Earnings management in the Dutch long–term care sector : before and after the 2015 long–term care reform

(1)

Earnings management in the Dutch long–term care sector

before and after the 2015 long–term care reform

Name: Arjan Arends

Student number: 10997539

Thesis supervisor: ir. drs. A.C.M. de Bakker

Date: 20 January 2018

Word count: 11,351

MSc Accountancy & Control, specialization Accountancy Faculty of Economics and Business, University of Amsterdam

(2)

2

Statement of Originality

This document is written by student Arjan Arends who declares to take full responsibility for the contents of this document.

I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it.

The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.

(3)

3

Abstract

This research analyses the influence of the 2015 long-term care reform on (accrual based) earnings management within the long-term care sector. We hypothesized that the long-term care reform and relating budget cut leads to an increase in uncertainty at managers which in turn lead to an increase in earnings management to maintain the current financing level from the municipalities after the reform.

Aside from the effect of the long-term care reform on the sector as a whole, we have also researched the influence on the intramural- (inpatient), extramural- (outpatient) and mixed care subsectors as well as the effect of firm size on this relationship. Regarding the effect of firm size we hypothesized that, based on the political cost hypothesis, there is a positive relation between firm size and the use of earnings management.

We have investigated the years 2013 up to and including 2016 which includes two years of data prior to the long-term care reform and two years subsequent to the long-term care reform. Our sample includes 1,819 observations of long-term care institutions in the Netherlands for the investigated period.

We have found evidence that, contrary to our expectation, there is a significant negative relationship between earnings management and firm size in the long term-care sector. Further research is needed to comprehensively interpret these results. Furthermore the results from this research indicate that there is no significant change in earnings management prior- and after the long-term care reform for both the sector as a whole or in one of the individual sub-sectors.

(4)

4

1. Introduction

1.1 Background

As the Netherlands faces the challenge of an ageing population and increasing costs for long–term care, it has introduced major reforms to its long–term care system in 2015. Long– term care prior to 2015 was arranged in the Exceptional Medical Expenses Act (Algemene Wet Bijzondere Ziektekosten (AWBZ)), an Act that came into force in 1968 to cover exceptional (uninsurable) health care costs. The AWBZ has since then been extended with different types of health care that do not necessarily meet that requirement, including long–term care, which has led to a situation in the Netherlands were increasingly more individuals were living in long–term care facilities. These developments are reflected in a huge increase of the budget required to keep the AWBZ in effect.

As a result a reform of the long–term care in the Netherlands came in to effect per the first of January 2015, by which the AWBZ has been repealed. The extramural (outpatient) long–term care entitlements of the AWBZ have been primarily transferred to the Social Support Act (Wet Maatschappelijke Ondersteuning 2015 (WMO 2015)) and the Health Insurance Act (Zorgverzekeringswet (Zvw)). The intramural (inpatient) long–term care entitlements, for people with severe long–term care needs, such as vulnerable elderly people and people with severe chronic illness or disability, have been transferred to the new Long–term Care Act (Wet langdurige zorg (Wlz)).

The intention of this reform with regard to long–term care is to focus on people’s capabilities rather than their disabilities and keeping them in their own homes longer, with support of their social network, while still providing a safety net for those individuals who are not able to support themselves. By reducing the number of individuals whom are eligible for intramural long–term care, the Dutch authorities hope to reduce the costs of the long–term care system to a sustainable level.

Where the AWBZ was previously executed by the Dutch care authority (Nederlandse Zorgautoriteit (Nza)) on a national level, the WMO 2015 is executed by the municipalities. Although the right of care is laid down in the WMO 2015, the municipalities can decide how they further implement and execute this act. This uncertainty together with a reduction of the available budget by approximately 25% could lead to a situation where managers of long–term care facilities feel the need to, instead of austerity measures such as cutting costs, manage

(7)

7 earnings to maintain the current financing level from the municipalities after the reform (Anheier et al. 1997).

1.2 Motivation and contribution

There is a lot of prior literature regarding earnings management; however the literature is predominantly focused on the profit sector. There is substantially less prior research done regarding earnings management in the non–profit sector. Specifically for the healthcare sector, the prior literature is primarily focused on hospitals. For example Leone et al (2005) finds evidence for earnings management at nonprofit hospitals and Eldenburg et al. (2011) find that in non–profit hospitals, managers have incentives to avoid high levels of net income to decrease the probability of scrutiny by government and other stakeholders. This research aims to contribute to the literature on earnings management in the healthcare sector by expanding the prior literature to the long–term care sector.

1.3 Research goals and research question

The aim of this research is to investigate how the 2015 reform of the long–term care sector affected earnings management in the Dutch long–term care sector.

The central research question is

“Has the 2015 reform of the Dutch long–term care sector led to an increase in earnings management?”

1.4 Structure

This research is organized as follows. In chapter two we will discuss prior literature on earnings management including the hypothesis development. The theoretical framework for answering the central research question will be presented in chapter two.

Chapter three will discuss the research methodology and an explanation of the research design and empirical model used. The data used is described in chapter four. In the fifth chapter we will present our empirical results and subsequently, we will discuss our conclusion and the limitations of our research in chapter six.

(8)

8

2. Theoretical framework

In this chapter we will discuss the existing theories and literature that define earnings management in general, the underlying economic theories that support the existence of earnings management and the incentives and strategies for earnings management in the Dutch long–term care sector. After outlining the phenomenon of earnings management, we will discuss the methods to empirically research the existence and occurrence of earnings management. We will conclude this chapter with the development of our hypotheses.

2.1 Defining earnings management

Even though the extent to which executives alter reported earnings for their own benefit is central to a large amount of studies in the field of accounting research, academics do not have consensus over the definition of earnings management (Messod, 2001). In this paragraph we will discuss the most predominant definitions of earnings management.

(1) Earnings management is “…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).” Schipper (1989, p.92).

(2) Earnings management is “…the use of accounting methods and the change of these methods towards the

accounting policies that are necessary for the strategy of creating financial reports.” Hoogendoorn (1997).

(3) Earnings management is “…the accounting policy choice to achieve some specific objective.” Scott (2003)

(4) ’’Earnings management occurs when managers use judgment in financial reporting and in structuring

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

Healy et al. (1999, p.368).

