Does State-Owned Enterprises’ Performance Evaluation Detect Earnings Manipulation?

Abstract

Performance evaluation systems serve as a crucial governance mechanism in enhancing operational efficiency and ensuring sustainable growth for state-owned enterprises (SOEs). Despite their significance, the effectiveness of these evaluation systems has received limited academic attention. This study examines how performance evaluations address earnings manipulation issues, focusing specifically on both accrual-based and real activity-based earnings management. Our empirical findings indicate that SOEs with higher accrual-based earnings management receive significantly lower ratings in performance evaluations. However, no significant relationship is observed between real activity-based management and performance evaluation ratings. These results suggest that while performance evaluations effectively account for accrual-based earnings manipulation, they fail to capture real activity-based earnings management. Our study emphasizes the need for a more nuanced approach to performance evaluation that not only detects accrual manipulation but also considers operational adjustments made by managers. Furthermore, these findings imply that performance evaluation committees and government regulators should integrate industry-specific expertise into the evaluation process to enhance the detection of real earnings manipulation, thereby strengthening governance tools in SOEs. This research contributes to the broader discourse on improving effectiveness in public sector performance assessments.

1. Introduction

The performance evaluation system for state-owned enterprises (SOEs) in South Korea serves as a critical governance mechanism to enhance operational efficiency, transparency, and alignment with sustainable development goals (SDGs). By setting clear performance targets and monitoring outcomes, this system aims to promote sustainable practices, improve financial soundness, and build public trust. Moreover, it offers an adaptable model for public enterprises globally. However, the effectiveness of this governance tool is potentially undermined by earnings management practices, which may obscure the true financial and operational performance assessment of SOEs. Earnings management, particularly through accrual manipulation and real activities, can distort the evaluation process, leading to misinformed policy decisions and compromised sustainable growth (Tabassum et al.) [1].

The central challenge lies in the potential distortion caused by earnings management practices, which may significantly impair the integrity and sustainability of performance evaluations. Accrual-based earnings management involves manipulating accounting estimates to influence reported earnings [2], while real earnings management entails altering operational decisions, such as cutting discretionary expenses or delaying investments, to meet short-term targets [3]. Both practices can mask the underlying economic reality, leading to overstated earnings and misleading assessments of an enterprise’s sustainability. Consequently, a critical question arises: does the current performance evaluation system effectively capture earnings quality, particularly the extent of accrual and real earnings management, to ensure the sustainable growth of SOEs?

The existing literature has extensively documented earnings manipulation in SOEs around the world [4,5,6,7,8]. These financial deceptions resulted in downgraded evaluation ratings and required employees to return performance-based incentives. These cases highlight significant concerns about the effectiveness of the current performance evaluation system in detecting and addressing financial fraud. This study extends the scope of earnings management research by examining local public enterprises in Korea, investigating whether the performance evaluation system accounts for earnings quality and how it influences sustainability assessments.

Our analysis results show that higher discretionary accruals are linked to lower performance ratings, indicating that evaluators tend to penalize firms for accrual-based earnings management. However, no significant relationship is found between real earnings manipulation and performance ratings. To validate these findings, we conducted additional tests, which consistently supported our results. We found that only upward earnings management led to lower ratings, particularly in qualitative evaluation measures. We also addressed potential endogeneity by using two-stage least squares (2SLS) and a lagged equation model, which confirmed our main findings. Additionally, we observed that discretionary accruals had a more negative impact on ratings for financially constrained firms, likely due to the government’s focus on the financial soundness of SOEs.

Our analysis focuses on the performance evaluation system in South Korea, which presents a unique institutional setting where annual evaluations are supervised by the central government according to relevant laws. The government plays a crucial role in guiding public enterprises towards efficient management by setting annual targets based on evaluation indicators. Moreover, it aims to enhance sustainability by improving the transparency and financial soundness of public enterprises. This institutional context provides a distinctive backdrop for exploring the impact of earnings management on performance evaluations and offers valuable insights for improving the sustainability and integrity of these assessments. This paper makes a unique contribution to the literature by analyzing how accrual and real earnings management practices influence the performance evaluations of Korean SOEs, offering policy implications for strengthening the evaluation system against financial manipulations. Additionally, the experiences of South Korean SOEs provide valuable case studies that can inform practices in SOEs globally. Given the relative scarcity of research on the performance evaluation of public enterprises, this paper makes a significant contribution to the field.

This study is structured into several key sections. Section 2 provides an overview of the performance evaluation system for local public enterprises. Section 3 reviews relevant literature and outlines the research hypotheses. Section 4 details the research model and the methodology used for analysis. Section 5 presents the empirical findings, followed by Section 6, which offers additional analyses to further support the study. Finally, Section 7 concludes the paper.

2. Institutional Background: Performance Evaluation Policy of Local Public Enterprises

State-owned enterprises (SOEs) play a significant role in supplementing market functions and augmenting national finances by consistently offering essential services to the public at affordable prices. However, issues such as lax and inefficient management practices within local public enterprises, encompassing irregular hiring procedures and misappropriation of funds, have emerged as notable social concerns in Korea. Consequently, many of these enterprises have been plagued by excessive fiscal deficits, exacerbating the financial condition of local governments. To address these challenges, the central government department, the Ministry of Interior and Safety (MOIS), has implemented a performance evaluation system under the Local Public Enterprises Act in Korea. The primary objective of this system is to foster competition and enhance the management efficiently of public enterprises. By establishing predefined goals and directives for local public enterprises, the performance evaluation aims to improve their overall management, the evaluation system compares, analyzes, and assesses performances based on the established standards, thereby providing incentives or questioning responsibilities as necessary [9,10].

The current performance evaluation system categorizes enterprises into five distinct grades: A, B, C, D and E. An enterprise that attains an A grade in the evaluation is eligible to award bonuses ranging from 180% up to 400% of employees’ monthly salaries. Conversely, enterprises that receive the lowest grade, E, forgo bonus payments to all employees and implement a salary reduction of 5% to 10% for the president and executives in the subsequent year. In addition, based on the evaluation outcomes, city and provincial governments have the authority to dismiss the president of the enterprise. Furthermore, the central government can mandate management improvements, such as downsizing projects or reorganizing the corporate structure, and in extreme cases, even order the liquidation of the enterprise.

The performance evaluation of local public enterprises is conducted by two entities: the Ministry of Interior and Safety (MOIS) and city/provincial governments. Notably, the majority of these enterprises undergo evaluation by MOIS, with exceptions being local-level WACs (Water Authority Corporations) and SCs (Sewage Corporations). For the purpose of this study, our sample is limited to the performance evaluation results of SOEs that are evaluated by MOIS. SOEs evaluated by city/provincial government are excluded from our sample due to the non-disclosure of detailed scores to the public. Consequently, our analysis focus on 159 local public enterprises that were evaluated by the MOIS as of 2020.

