1. The discretionary current accruals are significantly positive in the event
year-end for the newly listed firms, but there are problems related to them and make them unreliable.
2. The discretionary current accruals in event year -1 and -2 for the newly listed firms and the conditional conservatism test suggest that the sample firms do not engage in earnings management and even report more conservatively in the two years before newly listed.
3. There may be potential biases in discretionary current accruals when using the cross-sectional modified Jones model to estimate them due to the out-of-sample estimation problem.
This chapter also makes some suggestions for further study.
The flow chart of this study is as follows:
Figure 1.2: Research Flow Chart Research Motivations
Develop Research Topics
Relevant Earnings Management Literatures Review
Understanding the TSE/GTS Market in Taiwan Relevant IPO
Literatures Review
Develop Hypothesis
Sample Selection and Data Gathering
Examine Discretionary Current Accruals around the Event Year
Out-of-sample Estimation Test
Empirical Results and Analysis
Conclusion and Suggestion Examine
Individual Working Capital Components Changes in Year 0
Conditional Conservatism Test
Chapter 2. Literature Review
This study mainly examines whether the newly listed firms engage in earnings management before their listing in Taiwan Stock Exchange market and GreTai Securities market. The discretionary current accruals is the main proxy for earnings management. To focus on this topic, this chapter reviews the literatures in the following three sections. First, this chapter organizes the famous literatures about accruals estimation. Second, this chapter discusses the literatures about earnings management at IPO. The literatures about earnings management for newly listed firms in Taiwan are in the last section.
2-1 Accruals Estimation Related Literatures
Accruals is the most popular proxy for earnings management because it reflects the overall impact of the accounting policies adopted by firms. How to estimate accruals and distinguish the discretionary portion from the total accruals are popular researches in accounting literatures. This section discusses some famous accruals estimation related literatures. This paper uses the change in sales to control the effect of firms’ business fluctuations (Jones [1991]) and subtracts the change in accounts receivable from the change in sales when calculating the nondiscretionary current accruals (Dechow et al. [1995]) because Dechow et al. (1995) show that the modified version of the Jones model has the better power to detect earnings management.
Healy (1985)
Healy examines the relation between managers’ managerial accounting decisions and earnings-based bonuses due to the incentives for managers to increase their compensation. This study is different from preceding papers because Healy not only examines the accruals which increase the earnings but the accruals which decrease the earnings. Healy separates its sample into three subsamples based on the existence of bonus plan upper bounds and lower bounds and compares the average accruals in each
subsample to observe if managers engage in earnings management.
Healy assumes managers are likely to adopt income-decreasing accruals when their performance is over the upper bounds of the bonus plan or under the lower bounds of the bonus plan in order to delay the earnings to the next period, and to adopt income-increasing accruals when their performance is in the middle of bounds to increase their bonus. The results in this study support its hypothesis that the accruals in company-years without upper and lower bounds of the bonus plan are significantly lower than those with upper and lower bounds. Healy considers that the accounting earnings are composed of cash flow from operations, discretionary accruals, and nondiscretionary accruals. Accruals are the differences between net income and operating cash flow.
t,
t
t DAC NDAC
TAC = + (1)
where:
TAC : Total accruals in year t, t
DAC : Discretionary accruals in year t, t
NDAC : Nondiscretionary accruals in year t. t
Healy also mentions that the discretionary accruals would be biased if we cannot distinguish these two portions of accruals accurately.
DeAngelo (1986)
DeAngelo investigates the accounting decisions made by firms during the management buyouts period. Due to the incentives for the firms to reduce the buyout compensation, the author hypothesizes that sample firms understate their earnings in periods before the management buyout, but the results are not consistent with the hypothesis. The author thinks the most plausible explanation is that, public
stockholders and their financial advisors carefully examine the financial statements of these firms for evidence of income-decreasing accounting.
In this study, DeAngelo (1986) revises the Healy (1985) model and uses the change in accruals to reduce the nondiscretionary accruals’ effect on earnings
manipulation test. The formula is:
t TAC DAC DAC NDAC NDAC
TAC (2)
DeAngelo assumes the fluctuations in nondiscretionary accruals are random walk, so the expected value of the change in nondiscretionary accruals is zero. Under this assumption, the expected value of total accruals change represents the expected value of discretionary accruals change. So this paper uses the change in total accruals as a proxy for earnings management.
