The Effects of Charges on Price and Analyst Forecast Accuracy

A dissertation submitted to the Kent State University Graduate School of Management in partial fulfillment of the requirements for the degree of Doctor of Philosophy

by

Mary Hilston Keener

May 2007

ACKNOWLEDGEMENTS

I wish to thank my dissertation committee of Dr. Linda Zucca, and Dr. Michael Hu, and especially Dr. Ran Barniv, my committee chair, for their support and guidance in completing this dissertation. Also, I extend my thanks to Dr. Michael Ellis and Dr. Richard Kolbe who served as the graduate faculty representative and the moderator, respectively, during the final defense of this dissertation.

I also wish to express gratitude to my parents, Tom and Jane Hilston, for their love, support, and encouragement throughout my life and in particular during my PhD program. In addition, I wish to thank my in-laws, Bob and Althea Keener, for their support as I have worked on my dissertation. I wish to thank my older brother, John, his wife, Dana, my younger brother, Chuck, and the rest of my family and friends for being extremely supportive and caring throughout my many years in school.

I would like to extend special thanks to my husband, Jason, and my son, Peyton Robert. Peyton, thanks for understanding how busy Mommy has been since you arrived in April. Jason, without your endless encouragement, kindness, and love, I would never been able to complete this dissertation. This dissertation is dedicated to Jason and Peyton.

ii

TABLE OF CONTENTS

Page

Chapter 1 INTRODUCTION..………………………………………………... 1

1.1 Background and Purpose…………………………. 1

1.2 Types of ………………………….. 3

1.3 Hypotheses and Models………………………….. 3

1.4 The Importance of This Dissertation……………. 5

1.5 Summary…………………………………………. 6

Chapter 2 LITERATURE REVIEW………………………………………... 8

2.1 Introduction………………………………………. 8

2.2 (Essay 1) and Literature…………………………………………. 16

2.3 (Essay 2) Value Relevance of Restructuring Charges 21

2.4 (Essay 3) Analyst Forecast Literature……………. 29

2.5 Summary………………………………………… 38

Chapter 3 ESSAY 1: DIFFERENCES IN FINANCIAL HEALTH FOR FIRMS TAKING RESTRUCTURING CHARGES

3.1 Background……………………………………. 41

3.2 Importance of Examining Financial Health…….. 42

3.3 Hypothesis Development……………………….. 43

3.4 Data Sources and Methodology………………… 43

iii

3.5 Empirical Results……………………………….. 51

3.6 Conclusions……………………………………... 57

3.7 Bibliography……………………………………. 59

Tables…………………………………………... 61

Chapter 4 ESSAY 2: THE VALUE RELEVANCE OF RESTRUCTURING CHARGES FOR FIRMS WITH VARYING LEVELS OF FINANCIAL HEALTH

4.1 Introduction…….………………………………. 81

4.2 Importance of Restructuring Charges…….…….. 81

4.3 Hypothesis Development……………………….. 84

4.4 Data Sources and Methodology………………… 85

4.5 Empirical Results……………………………….. 90

4.6 Conclusions……………………………………... 98

Bibliography……………………………………. 100

Tables…………………………………………… 103

Chapter 5 ESSAY 3: THE IMPACT OF RESTRUCTURING CHARGES ON ANALYST FORECAST ACCURACY, BIAS, AND REVISIONS

5.1 Background……………………………………… 116

5.2 Impact of Restructuring Charges on Analyst Forecasts…….…………………………………. 118

5.3 Hypotheses Development……………………….. 122

5.4 Data Sources and Methodology………………… 123

iv

5.5 Empirical Results……………………………….. 134

5.6 Conclusions……………………………………... 143

Bibliography……………………………………. 145

Tables…………………………………………… 150

Chapter 6 CONCLUSIONS………………………………………………… 169

6.1 Background, Purpose and Results……………… 169

6.2 Limitations……………………………………… 171

6.3 Future Research………………………………… 172

BIBLIOGRAPHY……………………………………………….. 173

v

LIST OF TABLES

Essay 1 Tables

1 Final Sample Determination…………………………………………………………...... 61 2 Firms in Sample Split by Year and SIC Division……………………………….…….……… 62 3 Descriptive Statistics for Full Sample....………..……………………………………..…...... 64 4 Distress Classification Contingency Tables…………………………………………………... 66 5 Descriptive Statistics for Observations Classified the Same by Both Models..…………...... 67 6 Independent Sample t-tests and Wilcoxon Z-statistics for Mean Differences Between Sample Firms …………...... 69 7 Independent Sample t-tests and Wilcoxon Z-statistics for Mean Differences Between Sample Firm-Events …………...... 71 8 Ohlson (1980) Logistic Regression Model Results for Distressed and Non-Distressed Companies for All Firms……………………….…………………………………...... 73 9 Ohlson (1980) Logistic Regression Model Results for Distressed and Non-Distressed Companies for All Firm-Event Observations…………….…….………………………. 74 10 Predicting Bankruptcy Using Equation 1.4 for Those Restructuring Firms Classified as Distressed………………………..……………………………………………………… 75 11 Predicting Bankruptcy Using Equation 1.4 for Those Restructuring Firm-Event Observations Classified as Distressed………………….….…………………………… 76 12 Predicting Bankruptcy Using Equation 1.4 for All Restructuring Firms……………………..77 13 Predicting Bankruptcy Using Equation 1.4 for All Restructuring Firm-Event Observations.. 78 14 Modified Ohlson Regression Model Results for Multinomial Logistic Regression Model for All Firm-Event Observations with Different Classification Groups...... … 79

Essay 2 Tables

1 Final Sample Determination…………………………………...….…………………...... ….. 104 2 Firms in Sample Split by SIC Division ……...…………………………………..………….. 105 3 Descriptive Statistics………………...... ………..……………………………...…….……… 106 4 Price and Return Model Regressions for Firms……………………………………………... 108 5 Price and Return Model Regressions for Firm-Event Observations….……………………... 110 6 Regressions of Three Versions of the Price Model for Firms………………………………. 112 7 Regressions of Three Versions of the Return Model for Firms………………..……………. 114

Essay 3 Tables

1 Final Sample Determination…………………………………………...... …...……….. 151 2 Firms in Sample Split by SIC Division ……...…………………………………..………….. 153 3 Descriptive Statistics for Full Sample....………..…………………………………………… 154 4 Descriptive Statistics for the Analyst Forecast Accuracy and Bias Portion of the Essay..….. 155

vi

5 Univariate Data on Forecast Errors for the Year Before and the Year After Restructuring Charges..………………………………………...………………………………..…… 156 6 Regressions of Equation (3.5) where h=1 with Firm-Quarter Observations ...... … 157 7 Regressions of Equation (3.5) where h=2 with Firm-Quarter Observations ...... … 159 8 Regressions of Equation (3.5) where h=3 with Firm-Quarter Observations ...... … 160 9 Regressions of Equation (3.5) where h=5 with Firm-Quarter Observations ...... … 161 10 Regressions of Equation (3.7) for All Firm-Year Observations: Accuracy ...... … 162 11 Regressions of Equation (3.8) for All Firm-Year Observations: Bias ...... … 164 12 Regressions of Equation (3.8) for All Firm-Year Observations: Positive Bias ...... … 166 13 Regressions of Equation (3.8) for All Firm-Year Observations: Negative Bias ...... … 168

vii

CHAPTER 1

INTRODUCTION

1.1 Background and Purpose

This dissertation presents three essays that examine the usefulness of reported operational restructurings to corporations, financial analysts, and investors. The main theme that binds the

three essays is that, depending on the financial health of the firm taking the charge, each represents an empirical study of how restructuring charges differentially impact various user groups. While each of the three essays can be read independently, together they address specific

significant issues in the more general area of the effects of restructuring charges on firms.

Because restructurings have become “a staple of management life during the past decade”

(Bowman et al. 1999), this dissertation is quite relevant to the accounting profession. In just the first three months of 2006, many major companies including Hewlett-Packard, Ford, Kraft, Del

Monte, and Xerox have announced operational restructuring efforts, and this attests to the timeliness of the findings in this dissertation (Lawton 2006, McCracken and White 2006, Ellison

2006, Berman 2006, Bulkeley 2006).

Daniels et al. (1995) define a restructuring as typically involving a firm-level action where personnel are terminated, product lines are eliminated, and assets are disposed of. By examining the effects of restructuring announcements on analyst forecasts and the impact of restructuring charges on price, this dissertation sheds some light on the usefulness of operational

1 2

restructurings to firms. Kross et al. (2001) define operational restructurings as typically involving the asset side of the balance sheet and and restructurings as typically involving the right-hand side of the balance sheet. This dissertation examines only operational restructurings.

Operational restructurings are multi-dimensional corporate changes undertaken by companies usually to either improve efficiency or to avoid filing for bankruptcy. Operational restructuring projects typically include some combination of workforce reductions, asset writedowns, the disposal of certain assets and facilities, product line discontinuations, the reconfiguration of facilities, plant relocations, or the closing of certain plants and facilities (Lopez 2002).1

This dissertation separately examines several groups of firms. In particular, it compares firms in financial distress that operationally restructure in an attempt to avoid bankruptcy with financially healthy firms that restructure for improving efficiency. The sample for this dissertation contains firms undertaking operational restructuring efforts during the period 1993-

2003. While operational restructuring has not been linked to bankruptcy in prior literature, Herz et al. (1992) study debt restructuring as a means to avoid bankruptcy. Atiase et al. (2004) examine whether restructuring charges are associated with improved performance in earnings and cash flows. This dissertation also examines several groups of restructuring firms to determine the value relevance of restructuring charge information and the probability of financial distress.

Finally, the dissertation examines the differential impact of the magnitude of restructuring charges announcements by financially healthy firms and distressed firms on analysts’ forecast revisions, accuracy, and bias.

Many studies discuss the frequency of restructuring charges. For example, Khurana and

Lippincott (2000) document how widespread restructuring charges were during the nineties, citing that only four of the thirty Dow Jones Industrial Companies did not take a restructuring

1 EITF 94-3 (1995) provided three requirements for companies to follow when reporting restructuring charges. It also defines the different types of restructuring costs.

3

charge between 1991 and 1995. Also, the authors cite surveys indicating that restructuring charges are valuable to shareholders. The financial press has reported a large number of corporate restructurings while targeting restructuring charges as “muddying a company’s earnings picture” and creating confusion (Smith and Lipin 1996). Harlan (1994) states that the market’s positive reaction to restructuring charges may be one reason for their increased popularity with corporations. In addition, the financial press frequently provides announcements of operational restructuring efforts being taken by large corporations (Landers 2002; Martin 2002).

1.2 Types of Restructurings

There are several types of restructurings, but this dissertation focuses primarily on operational restructurings. Operational restructurings are quite different from equity restructurings, quasi-reorganizations, and troubled-debt restructurings. Usually to either improve efficiency or to avoid filing for bankruptcy, operational restructurings are multi-dimensional corporate changes undertaken by companies. Equity restructurings involve the altering of ownership via spin-offs, split-offs, or equity carve-outs (Kross et al. 2001). Quasi- reorganization occurs when a corporation reorganizes to eliminate its accumulated retained earnings deficit (Herz et al. 1992). Troubled-debt restructuring includes modification of debt covenants, the exchange of other assets for debt, and the exchange of equity for debt (Pirrong and

Koeppen 1993).

1.3 Hypotheses and Models

This dissertation examines several research hypotheses. The first essay, “Differences in

Financial Health for Firms Taking Restructuring Charges,” examines whether certain financial indicators can be used to predict how successful restructuring efforts will be for distressed

4

companies. Brickley and Van Drunen (1990) and Bens (2002) determine that companies utilize operational restructurings as last attempts to avoid bankruptcy. For restructuring companies, this essay hypothesizes that financially distressed firms differ from healthy firms and companies that subsequently file for bankruptcy within three years differ from other financially distressed restructured firms that are able to avoid bankruptcy following a successful restructuring.

The second essay, “The Value Relevance of Restructuring Charges for Firms with

Varying Levels of Financial Health,” examines the value relevance of restructuring charges for

several different groups of restructuring firms. Studies including Khurana and Lippincott (2000)

and Bens (2002) find that restructuring charges are value relevant, and this dissertation further

examines the value relevance of these charges for companies that are restructuring for different

reasons. The hypotheses state that restructuring charges are value relevant and that the value

relevance of restructuring charge information is smaller for non-distressed firms, greater for

distressed firms that file for bankruptcy within three years of restructuring, and greatest for

financially distressed firms that avoid filing for bankruptcy during the three years following the

restructuring. Further, the essay explicitly examines the impact of restructuring cost and the

probability of financial distress on stock prices and returns.

The third essay, “The Impact of Restructuring Charges on Analyst Forecast Accuracy,

Bias, and Revisions,” separately examines the effects of restructuring charge announcements on

analysts’ forecast revisions, accuracy and bias for financially healthy firms and for firms that are

identified as being in financial distress. Although Chaney et al. (1999) examine the effect of

restructuring charges on analysts’ forecast revisions and errors, the authors do not examine how

the effects of restructuring on analysts’ forecasts may be different for firms restructuring for

different reasons. Finally, the essay examines the impact of the restructuring cost and the

probability of financial distress on forecast revisions, forecast accuracy, and bias.

5

1.4 The Importance of This Dissertation

This dissertation provides several incremental contributions to the extant literature. First, although restructuring charges have been examined in prior research, past studies have not focused on examining restructuring charges separately for healthy companies restructuring to improve their efficiency and financially distressed companies restructuring to avoid potential bankruptcy. Therefore, this dissertation contributes to the literature by helping to close the gap between the current restructuring and bankruptcy literatures. Second, this dissertation contributes to the existing value relevance literature by examining the impact of restructuring charges for companies that are restructuring for different reasons on prices and returns. Third, the results of this dissertation contribute to the literature examining analysts’ reactions to restructuring charges by providing evidence on the impact of restructuring charges on analyst forecasts separately for financially healthy firms and for several groups of distressed firms. The primary incremental contribution arises because past research has failed to consider the differences in the primary reasons why companies restructure, where these reasons are generally either to improve efficiency or to avoid further financial distress or even bankruptcy. This dissertation considers the impact that the company’s reason for restructuring has on price and returns in Essay 2 and on analyst forecast revisions, accuracy, and bias in Essay 3. By adding an indicator variable indicating the company’s reason for restructuring to the models in Essays 2 and 3, I am able to see the impact that a company’s reason for restructuring has on its stock price and returns and also the impact of this motivation on analyst forecasts.

The results of this dissertation have implications for different groups. First, the results can help analysts and investors to determine why companies restructure and how successful restructuring efforts usually are. Second, the results provide some implications to managers of

6

companies that are considering undertaking operational restructurings. Third, the results of this dissertation have implications for standard setters and regulators who will be able to use the results to develop a more relevant standard relating to restructuring charges. To provide investors with more relevant information, standard setters and regulators can utilize the results for examining companies’ motivations for restructuring. If it is determined that restructuring efforts often result in subsequent bankruptcy for companies in financial distress, then companies may be required to disclose their reasons for undergoing operational restructurings.

Fourth, this dissertation has implications for analysts and investors because reported restructuring charges should provide investors and analysts with more information for determining the impact of restructuring on firm value. Finally, the results are likely to provide financial analysts with additional information that they can use in determining how to interpret an announcement of operational restructuring. Analysts may be able to use the financial health of a company to determine how the announcement of a restructuring charge should affect a forecast.

1.5 Summary

The remainder of the dissertation is organized as follows. Chapter 2 discusses the extant relevant literature. Chapter 2 is composed of several sections. The first section provides a background of the relevant literature. The second section, which relates to Essay 1, presents literature that involves the identification of firms in financial distress. The third section, which relates to Essay 2, outlines papers that discuss the many issues related to the value relevance of restructuring charges. The fourth section, which relates to Essay 3, discusses studies about the impact of the announcement of restructuring on analysts’ forecast accuracy and bias.

Chapter 3 presents the first essay, which relates to the financial health of firms taking restructuring charges. Chapter 4 presents the second essay, which relates to the value relevance

7

of restructuring charges. Chapter 5 presents the third essay, which relates to the effects of restructuring charge announcements on analyst forecast revisions, accuracy, and bias. Each essay presents the methodological discussion and provides the models that are utilized to test the hypotheses. Chapter 6 provides the conclusions obtained from this dissertation. While the three essays are presented independently, both the introduction and summary chapters relate the three essays to each other and address significant issues in the area of operational restructuring.

CHAPTER 2

LITERATURE REVIEW

2.1 Introduction

2.1.1 Background

The purpose of this chapter is to provide a discussion of the relevant accounting literature examining operational restructurings and the market’s response to these efforts. This chapter is separated into four primary sections. Section 2.1 provides a background and summary of the general restructuring literature. Section 2.2, which relates to Essay 1, presents literature that involves the identification of firms in financial distress. Section 2.3, which relates to Essay 2, outlines papers that discuss the many issues related to the value relevance of restructuring charges. Section 2.4, which relates to Essay 3, discusses studies about the impact of the announcement of restructuring on analysts’ forecast accuracy and bias.

2.1.2 Restructuring: Some Related Issues

Restructuring has been defined as a collection of activities designed to increase shareholder wealth by maximizing the value of corporate assets (Halperin and Bell 1992). A restructuring usually involves a reorganization of the firm where personnel are terminated and assets are sold off or scrapped (Daniels et al. 1995). Lai and Sudarsanam (1997) find that , ownership structure, and corporate governance influence firms’ decisions about the types of activities that comprise their restructuring plans. Frederikslust et al. (2003) find that firms which are declining in performance choose many different strategies for recovery and that these varied strategies have different implications for managers, shareholders, and lenders.

8 9

Corporate restructuring has become a common event in the professional lives of managers. Since 1980, more than a trillion US dollars have been spent by companies to restructuring, either in or out of court (Gilson 1998). Many successful companies undergo frequent restructurings to improve their overall efficiency (Herz and Abahoonie 1992). Many studies also examine firms restructuring in response to financial distress (Berger and Ofek 1999,

Atiase et al. 2004, Faccio and Sengupta 2006). Donaldson (1990) and Dial and Murphy (1995) present case studies on the restructuring activities of two large firms, General Mills and General

Dynamics, respectively. Radcliffe et al. (2001) conduct a case study of three large corporate restructurings to identify the role of management in the restructuring process.

Herz and Abahoonie (1991, 1992) discuss issues related to corporate restructurings in several studies. The authors demonstrate that the major goals of restructuring are to accurately reflect the dollar amounts involved in financial transactions and to optimize values in corporate business plans (Herz and Abahoonie 1992). Although restructuring was once considered nothing short of a major reworking of a firm’s to relieve the strain of and principal payments (Herz and Abahoonie 1991), it has become much more common for firms to restructure simply to improve their operating efficiency.

Hoskisson and Turk (1990) examine corporate restructuring caused by threats and suggest that the poor corporate monitoring caused by inadequate board of director governance may lead to higher levels of diversification.1 Markides (1992) discusses how many

firms reduced their diversification by refocusing during the 1980s.2 Bethel and Liebeskind

1 Hoskisson and Turk (1990) also find that if corporate diversification results in the loss of strategic control and poor performance, then the threat of takeover is likely to be related to the incidence of corporate restructuring. Corporate restructurings are likely to correct inadequate governance patterns, create a more focused diversification strategy, increase strategic control, reduce reliance on bureaucratic control through reduced staff size, and increase the performance of the firm and shareholder wealth. 2 Markides (1992) examines whether this refocusing created market value for the companies involved. The author demonstrates that refocusing announcements are associated with significant, positive abnormal returns, which implies that the firms were more diversified than was necessary.

10

(1993) examine the relationship between ownership structuring and restructuring, and they find that blockholder ownership is associated with corporate restructuring. This suggests that many

companies only restructure when pressured to do so by large shareholders. Zantout (1994)

suggests that downsizing, usually included in a restructuring plan, is often a defensive strategy

used to avoid hostile takeover, not a method used to improve efficiency. By way of illustration,

Jennings et al. (1998) focus on the equity implications that result from restructurings.

Various studies examine the manipulation of restructuring costs to manage earnings.

Pulliam and Berton (1994) discuss the increased concern of the Securities and Exchange

Commission (SEC) about the increased size and frequency of restructuring charges and the

potential use of these charges to manage earnings. The SEC suggested that restructuring charges

should be “legitimate rather than expected costs.” Bronson (1994) finds that when restructuring

charges are actually incurred, firms manipulate their earnings by reducing their restructuring

liability instead of reducing income. In their 2005 working paper, Bens and Johnston attempt to

determine whether discretionary restructuring charges are associated with earnings management,

other economic explanations, or both.3 The financial press and academics suggest that in order to manage future earnings firms systematically overstate restructuring charges. The results of this study demonstrate that restructuring firms make real changes in operations after the charge. The results provide no evidence that the discretionary restructuring charge is related to the proxies for earnings management. The authors find that the discretionary restructuring charges are positively associated with voluntary disclosure and recent changes in upper management. In addition, the authors find that discretionary restructuring charges are negatively associated with leverage.

3 Bens and Johnston (2005) conduct their study using a cross-sectional sample of 247 restructuring charges taken during the period from 1989 to 1992. First, the authors create an expectations model for restructuring charges and use it to identify the residual from this model as the discretionary portion of the accrual. Next, the authors estimate multivariate regressions with the discretionary portion as the dependent variable and several earnings management proxies as explanatory variables.

11

Several studies also examine the impact of executive compensation contracts on restructuring charges. Dechow et al. (1994) discuss the widespread use of earnings-based incentives for executive compensation contracts and their relation to operational restructuring plans. The authors suggest that executives may avoid undertaking restructuring efforts to maximize their current earnings-based compensation, regardless of the impact on the overall well-being of the firm. To prevent these opportunistic activities, the compensation plans are overseen by compensation committees that can adjust compensation to prevent executives from getting involved in opportunistic behavior. Dechow et al. (1994) provide evidence suggesting that compensation committees adjust earnings-based incentive compensation and help to ensure that executives are not deterred from undertaking valuable operational restructuring efforts. Adut et al. (2003) examine the relationship between restructuring charges and CEO cash compensation.

The authors determine that compensation committees appear to completely shield initial and subsequent restructuring charges for CEOs with long tenure, if the firm had not recorded a charge in the two years immediately prior to the restructuring. The results of this study suggest that compensation committees evaluate the context of a restructuring charge before determining the extent to which the committee shields executive compensation from the effect of these charges.

The extant body of restructuring literature includes several analytical studies related to corporate restructuring efforts. Shelley at al. (1998) discuss agency theory and relate it to restructuring. Agency theory suggests that the separation of ownership and management in corporations is Pareto optimal, despite the potential for conflicts of interest, if the corporate governance mechanism adequately safeguards shareholders’ collective interests. The authors suggest that the corporate restructuring activity that occurred frequently during the 1980s is thought by many observers to be a rational response to the breakdown of the firms’ internal control mechanisms to which the diversification mistakes of the 1960s and 1970s are attributed.

12

The goals of the restructuring efforts are usually to shift strategy, to infuse the company with new technology, to exploit the relatedness of the business units, to make follow-up acquisitions to build a critical mass, to sell off unneeded units, and to make efficient use of cash and leverage.

After restructuring occurs, the company should be more efficient in its current operations, more effective in managing risks, and more adaptive to its environment. The authors determine that voluntary corporate restructuring efforts are used as pre-emptive measures against external capital market intervention. Shelley et al. (1998) also determine that the market values the non-tax benefits and the costs that are associated with restructurings. Frantz (1999) introduces an analytical model that attempts to explain the discretionary write-downs, write-offs, and other provisions involved in restructuring. The model is comprised of a firm that is about to be restructured, a manager, and a financial market. The manager in this model has private information about the likelihood of success of restructuring efforts, and the manager may recognize all or a portion of the expenditure by reporting a discretionary restructuring provision.

The manager makes the decision whether or not to report a provision, thereby recognizing the impact that restructuring may ultimately have on compensation. The paper demonstrates how the manager can communicate his or her private information to investors through the provisions that he releases or decides not to release.

Denis and Kruse (2000) examine the incidence of corporate restructurings and managerial control-reducing disciplinary events including and board dismissals among

firms that have experienced a large decline in operating performance during a takeover period and

a less-active period. The authors determine that although some managerial disciplinary events are

related to overall takeover activity, the decline in takeover activity has not precluded

performance-enhancing restructurings subsequent to performance declines.

13

Atiase et al. (2004) determine whether restructuring charges are associated with improved performance. The authors examine a sample of firms undergoing operational restructurings from

1991-1993, and they find that the restructuring firms’ earnings increase compared to the earnings levels immediately before restructuring. However, the authors find that this result may be largely caused by firms with multiple restructurings and by firms reporting losses in the restructuring year. The authors find no association between restructuring charges and post-restructuring changes in earnings compared to the year before restructuring. Atiase et al. (2004) conclude that restructuring charges are associated with improved earnings but not necessarily with improved operating performance.

The recent accounting literature discusses a wide array of restructuring-related topics.

Muehlberger (2005) examines the impact of restructuring on organizational governance. Covitz et al. (2006) test the assumption that longer restructurings are more costly. Holmstrom (2006) examines the impact that restructurings have had on shareholder value. Renneboog and Szilagyi

(2006) provide an exhaustive overview of the extant research on how corporate restructuring impacts wealth. Many recent studies have examined international issues related to restructuring, focusing on the impact of restructuring in specific countries including Japan (Kim and Mody 2003), China (Kam et al. 2006), the UK (Hillier and McColgan 2005), or restructuring issues in many different countries (Bartolinli 1995, Couke et al. 2006).

The articles in this section examine some important issues related to restructuring. Herz and Abahoonie (1992) discuss the tax and accounting implications of firms restructuring to avoid bankruptcy and suggest that careful planning should be utilized by these companies to maximize the use of net operating loss carryovers. Dechow et al. (1994) and Adut et al. (2003) examine the relationship of operational restructuring plans to the use of earnings-based incentives for executive compensation contracts. Dechow et al. (1994) find that compensation committees

14

override certain provisions in executive incentive plans to avoid having executives behave opportunistically by not undertaking potentially value-enhancing restructurings. Adut et al.

(2003) determine that compensation committees evaluate each restructuring charge to establish the extent to which they will have to shield executive compensation from the effect of the charges. Denis and Kruse (2000) discuss the relationship between corporate takeovers and restructuring activities, and they determine that a decrease in takeover activity does not cause fewer restructurings after performance declines. Bens and Johnston (2005) examine the use of discretionary restructuring charges in operations management and find that restructuring charges are not used to manage earnings as often as generally portrayed. Companies often restructure either when threatened with external market intervention (Shelley et al. 1998) and in response to pressure from shareholders (Frantz 1999). Although this dissertation does not contribute directly to any of the lines of research discussed in this section, the presentation of these issues provides a framework for the types of research that have been done on restructuring issues.

2.1.3 Restructuring in the Professional Accounting Literature

Although restructuring charges have become common for many companies, the rules for reporting restructuring issues remain rather vague when it comes to handling specific scenarios.

APB Opinion No. 30 (1973) required that restructuring charges be included in income from operations. Emerging Issues Task Force (EITF) No. 86-22 examines how companies should classify the gain or loss that results from restructuring in the income statement.4 In December

1986, the SEC released Staff Accounting Bulletin (SAB) No. 67, which specifies that

4 The EITF was not able to reach a consensus on this issue and decided not to recommend any special income statement format and that management should be responsible for choosing the most meaningful income statement presentation. The EITF agreed that restructuring is most frequently associated with operations, but they also agreed that there is not necessarily one best location for disclosing restructuring information on all statements.

15

restructuring charges should be presented as a component of income from continuing operations and separately disclosed when material. In EITF 87-4, the Task Force decided that the requirements of SAB No. 67 are not required for nonpublic companies.5 Because there was little accounting guidance specifically for restructurings, most firms used FASB Statement 5,

Accounting for Contingencies, and SEC Staff Accounting Bulletin No. 67 for authoritative support (Daniels et al. 1995).

Between March 1994 and January 1995 the EITF discussed the recognition of liabilities for employee termination benefits and other costs to exit an activity including costs involved in a restructuring. In January 1995, the EITF released EITF 94-3, which provides three requirements for companies to follow when reporting restructuring charges.6 Lopez (2002) assesses whether the reporting of material components of a restructuring charge mandated by EITF 94-3 provides information to financial statement users beyond the information contained in the aggregate charge. The primary findings of the author suggest that the components of restructuring charges are informative. The author also indicates that analysts interpret certain components of restructuring charges across forecast horizons and other periods differently. Overall, Lopez

(2002) suggests that the EITF was justified in releasing statement 94-3, which mandates that the material components of restructuring charges be disclosed. SAB No. 100, released in November

1999, states that the costs and charges falling in the scope of EITF Issue 94-3 must be accounted for in accordance with the appropriate standard.7 The primary types of abuses that EITF Issue

5 As a result of the EITF’s discussions, the SEC later clarified that SAB No. 67 is not intended to apply to gains or losses that result from the simple sale of assets or of a portion of a business segment and that the SEC views restructuring charges as operating expenses. Also, the SEC subsequently required companies to reclassify incorrectly recorded charges from prior years. 6 First, firms must record the costs of restructuring during the period in which management commits to the plan. Second, costs classified as restructuring charges must provide no future benefit to the firm over and above the restructuring execution. Third, EITF 94-3 requires the disclosure of many details about the restructuring plan. EITF 94-3 was later nullified by the FASB’s release of SFAS No. 146 in June 2002. 7 The SEC released this statement primarily to crack down on companies who were using restructuring charges to take “big baths” to manage their earnings.

16

94-3 was designed to prevent are taking as current expenses the future costs of computers or software expected to take the place of terminated employees, charging expected future advertising costs attributed to the restructuring, and accruing costs of terminating employees because of future acquisitions that had not yet occurred (Pulliam and Berton 1994).

FASB Statement No. 146, released in June 2002, addresses the accounting and reporting for costs related to exit or disposal activities, and it nullifies EITF Issue 94-3. SFAS No. 146 improves financial reporting over EITF Issue 94-3 by requiring that a liability be recorded for costs related to exit or disposal activities and that this liability will be measured at fair value only when the liability is incurred. SFAS No. 146 helps to improve the comparability and representational faithfulness of financial information because the treatment for liabilities in this case is the same as other similar events. The preceding paragraphs examine the accounting pronouncements that have been promulgated in the restructuring area. This dissertation provides accounting standard-setting bodies with relevant information that can be used as restructuring standards are created in the future.

2.2 (Essay 1) Financial Distress and Bankruptcy Literature

Because the primary focus of this study is to examine the differences in analyst reactions to restructuring charges and the value relevance of restructuring charges for healthy and distressed firms, it is extremely important to use care in identifying financially distressed firms.

Studies have used various methods to determine whether a firm is distressed. A brief discussion of the bankruptcy literature follows.

One of the first landmark papers in this area is Altman’s (1968) paper. The purpose of this study is to attempt to assess the quality of ratio analysis as an analytical technique for predicting bankruptcy. Specifically, Altman examines a set of financial and economic ratios in

17

the context of bankruptcy prediction using multiple discriminant analysis. Ohlson (1980) re- examines the issue of predicting bankruptcy using financial and economic ratios. Ohlson finds that it is possible to identify four basic factors that are statistically significant in identifying the probability of bankruptcy. The factors identified by Ohlson include the size of the company, a measure of the financial structure, a measure of performance, and a measure of current liquidity.

Ohlson utilizes the methodology of logistic analysis to avoid some problems associated with the discriminant analysis method applied by Altman (1968).8 Ohlson’s (1980) model includes nine

explanatory variables, and six of the nine variables are determined to be significant in predicting

bankruptcy.

Gilbert et al. (1990) determine whether traditional bankruptcy prediction models like

Altman (1968) and Ohlson (1980) are able to distinguish distressed firms filing for bankruptcy

from other distressed firms avoiding filing for bankruptcy. The authors find that a bankruptcy

prediction model developed using a bankrupt/random sample is not able to distinguish firms that

fail from other financially distressed firms. The authors demonstrate that a financial ratio-based

model comprised of distressed firms also performs poorly, which suggests that the resolution of

distress is influenced by other nonfinancial factors. The results of this study further suggest that

cash flow variables add to the explanatory power of bankruptcy prediction models. Finally, the

results indicate that financial variables that show an ability to distinguish between bankrupt and

8 In recent years, accounting studies suggest that logistic regression has become the preferred technique to MDA for several reasons (Barniv and McDonald 1999). First, there are several requirements imposed on the distributional properties of the predictors including the fact that the variance-covariance matrix should be the same for the two groups. Second, the output of MDA is a score that does not provide intuitive information because it is basically an ordinal ranking device that does not designate the likelihood of bankruptcy to each company. Third, the matching criteria used for MDA including size and industry cause problems with matching. Ohlson (1980) concludes that using logistic analysis avoids all of these problems associated with the use of MDA. Webster and Coyne (1991) compare the multiple discriminant analysis method used by Altman (1968) to the logistic regression analysis used by Ohlson (1980) to determine if one technique is preferable to the other in predicting small firm bankruptcy. The findings indicate that neither method results in lower type I errors.

18

distressed firms are different from those that discriminate between bankrupt and non-bankrupt firms.

Begley et al. (1996) examine Altman’s (1968) model and Ohlson’s (1980) model and determine that the models do not perform as well in more recent periods, even when the coefficients are updated. Begley et al. (1996) re-estimate both Altman’s and Ohlson’s models

using data from the 1980s. The authors find that when they compare the original models with

their re-estimated models, Ohlson’s original model displays the strongest overall performance.

The authors’ results support the use of the Ohlson (1980) model as the preferred bankruptcy

prediction model.

Grice and Dugan (2001) examine the potential problems related to the use of bankruptcy

prediction models in current research, and they find that the majority of problems arise when the

bankruptcy prediction models are not applied correctly. Using the Ohlson (1980) model and the

Zmijewski (1984) model, the authors determine that researchers should be very careful when

using bankruptcy prediction models. Particularly when the models are applied to time periods

and industries other than those used to develop the models, problems may arise. Also, various

bankruptcy prediction models may be more useful for evaluating various forms of financial distress other than just bankruptcy. The authors suggest that researchers attempting to use established bankruptcy prediction models need to carefully examine the limitations associated with the use of the models.

Grice and Ingram (2001) examine Altman’s (1968) model to determine whether it remains useful for predicting bankruptcy in recent periods. Also, the authors attempt to determine whether the Altman model is as useful for non-manufacturing firms as it is for manufacturing firms. Finally, the authors examine the Altman model to determine whether the

model is as useful for predicting financial stress conditions other than bankruptcy as it is for

19

predicting bankruptcy. The results suggest that the Altman model is not as useful for predicting bankruptcy in recent periods as it was when it was created and that the Altman model is not as useful for non-manufacturing firms as it is for manufacturing firms. The authors are able to find evidence consistent with the conjecture that the Altman model is useful for predicting stress conditions other than bankruptcy. The results of this study further suggest that those relying on

Altman’s (1968) model should re-estimate the model’s coefficients rather than relying on those originally calculated by Altman.

Grice and Ingram (2001) utilize a unique method from those introduced by Altman and

Ohlson for determining which firms are financially distressed. Grice and Ingram (2001) define distressed companies as those reported by COMPUSTAT as meeting one or more of the following conditions: Chapter 11 bankruptcy, Chapter 7 liquidation, bonds vulnerable to , or low stock ratings. Companies with Standard and Poor’s (S&P) ratings of CCC or below

(COMPUSTAT item 280) or whose stock was rated “lower B” or below (COMPUSTAT item

282) were classified as distressed. In this study, the latter method of examining stock and bond ratings is utilized as a method of identifying financially distressed companies.

Astebro and Winter (2003) discuss several methodological issues associated with predicting the outcome of financial distress. The authors are able to arrive at three distinct conclusions. First, the probabilities of survival and acquisition of financially distressed firms are explained differently by the same set of explanatory variables. Second, the authors determine that relative to multinomial models, binary models are misspecified. Finally, the authors determine that industry specific intercepts have greater explanatory power than industry-adjusted regressors.

Hillegeist et al. (2004) examine Altman’s (1968) Z-score and Ohlson’s (1980) O-score to

determine whether these variables summarize publicly available information about the likelihood

of bankruptcy. The authors determine that the traditional utilization of accounting-based

20

measures of bankruptcy risk is inadequate. The authors suggest that the BSM-PB, a market-based bankruptcy prediction measure based on Black and Scholes (1973) and Merton (1974), has relatively more explanatory power than the scores from either the Ohlson (1980) or the Altman

(1968) models. However, the authors also determine that the Z-score and the O-score provide significant, incremental information. Therefore, they suggest that BSM-PB should be used in conjunction with the use of the Altman (1968) and Ohlson (1980) models. Finally, the authors indicate that BSM-PB does not reflect all available market-based information regarding the probability of bankruptcy.

Several recent studies have again addressed the issue of the success rates of commonly used Z-Score prediction models like Altman (1968). Agarwal and Taffler (2006) examine statistical failure prediction models including Altman (1968) to test whether these methodologies actually work in practice. By examining many well-known UK-based z-score models over twenty-five years, the authors determine that older Z-score models including Altman (1968) have high predictive value over the extended time-period they examine. Weiss and Capkun (2006) develop a methodology for incorporating Type I error costs (the amount lost from lending to a firm that goes bankrupt) and Type II error costs (the opportunity cost of not lending to a firm that does not go bankrupt) into bankruptcy prediction models. The results indicate that a lending model only accounting for error costs and firm size will yield higher profits than a model that relies only on the percentage of firms correctly predicted. Janes (2006) examines the relation between accounting accruals and financial distress, and he determines that financial distress occurs much more frequently in companies with extremely high or low accruals. Also, Janes

(2006) demonstrates that extremely high or low accruals are related to financial distress over the long-term.

21

This line of research provides a description of the prediction of bankruptcy using various financial indicators. Altman (1968) and Ohlson (1980) develop useful bankruptcy prediction models that have since been examined and modified extensively by many studies. Various more recent studies have tested Altman’s and Ohlson’s models to determine if they are still relevant.

Begley et al. (1996) examine Altman’s (1968) model and Ohlson’s (1980) model and determine that the models do not perform as well in more recent periods, even when the coefficients are updated. Begley et al. (1996) find that when they compare the original models with their re- estimated models, Ohlson’s original model displays the strongest overall performance. Grice and

Ingram (2001) suggest that those relying on Altman’s (1968) model should re-estimate the model’s coefficients rather than relying on those originally calculated by Altman.

Hillegeist et al. (2004) examine Altman’s (1968) model and Ohlson’s (1980) model to determine whether these models summarize publicly available information about the likelihood of bankruptcy, and they find that the traditional utilization of accounting-based measures of bankruptcy risk is inadequate. Agarwal and Taffler (2006) determine that older bankruptcy prediction models including Altman (1968) have high predictive value over the extended time- period they examine. This essay contributes to this literature by developing a new model that can be used to identify firms in financial distress. The model developed in this study is used to determine whether each sample firm is financially healthy or distressed instead of attempting to predict which firms will file for bankruptcy. The value relevance of restructuring charges and analyst forecast accuracy are examined separately for the firms classified as healthy and distressed in additional essays.

2.3 (Essay 2) Value Relevance of Restructuring Charges

2.3.1 Market Response to Restructuring

22

Many studies examine the stock market’s response to the announcement of an operational restructuring, and the results of these studies are mixed. Some proponents of restructuring argue that leaner, more efficient organizations result from restructuring, while opponents of restructuring suggest that the organizational disruption accompanying restructurings exceeds any benefits from such transactions (Bowman and Singh 1993). Brickley and Van Drunen (1990)

examine 222 internal corporate restructurings made by 179 firms. Internal corporate

restructurings typically involve the splitting or merging of divisions and unit formations and

liquidations. The authors find positive average stock price reactions to restructuring

announcements but a decline in earnings performance in the 3-year period after the restructuring

announcement. Also, the authors find that firms that alter their divisional configurations tend to

increase their shareholders’ wealth as a result of information about investment opportunities and

improvements in efficiency. The contemporaneous decline in earnings that the authors document

is caused by increased expenses, and this result is inconsistent with the idea that the market

pressures managers into focusing on short-term earnings. An examination by the authors of pre-

restructuring performance shows that poorly organized firms are motivated by market pressures

to change their organizations while change occurs in healthy firms as part of the growth process.

Carter (1998) compares a sample of restructuring firms to a sample of similarly performing non-restructuring firms. Using return on assets as a performance measure, Carter

(1998) finds that the market reacts positively to restructurings and that operating performance improves in years three through five after a restructuring. When return on sales is used as a performance measure, she finds no difference between the two groups of firms. Several other studies also find that the market reacts positively to restructuring charge announcements [Martin

23

and Kensinger (1990); Francis et al. (1996), Bunsis (1997), Ballester et al. (1999), and Kross et al. (2001)9].

In contrast, Bens (2002) finds that for a sample of firms announcing restructuring charges

in the fourth quarter from 1990 to 1993, the market reacts negatively to the operational

restructuring announcements.10 Several other studies also find that the market reacts negatively to restructuring charge announcements [Blackwell et al. (1990), Elliott and Hanna (1996), Carter

(2000), Lopez et al. (2002), Poon et al. (2001)]. Bartov et al. (1997) demonstrate that, even in cases where the reaction is statistically significant, the market’s reaction is very small for many prior studies. The mixed findings of these studies demonstrate the difficulty in interpreting the performance and market effects of an operational restructuring.

Other studies in the restructuring charge area examine the market’s response to several components that typically comprise an operational restructuring plan. Blackwell et al. (1990) and

Lin and Rozeff (1993) find negative market reactions to plant-closing announcements. Worrell et al. (1991), Lin and Rozeff (1993) and Elayan et al. (1998) find that the market reacts negatively to announcements of layoffs.11 Francis et al. (1996) determine that the market reacts negatively to inventory write-offs. Lopez (2002) determines that restructurings are multi-dimensional efforts that may require disaggregation into components for a complete understanding of their effect on the market. John et al. (1992) examined firms’ responses to losses, and they determine that firms

9 Francis et al. (1996) suggest that the positive market reaction to restructuring charges is consistent with the view that the greater flexibility in measuring and recognizing restructuring charges allows management to utilize these items to signal information about the company’s expected future performance. Kross et al. (2001) find positive abnormal returns of an average of 1.5 percent in the 2-day period surrounding the restructuring announcement. 10 Bens (2002) finds that restructuring charge announcements are viewed by the market as bad news because the mean three day market adjusted return at the announcement was –3.0%, which is significant. 11 Elayan et al. (1998) document that the market reaction to layoff announcements depends on many other factors including the size of the layoff, the industry of the firm, the information set available to shareholders, and the financial performance of the firm before the announcement. Worrell et al. (1991) determine that announcements of large or permanent layoffs result in stronger negative responses than other announcements.

