NEW NEW equals to 1 if the earnings announcement date is within 120

calendar days from the beginning of CEO tenure, and 0 otherwise.

NORM/CHAL NORM takes the value of 1 (CHAL takes of value of 0) if the CEO

turnover occurs in a normal (challenging) situation. All CEO turnovers

occur in either a normal situation or a challenging situation. We use six

alternative proxies for a challenging turnover situation: the CEO

turnover is non-routine; the new CEO is appointed externally; the

company’s ROA over the 12 months prior to the CEO turnover falls in

the lowest 1/3 among its industry peers; stock return over the 12 months

prior to CEO changes falls in the lowest 1/3 among its industry peers;

the volatility (VOL) over the 12 months prior to CEO turnover is among

the highest 1/3 in the industry; the 12 month sales growth prior to CEO

change falls in the lowest 1/3 among its industry peers. An industry

includes all firms sharing the same two-digit SIC head.

UE Unexpected earnings is the difference between actual quarterly earnings

per share and expected earnings per share, scaled by the closing stock

price five trading day before the earnings announcement (as in

DellaVigna and Pollet, 2009). We use the median of all valid consensus

analyst forecasts as a proxy for expected earnings. A valid forecast has

to be revised or confirmed to be valid during the 30 day period prior to

the earnings announcement. Analyst forecast earnings per share are

adjusted for share splits and other corporate events between the forecast

date and the earnings announcement date.

OUT OUT equals to 1 if the earnings announcement date is within 120

calendar days prior to the end of CEO tenure, and 0 otherwise.

PERS Earnings persistent is the coefficient from the first-order auto-regression

model, where we regress (𝐸𝑡 − 𝐸𝑡−4) on (𝐸𝑡−1 − 𝐸𝑡−5) using financial

results over the past 8 quarters. 𝐸𝑡 is net profit in the current quarter

(niq).

PPE Gross book value of property, plant and equipment (ppegtq). We assume

a linear growth rate of PPE and fill in the missing PPE observations if

needed.

RET12M Accumulated stock returns over previous 12 months.

ROA Net profit (niq) over the four quarters ending in quarter 𝑡 divided by total

assets (atq) as the end of quarter 𝑡.

SALESGRWTH12M Sales growth is the percentage growth in rolling four-quarter sales over

the rolling four-quarter sales ending in the same fiscal quarter of last

year.

SIZE The size decile number of a firm. At the end of each calendar quarter,

all firms are sorted into deciles based on their size. Size is the market

value is of equity (prcc ×cshoq).

VOL Annualized daily stock return volatility, or standard deviation, in a

calendar year.


36

Appendix II Using Modified Jone’s Model to Estimate Discretionary Accruals

We run the following cross-sectional model for each two-digit SIC-quarter group:

𝑻𝒐𝒕𝒂𝒍 𝑨𝒄𝒄𝒓𝒖𝒂𝒍𝒔𝒕

𝑨𝒕−𝟏= 𝜶𝟎 + 𝜶𝟏

𝟏

𝑨𝒕−𝟏+ 𝜶𝟐

∆𝑺𝒕

𝑨𝒕−𝟏+ 𝜶𝟑

𝑷𝑷𝑬𝒕

𝑨𝒕−𝟏+ 𝜺𝒕 , (1)

where Total Accrualst is the earnings before extraordinary items and discontinued operations minus the

operating cash flows in quarter t; At−1 is the total assets in quarter t − 1; ∆St is the change in revenues from

the preceding quarter; and PPEt is gross property, plant, and equipment.1 We require at least 15 observations

for each cross-sectional estimate. Coefficient estimates from equation (1) are used to estimate the normal levels

of accruals as below:

𝑁𝑜𝑟𝑚_𝑎𝑐𝑐𝑟𝑢𝑎𝑙𝑠𝑡 = �̂�0 + �̂�11

𝐴𝑡−1+ �̂�2

∆𝑆𝑡−∆𝐴𝑅𝑡

𝐴𝑡−1+ �̂�3

𝑃𝑃𝐸𝑡

𝐴𝑡−1, (2)

where ∆𝐴𝑅𝑡 is the change in accounts receivable. Discretionary accruals (𝐷𝐴𝑡) is the difference between total

accruals and the fitted normal accruals.

1 Missing quarterly gross PPE values are filled in by linear interpolation.


37

Appendix III Important variables

This table presents summary statistics of important variables for sample firm-quarters from 2005 to 2012.

Unexpected earnings are winsorized at 1% on both tails. Appendix I includes definitions of variables.

