MANAGEMENT EARNINGS FORECASTS, CASH FLOW FORECASTS AND EARNINGS MANAGEMENT by Hanmei Chen

A Dissertation Presented in Partial Fulfillment of the Requirements for the Degree Doctor of Philosophy

ARIZONA STATE UNIVERSITY August 2008

UMI Number: 3318443

INFORMATION TO USERS

The quality of this reproduction is dependent upon the quality of the copy submitted. Broken or indistinct print, colored or poor quality illustrations and photographs, print bleed-through, substandard margins, and improper alignment can adversely affect reproduction. In the unlikely event that the author did not send a complete manuscript and there are missing pages, these will be noted. Also, if unauthorized copyright material had to be removed, a note will indicate the deletion.

®

UMI UMI Microform 3318443 Copyright 2008 by ProQuest LLC. All rights reserved. This microform edition is protected against unauthorized copying under Title 17, United States Code. ProQuest LLC 789 E. Eisenhower Parkway PO Box 1346 Ann Arbor, Ml 48106-1346

MANAGEMENT EARNINGS FORECASTS, CASH FLOW FORECASTS AND EARNINGS MANAGEMENT by Hanmei Chen

has been approved April 2008

Graduate Supervisory Committee: James Boatsman, Chair Michael Mikhail Molly Mercer

ACCEPTED BY THE GRADUATE COLLEGE

ABSTRACT This study examines the relationship between discretionary accruals and managers' choice of whether to disclose cash flow forecasts with earnings forecasts. Empirically, this paper conducts both cross-sectional analysis and time-series analyses. In a cross-sectional analysis, it finds a significant negative relation between discretionary accruals and the choice of issuing cash flow forecasts along with earnings forecasts. The result is consistent with the hypothesis that companies who issue both management earnings forecasts and cash flow forecasts would be more reluctant to manipulate earnings through discretionary accruals than companies who only issue management earnings forecasts. Similarly, in a time-series analysis, this study also finds a significant negative relation between discretionary accruals and the choice of issuing both cash flow forecasts and earnings forecasts. The result supports the hypothesis that the discretionary accruals of the same company are lower at the time it issues both cash flow forecasts and earnings forecasts than at the time it only issues earnings forecasts. This research contributes to literature by providing new evidence how accounting information from voluntary disclosure impacts earnings management. It also enriches the literature on the consequences of management forecasts.

iii

ACKNOWLEDGMENTS I gratefully thank my advisor, Dr. James Boatsman, for his guidance and encouragement throughout my five-year doctoral study at Arizona State University. I believe that what I have learned from him will significantly benefit my future career. Special thanks to my dissertation committee members, Dr. Michael Mikhail and Dr. Molly Mercer for their guidance and encouragement on my dissertation. I also thank all other faculty members for their support. I would like to thank my good colleagues and friends, Jian Zhang, Zhan Gao, J.K Aier, Ryan Casey, David Erkens, Rick Laux and Wan-Ting Wu for their support to my studies.

iv

TABLE OF CONTENTS Page LIST OF TABLES

vii

I. INTRODUCTION

1

II. LITERATURE REVIEW

5

Background

5

Incentives for Voluntary Disclosure

5

Consequences of Voluntary Disclosure

6

Management Forecasts and Earnings Management

8

III. HYPOTHESIS DEVELOPMENT

11

Management Choices to Avoid Negative Earnings Surprises

11

Management Incentives to Issue Cash Flow Forecasts

12

Cash Flow Forecasts and Earnings Management

14

Penalties upon the Detection of Earnings Management

16

Hypotheses

17

IV. RESEARCH DESIGN AND SAMPLE SELECTION

19

Cross-Sectional Analysis

19

Cross-Sectional Analysis Using an Alternative Approach

21

Time-Series Analysis

23

Control Variables

24

Sample Selection

25

V EMPIRICAL RESULTS

27

Descriptive Statistics

27 v

Cross-Sectional Analysis

28

Cross-Sectional Analysis Using an Alternative Approach

28

Time-Series Analysis

29

VI. SELECTION BIAS CONCERNS

30

VII. CONCLUSION

32

REFERENCES

33

APPENDIX A FRAMEWORK FOR ANALYZING MANAGEMENT EARNINGS FORECASTS

38

VI

LIST OF TABLES Table

Page

1. Definition of Variables

40

2. Descriptive Statistics

41

3. Correlation between Variables

43

4. Results from Cross-Sectional Analysis

45

5. Results from Alternative Approach of Cross-Sectional Analysis

47

6. Results from Time-Series Analysis

49

7. Test of Selection Bias by Using Heckman's Selection Model

51

vii

I. INTRODUCTION This dissertation investigates a specific form of voluntary disclosure, namely management cash flow forecasts. While previous studies have provided evidence as to why management issues cash flow forecasts, I focus on one consequence of management choice of issuing cash flow forecasts, which is ex post earnings management. Management issues company guidance on earnings and/or cash flows through conference calls and other press releases.