Jones (1991)
Jones uses empirical research to find out if firms try to decrease earnings by earnings manipulation during import relief investigation by the United States
International Trade Commission (ITC) in order to get benefit. Because ITC considers accounting numbers as the factors to decide which firm can get import protection, this provides an incentive for firms to manage their earnings. There are 25 companies in five different industries in the sample. The result of this study supports the hypothesis that managers make income-decreasing accruals during investigation, especially through the discretionary accruals.
Jones relaxes the assumption that nondiscretionary accruals are constant from period to period and develops the following model to estimate non-discretionary accruals:
Jones assumes that nondiscretionary accruals of a firm would fluctuate from year to year because of the change in the economic circumstances and uses revenues to control the effect. Jones considers that the revenues can measure the firms’ operation conditions objectively. The reason the Gross PPE is included in this model instead of change in this account is that the total depreciation expense is included in the total accruals. The error term is the discretionary portion of the total accruals.
Friedlan (1994)
Friedlan assumes the IPO firms inflate their earnings in order to obtain the higher IPO prices and examines the accounting decisions made by issuers. The empirical results are consistent with the hypothesis that firms make income-increasing discretionary accruals before IPO.
Friedlan considers that the DeAngelo (1986) model is not suitable for the IPO firms because the IPO firms usually have high growth rates and the unusual growth affects the IPO firms’ operating performances, including accruals. Using the
DeAngelo model will contribute all the changes in total accruals to changes in discretionary accruals, but in fact, part of the changes should be contributed to the growth factor. The revised model in this study is:
− •
Friedlan revises the DeAngelo (1986) model and uses the sales growth rate to adjust the lagged total accruals, then standardizes the change in total accruals by this year’s sales.
Dechow, Sloan, and Sweeney (1995)
There are several popular models in earnings management literatures to estimate the discretionary and non-discretionary accruals. Dechow, Sloan, and Sweeney compare the specifications and power of test to evaluate these alternative models,
including the Healy Model, the DeAngelo Model, the Jones Model, and the Industry Model. They also develop a modified version of the Jones Model that the
nondiscretionary accruals are estimated as following:
), / ( ] / ) [(
) / 1
( −1 + 1 ∆ −∆ −1 + 2 −1
= i it i it it it i it it
it A REV REC A PPE A
NDA α β β (5)
where:
RECit
∆ = Net receivable in year t less net receivable in year t-1 for firm i.
The parameters used to calculate nondiscretionary accruals are still estimated from original Jones (1991) Model. The only difference between the Jones Model and the modified version is that when calculating nondiscretionary accruals, the modified Jones Model subtracts the change in net receivables from the change in revenues. This process implicitly assumes that all changes in credit sales are due to earnings
management. The reason is that it is easier to manipulate earnings through recognition of revenues on credit sales than on cash sales.
After a series of experiments, this study shows that all the models mentioned above are well specified for a random sample of event-years, but all lead to
mis-specified tests when firm-years having extreme financial performances. Besides, if variables used to detect earnings management are correlated with firm’s
performances, all the models considered are potentially mis-specified. So it is important to control these factors. Finally, the modified version of the Jones Model developed by Dechow, Sloan, and Sweeney shows the most power to detect earnings management.
Hribar and Collins (2002)
Estimating accruals is a very common procedure in accounting literatures. Due to the data availability problem of cash flow statements, many people use the successive balance sheets to estimate accruals. But this method is based on a presumption that the changes in working capital components are related to the accrual component of
revenues and expenses on the income statement. Once there are non-operating events, this presumption does not stand. The authors examine the accruals estimated both
from balance sheet changes and cash flow statements and demonstrate the errors caused by using balance sheets to estimate accruals.
The authors use three main non-operating events to examine the estimated accruals, which are mergers/acquisitions, divestitures, and foreign currency
translations. They divide their sample into four subsamples: firm-years with a merger or acquisition, firm-years with discontinued operations greater than $10,000,
firm-years with gain or loss on foreign currency translations greater than $10,000, and firm-years with none of the above events.