If we compare these four definitions, we note that all four definitions act within the constraints of management’s role in the financial reporting process with a focus on the ability of managers to intervene in this process. The definitions of Hoogendoorn (1997) and Healy (1999) focus predominantly on accounting choices (accrual based earnings management) with the distinction that Healy (1999) explicitly notes the intent to mislead stakeholder to influence contractual outcomes. Considering fraud is defined as “one or more intentional acts designed to deceive

other persons and cause them financial loss.” (National Association of Certified Fraud Examiners, 1993,

p.6). Earnings management potentially also includes elements of fraud under the definition of Healy at al. (1999).

(9)

9 The definition of Schipper (1989) is broader and also allows earnings management to occur through real activities (real activities earnings management).

This research focuses on the contractual engagement between long-term care institutions and municipalities and the choices managers of these institutions make to influence this relationship. We therefore consider the definition of Healy et al. (1999) to fit best within the confines of our research.

2.2 Agency theory

Economic theories aim to provide a coherent and systematic framework for researching, understanding and developing various accounting practices. The economic phenomena of earnings management is mainly supported by the agency theory.

Agency theory was developed in the 1960’s and early 1970’s (e.g., Arrow, 1971; Wilson, 1968). It describes the relationship between two parties in which one party (the principle) delegates responsibilities to the second party (the agent). Agency theory can be defined as follows (Jensen et al., 1976):

Agency theory is “a contract under which one or more (principals) engage another person (the agent) to

perform some service on their behalf which involves delegating some decision making authority to the agent’’

According to the agency theory we can identify two main problems in this relationship (Eisenhardt, 1989):

(1) The goals of the principal and agent may not be aligned; and

(2) It is difficult and expensive for the principal to monitor the agent as a result of information asymmetry.

In essence, agency theory indicates that, due to information asymmetry between the agent and the principal, agents are able to use judgment in financial reporting to manage earnings in such a way that it is not noticeable by the principal. This risk increases if there is a misalignment of interests between the agent and the principal (Deegan et al, 2006; Eisenhardt, 1989).

To minimize these problems, Eisenhardt (1989) suggests that the contract between the two parties should be as efficient as possible. To achieve this, Eisenhardt (1989) makes a distinction between behavior–oriented contracts (e.g., salary, hierarchical governance) and outcome oriented contracts (e.g., stock options, market governance).

(10)

10 Within the confines of this research the governmental bodies which largely fund the institutions through subsidies are the principals and the management of the long–term care institutions acts as the agents: The municipalities have the legal obligation to provide care to their citizens and have contractually delegated these obligations to various institutions. The contract between these parties can be classified as an outcome oriented contract as these contracts typically focus on the quantity and level of care to be provided by the institution and the monetary compensation to be received in return (e.g., NZa, 2015; Gemeente Zwolle, 2017).

2.3 Incentives for earnings management

Although there is no prior research done regarding the incentives for earnings management specifically in the long-term care sector, there is, although limited, prior research regarding earnings management in the not–for–profit sector in a broader sense. Some incentives for earnings management in the for–profit sector are applicable to the not–for–profit sector as well (Wilcox, 2002). Based on the available body of knowledge we have identified the following incentives for earnings management in the Dutch long-term care sector:

(1) Contracting incentive; and the (2) Reputation incentive.

2.3.1 Contracting incentive

Healy et al. (1998, p.18) note that “accounting data is used to help monitor and regulate the contractual

relations between many of the firm’s stakeholders. Explicit and implicit management compensation contracts are used to align the incentives of management and external stakeholders. Lending contracts are written to limit managers’ actions that benefit the firm’s stockholders at the expense of its creditors”. In the long-term care

sector contracts are put in place to align the incentives of the municipalities and the long-term healthcare institutions. Previous research by Jegers (2010, p.408) indicates that the contracting incentive is applicable to not-for-profit organizations as managers have the incentive to gain or retain government subsidies.

The contracting incentive can be linked to the political cost hypothesis (Watts et al., 1990) of the positive accounting theory. The positive accounting theory is based on two assumptions (Deegan et al., 2006):

(1) It is assumes that all actions by individuals are driven by self-interest; and (2) it is assumes that individuals will always act in an opportunistic way.

(11)

11 The political cost hypothesis states that there is a positive relationship between the level of profit and political attention. Research by Zimmerman (1983) also indicates there is a positive relationship between firm size and political attention. Applying this to long-term care institutions gives managers an incentive to prevent high profits to be reported as high profits would lead to increases political attention which could subsequently lead to a downward adjustment of government subsidies for the subsequent year (i.e.; high profits interfere with the managers ability to retain government subsidies).

Long-term care institutions can either implement austerity measures such as cutting costs to meet the goal of the municipalities (high–quality long–term care for its citizens at a reduced cost) or they can strategically manipulate accounting choices to affect earnings, signaling to the municipalities that they need the current level of funding to provide the quantity and quality level of care they have contracted.

2.3.2 Reputation incentive

According to Leone et al. (1999) managers of non-profit hospitals have an incentive to manage earnings based on reputation concerns. Executives in the healthcare sector typically do not have bonus plans in place that are based on financial performance, however an executive of a larger company will usually earn more than an executive of a smaller company. From this perspective, an executive has an incentive to manage earnings to signal to the market he is a good manager and thereby furthering his career by transitioning to a larger institution (Leone et al., 1999). This is substantiated by the empirical research of Leone et al. (1999) that reports a 4% higher turnover rate of executives in the not-for-profit sector in comparison to the for-profit sector.

Although that research was performed in the hospital (cure) sector, its results are applicable to the care sector as well. In the Netherlands the maximum income of an executive in the healthcare sector is limited by the Standards for Remuneration Act (Dutch: Wet Normering Topinkomens, WNT). This Act is applicable for both the care and cure sector and includes a maximum remuneration that is partially dependent on the size of the institution the executive is working for (Ministry of the Interior and Kingdom Relations, 2015).

2.4 Strategies for earnings management

Leone et al. (2005) describe that organizations can either manage earnings through accruals (accrual based earnings management) or by influencing (the timing of) real transactions (real activities earnings management). Within the Dutch long-term care sector it is difficult to

(12)

12 influence actual expenses. These institutions do not have material stock balances and there is a relatively low level of purchasing of goods. The income statement generally comprises of revenues which have been contracted, depreciation charges and payroll costs.