Company issuance guidance is also called

management forecasts in practice and research.

Among different forms of management

forecasts, earnings forecasts are most commonly used.

Since the year 2000, the number

of companies providing both cash flows and earnings guidance has increased dramatically.

It is believed that the increase is a part of the impact from Regulation FD

(2000), which requires management to provide forecasts to all parties or curtail their forecasts completely.

This study investigates managers' behavior toward earnings

management as a consequence of management forecasts.

Specifically it investigates

whether the managers' choice of issuing both earnings and cash flow forecasts ex ante will reduce the level of earnings management. Prior research has revealed that managers who provide earnings forecasts may later engage in earnings management.

Evidence shows a significant positive association

between management forecasts and earnings management, in which managers usually use discretionary accruals to meet their own earnings forecasts.

Kasznik (1999) suggests

that when management issues some form of earnings forecasts, it sets a self-imposed target to be met or beaten when realized earnings are announced.

2 However, managers tend to reduce earnings management when investors have more detailed information.

Dutta and Gigler (2002) develop an analytical model to explain

how management is more likely to manage earnings when they do not provide an earnings forecast than when they do.

Collins and Mclnnis (2006) investigate how

analysts' cash flow forecasts deter earnings management.

Their results show that cash

flow forecasts from analysts will reduce accrual manipulation and the likelihood that firms will meet their earnings targets.

Richardson (2000) suggests a systematic

relationship between the level of information asymmetry and the level of earnings management.

If investors have sufficient resources, incentives, or access to relevant

information to monitor managers' actions, the level of earnings management will be reduced.

Management forecasts are an important source of incremental information

for investors beyond financial statements required by regulations.

When management

issues both an earnings forecast and a cash flow forecast, it is possible for investors to decompose earnings into operating cash flows and accruals.

The decomposition gives

investors a clearer picture of a company's future earnings, cash flows and accruals. Therefore, management may be more reluctant to manipulate earnings through accruals in subsequent earnings. In this dissertation, I investigate how management's choice of disclosing forecasts (earnings and cash flows) ex ante may impact their earnings management behavior. Both cross-sectional and time-series analyses are conducted.

Cross-sectionally, I find

that the magnitude of discretionary accruals, which is the proxy of earnings management, is lower in companies issuing both earnings forecasts and cash flow forecasts during the same forecasting period than it is in companies issuing only earnings forecasts.

I also

3 find significant change in the magnitude of discretionary accruals in the same company when forecast behavior is charged by starting/stopping issuing a cash flow forecast with an earnings forecast. In addition, to allay the concern of sample selection bias, I conduct a test using Heckman's selection model (Heckman 1979). In the test, I identify the determinants of management choice of issuing cash flow forecasts and examine the correlation between the determinants and the level of earnings management proxies. I don't find a significant relation between them.

Therefore, sample selection bias is not a serious concern in this

study. This dissertation contributes to the literature in two important ways.

First, it

provides evidence of how voluntary disclosure of accounting information impacts earnings management.

Second, this study enriches the literature about the consequences

of management forecasts by providing information about a specific consequence, i.e., earnings management.

In the Hirst et al. (2007) framework of management forecasts,

there are three stages.

The first stage is the incentive of management to voluntarily

issue a forecast (either earnings forecast or cash flow forecast). forecast characteristics.

Since management forecasts are a voluntary disclosure, they

involve a lot of management discretions. management forecast.

The second stage is

The third stage is the consequence of a

The majority of prior research on consequences discusses stock

market reaction to management forecasts.

Very few papers discuss how management's

behavior of issuing forecasts ex ante impacts earnings management.

This paper

contributes to the literature by providing new evidence of the impact of cash flow forecasts on management's choice to engage in earnings management.

4

The rest of this paper is organized as follows. Section 2 introduces background relating to management forecasts of earnings and cash flows and reviews prior literature. Section 3 develops hypotheses. Section 4 describes sample selection and research design. Main results are discussed in Section 5. Section 6 addresses self-selection concerns. Conclusions are in Section 7.

II. LITERATURE REVIEW Background Management forecasts have become common within companies.

Some form of

financial guidance facilitates communication between the company and investors.

A

survey on earnings guidance by the National Investor Relations Institute (NIRI 2006) reports that over 66% of companies provide some form of earnings guidance, which evidences the popularity of management forecasts.

Companies can choose to provide

only earnings guidance, or choose to provide guidance related to operating cash flows and accruals components of earnings.

Since the year 2000, the number of companies

providing both cash flows and earnings guidance has increased dramatically.

Many

would attribute that increase to Regulation FD in 2000, which requires management to provide forecasts to all parties simultaneously or curtail forecasts completely. Incentives for Voluntary Disclosure Prior studies investigate incentives for voluntary disclosure, including management forecasts.

Managers have superior information to outside investors about their

companies' expected future performance. Why do managers voluntarily provide private information to the market?

Various explanations are provided in prior studies.

