Investigation of different aspects of the abnormal accruals phenomena in emerging countries

Размещено на http://www.allbest.ru

Размещено на http://www.allbest.ru

Introduction

In the IPO process there are generally three stages: preparation, execution and post-deal performance. During the preparation period companies are trying to weight all the pros and cons of the IPO and decide on its timing. The IPO is most beneficial for the firms with strong financials, growth potential and investors' interest. The last factor is usually determined by the previous two characteristics what makes them to be of the highest attention for the managers. The easiest way to create a “shining” image for the company is to improve its earnings performance.

Managers of IPO firms are actively using accruals as a tool of earnings management (also termed as window dressing) to improve their reported earnings as it helps to offer higher prices during stock placements. In general accrual accounting involves making adjustments in current statements to reflect earnings or expenses that are relevant for current period but will be correspondingly received or incurred in the next period. This helps to increase current profits despite no cash inflows but will impact firm's performance in the next periods.

“Buffet recommends that individual investors use cash flow from operations as a check on the quality of a company's earnings. In situations where there is a large discrepancy between a firm's reported earnings and cash flow from operations, the firm is likely engaging in a certain degree of earnings management”. The difference between reported earnings and cash flow from operations is nothing than accruals.

Earnings management becomes attractive for several factors (as discussed in previous research):

1) Information asymmetry that misleads investors resulting in overpriced stock issues;

2) Possibility to apply higher industry multiples with better financials;

3) International accounting regulation that allows managing of accounting numbers before stock issue (retroactive restatement of the financial statements).

On the other hand, companies are limited in earnings management by the obligatory audit checks and possibility to be sued in the court for mispricing caused by the manipulated financial information. Still, according to Teoh et al. (1998) “based on discussions with investment bankers it appears that the underwriters' due diligence process generally does not include auditing the firm's financial statements”.

During the execution stage stock price is determined by the market and afterwards several factors influence its further pricing. Among them - general macroeconomic situation, market-timing effect and window dressing (if they exist). Artificially improved performance cannot be sustained for a long time and as practice shows earnings after IPO drop. In this way IPO firm's last stage of the post-deal performance is formed.

The problem of post-issue underperformance was broadly highlighted in many articles. One of the most well known in this field of research authors Ritter (1991) stated that “investors are periodically overoptimistic about the earnings potential of young growth companies”. We are trying to extract window dressing from the range of factors that lead to such an overoptimism.

The problem of earnings management caused a debate in the literature on how to measure this phenomenon and if it is really significant. Several authors performed the research using abnormal accruals as a proxy. The opponents argued that the use of such a proxy causes biased results. We discuss these views in the next section (Literature Review).

While there is literature on earnings management in developed countries, little has been done for emerging markets. At the same time the problem can actually be more relevant for such countries because institutions (constraining misbehavior) are weak and investors may be less experienced, so earnings management may be both easier and more tempting.

In this paper we would like to adress three major questions:

1) If the phenomenon of earnings management exists in the emerging countries and has any impact on post-IPO performance?

2) What are specific factors that may determine the use of abnormal accruals in emerging countries?

3) Does earnings management have any statistically significant relation to the post-IPO long-term stock underperformance?

We will use abnormal accruals as a proxy of earnings management. Thus, our research will be based on a four-stage procedure.

First, we are going to extract discretionary component of accounting accruals with the help of regression model. We apply the widely used Jones model (and its modification).

Second, we will analyze whether the following factors have an effect on abnormal accruals:

o state ownership (as a proxy - government share in a company);

o choice of stock exchange (local versus international, measures investors sophistication and disclosure requirements);

o choice of auditor (lower versus higher quality),

controlling for:

o size and age of a company (as a proxy on information asymmetry);

o market timing (average market-to-book ratio as a proxy of investors valuation of a company);

o country, timing and industry effects (corresponding dummy variables).

At the third stage we are going to answer the question: does manipulation with accruals influence post-IPO earnings performance? In this scope we will use return on sales as a proxy for earnings performance and run the regression on abnormal and expected accruals.

Finally we would like to address the issue if the manipulation with accruals influences post-IPO stock performance. For this purpose we will be using post-IPO 3-years return on stock calculated as buy and hold return and as cumulative abnormal return aggregated in time.

Our initial sample consists of 2471 companies listed on the national Stock Exchanges (or on several at one time) of 19 countries from the MSCI emerging markets index. There was clustering in geography, industry and time. The largest number of deals was executed in Asia (China, Korea and India). Most active years were 2007 and 2010 due to the cyclicality of the economies. Industrial sector firms went public more often than others, especially in machinery&equipment sector or chemical industry. We account for these facts when creating the regression models.

