Do investors fully understand the economic implications of cash flows from operations

... ission of the copyright owner Further reproduction prohibited without permission ABSTRACT Do Investors Fully Understand the Economic Implications of Cash Flows from Operations? by Mei Luo Doctor... reporting them as part of operating cash flows or financing cash flows If the same cash effects are not classified in cash from operating activities’ but in other sections of the statement of cash flows, ... eproduced with perm ission of the copyright owner Further reproduction prohibited without permission Do Investors Fully Understand the Economic Implications of Cash Flows from Operations? Copyright

Do Investors Fully Understand the Economic Implications of Cash

Flows from Operations?

by Mei Luo

B.ECON (Tsinghua University) 1998 M.S (University o f California, Berkeley) 20 0 2

A dissertation submitted in partial satisfaction o f the

requirements for the degree o f

Doctor o f Philosophy

inBusiness Administration

in theGRADUATE DIVISION

o f theUNIVERSITY OF CALIFORNIA, B E R K E L E Y

UMI Number: 3165474

Copyright 2004 by Luo, Mei

All rights reserved

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 3165474 Copyright 2005 by ProQuest Information and Learning Company All rights reserved This microform edition is protected against unauthorized copying under Title 17, United States Code

ProQuest Information and Learning Company

300 North Zeeb Road P.O Box 1346 Ann Arbor, Ml 48106-1346

R eproduced with perm ission of the copyright owner Further reproduction prohibited without perm ission.

Do Investors Fully Understand the Economic Implications of Cash

Flows from Operations?

Copyright © 2004

by Mei Luo

Doctor o f Philosophy in Business Administration

University o f California, Berkeley Professor Xiao-Jun Zhang, Chair

This dissertation investigates the forecasting ability and persistence with respect to future cash flows o f four cash components, the inclusion o f which with Cash Flows from Operating Activities have the potential to generate misleading signals about the

company’s financial picture It also examines whether market participants fully reflect the cash components’ respective implications for future cash flows Current operating cash flows play an important role in assessing future economic conditions and security values The GAAP-based rules or flexibility faced by managements for reporting operating cash flows can potentially mislead investors in their assessments Four components o f

operating cash flows are collected from fiscal years 1988-2000 for firms in the Fortune

1

R eproduced with perm ission of the copyright owner Further reproduction prohibited without perm ission.

500 index as o f 2001: (1) nonrecuning cash flows, (2) tax benefits realized from nonqualified employee stock options, (3) investment-type cash outflows - R&D expenses and cash outlays involved in restructuring activities, and (4) cash proceeds from selling or securitizing accounts receivables The dissertation documents that these operating cash components possess incremental value in predicting future cash flows over total operating cash flows and accrual components They also differ in persistence from other operating cash flows coming from companies’ core and continuing operations

Furthermore, hedge portfolios using the information in tax benefits realized prior to the year 1999, research and development expenses and transactions o f selling or securitizing accounts receivables can separately earn positive abnormal returns over the subsequent six months up to three years Their return predictive abilities persist after controlling for factors previously documented to predict returns The empirical findings indicate that the stock market may not fully appreciate future economic implications o f components o f current operating cash flows Further analysis verifies that the market mispricing is partially due to failures to fully impound the future cash flow information (not necessarily future earnings information) contained in the operating cash flow components

This dissertation is dedicated to my parents: Luo Wanxin W 7 I M and Yu Dan ^

for their eternal love and faith in me, to my husband Zheng Yi ^ for his indispensable encouragement and support, and my sister Luo Yan ^ ffe for her company

R eproduced with perm ission of the copyright owner Further reproduction prohibited without perm ission.

TABLE OF CONTENTS

1 Introduction 1

2 Motivation and Hypotheses 9

3 Research Methodology 17

3.1 T est o f Hypotheses H I (1) - H I (4) 17

3.1.1 F orecasting Im p lica tio n s 17

3.L2 Persistence of C u rre n t C ash C om ponents 21

3.2 Tests o f Hypothesis I I 22

3.2.1 Portfolio A nalysis 22

3.2.2 Regression A nalyses 23

3.3 F u tu re E arn in gs Im plications V ersus F u tu re C ash Flow Im plications

26

3.3.1 Persistence of the C ash C om ponents for F u tu re E a rn in g s 26

3.3.2 A b n o rm al R etu rn s Associated w ith F u tu re C ash N ew s 26

4 Data and Sam ples 29

4.1 Sam ple Selection and D escriptive Statistics 29

4.2 D ata Collection P ro ced u res 31

5 Empirical Results 35

5.1 F u tu re C ash Flow Im p lica tio n s 35

5.2 R esults of H ypothesis H I I 37

5 3 F u tu re E arn in g s Implications V ersus Future C ash Flow Im plicatio n s 41

Bisciissions and Sensitivity Analyses 43

6.1 A b n o rm al R etu rn s A ssociated W ith Tax Benefits F ro m Stock O ptions 4 3 6.2 Sensitivity Analyses 44

7 Conclusions and Implications 46

References 48

Appendixes 51

Figures 59

Tables 63

m

R eproduced with perm ission of the copyright owner Further reproduction prohibited without perm ission.

I would like to than my dissertation chair Xiao-Jun Zhang, for his continual encouragement and invaluable guidance that is indispensable for the development and completion o f this work, the dissertation committee members Daniel L McFadden and Maria Nondorf, for many great comments and suggestions, and especially to committee member Sunil Dutta for his support and mentoring to m y endeavors pursuing the academic career I have benefited greatly from discussions with William Beaver, Qintao Fan, Dwight Jaffee, Brett Trueman, and workshop participants at University o f

California, Berkeley and Stanford University

The generous help from m y fellow doctoral students is gratefully acknowledged Special thanks are due to Donglin Li, Haifeng You and Katherine Gunny I would also like to thank Qing Yang, Jinwen Xiao, Jennie Jiang and Yan Liu, whose friendship made

m y study at Berkeley enriching and enjoyable Finally, thanks to warm-hearted dean Campbell, faculty and staff o f the Haas School o f Business for making m y stay in Berkeley smooth

My greatest debt is to m y parents, Yu Dan and Luo Wanxin, whose love and pride in

me made the completion o f this degree possible I am also indebted to m y husband, Yi Zheng, whose constant support, help and encouragement accompanied me through the whole process I am grateful to my sister Luo Yan who grew up with me and helped me become the person I am today

