Are Dutch companies conservative in their allowance for doubtful accounts? : evidence from listed companies in the Netherlands

(1)

- 1 -

University of Amsterdam

Amsterdam Business School

Master in International Finance

Master Thesis

Are Dutch companies conservative in their allowance for

doubtful accounts? Evidence from listed companies in the

Netherlands.

Author:

Kilian van Loenen

Thesis Supervisor:

Dhr. prof. dr. David Veenman

Date:

(2)

- 2 - Abstract

This paper examines the existence of conservatism in the financial statements of listed companies in the Netherlands. To test whether companies are conservative in taking an allowance for accounts receivable, and whether there is evidence that this could facilitate earnings management, I examine the movement of accounts receivable allowances among listed companies in the Netherlands during the period 2007-2016. The information obtained was analyzed in order to test if there was conservatism in the companies. The outcome of the test for conservatism was further analyzed with regressions to control for the size of the companies in the sample, and to see if conservatism was influenced by time, bankruptcy, or the outcome of the income statement. The main finding in this research is that listed companies in the Netherlands are conservative in allowance taking for accounts receivable, but that this conservatism does not increase over time, is not influenced by bankruptcy, and the impact on the income statement has no influence on levels of bankruptcy. These results support the conclusion that there is no evidence of earnings management.

(3)

- 3 -

3.1.4 H4: “The outcome of the profit and loss statement has no effect on the allowance for accounts receivable.” ... - 18 - 4. Methodology ... - 19 - 4.1 Methodology ... - 19 - 4.2 Sample ... - 22 - 5. Results ... - 24 - 5.1 Bootstrapping method ... - 25 - 5.2 Hypothesis 1 ... - 26 - 5.3 Hypothesis 2 ... - 27 - 5.4 Hypothesis 3 ... - 29 - 5.5 Hypothesis 4 ... - 31 - 5.6 Complete regression ... - 32 -

6. Conclusions and implications ... - 34 -

(4)

- 4 -

1. Introduction

In the last few decades the global economy has changed. Undoubtedly, major factors driving the change were increasing globalization, business integrity, and technology. One of the topics discussed within these last decades has been the influence of conservatism on earnings quality (Jackson & Lui (2010), Watts (2003), Dechow, et al. (2010), and many more). Companies around the world prepare financial statements to inform their stakeholders. A stakeholder could be the shareholder, the financier, or the analyst who prepares a buy or sell recommendation. For these stakeholders, it is important that financial statements represent a trustworthy view of the company. However, some have concluded that the influence of conservatism on financial statements could lead to decreased value for stakeholders through lower earnings (Cyert, et al. 1962), and that conservatism could eventually, facilitate earnings management.

This study will examine whether conservatism is present in the financial statements of listed companies in the Netherlands. I define conservatism in accounts as policies or tendencies that result in the downward bias of accounting net asset value relative to economic net asset value. Conservatism embedded in the financial statements of companies can influence stock prices, and may be used by managers to misguide equity markets via the management of earnings. Therefore, it is important to determine to what extent companies in the Netherlands are conservative in their financial statements. Previous studies show that conservatism in financial statements can influence the stock price of a company (Kasznik & McNichols 2002), and can be used by managers to conduct earnings management (Jackson & Lui 2010 and Dechow, et al. (2010). Moreover, it can reduce the accuracy of analysts’ forecasts (Ruch & Taylor 2014), and create hidden reserves (Penman & Zhang 2002).

To examine conservatism, I will start by modeling the movements in the allowance of accounts receivable. Many researchers have studied conservatism in financial statements, but only a few have focused on this specific item in the annual report (Jackson & Lui 2010). The advantage of assessing the allowance is its simplicity in verifying whether the company is conservative. This is done with the following test: Company X takes an allowance during period zero, and the expectation is that the full allowance will be written off in the period afterwards. If this is not the case, and the allowance of period zero exceeds this write-off in the period afterwards, the allowance taken in the previous period was too conservative (Jackson & Lui 2010). Secondly, I will examine whether the level of conservatism increases over the time. According to Watts

(5)

- 5 -

(2003a), companies have a tendency to become more conservative over the years, which leads to higher reserves on their balance sheets, resulting in lower accounts receivable. The research design of Jackson & Lui (2010) will be used to determine whether the percentage of provision for doubtful accounts (compared with accounts receivable) increases over the years. Thirdly, I will examine the influence of bankruptcy and earnings on conservatism in the allowance of accounts receivable. The evidence found in my study can help to verify the presence of conservatism within companies and whether the allowance is used to manage earnings. When companies are too conservative in taking allowances, they understate accounts receivable, thereby reducing earnings and creating a reserve, which gives managers the flexibility to release the reserve and to manage earnings upwards in the future (Jackson & Lui 2010 and Dechow et al. 2010).

To examine all the issues described above, I use a sample of companies listed in the Netherlands over the period 2007–2016. The sample represents companies engaged in a wide variety of activities. This study differs from previous studies on conservatism given the more recent period of investigation and the particular market, the Netherlands. Previous research found evidence of conservatism in financial statements, but focused on samples that reflected periods mainly before the crisis of 2009 (Ruch & Taylor 2014). With my sample, a more recent period will be examined to see if such theories of conservatism remain valid in today’s world. Previous research has almost always studied samples in the United States. Based on the research of Ball, et al. (2000), which found differences in conservatism between common and code law countries, there may be differences between the United States and the Netherlands, in conservatism. In the Netherlands, it is common to insure accounts receivable against bankruptcy, which means that the allowance taken during a year will be lower, given that the accounts are insured. In the United States, however, such insurance is utilized far less often. Therefore, my study will contribute to the academic literature by examining whether conservatism is present in the Netherlands, and whether conservative theories still hold true in the recent past.

This study finds evidence of conservatism in the accounts of listed companies in the Netherlands. Based on my findings, there is no evidence to support the conclusion that the level of conservatism grows over the years. Neither is there evidence that companies use their accounts receivable allowance to conduct earnings management. The implications of these

(6)

- 6 -

results are that equity investors may misjudge the existence of conservatism in financial accounts, which could lead to lower stock valuations, and eventually, poor investment decisions by equity investors. This study supports the theory of Watts (2003a), that all companies are conservative. However, it opposes the other theory of Watts (2003a) & Basu (1997), that conservatism has the tendency to grow over the years. This study is closely linked to the research of Jackson & Lui (2010), which argues that conservatism in the allowance of accounts receivable can facilitate earnings management. However, my study disputes this theory, finding no evidence of significant difference between the level of conservatism and provision taken by profitable and unprofitable companies.

This paper first reviews the literature regarding the topics discussed herein, then describes the sample, explains the hypotheses and the variables used, performs the regressions to test the hypotheses, and, finally, discusses the conclusions of this research.

2. Literature review

Many research papers study conservatism in financial statements. Most of these papers focus on companies in the United States (U.S.). My study closely follows the work of Jackson & Lui (2010), which focused on companies in the U.S. in order to determine whether conservatism in accounts receivable leads to earnings management. More specifically, I will focus on three main topics: conservatism in financial statements, earnings management, and the allowance of accounts receivable, all of which have been intensively discussed in previous literature.