We find that the phenomena of earnings management exists in the emerging countries and very high IPO abnormal accruals continuously drop in long-term together with the returns on sales and on the assets.

We document statistically that earnings management has significant negative impact on the earnings performance. Thus, our hypothesis that window dressing of financials from which companies benefit at the IPO stage leads to the poor long-term performance.

Moreover, we show that accruals management that misleads investors before the stock issue is further priced by the market so that return on stock for the firms aggressively managing their financials underperforms for the next at least 2,5 years.

Finally, we prove that the choice of stock exchange and of the auditor may signal about the higher possibility that the IPO-company manages its earnings. It also supports the idea that in emerging markets with the underdeveloped local markets there are more incentives for the firms in window dressing their performance and misleading unsophisticated investors.

1. Literature review

Jones (1991) published one of the fundamental articles concerning the problem of accruals. She tested whether firms that benefit from the import relief (from tariff increases or quota reductions) are interested in decreasing their earnings with the help of earnings management. The author used abnormal (discretionary) total accruals as a measure of earnings management. She introduced a two-step procedure to estimate abnormal accruals. First, she found normal total accruals with regression analyses and then extracted residuals serving as a proxy for discretionary component of accruals. With this models Jones showed that firms do use abnormal accruals in the export-import operations.

The model created by Jones was further developed by Dechow et al. (1995) and received a name of modified Jones model. The authors argued that Jones model does not capture manipulations with the sales itself, so they added an adjustment to the classic Jones model by subtracting trade receivables at the second stage. This was supposed to improve the precision of the proxy for earnings management. However, both approaches received some criticism and were reviewed in later articles (Teoh et al. (1998 a, b) and Peasnell et al. (1999)).

Sloan (1996) was among researchers who investigated the influence of accrual and cash accounting on the stock performance. In fact some practitioners (mostly research analysts) already in late 1980s paid attention to the ways of earnings quality detections when forecasting stock performance. Sloan described a model for stock prices prediction based on the peculiarities of the accounting processes. He introduced accruals in the model assuming that they influence stock performance and showed that stock prices act as if investors fail to identify correctly the different properties of accrual and cash components of earnings.

Subramanyam (1996) made analyses from another point of view. He looked at how discretionary accruals are reflected in the stock prices and found that market incorporated abnormal accruals in the stock value that is a sign of the market inefficiency. The author used simple and modified Jones models to calculate abnormal accruals following the two-step procedure. However, instead of time-series analyses he executed cross-sectional analyses, explaining this choice by the significantly larger sample what should increase the precision of the estimates. In the paper it was shown that cross-sectional regression in general gives a better specification.

Teoh et al. (1998) examined relationship between IPO-year, long-run post-issue accruals and earnings performance. They mention that total accruals may contain some objectively or economically determined component and random part called abnormal accruals. Teoh et al. used a correction to the Jones model to calculate discretionary accruals by focusing on working capital component of the total accruals. The authors discovered that return on sales declines significantly after IPO and is influenced by the earnings management. It may be driven by the increase in sales due to new investments. However it was shown that the same tendency exists when using return on assets as a performance measure (assets incorporate all new investments right after IPO). Decrease in cash flows (non-accrual indicator) may also influence the underperformance, but in fact they increase when return on sales drops. This means that post-IPO earnings underperformance is due to the drop in abnormal accruals that reverse after being managed before the stock issue.

The behavior of expected and abnormal accruals after IPO was also investigated. For the US companies abnormal accruals decrease continuously and more significantly than normal accruals. If firms are ranked by the size of their abnormal accruals (conservative versus aggressive) then only “aggressive” firms underperform.

Additionally Teoh et al. (1998b) examine relation of accruals to long-term stock return underperformance. They used data from the first public financial statement issued after IPO that according to the authors' opinion should include information on both pre- and post-IPO months. The reason is that pre-IPO accounting information is not always available, while in the first statement there cannot be big drops in the accruals due to litigation reasons.

What the authors found is that “depending on benchmark specification, IPO firms that are ranked in the highest quartile based on IPO-year discretionary current accruals (“aggressive” IPOs) earn a cumulative abnormal return of approximately 20 to 30 percent less than the cumulative abnormal return of IPO firms ranked in the lowest quartile (“conservative” IPOs)” within 3 years period.

Peasnell et al. (1999) were among the opponents of the classic Jones model having found some problems with its relevance. They proposed another type of modified Jones model called marginal model. The authors excluded depreciation from the accruals measurement as considered it to be inappropriate for the discretionary earnings management. As Teoh et al. (1998) Peasnell et al. used working capital accruals instead of total accruals at the first stage. Second step was to calculate normal accruals with the help of regression model (similar to Jones model), but using other variables (components of working capital accruals): change in stocks, in receivables net of bad debt allowances and change in payables.

