The Influence of Earnings Management on Long-Run Stock Performance of IPO Firms

The Influence of Earnings Management on

Long-Run Stock Performance of IPO Firms

The Influence of Earnings Management on Long-Run Stock

Performance of IPO Firms

Karin Rozendal

Master’s Thesis

University of Groningen

Faculty of Economics and Business

MSc Business Administration, specialization Finance

October 2010

Supervisor: Dr. C.A. Huijgen, Associate Professor of Financial Accounting, Faculty of Economics Rijksuniversiteit Groningen

ACKNOWLEDGEMENTS

This thesis is the final product for my Master’s in Business Administration, with a specialization in Corporate Finance. Unfortunately, this means that an end has come to my days as a student at the University of Groningen, a period in which I have grown both intellectually and personally, met new friends, and had a great time. However, after several years of study, I look forward to the new things to come.

At this place, I want to thank all the people who assisted me writing my master’s thesis. First of all, I want to thank my supervisor dr. C. Huijgen who assisted me in a challenging way. He has given me valuable feedback and input for my thesis through the entire process. I also want to thank my parents for all their help and support. They made it possible for me to study. Finally I would like to thank my boyfriend, family and friends for their support and interest during the completion of this thesis.

ABSTRACT

Issuers of initial public offerings (IPO) have incentives and opportunities to inflate reported earnings by taking large positive accruals. This paper examines whether IPO firms, on average, have indeed high positive issue-year abnormal accruals, followed by a reversal of the earnings-manipulating accruals. Moreover, this paper examines whether aggressive IPO-year earnings management through income-increasing accounting adjustments leads investors to misinterpret reported earnings and be overly optimistic about the issuing firms’ prospects. If investors interpret accruals naively, aggressive earnings manipulating issuing firms are consequently overvalued, which might explain the long-run underperformance of stock returns of IPO firms. Results of this study indicate that accounting accruals are significantly higher in the year of the initial public offering than in one year before and in the four years after the offering, which suggests that firm management manipulate earnings upwards during the IPO year. Furthermore, results show that accruals in the first and fourth year after the IPO are negatively related to IPO year accruals, indicating that accounting accruals reverse. However, no evidence is found that market-adjusted buy-and-hold stock returns in one, two, three and four years after the IPO are related to the level of earnings management during the IPO year, using total accruals, Jones model discretionary accruals and discretionary current accruals, and DeFond and Park model discretionary working capital accruals as measure for earnings management. Including an interaction term of a dummy for a decrease in accruals and the accruals variables indicate that there is also no significant relationship between high IPO year accruals and firm stock performance for firms experiencing decreases in accruals in the second to fourth year following the IPO. Only in the first year after the offering, I find weak evidence that stock performance is related to the IPO year total accruals and Jones model discretionary accruals interaction terms, using a significance level of 10 per cent. I conclude that there is no robust evidence that investors misinterpret the implications of earnings management, which causes stock prices of related firms to deviate from their true values.

Keywords:, initial public offerings; long-run stock performance; market efficiency; earnings management; accounting accruals; German stock market

TABLE OF CONTENTS

1. INTRODUCTION...7

2. THEORETICAL AND EMPIRICAL BACKGROUND ON EARNINGS MANAGEMENT IN IPOs ...10

2.1 Initial public offerings in Germany ...10

2.2 The mechanics of going public ...11

2.3 Valuing IPOs...13

2.4 IPO pricing and long run performance ...13

2.5. Earnings management in Initial Public Offerings ...14

2.6 Stock prices and earnings management ...15

2.7 Summary and conclusions ...17

3. HYPOTHESES DEVELOPMENT, SAMPLE SELECTION, DATA AND MEASURES FOR EARNINGS MANAGEMENT ...18

3.1 Hypotheses development ...18

3.2 Sample Selection and Data ...19

3.3 Measures for earnings management ...20

- 3.3.1 The Modified Jones model ...20

- 3.3.2 DeFond and Park model ...23

3.4 Data for IPOs ...24

3.5 Summary and conclusions ...25

4. EMPIRICAL RESULTS ...26

4.1 Earnings management and future accruals...26

4.2 Earnings management and stock returns...27

4.3 Univariate analysis ...29

4.4 Multivariate analysis...30

4.5 Summary and conclusions ...34

5. ROBUSTNESS CHECK ...35

6. SUMMARY AND CONCLUSION ...37

1.

INTRODUCTION

INITIAL PUBLIC OFFERINGS (IPOs) provide incentives for firms to manage reported earnings opportunistically in several ways. First, high reported earnings may raise the stock price at which the firm goes public, which has a significant impact on the wealth of the issuing company. Second, the lock-up period provides an incentive to manage reported earnings opportunistically (Seger, 2010). This is the period after the IPO (usually 180 or 360 days), in which managers can not sell their own shares in the IPO firm. The manager has incentives to ensure that profits remain high during this period, because he/she can sell the shares for a higher price when the lock-up period is over. Third, it may be difficult for investors to correctly judge the market value of the firm, due to high information asymmetries between the issuing company and potential investors.

There is generally little information available about the issuing firm at the time of the initial public offering besides the information in the offering prospectus. This scarcity of information forces investors to rely on this offering prospectus, which may contain only one to three years of financial statements. Due to the motivation and opportunities that managers have to manage earnings, the IPO process is thus particularly susceptible to earnings manipulation (Teoh, Welch and Wong, 1998a). According to DuCharme, Malatesta and Sefcik (2001) managers have considerable room under Generally Accepted Accounting Principles (GAAP) to opportunistically manage earnings by accounting adjustments to cash flows (collectively called accruals). Different techniques to inflate reported earnings can be grouped in one of three categories: (i) the choice of accounting methods; (ii) revision of estimates; and (iii) acceleration or deferral of revenues and expenses (DuCharme, Malatesta and Sefcik, 2001). The flexibility under GAAP is meant to allow statements to reflect information about underlying economic conditions more accurately than would be possible with strict reporting rules. However, if managers wish to mislead investors, as might be the case with initial public offerings, discretion provides a greater scope for misrepresenting true underlying firm performance.

stock prices to decrease. The greater the manipulation in earnings at the time of the initial offering, the larger the ultimate price correction (Teoh, Welch and Wong, 1998a). Therefore, stock returns in the years following an IPO can potentially indicate whether or not investors correctly judge the implications of earnings management. If, on average, investors correctly interpret financial statements, earnings management should be unrelated to stock returns. However, if investors ignore the lower quality of earnings, then earnings management may lead to a temporary deviation of true firm value which will lead to a price correction in subsequent years.

