Earnings management practices for venture IPO firms.

Yoon, Soon Suk ; Kim, Hyo Jin

INTRODUCTION

The purpose of the study is to examine the earnings management practices for the Kosdaq venture firms in comparison to the Kosdaq non-venture firms and the KSE firms when they go public. The Kosdaq exchange, the equivalent of the Nasdaq exchange in the USA, was formally established in 1996 by combing formerly less structured and relatively dormant over-the-counter (OTC) stocks into a more lively market system. The Kosdaq exchange grew slowly until 1998. As of December 31, 2001, the number of Kosdaq firms surpassed the number of firms listed on the Korea Stock Exchange (KSE). (1)

There are three different listing regimes in two stock exchanges for IPO firms to choose from in Korea: the KSE, the Kosdaq non-venture section and the Kosdaq venture section. Firms must satisfy stringent listing criteria to be listed on the KSE. (2) The Kosdaq non-venture section requires less stringent listing requirements (3) as compared to the KSE. In contrast, the Kosdaq venture section imposes virtually no restrictions if a firm is designated as a venture firm by the Korean government. Specifically, no restrictions are imposed on financial requirements including paid-in-capital, stockholders' equity, total assets, profits or debt ratios as long as it is a venture firm. Venture firms are so-called 'new economy' firms consisting mostly of technology or information oriented firms.

The rapid growth of the Kosdaq market with lenient listing regulations can in part explain the resulting high delisting rate. Over the five-year period from 1997 to 2001, the delisting rate for the Kosdaq exchange firms was 28.4% as compared to 17.9% for the KSE exchange.

The Kosdaq firms may face delisting because of the following events:

* false or serious omissions in the listing application and the attached documents,

* designation as an unfaithful disclosure firm more than 3 times during the past 2 years,

* default, suspension of primary business for more than 6 months, or consecutive full encroachment of paid-in capital for more than 2 years,

* inactive stock transactions for more than 3 months,

* respective stock prices are less than 20% of its par value for a certain period.

Strict listing regulations, on the one hand, may help ensure that investors have opportunities to invest in relatively safer firms. On the other hand, however, small firms with good growth potential may have limited access to public capital providers merely because they were not able to satisfy the strict KSE listing requirements. These small firms may face additional financial difficulties or be forced to forgo good investment opportunities. Therefore, there is a delicate trade-off between strict listing regulations and promotion of small firms with good growth potentials. Because the KSE and the Kosdaq market employ different listing regulations, firms may self-select stock exchanges depending on their respective firm characteristics.

Earnings management may occur when managers use their discretion in financial reporting and in structuring transactions to alter financial reports. These actions may either mislead some stakeholders about the underlying performances or may influence contractual outcomes that depend on reported accounting numbers. Incentives to manage earnings have been identified when earnings directly or indirectly affect share prices, equity issues, proxy fights, labor contracts, management compensation or government regulations.

Prior studies document that equity issuing firms tend to overstate earnings around the equity issues (Aharony, Lin & Loeb, 1993; Friedlan, 1994; Loughran & Ritter, 1995; Loughran & Ritter, 1997; Rangan, 1998; Teoh, Welch & Wong, 1998; Shivakumar, 2000; Yoon, 2001; Yoon & Miller, 2002a). Yoon (2005) compared earnings management for the KSE firms with the Kosdaq firms in general. He concluded that the Kosdaq firms tend to manage their earnings more aggressively than KSE firms when the operating cash flows are either significantly poor or good. The current study is different from Yoon (2005) in that the current study focuses on the differences in earnings management practices motivated by differential institutional listing arrangements for IPO firms in Korea.

In comparing the earnings management practices of the IPO firms in Korea, we will focus on the comparison of "levels of earnings management" rather than actual accounting policy choice for earnings management. Firms may use diverse methods to manage earnings. Aggregate accruals approach like the current study is not concerned about actual accounting policy choices. Specific accrual approach mainly examines the choice of specific accounting method to compare earnings management practices. McNichols & Wilson (1988), for example, use the specific accrual approach in examining earnings management.

Our study examines the earnings management practices for the Kosdaq venture IPO firms in comparison to the Kosdaq non-venture IPO firms and the KSE IPO firms from 1996 to 2000. We expect that firms going public will have strong incentives to manage earnings to make sure that they pass the listing requirements. The major reason for the incentives to manage earnings is that the IPO firms hope to sell out their new shares at good prices.

Earnings management studies require models that will properly separate discretionary accruals from total accruals. This study uses two methods for discretionary accrual estimation. One is the modified Jones model (Dechow et al., 1995) and the other is a modified version of Yoon & Miller (2002). Yoon and Miller indicate that the model fits relatively well for Korean firms as compared to the modified Jones model.

We use two methods to test differences in earnings management practices for IPO firms. The first method is the tests for mean accrual differences for earnings management practices between the Kosdaq venture firms and their control samples of the Kosdaq non-venture firms and the KSE firms. The second method is a graphic approach that is easier to understand by readers without strong knowledge on statistics. The graphic approach reveals a richer spectrum of average accruals across different portfolios. For the graphic approach, we rank the treatment and control samples based on CFOs to form quintile portfolios. We then analyze the mean accruals for each CFO portfolio.

RESEARCH DESIGN

Hypotheses

We assert that earnings management of IPO firms can be associated with different institutional listing regulations. Differential listing criteria will lead the would-be IPO firms to self-select the stock exchanges which will best serve their strategies. Particularly when listing requirements are less strict and the possibility of getting penalized is low, firms contemplating IPOs will have incentives to overstate their values so that stocks prices can be maximized.

Given the presence of information asymmetry, less stringent listing regulations for the IPO firms are likely to be associated with increased adverse selection problem. The cost of the adverse selection problem for the IPO firms will be wealth transfers from the new shareholders to existing inside shareholders. Wealth transfers amount to the premium paid by the new shareholders over the underlying values. Wealth transfers of this nature can be more serious for the Kosdaq IPO firms since less strict regulations are imposed on them, resulting in lower levels of transparency. Therefore, we expect that the Kosdaq firms will manage earnings more aggressively when they go public than their KSE counterparts do.