Earnings can however be managed with the use of provisions and accruals. Under Dutch GAAP provisions can be formed for cost-equalization purposes (e.g.; maintenance provisions) and for legally enforceable obligations, for which the amount can reliably be estimated but for which the moment of settlement is uncertain. Provisions and accruals do not directly lead to an outflow of cash but any changes are still accounted for in the income statement.

Prior research describes different ways for a manager to apply earnings management of which the following are the most relevant to the Dutch long-term care sector:

(1) Income smoothing (2) Big bath accounting (3) ‘Cookie jar’ reserves

2.4.1 Income smoothing

Albrecht et al. (1990, p. 713) defines income smoothing as: “the deliberate dampening of

fluctuations about some level of earnings in which is considered to be normal for the firm”.

Managers can use income smoothing to reduce fluctuations in the result for the year to reflect a more stable image in the financial reporting of an organization. Income smoothing can take on the form of earnings minimization or earnings maximization (Healy et al, 1999). Managers will apply income minimization in periods where income is above optimum and will apply income maximization in periods where income is below optimum (Fudenberg et al,1995). Leone et al. (2005) conclude that, for non-profit hospitals, earnings are managed towards breakeven through the use of discretionary accruals.

Based on the reputation incentive, executives are driven to manage earnings from a reputation perspective. Considering healthcare institutions are not-for-profit and the career opportunities of the executive are positively related to his track record, the executive has an incentive to smooth income about zero to reflect stability (Leone et al., 1999). From the perspective of the contracting incentive, income smoothing is an instrument to prevent reporting large profits to retain the current level of government subsidies.

(13)

13

2.4.2 Big bath accounting

Big bath accounting can be described as large artificial profit reducing write-offs or accruals in income statements (Healy, 1985; Watts et al., 1990).

Based on the research by Mohanram (2003), earnings management includes both upwards and downwards manipulation. When a firm performs below their target, there is an incentive to artificially make thinks look even worse than they actually are. Mohanram (2003) explains this by indicating that if a firm performs below target, the costs of being even worse are typically minimal.

By forming unrealistic provisions and accruals, management can artificially worsen the earnings in a certain year and enhance earnings of the subsequent years. Leone et al. (2005) have found evidence of this phenomenon in non-profit hospitals in the United Sated of America particularly in relation to a change of management. The incentive for the new management to apply big bath accounting is related the reputation incentive; the losses in the current year will be related to the previous management by the principal while the future profits will be accredited to the new management. From this perspective this method could be applied in the Dutch long-term care sector as well.

2.4.3 ‘Cookie jar’ reserves

Cookie jar reserves are closely related to big bath accounting and earnings smoothing. Levitt (1998) notes that companies could use unrealistic assumptions while estimating provisions and accruals. These unnecessarily high accruals can be seen as cookie jars to be filled in goods times and to be released when results are below par.

Cookie jars reserves relate closely to the contracting incentive as it is an earnings management method that enables management to report earnings around break-even to retain their current level of government subsidies.

2.5 Measuring earnings management

Prior research has revealed several methods to empirically research accrual based earnings management. To understand these methods it is important to have a basic understanding of accrual accounting. The nature of an accrual is such that the effect on the income statement is recorded in a different financial period than the actual cash flow. Non-discretionary accruals are part of the normal business activities used to better match income and expenses to a financial period as to increase the relevance of the earnings figure (Bauwhede, 2003, p. 198). Discretionary

(14)

14 accruals are influenced by managers to apply earnings management (Bauwhede, 2003, p. 198). According to Dechow et al. (1995) who reviewed accrual based models for detecting accrual based earnings management, the following models can detect accrual based earnings management:

(1) Healy model; (2) DeAngelo model; (3) (modified) Jones model; (4) Industry model.

Through the use of discretionary accruals the management has the ability to misinform its stakeholders regarding the performance of the company. These models therefore use discretionary accruals as a proxy for measuring earnings management. The method of how these models do this differs. The following subparagraphs will discuss each of these models’ measurement approach in detail.

2.5.1 Healy model

The first attempt to measure earnings management by using non-discretionary accruals as a proxy for earnings management is made by Healy (1985). The model used in this model is referred to as the Healy model. Healy (1985) states that systematic earnings management occurs every period and that accruals can be estimated by the difference between the reported earnings and the operating cash flow. The model measures the non-discretionary part of accruals by comparing the accruals in a certain year to the average accruals in previous years and assumes that the mean of the non-discretionary accruals is 0 (i.e.; if there is no earnings management, there should be no difference between the total accruals between the years). Cash flow is calculated as the operating cash flow minus changes in accruals for inventory, receivables and payables (Healy 1985).

The Healy model uses the following formula to calculate the total accruals for the year:

(1)

Where:

= Total Accruals in year t

= Change in current assets in year t = Change in current liabilities in year t

(15)

15 = Change in short term debt included in current liabilities in year t

= Depreciation and amortization in year t

= 1,2, …. t is a year subscript for years included in the estimation period

Subsequently, the non-discretionary part of the total accruals can be calculated by comparing the total accruals of a certain year to the average accruals in prior years (Dechow, 2015). To do this the Healy model uses the mean of total accruals ( ) scaled by lagged total assets ( ) from the estimation period as the measure of nondiscretionary accruals:

∑

(2)

Where:

= Estimated non-discretionary accruals in year t = Total accruals in year t

= Lagged total assets

= The number of years in the estimation period = A year subscript indicating a year in the event period

The discretionary accruals can subsequently be calculated as follows:

(3)

Where:

= Discretionary accruals in year t

= Estimated non-discretionary accruals in year t = Total accruals scaled by lagged assets in year t

A flaw in the Healy model is that it assumes that non-discretionary accruals constant over time (Dechow, 1995) which according to research performed by Kaplan (1985) is not correct as Kaplan (1985) concludes that non-discretionary accruals are influenced by changes in economic circumstances.

(16)

16

2.5.2 DeAngelo model

According to Dechow (1995), the DeAngelo model (DeAngelo, 1986) is an addition to the research done by Healy (1985). DeAngelo (1986) adopts the measurement approach for total accruals as used by Healy (1985) but her research includes the effect of prior year accruals on the current year accruals by measuring the nondiscretionary accruals as the difference between the total accruals in the current year and the total accruals in the prior year. The total assets and the discretionary accruals are calculated using the same formulas as in the Healy model. In effect the DeAngelo model measures the non-discretionary accruals as the last period’s total accruals ( ) scaled by the lagged total assets ( ):

(4)

Where:

= Estimated non-discretionary accruals in year t = Total accruals in year t-1

= Lagged total assets

= A year subscript indicating a year in the event period

Like the Healy model, the DeAngelo model assumes that if there is no earnings management, there should be no difference between the total accruals between the years.