Figure 1 depicts the comprehensive timeline of the detailed evaluation process. Initially, enterprises are required to submit their evaluation data to the management information system by the middle of March of the year subsequent to the evaluation year. Then, an evaluation team appointed by the MOIS undertakes the evaluation process from mid-March to June of the following year, utilizing all the information provided by the enterprise. Typically, these evaluation teams comprise university professors, certified public accountants, and other experts who possess professional knowledge and maintain independence from any evaluation firms. Ultimately, the performance evaluation results are uploaded onto the management information system (CleanEye) by August of the year following the evaluation year.

The performance evaluation indicators encompass a range of financial and accounting outcomes. However, these outcomes are qualitatively distinct from earnings management, as the indicators do not directly account for abnormal levels of accruals or real activities, such as unusual cash flows, production costs, or discretionary expenses. This study shifts the focus to these abnormal levels of accruals and real activities, which serve as proxies for earnings manipulation driven by financial reporting incentives in public enterprises. This raises an unexplored empirical question: to what extent do abnormal accruals and real earnings management are reflected in the performance evaluation systems of SOEs?

3. Literature Review and Research Hypothesis

3.1. Performance Evaluation and Sustainable Growth of SOEs

In recent years, the performance evaluation mechanisms of SOEs have become a focal point of research due to their critical role in promoting accountability, efficiency, and transparency in the public sector. A growing body of literature emphasizes the need to integrate both financial and non-financial indicators in performance assessments. Klovienė and Gimžauskienė [11] delve into the construction of performance measurement systems for SOEs, advocating for the inclusion of sustainability and transparency as key elements.

Nations around the world employ a variety of methodologies to evaluate the performance of their SOEs in their pursuit of sustainable growth. The Asian Development Bank’s report, which examines the performance evaluation and management practices across 11 Asian economies including South Korea, underscores the wide array of approaches and the importance of utilizing both financial and non-financial indicators to assess SOE performance, as noted by the OECD in 2016 [12]. Turning to China, the State-owned Assets Supervision and Administration Commission (SASAC) of the State Council introduced an economic value added (EVA) performance evaluation system in 2010, specifically targeting the cash holdings value of Central State-Owned Enterprises (CSOEs), according to [13]. Meanwhile, Indonesia employs the Criteria for Performance Excellence (KPKU) assessment framework, and research by Ratri et al. [14] shows a positive correlation between KPKU assessments and the performance of Indonesian SOEs. These examples underscore the global trend toward diverse yet integrated approaches to SOE performance evaluation. In Korea, government performance evaluations (GPE) serve as a key mechanism for assessing the performance of SOEs, reflecting the country’s commitment to sustainable and transparent governance practices [15].

There is no universal evaluation model suitable for all scenarios. The existing literature primarily focuses on political factors that can undermine the effectiveness of performance evaluation systems in state-owned enterprises (SOEs). However, limited attention is given to accounting factors, such as earnings management, which can distort sustainable performance evaluation mechanisms in SOEs. Critics argue that the evaluation of SOEs may be biased and fail to objectively reflect their actual operational efficiency. For example, Du et al. [16] identify several political influences on the evaluation process, including the political connections of SOE Chief Financial Officers, the proximity of SOE headquarters to the central office of the State-owned Assets Supervision and Administration Commission (SASAC), and the political rank of the enterprise. These factors can skew evaluations and undermine objectivity. Subedi and Farazmand [17] suggest that using Economic Value Added (EVA) as a performance metric can motivate public administrators to improve the performance of public organizations, as shown in Chinese SOEs. However, concerns about fairness persist. Du et al. [18] find that adopting EVA measures may introduce fairness issues, as supervisors may incorporate personal biases into their assessments. For example, SOEs that perform poorly on EVA metrics may not face penalties if they achieve strong Return on Equity (ROE) outcomes, reflecting potential leniency in the evaluation process. Similarly, Kim, Shin and Yu [15] observe that government performance evaluation (GPE) scores for SOEs are significantly lower during public election years, as these scores play a key role in decisions about CEO replacement. This highlights the influence of external political factors on the fairness and reliability of performance evaluations.

3.2. Performance Evaluation and Earnings Management of SOEs

Previous research examines the differences in earnings management between state-owned enterprises (SOEs) and private companies, offering mixed findings across countries. Early theoretical perspectives emphasize that state ownership may correlate with corporate inefficiency due to governance deficiencies [19,20]. However, recent empirical studies challenge this conventional view. In China, SOEs demonstrate more restrained earnings management practices compared to private firms [6]. This finding is aligned with Dantas et al. [21], arguing that lower financial distress risk—stemming from implicit government guarantees—reduces incentives for accrual-based earnings smoothing in such entities.

Cross-national institutional variations, however, yield divergent outcomes. Capalbo et al. [22] find no statistically significant relationship between state ownership and accrual earnings management in Italian companies. Their study identifies firm size as a mitigating factor for SOEs’ earnings management, whereas profitability exacerbates it. Notably, political dynamics emerge as a critical explanatory variable in municipal SOEs. Müller and Sidki [23] highlight partisan ideological conflicts in local governments, election cycles, and political budget pressures substantially influence the extent to which municipal enterprise managers engage in financial statement manipulation. This politicized dimension enriches theoretical frameworks for interpreting localized SOE behaviors.

Research on Korean public enterprises provides further insights into earnings management practices and their determinants. Lee and Lee [24] show that firm size and return on assets are negatively correlated with earnings management in SOEs, while debt ratios exhibit a positive correlation. Cha and Kim [25], building on the work of Roychowdhury [26], identify strategies employed by SOEs to avoid reporting losses or declines in profits, confirming the existence of these practices. Han [27] explores the effects of earnings management by SOEs, finding that it boosts short-term financial performance but negatively affects long-term outcomes.

Additional studies highlight the influence of performance evaluation systems on corporate behavior. Kim and Lee [28] demonstrate that such systems significantly enhance corporate value, as measured by operating and net profits. Yoon and Moon [29] find that poor performance evaluations lead to substantial reductions in both accrual-based and real earnings management in the following year. These findings underscore the critical role of performance evaluation systems in improving financial reporting quality through effective monitoring.

3.3. Earnings Management Through Discretionary Accruals

Accruals represent financial obligations that have not been fully realized by the end of the fiscal year but are anticipated to be recognized in the future. This includes items such as accounts receivable, accounts payable, depreciation expenses, and asset impairment. Accruals can be divided into non-discretionary and discretionary components. Non-discretionary accruals naturally arise from routine business activities, while discretionary accruals involve management’s judgment and may deviate from typical operational needs. Current studies on earnings management primarily focus on discretionary accruals, as they can be intentionally adjusted by managers seeking to either inflate or deflate reported earnings [30,31]. Recent studies have expanded this focus to SOEs. Bisogno and Donatella [32] systematically review earnings management patterns in public sector entities, while Columbano et al. [33] analyze 302 Italian National Health Service organizations, revealing two key findings: (1) net income exhibits greater stability than cash flows due to accrual accounting, and (2) accrual components enhance the predictive value of future cash flow projections.