Healy

and Palepu (2001) categorize them into six forces: capital market transactions, corporate control contests, stock compensation, litigation, proprietary costs, and management talent signaling. Capital market transactions.

Companies make private information available to

reduce the information asymmetry between managers and investors and to reduce the cost of external financing.

Evidence shows that the level and analyst ratings of

6 disclosure are higher when companies are preparing to issue securities (Myers and Majluf 1984, Lang and Lundholm 2000, Healy et al. 1999). Corporate control contests.

Investors and the boards of directors hold managers

accountable for companies' performance.

To avoid losing their positions, managers use

disclosures to reduce undervaluation and to explain away poor earnings performance (Warner et al. 1988, DeAngelo 1988). Stock compensation.

Healy and Palepu (2001) suggest that managers engaging in

insider trading would use disclosure of private information to reduce the undervaluation of the firm's stock and increase stock liquidity.

They also suggest that voluntary

disclosure could reduce contracting costs associated with stock compensation. Litigation.

Managers with bad earnings news tend to pre-disclose information to

reduce the risk of litigation because it spreads the stock price shock over multiple dates (Skinner 1994, Kasznik and Lev 1995). Proprietary costs.

Companies have concerns that the disclosure of private

information may reduce their competitive position in product markets.

These concerns

would reduce the desire for voluntary disclosure (Verrechia 1983, Feltham and Xie 1992). Management talent signaling.

Trueman (1986) suggests that managers provide

private information to signal their ability to recognize changes as they arise in the future. Consequences of Voluntary Disclosure The majority of prior studies on voluntary disclosure consequences focus on capital market reactions.

The increase of both quantity and quality of disclosed information

will be rewarded by the stock market.

Healy et al. (1999) find that companies with

7 higher levels of disclosure experience significant increases in stock prices that are unrelated to current earnings performance. Gelb and Zarowin (2000) find that companies with higher disclosure ratings will increase the speed with which information gets into price.

Several studies investigate the relationship between liquidity and voluntary

disclosure. They find a significant relationship between analysts' ratings of disclosure and companies' bid-ask spreads (Welker 1995, Healy et al. 1999). The increase in quantity and quality of disclosure information can help companies reduce information risk and, therefore, reduce the cost of capital.

Botosan (1997) finds

a negative relationship between the cost of equity capital and the extent of companies' voluntary disclosure.

Botosan and Plumlee (2002) extend that study and show a

negative cross-sectional relationship between cost of capital and analysts' ranking of annual report disclosures. Besides the studies of market reactions, other studies provide evidence of analysts and investors' behavior as consequences of voluntary disclosure.

Jennings (1987)

suggests that analysts update their forecasts in response to the release of firms' earnings forecasts.

And when a company increases the quantity of its disclosure of earnings

information, the number of analysts following this company increases as well (Graham et al. 2005, Wang 2006). W illiams (1996) finds that analysts will revise their forecasts more often when companies provide more accurate information.

Evidence shows that

increased disclosure by companies will be associated with increased investment in the companies' stocks (Diamond and Verrecchia 1991).

Healy et al. (1999) find the

increase in companies' disclosure leads to increased institutional ownership.

8 Management Forecasts and Earnings Management Prior literature points out that forecasts are issued for a variety of reasons and obtain a variety of consequences.

Recall that the Hirst et al. (2007) framework involves three

sequential components of management forecasts: antecedents, characteristics, and consequences.

Aspects of the first component affect the second component. And

aspects of the second component in turn affect the third component.

For example,

environmental fore cast characteristics, such as legal and regulatory change, and the institutional environment affect forecast characteristics like forecast form and forecast disaggregation.

And forecast characteristics affect forecast consequences. Management

forecast consequences, the third component, include a variety of outcomes associated with the issuance of a forecast.

The common forecast consequences are stock market

reaction, cost of capital, earnings management, litigation risk, analyst and investor behavior, and reputation for accuracy and transparency. Some forecast consequences are reactions from outsiders like investors or analysts. Prior studies find that management earnings forecasts indeed have information content as they influence stock prices (Penman 1980). Coller and Yohn (1997) also indicate that information asymmetry (proxied as bid-ask spread) is reduced as a consequence of management earnings forecasts. Many studies find that analysts update their forecasts after the issuance of management earnings forecasts (Cotter et al. 2006). Other consequences are directly related to managers' behaviors.

Cheng et al. (2005)

find that companies issuing management earnings forecasts regularly tend to invest more in short-sighted business options and sacrifice their long-term growth to smooth current earnings.

9 Managers have substantial discretion over the magnitude of realized earnings. To influence investors' and analysts' decisions, management can control the news release of realized earnings.

Prior research examines whether management forecasts induce

subsequent earnings management to meet self-imposed earnings targets.

If managers

try to meet or beat expectations, they can do it in two ways: (1) managing earnings, and (2) managing forecasts. Matsumoto (2002) examines both mechanisms of avoiding a negative earnings surprise.

She finds that both abnormal accruals and lower management expected

forecasts play important roles in avoiding negative earnings surprises. studies provide evidence of the use of each method.

