7/31/2019 SSRN-id918421

1/59

Earnings Quality at Initial Public Offerings

Ray Ball*

and Lakshmanan Shivakumar**

*Graduate School of Business

University of Chicago1101 East 58th Street

Chicago, IL 60637

Tel. (773) 834 5941ray.ball@gsb.uchicago.edu

**London Business School

Regents Park

London NW1 4SA

United Kingdom

Tel. (44) 207 262 5050lshivakumar@london.edu

First version: 11 February 2004

Current version: 14 July 2006

Acknowledgments

We are grateful for comments from Mary Barth, Alan Jagolinzer, Emre Karaoglu, S.P.

Kothari (editor), Maureen McNichols, Zvi Singer, Florin Vasvari, Ivo Welch, an anonymous

referee and seminar participants at Stanford University, University of Chicago, University ofCalifornia at Berkeley and University of Southern California. Ball gratefully acknowledges

financial support from the University of Chicago, Graduate School of Business.

7/31/2019 SSRN-id918421

2/59

Abstract

Financial reporting around the time of IPOs is consistent with listed firms reporting more

conservatively than previously as private firms, consistent with the results in Ball andShivakumar (2005). We hypothesize that IPO firms supply the higher quality financial reports

demanded by public investors, who face higher information asymmetry than private

investors. The market mechanisms for enforcing this demand include monitoring by internal

and external auditors, boards, analysts, rating agencies, the press and other parties. Oncepublic, firms are subject to greater regulatory scrutiny and penalties. From the point of

releasing the public prospectus document onwards, IPO firms face a greater threat of

shareholder litigation and regulatory action if they do not meet higher reporting standards.

The evidence is overwhelmingly in favor of this hypothesis. We show that the evidence

reported by Teoh, Welch and Wong (1998) in support of the alternative hypothesis, that IPOfirms opportunistically inflate earnings to influence the IPO price, is unreliable for a variety

of reasons. We provide cleaner evidence, from samples of U.K. and U.S. IPOs, that IPOprospectus financials are conservative by several standards. We conjecture that the types of

bias we observe in conventional estimates of discretionary accruals occur in a broad genre

of studies on earnings management around large transactions and events.

7/31/2019 SSRN-id918421

3/59

1

Earnings Quality at Initial Public Offerings

We study earnings management around the time of initial public offerings (IPOs), for

two primary reasons. One motivation is to isolate the effect on financial reporting quality of

the change in the market and regulatory environment when a firm transits from private to

public status. This motive stems from the hypothesis and evidence in Ball and Shivakumar

(2005) that, in an important dimension of financial reporting quality the use of accounting

accruals to recognize losses in a timely fashion, referred to as conditional conservatism

listed companies meet a higher reporting standard than private companies.1

This quality

difference is attributed to differences between private and public firms in both the market for

and the regulation of their financial reporting. Public-firm investors, lenders and other

financial statement users are at greater arms length from the company than private-firm

users. They therefore face higher information asymmetry, and demand higher quality

financial reporting to help resolve it. Elaborate market mechanisms have evolved to monitor

public firm financial reporting quality. These include boards, company lawyers, internal

auditors, independent auditors, analysts, rating agencies, reputation effects, the press, short-

sellers, law courts and trial lawyers. In addition, public-firm financial reporting is subject to

greater regulatory scrutiny than private-company reporting. Relative to private firms, listed

firms consequently face market and regulatory incentives to meet higher reporting standards.

Differences between the market and regulatory environments of private and public

firms provide a rare opportunity to study the effect of these variables on reporting quality.

The market and regulatory environment varies little among the public firms in a single

country, so it is difficult to observe market and regulatory effects on reporting behavior by

1 Ball and Shivakumar (2005) study U.K. companies. The higher reporting standard of public companies is

replicated in other countries by Burgstahler, Hail and Leuz (2006) and Peek, Cuijpers and Buijink (2006).

7/31/2019 SSRN-id918421

4/59

2

studying public firms alone. One solution is to study the effect of international variation in

market and regulatory regimes.2 However, it is well-known that correlated omitted variables

are a concern in this literature. Differences between private and public firms, controlling for

size and industry as in Ball and Shivakumar (2005), are comparatively free of this concern.

Firms that make initial public offerings (IPOs) provide an alternative research design

for investigating market and regulatory effects on public reporting quality. In the period

leading up to their IPOs, the economic role and hence the required quality of their

financial reporting is changing. A direct implication of the hypothesis and evidence in Ball

and Shivakumar (2005) is that IPO firms should begin to report higher quality earnings when

they change status from private to listed, but exactly when they can be expected to change

their reporting practices has not been documented to our knowledge. The omitted variables

problem is somewhat mitigated because it is the same firm in the same industry that

undergoes a transition between private and public status. 3 In one of the samples we study

[due to Teoh, Welch and Wong (1998b)], the explosive growth associated with the receipt of

IPO proceeds is a correlated variable that requires extreme care when estimating abnormal

accruals.4

Firms going public also experience substantial pre-IPO growth. In our other sample

[U.K. IPOs, described below], we have both public and private financial data for the same

firm and the same year, which finesses the growth problem in comparing public and private

reporting. Our first motivation for studying IPO financial reporting therefore is to provide a

robust test of the hypothesis that, as their IPO approaches and they encounter different market

and regulatory demands, private companies begin to increase their financial reporting quality.

2 For example: Ali and Hwang (2000), Ball, Kothari and Robin (2000), Ball, Robin and Wu (2003), Leuz

(2003), Leuz, Nanda and Wysocki (2003), Bushman, Piotroski and Smith (2004, 2006), Bushman and Piotroski

(2006), Ball, Robin and Sadka (2006), and Leuz and Oberholzer (2006).3 This design is mirrored in studies of voluntary delistings, including Leuz, Triantis and Wang (2006), who

study firms that voluntarily go dark.

7/31/2019 SSRN-id918421

5/59

3

The second motivation for studying earnings management around the time of IPOs is

that the influential research of Teoh, Welch and Wong (1998b) [hereafter, TWW] reaches

almost the opposite conclusion. TWW hypothesize (1998b, p. 1936) that managers

opportunistically make generous use of accruals to overstate earnings, in order to inflate the

price of IPO shares offered to unaware investors. In support of this argument, they document

economically and statistically significant abnormal current (i.e., working capital) accruals in

the IPO year.5

Increased earnings management at the time of an IPO is inconsistent with the

Ball and Shivakumar (2005) hypothesis that firms going public encounter market and

regulatory incentives at that time to increase their reporting quality. We question both the

hypothesis and the evidence in the TWW study.

We question the hypothesis of widespread and substantial earnings management by

IPO firms, in part because this would attract enhanced scrutiny from market monitors such as

analysts, underwriters, auditors, boards, the press and the other parties to the transaction, as

well as enhanced regulatory scrutiny.6

The resulting litigation and regulatory risk from

inflating earnings would be accentuated by the fact that earnings management can only

borrow earnings from other periods, so its effects are temporary at best: earnings inflation

causes subsequent earnings deflation. Further, poor reporting quality could increase the cost

of capital, which is particularly worrisome for firms needing external financing, and create

adverse reputational effects for firms managers, board members and auditors. Relative to

private firms, listed firms face higher expected market costs of inflating earnings (such as risk

of shareholder and lender litigation, increased cost of capital and loss of reputation), as well

4 Growth is correlated with accruals for firms in general (Fairfield, Whisenant and Yohn, 2003).5 TWW also argue that post-IPO information on average does not confirm the inflated earnings, and hence pricesrevert over time. Consistent with this hypothesis, they show that IPO-period abnormal accruals predict post-IPO

returns. We do not address stock returns in this paper.6

This point is also noted in DeAngelo et al. (1994).

7/31/2019 SSRN-id918421

6/59

4

as higher expected regulatory enforcement costs. Consequently, we are hesitant to assume

there is an unequivocal incentive to overstate earnings around IPOs.

Nor is the TWW evidence immune from suspicion. Their evidence of earnings

management is based on estimates of discretionary current accruals, which consist of

unusual changes in firms working capital assets (such as inventories, accounts receivable and

prepayments) and working capital liabilities (such as accounts payable), controlling for

normal or non discretionary changes estimated using the Jones (1991) model. The TWW

estimates of discretionary accruals are likely to be particularly unreliable and biased in

favor of apparent earnings inflation for at least five reasons.

First, we show that the TWW estimates of discretionary current accruals are too

large to be credible. For example, we report (Table 5, Panel B) there is a 600.39% increase in

Accounts Receivable for the average firm in the quartile of IPO firms classified by TWW as

having the most overstated earnings. There is no credible within-GAAP means of over-

valuing receivables by this magnitude, and the only alternative faking credit sales and

uncollected receivables by such an enormous magnitude surely would be detected when

carried out by an entire quarter of all IPO firms.