In addition to the different institutional arrangement for the IPO firms, there is one more factor that is likely to motivate firms to overstate earnings to sustain boosted share prices. One of the common features of the IPO firms in Korea is that the underwriters of IPOs are required to maintain the prices of newly offered stocks above certain levels to protect new shareholders up to three months after the IPOs. Also required is that the original shareholders are not allowed to sell their stocks until six months after the IPO. This will lead to earnings management practices by all IPO firms particularly in the IPO year. In other words, all IPO firms will have incentives to manage earnings, but the differential listing regulations will induce firms to self-select the stock exchange that will best serve their needs. Furthermore, when IPO firms choose the exchange with less strict regulations, they will be likely to exercise more opportunistic behavior.

Deriving from the above discussion, we state our first research hypothesis as follows:

H1: Kosdaq firms will manage earnings more aggressively than their KSE counterparts when firms go public.

The listing regulations within the Kosdaq market are different depending on the types of firm. Venture firms are less strictly regulated than non-venture firms in many respects including IPO regulations. Therefore, we expect that the earnings management practices for the venture firms will be more aggressive than the non-venture firms especially when they go public. Hence, we can set up our second research hypothesis as follows:

H2 Kosdaq venture firms will manage earnings more aggressively than their Kosdaq non-venture counterparts when firms go public.

Sample Selection

As discussed earlier, the Kosdaq exchange was formally established in July 1996 by organizing formerly less structured and relatively dormant 343 over-the-counter (OTC) stocks into a more lively market system. The Kosdaq exchange began to draw investors' attention slowly in 1996 and 1997. Only 331 firms were listed as of December 31, 1996. Ten firms went public in 1996. However, forty-five firms got delisted during 1996. This indicates that the Kosdaq market was very unstable in the early years of inception.

As of December 31, 2001, 721 firms including 80 mutual funds were listed on the Kosdaq exchange. During our study period, 224 new firms were listed on the Kosdaq market and 81 on the KSE. A total of 16 Kosdaq firms and 10 KSE firms were excluded from the sample because some major financial variables were not available for those firms. As a result, the final sample consists of 208 Kosdaq firms (75 venture firms and 133 non-venture firms) and 71 KSE firms. For some firms with outlying values, we have winsorized them to the 1st and 99th percentiles for individual variables rather than eliminating them from the sample because the original sample was small already.

Two reasons can be provided for the selection of our study period. First, the Kosdaq market was launched in 1996. Even though the former OTC firms were merged into the Kosdaq market, the new market setting for the Kosdaq market is different from its predecessor because the market operates in a more structured way than its predecessor. Therefore, we have not included the OTC market period in our study period. Second, the Korean accounting standards mandated cash flow statements starting in 1995. Some prior studies use a balance sheet approach to estimate total accruals. However, Collins & Hribar (2002) shows that the error resulting from using the balance sheet approach not only reduces the discretionary accrual model's power to detect earnings management, but also has the potential to generate incorrect inferences about earnings management. Therefore, we use cash flows from operations as given in the cash flow statements to avoid measurement errors resulting from the balance sheet approach.

Estimation of Discretionary Accruals

We need to identify a model that most appropriately estimates the discretionary accruals. Researcher' ability to accurately estimate discretionary accruals critically affects the success of earnings management tests. In other words, all tests are joint tests of the researcher's model of discretionary accruals and earnings management. Therefore, the development of a well fitting model is very important for this line of research. Prior research documents that the modified Jones model (Dechow et al., 1995) is generally effective. Others like Kothari et al. (2005) document that the modified Jones model ('MJ model' hereafter) is severely misspecified and tends to over reject the null hypothesis of earnings management. Yoon & Miller (2002a) also document that the MJ model does not fit well particularly for Korean firms. In this research, we use two alternative models. The first one slightly modifies Yoon & Miller model ('YM model' hereafter) and the second one is the MJ model. Since the latter has been widely used and described in prior studies, we will not describe it in this study. The YM model is described below.

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]

Where,

TA (total accruals) = NI (accounting earnings) - CFO (cash flow from operations)

REV = net sales revenue

REC = trade receivables

EXP = sum of cost of goods sold and selling and general administrative expenses excluding

non-cash expenses

PAY = trade payables

DEP = depreciation expenses

PEN = retirement benefits expenses

[DELTA]) = change operator]

The YM model posits that total accruals will normally depend on changes in sales revenue, changes in expenses and some non-cash expenses including depreciation and retirement benefits expenses. We use the first two variables as proxies for current accruals (or working capital accruals) and the third variable as proxy for non-current accruals. The two components of accruals have both aspects of non-discretionary and discretionary nature. The model tries to separate non-discretionary portion from total accruals and uses residuals as discretionary accruals by regressing total accruals on the three variables. According to Dechow & Dichev (2002), current accruals are mostly positive since most firms are growing and increasing in their working capital while non-current accruals and total accruals are negative primarily because of depreciation.

The first explanatory variable, (DREV-DREC)/REV, represents changes in non-discretionary revenues because we subtract changes in receivables from changes in revenue. This variable should capture a firm's tendency to increase reported earnings by increasing the front-loading of credit sales. Front-loading of sales would tend to decrease this variable since it would not increase the numerator while simultaneously increasing the denominator. In other words, the change in cash sales should not be affected by the frontloading of credit sales.