Firms engage in earnings management for various reasons, supported by empirical evidence. First, research by Burgstahler and Dichev [34] identify a discontinuity in earnings distributions near zero, suggesting firms often manipulate earnings to avoid reporting small losses. Similarly, firms may slightly inflate earnings to exceed financial analysts’ forecasts [35]. Public enterprises, like private ones, also manage earnings to avoid deficits [25]. Second, managers often manage earnings to maximize compensation, especially when incentives are tied to financial performance [36]. Public enterprises face similar pressures, as their performance evaluations influence managerial rewards [37]. Third, firms with high financial leverage manipulate earnings to prevent breaches of debt covenants [38]. Likewise, public enterprises manage earnings to avoid stricter oversight under debt management policies [39]. Finally, private firms frequently smooth earnings over time [40], and Korean public enterprises similarly prefer to report stable earnings [41].

3.4. Earnings Management Through Real Activities

Real earnings management involves adjustments to a firm’s actual operations that directly affect cash flow. Managers may delay economically significant actions or reduce investments to influence reported earnings [42]. Studies on real earnings management show that managers often cut R&D investments [43,44,45] or sell fixed assets to boost short-term earnings [46]. Roychowdhury [26] introduces a model to estimate real earnings management by analyzing deviations from expected levels of normal operating activities, such as current sales, changes in sales, and prior-period sales. Deviations from these expectations are seen as evidence of real earnings management. Metrics like sales, production costs, and selling and general expenses are commonly used to measure this type of manipulation. For instance, Burgstahler and Dichev [34] explore whether firms nearing zero profit levels use real earnings management to avoid losses. Roychowdhury [26] provides evidence that firms at risk of deficits manipulate real activities to artificially improve reported earnings.

Following Roychowdhury [26]’s work, research on real earnings management expands to explore its relationship with accrual management and its impact on future performance. Studies find that higher levels of real earnings management often correlate with lower levels of accrual earnings management. After the enactment of the Sarbanes–Oxley Act (SOX), firms reduce accrual manipulation but increase real management, indicating a shift in preference toward real activities as a method of earnings manipulation [2]. Furthermore, real earnings management negatively affects long-term performance [26,42,47]. For example, firms engaging in real manipulation during seasoned equity offerings experience greater post-offering declines compared to those relying on accrual management [48]. Recent studies document a trade-off between real and accrual management, where a decrease in accrual manipulation often coincides with an increase in real earnings management [49,50].

3.5. Hypothesis Development

The interplay between performance evaluation and accounting transparency in public enterprises is a critical area of investigation. From the theoretical framework, the agency theory serves as a foundational lens for understanding the principal-agent relationship inherent in public enterprises. Agency theory highlights the principal-agent relationship, wherein managers may engage in earnings management to align reported performance with their interests, especially in private companies where avoiding losses is a paramount objective [26,34]. Similarly, managers of SOEs also have motivations to adjust the financial income for evaluators despite their operational goals of fulfilling public service responsibilities of priority.

The performance evaluation systems in public enterprises function as critical governance mechanisms designed to enhance operational efficiency and accountability leading to sustainable growth. Prior research indicates that these systems play a significant role in improving both managerial performance and financial reporting quality. For instance, studies on Korea’s performance evaluation system reveal that poor evaluations in one period often lead to notable improvements in earnings quality in subsequent periods [28,29]. This underscores the capacity of these systems to detect and discourage manipulative practices, reinforcing their role in promoting transparency. The performance evaluation system is designed to identify and penalize accrual-based manipulations. Given the governance role of these evaluations, it is anticipated that their structured criteria and the expertise of evaluation teams will enable them to detect discrepancies resulting from earnings manipulation. Thus, effective evaluations will be inversely related to instances of accrual-based earnings management, aligning with the expectation that these systems promote higher earnings quality. Therefore, we posit the following:

H1. 

Accrual-based earnings management level is negatively associated with performance evaluation ratings in Korean public enterprises.

While SOEs possess the potential to engage in both AEM and REM, our framework indicates that performance evaluation systems are systematically more effective in identifying AEM than REM. This asymmetry in detectability stems from two interrelated factors: the substitutive relationship between AEM and REM [2,51] and the inherent challenges in identifying REM due to the absence of objective benchmarks [52]. In SOEs, financial performance is closely scrutinized under strict accounting standards and evaluative frameworks (e.g., MOIS in Korea or SASAC metrics in China). Because AEM involves deliberate adjustments to accounting policies or estimates (for example, changes in revenue recognition or depreciation methods) that directly alter key financial metrics, these modifications leave discernible traces in the financial statements. Evaluators can readily compare these accounting figures against cash flows or historical patterns, making deviations caused by AEM immediately conspicuous.

In contrast, REM involves strategic operational decisions aimed at influencing reported financial outcomes—such as altering production schedules or delaying expenditures [42]—which are an integral part of normal business practices. Given the complexity of REM practices, performance evaluation systems may struggle to determine when adjustments have crossed from legitimate operational variability into the realm of earnings manipulation [53]. Moreover, the operational changes inherent in REM—such as altering production schedules or de-laying expenses—tend to blend with everyday business activities and therefore often evade the scrutiny of evaluators [26]. Although managers might substitute AEM with REM to lower detection risk, REM typically imposes longer-term operational costs (e.g., reduced investment in R&D that may harm future profitability) that are not immediately reflected in annual performance metrics.

Additionally, SOEs operate under dual mandates to promote both economic performance and public welfare, which restricts the flexibility to engage in REM. The tension between fulfilling social objectives (such as maintaining employment or stabilizing prices) and achieving short-term financial targets means that even when REM occurs, its effects are diluted by the institution’s multifaceted operational goals. Furthermore, studies have shown that while stricter accounting rules help to reduce AEM, they may inadvertently push firms toward REM as an alternative strategy [54]. Given that performance evaluation systems predominantly focus on the immediate financial outcomes, the indirect and time-delayed impacts of REM are less likely to trigger robust evaluative responses.

Taken together, these factors indicate that while AEM is likely to influence performance evaluation outcomes due to its overt manipulation of accounting figures, REM remains largely undetected because of its operational embeddedness and the absence of clear detection benchmarks. Therefore, we propose the following:

H2. 

Real earnings management is insignificantly associated with performance evaluation outcomes in Korean public enterprises.