2.1 Conservatism

Conservatism in the financial world has been investigated widely, for example, by Jackson & Lui (2010), Basu (1997), Ruch & Taylor (2014), Cyert, et al. (1962), Cimini (2015), Dechow, et al. (2010), and Watts (2003). Conservatism has influenced accounting for a long time. Basu (1997) found evidence of conservatism in 15th century trading partnerships. However, accounting conservatism remains a controversial topic within academia—on one side there is the value relevance approach, and on the other side, the contracting efficiency approach (Mora & Walker 2015). Conservatism in accounts can be defined as accounting policies or tendencies that result in the downward bias of accounting net asset value relative to economic net asset value. Conservatism can be seen as the cautionary principle, whereby companies anticipate no

(7)

- 7 -

profits and provide for all probable losses (Bliss 1924), or as a way to report low values for assets and the highest value for liabilities (Watts & Zimmerman 1986).

There are many studies that indicate the existence of conservatism. But the first question that we need to pose is why conservatism exists in financial statements? According to Watts (2003a), all explanations of conservatism seem to benefit the users of financial statements. He argues that conservatism can be explained through contracting, shareholder litigation, taxation, and the litigation of regulators.

The first explanation is contracting. This says that conservative accounting is a way of addressing moral hazard caused by parties to the firm having asymmetric information, asymmetric payoffs, limited horizons, and limited liability. Most studies use this explanation because it represents an early source of conservatism, and its arguments are fully developed and support evidence of the contracting hypothesis. Watts (2003a) comprised this explanation of conservatism in three distinct theories: debt, compensation and governance. Blunck (2007) focused his research on these three theories to find evidence of conservatism.

Blunck (2007) argued that the debt theory comes from the asymmetric payoff of debt-holders; therefore the debt-holders want managers to be conservative in accounting. The reason for this is that debt-holders have little to gain when a company has better than expected performance, but have much to lose when company performance is poor. Therefore, debt-holders place covenants in debt contracts that transfer assets to debt-holders when such covenants are breached. With these covenants, debt holders try to limit their losses and protect their interests in the event of liquidation. For example, this can be illustrated by looking at a factoring company that finances two companies (Over V and Under V), based on their accounts receivable. The first company, Over V, writes off more receivables than their allowance during the year. The second company, Under V, takes a higher allowance than write-off (conservative). Both companies have accounts receivable at year-end of 90. In the case of bankruptcy, the factoring company will have a larger residual debt with Over V than with Under V, which has upside related to its allowance for accounts receivable, which can still be retrieved. Another way that conservatism affects debt, according to Blunck (2007), is that debt holders place restrictions in debt contracts that restrict managers from paying liquidating dividends to

(8)

- 8 -

shareholders. Conservative accounting also ensures that debt holders receive bad news in a timely manner so they can exercise their rights.

The compensation theory according to Blunck (2007) is based on the ex-post settling up problem. This comes from managers and shareholders having different priorities related to time horizon, and incentivizes managers to focus on current performance instead of the long-term value of the company (Blunck 2007) (Graham et al. 2005). Managers’ payoffs are normally based on short-term performance, while shareholders are typically more focused on the long-term value of the company. Conservative accounting allows companies to recognize bad news as it arises, while recognizing good news only once it is realized. In accounting this is called the prudence principle. For example, conservatism can limit opportunistic behavior in reporting among managers because good news has to be realized before it can be inserted in the accounts, while bad news has to be recognized in the accounts, immediately.

The last theory of contracting conservatism is the governance theory. This implies that asymmetric timeliness of earnings helps shareholders monitor managers more effectively. Since bad news has to be reported immediately, shareholders receive this information in a way that gives them time to respond to managers, make changes, or even replace management (Blunck 2007).

A second explanation for conservatism is shareholder litigation, which also produces asymmetric payoffs by overstating company assets. By overstating assets, companies stand a higher chance of generating litigation costs than when assets are understated. In other words, financial statements will be undervalued by conservatism, which will reduce the likelihood of lawsuits against the shareholders. Following this explanation, a study from LaFond & Watts (2008) investigated the effect of information asymmetries between managers and outside equity investors for conservatism. They found evidence that supports the claim that equity investors demand more conservative earnings as a means of mitigating agency problems and that conservatism increases following increases in information asymmetries.

According to Watts (2003a), a third explanation for conservatism comes from the link between taxation and reporting, which can generate conservatism in financial reporting. With conservative accounting, the additional allowance taken will lower the earnings. This allows

(9)

- 9 -

companies to spread their earnings over multiple years, increasing the company’s forecasting power, which is beneficial for the stakeholders.

A fourth and final explanation for conservatism arises from institutional financial reporting standards and regulators (Watts 2003a), where conservatism can reduce the political costs imposed on them. For example, asymmetries in litigation costs are consistent with non-contracting parties (regulators and auditors) valuing conservatism’s constraint on opportunistic payments to managers and other parties.

In the literature, two types of conservatism are distinguished: unconditional and conditional. Unconditional conservatism is the systematic reduction of an accounting number without regarding information or an event. This type of conservatism is also called balance sheet conservatism (Mora & Walker 2015). Examples of this type of conservatism include accelerated depreciation methods and expensing research and development costs (Brinck 2007). This type of conservatism is mainly found in the explanations of conservatism as shareholder litigation, taxation, and the litigation of regulators (Qiang 2007). Unconditional conservatism leads to understated assets, but does not always lead to lower earnings because accelerating deprecation only affects the timing of the earnings (Mora & Walker 2015). The other type of conservatism is conditional conservatism, which is defined by Brinck (2007) as asymmetric timeliness or asymmetric verification, conditional on an event. Both definitions require managers to postpone good news until it is realized, and to realize bad news when it is known. Conditional conservatism can be explained by looking at a company that receives new information regarding the quality of its accounts receivable. Under conditional conservatism, when the company receives bad news about the collection of its accounts receivable, a loss will be reported immediately. In the case of good news, the announcement is postponed until the receivables are realized. Both these types of conservatism tend to reduce earnings and book value by inflating additional costs during the year. Based on their research, Ruch & Taylor (2014) argue that conditional conservatism has a negative effect on earnings persistence, while unconditional conservatism could facilitate earnings management.

Another way to view conservatism is to examine its effects on earnings quality. In their research, Ruch & Taylor (2014) reviewed the literature on this subject by grouping the earnings quality proxies identified in Dechow et al. (2010) into three categories: (1) the time series

(10)

- 10 -

properties of earnings, (2) accrual quality, and (3) the presence of earnings management. For each of the three groups they explained the effects through unconditional and conditional conservatism.