Some other

Kasznik (1999) shows that

managers who issue earnings forecasts use discretionary accruals to meet or beat their own forecasts.

Ke and Yu (2006) indicate that managers provide pessimistic forecasts

to lower analyst expectations, in order to meet or beat ex post expectations. My research involves a unique setting in which managers are constrained to use one method.

When managers issue both earnings and cash flow forecasts, users can

decompose earnings into operating cash flows and accruals.

Cash flow information

helps users evaluate firm viability by providing information about solvency and liquidity. The inherent uncertainty associated with accruals increases the risk of misstatement. And cash flow information can help to reduce uncertainty and validate earnings (Defond and Hung 2003).

Thus, providing cash flow forecasts may make management more

reluctant to manage earnings. Some prior studies investigate how the decomposition of earnings forecasts (either from management or analysts) would impact earnings management.

Hirst et al. (2006)

10 indicate the disaggregation of management forecasts would decrease the incentive to manage earnings in an experimental setting. Collins and Mclnnis (2006) examine the relationship between the forms of analyst forecasts and meeting or beating an earnings expectation.

They find that analysts' cash flow forecasts will reduce accruals

manipulation and the likelihood that firms will meet or beat earnings targets.

III. HYPOTHESIS DEVELOPMENT This research is aimed at the relationship between management choice of a specific form of voluntary disclosure (cash flow forecast) and their subsequent earnings management decisions.

In order to do so, I develop hypotheses based on the following

discussion. Management Choices to Avoid Negative Earnings Surprises Managers have various incentives to avoid negative earnings surprises. These incentives include influencing capital market expectations, compensation, and other benefits related to accounting numbers. Such incentives influence managers' behavior in disclosing earnings forecasts and realized earnings.

In attempts to meet or beat

expectations created through either analyst or management forecasts, managers have several choices.

One choice is to overstate realized earnings with discretionary accruals

such as allowances and reversals (Kasznik 1999).

For managers, this choice is

relatively easy because the estimate of these accruals involves substantial management discretion.

If the reported earnings lack transparency and investors cannot detect

manipulation, the costs associated with manipulation of discretionary accruals are small as compared to the potential stock market benefit. If managers can anticipate that realized earnings will fall short of expectations, the second choice involves issuing a lower earnings forecast in advance to reduce the negative earnings surprise later (Matsumoto 2002, Ke and Yu 2006). second choice has disadvantages.

However, the

There is a time gap of one to three months between

the issuance of quarterly management forecasts and the announcement of realized earnings.

The uncertainty inherent in a company's operations does not allow managers

12 to make a very precise estimate of the next period's earnings.

They need to consider the

costs associated with the loss of credibility and negative stock price shock caused by intentionally released bad news.

Hence, it is difficult for managers to estimate the

benefits from an "understated" earnings forecast to reduce negative earnings surprises. This makes managing forecasts a hard and costly method.

Finally, managers may

choose to use both methods at the same time (Matsumoto 2002).

Of course, real

earnings management remains a possibility here. Management choices always involve assessing the tradeoff between the costs and benefits associated with each method.

If a choice involves large costs and small

benefits, rational managers will certainly avoid it.

In this study, I predict that when

managers provide both earnings and cash flow forecasts, their incentive to engage in earnings management incentives will be reduced1.

The underlying rationale is that

providing both earnings and cash flow forecasts enables users to better detect when discretionary accruals are used to meet or beat expectations.

Thus, using discretionary

accruals to manipulate earnings becomes very costly and therefore less likely to occur. Management Incentives to Issue Cash Flow Forecasts When management chooses to issue a forecast, a lot of factors may impact their decisions.

These factors, including firms' characteristics, regulation environment, and

law changes, may influence management decisions such as the form and timing of forecasts, forecast accuracy, etc.

1

Management forecasts are an important form of

In this paper, I only discuss the situation when companies disaggregate earnings into operating cash flows and accruals. Hirst et al. (2006) provide examples of more key line items and examine the impact from disaggregation of earnings forecasts on credibility of management forecasts.

13 voluntary disclosure. Accordingly, managers enjoy discretion in deciding the form, timing, and other characteristics of management forecasts. They can also decide to what extent to disclose detailed information.

The factors which may impact

management to disclose both earnings and cash flow forecasts can be categorized into external ones and internal ones. The external factors include regulation environment, law changes and peer pressure from competitors. First, Regulation FD (2000) requires management to either provide forecasts to all parties simultaneously or provide no forecasts at all. The regulation forbids providing private cash flow information to analysts exclusively. A direct consequence of this regulation was a rapid increase in the number of cash flow forecasts after its issuance in 2000. Second, peer pressure is another reason companies issue cash flow forecasts. The National Investors Research Institute surveyed 1,000 companies and found that about 30% of companies provide cash flow forecasts because other companies in the same industry do it. On the other hand, the internal reasons which impact management to disclose cash flow forecasts are to add credibility to disclosed information, to be able to mitigate bad news in earnings, and to signal economic viability (Wasley and Wu 2006).