24

were able to increase their focus and become more efficient after restructuring efforts. Smart and

Waldfogel (1994) utilize a “surprise” variable to determine what would have happened at the restructuring firm in the absence of the restructuring.12

Several studies including Blackwell et al. (1990) and Elayan et al. (1998) demonstrate that there is a slight improvement in return on equity (ROE) in the years subsequent to plant- closing announcements. Elayan et al. (1998) suggest that the increase in ROE after the announcement of layoffs is a signal of the increased efficiency of a firm and its labor force.

Morton and Neill (2001) examine the correlation of market prices with an accounting- based fundamental value after a restructuring has taken place. In the restructuring process, a firm usually accrues operating costs and consolidates them into one charge to income. Although the charge is included in income from operations, the results of operations before these non-recurring costs are discussed in both management disclosures and the financial press. The discussions of operations before restructuring costs have caused some concern that the reporting of these restructuring charges may influence market prices and cause some securities to potentially be mispriced. Using the Feltham-Ohlson model, Morton and Neill (2001) also examine the effect of restructuring charges on firm value based on fundamental accounting value. The authors find that the observed price changes are strongly associated with the estimated change in firm value implied by the accounting-based model. Also, the authors discover that there is a negative relation between the change in value and future accounting returns.

Black et al. (2000) examine a sample of firms that report either single or multiple occurrences of discontinued operations, special items, or extraordinary items over a rolling six- year period from 1977 to 1996. The authors determine that, contrary to prior research, these

12 The Smart and Waldfogel (1994) “surprise” measure is the operating margin and return on equity surprise at restructuring firms, less the operating margin and return on equity surprise at similar non- restructuring firms.

25

unusual items are value relevant and that their finding is consistent with one of several explanations. They hypothesize that multiple occurrences of these items indicate firms in financial distress or multiple occurrences indicate firms whose managers have engaged in repeated attempts at earnings management and that the most recent attempt is being devalued by the market. The authors discover patterns of discretionary accruals consistent with managers engaging in upward earnings management before taking multiple write-downs using special items. Black et al. (2000) also find that single occurrences of special items, including restructuring charges, are value-relevant and are positively correlated with market values.

Moehrle (2002) and Levitt (1998) also find that managers use restructuring charges to manage earnings from one period to another.

Boone and Mulherin (2003) attempt to value the process of corporate restructuring on an ex ante basis. The authors examine a sample of 298 firms that announced during the period from

1989-1998 that they were considering restructuring alternatives. The authors find that restructuring is a lengthy process and that the majority of the restructuring period occurs before any definitive proposals for corporate change are made. Boone and Mulherin (2003) also determine that the market underestimates the full wealth effects of completed restructurings.

Based on a performance review, Bens (2002) finds that restructurings are often corporate responses to significant operational problems. Restructuring charges are found to be frequently preceded by poor accounting and stock price performance. Bens (2002) also finds that the market fails to fully capitalize on the extent of these operational problems. Eames and Sape (2003) use a price model to examine the value relevance of special items including restructuring charges, and they determine that restructuring charges are not value relevant, contrary to the findings in most research.

26

For a global sample of firms during the period between 1989 and 1997, Ballester et al.

(1999) examine the link between restructuring and financial performance. The authors examine this topic because restructuring became a global phenomenon in the early to mid 1990s, with many firms announcing layoffs, reductions in pay, plant closings, and other money-saving ventures. Ballester et al. (1999) suggest that the increase in the number of restructurings during

the 1990s was caused by the quick incorporation of technology into firms. The authors create a

new measure, changes in labor expense intensity, which captures a firm’s restructuring efforts in general terms, and they use this variable to examine the financial performance of restructuring firms separately for firms with increasing and decreasing sales.13

The articles in this section examine the market’s response to operational restructurings.

Although many of these studies find that the market reacts positively to restructuring announcements [Ballester et al. (1999), Kross et al. (2001)], Bens (2002) finds that the announcement of an operational restructuring is viewed by the market as bad news. Other restructuring-related studies examine the market’s response to several common components of an operational restructuring. These studies find that the market reacts negatively to announcements of events including layoffs and plant closings [Lin and Rozeff (1993); Elayan et al. (1998);

Francis et al. (1996)]. Chalos and Chen (2002) find that the market reacts negatively to layoffs related to plant closings and positively to layoff announcements related to product line revenue refocusing.

13 Ballester et al. (1999) find that during the sample period, sample firms reorganized in a way that often emphasized a reduction in their labor expenses relative to sales, even when firm sales were growing. Also, the results show that while reorganizing firms that reduced their labor expenses had lower profitability, they ultimately had higher stock market returns. The analysis in Ballester et al. (1999) illustrates the importance of knowing the economic context of restructuring firms when interpreting their financial performance.

27

This dissertation contributes to the literature by examining the value relevance of restructuring charges using price and return models for the full sample of restructuring firms. The incremental contribution of this study is the examination of the value relevance of restructuring charge information using a sample of restructuring firms hand-collected from newswires and other news announcements. Based on the findings of studies including Brickley and Van Drunen

(1990) and Bens (2002), this essay hypothesizes that restructuring charge information is value relevant. Also, this essay suggests that the market will react positively to restructuring charge announcements as determined previously by Ballester et al. (1999) and Kross et al. (2001).

2.3.2 Effect of Restructurings on Earnings

Kang and Shivdasani (1997) examine a sample of Japanese firms that restructured after a decline in operating performance and determine that there is approximately a 1.5 percent increase in the industry-adjusted ratio of operating income to assets in the three years after a performance shock. Brickley and Van Drunen (1990) find that earnings decline in the three years following an operational restructuring. As discussed in Kross et al. (2001), the financial press still believes that the market reacts positively to restructuring even though prior studies, including Brickley and

VanDrunen (1990), show very little improvement in firms three to five years after they restructure.

For a sample of firms reporting restructuring charges in their financial statements between 1985 and 1995, Khurana and Lippincott (2000) examine the earnings-return relation during the restructuring charge year. The authors focus on the association between restructuring charges and common stock returns during the restructuring year and assess whether the profitability of the firm influences the decisions made by capital market investors. The authors determine that there is a significant positive relation between restructuring charges and common

28

stock returns in the restructuring year. When Khurana and Lippincott (2000) divide the sample into profit and loss firms, they obtain different results for the two groups. For profitable firms, the coefficient on the restructuring charge variable is negative and insignificant, while the coefficients on both the level and change in earnings are both positive and highly significant.

This finding suggests that the restructuring charges are temporary items for profitable firms and that these charges are not expected to be value increasing. The level of earnings and the change in earnings are both determined to be value relevant. Khurana and Lippincott (2000) conclude that restructuring charges have limited cash flow implications for profitable firms.

In addition, Khurana and Lippincott (2000) examine the restructuring charges of unprofitable firms. For these firms, the coefficient on the restructuring charge variable is positive and highly significant. The coefficient on the level of earnings variable is insignificant, and the coefficient on the change in earnings variable is only slightly significant. The results for loss firms suggest that current losses are viewed as being temporary and not value-relevant while restructuring activities are expected to have a permanent positive effect on future performance.

The authors further separate the restructuring firms into groups based on the primary purpose for the charge. The three main reasons for taking restructuring charges, as identified by Khurana and

Lippincott (2000), are restructuring with the primary purpose of exiting a line of business, restructuring where the primary purpose is to eliminate personnel, and restructuring where the primary purpose cannot be discerned. The authors find that both of the first types of restructurings are positively associated with returns.

Ellis and Gebhardt (2006) examine the finding that prices respond quickly to the information in quarterly earnings announcements. They demonstrate that companies that have conditioned on past earnings surprises will exhibit positive (negative) returns following a prior negative (positive) surprise. The authors attribute this reaction to the company’s expectations that

29

past earnings surprises will continue into the future, and investors respond to these expectations by reversing the price trend. Ellis and Gebhardt (2006) determine that the price response to meeting earnings forecasts is caused by the overreaction of the investors, and future returns are expected to undo the overreaction.

The articles in this section examine the effects of operational restructurings on earnings.

Brickley and Van Drunen (1990) find that earnings decline in the three years following an operational restructuring and that for profitable firms restructuring charges have limited cash flow implications. Based on their results for unprofitable firms, Khurana and Lippincott (2000) suggest that current losses are viewed as being temporary and not value-relevant while restructuring activities are seen as having a permanent positive effect on future performance.

Although Khurana and Lippincott (2000) separately document the impact of restructuring charges on earnings for profit and loss firms, prior studies have not separately examined firms restructuring for different reasons. This dissertation contributes to the literature by separately examining the impact of the magnitude of a restructuring charge on stock price and returns for firms restructuring for different reasons, primarily either to improve efficiency or to avoid filing for bankruptcy.

2.4 (Essay 3) Analyst Forecast Literature

2.4.1 Analyst Forecasts and Restructuring

A major objective of this dissertation is to estimate the impact of restructuring on analyst forecast revisions, analyst forecast accuracy, and analyst forecast bias. Although the literature on restructuring charges is plentiful, very few studies have examined the effects of restructuring announcements on analysts’ forecasts. Among the first studies to examine the effect of reported restructuring charges on analysts’ forecasts is Chaney et al. (1999). Chaney et al. (1999) examine

30

restructuring charges from the perspective of financial analysts. The authors provide evidence that analysts expect performance to decline for restructuring firms over the short-run with possible improvement over the long term. The authors examine forecast errors in the year immediately following a charge, and they find evidence that analysts’ accuracy has declined.

Also, the authors find that even though analysts forecast a downward revision after a charge, the

analysts are still optimistically biased. This essay extends Chaney et al. (1999) by adding specific

variables to determine the impact of the restructuring event and the likelihood of financial distress

on analysts’ forecast accuracy for firms with varying levels of financial health.

By investigating both analysts’ forecast accuracy and dispersion, Clement and Lopez

(2000) examine the effects of prior and multiple restructuring charges on earnings forecasts. The

authors predict that analysts learn from prior restructuring charges. The authors describe that

they use the word “learn” to mean that current restructuring charges impair forecast accuracy to a

lesser extent when prior restructuring charges are present. The authors find that analysts are able

to learn from prior restructuring events. Also, the authors determine that the relative magnitude

of restructuring charges is associated with a decrease in forecast accuracy and an increase in

dispersion for up to two years after the announcement of the restructuring. Overall, the results of

this study suggest that restructuring creates uncertainty for analysts that lasts for at least two years

after the announcement.

Lin and Yang (2006) examine the effect of previous restructuring charges on analyst

forecast revisions and accuracy. The authors determine that analysts react differently to first-time

restructuring firms than to repeat restructuring firms. The authors find that analysts revise their

forecasts of both one-year-ahead earnings and five-year earnings growth more negatively for

first-time restructuring firms than for firms that have taken charges in the past. When the authors

examine forecast errors in the year following the restructuring, they find that restructuring

31

charges in the current period complicate the earnings forecast process for analysts. Finally, the authors determine that the documented decrease in analyst forecast accuracy is reduced if a company has prior charges in the previous two years.

The incremental contribution of this essay is to separately examine the impact of restructuring charges on analyst forecast revisions, analyst forecast accuracy, and analyst forecast bias for firms restructuring for several different reasons. This study will compare the impact of restructuring charges on analysts for healthy firms restructuring to improve their efficiency to financially distressed firms restructuring to avoid bankruptcy.

2.4.2 Research Issues Related to Analyst Forecast Revisions

Many studies examine issues related to analyst forecast revisions, an area discussed in

Essay 3. Mest and Plummer (2000) use a linear model to demonstrate how analysts tend to revise their forecasts of future earnings in response to current forecast errors. The authors find that a non-linear model actually better describes the association between analysts’ forecast revisions and their forecast errors. Huai (2000) investigates the extent to which investors exhibit rational behavior in response to analyst optimism and auto-correlated analyst forecast revisions. The author finds that although investors are aware of the optimism in analyst forecasts, they do not adjust completely for it. Duru and Reeb (2002) examine the relationship between corporate international diversification and the accuracy and bias of analysts’ earnings forecasts, and they determine that greater corporate international diversification leads to less accurate and more optimistic forecasts.

Barron et al. (2002) examine changes in the precision and commonality of information contained in individual analysts’ earnings forecasts, and they determine that the idiosyncratic information in individual analysts’ forecasts increases after earnings announcements. The authors

32

also find that this increase is more significant as more forecast revisions occur. Clement and Tse

(2003) examine whether investors extract all of the information that analysts’ characteristics provide about forecast accuracy, and the authors determine that only some of the characteristics that are related with future forecast accuracy are also associated with return responses to forecast revisions. Lim and Kong (2004) obtain evidence from four Asia-Pacific markets that indicates that abnormal returns are related to the latest forecast revisions.

Hollie et al. (2005) determine that analysts revise their forecasts upon the preliminary earnings announcement and ignore the new information in SEC filings. Claus (2000) examines analysts’ earnings forecasts for six countries and determines that analysts’ revisions occur most frequently around periods when financial statement information is released. Lim and Kong

(2004) obtain evidence from four Asia-Pacific markets that indicates that abnormal returns are related to the latest forecast revisions. Many other studies have also examined issues related to analyst forecast revisions (Ivkovic and Jegadeesh 2002, Barth and Hutton 2003, Asquith et al.

2004, Pinello 2005, Yan 2006).

2.4.3 Research Issues Related to Analyst Forecast Accuracy

Many studies examine the impact of various items on analyst forecast accuracy. Alford and Berger (1999) use a simultaneous equations model to examine forecast accuracy, analyst following, and trading volume. The authors determine that special items including restructuring charges impair analysts’ ability to predict future earnings. The negative effect that special items have on accuracy is consistent with concerns that standard-setters have raised that unusual events impair investors’ ability to interpret and predict future earnings. By investigating analysts’ forecast accuracy, Clement and Lopez (2000) examine the effect of prior and multiple restructuring charges on earnings forecasts. The authors determine that the relative magnitude of

33

restructuring charges is associated with a decrease in forecast accuracy for up to two years after the announcement of the restructuring.

Ashbaugh and Pincus (2001) determine whether, relative to the International Accounting

Standards (IAS), the variation in accounting standards across countries has an impact on the

ability of analysts to forecast non-U.S. firms’ earnings accurately and whether analysts’ forecast

accuracy changes after firms adopt IAS. Hope (2003) examines the relationship between forecast

accuracy and the degree of enforcement of accounting standards and finds that strong

enforcement is associated with higher forecast accuracy. Duru and Reeb (2002) examine the

relationship between corporate international diversification and the accuracy of analysts’ earnings

forecasts. They determine that greater corporate international diversification leads to less

accurate forecasts.

Kwon (2002) examines the difference in forecast accuracy between high- and low-tech

firms and determines that there is lower unsigned error for high-tech firms than for low-tech

firms. Bonner et al. (2003) examine whether there are differences in how sophisticated and

unsophisticated investors use factors including characteristics of the analysts and the age of the

forecast to predict the relative accuracy of forecast revisions. The results suggest that

sophisticated investors have more knowledge about the relationship of the factors to forecast

accuracy. Gu and Wu (2003) suggest that if the analyst’s goal is to provide the most accurate

forecast by reducing the mean absolute forecast error, then the optimal forecast is the median

instead of the mean earnings. The authors find that earnings skewness is significantly related to

analysts’ forecast bias.

Chen et al. (2005) demonstrate that investors’ reactions to forecast news are increasing in

the product of the accuracy and the length of analysts’ forecast records. Chiang (2005) examines

the relationship between analysts’ forecast accuracy and corporate transparency, and the results

34

indicate that analysts’ forecast accuracy is strongly associated with corporate transparency.

Easton and Monahan (2005) evaluate the reliability of an expected return proxy using its relationship with realized returns. The results demonstrate that the return proxies examined are unreliable. Easton and Monahan (2005) demonstrate that some proxies are reliable when the long-term growth forecasts are low or when analysts’ forecast accuracy is high.

Hope and Kang (2005) determine whether macroeconomic uncertainty affects the forecast accuracy of analysts in an international setting, and they establish that forecast accuracy declines in the level of macroeconomic uncertainty. Barniv et al. (2005) examine the ability of analysts’ characteristics to explain the relative forecast accuracy of analysts across legal origins.

The authors determine that analysts in common-law countries outperform their peers in civil-law countries because market-based incentives exist. Also, Barniv et al. (2005) determine that analysts with superior ability in civil-law countries less consistently provide superior forecasts.

Fan et al. (2006) investigate the accuracy of the earnings forecasts of financial analysts from investors and financial analysts. The findings indicate that analysts’ forecasts outperform random walk time-series forecasts. Many other studies have also discussed issues related to analyst forecast accuracy (Gilson et al. 1998, Kini et al. 2005, Harjoto and Zaima 2005, Liu and Su 2005,

Bae et al. 2005, Bae et al. 2006, Koga and Uchino 2006, Ertimur et al. 2006, Clement et al. 2006)

2.4.4 Research Issues Related to Analyst Forecast Bias

Studies also examine issues related to analyst forecast bias, an issue examined in Essay 3.

Lim (2001) finds that positive and predictable forecast bias may be a rational property of all optimal earnings forecasts. Mest and Plummer (2003) extend Lim (2001) by examining analysts’ earnings and sales forecasts, and they determine that analysts’ optimistic bias is greater for earnings forecasts than for sales forecasts. Dechow et al. (2000) examine the role of sell-side

35

analysts’ earnings growth forecasts in the pricing of common stock and determine that these analysts’ forecasts are overly optimistic around stock offerings. Further, the authors identify a positive relation between the fees paid to analysts’ employers and the level of the analysts’ growth forecasts.

Han et al. (2001) demonstrate that publicly-available information can be used to establish estimates of analysts’ optimistic bias in earnings forecasts. Duru and Reeb (2002) examine the relationship between corporate international diversification and the bias of analysts’ earnings forecasts, and they determine that greater corporate international diversification leads to more optimistic forecasts. Gu and Wu (2003) find that earnings skewness is significantly related to analysts’ forecast bias. Additional studies have examined issues related to the bias exhibited in analyst forecasts (Ciccone 2001, Abarbanell and Lehavy 2003, Cowen et al. 2003, Scherbina

2004, Friesen and Walker 2005, Pastor et al. 2006).

2.4.5 Research Issues Related to Analyst Forecast Dispersion

Athanassakos and Kalimipalli (2003) examine the relationship between analysts’ forecast

dispersion and future stock return volatility and determine that there is a strong positive

relationship between analysts’ forecast dispersion and future return volatility. By investigating

analyst dispersion, Clement and Lopez (2000) examine the effect of prior and multiple

restructuring charges on earnings forecasts. The authors determine that the relative magnitude of

restructuring charges is associated with an increase in dispersion for up to two years after the

announcement of the restructuring. Kwon (2002) examines the difference dispersion between

high- and low-tech firms and determines that there is less dispersion for high-tech firms than for

low-tech firms. Ang and Ciccone (2001) examine the relation between returns and analyst

dispersion, and they find that firms with low dispersion outperform firms with high dispersion.

36

Dische (2001) examines analyst dispersion and determines that higher abnormal returns can be achieved by applying an earnings momentum strategy to with low dispersion. Koga and

Uchino (2006) examine the dispersion in analyst forecasts, and they determine that, for Japanese firms, analyst forecasts are more dispersed (less concentrated) for firms that have established associations with banks.

2.4.6 Research Issues Related to Analyst Consensus Forecasts

Irvine (2004) examines analysts’ earnings forecasts and stock recommendations to determine whether they affect their brokerage firm’s share of trading in the forecast stocks. The author determines that individual analyst forecasts that differ from the consensus forecast generate significant brokerage-firm trading in the forecast stocks in the two weeks after the forecast release date. This finding suggests that analysts’ forecasts affect their brokers’ commission revenue. Galenti (2004) determines how analysts’ consensus can constrain the forecasts of an analyst. The authors show that the forecast release depends on consensus dispersion and upon the analyst’s private information. Lin and Shih (2006) examine the increasing number of firms reporting earnings that meet or narrowly beat analyst earnings forecasts, resulting in very small or no error.

2.4.7 Earnings Management and Analysts

Bernhard and Campello (2002) investigate whether firms manage analyst forecasts to generate positive earnings surprises. Roychowdhury (2005) finds evidence suggesting that managers manipulate real activities to meet analyst forecasts. Liu (2005) examines the role that sell-side analysts and buy-side institutional investors have on earnings management. The author finds that firms with higher analyst following are more likely to participate in upward earnings

37

management. Koh (2006) examines the changes in earnings management in the post-scandal period, and he finds that managers do not meet or just beat analyst forecasts as often as in the past.

2.4.8 Other Analyst Research Studies

Other issues related to analysts have also been discussed in recent research. Using their past history for a specific firm’s earnings, Park and Stice (2000) identify superior analysts, and they demonstrate that subsequent forecast announcements by the superior analysts have more impact on price than the forecasts of other analysts. Elgers et al. (2001) suggest that the weighing of analysts’ annual earnings forecasts implicit in prices is lower than the historical relationship between financial analysts’ forecasts and realized earnings. Stevens and Williams (2004) analyze forecast reactions to positive versus negative information. The forecast data demonstrate that there is a systematic underreaction to both positive and negative information. The documented

underreaction is determined to be larger for positive information than for negative information.

Lacina and Karim (2004) determine whether a negative stock market reaction related to a

management forecast of near term low earnings is reduced by a concurrent management forecast

of improved longer term earnings expectations. The results demonstrate that analysts react less

negatively to management forecasts of improved earnings expectations than to management

forecasts of low earnings. Shane and Brous (2001) provide evidence that, using the next earnings

announcement and other information available between earnings announcements, analysts and

investors correct the documented underreactions of earnings forecasts and stock prices to earnings

news. Rammath et al. (2006) review research related to the role of financial analysts in capital

markets.

38

2.4.9 Summary

The line of analyst research examining the impact of the announcement of restructuring charges on analysts’ forecast accuracy is the primary basis for this dissertation. Chaney et al.

(1999) find that after an operational restructuring firm performance may decline over the short term and improve gradually over the long-term. Chaney et al. (1999) also examine analyst forecast errors in the year immediately following a charge, and they find evidence that analysts’ accuracy declines. This essay extends this research by separately examining analysts’ forecast accuracy for healthy and distressed firms.

Essay 3 considers the impact that the company’s reason for restructuring has on analyst

forecast revisions as in Hollie et al. (2005), forecast accuracy as in Fan et al. (2006), and forecast

bias as in Gu and Wu (2003). Although various other issues including analyst forecast dispersion

(Kwon 2002), analyst consensus forecasts (Lin and Shih 2006), and earnings management (Koh

2006) are briefly discussed in this literature review, these issues are beyond the scope of this

dissertation. Prior studies have failed to consider the differences in the primary reasons why

companies restructure, generally either to improve efficiency or to avoid further financial distress

or even bankruptcy, and the effect of these motivations on analyst forecasts. This is the primary

contribution of Essay 3. I expect that analysts will revise their forecasts upward for healthy firms

and downward for firms in financial distress. I also expect that analysts’ forecasts are

optimistically biased for financially distressed firms and less optimistically biased for firms that

are financially healthy and restructure to improve their operational efficiency.

2.5 Summary

The purpose of this literature review is to provide an overview of the prior literature

studying restructuring charges separately for firms in financial distress and financially healthy

39

firms and to highlight the incremental contribution of this dissertation. The first section in this chapter discusses some general issues related to restructuring to provide a background for the

dissertation. Various methods of identifying firms that are in financial distress are discussed in

the second section. The third section in this chapter discusses studies on the value relevance of

restructuring charges. Many of the restructuring charge studies examine the value relevance of restructuring charges, and many of these studies find that the market tends to positively value

announcements of operational restructurings [Brickley and Van Drunen (1990); Martin and

Kensinger (1990); Kross et al. (2001)]. However, studies examining the announcement of

individual components of many operational restructurings find different results. For example, the

stock market reacts negatively to corporate layoff announcements (Worrell et al. 1991; Elayan et

al. 1998). Finally, the fourth section discusses the impact that restructuring announcements have

on analysts’ forecast errors and dispersion.

This dissertation provides several incremental contributions to the extant literature. First,

although restructuring charges have been examined in prior research, past studies have not

focused on examining restructuring charges separately for healthy companies and distressed companies. Therefore, this dissertation contributes to the literature by helping to close the gap between the current restructuring and bankruptcy literatures. Second, this dissertation contributes to the existing literature related to the value relevance of restructuring charges by examining the value relevance of restructuring charges for companies that are restructuring for different reasons and with different results. Third, by providing evidence on the impact of restructuring charges on analysts’ forecasts separately for financially healthy firms and for several types of distressed

firms, the results contribute to the literature examining analysts’ reactions to events including

restructuring charges (Cheney et al. 1999).

CHAPTER 3 LITERATURE REVIEW

Essay 1

Differences in Financial Health for Firms Taking Restructuring Charges

Chapter 3

Essay 1: Differences in Financial Health for Firms Taking Restructuring Charges

3.1 Background

This essay examines three different groups of restructuring firms. The purpose is to investigate the differences between healthy and distressed firms, and then within the distressed group the differences between companies that restructure and subsequently file for bankruptcy within three years and other financially distressed restructured firms that are able to avoid bankruptcy following a restructuring. Before examining the differences in these groups, it is necessary to use several proven methods including Altman’s (1968) and Ohlson’s (1980) bankruptcy prediction models to identify the financial health for each sample firm. Agarwal and

Taffler (2006) examine statistical failure prediction models including Altman (1968) to test whether these methodologies still work, and they determine that older Z-score models including

Altman (1968) have high predictive value over their twenty-five year sample period. This indicates the usefulness of Altman’s (1968) prediction model, even almost forty years after it was developed. The first steps in this essay develop hypotheses and provide a description of the methodology used to test those hypotheses.

Sudarsanam and Lai (2001) examine a sample of distressed UK firms for the period from

1985 to 1993 to determine the differences in restructuring strategies adopted by recovered and non-recovered firms. The results demonstrate that recovered and non-recovered firms adopt similar sets of restructuring activities and that the managers of non-recovered firms restructure more intensively than recovered firms. The authors find that non-recovered firms are much less effective in implementing their restructuring strategies than the recovered firms. Sudarsanam and

41 42

Lai (2001) find that recovered firms1 normally choose growth-oriented and external-market focused strategies while non-recovered firms often choose “fire-fighting” strategies. Kane and

Richardson (2000) determine that when higher earnings accompany the restructuring of a firm’s assets, the likelihood of emergence from financial distress increases, thus suggesting that asset restructuring enhances the impact of the earnings’ signal. The incremental contributions of this essay over the extant studies include the method developed and utilized to identify firms in financial distress, the comparisons of univariate statistics for firms restructuring for different reasons, and the method used to identify the distressed firms that will be able to avoid filing bankruptcy for at least three years after restructuring.

The first section of this essay discusses the importance of carefully identifying firms in financial distress. In the second section, the hypotheses are developed. In the third section, the data set and the methodology utilized to test the hypotheses are discussed. The empirical results are described in the fourth section of this essay. Finally, the fifth section summarizes and draws conclusions from this essay.

3.2 Importance of Examining Financial Health

Because the purpose of this essay is to examine the effects of restructuring charge announcements on stock price and analysts’ forecast accuracy for firms in different financial situations, it is necessary to choose methods to identify firms in financial distress. This essay first uses the Altman (1968) model and the Begley (1996) model to identify firms in financial distress.

The results of these two models are then compared to determine which method is more accurate.

Also, these two proven methods are used to construct a bankruptcy prediction model based on the

1 Sudarsanam and Lai (2001) define recovered firms as distressed firms having a positive Altman Z-score over the two years after the distress year. Non-recovered firms still have negative Altman Z-scores two years after the distress year.

43

results of Ohlson (1980) to determine the method useful for identifying firms that are likely to file for bankruptcy.

3.3 Hypotheses Development

The hypotheses examine the differences between healthy firms and two groups of financially distressed firms. Specifically, the differences between healthy and distressed firms and the differences between financially distressed companies that restructure and subsequently file for bankruptcy within three years and other financially distressed restructured firms that are able to avoid bankruptcy following a successful restructuring are examined.

Thus, the first hypotheses are stated as:

H1a: Financially distressed companies that restructure differ from healthy companies that

restructure.

H1b: Financially distressed companies that restructure and subsequently file for

bankruptcy within three years differ from other financially distressed restructured

firms that are able to avoid bankruptcy following a restructuring.

3.4 Data Sources and Methodology

3.4.1 Data

The full sample for this essay includes data for the period from 1992 to 2003 and firms announcing operational restructurings during the period from 1993 to 2003.2 An initial keyword

search was conducted of the newswires available on the Lexis-Nexis Database to identify the

companies that restructured over the period from 1993 through 2003. Once it was determined

that a company restructured, it was then necessary to obtain information about the type of

2 Because one variable requires data from period t-1, some 1992 data are also used.

44

restructuring, the dollar amount of the restructuring charge, and the date of the announcement.

Information on the amount of the restructuring charge was also obtained from the newswire announcements and financial statements available on Lexis-Nexis. Companies that underwent debt restructurings or management restructuring are not deemed relevant and are therefore excluded from this essay. Only companies undergoing operational restructurings are examined.

Operational restructuring projects typically include some combination of workforce reductions, asset writedowns, the disposal of certain assets and facilities, product line discontinuations, the reconfiguration of facilities, plant relocations, or the closing of certain plants and facilities (Lopez

2002). Bankrupt companies were identified by using The Bankruptcy Yearbook and Almanac

(1993-2003).

Next, financial and market data were obtained from the COMPUSTAT database. The

COMPUSTAT data were read and analyzed for the year of the restructuring. The expected signs

for all of the variables are presented in the tables.3

Panel A of Table 2 shows the number of times during the period from 1993 to 2003 that

the firms in the sample restructured. The first set of columns shows that 1,245 firms restructured

only once during the sample period, while 411 firms restructured more than once. The middle

columns of Panel A show the number of sample firm-event observations taking place in each

year, with the most observations being from 2003 and the fewest being from 1994. The third set

of columns in Panel A shows the number of firms in each year that restructured for the first time

during the sample period (1993-2003).

3 Table 1 provides the steps used to arrive at the final sample of 1,656 firms (in Panel A) which comprised the 2,202 firm-event observations (in Panel B). Some firms completed more than one operational restructuring over the period from 1993 through 2003; the firms are included more than once in the sample as different firm-event observations. If a firm restructured more than once in the same year, only the first- time restructuring is included in the sample. Most firm-event observations occurred several years apart.

45

As shown in Panel B of Table 2, the sample is comprised primarily of manufacturing firms, with 55.8 (58.6) percent of the sample firms (firm-event observations) coming from this category. The next largest sample group is service firms, which comprise 23.7 (21.7) percent of the firms (firm-event observations). The sample also contains smaller percentages of firms from the transportation, communication, gas and electric category (6.2%); the wholesale and retail trade categories (4.6 and 5.6%); and the financial, insurance, and real estate category (1.8%).

3.4.2 Methodology

Prior to testing the hypotheses, it is necessary to create a sample of firms that have restructured. The full sample for this essay contains firms undertaking operational restructuring efforts during the period from 1993 through 2003 that have data availability for the required variables. Equations 1.1 through 1.3 are used to determine whether or not each firm was in financial distress. In order to determine a “distress” value for each firm in the sample, Altman’s

(1968) original Z-score model and Begley et al.’s (1996) updated version of the Altman model are used. The Begley et al. (1996) study is one of the only papers done since 1968 to even attempt to test the predictive ability of the original Z-score model (See Agarwal and Taffler

2006).

Although the Altman and Begley models were originally intended as bankruptcy

prediction models, Grice and Dugan (2001) indicate that bankruptcy prediction models like

Altman’s are actually more useful for identifying firms that are financially distressed, as opposed

to identifying the more limited bankruptcy condition. Because these models have been proven

successful, the linear Z-score equations are used and the numbers for each variable for the firms

in the sample are substituted. These models are used to determine a distress value for each firm,

46

and then a cutoff point can be used to classify firms as either distressed or healthy. Altman’s

(1968) Z-score model is as follows:

Z = 0.012 X1 + 0.014 X2 + 0.033 X3 + 0.006 X4 + 0.999 X5, (1.1) where

Z is used to determine whether each company is in financial distress,

X1 is working capital divided by total assets * 100,

X2 is retained earnings divided by total assets * 100,

X3 is earnings before interest and taxes divided by total assets * 100.

X4 is the market value of equity divided by the book value of debt * 100, and

X5 is sales divided by total assets.

Working capital divided by total assets is a measure of the net liquid assets of a firm relative to the overall capitalization; firms with losses are likely to also have shrinking current assets compared to total assets.4 The market value of equity divided by book value of debt

variable shows how much the firm’s assets can decline in value before the firm becomes

insolvent.5 Retained earnings divided by total assets is included because it implicitly considers

the age of a firm, and financial distress is much more common in the early years of a firm’s life.

Sales divided by total assets is a measure of firm size.6 Earnings before interest and taxes,

divided by total assets is a measure of the true productivity of a firm’s assets, ignoring tax and

leverage factors. Because a firm’s existence is based on the earning power of the firm’s assets,

4 Altman (1968) finds that working capital divided by total assets is the most valuable measure of the liquidity. 5 Including the market value of equity divided by the book value of debt adds a market value dimension not considered before Altman (1968), and this variable is determined to be a better predictor of bankruptcy than net worth/total debt. 6 Although Altman (1968) finds that sales divided by total assets is the least significant variable on its own, it is important to include this variable because of its unique relationship to the other variables included in the model.

47

this ratio is especially important.7 All X-values are included in the calculation of Z for each firm

or firm-event observation, even when the values are negative.

Begley et al. (1996) re-estimate Altman’s model using data from the 1980s, and their

updated model is as follows:

Z = 0.104 X1 + 1.010 X2 + 0.106 X3 + 0.003 X4 + 0.169 X5, (1.2)

using the same variables and variable definitions as Altman’s model.

Altman finds that for his sample firms, firms with Z-scores greater than 2.99 were mostly not in financial distress and many of the firms with Z-scores less than 1.81 went bankrupt.

Altman further finds that using a Z-score of 2.675 as a cutoff minimizes the number of firms that are misclassified by the model. Therefore, this essay uses 2.675 as the cutoff point for the

Altman model results. Begley et al. (1996) find that the most appropriate cutoff point for their model is 0.545. Firms with Z-scores less than 0.545 are classified as financially distressed and are assigned a value of 1, and firms with Z-scores greater than 0.545 are classified as being non- distressed and are assigned a value of 0.

After each firm is classified as either being financially distressed or non-distressed using both the Altman (1968) model and the Begley et al. (1996) model, Ohlson’s (1980) logistic regression model is used to generate the probability of financial distress for each company and for classification procedures.8 Also, the results of the Ohlson logistic regression model provide a

probability value between 0 and 1 for each firm that indicates the likelihood of a firm being in

financial distress. These probability values are included as additional predictor variables in the

7 It is important to note that because of Altman’s original computer format arrangement, variables X1 to X4 are included in calculations as absolute percentage values (10% as opposed to .10). Because of its extremely high relative discriminant coefficient, only X5 is expressed as a decimal instead of a percentage. 8 The use of non-complete defined groups (distressed vs. non-distressed firms) is an issue in this study because the groups are assigned using a combination of several models (Altman 1968, Begley 1996, Ohlson 1980) that have not been used together before to identify the distress condition. The use of these models to determine a firm’s financial condition may be a limitation for this essay, and other methods for identifying distressed firms may be developed in the future.

48

price and return models discussed in Essay 2 and the analyst forecast models discussed in Essay

3. Ohlson’s (1980) model is as follows:

DISTRESSi,t = a0 + a1 SIZE i,t + a2 TLTA i,t + a3 WCTA i,t + a4 CLCA i,t + a5 NITA i,t (1.3)

+ a6 FUTL + a7 INTWOi,t + a8 OENEGi,t + a9 CHIN + ui,t, where

DISTRESS equals 1 if a firm is determined to be in financial distress, 0 otherwise;

SIZE is the log of total assets;

TLTA is total liabilities divided by total assets;

WCTA is working capital divided by total assets;

CLCA is current liabilities divided by current assets;

OENEG equals 1 if owners’ equity is negative, 0 otherwise;

NITA is net income divided by total assets;

FUTL is cash flows from operations divided by total liabilities;

INTWO equals 1 if net income was negative over the last two years, 0 otherwise;

CHIN = (NIt – NIt-1) / (| NIt | + | NIt-1 |).

Ohlson’s (1980) model includes nine explanatory variables, and even though Ohlson finds that only six of them are significant, all of them are included in this essay. The log of total assets, total liabilities divided by total assets, net income divided by total assets, cash flows from operations divided by total liabilities, FUTL, and CHIN are all significant predictors of bankruptcy in Ohlson’s (1980) model.9 As an additional sensitivity test, the model in equation

1.3 is run a second and third time using a multivariate logistic regression, with dependent variable

DISTRESS obtaining three possible values. For the multivariate logistic regression, DISTRESS

9 The variable TLTA is included as a measure of firm leverage, and NITA and FUTL are included as measures of firm performance. The variable OENEG is used as a discontinuity correction for TLTA. The variables WCTA and CLCA are included as measures of current liquidity. Because of its importance in McKibben (1972), the CHIN variable is a measure of the change in net income that is included

49

equals two for firms that ultimately file for bankruptcy, one for firms defined by both Altman

(1968) and Begley (1996) as distressed that do not file within three years of restructuring, and zero for firms classified as non-distressed by both the Altman (1968) and Begley (1996) models.

Several methods are implemented to test the hypotheses. First, independent sample t-

tests and Wilcoxon tests for independents samples are used to see whether there is a significant

difference between the financial characteristics for the two samples, healthy firms that

restructured to improve efficiency are compared with those that restructured because of financial

distress. Significant differences in the univariate statistics for the firms classified as distressed

and those classified as non-distressed by both the Altman (1968) and Begley (1996) models will

provide support for the use of these models to identify the financial condition of firms. The

financial characteristics examined for the sample firms were chosen based on their results in the

Ohlson and Altman models identified in equations 1.1 and 1.3. The second set of tests examines

whether firms that are classified as being in financial distress but are able to avoid filing for

bankruptcy for at least three years subsequent to restructuring are significantly different from the

distressed firms that ultimately file. The proportion of the firms in financial distress that do not

file for bankruptcy within three years of restructuring is determined. Again using independent

sample t-tests and Wilcoxon tests on various financial variables, the differences between the

distressed firms that file in subsequent years and the distressed firms that avoid bankruptcy are

further examined to determine whether there is a significant difference between firm-specific

characteristics for companies that file and companies that avoid bankruptcy.

Also, a multivariate logistic regression model is used to test the second hypothesis. To

determine factors that increase the likelihood that a distressed firm can avoid filing for

bankruptcy during the three years subsequent to restructuring, the model includes various

financial variables different from those in Ohlson’s and Altman’s models. Gilbert et al. (1990)

50

demonstrate that financial variables that possess the ability to distinguish between distressed and bankrupt firms should be different from those used to distinguish between bankrupt and randomly selected non-bankrupt firms, and this provides support for using variables in this model that have not previously been incorporated in other bankruptcy prediction models. The final multivariate logistic regression model used in this essay for testing the second hypothesis is as follows

(Keener 2003):

BANKRUPTi,t = a0 + a1 COVERAGEi,t + a2 QR i,t + a3 REC_TURN i,t + (1.4)

a4 BVPSi,t + a5 GPRATIO i,t + a6 NPRATIO i,t + a7 ROE i,t + ui,t,

where

BANKRUPT equals 1 if the distressed firm goes bankrupt in the three years subsequent

to restructuring, 0 otherwise;

COVERAGE is the current cash debt coverage ratio (net cash from operations divided by

average current liabilities);

QR is the quick ratio (cash, marketable securities, and receivables divided by current

liabilities);

REC_TURN is accounts receivable turnover;

BVPS is book value per share;

GPRATIO is gross profit divided by sales;

NPRATIO is net income divided by sales;

ROE is return on equity (net income less preferred dividends divided by average common

stockholders’ equity).

These variables are different from the variables included in the Ohlson (1980) and Altman (1968) models, and they are used to proxy for various aspects of companies in financial distress. The quick ratio, previously used in Sung et al. (1999) and Shin and Lee (2002), and the current cash

51

debt coverage ratio, previously used in Frydman et al. (1985) and Sung et al. (1999), are proxies for liquidity. The current cash debt coverage ratio is expected to be a significant predictor because prior research has found that cash flow variables add to the predictive power of

bankruptcy prediction models for distressed firms (Gilbert et al. 1990). The receivable turnover

ratio (Dhumale 1998) is included in the model as a proxy for activity. Return on equity

(Neophytou and Molinero 2004), the gross profit ratio (Cinca et al. 2001), the net profit ratio

(Cinca et al. 2001), and book value per share (Ko and Lin 2006) are included in the model as proxies for performance.

3.5 Empirical Results

Detailed descriptive statistics for the full sample of restructuring firms are provided in

Table 3. Panel A provides the statistics for the 1,656 sample firms, while Panel B provides the statistics for the full sample of 2,164 firm-event observations.10 Several of the statistics from

Panel A provide interesting information about the sample. First, the mean (median) value for

total assets for the full sample of firms is $2.134 billion ($241 million), which indicates that the

sample is composed primarily of large companies.11 The mean (median) values for the current

ratio and the quick ratio of 2.48 (1.71) and 1.79 (1.05), respectively, indicate that the sample

firms are fairly liquid. The median restructuring charge amount is $8.26 million, which indicates

that most of the companies in the sample are undergoing fairly significant restructuring efforts.

The statistics for the firm-event observations sample provided in Panel B are very similar to those

for the sample of 1,656 firms in Panel A, so additional discussion is not required.

10 The observations have been trimmed to delete the top and bottom one percent of the data points to avoid having these outlier observations skew the results. 11 The minimum ($54,000) and maximum ($87.9 billion) values for total assets indicate the diversity of the firms in the sample. The median value for net income indicates that the average sample firm experienced a net loss of $3.810 million. The median for the receivable turnover ratio indicates that the firms turn over their receivables about six times per year.

52

To determine a “distress” value for each firm in the sample, Z-score values were calculated for each firm using both equations 1.1 and 1.2. Established cutoff points were used to classify each firm as either financial distressed or non-distressed. Firms with Z-scores larger than the cutoff points previously identified by Altman and Begley (2.675 and 0.545, respectively) were classified as not being in financial distress while firms with Z-scores below the cutoff points were classified as being in financial distress.