Under conditional conservatism, the time-series properties of earnings is the timely recognition of bad news and the deferred recognition of good news (Ruch & Taylor). Given that this comes from an unexpected event, it cannot be predicted. Basu (1997) argues that operating losses are less persistent than operating gains, which arise from the different timing of good and bad news. Ruch & Taylor (2014) found that research (Dichev & Tang 2008 and Chen et al. 2013) related to the effects of conditional conservatism on the time-series properties of earnings, consistently argues that this type of conservatism reduces earnings persistence and predictability. On the other hand, the impact of unconditional conservatism is more difficult to assess, and depends on the conservatism practice used. For example, straight-line depreciation will lead to more persistent earnings than accelerated depreciation. Accelerated depreciation results in less persistent earnings because higher depreciation is taken during the first couple of years, reducing earnings. In the years afterward, the depreciation will be lower, resulting in higher earnings. (Ruch & Taylor 2014).

Accrual quality of earnings under conditional conservatism is characterized by asset write-downs in bad news periods and gain deferral in good news periods. The magnitude of accrual earnings is likely to increase during bad new periods, and to decrease during good news periods (Ruch & Taylor 2014). Under unconditional conservatism, the methods that influence the impact of accrual quality are accelerated depreciation and the LIFO (last in, first out) accounting principle. This can be seen in the example above, as the LIFO reserve may be released into accrual earnings, reducing earnings persistence.

Presence of earnings management is the third and last effect of earnings quality as described in the research of Dechow et al. (2010). According to Ruch & Taylor (2014), it is unlikely that conditional conservatism would have a significant relationship with earnings management. Due to the nature of conditional conservatism, managers have no influence over news events. However, when there is a provision created, managers may have influence in the release of the provision, which can lead to earnings management. Under unconditional conservatism, managers can steer results through the accumulation of reserves from accelerated depreciations,

(11)

- 11 -

which appear on the balance sheet, which can be released into earnings when targets need to be met (Ruch & Taylor 2014). Evidence of this claim can be found in the study of Jackson & Lui (2010).

In the research of Ruch & Taylor (2014), a distinction is made between financial users, who are affected by conservatism in accounts and equity market users, debt market users, and corporate governance users. The literature on conservatism is inconclusive as to whether the effects of conservatism on financial statements are positive or negative for equity market users. There is a lack of consensus among accounting researchers about the cost and benefits for this group. For debt market users, Ruch & Taylor (2014) concluded that conservatism is a benefit for both lenders and borrowers. As for corporate governance users, Ruch & Taylor (2014) found this group to consistently demonstrate that conditional conservatism benefits corporate governance by providing earnings that are more relevant to compensation decisions and inducing management to make better investment decisions.

Even between markets there are differences in the conservatism of financial statements. A study from Ball, et al. (2000) found differences in the conservatism of financial statements in common-law countries (Australia, Canada, United Kingdom and United States) and code-law countries (France, Germany and Japan). The study found evidence in forty thousand accounting years of higher conditional conservatism in common law countries compared with code-law countries. It arises from the monitoring of managers, which is an important feature in common-law corporate governance, and this increases incentives for managers to attend to the sources of losses more quickly (Ball, et al. 2000). Later, other studies more focused on Europe highlighted the same conclusion—common law countries exhibit higher levels of conditional conservatism than code law countries (Mora & Walker 2015). Mora & Walker (2015) derived from their literature research that higher levels of conditional conservatism are associated with countries with more developed stock markets, higher investor protection, and litigation risk. However, higher levels of unconditional conservatism are associated with less developed stock markets, and/or countries in which banks are the main source of ﬁnance, and where accounting choices are affected by tax consideration.

A study by D'Augusta, et al. (2015) demonstrated that conservatism does not only affect financial statements. In their research, they showed that conservatism also affects investor

(12)

- 12 -

disagreement around annual earnings announcement dates. This effect is an overall negative association between a ﬁrm’s level of conservatism and proxies of disagreement based on abnormal volume on announcement days. This disagreement creates movement in the equity market, which is, among other bodies, influenced by conservatism.

One of the first studies to claim that conservatism facilitates earnings management was Jackson & Lui (2010), who tested the allowance of accounts receivable for listed companies in the US during the period 1980-2004. They found evidence of conservatism in their sample. By looking at the fluctuation in bad debt expense, they found evidence that companies manage their bad debt expense downward to meet or beat analysts’ earnings-per-share forecasts. With these two findings, they could support the claim that conservatism in financial statements could be used for earnings management by utilizing the allowance for accounts receivable. Earnings management is a subject closely linked to conservatism in academia. When a firm manager understates their financial accounts, they have the ability to steer the result.

2.2 Earnings management

Earnings management is the use of accounting techniques to produce financial reports that present an overly positive view of a company's business activities and financial position. Schipper (1989) defines earnings management as a purposeful intervention in the external financial reporting process, with the intent of obtaining some private gain (as opposed to, say, merely facilitating the neutral operation of the process). Over the last decade, literature on this topic increased significantly. Earlier studies focused on describing earnings management, but in the last decade these focused more on finding evidence of earnings management.

In the current environment, where financial analysts can share their opinion with the entire world, earnings management has become more important. Earnings management gives managers opportunities to influence and manage earnings to meet the expectations of analysts, who will be rewarded with a higher stock price (Kasznik & McNichols 2002). Still, despite the available literature on earnings management, it is difficult for researchers to document its existence. The frequently cited research of Healy & Wahlen (1999) reviews the academic evidence for earnings management and its implications for accounting standard setters and regulators. They found that past studies used specific accruals to find evidence of earnings management, such as bank loan loss provisions, claim loss reserves for property-casualty

(13)

- 13 -

insurers, and deferred tax valuation allowances. The conclusion was that at the moment, literature on earnings management only provides modest insights for standard setters. This derives from the fact that previous academic research tries to document earnings management, without providing evidence.

One way to find evidence of earnings management is through models. During the last decades, researchers made models to test for earnings management. In the study of Dechow, et al. (1995), the performance of some of these models where tested by comparing the specification and power of commonly used test statistics. The models that are tested in this study are those from Healy (1985), DeAngelo (1986), and Jones (1991). After testing, they concluded that the models misspecified tests because of the conditions in which the tests were taken (not in line with their own examination, as discussed in their own research). However, the modified version of the model from Jones (1991) is the most powerful test for earnings management. This led to the following conclusion—the ability of the models to test for earnings management needed to be further developed and improved. The research of Dechow, et al. (1995) has been renewed by Dechow, et al. (2011), providing a new approach to testing for accrual-based earnings management. Older models are discussed to explain their limitations because current models (for measuring earnings management) lack power and are often misspecified.

A model similar to the one I used is the one from McNichols & Wilson (1988), which tested the allowance of accounts receivable to test for earnings management. The main difference between the McNichols and Wilson model and the other models is that it is based on one single accrual and uses GAAP (General Accepted Accounting Principles) to model for the right value of the single accrual without earnings management. As a result, they found evidence that companies manage their earnings by choosing income-decreasing accruals when income is extreme, which can be explained as earnings management.