Second, TWW estimate accruals from changes in working capital reported on

successive balance sheets, which we show to be biased in favor of the earnings inflation

hypothesis, relative to accruals estimated directly from cash flow data. One reason for the

difference is that 16.7% of the firms have an acquisition or divestiture in the IPO year, which

affects post-IPO balance sheet working capital items (Hribar and Collins 2002).

Third, TWW study changes in working capital from the last balance sheet prior to the

IPO to the first balance sheet after the event. Post-IPO increases in working capital (relative

to the control model) are taken as indicating income-increasing earnings management. As

7/31/2019 SSRN-id918421

7/59

5

TWW acknowledge, this occurs too late to influence the IPO issue price. Any inflation of

pre-IPO earnings via current accruals would inflate pre-IPO working capital, not post. We

therefore have difficulty relating the TWW research design to the hypothesis that IPO price

inflation provides an incentive to inflate earnings.7

Fourth, even in cash flow data, we show there are substantial upward biases in

estimating the discretionary component of accruals in the context of IPOs, due to firms use

of IPO proceeds to adjust their working capital, frequently drastically. One motive for going

public is to relieve a resources constraint. IPO firms thus are likely to have under-invested in

receivables and inventory, conditional on sales, and using IPO proceeds to alleviate these

constraints most likely shows up in the TWW metric as positive discretionary accruals in the

IPO year. Similarly, IPO firms are likely to previously have made sub-optimal use of trade

credit and other operating liabilities. The only way this selection problem would be avoided

would be if the entire IPO proceeds were kept in cash, invested in long term assets or used to

repay long term liabilities. Any use of the proceeds to bolster working capital (other than

cash), conditional on sales, is falsely identified by TWW as income-increasing earnings

management in the IPO year.8,9

Any use of the firms new public status to increase working

capital liabilities is falsely identified as income-decreasing earnings management, and any use

of IPO proceeds to pay them off is falsely identified as income-increasing management.

Fifth, Dechow, Sloan and Sweeney (1995), Kothari, Leone and Wasley (2005) and

Ball and Shivakumar (2006) find that the Jones model, used by TWW to estimate earnings

7 In other words, the TWW research design controls for pre-IPO earnings inflation, by using the pre-IPO balance

sheet as the base for computing IPO-year discretionary accruals, so it cannot possibly address pre-IPO earnings

inflation. Other studies have made the same mistake: for example, DuCharme et al. (2004).8 For example, this helps to explain the enormous 600.39% increase in Accounts Receivable for the average

firm in the quartile of firms classified by TWW with the most overstated earnings (Table 5, Panel B).9 In addition, large transactions such as IPOs typically involve substantial expenses, including legal, accountingand investment banking fees. To the extent these expenses are unpaid and hence are accrued at the balance date,

7/31/2019 SSRN-id918421

8/59

6

management via discretionary accruals, is substantially mis-specified. The model ignores

the roles of accruals in reducing noise in earnings (Dechow 1994) and in timely loss

recognition (Ball and Shivakumar 2006).

We report two sets of evidence to distinguish between the hypotheses of opportunism

and enhanced public-firm reporting quality. We initially examine U.K. IPOs of non-financial

firms over 1992-1999. While the results do not necessarily generalize completely to non-U.K.

settings, the data permit a direct analysis of financial reporting behavior that does not rely on

discretionary accruals estimates. The evidence from the U.K. sample is that IPO firms begin

to report more conservatively (in both its conditional and unconditional senses) several years

in anticipation of an IPO. There is no evidence these firms engage in earnings management.

Overall, the IPO firms report consistently with the additional market demands and regulatory

incentives of their new public status.

We then turn to resolving the apparent inconsistency with the TWW results. We

conduct a detailed examination of the individual working capital components of the accruals

in the TWW sample. We also compare discretionary accruals as estimated by TWW from

balance sheet data with equivalent estimates from cash flow data. These analyses show that

the discretionary accrual estimates in their study contain substantial endogenous effects of the

IPO and do not constitute reliable evidence of earnings management. We also show that the

average TWW sample member with available data from financial statements issuedpriorto

the IPO (recall that TWW analyze post-IPO financials) exhibits negative current accruals as

well as conditional conservatism. This is inconsistent with opportunistic earnings inflation in

the last financial statements issued before going public.

could affect estimates of discretionary accruals. We are aware of no evidence on the existence and magnitude of

such effects.

7/31/2019 SSRN-id918421

9/59

7

We believe the results are of interest for several reasons. Because IPO firms undergo

a transition between private and public status, they offer unique insight into the enhanced

market as well as regulatory standards expected of public firm reporting. In addition, our

results cast doubt on the TWW hypothesis that earnings inflation plays a substantial role in

apparent IPO over-pricing. More generally, our results suggest caution when interpreting the

substantial literature on earnings management around the time of large transactions and other

large events, many of which can be expected to exhibit substantial endogenous working

capital changes similar to those we observe at IPOs.

The rest of the paper is structured as follows. The following section reviews the

literature on earnings management around large transactions and events. In section 2, we

examine a sample of U.K. IPOs which permits more direct tests of reporting behavior,

without relying on discretionary accruals estimates. Section 3 re-examines the TWW sample

and problems with their discretionary accrual estimates. We offer conclusions in section 4.

1. Literature on earnings management around large transactions and events.

Commencing with DeAngelo (1986), there is a substantial literature on earnings

management around the time of large transactions and events. Large transactions studied

include management buyouts (DeAngelo 1986; Perry and Williams 1994), initial public

offerings (Aharoney, Lin and Loeb 1993; Friedlan 1994; Teoh, Welch and Wong 1998b;

Teoh and Wong 2002; Xiong, Stammerjohan and Gill 2005), seasoned equity offerings

(Teoh, Welch and Wong 1998a; Rangan 1998; Shivakumar 2000; Teoh and Wong 2002),

convertible debt issuance (Urcan and Kieschnick 2006), mergers and acquisitions (Erickson

and Wang 1999; Louis 2004; Powell and Stark 2005), stock splits (Louis and Robinson

2005), and employee stock option reissues (Coles, Hertzel and Kalpathy 2006). Large events

7/31/2019 SSRN-id918421

10/59

8

studied included debt covenant violations and financial distress (DeAngelo, DeAngelo and

Skinner 1994; DeFond and Jiambalvo 1994; Sweeney 1994), SEC enforcement actions

(Dechow, Sloan and Sweeney 1996; Jiambalvo 1996), antitrust investigations (Cahan 1992),

political events (Jones 1991; Han and Wang 1998) and labor negotiations (Liberty and

Zimmerman 1986).

A typical hypothesis in this genre is that substantial earnings management occurs

around large events, because managers then have unusually strong incentives to influence

their reported performance. Few authors note that these settings also are characterized by

higher than usual litigation and regulatory risk from inflating earnings, and higher than usual

scrutiny by market monitors such as analysts, underwriters, auditors, boards, the press and

other parties to the transaction, as well as by regulators. Few note that litigation and

regulatory risks are accentuated by the fact that earnings management reverses over time, or

that poor reporting quality could lead to an increased cost of capital or adverse reputational

effects. While this is not true for all papers DeAngelo (1986) and DeAngelo, DeAngelo and

Skinner (1994) being exceptions this literature is notably free of countervailing arguments.

As in TWW, the evidence of earnings management offered in this literature typically

is based on estimates of discretionary accruals using variants of the Jones (1991) or

Dechow and Dichev (2002) accruals models. Like IPOs, many of these events are associated

with endogenous changes to cash flows and to working capital, and thus seem likely to create

the appearance of abnormal (discretionary) accruals, even in the absence of earnings

management. Many studies estimate accruals from balance sheet changes in working capital

(excluding cash), not by directly taking them from cash flow statements (which did not

become generally available until 1987). This procedure is well-known to introduce error and

bias (Hribar and Collins 2002). As noted above, Dechow, Sloan and Sweeney (1995),

7/31/2019 SSRN-id918421

11/59

9

Kothari, Leone and Wasley (2005) and Ball and Shivakumar (2006) find that commonly-used

accrual models are substantially mis-specified. Consequently, we believe that conventional

estimates of discretionary accruals are unreliable around large transactions and events

generally, IPOs included.

2. Financial Reporting by UK Companies Going Public

A useful feature of the UK setting is that we can obtain two sets of financial datafor

the same firms and fiscal years, but prepared at different points in time and under different

market and regulatory circumstances. One set of financials was prepared when firms were

private. The other set contains the restated financials for the same years and the same firms

that they subsequently reported in prospectuses issued in contemplation of going public.

The difference between the two sets of financials is the transition from a private to a public

market and regulatory environment for financial reporting.