Teoh et al. (2002) continued investigation of the problems caused by usage of earnings management before the IPO by expanding it to the role of analysts' credulity. They argue that financial analysts interpret accounting information on behalf of investors and present forecasts of IPO firms' earnings what influences IPO stock price. If analysts regularly misinterpret companies' financials and believe that their high earnings represent good news while a portion of this great performance is due to window dressing, this may result in the overvalued IPO firms and further stock underperformance.

The authors also find that misinterpretation of earnings occurs both in case of equity-issuing and non-issuing firms what helps to explain the general investors' misevaluation of companies and further downward revaluation of stock.

Xie (2001) investigated how the market prices abnormal accruals in more general setting than just in case of IPO and found that market not only priced but also overpriced abnormal accruals. He used hedge-portfolio test, first performed by Sloan (1996) and Mishkin test (1983). Xie points out that Jones model measuring “discretionary accruals” suffers from the measurement error that reflects unusual business circumstances besides the managerial discretion that we are looking for. He proposes to make an adjustment for this by controlling for major unusual accruals and non-articulation events such as mergers or divestitures.

Another study performed by Li et al. (2005) was aimed at analyzing the reasons for delisting of firms. They showed that one of the major determinants is the earnings management in IPO year and thus try to find some equilibrium level between earnings management benefits and disadvantages. Li et al. extend the list of negative effects or `costs of earnings management', including poor timing of sales, disruption of suppliers' and customers' delivery schedules and memory about low quality IPOs.

An alternative view on the earnings management was presented by Ball et al. (2008). They claim that IPO report more conservatively that maybe explained by the higher quality reporting standards for public companies. Authors question Teoh et al. (1998) conclusion on the existence of the window dressing reflected in abnormal accruals. They state the following arguments:

· regulatory institutions and other type of controllers may discover illegal actions followed by litigation;

· earnings inflation can result in the further earnings deflation;

· low reporting quality may lead to increase in the cost of capital unfavorable for the management.

Moreover, Ball et al. (2008) list several reasons why Teoh et al. (1998) definition of abnormal accruals is unreliable:

· estimates of abnormal returns are too high that cannot be allowed by GAAP except for “faking sales and uncollected receivables” that are to be discovered;

· changes in working capital should be taken directly from the cash flow statement rather than from balance sheets;

· “any inflation of pre-IPO earnings via current accruals would inflate pre-IPO working capital, not post” and thus wouldn't reflect earnings management increase attempts.

· “DCA estimates are biased by unusually high IPO proceeds”;

· Jones model according to the authors opinion is mis-specified;

· low value of deflator as a source of extreme values in accruals.

In their research authors used a sample of UK firms and showed that prior to the IPO they report more conservatively.

We would like to perform our study on the sample of emerging countries. Among our assumptions on the presence of earnings management there is the underdevelopment of local financial markets and some specific characteristics of doing business (for example high rate of corruption that may help to avoid the punishment). One of such characteristics is the quality of accounting procedures.

In the research by Ball et al. (2003) it was shown that the East Asian countries (Hong Kong, Malaysia, Singapore and Thailand) have managerial incentives to perform lower quality accounting that is reflected in the timing of the economic income or recognition of losses. One of the arguments is that accounting quality depends on the “interplay between market and political forces in the reporting jurisdiction”. Economic income is understood in this research as “change in the market value of equity, adjusted for dividends and capital transactions with shareholders” and “timely recognition is the timely revision of the book values of equity and of asset and liability reported”.

It is also stated that information asymmetry is usually overcome by “insider” communication that is why there is lower demand for high quality reporting. Asian economies are known for their dominance of family ownership. A similar phenomenon may be observed in other countries of the emerging world where there is a strong government influence and importance of the major shareholder decisions.

In addition, verification of the number of litigation cases in Asia post low quality reporting in comparison with the developed markets also reflects existence of the incentives to manipulate financials. “Saudagaran and Diga (2000) reports that there have been no cases of judicial actions against auditors in Singapore and Hong Kong”.

Xiong et al. (2014) proposed a model to examine the association among auditor litigation, earnings management and IPO underpricing. Authors mentioned Stice (1991) finding on that “higher levels of accounts receivable and inventory account balance are associated with increased litigation risk…”. This means that generally in case of earnings management there should be auditor litigation. However, for emerging countries it is more difficult to observe this situation because the low number of litigation cases reflects low quality of reporting standards in these countries rather than the efficiency of the monitoring of financials manipulation followed by auditor litigation.

Therefore, based on the literature review and evidence obtained we will focus on measuring the impact of earnings management on the post-IPO earnings and stock performance in emerging countries taking into account specific features of these countries' economies.