The second explanatory variable, (DEXP-DPAY)/REV, represents changes in non-discretionary expenses. Management may utilize not only sales but also expenses in managing reported earnings. Hence, unless we properly take into account both sales changes and expense changes, we may not properly capture the dual aspects of current accruals. One of the major weaknesses of the MJ model is that the first variable does not have a predicted relationship with total accruals. Total accruals are expected to have a positive relationship with current accruals. However, it is difficult to predict the relationship that the changes in sales will have with total accruals. Sometimes, sales and receivables will be utilized to manage earnings, whereas in other times expenses and payables can be utilized for the same purpose. If we include the first variable only in our model, we may still capture the impact of expense changes on the current accruals because revenue changes and expense changes are correlated to a certain degree. (4)

The third variable associates non-cash expenses with non-current accruals. A non-discretionary level of non-cash expense is represented by the sum of depreciation expenses and retirement benefits expenses. When retirement benefits liabilities are less than fully funded which is the case for most of the Korean firms, the unfunded portion of the retirement benefits expenses are non-cash expenses. In the cash flow statements, this information is provided together with other non-cash expenses. By construction, the third variable will have a negative relationship with total accruals since non-cash expenses are subtracted from cash flows from operations to arrive at reported earnings.

We assert YM model is better than the MJ model for several reasons. First, YM model does not impose the intercept term to go through the origin while the MJ model does so by suppressing the intercept term. (5) Kothari et al. (2005) also point out that failure to include a constant magnifies misspecification of the MJ model. Second, as mentioned above, the YM model will capture the fact that firms tend to use not only revenues but also expenses in making accounting choices for earnings management. Third, the YM model better ensures a proper matching between the dependent variable and the independent variables since the YM model uses the flow variables for the independent variables as well as for the dependent variable. In contrast, the MJ model uses a level variable (Property, Plant &Equipment/Total Assets) to proxy for the non-current accrual. (6) Matching a level independent variable with the flow dependent variable is expected to lower the explanatory power of the level variable.

We used panel data in estimating the discretionary accruals to ensure statistical robustness. We combined the financial statements for the listed firms from the period of 1995 through 2001 to make the panel data. All the variables were electronically retrieved from the cash flow statements that are available from KISFAS, the Korean Compustat data set. We maintained as many firm-year observations as possible by winsorizing outliers instead of eliminating the respective firms. We also tried to keep the winsorization to a minimum level to maintain the characteristics of the original data.

Discretionary accruals are obtained by subtracting fitted values of accruals (non-discretionary accruals) from total accruals as follows:

[DA.sub.i] = [TA.sub.i] / [RE.sub.Vi] - [[b.sub.0] + [b.sub.1] ([DREV.sub.i] - [DREC.sub.i)/] [REV.sub.i] + [b.sub.2] ([DEXP.sub.i] - [DPAY.sub.i)/] [REV.sub.i] + [b.sub.3] ([DEP.sub.i] + [PEN.sub.i])/ [REV.sub.i]] (2)

Here, [b.sub.k] represents the estimated coefficients of [b.sub.k] in equation (1). The discretionary accruals (DA) obtained from the equation (2) represent the differences between actual total accruals and expected (non-discretionary) total accruals for each observation.

Estimation Results for Discretionary Accruals

We ran regressions for the YM model and the MJ model by industry for 32 two-digit industries. A total of 6,429 firm-year observations from the period of 1995 to 2001 were used to estimate the discretionary accruals. The data set used for the estimation of discretionary accruals is much larger than the sample data set (279 observations: 75 Kosdaq venture IPO firms, 133 Kosdaq non-venture IPO firms and 71 KSE IPO firms). The large data set helps ensure robust estimations for the discretionary accruals.

The number of observations used in the estimation for each industry ranges from a low of 28 for the Fishing industry to a high of 798 for the Electronics industry. The adjusted R2 ranges from a low of 0.000 for the Film industry to a high of 0.683 for the Education and Business Support industry for the YM model. The adjusted R2 for the 32 industries averages 0.237 for the YM model, a modest goodness of fit. However, the average adjusted R2 for the MJ model is 0.042, which is much poorer than the YM model. Detailed explanations for the comparison of the two models are omitted here. Instead, the current study will briefly interpret the results of the YM model.

The explanatory variables for the YM model had expected signs with strong statistical significance in general. More specifically, the first variable, (DREV-DREC)/REV, generally had negative relationships with the total accruals. The proportions of negative coefficients for the first variable were 75% (24/32). 31% (10/32) were statistically significant at the 5% significance level. The second variable, (DEXP-DPAY)/REV, had predominantly positive relationships with total accruals. 78% (25/32) of the coefficients were statistically significantly positive at the 5% significance level. 94% (30/32) were positive. The third variable, (DEP+PEN)/REV, also had generally negative relationships with total accruals. The proportion of negative coefficient for the third variable was 63% (20/32). 28% (9/32) were statistically significant at the 5% significance level.

Test for Earnings Management

This study's test period covers 1996-2000 while the estimation for discretionary accruals uses seven-year data from 1995 to 2001. The discretionary accruals are estimated by running the regressions for the YM model and the MJ model by industry for 32 industries.

Two methods are used to test differences in earnings management practices between the Kosdaq venture IPO firms and the two control groups of IPO firms. First, the study uses difference tests for accruals. We do not expect that there will be a difference in average accruals between the Kosdaq venture IPO firms and the control sample firms unless they employ differential earnings management practices. Any systematic differences in accruals may be attributed to the differential listing requirements. We hypothesize that the Kosdaq venture IPO firms will have stronger incentives to manage earnings because of the differences in institutional arrangements of the IPO markets.

Second, the current study uses a graphic approach that is a similar method suggested by Burgstahler & Dichev (1997). The difference of our approach from Burgstahler & Dichev's lies in the fact that we use a histogram method while the other uses a scatter diagram method. This method presents a broader picture for earnings management practices by graphically portraying the average accruals across more narrowly classified cash flows from operations ('CFO' hereafter) portfolios. McNichols (2000) argues that the graphic approach allows researchers to make a strong prediction about the frequency of earnings realizations that is unlikely to be due to the non-discretionary component of earnings.

