Mimicking and Herding Behaviors among U.S. Investment Analy - PowerPoint Presentation

Slide1

Mimicking and Herding Behaviors among U.S. Investment Analysts: Implications for Market Reactions to Actual Earnings Announcements

Liem

NguyenWestfield State UniversityJudy K. Beckman and Henry Oppenheimer University of Rhode Island

AAA Atlanta Annual Meetings August 6, 2014Slide2

Introduction

CVS

Institutional Investors, Regulators, Credit Rating Agencies, Suppliers, etc.Analysts

Management

Background

IntroductionSlide3

Management Voluntary Disclosure

“CVS Caremark (CVS) issued weak 2011 earnings guidance, and its shares fell 6% in morning trading. CVS posted fourth quarter earnings of 80 cents, versus expectations for 81cents” Barron’s February 3, 2011Tyco (TYC) now expects earnings of 30 cents to 33 cents a share instead of the 45 cents to 47 cents guidance it gave in July, chief executive Ed Breen said in his first conference call with the company's investors.” AP September 2005.

BackgroundIntroductionSlide4

Introduction

The Sequence of Management Annual Earnings Guidance and Analyst’s Estimates

Analyst’s Consensus Number

(PRIOR)

First Analyst’s Consensus Number

(FIRST)

Mimicking occurs when the amount is within 1 cent of management guidance

Management Announces Annual Earnings Guidance

Actual Annual Earnings Is Released

(GUI

)

(ACT

)

Last Analyst’s Consensus Number

(LAST)

$3.50

$3.45

$3.40

$3.40

$3.42

Herding occurs when the subsequent analysts issue forecasts within 1 cent of the firstSlide5

Literature: Meetable/Beatable Forecasts (Mimicking)

Literature

Skinner (1994) Managers may preempt announcement of negative earnings surprises to reduce the potential costs of shareholder suitsSoffer et al. (2000) - short-term management earnings guidance tends to be downwardly biased

Matsumoto (2002) and Richardson, Teoh, and Wysocki (2004)

- firms “walk” analysts toward beatable earnings forecasts via their earnings guidance disclosuresSlide6

Literature: Management Incentives

Baik and Jiang (2006)Management dampens analysts’ earnings expectations with pessimistic quarterly guidance when their firms:

Have high growth opportunities,operate in a high-litigation industry,have transient institutional ownership.

All factors above are associated with pessimistic management guidance because of the costs of missing the consensus of analysts forecasts.

LiteratureSlide7

Literature: Herding Behaviors

Trueman (1994) - Math models predict that an analyst prefers to release an earnings forecast that is close to prior earnings expectations even when his/her own information justifies a more extreme forecast.Hong, Kubik, and Solomon (2000)

- find that more experienced analysts are more likely to issue bold forecasts than are less experienced analysts and that brokerage firms are more likely to discharge the less experienced analysts for inaccurate or bold forecastsLiteratureSlide8

Literature: Herding Behaviors

LiteratureClement and Tse (2003, 2005)

- find that analyst characteristics are associated with forecast accuracy for both bold and herding forecastsGleason and Lee (2003) - bold forecast revisions generate stronger return responses than do herding forecast revisionsSlide9

Literature: Herding Behaviors and Extreme Earnings Surprises Trueman (1994, p. 98)

When herding is observed among analysts:Extreme surprise announcements of actual earnings => smaller stock price reactions than would be expected from analysts using information in an an unbiased manner to produce their forecasts.

Reasoning: Investors recognize the possibility that an analyst actually had information that justified a more extreme forecast. Consequently, large earnings “surprises” do not surprise investors as much as if they had taken the announced forecast at face value.Slide10

Literature: Influence of Regulation FD

Findlay and Mathew (2006)After implementation in October 2000 of US. SEC Regulation Fair Disclosure (Reg. FD), analyst forecast accuracy declines overall, but analysts that were less accurate (more accurate) before Reg. FD improve (become less accurate) after its implementationBailey, Li, Mao, and Zhong (2003) find that after the passage of Regulation FD, analyst forecast dispersion increases

LiteratureSlide11

Research Motivation

How do analysts revise their estimates conditional on management earnings guidance releases?How do following analysts revise their estimates when faster analysts revise their estimates following management earnings guidance?Does Regulation Fair Disclosure change the way analysts revise their estimates?Do analyst herding behaviors result in market reactions to actual earnings announcements as predicted by Trueman (1994)?