1 Introduction

The neo-classical equity valuation model well establishes that the equity value equals the discounted value o f expected future free cash flows Three versions o f the valuation approaches are commonly used: free cash flows model, dividends model, and earnings model Under the premise that cash flow prediction is fundamental to assessing firm value, the investment community and academic research have studied the prediction o f future cash flows The Financial Accounting Standards Board (FASB) also indicates that

a primary objective o f financial reporting is to provide information to help investors, creditors and others assess the amount and timing o f prospective cash flows.' The ability

o f current operating cash flows to predict firms’ future cash flow performance is evident

in prior studies The general conclusion is that earnings components - aggregate cash flows from operations and accruals, have significant predictive ability for future cash flows from operations, a component o f free cash flows (e.g., Dechow et al 1998, Barth et

al 2001 and Finger 1994) Using share prices as an implicit proxy for expected future cash flows, studies have shown that current cash flows from operations has information content incremental to current earnings and accruals (e.g., W ilson 1986, 1987) These studies tend to treat each dollar o f current cash flows from operations as having the same persistence or the same predictive ability for future cash flows, ignoring the composition

o f the cash flows.^

FASB 1978, 37-39.

^ In the presence of other explanatory variables that may capture information in the cash composition, the same predictive ability is still imposed on each dollar of cash after controlling for the other variables in predicting future cash flows.

1

R eproduced with perm ission of the copyright owner Further reproduction prohibited without perm ission.

“Cash is the King”, which has been proposed as the underlying valuation metric, and often times the closeness to cash figure is used to diagnose earnings quality Operating cash flows particularly measure the firm’s ability to generate cash flows internally, and they can help investors gain important insights into a company’s core business, especially when the faith in earnings figures is shattered by many alleged accounting irregularities However, the Wall Street Journal (C l, M ay 8, 2002) has one article illustrating how the sums o f cash flow reported in financial filings are less than they appear at first blush

“Long viewed as the most reliable, least manipulated o f the financial documents filed by a company, the ‘statement o f cash flows’ has its shortcomings, too

If you think operating cash flow as reported actually gives you operating cash flow, you are kidding yourself.”

W hen certain cash flows are not generated by firms’ continuing and core operations, or are subject to management discretion or conditions beyond managem ent’s control, they would behave differently from other operating cash flows in terms o f abilities to map into

future cash flows “The Financial Numbers Game” (Mulford and Comiskey 2002) lists various problems with the GAAP-based rules and flexibility faced by management for reporting operating cash flows The inclusion o f cash effects whose recurrence is doubtful due to various economic characteristics can potentially generate misleading signals about the company’s cash generating power if investors fixate on the face value o f reported operating cash flow Among the transactions causing concerns to investors and academics are the following cash events: ( 1) nonrecurring cash flows, (2) tax benefits realized from the exercise o f nonqualified employee stock options,^ (3) investment-type

^ The Emerging Issues Task Force (EITF) Issue 00-15 requires the income tax benefits due to stock options

be classified as an operating cash flow in the cash flow statement, effective for financial statements after

cash outflows - R&D expenses and cash outlays involved in restructuring activities, and (4) cash proceeds from selling or securitizing accounts receivable.'^

To what extent investors distinguish among various cash flow components and unravel the additional information contained in reported current operating cash flows about future cash flows remains an interesting question This dissertation investigates the future economic implications o f these four components o f operating cash flows that may behave differently from other core and continuing operating cash flows and have the potential to generate misleading signals It also examines whether market participants fully incorporate the information contained in these components I perform the analyses

on firms in the Fortune 500 index as o f year 2001 The sample period spans 13 years from 1988 to 2000 Heavy scrutiny focused on mature and large firms ensures cash flow statements are an important piece o f information used to assess future prospects

I first investigate the future cash flow implications, i.e., forecasting abilities and persistence, o f each o f the four identified cash components Empirically shown to possess superior cash forecasting ability, the cash forecast model in Barth, Cram and Nelson (2001) documents the significant predictive ability o f total operating cash flow and accruals Therefore, I test the incremental predictive ability o f each cash component controlling for the free cash flow (operating cash flow minus capital expenditures) and accrual components The results show that the four cash flow components differ in

July 20, 2000 Prior to the effective date, companies have choice o f reporting them as part of operating cash

flows or financing cash flows.

If the same cash effects are not classified in ‘cash from operating activities’ but in other sections of the statement of cash flows, this study does not deem it as problematic operating cash reporting, and those firms would have zero of such operating cash components.

R eproduced with perm ission of the copyright owner Further reproduction prohibited without perm ission.

persistence from the other operating cash flows^ and possess significant incremental value in the prediction o f cash flows in the next three years (both operating cash flows and free cash flows) Speciflcally, nonrecurring cash flows lead to relatively fewer future cash flows and they are relevant for forecasting cash flows o f the following two years, in contrast with the nonrecurring (transitory) items called by the companies Tax benefits realized prior to 1999 from employee stock option exercises are a strong positive indicator o f future cash flows, but the benefits realized thereafter are less persistent and even capture negative news about multiple-year-ahead cash flows Research and development cash investment outflows (measured as negative cash flows) have a lower coefficient in predicting long-term cash flows, possibly due to offsetting o f positive cash benefits with the persistent investment cash outflows Contrary to investors’ concerns, the cash effects o f selling or securitizing accounts receivable that are subject to managerial discretion, tend to have the same likelihood o f recurrence as other operating cash flows However, firms engaging in these inherently financing transactions will systematically experience lower cash flows in the following three years than firms not using the financing vehicle, implying that firms can boost current cash flows by sacrificing future cash flows.