For the graphic approach, we rank sample firms based on their standardized CFOs. We then form five equal size CFO portfolios. Mean accrual for each CFO portfolio for three comparative groups is presented in a histogram form. Systematic patterns of the differences for average accruals among three groups will suggest differential earnings management practices. We expect that the average accruals will show a monotonic decrease from the worst CFO quintile portfolio to the best CFO quintile portfolio, given the fact that the degree of earnings management practices hinges upon the level of CFO. In addition, we expect that the pattern for monotonic decreases for average accruals will be most salient for the Kosdaq venture IPO firms.

One advantage for the graphic approach is that this approach presents richer information about the distribution for average accruals across the different CFO portfolios. The mean difference method tests show only that the central tendency of accruals, i.e., the difference of the overall mean of accruals among three groups. In the mean difference test, any positive and negative deviations from the overall mean will cancel each other out. As a result, we may fail to see differences in discretionary accruals as long as the overall mean accruals are similar even if mean accruals for each CFO portfolio may exhibit significant differences across the CFO portfolios.

The graphic approach mitigates misspecification of regression approach in estimating discretionary accruals since the graphic approach provides a proper controlling for performance. Kothari et al. (2005) shows that a performance-matched accrual is useful in mitigating errors where researchers' partitioning variable is correlated with performance. Our graphic approach also controls performance. Performance matching can remove the portion of earnings management that is motivated by poor or superior performance since both treatment and control firms by design experience similar CFOs.

EMPIRICAL RESULTS

Descriptive Statistics

Table 1 reports the descriptive statistics for some key variables for the three year period surrounding the year of IPO in a comparative form for the Kosdaq venture firms, the Kosdaq non-venture firms and the KSE firms. We provide descriptive statistics for the alternative discretionary accruals (DA) based on the YM model and the MJ model. Furthermore, we also provide descriptive statistics for total accruals (TA), net income (NI) and cash flows from operations (CFO) using two different deflators of net sales and beginning total assets (BTA) for the sake of comparison. However, we will focus on the discretionary accruals in the analysis and interpretation of the empirical results for accruals. Panels A, B and C respectively show the descriptive statistics for those variables for one year before the IPO, the IPO year and one year after the IPO.

We get much wider ranges and standard deviations when BTA is used than when net sales are used as deflators. This is particularly the case for the Kosdaq venture firms in the year before IPO and the IPO year, for the Kosdaq non-venture firms in the year before IPO and the IPO year, and for the KSE firms in the IPO year. In the same vein, mean accruals (both total accruals and discretionary accruals), net income and cash flows from operations are more inflated when BTAs are used than when net sales are used as deflators, especially in the year before IPO and the IPO year.

DA averages 0.0178 for the YM model (0.0637 for the MJ model) for the Kosdaq venture firms in the year before IPO. Average DAs for the two control samples are -0.0032 (-0.0134) for the Kosdaq non-venture firms and 0.0195 (0.0212) for the KSE firms during the same period.

We can also observe that TA is negatively related to CFO so that NI is the highest for the Kosdaq venture firms at 0.1074 (0.2104) while its CFO is the lowest at 0.0472 (0.0882) among the three groups. This may indicate that the Kosdaq venture firms tend to increase reported earnings more aggressively than their counterparts when they prepare to go public in the following year.

Panel B shows that DA for the Kosdaq venture firms (0.1006 and 0.1792) is much larger than those of the control groups (0.0224 and 0.0539 for the Kosdaq non-venture and 0.0419 and 0.0639 for the KSE). Shown in Panel B is that CFO is the lowest and yet NI is the highest for the Kosdaq venture firms in the IPO year. This indicates that the Kosdaq venture firms tend to manage earnings heavily in the IPO year as well.

Panel C reveals that DA (-0.0162 and -0.0427) and NI (-0.0418 and -0.0130) for the Kosdaq venture firms is lowered drastically as compared to the previous two years. NI is also lower as compared to the two control samples. Perhaps the Kosdaq venture firms might have exhausted accruals to increase their NI in the year preceding IPO and in the IPO year. As a result, NI decreases drastically in the year following IPO.

Results for Mean Difference Tests

Table 2 reports the results for the difference tests between the Kosdaq IPO firms and the KSE IPO firms for discretionary accruals (DA) and total accruals (TA) as well. The mean difference tests for TA support H1 that the Kosdaq IPO firms tend to manage earnings more aggressively than its KSE counterparts in the year preceding IPO and in the IPO year. However, the same test for DA by and large fails to support H1 except for the MJ model in the IPO year. The lack of statistical significance for DA may be related to the possible misspecification of the DA estimation. However, the lack of statistical significance may be more likely related to the fact that the Kosdaq non-venture firms are not significantly different from the KSE firms in terms of major firm characteristics, maybe with the exception of firm size. Since we get inconsistent evidence from the two different accruals, we are not warranted to conclude that the Kosdaq firms employ more aggressive earnings management strategies in the year preceding IPO.

Table 3 reports the results for the mean accrual difference test between the Kosdaq venture IPO firms and the Kosdaq non-venture IPO firms. Table 5 reveals that the Kosdaq venture firms increase reported earnings more aggressively than the Kosdaq non-venture firms, definitely in the IPO year and probably in the year preceding IPO. Statistically significant differences are evidenced by both DA and DA in the IPO year and by TA and DA from the MJ model in the preceding year. In conjunction with the results reported in Table 2, we can document that most of the accrual differences between the Kosdaq IPO firms and the KSE IPO firms come from the aggressive earnings management practices by the Kosdaq venture IPO firms. We can observe that the magnitudes for DA and TA for the Kosdaq venture IPO firms in the IPO year are many times of those of the Kosdaq non-venture IPO firms. These results clearly indicate that the Kosdaq venture firms tend to increase reported earnings very aggressively in the IPO year and also probably so in the year preceding IPO. The results, hence, support the H2 that the Kosdaq venture firms manage earnings more aggressively than their Kosdaq non-venture counterparts when they go public. The clear differences in earnings management practices of IPO firms within the market seem to be driven primarily by the differences in listing regulations.