4. Research Design

4.1. Earnings Management Measures

To measure accrual earnings management, we employed discretionary accruals models of the performance-matched model by [55] in Equation (1). The residual terms estimated by industry and year are used as the level of discretionary accruals.

$${\displaystyle \frac{{TAC}_t}{A_{t-1}}}={\beta }_0+{\beta }_1{\displaystyle \frac{1}{A_{t-1}}}+{\beta }_2{\displaystyle \frac{∆{REV}_t-∆{AR}_t}{A_{t-1}}}+{\beta }_3{\displaystyle \frac{∆{PPE}_t}{A_{t-1}}}+{\beta }_{4}ROA+{\epsilon }_t$$

(1)

where ${TAC}_t$ is total accrual; ${∆REV}_t$ is change in sales; ${∆AR}_t$ is change in account receivable; $∆{PPE}_t$ is change in tangible asset (except construction in progress and land); $A_t$ is total asset; ${ROA}_t$ is net income divided by total asset.

To measure the degree of real earnings management, we followed the methodology of Roychowdhury [26], suggesting three activities that may boost current earnings by measuring the following as proxy variables for real earnings management: abnormal cash flow from operations (AbCFO), abnormal production costs (AbPC), and abnormal discretionary expenses (AbDE). We used the Roychowdhury [26] framework in our context of Korean local SOEs (e.g., water/gas utilities) for two reasons: first, these entities operate under standardized service mandates with homogeneous cost structures, satisfying the model’s intra-industry homogeneity assumption. Second, their non-market-oriented nature—where operational costs are determined by budget allocations rather than competitive strategies—aligns with the model’s premise that current/past revenues are the primary determinant of “normal” costs. Specifically, AbCFO, AbPC, and AbDE are calculated as the differences between the actual values of cash flow from operating activities, production costs, and discretionary expenses and their normal levels (i.e., the predicted values of Equations (2)–(4), respectively). The residual values from estimation models (2), (3), and (4) below are AbCFO, AbPC, and AbDE, respectively.

$${\displaystyle \frac{{CFO}_t}{A_{t-1}}}={\beta }_0+{\beta }_1{\displaystyle \frac{1}{A_{t-1}}}+{\beta }_2{\displaystyle \frac{S_t}{A_{t-1}}}+{\beta }_3{\displaystyle \frac{∆S_t}{A_{t-1}}}+{\epsilon }_t$$

(2)

$${\displaystyle \frac{{PC}_t}{A_{t-1}}}={\beta }_0+{\beta }_1{\displaystyle \frac{1}{A_{t-1}}}+{\beta }_2{\displaystyle \frac{S_t}{A_{t-1}}}+{\beta }_3{\displaystyle \frac{∆S_t}{A_{t-1}}}+{\beta }_4{\displaystyle \frac{∆S_{t-1}}{A_{t-1}}}+{\epsilon }_t$$

(3)

$${\displaystyle \frac{{DE}_t}{A_{t-1}}}={\beta }_0+{\beta }_1{\displaystyle \frac{1}{A_{t-1}}}+{\beta }_2{\displaystyle \frac{S_{t-1}}{A_{t-1}}}+{\epsilon }_t$$

(4)

where ${CFO}_t$ is cash flow from operations; $S_t$ is sales revenue; Δ$S_t$ is the changes in sales revenue; ${PC}_t$ is production costs; and ${DE}_t$ is discretionary expenses; $A_t$ is total assets. AbCFO and AbDE are multiplied by −1 to make their signs consistent with that of AbPC, indicating that a higher value represents more real earnings management. The sum of the three variables denotes REM variable, which is our main interest.

4.2. The Research Model

To investigate the impact of earnings management on performance evaluation results, we constructed the research model as follows:

$$\begin{array}{ll}{Ratings}_t=& {\beta }_0+{\mathit{\beta }}_1{AEM}_t+{\mathit{\beta }}_2{REM}_t+{\beta }_3{SIZE}_t+{\beta }_4{LEV}_t+{\beta }_5{OCF}_t+{\beta }_6{ROA}_t+{\beta }_7{TYPE}_t\\ & +{\beta }_8{GRW}_t+{\beta }_9{LOSS}_t+{\displaystyle \sum }IND+{\displaystyle \sum }YR+{\epsilon }_t\end{array}$$

(5)

Performance evaluations results are in the forms of scores and grades. The dependent variable, ${Ratings}_t$, includes the following: (1) ${Ratings}_{dummy}$, a dummy variable with a value of 1 if the evaluation results are higher than the average of each type of local public enterprises, or 0 (zero) if otherwise; (2) ${Ratings}_{rank}$, a rank variable created by modifying the five ratings (A, B, C, D, and E) by 0.25 for each level (from 1 point for A to 0 points for E); and (3) $Ratings\_100$, estimated by dividing the evaluation scores by 100. Next, AEM, the independent variable of interest, denotes discretionary accruals for accrual earnings management. REM, another independent variable of interest, represents the composite variable of AbCFO, AbDE, and AbPC, which are calculated based on Roychowdhury [26]. According to the hypothesis, the coefficients of AEM and REM are predicted to be all negative, suggesting that the degrees of earnings management are negatively reflected in performance evaluations.

As control variables, we included the following: the size of the enterprise (SIZE), in that larger enterprises may have a positive relationship with the evaluation; debt ratio (LEV), in that a lower debt ratio showed better performance evaluations [56]; cash flow from operating activities (OCF) as important information when evaluating management performance [57]; ROA (return on assets [44]); the three types of local public enterprises (TYPE) based on previous study in Korea documenting that the degrees of discretionary accruals vary with the type of public enterprise; sales growth (GRW), assuming that the higher the growth rate, the better the performance evaluation; LOSS, a dummy variable, assuming that the incentive of earnings management to avoid reporting a deficit affect the performance evaluation [34]; and Industrydum, representing the seven business type classification of local public enterprises.

4.3. Sample Selection

This paper targeted local public enterprises in Korea under the performance evaluation by the Ministry of the Interior and Safety (MOIS), using the performance evaluation results during the period from 2011 to 2020. The financial data and performance evaluation results used in this paper were manually obtained from Cleaneye, the management information disclosure system for local public enterprises.

Table 1. Sample construction and distribution.

Panel A: Sample Construction
Collecting local public enterprises that conducted performance evaluation directly by the MOIS	1173
Excluding data not having both grade and score of performance evaluation report of local public enterprises	1086
Excluding data not having financial data to measure the variable, REM, real earnings management	971
Excluding data not having financial data to measure the variables, AEM, discretionary accruals	735
Panel B: Sample Distribution by Business Types
Metropolitan Facilities Management Corp.	282
Comprehensive Corp.	173
Metropolitan City Development Corp.	145
Metropolitan Rapid Transit Corp.,	57
Water Authority Corp.	40
Sewage Corp.	28
Other	10
Total	735

presents data construction and sample distribution by business types of local public enterprises. The sample is winsorized the top and bottom 1% of independent and dependent variables to resolve the effect of outliers, producing a final sample of 735 enterprise-years.