According to Burgstahler & Dichev (1997), it is common for companies to conduct earnings management to avoid small losses and decrease earnings. They found evidence of earnings management in a broad range of firms, time periods, and previous earnings history. Burgstahler & Dichev (1997) also found that analysts incorporate anticipated earnings management in their earnings forecasts. Jackson & Lui (2010) found evidence of the link between conservatism and earnings management in their research. Other critics have claimed the existence of this link, but

(14)

- 14 -

without evidence supporting the claim. My research is another step in support of the link between conservatism and earnings management based on Jackson & Lui’s (2010) study of the allowance of accounts receivable.

2.3 Allowance of accounts receivable

My research is mainly focused on the movement of the allowance of accounts receivable, on which academic literature is limited given it is a small item in the annual report. Research from Cyert, et al. (1962) indicates that this small item can have a large influence on company earnings. The allowance for accounts receivable is a contra-asset account, which reduces the total receivables reported to reflect only the accounts receivable expected to be collected. The allowance is established by recognizing a bad debt loss on the financial statements within the same accounting period that the associated sale is reported.

In then past, the method for determining the allowance was a process that was done in two steps. First, the accounts receivable were placed in age categories, the second step consisted in an estimation of the needed provision based on loss expectancy and past experience (Cyert, et al. 1962). In order to improve this process a model was developed using Markov’s chains to describe the loss process. With this model, it was easier to calculate the allowance, but even this model needed improvement. Kuelen, et al. (1981) followed-up on this, and found a small flaw in the aging of partial payments.

To analyze the estimation of the allowance of accounts receivable, Riley & Pasewark (2009) tested three methods that can be used by auditors to see if the allowance represents a true value in line with previous write-offs. The first method is comparing bad debt expense to write-offs. This means that cumulative bad debt expense would be compared to cumulative write-offs over multiple years. The benchmark used for this method (∑ Bad debt expense ÷ ∑ Write-off) is 1.0 for a multiple year method. The second method is comparing the beginning allowance for accounts receivable to write-offs. This means that each year, beginning-of-year allowance will be compared to that year’s write-offs. The benchmark for this method (Allowance beginning of the year ÷ write-off during the year) is between 1 and 2. The last method discussed is assessing the allowance exhaustion rate. This is done by calculating the number of years it will take until the beginning of the year allowance is completely utilized in the write-offs. The benchmark for this method is 1 or 2 years. All these benchmarks are based on researchers’ opinions, but they

(15)

- 15 -

provide a fair indication of the normal level of allowance. These methods have been used to test the past allowance of U.S. listed companies. In the research of Riley & Pasewark (2009), evidence was found of higher outcomes, which indicated conservatism in the accounts could be used for earnings management.

3. Hypotheses

Existing studies reveal that conservatism in financial statements might benefit the users of these statements, influence the value of a company, and potentially facilitate earnings management. Most of these studies are done in one geographic area in the world and are usually focused on older samples. This makes it interesting to see if the results and theories of past studies still hold in another time period and geographic area. One can argue that conservatism is something that can be found in all financial statements. However, after looking at the studies, I see that auditors, shareholders and analysts have become stricter regarding conservatism in financial statements. I therefore argue that the results found in the past can differ when testing for conservatism in a new market and different time period. This argument is supported by the study of Ball, et al. (2000), which indicated that results in the U.S. may differ from results in the Netherlands.

Looking at the explanation of conservatism, I think that company managers have incentives to increase their earnings as much as possible, which will decrease conservatism in financial statements and increase manager income. Moreover, conservatism in financial statements leads to lower earnings because of the higher expense taking in the year, which can reduce the value of the company. Yet at the same time, shareholders and debt holders do not want to be exposed to litigation costs, which increase when conservatism is low in financial statements. So, it remains beneficial for managers to be conservative in financial statements to maintain their positions. In order to maintain their positions, managers want room to manage earnings in order to meet analysts’ forecasts in other periods. One way to do this is through the allowance of accounts receivable. At the moment, it remains difficult for companies and outsiders to estimate this allowance, which makes it easier for managers to overestimate it. All of these theories and studies lead to the following null hypotheses that are the focus of my research.

H1. “Dutch companies are not conservative in their allowance for accounts receivable.” H2. “There is no evidence of higher conservatism in the allowance over multiple years in

(16)

- 16 -

H3. “The bankruptcy level in the Netherlands has no effect on the level of conservatism in the allowance of accounts receivable.”

H4. “The outcome of the profit and loss statement has no effect on the allowance for accounts receivable.”

3.1.1 H1: “Dutch companies are not conservative in their allowance for accounts receivable.”

Previous literature studying conservatism shows that conservatism can come from contracting, shareholder litigation, taxation, and the litigation of regulators (Watts 2003a). In my research, I look for unconditional conservatism, which means that the taken allowance comes from bad news signals. These signals must be realized one year afterward because accounts receivable are due within one year. To detect conservatism in my sample, I investigate the allowance against write-offs over consecutive years. I will use the model described by Jackson & Lui (2010), which allows me to obtain a level of conservatism for every company for all years. The reason that conservatism can be found in this allowance can be explained by the debt theory. Almost all companies used in the sample pledge their accounts receivable to secure their debt financing. Correspondingly, their lenders then demand that the accounts receivable are correctly estimated in order to minimize negative outcomes during bankruptcy, thereby giving the manager an incentive to be conservative in taking an allowance. My expectation is that companies are conservative in taking the allowance based on the fact that they do not want negative surprises, which can jeopardize the continuation of the company. In this case, lenders could initiate a lawsuit against the management of the company, something managers would want to prevent. Therefore, I expect that companies in the Netherlands are conservative in taking an allowance for accounts receivable. My expectation supports the alternative hypothesis, which is stated as follows:

(17)

- 17 -

3.1.2 H2: “There is no evidence of higher conservatism in the allowance over multiple years in Dutch companies.”

The research of Watts (2003a) found that conservatism evolved over the years. In his research, evidence suggested that accounting had become significantly more conservative (Watts 2003b). One study that supports this theory is Jackson & Lui (2010), who found evidence of higher levels of conservatism over the years in companies in the United States. This increase in the size of the allowance comes from the fact that companies are conservative year after year. To better understand how the allowance becomes more conservative over time, look at this imaginary company. The company has a write-off of 50 from year 1 to 10 and takes an allowance of 100 in those years. Every year the company overstates the write-off by 50, which is taken as a provision on the balance sheet. In year 1, the allowance is 50, and the year afterward this increases to 100, until it ultimately reaches 500 after ten years. Given that the overstatement is only 50 during any one-year, we see that in ten years this eventually leads to an allowance that can be used for write-offs over the next ten years. My expectation is that the allowance becomes more conservative over time, supporting the alternative hypothesis. The alternative hypothesis is stated as follows:

𝐻 : The level of conservatism increases over the years.

3.1.3 H3: “The bankruptcy level in the Netherlands has no effect on the level of conservatism in the allowance of accounts receivable.”