Additional tests for the accrual differences in the three years surrounding the IPO year between the Kosdaq non-venture IPO firms and the KSE IPO firms (which is not shown in tabular form) revealed that the differences are not statistically significant. This indicates once again that the Kosdaq venture firms manage earnings much more aggressively than their counterparts of the Kosdaq non-venture IPO firms or the KSE IPO firms.

Graphic Approach

[FIGURE 1 OMITTED]

The mean difference tests focus on the differences in the central tendency for the treatment sample (the Kosdaq venture IPO firms) and the control samples (the Kosdaq non-venture IPO firms and the KSE IPO firms). The mean difference test is related to a point estimate. Therefore, any deviations from the mean may cancel each other out so that the test may fail to show rich distributional characteristics. However, with the graphic approach we are using, each sample for each year (the preceding year, the IPO year and the following year) is divided into five equal size portfolios that are formed based on CFO. A clear advantage for the graphic approach over the mean difference tests is that the approach presents richer information about the distribution for average accruals across the different CFO portfolios.

Figures 1 through 3 show the average DA for each CFO portfolio, one for each year. (8) The first thing we can notice from the comparison of graphs is that the DA estimation approaches, even though they are different in terms of design and results, does not affect the empirical results very differently. The YM model has lower range of DA between the minimum and the maximum values than the MJ model. However, the overall pictures of the two different approaches show very similar patterns.

Figure 1 and Figure 2 reveal that firms in general tend to incur accruals in the opposite directions of CFO. The relationships are generally monotonically negative. We observe that the average DA across the CFO portfolios decreases gradually. Therefore, accruals and CFOs are strongly negatively correlated. We don't expect that there should be negative relationships if firms do not manage earnings. For example, firms in the bottom quintile portfolio (Portfolio 1) have negative CFO and hence employ high accrual strategies. This is consistent with the findings of Yoon & Miller (2002b) and Kothari et al. (2005). While we fail to find any systematic differences in DA among the three IPO groups in the preceding years from the YM model, we find more pronounced differences in DA from the MJ model. (9)

Given the general tendency of negative relationships between accruals and CFOs, we observe from Figure 2 that the Kosdaq venture firms tend to employ more aggressive earnings management strategies in the IPO year. The average DA for each of the five CFO portfolios is consistently the highest for the Kosdaq venture firms among the three groups in the IPO year. The average DA is as high as 0.280 for the YM model and 0.495 for the MJ model for the bottom quintile CFO portfolio in the IPO year. In comparison, the average DA for the bottom quintile CFO portfolio in the IPO year is 0.137 for the YM model and 0.345 for the MJ model for the Kosdaq non-venture firms and 0.117 and 0.181 respectively for the KSE firms. We can infer that the degree of earnings management practices is non-trivial for all three groups when CFO is poor, and that earnings management practices appear to be the most significant for the Kosdaq venture firms.

Figure 3, however, indicates that the earnings management practices are not consistently related to CFO in the year following IPOs. The Kosdaq venture firms which implemented aggressive earnings management strategies in the IPO year don't appear to employ income-increasing strategies in the following year. For the Kosdaq venture firms, the average accruals are mostly negative. In contrast, the average accruals for the Kosdaq non-venture firms and the KSE firms do not show significant changes from the previous years.

[FIGURE 2 OMITTED]

[FIGURE 3 OMITTED]

One possible reason for the turn-around in accounting strategies is that the Kosdaq venture firms might have exhausted all possible means for managing earnings in the IPO year. Another reason may be that the venture firms no longer need to manage earnings to maintain the stock prices at a certain level. As discussed earlier, IPO firms and underwriters are normally expected to manage prices for newly issued shares for a sustained period of time up to three months after the IPO. Furthermore, shareholders are not allowed to sell their respective stocks until six months after the initial public offerings.

CONCLUSIONS

We expect that firms going public will have incentives to manage earnings to make sure that they pass the specific listing requirements. Also, firms hope to sell out their IPO shares at good prices. Two stock exchanges, the Kosdaq exchange and the Korea Stock Exchange (KSE), with differential listing regulations are available for would-be public firms to self-select their initial public offerings. If a firm is qualified as a venture firm, then the firm can choose the Kosdaq venture listing alternative, for which the listing regulations are even more lenient than the Kosdaq non-venture firms within the same Kosdaq exchange. The differential listing regulations will encourage firms to choose the stock exchange that is most advantageous for them. These less strict listing regulations for the venture firms may induce the firms to manage earnings aggressively. The study examined the earnings management practices for the Kosdaq venture IPO firms in comparison to the Kosdaq non-venture IPO firms and the KSE IPO firms during the period 1996 to 2000.

We selected a treatment sample of 75 Kosdaq venture IPO firms and two control samples of 133 Kosdaq non-venture IPO firms and 71 KSE IPO firms during the five year period. The study tested the accrual differences between the treatment sample and the control samples. The study also used a graphic analysis approach to investigate the differences in earnings management practices for the Kosdaq venture IPO firms and the two control samples.

The results for the tests consistently reveal that the Kosdaq venture IPO firms employ more aggressive earnings management practices than the two control samples in the IPO year and probably in the preceding year as well. However, the Kosdaq venture firms drastically reduce accruals and report relatively lower earnings in the year following the IPOs.

The graphic approaches for accruals also reveal that the Kosdaq venture firms employ more aggressive earnings management practices in the IPO year. In addition, the graphic approach shows that the degree of earnings management practices depend very strongly on the cash from operations in a monotonically negative manner.

The current study suffers from possible misspecification problems for the discretionary accrual estimation. Even though we used two alternative models in estimating discretionary accruals, the goodness of fit is still far from satisfaction. Future research needs to develop a more refined and better fitting model.

ACKNOWLEDGMENT

Dr. Kim is grateful for her post-doctoral research fellowship supported by the government of Canada.