Private-firm financial data are available from filings with the Companies House, the

government agency that administers all limited liability companies in the United Kingdom.

Ball and Shivakumar (2005) REPORT that these financials are of lower quality than those of

public firms of equivalent size and industry. These data are available because the U.K.

Companies Act requires all limited liability companies to file annual financial statements,

regardless of whether they are private or public. The other dataset is hand collected from the

prospectuses subsequently issued when the same companies were preparing to go public.

When firms go public, they are allowed to restate their private-company financials for

previous years before including them in the public prospectus document. Prospectuses

generally include financial statements for the past three years, though some young firms have

7/31/2019 SSRN-id918421

12/59

10

less than three years past data to report. When the financials are restated, this is clearly

identified in the auditors report included in the prospectus.

Thus, for a given firm and fiscal year, we have financial data that were prepared

without full knowledge of an impending IPO (i.e., the private-company financial statements

initially filed with the Companies House) and equivalent data that subsequently were

prepared with full knowledge of the IPO (i.e., financial statements included in the public

prospectus). A comparison of the financial statements from these two sources provides

evidence on how knowledge of the public offering affects managers reporting strategies.

This comparison provides direct insight into financial reporting behavior when firms go

public, without relying on discretionary accruals estimates, as in prior literature.

We expect the financial reporting of private companies to increasingly resemble

public-company reporting as their IPO approaches. We do not expect a discrete change at the

time of the IPO. It is safe to assume that, on average in our sample, the probability of a future

IPO event rose monotonically over the three years preceding its actual date. Immediately

before the event, when the probability is higher, private companies are more likely to be

adapting their financial reporting to going public than they are in previous years, when the

probability is lower. We therefore expect restatement of the private-company financial

statements subsequently included in the public prospectus to be larger in magnitude in event

year -3 or -2, than in event year -1 (i.e., than in the last fiscal year-end before the IPO).

Although the sample is from the U.K., the results are relevant to the U.S. Accounting

Principles Board Opinion20 explicitly allows IPO firms to retroactively change all financial

statements presented in the offering prospectus, so we would expect at least some restatement

in contemplation of going public in the U.S. We do not have access to the financial

7/31/2019 SSRN-id918421

13/59

11

statements prepared by US firms for private use, prior to going public, so we are unable to

study the effect of public-firm incentives on IPO reporting restatements in the U.S. context.

We study the three years prior to the IPO. We intentionally do not focus on the IPO

year, as in TWW. The motive TWW ascribe to earnings inflation is to increase the offering

price. This motive would only cause inflation in earnings reported prior to the IPO, and not in

earnings reported later in the IPO year or in subsequent years. The offering price is

determined during the IPO year, but earnings for that year are not determined until year-end,

which is too late to influence the IPO price. Thus, the TWW focus on IPO-year earnings

offers limited insight into whether IPO firms inflate earnings to influence their IPO proceeds.

For U.K. IPOs non-financial between 1990 and 1999, we compare financial

statements initially submitted as private companies to the Companies House during the three

years before the IPO with those subsequently included in their IPO prospectuses. The

objective here is to determine whether they subsequently inflate their results, or whether they

report more conservatively. We also examine whether IPO firms report earnings more

conservatively than non-IPO firms, using two accruals-based tests from the prior literature.

First, using the Ball and Shivakumar (2005, 2006) approach, we test whether IPO firms are

more conditionally conservative than non-IPO firms, in the sense that they report losses in a

timelier manner. Second, we examine abnormal accruals of IPO firms estimated from linear

and piecewise-linear versions of the Jones (1991) model, for the two years prior to the IPO.

Both of the above accruals-based tests use accruals obtained from cash flow statements,

which avoids the criticisms of balance sheet data raised by Hribar and Collins (2002).

2.1 Sample and summary statistics

All U.K. IPOs between 1992 and 1999 are identified from the Securities Data

Corporation database. The IPO prospectus data are manually collected. The financial data

7/31/2019 SSRN-id918421

14/59

12

firms reported as private companies are obtained from the March 2000 version of the FAME

database supplied byBureau Van Dijk, which covers U.K. firms filing with the Companies

House.10

The FAME database also provides financials for the control samples we use,

comprising public and private firms that did not make an IPO during the period.

Sample details are presented in Table 1, Panel A. There were 720 IPOs on the London

Stock Exchange between 1992 and 1999. We exclude offerings of non-ordinary shares and

offerings by financial firms. We also exclude privatizations of state-owned enterprises and

IPOs of non-UK firms, because their pre-IPO financials were not filed with the Companies

House. These exclusions limit the sample to 496 IPOs. Requiring both private-status FAME

data and public-status prospectus data reduces the final sample to 393 IPO firms. Accruals-

based tests require cash flow statement data, which are available only from 1995 onwards.

Panel B of Table 1 provides the annual breakdown of the 393 sample IPOs and the

private and listed non-IPO firms used as controls. The year shown is event year -1, the year

of the last financial statements reported prior to the offering (for comparability, the control

sample dates are aligned with this convention). Fiscal years are converted to calendar years

using the COMPUSTAT approach: fiscal years ending before May 31st

are classified into the

previous calendar year, while those ending on or after June 1st

are classified into the current

calendar year. The small sample in 1999 is largely due to the way fiscal years are converted

to calendar years. IPOs are clustered in the mid 1990s; for almost half the sample the event

year 1 is in 1995 or 1996.

Panel C of Table 1 presents summary statistics for the 393 IPO firms. Prior to the

IPO, the firms have average (median) total assets of 52 million (11 million) and average

(median) debt of 92% (74%) of total assets. The IPOs raise an average (median) of 47

10

The dataset, and procedures we followed to verify it, is described in Ball and Shivakumar (2005).

7/31/2019 SSRN-id918421

15/59

13

million (18 million), an approximate doubling of total assets for the average firm. Firms

likely deployment of the proceeds in working capital makes us suspicious of discretionary

accruals, especially when scaled by total assets. Not surprisingly, the IPO firms are high-

growth, with average growth in total assets of 46% in the pre-IPO year, and average growth

in sales of 57%. The (unreported) growth in sales and total assets in event year 2 is similar.

Rapid growth adds further suspicion to the reliability of discretionary accruals estimates.

2.2 Comparison of private-status and prospectus financials

Of the 393 IPO firms in the sample, 109 restate their financials for inclusion in the

prospectus. While restatements are flagged in the prospectus audit reports, the prospectuses

seldom give the details necessary to reconcile their financials with those previously reported.

However, some insight can be obtained by comparing the two financial statements. For

comparability, we exclude all firm-years for which the two financial statements differ in

either the reporting unit or the fiscal year, and require the balance sheet cash balance and the

cash flow from operations to be the same in both. This results in a loss of 140 firms in event

year 1, 198 firms in event year 2, and 245 firms in event year 3. This leaves 253 firms in

event year 1, 195 in event year 2, and 148 in event year 3.

Panels A to C of Table 2 report sample mean and median statistics for income

statement and balance sheet variables in event years 3 to 1, respectively. The table reports

the original private-company variables, their equivalents as reported in the public

prospectuses, and tests of the differences between them. In the discussion below, we focus on

the semi-parametric Wilcoxon rank test because the variables are unlikely to be normally

distributed. For completeness the table also reports the t-statistics for the unstandardized

differences and for the differences standardized by the absolute amount of the private-

company version of the variable.

7/31/2019 SSRN-id918421

16/59

14

IPO firms tend to write-down the tangible and intangible fixed assets reported

previously in year 3, before including them in the public prospectus. Eighteen firms reduce

reported intangibles, compared with only three that increase them. The difference for fixed

assets is significant only at the 10% level. A significantly larger number of firms appear to

increase their current assets as compared to those that decrease their current assets. The

overall effect is to decrease total assets by approximately 1.3%. There is little restatement of

current or long-term liabilities. The decrease in total assets is largely achieved by writing

down assets directly against retained profit (retained earnings), rather than as a charge to net

income, which is largely unchanged. Due to the average firms high pre-IPO leverage, the

average asset write-down is large relative to retained profit and shareholders funds (book

value of equity), which fall on average by 28% and 17% respectively.