In this study, I examine whether post-IPO stock return performance is explained by firm’s systematically selecting accounting methods to increase reported earnings in periods prior to going public, as measured by discretionary accruals, using a sample of 240 initial public offerings (IPOs) in Germany that went public between 1998 and 2004. The research question I formulate for this study is:

“Is the use of earnings management prior to initial public offerings (measured by high accounting accruals) related to underperformance in the years following the IPO, indicating that investors do not correctly interpret financial statements?”

Using a modified version of the model by Jones (1991), I decompose the reported earnings into three components: cash flow from operations, expected accruals, and discretionary (or managed) accruals.

IPOs. Investigating the IPO market in Germany is of special interest for a number of reasons. As indicated by Bessler and Thies (2007), the banking system, the legal system, as well as the corporate governance structure in Germany are viewed as different from that of the US and other countries. Therefore, using a sample of German initial public offerings may lead to some additional insights. According to Bessler and Thies, the German universal banking system is different from that of other countries, because financial institutions traditionally hold equity stakes in firms, provide venture capital financing and loan services before and after the IPO, but also offer full or partial underwriting services and securities trading. Furthermore, banks in Germany are involved in mutual fund management as well as in investment research and analyst recommendations. With respect to the corporate governance system, Bessler and Thies indicate that the fact that Germany traditionally had a bank oriented system, where banks are firm’s key source of financing. The different roles of the bank in investment banking on one side and equity research and trading on the other, may lead to severe conflicts of interest. In fact, some earlier empirical studies for seasoned equity offerings (SEOs) find different empirical results for Germany than usually found in studies for the US (Bessler and Thies, 2007). Another reason why the German IPO market differs from that of other countries is the relatively small number of companies that went public and the relatively small amount of equity raised through IPOs in Germany compared to the US and some other countries, despite the size of the German economy. Hence, some different results may be expected due to a less competitive and less efficient primary market. No previous research on the impact of earnings management on the long run stock performance of IPO firms in Germany has been performed, to my knowledge.

The results of this study suggest that accruals are managed upwards in the year of the initial public offering, and reversed in the four years after the offering. However, mixed evidence is found that market-adjusted buy-and-hold stock returns in one, two and three years after the IPO are related to the level of earnings management during the IPO year. This study shows indications that market-adjusted stock returns in the four years after the IPO are negatively influenced by high IPO year accounting accruals, using the DeFond and Park (2001) model to estimate discretionary accruals. The relationship disappears, however, when the Jones (1991) model to discriminate between normal and abnormal accruals is utilized. Results are thus dependent on the method of estimating discretionary accounting accruals. Therefore, I must conclude that this study can not find robust evidence that investors misinterpret the implications of earnings management.

2.

THEORETICAL AND EMPIRICAL BACKGROUND ON EARNINGS

MANAGEMENT IN IPOs

In this chapter, I elaborate on the theoretical and empirical background of earnings management in initial public offerings. First, I will describe the market for initial public offering in Germany (section 1) and the mechanics of offering (section 2). In section 3, I describe the approaches of valuing IPOs and section 4 summarizes on the empirical evidence on the long run performance of IPO firms. Section 5 presents the empirical findings of earnings management in initial public offerings, and section 6 summarizes previous research on the relationship of earnings management with subsequent stock performance. Finally, the findings of this section are summarized in section 7.

2.1 Initial public offerings in Germany

Ritter (1998) defined an IPO as: “An IPO occurs when a security is sold to the general public

for the first time, with the expectation that a liquid market will develop”. Most firms start out by

raising capital from a small number of investors. However, if the company grows further and needs additional equity capital, it may be desirable to ‘go public’ by selling shares to a large number of diversified investors. Such an initial public offering enhances liquidity, which allows the company to raise capital on more favourable terms. However, there are substantial costs associated with going public. For example, publicly traded firms need to supply information on a regular basis to investors and regulators, and legal, auditing, and underwriting fees must be paid. Moreover, management time and effort must be devoted to complete the offering. These direct and indirect costs affect the cost of capital for firms going public (Ritter, 1998). The high costs associated with going public makes it even more important to receive a high stock price at the initial offering.

In Germany, companies have traditionally relied primarily on internal financing methods and bank loans. Between 1945 and 1983 IPOs were very rare in Germany, but after 1983, the number of IPOs started to increase. Compared to the US and the UK, however, IPO activity still remained rather low. For example, in the years between 1988 and 1995 a total of 151 IPOs were carried out in Germany, compared to more than 1,000 in the UK and nearly 2,500 new listings on the NYSE and the American Stock Exchanges and 3,000 at Nasdaq (Schuster, 2003). Starting in 1997, however, Germany witnessed an unprecedented increase in IPO activities, which coincided with what was seen as a major innovation for the German equity market: the introduction of the New Market (Neuer

Markt) by the Deutsche Börse. The New Market was established as a trading segment for young

innovative growth stocks that had to meet international standards of transparency and publicity. Previously, the German equity market consisted of different segments. The Official Market (Amtlicher

industry sectors are traded. The Regulated Market (Geregelter Markt) was introduced in 1987 as a segment for small to medium sized high-tech growth firms, with a lower market capitalization and a lower trading volume. Reducing the listing requirements accommodated the needs of this segment (Franzke, Grohs and Laux, 2004). The Unofficial new market (Freiverkehr) is the third segment and has correspondingly lower regulations, operating under private law. Recognizing a need to promote growth stocks, the Deutsche Börse established the New Market within the Unofficial Regulated Market (Macy and Terry, 2002). The main differences between the Official Market, the Regulated Market and the Unofficial Regulated Market are the stringency of the listing, reporting, and disclosure requirements. While officially listed in the Unofficial Regulated Market, lower requirements for firms listed on the New Market in terms of capital and operating history are offset by more stringent transparency and reporting rules after the IPO date (Schuster, 2003). For example, firms trading on the New Market must publish reports quarterly and annually in both German and English compared to biannual reports in German for domestic firms in the Official Market. The reports must be in accordance with either International Accounting Standards or U.S. Generally Accepted Accounting Principles (US GAAP). The additional requirements of the reporting of company information increase the ability to compare both domestic and international firms and therefore increases competition for investment funds. The increased transparency and liquidity lowers the risk and required return of the firms (Macy and Terry, 2002).