This paper shows that the accruals estimated from balance sheets in the “merger”
subsample are positively biased, and negatively biased in the “discontinued operations” subsample. So the authors suggest using the cash flow statements to estimate accruals, especially when there are non-operating events in the sample firm-years.
2-2 Earnings Management around IPO process Literatures
Literatures about earnings management around large transactions and events are countless. This section reviews some earnings management around IPO process literatures because the topic of this study is earnings management for newly listed firms in Taiwan. This study observes the significantly positive discretionary current accruals in the event year-end for newly listed firms as the evidence in Teoh et al.
(1998b) study. The difference between my study and Teoh et al. (1998b) is that they interpret these positive DCAs directly into earnings management for IPO firms. But this paper illustrates the potential biases related to the discretionary current accruals in the event year mentioned by Ball and Shivakumar (2007) and uses the DCAs in the event year -1 and -2 to examine the earnings management hypothesis instead of those in the event year.
Teoh, Welch, and Wong (1998b)
Based on Ritter (1991), “investors are periodically overoptimistic about the
earnings potential of young growth companies,” this study examines whether issuers of initial public offerings inflate their earnings by accruals in the IPO year and the subsequent stock returns. The authors assume that the IPO process is a good timing for issuers to manage their earnings due to the asymmetry information between investors and issuers. They also assume that the investors cannot identify which IPO firms manipulate earnings, so the high earnings on the financial statements included in the prospectus directly translate into a higher offering price. When these IPO firms start to trade on the stock market and are unable to maintain the earnings level, the investors begin to realize they were too optimistic about these companies and start to realize the real values of these firms. This would be reflected on the stock prices and caused the poor stock return performances thereafter.
The sample in this study includes 1,649 IPO firms in the U.S. between 1980 and 1992. The financial statement information is taken from the IPO year-end (year 0), so the numbers on the financial statements include both pre- and post-IPO periods.
Although the discretionary long-term accruals also represent earnings management, the authors emphasize the discretionary current accruals as the main variable because it is easier to control current accruals than long-term accruals. The nondiscretionary portions of total accruals reflect the change in business conditions of firms, so they are not considered proxies of earnings management. This paper uses the modified Jones Model to estimate the DCA at year 0 and divides the sample into four groups by DCA quartiles. Firms with highest DCA are included in the “aggressive” group and the group with lowest DCA is called “conservative”.
This study examines the relation between IPO firms’ earnings management and the long-term post-IPO stock underperformance. The empirical results show that the discretionary current accruals of IPO firms in year 0 are higher than non-issuers, and on average, firms classified in the most aggressive group have a 15% to 30% inferior performance than those classified in the most conservative group in the subsequent three years. In summary, the empirical results in the paper support the hypothesis that IPO firms inflate their earnings during the IPO process and the earnings management is related to the subsequent stock price performance.
Teoh, Wong, and Rao (1998c)
This paper examines mainly the issue-year earnings and the abnormal accruals in the IPO firms and the relation between the issue-year abnormal accruals and the subsequent stock returns.
The sample in this study consists of 1682 IPO firms going public between 1980 and 1990. IPO firms must meet the following criteria to be included: offer price > $1, gross proceeds > $1M, only common stock is offered, and the offering is handled by an investment banker. The methods used to estimate abnormal accruals in this paper include cross-sectional modified Jones (1991) model and Beneish (1994) M-score model. The operating performance is measured by three variables, return on sales of the IPO firms, industry-adjusted return on sales, and return on sales in relation to matched firms.
The empirical results reveal that the IPO firms report high earnings during the IPO process by reporting abnormal accruals; after the IPO, their earnings are significantly lower than non-issuing industry peers and matched non-issuers. Those IPO firms with the highest abnormal current accruals in the event year
underperformed the most in the following three years; but the expected current accruals, abnormal long-term accruals, and expected long-term accruals do not have the forecast power. Besides, the IPO firms are more likely than their peers to adopt accounting policies which could increase their earnings. Finally, the evidence in this paper is consistent with Teoh et al. (1998b) that the more abnormal accruals of IPO firms have, the worse their subsequent stock returns.