The results for event year 2 are qualitatively similar to, but smaller than, those

obtained for event year 3. Intangible fixed assets, total assets, retained profit and

shareholders funds continue to be significantly lower in the prospectus than in the private-

status financials. One might expect the write-offs in event years 3 and 2 to induce a

mechanical increase in year -2 net income, due to reduced depreciation and amortization, but

this is not observed. Either new, offsetting losses are being recognized in net income, or the

assets were not being depreciated or amortized earlier. The results for event year 1 are

weaker still, with the only significant difference being for intangible assets.11

The narrowing difference between the prospectus and private-status data from year 3

to year 1 is consistent with the IPO increasing in likelihood closer to its date. By year -1,

many of the private companies presumably were aware of a high likelihood of going public,

and were adapting their financial reporting to the increased market and regulatory demands

7/31/2019 SSRN-id918421

17/59

15

on public companies. To check that this pattern is not due to differences in samples across the

three years, we repeat the analysis for a constant sample across years, and obtain results

(available upon request) that are qualitatively similar to those reported in Table 2.12

In summary, many IPO firms restate their private-status financials before including

them in the IPO prospectus and the restatements generally involve writing down or writing

off of assets, particularly subjective assets such as intangible assets. Such write-downs lower

the book value of shareholders equity, making the balance sheet more conservative (Watts

1993, 2003a,b). Differences between the prospectus data and the Companies House data

narrow as the firms move closer to the IPO date, consistent with them adapting to the

increasing likelihood of a future IPO. Although the asset write-downs in prospectus data

could be expected to increase reported net income, by reducing depreciation and amortization

expense, no such evidence is observed. In all three pre-IPO years, average net income

reported in the prospectus is not significantly different from that previously reported to the

Companies House. There is no evidence in any of the three years prior to the IPO that

managers systematically inflate reported net income.

In the following two sub-sections, we report accruals-based tests of conservatism and

earnings management. These test the robustness of our conclusions.

2.3 Conditional conservatism

We compare the conditional conservatism in the restated prospectus accruals of IPO

firms with that of UK private and public firms that did not go public during the sample

11 The above statistics are for all firms, including those that did and did not restate financials. Stronger results

are obtained from analyzing only restating firms.12 The fact that private firms are required to file their financials in the U.K. could influence their reporting. Wedo not believe this invalidates our conclusions, for three reasons. First, Ball and Shivakumar (2005) show that,

despite this fact, private firms report lower quality earnings. Second, any bias due to the filing requirement

would seem to be against our hypothesis, by making U.K. private firms more likely to report conservatively atall times, not simply when they are going public. However, we observe less conservatism in their reporting three

7/31/2019 SSRN-id918421

18/59

16

period. The test modifies the Jones (1991) model to incorporate conservatively asymmetric

accruals as in Ball and Shivakumar (2005, 2006), as follows:

, 0 1 , 2 , 3 , 4 , 5 , ,

10 , 11 , , 12 , ,

13 , , 14 , , 15 , , ,

20 , 21

*

* *

* * * *

j t j t j t j t j t j t j t

j t j t j t j t j t

j t j t j t j t j t j t j t

j t

ACC CFO Sales FASSET DCFO DCFO CFO

DPUB DPUB CFO DPUB Sales

DPUB FASSET DPUB DCFO DPUB DCFO CFO

DPVT

= + + + + +

+ + +

+ + +

+ +, , 22 , ,

23 , , 24 , , 25 , , , ,

* *

* * * *

j t j t j t j t

j t j t j t j t j t j t j t j t

DPVT CFO DPVT Sales

DPVT FASSET DPVT DCFO DPVT DCFO CFO

+

+ + + +

(1)

where the variables are as defined in Table 3. Continuous variable are trimmed by 1% at

each extreme to mitigate influential observations.

The two-part control sample against which IPO prospectus accruals are benchmarked

consists of all the non-IPO firms with available data from 1995 (when cash flow statement

data first become available on FAME) to 1999. The two parts of the control are the 3,664 and

50,659 available firm/year observations for listed and private firms respectively over the

period.13

To ensure that economically insignificant firms do not drive our results, the

regressions include only private firms with at least 1 million in beginning total assets. The

172 observations comprising the year immediately prior to the IPO (event year 1) for all the

172 IPO firms with available prospectus data then are added to this control sample, and the

pooled regression statistics are reported in the first set of columns in Table 3.

This procedure is repeated for event year -2. Separately, the 95 observations with

available prospectus data for the second year prior to the IPO (event year 2) are added to the

control sample, and the pooled regression statistics are reported in the second set of columns

in the table. The control sample is identical for both IPO event years, since both are spread

years before going public (relative to their subsequently restated prospectus financials). Third, the change inreporting as the IPO year approaches implies that prior financials were not prepared for public use.13 The control sample for event year -2 is slightly smaller, because the outlier deletion rule is applied to the

pooled sample including IPO firms, and the number of the latter is smaller than in event year -1.

7/31/2019 SSRN-id918421

19/59

17

over 1995-1999. Event year -3 is not analyzed because the prospectuses include only three

years of data, and the accruals model (1) requires one lagged observation forSales and for

the scaling variable, lagged total assets.

Each pooled regression tests whether the model parameters for public non-IPO firm

accruals over the sample period (10 through 15), and for private non-IPO firm accruals (20

through 25), differ incrementally from the equivalent parameters for accruals reported by

IPO firms in their public prospectuses (0 through 5). If the prospectus accruals are more

(less) conditionally conservative in the years prior to the IPO than those of listed firms, then

we expect 15, the incremental sensitivity of accruals to negative cash flows for listed

companies, to be negative (positive). Similarly, a negative (positive) value for25 would

indicate IPO prospectus accruals are more (less) conditionally conservative than those of

private companies.

Results from the pooled regression (1) are reported in Table 3.14

The estimated

incremental coefficients 15 onDPUBj,t*DCFOj,t*CFOj,tand 25 onDPVTj,t*DCFOj,t*CFOj,t

generally are negative, and both economically and statistically significant, consistent with the

IPO prospectus accruals for the two prior years being more conditionally conservative than

public and private firms accruals. The large incremental coefficients 25 (-0.87 and -1.28 for

event years -1 and -2 respectively) indicate substantially more conditional conservatism in the

IPO-prospectus data than in private firm financials, consistent with greater market and

regulatory demand for timely loss recognition upon going public. However, the similarly

large and significant 15 coefficients (-0.60 and -1.02) for the comparison with public firms

14 The results are qualitatively insensitive to standardizing the intercept and the dummy variables (DCFO, DPUB

and DPVT) by beginning total assets, and to adding size as an independent variable (defined as beginning total

assets).

7/31/2019 SSRN-id918421

20/59

18

are not expected. One possibility is that there is such greater scrutiny of IPO firms at the time

of their public offering that they even are conservative by normal public firm standards.

The coefficients 10 onDPUBj,tand 20 onDPVTj,t,which capture any incremental

accruals of IPO firms that are not explained by the economic variables in the model, are not

significant. If IPO firms were engaging in earnings inflation, one would expect the

incremental intercepts 10 and 20 for listed and private firms to be large and significantly

negative. We conclude that during the two years prior to their IPO, firms going public exhibit

accruals that are more conditionally conservative than both public and private firms over the

period and that there is little evidence to suggest earnings overstatements in these years.

2.4 Analysis of discretionary accruals in IPO prospectuses

Our final analysis of the U.K. data investigates the sign and magnitude of IPO firms

estimated abnormal (discretionary) accruals. Normal accruals are estimated by fitting

accruals models to all the non-IPO public firms in the same industry as the IPO firm.

Abnormal accruals then are estimated by applying the fitted model parameters to the IPO

firm. The accruals models used are the Jones model and a piecewise linear variant suggested

by Ball and Shivakumar (2006):

t,it,it,i5jt,i4jt,i3jt,i2jt,i1j0jt,i CFO*DCFODCFOCFOFASSETSalesACC ++++++= (2)

Model parameters are estimated separately for each IPO firmj from a cross-section of all the

non-IPO listed firms i in its 2-digit SIC with data for the contemporaneous yeart. Only

industry-years with at least 10 observations are considered.ACCi,tis total accruals for IPO

firm i in yeart, CFOi,tis cash flow from operations from cash flow statements, SALESj,tis

change in sales andFASSETj,t is book value of fixed assets. The above variables are

standardized by beginning total assets. DCFOj,t is a dummy indicator for negative cash flows

7/31/2019 SSRN-id918421

21/59

19

that takes the value 1 ifCFOj,t

7/31/2019 SSRN-id918421

22/59

20

assets and liabilities) accruals in the IPO year. They estimate discretionary current accruals

as the difference between the actual accruals and a control for non-discretionary accruals

estimated out-of-sample from balance sheet data, using the Jones (1991) model:

, 0 1 , ,

, 1

1i t j j i t i t

i t

CA SalesTA

= + +

(4)

where CAi,t is current accruals for firm i in yeart, defined as (Compustat annual data items

are in parentheses): [Accounts Receivable (2) + Inventory (3) + Other Current Assets (68)]

[Accounts Payable (70) + Tax Payable (71) + Other Current Liabilities (72)]. Sales is

change in Revenues (12). All accounting variables are standardized by TAi,t-1, Total Assets (6)

at the beginning of period t. For each IPO firmj, the parameters 0j and 1j are estimated out-

of-sample from a cross-section of all its two-digit SIC code peers that did not issue equity in

the year. Discretionary current accrualsDCAj,tfor firmj that IPOs in yeartare calculated as:

( ) ( ), , 0 , 1 1 , , 1/j t j t j j t j j t j tDCA CA TA Sales AR = + (5)

where ARj,tis change in accounts receivable for firmj in yeart, standardized by total assets

at the beginning oft. Although this approach to estimating discretionary accruals is popular,

there are several problems with applying it in an IPO year that we discuss below.17

3.1 Credibility of discretionary accruals magnitudes

TWW classify IPO firms into quartiles based on their signed DCA estimates for the

year that contains the IPO. The financial reporting of the quartile with the smallest DCA

estimates is labeled conservative. The largest DCA quartile is labeled aggressive

reporters, with the most overstated earnings. In their Table II, Panel B, TWW report the mean

17 TWW also document that their estimated discretionary accruals predict future returns of IPO firms. We do not

examine this issue, because it is not unique to IPO firms and has been widely studied in the more general context

of the Sloan (1996) accruals anomaly, that accruals are negatively related to future returns. Xie (2000)

concludes this is primarily due to the discretionary component of accruals. Fairfield et al. (2003) and Desai et

7/31/2019 SSRN-id918421

23/59

21

and median DCAs for each of the earnings management quartiles (henceforth EM quartiles).

Firms in EM quartile 1 have mean (median) discretionary current accruals of -24.33% (-

14.93%) of beginning total assets, while firms in EM quartile 4 have mean discretionary

current accruals of +53.92% (+39.76%) of beginning total assets.

The magnitudes of the discretionary accruals estimates seem too large to credibly

represent earnings management, particularly in the aggressive reporting quartile. For one

quarter of all IPO firms to inflate currentassets by an average of more than 50% oftotal

assets would require a proportionally very large over-valuation of the firms actual working

capital assets, such as inventories and receivables, or faking the existence of a proportionally

very large quantity of such assets. Furthermore, working capital assets such as inventories

and receivables are among the more easily verifiable assets on the balance sheet, in terms of

both their existence and their GAAP valuation, so such a large discrepancy would seem

difficult for one quarter of all IPO firms to disguise from their internal and external auditors.

Finally, earnings management of more than 50% of total assets is an order of magnitude

larger than the population ROA (earnings before interest, as a percent of total assets), and

would almost certainly be identified by financial analysts, if not by nave investors.

Moreover, if the estimated discretionary accruals represent earnings management,

then it is unclear why the 25% of all IPO firms in EM quartile 1 would understate their

earnings (and hence reduce their ROA) by an average of almost 25% of total assets. Here too,

it does not seem credible that such an extreme outcome, if it occurred, would go unnoticed.

We therefore conclude that the magnitude of the average TWW estimate of

discretionary current accruals in the aggressive reporting quartile is too large to credibly

al. (2004) conclude the anomaly is a manifestation of the well-known growth-value anomaly, not of investors

failing to undo earnings management.

7/31/2019 SSRN-id918421

24/59

22

represent earnings management. In the following sections, we describe reasons why the DCA

estimates are unreliable.

3.2 Errors and bias in estimating discretionary current accruals in IPO years

There are several sources of error and bias in conventional estimates of discretionary

current accruals when used in IPO years. Errors arise from endogenous, IPO-related changes

in working capital that are not associated with earnings management. Any use of the proceeds

to bolster working capital (other than cash) is likely to be falsely identified as income-

increasing earnings management. For example, if a firm invests IPO proceeds in accounts

receivable, inventory or any other non-cash current asset during the IPO year, then its current

accruals would appear to be positive. This error occurs in accruals calculated from successive

balance sheets, because a non-cash working capital asset increases. But the error also occurs

in accruals calculated from cash flow statements, because operating cash flow is reduced

relative to earnings. Either way, there is an appearance of income-increasing current accruals.

Current accruals of this sort have nothing to do with managerial manipulation but merely

reflect the firms decision to invest some of the IPO proceeds in operating activities.

Similarly, any use by firms of their new public status to increase operating liabilities is likely

to be falsely identified as income-decreasing earnings management, and any use of IPO

proceeds to pay them off is likely to be falsely identified as income-increasing management.

The only circumstances in which no such errors arise would be if the entire IPO

proceeds were kept in cash, invested in long-term assets or used to repay non-operating

liabilities but not used to alter the firms net investment in non-cash working capital. The

likelihood of this occurring around any large transaction or event and especially around the

time of going public and receiving IPO proceeds seems remote.

7/31/2019 SSRN-id918421

25/59

23

Bias arises in the IPO context because the endogenous, IPO-related changes in

working capital on average are likely to be positive. In addition to relieving a prior resource

constraint by receiving the IPO proceeds, IPO firms tend to be high-growth. Relative to other

listed firms, their changes in inventories and receivables are likely to be larger, even in the

absence of any earnings management. The resource constraint and growth effects likely are

compounded. If IPO firms are limited in growth by financial constraints in the pre-IPO

period, then the post-IPO growth in inventories and receivables (even conditional on changes

in sales) would accelerate as the financial constraints are loosened.18

The use of IPO proceeds to adjust working capital is consistent with the magnitude of

the TWW discretionary current accruals estimates, described above. The magnitude of the

DCA estimates for the aggressive quartile is consistent with these firms investing IPO

proceeds in working capital. We therefore investigate whether the firms in TWWs highest

EM quartile tend to have the highest growth in total assets. We also examine each

component of IPO firms working capital changes, to get closer to the source of the large

DCA estimates reported in TWW.

We examine the growth in total assets and changes in working capital components for

the TWW sample of 1649 firms going public between 1980 and 1992.19

We lose 13 firms we

cannot match with the 2005 version of COMPUSTAT, based on the matches provided in the

merged CRSP-COMPUSTAT database. We lose two additional firms for which

18 One might argue that this effect of growth on current accruals will not affect discretionary current accrual

estimates, which control for the relation between current accruals and changes in sales. However, TWW control

for changes in sales minus changes in accounts receivable, which also will grow with any investment of IPO

proceeds in operations. Subtracting changes in accounts receivable ensures a mechanical effect of growth ondiscretionary accrual estimates.19 TWW kindly make their dataset available at http://www.afajof.org/Pdfs/datasets/ms4367.html. It includes

individual- firm DCA estimates and earnings management quartile classifications.

7/31/2019 SSRN-id918421

26/59

24

COMPUSTAT data is not available in the IPO year.20

This reduces the TWW sample we

analyze to1634 firms.21

Panel A of Table 5 reports mean and median IPO-year changes in total assets and in

components of working capital items for the TWW IPO sample, sorted into EM quartiles.

Firms in the highest EM quartile increase their mean total assets by an enormous 275% in the

IPO year (their median increase is 167%). This is larger than the increase for any of the

remaining three EM quartiles.22

The null hypothesis that the average growth in assets for the

highest EM quartile is the same as the averages for the other three quartiles is rejected at the

1% significance level.

The magnitudes of the changes in the individual components of working capital also

are revealing. Firms in the highest EM quartile increased their accounts receivables in the

IPO year on average by 47.57% of pre-IPO total assets. Their average pre-IPO accounts

receivables were 30.44% of total assets. Similarly, the inventories of these aggressive firms

increased on average by 31.86% of pre-IPO total assets, even though inventory was only

24.32% of pre-IPO total assets. These IPO-year increases in inventories and receivables

appear in the financial statements as current (working capital) accruals. The TWW procedure

would interpret these increases in working capital as income-increasing accruals. They would

be classified as discretionary current accruals unless they were accompanied by

proportionally enormous IPO-year increases in sales.23

We do not find it credible that,

20 For 65 firms, some of the data needed to compute current accruals from balance sheet changes are reported as

combined data items (SAS missing code C in Wharton Research Database Service). We set these missingdata items to 0.0.21 The magnitude of discretionary accruals for the 1634 IPOs that we are able to match with COMPUSTAT are

very similar to those for the entire sample of 1649 IPOs in the TWW database.22 Stronger results are obtained when non-cash current assets (in particular, accounts receivable, inventory and

other current assets) are excluded. On average, firms in the highest quartile increase their total assets other thannon-cash current assts by nearly 650%, the highest of all DCA quartiles.23 Strictly speaking, this would require proportional increases in IPO-yearcash sales (i.e., sales less accounts

receivable), which is the growth variable controlled for in the Jones current accruals model (4).

7/31/2019 SSRN-id918421

27/59

25

through aggressive accounting policies, this quartile of firms could inflate their receivables

and inventories by such magnitudes.