Table 4 provides a contingency table to illustrate the similarities in the Altman (1968)

and the Begley (1996) classification procedures. The two classification schemes provide

strikingly similar results. Panel A provides the contingency table for firms, and Panel B provides

the contingency table for firm-event observations. In Panel A, 899 firms (54.3%) were classified

as being financially distressed using Altman’s Z-score model, and the remaining 757 firms

(45.7%) were classified as not being distressed. Using Begley’s Z-score model, 767 firms

(46.3%) were classified as financially distressed, and the remaining 889 firms (53.7%) were

classified as non-distressed. Roughly seventy-three percent of the firms are classified the same

by both the Altman (1968) model and the Begley (1996) model, and this study uses only these

firms in further models. In Panel B, 1,112 firm-event observations (51.4%) were classified as

being financially distressed using Altman’s Z-score model, and the remaining 1,052 firm-event

observations (48.6%) were classified as not being distressed. Using Begley’s Z-score model, 917

firm-event observations (42.4%) were classified as financially distressed, and the remaining 1,247

firm-event observations (57.6%) were classified as non-distressed. Only firms and firm-event

observations classified into the same health classification group by both the Altman (1968) model

and the Begley (1996) model are used in the final sample.

Panel A (Panel B) of Table 5 provides the descriptive statistics for the firms (firm-event

observations) classified into the same financial health category by both the Altman (1968) and

53

Begley (1996) models. The firms in Panel A and the firm-event observations in Panel B are fairly similar in most cases, so only the firm statistics from Panel A are discussed here. The median values for current assets and sales of $124 million and $261 million, respectively, indicate that the sample of firms classified into the same distress category by Altman and Begley are large firms. However, the median net loss value of $4.8 million indicates that these firms are not realizing as much profit as their shareholders would like. Again, the median current ratio (1.76) and quick ratio (1.07) values indicate that the sample of firms has fairly liquid assets.

Panel A of Table 6 (Table 7) provides the mean values, the independent sample t- statistics, and the Wilcoxon Z-statistics for comparing the firms (firm-event observations) classified by both Altman’s model and Begley’s model as distressed with the firms (firm-event observations) classified by both models as non-distressed. In Panel A, the significance of the Z- statistics for the firms classified as non-distressed and those classified as distressed by both the

Altman (1968) and Begley (1996) models helps to demonstrate the success of the models in identifying financially distressed firms. A comparison of the means using the two-tailed t- statistics demonstrates that the firms classified as non-distressed are financially healthier than the distressed firms for all of the variables except for returns. A comparison of the means using

Wilcoxon Z-statistics demonstrates that the firms classified as non-distressed are financially healthier than the distressed firms for all of the variables except for the quick ratio and the gross profit ratio. The mean charge amounts for the non-distressed and distressed firms, respectively, are $53.61 million and $20.41 million, and this demonstrates that, on average, non-distressed

firms spend more on restructuring costs than distressed firms.

The two-tailed t-statistics and Wilcoxon Z-statistics reported in Panel A of Table 6 indicate that almost all of the balance sheet, income statement, and cash flows statement accounts examined are significantly greater for the non-distressed firms than for the distressed firms. An

54

examination of the Z-statistics for the working capital, current cash debt coverage, and current ratio variables also indicates that the two groups are significantly different. However, the distressed firms actually have significantly higher current ratios than the firms classified by both

Altman (1968) and Begley (1996) as non-distressed. This suggests that the firms classified as distressed remain highly liquid. All the variables in Panel A are significantly different at the 0.01 level for the distressed firms and the non-distressed firms except for the quick ratio and the gross profit ratio. Therefore, the results in Panel A of Table 6 provide strong support for H1a.

Panel B of Table 6 provides the mean values and the statistical tests for the differences between the firms classified by both the Altman (1968) and Begley (1996) models as distressed that file for bankruptcy in the three years subsequent to restructuring and those firms that avoid filing for bankruptcy for at least three years. Of the 613 firms classified as distressed, 545 firms

(89% of the distressed firms) are able to avoid filing for bankruptcy for at least three years after restructuring. Only sixty-eight (11%) of the distressed sample firms file for bankruptcy in the three years subsequent to restructuring, and this suggests that there is a fairly high short-term success rate for corporate restructuring efforts.12

Overall, the results reported in Tables 6 and 7 demonstrate that the firms classified as distressed are significantly different from non-distressed firms. The significance of the differences helps to validate the combined use of the Altman (1968) and Begley (1996) classification procedures to differentiate between healthy and distressed firms. Further, the results illustrate that the firms that ultimately file for bankruptcy are significantly different in terms of a few financial variables from firms that are able to avoid filing for bankruptcy after restructuring.

12 Annual bankruptcy rates among publicly traded firms were usually less than one percent.

55

To further examine the accuracy of the classification method used in this essay and to obtain the probability that each sample firm is in financial distress, I use Ohlson’s (1980) bankruptcy prediction model, which is provided in equation 1.3. As previously mentioned,

Ohlson’s logistic regression model is used to identify the likelihood that a firm is healthy or in financial distress. The classification results of both Begley’s (1996) and Altman’s (1968) models

are used to assign the value of 1 to companies that are identified in financial distress and the value

of 0 for healthy companies. Then, they are used as the group classifications for the Ohlson model.

Tables 8 and 9 provide the results from running Ohlson’s model where the distress- classification result is obtained from Altman’s (1968) model and Begley’s (1996) model. The results in Tables 8 and 9 are for the 1,202 firms and 1,581 firm-event observations, respectively.

Panel A of each table provides the results of Ohlson’s logistic regression model. The p-values for

the Wald statistics in both panels show that all of the variables are significant predictors of

financial distress except for WCTA and TLTA. The WCTA variable is not significant in

Ohlson’s (1980) paper either. The classification matrix presented in Panel B of Tables 8 and 9

shows that Ohlson’s model is accurate 89.1 percent and 89.2 percent of the time, respectively,

when the DISTRESS dummy variable is used as the dependent variable in the logistic regression.

The results in Tables 8 and 9 demonstrate the high success rate of the classification method used

for identifying firms in financial distress.

Tables 10 and 11 provide the results of the regression reported in equation 1.4 developed

to predict bankruptcy for firms and firm-event observations, respectively. These financially

distressed firms are classified as bankrupt or non-bankrupt. Using variables that were not

included in the Begley (1996) and Ohlson (1980) models, equation 1.4 is used to predict which of

56

the restructuring firms classified as being in financial distress will file for bankruptcy during the three years subsequent to restructuring.

Panel A of Tables 10 and 11 provides the results of the logistic regression. Although

some of the variables are not significant predictors of bankruptcy, the overall model is highly

significant. For the sample of 430 firms (522 firm-event observations), the Cox & Snell R-

Square for the model is 0.064 (0.054), and the Nagelkerke R-Square is 0.142 (0.121). The Wald

statistic is significant at the 0.10 level for BVPS, ROE, and the Net Profit Ratio. This indicates that firms with lower net profit ratios and lower ROE are more likely to file for bankruptcy in the three years subsequent to restructuring. The positive sign for book value per share suggests that firms with higher book values are more likely to file for bankruptcy in the three years subsequent to restructuring, a finding that is contrary to expectations. The classification tables in Panel B of

Tables 10 and 11, respectively, demonstrate that this model is accurate at predicting bankruptcy for distressed, restructuring firms 77.9 and 76.1 percent of the time, when a cutoff point of 0.1 is used. The results in Tables 10 and 11 provide additional support for H1b, which states that financially distressed companies that restructure and file for bankruptcy within three years of restructuring are different from other financially distressed restructured firms that avoid

bankruptcy during that period.

Tables 12 and 13 examine the model’s ability to classify bankrupt versus non-bankrupt

firms. I obtain over 94 percent correct classifications, suggesting the model is very useful for

predicting bankruptcy. This analysis is supplemental, and it is not used to test any hypotheses.

As an additional-supplemental analysis, I regress Ohlson’s (1980) nine-variable model using a three-group partition. In Table 14, the DISTRESS classification has three possible values. DISTRESS is assigned a value of two for firms that ultimately file for bankruptcy, one for distressed firms that do not file within three years of restructuring, and zero for non-distressed

57

restructuring firms. Panel B of Table 14 shows the model's very high classification accuracy, demonstrating that 87 percent of the firm-event observations are correctly classified.

3.6 Conclusions

This essay provides the link between restructuring efforts by corporations and subsequent bankruptcy. Thirty-eight percent of the firms reporting operational restructurings were determined to be in financial distress using both the Altman (1968) and Begley (1996) model for the period from 1993 to 2003. This indicates that many companies undertake restructuring efforts as an attempt to prevent impending bankruptcy. Independent sample t-tests and Wilcoxon Z-tests demonstrate that firms classified as non-distressed are significantly different from those firms classified by both the Altman (1968) and Begley (1996) model as financially distressed, providing support for H1a. Overall, over 89 percent of the firms and firm-event observations are correctly classified as either healthy or distressed. This high success rate for companies determined to be in financial distress may provide statistical support for using restructuring to improve financial health.

Further, the results of this essay demonstrate that over ninety percent of the firms in

financial distress were able to avoid filing for bankruptcy for at least three years after

restructuring. The bankruptcy prediction model described in equation 1.4 is successful 78 percent

of the time at determining which financially distressed firms will file for bankruptcy within three

years of restructuring, and this provides support for H1b.

Several limitations on the findings in this essay should be noted. First, future research

may be able to improve the process by which the list of corporate restructurings was obtained.

Because the initial list of restructuring firms used in this essay was obtained through a key word

search of the newswires available on the Lexis-Nexis Database, it is possible that some

58

restructuring firms were left out of the sample. If an announcement of the restructuring did not take place on a newswire, then it may have been inadvertently left out the sample. Perhaps by looking at the 10-Ks and 8-Ks of firms with special item dollar amounts available on

COMPUSTAT, fewer firms that potentially have taken restructuring charges would have been eliminated from the sample. Second, the results of the bankruptcy prediction model in equation

1.4 may be affected by the partial predictive ability of the model. Because equation 1.4 has only

been used in Keener (2003), the results of this model may not be generalizable to other samples.

Third, the results may be adversely affected by including some companies that restructured multiple times as firm-event observations. Therefore, the analysis also focuses on firms that announced restructuring efforts only once during the research period (the majority of the firms).

Finally, a major limitation of this study is the “ad-hoc” statistical identification of healthy and financially distressed firms.

BIBLIOGRAPHY

Agarwal, V., and R. Taffler. 2006. Twenty-five years of Z-scores in the UK: Do they really work? Working paper presented at the 2006 AAA National Conference, Cranfield University, Bedfordshire, England.

Altman, E. I. 1968. Financial ratios, discriminant analysis, and the prediction of corporate bankruptcy. Journal of Finance 23 (4): 589-609.

Begley, J., J. Ming, and S. Watts. 1996. Bankruptcy classification errors in the 1980s: An empirical analysis of Altman’s and Ohlson’s models. Review of Accounting Studies 1: 267-284.

Cinca, C. S., C. M. Molinero, and J. L. G. Larraz. 2001. Country and size effects in financial ratios: A European perspective. Working paper, University of Zaragoza, Spain.

Dhumale, R. 1998. Earnings retention as a specification mechanism in logistic bankruptcy models: A test of the theory. Journal of Business Finance & Accounting 25 (7/8): 1005-1023.

Frydman, H., E. I. Altman, and D. Kao. 1985. Introducing recursive partitioning for financial classification: The case of financial distress. Journal of Finance 40 (1): 269-291

Gilbert, L. R., K. Menon, and K. B. Schwartz. 1990. Predicting bankruptcy for firms in financial distress. Journal of Business Finance and Accounting 17 (1): 161-171.

Grice, J. S., and M. T. Dugan. 2001. The limitations of bankruptcy prediction models: Some cautions for the researcher. Review of Quantitative Finance and Accounting 17: 151- 166.

_____ and R. W. Ingram. 2001. Tests of the generalizability of Altman’s bankruptcy prediction model. Journal of Business Research 54: 53-61.

Kane, G. D., and F. M. Richardson. 2000. Corporate restructuring and the likelihood of emergence from financial distress. Working paper. University of Delaware.

Keener, M. H. 2003. How successful are restructuring efforts for firms trying to improve efficiency and for firms in financial distress? Working paper. Kent State University.

Ko, P., and P. Lin. 2006. An evolution-based approach with modularized evaluations to forecast financial distress. Knowledge-Based Systems 19 (1): 84-91.

McHugh, Christopher M., ed. 1993-2003. The Bankruptcy Yearbook & Almanac. Boston: New Generation Research, Inc.

59 60

McKibben, W. 1972. Econometric forecasting of common stock investment returns: A new methodology using fundamental operating data. Journal of Finance 27 (May): 371-380.

Neophytou, E. and C. M. Molinero. 2004. Predicting Corporate Failure in the UK: A Multidimensional Scaling Approach. Journal of Business Finance and Accounting 31 (5/6): 677-710.

Ohlson, J. A. 1980. Financial ratios and the probabilistic prediction of bankruptcy. Journal of Accounting Research 18 (1): 109-131.

Shin, K, and Y. Lee. 2002. A genetic algorithm application in bankruptcy prediction modeling. Working paper. Ewha Womans University, Seoul, South Korea.

Sudarsanam, S., and J. Lai. 2001. Corporate financial distress and turnaround strategies: An empirical analysis. British Journal of Management 12: 183-199.

Sung, T. K., C. Namsik, and G. Lee. 1999. Dynamics of modeling in data mining: Interpretive approach to bankruptcy prediction. Journal of Management Information Systems 16 (1): 63-85.

61

TABLE 1 Final Sample Determinationa Panel A: Firms

Number of Firms

Sample of firms announcing restructuring from 1993 through 2003 3,155

Observations with missing data (IBES and Compustat) (1,471)

Less outliers deleted (28)

Final number of firms included in this essay 1,656

Panel B: Firm-Event Observations Number of Firm- Event Observations

Sample firm-event observations announcing restructuring from 1993 through 2003 3,867

Observations with missing data (IBES and Compustat) (1,627)

Less outliers deleted (38)

Final number of firm-event observations included in this essay 2,202

a Some firms completed more than one operational restructuring over the period from 1993 through 2003; the firms are included more than once in the sample as different firm-event observations. If a firm restructured more than once in the same year, only the first-time restructuring is included in the sample. Most firm event observations occur several years apart.

62

TABLE 2: Panel A # of Firms Restructuring for # of Times Firms Restructure # of Firm-Event Observations the First Time During the a b Between 1993 and 2003 Occurring Each Year Sample Period 1 1,248 1993 163 1993 162 2 292 1994 77 1994 69 3 79 1995 150 1995 126 4 22 1996 167 1996 129 5 11 1997 163 1997 120 c > 5 4 1998 261 1998 204 Total 1,656 1999 202 1999 147 2000 168 2000 127 2001 268 2001 196 2002 186 2002 104 d2003 397 2003 272 Total 2,202 Total 1,656

a The average number of years between firm-event observations is 4.45, which indicates that the firm-event observations included in the sample actually represent different restructuring programs for the companies that occurred several years apart. b If a company took restructuring actions more than once in the same year, only one firm-event observation is included for the company for that year. Also, only one firm event is included in the sample if a firm restructured more than once in the same fiscal year. c This number indicates that 4 sample firms restructured more than five times between 1993 and 2003. For example, AT&T restructured in nine different years during the period from 1993 to 2003. d The 397 firm-event observations occurring in 2003 include 122 companies that had restructured previously at least once.

63

TABLE 2: Panel B # of Firm- Firm-Event # of Event Firms as Observations SIC Division Firms Observations % of Total as % of Total Agriculture, Forestry, and Fishing 2 2 0.12 0.09 Mining 20 27 1.21 1.23 Construction 12 13 0.72 0.59 Manufacturing 924 1,290 55.80 58.58 Transportation, Communications, Electric, & Gas 102 138 6.16 6.27 Wholesale Trade 76 97 4.59 4.41 Retail Trade 92 116 5.56 5.27 Finance, Insurance, & Real Estate 29 34 1.75 1.54 Services 392 477 23.67 21.66 Public Administration 7 8 0.42 0.36 Total 1,656 2,202

64

Table 3: Descriptive Statistics Panel A: Full Sample of Firms Standard Variable Firms Mean Median Deviation Minimum Maximum Balance Sheet Accounts: Current Assets ($Millions) 1,656 759.7 111.1 2,694.1 0.03 48,489.0 Total Assets ($Millions) 1,656 2,134.3 241.2 6,993.7 0.05 87,862.0 Current Liabilities ($Millions) 1,656 581.5 63.0 2,198.7 0.17 33,150.0 Total Liabilities ($Millions) 1,656 1,403.8 119.3 4,686.1 0.27 61,375.0 Retained Earnings ($Millions) 1,656 278.1 0.89 2,510.6 -45,454.0 51,669.0 Stockholders' Equity ($Millions) 1,656 730.4 91.84 2,593.1 -5,916.0 37,746.0 Book Value of Equity ($Millions) 1,656 714.0 90.46 2,572.7 -5,916.0 37,746.0 Common Shares Outstanding (Millions) 1,656 95.7 24.37 303.8 0.007 4,968.0 Book Value Per Share ($/share) 1,656 6.21 4.63 20.00 -618.6 217.5

Income Statement Accounts: Sales ($Millions) 1,656 2,151.7 245.1 7,403.8 0.000 112,937.0 EBIT ($Millions) 1,656 134.4 3.31 642.7 -7,613.0 11,595.0 Net Income ($Millions) 1,656 -28.19 -3.81 1,133.8 -38,468.0 7,230.0

Cash Flow Statement Accounts: Cash Flows from Operations ($Millions) 1,656 168.8 6.80 733.5 -3,657.0 12,315.0

Liquidity Analysis: Working Capital ($Millions) 1,656 178.2 38.66 808.8 -6,723.2 17,392.0 Current Cash Debt Coverage Ratio 1,656 0.12 0.16 8.35 -10.36 335.7 Current Ratio 1,656 2.48 1.71 3.38 0.01 64.67 Quick Ratio 1,656 1.79 1.05 3.21 0.01 63.51

Activity Analysis: Receivable Turnover Ratio 1,656 10.23 5.95 28.27 0.000 962.0

Profitability Analysis: Return on Equity (ROE) 1,656 -86.97 -3.35 652.5 -17,434.1 623.3 Gross Profit Ratio 1,656 -21.55 32.38 997.2 -34,394.3 100.0 Net Profit Ratio 1,656 -1.78 -0.02 27.34 -1,005.6 1.72

Other Values: Restructuring Charge Amount ($Millions) 1,656 69.24 8.26 496.1 0.03 14,000.0 Price 3 mos. after fiscal year-end ($) 1,656 14.45 8.60 32.16 0.002 884.4 Returns 3 mos. after fiscal year-end (%) 1,656 0.19 -0.09 2.86 -1.00 70.19 Begley Z-Score 1,656 -118.3 3.98 812.4 -26,140.5 130.7 Altman Z-Score 1,656 0.91 2.50 16.03 -413.0 203.3 Ohlson_Prob 1,656 0.47 0.33 0.38 0 1

65

TABLE 3: Descriptive Statistics a Panel B: Full Sample of Firm-Event Observations Firm-Event Standard Variable Obsv Mean Median Deviation Minimum Maximum Balance Sheet Accounts: Current Assets ($Millions) 2,164 1,054.3 175.7 3,401.0 0.03 48,489.0 10,565. Total Assets ($Millions) 2,164 3,078.3 370.6 4 0.93 242,223.0 Current Liabilities ($Millions) 2,164 853.0 93.67 6,896.8 0.27 139,025.0 Total Liabilities ($Millions) 2,164 2,077.7 202.7 6,896.8 0.27 139,025.0 Retained Earnings ($Millions) 2,164 320.4 7.62 3,384.7 -99,856.0 51,669.0 Stockholders' Equity ($Millions) 2,164 1,010.0 137.3 4,032.8 -5,916.0 103,198.0 Book Value of Equity ($Millions) 2,164 989.3 132.4 4,013.6 -5,916.0 103,198.0 Common Shares Outstanding (Millions) 2,164 115.6 30.31 359.2 0.01 7,629.0 Book Value Per Share ($/share) 2,164 6.92 5.21 8.31 -37.18 75.01 Income Statement Accounts: Sales ($Millions) 2,164 2,883.0 418.1 8,571.0 0.000 112,937.0 EBIT ($Millions) 2,164 223.3 9.94 959.6 -7,613.0 13,044.0 Net Income ($Millions) 2,164 15.24 -3.38 1,270.2 -38,468.0 24,730.0 Cash Flow Statement Accounts: Cash Flows from Operations ($Millions) 2,164 265.9 14.34 1,075.1 -3,657.0 24,730.0 Liquidity Analysis: Working Capital ($Millions) 2,164 201.4 53.96 1,164.2 -33,780.0 17,392.0 Current Cash Debt Coverage Ratio 2,164 0.15 0.19 7.31 -10.36 335.67 Current Ratio 2,164 2.38 1.71 2.91 0.01 47.21 Quick Ratio 2,164 1.67 1.04 2.68 0.01 47.21 Activity Analysis: Receivable Turnover Ratio 2,164 10.03 5.95 27.11 0.000 962.0 Profitability Analysis: Return on Equity (ROE) 2,164 -43.87 -0.88 169.1 -1,940.2 188.0 Gross Profit Ratio 2,164 28.36 32.57 71.74 -1,302.1 97.33 Net Profit Ratio 2,164 -0.44 -0.01 2.12 -45.62 0.96 Restructuring Charge Amount ($Millions) 2,164 88.70 10.00 542.7 0.07 14,000.0 Price 3 mos. after fiscal year-end ($) 2,164 16.19 10.22 30.38 0.01 884.4 Returns 3 mos. after fiscal year- end (%) 2,164 0.09 -0.05 0.89 -0.97 8.23 Begley Z-Score 2,164 -84.60 7.57 689.2 -26,140.5 130.7 Altman Z-Score 2,164 1.61 2.62 13.71 -413.0 209.3 Ohlson_Prob 2,164 0.43 0.25 0.37 0 1

a The 2,164 firm-event observations in this table are the portion of the total 2,202 firm-event observations for which full data was available for all the financial variables included in this table.

66

TABLE 4 a Panel A: Distress Classification Contingency Table for Firms

Begley Distressed Non-Distressed

b Distressed 613 (37.02%) 286 (17.27%) 899 (54.29%) Altman

Non-Distressed 154 (9.30%) 603 (36.41%) 757(45.71%)

767 (46.32%) 889 (53.68%) 1,656 firms

TABLE 4 Panel B: Distress Classification Contingency Table for Firm-Event Observations

Begley Distressed Non-Distressed

Distressed 723 (33.41%) 389 (17.98%) 1,112 (51.39%c) Altman

Non-Distressed 194 (8.96%) 858 (39.65%) 1,052 (48.61%) 2,164 firm-event d 917 (42.38%) 1,247 (57.62%) observations

a These tables are used to determine the percentage of firms (firm-event observations) that are classified as being in financial distress by the Altman model and by the Begley model. The model demonstrates the number of observations that are classified differently by the two models. b The Altman and Ohlson models classified 1,216 firms into the same category. The combined models identified 613 distressed firms and 603 non-distressed firms. This 1,216 firms classified the same by the two models comprise the sample to be used in the Ohlson model to calculate the probability of distress (p- value). c The Altman model classifies 52% of the firms (51% of the firm-event observations) as distressed and 46% of the firms (49% of the firm-event observations) as non-distressed. The Begley model classifies 46% of the firms (42% of the firm-event observations) as distressed and 54% of the firms (58% of the firm-event observations) as non-distressed. d Thirty-eight firm-event observations were lost due to missing data for one or more of the variables needed for the Altman and Begley models.

67

TABLE 5: Descriptive Statistics Panel A: Firms Classified the Same by the Begley (1996) and Altman (1968) Models Standard Variable Firms Mean Median Deviation Minimum Maximum Balance Sheet Accounts: Current Assets ($Millions) 1,202 843.4 123.9 2,698.3 0.03 40,996.0 Total Assets ($Millions) 1,202 2,363.8 249.7 8,350.8 0.93 121,783.0 Current Liabilities ($Millions) 1,202 640.4 67.78 2,206.1 0.17 33,507.0 Total Liabilities ($Millions) 1,202 1,546.4 126.1 5,127.2 0.27 65.75 Retained Earnings ($Millions) 1,202 204.2 -3.37 4,224.9 -99,586.0 51,669.0 Stockholders' Equity ($Millions) 1,202 826.1 89.91 3,587.4 -5,916.0 65,377.0 Book Value of Equity ($Millions) 1,202 807.9 87.82 3,570.9 -5,916.0 65,158.0 Common Shares Outstanding (Millions) 1,202 115.5 27.43 408.4 0.01 7,629.0 Book Value Per Share ($/share) 1,202 5.64 4.25 7.19 -37.18 55.63

Income Statement Accounts: Sales ($Millions) 1,202 2,460.6 261.0 7,938.4 0.000 112,937.0 EBIT ($Millions) 1,202 177.5 4.24 879.3 -7,613.0 13,044.0 Net Income ($Millions) 1,202 -12.15 -4.82 1,331.3 -38,468.0 7,230.0

Cash Flow Statement Accounts: Cash Flows from Operations ($Millions) 1,202 199.8 7.13 912.8 -3,657.0 12,315.0

Liquidity Analysis: Working Capital ($Millions) 1,202 203.0 44.20 805.2 -5,864.0 14,366.0 Current Cash Debt Coverage Ratio 1,199 0.14 0.16 9.77 -10.36 335.7 Current Ratio 1,202 2.60 1.76 3.49 0.01 47.21 Quick Ratio 1,202 1.90 1.07 3.23 0.01 47.21

Activity Analysis: Receivable Turnover Ratio 1,179 10.87 5.90 34.12 0.000 962.0

Profitability Analysis: Return on Equity (ROE) 1,061 -49.58 -2.86 159.8 -1,840.4 188.0 Gross Profit Ratio 1,179 26.35 32.58 76.81 -1,154.7 97.33 Net Profit Ratio 1,183 -0.62 -0.02 2.58 -45.62 0.76

Restructuring Charge Amount ($Millions) 1,202 50.68 8.21 160.8 0.07 2,100.0 Price 3 mos. after fiscal year-end ($) 1,202 15.07 8.41 36.09 0.01 884.4 Returns 3 mos. after fiscal year-end (%) 1,202 0.09 -0.03 0.88 -0.97 8.23 Begley Z-Score 1,202 -140.7 1.01 912.6 -26,140.5 130.7 Altman Z-Score 1,202 -0.03 2.69 16.16 -413.0 60.83 Ohlson_Prob 1,202 0.50 0.33 0.41 0 1

68

TABLE 5: Descriptive Statistics Panel B: Firm-Event Observations Classified the Same by the Begley and Altman Models Firm- Event Standard Variable Obsv Mean Median Deviation Minimum Maximum Balance Sheet Accounts: Current Assets ($Millions) 1,581 1,042.2 172.3 3,249.2 0.03 41,338.0 Total Assets ($Millions) 1,581 2,821.6 344.9 9,104.8 0.93 121,783.0 Current Liabilities ($Millions) 1,581 812.5 89.36 2,745.2 0.17 39,372.0 Total Liabilities ($Millions) 1,581 1,892.6 187.1 5,971.5 0.27 71,610.0 Retained Earnings ($Millions) 1,581 301.7 3.61 3,863.2 -99,586.0 51,669.0 Stockholders' Equity ($Millions) 1,581 940.7 120.4 3,534.2 -5,916.0 65,377.0 Book Value of Equity ($Millions) 1,581 920.8 115.9 3,511.9 -5,916.0 65,158.0 Common Shares Outstanding (Millions) 1,581 125.2 30.80 398.3 0.01 7,929.0 Book Value Per Share ($/share) 1,581 6.09 4.71 7.53 -37.18 55.63 Income Statement Accounts: Sales ($Millions) 1,581 2,954.9 412.2 8,700.2 0.000 112,937.0 EBIT ($Millions) 1,581 234.6 9.99 1,010.0 -7,613.0 13,044.0 Net Income ($Millions) 1,581 9.23 -3.47 1,288.1 -38,468.0 7,690.0 Cash Flow Statement Accounts: Cash Flows from Operations ($Millions) 1,581 266.0 12.48 1,092.9 -3,657.0 12,315.0 Liquidity Analysis: Working Capital ($Millions) 1,581 229.70 55.40 837.2 -5,864.0 14,366.0 Current Cash Debt Coverage Ratioa 1,359 0.14 0.18 8.54 -10.36 335.7 Current Ratio 1,581 2.49 1.74 3.21 0.01 47.21 Quick Ratio 1,581 1.78 1.05 2.98 0.01 47.21 Activity Analysis: Receivable Turnover Ratio 1,359 10.27 5.95 30.10 0.000 962.0 Profitability Analysis: Return on Equity (ROE) 1,359 -42.09 0.09 147.9 -1,840.4 188.0 Gross Profit Ratio 1,359 28.93 32.80 75.94 -1,154.7 97.33 Net Profit Ratio 1,359 -0.57 -0.01 2.42 -45.62 0.86 Restructuring Charge Amount ($Millions) 1,581 82.90 10.00 419.1 0.07 8,000.0 Price 3 mos. after fiscal year-end ($) 1,581 16.54 10.13 34.06 0.01 884.4 Returns 3 mos. after fiscal year-end (%) 1,581 0.06 -0.05 0.83 -0.97 8.2 Begley Z-Score 1,581 -111.2 6.28 800.1 -26,140.5 130.7 Altman Z-Score 1,581 0.76 2.83 14.49 -413.0 76.6 Ohlson_Prob 1,581 0.46 0.23 0.41 0 1

a All variables with 1,359 firm-event observations are only used in equation 1.4.

69

TABLE 6: Independent Sample t-tests for Mean Differences Panel A: Comparison of Non-Distressed Firms with Distressed Firmsa 2-tailed Non- t- Wilcoxon Variable Distressed n Distressed n statistic p-value Z p-value Means Means Balance Sheet Accounts: Current Assets 1,243.7 600 246.41 613 7.51 <0.001 -18.38 <0.001 Total Assets 3,082.7 600 800.5 613 6.45 <0.001 -17.70 <0.001 Current Liabilities 913.3 600 184.9 613 6.87 <0.001 -16.66 <0.001 Total Liabilities 1,905.3 600 633.1 613 5.47 <0.001 -16.09 <0.001 Retained Earnings 917.2 600 -411.6 613 8.53 <0.001 -29.48 <0.001 Stockholders' Equity 1,177.3 600 167.3 613 7.22 <0.001 -20.32 <0.001 Book Value of Equity 1,162.1 600 152.1 613 7.27 <0.001 -20.56 <0.001 Common Shares Outstanding 152.9 600 51.61 613 5.38 <0.001 -10.31 <0.001 BVPS 9.13 600 1.42 613 6.24 <0.001 -19.72 <0.001 Income Statement Accounts: Sales 3,846.9 600 524.5 613 7.97 <0.001 -21.04 <0.001 EBIT 314.3 600 -37.06 613 9.25 <0.001 -24.00 <0.001 Net Income 121.6 600 -190.4 613 4.28 <0.001 -21.26 <0.001 Cash Flows Statement Amounts: Cash Flows from Operations 310.9 600 5.78 613 7.61 <0.001 -21.39 <0.001 Liquidity Analysis: Working Capital 330.4 600 61.55 613 6.79 <0.001 -16.24 <0.001 Current Cash Debt Coverage Ratio 0.96 600 -0.74 613 3.02 0.003 -21.18 <0.001 Current Ratio 2.26 600 2.91 613 -3.04 0.002 -3.49 <0.001 Quick Ratio 1.42 600 2.39 613 -4.80 <0.001 -0.88 0.38 Activity Analysis: Receivable Turnover Ratio 12.71 600 8.39 613 2.40 0.02 -7.98 <0.001 Profitability Analysis: ROE 5.17 600 -144.9 613 6.21 <0.001 -23.80 <0.001 Gross Profit Ratio 35.26 600 -102.5 613 2.09 0.04 -1.14 0.25 Net Profit Ratio 0.02 600 -2.98 613 3.94 <0.001 -25.20 <0.001 Other Values: Charge Amount 53.61 600 20.41 613 4.42 <0.001 -9.95 <0.001 Price 3 mos. after fiscal year-end 21.84 600 7.24 613 7.18 <0.001 -21.43 <0.001 Returns 0.14 600 0.07 613 0.54 0.59 -10.83 <0.001 Begley Z-Score 33.87 600 -334.8 613 7.04 <0.001 -30.15 <0.001 Altman Z-Score 4.50 600 -4.99 613 10.24 <0.001 -30.15 <0.001 Ohlson_Prob 0.18 600 0.82 613 -48.43 <0.001 -27.19 <0.001

a All t-tests are two tailed because the expected signs are not known in all cases. Data are for the year of restructuring for each firm. The sample in this table includes only observations classified into the same distress classification (0 or 1) by both the Altman (1968) and Begley (1996) models.

70

TABLE 6: Panel B a Comparison of Distressed, Non-Bankrupt Firms with Distressed, Bankrupt Firms Distressed, Two- Non- Distressed, tailed t- Wilcoxo Variable Bankrupt n Bankrupt n statistic p-value n p-value Means Means Balance Sheet Accounts: Current Assets 252.13 545 200.6 68 0.68 0.50 -0.92 0.36 Total Assets 829.49 545 567.9 68 1.09 0.28 -1.88 0.06 Current Liabilities 184.46 545 188.1 68 -0.06 0.95 -3.19 0.001 Total Liabilities 637.11 545 601.3 68 0.15 0.89 -3.91 <0.001 Retained Earnings -412.7 545 -403.4 68 -0.06 0.95 -1.95 0.05 Stockholders' Equity 192.38 545 -33.46 68 4.40 <0.001 -4.56 <0.001 Book Value of Equity 176.0 545 -39.39 68 4.38 <0.001 -4.82 <0.001 Common Shares Outstanding 53.52 545 36.31 68 1.51 0.13 -0.10 0.92 BVPS 1.63 545 -0.23 68 0.73 0.47 -3.63 <0.001 Income Statement Accounts: Sales 532.5 545 460.7 68 0.49 0.63 -3.03 0.002 EBIT -35.44 545 -49.98 68 0.53 0.60 -1.34 0.18 Net Income -187.6 545 -212.6 68 0.24 0.81 -5.27 <0.001 Cash Flows Statement Amounts: Cash Flows from Operations 8.09 545 -12.74 68 1.53 0.13 -0.23 0.82 Liquidity Analysis: Working Capital 67.68 545 12.43 68 1.22 0.23 -2.53 0.01 Current Cash Debt Coverage Ratio -0.74 545 -0.70 68 -0.24 0.82 -0.42 0.67 Current Ratio 3.04 545 1.83 68 2.91 0.004 -3.77 <0.001 Quick Ratio 2.52 545 1.35 68 2.96 0.004 -4.56 <0.001 Activity Analysis: Receivable Turnover Ratio 8.01 545 11.36 68 -1.31 0.20 -1.17 0.24 Profitability Analysis: ROE -111.2 545 -539.4 68 1.55 0.13 -3.48 0.001 Gross Profit Ratio -99.75 545 -123.7 68 0.18 0.86 -2.90 0.004 Net Profit Ratio -2.74 545 -4.86 68 1.14 0.26 -1.57 0.12 Other Values: Charge Amount 17.47 545 43.98 68 -0.85 0.40 -0.50 0.62 Price 3 mos. after fiscal year- end 7.66 545 3.92 68 2.33 0.02 -4.10 <0.001 Returns 0.16 545 -0.63 68 5.31 <0.001 -6.63 <0.001 Begley Z-Score -329.3 545 -378.7 68 0.44 0.66 -0.01 0.99 Altman Z-Score -4.76 545 -6.83 68 0.97 0.33 -30.15 <0.001 Ohlson_Prob 0.81 545 0.88 68 -2.62 0.01 -27.19 <0.001

a All observations in this table are classified into the same category by the Altman and Begley models. All t-tests are two tailed because the expected signs are not known in all cases. Data are for the year of restructuring for each firm. The item names in bold and italics are all included directly in regression models in this essay.

71

TABLE 7: Independent Sample t-tests and Wilcoxon Z-statistics for Mean Differences a Panel A: Comparison of Non-Distressed Firm-Event Observations with Distressed Firm-Event Observations Non- 2-tailed t- p- Wilcoxon Z- p- Variable Distressed n Distressed n statistic value Statistic value Means Means Balance Sheet Accounts: Current Assets 1,598.7 858 381.7 723 8.09 <0.001 -19.67 <0.001 Total Assets 4,066.8 858 1,343.9 723 6.26 <0.001 -18.94 <0.001 Current Liabilities 1,237.8 858 307.8 723 7.29 <0.001 -17.64 <0.001 Total Liabilities 2,604.6 858 1,040.2 723 5.44 <0.001 -16.68 <0.001 Retained Earnings 1,112.5 858 -660.6 723 9.11 <0.001 -33.62 <0.001 Stockholders' Equity 1,468.2 858 314.7 723 6.84 <0.001 -22.85 <0.001 Book Value of Equity 1,451.3 858 291.2 723 6.93 <0.001 -23.08 <0.001 Common Shares Outstanding 173.6 858 67.76 723 5.56 <0.001 -11.23 <0.001 BVPS 9.15 858 2.43 723 19.55 <0.001 -22.67 <0.001 Income Statement Accounts: Sales 4,708.1 858 874.3 723 9.61 <0.001 -22.36 <0.001 EBIT 441.01 858 -10.33 723 9.67 <0.001 -26.23 <0.001 Net Income 190.7 858 -206.1 723 5.84 <0.001 -24.22 <0.001 Cash Flows Statement Amounts: Cash Flows from Operations 456.0 858 41.05 723 8.17 <0.001 -23.97 <0.001 Liquidity Analysis: Working Capital 360.9 858 73.97 723 7.32 <0.001 -17.46 <0.001 Current Cash Debt Coverage Ratio 0.80 858 -0.65 723 3.66 <0.001 -23.98 <0.001 Current Ratio 2.16 858 2.88 723 -4.23 <0.001 -2.71 0.007 Quick Ratio 1.35 858 2.30 723 -6.09 <0.001 -1.93 0.05 Activity Analysis: Receivable Turnover Ratio 12.31 858 7.79 723 3.19 <0.001 -8.43 <0.001 Profitability Analysis: ROE 5.37 858 -114.9 723 13.27 <0.001 -26.50 <0.001 Gross Profit Ratio 35.57 858 16.40 723 4.54 <0.001 -0.90 0.370 Net Profit Ratio 0.02 858 -1.28 723 9.94 <0.001 -27.92 <0.001 Other Values: Charge Amount 114.3 858 37.77 723 3.60 <0.001 -10.08 <0.001 Price 3 mos. after fiscal year-end 24.34 858 7.26 723 10.43 <0.001 -25.25 <0.001 Returns 0.17 858 -0.07 723 5.48 <0.001 -12.81 <0.001 Begley Z-Score 34.82 858 -284.6 723 7.40 <0.001 -34.30 <0.001 Altman Z-Score 4.71 858 -3.94 723 11.47 <0.001 -34.30 <0.001 Ohlson_Prob 0.15 858 0.82 723 -54.46 <0.001 -29.97 <0.001

a All t-tests are two tailed because the expected signs are not known in all cases. Data are for the year of restructuring for each firm. The sample in this table includes only observations classified into the same distress classification (0 or 1) by the Altman (1968) and Begley (1996) models.

72

TABLE 7: Panel B a Comparison of Distressed, Non-Bankrupt Firm-Event Observations with Distressed, Bankrupt Firm-Event Observations Non- Bankrupt, Bankrupt, Two-tailed Wilcoxon p- Variable Distressed n Distressed n t-statistic p-value Z-Statistic value Means Means Balance Sheet Accounts: Current Assets 401.9 631 243.4 92 2.03 0.04 -1.13 0.26 Total Assets 1,438.0 631 698.6 92 2.36 0.02 -2.02 0.04 Current Liabilities 315.3 631 256.2 92 0.83 0.41 -3.54 <0.001 Total Liabilities 1,082.1 631 753.2 92 1.26 0.21 -4.20 <0.001 Retained Earnings -693.2 631 -437.0 92 -1.24 0.21 -1.55 0.12 Stockholders' Equity 367.4 631 -47.01 92 3.89 <0.001 -5.54 <0.001 Book Value of Equity 342.4 631 -59.47 92 3.83 <0.001 -5.73 <0.001 Common Shares Outstanding 72.30 631 36.66 92 2.87 0.004 -0.35 0.73 BVPS 2.73 631 0.39 92 2.10 0.04 -4.84 <0.001 Income Statement Accounts: Sales 911.4 631 620.0 92 1.62 0.11 -3.78 <0.001 EBIT -6.79 631 -34.59 92 0.99 0.32 -0.98 0.33 Net Income -209.6 631 -181.8 92 -0.31 0.76 -4.37 <0.001 Cash Flows Statement Amounts: Cash Flows from Operations 48.75 631 -11.79 92 2.61 0.009 -0.38 0.70 Liquidity Analysis: Working Capital 86.63 631 -12.84 92 2.40 0.02 -3.25 0.001 Current Cash Debt Coverage Ratio -0.66 631 -0.55 92 -0.72 0.47 -0.45 0.66 Current Ratio 3.05 631 1.70 92 4.22 <0.001 -4.88 <0.001 Quick Ratio 2.47 631 1.19 92 4.21 <0.001 -6.00 <0.001 Activity Analysis: Receivable Turnover Ratio 7.46 631 9.94 92 -1.92 0.06 -2.59 0.01 Profitability Analysis: ROE -100.7 631 -265.7 92 2.95 0.005 -3.67 <0.001 Gross Profit Ratio 19.35 631 -3.62 92 1.43 0.16 -3.99 <0.001 Net Profit Ratio -1.15 631 -2.18 92 1.48 0.14 -0.72 0.47 Other Values: Charge Amount 33.77 631 65.09 92 -0.89 0.38 -0.55 0.58 Price 3 mos. after fiscal year-end 7.78 631 3.63 92 3.04 0.002 -5.62 <0.001 Returns 0.01 631 -0.60 92 10.04 <0.001 -7.41 <0.001 Begley Z-Score -264.1 631 -425.2 92 1.01 0.32 -0.54 0.59 Altman Z-Score -3.55 631 -6.67 92 1.24 0.22 -1.45 0.15 Ohlson_Prob 0.81 631 0.89 92 -2.88 0.005 -3.87 <0.001

a All observations in this table are classified into the same category by the Altman and Begley models. All t-tests are two tailed because the expected signs are not known in all cases. Data are for the year of restructuring for each firm. The item names in bold and italics are all included directly in regression models in this essay.