Cash flow forecasts may be

perceived to be more credible than earnings forecasts because of the widely held perception that it is harder to manage cash flows than earnings. Thus, companies could use the issuance of cash flow forecasts to signal the high credibility of the earnings information provided to investors. At the same time, if cash flow forecasts convey good news for the company, they can be used to mitigate bad news in expected earnings. Evidence suggests that bad news on earnings has less predictive power for a company's future performance (Hayn 1995). Thus, if cash flow forecasts convey good news while

14 earnings forecasts convey bad news, management would issue the good news with the bad news to mitigate the effect brought by its impact. Cash Flow Forecasts and Earnings Management In this dissertation, I expect that the ex ante issuance of cash flow forecasts along with earnings forecasts will restrain the management's ability to engage in ex post earnings management through accruals. Issuance of cash flow forecasts and earnings management through accruals are two separate decisions made at different times. incentives to issue cash flow forecasts.

As discussed above, managers have various However, at the same time, issuance of cash

flow forecasts also adds a self-constraint to earnings management and limits management's ability to use discretionary accruals.

Einhorn and Ziv (2008)

demonstrate that by voluntarily disclosing private information, firms make an implicit commitment to provide similar disclosures in the future.

Managers will be less willing

to voluntarily disclose cash flow forecasts in the first place if they think this constraint is too costly.

But managers still have the incentives discussed above to disclose cash flow

forecasts despite the constraint brought by this voluntary disclosure. And once managers start disclosing cash flow forecasts, they would likely keep doing so in all time periods. Prior studies provide evidence that discretionary accruals can be used as an earnings management tool (Jones 1991, Dechow et al. 1995). Managers can choose to manipulate different accruals accounts to meet or beat an earnings target.

The issuance of both

cash flow forecasts and earnings forecasts enables investors and other information users to decompose earnings into cash flows and accruals.

This decomposition actually gives

investors an estimate of accrual forecasts, because it is easy to calculate an accrual

15 forecast using forecasts of earnings and cash flows. Therefore, if managers manipulate one or more accruals accounts later, it is easier for investors to detect earnings management with the expectation from accrual forecasts than without any expectation of accruals. Sometimes, the manipulation occurs in some pretty obvious accounts. For example, managers could decrease the allowance of doubtful accounts as a percent of receivables and decrease bad debt expense. In this case, with or without the help of accrual forecasts, investors could easily detect the manipulation. However, sometimes the manipulation can be more subtle.

For example, if a company increases its

production to defer fixed cost in inventory, a higher earnings number results. For those investors without adequate knowledge in accounting, it is relatively hard to detect this manipulation. But with the help of accrual forecasts, they may notice the abnormal size of realized accruals relative to forecasted accruals. That may spur them to dig deeper, ask more questions, or turn to professionals.

And these actions will increase the

probability that earnings management will be detected. Some manipulations can get very subtle and even analysts have difficulty detecting them. Picconi (2006) finds that analysts are seemingly unaware of the earnings implications of various aspects of pension accounting, e.g., manipulation of the expected rate of return on plan assets. The issuance of cash flow forecasts along with earnings forecasts highlights the difference between forecasted accruals and realized accruals. Therefore, users may take a closer look at subtle manipulations which might be ignored if there are no accrual forecasts. Any resulting deeper digging into financial statements and questioning of management on unclear accounts may make earnings management more likely to be detected.

16 Penalties upon the Detection of Earnings Management There are various incentives for managers to engage in earnings management behavior, such as maximizing bonuses, avoiding detection of managerial shirking and meeting a market target.

However, if earnings management behavior is detected by the

market, penalties, including stock price falling and legal recourse, are often immediate. Managers engaging in earnings management will suffer significant reputational penalties when their behaviors are detected.

These reputational penalties include

management turnover and subsequent ex post settling up in the managerial labor market. Desai et al. (2006) find that a high management turnover rate will follow the revelation of aggressive accounting or fraud.

Managers of firms that restated their earnings are

significantly more likely to lose their jobs2. And the rehire rate of those managers is significantly lower than their counterparts at firms that did not restate their earnings. Firms that engage in earnings management will be penalized by the market once the earnings management behavior is detected.

Palmrose et al. (2004) find that

restatements involving fraud and reduced earnings are associated with more negative returns. They consider the extra market penalty as investors' skepticism regarding management competence and integrity.

Rational managers would be aware of these potential reputational and market penalties. They will balance the tradeoff between the benefit brought by earnings

2

Research ers and regulators consider restatements of financial statements as the detection of a previously issued financial statement containing misstatements. A large amount of restatements involve earnings. (GAO 2006, Panel of Audit Effectiveness 2000, and Palmrose et al. 2004).

17 management and the penalties associated with future detection. If they think the chance of being detected by the market is high, they will be more reluctant to engage in earnings management. Hypotheses If a company chooses to disclose both earnings and cash flow forecasts during the same forecasting period, that choice gives investors an early opportunity to decompose announced earnings into cash flows and accruals.