Consistent with this belief, we note that the magnitude of discretionary working

capital (current) accruals computed by TWW exceeds 100% of total assets for 44 of the firms

in their sample. For this to constitute earnings management, an enormous proportional over-

statement of current assets and/or under-statement of current liabilities would be required. It

is counter-intuitive that an overstatement of both earnings and working capital of this

magnitude occurs and goes undetected by auditors, boards, analysts, investors, the press, trial

lawyers, regulators and other monitors.

Further evidence is provided in Panel B of Table 5, which reports the mean and

median growth rates in individual components of working capital in the IPO year. Growth

rates are calculated only for firms with a non-zero pre-IPO beginning balance. The percent of

firms with a zero beginning balance also is reported. The firms classified as the most

aggressive reporters are those with the largest IPO-year increases in working capital assets.

On average, accounts receivable for firms in EM quartile 4 grow in the IPO year by 600%.

Their median growth rate is over 100%. Their average inventories grow by 160% (median

99%) and average other current assets grow by 333% (median 133%). The enormous growth

in working capital assets for EM quartile 4 seems more likely to arise from these firms

investing IPO proceeds in working capital assets than from aggressive financial reporting.

The growth in working capital assets is considerably smaller for the other EM

quartiles, but substantial nevertheless. Working capital liabilities for EM quartile 1 increase

substantially: average increases are 159% for accounts payable, 1090% for taxes payable and

340% for other current liabilities. These increases are substantial in magnitude and generally

7/31/2019 SSRN-id918421

28/59

26

larger than for the other quartiles, suggesting that firms classified by TWW as conservative

reporters are more likely to be funding their growth in part through short-term borrowings.

Finally, we note from Table 1 Panel B that approximately 3% of firms classified in

EM quartile 4 have zero opening balances in the IPO year for individual working capital

assets. If their increases in working capital assets are due to earnings management and can

contribute to them being labeled as aggressive reporters, then these firms must be able to

create earnings out of nothing.

The magnitude of the increases in working capital assets, particularly for EM quartile

4, seem too large to be credibly interpreted as evidence of aggressive accounting policies.

Increases in working capital items of such large magnitudes would require fraud on a large

scale, and would be easily detectable. We believe these large increases in working capital

assets are more consistent with firms investing their IPO proceeds in working capital assets

than with them inflating their earnings by overstating assets.

Panel C of Table 5 reports individual components of current accruals for the ten firms

on each extreme of TWWs discretionary accruals measure. Most of these firms exhibit an

unusually large change in a current asset or current liability. Firms classified by TWW as

extremely aggressive earnings managers mainly are firms that aggressively grow their

current assets in the IPO year, possibly due to receiving the IPO proceeds. In the firms with

the two largest positive estimated discretionary accruals (permnos 78018 and 77245),

inventory increased by 302% and 425% of total assets respectively. The classification of

such firms as aggressive inflators, rather than as aggressively growing firms, seems

inappropriate. It is implausible that IPO firms fake the existence of (or over-value) inventory

by this magnitude.

7/31/2019 SSRN-id918421

29/59

27

Panel C also shows that firms classified by TWW as extremely conservative

earnings managers typically are firms that aggressively grow their financing via current

liabilities in the IPO year, possibly due to an IPO-enhanced credit rating. In the firm with

their largest negative estimate DCA (permno 77913), accounts payable and other current

liabilities increased by 169% and 114% of total assets, respectively.

Another source of extreme values in the TWW discretionary accruals estimates is low

values of the deflator, pre-IPO total assets. This helps explain the large values for expected

or non discretionary accruals estimated from deflated variables using the Jones model. For

example, permno 76259, which has the second most negative DCA in the TWW sample, has

current accruals totaling 31.7% of total assets. But the discretionary component computed

by TWW for this firm is 211%, implying expected working capital accruals of 242.7% of

total assets.

As an alternative test for whether the 20 firms with extreme DCA actually report

extremely aggressively or extremely conservatively, we track some adverse post-IPO events

that could be triggered by extreme earnings inflation. We search press coverage during the

five years after the IPO for the following four events: Forced senior management turnover or

senior management resignations on account of poor firm performance; Management releases

about earnings disappointments; Financial restatements; and Litigation against the firm, its

management or directors on financial reporting issues. If the firms with extreme DCA are

actually firms that have heavily manipulated their earnings, then we expect these events to

occur more frequently for the aggressive reporters than for conservative reporters. Our search

results do not support this. Among the 10 firms with extreme negative DCA (conservative

reporters), there are 2 subsequent occurrences of senior management turnover, 4 occurrences

of earnings disappointment releases and 2 occurrences of financial restatements. The

7/31/2019 SSRN-id918421

30/59

28

corresponding number of subsequent adverse events for the 10 firms with extremely positive

DCA (aggressive reporters) are slightly lower. There are zero occurrences of subsequent

senior management turnover, 3 occurrences of earnings disappointment releases and 1

occurrence of financial restatement. In none of the 20 firms with extreme DCA is there

litigation during the following five years relating to financial reporting issues.

3.3 Errors in estimating discretionary current accruals from balance sheet data

Hribar and Collins (2002) show that current (working capital) accruals are biased

when estimated from changes in balance sheet data. The bias is larger around major financing

events because these firms tend to have acquisitions or divestitures that affect the numbers in

consecutive balance sheets. In the TWW IPO sample, COMPUSTAT identifies (in annual

footnote 1) 16.7% of the firms as having an acquisition or divestiture in the IPO year.

To gain appreciation of the amount of error in the TWW IPO sample, we compare

estimates of discretionary current accruals using balance sheet changes data with that

estimated using cash flow statement data. Because cash flow statement data are available

only from 1987, only 478 firms of the sample firms have cash flow statement data available

to compute current accruals. Of these, 462 firms meet the additional data requirements

needed to estimate discretionary current accruals. For these IPOs, we initially replicate the

TWW estimates of discretionary current accruals using balance sheet changes.24

We then re-

estimate them using current accruals taken directly from cash flow statement data. A

comparison of the two estimates provides insight into the error introduced by using balance

sheet changes.

In unreported results, we find that the correlation between the DCA estimates reported

in TWW database and those obtained in our replication (using current accruals from balance

7/31/2019 SSRN-id918421

31/59

29

sheet changes) is 0.84. This correlation drops to 0.23 when discretionary current accruals are

estimated from cash flow statement data, indicating that discretionary current accruals

estimated using balance sheet changes contain substantial noise. Moreover, if stocks were

sorted into EM quartiles based on discretionary accrual estimates of the 462 firms only, then

for 177 firms the quartile classification would vary depending on whether current accruals are

measured from cash flow statement data or from balance sheet changes. In other words, for

over a third of the sample, the earnings management classification depends on the source of

the data for current accruals.

Table 6 reports the mean and median of the discretionary current accruals estimated

by us using data from balance sheet changes in column (1) and re-estimated using cash flow

statement data in column (2). Differences between the two estimates are reported in column

(3). For both extreme quartiles (conservative and aggressive), the mean discretionary

current accrual is significantly lower in magnitude when estimated using cash flow statement

data instead of balance sheet data. For firms classified as the most conservative reporters, the

mean discretionary current accrual estimated from balance sheet changes is 9% of total assets

less than when estimated from cash flow statement data. For firms classified as the most

aggressive reporters, balance sheet estimates of accruals are over-stated relative to cash flow

estimates by 7% of total assets. The differences in median discretionary accruals are of the

same sign, but lower in magnitude, and are significant for only the lowest earnings

management quartile.

The differences between balance sheet and cash flow estimates of accruals suggest

that the extreme EM quartiles in TWW are more likely to include firms engaging in

comparatively large IPO-year transactions such as acquisitions and divestitures. As noted by

24

In regressions estimating Equation (4), extreme observations are trimmed at the 1% and 99% levels.

7/31/2019 SSRN-id918421

32/59

30

Hribar and Collins (2002), these transactions affect balance-sheet working capital accounts,

but are not reported as accruals in cash flow statements.

3.4 Bias in applying the Jones model fitted to out-of-sample data

To estimate discretionary accruals for IPO firms, TWW first fit the Jones model of

non-discretionary accruals (4) to data for non-IPO firms from 1980 to 1993. They then apply

the fitted model to non-IPO firms in order to separate their accruals into discretionary and

non discretionary components, as in (5). This out-of-sample estimation procedure assumes

that non-discretionary accruals in IPO firms are determined in the same way as in non-IPO

firms, which seems unlikely. The sensitivity of accruals to changes in revenue is likely to

depend on a firms stage in the life cycle. In addition, IPO firms can be expected to behave

differentially during the IPO year: they are likely to have been resource-constrained prior to

the IPO and thus to have under-invested in working capital assets such as inventories and

receivables, and to have over-utilized trade credit, conditional on their sales. The IPO

proceeds relax such resource constraints, and hence the IPO year is likely to exhibit

comparatively large increases in working capital, both unconditionally and conditionally on

sales. This suggests that both the intercept and the coefficient on change in sales in the Jones

model are likely to be higher for IPO firms than for non-IPO firms, even in the absence of

earnings management.