The boom on the New Market between 1997 and 2000, when many firms entered the market, provides a striking contrast to the preceding era. In the year 1999 alone, more IPOs were carried out than in the 10 years from 1988 to 1997 (Frankze, Grohs and Laux, 2004). However, by US standards, the levels of both IPO and venture capital activities remained rather low even in this boom phase. Of course, the world-wide stock market downturn after 2000 also resulted in a rather sudden end of the IPO boom in Germany.

2.2 The mechanics of going public

In Germany, as well as in most other countries in Europe, the regulations regarding an initial public offering are set and maintained by the relevant exchange itself, with the consent of the Secretary of the Treasury, and must also be in accordance with legal guidelines that are set under the European Investment Services Directive (Schuster, 2003).

Before a firm can establish an IPO, it must first obtain permission from the German Federal Financial Supervisory Authority (Bundesanstalt für Finanzdienstleistungsaufsicht, BaFin) to have its shares listed on an exchange. The request for a listing must be made on the basis of a notice of introduction, whose contents are subject to detailed regulation (Schuster, 2003).

the public offer of securities was created. Since its founding in 1995, the Federal Supervisory Office for Securities Trading (Bundesaufsichtsamt für den Wertpapierhandel or BAWe) in Frankfurt am Main, which merged in 2002 with the Federal Financial Supervisory Authority (Bundesanstalt für Finanzdienstleistungsaufsicht or BaFin), is the central depository for prospectuses. The Financial Market Promotion Act (Finanzmarktförderungsgesetz or FMFG) of April 1998 introduced the examination of prospectuses by the BAWe; every prospectus was examined on completeness, especially on whether it contained the minimum information requirements by the Verkaufsprospekt Gesetz and the Prospectus Law Regulations. The BAWe is thus explicitly concerned with full disclosure of material information, but does not attempt to determine whether a security is fairly priced or not. Other criteria include publishing annual accounts, specifying to which use the proceeds of the offering will be put, and disclosing the shareholdings of the management and board of directors. Moreover, a chartered accountant must certify the firm’s annual accounts (Schuster, 2003).

A universal or an investment bank typically underwrites the IPO. Particularly in Germany, companies seeking a listing have usually been engaged in a long-run relationship with its underwriter (Hausbankbeziehung). The underwriter is not only involved in working out the registration statement, but is also responsible for managing the underwriting and floatation process (Schuster, 2003).

As described in section 2.1, German companies have, in principle, a choice between three market segments in which to list their shares: The Official Market, the Regulated Market or the New Market. The choice of market segment is mainly based upon the minimum size of the issue. Moreover, firms that intend to go public have a choice of which method their shares are offered to the public. A private placing, an offer for sale by tender (also referred to as bookbuilding), and an offer for sale at a fixed price can be used to obtain a listing and issuing equity (Schuster, 2003). In an offer for sale at a fixed price, the prospectus states the number of shares being offered for sale and the price per share. The fixed price element is designed to eliminate price uncertainty, which makes the issue more appealing to investors. Investors can then submit bids for the number of shares they wish to take up at the fixed price. In offers for sale at a fixed price, the issue is sub-underwritten, at the same price, by a group of financial institutions. Once the price of the issue is fixed, it can neither be changed in response to emerging demand, nor withdrawn (Schuster, 2003).

range of German IPOs is typically set after bookbuilding has started, with the pricing typically occurring seven trading days later. The price range is frequently more than 2 Euros, but once set, IPOs never price above the maximum in Germany (Ritter, 2003).

2.3 Valuing IPOs

In principal, valuing IPOs is no different from valuing other stocks (Schuster, 2003). Common approaches like the discounted cash flow (DCF) method and comparable firms analysis can be used. In practice, historical accounting information is of limited use in predicting future profits or cash flows because many IPOs are of young growth companies in high-tech industries. Valuing IPO firms may therefore rely heavily on how the market is valuing comparable firms. In some cases, it may be difficult to find comparable publicly-traded firms to use for valuation purposes.

2.4 IPO pricing and long run performance

Several studies have found that IPOs underperform after the stock issue. A well-cited study on this topic is performed by Ritter (1991), who reports substantially lower stock returns over a three-year holding period after the initial offering for a sample of 1,526 IPOs going public between 1975 and 1984 than for a set of comparable firms matched by size and industry. On average, stock returns were 27 per cent lower for initial public offerings than for seasoned offerings in the same industry and for companies of the same size. He concludes that the patterns he finds are consistent with investors being “periodically overoptimistic about the earnings potential of young growth companies" and “firms taking advantage of these windows of opportunity”. As is described in the introduction, manipulations of earnings might be one possible source for this overoptimism. Aggarwal and Rivoli (1990) support these results. Using a sample of 1,598 IPOs issued between 1977 and 1987, they report a negative abnormal return of 13.73% for investors purchasing all IPOs in the open market at the end of the first trading day and holding them for a period of 250 trading days. Loughran and Ritter (1995) find that companies issuing both initial and seasonal equity offerings during 1970 to 1990 significantly underperform relative to non-issuing firms for five years after the offering date. They show that investors would have had to invest 44% more money in the issuers than in non-issuers of the same size to have the same wealth five years after the offering date.

Finally, there are some studies who do not find evidence for the long-run underperformance of IPO firms. One example is the study by Brav, Geczy, and Gompers (2000), who find that post-issue IPO returns are similar to those of firms with similar size and book-to-market characteristics. Furthermore, they find that seasoned equity offering returns covary with those of similar non-issuing firms. Gompers and Lerner (2003) show that IPOs issued between 1935 and 1972 performed poorly in the years after issue when event-time buy-and-hold abnormal returns are used. However, they suggest that this underperformance disappears when other methods are used to measure excess returns.

In this study, I want to examine whether the long-run underperformance of IPO firms after the issue might be explained by investors naively interpreting the implications of earnings management. In the next sections, I will describe the findings of current literature on the presence of earnings management in initial public offerings and its influence on stock performance.