2. Research design

2.1 Sample selection and data

Our sample is comprised of data on the companies from the countries in emerging markets group according to the MSCI Index country membership (see Table 1). We need to collect information on the initial public offerings for the period since 1997 and up to now. The start year is determined by the time when the majority of countries were included in the MCSI Index that indicates both that they entered emerging countries group and that they established their financial markets. Thus we would exclude from our sample Egypt, Qatar and United Arab Emirates to obtain a more homogeneous data.

Table 1. List of emerging countries according to the MSCI Index country membership

Number	Country	Date of inclusion into the MSCI Index	Included in sample
1	BRAZIL	1988	+
2	CHILE	1988	+
3	CHINA	1996	+
4	COLOMBIA	1994	+
5	CZECH REPUBLIC	1996	+
6	EGYPT	2001	-
7	GREECE	1988	+
8	HUNGARY	1996	+
9	INDIA	1994	+
10	INDONESIA	1989	+
11	KOREA	1992-1998	+
12	MALAYSIA	1988	+
13	MEXICO	1988	+
14	PERU	1994	+
15	PHILIPPINES	1988	+
16	POLAND	1995	+
17	QATAR	2006	-
18	RUSSIA	1997	+
19	SOUTH AFRICA	1995	+
20	TAIWAN	1996-2005	+
21	THAILAND	1988	+
22	TURKEY	1989	+
23	UNITED ARAB EMIRATES	2006	-

From the preliminary data analyses we can conclude that in the emerging markets companies started to use more or less actively equity financing since 2000s. Thus for our further analyses we will use a tighter time frame than previously expected: since 2000 up to 2012 (as the last three years are needed to observe post-issue earnings and returns).

Another important issue that we take into account for choosing the most appropriate sample is the type of financing. To address our research questions we should use only new listings of the ordinary stock as they may signal more strongly about earnings management (investors are less informed in the case of the first placement, especially when long history of the accountancy information is not publicly available and they have to rely primarily on the prospectus and analysts estimates).

Sector wise we exclude financial institutions and utility companies (electric, gas and water firms, integrated providers) as they have specific requirements and accounting rules. For example, banks must publish their financials independently of their stock market status that prevents them more from earnings management. Further, we will have to exclude firms if data is not available in the year prior to the IPO as well as if any other information necessary for computing models is unavailable.

Data on IPO deals, including company name, sector, dates, advisor and some other deal information is taken from the Zephyr Database (Bureau Van Dijk). Other financial data for the companies is received from the DataStream (Thomson Reuters).

The initial sample consists of 2471 companies listed on the national Stock Exchanges (or on several at one time) of 19 countries (we excluded Check Republic firms due to the lack of he necessary information on the IPO companies).

2.2 Methodology

In accounting standards (GAAP and ICFR) one can find a set of general principles such as materiality, verifiability, reliability and relevancy. Two last principles often contradict one another. For example, when reflecting value of a building on the Balance Sheet at historical price reliability principle works (as this is a confirmed price), but relevancy does not (because market price may be more relevant though not precisely known).

Thus there are two general methods to reflect company's performance: cash accounting (corresponds to reliability) and accrual accounting (corresponds to relevancy).

Cash accounting is the methodology of recording transactions only if the real cash flows in or out of the firm. The main disadvantage of this approach is that such a measurement is not stable and gives low relevance for the investors that want to value the firm's stock. For example, in case of a large capital expenditure real outflow is reflected only in the first year when it occurs while the equipment where this capital was spent on continues to produce cash in the next periods. It may create a very unstable picture of a firm in different points of time.

That is why another approach is usually used by the firms - accrual accounting (cash method is mostly applicable for very small firms that is not the case for IPO). It helps to create a more realistic description of the company's financial standing. Revenues are reflected in the period in which they were actually earned and only corresponding to these revenues costs are recorded. However, this method has its own drawback: it is not so reliable as cash accounting, because decisions to reflect that or this number are based on the subjective management expectations and judgments. This causes information asymmetry issue for the investors.

Managers have a unique knowledge of the business structure and real plans of a company, so they can work out accounting information based on the accrual method according to their own preferences. For sure, they have constraints such as audit checks and possibility of litigation, but there is a number of ways how earnings may be just window dressed in short-term not causing any questions from the regulators. The problem is that any manipulations should reflect negatively on the future company's performance.

In fact, “clever” investors may even gain from their ability to discover firms that use earnings management. This is one of the strategies used by Warren Buffet: short stocks of the companies with bad earnings quality and buy shares of firms with high quality accounting. High quality accounting means that financial statements presented to the public show the realistic firm's position in the market and allow accurate forecasting of future development of a firm. As we could see neither cash accounting nor accrual accounting fully conform high quality requirements, but what investors can do is that they could compare both methods to discover if there is any significant differences between them which may signal about presence or absence of earnings management.