To test this, we re-estimate the Jones model using data for all firms over the same

period (1980 to 1993), but including the IPO firms in the sample and allowing the coefficient

on change in revenues to vary between IPO and non-IPO firms.25

The IPO firms are included

during their IPO year only, to focus on the effect of that event on accruals. The accrual model

estimated is:

7/31/2019 SSRN-id918421

33/59

31

, 0 1 , 2 , 3 , , ,*j t j t j t j t j t j tCA Sales DIPO DIPO Sales = + + + + (6)

where CAj,tis working capital (current) accruals, estimated either from balance sheet

changes or from cash flow statement data,DIPOj,t is a dummy variable to identify an IPO

firm, and Salesj,t is the change in revenues in yeart, scaled by beginning total assets. The

number of IPO years added to the dataset is 1461 when balance sheet data are used and 431

when accruals are taken directly from cash flow statements. We delete the extreme 1% of

observations for current accruals and forSalesj,t. We estimate the above regression

separately for each 2-digit SIC code industry in which there are at least 5 IPO firms (to

provide reliable estimates of the incremental IPO-firm coefficient on change in sales, 3 ), and

also from pooled data. For industry-specific regressions, t-statistics are obtained from the

cross-sectional distribution of industry-specific coefficients. Our prediction is that 2 and 3

are positive, because the IPO proceeds relax the resource constraint and permit increases in

working capital, both unconditionally and conditionally on sales. Table 7 reports estimates

from both the pooled and industry-specific regressions.

The incremental coefficient 3 onDIPOj,t*Salesj,t is positive in all four

specifications, irrespective of whether current accruals are estimated from changes in balance

sheet figures or from cash flow statements, and of whether industry or pooled regressions are

employed. For non-IPO firm/years, current accruals increase at the margin by approximately

11% - 13% of changes in sales.26

For IPO firms during the IPO year, the sensitivity of current

accruals to changes in sales increases to approximately 14% -18%.27

The IPO firms IPO-year

incremental slope of current accruals on sales, 3 , is significant in three of the four

25When cash flow statement data are required, the period is restricted to 1987 to 1993.

26 In the four specifications, the slopes are: 0.126, 0.111, 0.112 and 0.108.

7/31/2019 SSRN-id918421

34/59

32

specifications: in both pooled and industry regressions using balance sheet data (with a total

of 1461 IPO observations); and in the pooled but not the industry regressions using cash flow

data (431 IPO firms and 27 industry-specific observations). We conclude that out-of-sample

estimates of Jones model coefficients are biased in relation to IPO years.28

Similarly, the incremental intercept2 for IPO firm/years is positive in all four

specifications, varying between 0.033 and 0.050. It is significant in three (again, from 431

observations). This indicates that IPO firms on average have significantly higher accruals of

approximately 3% to 5% of total assets in the IPO year, independent of their sales growth.

This is consistent with earnings management in the IPO year, but it is also consistent with

IPO firms investing part of the IPO proceeds in operating activities and with other IPO-

induced effects.

We have several reasons to view the latter interpretation endogenous consequences

of the IPO as the more plausible explanation for IPO firms increases in working capital,

both conditional and unconditional on sales growth. First, resource constrained private firms

presumably under-invest in all assets, including working capital assets such as inventories,

receivables and prepayments, and over-utilize short term debt. Relaxing the constraint upon

going public should lead to all working capital items being adjusted at the margin. The TWW

procedure would falsely identify increases in working capital assets and decreases in current

liabilities as income-increasing accruals. Second, the fact that the incremental coefficient

3 on change in sales is different between IPO and non-IPO firm/years leads us to be wary of

all Jones model parameters that are estimated out of sample. Third, working capital assets

27 In the four specifications, the slopes are: 0.126 + 0.018 = 0.144; 0.111 + 0.065 = 0.176; 0.112 + 0.024 =0.136; and 0.108 + 0.057 = 0.165.28 The incremental coefficient on Sales for IPO firms could alternatively reflect earnings management by IPO

firms that is correlated with changes in sales.

7/31/2019 SSRN-id918421

35/59

33

such as receivables and inventories are among the more verifiable items on a balance sheet.

Fourth, public firms are held to higher standards of financial reporting than private firms

(Ball and Shivakumar 2005), they are subject to increased litigation and regulatory action,

and we expect their IPO-year financials are monitored closely by a variety of parties, so we

are skeptical of the interpretation that they are inflating earnings.

3.5 Analysis of pre-IPO accruals

TWW study accruals in the year that includes the IPO: that is, in the period between

the year before the IPO and the year after. We have described two principal problems that

arise. First, these accruals influence the post-IPO financial statements, and therefore occur too

late to inflate the IPO price, which is the prime objective they ascribe to IPO firms earnings

management. Second, the post-IPO balance sheet is affected by the endogenous firms use of

the IPO proceeds to adjust their working capital.

We therefore investigate IPO firms accruals in event year -1. These accruals affect

the last annual earnings report and balance sheet issued before the IPO. The pre-IPO data

needed to estimate discretionary accruals in event year -1 from balance sheet changes are

available from COMPUSTAT for a subsample of 131 firms in the TWW sample. If current

accruals are estimated from cash flow statement data the available subsample is only 43

firms.

We analyze the abnormal accruals for the IPO firms in event year -1, defined as the

actual current accruals less normal accruals estimated from either a linear or piecewise linear

version of the Jones model. The piecewise linear model is given by:

ititit4jit3jit2jit1j0jit CFO*DCFODCFOCFOSalesCA +++++= (7)

where CFOitis cash flow from operations for firm i in period t,DCFOitis a dummy indicator

for negative cash flows that takes the value 1 ifCFOit

7/31/2019 SSRN-id918421

36/59

34

variables are as defined above. All accounting variables are standardized by beginning total

assets.29 In the linear Jones model, j2 to j4 are constrained to zero. The parameters, j0 to,

are estimated separately for each IPO firm j using contemporaneous data from a cross-section

of all the non-IPO listed firms in its 2-digit SIC. In each regression, we trim the extreme 1%

on either side of each continuous variable. Only industry-years with at least 10 non-IPO

firms for estimation are considered.

In order to maximize the number of observations, we obtain data on current accruals

from cash flow statement data when available, but otherwise estimate it from balance sheet

changes. Nonetheless, for comparison, we also report results that use current accruals

obtained only from cash flow statements.

Panel A of Table 8 reports summary statistics for abnormal accruals. Both the mean

and the median abnormal accruals are significantly negative, irrespective of whether the

linear or the piecewise linear Jones model is applied. When we restrict the sample to current

accruals obtained from cash flow statement data, the abnormal accruals continue to be

negative but are no longer statistically significant at the normally accepted levels. The mean

abnormal accruals, when current accruals are obtained from both cash flow statement data

and from balance sheet changes, are relatively high at -12% or -15% of beginning total assets.

However, the means appear to be largely driven by noise from use of balance sheet changes

data for current accruals. The median of abnormal accruals is less than 2% of beginning total

assets in magnitude. 30

29 Using the Modified Jones model to estimate abnormal accruals, or standardizing the intercept by lagged total

assets as in TWW, does not qualitatively change the results.30 The above results are consistent with: Xiong et al. (2005) who, in a sample of 284 U.S. IPOs, find little

evidence of positive pre-IPO abnormal accruals; and Venkataraman, Weber and Willenborg (2004), who report

significantly negative pre-IPO abnormal accruals for a sample of 113 U.S. IPOs.

7/31/2019 SSRN-id918421

37/59

35

There is no evidence that abnormal accruals in event year -1 are positive, which is

inconsistent with the earnings management hypothesis. These results add support to our

contention that the positive abnormal accruals observed in the IPO year are affected by firms

use of IPO proceeds, and, hence are not reliable for drawing inferences about earnings

management by IPO firms.

We also test the prediction that IPO firms will report conservatively, by estimating the

following regression, adapted from Ball and Shivakumar (2005, 2006):

itititit

itititititit

ititititititit

CFODCFODIPO

DCFODIPOCFODIPOSalesDIPO

DPUBCFODCFODCFOCFOSalesCA

++

++++++++=

**

***

*

14

131211

1043210

(8)

where DIPOit takes the value 1 for event year 1 for IPO firms and 0, otherwise. All other

variables are as defined earlier. The above regression is estimated using a pooled sample of

all event year 1 data for IPO firms and all non-IPO firms during the TWW sample period

(i.e., 1980 to 1992) with data available on COMPUSTAT. To ensure that our IPO firms and

non-IPO firms are comparable, we restrict the regression to firms in industries with at least 5

IPOs.31

To control for influential observation we trim 1% of both extremes for each

continuous variable. Current accruals are estimated from cash flow statement data when this

is available and otherwise, estimated from balance sheet changes. Due to lack of sufficient

observations, we are unable to estimate the regressions when current accruals are taken from

cash flow statement data alone.