This early opportunity, in turn, may

trigger deeper investigation of discretionary accruals and make earnings management more detectable.

Once earnings management is detected by investors, management

faces reputational and market penalties.

To avoid those penalties, rational managers

will restrain their ex post earnings management behavior with the ex ante issuance of cash flow forecasts along with earnings forecasts. Based on the above discussion, I develop two hypotheses that address the relationship between managers' choice of issuing cash flow forecasts and subsequent earnings management.

The first hypothesis relates to cross-sectional variation in the

choice to provide both cash flow and earnings forecasts or earnings forecasts alone.

In

particular, it examines the relationship between those choices and a discretionary accruals proxy for earnings management.

This cross-sectional relationship is stated in

the following hypothesis. HI: Discretionary accruals are lower in companies that issue both earnings and cash flow forecasts than in companies that only issue earnings forecasts. For reasons discussed earlier, companies sometimes change forecasting policy. Companies can change from issuing only earnings forecasts to issuing both earnings and

18 cash flow forecasts, or they can change from issuing both forms of forecasts to issuing only an earnings forecast. The second hypothesis relates to time-series variation in a specific company's forecasting choices.

In particular, it involves the relationship between those choices

and a discretionary accruals proxy for earnings management.

I predict that a company's

discretionary accruals will decrease when it starts issuing cash flow forecasts and increase when it stops.

This time-series relationship is stated in the following

hypothesis. H2: Discretionary accruals will be lower in the periods when a company issues both earnings and cash flow forecasts than in the periods when it only issues earnings forecasts.

IV. RESEARCH DESIGN AND SAMPLE SELECTION Cross-Sectional Analysis To test Hypothesis 1,1 conduct a cross-sectional analysis. sample with two subgroups of companies.

In this analysis, I use a

One subgroup, the cash flow forecasts

sample consists of companies that issue both earnings and cash flow forecasts.

The

earnings forecasts sample consists of companies that only issue earnings forecasts.

The

purpose of cross-sectional analysis is to investigate the effect of issuing both earnings and cash flow forecasts on the level of earnings management.

Assuming other factors

which may impact earnings management remain constant, the levels of earnings management in different companies will be different when they choose between issuing cash flow forecasts with earnings forecasts or just earnings forecasts.

The specification

used in this analysis is: DA, =b0 + bxxCFFt+b2xSizet

+ b3xCFO, + b^xLEVl+b5 xMTB, +b6xROA, +et

(1)

where DA = discretionary accruals developed from Jones Model (1991); CFF = dummy variable, coded 1 if a company chooses to issue cash flow forecasts along with earnings forecasts, 0, otherwise; CFO = cash flows from operating activities; LEV- leverage; MTB = market-to-book ratio; ROA = return on assets. In this analysis, the dependent variable is discretionary accruals.

Both anecdotal

evidence and academic research confirm its popularity as a proxy for earnings

20 management (Jones 1991, DeAngelo 1986, Dechow et al. 1995).

Discretionary

accruals are first introduced as the proxy for earnings management in the Jones Model (Jones 1991).

Dechow et al. (1995) test the validity of several popular forms of

discretionary accruals proxies which enjoy popularity in empirical research.

They find

that the discretionary accruals proxies from the Jones model (Jones 1991), modified Jones model, and industry model (Dechow and Sloan 1991) are all quite accurate. Considering the similarity among the proxies and the limited sample size required by the Jones model (Jones 1991). I choose to use the discretionary accruals proxy developed from that model. The two steps in the Jones model control the effect of changes in a company's economic characteristics on nondiscretionary accruals.

In the first step, the model

estimates the nondiscretionary component as the amount of total accruals predicted by prior year assets, current revenues, and current PP&E. Here the firm-specific parameters are estimated from the following model: TA, = a, (1 / 4_,) + a2 (AREV,) + a, (PPE,) + u,

(2)

where TA = total accruals in the current year scaled by total assets in the previous year A - total assets in the previous year AREV= change in revenues in the current year scaled by total assets in the previous year 3

I also run the regressions in both hypotheses by using the modified Jones model. The results show no significant difference between the two measures. The modified Jones model requires data on net receivables. This data requirement reduces my sample from 595 observations to 427 observations. Considering the 30% decrease of sample size, I choose to report the original Jones model results. However, results are qualitatively similar.

21 PPE = property plant and equipment in the current year scaled by total assets in the previous year Then the model specifies nondiscretionary accruals (NDA) as the predicted value of TA from equation (2): NDA, = a, (1 / A,_x) + a2 (AREV,) + a3 (PPEt)

(3)

where a = firm-specific parameters derived from equation (2). The second step is to derive the discretionary accruals by subtracting NDA from TA, or: DAt=TAt-NDAt

(4)

CFF in equation (1) is a dummy variable coded 1 or 0 depending upon whether companies issue both earnings and cash flow forecasts or only earnings forecasts. Because disaggregation of earnings forecasts will give investors an earlier opportunity to decompose announced earnings into cash flows and accruals and to detect earnings management, I expect a negative relationship between CFF and DA. Cross-Sectional Analysis Using an Alternative Approach In order to further validate my results, I use an alternative approach to conduct the cross-sectional analysis.