73

TABLE 8: Ohlson (1980) Logistic Regression Model Results for Distressed and Non-Distressed Companies for All Firmsa Panel A: Ohlson Regression Model Results Coefficient Variable Expected Sign Estimate Wald Statistic P-value Intercept 7.876 24.01 <0.001 SIZE (-) -0.365 34.02 <0.001 TLTA (+) -0.626 1.13 0.29 WCTA (-) -1.153 1.35 0.25 CLCA (+) 1.009 2.78 0.096 OENEG (+/-) -5.627 19.09 <0.001 NITA (-) -8.040 73.82 <0.001 FUTL (-) -1.057 17.02 <0.001 INTWO (+) -1.496 39.91 <0.001 CHIN (-) 0.263 21.33 <0.001

-2 Log Likelihood 682.4 Cox & Snell R Square 0.56 Nagelkerke R Square 0.74 N = 1,202 firms

Panel B: Classification Table for Predicting Financial Distressb Predicted DISTRESS Observed 0 1 Percentage Correct 0 567 46 92.3% 1 84 505 85.7% Overall % 89.1%

a Equation 1.3: DISTRESSi,t = a0 + a1 SIZE i,t + a2 TLTA i,t + a3 WCTA i,t + a4 CLCA i,t + a5 NITA i,t + a6 FUTL + a7 INTWOi,t + a8 OENEGi,t + a9 CHIN + ui,t, where DISTRESS equals 1 if a firm is in distress and 0 otherwise, SIZE is the log of total sales, TLTA is total liabilities divided by total assets, WCTA is working capital divided by total assets, CLCA is current liabilities divided by current assets, OENEG equals 1 if owners’ equity is negative and equals 0 otherwise, NITA is net income divided by total assets, FUTL is cash flows from operations divided by total liabilities, INTWO equals 1 if net income was negative over the last two years and equals 0 otherwise, and CHIN = (NIt – NIt-1) / ( | NIt | + | NIt-1 | ). The sample in this table includes only observations classified into the same distress group (0 or 1) by the Altman (1968) and Begley (1996) models. b An item is classified as correct if Ohlson’s model predicts the same distress group as Begley’s model. The cut-off point for this table is 0.5.

74

TABLE 9: Ohlson (1980) Logistic Regression Model Results for Distressed and Non-Distressed a Companies for All Firm-Event Observations Panel A: Ohlson Regression Model Results Coefficient Variable Expected Sign Estimate Wald Statistic P-value Intercept 8.667 29.59 <0.001 SIZE (-) -0.307 30.52 <0.001 TLTA (+) -0.537 0.91 0.34 WCTA (-) -1.336 2.05 0.15 CLCA (+) 0.917 2.79 0.095 OENEG (+/-) -6.791 26.17 <0.001 NITA (-) -9.199 98.35 <0.001 FUTL (-) -1.251 22.93 <0.001 INTWO (+) -1.451 45.24 <0.001 CHIN (-) 0.303 27.31 <0.001

-2 Log Likelihood 865.0 Cox & Snell R Square 0.561 Nagelkerke R Square 0.750 N = 1,581 firm-event observations

Panel B: Classification Table for Predicting Financial Distressb Predicted DISTRESS Observed 0 1 Percentage Correct 0 806 54 93.6% 1 114 607 83.9% Overall % 89.2%

a Equation 1.3: DISTRESSi,t = a0 + a1 SIZE i,t + a2 TLTA i,t + a3 WCTA i,t + a4 CLCA i,t + a5 NITA i,t + a6 FUTL + a7 INTWOi,t + a8 OENEGi,t + a9 CHIN + ui,t, where DISTRESS equals 1 if a firm is in distress and 0 otherwise, SIZE is the log of total sales, TLTA is total liabilities divided by total assets, WCTA is working capital divided by total assets, CLCA is current liabilities divided by current assets, OENEG equals 1 if owners’ equity is negative and equals 0 otherwise, NITA is net income divided by total assets, FUTL is cash flows from operations divided by total liabilities, INTWO equals 1 if net income was negative over the last two years and equals 0 otherwise, and CHIN = (NIt – NIt-1) / ( | NIt | + | NIt-1 | ). The sample in this table includes only observations classified into the same distress classification (0 or 1) by the Altman (1968) and Begley (1996) models. b An item is classified as correct if Ohlson’s model predicts the same distress group as Begley’s model. The cut-off point for this table is 0.5.

75

TABLE 10: Predicting Bankruptcy Using Equation 1.4 for Those Restructuring Firms Classified as Distresseda Panel A: Bankruptcy Prediction Resultsb

Coefficient Variable Expected Sign Estimate Wald Statistic P-value Intercept -2.652 60.31 <0.001 Receivable Turnover (-) 0.011 1.64 0.20 Quick Ratio (-) -0.173 1.96 0.16 Current Cash Debt Coverage Ratio (-) 0.314 2.01 0.16 BVPS (-) 0.069 7.36 0.007 Gross Profit Ratio (-) -0.002 1.36 0.24 ROE (-) -0.002 7.71 0.005 Net Profit Ratio (-) -0.095 2.95 0.09

-2 Log Likelihood 228.5 P = 0.000 Cox & Snell R Square 0.06 Nagelkerke R Square 0.14 N = 430 firmsc

Panel B: Bankruptcy Prediction (n = 430 firms) Predicted BANKRUPT Observed 0 1 Percentage Correct 0 313 79 79.8% 1 16 22 57.9% Overall % 77.9%

a Equation 1.4 is as follows: BANKRUPTi,t = a0 + a1 COVERAGEi,t + a2 QR + a3 REC_TURN i,t + a4 BVPS i,t + a5 GPRATIO + a6 NPRATIO i,t + a7 ROE i,t + ui,t, where BANKRUPT equals 1 if the firm goes bankrupt in the three years subsequent to restructuring and 0 otherwise, COVERAGE is the current cash debt coverage ratio (net cash from operations divided by average current liabilities), QR is the quick ratio, REC_TURN is accounts receivable turnover, BVPS is book value per share, GPRATIO is gross profit divided by sales, NPRATIO is net income divided by sales, and ROE is return on equity (net income less preferred dividends divided by average common stockholders’ equity). The appropriate cutoff point for the contingency table in panel B is 0.1. This cutoff point is chosen to minimize the number of observations misclassified. The sample in this table includes only observations classified into the same distress classification (0 or 1) by the Altman (1968) and Begley (1996) models. b The data used in this model for each firm are from the most recent year prior to the firm’s restructuring. c The sample in this table consists of the 430 restructuring firms classified by both Altman (1968) and Begley (1996) as distressed that have all the data necessary for Equation 1.4.

76

TABLE 11: Predicting Bankruptcy Using Equation 1.4 for Those Restructuring Firm-Event a Observations Classified as Distressed Panel A: Bankruptcy Prediction Resultsb

Variable Expected Sign Coefficient Estimate Wald Statistic P-value Intercept -2.583 65.46 <0.001

Receivable Turnover (-) 0.012 2.15 0.14 Quick Ratio (-) -0.207 2.75 0.097 Current Cash Debt Coverage Ratio (-) 0.306 2.19 0.14 BVPS (-) 0.049 5.10 0.02 Gross Profit Ratio (-) -0.001 0.83 0.36 ROE (-) -0.002 9.05 0.003 Net Profit Ratio (-) -0.084 2.48 0.12

-2 Log Likelihood 277.8 P = 0.000 Cox & Snell R Square 0.054 Nagelkerke R Square 0.121 N = 522 firm-event observationsc

Panel B: Bankruptcy Prediction (n = 522 firm-event observations) Predicted BANKRUPT Observed 0 1 Percentage Correct 0 372 105 78.0% 1 20 25 55.6% Overall % 76.1%

a Equation 1.4 is as follows: BANKRUPTi,t = a0 + a1 COVERAGEi,t + a2 QR + a3 REC_TURN i,t + a4 BVPS i,t + a5 GPRATIO + a6 NPRATIO i,t + a7 ROE i,t + ui,t, where BANKRUPT equals 1 if the firm goes bankrupt in the three years subsequent to restructuring and 0 otherwise, COVERAGE is the current cash debt coverage ratio (net cash from operations divided by average current liabilities), QR is the quick ratio, REC_TURN is accounts receivable turnover, BVPS is book value per share, GPRATIO is gross profit divided by sales, NPRATIO is net income divided by sales, and ROE is return on equity (net income less preferred dividends divided by average common stockholders’ equity). Note that 97 of the sample firms restructured more than once during the period from 1993 to 2003. The appropriate cutoff point for the contingency table in panel B is 0.1. This cutoff point is chosen to minimize the number of observations misclassified. The sample in this table includes only observations classified into the same distress classification (0 or 1) by the Altman (1968) and Begley (1996) models. b The data used in this model for each firm are from the most recent year prior to the firm’s restructuring. c The sample in this table consists of the 522 restructuring firm-event observations, classified by both Altman (1968) and Begley (1996) as distressed that have all the data necessary for Equation 1.4.

77

TABLE 12: Predicting Bankruptcy Using Equation 1.4 for All Restructuring Firms a Panel A: Bankruptcy Prediction Results Coefficient Variable Expected Sign Estimate Wald Statistic P-value Intercept -3.346 116.2 <0.001 Receivable Turnover (-) 0.004 0.43 0.51 Quick Ratio (-) -0.154 1.85 0.17 Current Cash Debt Coverage Ratio (-) -0.064 0.14 0.71 BVPS (-) 0.040 4.69 0.03 Gross Profit Ratio (-) -0.001 0.97 0.32 ROE (-) -0.002 14.02 <0.001 Net Profit Ratio (-) -0.081 2.31 0.13 -2 Log Likelihood 354.1 P = 0.000 Cox & Snell R Square 0.032 Nagelkerke R Square 0.101 N = 1,022 firmsb

Panel B: Bankruptcy Prediction (n = 1,022 firms) Predicted BANKRUPT Observed 0 1 Percentage Correct 0 952 22 97.7% 1 36 12 25.0% Overall % 94.3%

a Equation 1.4 is as follows: BANKRUPTi,t = a0 + a1 COVERAGEi,t + a2 QR + a3 REC_TURN i,t + a4 BVPS i,t + a5 GPRATIO + a6 NPRATIO i,t + a7 ROE i,t + ui,t, where BANKRUPT equals 1 if the firm goes bankrupt in the three years subsequent to restructuring and 0 otherwise, COVERAGE is the current cash debt coverage ratio (net cash from operations divided by average current liabilities), QR is the quick ratio, REC_TURN is accounts receivable turnover, BVPS is book value per share, GPRATIO is gross profit divided by sales, NPRATIO is net income divided by sales, and ROE is return on equity (net income less preferred dividends divided by average common stockholders’ equity). The appropriate cutoff point for the contingency table in panel B is 0.1. This cutoff point is chosen to minimize the number of observations misclassified. The sample in this table includes only observations classified into the same distress classification (0 or 1) by the Altman (1968) and Begley (1996) models. b The sample in this table consists of the 1,022 restructuring firms that have all the data necessary for Equation 1.4.

78

TABLE 13: Predicting Bankruptcy Using Equation 1.4 for All Restructuring Firm-Event Observationsa Panel A: Bankruptcy Prediction Results

Variable Expected Sign Coefficient Estimate Wald Statistic P-value Intercept -3.161 122.9 <0.001 Receivable Turnover (-) 0.004 0.63 0.43 Quick Ratio (-) -0.207 3.06 0.08 Current Cash Debt Coverage Ratio (-) -0.092 0.33 0.57 BVPS (-) 0.020 1.28 0.26 Gross Profit Ratio (-) -0.001 0.87 0.35 ROE (-) -0.002 14.25 <0.001 Net Profit Ratio (-) -0.068 1.76 0.19 -2 Log Likelihood 465.5 P = 0.000 Cox & Snell R Square 0.026 N = 1,359 firm-event Nagelkerke R Square 0.084 observationsb

Panel B: Bankruptcy Prediction (n = 1,359 firm-event observations) Predicted BANKRUPT Observed 0 1 Percentage Correct 0 1274 25 98.1% 1 48 12 20.0% Overall % 94.6%

a Equation 1.4 is as follows: BANKRUPTi,t = a0 + a1 COVERAGEi,t + a2 QR + a3 REC_TURN i,t + a4 BVPS i,t + a5 GPRATIO + a6 NPRATIO i,t + a7 ROE i,t + ui,t, where BANKRUPT equals 1 if the firm goes bankrupt in the three years subsequent to restructuring and 0 otherwise, COVERAGE is the current cash debt coverage ratio (net cash from operations divided by average current liabilities), QR is the quick ratio, REC_TURN is accounts receivable turnover, BVPS is book value per share, GPRATIO is gross profit divided by sales, NPRATIO is net income divided by sales, and ROE is return on equity (net income less preferred dividends divided by average common stockholders’ equity). Note that 364 of the sample firms restructured more than once during the period from 1993 to 2003. The appropriate cutoff point for the contingency table in panel B is 0.1. This cutoff point is chosen to minimize the number of observations misclassified. The sample in this table includes only observations classified into the same distress classification (0 or 1) by the Altman (1968) and Begley (1996) models. b The sample in this table consists of the 1,359 restructuring firm-event observations that have the data needed for Equation 1.4.

79

TABLE 14: Modified Ohlson Regression Model Results for Multinomial Logistic Regression Model for All Firm-Event Observations with Different Classification Groupsa Panel A: Ohlson Model Results

Variable Coefficient Estimate Wald Statistic P-value Intercept -20.587 602.0 <0.001 SIZE 0.183 5.88 0.02 TLTA 0.501 0.65 0.42 WCTA 1.253 1.70 0.19 CLCA -0.897 2.74 0.098 OENEG 1.120 13.87 <0.001 NITA 9.940 103.62 <0.001 FUTL 1.132 16.50 <0.001 INTWO -0.120 0.23 0.63 CHIN -0.152 2.50 0.11

N = 1,581 firm-event Pseudo R-Square Values: observations Cox & Snell 0.568 Nagelkerke 0.685 McFadden 0.474

-2 Log Likelihood 1,451.6 P-value = 0.000

Panel B: Classification Table for Predicting Financial Distress Predicted

Observed 0 1 2 Percent Correct 0 782 51 0 93.9% 1 112 523 6 81.6% 2 19 85 3 2.8% Overall % 57.7% 41.7% 0.6% 82.7%

a Equation 1.3 is as follows: DISTRESSi,t = a0 + a1 SIZE i,t + a2 TLTA i,t + a3 WCTA i,t + a4 CLCA i,t + a5 NITA i,t + a6 FUTL + a7 INTWOi,t + a8 OENEGi,t + a9 CHIN + ui,t, where the variables are as described in table 5, except for DISTRESS, which equals 2 for firms that ultimately file for bankruptcy, equals 1 for distressed firms that do not file within three years of restructuring, and equals 0 for non-distressed restructuring firms. The sample in this table includes only observations classified into the same distress classification (0 or 1) by the Altman (1968) and Begley (1996) models.

CHAPTER 4

Essay 2

The Value Relevance of Restructuring Charges for Firms with Varying Levels of Financial Health

CHAPTER 4

Essay 2: The Value Relevance of Restructuring Charges for Firms with Varying Levels of Financial Health

4.1 Introduction

The purpose of this essay is to examine the value relevance of restructuring charges for several different groups of firms undergoing the process. Prior studies find that restructuring charges are value relevant (e.g., Bens 2002), and this essay further examines the value relevance of these charges for companies restructuring for different reasons. Specifically, the primary incremental contribution of this essay is that it separately examines the value-relevance of restructuring charges for healthy firms utilizing the process to improve their efficiency and distressed firms restructuring to avoid bankruptcy filing.

The first section of this essay discusses the importance of restructuring charges in determining a firm’s stock price. In the second section, the hypotheses are developed. In the third section, the data set and the methodology utilized to test the hypotheses are discussed. The empirical results are described in the fourth section. Finally, the fifth section summarizes and draws conclusions.

4.2 Importance of Restructuring Charges

The research paradigm examined in this essay is the value relevance of restructuring charges. Carter (1998) compares a sample of restructuring firms to a sample of similarly performing non-restructuring firms and finds that operating performance improves in firms during years three through five after a restructuring. Many studies have determined that restructuring charges have a positive impact on price [Brickley and Van Drunen (1990), Martin

81 82

and Kensinger (1990), Francis et al. (1996), Bunsis (1997), Ballester et al. (1999), and Kross et al. (2001)]. However, some recent studies determine that restructuring charges may actually have a negative impact on price [Blackwell et al. (1990), Elliott and Hanna (1996), Carter (2000), Poon et al. (2001), Bens (2002), Lopez et al. (2002)]. Bartov et al. (1997) demonstrate that even in cases where it is statistically significant, the market’s reaction is very small for many prior studies. The mixed findings of these studies demonstrate the difficulty in interpreting the performance and market effects of an operational restructuring.

Other studies in the restructuring charge area examine the market’s response to several components that typically comprise an operational restructuring plan. Blackwell et al. (1990) and

Lin and Rozeff (1993) find negative market reactions to plant-closing announcements. Worrell et al. (1991), Lin and Rozeff (1993) and Elayan et al. (1998) find that the market reacts negatively to announcements of layoffs.1 Francis et al. (1996) determine that the market reacts negatively to

inventory write-offs. Lopez (2002) determines that restructurings are multi-dimensional efforts

that may require disaggregation into components for a complete understanding of their effect on

the market. John et al. (1992) examined firms’ responses to losses, and they determine that firms

were able to increase their focus and become more efficient after restructuring efforts. Smart and

Waldfogel (1994) utilize a “surprise” variable to determine what would have happened at the

restructuring firm in the absence of the restructuring. This essay further examines the value

relevance of restructuring charges to determine the reasons why the extant research contains

studies with contradictory findings that the market reacts positively to restructuring charges and

that the market reacts negatively to restructuring charges.

1 Elayan et al. (1998) document that the market reaction to layoff announcements depends on many other factors including the size of the layoff, the industry of the firm, the information set available to shareholders, and the financial performance of the firm before the announcement. Worrell et al. (1991) determine that announcements of large or permanent layoffs result in stronger negative responses than other announcements.

83

Further, this essay examines the value relevance of restructuring charges for firms with varying levels of financial health. Khurana and Lippincott (2000) separately examine restructuring charges for firms posting losses. For the loss firms, the restructuring charge is positive and highly significant. The level of earnings is not significant for loss firms, and the change in earnings is only slightly significant. The results for loss firms suggest that current losses are viewed as being temporary and not value-relevant, but restructuring activities are seen as having a permanent and positive effect on future performance. The authors further separate the restructuring firms into groups based on the primary purpose for the charge. The three main categories of reasons for taking restructuring charges are restructuring with the primary purpose of exiting a line of business, restructuring where the primary purpose is to eliminate personnel, and restructuring where the primary purpose cannot be discerned. The authors find that both of the first types of restructurings are positively associated with returns.2

Based partially on Khurana and Lippincott’s (2000) findings for profit and loss firms, this

essay suggests that the value relevance of restructuring charge information for distressed firms

restructuring to avoid bankruptcy is greater than the value relevance of restructuring charges for

healthy firms restructuring to improve their efficiency. Because distressed firms have several

different possible outcomes from restructuring, these groups of firms with different results during

the three years after restructuring are examined separately. Distressed firms that avoid

bankruptcy in the three years following a restructuring charge are likely to have had more well-

developed, organized plans for their restructuring activities than firms that file for bankruptcy

after restructuring. Therefore, this essay predicts that the value relevance of restructuring charge

information for distressed firms that are able to avoid filing for bankruptcy for at least three years

2 In January 1995, the EITF released EITF 94-3, which requires firms to record the costs of restructuring during the period in which management commits to the plan and to disclose many details about the restructuring plan. Also, costs classified as restructuring charges must provide no future benefit to the firm over and above the restructuring execution. EITF 94-3 was later nullified by SFAS No. 146.

84

after restructuring is greater than the value relevance of restructuring charge information for distressed firms that ultimately file for bankruptcy in the three years after restructuring.

4.3 Hypotheses Development

The hypotheses examine the effect of restructuring charges on price. H2a and H2b examine the impact of restructuring charges on price in order to address the contradictory findings related to the value relevance of restructuring charges documented by prior studies including Kross et al.

(2001), Bens (2002), and Lopez et al. (2002). H2c examines the value relevance of restructuring charges for several different groups of restructuring firms to determine whether the charge has a different effect on price depending on a corporation’s reason for restructuring. Based on prior studies, including the Khurana and Lippincott (2000) paper that separately examines value relevance for firms in different financial health categories, the predictions described in H2c are developed.

This prior literature leads to the three hypotheses stated as follows:

H2a: Restructuring charges are value relevant and the stock market reacts positively to

the magnitude of restructuring charges.

H2b: The value relevance of restructuring charge information is smaller for non-

distressed firms, greater for distressed firms that file for bankruptcy within three

years of restructuring, and greatest for financially distressed firms that avoid

filing for bankruptcy during the three years following the restructuring.

H2c: Restructuring costs (financial distress) have positive (has negative) impact on

price and returns.

85

4.4 Methodology

4.4.1 Methodology

Prior to testing the hypotheses, it is necessary to create a sample of firms that have restructured. The full sample for this essay contains data from 1992 to 2004 for firms undertaking operational restructuring efforts during the period from 1993 through 2003 that have data availability for the required variables. Equations 2.1 through 2.3 are used to determine whether or not each firm was in financial distress. In order to determine a “distress” value for each firm in the sample, this study uses Altman’s (1968) original Z-score model and Begley et al.’s (1996) updated version of the Altman (1968) model.

Although the Altman (1968) and Begley (1996) models were originally intended as bankruptcy prediction models, Grice and Dugan (2001) indicate that bankruptcy prediction models like Altman’s (1968) are actually more useful for identifying firms that are financially distressed, as opposed to identifying the more limited bankruptcy condition. Because these models have been proven successful, this essay uses the linear Z-score equations and substitutes the numbers for each variable for the firms in the sample. These models are used to determine a distress value for each firm, and then a cutoff point can be used to classify firms as either distressed or healthy. Altman’s (1968) Z-score model is as follows:

Z = 0.012 X1 + 0.014 X2 + 0.033 X3 + 0.006 X4 + 0.999 X5, (2.1) where

Z is used to determine whether each company is in financial distress3,

4 X1 is working capital divided by total assets * 100 ,

3 All X-values are included in the calculation of Z for each firm or firm-event observation, even when the values are negative. 4 Working capital divided by total assets is a measure of the net liquid assets of a firm relative to the overall capitalization; firms with losses are likely to also have shrinking current assets compared to total assets.

86

5 X2 is retained earnings divided by total assets * 100 ,

6 X3 is earnings before interest and taxes divided by total assets * 100 ,

7 X4 is the market value of equity divided by the book value of debt * 100 , and

8 X5 is sales divided by total assets .

Begley et al. (1996) re-estimate Altman’s (1968) model using data from the eighties, and their

updated model is as follows:

Z = 0.104 X1 + 1.010 X2 + 0.106 X3 + 0.003 X4 + 0.169 X5, (2.2)

using the same variables and variable definitions as Altman’s (1968) model.

Altman (1968) finds that for his sample firms, firms with Z-scores greater than 2.99 were

mostly not in financial distress and many of the firms with Z-scores less than 1.81 went bankrupt.

Altman (1968) further finds that using a Z-score of 2.675 as a cutoff minimizes the number of

firms that are misclassified by the model. Therefore, this study uses 2.675 as the cutoff point for

the Altman (1968) model results. Begley et al. (1996) find that the most appropriate cutoff point

for their model is 0.545. Firms with Z-scores less than 0.545 are classified as financially

distressed and are assigned a value of 1, and firms with Z-scores greater than 0.545 are classified

as being non-distressed and are assigned a value of 0.

Altman (1968) finds that working capital divided by total assets is the most valuable measure of the liquidity. 5 Retained earnings divided by total assets is included because it implicitly considers the age of a firm, and financial distress is much more common in the early years of a firm’s life. 6 Earnings before interest and taxes, divided by total assets is a measure of the true productivity of a firm’s assets, ignoring tax and leverage factors. Because a firm’s existence is based on the earning power of the firm’s assets, this ratio is especially important. 7 The market value of equity divided by book value of debt variable shows how much the firm’s assets can decline in value before the firm becomes insolvent. Including the market value of equity divided by the book value of debt adds a market value dimension not considered before Altman (1968), and this variable is determined to be a better predictor of bankruptcy than net worth/total debt. 8 Sales divided by total assets is a measure of firm size. Although Altman (1968) finds that sales divided by total assets is the least significant variable on its own, it is important to include this variable because of its unique relationship to the other variables to be included in the model.

87

After each firm is classified as either being financially distressed or non-distressed using the Altman (1968) model and the Begley et al. (1996) model, Ohlson’s (1980) logistic regression model is used to confirm the accuracy of the classification procedure. Also, the results of the

Ohlson logistic regression model provide a probability value between 0 and 1 for each firm that indicates the likelihood of a firm being in financial distress. These probability values are included as additional predictor variables in the price and return models discussed in Essay 2. Ohlson’s

(1980) model is as follows:

DISTRESSi,t = a0 + a1 SIZE i,t + a2 TLTA i,t + a3 WCTA i,t + a4 CLCA i,t + (2.3)

a5 NITA i,t + a6 FUTL + a7 INTWOi,t + a8 OENEGi,t + a9 CHIN + ui,t, where

DISTRESS equals 1 if a firm is determined to be in financial distress, 0 otherwise;

SIZE is the log of total assets;

TLTA is total liabilities divided by total assets;9

WCTA is working capital divided by total assets;10

CLCA equals current liabilities divided by current assets;

OENEG equals 1 if owners’ equity is negative, 0 otherwise;11

NITA is net income divided by total assets;12

FUTL is cash flows from operations divided by total liabilities;

INTWO equals 1 if net income was negative over the last two years, 0 otherwise;

13 CHIN = (NIt – NIt-1) / ( | NIt | + | NIt-1 | ).

9 The variable TLTA is included as a measure of firm leverage. 10 The variables WCTA and CLCA are included as measures of current liquidity. 11 The variable OENEG is used as a discontinuity correction for TLTA. 12 NITA and FUTL are included as measures of firm performance. 13 The CHIN variable is a measure of the change in net income and is included as suggested by McKibben (1972).

88

Ohlson’s (1980) model includes nine explanatory variables, and all of them are included in this essay even though Ohlson finds that only six of them are significant. The log of total assets, total liabilities divided by total assets, net income divided by total assets, cash flows from operations divided by total liabilities, FUTL, and CHIN are all significant predictors of bankruptcy in Ohlson’s (1980) model.

The hypotheses in this essay are tested by examining the value-relevance of restructuring charge information using two methods: associating stock returns with contemporaneous financial data and associating prices with financial data (Aboody and Lev 1998). Prior research (Ballester et al. 1999) has primarily found that prices tend to rise in response to restructuring, and therefore

RSTRi,t and RSTRi,t / Pi,t-1 in equations 2.4a and 2.5a, respectively, are expected to be positive and

significant. The following regression models are used:

Pi,t = ∑ a0 Ct + a1 BVPSi,t + a2 EPSi,t + a3 RSTRi,t + ui,t (2.4a)

Ri,t = ∑ b0 Ct + b1 EPSi,t / Pi,t-1 + b2 ΔEPS / Pi,t-1 + b3 RSTRi,t / Pi,t-1 + ui,t, (2.5a) where

Pi,t is firm i’s monthly stock price three months after fiscal year-end;

Ct equals 1 if the observation is from year t, where t represents a year between 1993 and

2003, 0 otherwise;

BVPS is firm i’s book value per share for year t;

EPS is firm i’s earnings per share for year t;

RSTR is the dollar amount of restructuring charges scaled by the number of outstanding

shares;

Ri,t is the firm’s annual stock return, which is calculated using the following, a simple

return model (Strong 1992), and Rt = (Pt – Pt-1) / Pt-1, where Rt is current year

89

returns, Pt is the price three months after fiscal year-end, and Pt-1 is the price from

the prior year;

ΔEPS is the change in earnings per share;

Pi,t-1 is the price at the beginning of period.

In addition, this essay uses the following variations of equations 2.4a and 2.5a14:

Pi,t = ∑ a0 Ct + a1 BVPSi,t + a2 EPSi,t + a3 RSTRi,t + a4 OHLSON_PROBi,t + ui,t (2.4b)

Pi,t = ∑ a0 Ct + a1 BVPSi,t + a2 EPSi,t + a3 RSTRi,t + a4 DISTRESSi,t + ui,t (2.4c)

Ri,t = ∑ b0 Ct + b1 EPSi,t / Pi,t-1 + b2 ΔEPS / Pi,t-1 + b3 RSTRi,t / Pi,t-1 (2.5b)

+ b4 OHLSON_PROBi,t-1 + ui,t

Ri,t = ∑ b0 Ct + b1 EPSi,t / Pi,t-1 + b2 ΔEPS / Pi,t-1 + b3 RSTRi,t / Pi,t-1

+ b4 DISTRESSi,t + ui,t, (2.5c) where

OHLSON_PROB is the probability of financial distress obtained from running Ohlson’s

logistic regression model shown in equation 2.3;

DISTRESS equals 1 for firms classified by the Altman (1968) and Begley (1996) model

to be distressed, 0 otherwise.

The three versions of equations 2.4 and 2.5 are estimated during the restructuring event year for all sample firms. Also, equations 2.4a and 2.5a are estimated separately for the financially healthy restructuring firms, for the firms in financial distress that did not subsequently file for bankruptcy, and for the firms in financial distress that did subsequently file. The results of the Ohlson logistic regression model in equation 2.3 provide a probability value between 0 and

1 for each firm that indicates the likelihood of a firm being in financial distress. These probability values are included in equations 2.4b and 2.5b above. The distress values assigned to

14 See Aharony and Barniv (2004) for using a probability value measure to assess value relevance of accounting information in valuation models for .

90

each firm by using both the Altman (1968) and Begley (1996) models in equations 2.1 and 2.2 are included as additional explanatory variables in equations 2.4c and 2.5c above. Only firms classified as distressed or non-distressed by both models are included in the final sample.

4.4.2 Data

The full sample for this essay includes data for the period from 1992 to 2004 and firms announcing operational restructurings during the period from 1993 to 2003.15 Financial and market data were obtained for this essay from the COMPUSTAT database. Table 1 provides the steps used to arrive at the final sample of 1,207 firms and 1,562 firm-event observations.

The preliminary sample for this essay contains 1,656 firms and 2,202 firm-event observations. As shown in Table 2, the sample is comprised primarily of manufacturing firms, with 55.8 (58.6) percent of the sample firms (firm-event observations) coming from this category.

The next largest sample group is service firms, which comprise 23.7 (21.7) percent of the sample firms (firm-event observations). The sample also contains smaller percentages of firms from the transportation, communication, gas and electric category (6.2%), the wholesale and retail trade categories (4.6 and 5.6%), and the financial, insurance, and real estate category (1.8%).

4.5 Empirical Results

4.5.1 Univariate Analysis

Table 3 provides the descriptive statistics obtained after the outliers are deleted for each variable. Panel A of Table 3 provides the descriptive statistics for the full sample of firms. The mean (median) restructuring charge amount is $53.31 million ($8.5 million), which indicates that most of the companies in the sample are undergoing fairly significant restructuring efforts. The

15 Because one variable requires data from period t-1, some 1992 data are also used.

91

median values for EPS and BVPS are -$0.31 and $4.16 per share, respectively. Panel B of Table

3 provides the descriptive statistics for the full sample of firm-event observations. The mean

(median) restructuring charge amount is $82.9 million ($10 million). The median values for EPS and BVPS are -$0.21 and $4.71 per share, respectively.

4.5.2 Value Relevance across Three Types of Firms

To test hypothesis H2a, the models shown in equations 2.4a and 2.5a are each estimated three times, once for the companies not in financial distress, once for the companies in financial distress that avoid bankruptcy for at least three years subsequent to restructuring, and once for the companies in financial distress that eventually file for bankruptcy. The results of regressions using firm observations are presented in Table 4.

Panel A of Table 4 includes the results for the estimations of equation 2.4a, the price model, for firms classified into the same health category by both the Begley and Altman models.16 The three columns in each panel provide the results for the non-distressed firms, the

distressed and non-bankrupt firms, and the distressed and bankrupt firms. The price models for

the non-distressed group, the distressed and non-bankrupt group and the distressed and bankrupt

group have R-square values of 0.65, 0.93 and 0.44, respectively. The much higher R-square for

the distressed and non-bankrupt firms indicate that book value per share, earnings per share, and

charge per share are more value relevant to price for distressed, non-bankrupt firms. Overall, the

results of the price models support H2b.

16 The price and return models in the remainder of the essay were all tested for various specification problems including multicollinearity, heteroscedasticity, and autocorrelation. To test for multicollinearity, I examined variance inflation factors (VIFs) and found that they were less than five in all cases. White’s Chi-Square test was used to test for problems with heteroscedasticity. Durbin-Watson statistics were examined, and no signs of autocorrelation were found.

92

For the price models in Panel A of Table 4, the earnings per share coefficient is negative

and significant for both groups of distressed firms. The coefficient on book value per share is

positive and significant for the non-distressed firms. The coefficient for restructuring charge per

share is positive and significant for the distressed firms that filed for bankruptcy within three

years of restructuring. This suggests that restructuring charge dollar amounts provide value-

relevant information to investors for distressed firms that file for bankruptcy within three years of

restructuring when price models are used. I ran the Chow Test to determine whether the

coefficients for the three regression models examined are significantly different from each other,

and the significant Chow F-values in Panel A of Table 4 show that the impact of the independent

variables on price and returns is significant and different across the three groups.

Panel B of Table 4 provides the results of the estimation of equation 2.5a, the returns

regression model. The returns model for the non-distressed firms, the distressed and non-

bankrupt firms, and the distressed and bankrupt firms have adjusted R-square values of 0.11, 0.85

and 0.25, respectively.17 The much higher R-square value for the distressed and non-bankrupt

group of firms suggests that book value per share, earnings per share, and restructuring charge per

share provide more value relevant information for distressed firms that do not file for bankruptcy

within three years than for the distressed firms that ultimately file for bankruptcy.18

The results in Panel B of Table 4 demonstrate that the variable of interest, charge per share, is not significant for any of the returns models, which indicates that restructuring charge

17 The adjusted R-squares are very high relative to the prior studies testing annual return models. 18 The earnings per share variable is negative and significant for the non-distressed firms and the distressed non-bankrupt firms, which indicates that as the level of earnings per share decreases, the level of returns increases for healthy firms and distressed non-bankrupt firms. The negative significant coefficient on the EPS variable for the non-distressed firms and the distressed, non-bankrupt firms indicates that returns increase as earnings per share decreases. The increase in returns that occurs when EPS decreases is an unexpected finding. The change in earnings per share variable is positive and highly significant for the non-distressed firms and negative and significant for the distressed, non-bankrupt firms. This indicates that as the change in earnings per share increases, the level of returns increases for healthy firms and decreases for distressed non-bankrupt firms.

93

information is not value relevant when returns models are used. Again as an additional analysis, I ran the Chow Test to determine whether the coefficients for the three regression models examined are significantly different from each other, and the significant Chow F-values in Panel B of Table

4 show that the distressed and non-bankrupt firms do not produce significantly different results from the non-distressed firms. However, there are significant differences between the non- distressed firms and the distressed, bankrupt firms and between the distressed, non-bankrupt firms and the distressed, bankrupt firms.

The results of the regressions of equations 2.4a and 2.5a for the three health conditions using firm-event observations classified the same by the Altman and Begley models are presented in Table 5. Panel A of Table 5 includes the results for the estimations of equation 2.4a, the price model.19 The three columns in each panel provide the results for the non-distressed firms, the

distressed and non-bankrupt firms, and the distressed and bankrupt firms. The price models for

the non-distressed group, the distressed and non-bankrupt group, and the distressed and bankrupt

group have R-square values of 0.44, 0.10 and 0.30, respectively. The higher R-square for the

non-distressed firms indicate that book value per share, earnings per share, and charge per share

are more value relevant to price for non-distressed firms. Overall, the results of the price models

in Panel A tend to support H2b.

For the price models in Panel A of Table 5, the earnings per share coefficient is positive

and significant for the non-distressed group of firms and negative and significant for the

distressed, non-bankrupt firms. The coefficient on book value per share is positive and

19 The price and return models in the remainder of the essay were all tested for various specification problems including multicollinearity, heteroscedasticity, and autocorrelation, and the problems were corrected whenever they were found. For example, the test for multicollinearity examined variance inflation factors to ensure that they were less than 5 and by confirming that there are not high pair-wise simple correlations among regressors. Although there were a few relatively high correlations among the regressors for the price and return models used, the VIF factors are in all cases below 5. The test for heteroscedasticity and autocorrelation used White’s chi-square test and Durbin-Watson statistics, respectively.

94

significant for the non-distressed firms and the distressed and non-bankrupt firms. The coefficient for restructuring charge per share is positive and significant for the two groups of distressed firms. This suggests that when price models are used, restructuring charge dollar amounts provide positive value-relevant information to investors for distressed firms when firm- event observations are examined. The results also support H2b and Khurana and Lippincott’s

(2000) findings. I also ran the Chow Test to demonstrate that the variables in the models are significantly different across the three groups, and the Chow F-values shown in Panel A of Table

5 are significant.

Panel B of Table 5 provides the results of the estimation of equation 2.5a, the returns regression model, when firm-event observations are used. The return models for the non- distressed group, the distressed and non-bankrupt group, and the distressed and bankrupt group have R-square values of 0.14, 0.31, and 0.30, respectively. The higher R-square for the distressed firms indicate that book value per share, earnings per share, and charge per share are more value relevant for distressed firms when the results of returns models are examined.20 The results in

Panel B of Table 5 demonstrate that the coefficient on the variable of interest, charge per share, is

negative and significant for all three groups of firms for the returns model. This indicates that

restructuring charge information is value relevant for returns models and that returns decrease as

the dollar amount of the restructuring charge per share increases. The significance of the

restructuring charge variable provides support for hypothesis H2a and H2b. The negative and

significant coefficient on the restructuring charge variable using the returns model confirms

Khurana and Lippincott’s (2000) expectations for the healthy firms. However, the negative and

20 The earnings per share variable is negative and significant for the non-distressed firms, which indicates that as the level of earnings per share decreases, the level of returns increases for healthy firms. The change in earnings per share variable is positive and significant for all three groups of firms, which indicates that as the change in earnings per share increases, the level of returns also increases.

95

significant coefficient for the restructuring charge variable for the two groups of distressed firms is contrary to expectations and to the findings of Khurana and Lippincott (2000). The significant

Chow F-values in Panel B of Table 5 show that there are significant differences across the hypotheses between the impact of the independent variables on price and returns. Overall, the results in Tables 4 and 5 provide some support for H2a and H2b.

4.5.3 The Effect of Restructuring Charges and the Likelihood of Financial Distress

The next tables are used to test H2c, focusing on the joint impact of the magnitude of

restructuring charges and the likelihood of financial distress on price and returns. Table 6

provides the price model results when several alternative specifications for equation 2.4 are run.

Panel A provides the results of the price model for firm observations, and Panel B provides the

results for the price model for firm-event observations. The results in the first column of Panel A

of Table 6 show that the R-squared value for the price model is 0.505. The coefficient on

earnings per share is negative and significant in each model, which suggests that price declines as

earnings per share increases, an unexpected result. The coefficient on the restructuring variable is

also positive and significant for all the models in Table 6, which suggests that restructuring

charges are value relevant when price models are examined and that price increases in response to

the magnitude of the restructuring charge.

In the second column of Table 6, the Ohlson probability value (OHLSON_PROB)

obtained from equation 2.3 is included as an independent variable in the price model. As

expected, the coefficient of the OHLSON_PROB variable is negative and highly significant. The

significant, negative coefficient on the OHLSON_PROB variable indicates that, as expected, firm

value declines as the probability of financial distress increases. Also, the strong significance and

the negative sign of the OHLSON_PROB variable for the price model indicate that firm value is

96

inversely related to the likelihood of distress. It is also important to note that the adjusted R- square value improves slightly to 0.526 when the OHLSON_PROB variable is included in the model, indicating an improvement to the basic price model. In the third column of Table 6, the dummy variable indicating the health classification from the Altman and Begley models is included in equation 2.4. The R-square value for the third column is the same as the value when the OHLSON_PROB variable is included. The coefficient on the DISTRESS dummy variable is negative and significant, which suggests that price decreases as the likelihood of financial distress increases.

The results for the firm-event observations in Panel B of Table 6 provide similar results to those reported in Panel A, except that the three price models in Panel B all have lower R- square values than the corresponding models in Panel A. Overall, the results in Table 6 demonstrate that the dollar amount of restructuring charges is value relevant when price models are examined using firm observations.

Table 7 provides the return model results for several alternative specifications for equation 2.5. In Panel A, the R-square values are higher than the corresponding R-square value from the price model. The lack of a significant coefficient on the charge per share variable for all versions of equation 2.5 examined suggests that restructuring charges are not value relevant when a returns model is examined using firm observations. The earnings per share variable also lacks significance for all versions of equation 2.5.

In the second column of Table 7, the Ohlson distress probability value

(OHLSON_PROB) obtained from equation 2.3 is included as an additional independent variable.

The coefficient on the OHLSON_PROB variable is negative and significant. The strong significance and the negative sign of the OHLSON_PROB variable for the returns model indicate that firm value declines as the likelihood of distress increases. In the third column of Table 7,

97

instead of including the OHLSON_PROB variable, the dummy variable indicating the health classification from the Altman and Begley models is included in equation 2.4. The coefficient on this DISTRESS dummy variable is negative and significant, which also indicates that returns decrease as financial health declines. Comparing the adjusted R-square values indicates that the best specified model is the one that includes the DISTRESS dummy variable as an additional explanatory variable in the return model.

The firm-event observation results in Panel B of Table 7 are consistent with many of the firm results found in Panel A, but several differences from Panel A should be noted. First, the return models in Panel B for the firm-event observations all have lower R-square values than the

corresponding models in Panel A. Second, the coefficient on the charge per share variable is

negative and significant for all the versions of the returns model when firm-event observations are

examined. The results in Panel B of Table 7 demonstrate that the dollar amount of restructuring

charges is value relevant when returns models are examined for all sample firm-event

observations and that firm value decreases as restructuring charge per share increases. Overall,

the results reported in Tables 6 and 7 tend to support H2c. In particular, the findings provide

strong support for the negative impact of the likelihood of financial distress on price and returns.

Robustness Tests

To determine potential problems because of model specification and OLS assumptions, this essay also examines potential problems with autocorrelation, multicollinearity, and heteroscedasticity. An examination of the Durbin-Watson statistics indicates that there is no autocorrelation of the residuals. It seems that there is no major problem with multicollinearity because there are no high correlations between the independent variables. Although a few pairwise correlations between the coefficients for the price and returns models are fairly high, the

98

variance inflation factors (VIFs) are below two for most variables and are below five for all of the variables, which roughly indicates that there is no material multicollinearity problem. An examination of the White’s Chi-squared statistic for the models indicated that there were initially some heteroscedasticity problems, so White-adjusted t-statistics are calculated and reported for all regressions.