I investigate whether stock prices fully incorporate the components’ respective implications for future cash flows by performing portfolio return analyses over the three years subsequent to financial statement releases and conducting cross-sectional

regressions o f future retums on the cash flow components The results indicate that hedge portfolios exploiting the information in tax benefits o f stock options prior to 1999, R&D

^ The other operating cash flows generally include operating cash flows from core and continuing operations, the difference between total operating cash flows and each o f the cash components o f interest identified in this study.

expense, or transactions o f selling or securitizing part o f accounts receivables can separately earn positive abnormal retum s over the subsequent one-year, each o f the three years, or six months The cash transactions’ associations with future abnormal retums are consistent with their incremental implications for future cash flows and are robust to the inclusion o f previously documented retum predictors The cash components also fully capture the retum predictive ability o f the accmal anomaly first found in Sloan (1996) The reverse relation between tax benefits realized after 1998 and future retums seems to reflect risk changes correlated with factors documented in prior studies or the wealth transfer effects o f the stock options Additionally, the stock market fully reflects the additional information in nonrecurring cash flows.

This dissertation focuses on the prediction o f future cash flows, a theoretical valuation input, to test whether the stock market correctly reflects the cash flow components’ implications for future cash flows There is a body o f literature on m arket’s naive expectations about future eamings, with the underlying premise that earnings are used in assessing equity values Penman and Zhang (2002) suggest investors fixate on eamings and fail to differentiate eamings quality due to conservative accounting, and Sloan (1996) shows that stock retums do not fully reflect information about future eamings contained in the accrual and cash flow components.® In a supplementary test, I examine the eamings implications o f the four cash flow components Except that tax benefits o f stock options have a strong negative effect on eamings persistence for the whole sample period, the other cash flow components consistently have the same directional impact on future eamings as on future cash flows To verify that the market

* Some o f the other studies that relate market efficiency to the understanding o f certain variables’

implications for future eamings include Beaver et al (2001), Xie (2001), and Rajgopal et al (2003).

5

R eproduced with perm ission of the copyright owner Further reproduction prohibited without perm ission.

mispricing is at least partially attributable to failures to fully impound future cash flow implications, I decompose the subsequent year abnormal retums into the component arising from one-year-ahead cash flow news and the component associated with other factors, including eamings news The results are consistent with price corrections in response to the one-year-ahead cash flow news that are predictable from the various cash components I also find that the cash news-related abnormal retums are significantly associated with R&D, tax benefits o f stock options realized prior to 1999 and selling or securitizing accounts receivable, in the direction consistent w ith their future cash flow implications The significant abnormal retums over multiple future years predictable by R&D expenses indicate that price corrections are incomplete as to the realization o f short-term cash flows However, I do not exclude the possibility that some unknown risk factor not reflected in factors suggested by Fama and French (1992) accounts for the persistent significant retums.

My study makes the following contributions Foremost, this dissertation contributes

to the literature on forecasting cash flows Until recently, this literature has largely focused on comparing the ability o f eamings and operating cash flows to predict future cash flows Dechow, Kothari and Watts (1998) and Barth et al (2001) find the accraal components enhance the predictive ability M y study constitutes an attempt to investigate the incremental predictive value o f selected components o f operating cash flows over accraals and aggregate operating cash flows Second, it addresses public concerns that the reported operating cash flow from the statements o f cash flows can potentially mislead investors when the cash flow effects o f certain transactions are included The large tax benefits realized from employee stock options in the period o f high-flying stock prices do

not sustain, but those realized prior to 1999 provide a positive signal about future cash flows The common practice o f sale or securitization o f accounts receivables does not generate less persistent cash flows and it apparently sacrifices future cash flows while increasing current cash flows Third, this dissertation adds to the market efficiency literature that investigates whether accounting information is fully impounded in stock prices Different from most prior studies that assume investors fixate on eamings prediction, this study is conducted under the premise that investors also predict cash flows to assess firm value Failing to fully impound the future cash flow information contained in the current statements will result in predictable abnormal retums The results corroborate the notion that cash flow prediction is the ultimate fundamental valuation attribute Prior studies have documented a significant relation between future abnormal retums and accmals (or total cash flows) While Desai, Rajgopal and Venkatachalam (2004) asserts that the accmal anomaly can be a manifestation o f the glamour stock phenomenon, m y study potentially provides an altemative explanation in that the market fails to fully reflect the information in the various cash flow components Finally, this study provides subtle implications for standard setting bodies about cash flow reporting and the disclosure o f cash effects from significant transactions Investors as a whole do not appear to fully understand the implications o f certain cash flow components, possibly due to high costs (e.g., lack o f detailed disclosures or skill requirements) of analyzing the statement o f cash flows reported under the current disclosure regime Prior studies have documented the widespread nonarticulation in the cash flow statements and listed some factors causing the nonarticuiated differences, e.g., Bahnson et al (1996) and Drtina and Largay (1985).’ The results o f this dissertation should also be o f interest to financial

Under the indirect method of presenting the statement o f operating cash flows, net income is adjusted by

7

R eproduced with perm ission of the copyright owner Further reproduction prohibited without perm ission.

statement users, such as asset managers, creditors, and analysts, who wish to understand cash flow reporting and cash forecasting.

The remainder o f the dissertation is organized as follows The next section provides motivation and hypotheses development Section 3 explains the estimation models and research methodology Section 4 describes the samples and data Section 5 presents the results and section 6 provides further discussion and sensitivity analyses Section 7 concludes the dissertation with implications for future research

changes in current accounts that would equal the differences between the beginning and ending balances on the balance sheet This adjustment assumes that changes in a noncash current account relate an operating source or use of cash to an income statement account When current accounts change without a link to the income statement, errors occur when mechanically applying the indirect method For example, if stocks are issued to settle a significant portion o f accounts payable, a nonarticulated difference occurs when net income is adjusted by changes in the balance of accounts payable that obviously are not related to income statement.