Ball and Shivakumar (2007)
Unlike popular literatures in the earnings management territory, the authors hypothesize the IPO firms report their financial statements more conservatively, because when one firm is going to initial public offering, it faces a higher demand for reporting quality from a broad range of stakeholders, like internal auditors,
independent auditors, boards, corporate lawyers, analysts, etc.
They also question the hypothesis and evidence of Teoh et al. (1998b) that firms inflate earnings during IPO year in order to get higher IPO prices. First, if IPO firms
inflate earnings by earnings management, they would attract the scrutiny from many market monitors, like auditors, analysts, as well as the regulatory, and might be detected. Second, poor reporting quality would cause the cost of capital to rise, which is not good for a growing firm. Third, the evidence from Teoh et al. (1998b) study is based on discretionary current accruals, which include the changes in working capital, estimated from modified Jones Model. The authors find that these accruals are
unreliable due to the following reasons:
The discretionary current accruals are too large to be reliable in Teoh et al.
(1998b) sample and would be detected by market monitors.
The accruals are estimated from balance sheet changes and are biased in the earnings inflation hypothesis.
Using accruals in year 0 is too late to influence the IPO price.
The DCA estimates are biased because of the high growth of these IPO firms and the IPO proceeds.
Some unusually high discretionary accruals are caused by the low value of deflator, pre-IPO total assets.
This study re-examines the Teoh et al. (1998b) sample of 1,649 U.S. IPO firms from 1980 to 1992, calculates current accruals both from balance sheet and cash flow statement, and uses either the Jones Model or the following piecewise linear Jones model adapted from Ball and Shivakumar (2005, 2006) to estimate normal current accruals:
,
4 *
3 2
1
0 it it it it it it
it Sales CFO DCFO DCFO CFO
CA =α +α ∆ +α +α +α +ν (6)
where:
CFO =Cash flow from operations, it
DCFO =Dummy indicator for negative cash flows that takes the value 1 if it
CFO <0 and 0 otherwise. it
The authors use the discretionary current accruals in year-1 instead of year 0 to run regressions and to do analysis. The empirical results of this paper are consistent with the higher reporting quality hypothesis. The authors also conclude that using the
traditional ways to estimate discretionary accruals around large transactions and events are unreliable.
2-3 Earnings Management for Newly Listed Firms in Taiwan
This section reviews the related literatures about earnings management for newly listed firms in Taiwan because the sample in this paper includes newly listed firms in TSE market and GTS market. The differences between this paper and the others below are as follows. This paper assumes the discretionary current accruals have the better power to test earnings management hypothesis because the managers have greater flexibility and control over current than long-term accruals (Teoh et al.
[1998b]). So this study uses the discretionary current accruals as the proxy for earnings management (Tai [1999]) instead of discretionary accruals (Jeng [1992], Su [1992], Lien [1993], Chen [1993], and Huang [1995]). This study finds the
significantly positive discretionary current accruals in the event year-end, consistent with Huang (1995) and Tai (1999), and further examines the individual working capital components changes. This study assumes the discretionary current accruals in the event year -1 and -2 can better test the earnings management hypothesis for newly listed firms because the potential biases with the discretionary current accruals in the event year (not mentioned by relevant literatures in Taiwan) and the criteria for newly listed firms to qualify (Table 3.1). The studies below only focus on the newly listed firms in the TSE market, but the sample in my study includes the newly listed firms both in the TSE market and GTS market. The conclusions in this paper are consistent with Huang (1995) that firms do not engage in earnings management before newly listed but not with Su (1992), Lien (1993), Chen (1993), and Tai (1999). Due to the preceding reasons why this paper is different from following studies, I believe the variables, measurement periods, and methodology this paper adopts provide more reliable evidence to test whether firms engage in earnings management before newly listed.
Jeng (1992)
The sample in Jeng (1992) study includes 60 newly listed firms between 1986 and 1990. This study examines whether newly listed firms manipulate their earnings
The sample in Jeng (1992) study includes 60 newly listed firms between 1986 and 1990. This study examines whether newly listed firms manipulate their earnings