One of the explanations of conservatism is shareholder litigation. An example of this is the higher litigation costs for companies with largely overstated assets, because these overstatements can lead to lawsuits. In order to protect the company against this, managers understate accounts to decrease litigation costs. With this hypothesis, I will test whether the level of bankruptcy has a considerable negative influence on the conservatism of companies. One can argue that managers will increase the allowance in a time with more bankruptcies in order to reduce litigation costs and chances of a negative surprise in their receivables. On the other hand, when the level of bankruptcies increase, I would expect that there would be a higher use of the allowances, resulting in a negative influence of bankruptcy on conservatism. If this test provides evidence that companies have a high level of conservatism, even when the level

(18)

- 18 -

of bankruptcy is low, it will support the claim of conservatism. Another argument is that it is common for companies in the Netherlands to insure their accounts receivable against bankruptcy. If this is true, the level of bankruptcy will not influence the level of conservatism, because companies secured by insurance will not feel as pressured by bankruptcy. Therefore, I expect that the level of bankruptcy has no influence on the level of conservatism. My expectation supports the alternative hypothesis, which is stated as follow:

𝐻 : The level of bankruptcy has no influence on the level of conservatism.

3.1.4 H4: “The outcome of the profit and loss statement has no effect on the allowance for accounts receivable.”

In the research leading up to this hypothesis, I tested for conservatism in the allowance, looked at the change in the allowance, and tested the influence of the level of bankruptcy. In the literature discussing conservatism there are many researchers who found evidence of conservatism in financial statements. Some of them argue that this conservatism is linked to earnings management. Earnings management states that managers manage financials to their own benefit and adjust them to achieve a certain goal. One of these explanations is that companies use conservatism in financial statements to reduce earnings and lower tax. Others found evidence that conservatism was used to meet analysts’ predictions (Jackson & Lui 2010), which can eventually lead to higher valuations in the equity market (Kasznik & McNichols 2002).

Burgstahler & Eames (1997) argued that companies will most often use earnings management to avoid losses and decrease earnings. When equity investors hold shares of many companies they do not have the time, resources, experience, or knowledge to review all companies. On average, shareholders will probably look only at earnings and forecasts. Therefore, it can be beneficial for managers to manage earnings and “fool” these investors. With this hypothesis, I will test if conservatism in the allowance is influenced by the outcome of the income statements of a company. In other words, are there lower allowances taken during a loss-making year to avoid a loss or increase in earnings? I argue that companies change their allowance in order to manage earnings. Companies will benefit from the allowance with steady earnings, better

(19)

- 19 -

predictability for the analyst, and—in extreme years—lower the taxation towards lower performing years. Thus, I state the alternative hypothesis as follow:

𝐻 : There is difference between the allowance taken when a company makes a profit or a loss.

4. Methodology

4.1 Methodology

There are many different measures used to assess conservatism in the existing literature. In my research, I will use the calculation described by Jackson & Lui (2010), comparing the allowance against write-offs in consecutive years. In order to assess conservatism in the allowance of listed companies in the Netherlands, a company’s ratio of conservatism is modeled for every company and every year resulting in the following new variables:

𝐶𝑜𝑛 =𝐴𝑙𝑙𝑜𝑤 𝐴𝑑𝑑

𝑊𝑂 𝑎𝑛𝑑 𝐶𝑜𝑛_𝐻2 =

𝐴𝑙𝑙𝑜𝑤 𝐴𝐶𝐶 𝑅𝑒𝑣

In the calculation, Allow Add is the addition to the allowance in year zero (all additions are positive in the data set), WO is the write-off during the year afterward, Allow is the allowance of the doubtful accounts on the balance sheet, and ACC Rev is the net amount of accounts receivable. Variable 𝐶𝑜𝑛_𝐻1 will be used to measure the level of conservatism, and variable 𝐶𝑜𝑛_𝐻2 will measure the level of the provision compared to the accounts receivable. These two new variables are complementary to each other because companies can be conservative, but if the provision does not increase, the company could be conducting earnings management. With these two variables, both sides of the annual report will be tested by looking at the income statement through allowance (𝐶𝑜𝑛_𝐻1) and the balance sheet through provision (𝐶𝑜𝑛_𝐻2). The outcome can only be reliable when the allowance addition and write-off meet these three requirements: first, the receivables need to be classified as current, because accounts receivable should be received within one year (as given in the outline of the accounting policy). Second, the addition needs to be cleared from other items like foreign exchange rate changes or acquisitions. Third, the write-offs need to be clear of releases. In the sample, this has been done by using the specification given in the annual report.

(20)

- 20 -

As an illustration, I will test if the allowance of Akzo Nobel is conservative: The allowance taken in 2014 is 35, the write-off the next year is 18

The allowance taken in 2013 is 40, the write-off the next year is 19 The allowance taken in 2012 is 46, the write-off the next year is 19 The allowance taken in 2011 is 36, the write-off the next year is 25 The allowance taken in 2010 is 33, the write-off the next year is 32 Using the formula results in the following calculation per year: 2014 35/18=1.94

2013 40/19=2.10 2012 46/19=2.42 2011 36/25=1.44 2010 33/32=1.03

Based on this simple model, the conclusion is that the allowance taken by Akzo Nobel is conservative in the sample period. This can be seen in that the ratio is higher than the benchmark. When the company has an outcome of 1, it indicates that the allowance taken in period is the same as the write-off during the following period.

In order to test whether my findings are conservative, the value of Con_H1 needs to exceed 1; this required value is the benchmark. A value that is higher than 1 indicates that the allowance taken in year zero is higher than the write-off during the following year, which can be seen as conservatism because the allowance was too high. In order to see if the values found in my sample are significantly different compared with the benchmark, a mean and standard deviation per year for the full sample is calculated. The mean and standard deviation will be tested with a t-test with a 95% confidence level.

𝑥̅ − 𝜇 𝑆 √𝑛

𝐻 : 𝜇 = 𝜇 𝑜𝑟 𝐻 : 𝜇 ≠ 𝜇

With this t-test, I will test my findings of Con_H1 to see if there is a significant difference between these values and the benchmark of 1 with a confidence level of 95%. If Con_H1 is significant, it means that I can say with a confidence level of 95% that there are outcomes of Con_H1 that exceed 1, indicating that the companies are conservative. When the outcomes of Con_H1 are not significant, it means that there was no evidence found that Con_H1 has values above 1, indicating that the companies are not conservative. In the t-test, x̄ is the mean of the

(21)

- 21 -

sample, S is the standard deviation of the sample, n is the number of findings in my sample, and 𝜇 is the test value, in my case, the benchmark of 1. If the findings of Con_H1 do not differ significantly from 1 in the t-test, the companies are not conservative which means I accept the null hypothesis. If the findings are greater than 1 in the t-test, there is a significant difference between the outcomes of Con_H1 and the benchmark. Therefore, I would reject the null hypothesis and accept the alternative hypothesis, meaning that the companies are conservative. Afterwards, the variable that obtains the level of conservatism and the level of provision will be used in the regression as the dependent variable.