5. Empirical Results

5.1. Descriptive Statistics and Correlation Analysis

Table 2. Descriptive statistics.

Variable.	N	Mean	Std. Dev.	Q1	Median	Q3
AEM	735	−0.035	0.624	−0.157	−0.001	0.089
REM	735	0.055	2.321	−0.312	0.001	0.474
Ratings_dum	735	0.570	0.495	0.000	1.000	1.000
Ratings_rank	735	0.596	0.221	0.500	0.500	0.750
Ratings_100	735	0.854	0.043	0.828	0.861	0.886
SIZE	735	24.726	3.200	21.325	25.340	27.539
LEV	735	1.402	1.968	0.185	0.806	1.894
OCF	735	0.056	0.296	−0.015	0.019	0.090
ROA	735	0.003	0.034	0.000	0.000	0.007
TYPE	735	0.346	0.476	0.000	0.000	1.000
GRW	735	0.082	0.317	−0.017	0.051	0.137
LOSS	735	0.561	0.497	0.000	1.000	1.000

This table presents descriptive statistics of the variables used in this paper. The sample includes a total 735 firm-year observations for SOE in Korea from 2011 to 2020. The variables are defined as follows: AEM = Discretionary Accrual calculated according [55]; REM = Proxy for real earning management calculated by the Roychowdhury [26] model per type-year. REM is the sum of AbCFO (abnormal cash flow from operating activities), AbDE (abnormal discretionary expenses), and AbPC (abnormal production costs). AbCFO and AbDE are multiplied by −1 to make higher REM represent more real earnings management; Ratings_dum = A dummy variable with a value of 1 if the evaluation results are higher than the SOE business type average and 0 otherwise; Ratings_rank = A rank variable created by modifying the five ratings (A, B, C, D, and E) by 0.25 for each level, from 1 point for A to 0 points for E; Ratings_100 = A variables for which evaluation scores are divided by 100; Where ${SIZE}_t$ is natural logarithm of toal assets; ${LEV}_t$ is total liabilities divided by total equity; ${OCF}_t$ is the value of operating cash flows divided by assets; ${ROA}_t$ is net income divided by total asset; ${TYPE}_t$ is a dummy variable with a value of 1 for a public authority and public company and 0 for a government direct management enterprise; ${GRW}_t$ is change in sales (${sales}_t-{sales}_{t-1}$) divided by sales of t − 1 year; ${LOSS}_t$ is a dummy variable with value of 1 for net loss and 0 otherwise.

shows descriptive statistics on the variables used in this paper. The mean and median of REM are 0.055 and 0.001, respectively. AEM has negative mean and median values, showing that our sample firms of local public enterprises had negative discretionary accruals, on average. Regarding the performance evaluation results, the mean (median) values are 0.854 (0.861) for Ratings_100 and 0.596 (0.500) for Ratings_rank, respectively, suggesting that the number of public enterprises with A and B ratings is larger than those with C and D ratings. The mean and median of LOSS are 0.561 and 1.000, respectively, indicating that more than 50% of local public enterprises suffer operating losses. These statistics confirm that local public enterprises have been operated for the public interest, rather than to maximize profits, but nonetheless question the efficiency of their operations.

Table 3. Correlation analysis.

		(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)	(10)	(11)
(1)	AEM	1.000
(2)	REM	0.290	1.000
(3)	Ratings_dum	0.014	0.050	1.000
(4)	Ratings_rank	−0.015	0.004	0.639	1.000
(5)	Ratings_100	0.012	0.052	0.702	0.865	1.000
(6)	SIZE	0.013	−0.038	0.045	−0.021	−0.089	1.000
(7)	LEV	−0.147	−0.021	−0.074	−0.025	−0.026	−0.213	1.000
(8)	OCF	−0.092	−0.181	0.032	0.080	0.070	−0.142	0.112	1.000
(9)	ROA	0.228	−0.025	0.167	0.196	0.187	0.037	−0.019	0.101	1.000
(10)	TYPE	0.062	−0.019	−0.034	−0.013	−0.067	0.460	0.042	−0.019	0.230	1.000
(11)	GRW	0.065	0.077	0.123	0.096	0.077	0.083	0.029	0.032	0.149	0.133	1.000
(12)	LOSS	−0.109	0.007	−0.038	−0.020	0.030	−0.477	0.089	0.067	−0.505	−0.406	−0.106

This table presents the Pearson correlation coefficients for the variables used in our analysis. These data include a sample of local public enterprises in Korea from 2011 to 2020. Correlation coefficients in bold are all significant at a 5% or 1% level. Refer to Table 2 for variable definitions.

presents the Pearson correlation analysis. The results show that the variables of AEM and REM are not significantly correlated with Ratings_dum, Ratings_rank, and Ratings, which is inconsistent with our hypothesis. Notably, it shows that leverage (LEV) and operating cash flow (OCF) are significantly and negatively correlated with evaluation results, but return on asset (ROA) is positively correlated with them.

5.2. The Effect of Earnings Management on Performance Evaluation

Table 4. The effect of earnings management on performance evaluation results.