Accruals in accounting are made at the end of accounting period and before the preparation of the financial statements. Accounting rules suggest that income and expenses must be reflected in the period of time when they were incurred. In case some revenue was earned but is expected to be received in the next period accruals may help to create a more realistic picture of current firm's performance by taking into account this part of revenue.

The most used methods of accruals management are:

· Revenue and expenses recording before invoice is received. This could be made with the help of accounts payable item. Company can write the checks late or use overdrafts with the banks that could refuse to show the auditors their clients' accounts. Then a firm won't deduct these amounts in payables what increases its income. Also company could recognize revenues too early.

· Depreciation. Start recording costs later arguing this by production timing matters.

· Capital expenditures. Capitalizing some expenditures that could be in fact expensed in the current period.

· “Big bath”. This is less relevant for pre-IPO tool, but still interesting to know. Company can have not a very good performance today (or even really bad). Say it is 3 years before the IPO. It could make large write-offs that are relevant for future periods. This leads to even much worse results but suddenly a firm rebounds in the next three years right before the IPO. It is not obliged to show its reports before those three prosperous years but can get worse post IPO.

· Selling accounts receivable. This could be made in the following way: a firm sells its receivables to an agency but does not reflect the cash earned (because it is lower then the actual amount), but reflects receivables in the income side.

· Accounting for deferred revenue and expense. For example, we take a firm that provides a bonus program to the clients and allows them to buy products in the next period with these bonuses. The company should reflect on its Income Statement (discounts to the clients) its expectation about the loss from this “free selling”. To make the short-term situation better it can lower its expectation about the loss, while in the next period this loss will realize making a significant difference between two periods.

Deferred revenue artificially increased may be just reflected in the other income with the footnotes at which many investors do not look. And this footnotes will tell that the additional revenues are not from the core business and do not reflect the actual financial health of the company. On the opposite side, it could not account for the expenses that are not connected to the main operations.

· Deferred tax. This is the difference between tax accounting and business accounting. Manager can reduce some costs that are not tax deductible that will improve income.

· Other income. These could be case, for example, of the lawsuits. Company should reflect the costs or benefits that it can incur after the decision is made. It could expect to “loose the game” but reflect the opposite situation in its figures. The same discrepancies could appear during the sell of property or in the exchange differences recording (in fact, accountants use this item very often to cover the gaps).

· Inventory reserves. Company records an estimated charge for inventory that is not identified yet, but is expected to be present and which amount should be changed due to write-off. But the firm may fail to write down impaired assets that will be done post IPO.

· Holding structures. This is a much more complicated way of manipulation as it touches upon some criminal content. It concerns different way of costs shading, including taxation matters via offshore structures that could impact income statement as it is consolidated.

Also some companies create sister companies to move significant amounts of debt. For example, a company with huge R&D costs takes an order to perform research from another firm which is in fact its sister company. R&D is performed but the expenses paid by loans are left on the sister's balance.

· Allowance for bad debt. This is adjustment of the value of the accounts receivable and possibly is one of the most widely used methods of manipulation. If the company expects a large loss from the

· Pension plans. Company lowers its pension obligations if it makes a “judgment” about a bigger assumed return on pension assets. Also it could not account for the stock option that reduces the costs.

Thus, there are two major areas of manipulation: with the timing of records and with the “dishonest” reflection on judgments.

A general approach may be used (and tested by several researches) that measures total net operating accruals as just difference between net income (approximation of accrual earnings) and cash flow from operations (approximation of cash earnings).

However, it should be taken into the account that managers not always want to mislead investors but can have mistakes in their expectations or reflect realistic though unusually large amounts of accruals. A good way to discover the management's intentions to be dishonest is to compare a firm with industry peers, but in the scope of our research it is difficult to perform because information about sample firm's peers may be unavailable due to their private status. Therefore, we will be using another approach that is based on the Jones model aimed at estimating abnormal accruals of the individual firms as a proxy for earnings management. The idea is that the higher the total accruals are the greater is the possibility that the quality of earnings is low.

We have found in the literature that several proxies to measure abnormal accruals exist (for example, described by Teoh et al., 1998):

1) Cross-sectional Term-adjusted Jones Accruals: total accruals are regressed on PPE and change in sales to estimate expected and abnormal accruals. The idea is that PPE and change in sales should reflect economically determined accruals (not discretionary), i.e. before manipulations. Modified Jones model (by Dechow et al. (1995)) incorporates an adjustment by subtracting increases in trade receivables from change in sales (because credit sales may be also manipulated). The residuals are considered to be abnormal accruals.