Con_H1_base 𝐴𝑙𝑙𝑜𝑤 𝑎𝑑𝑑 𝑊𝑂 = 𝛽 + 𝛽 𝑇𝑖𝑚𝑒 + 𝛽 𝐵𝑎𝑛𝑘𝑟𝑢𝑝𝑡𝑐𝑦 + 𝛽 𝑅𝑒𝑠𝑢𝑙𝑡 + 𝛽 𝐴𝑠𝑠𝑒𝑡𝑠 + 𝜀 Con_H2_alternative 𝐴𝑙𝑙𝑜𝑤 𝐴𝐶𝐶 𝑅𝑒𝑣 = 𝛽 + 𝛽 𝑇𝑖𝑚𝑒 + 𝛽 𝐵𝑎𝑛𝑘𝑟𝑢𝑝𝑡𝑐𝑦 + 𝛽 𝑅𝑒𝑠𝑢𝑙𝑡 + 𝛽 𝐴𝑠𝑠𝑒𝑡𝑠 + 𝜀

These two regressions will be used to see if the independent variables: β Result, β Bankruptcy, and β Time have a significant relation after controlling for β Assets with the dependent variable, the level of conservatism, and the level of provision. This is done by two different samples that will be explained in a paragraph about the sample. The control variable is added to the regression in order to control for the size of the companies. Due to the nature of the sample, there is a large discrepancy between the sizes of the companies, which is controlled for with the natural logarithm of the assets. For the second hypothesis, I will test if the allowance in companies increases by plotting the outcome of the t-test in a graph and a regression using both samples. This will indicate whether the full allowance increases, rather than only the allowance taken during one year. The last two hypotheses will be tested with a regression in order to see if there is a significant influence on the level of conservatism and provision from the level of bankruptcy and the outcome of the income statements. The measure of the different variables is as follows: variable β Time is the movement in time with years, variable β Bankruptcy is the percentage of bankruptcy in the Netherlands in a certain year, and variable β Result shows if a company posted a profit or loss during the year. Control variable β Assets is the natural logarithm of the assets of the company at the end of year.

(22)

- 22 - 4.2 Sample

For the research, a sample has been created of the change in allowance from the beginning to the end of the year for the listed companies in the Netherlands in order to test for

conservatism.

The sample of this research will use panel data to examine conservatism in the allowance of accounts receivable of Dutch companies to see whether this facilitates earnings management. The period of the sample is fiscal years 2007-2016. Assuming that almost all companies have a fiscal year ending in December, this will result in eight observations per company. The sample has been created by obtaining information from the annual reports of listed companies in the Netherlands, and I will also use information for the end of the fiscal year. The variables taken from the financial statements are:

1. Amount of gross accounts receivable before allowance at the end of the fiscal year; 2. Addition of the accounts receivable taken during the year in the income statement; 3. Write-off on the allowance of accounts receivable during the year;

4. Allowance of accounts receivable at the end of the fiscal year; 5. Net result of the companies according to the income statement; 6. Total amount of assets at the end of the fiscal year;

7. The level of bankruptcies compared with the total amount of companies during the year in the Netherlands.

The output of these variables results in quantities that will be used in this empirical research. Companies selected for the sample are stocks listed in the Netherlands. I will use companies from three indices for my sample: the first is the AEX (Amsterdam Exchange Index), which is the index with the 25 largest companies; second is the AMX (Amsterdam Midkap Index), which is comprised of companies that are smaller than the AEX companies; and finally, the AScX (Amsterdam Small cap Index) which includes the smallest listed companies in the Netherlands.

The selection of the companies for this research was made based on their activities. Several companies that are listed in the Netherlands have different types of receivables or have no receivables. Therefore, retailers were eliminated because they have no material accounts receivable, as were financial institutions and insurance companies, because their accounts receivable do not match our description of receivables on companies. Their receivables have a

(23)

- 23 -

maturity of more than one year, can be secured, and can be derived from consumer business. Excluding retailers and financial institutions resulted in a sample of 73 companies.

Some of these companies do not supply a specification of the allowance. These are placed in an alternative sample because I cannot assess the conservatism of these companies in detail. For this research, I have included listed companies based on their having freely available annual reports. These reports are more strictly regulated—therefore almost all of these companies should specify the allowance—and information for these companies is available from Bureau Van Dijk (an external data supplier). In the past, few researchers have considered this topic, therefore specific information was limited. Knowing that the information for accounts receivable allowances is not available from information suppliers, I created my own sample by manually obtaining variables 1 through 4 from the annual reports of the listed companies.

In order to verify the sample, copies are made from the values included in the sample. The fifth and six variables will be obtained from the database of Bureau van Dijk. This company is one of the only data suppliers for the Dutch financial market. Access to the Bureau van Dijk database was obtained through the University of Amsterdam, and will be used to obtain earnings, revenue, and balance sheet values. The last variable will be obtained from the Centraal Bureau voor de Statistiek, a government department in the Netherlands that supplies information on the economy. The period of observation was selected due to its widely available information.

With this information, a base sample and alternative sample were created. The base sample is the sample with the allowances and write-offs during the year and reflects a period from 2010-2016. This sample will be used to test the level of conservatism of the allowances. The alternative sample is the sample with the changes in the level of provision compared with the accounts receivable for the period from 2007-2016. The alternative sample will be used to test the fluctuations in provisioning and whether these are variables that influence the level of provision. The hypotheses will be tested with t-tests and regressions done in SPSS. In order to test the hypotheses, two dependent variables will be created, Con_H1 & Con_H2, as well as a dummy variable for earnings.

(24)

- 24 -

5. Results

Before proceeding with the formal results of the study, it is important to explain the layout of the results and the problems encountered after the sample was constructed. A control variable called Assets was used in the regressions to control for the size of the companies. The results of the regressions are displayed in two separate tables: one testing the hypotheses by the level of conservatism in the base sample (with dependent variable Con_H1), and one the level of provision in the alternative sample (with dependent variable Con_H2).

In total, 338 observations are obtained for the base sample and 633 observations are obtained for the alternative sample. First, in some cases, no write-off was taken during a year, which resulted in observations without a value. To maintain an adequate sample size, these values are changed to 5 because this value is in line with the study from Jackson & Lui (2010, p. 577). By changing the value accordingly, I maintain a certain level of comparison between the two studies. Second, the sample contained values of 𝐶𝑜𝑛_𝐻1 in the base sample that were too extreme. These extreme values in the base sample led to a less accurate measurement of conservatism. By graphing the data in a boxplot, all values above five were marked extreme. Therefore, all values have been winsorized to 5, which minimizes the number of outliers in the base sample. When the variables are not winsorized, the standard deviation of the sample is too extreme and leads to inaccurate test results. For example, a company that has a conservatism level (𝐶𝑜𝑛_𝐻1) of 115 (meaning the allowance in year zero is 115 times the write-off in year one) is far too extreme. In a case like this, there was most likely a problem with one large project or customer that led to this extreme level of conservatism.

In the base sample, no observations were removed, only winsorized. In the alternative sample, 612 of the 633 collected observations were analyzed. The descriptive statistics from both samples can be seen in Table I. They exclude observations representing values that were far too extreme for the sample. One example is Takeaway, which had a provision that represented a value of 70% of the accounts receivable. To begin with, the distribution of the sample was tested in order to see if normal parametric tests were possible. This led to the discovery that neither sample had a normal distribution. The sample was adjusted but failed to achieve a normal distribution. Therefore, other options such as non-parametric models or bootstrapping were considered. In order to obtain reliable information, bootstrapping, which has the benefit of allowing the use of normal parametric tests, was utilized.