Panel A: The Effect of Accrual Earnings Management on Performance Evaluation Results.
Dep. Var.=	Ratings_dum			Ratings_rank			Ratings_100
Variables	Coeff.	Chi- Square		Coeff.	t-Value		Coeff.	t-Value
Intercept	−12.606	42.86	***	−0.195	−0.96		0.064	16.93	***
AEM	−0.365	5.75	**	−0.032	−2.59	**	−0.006	−2.51	**
SIZE	0.497	43.59	***	0.031	3.98	***	0.008	5.02	***
LEV	−0.223	16.66	***	−0.013	−2.70	***	−0.003	−3.89	***
OCF	0.084	0.09		0.032	1.17		0.003	0.62
ROA	23.981	24.40	***	1.934	5.00	***	0.413	7.31	***
TYPE	−1.396	9.81	***	−0.053	−0.91		−0.006	−0.49
GRW	0.751	6.81	***	0.055	2.30	**	0.010	2.22	**
LOSS	0.754	7.69	***	0.043	1.50		0.006	1.12
∑Industrydum	Yes			Yes			Yes
∑Yeardum	Yes			Yes			Yes
R2	0.1508			0.1568			0.2875
N	735			735			735
Panel B: The Effect of Real Earnings Management on Performance Evaluation Results.
Dep. Var.=	Ratings_dum			Ratings_rank			Ratings_100
Variables	Coeff.	Chi- Square		Coeff.	t-Value		Coeff.	t-Value
Intercept	−0.177	−0.87		−0.177	−0.87		0.668	17.05	***
REM	−0.001	−0.26		−0.001	−0.26		0.000	0.46
SIZE	0.031	3.91	***	0.031	3.91	***	0.007	4.95	***
LEV	−0.011	−2.32	**	−0.011	−2.32	**	−0.003	−3.49	***
OCF	0.040	1.40		0.040	1.40		0.005	1.13
ROA	1.791	4.79	***	1.791	4.79	***	0.388	6.74	***
TYPE	−0.054	−0.94		−0.054	−0.94		−0.006	−0.51
GRW	0.054	2.22	**	0.054	2.22	**	0.009	2.05	**
LOSS	0.045	1.57		0.045	1.57		0.007	1.20
∑Industrydum	Yes			Yes			Yes
∑Yeardum	Yes			Yes			Yes
R2	0.1446			0.1498			0.2811
N	735			735			735
Panel C: The Effect of Accrual and Real Earnings Management on Performance Evaluation Results.
Dep. Var.=	Ratings_dum			Ratings_rank			Ratings_100
Variables	Coeff.	Chi- Square		Coeff.	t-Value		Coeff.	t-Value
Intercept	−12.614	42.72	***	−0.195	−0.96		0.664	16.91	***
AEM	−0.419	7.03	***	−0.034	−2.57	**	−0.007	−2.75	***
REM	0.048	1.57		0.002	0.44		0.001	1.15
SIZE	0.498	43.45	***	0.031	3.98	***	0.008	5.01	***
LEV	−0.222	16.45	***	−0.013	−2.69	***	−0.003	−3.88	***
OCF	0.145	0.27		0.034	1.23		0.004	0.85
ROA	24.795	24.95	***	1.946	4.98	***	0.420	7.33	***
TYPE	−1.401	9.87	***	−0.053	−0.91		−0.005	−0.48
GRW	0.721	6.24	**	0.054	2.26	**	0.009	2.08	**
LOSS	0.770	7.95	***	0.044	1.50		0.007	1.14
∑Industrydum	Yes			Yes			Yes
∑Yeardum	Yes			Yes			Yes
Test of ${\beta }_1$ of AEM = ${\beta }_2$ of REM	chi-square = 7.34 ***			F-value = 5.54 ***			F-value = 8.38 ***
R2	0.1527			0.1570			0.2892
N	735			735			735

This table (Panel A, B and C) presents the OLS regression results from the effect of earnings management measured by AEM and REM on performance evaluation results measured by Ratings_dum, Ratings_rank, Ratings_100. ** and *** indicate significance at the 5% and 1% levels, respectively. Refer to Table 2 for variable definitions.

presents the main regression analyses that test the effects of earnings manipulation on performance evaluation results. Panel A shows that the coefficients of AEM are all significant and negative, −0.365 (chi-square = 5.75), −0.032 (t-value = −2.59) and −0.006 (t-value = −2.51) in three dependent variables. The coefficients of AEM in Panel C shows similar results in all three performance ratings measurements. These results suggest that accrual earnings management by local public enterprises are negatively reflected in performance evaluation results, which strongly supports our first hypothesis. This implies that the performance evaluation system can effectively capture accrual-based earnings management. Meanwhile, the coefficients of REM in Panel B and C show no significant effects on performance evaluation ratings in all three dependent variables. This is consistent with our second hypothesis, suggesting that current performance evaluation system should be revised to reflect real activities earnings management.

6. Additional Analyses

6.1. Additional Analyses Using Absolute Values of AEM and REM

There are two types of earnings manipulation: income-increasing versus income- decreasing behaviors. The incentives of income-increasing manipulation are to avoid negative influence from profit loss or/and receiving incentives. Meanwhile, the income-decreasing management is conducted when senior managers need to “preserve” their financial performance for the subsequent periods. As a further investigation, we distinguish AEM and REM into positive, negative and absolute values. Panel A of

Table 5. Additional analyses using absolute values of AEM and REM.

Panel A: Absolute Value of AEM and REM
Dep. Var.=	Ratings_dum			Ratings_rank			Ratings_100
Variables	Coeff.	Chi-Square		Coeff.	t-Value		Coeff.	t-Value
Intercept	−11.890	36.52	***	−0.154	−0.75		0.672	17.17	***
AbsAEM	0.055	0.11		0.012	0.76		0.003	0.86
AbsREM	−0.037	0.53		−0.006	−1.20		−0.001	−1.24
SIZE	0.473	38.29	***	0.030	3.75	***	0.007	4.83	***
LEV	−0.195	13.33	***	−0.011	−2.36	**	−0.003	−3.51	***
OCF	0.183	0.44		0.044	1.54		0.005	1.17
ROA	21.313	21.73	***	1.830	4.82	***	0.396	6.75	***
Type	−1.392	9.80	***	−0.057	−1.01		−0.007	−0.59
GRW	0.755	6.88	***	0.055	2.27	**	0.010	2.21	**
LOSS	0.747	7.70	***	0.046	1.60		0.007	1.22
∑Industrydum	Yes			Yes			Yes
∑Yeardum	Yes			Yes			Yes
R2	0.1448			0.1517			0.2835
N	735			735			735
Panel B: Absolute Values of AEM and REM by Positive and Negative Value
Dep. Var.=	Ratings_dum			Ratings_rank			Ratings_100
Variables	Coeff.	Chi-Square		Coeff.	t-Value		Coeff.	t-Value
Intercept	−12.495	38.35	***	−0.207	−1.01		0.663	16.9	***
AEM+	−0.447	2.99	*	−0.033	−1.98	**	−0.005	−1.80	*
AbsAEM-	0.363	3.34	*	0.046	2.80	***	0.007	1.71	*
REM+	0.037	0.34		−0.001	−0.13		0.001	0.56
AbsREM-	−0.069	1.19		−0.006	−1.06		−0.002	−1.28
SIZE	0.495	40.14	***	0.032	4.01	***	0.008	5.08	***
LEV	−0.223	15.67	***	−0.014	−2.96	***	−0.003	−3.61	***
OCF	0.154	0.30		0.036	1.27		0.000	−0.05
ROA	25.297	25.00	***	2.642	5.83	***	0.485	5.48	***
Type	−1.403	9.76	***	−0.059	−1.05		−0.006	−0.57
GRW	0.726	6.30	***	0.052	2.14	**	0.010	2.10	**
LOSS	0.781	8.15	***	0.062	2.11	**	0.009	1.57
∑Industrydum	Yes			Yes			Yes
∑Yeardum	Yes			Yes			Yes
R2	0.1540			0.1637			0.2884
N	735			735			735

Panel A shows the regression results using absolute value of AEM (AbsAEM) and REM (AbsREM). Panel B shows the regression results using absolute value of AEM and REM by positive and negative values, respectively. AEM+(REM+) is the value of AEM (REM) when AEM (REM) is positive, and zero value when AEM (REM) is negative; AbsAEM- (AbsREM-) is the absolute value of AEM (REM) when AEM (REM) is negative, and zero value when AEM (REM) is positive. *, ** and *** indicate significance at the 10%, 5% and 1% levels, respectively. Refer to Table 2 for variable definitions.

documents that there are no significant relations between the absolute values of AEM (AbsAEM) and REM (AbsREM) and the three types of evaluation ratings, proposing that the upward and downward earnings adjustments offset each other. Panel B shows that the positive value of AEM (upward adjustment of AEM) is negatively reflected in the evaluation system, consistent with our main regression results in Table 4. It is notable that negative values of AEM (downward adjustment of AEM) are positively significantly associated with evaluation results, suggesting that SOEs managers preserve the earnings and the evaluation system perceives the downward adjustment in a positive light. In line with our main analysis, upward adjustment of REM (REM+) as well as downward adjustment of REM (AbsREM-) have no significant impact on evaluation ratings.