4.6 Conclusions

The results of this essay demonstrate that the dollar amount of corporate restructuring charges tend to provide value relevant information to investors when both price and return models are used, which supports H2a. As expected, earnings per share, book value per share, and the change in earnings per share tend to provide value relevant information to the market for most of the samples. Also, the value relevance of restructuring charge information is determined to be smaller for firms that are not financially distressed, greater for financially distressed firms that file for bankruptcy during the three years following the restructuring, and greatest for firms in distress that do not file for bankruptcy in the three years subsequent to the restructuring. Thus, the results also provide partial support for H2b.

This information may have implications for investors and analysts as they determine how they should react to restructuring charge information. The results further indicate that firm value declines as the probability of financial distress increases. Statistical tests also demonstrate that the dollar amount of restructuring charges is value relevant to investors for all firms based on the results of the price model. Finally, the results reported in Tables 6 and 7 support H2c.

Several limitations of this essay should be noted. First, future research may be able to improve the process for obtaining a better list of restructuring firms. Because the initial list of restructuring firms used in this essay was obtained through a key word search of the newswires

99

available on the Lexis-Nexis Database, it is possible that some restructuring firms were left out of the sample. If an announcement of the restructuring did not take place on a newswire, then it may have been inadvertently left out the sample. Perhaps by examining all the 10-Ks and 8-Ks of firms with special item dollar amounts available on COMPUSTAT, fewer firms that potentially have taken restructuring charges would have been eliminated from the sample. Second, the results may be adversely affected by including some companies that restructured during multiple years as firm-event observations, which is why the results are also reported using firms that have restructured only once during the research period. Third, another limitation of this study is the

“ad-hoc” statistical identification of healthy versus financially distressed firms. Finally, this essay is subject to potential criticism on price and return models. This includes the efficient market explanation that the restructuring is already impounded in prices. The empirical results of the price models tend to suggest the potential existence of this limitation.

Some of the limitations of this essay can be addressed in future research. For example,

the sampling period can be extended to include firms restructuring between 2004 and 2006.

Increasing the sample period would increase the overall sample size and the number of firms in

financial distress that file for bankruptcy within three years of restructuring. Further statistical

comparisons can be made between the value relevance of restructuring charge information for the

sub-samples examined. Future studies may be able to improve upon the method used to

determine which firms are healthy and which firms are distressed. Also, future research can

extend the results of this study by determining whether the results of this study are generalizable

to other time periods or to specific industries or analyzing repeated restructurings by the same

firms (for example, see Lin and Yang 2006).

BIBLIOGRAPHY

Aboody, D., and B. Lev. 1998. The value relevance of intangibles: The case of software capitalization. Journal of Accounting Research 36 (Supplement): 161-191.

Aharony, J., and R. Barniv. 2004. Using financial accounting information in the governance of takeovers: An analysis by type of acquirer. Journal of Accounting and Public Policy 23 (5): 321-349.

Altman, E. I. 1968. Financial ratios, discriminant analysis, and the prediction of corporate bankruptcy. Journal of Finance 23 (4): 589-609.

Ballester, M., J. Livnat, and N. Sinha. 1999. Corporate reorganizations: Changes in the intensity of labor and capital expenditures. Journal of Business Finance and Accounting 26 (9/10): 1205-1238.

Bartov, E., F. W. Lindahl, and W. E. Ricks. 1998. Stock price behavior around announcements of write-offs. Review of Accounting Studies 3: 327-346.

Begley, J., J. Ming, and S. Watts. 1996. Bankruptcy classification errors in the 1980s: An empirical analysis of Altman’s and Ohlson’s models. Review of Accounting Studies 1: 267-284.

Bens, D. A. 2002. The determinants of the amount of information disclosed about corporate restructurings. Journal of Accounting Research 40 (1): 1-20.

Blackwell, D. W., M. W. Marr, and M. F. Spivey. 1990. Plant-closing decisions and the market value of the firm. Journal of Financial Economics 26 (2): 277-288.

Brickley, J. A., and L. D. Van Drunen. 1990. Internal corporate restructuring: An empirical analysis. Journal of Accounting and Economics 12 (1-3): 251-280.

Bunsis, H. 1997. A description and market analysis of write-off announcements. Journal of Business Finance and Accounting 24 (9 & 10): 1385-1400.

Carter, M. E. 1998. Does operating performance improve after corporate restructuring? Working paper. MIT.

______. 2000. Operating performance following corporate restructurings. Working paper. Columbia University.

Elayan, F. A., G. S. Swales, B. A. Maris, and J. R. Scott. 1998. Market reactions, characteristics, and the effectiveness of corporate layoffs. Journal of Business Finance & Accounting 25 (3/4): 329-351.

100 101

Elliott, J. A., and J. D. Hanna. 1996. Repeated accounting write-offs and the information content of earnings. Journal of Accounting Research (Supplement): 117-134.

Francis, J., J. D. Hanna, and L. Vincent. 1996. Causes and effects of discretionary write-offs. Journal of Accounting Research 34 (Supplement): 117-134.

Grice, J. S., and M. T. Dugan. 2001. The limitations of bankruptcy prediction models: Some cautions for the researcher. Review of Quantitative Finance and Accounting 17: 151- 166.

John, K., L. H. Lang, and J. Netter. 1992. The voluntary restructuring of large firms in response to performance decline. Journal of Finance 47 (3): 891-917.

Khurana, I. K., and B. Lippincott. 2000. Restructuring and firm value: The effects of profitability and restructuring purpose. Journal of Business Finance and Accounting 27 (9/10): 1106-1129.

Kross, W., T. Park, and B. Ro. 2001. Is ‘operational restructuring’ a magic phrase? An examination of market responses to restructuring announcements. Working paper, Purdue University, Purdue, IN.

Lin, B., and R. Yang. 2006. The effect of repeat restructuring charges on analysts’ forecast revisions and accuracy. Review of Quantitative Finance and Accounting 27 (3): 267- 283.

Lin, J. C., and M. S. Rozeff. 1993. Capital market behavior and operational announcements of layoffs, operation closings, and pay cuts. Review of Quantitative Finance and Accounting 3: 29-45.

Lopez, T. J. 2002. Evidence on the incremental information contained in the components of restructuring charges. Journal of Business Finance and Accounting 29 (5/6): 613-659.

Lopez, T. J., P. R. Regier, and L. Holder-Webb. 2002. Do restructurings improve operating performance? Working paper. University of South Carolina.

Martin, J. D., and J. W. Kensinger. 1990. Exploring the controversy over corporate restructuring. Financial Executives Research Foundation: Morristown, New Jersey.

McKibben, W. 1972. Econometric forecasting of common stock investment returns: A new methodology using fundamental operating data. Journal of Finance 27 (May): 371-380.

Ohlson, J. A. 1980. Financial ratios and the probabilistic prediction of bankruptcy. Journal of Accounting Research 18 (1): 109-131.

Poon, P. S., G. D. Newbould, and C. Durtschi. 2001. Market reactions to corporate restructurings. Review of Qualitative Finance and Accounting 16: 269-290.

102

Smart, S. B., and J. Waldfogel. 1994. Measuring the effect of restructuring on corporate performance: The case of management . Review of Economics and Statistics 76: 503-511.

Strong, N. 1992. Modeling abnormal returns. Journal of Business Finance and Accounting 19 (4): 533-553.

103

TABLE 1 Final Sample Determinationa Panel A: Firms Number of Firms

Sample of firms announcing restructuring from 1993 through 2003 3,155

Observations with missing data (IBES and Compustat) (1,471)

Less outliers deleted (28)

Less firm-event observations not classified into the same health category by Altman (1968) and Begley (1996) (449)

Final number of firms included in this essay 1,207

Panel B: Firm-Event Observations Number of Firm-Event Observations

Sample firm-event observations announcing restructuring from 1993 through 2003 3,867

Observations with missing data (IBES and Compustat) (1,627)

Less outliers deleted (38)

Less firm-event observations not classified into the same health category by Altman (1968) and Begley (1996) (522)

Final number of firm-event observations included in this essay 1,562

a Several firms completed more than one operational restructuring over the period from 1993 through 2003; the firms are included more than once in the sample as different firm-event observations.

104

TABLE 2 Firm-Event # of Firm-Event Firms as % Observations SIC Division # of Firms Observations of Total as % of Total Agriculture, Forestry, and Fishing 1 1 0.1 0.1 Mining 17 22 1.4 1.4 Construction 233 350 19.3 22.4 Manufacturing 437 585 36.2 37.5 Transportation, Communications, Electric, & Gas 55 72 4.6 4.6 Wholesale Trade 124 163 10.3 10.4 Retail Trade 33 21 2.7 1.3 Finance, Insurance, and Real Estate 239 276 19.8 17.7 Services 62 69 5.1 4.4 Public Administration 6 3 0.5 0.2 Total 1,207 1,562

105

TABLE 3: Descriptive Statisticsa b Panel A: Firm Observations Variable Name N Mean Median Std. Dev. Minimum Maximum Price Model Variables: BVPSi,t ($) 1,207 5.23 4.16 22.00 -618.6 217.45

EPSi,t ($) 1,207 -3.29 -0.31 49.71 -1,677.0 52.26 Charge Amount ($Millions) 1,207 53.31 8.50 155.50 0.03 2,100.0

Pi,t ($) 1,207 14.46 8.42 35.96 0.002 884.4

Return Model Variables: Ri,t (%) 1,207 0.10 -0.09 2.25 -1.00 70.19

Pi,t-1 ($) 1,207 19.15 10.00 61.35 0.01 1,625.0 RSTRi,t ($) 1,207 1.25 0.33 7.87 0.002 199.3

EPSi,t / Pi,t-1 1,194 -0.34 1.63 1.63 -30.28 5.25 ∆EPS / Pi,t-1 1,194 -0.36 20.66 20.66 -702.1 89.16

RSTRi,t / Pi,t-1 1,194 0.09 0.23 0.23 <0.001 3.27

Financial Health Variables: OHLSON_PROB 1,207 0.50 0.38 0.39 0 1 DISTRESS 1,207 0.51 1.00 0.50 0 1

a Pi,t is the share price three months after fiscal year end. BVPSi,t is book value per share. EPSi,t is earnings per share. RSTRi,t is the dollar amount of restructuring charges per common share outstanding. Ri,t is simple returns calculated using the equation Rt = (Pt – Pt-1) / Pt-1. Pi,t-1 is the price three months after fiscal year end from the previous year. ΔEPS is the change in earnings per share between year t-1 and year t. OHLSON_PROB is the probability of financial distress obtained from running Ohlson’s regression model (equation 2.3). DISTRESS equals 1 for firms classified by both Altman and Begley as distressed, and 0 otherwise. The sample in this table includes only observations classified into the same distress classification (0 or 1) by the Altman (1968) and Begley (1996) models. b The data has been trimmed to delete the outliers.

106

TABLE 3: Descriptive Statistics a Panel B: Firm-Event Observations Variable N Mean Median Std. Dev. Minimum Maximum Price Model Variables:

BVPSi,t ($) 1,562 6.09 4.71 7.53 -37.18 55.63

EPSi,t ($) 1,562 -1.05 -0.21 4.55 -58.62 16.10 Charge Amount ($Millions) 1,562 82.90 10.00 419.1 0.07 8,000.0

Pi,t ($) 1,562 16.54 10.13 34.06 0.01 884.4

Return Model Variables:

Ri,t (%) 1,562 0.06 -0.05 0.83 -0.97 8.23

RSTRi,t ($) 1,562 1.17 0.34 6.93 0.002 199.3

Pi,t-1 ($) 1,562 19.82 11.38 54.85 0.01 1,625.0

EPSi,t / Pi,t-1 1,528 -0.24 -0.02 0.76 -10.24 1.62

∆EPS / Pi,t-1 1,528 0.17 -0.01 1.77 -3.90 32.23

RSTRi,t / Pi,t-1 1,528 0.24 0.04 1.80 <0.001 56.06

Financial Health Variables: OHLSON_PROB 1,540 0.45 0.27 0.39 0 1 DISTRESS 1,562 0.46 0 0.50 0 1

a The data have been trimmed to delete the outliers. The sample in this table includes only observations classified into the same distress classification (0 or 1) by both the Altman (1968) and Begley (1996) models.

107

TABLE 4: Price and Return Model Regressions:a b Firm Observations Panel A: OLS Regressions of Equation 2.4a: Price Modelc

(2.4a) Pi,t = ∑ a0 Ct + a1 BVPSi,t + a2 EPSi,t + a3 RSTRi,t + ui,t Distressed and Distressed and Non-Distressed Non-Bankrupt Bankrupt Variable Name (Expected Sign)d Coefficients Coefficients Coefficients BVPS (+/-) 2.500 -0.083 0.013 (2.71***) (-0.66) (0.40) EPS (+/-) 1.489 -0.360 -0.191 (1.29) (-7.40***) (-6.30***) Charge Per Share (+/-) -1.094 0.354 2.015 (-0.69) (1.47) (1.97*)

Adjusted R2 0.648 0.931 0.443 N 597 firms 542 firms 68 firms F-value 85.58*** 558.8*** 5.44***

Chow Test Results for Price Model for Firm Observations Distressed, Non- Non-Distressed Firms vs. Non-Distressed Firms Bankrupt Firms vs. Distressed, Non- vs. Distressed, Bankrupt Distressed, Bankrupt Bankrupt Firms Firms Firms F-value 252.439*** 56.750*** 5.375***

a The coefficients on the yearly intercepts are not reported. b The data has been trimmed to delete the outliers. The sample in this table includes only observations classified into the same distress classification (0 or 1) by the Altman (1968) and Begley (1996) models. c Pi,t is the share price three months after fiscal year end, BVPSi,t is book value per share, EPSi,t is earnings per share, and RSTRi,t is the dollar amount of restructuring charges per common share outstanding. Corrections for heteroscedasticity, multicollinearity, and autocorrelation are made whenever problems are detected. d The White adjusted t-statistics are shown in parentheses. ***Significant at the 0.01 level; **Significant at the 0.05 level; *Significant at the 0.10 level

108

TABLE 4: Price and Return Model Regressions:a b Firm Observations Panel B: OLS Regressions of Equation 2.5a: Return Modelc

(2.5a) Ri,t = ∑ b0 Ct + b1 EPSi,t / Pi,t-1 + b2 ∆EPS / Pi,t-1 + b3 RSTRi,t / Pi,t-1 + ui,t Distressed and Non- Distressed and Non-Distressed Bankrupt Bankrupt Variable Name (Expected Sign)d Coefficients Coefficients Coefficients

EPS/Pi,t-1 (+/-) -0.951 -0.053 0.006 (t-statistic) (-1.57) (-1.18) (1.47)

Change in EPS/Pi,t-1 (+/-) 0.919 -0.095 -0.009 (1.90*) (-23.66***) (-1.77*)

Charge Per Share/Pi,t-1 (+/-) 0.140 0.106 0.044 (0.28) (0.69) (1.20)

Adjusted R2 0.114 0.854 0.245 N 594 firms 533 firms 67 firms F-value 6.87*** 240.16*** 2.79***

Chow Test Results for Returns Model for Firm Observations Non-Distressed Firms Non-Distressed Firms vs. Distressed, Non-Bankrupt vs. Distressed, Non- Distressed, Bankrupt Firms vs. Distressed, Bankrupt Firms Firms Bankrupt Firms F-value 0.817 19.990*** 5.016***

a The coefficients on the yearly intercepts are not reported. The coefficients are all negative and significant. b The data has been trimmed to delete the outliers. The sample in this table includes only observations classified into the same distress classification (0 or 1) by the Altman (1968) and Begley (1996) models. c RSTRi,t is the dollar amount of restructuring charges per common share outstanding. Ri,t is simple returns calculated using the equation Rt = (Pt – Pt-1) / Pt-1, Pi,t-1 is the price three months after fiscal year end from the previous year, ΔEPS is the change in earnings per share between year t-1 and year t, and the other variables are as previously described. Corrections for heteroscedasticity, multicollinearity, and autocorrelation are made whenever problems are detected. d The White adjusted t-statistics are shown in parentheses. ***Significant at the 0.01 level; **Significant at the 0.05 level; *Significant at the 0.10 level

109

TABLE 5: Price and Return Model Regressions:a b Firm-Event Observations Panel A: OLS Regressions of Equation 2.1a: Price Modelc

(2.4a) Pi,t = ∑ a0 Ct + a1 BVPSi,t + a2 EPSi,t + a3 RSTRi,t + ui,t Distressed and Distressed and Non-Distressed Non-Bankrupt Bankrupt Variable Name (Expected Sign)d Coefficients Coefficients Coefficients BVPS (+/-) 1.897 0.318 -0.025 (t-statistic) (7.61***) (3.51***) (-0.25) EPS (+/-) 1.695 -0.161 -0.110 (3.27***) (-1.85*) (-0.91) Charge Per Share (+/-) 0.359 0.269 1.951 (1.08) (1.47) (2.56**)

Adjusted R2 0.442 0.103 0.304 853 firm-event 622 firm-event 87 firm-event N observations observations observations F-value 52.90*** 6.50*** 4.12***

Chow Test Results for Price Model for Firm-Event Observations Distressed, Non- Non-Distressed vs. Non-Distressed vs. Bankrupt vs. Distressed, Distressed, Non-Bankrupt Distressed, Bankrupt Bankrupt F-value 80.71*** 18.32*** 4.91***

a The coefficients on the yearly intercepts are not reported. The coefficients are negative and significant. b The data has been trimmed to delete the top and bottom one percent of the observations. The sample in this table includes only observations classified into the same distress classification (0 or 1) by the Altman (1968) and Begley (1996) models. c Pi,t is the share price three months after fiscal year end, BVPSi,t is book value per share, EPSi,t is earnings per share, and RSTRi,t is the dollar amount of restructuring charges per common share outstanding. Corrections for heteroscedasticity, multicollinearity, and autocorrelation are made whenever problems are detected. d The White adjusted t-statistics are shown in parentheses. ***Significant at the 0.01 level; **Significant at the 0.05 level; *Significant at the 0.10 level

110

TABLE 5: Price and Return Model Regressions:a b Firm-Event Observations Panel B: OLS Regressions of Equation 2.5a: Return Modelc

(2.5a) Ri,t = ∑ b0 Ct + b1 EPSi,t / Pi,t-1 + b2 ∆EPS / Pi,t-1 + b3 RSTRi,t / Pi,t-1 + ui,t Distressed and Distressed and Non-Distressed Non-Bankrupt Bankrupt Variable Name (Expected Sign)d Coefficients Coefficients Coefficients

EPS/Pi,t-1 (+/-) -1.195 -0.037 0.022 (t-statistic) (-2.75***) (-0.80) (0.54)

Change in EPS/Pi,t-1 (+/-) 1.200 0.057 0.020 (3.26***) (2.86***) (0.66)

Charge Per Share/Pi,t-1 (+/-) -0.846 -0.119 -0.070 (-3.05***) (-3.52***) (-2.06**)

Adjusted R2 0.136 0.311 0.298 N 846 600 77 F-value 11.20*** 21.83*** 3.70***

Chow Test Results for Price Model for Firm-Event Observations Non-Distressed vs. Distressed, Non- Non-Distressed vs. Distressed, Non-Bankrupt Bankrupt Distressed, Bankrupt vs. Distressed, Bankrupt F-value 10.17*** 24.76*** 115.1***

a The coefficients on the yearly intercepts are not reported. b The data has been trimmed to delete the top and bottom one percent of the observations. The sample in this table includes only observations classified into the same distress classification (0 or 1) by the Altman (1968) and Begley (1996) models. c RSTRi,t is the dollar amount of restructuring charges per common share outstanding. Ri,t is simple returns calculated using the equation Rt = (Pt – Pt-1) / Pt-1, Pi,t-1 is the price three months after fiscal year end from the previous year, ΔEPS is the change in earnings per share between year t-1 and year t, and the other variables are as previously described. Corrections for heteroscedasticity, multicollinearity, and autocorrelation are made if problems are detected. d The White adjusted t-statistics are shown in parentheses. ***Significant at the 0.01 level; **Significant at the 0.05 level; *Significant at the 0.10 level

111

a TABLE 6: Panel A Regressions of the Price Model (Equation 2.4) b Firm Observations

(2.4a) Pi,t = ∑ a0 Ct + a1 BVPSi,t + a2 EPSi,t + a3 RSTRi,t + ui,t

(2.4b) Pi,t = ∑ a0 Ct + a1 BVPSi,t + a2 EPSi,t + a3 RSTRi,t + a4 OHLSON_PROBi,t + ui,t

(2.4c) Pi,t = ∑ a0 Ct + a1 BVPSi,t + a2 EPSi,t + a3 RSTRi,t + a4 DISTRESSi,t + ui,t

Variable Name (Expected Sign)c Equation 2.4a Equation 2.4b Equation 2.4c BVPS (+) 1.122 1.012 1.035 (1.63) (-1.38) (1.45) EPS (+) -0.492 -0.492 -0.496 (-2.24**) (-2.37**) (-2.39**) RSTR (+) 2.624 2.384 2.409 (1.95*) (1.75*) (1.79*) OHLSON_PROB (-) -14.546 (-2.74***) DISTRESS (-) -11.211 (-3.49***)

Adjusted R2 0.505 0.526 0.526 N 1,207 firms 1,207 firms 1,207 firms F-value 95.67*** 96.69*** 96.43***

a Pi,t is the share price three months after fiscal year end, BVPSi,t is book value per share, EPSi,t is earnings per share, and RSTRi,t is the dollar amount of restructuring charges per common share outstanding. OHLSON_PROB is the probability of financial distress obtained from running Ohlson’s regression model (equation 2.3). DISTRESS equals one when a firm is classified as distressed by the Altman model and the Begley model, and 0 otherwise. Corrections for heteroscedasticity, multicollinearity, and autocorrelation are made whenever problems are detected. The data has been trimmed to delete the top and bottom one percent of the observations. The sample in this table includes only observations classified into the same distress classification (0 or 1) by the Altman (1968) and Begley (1996) models. b The coefficients on the yearly intercepts are not reported. c The White adjusted t-statistics are shown in parentheses. ***Significant at the 0.01 level; **Significant at the 0.05 level; *Significant at the 0.10 level

112

TABLE 6: Panel Ba b Regressions of the Price Model (Equation 2.4) Firm-Event Observations

(2.4a) Pi,t = ∑ a0 Ct + a1 BVPSi,t + a2 EPSi,t + a3 RSTRi,t + ui,t

(2.4b) Pi,t = ∑ a0 Ct + a1 BVPSi,t + a2 EPSi,t + a3 RSTRi,t + a4 OHLSON_PROBi,t + ui,t

(2.4c) Pi,t = ∑ a0 Ct + a1 BVPSi,t + a2 EPSi,t + a3 RSTRi,t + a4 DISTRESSi,t + ui,t

c Variable Name (Expected Sign) Equation 2.4a Equation 2.4b Equation 2.4c BVPS (+) 1.399 1.144 1.199 (t-statistic) (9.43***) (5.66***) (6.69***) EPS (+) 0.508 0.218 0.282 (3.10***) (1.22) (1.72*) RSTR (+) 0.512 0.529 0.519 (1.86*) (1.91*) (1.91*) OHLSON_PROB (-) -10.215 (-4.43***) DISTRESS (-) -7.547 (-5.59***)

Adjusted R2 0.392 0.418 0.420 N 1,562 1,540 1,562 F-value 78.57*** 80.05*** 81.653***

a Pi,t is the share price three months after fiscal year end, BVPSi,t is book value per share, EPSi,t is earnings per share, and RSTRi,t is the dollar amount of restructuring charges per common share outstanding. OHLSON_PROB is the probability of financial distress obtained from running Ohlson’s regression model (equation 2.3). DISTRESS equals 1 for firms classified by both Altman and Begley as distressed, and 0 otherwise. Corrections for heteroscedasticity, multicollinearity, and autocorrelation are made whenever problems are detected. The data has been trimmed to delete the top and bottom one percent of the observations. The sample in this table includes only observations classified into the same distress classification (0 or 1) by the Altman (1968) and Begley (1996) models. b The coefficients on the yearly intercepts are not reported. c The White adjusted t-statistics are shown in parentheses. ***Significant at the 0.01 level; **Significant at the 0.05 level; *Significant at the 0.10 level

113

a TABLE 7: Panel A Regressions of the Return Model (Equation 2.5)b Firm Observationsc

(2.5a) Ri,t = ∑ b0 Ct + b1 EPSi,t / Pi,t-1 + b2 ∆EPS / Pi,t-1 + b3 RSTRi,t / Pi,t-1 + ui,t

(2.5b) Ri,t = ∑ b0 Ct + b1 EPSi,t / Pi,t-1 + b2 ∆EPS / Pi,t-1 + b3 RSTRi,t / Pi,t-1 + b4 OHLSON_PROBi,t-1 + ui,t

(2.5c) Ri,t = ∑ b0 Ct + b1 EPSi,t / Pi,t-1 + b2 ∆EPS / Pi,t-1 + b3 RSTRi,t / Pi,t-1 + b4 DISTRESSi,t + ui,t

Variable Name (Expected Sign) Equation 2.5a Equation 2.5b Equation 2.5c

EPS/Pi,t-1 (+/-) 0.023 0.013 0.014 (t-statistic) (0.70) (0.39) (0.44)

Change in EPS/Pi,t-1 (+/-) -0.095 -0.095 -0.095 (-21.85***) (-21.81***) (-21.67***)

Charge Per Share/Pi,t-1 (+/-) -0.050 -0.023 -0.014 (-0.30) (-0.13) (-0.08) OHLSON_PROB (-) -0.179 (-2.38**) DISTRESS (-) -0.153 (-2.54**)

Adjusted R2 0.809 0.809 0.810 N 1,194 firms 1,194 firms 1,194 firms F-value 389.0*** 362.9*** 363.2***

a RSTRi,t is the dollar amount of restructuring charges per common share outstanding. Ri,t is simple returns calculated using the equation Rt = (Pt – Pt-1) / Pt-1, Pi,t-1 is the price three months after fiscal year end from the previous year, ΔEPS is the change in earnings per share between year t-1 and year t, and the other variables are as previously described. OHLSON_PROB is the probability of financial distress obtained from running Ohlson’s regression model (equation 2.3). DISTRESS equals 1 for firms classified by both Altman and Begley as distressed, and 0 otherwise. Corrections for heteroscedasticity, multicollinearity, and autocorrelation are made whenever problems are detected. The data has been trimmed to delete the top and bottom one percent of the observations. The sample in this table includes only observations classified into the same distress classification (0 or 1) by the Altman (1968) and Begley (1996) models. b The White adjusted t-statistics are shown in parentheses. c The coefficients on the yearly intercepts are not reported. The coefficients were all negative and significant. ***Significant at the 0.01 level; **Significant at the 0.05 level; *Significant at the 0.10 level

114

TABLE 7: Panel Ba Regressions of the Return Model (Equation 2.5) b Firm-Event Observations

(2.5a) Ri,t = ∑ b0 Ct + b1 EPSi,t / Pi,t-1 + b2 ∆EPS / Pi,t-1 + b3 RSTRi,t / Pi,t-1 + ui,t

(2.5b) Ri,t = ∑ b0 Ct + b1 EPSi,t / Pi,t-1 + b2 ∆EPS / Pi,t-1 + b3 RSTRi,t / Pi,t-1 + b4 OHLSON_PROBi,t-1 + ui,t

(2.5c) Ri,t = ∑ b0 Ct + b1 EPSi,t / Pi,t-1 + b2 ∆EPS / Pi,t-1 + b3 RSTRi,t / Pi,t-1 + b4 DISTRESSi,t + ui,t

c Variable Name (Expected Sign) Equation 2.5a Equation 2.5b Equation 2.5c

EPS/Pi,t-1 (+/-) 0.025 -0.027 -0.037 (t-statistic) (0.74) (-0.70) (-0.86)

Change in EPS/Pi,t-1 (+/-) 0.061 0.064 0.064 (3.29***) (3.24***) (3.22***)

Charge Per Share/Pi,t-1 (+/-) -0.150 -0.142 -0.139 (-5.73***) (-6.01***) (-6.11***) OHLSON_PROB (-) -0.238 (-4.34***)

DISTRESSi,t (-) -0.255 (-6.37***)

Adjusted R2 0.202 0.214 0.224 N 1,528 1,513 1,513 F-value 30.67*** 30.36*** 32.13***

a RSTRi,t is the dollar amount of restructuring charges per common share outstanding. Ri,t is simple returns calculated using the equation Rt = (Pt – Pt-1) / Pt-1, Pi,t-1 is the price three months after fiscal year end from the previous year, ΔEPS is the change in earnings per share between year t-1 and year t, and the other variables are as previously described. OHLSON_PROB is the probability of financial distress obtained from running Ohlson’s regression model (equation 2.3). DISTRESS equals 1 for firms classified by both Altman and Begley as distressed, and 0 otherwise. Corrections for heteroscedasticity, multicollinearity, and autocorrelation are made whenever problems are detected. The data has been trimmed to delete the top and bottom one percent of the observations. The sample in this table includes only observations classified into the same distress classification (0 or 1) by the Altman (1968) and Begley (1996) models. b The coefficients on the yearly intercepts are not reported. The coefficients were all negative and significant. c The White adjusted t-statistics are shown in parentheses. ***Significant at the 0.01 level; **Significant at the 0.05 level; *Significant at the 0.10 level

CHAPTER 5

Essay 3

The Impact of Restructuring Charges on Analyst Forecast Accuracy, Bias, and Revisions

CHAPTER 5

Essay 3: The Impact of Restructuring Charges on Analyst Forecast Accuracy, Bias, and Revisions

5.1 Background

The purpose of this essay is to examine the effects of restructuring charge announcements on analysts’ forecast revisions and analyst errors for financially healthy firms and for firms in financial distress. Specifically, this essay examines three types of firms, non-distressed restructuring firms, distressed-non-bankrupt restructuring firms, and distressed restructuring firms that file for bankruptcy within three years of restructuring. Although Chaney et al. (1999) examine the effect of restructuring charges on analysts’ forecast revisions and errors, the authors do not examine how the effects of restructuring on analysts’ forecasts by the three types of firms.

Therefore, one incremental contribution of this essay is to separately examine analyst forecast revisions, accuracy, and bias for three groups of restructuring firms, non-distressed firms, distressed non-bankrupt firms, and distressed bankrupt firms.

The primary contribution of this paper arises because of the lack of extant studies that consider why companies restructure, generally either to improve efficiency or to avoid further financial distress, and the effect of these types of restructuring on analysts. This essay considers the impact that the restructuring has on analyst forecast revisions, accuracy, and bias. I expect that analysts will revise their forecasts upwards for healthy firms and downwards for firms in financial distress. I also expect that analysts’ forecasts are optimistically biased for financially distressed firms and less biased for firms that are financially healthy and restructure to improve their operational efficiency.

116 117

There are several types of restructurings, but this dissertation focuses primarily on operational restructurings. Operational restructurings are quite different from equity restructurings, quasi-reorganizations, and troubled-debt restructurings. Usually to either improve efficiency or to avoid filing for bankruptcy, operational restructurings are multi-dimensional corporate changes undertaken by companies. Equity restructurings involve the altering of ownership interests via spin-offs, split-offs, or equity carve-outs (Kross et al. 2001). Quasi- reorganization occurs when a corporation reorganizes to eliminate its accumulated retained earnings deficit (Herz et al. 1992). Troubled-debt restructuring includes modification of debt covenants, the exchange of other assets for debt, and the exchange of equity for debt (Pirrong and

Koeppen 1993).

Daniels et al. (1995) define operational restructuring as typically involving a firm-level action where personnel are terminated, product lines are eliminated, and assets are disposed of.

By examining the effects of restructuring announcements on analyst forecasts and the impact of restructuring charges on price, this dissertation sheds some light on the usefulness of operational restructurings to firms. Kross et al. (2001) define operational restructurings as typically involving the asset side of the balance sheet and debt and equity restructurings as typically involving the right-hand side of the balance sheet. In this dissertation, operational restructurings are defined as multi-dimensional corporate changes undertaken by companies usually to either improve efficiency or to avoid filing for bankruptcy. EITF 94-3 (1995) required firms to record the costs of restructuring during the period in which management commits to the plan and to disclose many details about the restructuring plan. Also, costs classified as restructuring charges were not allowed to provide future benefit to the firm over and above the restructuring execution. EITF

94-3 was later nullified by SFAS No. 146. Operational restructuring projects typically include some combination of workforce reductions, asset writedowns, the disposal of certain assets and

118

facilities, product line discontinuations, the reconfiguration of facilities, plant relocations, or the closing of certain plants and facilities (Lopez 2002).

The first section of this essay discusses the impact of restructuring charges on analysts’ forecast revisions, bias, and accuracy. In the second section, the hypotheses are developed. In the third section, the data set and the methodology utilized to test the hypotheses are discussed.

The empirical results of this essay are described in the fourth section of this essay. Finally, the fifth section summarizes and draws conclusions.

5.2 Impact of Restructuring Charges on Analyst Forecasts

5.2.1 Analysts and Restructuring

One of the first studies to examine the effect of reported restructuring charges on analysts’ forecasts is Chaney et al. (1999). They examine financial analysts’ reaction to the announcement of a restructuring. The authors provide evidence that analysts expect performance to decline for restructuring firms over the short-run with possible improvements in stock price and other financial indicators over the long term, three to five years after the restructuring. This essay examines the reaction of analysts to restructuring announcements to determine whether analysts react differently depending on the financial health of the firm. As in the many other cited studies, the essay examines analyst forecast revisions, accuracy and bias to determine the impact of a company’s restructuring efforts on analysts.

Lin and Yang (2006) examine the effect of previous restructuring charges on analyst forecast revisions and accuracy. The authors determine that analysts react differently to first-time restructuring firms than to repeat restructuring firms. The authors find that analysts revise their forecasts of both one-year-ahead earnings and five-year earnings growth more negatively for first-time restructuring firms than for firms that have taken charges in the past. When the authors

119

examine forecast errors in the year following the restructuring, they find that restructuring charges in the current period complicate the earnings forecast process for analysts. Finally, the authors determine that the documented decrease in analyst forecast accuracy is reduced if a company has prior charges in the previous two years.

5.2.2 Analyst Forecast Revisions

One research paradigm within the analyst forecast literature relates to analyst forecast revisions surrounding various events. Huai (2000) investigates the extent to which investors exhibit rational behavior in response to auto-correlated analyst forecast revisions. Mest and

Plummer (2000) use a linear model to demonstrate how analysts tend to revise their forecasts of future earnings in response to current forecast errors.1 Claus (2000) examines analysts’ earnings

forecasts for six countries and determines that analysts’ revisions occur most frequently around

periods when financial statement information is released. Barron et al. (2002) examine changes

in the precision and commonality of information contained in individual analysts’ earnings

forecasts.2 Lim and Kong (2004) obtain evidence from four Asia-Pacific markets that indicates

that abnormal returns are related to the latest forecast revisions. Hollie et al. (2005) determine that

analysts revise their forecasts upon the preliminary earnings announcement and ignore the new

information in SEC filings. Many other studies have also examined issues related to analyst

forecast revisions (Ivkovic and Jegadeesh 2002, Barth and Hutton 2003, Asquith et al. 2004,

Pinello 2005, Yan 2006).

5.2.3 Analyst Forecast Accuracy

1 Mest and Plummer (2000) find that a non-linear model actually better describes the association between analysts’ forecast revisions and their forecast errors. 2 Barron et al. (2002) determine that the idiosyncratic information in individual analysts’ forecasts increases after earnings announcements.

120

Much of the accounting literature published in the last five years relates to analysts and

their forecast accuracy. Duru and Reeb (2002) examine the relationship between corporate

international diversification and the accuracy of analysts’ earnings forecasts.3 Kwon (2002)

examines the difference in analyst forecast accuracy between high- and low-tech firms and

determines that there is lower unsigned error for high-tech firms than for low-tech firms.

Clement and Tse (2003) examine whether investors extract all of the information that analysts’

characteristics provide about forecast accuracy.4 Bonner et al. (2003) examine whether there are

differences in how sophisticated and unsophisticated investors use factors including

characteristics of the analyst and the age of the forecast to predict the accuracy of forecast

revisions.5

Barniv et al. (2005) examine the ability of analysts’ characteristics to explain the relative

forecast accuracy of analysts across legal origins. Chen et al. (2005) demonstrate that investors’

reactions to forecast news are increasing with forecast accuracy and the length of analysts’

forecast records. Chiang (2005) examines the relationship between analyst forecast accuracy and

corporate transparency and determines that analysts’ forecast accuracy does relate strongly to

corporate transparency. Easton and Monahan (2005) evaluate the reliability of an expected return

proxy using its relationship with realized returns.6 Fan et al. (2006) investigate the accuracy of

the earnings forecasts of financial analysts from investors and financial analysts and determine

that analysts’ forecasts outperform random walk time-series forecasts. Many other studies have

3 Duru and Reeb (2002) find that greater corporate international diversification leads to less accurate forecasts. 4 Clement and Tse (2003) determine that only some of the analysts’ characteristics associated with future forecast accuracy are also associated with return responses to forecast revisions. 5 Bonner et al. (2003) determine that sophisticated investors have more knowledge about the relationship of the factors affecting forecast accuracy. 6 Easton and Monahan’s (2005) results demonstrate that the return proxies examined are unreliable and demonstrate that some proxies are reliable when the long-term growth forecasts are low or when analysts’ forecast accuracy is high.

121

also discussed issues related to analyst forecast accuracy (Gilson et al. 1998, Harjoto and Zaima

2005, Kini et al. 2005, Liu and Su 2005, Clement et al. 2006, Ertimur et al. 2006).

5.2.4 Analyst Forecast Bias

Analyst forecast bias is another area that has often been examined by the extant literature.

Dechow et al. (2000) examine the role of sell-side analysts’ earnings growth forecasts in the pricing of common stock and determine that these analysts’ forecasts are overly optimistic around stock offerings. Lim (2001) finds that positive and predictable analysts’ forecast bias may be a rational property of all optimal earnings forecasts. Mest and Plummer (2003) extend Lim’s

(2001) study by examining analysts’ earnings and sales forecasts, and they determine that analysts’ optimistic bias will be greater for earnings forecasts than for sales forecasts.

Han et al. (2001) demonstrate that publicly-available information can be used to establish estimates of analysts’ optimistic bias in earnings forecasts. Duru and Reeb (2002) examine the relationship between corporate international diversification and the bias of analysts’ earnings forecasts and determine that greater corporate international diversification leads to more optimistic forecasts. Gu and Wu (2003) find that earnings skewness is significantly related to analysts’ forecast bias. Many other studies have also examined issues related to analyst forecast bias over the last few years (Ciccone 2001, Abarbanell and Lehavy 2003, Cowen et al. 2003,

Scherbina 2004, Friesen and Walker 2005).

5.2.5 Market Reaction to Restructuring and Bankruptcy

Park and Stice (2000) identify superior analysts using their past history for a specific firm’s earnings, and they demonstrate that subsequent forecast announcements by the superior analysts have more impact on price than the forecasts of other analysts. Shane and Brous (2001)

122

provide evidence that analysts and investors correct the documented underreactions of earnings forecasts and stock prices to earnings news using the next earnings announcement and other information available between earnings announcements. Lacina and Karim (2004) demonstrate that analysts react less negatively to management forecasts of improved earnings expectations than to management forecasts of bad earnings.

5.2.6 Other Issues

Elgers et al. (2001) suggest that the weighing of analysts’ annual earnings forecasts implicit in prices is lower than the historical relationship between financial analysts’ forecasts and realized earnings. Athanassakos and Kalimipalli (2003) examine the relationship between analysts’ forecast dispersion and future stock return volatility and determine that there is a strong positive relationship between analysts’ forecast dispersion and future return volatility. Irvine

(2004) examines whether analysts’ earnings forecasts and stock recommendations affect their brokerage firms’ share of trading in the forecast stocks.7 Stevens and Williams (2004) analyze

forecast reactions to positive versus negative information, and the forecast data demonstrate that

there is a systematic underreaction to both positive and negative information.8

5.3 Hypotheses Development

The hypotheses separately examines the effects of restructuring charge announcements on analysts’ forecast revisions and errors for financially healthy firms and for firms that are determined to be in financial distress.

7 Irvine (2004) determines that individual analysts’ forecasts that differ from the consensus forecast generate significant brokerage-firm trading in the forecast stocks in the two weeks after the forecast release date. 8 The documented underreaction is determined to be larger for positive information than for negative information.

123

This leads to the following hypotheses:

H3a: Analysts revise their forecasts downward over the short-run and slightly upward

over the long-run after learning of a restructuring announcement.

H3b: Analyst forecast accuracy decreases with the increasing probability of financial

distress during the year of the restructuring.

H3c: Analyst forecast bias increases with restructuring announcements and with the

increasing probability of financial distress.

5.4 Data Sources and Methodology

5.4.1 Data and Sample

The full sample for this essay includes data for the period from 1992 to 2004 and firms announcing operational restructurings during the period from 1993 to 2003.9 Companies that

underwent debt restructurings or management restructuring are not relevant and are therefore

excluded. The companies undergoing operational restructurings are examined. Financial and

market data are obtained from COMPUSTAT, and financial analyst data are obtained from the

I/B/E/S Summary-File Database. The expected signs for all of the equations are provided in the

tables at the end of this essay. Table 1 provides the steps used to arrive at the final sample of

1,172 firm-year observations and 33,811 firm-quarter observations. The analyst forecast revision

portion of this essay uses firm-quarter observations, and the accuracy and bias portion of the

essay uses firm-event observations as in Chaney et al. (1999). The sample of firm-event

observations for the analyst forecast accuracy and bias portion of the study is smaller than the

total number of firm-event observations in Panel A because additional observations are lost due to

missing data from I/B/E/S needed to calculate AFEt+1 and AFEt-1. While firm-quarter

9 Because this essay requires data from periods t+1 and t-1, some 1992 and 2004 data are also used.

124

observations are used for the analyst revision portion of this essay, annual forecasts and firm-year observations for year t, t+1, and t-1 are used for the accuracy and bias portion.

As shown in Table 2, the initial sample is comprised primarily of manufacturing firms, with 55.8 percent of the sample firms coming from this category. Service firms comprise 23.7 percent of the sample firms. The sample also contains smaller percentages of firms from the transportation, communication, gas and electric category (6.2%), the wholesale and retail trade categories (4.6 and 5.6%), and the financial, insurance, and real estate category (1.8%).