ENDNOTES

(1.) The Korea Stock Exchange, the Kosdaq Market and the Korea Futures Exchange have been merged into the Korea Exchange (KRX) in January 2005. However, the KRX is still divided into three divisions and the operation system of individual divisions is still the same as before. On the other hand, the operation system of the KRX is virtually same as other stock exchanges like NYSE in terms of disclosure regulation, listing procedures and other policies. Furthermore, Korean accounting standards are very much similar to those of IASB and/or the FASB thanks to the increased efforts for the convergence of international standards. Since our study's main objective is to compare earnings management of IPO firms in differential listing requirements, we will not elaborate on the operational and institutional features in detail. Detailed information on the disclosures, listing procedure, and other operation system of the KRX can be obtained from www.krx.co.kr.

(2.) In order to be listed on the KSE, an IPO firm must satisfy the following requirements: It should get a public auditor designated by the Financial Supervisory Service before going public; It should be at least three year old; It should have an equity greater than 5 billion Won (about $5 million) or sales greater than 20 billion Won (about $20 million); It should have an unqualified audit opinion; It should have no material pending lawsuits; It should have a debt ratio less than 1.5 times the industry average; and it should have a return on equity higher than 10%.

(3.) Firms with total assets greater than 50 billion Won (about $50 million) or stockholders' equity greater than 10 billion Won (about $10 million) can be listed even if they experience net losses. Firms with paid-in-capital greater than 0.5 billion Won (about $0.5 million) can be listed as long as income from continuing operations is positive. Firms with qualified audit opinion can also be listed.

(4.) We may have a multicollinearity problem when we include explanatory variables that are correlated. We need to take into account the trade-off between the adverse effect of multicollinearity and the problems of omitted variables in this case. Changes in revenues and changes in expenses are highly correlated. However, the correlation between changes in non-discretionary revenues and changes in non-discretionary expenses decreased significantly. Furthermore, when we use both of the variables, we get significant and consistent signs for both of the variables from the regressions (Yoon & Miller, 2002a).

(5.) The MJ model uses the inverse of firm size (1/beginning total assets) to proxy for the intercept and does not allow a free intercept term. This will lower the goodness of fit of the model. Using our sample, we find out that the adjusted R2 on average drops from 0.082 to 0.042. The marked difference in adjusted R2 between the suppressed and unsuppressed constant terms alone indicates how serious the misspecification problem is for the MJ model.

(6.) For some aged capital intensive firms, this variable becomes greater than one when property plant and equipment is fully or nearly fully depreciated. In this unusual case, the variable distorts the expected negative relationship between total accruals and non-current accruals.

(7.) TA should be equal to NI less CFO by definition. However, the relationship is slightly distorted by the fact that some variables are winsorized as a way to adjust for the outliers.

(8.) For the graphic approach, we report only discretionary accruals. The graphic analysis of total accruals shows stronger support of the earnings management hypothesis of the Kosdaq venture firms than the discretionary accruals.

(9.) For both the YM model and the MJ model, IPO year differences in discretionary accruals are statistically significant for the portfolios 1 though 3 between the Kosdaq venture firms and the Kosdaq non-venture firms, and for the first two portfolios between the Kosdaq venture and the KSE firms. In the year preceding IPO, discretionary accruals from the modified Jones model are also statistically different for portfolios 2 and 5 between the Kosdaq venture firms and the Kosdaq non-venture firms as well as for portfolio 2 between the Kosdaq venture firms and the KSE firms. In the following year, discretionary accruals from the modified Jones model is statistically significantly lower for the Kosdaq firms than for the KSE firms for portfolio 2 only. In sum, the Kosdaq venture firms tend to increase reported earnings more aggressively than the Kosdaq non-venture firms and the KSE firms when operating cash flows are poor in the IPO year and the preceding year.

REFERENCES

Aharony, J., C. Lin & M. Loeb (Fall 1993). Initial public offerings, accounting choices, and earnings management, Contemporary Accounting Research, 61-81.

Burgstahler, D. & I. Dichev (1997). Earnings management to avoid earnings decreases and losses, Journal of Accounting and Economics 24, 99-126

Choi, K. & M. Kim (1997). An empirical study on the earnings management of IPO firms, Korean Accounting Review, 1-28.

Collins, D. & P. Hribar (2002). Errors in estimating accruals: Implications for empirical research, Journal of Accounting Research 40, 105-135

Dechow, P. & I. Dichev (2002). The quality of accruals and earnings: The role of accrual estimation errors," The Accounting Review Vol.77 (Supplement), 35-59

Dechow, P., R. Sloan & A. Sweeny (1999). Detecting earnings management, The Accounting Review 193-225.

Friedlan, J. (1994). Accounting choices of issuers of initial public offerings, Contemporary Accounting Research (Summer), 1-31.

Kothari, S., A. Leone & C. Wasley (2005). Performance matched discretionary accrual measures, Journal of Accounting and Economics 39, 163-197 Loughran, T. & J. Ritter (1995). The new issues puzzle, The Journal of Finance (March), 23-51.

Loughran, T. & J. Ritter (1997). The operating performance of firms conducting seasoned equity offerings, The Journal of Finance (December), 1823-1850

McNichols, M. (2000). Research design issues in earnings management studies, Journal of Accounting and Public Policy, 313-345

McNichols, M. & P. Wilson (1998). Evidence of earnings management from provision for bad debts, Journal of Accounting Research (Supplement), 1-31

Rangan, S.(1998). Earnings management and the performance of seasoned equity offerings, Journal of Financial Economics, 101-122.

Shivakumar, L. (2000). Do firms mislead investors by overstating earnings before seasoned equity offerings? Journal of Accounting and Economics 29, 339-371.

Teoh, S., I. Welch & T. Wong (1998). Earnings management and the underperformance of seasoned equity offerings, Journal of Financial Economics, 63-99.