6.2. Qualitative and Quantitative Scores in Performance Ratings

We reexamine our research hypothesis by distinguishing performance evaluation results as qualitative (non-quantitative) and quantitative measures. In

Table 6. The logistics regression of qualitative and quantitative measures on earnings management.

Dep. Var.=	Qualitative Ratings_dum			Quantitative Ratings_dum
Variables	Coeff.	Chi-Square		Coeff.	Chi-Square
Intercept	−9.954	20.94	***	−6.645	11.59	***
AEM	−0.385	4.42	**	−0.315	2.98	*
REM	0.069	2.70		0.093	4.50	**
SIZE	0.445	26.74	***	0.285	13.78	***
LEV	−0.143	7.07	***	−0.124	5.56	**
OCF	0.054	0.03		−0.296	0.56
ROA	7.964	4.68	**	16.805	15.49	***
TYPE	0.422	0.68		−0.092	0.04
GRW	−0.488	2.77		0.648	4.57	**
LOSS	0.136	0.21		0.534	3.17	*
∑Industrydum	Yes			Yes
∑Yeardum	Yes			Yes
R2	0.1410			0.2356
N	735			735

This table presents the logistic regression results from the effects of earnings management on performance evaluation results of qualitative and quantitative measures, respectively. Qualitative (Quantitative) Ratings_dum is a dummy variable with a value of 1 if the evaluation results of Qualitative (Quantitative) measure are higher than the average of each type of enterprises and 0 otherwise. There are three types of local public enterprises such as public authority, public company, government direct management enterprise. *, ** and *** indicate significance at the 10%, 5% and 1% levels, respectively. Refer to Table 2 for variable definitions.

, we find a significant negative association between AEM and both qualitative and quantitative evaluation results. It is notable that the magnitudes of coefficients of qualitative measures are larger than those of quantitative measures, suggesting that the negative relation between discretionary accruals and evaluation results are more pronounced for qualitative than quantitative evaluation results. It can be interpreted that accrual management is more reflected in qualitative than quantitative measure, which is consistent with the notion that qualitative measures that are more subjective and easily adjusted have more room for reflection than quantitative measures that are more objective and determined by an established formula [58].

On the other hand, the coefficient of REM, which is another variable of interest, shows insignificant result for qualitative measure, but significant and positive result for quantitative ratings. This is inconsistent with our prediction, suggesting that the performance evaluation system is rather deceived by real activities manipulation in quantitative evaluation measure. We believe this may cause more problems in the assessment of quantitative measure in that public enterprises are likely to be motivated to perform more real activity manipulation.

6.3. Endogeneity Issues

The core finding of this study is that performance evaluations, on average, reflect accrual earnings management in the performance evaluation results. To address any potential endogeneity problems such as reverse causality in the relation between discretionary accruals (AEM) and evaluation results, we first re-design our research model by testing whether current performance evaluations are associated with prior year discretionary accruals. This model can alleviate the reverse causality problem in the relation between discretionary accruals and performance results by mitigating the possibility that evaluation ratings influence discretionary accruals reversely. As predicted in our hypothesis, Panel A of

Table 7. Additional analyses for endogeneity.

Panel A: The Effect of Prior AEM on Current Evaluation Ratings
Dep. Var.= Variables	Current Ratings_dum			Current Ratings_rank			Current Ratings_100
	Coeff.	Chi-Square		Coeff.	t-Value		Coeff.	t-Value
Intercept	−12.510	28.74	***	−0.169	−0.74		0.641	14.850	***
$\mathit{P}\mathit{r}\mathit{i}\mathit{o}\mathit{r}\mathit{A}\mathit{E}\mathit{M}$	−0.376	3.31	*	−0.023	−1.99	**	−0.005	−2.050	**
$PriorREM$	0.060	1.51		0.000	0.04		0.001	1.330
$PriorSIZE$	0.491	29.20	***	0.030	3.41	***	0.008	5.080	***
$PriorLEV$	−0.231	11.16	***	−0.018	−3.10	***	−0.004	−3.050	***
$PriorOCF$	0.111	0.12		−0.002	−0.07		0.005	0.920
$PriorROA$	11.111	8.06	***	1.620	3.70	***	0.357	3.730	***
$PriorTYPE$	−1.566	10.33	***	−0.090	−1.49		−0.021	−1.690	*
$PriorGRW$	0.331	1.03		0.030	1.05		0.003	0.510
$PriorLOSS$	0.528	3.17	*	0.038	1.28		0.011	1.830	*
∑Industrydum	Yes			Yes			Yes
∑Yeardum	Yes			Yes			Yes
R2	0.1307			0.1085			0.2384
N	562			562			562
Panel B: Regression Analysis Using 2SLS Analysis—2nd Stage
Dep. Var.=	Ratings_dum			Ratings_rank			Ratings_100
Variables	Coeff.	Chi-Square		Coeff.	t-Value		Coeff.	t-Value
Intercept	−12.565	40.51	***	−0.157	−0.79		0.679	17.44	***
AEM_HAT	−0.539	11.04	***	−0.034	−2.02	**	−0.009	−2.63	***
REM	0.010	0.07		0.001	0.29		0.001	1.58
SIZE	0.593	48.69	***	0.036	4.43	***	0.009	5.84	***
LEV	−0.207	14.94	***	−0.011	−2.32	**	−0.003	−3.48	***
OCF	0.02	0.01		0.029	1.13		0.003	0.70
ROA	19.424	17.61	***	1.673	4.44	***	0.360	6.03	***
Type	−1.491	11.09	***	−0.060	−1.06		−0.008	−0.75
GRW	0.685	5.81	**	0.051	2.14	**	0.009	2.03	**
LOSS	0.777	7.96	***	0.046	1.61		0.007	1.28
∑Industrydum	Yes			Yes			Yes
∑Yeardum	Yes			Yes			Yes
R2	0.1583			0.1572			0.2956
N	735			735			735
Panel C: Regression Analysis Using 2SLS analysis—1st Stage
Dep. Var. Variables				Discretionary Accruals
				Coeff.			t-Value
Intercept				0.559			−0.27
SIZE				0.023			0.74
LOSS				0.113			−1.94	*
OCF				0.055			−1.60
LOCAL				0.138			−1.81	*
R2				0.0125
N				876