2, Motivation and Hypotheses

Prior studies have examined the role o f current aggregate cash flows and eamings in predicting future cash flows and report mixed results about their relative predictive ability (e.g., Bowen et al 1986, Finger 1994, and Greenberg et al 1986) Barth et al (2001) extend prior work and find that eamings, when disaggregated into cash flows from operations and accmal components, perform better in predicting future cash flows than current and past cash flows or current and past aggregate eamings Using stock prices as

a measure o f information content, research has shown that the cash flow and accmal components o f eamings have incremental information content beyond eamings Some examples are W ilson (1986, 1987), Bowen et al (1987), and Raybum (1986) Other

studies examine the association with contemporaneous security retums Examples include Lipe (1986) on six eamings components, Barth et al (1999) on accmals and cash flows, and Livnat and Zarowin (1990) on operating, financing and investing cash flows The current aggregate operating cash flows have incremental value over eamings or accmals

in predicting future cash flows and its incremental information content is reflected in stock values

Reported cash flows are generally regarded as being less subject to distortion than reported eamings In recent years investors have seen outbreaks o f myriad alleged accounting irregularities, the proliferation o f pro-forma eamings and the SEC’s broadened investigations into accounting malpractices to boost eamings figures across a

R eproduced with perm ission of the copyright owner Further reproduction prohibited without perm ission.

broad range o f industries.® However, holding the cash flow statement as a beacon o f truth can also be imprudent Due to different composition characteristics, the same amount o f cash from operations may imply varying future cash prospects Cash flows sustain only when the underlying operating activity is likely to recur For example, if current cash flows comprise a large amount o f nonrecurring cash items, the current cash figure will not be a good indicator for next period’s cash flows Some studies (e.g., Barth et al 1999) find the cross-industry variation in the persistence parameter o f cash from operations, but they still impose restrictions on the components o f cash flows to have the same ability to forecast future cash flows Taking the reported cash flows at face value forsakes

information contained in the components that is relevant for predicting future cash flows.Several operating cash components may inform about future cash flows differently from the other operating cash flows B y definition, nonrecurring cash items, ex-ante, are not expected to recur in the future in the same way as other routinely generated cash Operating cash flows include some investment-type expenditures such as R&D expenses that are expected to generate cash flows over multiple future periods The lack o f

matching between the cash costs o f the investments and their benefits can result in a different predictive ability from other periodic cash outflows The cash flow effects from some significant special transactions can be either subject to managerial discretion or to conditions not directly controlled by management The items emphasized by M ulford and Comiskey (2002) and the previously cited WSJ article are the tax benefits from the

* Among the investigations by SEC into the false accounting practices are Enron Corp ’s various wrongdoings, Adelphia Communications Corp.’s participation in related party transactions (WSJ, May 17, 2002), illusory “round-trip” trades to boost revenues by energy-trading companies like Dynegy Inc., Reliant Resources Inc and CMS Energy Coip (WSJ, May 16, 2002), and Computer Associates International Inc.’s aggressive revenue-recognition policies and practices (WSJ, May 16, 2002).

exercise o f employee stock options (when reported as part o f operating cash flow) and cash proceeds from sale or securitization o f accounts receivable The following

paragraphs provide the reasoning to examine the potential incremental information contained in these four cash categories and develop the first hypothesis regarding their forecasting abilities and persistence.®

Nonrecurring cash flows: Under the indirect method o f preparing the statement o f cash flows from operating activities, the starting point is the net income Included in net income are several nonrecurring items such as Income from Discontinued Operations and Extraordinary Items, Cumulative Effects o f Change in Accounting Principle, special items like asset w rite -o f fs a n d others The associated cash effects o f these unusual or infrequent income components may possess less predictive ability for future cash flows than the other cash flows One example is cash from the operating income or loss o f discontinued operations If these items are really nonrecurring as claimed by companies

in the financial statements, they should be irrelevant for forecasting future cash flows.” However, firms have started to include more items in the special item or nonrecurring

® This is not to claim the rest of the cash flow components do not provide incremental information about future cash flows Rather, I analyze the cash components that have caught the attention of investors and academia but have not been formally examined These cash flow items are also available by inexpensive and skillful search in companies’ financial statements Extracting every interesting cash flow item would incur high costs due to nonarticulation in the operating cash flow statement under the indirect method Some other components, such as interest expense, are debated on the reasonableness to be classified as operating cash flows, which is beyond the scope o f this study This dissertation analyzes the operating cash components based on the sample-year’s effective format and classifications of cash flows.

See “Nonrecurring items ” Financial Statement Analysis, 7* Edition, Wild, Bernstein and Subramanyam (2001, P417).

" Ohlson (1999) uses the term “forecasting irrelevancy” to label the condition that transitory eamings items are forecasting irrelevant for next period’s abnormal eamings Nonrecurring (or transitory) cash flows are interpreted here as forecasting irrelevant with respect to future cash flows.

11

R eproduced with perm ission o f the copyright owner Further reproduction prohibited without perm ission.

category Gu and Chen (2003) find that both o f the ‘nonrecurring items’ analysts decide

to keep in and exclude from reported street eamings are significantly associated with future performance The cash effects involved in the nonrecurring items may well be relevant in forecasting future cash flows The altemative hypothesis regarding nonrecurring cash flows follows:

H I (1): Cash flows claimed to he nonrecurring by companies are less recurring than

other operating cash flows and are relevant in forecasting future cash flows (i.e., not nonrecurring)

Tax benefits from employee stock options: When nonqualified employee stock

options are exercised, the issuing company will receive a tax benefit that equals the tax rate times the difference between the exercise price and market price on the date options are exercised I refer to the reduction in taxes payable from this transaction as the tax benefits from employee stock options In 2000, the EITF reached a consensus and decided that such tax benefits should be included with operating cash flows This treatment is viewed as somewhat problematic cash flow reporting (Penman 2002) because there is no recognition o f the matching option expenses that produced the tax benefits In the market boom period, option exercises have created large corporate tax

This tax treatment applies in situations when options do not have a readily ascertainable fair market value

at the date of grant Most companies that issue nonqualified stock options do not have options with a readily ascertainable fair market value If the options are considered subject to substantial risk o f forfeiture, the issuing company may not receive the tax benefit at the date of exercise, rather when restrictions on the options lapse, depending on the employee’s choice o f income recognition date (correspondingly, the entitled compensation deduction is the difference between market price o f the stock on the date the restrictions lapse and the exercise price) See more details in “Concepts in Federal Taxation”, Murphy & Higgins (p 655), 2003 edition.

deductions and boosted operating cash flows, while in the recent recession period, not

as many benefits have accrued to those firms because o f falling stock prices The tax benefits are mostly influenced by employee exercise decisions and market factors, and thus are outside o f the corporation’s direct control The cash savings are likely to be more transitory than other operating cash flows, and less related to the core operations o f the firm However, large employee option exercises induced by good market conditions may reflect the strong underlying operations that can persist promisingly into future periods The likelihood o f recurrence o f tax benefits from stock options is ambiguous; however, the high levels observed during the market boom period are less likely to represent recurring cash flows Stated formally;

HI (2): Tax benefits from exercising non-qualified employee stock options differ

from other operating cash flows in their likelihood o f recurrence and the benefits realized in the market boom period are less likely to recur than those realized in the prior period

Cash expenses related to investing activities: Investment costs that are expensed as

incurred in the income statement are treated as uses o f cash in the operating section o f the statement o f cash flows Two items o f interest are research and development costs and cash expenses involved with restructuring charges Expensing research and development costs is criticized in some academic research for the mismatching o f costs and related future revenues Lev and Sougiannis (1996) show that the R&D capitalization process

Ciprianao, Collins and Hribar (2001) report that the tax savings from employee stock option deductions for the S&P 100 and the Nasdaq 100 averaged 32 percent of operating cash flows in 2000, up from 8 percent in 1997.