2.5. Earnings management in Initial Public Offerings

Healy and Wahlen (1999) define earnings management as follows:

“Earnings management occurs when managers use judgement in financial reporting and in

structuring transactions to alter financial reports to either mislead some stakeholders about the underlying economic performance of the company, or to influence contractual outcomes that depend on reported accounting numbers.”

As is described in the introduction, the IPO process is particularly susceptible to earnings manipulations, since there are high information asymmetries between investors and the issuing firm. Potential investors have to rely on the information provided in the offering prospectus, which usually only contains information on the financial statements of one to three years ago. Since accounting standard setters (for example Accounting Principles Board Opinion 20) allow IPO firms to change their accounting choices for all the financial statements presented in the offering prospectus, issuing firms have an excellent opportunity to present the earnings of the pre-issue fiscal years as positively as possible under GAAP (Teoh, Welch and Wong, 1998a). The firm has an incentive to show increases in the level of reported earnings to maintain a high equity price.

were engaged by Neill, Pourciau and Schaefer (1995), DuCharme (1994), Friedlan (1994), Teoh, Welch and Wong (1998a), Teoh, Wong and Rao (1998) and Darrough and Rangan (2005). Furthermore, Jain and Kini (1994) find empirical evidence that firms exhibit a decrease in operating performance in the years subsequent to the IPO. This suggests that investors may initially have high expectations for future growth in earnings that might not be fulfilled subsequently. In contrary to these findings, Ball and Shivakumar (2008) find that initial public offerings in the UK report more conservatively, and that prospective financials do not reflect systematic earnings inflation. They attribute this to the higher demand in reporting quality and higher monitoring by auditors, boards, analysts and other stakeholders.

2.6 Stock prices and earnings management

The question of whether investors take into account the temporary deviations in reported earnings caused by financial statement manipulation is of great interest to both participants in capital markets and academics. Available literature provides evidence that investors do not fully anticipate on the implication of current accounting choices for the development of future earnings. Sloan (1996) is one of the first to investigate whether future earnings and stock returns depend on current earnings management. In order to measure the extent of earnings management, he divides total earnings in a cash flow component and an accrual component. Using data from the NYSE and AMEX in the period 1962-1993, he finds evidence that the development of earnings performance indeed depends on the relative magnitudes of the cash and accrual components of earnings. Moreover, he shows that stocks with high accruals have lower stock returns and underperform relative to stocks with lower accounting accruals. He concludes that stock prices act as if investors initially fail to identify correctly the different properties of these two components of earnings. These results are confirmed by the study of Bradshaw, Richardson and Sloan (2001) who also show that firms with high working capital accruals have a higher probability of experiencing declines in reported earnings. They point out that a limitation of the earlier work of Sloan was that the relationship between predictable earnings declines and the prices of stocks could be attributable to “unidentified risk factors or research design flaws”. Bradshaw, Richardson and Sloan control for these factors by incorporating whether the forecasts of analysts anticipate on the level of accruals and whether independent auditors signal future earnings problems when faced with high current accruals. They conclude that analysts and auditors do not alert investors to future earnings problems associated with high accruals, which may imply that investors do not anticipate on high accruals.

consistent with investors naively extrapolating pre-issue earnings without fully adjusting for the potential manipulation of reported earnings.

Finally, another example of research examining whether future earnings depend on earnings management is the study by Chan, Chan, Jegadeesh and Lakonishok (2006), who examine companies listed on the NYSE, AMEX and NASDAQ for the period 1971 to 1995. Unlike Sloan (1996) and Bradshaw et al. (2001), Chan et al. distinguish between the discretionary, or unexpected, and the non-discretionary part of accruals. They find that the non-discretionary part of accounting accruals are reliably, negatively related to future stock returns. However, their results indicate that the nondiscretionary part of accruals do not predict future returns.

Despite the significant outcomes of the studies described above, literature on the disability of investors to value accruals appropriately is rather mixed. For example, Subramanyan (1996) finds some evidence that discretionary accruals are in fact priced by the stock market. He concludes that discretionary accruals communicate information about future firms’ profitability and that this information is picked up by investors. Furthermore, Healy and Wahlen (1999) state in their review of earnings management research that “we clearly need to better reconcile conflicting findings on the effect of earnings management on stock prices and resource allocation in the economy.”

2.7 Summary and conclusions

3.

HYPOTHESES DEVELOPMENT, SAMPLE SELECTION, DATA AND

MEASURES FOR EARNINGS MANAGEMENT

This chapter describes the development of the hypotheses (section 1), as well as the sample selection procedure and the sources of data used (section 2). Section 3 presents the descriptive statistics of the sample firms and section 4 presents the measures for calculating earnings management. Finally, I summarize the findings of this section in section 5.

3.1 Hypotheses development

As is described in the introduction, managers may opportunistically increase their stock prices through accounting techniques that increase reported earnings. Consequently, investors may be temporarily deceived about the fundamental values of the firm. However, at some point in time managers must reverse any excessive earnings-manipulating accruals made in the past, which will have a negative influence on earnings. If the prior earnings inflating techniques have caused investors to form overly optimistic expectations of future earnings, they will be disappointed by the post-IPO earnings performance and stock prices will tend to decline. Three hypotheses can be derived from this reasoning, which will be examined in this paper.

First, if firms indeed seek to boost earnings before the initial public offering, the financial statements in the years surrounding the issue should contain unusually high positive discretionary accruals. The first hypothesis therefore states:

1

H

Second, the reasoning as described above ascertains that high current accruals can not be maintained in the future. An upwards manipulation by accounting accruals has to be reversed at some point in time. Therefore, the accrual component and cash flow component of current earnings should have different implications for future earnings. More specifically, current earnings performance is less likely to persist if it is attributable to discretionary accruals than to operating cash flows. This leads to the second testable hypothesis:

2

H

investor’s expectations, stock prices are adversely affected. The last issue examined in this paper is whether or not investors interpret accruals naively and do or do not detect and adjust for all accounting choices made by firms going public, as measured by the development of post-issue stock performance. The last hypothesis, expecting that investors naively interpret accruals, thus states:

3

H

3.2 Sample Selection and Data

To test whether post-issue stock performance is related to earnings management during the IPO year, I analyze firms that made IPOs from 1998 through 2004. Firms that went public in this period were identified from Thomson Research. Post-offering accounting and stock data were taken from Worldscope, provided by Thomson Research. Detailed data on the issue were extracted from the German exchange website. For a more specific explanation on the source of data used in this paper, see table 1.