Research MotivationIntroductionSlide12

Data and Methodology

Data and Methodology

First Call (1995 – 2009)Annual earnings per share (EPS) management guidance

Omit guidance events that occur after the fiscal year end but before the earnings announcement

Mid-point of the earnings guidance if given as a range

Omit observations without CUSIP

Neutral/Negative/Positive Surprises:

t

he guidance qualifies as a positive (negative) surprise when it is higher (lower) than the current

expectation

based on

the last analyst’s consensus update prior to the guidance

Examine

analysts

forecast revisions issued within 7 days of management guidance

Slide13

Table 1 – Comparison of Management Guidance Events for Annual v. Quarterly Earnings

Year

ANNUAL

QUARTERLY

Negative

Neutral

Positive

Total

Negative

Neutral

Positive

Total

1995

29

321

15

365

231

407

33

690

1996

76

406

25

507

538

522

88

1,183

1997

121

573

50

744

775

756

146

1,681

1998

314

978

112

1,404

1,210

1,022

217

2,843

1999

358

990

114

1,461

1,067

721

239

3,143

2000

415

1,105

187

1,707

1,251

1,026

350

3,360

2001

855

2,246

509

3,609

2,482

2,246

595

6,260

2002

722

3,022

760

4,502

1,719

2,517

830

5,880

2003

766

3,404

696

4,866

1,568

2,341

653

5,095

2004

876

3,930

812

5,617

1,588

2,449

892

5,432

2005

860

3,876

733

5,469

1,629

2,036

684

4,651

2006

1,047

3,864

788

5,696

1,621

1,853

732

4,574

2007

732

2,898

1,770

5,389

1,224

1,716

777

3,856

2008

31

260

4,985

5,276

121

711

2,479

3,311

2009

25

178

3,847

4,050

110

482

2,065

2,657

Total

7,277

28,051

15,403

50,662

17,134

20,805

10,780

54,616Slide14

Measuring Mimicking and Herding

The Sequence of Management Annual Earnings Guidance and Analyst’s Estimates

Analyst’s Consensus Number

(PRIOR)

First Analyst’s Consensus Number

(FIRST)

Mimicking occurs when the amount is within 1 cent of management guidance

Management Announces Annual Earnings Guidance

Actual Annual Earnings Is Released

(GUI

)

(ACT

)

Last Analyst’s Consensus Number

(LAST)

$3.50

$3.38

$3.40

$3.40

$3.42

Herding occurs when the subsequent analysts issue forecasts within 1 cent of the first

$3.39Slide15

Clement and Tse (2005, JF) and

Gleason and Lee (2003, TAR) MeasurementSlide16

Hypotheses Related to Mimicking

Hypotheses

H1-1: Analysts are less likely to mimic management’s earnings guidance for firms which ultimately experience losses than for other firms.H1-2: Analysts are less likely to mimic management’s earnings forecasts when following firms with higher potential litigation costs than for firms without this potential.H1-3: Analysts are less likely to mimic management’s earnings guidance for growth

firms than for other firms.H1-4: Analysts are less likely to mimic management’s earnings

guidance

after the implementation of Regulation Fair Disclosure than they were before the implementation. Slide17

Hypotheses Related to Herding

HypothesesH2-1: Following analysts are less likely to herd if the first analyst mimics management’s earnings guidance.

H2-2, 3, and 4: Analysts are less likely to herd together for-2 loss firms, -3 high litigation firms, -4 growth firms

H2-5: Analysts are more likely to herd after the implementation of Regulation Fair Disclosure than before.Slide18

Hypothesis Related to Herding and Market Reaction to Earnings Announcements

H3: Under extreme surprise outcomes from actual earnings announcements, the overall market reaction is less intense if analysts previously herded their forecasts following the management earnings guidance than it is when herding was not observed among analysts in response to management earnings guidance.Slide19

Empirical Results

Empirical Results

A

nalyst’s

revision dates after the release

of

Management Earnings

Guidance

(1995-2009)Slide20

Empirical Results

Empirical Results

Figure 3A:

Earnings difference

between analyst’s estimates and management earnings guidance

(1995-2009)Slide21

Empirical Results

Empirical Results

Figure 3B:

Earnings difference

between management earnings guidance and actual earnings

(1995-2009)Slide22

Descriptive Statistics (n = 67,595)Slide23

Logit

regression of analysts’ mimicking reaction to management annual earnings guidance (1995-2009)

Empirical ResultsSlide24

Table 6 - Logit regression of following analysts’ herding reaction to first analyst’s estimates after management annual earnings guidance (1995-2009)

Empirical ResultsSlide25

Empirical Results

Empirical Results

Model 3 -

Linear regression of market reactions to actual earnings announcement for extreme earnings

surprise of 10% or more

(1995-2009)Slide26

Table 7 - Linear regression of market reactions to actual earnings announcement for extreme earnings surprise (1995-2009)

Empirical ResultsSlide27

Conclusion

Analysts are less likely to mimic their estimateshigher litigation potential firmsgrowth firmsloss firmsAfter Regulation FD

Analysts are less likely to herd their estimatesWhen the first analyst mimicsBefore Regulation FDConfirmation of Trueman (1994) theory: Market reacts less intensively to actual earnings announcements of extreme (top and bottom 10%) surprises after observing analysts’ prior herding behavior.