N Mean Median SD

25th

Pctile

75th

Pctile

Cumulative abnormal return

CAR(-1,1) 115,144 -0.07 -0.16 8.80 -4.26 4.07

CAR(-1,1)×GOOD 47,546 1.99 1.38 8.54 -2.30 5.94

CAR(-1,1)×BAD 26,482 -3.40 -2.60 8.91 -7.24 1.05

Unexpected earnings

UE(t) 78,686 -0.01 0.05 1.11 -0.08 0.25

UE(t)×GOOD 50,381 0.38 0.17 0.60 0.06 0.42

UE(t)×BAD 28,305 -0.72 -0.21 1.41 -0.67 -0.06

BAD(t) 80,341 0.36 0.00 0.48 0.00 1.00

Firm characteristics

MB(t-1) 114,839 5.50 5.00 2.86 3.00 8.00

SIZE(t-1) 118,546 5.47 5.00 2.83 3.00 8.00

EVOL(t) 111,081 0.05 0.01 0.13 0.00 0.03

LOSS(t) 124,763 0.30 0.00 0.46 0.00 1.00

NEGDA(t) 81,457 0.47 0.00 0.50 0.00 1.00 PERS(t-1) 114,593 0.15 0.11 0.45 -0.13 0.42


38

Appendix IV Routine CEO turnover and external appointment

This appendix summarizes the procedure to identify routine CEO turnover and externally appointed

CEOs.

Audit Analytics records a CEO change event as an “action” with a “reason.” It also reports whether

the departing CEO has retained any position within the company. A CEO change is routine when the departing

CEO retires after 60 or the departing CEO retains a position within the company. If the departing CEO retires

before 60 or the reason for retirement is non-routine (for example, investigation, health, change in control and

disagreement), the CEO change is classed as non-routine. If the action of CEO departure is recorded as death,

dismissal, resignation or ceasing employment of a CEO younger than 60, the CEO change is classed as non-

routine. For CEO turnovers with insufficient information, we rely on the gap between CEO departing and new

CEO in position to determine whether a CEO turnover is non-routine. If a firm appoints a new CEO more than

30 days after his/her predecessor leaves, we treat the CEO change as non-routine. However, 356 CEO

appointments (or 13% of total CEO turnovers) cannot be matched with a CEO departure, and we do not have

sufficient information to determine whether the CEO turnover is routine. After reading a sample of related SEC

filings, we believe that they tend to be genuine CEO turnover events and label these as other.2 Our sample

contains more routine than non-routine turnovers – each type accounts for 56% and 44% of CEO turnovers,

respectively.3

2 For unmatched CEO appointments, we randomly sampled 20 cases and reviewed their filings. These

appointments are mostly genuine CEO turnovers. They are neither predominantly routine nor non-routine

turnovers. These appointments cannot be matched with a departure for the following reasons: (1) the departure

of the preceding CEO is reported in the news or included in the filings but not recorded in the database; (2) the

CEO departure occurred in 2004, before the executive turnover data become available in Audit Analytics; (3)

data entry errors and not CEO turnover events.

3 Hazarika et al. (2012), a recent study on earnings management, board discipline and forced CEO

turnovers, record 20% of CEO turnovers as forced turnovers and 80% as voluntary turnovers. We are aware

that the percentage of forced turnover is smaller than our 44% of non-routine CEO turnovers. Two reasons can

be driving the difference: (1) Forced turnover is a sub-category of non-routine turnover; for example, a sudden

CEO resignation is a voluntary change (not a forced change) in their paper but can be non-routine in our paper;

(2) Our paper includes all CEO turnovers events in CRSP/Compustat/IBES companies instead of the S&P 1000

companies in their paper. Small companies have less scope than large companies to implement a planned CEO

succession; thus, our sample can contain a higher percentage of non-routine CEO changes than their S&P 1000

sample.


39

We examine the reason for a CEO appointment to decide whether the new CEO is promoted internally

or hired externally. If the new CEO is appointed to assume an additional position or as a result of a position

change within company, the new CEO is an internal hire; otherwise, the new CEO is an external hire. A total

of 220 CEO departures (or 8% of total CEO turnovers) cannot be matched with a CEO appointment, and we

lack the information to determine whether the firms promote new CEOs internally or hire externally after these

CEO departures. We reviewed a random sample of related SEC filings and found that they mostly tended to be

genuine CEO turnover events. Thus, we categorize them as the other type of CEO turnover.4 About 63% of the

appointments are internal promotions and about 37% of new CEOs are hired externally.5

Routine CEO succession does not tend to involve internal promotion – only 64% of CEOs are

promoted internally as part of routine successions.

4 For unmatched CEO departures, we randomly sampled 20 cases and reviewed their filings. These

departures are mostly genuine CEO turnovers. These departures cannot be matched with an appointment for

the following reasons: (1) the CEO departure occurred in 2004, before the executive turnover data became

available in Audit Analytics; (2) data entry errors.

5 In Parrino (1997), external new appointments account for 15% of CEO appointments in the Forbes

sample from 1969 to 1989. This percentage is lower than the percentage of external appointments, 37%, in our

study. Parrino (1997, Table 3) also shows that large companies are more likely to promote internally to fill the

CEO position. Our sample covers CEO changes in all CRSP/Compustat/IBES firms, while Parrino (1997)

focuses on large firms. The inclusion of small firms could increase percentage of external succession in our

study.

Documents

Stock Price Response to New CEO Earnings News · Stock Price Response to New CEO Earnings News 4 recorded by Audit Analytics between 2005 and 2012. Using panel regressions, we compare