The DeAngelo model inherits the flaw that non-discretionary accruals are constant over time from the Healy model (Dechow, 1995).

2.5.3 (Modified) Jones model

In contrast to Healy (1985) and DeAngelo (1986), the research published in 1991 by Jones rejects the assumption that non-discretionary accruals do not change over time. Jones (1991) controls for the effect of economic circumstances on a firm’s non-discretionary accruals as mentioned by Kaplan (1985) in what is referred to as the Jones model.

The Jones model assumes that fluctuations in revenue will lead to fluctuations in accruals and that depreciation of property, plant and equipment will decrease accruals and uses de variance in revenue and property, plant and equipment as independent variables to predict the non-discretionary part of accruals (Jones, 1991; Dechow, 1995):

(17)

17 ̂ ( ) ̂ ( ) ̂ ( ) (5) Where:

= Estimated non-discretionary accruals in year t = Revenues in year t less revenues in year t-1

= Gross property plant and equipment at the end of year t

= Total assets at the end of year t-1

= A year subscript indicating a year in the event period ̂ ̂ ̂ = Estimated year specific parameters

The estimates of the year-specific parameters are generated by means of an ordinary least squares regression (OLS) using the following formula:

( ) ( ) ( ) (6) Where:

= The year specific regression coefficients = Measurement error in year t

The Jones model operates under the underlying assumption that revenue is by definition non-discretionary (Dechow 1995) but according to Jones (1991) earnings management through the manipulation of revenues (e.g. credit sales) is possible.

Dechow (1995) modified the Jones model, in what is referred to as the modified jones model, to overcome this limitation by deducting the variance in receivables:

̂ ( ) ̂ ( ) ̂ (7)

The estimates of ̂ , ̂ , ̂ in the modified Jones model are obtained the same way as in the original Jones Model. The Jones model (Jones, 1991) and Modified Jones model (Dechow,

(18)

18 1995) both follow the measurement approach of the Healy model (Healy, 1985) for total assets (Dechow, 1995) and discretionary accruals.

2.5.4 Industry model

The Industry model as developed by Dechow et al. (1991), like the modified Jones model, is a variation on the Jones model. This model also operates under the assumption that the non-discretionary accruals change from period to period but controls for the effect of economic circumstances under the assumption that the determinants of non-discretionary accruals are the same for all firms within a certain industry (Dechow, 1995):

̂ ̂ (

) (8)

Where:

= Estimated non-discretionary accruals in year t

= Total accruals in year t

= Total assets at the end of year t-1

(

) = The median value of total accruals in year t for industry j scaled by lagged total assets

̂ ̂ = year specific parameters

This approach however has its disadvantages as the Industry-model as there may be correlation between discretionary accruals within a specific sector which is neglected by the Industry model (Dechow et al., 1995). Secondly, under the assumption that variation in non-discretionary accruals reflects the firms’ response to changing economic circumstances, the Industry Model is not able to extract all non-discretionary accruals from the discretionary accrual proxy as it only removes variation in non-discretionary accruals across firms (Dechow et al., 1995).

2.6 Hypotheses development

In subparagraph 2.3.1 we discussed the influence of the political cost hypothesis on the long-term care sector. We noted that there is an expected positive relationship between firm size and the level of profit and political attention.

This could be an incentive for managers to prevent reporting high profits as these may lead to lower subsidy income in consecutive years. Managers in the long-term care sector can apply

(19)

19 earnings management to manage earnings downwards to signal to the municipalities that they need the current level of funding to provide the quantity and quality level of care they have contracted.

Based on these assumptions we assume a positive relationship between firm size and earnings management. We have developed hypothesis 1 to test these assumptions:

H1 : Earnings management is positively related to the size of the long-term care institution

In chapter two we have defined a theoretical framework based on prior research regarding earnings management in the not–for–profit sector and have asserted that there are multiple incentives for managers of long-term care institutions to apply earnings management.

Most of these can be traced back to the agency theory. Within the confines of this research the governmental bodies which largely fund these institutions through subsidies are the principals and the management of the long–term care institutions acts as the agents.

The long-term care reform which took place in 2015 caused a shift from the national government to local municipalities as principal. This reform also caused further uncertainty for the agents as, although the right of care is laid down in the new legislation, the municipalities can decide how they further implement and execute this legislation.

This uncertainty together with a reduction of the available budget by approximately 25% could lead to a situation where managers of long–term care facilities feel the need to, instead of austerity measures such as cutting costs, manage earnings to maintain the current financing level from the municipalities after the reform (Anheier et al. 1997).

With our second and final hypothesis we will therefore attempt to answer the main research question by measuring the effect of the long-term care reform on earnings management.

H2 : Earnings management within the Dutch long–term care sector increases after the 2015 long–term care

reform

The long-term care sector in the Netherlands can be further divided in two subsectors, namely intramural (inpatient) long–term care and extramural (outpatient) long-term care. For our final hypothesis, we will elaborate on our second hypothesis by measuring the effect of the 2015 long-term care reform for both of these subsectors. As some institutions provide both intramural and extramural care we will also include a mixed category. Reference is made to subparagraph 4.1.2 for the opted method to distinguish institutions between these subsectors.

(20)

20

H3: Earnings management within the Dutch long–term care sector increases for both the intramural- and

(21)

21

3. Methodology

This research is a form of quantitative, positive research as it attempts to explain economic behavior by using statistical analysis to evaluate quantitative data. This research can be further classified as causal-comparative research as it tries to establish a cause and an effect relationship between the long-term care reform and earnings management.

3.1 Research design

We will first measure the non-discretionary accruals per firm per year by using the Modified Jones Model as described in subparagraph 2.5.3. The discretionary accruals per firm per year, which will be used as a proxy for earnings management, will be calculated by subtracting the non-discretionary accruals from the total accruals using the method described in subparagraph 2.5.1.

Based on our review in the previous chapter, we consider the Modified Jones Model the most proficient in detecting earnings management. This statement is backed by the research of Dechow et al. (2015), Guay et al (1996) and Stolowy et al. (2004). It has been proven that the Modified Jones model is able to identify discretionary accruals (Dechow et al., 2015; Guay et al., 1996) and the model has the best explanatory value with the least systematically errors (Stolowy et al., 2004).