For inclusion in the final sample, IPOs must have available Worldscope financial data to calculate the accrual measures in the year of the offering, and Worldscope stock return data for at least one of the four years after the initial public offering. Furthermore, I exclude all financial companies from the sample for two reasons. First, the financial turmoil in the savings and loan industry in the 1980s led to specific changes in regulatory accounting standards for banks and other financial institutions to mitigate earnings management (Healy and Wahlen, 1999) and second, current models for estimating accounting accruals are not suitable for the financial sector. The final sample for the test consists of 240 firms, but actual sample size varies in different tests according to data availability. For a more specific explanation on the sample selection process, see table 2. The sample has some clustering by industry and year of offer. For example, the computer equipment and services industry (SIC codes 35 and 73) accounts for 48 per cent of the sample firms. Furthermore, the years 1998 through 2000 seem to be hot issue periods, as about 91 per cent of the sample firms went public during these years.

The fiscal year in which the offering occurs is called year 0. Thus, fiscal year 0 includes both pre- and post-IPO information, and fiscal year -1 ends before the date of the IPO.

market value of equity is about 40 per cent and 25 per cent of the mean and median book values, respectively, for the IPO sample, and about 74 per cent and 60 per cent of the mean and median book values, respectively, for the non-IPO sample. The sample firms experience an increase in sales as a percentage of total assets in the offering year of 1 per cent on average, which is a key variable in the regression to estimate discretionary accruals, compared to 0.89 per cent on average for the non-IPO sample. IPO firms are generally young firms, having existed for a mean of 21 years, and a median of 10 years before going public. The offering price is €31.02 on average per share, with a median of €23.00 per share. Finally, the sample IPOs are underpriced, on average, by 43.5 per cent, with a median of 13.1 per cent.

Panel B reports the time distribution of the IPO samples. There is a concentration of IPOs in the years 1999 and 2000, representing almost 70 per cent of the total sample. Furthermore panel C, presenting the industry distribution of the sample, shows there is a concentration of IPOs in the computer equipment and services industry, which is not surprising for the sample period.

3.3 Measures for earnings management

A fundamental element of any test for earnings management is a measure of management's possibilities to manipulate earnings (McNichols, 2000). As mentioned in the introduction, reported earnings can be distinguished in cash flows from operations and accounting adjustments to cash flows, which are called accruals. Recent studies on earnings management have relied primarily on accruals-based measures to estimate the degree of manipulation by firm management (Teoh, Wong and Rao, 1998). Accruals thus represent the difference between the accounting earnings of a firm and its underlying operating cash flow. Large positive accruals, for instance, indicate that earnings are much higher than the cash flow generated by the firm. This might be an indication that management is manipulating financial results. Therefore, accruals can be used as a proxy to measure earnings management. In this study, I use two different models to estimate accrual levels that are believed to reveal earnings manipulations.

- 3.3.1 The Modified Jones model

regression approach to measure the discretionary components influencing accruals, by specifying a relationship between total accruals and change in sales and property, plant and equipment.

Following previous literature I use the modified Jones model for this study. The original Jones model is one of the most commonly used models in literature to distinguish the discretionary from the non-discretionary part of accruals (McNichols, 2000). However, according to Dechow, Sloan and Sweeney (1995), the original Jones model measures the discretionary accruals incorrectly when firm management tries to manipulate sales. They advocate that a modified version of the model developed by Jones (1991) exhibits the most power in detecting earnings management. According to Dechow, Sloan and Sweeney, the original Jones model implicitly assumes that sales are nondiscretionary. Therefore, part of managed earnings will be removed from the discretionary accrual estimate if earnings are managed through discretionary revenues, causing the estimate of earnings management to be biased toward zero. The modified Jones model is designed to eliminate this limitation of the original Jones model. Nondiscretionary accruals are estimated in the same way as in the original Jones model, except that the change in sales is adjusted by the change in accounts receivable in the specified period. The modified Jones model implicitly assumes that all changes in credit sales in the specified period result from earnings management, based on the reasoning that it is easier to manage earnings by manipulating the recognition of credit sales than it is to manage earnings by manipulating cash sales. The modified Jones model can be described as follows:

it it it it it it it it

it A A SALES REC A PPE A

TACC / _1 









Where: TACC = Total accruals (∆ noncash working capital – depreciation expense);

ΔSALES = Change in sales revenue;

REC = Change in accounts receivable;

PPE = Property, plant and equipment; A = Total assets;

calculated as the change in noncash working capital less total depreciation expense. The change in noncash working capital is defined as the change in current assets other than cash and short-term investments less the change in current liabilities other than short-term debt and the current portion of long-term debt. All variables are scaled by lagged total assets.

The results of this regression are reported in table 4, panel A, for each fiscal year as well as for the sample period as a whole. The fitted accruals are considered to be the level necessary to support current economic conditions and are termed nondiscretionary accruals. The prediction error, termed discretionary accruals (DACC), is presumed not to be dictated by the economic conditions of the firm, and is therefore a good proxy for the use of earnings management. DACC is calculated using the following model (Jones, 1991):

]

/

)

[(

]

/

1

[

(

]

/

[

/













it

A

TACC

A

A

SALES

REC

A

DACC





])

/

[





PPE

A

In table 4, panel A, we can see that the regression outcomes vary significantly in each fiscal year. Coefficients on property, plant and equipment (PPE) are all negative, which is the expected sign because PPE are related to an income-decreasing accrual (i.e. depreciation expense). However, coefficients on the change in revenues are both negative and positive. Due to this variability in the regression outcomes for each fiscal year in table 4, panel A, I use the outcomes estimated in the cross-sectional regression for the whole period 1998 – 2004 to calculate discretionary accruals.

Besides the separation of discretionary and non-discretionary accruals, I distinguish accruals into their current and long-term component. These two components are evaluated separately, since some studies argue that managers have more discretion over short-term than over long-term accruals. For example, Jones (1999) argues that the estimated discretionary portion of noncurrent accruals is less likely to reflect year-specific discretion, and that current accruals thus provide a more accurate basis for estimating discretionary behavior than total discretionary accruals. Following Teoh, Wong and Rao (1998) and Teoh, Welch and Wong (1998a), I measure expected and abnormal current and long-term accruals using an extension to the Jones model as described below.