Panel A of Table 7 presents the effect of the prior AEM and REM on current evaluation results measured by Ratings_dum, Ratings_rank, Ratings 100. Panel B of Table 7 presents the second stage regression results in the 2SLS methodology. AEM_HAT is calculated as a residual in the first stage in the regression model as follows: ${AEM}_t={\beta }_0+{\beta }_1{SIZE}_t+{\beta }_2{LOSS}_t+{\beta }_3{OCF}_t+{\beta }_4{LOCAL}_t+{\epsilon }_t$. ${LOCAL}_t$ is a dummy variable that is 1 if SOEs are operated by city/provincial government, and 0 if SOEs are operated by other local government. *, ** and *** indicate significance at the 10%, 5% and 1% levels, respectively. Refer to Table 2 footnote for variable definitions.

shows that the coefficients on prior AEM in three dependent variables are all negative and significant at the conventional levels, suggesting that current year performance ratings on average reflect prior year discretionary accruals.

As a second method to resolve this endogeneity concern, we implemented the two-stage least squares (2SLS) methodology. We generated and included the predicted value of discretionary accruals, AEM_HAT, in our main research model of Equation (5) as a second step. Panel B of Table 7 reports the 2SLS method result, showing that the coefficients of AEM_HAT are all negative and significant at the conventional levels. These statistics strongly support our hypothesis, suggesting that the negative relation between evaluation results and AEM is established even after controlling for the endogeneity problems.

6.4. The Effect of Earnings Management on Performance Evaluation Results, Conditional on Financial Constraints

In order to strengthen the financial soundness of public enterprises, the MOIS implemented intensive monitoring policies for efficient financial management of public enterprises. Specifically, the MOIS selects the public enterprises with financial risks, requires them to submit a report on financial management planning and related improvements. As prior studies documented that financially constrained SOEs tend to execute earnings management [23], we focus on the evaluation system reflection for the financially constrained SOEs in the ratings process. In this regard, we posit that evaluators more negatively incorporate earnings manipulation behaviors when public enterprises are more financially constrained.

Table 8. The effect of earnings management on performance evaluation results, conditional on financial constraints.

Dep. Var.=	Ratings_dum			Ratings_rank			Ratings_100
Variables	Coeff.	Chi-Square		Coeff.	t-Value		Coeff.	t-Value
Intercept	−14.181	48.13	***	−0.194	−0.95		0.662	16.94	***
AEM	−0.313	3.55	*	−0.031	−2.32	**	−0.006	−2.37	**
REM	0.046	1.01		0.002	0.52		0.001	1.34
FC	−0.420	4.56	**	−0.015	−0.59		−0.007	−1.47
AEM*FC	−0.603	6.13	**	−0.050	−1.8	*	−0.012	−2.12	**
REM*FC	0.037	0.21		0.003	0.35		0.000	0.18
SIZE	0.570	49.85	***	0.032	3.87	***	0.008	5.08	***
LEV	−0.252	19.20	***	−0.013	−2.57	**	−0.003	−3.92	***
OCF	0.118	0.17		0.027	0.99		0.002	0.38
ROA	24.304	23.45	***	1.997	4.95	***	0.428	7.34	***
TYPE	−1.377	9.34	***	−0.051	−0.87		−0.004	−0.39
GRW	0.736	6.35	**	0.056	2.29	**	0.010	2.15	**
LOSS	0.804	8.45	***	0.045	1.57		0.007	1.25
∑Industrydum	Yes			Yes			Yes
∑Yeardum	Yes			Yes			Yes
R2	0.1540			0.1637			0.2884
N	735			735			735

This table presents the OLS regression results from the effect of earnings management on performance evaluation results, conditional on financial constraint. FC is a dummy variable indicating 1 if cash divided by total asset is lower than the median and 0 otherwise. *, ** and *** indicate significance at the 10%, 5% and 1% levels, respectively. Refer to Table 2 footnote for variable definitions.

shows that the coefficients of AEM*FC are all significant and negative in all three dependent variables, but the coefficients of REM*FC are all insignificant, which is in line with our previous results. These results suggest that evaluation results reflect AEM more negatively when firms are under financial constraints, but do not reflect REM regardless of financial constraints.

7. Conclusions

This paper explores the relationship between earnings management practices and performance evaluation outcomes within Korean local public enterprises. The empirical analyses show that accrual-based earnings management negatively impacts performance evaluations, as enterprises engaging in such practices tend to receive lower evaluation scores. In contrast, real activities-based earnings management does not significantly affect evaluation results, suggesting that evaluation teams may overlook these practices during their assessments. Moreover, a positive correlation emerges between real earnings management and quantitative performance indicators, indicating that the current evaluation system may inadvertently incentivize profit inflation through real activities manipulation.

Given this potential for encouraging real earnings management, we recommend enhancing the design of performance evaluation measures. Our research makes a meaningful contribution to the literature by examining Korea’s unique institutional context, where annual evaluations are systematically supervised by the central government in accordance with established laws for more than a decade. This governmental role is pivotal in steering public enterprises towards effective management and sustainable growth, emphasizing the need for transparency and financial soundness.

Additionally, our findings have significant implications for central and local governments, as well as performance evaluation committees, as they navigate the complexities of assessing SOEs. By illuminating how both accrual and real earnings management practices affect performance evaluations, we provide actionable insights for strengthening the evaluation framework against potential financial manipulations. However, two methodological limitations should be noted: first, following Chen et al. [59], our reliance on residual-based earnings management proxies (e.g., abnormal accruals) may introduce measurement errors that attenuate the observed relationships. Second, while accruals volatility in our sample of traditional utilities is minimal due to stable sectoral practices, studies of SOEs undergoing technology-driven business model transformations [60] may require adjusted measurement frameworks. The experiences of Korean SOEs also offer valuable exemplary cases that can inform practices in SOEs worldwide, contributing to a deeper understanding of performance evaluation dynamics in the public sector.

Subsequent investigations could broaden the scope by examining performance evaluations in different countries or sectors, thus providing comparative insights. Future research should also address the limitations noted above by (1) applying Chen et al.’s [59] instrumental variable approaches to mitigate generated regressor bias, and (2) integrating Srivastava’s [60] technology-adoption metrics into accruals models for SOEs in innovation-driven industries. Furthermore, longitudinal studies could offer deeper insights into the evolving nature of earnings management practices and their implications over time. Future studies could further elucidate the relationship between performance evaluation systems and earnings management, enhancing the sustainable growth and integrity of public enterprise assessments globally.