It should be additionally commented why Jones (1991) included change in sales and stationary value of PPE. Change in sales reflects the change in economic activity and is an objective measure of the firm's operations before managers' manipulations. However, depreciation is also one of the items to manipulate, and as it is accounted in the cash flow statement not as a change but as an end period number (unlike receivables, payables and inventories), it should be incorporated in the PPE in the same way (not change in this account).

2) Matched-Pair Proxy: this model is used to overcome a drawback of the Jones approach that overrejects for extreme performing firms. Here every company going public is matched to a comparable non-issuing firm (main criteria is that return on sales cannot be lower than 80% in matched firm). Then abnormal accruals for both types of firms are calculated according to the Jones model and non-IPO company's abnormal accruals are subtracted from IPO firm's abnormal accruals.

3) Beneish M-score: calculates a score for the likelihood that the firm is a manipulator using sample of GAAP violators investigated by SEC (Teoh et al., 1998a):

M= -4.840+0.920 Days sales in receivable+0.528 Gross margin + 0.404 Asset quality + 0.892 Sales growth +0.115 Depriciation - 0.172 SGA +4.679 Accruals - 0.327 Leverage index.

This model is representative for US firms, while for the sample of emerging countries it may be irrelevant. With the lack of access to the violators investigations in emerging countries it is impossible to find such score for emerging countries sample.

We have investigated different approaches applied in the academic literature and criticism made towards simple or modified Jones model. However, other models are just developments of the Jones pattern and do not improve the specification much. Moreover, they are mostly incorporated in the analyses of the general accounting matters without connection to some events. Thus, for our research we would like to use as the platform for abnormal accruals detection modified Jones model.

This means that we will calculate normal total accruals with the modified Jones approach and extract abnormal accruals from the fitted regression. It is the first step in our modelling section and our hypothesis 1 is that companies in emerging countries do use window dressing of their financial before flotation.

The second step will be testing the following hypotheses:

2) high state ownership (as a proxy - government ownership in a company) increases the probability of earnings manipulation before the IPO;

3) choice of stock exchange (measures investors sophistication) impacts the probability of earnings management. I.e. if company issues stock only on the local market in emerging region than it has higher incentives to manipulate financials due to lower investors sophistication on these markets and vice versa;

4) choice of the auditor should signal about higher probability of managed earnings. We classify auditors of the companies in 4 subgroups: BIG 4 auditors (KPMG, PWC, Earnst&Young and Deloitte), other auditor companies (such as Grand Thornton, Mazars or BDO), not disclosed and unknown in the descending order of their relative quality (as well as another variant - uncorporating dummy variables).

At the third stage we will be answering the question: if abnormal accruals impact post-IPO underperformance of the company? Our hypothesis 5 is that high abnormal accruals negatively influence further company's earnings performance. As a measure of earnings performance we will use return on sales.

Fourth step will be to test hypothesis 6 of the negative correlation between abnormal accruals in IPO year and stock underperformance in the post-IPO period. The proxies for stock performance are CAR and BH returns with a benchmark of the local stock exchange Index.

3. Results

3.1 First step: modeling normal and abnormal accruals

We used Jones modified model to calculate normal or expected level of accruals that the companies in each industry should have had in the year when they did IPO. For this purpose we separated our initial sample in two subsamples: one for estimation and another one to fit the data in equation describing normal accruals.

For estimation we used only firm's financials for at least 3 years post their IPO to be sure that we use not manipulated data. Therefore, we did 8 regressions for each fiscal year starting from 2007 up to 2014. We didn't find expected accruals for years 2000 - 2006 due to the lack of necessary information (i.e. to make regression for 2000 fiscal year we should have used financial data of companies that became public not later than in 1997, however, this data is unavailable as all activity on financial markets in emerging countries started only in 1997. To get regression for 2006 we had only available and relevant data for years 2000-2003, but this was a very small sample to get some confident result). We are satisfied with obtaining normal accruals for the period since 2007 as this was the first pick year for the IPO market with a lot of observations as well as good data existing for the next years up to 2012.

For example, to receive coefficients of the equation of normal accruals in 2014 we obtained financial data for all the companies that issued stock in years 2000 up to 2011. We regressed total accruals (TAC) on the change in sales in the corresponding fiscal year in comparison to the prior year (delSales) and on the property, plant&equipment in the corresponding fiscal year (PPE). To calculate total accruals we used widely accepted method of finding the difference between IPO-year Net income and Cash flow from operations. To take into the account the variability in financials of companies in different industries as well as possible cyclicality in these industries we adjusted each parameter for the industry mean value before performing the regression. All the values in regression were normalized by the total assets of the company in the year prior to the IPO.