5.4.2 Methodology

Prior to testing the hypotheses, it is necessary to create a sample of firms that have restructured. The full sample for this essay contains firms undertaking operational restructuring efforts during the period from 1993 through 2003 that have data availability for the required variables. Equations 2.1 through 2.3 are used to determine whether each firm is in financial distress. In order to determine a “distress” value for each firm in the sample, Altman’s (1968) original Z-score model and Begley et al.’s (1996) updated version of the Altman model are used.

Although the Altman and Begley models were originally intended as bankruptcy

prediction models, Grice and Dugan (2001) indicate that bankruptcy prediction models like

Altman’s are actually more useful for identifying firms that are financially distressed, as opposed

to identifying the more limited bankruptcy condition. Because these models have been proven

successful, this study uses the linear Z-score equations and substitutes the numbers for each

variable for the firms in the sample. These models are used to determine a distress value for each

125

firm, and then a cutoff point can be used to classify firms as either distressed or healthy.

Altman’s (1968) Z-score model is as follows:10

Z = 0.012 X1 + 0.014 X2 + 0.033 X3 + 0.006 X4 + 0.999 X5, (3.1) where11

Z is used to determine whether each company is in financial distress;12

13 X1 is working capital divided by total assets * 100;

14 X2 is retained earnings divided by total assets * 100;

15 X3 is earnings before interest and taxes divided by total assets * 100;

16 X4 is the market value of equity divided by the book value of debt * 100;

17 X5 is sales divided by total assets.

Begley et al. (1996) re-estimate Altman’s model using data from the 1980s, and their updated model is as follows:

Z = 0.104 X1 + 1.010 X2 + 0.106 X3 + 0.003 X4 + 0.169 X5, (3.2)

10 It is important to note that because of Altman’s original computer format arrangement, variables X1 to X4 are included in calculations as absolute percentage values (10% as opposed to .10). Only X5 is expressed as a decimal instead of a percentage because of its extremely high relative discriminant coefficient. 11 Altman finds that for his sample firms, firms with Z-scores greater than 2.99 were mostly not in financial distress and many of the firms with Z-scores less than 1.81 went bankrupt. Altman further finds that using a Z-score of 2.675 as a cutoff minimizes the number of firms that are misclassified by the model. Therefore, this study uses 2.675 as the cutoff point for the Altman model results. 12 All X-values are included in the calculation of Z for each firm or firm-event observation, even when the values are negative. 13 Working capital divided by total assets is a measure of the net liquid assets of a firm relative to the overall capitalization; firms with losses are likely to also have shrinking current assets compared to total assets. Altman (1968) finds that working capital divided by total assets is the most valuable measure of the liquidity. 14 Retained earnings divided by total assets is included because it implicitly considers the age of a firm, and financial distress is much more common in the early years of a firm’s life. 15 Earnings before interest and taxes, divided by total assets is a measure of the true productivity of a firm’s assets, ignoring tax and leverage factors. Because a firm’s existence is based on the earning power of the firm’s assets, this ratio is especially important. 16 The market value of equity divided by book value of debt variable shows how much the firm’s assets can decline in value before the firm becomes insolvent. Including the market value of equity divided by the book value of debt adds a market value dimension not considered before Altman (1968), and this variable is determined to be a better predictor of bankruptcy than net worth/total debt. 17 Sales divided by total assets is a measure of firm size.

126

using the same variables and variable definitions as Altman’s model.18

The observations used in this essay are those classified by the Altman and Begley models into the same groups.

After each firm is classified as either being financially distressed or non-distressed using both the Altman (1968) model and the Begley et al. (1996) model, Ohlson’s (1980) logistic regression model is used to confirm the accuracy of the classification procedure. Also, the results of the Ohlson logistic regression model provide a probability value between 0 and 1 for each firm that indicates the likelihood of a firm’s being in financial distress. These probability values are included as additional predictor variables in the following models. Ohlson’s (1980) model is as follows:19

DISTRESSi,t = a0 + a1 SIZE i,t + a2 TLTA i,t + a3 WCTA i,t + a4 CLCA i,t + (3.3)

a5 NITA i,t + a6 FUTL + a7 INTWOi,t + a8 OENEGi,t + a9 CHIN + ui,t,

where

DISTRESS equals 1 if a firm is determined to be in financial distress, 0 otherwise;

SIZE is the log of total assets;

TLTA is total liabilities divided by total assets;

WCTA is working capital divided by total assets;

CLCA is current liabilities divided by current assets;

OENEG equals 1 if owners’ equity is negative, 0 otherwise;

18 Begley et al. (1996) find that the most appropriate cutoff point for their model is 0.545. Firms with Z- scores less than 0.545 are classified as financially distressed and are assigned a value of 1, and firms with Z-scores greater than 0.545 are classified as being non-distressed and are assigned a value of 0. 19 Ohlson’s (1980) model includes nine explanatory variables, and all of them are included in this essay even though Ohlson finds that only six of them are significant. SIZE, TLTA, NITA, FUTL, and CHIN are all significant predictors of bankruptcy in Ohlson’s (1980) model. The variable TLTA is included as a measure of firm leverage, and NITA and FUTL are included as measures of firm performance. The variable OENEG is used as a discontinuity correction for TLTA. The variables WCTA and CLCA are included as measures of current liquidity. The CHIN variable is a measure of the change in net income that is included because of its importance in McKibben (1972).

127

NITA is net income divided by total assets;

FUTL is cash flows from operations divided by total liabilities;

INTWO equals 1 if net income was negative over the last two years, 0 otherwise;

CHIN = (NIt – NIt-1) / ( | NIt | + | NIt-1 | ).

H3a and H3b are tested by examining analysts’ earnings forecasts revisions surrounding the announcement of a restructuring charge. Based on Chaney et al. (1999), the first examination is the analysts’ forecast revisions in the period surrounding the announcement for the healthy and distressed firms to determine whether analysts view restructuring charges as good news, bad news, or uninformative. Forecast revisions are defined as follows:

h h h AFREVi,j = AFi+1 – AFi-1 , (3.4)

where

h AFREVi,j is the analyst forecast revision of h-year ahead earnings (h = 1, 2, or 3) or

earnings growth (h = 5) around a restructuring charge announced in month i or

quarter j;

h AFi-1 is the mean forecast of h-year ahead earnings (h = 1, 2, or 3) or earnings growth (h

= 5) in the month i-1, the month prior to the announcement of the restructuring

charge, deflated by price at the beginning of the period i-1;

h AFi+1 is the mean forecast of h-year ahead earnings (h = 1, 2, or 3) or earnings growth (h

= 5) in month i+1, the month subsequent to the announcement of the

restructuring charge, deflated by price at the beginning of the period i-1.

Chaney et al. (1999) examine analysts’ forecast accuracy surrounding a restructuring charge announcement, and their research design is adapted here to separately examine analysts’ forecast accuracy around the announcement of the restructuring charge for distressed and healthy firms. To test the relationship between analysts’ forecast revisions and restructuring charges, the

128

following regression also used by Chaney et al. (1999) is utilized. Analysts’ forecasts revisions are regressed on a dummy variable that equals one if the firm announced a restructuring charge in that quarter and zero otherwise.

h AFREVi,j = B0 + B1 LOSSj + B2 UEWOj + B3 RESTj + B4 LOSS * RESTj + (3.5a)

B5 LOSS * UEWOj + u,

h AFREVi,j = B0 + B1 LOSSj + B2 UEWOj + B3 RESTj + B4 LOSS * RESTj + (3.5b)

B5 LOSS * UEWOj + B6 Ohlson_Prob + u, and

h AFREVi,j = B0 + B1 LOSSj + B2 UEWOj + B3 RESTj + B4 LOSS * RESTj + (3.5c)

B5 LOSS * UEWOj + B6 DISTRESS + u.

The following equations, (3.5d) and (3.5e), are used for unobservable ex-post analysis that includes the bankruptcy event:

h AFREVi,j = B0 + B1 LOSSj + B2 UEWOj + B3 RESTj + B4 LOSS * RESTj + (3.5d)

B5 LOSS * UEWOj + B6 D1 + B7 D2 + B8 D3 + u and

h AFREVi,j = B0 + B1 LOSSj + B2 UEWOj + B3 RESTj + B4 LOSS * RESTj + (3.5e)

B5 LOSS * UEWOj + B6 BANKRUPT + u, where

h AFREVi,j is the revision in the mean forecasts for h-year ahead earnings (h=1, 2, or 3) or

earnings growth (h=5) subsequent to a restructuring charge and/or earnings

announcement in month i of quarter j;

LOSSj equals 1 if income before restructuring charges and extraordinary items in quarter

j is less than zero, 0 otherwise;

UEWOj is unexpected earnings before restructuring charges in quarter j, defined below;

RESTj equals 1 if the firm announced a restructuring charge in quarter j, 0 otherwise;

LOSS * RESTj is the interaction of LOSS and REST;

129

LOSS * UEWOj is the interaction of LOSS and UEWO;

OHLSON_PROB is the probability of financial distress obtained from running Ohlson’s

logistic regression model shown in equation 3.3;

DISTRESS equals 1 for firms classified by the Altman and Begley model to be

distressed; 0 otherwise,

D1 equals 1 for non-distressed firms, 0 otherwise;

D2 equals 1 for distressed firms that do not file for bankruptcy for at least three years of

restructuring, 0 otherwise;

D3 equals 1 for firms that file for bankruptcy within three years of restructuring, 0

otherwise;

BANKRUPT equals 1 for firms that file for bankruptcy within three years of

restructuring, 0 otherwise.

Unexpected earnings before restructuring charges (UEWOj) is calculated by subtracting

AFj from EWOj, where AFj is the most recent mean analyst forecast for quarter j and EWOj is earnings before restructuring charges and extraordinary items in quarter j. If analysts view restructuring charges as good news for future earnings, a positive coefficient on REST is expected. Alternatively, if analysts interpret restructuring charges as a sign of declining performance, the coefficient on REST should be negative. Also, if unexpected earnings persist into the future, there should be a positive coefficient on UEWO. As in Chaney et al. (1999), this study includes an interaction term for loss and unexpected earnings because prior research suggests that the market’s response to unexpected earnings is different in the presence of a loss

(Hayn 1995). Also, Hogan and Jeter (1998) suggest that the market’s reaction to a restructuring charge is very content-specific and that this reaction is altered in the presence of a loss. This

130

paper also includes an interaction term between the presence of a loss and the presence of a restructuring charge as utilized by Chaney et al. (1999).

The results of the Ohlson logistic regression model in equation 3.3 provide a probability value between 0 and 1 for each firm that indicates the likelihood of a firm being in financial distress. These probability values are included in equation 3.5b above. The distress values assigned to each firm by using the Altman (1968) and Begley (1996) models in equations 3.1 and

3.2 are included as additional explanatory variables in equation 3.5c above.

To test H3b and H3c, the study assesses analysts’ forecast accuracy and bias in forecasting future earnings subsequent to a restructuring charge. Following the procedure used by Chaney et al. (1999), this study defines the forecast error and forecast bias in the year subsequent to the restructuring charge as the difference between actual earnings and the most recent mean forecast prior to the annual earnings announcement as follows. The year of the restructuring charge, t, is eliminated as in Chaney et al. (1999) because it is not clear whether or not analysts’ forecasts for that period include any expectation of a restructuring charge. AFEt+1 is the forecast error (i.e., a measure of forecast accuracy), and it is calculated in equation 3.6.

AFEt+1 = (EPSt+1 – AFt+1) / Pt+1, (3.6) where

AFEt+1 is the analyst forecast error for year t+1;

EPSt+1 is actual earnings per share for year t+1;

AFt+1 is the mean earnings forecast for year t+1 measured in the month preceding the

earnings announcement for year t+1;

Pt+1 is the market price of stock at the beginning of year t+1.

To consider the accuracy and bias of analysts’ forecasts subsequent to a restructuring charge as in

Chaney et al. (1999), the following regression equations are estimated. The study estimates the

131

regression using both the absolute values of forecast errors and the actual forecast bias (without the absolute value) to assess accuracy and bias, respectively, in equations 3.7 and 3.8.

⏐AFEt+1⏐ = ∑ γt Ct + γ1 ⏐AFEt-1⏐ + γ 2 RESTt + γ 3 ⏐RETt,t+1⏐ + γ4 LNUMBt+1 (3.7a)

+ γ5 RECESSt+1 + v,

⏐AFEt+1⏐ = ∑ γt Ct + γ1 ⏐AFEt-1⏐ + γ 2 RESTt + γ 3 ⏐RETt,t+1⏐ + γ4 LNUMBt+1 (3.7b)

+ γ5 RECESSt+1 + γ6 OHLSON_PROB + v, and

⏐AFEt+1⏐ = ∑ γt Ct + γ1 ⏐AFEt-1⏐ + γ 2 RESTt + γ 3 ⏐RETt,t+1⏐ + γ4 LNUMBt+1 (3.7c)

+ γ5 RECESSt+1 + γ6 DISTRESS + v.

The following equations, (3.7d) and (3.7e), are used for unobservable ex-post analysis that

includes the bankruptcy event:

⏐AFEt+1⏐ = ∑ γt Ct + γ1 ⏐AFEt-1⏐ + γ 2 RESTt + γ 3 ⏐RETt,t+1⏐ + γ4 LNUMBt+1 (3.7d)

+ γ5 RECESSt+1 + γ6 D1 + γ7 D2 + γ8 D3 + v,

⏐AFEt+1⏐ = ∑ γt Ct + γ1 ⏐AFEt-1⏐ + γ 2 RESTt + γ 3 ⏐RETt,t+1⏐ + γ4 LNUMBt+1 (3.7e)

+ γ5 RECESSt+1 + γ6 BANKRUPT + v.

The following equations are related to forecast bias:

AFEt+1 = ∑ γt Ct + γ1 AFEt-1 + γ 2 RESTt + γ 3 RETt,t+1 + γ4 LNUMBt+1 (3.8a)

+ γ5 RECESSt+1 + v,

AFEt+1 = ∑ γt Ct + γ1 AFEt-1 + γ 2 RESTt + γ 3 RETt,t+1 + γ4 LNUMBt+1 (3.8b)

+ γ5 RECESSt+1 + γ6 Ohlson_Prob + v, and

AFEt+1 = ∑ γt Ct + γ1 AFEt-1 + γ 2 RESTt + γ 3 RETt,t+1 + γ4 LNUMBt+1 (3.8c)

+ γ5 RECESSt+1 + γ6 DISTRESS + v.

The following equations, (3.8d) and (3.8e), are used for unobservable ex-post analysis that

includes the bankruptcy event:

132

AFEt+1 = ∑ γt Ct + γ1 AFEt-1 + γ 2 RESTt + γ 3 RETt,t+1 + γ4 LNUMBt+1 (3.8d)

+ γ5 RECESSt+1 + γ6 D1 + γ7 D2 + γ8 D3 + v, and

AFEt+1 = ∑ γt Ct + γ1 AFEt-1 + γ 2 RESTt + γ 3 RETt,t+1 + γ4 LNUMBt+1 (3.8e)

+ γ5 RECESSt+1 + γ6 BANKRUPT + v. where

⏐AFEt+1⏐ is the absolute value of the analyst forecast error for year t + 1, the year after

the restructuring charge;

AFEt+1 is the analyst forecast error for year t + 1, the year after the restructuring charge;

⏐AFEt-1⏐ is the absolute value of the analyst forecast error for year t – 1, the year before

the restructuring charge

AFEt-1 is the analyst forecast error for year t – 1, the year before the restructuring charge;

Ct equals 1 if the observation is from year t, where t represents a year between 1993 and

2003, 0 otherwise;

RESTt equals 1 if the firm announced a restructuring charge in year t, 0 otherwise;

⏐RETt,t+1⏐ is the absolute value of the market return from the beginning of period t to the

end of year t + 1;

RETt,t+1 is the market return from the beginning of period t to the end of year t + 1;

LNUMBt+1 is the log of the number of analysts forecasting earnings for a given firm in

year t + 1;

RECESSt+1 equals 1 if the observation occurred during a time of recession as defined by

the National Bureau of Economic Research, 0 otherwise;

DISTRESS equals 1 for firms classified by the Altman and Begley model to be

distressed, 0 otherwise;

133

OHLSON_PROB is the probability of financial distress obtained from running Ohlson’s

logistic regression model shown in equation 3.3;

D1 equals 1 for non-distressed firms, 0 otherwise;

D2 equals 1 for distressed firms that do not file for bankruptcy for at least three years of

restructuring, 0 otherwise;

D3 equals 1 for firms that file for bankruptcy within three years of restructuring, 0

otherwise;

BANKRUPT equals 1 for firms that file for bankruptcy within three years of

restructuring, 0 otherwise.

As in Chaney et al. (1999), this paper includes lagged forecast errors based on the view that analysts fail to learn completely from their own prior errors. Firm-market returns are included as a control variable as in prior studies including Ali et al. (1992). However, if the analysts’ forecasts are efficient, they should reflect all information revealed in past stock returns about future earnings, and thus the coefficient on the returns variable should be 0. Also included is a recession variable to control for the impact of economic contraction on forecast errors as in

Chaney et al. (1999). Because prior studies including Lys and Soo (1995) find that the number of analysts’ following a firm is highly correlated with analysts’ forecast accuracy, this study includes the log of the number of analysts as a control variable in equations 3.7 and 3.8. Also,

Chaney et al. (1999) suggest that this variable be included to control for the relation between forecast errors and analyst following documented by Alford and Berger (1999). Although Lang and Lundholm (1996) suggest that a firm’s size may impact analyst forecast accuracy, Chaney et al. (1999) demonstrate that including market value of equity in the equations as a measure of size in the equations does not impact the results. As an additional sensitivity analysis, this study also

134

runs equation 3.8 once for positive analyst forecast errors (positive bias) and separately for negative analyst forecast errors (negative bias).

The results of the Ohlson logistic regression model in equation 3.3 provide a probability value between 0 and 1 for each firm that indicates the likelihood of a firm being in financial distress. These probability values are included in equations 3.7b and 3.8b. The distress values assigned to each firm by using the Altman (1968) and Begley (1996) models in equations 3.1 and

3.2 are included as additional explanatory variables in equation 3.7d and 3.8d above.

5.5 Empirical Results

5.5.1 Univariate Analysis

Detailed descriptive statistics for the sample are provided in Tables 3 and 4. Table 3 provides the results obtained after deleting the outlier observations for each variable.20 Panels A and B provide descriptive statistics for the full sample of firm-quarter observations and the firm- quarter observations classified into the same group by the Altman and Begley models, respectively, for the analyst forecast revision portion of the essay. Table 4 provides descriptive statistics for the full sample of firm-years in Panel A and firm-years classified into the same group by both the Altman and Begley models in Panel B for the analyst forecast error and bias portion of the essay. Several of the statistics provide interesting information about the sample.

Table 3 shows that the analysts’ forecast revision of the next period’s earnings

h=1 (AFREVi,j ) is positive, on average (+0.002). The mean analysts’ forecast revision for the 2-

h=2 year ahead earnings (AFREVi,j ) is also positive, on average (+0.001). The mean forecast revisions are also positive for the 3-year and 5-year ahead forecasts (both +0.0001). In Panel B of

Table 4, the mean analysts’ forecast errors (AFE), both before and after the restructuring and/or

20 The top and bottom one percent of observations were trimmed.

135

earnings announcement period, are negative (-0.03 and -0.02, respectively), which indicates that the analysts are optimistic over the sample period.

Table 5 provides a univariate analysis of forecast errors. This analysis makes it possible to provide a discussion of how restructuring charges affect analysts’ accuracy and cause possible bias in analyst forecasts. Two comparisons are provided in Table 5. First, the study compares the absolute value of analysts’ forecast errors measured in the year preceding the restructuring charge and the absolute value of the forecast errors measured in the year after the charge. An increase in this absolute value indicates that there is a decline in accuracy, but the results are not significant in Panel B. Next, the actual errors before and after the restructuring charge are compared to see if analysts’ forecasts are biased and whether this bias is different if there is a restructuring charge.

The results in Panel A of Table 5 are for the firm data while the results in Panel B of Table 5 are for the firm-event data.

The results in Panel B of Table 5 demonstrate that the mean absolute forecast error for the restructuring sample declines from 6.5% in the year before the restructuring charge to 3.6% in the year after the restructuring charge for the firm-event data. The decrease is significant at the

0.01 level, based on a t-test for the difference in the values. Contrary to the findings of Chaney et al. (1999), this indicates that for this sample, restructuring charges actually improve the analysts’ abilities to forecast earnings. The comparison of actual forecast errors demonstrates that the mean forecast error for the restructuring sample changes from -4.6% in the year before the restructuring charge to -2.6% in the year after the restructuring charge, and this change is significant at the 0.10 level. Again contrary to the findings of Chaney et al. (1999), this indicates that analysts are less optimistic in their forecasts after restructuring charges. The results in Panel

A for the firm data are insignificant.

136

5.5.2 Analyst Forecast Revisions

Tables 6 presents the results of estimating equations 3.5a, 3.5b, and 3.5c with h set equal to one, where h represents the number of years ahead for the forecast. The tables for this portion of the essay contain firm-quarter observations as in Chaney et al. (1999). The firm-quarter observations included in this paper are classified into the same health category by the Begley

(1996) model and the Altman (1968) model. The first column in Panel A of Table 6 presents the effect of restructuring charges on the h-year ahead forecast revisions. The second and third columns of each table examine the incremental effect of adding the financial health indicator variables, Ohlson_Prob and DISTRESS, to the equation. The coefficient on the variable of interest, REST, is negative and significant for h = 1 when the Ohlson probability variable is included as an additional predictor variable in column 2 of Table 6. This suggests that analysts revise their forecasts downward after learning of a restructuring charge, at least over the short run. This result is consistent with the findings of Chaney et al. (1999) and indicates that analysts believe any incremental information about changes signaled by the restructuring charge leads to a decrease in earnings. These findings suggest that analysts view restructuring charges as

moderately bad news over the short term. The results support H3a and suggest that analysts may

view restructuring as indicating a short-term decrease in profits.

The coefficient on unexpected earnings (UEWO) is positive and significant, as expected,

and this suggests that the unexpected component of earnings is expected to persist into the

following year. The coefficient on the variable LOSS is negative and significant, suggesting that

analysts revise their one year ahead forecasts downward in the presence of a loss in the current

period. The coefficient on the interactive variable LOSS*REST is negative and significant for

equation 3.5a, and this suggests that analysts revise their forecasts down to a greater extent in the

presence of both a loss and a restructuring charge. The LOSS*UEWO coefficient is negative for

137

equations 3.5a, 3.5b, and 3.5c, suggesting that analysts react to unexpected earnings by further revising forecasts downward when there is a loss over forecast horizons of one year.

For equation 3.5b, where the probability of distress as calculated with the Ohlson (1980) model is included as an explanatory variable, the coefficient on the Ohlson probability variable is positive and significant when h=1. The positive and significant coefficient on the Ohlson probability variable seems to indicate that analysts revise their short-term forecasts upward when the likelihood of distress as determined by the Ohlson (1980) model is greater, an unexpected finding. When the DISTRESS dummy variable is included in equation 3.5c, the coefficient on this variable, which equals one for firms classified by both the Altman and Begley model to be distressed, and 0 otherwise, is positive and significant. This suggests that analysts revise their forecasts for distressed firms upward over the short-run. This provides support for H3b. The model with the highest adjusted R-square, and therefore the highest predictive value, is equation

3.5b, the model that includes OHLSON_PROB as an additional predictor variable.

Panel B of Table 6 provides the results of ex post analysis that includes the bankruptcy event. The results of the regressions of equations 3.5d and 3.5e, which include bankruptcy indicator variables, are similar to the results of the regressions of equations 3.5b and 3.5c.

Equation 3.5d includes three dummy variables, where D1 equals one for non-distressed firms, D2 equals one for distressed firms that avoid filing for bankruptcy for three years after restructuring, and D3 equals one for bankrupt firms. Equation 3.5e contains one dummy variable that equals one for bankrupt firms, and zero otherwise. For both equations 3.5d and 3.5e, the coefficients on

LOSS and LOSS*REST are both negative and significant. For equation 3.5d, the coefficient on

D2 is positive and significant, which suggests that analyst forecast revisions are greater for distressed firms that avoid filing for bankruptcy for at least three years. The negative and significant coefficient on D3 suggests that analysts forecast revisions are smaller for firms filing

138

for bankruptcy within three years of restructuring. For equation 3.5e, the negative and significant coefficient on BANKRUPT confirms the finding that analysts forecast revisions are smaller for firms filing for bankruptcy within three years of restructuring.

Tables 7 through 9 present the results of estimating equations 3.5a, 3.5b, and 3.5c with h set equal to two, three, and five, respectively, where h represents the number of years ahead for the forecast. The first column of each table presents the effect of restructuring charges on the h- year ahead forecast revisions. The second and third columns of each table examine the incremental effect of adding the financial health indicator variables, Ohlson_Prob and

DISTRESS, to the equations. The coefficient on the variable of interest, REST, is insignificant over forecast horizons of two, three and five years. The coefficient on unexpected earnings

(UEWO) is positive and significant over the two-year forecast horizon, which suggests that the unexpected component of earnings persists into the following year for forecast horizons of two

years. The coefficients on the interactive variable LOSS*UEWO are also negative and significant for all three models over the two-year forecast horizon, suggesting that analysts react to unexpected earnings by further revising forecasts downward when there is a loss over forecast horizons of two years.

For equation 3.5b, where the probability of distress as calculated with the Ohlson (1980) model is included as an explanatory variable over each forecast horizon, the coefficient on the

Ohlson probability variable is positive and significant when h equals one. The positive and significant coefficient on Ohlson_Prob when h equals one indicates that analysts revise their short-term forecasts upward when the likelihood of distress as determined by the Ohlson (1980) model is greater. The coefficient on the DISTRESS variable, which equals one for firms classified by both the Altman and Begley model to be distressed, and zero otherwise, is

139

insignificant for all models in Tables 7, 8 and 9.21 Overall, the results reported in Tables 6-9 tend to support H3a.

5.5.3 Analyst Forecast Accuracy

Table 10 presents the results of estimating several variations of equation 3.7, which examines the accuracy of analyst forecasts. Panel A provides ex ante analysis while Panel B provides the ex post analysis. In Panel A, the coefficient on the variable of interest, REST, is negative and significant. This suggests that analysts’ absolute forecast errors decline and that accuracy increases after a restructuring charge when compared to a similar period not following a charge, and this finding fails to support H3b. The coefficient on the lagged forecast errors is positive and significant for equations 3.7a and 3.7b. The coefficient on returns is negative and insignificant for equations 3.7b and 3.7c, which indicates that analysts are more accurate and make fewer errors when there are greater returns over the past two years if a company has a greater likelihood of being in financial distress. As predicted, the coefficient on the analysts’ following variable is negative and significant for equation 3.7c, and this indicates that absolute forecast errors increase when there are fewer analysts following a financially distressed firm. The yearly intercept variables (not shown in Table 10) are negative for all years.

The Panel A results for equation 3.7b show that the coefficient on the Ohlson probability variable is positive and insignificant, and this suggests that analyst forecast errors increase

(accuracy decreases) slightly when firms are more likely to be in financial distress. The results for equation 3.7c show that the coefficient on the DISTRESS variable is positive, and this provides weak evidence that analyst forecast errors increase (accuracy decreases) for firms identified as being financially distressed using both Altman’s (1968) and Begley’s (1996) models.

21 The coefficients not specifically mentioned in the preceding discussion for tables 7-9 are insignificant.

140

The results in Panel B of Table 10 provide the ex post analysis for the analyst forecast accuracy models. The coefficients on D1 and D2 in equation 3.7d are negative and significant, and this suggests that analyst forecast accuracy improves for healthy firms and firms identified as distressed using the Altman and Begley models that avoid filing for bankruptcy within three years of restructuring. The results for the regression of equation 3.7e demonstrate that the coefficient on the bankruptcy dummy variable is positive and highly significant, which suggests that analyst forecast errors increase and accuracy declines for firms filing for bankruptcy within three years of restructuring. Overall, the results reported in Table 10 tend to support H3b.

5.5.4 Analyst Forecast Bias

Table 11 presents the results of estimating equation 3.8, which uses actual analyst forecast errors to examine analyst forecast bias. The coefficient on the restructuring indicator variable is negative and significant for equation 3.8b, and this provides evidence that analyst forecasts are more upwardly biased after a restructuring charge than in other periods for the same firms when the likelihood of distress is greater. This also suggests that analysts fail to react fully to financial statement information. The coefficient on returns is positive for all versions of equation 3.8 in both Panels A and B, which suggests less upward analyst forecast bias as returns increase. The coefficient on the analyst following variable (LNUMB) is positive and significant for equation 3.8a, which suggests that there may be less upward bias in the analysts’ forecast when the analysts’ following is large.

The coefficient on the OHLSON_PROB variable is positive and moderately significant for equation 3.8b, and the coefficient on DISTRESS is also positive and significant for equation

3.8c. This suggests that analyst forecast bias increases for companies identified as financially distressed. For equation 3.8d, D1 and D2 have negative and significant coefficients, suggesting

141

that analyst forecast bias decreases for healthy firms and distressed firms that are able to avoid filing for bankruptcy for three years after restructuring. For equation 3.8d (3.8e), the coefficient on D3 (BANKRUPT) is insignificant. The results in Table 11 provide weak evidence that there is an increase in analyst forecast bias subsequent to a restructuring charge, consistent with the findings reported by Chaney et al. (1999). This indicates that analysts do not benefit from the information provided by restructuring charges, and these charges may cause analysts to be more biased and have decreased accuracy. The results also demonstrate that, contrary to expectations, analyst forecasts are more biased for healthy firms and distressed firms that avoid filing for bankruptcy and less biased for distressed firms that ultimately file for bankruptcy within three years of restructuring. Overall, the results reported in Table 11 tend to support H3c.

Additional analyses are reported in Tables 12 and 13. The five variations on equation 3.8 are presented separately for firms with positive and negative actual analyst forecast errors. The results where the analysts’ forecast bias is positive are presented in Table 12. For all five versions of equation 3.8 in Panels A and B of Table 12, the coefficient on the restructuring variable is negative and significant, and this suggests that analyst forecasts are more optimistic after a restructuring charge than in other periods for the same firms. The coefficients on the number of analysts following a firm and returns are negative and significant for equation 3.8 when analysts’ forecast errors are positive. This suggests less optimistic analyst forecasts when analyst following is large and when returns are high. The signs for the yearly intercepts (not reported) are mixed.

The coefficient on the DISTRESS variable in equation 3.8c is positive and significant, indicating that analyst forecast optimism increases as the likelihood of distress increases. This result tends to support H3c. The coefficient on the Ohlson probability variable is insignificant for equation 3.8b. The coefficients on D1 and D2 from equation 3.7d are negative and significant in

142

Panel B of Table 12, and this suggests that analysts are less optimistic for firms that do not file for bankruptcy within three years of restructuring. The positive and significant coefficient on the

BANKRUPT dummy variable in equation 3.7e suggests that there is more analyst optimism for firms filing for bankruptcy within three years of restructuring. Overall, the model including the three dummy variables has the greatest predictive power of the five variations of equation 3.8 in

Table 12.

The results when there is negative analyst bias are presented in Table 13. The coefficients for the returns variable are positive and significant for all five versions of equation

3.8 when there is negative bias, and this suggest that analysts are more pessimistic for firms with high levels of returns. The coefficients for the analyst following variable are positive and significant for equations 3.8a, 3.8b, and 3.8c, which provides weak evidence that analysts are more pessimistic for firms with a large analyst following. The coefficients on the DISTRESS variable and the OHLSON_PROB variable are both positive and significant when there is negative bias, and this indicates that analysts are more pessimistic for firms with a greater likelihood of financial distress. The coefficient on the BANKRUPT dummy variable in equation

3.8e is negative and significant, and this indicates that analyst forecasts are less pessimistic for firms that file for bankruptcy within three years of restructuring. The coefficients on the yearly intercepts (not reported) have mixed signs and levels of significance.

Robustness Tests

To determine whether the significant results in this essay are caused by model specification problems, tests are run on the full equation to detect problems with autocorrelation, multicollinearity, and heteroscedasticity. An examination of the Durbin-Watson statistics for the full models indicates that there is no autocorrelation of the residuals. The Durbin-Watson

143

statistics are in the appropriate range in all samples. Several tests are run to uncover potential problems with multicollinearity. First, there does not appear to be a major problem with multicollinearity because there is not a high adjusted R-square with few significant t-statistics, and there are no correlations that are higher than 0.5. The variance inflation factors (VIFs) are also examined to test for multicollinearity problems, and the values show that there is no indication for multicollinearity problem because the VIF values are less than five in all cases.

The significance of the White’s Chi-square statistics indicated a problem with heteroscedasticity, so White adjusted t-statistics are reported for all of the models in Essay 3.

5.6 Conclusions

The results of this essay provide several important conclusions. First, the results demonstrate that, after a restructuring charge announcement, analysts actually revise their forecasts downward. The results of this essay also suggest a leveling effect or a small increase in analysts’ forecasts over longer forecast horizons. By examining analysts’ forecast errors in the year after a restructuring charge, the results of this essay demonstrate that analysts’ accuracy declines and that analysts are still optimistically biased after a restructuring charge. Also, the results of this essay suggest that analyst forecast errors tend to increase for non-distressed firms and for distressed firms that avoid filing for bankruptcy. The results provide fairly strong support for the hypotheses. The findings demonstrate that analysts are optimistically biased for distressed, bankrupt firms, and this is demonstrated by the fact that the coefficients on the financial health variables (OHLSON_PROB and DISTRESS) included in equation 3.8 are both positive and highly significant. This also suggests that optimistic analyst forecast bias increases as the financial health of a firm declines. Overall, the results presented in this essay tend to support H3a, H3b, and H3c.

144

Several limitations on the results of this essay should be noted. First of all, there may be other explanations for the changes in analysts’ forecasts other than just the occurrence of the restructuring charge. There may be other confounding events that have not been considered in this essay. Also, the underlying conditions that cause a firm to decide to restructure may lead to some of the increases in bias and declines in accuracy. Finally, the empirical results are limited by the estimation of health and distress using three models from the extant literature, Altman

(1968), Ohlson (1980), and Begley et al. (1996). The use of a different procedure to estimate the financial condition of each firm may produce slightly different results.

BIBLIOGRAPHY

Abarbanell, J. S. and R. Lehavy. 2003. Biased forecasts or biased earnings? The role of reported earnings in explaining apparent bias and over/underreaction in analysts' earnings forecasts. Journal of Accounting and Economics, Boston Conference October 2002.

Alford, A. W., and P. G. Berger. 1999. A simultaneous equations analysis of forecast accuracy, analyst following, and trading volume. Journal of Accounting, Auditing & Finance 14 (3): 219-240.

Ali, A., A. Klein, and J. Rosenfeld. 1992. Analysts’ use of information about permanent and transitory earnings components in forecasting annual EPS. The Accounting Review 67: 183-198.

Altman, E. I. 1968. Financial ratios, discriminant analysis, and the prediction of corporate bankruptcy. Journal of Finance 23 (4): 589-609.

Asquith, P., M.B. Mikhail, and A. S. Au. 2004. Information content of equity analyst reports. MIT Sloan Working Paper 4264-02, Boston, MA.

Athanassakos, G., and M. Kalimipalli. 2003. Analyst forecast dispersion and future stock return volatility. Quarterly Journal of Business and Economics 42 (1/2): 57-78.

Barniv, R., M. J. Myring, and W. B. Thomas. 2005. The association between the legal and financial reporting environments and forecast performance of individual analysts. Contemporary Accounting Research 22 (4): 727-758.

Barron, O. E., D. Byard, and O. Kim. 2002. Changes in analysts’ information around earnings announcements. The Accounting Review 77 (4): 821-846.

Barth, M. E., and A. P. Hutton. 2003. Analyst earnings forecast revisions and the pricing of accruals. Tuck School of Business Working paper 03-02; Stanford University, GSB Research Paper No. 1693R, Boston, MA.

Begley, J., J. Ming, and S. Watts. 1996. Bankruptcy classification errors in the 1980s: An empirical analysis of Altman’s and Ohlson’s models. Review of Accounting Studies 1: 267-284.

Bonner, S. E., B. R. Walther, and S. M. Young. 2003. Sophistication-related differences in investors’ models of the relative accuracy of analysts’ forecast revisions. The Accounting Review 78 (3): 679-706.

Chaney, P. K., C. E. Hogan, and D. C. Jeter. 1999. The effect of reporting restructuring charges on analysts’ forecast revisions and errors. Journal of Accounting and Economics 27: 261-284.

145 146

Chen, Q., J. Francis, and W. Jiang. 2005. Investor learning about analyst predictive ability. Journal of Accounting and Economics 39: 3-24.

Chiang, H. 2005. Analysts’ financial forecast accuracy and information transparency. Journal of American Academy of Business 7 (2): 164-167.

Ciccone, S. J. 2001. Analyst Forecast Properties, Financial Distress, and Business Risk. Working Paper, University of New Hampshire, Durham, NH.

Claus, J. J. 2000. Individual analyst forecast accuracy: An international analysis. Dissertation. Columbia University, New York City, NY.

Clement, M. B., L. L. Koonce, and T. J. Lopez. 2006. The roles of task-specific experience and innate ability in understanding analyst performance. Working paper, University of Texas at Austin, Austin, TX.

Clement, M. B., and S. Y. Tse. 2003. Do investors respond to analysts’ forecast revisions as if forecast accuracy is all that matters? The Accounting Review 78 (1): 227-249.

Cowen, A., B. Groysberg, and P. M. Healy. 2003. Which types of analyst firms make more optimistic forecasts? Harvard NOM Working paper 03-46, Harvard University, Boston, MA.

Dechow, P. M., A. P. Hutton, and R. G. Sloan. 2000. The relation between analysts’ forecasts of long-term earnings growth and stock price performance following equity offerings. Contemporary Accounting Research 17 (1): 1-32.

Duru, A., and D. M. Reeb. 2002. International diversification and analysts’ forecast accuracy and bias. The Accounting Review 77 (2): 415-433.

Eames, M. J., and J. Sepe. 2005. The valuation of special items. Journal of Applied Business Research 21 (3): 61-70.

Easton, P. D., and S. J. Monahan. 2005. An evaluation of accounting-based measures of expected returns. The Accounting Review 80 (2): 501-538.

Elgers, P. T., M. H. Lo, and R. J. Pfeiffer, Jr. 2001. Delayed price adjustments to financial analysts’ forecasts of annual earnings. The Accounting Review 76 (4): 613-632.

Ertimur, Y., J. Sunder, and S. V. 2006. Measure for measure: The relation between forecast accuracy and recommendation profitability of analysts. Working paper, Stanford University, Stanford, CA.

Fan, D. K., R. W. So, and J. J. Yeh. 2006. Analyst earnings forecasts for publicly traded insurance companies. Review of Quantitative Finance and Accounting 26: 105-136.

Friesen, G. C., and P. A. Weller. 2005. Quantifying Cognitive Biases in Analyst Earnings Forecasts. Working Paper, University of Nebraska-Lincoln, Lincoln, NE.

147

Gilson, S. C. 1998. Course overview: Creating value through corporate restructuring. Working Paper, Harvard University, Boston, MA.

Grice, J. S., and M. T. Dugan. 2001. The limitations of bankruptcy prediction models: Some cautions for the researcher. Review of Quantitative Finance and Accounting 17: 151- 166.

Gu, Z., and J. S. Wu. 2003. Earnings skewness and analyst forecast bias. Journal of Accounting and Economics 35: 5-29.

Han, B. H., and D. Manry, and W. Shaw. 2001. Improving the precision of analysts’ earnings forecasts by adjusting for predictable bias. Review of Quantitative Finance and Accounting 17: 81-98.

Harjoto, M. A., and J. Zaima. 2005. Conflict in Whispers and Analyst Forecasts: Which One Should Be Your Guide? Working Paper, San Jose State University, San Jose, CA.

Hayn, C. 1995. The information content of losses. Journal of Accounting and Economics 20 (2): 125-153.

Hogan, C. E., and D. C. Jeter. 1998. The information content of restructuring charges: A contextual analysis. Working paper. Vanderbilt University.

Hollie, D., J. Livnat, and B. Segal. 2005. Oops, Our earnings were indeed preliminary. Journal of Portfolio Management 31 (2): 94-104.

Huai, Z. 2000. How rational is the stock market towards properties of analyst consensus forecasts? Ph.D. dissertation. Columbia University.

Irvine, P. J. 2004. Analysts’ forecasts and brokerage firm trading. The Accounting Review 79 (1): 125-149.

Ivkovic, Z. and N. Jegadeesh. 2002. The timing and value of forecast and recommendation revisions: Do analysts receive early peek at good news? Working paper, AFA 2004 San Diego Meetings, San Diego, CA.

Kini, O., S. L. Mian, M. Rebello, and A. Venkateswaran. 2005. Core Specialization, Forecast Accuracy, and the Market for International Analysts, Working Paper, Georgia State University, Atlanta, GA.

Kwon, S. S. 2002. Financial analysts’ forecast accuracy and dispersion: High-tech versus low- tech stocks. Review of Quantitative Finance and Accounting 19: 65-91.

Lacina, M. J., and K. E. Karim. 2004. Tests of market reaction and analysts’ forecast revisions to the disclosure of improved management earnings expectations: A case of concurrent bad news management earnings forecasts. Review of Quantitative Finance and Accounting 23: 123-148.

148

Lang, M. H., and R. J. Lundholm. 1996. Corporate disclosure policy and analyst behavior. The Accounting Review 71 (4): 467-492.

Lim, T. K. 2001. Rationality and analysts’ forecast bias. The Journal of Finance 56 (1): 369- 385.

______, and H. C. Kong. 2004. New evidence on price impact of analyst forecast revisions. International Review of Financial Analysts 13: 161-190.

Lin, B., and R. Yang. 2006. The effect of repeat restructuring charges on analysts’ forecast revisions and accuracy. Review of Quantitative Finance and Accounting 27 (3): 267- 283.

Liu, J., and W. Su. 2005. Forecasting analysts’ forecast errors. Working Paper, UCLA, Los Angeles, CA.

Lys, T., and L. G. Soo. 1995. Analysts’ forecast precision as a response to competition. Journal of Accounting, Auditing & Finance 10 (4): 751-765.

McKibben, W. 1972. Econometric forecasting of common stock investment returns: A new methodology using fundamental operating data. Journal of Finance 27 (May): 371-380.