First, I estimate expected current accruals by cross-sectionally regressing total current accruals on the change in sales revenues, where total current accruals (TCA) can be calculated as the increase in total current assets minus the increase in total current liabilities:

The results are reported in table 4, panel B. Again, the normal current accruals are calculated using the estimated coefficients in the fitted equation. The prediction error of current accruals are the abnormal or discretionary current accruals (DCA):

])

/

)

[(

]

/

1

[

(

]

/

[

/













it

A

TCA

A

A

SALES

REC

A

DCA





We can see in table 4, panel B, that coefficients on the change in revenues are both negative and positive. Due to this variability in the regression outcomes for each fiscal year, I use the outcomes estimated in the cross-sectional regression for the whole period 1998 – 2004 to calculate current discretionary accruals, in accordance with the calculation of the discretionary accruals. The discretionary current accruals (DCA) can be calculated by subtracting the non-discretionary current accruals from the total current accruals.

- 3.3.2 DeFond and Park model

While the modified Jones model estimates normal or expected accruals based on a coefficient from a pooled cross-sectional regression and includes the intercept term from this regression, the measure of normal accruals by the DeFond and Park method varies by quarter and year for each firm, and is thus specifically tailored to each observation in the sample (DeFond and Park, 2001). The intuition behind this model is that the difference between realized working capital and market’s expectations of the level of working capital needed to support current levels of sales is the portion of working capital accruals that is unlikely to be sustained. Therefore, this portion of working capital accruals is expected to reverse against future earnings (DeFond and Park, 2001). DeFond and Parks’ estimate for discretionary accruals therefore measures the difference between realized working capital, and a proxy for the market's expectations of the level of working capital needed to support current sales levels. The discretionary working capital accrual is defined as follows:

t t t t t

WC

WC

SALES

SALES

DWCA





(

/

)



DWCA = Discretionary working capital accruals;

WC = Noncash working capital, computed as (Total current assets – Cash and Short-term investments) – (Total current liabilities – Short-term debt and Current portion long-term debt);

To summarize, I use four different accrual measures to measure earning manipulations: total accruals (TACC), Jones-model discretionary accruals (DACC), and Jones-model discretionary current accruals (DCA), and discretionary working capital accruals (DWCA) according to the DeFond and Park method.

3.4 Data for IPOs

The lack of available data makes it very difficult to find a large sample using strictly pre-IPO data to measure earnings management. Following Teoh, Welch and Wong (1998a), I calculate the accrual variables over the fiscal year when the firm goes public, which includes both pre-IPO and post-IPO months. However, this should not necessarily bias the results since the incentives to manage earnings are likely to persist in the months immediately after the offering. According to Teoh, Welch and Wong (1998a), entrepreneurs usually cannot dispose of their personal holdings until at least several months after the IPO. Furthermore, they state that firms face ‘unusual legal and possibly reputational scrutiny in the IPO aftermath’. This implies that immediate accounting reversals will make earnings management activities transparent enough to bring on lawsuits against the firm and its management. Thus, issuers who aggressively manage their pre-IPO earnings probably also manage their first post-IPO earnings.

Table 5 presents the summary statistics of the IPO sample by accrual quartiles, ranging from the most conservative earnings managing quartile (Q1) to the most aggressive earnings managing quartile (Q4). At a first glance, it seems to be that the most aggressive earnings managing firms have lower book-to-market ratio compared to the more conservative earnings managing firms. There seems to be no relationship between the level of earnings manipulation and the market value of the sample firms.

Table 6 presents the time distribution from one year before to four years after the offering for the key measures of earnings management: total accruals (panel A), the Jones model discretionary accruals (panel B), the Jones model current discretionary accruals (panel C) and the DeFond and Park model discretionary working capital accruals (panel D). A T-test is used to determine if mean accruals are significantly different from zero, while Wilcoxon test is used to determine if median accruals are significantly from zero. The level of statistical significance is 5 per cent or lower, and the level of statistical reliability is 95%.

Panel B shows that mean discretionary accruals are negative in the year before the IPO. In the year of the offering, discretionary accruals are positive, and in the three years following the IPO, discretionary accruals are again negative. The accruals in one year before and one year after the IPO are, however, not significantly different from zero. Median discretionary accruals show similar results, except that the accruals are negative and significant in the year before the IPO.

Panel C shows that on average, discretionary current accruals are negative for one year prior to the offering and one, two and three years after the offering. Accruals only in the first and second year after the IPO are significantly different from zero. However, the median discretionary current accruals are negative and significantly different from zero in the year before the IPO until the fourth year after the IPO.

In panel D, we display the time distribution of the discretionary working capital accruals by the DeFond and Park model. On average, accruals are positive in the year before the IPO and in the year of offering. In the first, second, third and fourth year after the IPO, accruals are negative on average. In contrast to Jones’ model accruals, the mean discretionary working capital accruals are not significantly different from zero, using a significance level of 5 per cent, in the four years after the IPO. However, accruals are again positive and significant in the year of offering. The median discretionary accruals are significantly different from zero in all years, except for one and four years after the offering. Using the DeFond and Park model, I can not find strong evidence of any reversal of positive IPO year accruals in the years after the initial public offering.

The fact that all three Jones model accruals measures are positive and significant in the year of the offering, but negative and significant in subsequent years suggest that accruals are used by firm management to increase reported earnings in the year of offering. Summarized, table 6 shows some evidence of earnings management in the year of the initial public offering and reversal of accruals in the four years following the offering.

3.5 Summary and conclusions

4.

EMPIRICAL RESULTS

This chapter discusses the empirical results and the interpretation of these results. First, I will describe the findings on earnings management persistence, by analyzing the relationship between IPO year accruals and accrual levels in one to four years after the IPO. In section 2, I describe the time-distribution of the post-IPO stock performance, and I compare the post-IPO stock performance between quartiles based on different accrual measures. In section 3, I analyze the influence of IPO year accruals on stock performance using univariate regressions and in section 4 I estimate the multivariate regressions of the influence of accruals on stock performance, including relevant control variables. Finally, I will summarize on the findings in section 5.