We also excluded all the extreme values from the data, so our final sample consisted of 667 companies in year 2014. In different years there was variability in the amount of firms, so in 2013 we used 699 companies for regression, in 2012 - 677, in 2011 - 713, in 2010 - 659 and so on.

We performed all the necessary tests to verify the soundness of the results: test for the normality of residuals, Breusch-Pagan-Godfrey and White tests to check for the presence of heteroskedasicity, Breusch-Godfrey Serial Correlation LM Test to check for autocorrelation, Ramsey RESET Test as a stability diagnosis and Chow forecast test to be sure that we can use the equations obtained for the further forecasting of abnormal accruals from the fitted equation. All the tests were successful and showed that our equations are perfectly specified and can be used for measuring normal accruals.

Next we fitted data from the second subsample to obtain normal level of accruals for the companies of interest. Our second subsample consists of 1088 companies that issued stock for the first time in the years 2007-2012. We extracted abnormal accruals for these firms in their IPO year by taking the difference between their total accruals in IPO year and calculated from the regression expected accruals for the IPO year. This is the simple version of the Jones model, while we also computed modified Jones model. For this purpose we had to gather information about the change in trade receivables in IPO-year compared to the prior year. However this item is available in the DataStream for fewer companies than other financials we used, so in modified model we computed abnormal accruals for 932 companies.

Below we present results for both types of models: for simple model 5 years since IPO and for modified model 3 years since IPO. We also added to the graphs post-IPO earnings performance of the companies as the next goal of this research is to measure the impact of abnormal accruals on the post-IPO earnings performance.

We can observe at the picture 3 unusually high total abnormal accruals with a big difference to the expected value of these accruals in the IPO-year. However in the next 3 to 5 years abnormal accruals converge to the expected level. At the same time we could notice very high return on sales in the IPO year that extremely drops in the next year and continues decreasing in the further periods. As mentioned by Teoh et al. (1998) it could be so that Return on sales drops because new investment made with the obtained during IPO capital and it decreases net income in the nominator of the ratio. However, we also add return on assets for the reference and see that this ratio decreases in the same manner as return on sales. Moreover, we compare dynamics of the above mentioned figures with the cash flow from operations (CFO) and find that CFO has slight upward trend, but never drops, which means that decrease in cash flows (non-accrual indicator) cannot explain earnings underperformance.

3.2 Second step: analyzing impact of several factors of interest on the behavior of accruals

At this stage we are testing the hypotheses:

1) high state ownership (as a proxy 1 - government ownership in a company, 0 - only private ownership) increases the probability of earnings manipulation before the IPO;

2) choice of stock exchange impacts the probability of earnings management (as a proxy we use binary variable: 1 - if IPO only on local stock exchange and 0 - otherwise.

3) choice of the less “widely recognized” auditor should signal about higher probability of managed earnings. We classify auditors of the companies in 4 subgroups: BIG 4 auditors (KPMG, PWC Earnst&Young and Deloitte), other auditor companies (such as Grand Thornton, Mazars or BDO), not disclosed and unknown by the rating from 0 to 3 correspondingly (similar methodology used by Aharony et al., 1993).

As control variables we include:

- as a measure of company size: log of market capitalization;

- as a measure of market timing (average market-to-book ratio as a proxy of investors valuation of a company);

- as a measure of information asymmetry: 1+logAGE;

- as a measure of the company financial position: change in sales in the IPO year and property, plant and equipment scaled by the total sales;

- as control for clustering in geography, industry and time: country dummies, industry dummies and year dummies.

First, we would like to pay attention to the difficulty of obtaining the sound information on the government ownership in the company. We used the data from Database Zephyr where information on the company's GUO (Global Ultimate owner with the highest direct or total share in the company) is reported. We suggest that the real state ownership can be implemented via corporations with government share. Thus our opportunity in the judgement based on the statistical results is very limited and we suggest investigating this problem as a topic of a separate research. Nevertheless, we report our findings according to the available information. Another interesting observation concerns the choice of stock exchange. In this research we show that if the company goes public only on the local exchange it has more incentives to deviate by manipulating its financials. We suggest that this happens due to less strict disclosure rules, high corruption level in the emerging countries and low litigation enforcement standards. The coefficient of the third factor - state ownership confirmed our concerns regarding the quality of data as the initial hypothesis was not accepted. Negative relationship between state share in the company and level of the manipulation is an unexpected result as we suggested that high state participation is an indicator of the low development of emerging markets with the many other problems correlated such as corruption. Significance level is 5% what does not allow to reject the factor that is why we only consider further investigation of the matter.