Because the aim of this research is to focus on the effect of the long-term care reform, we will adopt a times-series approach to compares the use of non-discretionary accruals before and after the reform. Prior research by Bartov et al. (1993) indicates that a time-series approach can be used when applying the Modified Jones model.

3.1.1 Conceptual model

Based on our hypotheses as described in paragraph 2.6 and our research design in paragraph 3.1 we have developed the conceptual model as depicted in figure 3.1 on the next page.

(22)

22

Figure 3.1 – Conceptual model

3.1.2 Regression model for Hypothesis 1

As discussed in chapter two, prior research assumes that there is a positive relationship between firm-size and political attention (Davidson et al., 2005) which could be a motivation for management to apply accrual based earnings management.

Hypothesis 1 researches this effect on earnings management in the Dutch long-term care sector. Under H0 we assume that firm size has no effect on earnings management.

For testing hypothesis 1 we will use the absolute value of the discretionary accruals as dependent variable and firm size as independent variable. Firm size will be measured based on (the natural logarithm of) revenue as revenue is the dominant benchmark value within this industry. Based on these assumptions, we have developed the following regression model:

| | Where:

| | = Absolute value of the discretionary accruals = LN (total revenues) in year t

= OLS parameters.

= Measurement error in year t

For hypotheses 2 we will test the effect of the long-term care reform on earnings management with firm-size as control variable. Under H0 we assume that the long-term care

Earnings-management Reform Firm size Institution category H₁,H₂,H₃ H₃ H2, H3 H3

(23)

23 reform has no effect on earnings management. Furthermore we will use firm-size as a moderator.

As in hypothesis 1 we will use the absolute value of the discretionary accruals as proxy for earnings management. Firm size is also measured in line with our regression model for hypothesis 1.

Furthermore our analysis needs a dummy for the independent variable which indicates the period prior (2013-2014) and after (2015-2016) the long-term care reform. Additionally we’ve added an interaction term between firm size and the long-term care reform dummy to the model to measure if the relationship between earnings management and the long-term care reform is different for larger firms then for smaller firms. Based on these assumptions, we have developed the following regression model:

| | Where:

| | = Absolute value of the discretionary accruals scaled by firm size = Dummy variable (dREFORM=0 prior to the long-term care

reform, dREFORM=1 after the long-term care reform)

= LN (total revenues) in year t

= The interaction effect between firm size and the long-term care reform

, , = OLS parameters.

= Measurement error in year t

For hypotheses 3 we will test the effect of the long-term care reform on earnings management for both subsectors intramural- and extramural care. Similar to hypothesis 2, we assume that under H0 the long-term care reform has no effect on earnings management for both

subsectors. As with hypothesis 2, we will use firm-size as a moderator.

Firm size, the absolute value of the discretionary accruals as proxy for earnings management and the dummy variable which indicates the period prior (2013-2014) and after (2015-2016) the long-term care reform are operationalized in line with our regression model for hypothesis 2.

(24)

24 Additionally we need a dummy to indicate if an institution is an intramural- (inpatient) or extramural (outpatient) institution. As some institutions provide both intramural and extramural care we will also include a mixed category. Reference is made to subparagraph 4.1.2 for the opted method to distinguish institutions between the intramural, extramural and mixed care subsectors. Additionally we’ve added interaction terms between the long-term care reform dummy and the subcategory dummies as well as firm size to the model to measure the effect of these interactions on the relationship between earnings management and the long-term care reform.

Based on these assumptions, we have developed the following regression model: | |

Where:

| | = Absolute value of the discretionary accruals scaled by firm size = Dummy variable (dREFORM=0 prior to the long-term care

reform, dREFORM=1 after the long-term care reform) = Dummy variable (dEXTRA=1 extramural care institution,

dEXTRA=0 intramural- or mixed care institution)

dMIXED = Dummy variable (dMIXED=1 mixed care institution,

dMIXED=0 intramural- or extramural care institution)

= LN (total revenues) in year t

= The interaction effect between the long-term care reform and the extramural care subcategory

= The interaction effect between the long-term care reform and the mixed care subcategory

= The interaction effect between firm size and the long-term care reform

, , = OLS parameters.

(25)

25

4. Data

We will investigate the period from financial year 2013 up to and including 2016. This period has been chosen as it includes two years of data prior to the long-term care reform and two years subsequent to the long-term care reform. Our sample will include all long-term care institutions in the Netherlands for the investigated period.

4.1 Data Gathering

The financial data of Dutch long-term care institutions needed is readily available as a dataset in both Excel and SPSS format on a dedicated website of the Dutch Ministry of Healthcare1_{. These datasets include more than 3000 variables regarding financial data, quality and}

compliance of Dutch healthcare institutions. Not all variables are relevant for this research.

4.1.1 Available data

We have filtered the long-term care institutions from these datasets based on variable q0_010a#13 with title ‘Nursing, Care and Home Care’ (Dutch: Verpleging, Verzorging en Thuiszorg

(VVT)) which indicates if an institution provides long term care.

Although 2012 is not part of our investigative period, we will need data for this year to calculate the variables on which the modified Jones model depends as the modified Jones model uses the change in revenues and receivables in comparison to prior year. For this same reason we will filter out any record for which no record is available in the preceding year based on variable ‘ConcernCode’ in the dataset or if the entity has no revenue in the preceeding year based on variable jeu10800 with title ‘Sum of operating income’ (Dutch: Som der bedrijfsopbrengsten).

Table 4.1 shows an overview of observation per financial year for the period 2012 – 2016 both before and after elimination of records with no relevant value for the preceding year.

(26)

26

Table 4.1 - Observation per financial year with preceding year data

Financial year N (before) N (after)

2012 674 2013 856 435 2014 860 451 2015 896 464 2016 946 469 Total 3,558 1,819

As indicated in table 4.1, our total sample consists of 3,558 observations. This includes 674 observations from 2012. 2012 is not under observation for this research but the data is needed to calculate the variances in comparison to prior year for 2013. Excluding 2012, our sample consists of 2,884 observations prior to eliminating records with no relevant value for the preceding year.

We have subsequently eliminated 1,065 records from our total sample of 2,884 observations due to the lack of necessary information for the preceding year. Of these 1,065 records, 786 records are eliminated because revenue in the current- and preceding year is zero. We assumed that if no revenues are recorded, these institutions have no activities in those years and are therefore not relevant to our research.