4.1 Earnings management and future accruals

As described in the previous sections, an upwards manipulation in earnings by accounting accruals has to be reversed at some point in time. This means that current upward manipulation in earnings will inevitably lead to a decline in future accruals, resulting in lower earnings. This indicates that the accrual component of current earnings should have a negative impact on future earnings. Sloan (1996) finds that most of the reversion takes place in the first year, and reversion is essentially complete after the third year. To measure whether the data used indeed confirm the negative relationship between current and future accrual levels, I measure the influence of earnings management in the IPO year on the accruals after one, two, three, and four years, using the following equation: it it x it

ACCRUALS

ACCRUALS













discretionary accruals in the IPO year. However, none of the coefficients is significant. Panel D shows that discretionary working capital accruals in the first, third and fourth year after the IPO are negatively related to the discretionary working accruals in the year of the offering, while discretionary working capital accruals in the second year is positively related to the discretionary working capital accruals in the IPO year. Only the coefficients in the first and second year are significant, indicating that accruals reverse in the first year after the IPO, but increase in the second year after the IPO.

To summarize, table 7 shows, dependent on the accrual measure, some evidence that accruals in the first and fourth year following the IPO are negatively influenced by the level of accounting accruals in the year of the offering, indicating that accounting accruals reverse. However, dependent on the accrual measure, this table indicates that accruals in the second and third year after the offering are positively influenced by the level of accruals in the year of the offering.

4.2 Earnings management and stock returns

The presence of earnings management alone does not necessarily imply that IPO firms are overvalued and will underperform in subsequent years. If investors would adequately discount the stock price to reflect earnings management, prices would be at an appropriate level. Therefore, the hypothesis that accounting accruals may explain post IPO stock return underperformance requires the presence of both earnings management and investor naivety in interpreting financial statements. To measure the extent to which discretionary accruals have an influence on post-IPO stock return underperformance, a measure for long-term stock performance is needed. Following Teoh, Welch, and Wong (1998a), I measure the long-run performance by analysing the return of an initial public offering for a buy-and-hold investment strategy, since this is most relevant to an investor. This investment strategy presumes that the shares of the IPO firm are received at the first closing price and are kept over a certain period.

In order to test whether investors naively interpret high accounting accruals, I measure the influence of earnings management in the IPO year on the annual market-adjusted buy-and-hold return after one, two, three and four years. Table 8 reports the time-distribution of the post-IPO stock performance, calculated by the percentage market-adjusted annual buy-and-hold return. The table shows that market-adjusted buy-and-hold returns are negative and significantly different from zero on average in the first and second year after the offering. In the third and fourth year after the IPO, returns are positive, and significant in the fourth year. This table indicates that stock performance of IPO firms is significantly lower that the market performance in the first two years after the IPO. In the third and fourth year after the offering, IPO-firm underperformance seems to disappear.

measures. Firms in the first quartile have the smallest accruals, firms in the fourth quartile have the largest, and so on. Then, the average post-IPO market-adjusted stock returns across the quartiles are compared. Table 9 reports the results of the distribution of the post-IPO stock performance per quartile. In panel A of table 9, the quartiles are based on total accruals. In the first and second year after the offering, annual market-adjusted buy-and hold returns of the most aggressive earnings manipulating quartile are negative on average, as well as the returns of the more conservative firms. A t-test does not indicate that mean returns for the firms with the largest discretionary accruals (the fourth quartile) are significantly different than for firms with the smallest discretionary accruals (the first quartile). Only in the third year after the offering, the annual market-adjusted buy-and-hold investment return of the first quartile is significantly higher than in the fourth quartile. The same results hold when quartiles are based on discretionary accruals (panel B). Again, the mean returns for the firms in the fourth quartile are not significantly different than for firms in the first quartile, except in the third year after the offering, where the annual buy-and-hold investment return of the first quartile is significantly higher than in the fourth quartile. When quartiles are based on discretionary current accruals (panel C), the mean returns for the firms in the fourth quartile are not significantly different than for firms in the first quartile in all years following the IPO. The same results hold when quartiles are based on discretionary working capital accruals estimated by the DeFond and Park model (panel D).

Panel E to panel H further analyses the post-IPO stock performance for the most conservative and the most aggressive quartile, analysing only the companies that experience an increase in accruals for the most conservative quartile, and analysing only the companies that experience a decrease in accruals for the most aggressive quartile in the first, second, third and fourth year after the IPO. Even with this more stringent definition of ‘conservative’ and ‘aggressive’, results are qualitatively similar to that of panel A through panel D. Buy–and-hold investment return is significantly higher for the most conservative firms than for the most aggressive earnings managing firms only in the third year after the IPO, except when using Jones model discretionary current accruals as measure for earnings management.

4.3 Univariate analysis

To further test the influence of the accrual variables on post-issue stock return underperformance, I regress the stock performance measure (the IPO market-adjusted post-issue annual buy-and-hold return AR) on the total accruals, the Jones model discretionary total accruals, and the Jones model discretionary current accruals, as well as the DeFond and Park model discretionary working capital accruals, together with a dummy variable for the reversal of the accruals and an interaction term, after one, two, three and four years after the initial public offering. I use the following regression: it it x it x it it x

it

ACCRUAL

Dummy

Dummy

ACCRUAL

AR

















(

*

)





AR = Annual market-adjusted buy-and hold investment return.

ACCRUAL = total accruals, Jones model discretionary total accruals, Jones model discretionary current accruals of DeFond and Park model discretionary working capital accruals.

Dummy = 1 if the accrual has decreased, 0 otherwise.

X stand for the number of years after the IPO. Results for the univariate regressions are presented in table 10. I include a dummy for accrual decreases and an interaction term to extract firms that experience post-IPO decreases in accruals, and analyse whether high IPO year accruals of those firms is related to post-IPO stock performance.

discretionary current accruals as measure for earnings management, the dummy variable is positively related to buy-and-hold returns, and significant in the first and second year after the IPO. The coefficient of the interaction term is negative in most regressions, indicating a negative relationship between IPO year accruals and stock performance for firms with post-IPO decreases in accruals. However, this coefficient is only significant in the first year after the IPO using total accruals as measure for earnings management. Overall, the results in table 10 are rather mixed, and do not provide convincing support for the hypothesis that IPO earnings management by issuers is negatively related to subsequent stock performance.