13

R eproduced with perm ission of the copyright owner Further reproduction prohibited without perm ission.

yields value-relevant information to investors Penman and Zhang (2002) show that R&D expensing reflects a conservative accounting process that makes the eamings less o f an indicator for future eamings Furthermore, restmcturing charges (sometimes reported as part o f special items) are typically investments to streamline a company’s operations for improving future profitability.''^ They usually impact several future years and involve substantial cash flow commitments that are included with operating cash flows Including these cash investments can skew the reported cash from operations downward Some high-growth firms may not generate positive operating cash flows in the current period as

a result o f heavy investments, however they m ay have tremendous cash inflows when those investments tum out to be successful The ability o f these investment-type cash expenditures to indicate future cash flows probably differs from that o f periodic operating cash expenses that typically do not bring future benefits

H I (3): Investment-type expenditures included with operating cash flows have a

different ability to predict future cash flows from other periodic operating cash expenses

Cash proceeds from sale or securitization o f accounts receivable: Decreases in the

balances o f operating-related assets are added when reconciling net income to cash flow from operating activities Management can exercise discretion to decrease accounts receivable b y securitizing or selling accounts receivable, which would temporarily boost current operating cash flow at the expense o f future cash flows If the amounts involved are significant, the financial health o f the firm will be distorted The Wall Street Journal

Carter (2000) presents some evidence of operating performance improvement in the long run following a restmcturing.

(May 8, 2002, C3) illustrated that TRW Inc.’s eamings dropped to $ 6 8 million in 2001 from $438 million one year ago, but operating cash flow rose to $1.49 billion from $1.15 billion largely due to the $327 million received from accounts receivable securitization

W ith selling or securitizating accounts receivable, firms discount all future cash inflows into present, mechanically resulting in lower cash inflows in the future periods than everything-else-equal firms The cash effects o f such significant isolated events may not recur because they are subject to managerial discretion Nonetheless, the accounts receivables sold or financed are generated from the continuing and core operating activities Therefore, if firms engage in these transactions regularly, the cash effects could

be recurring The hypothesis regarding the persistence o f these cash flows is stated in altemative form:

HI (4): The cash flow effects from the sale or securitization o f accounts receivable

are less recurring than other operating cash flows

The second hypothesis concems the extent to which stock prices distinguish among various cash components and reflect their respective forecasting and persistence

properties for future cash flows If the cash components are not priced in a manner consistent with their respective cash flow implications, as new information about future cash flows arrives or actual cash flows are revealed, subsequent stock prices will gradually reflect the information and predictable abnormal retums will occur

HII: Stock prices fail to fully reflect the information contained in the four

identified cash components at the time o f financial statement release Firms reporting relatively high levels o f cash components providing positive

15

R eproduced with perm ission of the copyright owner Further reproduction prohibited without perm ission.

information and firms reporting relatively low levels o f cash components providing negative in formation will experience a higher abnormal return in the subsequent periods.

3 R esearch Methodology

3.1 T est o f H ypotheses H I (1) - H I (4)3.1.1 Forecasting Im plicationsBuilding on the Dechow, Kothari and Watts (1998) model o f the accrual process to predict future operating cash flows, Barth, Cram and Nelson (2001) find that each major accrual component - change in accounts receivable, change in accounts payable, change

in inventory, depreciation and amortization, and other accruals, reflect different information relating to future cash flows, incremental to current total cash flows

Disaggregating earnings into current aggregate operating cash flows and accrual components substantially enhances the predictive ability Appendix A details the model developed in Dechow et al (1998) and Barth et al (2001) I investigate the prediction o f two definitions o f future cash flows CF; operating cash flows (OCF) as used in

previously studied cash forecasting models, and free cash flows (FCF) considered as the fundamental valuation inputs In testing the incremental predictive ability o f each current cash flow component separately, I control for the accrual components as well as current free cash flows (OCF minus capital expenditures) as opposed to operating cash flow (OCF) in Barth et al (2001), and allow different coefficients on OCF and capital expenditures.^^ It is uncertain when the future cash flows predictable by the various cash

This benchmark cash forecast model is relatively more restrictive in testing the differential implications

of cash components, because the incremental information in cash components about future cash flows will

be mitigated to the extent that the accrual components or capital expenditures are correlated with cash components For example, the change in accounts receivable is highly correlated with cash proceeds received from accounts receivable sale or securitization Alternatively, a more relaxed cash forecast model using total accruals and operating cash flows is employed as the basic benchmark model for comparative purposes: Future Cash Flows = a + piOCF + P7total accruals + Scash component + s The results (not reported) indicate the information properties of each o f the cash components are not sensitive to the choice between the two models.