(25)

- 25 - Table I Descriptive statistics

Base sample

N Minimum Maximum Mean Deviation Std. Median Percentile 25% Percentile 75%

Con_H1 338 0.00 5.00 1.98 1.66 1.31 0.80 2.85 Time 372 1.00 6.00 3.50 1.71 3.50 2.00 5.00 Bankruptcy 360 0.01 0.02 0.02 0.00 0.02 0.02 0.02 Profit 332 0.00 1.00 0.81 0.40 1.00 1.00 1.00 Loss 332 0.00 1.00 0.19 0.40 0.00 0.00 0.00 Assets 353 10.71 24.68 20.79 2.14 21.00 19.47 22.33 Valid N (listwise) 319 Alternative sample

N Minimum Maximum Mean Deviation Std. Median Percentile 25% Percentile 75%

Con_H2 612 0.00 0.38 0.05 0.06 0.04 0.02 0.06 Time 678 1.00 10.00 5.49 2.87 5.00 3.00 8.00 Bankruptcy 678 0.02 0.04 0.03 0.01 0.03 0.02 0.03 Profit 571 0.00 1.00 0.82 0.39 1.00 1.00 1.00 Loss 571 0.00 1.00 0.18 0.39 0.00 0.00 0.00 Assets 647 13.34 26.74 20.88 2.26 20.92 19.38 22.50 Valid N (listwise) 543 5.1 Bootstrapping method

Given that the samples used in the research do not have a normal distribution, I searched for a way to conduct the parametric tests. The most compelling way to do this was with bootstrapping, which was easy to use and is implemented in SPSS. Bootstrapping is a tool for making statistical inferences when standard parametric assumptions are questionable. There are several books written about the theory behind bootstrap, for example Hall (1992), Van der Waart (2000) and Lahiri (2000).

Bootstrapping is a form of sampling from the data that tries to capture features in the distribution which the over simplified normal approximation cannot do. Hayes (2009, p 412) stated the process of bootstrapping in his research as follows: “Bootstrap generates an empirical representation of the sampling distribution of the variable by treating the obtained sample of size n as a representation of the population in miniature, one that is repeatedly resampled during analysis as a means of mimicking the original sampling process. The resampling of the sample is conducted with replacement, so that a new sample of size n is built by sampling cases from the original sample but allowing any case once drawn to be thrown back to be redrawn as the

(26)

- 26 -

resample of size n is constructed. Once a resample is constructed, a and b are estimated and this resampled data set and the product of the path coefficients, recorded. This process is repeated for a total of k times, where k is some enormous number (typically at one thousand; in my research five thousand will be used). Upon completion, the analyst will have k estimates of the indirect effect, the distribution of which functions as an empirical approximation of the sampling distribution of the variable when taking a sample of size n from the original population."

5.2 Hypothesis 1

This hypothesis predicts that the companies in the Netherlands are conservative in taking an allowance for accounts receivable. In order to test if companies in the Netherlands are conservative, I will test if the values for 𝐶𝑜𝑛_𝐻1 differ significantly from the benchmark of 1. Due to large extreme values, the base sample was adjusted to a maximum value of 5 in order to maintain some level of comparison with the study of Jackson & Lui (2010). Before conducting the one sample t-test, I first visually inspect the findings of the means in the different years. Table II shows the means found in the base sample. By looking at the table, we can conclude that the values found exceed the benchmark of 1 in every year.

Table II

Comparing means Con_H1a

Time N Mean Standaard Deviation

2010 56 1.98 1.86 2011 55 1.85 1.64 2012 57 1.90 1.70 2013 59 1.98 1.60 2014 55 1.94 1.57 2015 Total 56 2.24 1.61 338 1.98 1.66

a. This table shows the mean of the base sample for the years 2010-2015. During the years, the number of observations fluctuate, which can come from bankruptcies, newly listed companies, or missing information.

I have evaluated the means by comparing them to the benchmark of 1. However, in order to test whether the values of 𝐶𝑜𝑛_𝐻1 exceed 1, I use the t-test after bootstrapping. This test indicates that values of Con_H1 significantly differ from the benchmark of 1 (ρ-value < 0.05), see Table III. The findings are in line with prior research on conservatism in studies like Watts (2003a),

(27)

- 27 -

Jackson & Lui (2003) and Basu (1997). The null hypothesis for this test was that the means in my sample did not differ from 1, and the alternative hypothesis indicated that there was a significant difference between the means in the sample compared with 1. Based on the significant levels, we reject the 𝐻 : 𝜇 = 𝜇 and accept 𝐻 : 𝜇 ≠ 𝜇 . In other words, with a 95% confidence level I can say that the values of Con_H1 significantly exceed the benchmark of 1, meaning that companies in the Netherlands are conservative.

Table III

Test of conservatism with the one sample t-test Con_H1

Time Mean Difference Bootstrapa Bias Std. Error Sig. (2-tailed) BCa 95% Confidence Interval Lower Upper 2010 0.98 0.00 0.25 0.00* 0.54 1.44 2011 0.85 0.00 0.22 0.00* 0.42 1.28 2012 0.90 0.00 0.22 0.00* 0.48 1.33 2013 0.98 0.01 0.21 0.00* 0.58 1.42 2014 0.94 0.00 0.21 0.00* 0.56 1.36 2015 1.24 0.00 0.22 0.00* 0.85 1.63

a. Unless otherwise noted, bootstrap results are based on 5,000 bootstrap samples b. * value found is significant based on a confidence level of 95%

c. This table shows the one sample t-test after bootstrap. The test variable is Con_H1, which tells us if a company is conservative. The test value in this one sample t-test is 1. The results are displayed per year and as one sample.

5.3 Hypothesis 2

This hypothesis predicts that companies become more conservative over time. Before the regressions are made, the mean of the years are plotted in a line graph to visually inspect the movement over the years. Figure I shows that there is an upward trend noticeable in Con_H1, but not in Con_H2. In order to verify whether there is a significant effect from Time on Con_H1 and Con_H2, I will test this variable with a regression.

(28)

- 28 - Figure Ia

Mean values of Con_H1 & Con_H2

Equation (1) Con_H1 = (𝐴𝑙𝑙𝑜𝑤 𝑎𝑑𝑑 /𝑊𝑂 ) & Equation (2) Con_H2 = (𝐴𝑙𝑙𝑜𝑤 /𝐴𝐶𝐶 𝑅𝑒𝑣 )

a. This figure graphs the movement of Con_H1 and Con_H2 over time. The information from Con_H1 is from 2010-2015 and the information from Con_H2 is from 2007-2016. The Y-variable is the years.