4.4 Multivariate analysis

To examine the influence of the accrual variables on post- issue stock return underperformance, I regress the measure for stock performance (the IPO market-adjusted post-issue annual buy-and-hold return AR) on the total accruals, the Jones model discretionary accruals and the Jones model current discretionary accruals, as well as the DeFond and Park model discretionary working capital accruals, together with a set of control variables. This set of control variables consist of a set of year-dummies (YEAR) (for presenting reasons, the dummies are reported in equation (5) as one variable), the IPO firm's logged market capitalization (MV) and the book-to-market ratio (BTM), the IPO first-day underpricing return (UnderP), one plus the log of the firm's age (AGE), the change in capital expenditures (CapEx), and the asset- scaled net income growth in the IPO year (NI):

i i i x it x it i

it

ACCRUAL

Dummy

Dummy

ACCRUAL

YEAR

MV

AR

















(

*

)













MTB





Under

Pr





AGE







CapEx







NI





Next, the logarithm of one plus the age of the firm at the IPO date is used as a control variable to control for any systematic influence due to the age of the firm. It is believed that risk is diminished for more established firms and therefore, the long-run performance should be higher on average. For example, Ritter (1991) documents more pronounced long-run underperformance for younger IPOs. Therefore, the coefficient for AGE is expected to have a positive sign in the long-run return analysis. I include the change in net capital expenditures because Cheng (1995) reports empirical evidence that firms which invest the proceeds from equity offerings in fixed assets experience less post-offering underperformance. Capital expenditures will be calculated as the lagged asset-scaled mean capital expenditures in the years after the IPO minus the lagged asset-scaled capital expenditures in year 0. According to Cheng’s hypothesis, the estimated coefficient on the change in capital expenditures in our regression is predicted to be positive. Finally, I include the asset-scaled net income growth in the IPO year to control for increased profitability of the firm.

year after the IPO. Market capitalization (MV) has a positive coefficient, as expected, and the book-to-market ratio (BTM) has both positive and negative coefficients, depending on the accrual measure. Previous literature indicates that smaller initial public offerings are expected to underperform larger IPOs with lower book-to-market ratios. However, coefficients on both measures are not significant. In contrast to expectations, underpricing return has a negative coefficient, but coefficients are again not significant. The control variable AGE, has a positive on the stock performance measure, indicating that younger IPOs underperform the older ones one year after the initial offering. However, this relationship is only significant at a 10 per cent significance level. Furthermore, the coefficient of the change in capital expenditures is positive and significant for most of the regressions, as predicted. This indicates that firms which invest the proceeds from equity offerings in fixed assets experience less post-offering underperformance. Finally, change in net income has a both a positive as a negative coefficient, depending on the accrual measure, but not significant. The adjusted R-squared of these regressions is about 7 per cent.

Panel B shows the regression outcomes of the variables two years after the initial public offering. Total accruals have a negative estimated coefficient of -0.020 (t-statistic -0.986), and discretionary accruals have a negative estimated coefficient of -0.021 (t-statistic -1.012). Both coefficients are not significant. Current discretionary accruals have a negative estimated coefficient of -0.038 (t-statistic -1.743), which is significant at a significance level of 10 per cent. The DeFond and Park discretionary working capital accruals have a negative coefficient of -0.048 (t-statistic of -0.522), which is not significant. The dummy variable for a decrease in accruals has a negative coefficient, except when using the Jones model discretionary accruals as accruals measure. The negative relationship is, however, not significant. When using total accruals as earnings measure, the interaction term has a positive coefficient of 0.024 (t-statistic 0.403), but the relationship is not significant. Same results hold when using other measures for earnings management: the coefficient on the interaction term of the dummy and Jones-model discretionary accruals is 0.032 (t-statistic 0.498), and the coefficient on the interaction term of the dummy and DeFond and Park model discretionary working capital accruals is 0.109 (t-statistic 0.767). The coefficient on the interaction term of the dummy and Jones model discretionary current accruals is negative: -0.039 (t-statistic -0.548). None of the interaction terms is significant. These results indicate that there is no significant relationship between annual returns two years after the IPO and high IPO year accruals. The interaction term shows that even for firms that experience decreases in accruals in two years after the offering, I can not find greater stock underperformance.

discretionary accruals. However, the coefficients are not significant. The book-to-market ratio (BTM) has a positive coefficient, which is significant when using total accruals, discretionary accruals and discretionary working capital accruals as accrual measure. This indicates that initial public offerings with higher book-to-market ratios experience less stock underperformance after the IPO. Underpricing return has a negative coefficient, and AGE has a positive coefficient, as expected, but both not significant. Furthermore, the coefficient of the change in capital expenditures is negative and significant in all models, contrary to expectations. Finally, a change in net income has a negative coefficient, but not significant. The adjusted R-squared of these regressions is rather high: around 25 per cent.

Panel C shows the regression outcomes of the variables three years after the initial public offering. Total accruals have a positive estimated coefficient of 0.089 (t-statistic 1.473). However, this relationship is not significant. The other Jones model accrual measures show similar results: discretionary accruals have a positive estimated coefficient of 0.088 (t-statistic 1.366), and current discretionary accruals have a negative estimated coefficient of 0.075 (t-statistic 0.846). The DeFond and Park discretionary working capital accruals have a coefficient of 0.284 (t-statistic 0.353), which is not significant. These results indicate that firms with high accrual measures in the year of the IPO subsequently do not show significant more underperformance three years after the IPO. The dummy variable for a decrease in accruals has a negative coefficient when using total accruals and the Jones model discretionary accruals as accruals measures, and a positive coefficient for the other accrual measures. These relationships are not significant. When using total accruals as earnings measure, the interaction term has a negative coefficient of -0.512 (t-statistic -1.708), which is only significant at a significance level of 10 per cent. The coefficient on the interaction term of the dummy and Jones-model discretionary accruals is -0.604 (t-statistic -1.708), and the coefficient on the interaction term of the dummy and DeFond and Park model discretionary working capital accruals is 0.791 (tstatistic -1.041). The coefficient on the interaction term of the dummy and Jones model discretionary current accruals is negative: -0.063 (t-statistic -0.369). None of the interaction terms is significant. These results indicate that there is no significant relationship between annual returns three years after the IPO and high IPO year accruals. The interaction term shows that even for firms that experience decreases in accruals in two years after the offering, I can not find greater stock underperformance. Furthermore, none of the control variables have a significant effect on the stock performance measure. The adjusted R-squared is about 11 per cent in these regressions.