Mest, D. P., and E. Plummer. 2000. Revisions in analysts’ earnings forecasts: Evidence of non- linear adaptive expectations. Journal of Forecasting 19: 467-484.

_____ and _____. 2003. Analysts’ rationality and forecast bias: Evidence from sales forecasts. Review of Quantitative Finance and Accounting 21 (2): 103-122.

Ohlson, J. A. 1980. Financial ratios and the probabilistic prediction of bankruptcy. Journal of Accounting Research 18 (1): 109-131.

Park, C. W., and E. K. Stice. 2000. Analyst forecasting ability and the stock price reaction to forecast revisions. Review of Accounting Studies 5: 259-272.

Pinello, A. S. 2005. How Do Individual Investors Use Analyst Forecasts? The Implications for Investor Reaction to Forecast Revisions and Earnings Surprises. AAA 2006 Financial Accounting and Reporting Section (FARS) Meeting, Working Paper, Georgia State University, Atlanta, GA.

Scherbina, A. D. 2004. Analyst disagreement, forecast bias and stock returns. Working paper, Harvard Business School Publishing, Forthcoming.

Shane, P., and P. Brous. 2001. Investor and (Value Line) analyst underreaction to information about future earnings: The corrective role of non-earnings-surprise information. Journal of Accounting Research 39 (2): 387-404.

149

Stevens, D. E., and A. W. Williams. 2004. Inefficiency in earnings forecasts: Experimental evidence of reactions to positive vs. negative information. Experimental Economics 7 (1): 75-92.

Yan, A. 2006. Analyst Sentiment around Takeover Announcements, Working Paper, Fordham University, New York, NY.

150

Table 1: Final Sample Determination

Panel A: Firms for Essay 3 (Analyst Forecast Error/Bias Portion of Essay)

Number of Firms

Sample firms from 1993 through 2003 3,867

Complete Compustat data not available (1,627)

Complete IBES data not available (including AFEt+1 and AFEt-1) (1,633)

Observations lost through trimming of outliers (38)

Final number of firms included in this essay 569

Panel A: Firm-Year Observations for Essay 3 (Analyst Forecast Error/Bias Portion of Essay)a Number of Firm-Year Observations

Sample firm-year observations from 1993 through 2003 3,867

Complete Compustat data not available (1,627)

Complete IBES data not available (1,030)

Observations lost through trimming of the outliers (38)

Final number of firm-year observations included in this essay 1,172

a The final sample of 1,172 firm-year observations for the forecast revision portion of this essay includes data for years t, t+1, and t-1 for the 569 firms announcing restructuring efforts only once during the sample period (1993-2003). Several firms are not included in the final sample because of the need to use three consecutive years for each company.

151

TABLE 1: Final Sample Determination (Continued) Panel C: Firm-Quarter Observations for Essay 3 (Forecast Revision Portion of Essay)

Number of Firm-Quarter Observations

Sample firm-quarter observations from 1993 through 2003 134,464

Complete Compustat data not available (59,555)

Complete IBES data not available (40,408)

Observations lost through trimming outliers (690)

Final number of firm-quarter observations included in this essay 33,811

152

Table 2 Firms in Sample Split by SIC Division

SIC Division Total Firms Firms as % Agriculture, Forestry, and Fishing 1 0.2 Mining 10 1.8 Construction 118 20.7 Manufacturing 191 33.6 Transportation, Communications, Electric, and Gas 49 8.6 Wholesale Trade 69 12.1 Retail Trade 55 9.7 Finance, Insurance, and Real Estate 58 10.2 Services 16 2.8 Public Administration 2 0.3 Total 569

153

Table 3 a Descriptive Statistics for the Analyst Forecast Revision Portion of the Essay b Panel A: Full Sample of Firm-Quarter Observations Variable N Mean Median Std. Dev. Minimum Maximum h=1 AFREVi,j 33,811 0.002 <0.001 0.02 -0.10 0.16 h=2 AFREVi,j 30,354 0.001 <0.001 0.02 -0.08 0.10 h=3 AFREVi,j 8,402 0.001 <0.001 0.02 -0.09 0.11 h=5 AFREVi,j 1,879 0.001 <0.001 0.01 -0.06 0.11 UEWO 33,811 -0.40 -0.09 1.27 -10.84 3.82 LOSS * UEWO 33,811 -0.36 0 1.14 -10.83 3.80 LOSS 33,811 0.34 0 0.47 0 1 REST 33,811 0.04 0 0.19 0 1 LOSS * REST 33,811 0.01 0 0.10 0 1

Panel B: Firm-Quarter Observations Classified the Same by the Altman and Begley Models Variable N Mean Median Std. Dev. Minimum Maximum h=1 AFREVi,j 25,346 0.002 <0.001 0.02 -0.10 0.160 h=2 AFREVi,j 22,702 0.002 <0.001 0.01 -0.08 0.953 h=3 AFREVi,j 6,172 0.001 <0.001 0.01 -0.09 0.107 h=5 AFREVi,j 1,377 0.001 <0.001 0.01 -0.06 0.108 UEWO 25,346 -0.35 -0.06 1.22 -10.83 3.82 LOSS * UEWO 25,346 -0.32 0 1.10 -10.83 3.80 LOSS 25,346 0.23 0 0.42 0 1 REST 25,346 0.02 0 0.15 0 1 LOSS * REST 25,346 0.01 0 0.08 0 1

a h AFREVi,j is the revision in the mean forecasts for h-year ahead earnings (h = 1, 2, or 3) or earnings growth (h = 5) subsequent to a restructuring charge and/or earnings announcement in month i of quarter j. h h h h AFREVi,j = AFi+1 – AFi-1 , where AFi-1 is the mean forecast of h-year ahead earnings (h = 1, 2, or 3) or earnings growth (h = 5) in the month i-1, the months prior to the announcement of the restructuring charge, h deflated by price at the beginning of the period i-1. AFi+1 is the mean forecast of h-year ahead earnings (h = 1, 2, or 3) or earnings growth (h = 5) in month i+1, the month subsequent to the announcement of the restructuring charge, deflated by price at the beginning of the period i-1. LOSSj equals 1 if income before restructuring charges and extraordinary items in quarter j is less than zero, and 0 otherwise. UEWOj is unexpected earnings before restructuring charges in quarter j. RESTj equals 1 if the firm announced a restructuring charge in quarter j, and 0 otherwise. b The data have been trimmed to delete the outliers.

154

Table 4 a Descriptive Statistics for the Analyst Forecast Accuracy and Bias Portion of the Essay Full Sample of Firm-Year Observationsb Variable N Mean Median Std. Dev. Minimum Maximum c ⏐AFEt+1⏐ 1,172 0.58 0.005 5.92 0 100.0

⏐AFEt-1⏐ 1,172 0.26 0.006 3.26 0 55.86

⏐RETt,t+1⏐ 1,138 0.65 0.40 1 0 11.00

RESTt 1,172 0.73 1 0.44 0 1

LNUMBt+1 1,172 1.82 1.79 2.00 0.30 4.40

RECESSt+1 1,172 0.10 0 0.30 0 1

AFEt+1 1,172 -0.34 0 5.52 -100.0 39.75

AFEt-1 1,172 -0.01 -0.001 4.40 -55.86 87.00

RETt,t+1 1,138 0.18 0.10 1.95 -37.00 11.00

a ⏐AFEt+1⏐ is the absolute value of the analyst forecast error for year t + 1, the year after the restructuring charge. AFEt+1 = (EPSt+1 – AFt+1) / Pt+1, where EPSt+1 is the actual earnings per share for year t+1, AFt+1 is the mean earnings forecast for year t+1 measured in the month preceding the earnings announcement for

year t+1, and Pt+1 is the market price of stock at the beginning of year t+1. ⏐AFEt-1⏐ is the absolute value of the analyst forecast error for year t – 1, the year before the restructuring charge. RESTt equals 1 if the firm announced a restructuring charge in year t, and 0 otherwise. ⏐RETt,t+1⏐ is the absolute value of the market return from the beginning of period t to the end of year t + 1. Ct equals 1 if the observation is from year t, and 0 otherwise. LNUMBt+1 is the log of the number of analysts forecasting earnings for a given firm in year t + 1. RECESSt+1 equals 1 if the observation occurred during a time of recession as defined by the National Bureau of Economic Research, and 0 otherwise. b The data have been trimmed to delete the outliers. c The 1,138 firm-year observations in Panel A have all data available for the variables in equations 3.7 and 3.8.

155

Table 5 Panel A: Univariate data on forecast errors for the year before and the year after restructuring charges: Firm Observationsa t-statistic Year t-1 Year t+1 (p-value) Mean absolute value of forecast error (⏐AFE⏐): Accuracy 0.032 0.034 0.204 Std Deviation 0.121 0.115 (0.838) n 545 firms 545 firms

Mean actual forecast error (AFE): Bias -0.022 -0.025 -0.306 Std Deviation 0.124 0.118 (0.760) n 545 firms 545 firms

Panel B: Univariate data on forecast errors for the year before and the year after restructuring charges: Firm-Year Observations t-statistic Year t-1 Year t+1 (p-value) Mean absolute value of forecast error (⏐AFE⏐): Accuracy 0.065 0.036 2.565 Std Deviation 0.267 0.127 (0.010***) 1,172 firm-year 1,172 firm-year n observations observations

Mean actual forecast error (AFE): Bias -0.046 -0.026 -1.771 Std Deviation 0.266 0.129 (0.077*) 1,172 firm-year 1,172 firm-year n observations observations

a ⏐AFEt+1⏐ is the absolute value of the analyst forecast error for year t + 1, the year after the restructuring charge. AFEt+1 = (EPSt+1 – AFt+1) / Pt+1, where EPSt+1 is the actual earnings per share for year t+1, AFt+1 is the mean earnings forecast for year t+1 measured in the month preceding the earnings announcement for

year t+1, and Pt+1 is the market price of stock at the beginning of year t+1. ⏐AFEt-1⏐ is the absolute value of the analyst forecast error for year t – 1, the year before the restructuring charge. The data has been trimmed to delete the top and bottom one percent of the observations. The sample in this table includes only observations classified into the same distress classification (0 or 1) by the Altman (1968) and Begley (1996) models. ***Significant at the 0.01 level; **Significant at the 0.05 level; *Significant at the 0.10 level

156

Table 6: Regressions of Equation (3.5) where h=1a with Firm-Quarter Observationsb c Panel A: Ex Ante Analysis h (3.5a) AFREVi,j = B0 + B1 LOSSj + B2 UEWOj + B3 RESTj + B4 LOSS * RESTj + B5 LOSS * UEWOj + u, h (3.5b) AFREVi,j = B0 + B1 LOSSj + B2 UEWOj + B3 RESTj + B4 LOSS * RESTj + B5 LOSS * UEWOj + B6 Ohlson_Prob + u, h (3.5c) AFREVi,j = B0 + B1 LOSSj + B2 UEWOj + B3 RESTj + B4 LOSS * RESTj + B5 LOSS * UEWOj + B6 DISTRESS + u Variable Name (Expected Sign) Equation (3.5a) Equation (3.5b) Equation (3.5c)

LOSSj (-) -0.001 -0.008 -0.004 (t-statistic) (-2.01**) (-8.86***) (-4.17***)

UEWOj (+) 0.003 0.002 0.002 (9.63***) (5.57***) (8.40***)

RESTj (-) -0.001 -0.002 -0.001 (-1.26) (-1.89*) (-1.39)

LOSS * RESTj (+/-) -0.006 -0.005 -0.005 (-1.68*) (-1.30) (-1.47)

LOSS * UEWOj (+/-) -0.005 -0.004 -0.005 (-4.33***) (-3.82***) (-4.18***) OHLSON_PROB (+/-) 0.001 (10.41***) DISTRESS (+/-) 0.003 (4.08***) Adjusted R2 0.011 0.021 0.013 Number of Firm-Quarter Observations 25,346 25,346 25,346 F-value 23.80*** 38.11*** 22.64***

a h AFREVi,j is the revision in the mean forecasts for h-year ahead earnings (h = 1, 2, or 3) or earnings growth (h = 5) after a restructuring charge and/or earnings announcement in month i of quarter j. h h h h AFREVi,j = AFi+1 – AFi-1 , where AFi-1 is the mean forecast of h-year ahead earnings or earnings growth in the month i-1, the months prior to the announcement of the restructuring charge, deflated by price at the h beginning of the period i-1. AFi+1 is the mean forecast of h-year ahead earnings or earnings growth in month i+1, the month after the announcement of the restructuring charge, deflated by price at the beginning of period i-1. LOSSj equals 1 if income before restructuring charges and extraordinary items in quarter j is less than 0, and 0 otherwise. UEWOj is unexpected earnings before restructuring charges in quarter j. RESTj equals 1 if the firm announced a restructuring charge in quarter j, and 0 otherwise. OHLSON_PROB is the probability of financial distress obtained from running Ohlson’s model (equation (3.3)). DISTRESS equals 1 for firms classified by both Altman and Begley as distressed, and 0 otherwise. Corrections for heteroscedasticity, multicollinearity, and autocorrelation are made whenever problems are detected. b The data has been trimmed to delete the outliers. The sample in this table includes only observations classified into the same distress classification (0 or 1) by the Altman and Begley models. The coefficients on the yearly intercepts are not reported. c The White adjusted t-statistics are shown in parentheses. ***Significant at the 0.01 level; **Significant at the 0.05 level; *Significant at the 0.10 level

157

Table 6: Regressions of Equation (3.5) where h=1a with Firm-Quarter Observationsb Panel B: Ex Post Analysisc h (3.5d) AFREVi,j = B0 + B1 LOSSj + B2 UEWOj + B3 RESTj + B4 LOSS * RESTj + B5 LOSS * UEWOj + B6 D1 + B7 D2 + B8 D3 + u, h (3.5e) AFREVi,j = B0 + B1 LOSSj + B2 UEWOj + B3 RESTj + B4 LOSS * RESTj + B5 LOSS * UEWOj + B6 BANKRUPT + u Variable Name (Expected Sign) Equation (3.5d) Equation (3.5e)

LOSSj (-) -0.004 -0.003 (t-statistic) (-9.06***) (-8.29***)

UEWOj (+) -0.000 -0.000 (-1.45) (-1.30)

RESTj (-) -0.001 -0.001 (-1.06) (-1.02)

LOSS * RESTj (+/-) -0.005 -0.005 (-2.56***) (-2.61***)

LOSS * UEWOj (+/-) 0.000 0.000 (0.39) (0.32) D1 (?) 0.001 (0.85) D2 (+) 0.002 (4.16***) D3 (-) -0.003 (-3.42***) BANKRUPT (-) -0.002 (-3.03***) Adjusted R2 0.005 0.005 Number of Firm-Quarter Observations 25,346 25,346 F-value 20.29*** 20.77*** a h AFREVi,j is the revision in the mean forecasts for h-year ahead earnings (h = 1, 2, or 3) or earnings growth (h = 5) after a restructuring charge and/or earnings announcement in month i of quarter j. h h h h AFREVi,j = AFi+1 – AFi-1 , where AFi-1 is the mean forecast of h-year ahead earnings or earnings growth in the month i-1, the months prior to the announcement of the restructuring charge, deflated by price at the h beginning of the period i-1. AFi+1 is the mean forecast of h-year ahead earnings or earnings growth in month i+1, the month after the announcement of the restructuring charge, deflated by price at the beginning of period i-1. LOSSj equals 1 if income before restructuring charges and extraordinary items in quarter j is less than 0, and 0 otherwise. UEWOj is unexpected earnings before restructuring charges in quarter j. RESTj equals 1 if the firm announced a restructuring charge in quarter j, and 0 otherwise. D1 equals 1 for non-distressed firms, and 0 otherwise. D2 equals 1 for distressed firms that do not file for bankruptcy for at least three years after restructuring, and 0 otherwise. D3 equals 1 for firms filing for bankruptcy within three years of restructuring, and 0 otherwise. BANKRUPT equals 1 for firms filing for bankruptcy within three years of restructuring, and 0 otherwise. Corrections for heteroscedasticity, multicollinearity, and autocorrelation are made whenever problems are detected. b The data has been trimmed to delete the outliers. The sample in this table includes only observations classified into the same distress classification (0 or 1) by the Altman and Begley models. The coefficients on the yearly intercepts are not reported. c The White adjusted t-statistics are shown in parentheses. ***Significant at the 0.01 level; **Significant at the 0.05 level; *Significant at the 0.10 level

158

a b Table 7: Regressions of Equation (3.5) where h=2 with Firm-Quarter Observations h (3.5a) AFREVi,j = B0 + B1 LOSSj + B2 UEWOj + B3 RESTj + B4 LOSS * RESTj + B5 LOSS * UEWOj + u, h (3.5b) AFREVi,j = B0 + B1 LOSSj + B2 UEWOj + B3 RESTj + B4 LOSS * RESTj + B5 LOSS * UEWOj + B6 Ohlson_Prob + u, h (3.5c) AFREVi,j = B0 + B1 LOSSj + B2 UEWOj + B3 RESTj + B4 LOSS * RESTj + B5 LOSS * UEWOj + B6 DISTRESS + u, Variable Name (Expected Sign)c Equation (3.5a) Equation (3.5b) Equation (3.5c)

LOSSj (-) 0.000 -0.005 0.000 (t-statistic) (0.48) (-5.65***) (0.20)

UEWOj (+) 0.003 0.002 0.003 (11.80***) (8.23***) (11.39***)

RESTj (-) 0.000 -0.000 0.000 (0.38) (-0.12) (0.37)

LOSS * RESTj (+/-) 0.003 0.004 0.003 (1.02) (1.36) (1.03)

LOSS * UEWOj (+/-) -0.004 -0.003 -0.004 (-3.50***) (-3.09***) (-3.49***) OHLSON_PROB (+/-) 0.006 (8.50***) DISTRESS (+/-) 0.000 (0.26) Adjusted R2 0.015 0.022 0.015 Number of Firm-Quarter Observations 22,702 22,702 22,702 F-value 29.29*** 36.64*** 24.42***

a h AFREVi,j is the revision in the mean forecasts for h-year ahead earnings (h = 1, 2, or 3) or earnings growth (h = 5) after a restructuring charge and/or earnings announcement in month i of quarter j. h h h h AFREVi,j = AFi+1 – AFi-1 , where AFi-1 is the mean forecast of h-year ahead earnings or earnings growth in the month i-1, the months prior to the announcement of the restructuring charge, deflated by price at the h beginning of the period i-1. AFi+1 is the mean forecast of h-year ahead earnings or earnings growth in month i+1, the month after the announcement of the restructuring charge, deflated by price at the beginning of period i-1. LOSSj equals 1 if income before restructuring charges and extraordinary items in quarter j is less than 0, and 0 otherwise. UEWOj is unexpected earnings before restructuring charges in quarter j. RESTj equals 1 if the firm announced a restructuring charge in quarter j, and 0 otherwise. OHLSON_PROB is the probability of financial distress obtained from running Ohlson’s model (equation (3.3)). DISTRESS equals 1 for firms classified by both Altman and Begley as distressed, and 0 otherwise. Corrections for heteroscedasticity, multicollinearity, and autocorrelation are made whenever problems are detected. The coefficients on the yearly intercepts are not reported. b The data has been trimmed to delete the outliers. The sample in this table includes only observations classified into the same distress classification (0 or 1) by the Altman and Begley models. c The White adjusted t-statistics are shown in parentheses. ***Significant at the 0.01 level; **Significant at the 0.05 level; *Significant at the 0.10 level

159

a b Table 8: Regressions of Equation (3.5) where h=3 with Firm-Quarter Observations h (3.5a) AFREVi,j = B0 + B1 LOSSj + B2 UEWOj + B3 RESTj + B4 LOSS * RESTj + B5 LOSS * UEWOj + u, h (3.5b) AFREVi,j = B0 + B1 LOSSj + B2 UEWOj + B3 RESTj + B4 LOSS * RESTj + B5 LOSS * UEWOj + B6 Ohlson_Prob + u, h (3.5c) AFREVi,j = B0 + B1 LOSSj + B2 UEWOj + B3 RESTj + B4 LOSS * RESTj + B5 LOSS * UEWOj + B6 DISTRESS + u, Variable Name (Expected Sign) c Equation (3.5a) Equation (3.5b) Equation (3.5c)

LOSSj (-) -0.001 -0.001 -0.001 (-1.56) (-1.10) (-0.89)

UEWOj (+) -0.000 -0.000 -0.000 (-0.15) (-0.14) (-0.09)

RESTj (+/-) -0.002 -0.002 -0.002 (-1.23) (-1.23) (-1.22)

LOSS * RESTj (+/-) -0.006 -0.006 -0.006 (-1.36) (-1.36) (-1.37)

LOSS * UEWOj (+/-) 0.000 0.000 0.000 (0.66) (0.67) (0.66) OHLSON_PROB (+/-) -0.000 (-0.13) DISTRESS (+/-) -0.001 (-0.57) Adjusted R2 0.007 0.006 0.006 Number of Firm-Quarter Observations 6,172 6,172 6,172 F-value 3.70*** 3.50*** 3.51***

a h AFREVi,j is the revision in the mean forecasts for h-year ahead earnings (h = 1, 2, or 3) or earnings growth (h = 5) after a restructuring charge and/or earnings announcement in month i of quarter j. h h h h AFREVi,j = AFi+1 – AFi-1 , where AFi-1 is the mean forecast of h-year ahead earnings or earnings growth in the month i-1, the months prior to the announcement of the restructuring charge, deflated by price at the h beginning of the period i-1. AFi+1 is the mean forecast of h-year ahead earnings or earnings growth in month i+1, the month after the announcement of the restructuring charge, deflated by price at the beginning of period i-1. LOSSj equals 1 if income before restructuring charges and extraordinary items in quarter j is less than 0, and 0 otherwise. UEWOj is unexpected earnings before restructuring charges in quarter j. RESTj equals 1 if the firm announced a restructuring charge in quarter j, and 0 otherwise. OHLSON_PROB is the probability of financial distress obtained from running Ohlson’s model (equation (3.3)). DISTRESS equals 1 for firms classified by both Altman and Begley as distressed, and 0 otherwise. Corrections for heteroscedasticity, multicollinearity, and autocorrelation are made whenever problems are detected. The coefficients on the yearly intercepts are not reported. b The data has been trimmed to delete the outliers. The sample in this table includes only observations classified into the same distress classification (0 or 1) by the Altman and Begley models. c The White adjusted t-statistics are shown in parentheses. ***Significant at the 0.01 level; **Significant at the 0.05 level; *Significant at the 0.10 level

160

a b Table 9: Regressions of Equation (3.5) where h=5 with Firm-Quarter Observations h (3.5a) AFREVi,j = B0 + B1 LOSSj + B2 UEWOj + B3 RESTj + B4 LOSS * RESTj + B5 LOSS * UEWOj + u, h (3.5b) AFREVi,j = B0 + B1 LOSSj + B2 UEWOj + B3 RESTj + B4 LOSS * RESTj + B5 LOSS * UEWOj + B6 Ohlson_Prob + u, h (3.5c) AFREVi,j = B0 + B1 LOSSj + B2 UEWOj + B3 RESTj + B4 LOSS * RESTj + B5 LOSS * UEWOj + B6 DISTRESS + u, Variable Name (Expected Sign) c Equation (3.5a) Equation (3.5b) Equation (3.5c)

LOSSj (-) 0.002 0.004 0.004 (t-statistic) (1.48) (1.77*) (1.76*)

UEWOj (+) 0.000 0.000 0.000 (0.29) (0.44) (0.50)

RESTj (+) 0.001 0.001 0.001 (0.18) (0.21) (0.28)

LOSS * RESTj (+/-) -0.019 -0.019 -0.019 (-1.89*) (-1.92*) (-1.94*)

LOSS * UEWOj (+/-) -0.000 -0.000 -0.000 (-0.48) (-0.43) (-0.57) OHLSON_PROB (+/-) -0.003 (-1.18) DISTRESS (+/-) -0.002 (-1.14) Adjusted R2 0.007 0.009 0.009 Number of Firm-Quarter Observations 1,377 1,377 1,377 F-value 1.68** 1.74** 1.76**

a h AFREVi,j is the revision in the mean forecasts for h-year ahead earnings (h = 1, 2, or 3) or earnings growth (h = 5) after a restructuring charge and/or earnings announcement in month i of quarter j. h h h h AFREVi,j = AFi+1 – AFi-1 , where AFi-1 is the mean forecast of h-year ahead earnings or earnings growth in the month i-1, the months prior to the announcement of the restructuring charge, deflated by price at the h beginning of the period i-1. AFi+1 is the mean forecast of h-year ahead earnings or earnings growth in month i+1, the month after the announcement of the restructuring charge, deflated by price at the beginning of period i-1. LOSSj equals 1 if income before restructuring charges and extraordinary items in quarter j is less than 0, and 0 otherwise. UEWOj is unexpected earnings before restructuring charges in quarter j. RESTj equals 1 if the firm announced a restructuring charge in quarter j, and 0 otherwise. OHLSON_PROB is the probability of financial distress obtained from running Ohlson’s model (equation (3.3)). DISTRESS equals 1 for firms classified by both Altman and Begley as distressed, and 0 otherwise. Corrections for heteroscedasticity, multicollinearity, and autocorrelation are made whenever problems are detected. The coefficients on the yearly intercepts are not reported. b The data has been trimmed to delete the outliers. The sample in this table includes only observations classified into the same distress classification (0 or 1) by the Altman and Begley models. c The White adjusted t-statistics are shown in parentheses. ***Significant at the 0.01 level; **Significant at the 0.05 level; *Significant at the 0.10 level

161

Table 10: Regressions of Equation (3.7) for All Firm-Year Observationsa: Accuracyb Panel A: Ex Ante Analysisc

(3.7a) ⏐AFEt+1⏐ = Σ γt Ct + γ1 ⏐AFEt-1⏐ + γ 2 RESTt + γ 3 ⏐RETt,t+1⏐ + γ4 LNUMBt+1 + γ5 RECESSt+1 + v, (3.7b) ⏐AFEt+1⏐ = Σ γt Ct + γ1 ⏐AFEt-1⏐ + γ 2 RESTt + γ 3 ⏐RETt,t+1⏐ + γ4 LNUMBt+1 + γ5 RECESSt+1 + γ6 OHLSON_PROB + v,

(3.7c) ⏐AFEt+1⏐ = Σ γt Ct + γ1 ⏐AFEt-1⏐ + γ 2 RESTt + γ 3 ⏐RETt,t+1⏐ + γ4 LNUMBt+1 + γ5 RECESSt+1 + γ6 DISTRESS + v, Variable Name (Expected Sign) Equation (3.7a) Equation (3.7b) Equation (3.7c)

⏐AFEt-1⏐ (+) 0.145 0.550 -0.052 (2.33**) (2.27**) (-1.61)

RESTt (+/-) -0.900 -1.164 -1.173 (-2.30**) (-1.70*) (-1.77*)

⏐RETt,t+1⏐ (+/-) 0.101 -0.117 -0.216 (0.58) (-0.71) (-1.14)

LNUMBt+1 (-) 0.784 0.409 -0.276 (7.68***) (1.66*) (-2.42**)

RECESSt+1 (+) 0.169 0.093 -0.015 (0.16) (0.65) (-0.13) OHLSON_PROB (+) 0.279 (0.69) DISTRESS (+) 0.395 (1.20) Adjusted R2 0.104 0.139 0.008 N 1,137 1,008 940 F-value 9.82*** 11.14*** 1.47

a ⏐AFEt+1⏐ is the absolute value of the analyst forecast error for year t + 1, the year after the restructuring charge. AFEt+1 = (EPSt+1 – AFt+1) / Pt+1, where EPSt+1 is the actual earnings per share for year t+1, AFt+1 is the mean earnings forecast for year t+1 measured in the month preceding the earnings announcement for

year t+1, and Pt+1 is the market price of stock at the beginning of year t+1. ⏐AFEt-1⏐ is the absolute value of the analyst forecast error for year t – 1, the year before the restructuring charge. RESTt equals 1 if the firm announced a restructuring charge in year t, and 0 otherwise. ⏐RETt,t+1⏐ is the absolute value of the market return from the beginning of period t to the end of year t + 1. Ct equals 1 if the observation is from year t, and 0 otherwise. LNUMBt+1 is the log of the number of analysts forecasting earnings for a given firm in year t + 1. RECESSt+1 equals 1 if the observation occurred during a time of recession as defined by the National Bureau of Economic Research, and 0 otherwise. OHLSON_PROB is the probability of financial distress obtained from running Ohlson’s regression equation (equation (3.3)). DISTRESS equals 1 for firms classified by both Altman and Begley as distressed, and 0 otherwise. Corrections for heteroscedasticity, multicollinearity, and autocorrelation are made whenever problems are detected. The coefficients on the yearly intercepts are not reported, but they are negative in all cases. b The data has been trimmed to delete the outliers. The sample in this table includes only observations classified into the same distress classification (0 or 1) by the Altman and Begley models. c The White adjusted t-statistics are shown in parentheses. ***Significant at the 0.01 level; **Significant at the 0.05 level; *Significant at the 0.10 level

162

Table 10: Regressions of Equation (3.7) for All Firm-Year Observationsa: Accuracyb Panel B: Ex Post Analysisc (3.7d) ⏐AFEt+1⏐ = Σ γt Ct + γ1 ⏐AFEt-1⏐ + γ 2 RESTt + γ 3 ⏐RETt,t+1⏐ + γ4 LNUMBt+1 + γ5 RECESSt+1 + γ6 D1 + γ7 D2 + γ8 D3 + v, (3.7e) ⏐AFEt+1⏐ = Σ γt Ct + γ1 ⏐AFEt-1⏐ + γ 2 RESTt + γ 3 ⏐RETt,t+1⏐ + γ4 LNUMBt+1 + γ5 RECESSt+1 + γ6 BANKRUPT + v, Variable Name (Expected Sign) Equation (3.7d) Equation (3.7e)

⏐AFEt-1⏐ (+) -0.041 0.138 (-0.26) (2.25**)

RESTt (+/-) -1.166 -0.787 (-2.81***) (-2.04**)

⏐RETt,t+1⏐ (+/-) -0.235 0.047 (-1.35) (0.27)

LNUMBt+1 (-) -0.338 0.812 (-1.76*) (8.05***)

RECESSt+1 (+) 0.242 0.416 (0.23) (0.40) D1 (-) -5.745 (-2.36**) D2 (+/-) -6.329 (-2.56**) D3 (+) -1.126 (-0.57) BANKRUPT (+) 5.315 (5.646***) Adjusted R2 0.018 0.128 N 940 1,138 F-value 1.96*** 11.45*** a ⏐AFEt+1⏐ is the absolute value of the analyst forecast error for year t + 1, the year after the restructuring charge. AFEt+1 = (EPSt+1 – AFt+1) / Pt+1, where EPSt+1 is the actual earnings per share for year t+1, AFt+1 is the mean earnings forecast for year t+1 measured in the month preceding the earnings announcement for

year t+1, and Pt+1 is the market price of stock at the beginning of year t+1. ⏐AFEt-1⏐ is the absolute value of the analyst forecast error for year t – 1, the year before the restructuring charge. RESTt equals 1 if the firm announced a restructuring charge in year t, and 0 otherwise. ⏐RETt,t+1⏐ is the absolute value of the market return from the beginning of period t to the end of year t + 1. Ct equals 1 if the observation is from year t, and 0 otherwise. LNUMBt+1 is the log of the number of analysts forecasting earnings for a given firm in year t + 1. RECESSt+1 equals 1 if the observation occurred during a time of recession as defined by the National Bureau of Economic Research, and 0 otherwise. D1 equals 1 for non-distressed firms, and 0 otherwise. D2 equals 1 for distressed firms that do not file for bankruptcy for at least three years after restructuring, and 0 otherwise. D3 equals 1 for firms filing for bankruptcy within three years of restructuring, and 0 otherwise. BANKRUPT equals 1 for firms filing for bankruptcy within three years of restructuring, and 0 otherwise. Corrections for heteroscedasticity, multicollinearity, and autocorrelation are made whenever problems are detected. The coefficients on the yearly intercepts are not reported. b The data has been trimmed to delete the outliers. The sample in this table includes only observations classified into the same distress classification (0 or 1) by the Altman and Begley models. c The White adjusted t-statistics are shown in parentheses. ***Significant at the 0.01 level; **Significant at the 0.05 level; *Significant at the 0.10 level

163

Table 11: Regressions of Equation (3.8) for All Firm-Year Observationsa: Biasb Panel A: Ex Ante Analysisc

(3.8a) AFEt+1 = Σ γt Ct + γ1 AFEt-1 + γ 2 RESTt + γ 3 RETt,t+1 + γ4 LNUMBt+1 + γ5 RECESSt+1 + v, (3.8b) AFEt+1 = Σ γt Ct + γ1 AFEt-1 + γ 2 RESTt + γ 3 RETt,t+1 + γ4 LNUMBt+1 + γ5 RECESSt+1 + γ6 OHLSON_PROB + v,

(3.8c) AFEt+1 = Σ γt Ct + γ1 AFEt-1 + γ 2 RESTt + γ 3 RETt,t+1 + γ4 LNUMBt+1 + γ5 RECESSt+1 + γ6 DISTRESS + v, Variable Name (Expected Sign) Equation (3.8a) Equation (3.8b) Equation (3.8c)

AFEt-1 (+/-) 0.002 -0.041 -0.044 (0.09) (-1.76*) (-1.26)

RESTt (+/-) 0.328 0.517 0.576 (0.54) (0.78) (0.83)

RETt,t+1 (+/-) 0.183 0.104 0.130 (1.41) (0.82) (0.96)

LNUMBt+1 (+) 0.148 0.196 0.219 (1.38) (1.41) (1.69*)

RECESSt+1 (+) 0.077 -0.008 0.056 (1.28) (-0.09) (0.47) OHLSON_PROB (+) 0.649 (1.79*) DISTRESS (+) 0.937 (2.40**) Adjusted R2 -0.003 0.012 0.003 N 1,138 1,008 940 F-value 0.76 1.00 1.20

a AFEt+1 is the analyst forecast error for year t + 1, the year after the restructuring charge. The year of the restructuring charge, t, is eliminated as in Chaney et al. (1999) because it is not clear whether or not analysts’ forecast for that period include any expectation of a restructuring charge. AFEt-1 is the analyst forecast error for year t – 1. RESTt equals 1 if the firm announced a restructuring charge in year t, and 0 otherwise. RETt,t+1 is the market return from the beginning of period t to the end of year t + 1. LNUMBt+1 is the log of the number of analysts forecasting earnings for a given firm in year t + 1. RECESSt+1 equals 1 if the observation occurred during a time of recession as defined by the National Bureau of Economic Research, and 0 otherwise. OHLSON_PROB is the probability of financial distress obtained from running Ohlson’s regression equation (equation (3.3)). DISTRESS equals 1 for firms classified by both Altman and Begley as distressed, and 0 otherwise. Corrections for heteroscedasticity, multicollinearity, and autocorrelation are made whenever problems are detected. The coefficients on the yearly intercepts are not reported, but they are positive in all cases. b The data has been trimmed to delete the outliers. The sample in this table includes only observations classified into the same distress classification (0 or 1) by the Altman and Begley models. The t-statistics for each variable are reported in parentheses on the next line. c The White adjusted t-statistics are shown in parentheses. ***Significant at the 0.01 level; **Significant at the 0.05 level; *Significant at the 0.10 level

164

Table 11: Regressions of Equation (3.8) for All Firm-Year Observationsa: Biasb Panel B: Ex Post Analysisc

(3.8d) AFEt+1 = Σ γt Ct + γ1 AFEt-1 + γ 2 RESTt + γ 3 RETt,t+1 + γ4 LNUMBt+1 + γ5 RECESSt+1 + γ6 D1 + γ7 D2 + γ8 D3 + v (3.8e) AFEt+1 = Σ γt Ct + γ1 AFEt-1 + γ 2 RESTt + γ 3 RETt,t+1 + γ4 LNUMBt+1 + γ5 RECESSt+1 + γ6 BANKRUPT + v Variable Name (Expected Sign) Equation (3.8d) Equation (3.8e)

AFEt-1 (+/-) -0.052 0.004 (-0.33) (0.15)

RESTt (+/-) 0.770 0.293 (1.85*) (0.49)

RETt,t+1 (+/-) 0.161 0.168 (1.10) (1.58)

LNUMBt+1 (+) 0.283 0.130 (1.48) (1.31)

RECESSt+1 (+) 0.384 0.036 (0.36) (0.27) D1 (-) -4.15 (-2.38**) D2 (+/-) -3.43 (-2.07**) D3 (+) 1.99 (0.44) BANKRUPT (+) -0.780 (-0.30) Adjusted R2 0.018 0.011 N 940 1,138 F-value 1.94** 1.20 a AFEt+1 is the analyst forecast error for year t + 1, the year after the restructuring charge. The year of the restructuring charge, t, is eliminated as in Chaney et al. (1999) because it is not clear whether or not analysts’ forecast for that period include any expectation of a restructuring charge. AFEt-1 is the analyst forecast error for year t – 1. RESTt equals 1 if the firm announced a restructuring charge in year t, and 0 otherwise. RETt,t+1 is the market return from the beginning of period t to the end of year t + 1. LNUMBt+1 is the log of the number of analysts forecasting earnings for a given firm in year t + 1. RECESSt+1 equals 1 if the observation occurred during a time of recession as defined by the National Bureau of Economic Research, and 0 otherwise. D1 equals 1 for non-distressed firms, and 0 otherwise. D2 equals 1 for distressed firms that do not file for bankruptcy for at least three years after restructuring, and 0 otherwise. D3 equals 1 for firms filing for bankruptcy within three years of restructuring, and 0 otherwise. BANKRUPT equals 1 for firms filing for bankruptcy within three years of restructuring, and 0 otherwise. Corrections for heteroscedasticity, multicollinearity, and autocorrelation are made whenever problems are detected. The coefficients on the yearly intercepts are not reported, but they are positive in all cases. b The data has been trimmed to delete the outliers. The sample in this table includes only observations classified into the same distress classification (0 or 1) by the Altman and Begley models. The t-statistics for each variable are reported in parentheses on the next line. c The White adjusted t-statistics are shown in parentheses. ***Significant at the 0.01 level; **Significant at the 0.05 level; *Significant at the 0.10 level

165

a Table 12: Regressions of Equation (3.8) for All Firm-Year Observations : Positive Bias (AFEt+1 is positive.)b Panel A: Ex Ante Analysisc

(3.8a) AFEt+1 = Σ γt Ct + γ1 AFEt-1 + γ 2 RESTt + γ 3 RETt,t+1 + γ4 LNUMBt+1 + γ5 RECESSt+1 + v (3.8b) AFEt+1 = Σ γt Ct + γ1 AFEt-1 + γ 2 RESTt + γ 3 RETt,t+1 + γ4 LNUMBt+1 + γ5 RECESSt+1 + γ6 OHLSON_PROB + v

(3.8c) AFEt+1 = Σ γt Ct + γ1 AFEt-1 + γ 2 RESTt + γ 3 RETt,t+1 + γ4 LNUMBt+1 + γ5 RECESSt+1 + γ6 DISTRESS + v Variable Name (Expected Sign) Equation (3.8a) Equation (3.8b) Equation (3.8c)

AFEt-1 (+/-) -0.001 -0.013 0.363 (-0.03) (-0.19) (0.49)

RESTt (+/-) -0.626 -0.749 -0.768 (-2.92***) (-3.02***) (-2.83***)

RETt,t+1 (+/-) -0.190 -0.210 -0.200 (-3.02***) (-2.74***) (-2.39**)

LNUMBt+1 (-) -0.166 -0.163 -0.138 (-2.35**) (-1.36) (-1.12)

RECESSt+1 (+) 0.002 -0.024 -0.065 (0.04) (-0.04) (-0.09) OHLSON_PROB (+) 0.088 (0.23) DISTRESS (+) 0.648 (2.28**) Adjusted R2 0.033 0.040 0.049 N 602 530 486 F-value 2.39*** 2.36*** 2.56***

a AFEt+1 is the analyst forecast error for year t + 1, the year after the restructuring charge. The year of the restructuring charge, t, is eliminated as in Chaney et al. (1999) because it is not clear whether or not analysts’ forecast for that period include any expectation of a restructuring charge. AFEt-1 is the analyst forecast error for year t – 1. RESTt equals 1 if the firm announced a restructuring charge in year t, and 0 otherwise. RETt,t+1 is the market return from the beginning of period t to the end of year t + 1. LNUMBt+1 is the log of the number of analysts forecasting earnings for a given firm in year t + 1. RECESSt+1 equals 1 if the observation occurred during a time of recession as defined by the National Bureau of Economic Research, and 0 otherwise. OHLSON_PROB is the probability of financial distress obtained from running Ohlson’s regression equation (equation (3.3)). DISTRESS equals 1 for firms classified by both Altman and Begley as distressed, and 0 otherwise. Corrections for heteroscedasticity, multicollinearity, and autocorrelation are made whenever problems are detected. The coefficients on the yearly intercepts are not reported. b The data has been trimmed to delete the outliers. The sample in this table includes only observations classified into the same distress classification (0 or 1) by the Altman and Begley models. The t-statistics for each variable are reported in parentheses on the next line. c The White adjusted t-statistics are shown in parentheses. ***Significant at the 0.01 level; **Significant at the 0.05 level; *Significant at the 0.10 level

166

a Table 12: Regressions of Equation (3.8) for All Firm-Year Observations : Positive Bias (AFEt+1 is positive.)b Panel B: Ex Post Analysisc

(3.8d) AFEt+1 = Σ γt Ct + γ1 AFEt-1 + γ 2 RESTt + γ 3 RETt,t+1 + γ4 LNUMBt+1 + γ5 RECESSt+1 + γ6 D1 + γ7 D2 + γ8 D3 + v (3.8e) AFEt+1 = Σ γt Ct + γ1 AFEt-1 + γ 2 RESTt + γ 3 RETt,t+1 + γ4 LNUMBt+1 + γ5 RECESSt+1 + γ6 BANKRUPT + v Variable Name (Expected Sign) Equation (3.8d) Equation (3.8e)

AFEt-1 (+/-) -1.824 -0.006 (-1.70*) (-0.25)

RESTt (+/-) -0.427 -0.534 (-2.14**) (-2.69***)

RETt,t+1 (+/-) -0.160 -0.126 (-2.16**) (-2.15**)

LNUMBt+1 (-) 0.146 -0.094 (2.01**) (-1.43)

