Detecting Discontinuities in the Distribution of Earnings Around Zero as an Indication of Earnings Management ^†

Abstract

This manuscript aims to investigate whether Greek public hospitals engage in earnings management practices in order to avoid great deficits or surpluses. For this purpose, the bootstrap kernel density estimation was used in order to detect discontinuities at the distribution of earnings around zero. The sample consists of 1110 financial statements of Greek public hospitals for the period 2009–2020. By using the bootstrap kernel density estimation, we revealed that Greek public hospitals engage in EM techniques to report positive earnings close to zero in order to enforce their primary role, which is contribution to social welfare. This research expands the literature on EM by offering insights regarding the quality of net earnings that Greek public hospitals report.

1. Introduction

The implementation of accrual accounting in the public sector has led to improved recording of accounting information, offering a more “holistic view” of the entity’s financial position [1]. The accrual accounting system offers an array of benefits, such as full cost information [2], better asset and liability management [3], enhanced accountability and transparency [2,4], and improved recourse allocation and decision making [5]. Although there are advantages with this system, the subjective point of view regarding accounting information [6] remains a weakness of accrual accounting, primarily because it opens the door to practices like earnings management [7].

Earlier studies have provided evidence that earnings management can occur not only in the private sector but also in the public sector and non-profit entities [8]. In this realm, there are research works that have investigated the existence of earnings management in the local government context [7,9,10,11,12,13,14] and studies that have searched for this behavior in the healthcare sector [15,16,17,18,19,20,21,22].

The background literature shares the view that earnings management arises when managers try to keep in balance the satisfaction of stakeholders and the maximization of their own interest [16,18]. Hence, this creates an agency problem occurs due to the fact that the agents (i.e., the hospital) and the principals (i.e., the stakeholders) have conflicting interests.

The aim of this study was to analyze the financial outcome that Greek public hospitals report, in order to detect earnings management practices. For this purpose, this study used the bootstrap kernel density estimation approach developed by Lahr [23]. This methodology is regarded to be more accurate for detecting discontinuities in the distribution of earnings because it uses a bootstrap test to define the bandwidth for the kernel density estimation [23]. Building on the agency theory, this research work investigated whether Greek public hospitals adjust their earnings in order to present a specific financial outcome. Against this background, the annual financial statements (i.e., balance sheets and income statements) of Greek public hospitals for the period 2011–2020 were used.

This study adds to the existing literature on healthcare accounting by providing evidence that the healthcare sector engages in earnings management practices. Moreover, it expands the literature of agency theory by offering evidence that Greek public hospitals try to maintain the provision of social benefits while maintaining financial stability.

The remainder of this paper is organized as follows. Section 2 reviews the earnings management literature. Section 3 describes the Greek healthcare context, and Section 4 develops the theory and the hypothesis. Section 5 presents the research design and Section 6 the findings of the analysis. The conclusions of the paper are shown in the last section.

2. Literature Review

2.1. Earnings Management in Healthcare

Earnings management is a behavior that can take place when managers deliberately manipulate financial records to mislead stakeholders [24]. EM in the healthcare sector has been examined through different research designs.

Leone and Van Horn [15] and Ballantine et al. [16] relied on accrual accounts and on the distribution of earnings. More specifically, Leone and Van Horn [15] used specific accrual accounts, such as third-party settlements and the allowance for doubtful accounts, as proxies for earnings management. They found that adjustments to the third-party allowance and the allowance for doubtful accounts were made in order to report small surpluses. Moreover, they used the Burgstahler and Dichev [25] methodology, and their analysis offered evidence that supported the loss-avoidance hypothesis. Ballantine et al. [16] used two research methodologies to detect earnings management: the Burgstahler and Dichev [25] procedure and then the adapted model of Dechow and Dichev [26]. They found that NHS trusts managed accruals due to legal pressures to report earnings at breakeven.

Another stream of research focused on the aggregate accruals models to detect earnings management. Vansant [18] found that when charity care met or exceeded the external stakeholders’ normative expectations in U.S. hospitals, managers followed income-increasing practices through the use of discretionary accruals. Anagnostopoulou and Stavropoulou [21] noticed an upward discretionary accruals adjustment in English NHS hospitals in the period before their application for beneficial status. Finally, Malkogianni and Cohen [22] offered evidence that earnings management practices are used in Greek state-owned hospitals: either the pre-managed earnings are positive or negative. Moreover, they found that Greek state-owned hospitals were willing to adjust their financial image in order to continue receiving donations and subsidies, and also, they documented an association between the unexpected costs and revenues with earnings management.

A different research trajectory investigated real earnings management operations in the healthcare context. Eldenburg et al. [17] examined how expenditures were used in order to manage earnings. Their study offered evidence that managers adjusted expenditures and asset management to meet earnings benchmarks. Later, Heese [19] tried to determine if overbilling is an instrument for adjusting financial performance. He argued that UN nonprofit hospitals used overbilling instead of accruals manipulation or cutting discretionary expenditures in order to manage earnings.