The total 1,065 records which have been eliminated minus the 786 records with no activities leave 279 records which relate to entities with activities in one of the years in the period under review but no relevant information in the preceding year. For these 279 records, we have tested whether they are part of the same population as the remaining sample by performing a Mann-Whitney U test to tests if the distributions of the revenue of both groups are identical. The results are summarized in table 4.2.

Table 4.2 – Results Mann-Whitney U Test

Revenue

Mann-Whitney U 58,118

Asymp. Sig. (2-tailed) .000*** *** significant at 0.01 level

(27)

27 From table 4.2, it can be concluded that the revenues of the eliminated records statistically do not significantly differ from the remaining records. We can therefore assume that we can extrapolate the results of our statistical analysis in chapter five of the remaining records to the discarded records.

4.1.2 Categorization of the institutions

Subsequently, we have extracted the intramural (inpatient) revenues based on variable jeu82002 with title ‘Wlz: ZZPs and other achievements’ (Dutch: Wlz: ZZP's en andere prestaties). ZZP is a Dutch acronym for care package (Dutch: Zorgzwaartepakket). A care package indicates which care a patient is entitled to receive based on the Long-Term Care Act (Wlz). This care is by definition intramural (inpatient nursing and care) care in a healthcare facility. The extramural (outpatient home care) care revenues are extracted based on variable jeu82002 Zvw Home care (Dutch: thuiszorg en wijkverpleging).

For this research we have used a cutoff percentage of 70% to determine if an institution is an intramural, extramural or mixed care institution meaning that if an institution’s revenues for either intramural- or extramural care exceed 70% in comparison to total revenues, we will categorize the institution as either intramural or extramural. If neither intramural- nor extramural care revenues exceed 70% in comparison to total revenues, we will categorize the institution as a mixed care institution. In table 4.3 we have summarized our samples by category, divided by the period prior to the long-term care reform and after the long-term care reform.

Table 4.3 – Number of samples per category

Financial year Intramural Extramural Mixed Total

2013 103 49 283 435

2014 105 53 293 451

2015 107 55 302 464

2016 103 57 309 469

Total 418 214 1,187 1,819

To indicate these three categories in our research we have created a dummy variable for each category as indicated in subparagraph 3.1.2. As shown in table 4.3 our sample includes sufficient observations both prior to and after the long-term care reform.

(28)

28

4.1.3 Data for Modified Jones model

All variables needed to calculate the non-discretionary accruals based on the modified Jones model can be derived from the dataset. Table 4.4 indicates the mapping between the variables of the modified Jones model and the variables in the dataset.

Table 4.4 - Mapping of variables between dataset and modified Jones model

Modified Jones model Dataset variable

REV jeu10800 Sum of operating income

REC jeu12521 Sum of accounts receivable and other

receivables

PPE jeu12241 Carrying amount PP&E at year end

Dep jeu11000 Total amortization and depreciation for the year

CA jeu09200 Carrying amount of the current assets at

year-end

CL jeu12641 Carrying amount of the current liabilities at year-end

STD jeu12601 Short term debt included in current liabilities

Cash jeu09100 Cash & Cash equivalents at year-end

A jeu09300 Total assets at year-end

4.1.4 Other data

As discussed in chapter 3 our control variables are limited to firm-size which will be measured based on revenue as revenue is the dominant benchmark value within this industry. Revenue for each firm per year can be derived from the aforementioned dataset by using variable jeu10800, Sum of operating income. The natural logarithm of this variable is our variable SIZE as included in our regression models.

Furthermore we have manually created a fiscal year variable as the datasets are available per fiscal year but do not include a variable for fiscal year. Based on the fiscal year, we have subsequently created a dummy variable which indicates the period prior (fiscal year 2013 and 2014) and after (fiscal year 2015 and 2016) the long-term care reform. This variable is labeled

(29)

29

4.2 Estimating the discretionary accruals

As previously indicated in paragraph 2.5, the discretionary accruals are calculated by subtracting the non-discretionary accruals from the total accruals. We will therefore first calculate the total accruals using formula 1. Subsequently we will estimate the non-discretionary accruals using the Modified Jones model.

4.2.1 Estimation of total accruals

We have calculated the total accruals using formula 1 as described in subparagraph 2.5.1. Table 4.5 includes the descriptive of the total accruals, scaled by the lagged total assets.

Table 4.5 – descriptive statistics of total accruals (scaled by lagged total assets)

Financial year N Minimum Maximum Mean Std.

Deviation 2013 435 -6.082 8.242 -0.034 0.605 2014 451 -2.285 4.466 -0.079 0.345 2015 464 -10.622 2.516 -0.011 0.613 2016 469 -1.604 1.991 -0.045 0.309 Total 1,819

From table 4.5 we can derive that the mean of total accruals has a negative value for each year. This can be explained by the effect of the amortization and depreciation expense in the total accrual formula; amortization and depreciation expenses are in effect negative accruals (Ronen & Yaari, 2008).

4.2.2 Estimation of Non-discretionary accruals

We subsequently calculated the non-discretionary accruals using the Modified Jones model as described in subparagraph 2.5.3. The estimates of ( ̂ , ̂ , ̂ ) in the Modified Jones model are estimated by means of an ordinary least squares regression. An overview of the year-specific estimations is included in table 4.6.

(30)

30

Table 4.6 – Estimated year specific parameters

Financial year N ̂ ̂ ̂ 2013 435 0.191 0.059 0.025 2014 451 0.127 0.049 0.034 2015 464 0.164 0.011 0.069 2016 469 0.142 0.017 0.012 Total 1,819 4.3 Descriptive statistics

In this paragraph we will discuss the statistical cohesiveness of our dataset. In subparagraph 4.3.1 we will discuss the core statistics and subsequently we will analyze the correlation matrix in subparagraph 4.3.2.

4.3.1 Core statistics

The core statistics of our dependent and independent variables are included in table 4.7 below.

Table 4.7 – Descriptive statistics

N Mean Std.