RECESSt+1 (+) 2.556 0.803 (2.01**) (1.36) D1 (-) -22.013 (-2.54**) D2 (+/-) -22.256 (-2.56**) D3 (+/-) -0.957 (-1.96*) BANKRUPT (+/-) 6.29 (9.95***) Adjusted R2 0.548 0.172 N 486 602 F-value 33.65*** 8.80***

a AFEt+1 is the analyst forecast error for year t + 1, the year after the restructuring charge. The year of the restructuring charge, t, is eliminated as in Chaney et al. (1999) because it is not clear whether or not analysts’ forecast for that period include any expectation of a restructuring charge. AFEt-1 is the analyst forecast error for year t – 1. RESTt equals 1 if the firm announced a restructuring charge in year t, and 0 otherwise. RETt,t+1 is the market return from the beginning of period t to the end of year t + 1. LNUMBt+1 is the log of the number of analysts forecasting earnings for a given firm in year t + 1. RECESSt+1 equals 1 if the observation occurred during a time of recession as defined by the National Bureau of Economic Research, and 0 otherwise. D1 equals 1 for non-distressed firms, and 0 otherwise. D2 equals 1 for distressed firms that do not file for bankruptcy for at least three years after restructuring, and 0 otherwise. D3 equals 1 for firms filing for bankruptcy within three years of restructuring, and 0 otherwise. BANKRUPT equals 1 for firms filing for bankruptcy within three years of restructuring, and 0 otherwise. Corrections for heteroscedasticity, multicollinearity, and autocorrelation are made whenever problems are detected. The coefficients on the yearly intercepts are not reported. b The data has been trimmed to delete the outliers. The sample in this table includes only observations classified into the same distress classification (0 or 1) by the Altman and Begley models. c The White adjusted t-statistics are shown in parentheses. ***Significant at the 0.01 level; **Significant at the 0.05 level; *Significant at the 0.10 level

167

a Table 13: Regressions of Equation (3.8) for All Firm-Year Observations : Negative Bias (AFEt+1 is negative.)b Panel A: Ex Ante Analysisc

(3.8a) AFEt+1 = Σ γt Ct + γ1 AFEt-1 + γ 2 RESTt + γ 3 RETt,t+1 + γ4 LNUMBt+1 + γ5 RECESSt+1 + v (3.8b) AFEt+1 = Σ γt Ct + γ1 AFEt-1 + γ 2 RESTt + γ 3 RETt,t+1 + γ4 LNUMBt+1 + γ5 RECESSt+1 + γ6 OHLSON_PROB + v

(3.8c) AFEt+1 = Σ γt Ct + γ1 AFEt-1 + γ 2 RESTt + γ 3 RETt,t+1 + γ4 LNUMBt+1 + γ5 RECESSt+1 + γ6 DISTRESS + v Variable Name (Expected Sign) Equation (3.8a) Equation (3.8b) Equation (3.8c)

AFEt-1 (+/-) -0.065 -0.051 -0.057 (-0.29) (-0.70) (-0.27)

RESTt (+/-) 1.499 2.015 2.060 (1.89*) (1.35) (2.50**)

RETt,t+1 (+/-) 0.933 0.611 0.665 (2.31**) (2.56**) (1.63)

LNUMBt+1 (+) 0.268 0.637 0.566 (0.74) (1.92**) (1.48)

RECESSt+1 (-) 0.369 0.225 0.418 (0.19) (0.74) (0.21) OHLSON_PROB (+/-) 1.641 (1.77*) DISTRESS (+/-) 1.449 (1.82*) Adjusted R2 0.009 0.011 0.013 N 536 478 454 F-value 1.36 1.34 1.39

a AFEt+1 is the analyst forecast error for year t + 1, the year after the restructuring charge. The year of the restructuring charge, t, is eliminated as in Chaney et al. (1999) because it is not clear whether or not analysts’ forecast for that period include any expectation of a restructuring charge. AFEt-1 is the analyst forecast error for year t – 1. RESTt equals 1 if the firm announced a restructuring charge in year t, and 0 otherwise. RETt,t+1 is the market return from the beginning of period t to the end of year t + 1. LNUMBt+1 is the log of the number of analysts forecasting earnings for a given firm in year t + 1. RECESSt+1 equals 1 if the observation occurred during a time of recession as defined by the National Bureau of Economic Research, and 0 otherwise. OHLSON_PROB is the probability of financial distress obtained from running Ohlson’s regression equation (equation (3.3)). DISTRESS equals 1 for firms classified by both Altman and Begley as distressed, and 0 otherwise. Corrections for heteroscedasticity, multicollinearity, and autocorrelation are made whenever problems are detected. The coefficients on the yearly intercepts are not reported. b The data has been trimmed to delete the outliers. The sample in this table includes only observations classified into the same distress classification (0 or 1) by the Altman and Begley models. c The White adjusted t-statistics are shown in parentheses. ***Significant at the 0.01 level; **Significant at the 0.05 level; *Significant at the 0.10 level

168

a Table 13: Regressions of Equation (3.8) for All Firm-Year Observations : Negative Bias (AFEt+1 is negative.)b Panel B: Ex Post Analysisc

(3.8d) AFEt+1 = Σ γt Ct + γ1 AFEt-1 + γ 2 RESTt + γ 3 RETt,t+1 + γ4 LNUMBt+1 + γ5 RECESSt+1 + γ6 D1 + γ7 D2 + γ8 D3 + v (3.8e) AFEt+1 = Σ γt Ct + γ1 AFEt-1 + γ 2 RESTt + γ 3 RETt,t+1 + γ4 LNUMBt+1 + γ5 RECESSt+1 + γ6 BANKRUPT + v Variable Name (Expected Sign) Equation (3.8d) Equation (3.8e)

AFEt-1 (+/-) -0.057 -0.057 (-0.27) (-0.26)

RESTt (+/-) 2.139 1.360 (2.57***) (1.72*)

RETt,t+1 (+/-) 0.691 0.866 (1.67*) (2.14**)

LNUMBt+1 (+) 0.607 0.187 (1.57) (0.52)

RECESSt+1 (-) 0.421 0.348 (0.21) (0.18) D1 (+) 0.589 (0.13) D2 (+/-) 2.028 (0.44) D3 (-) 3.019 (0.78) BANKRUPT (-) -3.545 (-2.24**) Adjusted R2 0.010 0.017 N 454 536 F-value 1.27 1.61*

a AFEt+1 is the analyst forecast error for year t + 1, the year after the restructuring charge. The year of the restructuring charge, t, is eliminated as in Chaney et al. (1999) because it is not clear whether or not analysts’ forecast for that period include any expectation of a restructuring charge. AFEt-1 is the analyst forecast error for year t – 1. RESTt equals 1 if the firm announced a restructuring charge in year t, and 0 otherwise. RETt,t+1 is the market return from the beginning of period t to the end of year t + 1. LNUMBt+1 is the log of the number of analysts forecasting earnings for a given firm in year t + 1. RECESSt+1 equals 1 if the observation occurred during a time of recession as defined by the National Bureau of Economic Research, and 0 otherwise. D1 equals 1 for non-distressed firms, and 0 otherwise. D2 equals 1 for distressed firms that do not file for bankruptcy for at least three years after restructuring, and 0 otherwise. D3 equals 1 for firms filing for bankruptcy within three years of restructuring, and 0 otherwise. BANKRUPT equals 1 for firms filing for bankruptcy within three years of restructuring, and 0 otherwise. Corrections for heteroscedasticity, multicollinearity, and autocorrelation are made whenever problems are detected. The coefficients on the yearly intercepts are not reported. b The data has been trimmed to delete the outliers. The sample in this table includes only observations classified into the same distress classification (0 or 1) by the Altman and Begley models. c The White adjusted t-statistics are shown in parentheses. ***Significant at the 0.01 level; **Significant at the 0.05 level; *Significant at the 0.10 level

CHAPTER 6

CONCLUSIONS

6.1 Background, Purpose and Results

This dissertation investigates the link between restructuring efforts by corporations and subsequent recovery or filing for bankruptcy. Three essays examine three major aspects of the research topic. The first essay provides the link between restructuring efforts by corporations and subsequent bankruptcy filing. Thirty-three percent of the sample firms that reported operational restructurings during the period from 1993 to 2003 were determined to be in financial distress when both the Altman (1968) model and the Begley (1996) model are used. This indicates that many companies undertake restructuring efforts as an attempt to prevent impending bankruptcy.

Independent sample statistical tests demonstrate that firms classified as non-distressed are significantly different from those firms classified as financially distressed, providing support for the first research hypothesis. Further, the results of this essay demonstrate that over 90 percent of the firms in financial distress were able to avoid filing for bankruptcy for at least three years after restructuring. This high success rate for companies in financial distress provides statistical support for using restructuring to improve corporate financial health.

The results of the first essay also identify several factors including the quick ratio and the current cash debt coverage ratio that possess significant predictive power for determining which firms in financial distress are the most likely to file for bankruptcy. The bankruptcy prediction model is successful 81 percent of the time at determining which financially distressed firms will file or will not file for bankruptcy within three years of restructuring, and this provides moderate support for the second research hypothesis.

169 170

The results of the second essay demonstrate that the magnitude of corporate restructuring charges tends to provide value relevant information to investors. Two types of models are implemented. In the first type, the dollar amount of a corporate restructuring charge has a negative impact on price for all firm observations and a positive impact on price for all firm-event observations. The dollar amount of a restructuring charge is determined to have a negative impact on returns for all firm-event observations.1 In the second type of model, price and returns

are regressed on dollar restructuring charges and the probability of financial distress or a

distress/healthy dummy variable and control variables. Both the restructuring and financial

distress variables are highly significant and strongly support the research hypothesis that

restructuring costs (financial distress) have positive (has negative) impact on prices and returns.

Overall, the results of the second essay have implications for investors and analysts as they

determine how they should react to restructuring charge information.

The results of the third essay provide several important conclusions. First, the findings

demonstrate that analysts revise their forecasts downward after a restructuring charge

announcement, partially supporting the first hypothesis. Contrary to the findings of Chaney et al.

(1999), however the results of this essay do not suggest a leveling effect in analysts’ forecasts

over three- and five-year ahead forecasts. The results of the third essay demonstrate that analysts’

forecast accuracy decreases with the increasing probability of financial distress2 and show that

analysts are still optimistically biased after a restructuring charge3, which provides support for the

second and third research hypotheses. Also, the results demonstrate that there are fewer analyst

forecast errors (greater accuracy) for non-distressed firms and for distressed firms that avoid

filing for bankruptcy. The findings also demonstrate that analysts tend to be optimistically biased

1 See Panel B of both Table 5 and Table 7 (p. 112 and 116) in Essay 2. 2 See Table 10 (p. 163-164) in Essay 3. 3 See Tables 11-13 (p. 165-170) in Essay 3.

171

for distressed-bankrupt firms. Overall, the results reported in Essay 3 support the three hypotheses.

6.2 Limitations

Several limitations of this dissertation should be noted. First, future research may be able to improve the process by which the list of corporate restructurings was obtained. Because the initial list of restructuring firms used in this essay was obtained through a key word search of the newswires available on the Lexis-Nexis Database, it is possible that some restructuring firms were left out of the sample. If an announcement of the restructuring did not take place on a newswire, then it may have been inadvertently left out the sample. Perhaps by searching the 10-

Ks and 8-Ks of firms with special item dollar amounts available on COMPUSTAT, fewer firms that potentially have taken restructuring charges would have been eliminated from the sample.

Second, the results of the bankruptcy prediction model in equation 1.4 may be affected by the partial predictive ability of the model. Because equation 1.4 has only been used in Keener

(2003), the results of this model may not be generalizable to other samples. Third, the results may be adversely affected by including some companies that restructured during multiple years as firm-event observations, which is why the results are also presented at the firm level using companies that restructured only once. Fourth, a major limitation of this study is the “ad-hoc” statistical identification of healthy versus financially distressed firms based on well-established models published in leading journals.

In addition, the second essay is subject to potential criticism on price and return models.

This includes the efficient market explanation that the restructuring is already impounded in prices. The empirical results of the price models also suggest the potential existence of this limitation. Several limitations specifically apply to the results of the third essay. First, the small

172

number of firms with analysts’ forecasts 5-years ahead may limit the generalizability of the results from the third essay to other samples. Other explanations can be provided for the changes in analysts’ forecasts other than just the occurrence of the restructuring charge. There may be other confounding events that are not considered in this essay. Also, some other underlying conditions not discussed in the third essay may cause a firm to decide to restructure, which may lead to some increases in bias and declines in forecast accuracy.

6.3 Future Research

Some of the limitations may be addressed by future research. Further statistical comparisons could be made between the value relevance of restructuring charge information for sub-samples not examined in this dissertation. For example, future studies may conduct further analysis of first-time restructurings and consecutive restructurings by some companies may be applied. Future studies may be able to improve upon the bankruptcy or distress prediction model used to determine which of the firms in financial distress will file for bankruptcy by further

refining the variables to be included. Also, future research can extend the results of this study by

determining whether the results are generalizable to other time periods or to specific industries.

BIBLIOGRAPHY

Abarbanell, J. S., and R. Lehavy. 2003. Biased forecasts or biased earnings? The role of reported earnings in explaining apparent bias and over/underreaction in analysts' earnings forecasts. Journal of Accounting and Economics, Boston Conference October 2002.

Aboody, D., and B. Lev. 1998. The value relevance of intangibles: The case of software capitalization. Journal of Accounting Research 36 (Supplement): 161-191.

Accounting Principles Board. 1973. Reporting the results of operations—Reporting the effects of disposal of a segment of a business, and extraordinary, unusual and infrequently occurring events and transactions. APB Opinion No. 30.

Adut, D., W. H. Cready, and T. J. Lopez. 2003. Restructuring charges and CEO cash compensation: a reexamination. The Accounting Review 78 (1): 169-192.

Agarwal, V., and R. Taffler. 2006. Twenty-five years of Z-scores in the UK: Do they really work? Working paper presented at the 2006 AAA National Conference, Cranfield University, Bedfordshire, England.

Aharony, J., and R. Barniv. 2004. Using Financial Accounting Information in the Governance of Takeovers: An Analysis by Type of Acquirer. Journal of Accounting and Public Policy 23 (5): 321-349.

Alford, A. W., and P. G. Berger. 1999. A simultaneous equations analysis of forecast accuracy, analyst following, and trading volume. Journal of Accounting, Auditing & Finance 14 (3): 219-240.

Ali, A., A. Klein, and J. Rosenfeld. 1992. Analysts’ use of information about permanent and transitory earnings components in forecasting annual EPS. The Accounting Review 67: 183-198.

Altman, E. I. 1968. Financial ratios, discriminant analysis, and the prediction of corporate bankruptcy. Journal of Finance 23 (4): 589-609.

Ang, J. S., and S. J. Ciccone. 2001. Analyst forecasts and stock returns. Working paper, Florida State University, Tallahassee, FL.

Ashbaugh, H., and M. Pincus. 2001. Domestic accounting standards, International Accounting Standards, and the predictability of earnings. Journal of Accounting Research 39 (3): 417-434.

Asquith, P., M.B. Mikhail, and A. S. Au. 2005. Information content of equity analyst reports. Journal of Financial Economics 75 (2): 245-282.

173 174

Astebro, T., and J. K. Winter. 2001. More than a dummy: The probability of failure, survival and acquisition of firms in financial distress. Working paper, University of Toronto, Toronto, ON.

Athanassakos, G., and M. Kalimipalli. 2003. Analyst forecast dispersion and future stock return volatility. Quarterly Journal of Business and Economics 42 (1/2): 57-78.

Atiase, R. K., D. E. Platt, and S. Y. Tse. 2004. Operational restructuring charges and post- restructuring performance. Contemporary Accounting Research 3 (Fall): 493-522.

Baber, W. R., and S. Kang. 2002. The impact of split adjusting and rounding on analysts’ forecast error calculations. Accounting Horizons 16 (4): 277-289.

Bae, K., R. M. Stulz, and H. Tan. 2005. Do local analysts know more? A cross-country study of the performance of local analysts and foreign analysts. Working paper, Queen’s University, Kingston, Ontario, Canada.

______, H. Tan, and M. Welker. 2006. International GAAP differences—The impact on foreign analysts. Working paper, Queen’s University, Kingston, Ontario, Canada.

Ballester, M., J. Livnat, and N. Sinha. 1999. Corporate reorganizations: Changes in the intensity of labor and capital expenditures. Journal of Business Finance and Accounting 26 (9/10): 1205-1238.

Barniv, R., and J. McDonald. 1999. Review of categorical models for classification issues in accounting and finance. Review of Quantitative Finance and Accounting 13 (1): 39-62.

______, M. J. Myring, and W. B. Thomas. 2005. The association between the legal and financial reporting environments and forecast performance of individual analysts. Contemporary Accounting Research 22 (4): 727-758.

Barron, O. E., D. Byard, and O. Kim. 2002. Changes in analysts’ information around, earnings announcements. The Accounting Review 77 (4): 821-846.

Barth, M. E., and A. P. Hutton. 2004. Analyst earnings forecast revisions and the pricing of accruals. Review of Accounting Studies 9: 59-96.

Bartolini, L., A. Razin, S. Symansky, M. Burda, and F. Van der Ploeg. 1995. Fiscal Restructuring in the Group of Seven Major Industrial Countries in the 1990s: Macroeconomic Effects. Economic Policy 10 (20): 109-146..

Bartov, E., F. W. Lindahl, and W. E. Ricks. 1998. Stock price behavior around announcements of write-offs. Review of Accounting Studies 3 (4): 327-346.

Begley, J., J. Ming, and S. Watts. 1996. Bankruptcy classification errors in the 1980s: An empirical analysis of Altman’s and Ohlson’s models. Review of Accounting Studies 1 (4): 267-284.

175

Bens, D. A. 2002. The determinants of the amount of information disclosed about corporate restructurings. Journal of Accounting Research 40 (1): 1-20.

______, and R. Johnston. 2005. Accounting discretion: use or abuse? Restructuring charges – 1989-1992. Working paper, University of Chicago, Chicago, IL.

Berger, P. G., and E. Ofek. 1999. Causes and effects of corporate refocusing programs. The Review of Financial Studies 12 (2): 311-345.

Berman, D. K. 2006. Del Monte nears $700 million deal for Meow Mix. The Wall Street Journal (March 2).

Bernhardt, D., and M. Campello. 2002. The dynamics of earnings forecast management. Working paper, University of Illinois, Urbana-Champaign, IL.

Bethel, J. E., and J. Liebeskind. 1993. The effects of ownership structure on corporate restructuring. Strategic Management Journal 14: 15-31.

Black, E. L., T. A. Carnes, and V. J. Richardson. 2000. The value relevance of multiple occurrences of nonrecurring items. Review of Quantitative Finance and Accounting 15 (4): 391-411.

Blackwell, D. W., M. W. Marr, and M. F. Spivey. 1990. Plant-closing decisions and the market value of the firm. Journal of Financial Economics 26 (2): 277-288.

Bonner, S. E., B. R. Walther, and S. M. Young. 2003. Sophistication-related differences in investors’ models of the relative accuracy of analysts’ forecast revisions. The Accounting Review 78 (3): 679-706.

Boone, A. L., and J. H. Mulherin. 2001. Valuing the process of corporate restructuring. Working paper, Penn State University, State College, PA.

Bowman, E. H., and H. Singh. 1993. Corporate restructuring: Reconfiguring the firm. Strategic Management Journal 14: 5-14.

______, ______, M. Useem, and R. Bhadury. 1999. When does restructuring improve economic performance? California Management Review 41 (2): 33-54.

Brickley, J. A., and L. D. Van Drunen. 1990. Internal corporate restructuring: An empirical analysis. Journal of Accounting and Economics 12 (1-3): 251-280.

Bronson, G. 1994. Questionable restructurings. Global Finance 8 (8): 36-37.

Bulkeley, W. M. 2006. Xerox’s net income rises 18% amid cost cuts, but sales slip. The Wall Street Journal (January 26).

Bunsis, H. 1997. A description and market analysis of write-off announcements. Journal of Business Finance and Accounting 24 (9 & 10): 1385-1400.

176

Carter, M. E. 1998. Does operating performance improve after corporate restructurings? Working paper, MIT, Boston, MA.

______. 2000. Operating performance following corporate restructurings. Working paper, Columbia University, New York City, NY.

Chalos, P., and C. J. Chen. 2002. Employee downsizing strategies: Market reaction and post announcement financial performance. Journal of Business Finance & Accounting 29 (5 & 6): 847-870.

Chaney, P. K., C. E. Hogan, and D. C. Jeter. 1999. The effect of reporting restructuring charges on analysts’ forecast revisions and errors. Journal of Accounting and Economics 27: 261-284.

Chen, Q., J. Francis, and W. Jiang. 2005. Investor learning about analyst predictive ability. Journal of Accounting and Economics 39 (1): 3-24.

Chiang, H. 2005. Analysts’ financial forecast accuracy and information transparency. Journal of American Academy of Business 7 (2): 164-167.

Ciccone, S. J. 2001. Analyst forecast properties, financial distress, and business risk. Working paper, University of New Hampshire, Durham, NH.

Cinca, C. S., C. M. Molinero, and J. L. G. Larraz. 2001. Country and size effects in financial ratios: A European perspective. Working paper, University of Zaragoza, Spain.

Claus, J. J. 2000. Individual analyst forecast accuracy: An international analysis. Ph.D. dissertation, Columbia University.

Clement, M. B., and T. J. Lopez. 2000. Evidence on the effect of multiple corporate restructurings on analysts’ earnings forecasts: do analysts learn from prior restructuring events? Working paper, University of Texas, Austin, TX.

______, and S. Y. Tse. 2003. Do investors respond to analysts’ forecast revisions as if forecast accuracy is all that matters? The Accounting Review 78 (1): 227-249.

______, L. L. Koonce, and T. J. Lopez. 2006. The roles of task-specific experience and innate ability in understanding analyst performance. Working paper, University of Texas at Austin, Austin, TX.

Coucke, K., E. Pennings, L. E. Sleuwaegen. 2006. Employee lay-off under different modes of restructuring. Working paper, Vlerick Leuven Gent Management School, Leuven, Belgium.

Covitz, D. M., S. Han, and B. A. Wilson. 2006. Are longer really more costly? Working paper, Board of Governors of the Federal Reserve, Washington D. C.

177

Cowen, A., B. Groysberg, and P. M. Healy. 2003. Which types of analyst firms make more optimistic forecasts? Harvard NOM Working paper 03-46, Harvard University, Boston, MA.

Daniels, R. B., R. W. Rouse, and T. Weirich. 1995. New guidance in accounting for corporate restructuring: EITF 94-3. The CPA Journal 65 (6): 30-34.

Dechow, P. M., M. R. Huson, and R. G. Sloan. 1994. The effect of restructuring charges on executives’ cash compensation. The Accounting Review 69 (1): 138-156.

______, A. P. Hutton, and R. G. Sloan. 2000. The relation between analysts’ forecasts of long- term earnings growth and stock price performance following equity offerings. Contemporary Accounting Research 17 (1): 1-32.

Denis, D. J., and T. A. Kruse. 2000. Managerial discipline and corporate restructuring following performance declines. Journal of Financial Economics 55 (3): 391-425.

Dhumale, R. 1998. Earnings retention as a specification mechanism in logistic bankruptcy models: A test of the free cash flow theory. Journal of Business Finance & Accounting 25 (7/8): 1005-1023.

Dial, J., and K. J. Murphy. 1995. Incentives, downsizing, and value creation at General Dynamics. Journal of Financial Economics 37: 261-314.

Dische, A. P. 2001. Dispersion in analyst forecasts and the profitability of earnings momentum strategies. EFMA 2001 Lugano Meetings. Working paper, University of St. Gallen, Switzerland.

Donaldson, G. 1990. Voluntary restructuring: The case of General Mills. Journal of Financial Economics 27 (1): 117-141.

Duru, A., and D. M. Reeb. 2002. International diversification and analysts’ forecast accuracy and bias. The Accounting Review 77 (2): 415-433.

Easton, P. D., and S. J. Monahan. 2005. An evaluation of accounting-based measures of expected returns. The Accounting Review 80 (2): 501-538.

Elayan, F. A., G. S. Swales, B. A. Maris, and J. R. Scott. 1998. Market reactions, characteristics, and the effectiveness of corporate layoffs. Journal of Business Finance & Accounting 25 (3/4): 329-351.

Elgers, P. T., M. H. Lo, and R. J. Pfeiffer, Jr. 2001. Delayed security price adjustments to financial analysts’ forecasts of annual earnings. The Accounting Review 76 (4): 613-632.

Elliott, J. A., and J. D. Hanna. 1996. Repeated accounting write-offs and the information content of earnings. Journal of Accounting Research (Supplement): 117-134.

178

Ellis, K., and W. R. Gebhardt. 2006. Expecting to be surprised: An analysis of price responses to earnings announcements. Working paper, University of California Davis, Davis, CA.

Ellison, S. 2006. Kraft to cut 8,000 more jobs. The Wall Street Journal (January 31).

Ertimur, Y., J. Sunder, and S. V. Sunder. 2006. Measure for measure: The relation between forecast accuracy and recommendation profitability of analysts. Working paper, Stanford University, Stanford, CA.

Faccio, M., and R. Sengupta. 2006. Corporate response to distress: Evidence from the Asian financial crisis. Working paper, Vanderbilt University, Nashville, TN.

Fan, D. K., R. W. So, and J. J. Yeh. 2006. Analyst earnings forecasts for publicly traded insurance companies. Review of Quantitative Finance and Accounting 26: 105-136.

Financial Accounting Standards Board. 1986. Display of business restructuring provisions in the income statement. Emerging Issues Task Force Issue No. 86-22. Norfolk, CT: FASB.

______. 1987. Restructuring of operations: Implication of SEC Staff Accounting Bulletin No. 67. Emerging Issues Task Force Issue No. 87-4. Norfolk, CT: FASB.

______. 1994. Liability recognition for certain employee termination benefits and other costs to exit an activity (including certain costs incurred in a restructuring). Emerging Issues Task Force Issue No. 94-3. Norfolk, CT: FASB.

______. 1975. Accounting for contingencies. Statement of Financial Accounting Standards Statement No. 5. Stamford, CT: FASB.

______. 2002. Accounting for costs associated with exit or disposal activities. Statement of Financial Accounting Standards Statement No. 146. Stamford, CT: FASB.

Francis, J., J. D. Hanna, and L. Vincent. 1996. Causes and effects of discretionary write-offs. Journal of Accounting Research 34 (Supplement): 117-134.

Frantz, P. 1999. Discretionary write-downs, write-offs, and other restructuring provisions: A signaling approach. Accounting and Business Research 29 (2): 109-121.

Frederikslust, R. A., T. J. Werner, S. N. Soedito, and J. Van Dalen. 2003. Corporate restructuring of Dutch companies: An empirical analysis. Working paper, University of Rotterdam.

Friesen, G. C., and P. A. Weller. 2005. Quantifying Cognitive Biases in Analyst Earnings Forecasts. Working paper, University of Nebraska-Lincoln, Lincoln, NE.

Frydman, H., E. I. Altman, and D. Kao. 1985. Introducing recursive partitioning for financial classification: The case of financial distress. Journal of Finance 40 (1): 269-291.

179

Galanti, S. 2004. Distinct or Distant? Analysts' Consensus and Forecasts in the Stock Market. Working paper, University of Paris Nanterre, Paris, France.

Gilbert, L. R., K. Menon, and K. B. Schwartz. 1990. Predicting bankruptcy for firms in financial distress. Journal of Business Finance and Accounting 17 (1): 161-171.

Gilson, S. C. 1998. Course overview: Creating value through corporate restructuring. Working paper, Harvard University, Boston, MA.

______, P. M. Healy, C. F. Noe, and K. Palepu. 1998. Information effects of spin-offs, equity carve-outs, and targeted stock offerings. Working paper, Harvard University, Boston, MA.

Grice, J. S., and M. T. Dugan. 2001. The limitations of bankruptcy prediction models: Some cautions for the researcher. Review of Quantitative Finance and Accounting 17 (2): 151- 166.

______, and R. W. Ingram. 2001. Tests of the generalizability of Altman’s bankruptcy prediction model. Journal of Business Research 54 (1): 53-61.

Gu, Z., and J. S. Wu. 2003. Earnings skewness and analyst forecast bias. Journal of Accounting and Economics 35 (1): 5-29.

Halperin, M., and S. J. Bell. 1992. Research Guide to Corporate Acquisitions, Mergers, and Other Restructuring. New York: Greenwood Press.

Han, B. H., and D. Manry, and W. Shaw. 2001. Improving the precision of analysts’ earnings forecasts by adjusting for predictable bias. Review of Quantitative Finance and Accounting 17: 81-98.

Harjoto, M. A., and J. Zaima. 2005. Conflict in whispers and analyst forecasts: Which one should be your guide? Working paper, San Jose State University, San Jose, CA.

Harlan, C. 1994. SEC wants explanations from firms planning to post restructuring losses. The Wall Street Journal (February 25).

Hayn, C. 1995. The information content of losses. Journal of Accounting and Economics 20 (2): 125-153.

Herz, R. H., and E. J. Abahoonie. 1991. Maximizing the value of troubled companies. Mergers and Acquisitions 26 (November/December): 13-21.

_____, and _____. 1992. Restructuring businesses in the 1990s. Journal of Accountancy 173 (April): 28-33.

_____, _____, and L. Krule. 1992. Restructuring corporate debt: Finding your way through the tax and accounting maze. Business 94 (September): 10-12.

180

Hillegeist, S. A., E. K. Keating, D. P. Cram, and K. G. Lundstedt. 2004. Assessing the probability of bankruptcy. Review of Accounting Studies 9 (1): 5-34.

Hillier, D. J., and P. M. L. McColgan. 2005. Managerial discipline and firm responses to a decline in operating performance. Working paper, University of Strathclyde, Glasgow, UK.

Hogan, C. E., and D. C. Jeter. 1998. The information content of restructuring charges: A contextual analysis. Working paper, Vanderbilt University, Nashville, TN.

Hollie, D., J. Livnat, and B. Segal. 2005. Oops, Our earnings were indeed preliminary. Journal of Portfolio Management 31 (2): 94-104.

Holmström, B. R. 2006. Pay without performance and the managerial power hypothesis: A comment. Working paper, MIT, Boston, MA.

Hope, O., and T. Kang. 2005. The association between macroeconomic uncertainty and analysts’ forecast accuracy. Journal of International Accounting Research 4 (1): 23-38.

______. 2003. Disclosure practices, enforcement of accounting standards, and analysts’ forecast accuracy: An international study. Journal of Accounting Research 41 (2): 235-272.

Hoskisson, R. E., and T. A. Turk. 1990. Corporate restructuring: Governance and control limits of the internal capital market. Academy of Management Review 15 (3): 459-477.

Huai, Z. 2000. How rational is the stock market towards properties of analyst consensus forecasts? Ph.D. dissertation, Columbia University.

Irvine, P. J. 2004. Analysts’ forecasts and brokerage firm trading. The Accounting Review 79 (1): 125-149.

Ivkovic, Z., and N. Jegadeesh. 2002. The timing and value of forecast and recommendation revisions: Do analysts receive early peek at good news? Working paper, AFA 2004 San Diego Meetings, San Diego, CA.

Janes, T. D. 2006. Accruals and financial distress. Working paper presented at the 2006 AAA National Conference, SUNY Buffalo, Buffalo, NY.

Jennings, R., L. H. Martin, and R. B. Thompson II. 1998. The effects of corporate restructurings on the usefulness of the balance sheet. Journal of Financial Statement Analysis 3 (2): 44-56.

John, K., L. H. Lang, and J. Netter. 1992. The voluntary restructuring of large firms in response to performance decline. Journal of Finance 47 (3): 891-917.

Kam, A., D. B. Citron, and Y. G. Muradoglu. 2006. Restructuring strategies of Chinese firms in distress: Are they value enhancing? Working paper, City University, London, England.

181

Kane, G. D., and F. M. Richardson. 2000. Corporate restructuring and the likelihood of emergence from financial distress. Working paper, University of Delaware, Newark, DE.

Kang, J., and A. Shivdasani. 1997. Corporate restructuring during performance declines in Japan. Journal of Financial Economics 46 (1): 29-65.

Keener, M. H. 2003. How successful are restructuring efforts for firms trying to improve efficiency and for firms in financial distress? Working paper, Kent State University, Kent, OH.

Khurana, I. K., and B. Lippincott. 2000. Restructuring and firm value: The effects of profitability and restructuring purpose. Journal of Business Finance and Accounting 27 (9/10): 1106-1129.

Kim, S. 2004. Macro Effects of Corporate Restructuring in Japan. IMF Staff Papers 51 (3): 457-492.

Kini, O., S. L. Mian, M. Rebello, and A. Venkateswaran. 2005. Core specialization, forecast accuracy, and the market for international analysts, Working paper, Georgia State University, Atlanta, GA.

Koga, K., and S. Uchino. 2006. Bank relations and security analyst forecasts among Japanese firms, Working paper, University of Illinois at Urbana-Champaign, Urbana-Champaign, IL.

Ko, P., and P. Lin. 2006. An evolution-based approach with modularized evaluations to forecast financial distress. Knowledge-Based Systems 19 (1): 84-91.

Koh, K., D. A. Matsumoto, and S. Rajgopal. 2006. Meeting analyst forecasts in the post-Enron world: Early evidence on stock market rewards and managerial actions. Working paper, University of Washington, Seattle, WA.

Kross, W., T. Park, and B. Ro. 2001. Is ‘operational restructuring’ a magic phrase? An examination of market responses to restructuring announcements. Working paper. Purdue University, Purdue, IN.

Kwon, S. S. 2002. Financial analysts’ forecast accuracy and dispersion: High-tech versus low- tech stocks. Review of Quantitative Finance and Accounting 19: 65-91.

Lacina, M. J., and K. E. Karim. 2004. Tests of market reaction and analysts’ forecast revisions to the disclosure of improved management earnings expectations: A case of concurrent bad news management earnings forecasts. Review of Quantitative Finance and Accounting 23: 123-148.

Lai, J., and S. Sudarsanam. 1997. Corporate restructuring in response to performance decline: Impact on ownership, governance and lenders. European Finance Review 1 (2): 197- 233.

182

Landers, P. 2002. Sony, Hitachi plan cutbacks in staffing as tech slowdown spurs restructuring. The Wall Street Journal (March 1).

Lang, M. H., and R. J. Lundholm. 1996. Corporate disclosure policy and analyst behavior. The Accounting Review 71 (4): 467-492.

Lawton, C. 2006. H-P to split hand-held operation, notebooks. The Wall Street Journal (February 13).

Levitt, A. 1998. The numbers game. Remarks delivered at the NYU Center for Law and Business, New York, NY, September 28.

Lim, T. K. 2001. Rationality and analysts’ forecast bias. The Journal of Finance 56 (1): 369- 385.

______, and H. C. Kong. 2004. New evidence on price impact of analyst forecast revisions. International Review of Financial Analysts 13: 161-190.

Lin, B., and R. Yang. 2006. The effect of repeat restructuring charges on analysts’ forecast revisions and accuracy. Review of Quantitative Finance and Accounting 27 (3): 267- 283.

Lin, J. C., and M. S. Rozeff. 1993. Capital market behavior and operational announcements of layoffs, operation closings, and pay cuts. Review of Quantitative Finance and Accounting 3: 29-45.

Lin, Z., and M. S. H. Shih. 2006. Penalty for firms reporting zero or small positive earnings surprises. Working paper, National University of Singapore, Singapore.

Liu, J., and W. Su. 2005. Forecasting analysts’ forecast errors. Working paper, UCLA, Los Angeles, CA.

Liu, Z. 2005. The role of financial analysts and institutional investors in the 'Numbers Game', Working paper, University of California Irvine, Irvine, CA.

Lopez, T. J. 2002. Evidence on the incremental information contained in the components of restructuring charges. Journal of Business Finance and Accounting 29 (5/6): 613-659.

______, P. R. Regier, and L. Holder-Webb. 2002. Do restructurings improve operating performance? Working paper, University of South Carolina, Columbia, SC.

Lys, T., and L. G. Soo. 1995. Analysts’ forecast precision as a response to competition. Journal of Accounting, Auditing & Finance 10 (4): 751-765.

Markides, C. C. 1992. Consequences of corporate refocusing: Ex ante evidence. Academy of Management Journal 35 (2): 398-412.

183

Martin, J. D., and J. W. Kensinger. 1990. Exploring the controversy over corporate restructuring. Financial Executives Research Foundation: Morristown, New Jersey.

McCracken, J., and J. B. White. 2006. Ford will shed 28% of workers in North America; Car maker to close 14 plants as it joins GM in overhaul of Detroit’s Business Model. The Wall Street Journal (January 24).

McHugh, Christopher M., ed. 1993-2003. The Bankruptcy Yearbook & Almanac. Boston: New Generation Research, Inc.

McKibben, W. 1972. Econometric forecasting of common stock investment returns: A new methodology using fundamental operating data. Journal of Finance 27 (May): 371-380.

Mest, D. P., and E. Plummer. 2000. Revisions in analysts’ earnings forecasts: Evidence of non- linear adaptive expectations. Journal of Forecasting 19: 467-484.

______, and ______. 2003. Analysts’ rationality and forecast bias: Evidence from sales forecasts. Review of Quantitative Finance and Accounting 21 (2): 103-122.

Moehrle, S. R. 2002. Do firms use restructuring charge reversals to meet earnings targets? The Accounting Review 77 (2): 397-413.

Morton, R. M., and J. D. Neill. 2001. The value relevance of current and forward looking accounting information subsequent to a corporate restructuring. Advances in Accounting 18: 1-24.

Muehlberger, U. 2005. Hierarchies, relational contracts and new forms of outsourcing. ICER Working paper 22/2005, Vienna University, Vienna, Austria.

Neophytou, E. and C. M. Molinero. 2004. Predicting Corporate Failure in the UK: A Multidimensional Scaling Approach. Journal of Business Finance and Accounting 31 (5/6): 677-710.

Ohlson, J. A. 1980. Financial ratios and the probabilistic prediction of bankruptcy. Journal of Accounting Research 18 (1): 109-131.

Park, C. W., and E. K. Stice. 2000. Analyst forecasting ability and the stock price reaction to forecast revisions. Review of Accounting Studies 5: 259-272.

Pastor, L., M. Sinha, and B. Swaminathan. 2006. Estimating the Intertemporal Risk-Return Tradeoff Using the Implied . CRSP Working paper 607, University of Chicago, Chicago, IL.

Pinello, A. S. 2005. How do individual investors use analyst forecasts? The implications for investor reaction to forecast revisions and earnings surprises. AAA 2006 Financial Accounting and Reporting Section (FARS) Meeting. Working paper, Georgia State University, Atlanta, GA.

184

Pirrong, G. D., and D. Koeppen. 1993. How to avoid bankruptcy. Management Accounting 74 (February): 35-40.

Poon, P. S., G. D. Newbould, and C. Durtschi. 2001. Market reactions to corporate restructurings. Review of Qualitative Finance and Accounting 16: 269-290.

Pulliam, S., and L. Berton. 1994. A restructuring of write-offs is in the making. The Wall Street Journal (November 2).

Radcliffe, V. S., D. R. Campbell, and T. J. Fogarty. 2001. Exploring downsizing: A case study on the use of accounting information. Journal of Management Accounting Research 13: 131-157.

Ramnath, S., S. Rock, and P. Shane. 2006. A review of research related to financial analysts’ forecasts and stock recommendations. Working Paper, Georgetown University, Washington D. C.

Renneboog, L., and P. G. Szilagyi. 2006. Corporate restructuring and bondholder wealth. ECGI - Finance Working paper 124/2006, Tilburg University, Netherlands.

Roychowdhury, S. 2005. Earnings management through real activities manipulation. Working paper, MIT, Boston, MA.

Scherbina, A. D. 2004. Analyst Disagreement, Forecast Bias and Stock Returns. Working paper, Harvard Business School Publishing, Forthcoming, Boston, MA.

Securities and Exchange Commission. 1986. Income statement presentation of restructuring charges. Release No. SAB 67, 17 CFR Part 211. Washington D.C.: SEC.

______. 1999. Restructuring and Impairment Charges. Release No. SAB 100, 17 CFR Part 211. Washington D.C.: SEC.

Shane, P., and P. Brous. 2001. Investor and (Value Line) analyst underreaction to information about future earnings: The corrective role of non-earnings-surprise information. Journal of Accounting Research 39 (2): 387-404.

Shelley, M. K., T. C. Omer, and T. J. Atwood. 1998. Capital restructuring and accounting compliance costs: The case of publicly traded partnerships. Journal of Accounting Research 36 (2): 365-378.

Shin, K, and Y. Lee. 2002. A genetic algorithm application in bankruptcy prediction modeling. Working paper. Ewha Womans University, Seoul, South Korea.

Smart, S. B., and J. Waldfogel. 1994. Measuring the effect of restructuring on corporate performance: The case of management buyouts. Review of Economics and Statistics 76: 503-511.

185

Smith, R., and S. Lipin. 1996. Are companies using restructuring costs to fudge the figures? The Wall Street Journal (January 30).

Stevens, D. E., and A. W. Williams. 2004. Inefficiency in earnings forecasts: Experimental evidence of reactions to positive vs. negative information. Experimental Economics 7 (1): 75-92.

Strong, N. 1992. Modeling abnormal returns. Journal of Business Finance and Accounting 19 (4): 533-553.

Sudarsanam, S., and J. Lai. 2001. Corporate financial distress and turnaround strategies: An empirical analysis. British Journal of Management 12: 183-199.

Sung, T. K., C. Namsik, and G. Lee. 1999. Dynamics of modeling in data mining: Interpretive approach to bankruptcy prediction. Journal of Management Information Systems 16 (1): 63-85.

Webster, G. H., and C. Coyne. 1991. Comparing MDA to logistic analysis in predicting small firm bankruptcy. Atlantic Economic Journal 19 (1): 64-66.

Weiss, L. A., and V. Capkun. 2006. The impact of incorporating the cost of errors into bankruptcy prediction models. Working paper presented at the 2006 AAA National Conference, University of Lausanne, Switzerland.

Worrell, D. L., W. N. Davidson, and V. M. Sharma. 1991. Layoff announcements and stockholder wealth. Academy of Management Journal 34 (3): 662-678.

Yan, A. 2006. Analyst sentiment around takeover announcements, Working paper, Fordham University, New York, NY.

Yeung, P. 2003. Belief-revisions after earnings announcements: Evidence from security analysts’ forecast revisions. Ph.D. dissertation, University of Oregon.

Zantout, Z. 1994. External capital market control, corporate restructuring, and firm performance during the 1980s. Journal of Business Finance & Accounting 21 (1): 37-64.

Zmijewski, M. E. 1984. Methodological issues related to the estimation of financial distress prediction models. Journal of Accounting Research 22 (Supplement): 59-82.