Table IV shows the results of the regression of time on Con_H1 and Con_H2 with Assets as a control variable. The coefficients for 𝑇𝑖𝑚𝑒 in both samples are not significant (ρ-value > 0.05), but for the base sample, there is a positive coefficient (B = 0.05), and for the alternative sample, a coefficient of zero (B = 0.00). The control variable, which controls for the size of the companies, is significant in the base sample with a ρ-value < 0.05. This means that the size of the company has a significant influence on the outcome of the level of conservatism. The main result of this hypothesis indicates that time has no significant impact on the level of conservatism and the level of provision in either the base sample, which reflects the period from 2010-2015, or the alternative sample, which reflects the period between 2007-2016. In the base sample, the evidence shows that 𝑇𝑖𝑚𝑒 has a positive effect on the level of conservatism, but that this is not significant. The findings in this sample are not in line with prior studies from Watts (2003a) and Jackson & Lui (2010) and will be discussed in the conclusion.

0,045 0,047 0,049 0,051 0,053 0,055 0,057 0,059 0,061 1,5 1,6 1,7 1,8 1,9 2 2,1 2,2 2,3 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 Con_H1 Con_H2

(29)

- 29 - Table IV

Regressions of annual measures of conservatism and provision on time Base sampleb Model B Bootstrapa Bias Std. Error Sig. (2-tailed) BCa 95% Confidence Interval Lower Upper 1 Intercept 5.04 0.01 0.96 0.00 3.21 6.98 Assets -0.16 0.00 0.04 0.00 -0.24 -0.07 Time 0.05 0.00 0.05 0.33 -0.06 0.15 Alternative sampleb Model B Bootstrapa Bias Std. Error Sig. (2-tailed) BCa 95% Confidence Interval Lower Upper 1 Intercept 0.06 0.00 0.03 0.04 0.01 0.11 Assets -0.00 0.00 0.00 0.97 0.00 0.00 Time 0.00 0.00 0.00 0.46 0.00 0.00

a. Unless otherwise noted, bootstrap results are based on 5,000 bootstrap samples

b. This table shows the result of the regression on time with control variable, Assets. In the table of the base sample variable, Con_H1 is the dependent variable in the alternative sample variable; Con_H2 is the dependent variable. This table will test whether the level of conservatism (Con_H1) and provision (Con_H2) has become more conservative over time. Column Sig. (2-tailed) indicates whether the variable time has a significant influence on the independent variable (conservatism / provision). Column B indicates what the direction of the influence is, i.e., if time increases with one, what will happen to value of conservatism/provision.

5.4 Hypothesis 3

This hypothesis predicts that the level of bankruptcy in the Netherlands has a significant negative influence on the level of conservatism and level of provision. Meaning that when the level of bankruptcy increases the level of conservatism and provision decreases. To analyze whether bankruptcy has a significant influence on the level of conservatism and on the level of provision, I made a regression using bankruptcy and assets as a control variable. The outcome of the regression can be found in Table V. Based on the outcome of the regression, there is no significant negative influence from bankruptcy on the level of conservatism and the level of provision because the ρ-value is > 0.05 in both samples after controlling for the size of the company. The variable that controls for the size of the companies is significant in the base sample with a ρ-value < 0.05. In the alternative sample, the coefficient is positive (B = 0.27);

(30)

- 30 -

for the base sample the coefficient is negative (B= -20.94). The coefficient for the base sample seems enormous, but we need to keep in mind that a one-point increase in bankruptcy is a step of almost 100% difference (variable is in percent form). Decreasing it by 100 means that with every percentage increase in bankruptcy, the level of conservatism decreases by 0.2094. The outcome of the regression will be discussed in the conclusion, but it seems to be in line with my prediction that bankruptcy has no significant influence on the outcome of the level of conservatism and level of provision.

Table V

Regressions of annual measures of conservatism and provision on bankruptcy Base sampleb Model B Bootstrapa Bias Std. Error Sig. (2-tailed) BCa 95% Confidence Interval Lower Upper 1 Intercept 5.68 -0.01 1.11 0.00 3.53 7.86 Assets -0.16 0.00 0.04 0.00 -0.25 -0.07 Bankruptcy -20.94 0.13 30.45 0.48 -80.28 40.35 Alternative sampleb Model B Bootstrapa Bias Std. Error Sig. (2-tailed) BCa 95% Confidence Interval Lower Upper 1 Intercept 0.05 0.00 0.02 0.05 0.00 0.09 Assets -0.00 0.00 0.00 0.96 0.00 0.00 Bankruptcy 0.27 0.01 0.35 0.45 -0.39 0.98

a. Unless otherwise noted, bootstrap results are based on 5,000 bootstrap samples.

b. This table shows the result of the regression on bankruptcy with the control variable, Assets. In the table of the base sample variable, Con_H1 is the dependent variable; in the alternative sample, variable Con_H2 is the dependent variable. This table will test whether the level of conservatism (Con_H1) and provision (Con_H2) are influenced by the level of bankruptcy. Bankruptcy is a variable that represents the percentage of companies that went bankrupt during that year, compared with the total amount of companies during that same year. Column Sig. (2-tailed) indicates whether the variable bankruptcy has a significant influence on the movement independent variable (conservatism / provision). Column B indicates what the direction of the influence is, i.e., if bankruptcy increases with one, what will happen to the value of conservatism/provision.

(31)

- 31 - 5.5 Hypothesis 4

This hypothesis predicts that there is a significant negative difference between the levels of conservatism and provision for companies that post a profit vs. a loss. This means that companies that post a loss have a significantly lower level of conservatism and provision. In order to get the most accurate results in this regression, all the observations of earnings are brought back to 1 for a profit and -1 for a loss. From this I have created a dummy variable that will be used in the regression. The reference variable will be profit, and the variable with loss will be used in the regression. Therefore, the result will be interpreted as the difference between companies that post a profit and companies that post a loss. Just as in the previous regression, the control variable will be Assets, which controls for the size of the company. The outcome of the regression is presented in Table VI which indicates neither dependent variable is significant based on the ρ-value > 0.05. The only significant variable is Assets, which reflects the size of the company. The base sample has a negative coefficient (B = -0.30), and in the alternative sample, there is an opposite noticeable effect. Based on the regression, there is no significant difference in the level of conservatism and level of provision among

companies that make a profit vs. those that post a loss. This does not support my prediction, and means that I cannot find evidence of earnings management. The result will be further discussed in the conclusion.

Table VI

Regressions of annual measures of conservatism and provision on outcome of the income Base sampleb Model B Bootstrapa Bias Std. Error Sig. (2-tailed) BCa 95% Confidence Interval Lower Upper 1 Intercept 4.97 0.02 1.05 0.00 2.86 7.09 Assets -0.14 0.00 0.05 0.00 -0.24 -0.05 Loss -0.30 -0.01 0.23 0.17 -0.73 0.12 Alternative sampleb Model B Bootstrapa Bias Std. Error Sig. (2-tailed) BCa 95% Confidence Interval Lower Upper 1 Intercept 0.06 0.00 0.02 0.02 0.01 0.10 Assets 0.00 0.00 0.00 0.75 0.00 0.00 Loss 0.01 0.00 0.01 0.11 0.00 0.03