Deviation

Median Min Max

|DA| 1,819 0.184 0.3729 0.071 0.0 7.0 SIZE 1,819 16.318 1.7609 16.337 10.2 20.5 dREFORM 1,819 0.51 0.500 1.00 0.0 1.0 dEXTRA 1,819 0.23 0.419 0.00 0.0 1.0 dMIXED 1,819 0.64 0.479 1.00 0.0 1.0 dREFORM * dEXTRA 1,819 0.06 0.238 0.0 0.0 1.0 dREFORM * dMIXED 1,819 0.33 0.471 0.0 0.0 1.0 dREFORM * SIZE 1,819 8.32 8.219 12.658 0.0 20.4

We observe that the mean of dREFORM is close to 0.5, indicating the observations in our sample are fairly well distributed over the period before and and after the long-term care reform.

(31)

31

4.3.2 Correlation matrix

Table 4.8 is a bivariate Pearson Correlation matrix which measures direction and the strength of the mutual linear relationships between our dependent and independent variables. Correlations in a Pearson Correlation matrix range between -1 and 1 and the greater the absolute values of the correlation, the stronger the linear relationship.

Table 4.8 – Correlation matrix

|DA| SIZE dREFORM dEXTRA dMIXED dREFORM * dEXTRA dREFORM * dMIXED dREFORM * SIZE

|DA| ₁ _-0.275*** -.010 .013 .026 .020 .006 -.041 SIZE _-0.275*** 1 -0.049 ** -0.087 *** -0.062 ** -0.075 ** -0.067 ** 0.068 ** dREFORM _-.010 _-0.049** 1 .008 .005 0.25 *** 0.693 *** 0.987 *** dEXTRA _.013 _-0.087*** .008 1 -0.487 *** 0.702 *** -0.255 *** -.003 dMIXED _.026 _-0.062** .005 -0.487 *** 1 -0.342 *** 0.524 *** -.002 dREFORM * dEXTRA .020 -0.075 ** 0.25 *** 0.702 *** -0.342 *** 1 -0.179 *** 0.233 *** dREFORM * dMIXED .006 -0.067 ** 0.693 *** -0.255 *** 0.524 *** -0.179 *** 1 0.677 *** dREFORM * SIZE -.041 0.068 ** 0.987 *** -.003 -.002 0.233 *** 0.677 *** 1

*** Correlation is significant at the 0.01 level ** Correlation is significant at the 0.05 level

Based on this correlation matrix, we can conclude that there is a significant negative relationship between discretionary accruals and firm size. There are also significant correlations between SIZE and the other independent variables. Based on the absolute values of the correlations, these relations are weak (near 0). The absolute values of the correlations between the dependent- and the independent variable give no direct indication of multicollinearity.

(32)

32

4.3.3 Multicollinearity test

In addition to the correlation matrix, we have performed a test to measure the Variance Inflation Factor (VIF) to detect multicollinearity between the different independent variables in our models. The VIF shows the increase in the instability of the coefficient estimates due to multicollinearity (Freund et. al., 2000). There is no formal cutoff value for the tolerance or VIF, however a cutoff value of 10 is commonly used (O’Brien, 2007). Considering that our three hypotheses are all three fairly similar with the only distinction in which dummy variables are added, we have tested multicollinearity based on hypothesis 3 which includes all variables. Table 4.9 includes the multicollinearity results of all independent variables excluding the interaction effects.

Table 4.9 – Multicollinearity test (excl. interaction effects)

VIF

SIZE 1.025

dREFORM 1.002

dEXTRA 2.318

dMIXED 2.283

Based on the results presented in table 4.9 we conclude that the use of these independent variables does not pose a potential problem regarding multicollinearity as the VIF for all variables is below the threshold of 10. Subsequently we have tested for multicollinearity including the interaction effects. Refer to table 4.10 for a summary of our results.

Table 4.10 – Multicollinearity test (incl. interaction effects)

VIF SIZE 98.010 dREFORM 1.002 dREFORM * SIZE 90.080 dEXTRA 2.318 dREFORM * dEXTRA 3.031 dMIXED 2.283 dREFORM * dMIXED 5.331

(33)

33 We note very high multicollinearity between SIZE and dREFORM * SIZE which indicates these two variables supply redundant information and have therefore decided to eliminate dREFORM * SIZE from our models. Removing dREFORM * SIZE from our model lowers the maximum VIF value of SIZE to an acceptable level of 1.025.

(34)

34

5. Results

This chapter includes the results of our hypothesis testing. In paragraph 5.1 we will comment on the results of hypothesis 1 and in paragraph two we will discuss the results of hypothesis 2. Subsequently we will analyze the normality of the error term in paragraph 5.3

5.1 Results hypothesis 1

With our first hypothesis we tested the assumption that based on the political cost hypothesis, earnings management is more likely to occur in larger long-term care institutions.

We’ve tested this hypothesis using the following regression model as discussed in subparagraph 3.1.1:

| |

The outcome of the regression model is summarized in table 5.1:

Table 5.1 – Regression of Absolute Discretionary Accruals on firm size R Squared 0.075 Adjusted R Square 0.075 P-value 0.000 *** N 1,819 *** significant at 0.01 level

From the R Squared value in table 5.1 we conclude that there is a weak relationship between absolute discretionary accruals and firm size. A value of 0.075 indicates that the regression model explains 7,5% of the variance of the discretionary accruals.

The P-Value of the model is 0.000 and therefore lower than 0.05 which means the model is significant, indicating that there is a significant relationship between firm-size and earnings management. The regression coefficients are summarized in table 5.2.

(35)

35

Table 5.2 – Regression coefficients of Absolute Discretionary Accruals on firm size

Coefficient Standard Error T-statistic P-value

Constant 1.133 0.078 14.458 0.000***

SIZE -0.058 0.005 -12.177 0.000***

*** significant at 0.01 level

Based on the coefficient of variable SIZE, this relationship is significantly negative and we therefore reject hypothesis 1 which assumes a positive relationship.

From testing the first hypothesis we therefore learn that, contrary to our expectation based on the prior research discussed as part of our literature review performed in chapter 2, there is a significant negative relationship between earnings management and firm size in the long term-care sector. In this the long term-term-care sector deviates from other sectors (both profit and non-profit).

A reason for this could be that, although managers have an incentive to manage earnings to signal to the market that they are good managers (and thereby furthering their careers by possibly transitioning to a larger institution with a better remuneration scheme), the public opinion regarding accounting scandals and the misuse of public funds may stop managers from engaging in earnings management (Jiraporn et al., 2008) as the possible negative effects may outweigh the negative effects. In line with the political cost hypotheses this effect may be higher at larger firms than at smaller firms as larger firms are more often the subject to public scrutiny than smaller firms.