17

R eproduced with perm ission o f the copyright owner Further reproduction prohibited without perm ission.

components will arise, so one-year, two-year and three-year ahead cash flows are placed

as the dependent variable Following Barth et al (2001), I deflated all variables by average total assets The firm-specific subscript i is omitted

The first estimation equation provides a basis for comparison with prior research and testing the hypotheses regarding the incremental predictive ability o f each current cash component

C F , n = a + ydi X O C F , + p i x C A P I T A L , + J3 3 x A A R + J3 4 x A I N V ,

Where n= l,2 and 3

CFt+n = future one year, two-year, three-year ahead cash flow from operations

(OCF) or free cash flows (FCF)

OCFt = total cash flow from operations measured at the fiscal year-end t

CAPITALt = total capital expenditures in fiscal year t (FCF-OCF-CAPITAL)

The accrual variables are defined the same as in Barth et al (2001): AARt, change in

accounts receivable, AINVt, change in inventory, AAPt, change in accounts payable,

DEPAMOt, depreciation and amortization expenses, and GACCt, the other accruals that

are calculated as EAR (earnings) minus (OCF+AAR+AINV-AAP-DEPAMO)

The first sub-hypothesis is tested via a pooled regression o f the following model

C F , = a + / F x O C F , + p i x CAPI T AL , + S i x N O N R E C , + f } 3 x A A R ,

+ /?4X AI NV , + P s x A A P , + p<.x D E P A M O , + ft i x O A C C , + e , (I)

NONREC is defined as nonrecurring cash flows included as part o f total operating cash

flows (OCF) The coefficient on NONREC, 5i, reflects the difference in forecasting

abilities between the component NONREC and the other operating cash flows I expect it

to be less recurring and predict §i to be <0 Because each cash component examined is

itself a component o f OCF, the total coefficient on nonrecurring cash component is

jJ\ + 8 \ + is also tested to see i f the cash flows claimed as “nonrecurring” by

companies are irrelevant for forecasting future cash flows

The second sub-hypothesis is tested in a similar manner and I allow the incremental

predictive ability o f tax benefits to differ across time periods

C F r + « = a + /?! X OCR+ /?2 X CAPITAL, + 5 2 x Yx TAXBEN + S'2 xY' x TAXBEN + S’l x F x TAXBEN,

+ /I3XAAR + J3axMXVi + /Is X AAPt + /is X DEPAMOt + /LxOACC, +St (H)

TAXBENt are tax benefits realized in fiscal year t from exercises o f nonqualified stock options Y, Y' and Y" are year indicators: Y equals 1 if in the period between 1988 and

1998, and otherwise 0; Y ' equals 1 if in year 1999 and otherwise 0; Y" equals 1 if in year

2000 and otherwise 0 The cash flow variables and accrual components are defined the same as in model (I) I expect the lower likelihood o f the tax benefits to persist in the

market boom period (i.e., from 1999 and on) than in the prior period, i.e., 82 > 5'2 and 82

>5" 2 - As explained in section 2 , 1 do not make predictions about whether the tax savings

have higher or lower persistence than other operating cash flows, thus the null hypotheses

o f 82 - 0, 8'2 =0, and S'T =0 are tested

I estimate the following equations to examine the incremental predictive value o f long-term investment cash outflows included with operating cash flows: R&D expenses and the cash portion o f restructuring charges

19

R eproduced with perm ission of the copyright owner Further reproduction prohibited without perm ission.

C F , „ = a + Px x O C F , + j 3 i x C A P I T A L , + S3X RD, ( S ' l x R E S T R , ) + ^ 3 x EAR,

+ p ^ x AI NV, + P s x A A P , + p ex D E P A M O , + p i x O A C C , + s, ^ n (III)

RDt represents research and development expenses RESTRt represents cash expenses incurred in fiscal year t associated with prior and current restracturing activities I do not

have a prediction for the sign o f coefficient 8 3, because the forecasting implications o f

investment expenses depend on when the associated cash henefits materialize and how persistent the investments are If firms make consecutively huge investments but associated benefits materialize much later, the next period cash flows can be negative,

resulting in greater persistence (positive 83).'^ Conversely, when firms decrease

investments and profits emerge out o f past successful investments, persistence is lower

and the coefficient 8 3 can be negative The accrual variable DEPAMO (depreciation and

amortization) matches the operating performance o f other long-term investments, such as purchase o f property, plant and equipment, to the costs that are not included in operating cash flows (see Barth et al 2001)

The following equation tests the likelihood o f recurrence o f cash from selling or securitizing accounts receivable, relative to other cash flows, controlling for the systematic difference in future cash flows between firms using the financing vehicle and firms not using it

CFr +«= a + X OCF, + p2X CAPITAL, +Sax ARFIN, xDUM + SU xD U M + p i x M R ,

Note the investment-type operating cash flows are a component of total operating cash flows, so are negative numbers Greater persistence than other operating cash flows results in a positive coefficient.

ARFINt measures the cash proceeds received from sale or securitization o f accounts receivable in fiscal year t DUM, a dummy variable, equals 1 when the firm engages in the sale transactions, and 0 when firms do not have the financing vehicle or firms do not utilize it The coefficient on the interaction term represents the persistence difference between the cash from selling or securitizing receivables and the other cash flows, so I

test the alternative hypothesis that 5 4 < 0 against the null 8 4=0 Because discounting all

future cash inflows to the present will mechanically result in lower cash in future periods than everything-else-equal firms, I expect the coefficient on the dummy variable DUM to

be negative: 5'4<0

3.1.2 Persistence o f Current Cash Components

The inclusion o f accrual components and capital expenditures in the forecasting equations can obscure the inferences about the persistence o f operating cash components The following equation directly investigates the persistence o f identified operating cash components with respect to future operating cash flows, relative to the rest o f operating

cash flows 8k (k=l,2,3,4) will equal zero if the components do not differ in persistence

OCF,+ „ = a + / ? i x OCF, + 8 1 X NONREC, + 8 2 x 7 x TAXBEN, + S' 2 x 7' x TAXBEN, + 5 " 2 x 7" x TAXBEN,

+ 8 3 X RD, + S' 2 X RESTRt + 8 4 x ARFIN,x DUM+8 h x DUM + £■< + « (V)

21

R eproduced with perm ission of the copyright owner Further reproduction prohibited without perm ission.