3.3 Third step: analyzing influence of abnormal accruals on the post-IPO earnings underperformance

In this section we would like to present the results of the regression of abnormal accruals on the earnings of IPO companies. In the paragraph 4.1 we showed the graphs comparing the behavior of abnormal accruals and post-IPO return on sales of the sample firms. We suggested an existence of the correlation between high abnormal accruals in the IPO year and firm's earnings underperformance in the next several periods of time among the IPO firms in emerging countries. We hypothesize that there should be negative relation between these two variables.

To measure the relation statistically we perform the regression of the proxy of earnings performance post-IPO on abnormal accruals and a set of control variables. As a proxy for earnings underperformance we used average 3 years return on sales relative to the base year 0. Abnormal accruals are normalized by the prior to the IPO-year total assets.

As control variables we use cash flows from operations scaled by the sales of the IPO-year and change in capital expenditure (mean capex in the three years following IPO-year related to the base year capex). These two independent variables we included referring to the evidence presented in the paper by Teoh et al. (1998a) that mentioned that CFO indicator should control for correlations between abnormal accruals and earnings as CFO is used for the calculation of total accruals in IPO year. Capex controls for the firms that used capital raised in IPO for the further investment. As reported by Cheng (1995) coefficient of capex should have a positive sign. We also include expected accruals in the regression, though we suggest that it should not have any influence on the underperformance.

Finally, we also control for the clustering in geography, industry and time by introducing corresponding dummies in the regression (we do not report their coefficients).

3.4 Fourth step: analyzing influence of abnormal accruals on the post-IPO stock price underperformance

financial market сash transaction

The last goal of this research was to investigate possible correlation between abnormal returns and the stock price underperformance often observed on the markets post-IPO. This topic raised many questions in the literature and was explored from different points of view, for example in scope of the market timing approach. We would like to add to this literature our research of the impact of the factor of abnormal accruals as one of the reasons for firms' post-IPO underperformance in the emerging countries.

Following the idea developed in the article by Teoh et al. (1998a) we would like also to investigate the stock behavior in 4 groups of companies ranged by their abnormal accruals. Thus we split our sample in the most aggressive subsample, less aggressive, less conservative and conservative, meaning that aggressive firms actively manipulate their earnings while conservative companies have low abnormal accruals or zero.

It can be clearly noticed that firms aggressively managing their earnings underperform more severely than those who formed financial statements more realistically (or conservatively). This is also may be seen from the Picture 6 that presents time distribution of firm's BH returns for 4 subsamples grouped accordingly to the level of earnings management as described above. This graphs indicates that aggressive companies with high abnormal accruals underperform much more than conservative firms for at least 2,5 years, while conservatively managed earnings correspond to the outperformance by the firms at least from the 1,5 years.

This forces us to hypothesize that high earnings management by the firms before IPO is reflected in the market pricing of the misleading information in the stock prices.

To verify the hypothesis we perform statistical modelling by regressing firms' post-IPO stock performance on the abnormal accruals. As a proxy for stock performance we use Buy-and-Hold returns (BH returns) and CAR returns aggregated in time, calculated 6 months post-IPO and 30 months since this 6 months period.

As control variables we use (according to Teoh et al. (1998b)) three-year market returns from the stock exchange that listed the IPO and first-day IPO underpricing return to control, for the volatility and firm's initial point of the underperformance after that returns further go down, log capitalization as a measure of size, 1+log firm's age as a measure of investors sophistication, change in CAPEX calculated in the same way as in regression 4.3 and having the same rational and change in Net Income that “measures the incremental effect of the accrual variables over an income growth variable” [Teoh et al. (1998b), p. 1955]. We also control as in the previous regressions for clustering in time, industry and geography.

As previously we performed all the necessary tests to ensure correct specification of the model and received highly significant results confirming our hypothesis for negative correlation between earnings management measured by abnormal accruals and post-IPO long-term stock performance.

Conclusion

This research is devoted to the investigation of different aspects of the abnormal accruals phenomena in emerging countries. We developed analyses based on four steps to address the questions of interest. Firstly, we extracted discretionary accruals component from the well-known Jones model (as well as its modified version) and used these values to verify other ideas. Results of the second model provided the evidence on two important indicators characterizing emerging market countries: choice of the stock exchange (local versus international) and of the auditor (lower quality versus higher quality). Both factors are significantly positively correlated to the abnormal accruals. This fact highlights our suggestion that underdeveloped emerging markets give more incentives to the firms to window-dress their financials before IPO and create negative consequences of the information asymmetry for investors. These consequences are reflected in the post-IPO earnings and stock underperformance of the listed companies.

In this research it is shown that firms with high IPO-year abnormal accruals significantly underperform by both measures. Moreover, if sample of firms is ranged in subgroups by the level of earnings management (measured by discretionary accruals), than the subsample with the companies implementing highly aggressive policy regarding accruals management underperform much strongly than so called conservative firms (that have very low or no abnormal accruals).

...