Finally, there is a stream of research that has examined both accrual and real activity manipulation [20]. One study found that Italian hospitals used discretionary accruals, provisions, and non-operating expenses at the desired threshold.

2.2. The Greek Healthcare Sector

The Greek healthcare system is a system that includes both private and public services [27]. The national healthcare system is funded by the National Organization for the Provision of Health Services and by the government budget [28], and it offers an array of healthcare services, such as pre-hospital emergency care and primary, ambulatory, and inpatient hospital care, etc. [27].

With the presidential decree 146 passed in 2003 [29], Greek public hospitals adopted accrual accounting in order to enhance accounting information, accountability, decision making, and the management of the resources [30].

The period before the financial crisis in Greece, the healthcare system was suffering, among others, from high health spending [31]. According to Kalavrezou and Jin [32], Greek healthcare expenditures accelerated very quickly. During the fiscal crisis, the European Commission (EC), the International Monetary Fund (IMF), and the European Central Bank (ECB) set health policy reform as a priority [31]. The rationalization of expenditures and the replacement of the retrospective payment system with a new one based on the German version of diagnostics-related groups (DRGs) were some of proposed reforms. Health expenditures in 2015, in comparison with 2010, were reduced (from 9.6% to 8.4%) [27]. In 2022, the health spending per capita in Greece was USD 3,015 less than the OECD average (USD 4,986) [33].

3. Conceptual Framework and Hypothesis Development

This research work leverages the underpinnings of agency theory because an agency problem might exist in the public hospitals. Taking into consideration that in the healthcare context, there are two players, namely the hospitals and the stakeholders, if both of them are utility maximizers, then it is possible that the agent (i.e., the hospitals) will try to satisfy its own interests instead of those of the principal (i.e., the stakeholders) [34]. In this relationship, the principal is plagued with asymmetric information that favors agent’s interests [35] and leads the agent to act more opportunistically.

Prior literature investigating earnings management in the healthcare sector has built on agency theory, offering interesting findings. Leone and Van Horn [15] found that CEOs acted in an opportunistic manner to lead stakeholders astray regarding the financial position of the hospitals. Ballantine et al. [16] revealed that hospitals endeavored to report earnings close to zero in order to mitigate scrutiny and CEO dismissal.

Usually, in the public sector, entities avoid reporting great surpluses or deficits due to the fact that these entities are not profit-maximizers but are promoters of social welfare. In this sense, Leone and Van Horn [15], by using the Burgstahler and Dichev [25] methodology, documented discontinuities around zero and a tendency to avoid reporting deficits. Ballantine et al. [16] also followed the Burgstahler and Dichev [25] methodology. They found that English hospitals report income at breakeven, and they also detected discontinuities in the interval just above zero. Finally, the study of Malkogianni and Cohen [22] documented that Greek public hospitals try to report small positive earnings.

The aim of this study was to analyze the distribution of net earnings through the bootstrap kernel density estimation and to identify any discontinuities (as proxies of earnings management) that might exist in the distribution of earnings around zero. We did not follow the approach of Burgstahler and Dichev [25] because in this method, the selection of the bandwidth is arbitrarily chosen. The approach of the bootstrap kernel density estimation allows the definition of the bandwidth through the bootstrap approach to endogenize the selection procedure. According to Lahr [23], the main characteristic of this procedure is that it offers “a reference distribution that cannot be globally distinguished from the empirical distribution rather than assuming a correct reference distribution”. To identify any discontinuities, net earnings (scaled by lagged total assets) were used. Based on these, we examined the following research question:

RQ: 

Greek public hospitals report positive earnings above zero. Thus, there will be discontinuities in the distribution of earnings close to zero.

4. Materials and Methods

4.1. Sample

We used the annual financial statements of the Greek public hospitals for the period 2009–2020. All data were hand-collected from the official site of the “Transparency Program” (named “Diavgeia”).

Until 2012, there were 133 hospitals. From 2013 onwards, the number of hospitals decreased due to amalgamations in the health sector, and thus, from 2015 onwards, the number of hospitals was 124. The sample included 112 out of 133 hospitals for the period 2009–2012 and 116 out of 124 for the rest of the period. Our sample consists of 1110 financial statements of Greek public hospitals for the period 2009–2020.

4.2. Method

This paper uses the methodology developed by Lahr [23] in order to identify spikes in the distribution of earnings. These spikes are indications of earnings management.

According to this approach, it is important to compare the observed and the expected data. In a normal distribution, if there is a divergence between the actual and expected data, and this difference is statistically significant, then is considered as a discontinuity and an indication of earnings management [23].