With our second hypothesis we answered the main research question by measuring the effect of the long-term care reform on earnings management.

reform

We’ve tested this hypothesis using the following regression model as discussed in subparagraph 3.1.2 with the exception of eliminating independent variable dREFORM * SIZE to circumvent multicollinearity issues as discussed in subparagraph 4.3.3.

(36)

36 The outcome of the regression model is summarized in table 5.3:

Table 5.3 – Regression of Absolute Discretionary Accruals on control variable SIZE (firm size) and dummy variable dREFORM

R Squared 0.076

Adjusted R Square 0.075

Significance F 0.000 ***

N 1,819

As with our previous regression model we conclude that the explanatory power of the model is limited with an R squared of 0.076, indicating that the regression model explains 7,6% of the variance in the mean of the discretionary accruals. The P-Value of 0.000 is lower than 0.01 and therefore the model is significant. This result is in line with the previous regression model which is very similar to this model. To determine the effect of the long-term care reform, we need to analyze the regression coefficients which are summarized in table 5.4:

Coefficients Standard Error T-statistic P-value

Constant 1.146 0.079 14.442 0.000 ***

dREFORM -0.018 0.017 -1.041 0.298

SIZE -0.058 0.005 -12.214 0.000 ***

As expected, the results for SIZE are similar to the regression model for hypothesis 1. For dREFORM we notice a P-Value of 0.298 which is higher than 0.01 and therefore we conclude that the period prior to the reform and the period after the reform do not significantly differ from each other. We therefore reject hypothesis 2 as no increases in earnings management is noted in the years following the 2015 long–term care reform.

With our third and final hypothesis we will elaborate on hypothesis 2 and research if the effect of the long-term care reform on earnings management differs for intramural (inpatient), extramural (outpatient) and mixed care subsectors.

(37)

37

H3: Earnings management within the Dutch long–term care sector increases for both the intramural- and

extramural care after the 2015 long–term care reform

We’ve tested this hypothesis using the following regression model as discussed in subparagraph 3.1.3 with the exception of eliminating independent variable dREFORM * SIZE to circumvent multicollinearity issues as discussed in subparagraph 4.3.3.

| |

The outcome of the regression model is summarized in table 5.3:

Table 5.3 – Regression of Absolute Discretionary Accruals on control variable SIZE (firm size) and dummy variable dREFORM

R Squared 0.077

Adjusted R Square 0.074

Significance F 0.000 ***

N 1,819

As this regression model is an extension on the regression model of hypothesis 2, the results are fairly similar. With an R squared of 0.076, the explanatory power of the model is the same as the model for hypothesis 2. The P-Value of 0.000 is lower than 0.01 and therefore this model is significant as well. To determine the effect of the long-term care reform and the intramural- and extramural care subsectors, we need to analyze the regression coefficients which are summarized in table 5.4.

(38)

38

Coefficients Standard Error T-statistic P-value

Constant 1.142 0.080 14.292 0.000 *** SIZE -0.058 0.005 -12.076 0.000 *** dREFORM -0.027 0.028 -0.996 0.319 dEXTRA -0.040 0.044 -0.921 0.357 dMIXED -0.005 0.029 -0.188 0.851 dREFORM * dEXTRA 0.057 0.060 0.947 0.344 dREFORM * dMIXED 0.017 0.041 0.425 0.671 *** significant at 0.01 level

As expected, the results for SIZE are similar to the regression model for hypothesis 1 and 2. Consistent with our results for hypothesis 2, we notice a P-Value for dREFORM which is not significant and therefore we conclude that the period prior to the reform and the period after the reform do not significantly differ from each other. We draw the same conclusion for the industry subcategories based on the P-values of dEXTRA (representing extramural care), dMIXED (representing mixed care) and the interaction variable dREFORM * dEXTRA (representing intramural care) which all exceed 0.05.

We therefore reject hypothesis 3 as no increases in earnings management is noted in the years following the 2015 long–term care reform for both the intramural, extramural and mixed care subsectors.

5.4 Normality of the error-terms

To test the robustness of the results of our research as discussed in the previous paragraph, we will test the normality of the error terms (which is one of the conditions for an Ordinary Least Squares regression) for the aforementioned regression models.

We have saved the standardized residuals of the regression models for our three hypotheses and ran a one-sample Kolmogorov-Smirnov test to test for normality. The results are summarized in table 5.5 on the next page.

(39)

39

Table 5.5 – Kolmogorov-Smirnov test for regression model hypothesis 1

Statistic df P-Value

Standardized residual H₁ 0.049 1,819 0.000***

Standardized residual H2 0.045 1,819 0.000***

Standardized residual H₃ 0.043 1,819 0.000***

From these results we conclude that the error-terms for the all three regression models are not normally distributed. This means we will need to reject the normally distributed residuals assumption and extra caution is required when interpreting the results of our research.

(40)

40

6. Conclusion

The main goal of this research is to determine if the 2015 reform of the long-term care sector led to an increase in earnings management. The central research question was therefore formulated as follows:

“Has the 2015 reform of the Dutch long–term care sector led to an increase in earnings management?”

Based on prior research regarding earnings management in the not–for–profit sector we noted that there are multiple incentives for managers of long-term care institutions to apply earnings management and that most of these incentives can be traced back to the agency theory.

Within the confines of this research the governmental bodies which largely fund this industry are the principals and the management of the long–term care institutions acts as the agents. As a result of the long-term care reform, a shift from the national government to local municipalities as principal occurred and budgets were lowered which caused uncertainty for the agents. We assumed that this uncertainty could lead to a situation where managers of long–term care facilities feel the need to, instead of austerity measures such as cutting costs, manage earnings to maintain the current financing level from the municipalities after the reform.

Prior research also suggest that, based on the political cost hypothesis, there is an expected positive relationship between firm size and the level of profit and political attention. We noted that this could be an incentive for managers in the long-term care sector to apply earnings management to manage earnings downwards as these may lead to lower subsidy income in consecutive years.

To answer the main research question of this research, we have therefore formulated the following hypotheses:

reform

The long-term care sector in the Netherlands can be further divided in two subsectors, namely intramural (inpatient) long–term care and extramural (outpatient) long-term care which led us to formulate a third hypothesis to measure to effect within these subsectors.

Earnings management in the Dutch long–term care sector : before and after the 2015 long–term care reform