Yoon, S. (2005). A comparison of earnings management between KSE firms and Kosdaq firms, Journal of Business Finance & Accounting 32, 1347-1372.

Yoon, S. & G. Miller (2002a). Earnings management of seasoned equity offering firms in Korea, International Journal of Accounting 37, 57-78.

Yoon, S. & G. Miller (2002b). Cash from operations and earnings management of Korean firms, International Journal of Accounting 37, 395-412.

Soon Suk Yoon, Chonnam National University (Korea)

Hyo Jin Kim, York University (Canada)

Table 1: Descriptive Statistics (7)

Panel A: One year before IPO

 Stock Exchange Variables Mean Median S.D.

Kosdaq Venture DA-YM 0.0178 0.0033 0.1429
([n.sub.1]=59) DA-MJ 0.0637 0.0373 0.1929
 TA/Sales 0.0590 0.0362 0.1285
 TA/BTA 0.1109 0.0675 0.2443
 NI/Sales 0.1074 0.0908 0.0822
 NI/BTA 0.2104 0.1506 0.1855
 CFO/Sales 0.0472 0.0397 0.1267
 CFO/BTA 0.0882 0.0803 0.2560
Kosdaq Non-venture DA-YM -0.0032 -0.0076 0.1415
([n.sub.2]=124) DA-MJ -0.0134 -0.0230 0.1953
 TA/Sales 0.0016 -0.0075 0.1205
 TA/BTA 0.0111 -0.0165 0.2285
 NI/Sales 0.0629 0.0460 0.0715
 NI/BTA 0.1302 0.0788 0.1919
 CFO/Sales 0.0620 0.0612 0.1317
 CFO/BTA 0.1088 0.1040 0.2541
KSE DA-YM 0.0195 0.0121 0.0966
([n.sub.3]=69) DA-MJ 0.0212 0.0014 0.1178
 TA/Sales -0.0175 -0.0187 0.1015
 TA/BTA -0.0098 0.0014 0.1301
 NI/Sales 0.0708 0.0541 0.0606
 NI/BTA 0.1045 0.0720 0.0991
 CFO/Sales 0.0860 0.0881 0.1106
 CFO/BTA 0.1120 0.1154 0.1589

 Stock Exchange Variables Min. Max.

Kosdaq Venture DA-YM -0.3182 0.3969
([n.sub.1]=59) DA-MJ -0.3498 0.6972
 TA/Sales -0.2356 0.5651
 TA/BTA -0.3330 0.8763
 NI/Sales -0.1004 0.3329
 NI/BTA -0.0789 0.7110
 CFO/Sales -0.4027 0.3094
 CFO/BTA -0.7752 0.7775
Kosdaq Non-venture DA-YM -0.3256 0.6000
([n.sub.2]=124) DA-MJ -0.6968 0.5576
 TA/Sales -0.2351 0.5681
 TA/BTA -0.0688 0.8761
 NI/Sales -0.2161 0.3399
 NI/BTA -0.8014 0.7166
 CFO/Sales -0.6000 0.4603
 CFO/BTA -0.7753 0.7778
KSE DA-YM -0.2684 0.2988
([n.sub.3]=69) DA-MJ -0.2501 0.3890
 TA/Sales -0.4024 0.2543
 TA/BTA -0.3328 0.3735
 NI/Sales 0.0060 0.2772
 NI/BTA 0.0056 0.5443
 CFO/Sales -0.2353 0.3785
 CFO/BTA -0.3093 0.6850

Panel B: IPO year

 Stock Exchange Variables Mean Median S.D.

Kosdaq Venture DA-YM 0.1006 0.0844 0.1402
([n.sub.1]=75) DA-MJ 0.1792 0.0903 0.2639
 TA/Sales 0.1114 0.0839 0.1528
 TA/BTA 0.1928 0.0883 0.2659
 NI/Sales 0.0917 0.0803 0.1058
 NI/BTA 0.1509 0.1276 0.1880
 CFO/Sales -0.0194 -0.0154 0.1818
 CFO/BTA -0.0416 -0.0157 0.2787
Kosdaq Non-venture DA-YM 0.0224 0.0191 0.1320
([n.sub.2]=133) DA-MJ 0.0539 0.0033 0.1281
 TA/Sales 0.0200 0.0022 0.1300
 TA/BTA 0.0237 0.0071 0.1393
 NI/Sales 0.0576 0.0518 0.0917
 NI/BTA 0.0777 0.0531
 CFO/Sales 0.0353 0.0519 0.1422
 CFO/BTA 0.0539 0.0604 0.1417
KSE DA-YM 0.0419 0.0412 0.0817
([n.sub.3]=71) DA-MJ 0.0639 0.0400 0.2187
 TA/Sales 0.0059 0.0084 0.0979
 TA/BTA 0.0521 0.0203 0.2337
 NI/Sales 0.0585 0.0460 0.0677
 NI/BTA 0.1061 0.0933 0.1378
 CFO/Sales 0.0525 0.0584 0.1011
 CFO/BTA 0.0493 0.0201 0.2322

 Stock Exchange Variables Min. Max.

Kosdaq Venture DA-YM -0.2163 0.4903
([n.sub.1]=75) DA-MJ -0.4362 0.9255
 TA/Sales -0.1923 0.5631
 TA/BTA -0.3514 0.8744
 NI/Sales -0.5179 0.3360
 NI/BTA -0.8016 0.7095
 CFO/Sales -0.5939 0.3592
 CFO/BTA -0.7747 0.7767
Kosdaq Non-venture DA-YM -0.2577 0.2926
([n.sub.2]=133) DA-MJ -0.1850 0.7404
 TA/Sales -0.4314 0.5671
 TA/BTA -0.3596 0.7137
 NI/Sales -0.6000 0.3430
 NI/BTA 0.0935 -0.2740
 CFO/Sales -0.6000 0.4614
 CFO/BTA -0.6541 0.4194
KSE DA-YM -0.2007 0.2673
([n.sub.3]=71) DA-MJ -0.7984 0.8584
 TA/Sales -0.3315 0.3264
 TA/BTA -0.7669 0.8735
 NI/Sales -0.2577 0.2926
 NI/BTA -0.6063 0.7105
 CFO/Sales -0.2992 0.3514
 CFO/BTA -0.7740 0.7777

Panel C: One year after IPO

 Stock Exchange Variables Mean Median S.D.