The alternative approach uses an increasingly well-accepted

proxy for accruals quality. That proxy is substituted for the Jones model proxy.

The

specification of this approach is: AQt =bQ + bx xCFF,+b2xSIZE, +b3 xCFQ +64 xLEV,+b5 xMTB, +b6 xROA, +e, where AQ = Accruals quality.

(5)

22 The independent variables (CFF, SIZE, CFO, LEV, MTB and ROA) are defined the same as in equation (1). Similar to the earlier analysis, I expect a negative relationship between AQ and CFF, i.e., accruals quality will be higher when management issues cash flow forecasts. The measure of accruals quality is developed by Dechow and Dichev (2002) and modified in McNichols (2002) and Francis et al. (2005). This measure defines the quality of accruals as the extent to which accruals map into previous, current, and future cash flows.

It captures both biased discretionary accruals and unintentionally

mistakenly estimated accruals.

Evidence from prior studies shows accruals quality is

strongly related to earnings persistence and the quality of internal control.

The validity

tests provided by these studies also suggest a significant relationship between accruals quality and discretionary accruals derived from the Jones model (Dechow and Dichev 2002, Doyle et al. 2007). Specifically, the measure of accruals quality is derived from the following regression: AWC, = d0+d,x

CFOt_x + d2x CFOt + d3x CFOM + d,x AREV, +d5x PPEt +s,

(6)

where AWC = the change of working capital accruals; CFO = cash flows from operating activities; AREV= the change of sales; PPE = property, plant and equipments. The residuals from this regression represent the extent to which current accruals do not map into previous, current, and future cash flows. Following Doyle et al. (2007), I

23 run the above regression cross-sectionally by year.

Then I aggregate the residuals from

the regression by firm and calculate their firm-specific standard deviation. quality (AQ) is defined as that standard deviation.

Accruals

For each firm, the accruals quality

proxy is constant over the sample time period. Time-Series Analysis Companies have various reasons to change forms of management forecasts.

Prior

research indicates that regulation changes, corporate governance and pressure from competitors play important roles in the choice of forecasting form.

For a given

company, if managers start issuing cash flow forecasts with earnings forecasts, the costs of overstated earnings are increased, and that increased cost restrains the use of discretionary accruals.

So I expect to see that discretionary accruals are lower in the

periods when a given company issues both earnings and cash flow forecasts than in the periods when it only issues earnings forecasts. time-series analysis. two subsamples.

To test Hypothesis 2, I conduct a

Similar to the cross-sectional analysis, the sample also includes

One contains observations in the periods when companies issue both

earnings and cash flow forecasts.

The other contains observations in the periods when

companies only issue earnings forecasts.

The specification used in this analysis is:

DA, =d0 +d} xCFFT+d2 xSIZE+d^ xCFQ+dt xLEV + d5 xMTQ+d6 xR04 +st

(7)

where the definitions of the dependent variable (DA) and control variables (SIZE, CFO, LEV, MTB, and ROA) are the same as in cross-sectional analysis. CFFT is a dummy variable coded 1 or 0 depending upon whether a given company issues both earnings and cash flow forecasts or earnings forecasts only.

It represents the

choices made in different periods of whether to disclose cash flow forecasts with

24 earnings forecasts or not.

As in the cross-sectional analysis, a negative relationship

between CFFT and DA is expected. Control Variables Various factors may influence the magnitude of discretionary accruals.

Becker et al.

(1998) indicate that the more important of those are company size, operating cash flows, and leverage.

Accordingly, company size (SIZE), operating cash flows (CFO), and

leverage (LEV) are included as control variables in equations (1) and (5).

I also include

market-to-book ratio and return on assets as control variables based on the following discussion. Company size (SIZE) is measured as the log of the market value of equity.

It is

expected to be correlated with operating characteristics that cause large companies to have systematically larger accruals.

Operating cash flows (CFO) is measured as cash

flows from operating activities scaled by lagged total assets. CFO has been shown to vary inversely with discretionary accruals (Dechow et al. 1995). DeFond and Jiambalvo (1994) show that companies with higher debt levels have a greater incentive to use accruals to increase earnings due to closeness to debt covenant constraints.

Therefore, leverage (LEV) is incorporated as a control variable and is

measured by the ratio of total liabilities to total assets. Skinner and Sloan (2002) show that companies with higher growth prospects usually face greater pressure to meet or beat earning targets. Accordingly, market-to-book ratio (MTB) is included as an additional control variable in equations (1) and (5) as a proxy for growth opportunities. This measure of growth opportunities is expected to be positively correlated with the level of discretionary accruals.