3.2 Tests o f H ypothesis II3.2.1 P ortfolio Analysis

In testing the first hypothesis HI, §k (k=l,2,3,4) reflects the incremental information

in each o f the cash components Future cash flows are increasing in the magnitude o f

cash component with a positive 5k, and are decreasing in the magnitude o f cash

components with a negative 5k- Abnormal stock returns can be earned by exploiting

investors’ inability to correctly assess the cash components’ predictive ability and persistence Each year, firms are sorted on the magnitude o f NONREC, TAXBEN, RD, RESTR, and ARFIN and assigned to ten portfolios respectively For each portfolio, the size-adjusted and market-adjusted returns are calculated by averaging individual abnormal returns in the decile for the following 3 months, 6 months, and each o f the subsequent three years, where the return cumulation period starts three months after the fiscal year-end Significance tests are conducted on the time-series mean o f hedge portfolio returns, taking long and short position in the extreme portfolios for each cash component and for each return window I also compare the abnormal returns difference between firms selling or securitizing accounts receivable and firms not doing so The size-adjusted returns o f each stock control for the size effect in stock retums (Fama and French 1992), and are calculated by measuring the buy-and-hold return (inclusive o f dividends and distributions) in excess o f the buy-and-hold return (inclusive o f dividends and distributions) on a sized-matched and equal-weighted portfolio provided by CRSP The market-adjusted retums control the overall market effects, by subtracting the buy- and-hold return on a value-weighted market portfolio o f NYSE, AMEX and NASDAQ firms

As shown in table 3, some components have a large percent o f zero observations.The usual decile assignment malces some portfolios have an unbalanced number o f observations Following Gu and Chen (2003), the non-zero observations are ranked into quartiles with zero observations maintaining their ordinal position, and I calculate the hedge portfolio retums as the abnormal retum difference between top quartile and bottom quartile portfolios One limitation o f forming portfolios every year is that the retums in each portfolio are not aligned in calendar time, and thus are not implementable

Additionally they may violate the independence assumption needed in the statistical

a n a l y s i s A sub-sample test is performed for only firms with December year-ends to compare the results

An illustration o f the cross-sectional dependence between the retums of adjacent years can be found in Zach (2002).

23

R eproduced with perm ission of the copyright owner Further reproduction prohibited without perm ission.

Ai?r H-1 = a + PxOTHERs + p i ACC, + hNONREC + XiTAXBEN + XiRD, + XiRESTR,

+ X a ARFIN, + X a DUM, + s,Ai (1)

ARt+i is abnormal retums over the subsequent year, starting 3 months after fiscal year- end t OTHERt denotes the other cash flows, i.e., OCF-NONREC-TAXBEN-RD- RESTR-ARFIN ACCt is total accraals that equal earnings (EAR) minus total operating cash flows (OCF) The other variables are defined as in cash forecast models

To corroborate the relation between the cash components and future abnormal retums, I re-estimate the regression controlling for five previously documented predictors

o f stock retums in Fama and French (1992) and Jegadeesh and Titman (1993)

AR,A^=a + PxOTHER + Pi ACC, + XiNONREC, + XiTAXBEN + XiRD, + X'lRESm + X a TAXBEN

+ X' a DUM + axBETAt + a i \n{M), + ln(R / M), + ua MOM, + s , ax (2)

BET At is historical CAMP beta, estimated from a regression o f monthly retum premium

on the CRSP equal-weighted market retum premium, using 60-month retum period ending three months after fiscal year-end o f year t.’* ln(M)t measures size, defined as natural log o f the fiscal year-end market value o f equity Book-to-market ratio ln(B/M)t takes the natural log o f the ratio o f fiscal year-end book value over market value o f equity Momentum returns, MOMt, are six-month market adjusted retums immediately preceding the calculation o f ARt+i The other variables are defined as in model (1)

At least prior 30-month retum period is required for estimating f immediately preceding the abnormal retum cumulation If there is no enough retum data for early years, I use p o f the earliest year with available 30-month estimation period to replace those Results from model (2) are the same if I delete those firm-years without enough prior retum data to estimate p.

If the stock prices fail to recognize the incremental information contained in current cash components, the components with positive signals about future cash flows will have

a positive coefficient X, and the ones with negative signals will have a negative X.

25

R eproduced with perm ission of the copyright owner Further reproduction prohibited without perm ission.

3.3 F u tu re E arn in g s Im plications V ersus Future C ash Flow Implications

3.3.1 Persistence of th e C ash C om ponents fo r F u tu re E arn in g sThe condition for earning abnormal retums is that the cash components contain additional information about future economic variables that will eventually be priced by the market and I presume that the economic variable is cash flows A number o f studies have shown that the market fails to fully reflect future earnings information contained in certain financial variables, with the premise that earnings are used to assess firm value

Ex ante, the persistence o f the identified cash components seems to be more informative about future cash flows than about future eamings For example, the tax savings from employee option exercises can be recurring if employees keep exercising options, but their predictive ability for future eamings is less obvious due to the fact that the compensation expense producing the tax benefits may not recorded as part o f eamings In addition, selling or securitizing accounts receivables can easily convert accmals into cash, holding current eamings or future eamings constant To compare the future cash flow implications with future eamings implications, the impact o f cash components on eamings persistence is estimated using the following equation

EARt +1 = a + PiOCFt + p iA C C , + 5 i x NO NRECt + ^ 2XTAXBENt + J s x RDt

+ S 'i X RESTRt + 5 AX A R FIN t x D U M + S ' a x D U M +s t Ai (3)

3.3.2 A b n o rm al R etu rn s A ssociated w ith F u tu re Cash News

This section tries to gauge how much o f the subsequent abnormal returns are attributable to the m arket’s failure to impound the future cash flow information rather than future eamings information that is contained in current cash components If price

corrections take place in response to the resolution o f short-run cash flow information, part o f abnormal retums should be associated with one-year-ahead cash surprises.'^ Assuming investors do not differentiate between cash components in forming short-term cash or eamings expectations, I use estimated residuals from the benchmark cash forecast model and the benchmark eamings prediction model to measure unexpected cash flows- UCFt+i, and unexpected eamings-UEARt+i over the next year The benchmark cash forecast model as employed in section 3.1.1 is:

CR +1 = (2+/?i X OCA+ /?2 X CAPITAL, + /?3 x M R+y?4x MNV,+ x M P,+p 6 x DEPAMO,+PixOACG+s,*\

The benchmark eamings prediction model has only accraals and cash flows as

independent variables (from Sloan 1996): EAR, + i = a + P\OCF, + PiACC, + +1. 1 also use

analysts eamings forecast errors, (EPSt+i - Et[EPSt+i])/Pt, to proxy for the unexpected eamings observing analysts’ influence in setting investors’ eamings expectations

I first decompose total abnormal retums into the retums associated with unexpected cash flows and retums associated with other factors including eamings news

KRi + 1 = yo + y\UCFt + 1 + yi UEA Rt + i + s , + \ (4)

I calculate the retum s to future cash news for each firm-year ( AR, + i) by multiplying the

UCFi,t+i by the slope coefficient f i estimated from the equation, which measures the

market reaction to cash news incremental to eamings news This decomposition accounts for the possibly high correlation between eamings and cash flows and could capture a relatively pure cash-based retum Additional results show the above definitions o f

I use future one year instead of other periods because results in table 6 and 7 show that significant abnormal retums can be eamed over the subsequent year, indicating a large part of information about cash flows or eamings contained in current cash components is revealed by that time.