An advantage of this procedure is that the parameters for computing normal accruals

of IPO firms are not estimated out of sample. The regression allows the coefficient on

Salesit to vary between IPO and non-IPO firms. If firms engage in earnings management, as

7/31/2019 SSRN-id918421

38/59

36

hypothesized by TWW, we expect the coefficient on the dummy variable for IPO firms ( 10 )

to be significantly positive. Further, if non-IPO listed firms are conditionally conservative in

their reporting then as in Ball and Shivakumar (2006), 4 will be significantly positive. If the

IPO firms are more (less) conditionally conservative in event year -1 than comparable data

for non-IPO listed firms, then 14 , the incremental sensitivity of current accruals to negative

cash flows for IPO firms, will be positive (negative).

The regression results are reported in Panel B of Table 8. The coefficient onDIPOit is

insignificant in this regression, implying that IPO firms cannot be concluded to either

overstate or understate their earnings in the IPO year relative to comparable non-IPO firms.

The coefficient onDIPO*Salesitalso is insignificant, implying that the relation between

current accruals and change in sales in the pre-IPO year cannot be shown to differ between

IPO firms and listed firms. Consistent with Ball and Shivakumar (2006), the coefficient 4 is

significantly positive, indicating that non-IPO firms are conditionally conservative in this

period. Moreover, the significantly positive coefficient 14 suggests that IPO firms are more

conditionally conservative in the year prior to the IPO than non-IPO firms. Although the

latter result is not expected based on the Ball and Shivakumar (2005) hypothesis, it is

consistent with IPO firms facing greater litigation risks than seasoned firms and

consequently, being more conservative.

31 This requirement is the main reason we lose a lot of IPO observations in this regression compared to analyses

based on abnormal accrual estimates. Although our inference is not sensitive to this requirement, we believe

that it makes the incremental coefficients for non-IPO firms more meaningful and easier to interpret.

7/31/2019 SSRN-id918421

39/59

7/31/2019 SSRN-id918421

40/59

38

behavior of IPO firms more closely resembles that of public firms as the IPO date

approaches, even though they remain private at the time.

We conjecture that the types of bias we observe in conventional estimates of

discretionary accruals around the time of IPOs occur in a broad genre of studies on

earnings management around large transactions and events. The typical hypothesis in this

genre is that substantial earnings management occurs because, at the time of large

transactions and events, managers have unusually strong incentives to influence their reported

financial performance. The alternative view is that these are times of unusual scrutiny and

risk arising from earnings management, and that the appearance of discretionary accruals is

due to changes in companies working capital that is endogenous to the transactions and

events.

7/31/2019 SSRN-id918421

41/59

39

References

Aharoney, J., Lin, C. and Loeb, M., 1993. Initial public offerings, accounting choices and

earnings management. Contemporary Accounting Research 10, 6181.

Ali, A. and L. Hwang, 2000. Country specific factors related to financial reporting and therelevance of accounting data. Journal of Accounting Research 38, 1-21.

Ball, R., S. P. Kothari and A. Robin, 2000. The effect of international institutional factors on

properties of accounting earnings. Journal of Accounting & Economics 29, 1-52.

Ball, Ray, Ashok Robin and Gil Sadka, 2006. Are timeliness and conservatism due to debt or

equity markets? An international test of contracting and value relevance theories of

accounting. Manuscript, University of Chicago.

Ball, Ray, Ashok Robin and Joanna Shuang Wu, 2003. Incentives versus standards:

Properties of accounting income in four East Asian countries and implications for acceptanceof IAS. Journal of Accounting & Economics 36, 235-270.

Ball, R., Shivakumar, L., 2005. Earnings quality in U.K. private firms: comparative loss

recognition timeliness, Journal of Accounting & Economics 39, 83-128.

Ball, R., Shivakumar, L., 2006. The role of accruals in asymmetrically timely gain and loss

recognition. Journal of Accounting Research, forthcoming.

Basu, S., 1997. The conservatism principle and asymmetric timeliness of earnings. Journal ofAccounting & Economics 24, 3-37.

Burgstahler, D., Hail, L. and Leuz, C., 2006. The importance of reporting incentives:

Earnings management in European private and public firms. The Accounting Review,

forthcoming.

Bushman, R., Piotroski, J., 2006. Financial reporting incentives for conservative accounting:

The influence of legal and political institutions. Journal of Accounting and Economics(forthcoming).

Bushman, Robert M., Joseph Piotroski and Abbie J. Smith, 2004. What determines corporatetransparency?" Journal of Accounting Research 42, 207-252.

Bushman, Robert M., Joseph Piotroski and Abbie J. Smith, 2006. Capital allocation and

timely accounting recognition of economic losses, Working Paper, University of Chicago

(May).

Cahan, S., 1992. The effect of antitrust investigations on discretionary accruals: A refined testof the political cost hypothesis. The Accounting Review 65, 7795.

7/31/2019 SSRN-id918421

42/59

40

Coles, J. L., Hertzel M., and Kalpathy, S., 2006. Earnings management around employee

stock option reissues.Journal of Accounting and Economics, forthcoming.

DeAngelo, L., 1986. Accounting numbers as market valuation substitutes: A study ofmanagement buyouts of public stockholders. The Accounting Review 61, 400-420.

DeAngelo, H., DeAngelo, L. and Skinner, D., 1994. Accounting choice in troubled

companies. Journal of Accounting and Economics 17, 113143.

Dechow, P.M. and Dichev, I., 2002. The quality of accruals and earnings: The role of accrualestimation errors. The Accounting Review 77, 35-59.

Dechow, P.M., Sloan, R.G. and Sweeney, A.P., 1995. Detecting earnings management. The

Accounting Review 70, 193-225.

Dechow, P.M., Sloan, R.G. and Sweeney, A.P., 1996. Causes and consequences of earnings

manipulation: An analysis of firms subject to enforcement actions by the SEC. ContemporaryAccounting Research 13, 136.

DeFond, M. and Jiambalvo, J., 1994. Debt covenant violation and manipulation of accruals.Journal of Accounting and Economics 17, 145176.

Desai, H., S. Rajgopal and M. Venkatachalam, 2004. Value-glamour and accruals mispricing:

One anomaly or two? The Accounting Review 79, 355-385.

Ducharme, L.L., P.H. Malatesta and S.E. Sefcik, 2004. Earnings management, stock issuesand shareholder lawsuits. Journal of Financial Economics 71, 27-49.

Erickson, M. and Wang, S., 1999. Earnings management by acquiring firms in stock for stock

mergers. Journal of Accounting and Economics 27, 149-176.

Friedlan, J., 1994. Accounting choices by issuers of initial public offerings. Contemporary

Accounting Research 11, 131.

Han, J. and Wang, S., 1998. Political costs and earnings management of oil companies during

the 1990 Persian Gulf crisis. The Accounting Review 73, 103117.

Hribar, P., Collins, D.W., 2002. Errors in estimating accruals: Implications for empirical

research. Journal of Accounting Research 40, 105-134.

Jiambalvo, J., 1996. Discussion of causes and consequences of earnings manipulation: Ananalysis of firms subject to enforcement actions by the SEC. Contemporary Accounting

Research 13, 3747.

Jones, J., 1991. Earnings management during import relief investigations. Journal ofAccounting Research 29, 193228.

7/31/2019 SSRN-id918421

43/59

41

Kothari, S.P., A. Leone and C. Wasley, 2005. Performance matched discretionary accrual

measures. Journal of Accounting and Economics 39, 163197.

Leuz, C. 2003, IFRS versus US GAAP: Information Asymmetry-Based Evidence fromGermanys New Market. Journal of Accounting Research 41, 445-472.

Leuz, C., Nanda, D., Wysocki, P.D., 2003. Earnings management and investor protection: An

international comparison, Journal of Financial Economics 69, 505-527.

Leuz, Christian and Felix Oberholzer, 2006. Political relationships, global financing, andcorporate transparency. Journal of Financial Economics, forthcoming.

Leuz, C., A. Triantis and Y. Wang, 2006. Why do firms go dark? Causes and economic

consequences of voluntary SEC deregistrations. Working paper, University of Chicago,

March.

Liberty, S. and Zimmerman, J., 1986. Labor union contract negotiations and accountingchoices. The Accounting Review 61, 692712.

Louis, H., 2004. Earnings management and the market performance of acquiringfirms. Journal of Financial Economics 74, 121-148.

Louis, H. and Robinson, D., 2005.D