25 Return on the assets (ROA) is chosen to be one the control variables because it is highly correlated with discretionary accruals (DA). Dechow et al. (1998) developed a model to show how earnings shocks impact accruals. Given different levels of performance, companies may engage in different levels of earnings management. Empirical evidence from prior studies suggests a positive relationship between accruals and performance measures (Healy 1996, Dechow et al. 1998, Kothari et al. 2005). Sample Selection Companies can disclose management earnings and cash flow forecasts through various channels such as conference calls, press releases and required company announcements.

In this research, I identified companies that issue management

earnings forecasts and/or cash flow forecasts by using the First Call database. All the forecasts are from year 1994 to 2003 because First Call only provides management cash flow forecasts information during that time period. In the cross-sectional analysis, I use the following steps to obtain the sample.

I first

choose all the companies issuing quarterly earnings forecasts during the years from 1994 to 2003.

This big group contains 858 companies.

Then I divide this group into two

subgroups. One contains firm-quarter observations relating to companies that issue both earnings and cash flow forecasts during the same forecasting period, which is the cash flow forecasts sample in the cross-sectional analysis.

This subsample contains 302

observations relating to 112 companies issuing both earnings and cash flow forecasts during the 1994 to 2003 time period.

The other subgroup contains firm-quarter

observations relating to companies only issuing management earnings forecasts during the same time period.

To make the subsamples more comparable in cross-sectional

26 analysis, I match the earnings forecasts sample with the cash flow forecasts sample by size and industry4.

Some firms from both subsamples are eliminated because of the

lack of available data in the Compustat database.

A final sample contains 254

observations in the cash flow forecasts sample and 241 observations in the earnings forecasts sample. In the time-series analysis, the sample consists of companies that changed their forecast disclosure policies during the sample period.

This sample consists of 2,697

firm-quarter observations relating to 112 companies during the years 1994 to 2003. Similar to the cross-sectional analysis, this group is divided into two subsamples.

One

consists of firm-quarter observations relating to quarters when a given company issued both earnings and cash flow forecasts. I term this the cash flow forecasts sample in the time-series analysis. This subsample involves observations relating to the same companies as in the cash flow forecasts sample in the cross-sectional analysis.

The

other subsample consists of firm-quarter observations relating to quarters when companies only issue earnings forecasts. I term this the earnings forecasts sample in the time-series analysis.

The final sample contains 254 observations in the cash flow

forecasts sample and 2,253 observations in the earnings forecasts sample.

4

S ize is defined as market value of equity. Industry is defined as the two-digit SIC code. For each company in the cash flow forecasts sample, there is a matching company in the earnings forecasts sample that is in the same quartile of size and with the same two-digit SIC code.

V. EMPIRICAL RESULTS Descriptive Statistics Table 2 presents descriptive statistics. cross-sectional analysis.

Panel A compares the two subsamples in the

The levels of discretionary accruals (DA) are significantly

different between the two subgroups. The mean of discretionary accruals in the cash flow forecasts sample is -0.018 while the mean in the earnings forecasts sample is 0.002. Some company characteristics vary significantly across the two subgroups.

The mean

market value (SIZE) is 10,369 million dollars in the cash flow forecasts sample and 3,537 million dollars in the earnings forecasts sample.

Cash flows from operating

activities (CFO, which is scaled by total assets) is 0.041 in the cash flow forecasts sample and 0.007 in the earnings forecasts sample. is consistent with Wasley and Wu (2006).

That operating cash flows disparity

Companies with favorable cash flows

information are more likely to issue cash flow forecasts.

And investors also want more

information about cash flows from companies with greater operating cash flows.

Other

firm characteristics such as leverage (LEV), growth prospects (MTB, CAPINT), profitability (ROA, ROE, GM) and litigation risk (Z) are similar across the two subgroups. Panel B of Table 2 compares the two subsamples in the time-series analysis. Similar to Panel A, discretionary accruals (DA) are significantly different between the cash flow forecasts sample and the earnings forecasts sample.

The mean of DA in the

cash flow forecasts sample is -0.018 while the mean of DA in the earnings forecasts sample is 0.004. over time.

Not surprisingly, firm characteristics do not show significant change

That is consistent with the observation that external factors are important

28 factors impacting a company's decision to issue cash flow forecasts. The external factors include regulation changes, law changes and peer pressure from competitors in the same industry.

The characteristics are also consistent with the argument that

companies tend to provide similar voluntary disclosure continuously once they start doing so in the first place. Cross-Sectional Analysis Table 4 reports the results from cross-sectional analysis. This analysis examines the relationship between the choice made by different companies of whether to disclose cash flow forecasts with earnings forecasts or earnings forecasts alone (CFF) and the magnitude of discretionary accruals (DA).

The coefficient of CFF is -0.015 and

significant (pO.Ol). This result supports my Hypothesis 1, i.e., companies that issue both earnings and cash flow forecasts during the same forecasting period have smaller discretionary accruals than companies that only issue earnings forecasts. All control variables are highly significant (p