27

R eproduced with perm ission of the copyright owner Further reproduction prohibited without perm ission.

unexpected eamings and cash flows result in a low correlation o f 0.2 for my sample, suggesting that this measure o f cash surprise captures different information from

e a m i n g s A significant relation between the retum s to future cash news and the current cash components would be consistent with the existence o f price corrections based on the revelation o f future cash flow information that are predictable from the respective cash components (similar to the concept used in Abarbanell and Bushee 1998)

ARt +1 == OJ + X\NONKECt + XiTAXBENt + XiRDt + X'lRESIRi + X a ARFIN i + X' a DUAA + s t + i (5)

Using alternative measures o f cash surprises and eamings surprises assuming annual cash follows a random walk model, Bowen, Burgstahler, and Daley (1986, 1987) found low correlation between the two variables and suggest their measure of unexpected cash flows captures different information from earnings that is reflected in the stock market Their studies are based on data from 1972-1981, so it would be inappropriate to make a comparison.

4o D ata and Samples

4.1 Sam ple Selection an d D escriptive StatisticsThe sample includes firms in Fortune 500 index as o f 2001 that are listed on NYSE, AMEX and NASDAQ I choose the largest corporations and most mature firms to ensure cash flow statements are an important element in the information set used by investors to assess stock values The sample period begins in fiscal 1988, when most firms started adopting SFAS 95 and ends in fiscal 2000 because the abnormal retum calculation needs

at least two subsequent years for the retum window I hand-collect cash flow components from the 10-K filings obtained from websites ofEdgarScan, EdgarOnline, and the

Lexis/Nexis database Other financial statement data and stock retum data are from CRSP and Compustat Annual Industrial and Research Files Financial service firms (SIC codes

6000-6999) are excluded because their cash flow activities are quite distinct from other industrial firms Regulated firms (SIC codes 4000-4999) are removed because their cash flows are subject to regulatory intervention and restmcturing cash expenses often times are not available In testing Hypotheses HI, samples are restricted to firm-year

observations with sufficient data to estimate the cash forecast models I—IV Retum tests

on each cash component are restricted to firm-years with full data to estimate the coiresponding cash forecast models employing a basic benchmark model as explained in footnote 15 and with available monthly stock retum data from CRSP These criteria result

in a primary sample o f 3,172 firm-year observations for estimating the cash forecast models and 3,238 firm-years for the retum tests (a total o f 290 firms) In calculating analysts’ eamings forecast errors in year t+1,1 obtain from I/B/E/S database the actual

29

R eproduced with perm ission of the copyright owner Further reproduction prohibited without perm ission.

EPSt+1, analysts eamings forecasts for year t+1 three months after fiscal year t, and the price deflator as o f three months after year t The actual EPS and forecasted EPS are split-adjusted by IBES Except the unexpected eamings measured by analysts forecast errors, all financial statement variables are measured as o f fiscal year end, and scaled by average total assets Table 1 defmes all the variables.

Table 2 provides distributional statistics for the key financial and retum variables One-year ahead free cash flows have a much lower mean than operating cash flows Nonrecurring cash flows are negatively skewed, while the distribution o f tax benefits is positively skewed Investment-related cash expenses are measured as a negative amount, with mean values o f -0.022 and -0.002 for R&D and restracturing cash outflows

ARFIN_2 shows a higher mean than ARFIN l because the former measures the total cash proceeds received Size-adjusted and market-adjusted retums have similar distributions Table 3 describes the industry composition o f the whole sample and sub­samples containing only non-zero observations o f each cash component The industry classifications are the same as ones in Barth, Beaver, and Landsman (1998) Non-zero NONREC, TAXBEN and ARFIN count only 25%, 20% and 15%, respectively, o f the whole sample For each sample, industries w ith large concentrations o f firm-year observations are Durable Manufacturers, Computers, and Retail firms Table 4 presents the Pearson (Spearman) correlation matrix Current total operating cash flows and accraal components are significantly negatively correlated, consistent w ith prior research Cash components are significantly correlated with future cash flows as well as some o f the accraal components, highlighting the importance o f controlling for accraal components in the cash forecast models There are no consistently high correlations among cash

components except that R&D and restracturing expenditures are significantly correlated

at 0.17 (Spearman rank-order) Note that ARFIN has a significant Pearson correlation o f

-0.1 with AAR This is consistent with the notion that net accounts receivable sold or

securitized results in a negative change in the accounts receivable balance Interestingly, DUM has a more negative correlation with the one-year ahead cash flows than ARFIN (- 0.14 versus -0.03), which implies that the utilization o f the fmancing facility rather than the amount o f cash effects is more meaningfiil in assessing future performance Future free cash flows and operating cash flows yield qualitatively similar correlations with other variables

4.2 Data Collection Procedures

I use D atal24 identified in Compustat as the cash portion o f “extraordinary items and discontinued operations (including cumulative effect o f accounting changes)” to measure nonrecurring cash items If more nonrecurring items are explicitly specified in the income statement or in the notes to financial statements,^' I searched for the associated cash flow effects and combined them with datal24 to measure the nonrecurring cash (NONREC).The tax benefit realized as actual cash savings in the current period^^ ideally should

be explicitly labeled in the statement o f operating cash flows However, a detailed examination o f other parts o f the financial statements is required to determine the cash

The other nonrecurring cash items are the cash portions of litigation costs, special items, unusual items, one-time costs, merger-related costs and nonrecurring charges, and they have to be explicitly specified in the income statement or notes to financial statements.

I define the actual cash savings as the reduction in taxes payable in the current year from exercising non­ qualified employee stock options The actual realized cash savings may not come from the tax deductions taken from current period’s option exercises.

31

R eproduced with perm ission o f the copyright owner Further reproduction prohibited without perm ission.