A key element for the successful implementation of bootstrap kernel density estimation is the creation of a reference distribution. For the avoidance of biased results, the selection of a suitable bandwidth is critical. For the selection of the bandwidth, Silverman’s [36] rule of thumb was used for the bandwidth estimation. Then, the Epanechnikov kernel function was chosen to construct the kernel density estimation. The next procedure regarded the comparison between the reference distribution and the empirical cumulative distribution function in order to calculate the maximum difference (dmax). Afterwards, by using the bootstrap method, the confidence intervals for the empirical distribution were constructed. In the succeeding phase, we checked if the reference distribution, at the point with the maximum difference (dmax), met the confidence band. Finally, a z-test was used to test for discontinuities around zero.

5. Empirical Findings

5.1. Descriptives

Table 1. Descriptive statistics.

	Mean	p50	sd	Min	Max
Net earnings (scaled by lagged Total assets)	0.0617	0.0558	0.1084	−0.2736	0.5090
Total assets	92,000,000	61,700,000	98,600,000	793,024	612,000,000
Operating income	43,200,000	32,400,000	40,400,000	1,689,688	182,000,000
Operating cost	59,900,000	47,500,000	53,100,000	2,393,554	241,000,000

summarizes the descriptive statistics of Greek public hospitals for the period 2009–2020. As can be seen, the mean value of the net earnings is 0.062, which verifies that Greek public hospitals report positive surpluses close to zero.

Further, Table 1 shows that the average of total assets is EUR 92 million, while there are hospitals that have total assets of EUR 793 thousand. The mean value of the operating income is EUR 43.2 million, while the operating cost has a mean value of EUR 59.9 million. This reveals that during regular operation, public hospitals’ costs are greater than their income.

Figure 1 shows the distribution of net earnings (scaled by lagged total assets). The distribution has a range from −20% to +20%. As can be seen, the reported earnings are positive and close to zero.

5.2. Results

This study used the bootstrap kernel density estimation method in order to search for discontinuities at the distribution of net earnings, which might be an indication of earnings management. By selecting the Epanechnikov kernel function, we set the reference distribution for comparison with the cumulative distribution of net earnings. Following the methodology of Lahr [23], we defined the bandwidth (bandwidth = 0.0555) and the maximum difference (dmax = −0.0244) between the reference distribution and the empirical cumulative distribution.

In Figure 1, the histogram reveals that there are discontinuities at the intervals close to zero. These discontinuities are due to the differences between the actual data and the expected data. In order to assess whether these differences are statistically significant and to detect the exact location of these discontinuities, we used the z-test.

Table 2. Discontinuity test (bootstrap kernel density estimation).

	Intervals	Std. Diff.	Observed	Expected	p-Values	Sig.
1	−0.1110	−1.3323	0.0322	0.0407	0.1828
2	−0.0555	−3.2418	0.0603	0.0902	0.0012	***
3	0.0000	0.9303	0.1726	0.1615	0.3522
4	0.0555	3.1264	0.2464	0.2056	0.0018	***
5	0.1110	3.5265	0.2287	0.1846	0.0004	***
6	0.1665	−1.4565	0.1060	0.1214	0.1453

Level of significance *** = 0.01.

presents the actual and expected values for each interval. These intervals were set with the selected bandwidth (bandwidth = 0.0555). As can be seen, in the intervals [−0.111, −0.0555], [0, 0.0555], and [0.0555, 0.111], there are differences between the observed and the expected values, which are statistically significant at 1%. More specifically, in the interval [−0.111, −0.0555], the actual values are less than the expected ones, and in the intervals [0, 0.0555] and [0.0555, 0.111], the actual data are more than the expected data. This might be an indication of earnings management and also this confirms our hypothesis.

6. Conclusions

The aim of this paper was to analyze the financial outcome that Greek public hospitals report in order to detect earnings management practices. For this purpose, the methodology of bootstrap kernel density estimation was employed. The primary benefit of this approach is that it generates a reference distribution, which is not distinguished from the empirical distribution [23].

The detection of earnings management through the identification of discontinuities in the distribution of earnings is a common procedure in this field [37]. The benefit of this procedure is that it gives the opportunity to make strong predictions about the frequency of earnings, which might be a result of the discretionary portion of earnings [38].

By implementing the bootstrap kernel density estimation approach, we used Silverman’s [36] rule of thumb in order to select a first bandwidth estimate. Then, we compared the empirical cumulative distribution and the reference distribution to detect the location of maximum difference (dmax) between them. For this purpose, we used the Epanechnikov kernel function. The next procedure was performed to obtain the bootstrap confidence intervals for the empirical cumulative distribution. Then, we tested whether the reference distribution at the maximum difference point (dmax) met exact the confidence band. Finally, we employed a z-test to the discontinuities around zero.

The results from the analysis revealed that there are discontinuities around zero. More specifically, we found that there is a tendency within Greek public hospitals to report earnings above and close to zero.

This study adds to the existing literature on healthcare accounting by providing evidence that public hospitals engage in earnings management practices because there is a preference to report positive earnings close to zero. Moreover, it expands the literature of agency theory by offering evidence that Greek public hospitals try to maintain the provision of social benefits while maintaining their financial sustainability. The above implications are important for policy makers and oversight authorities.