Kosdaq Venture DA-YM -0.0162 0.0145 0.2402
([n.sub.1]=59) DA-MJ -0.0427 -0.0310 0.2118
 TA/Sales -0.0644 -0.0345 0.2634
 TA/BTA -0.0462 -0.0298 0.2074
 NI/Sales -0.0418 0.0403 0.2430
 NI/BTA -0.0130 0.0406 0.1959
 CFO/Sales 0.0070 0.0516 0.2227
 CFO/BTA 0.0324 0.0536 0.1585
Kosdaq Non-venture DA-YM 0.0254 0.0251 0.1587
([n.sub.2]=124) DA-MJ 0.0085 -0.0058 0.1687
 TA/Sales -0.0269 -0.0297 0.1737
 TA/BTA -0.0086 -0.0214 0.1694
 NI/Sales -0.0055 0.0264 0.1556
 NI/BTA 0.0196 0.0389 0.1507
 CFO/Sales 0.0121 0.0476 0.1681
 CFO/BTA 0.0281 0.0431 0.1511
KSE DA-YM 0.0501 0.0475 0.1273
([n.sub.3]=69) DA-MJ 0.0134 0.0200 0.1158
 TA/Sales -0.0146 -0.0101 0.1147
 TA/BTA -0.0148 -0.0090 0.1211
 NI/Sales 0.0236 0.0414 0.1223
 NI/BTA 0.0390 0.0373 0.0870
 CFO/Sales 0.0392 0.0530 0.1497
 CFO/BTA 0.0553 0.0428 0.1344

 Stock Exchange Variables Min. Max.

Kosdaq Venture DA-YM -0.6000 0.6000
([n.sub.1]=59) DA-MJ -0.7828 0.6846
 TA/Sales -0.6000 0.5691
 TA/BTA -0.7674 0.6104
 NI/Sales -0.6000 0.2989
 NI/BTA -0.8015 0.1987
 CFO/Sales -0.6000 0.4534
 CFO/BTA -0.4981 0.3990
Kosdaq Non-venture DA-YM -0.6000 0.5402
([n.sub.2]=124) DA-MJ -0.6954 0.7172
 TA/Sales -0.6000 0.5711
 TA/BTA -0.7670 0.6801
 NI/Sales -0.6000 0.3340
 NI/BTA -0.7934 0.5002
 CFO/Sales -0.6000 0.4219
 CFO/BTA -0.4886 0.3877
KSE DA-YM -0.2372 0.6000
([n.sub.3]=69) DA-MJ -0.4252 0.2560
 TA/Sales -0.2705 0.3175
 TA/BTA -0.4398 0.2283
 NI/Sales -0.6000 0.2985
 NI/BTA -0.2244 0.3278
 CFO/Sales -0.5420 0.2946
 CFO/BTA -0.2366 0.5553

Note (1) All variables are deflated by the same period net sales or
beginning total assets (BTA).

Note (2) TA= Total accruals, DA = Discretionary accruals, NI = Net
income, CFO = Cash flows from operations

Note (3) DA-YM represents DA from the YM model and DA-MJ
from the MJ model.

Table 2: Results for Mean Difference Tests between Kosdaq
IPO Firms and KSE IPO Firms

 DA Model/
Variables Deflator Year Kosdaq KSE

DA YM Model -1 0.0040 0.0195
 0 0.0506 0.0420
 +1 0.0037 0.0501
 MJ Model -1 0.0114 0.0212
 0 0.1055 0.0539
 +1 -0.0095 0.0134
TA Sales -1 0.0201 -0.0175
 0 0.0529 0.0060
 +1 -0.0648 -0.0146
 BTA -1 0.0433 -0.0098
 0 0.1028 0.0237
 +1 -0.0218 -0.0148

 DA Model/
Variables Deflator t-ratios p-levels

DA YM Model -0.987 0.325
 0.629 0.530
 -2.146 0.033
 MJ Model -0.481 0.631
 2.279 0.024
 -1.204 0.230
TA Sales 2.140 0.034
 3.050 0.003
 -2.013 0.045
 BTA 2.255 0.025
 3.274 0.001
 -0.363 0.717

Note) DA = Discretionary accruals, TA= Total accruals,
YM model = Yoon and Miller model; MJ model = Modified
Jones Model,

BTA = Beginning total assets

Table 3: Results for Mean Difference Tests between the
Kosdaq Venture IPO Firms and the Kosdaq Non-venture IPO Firms

 DA Model/
Variables Deflator Year Venture Non-venture

DA YM Model -1 0.0178 -0.0026
 0 0.1006 0.0224
 +1 -0.0324 0.0232
 MJ Model -1 0.0637 -0.0191
 0 0.1792 0.0639
 +1 -0.0427 0.0085
TA Sales -1 0.0590 0.0016
 0 0.1114 0.0200
 +1 -0.1098 -0.0404
 BTA -1 0.1109 0.0054
 0 0.1928 0.0521
 +1 -0.0462 -0.0086

 DA Model/
Variables Deflator t-ratios p-levels

DA YM Model 0.901 0.368
 3.946 0.000
 -1.497 0.136
 MJ Model 2.616 0.010
 3.214 0.002
 -1.768 0.080
TA Sales 2.880 0.004
 4.363 0.000
 -1.389 0.166
 BTA 2.751 0.007
 3.825 0.000
 -1.320 0.189

Note) DA = Discretionary accruals, TA= Total accruals,
YM model = Yoon and Miller model; MJ model = Modified
Jones Model,

BTA = Beginning total assets