Examining the efforts of government Internal Supervisory Apparatus (APIP) in enhancing bureaucratic reform, organizational integrity, and risk management: An efficiency analysis

: The efficiency level of APIPs should be a concern for the government to ensure that the resources allocated are optimally used to fulfill its role in preventing and detecting corruption. This study investigates the efficiency level of APIP using Data Envelopment Analysis (DEA). The number of employees and spending budget ratios of APIP are used as input, and ministries and regional government’s bureaucratic reform, risk management, and integrity perception score as output. Those outputs are selected since they represent the indicators of an effective anti-corruption effort. Applying a purposive sampling of 50 APIPs at the ministerial and regional levels, this study shows that only seven APIPs (14%) achieve maximum efficiency level. Furthermore, this study develops an algorithm to set those APIPs as a reference group by other similar APIPs based on the characteristics of inputs and output. Lastly, this study employs quadrant analysis to map the APIP’s efficiency level and its capabilities. It finds that 14 A PIPs are relatively efficient and capable, while 35 APIPs are capable but not relatively efficient, and one APIP is not capable or efficient. This study could be used to design a more detailed and tailor-made agenda for strengthening APIP as an essential element in combating corruption in Indonesia.


Introduction
The need for a more accountable, transparent, and efficient state finance has been the spirit that underlies the reform in Indonesia.The reform itself encompasses political reform through the implementation of decentralized government over centralized government and the bureaucratic reform that aspires to achieve good governance and clean government (Turner et al., 2022).Those reforms were followed by the state finance reform, which encompasses all phases of financial management, from the planning, budgeting, implementation, controlling, supervision, auditing, and reporting phases (Government of Indonesia, 2003).One of the fundamental agendas in the state finance reform that contributed to the achievement of good governance and clean government was the establishment of a government internal control system (SPIP).
The SPIP underlines the role of the Government Internal Supervisory Apparatus (APIP) in implementing internal audit functions for all ministries, agencies, and regional governments.Indeed, many studies emphasize the role of auditing and oversight in combating corruption as well as the role of auditors to protect the public interest (Paterson et al., 2019;van Brenk et al., 2022).In a more specific context, the empowerment of the government's internal audit function by public managers is needed as the urge for accountability and transparency increases (Diamond, 2002).In this case, the function of APIP to monitor the internal control system could provide an early warning system to prevent and detect irregularities and misuse (Jeppesen, 2019) and decrease information asymmetry between government and citizens (Hay & Cordery, 2018).This argument is also echoed by the Association of Certified Fraud Examiners (2022), which highlights that internal auditors are able to detect corruption more frequently than external auditors.Saptono & Purwanto (2022) also found that the lack of quality and capability of the internal audit is still an obstacle in the policy of eradicating corruption, especially in state-owned enterprises.Thus, it can be said that the role of APIP, which provides supervisory and oversight toward the government, became an important element of the anti-corruption agenda in Indonesia.
The critical role of APIP in combating corruption is acknowledged by the government.Strengthening the capacity of APIP became one objective of the National Anti-Corruption Strategy in 2018.Ministry of State Apparatus Utilization and Bureaucratic Reform (Kemenpan-RB) also points out the role of APIP as one component of building an Integrity Zone toward a Corruption Free Area and Clean Serving Bureaucratic (ZI-WBK/WBBM) (Kemenpan-RB, 2021).In order to measure the effectiveness of the implementation of the APIP role, the Financial and Development Supervisory Board (BPKP) has conducted a measurement of the APIP capability level as an indicator of how far APIP has contributed to the achievement of organizational goals.The BPKP Regulation Number 6 of 2015 on the Grand Design for Increasing APIP Capability in 2015-2019 states that the capability level of 85.23% of APIP in 2015 was still at level 1 on a scale of 5 (BPKP, 2015).This indicates that there is a risk that APIP has not been able to carry out its role and provide added value for the organization optimally.Fortunately, this condition has experienced a significant improvement in 2022.According to BPKP (2022), 54.74% of APIP have reached a capability level above 3, 38.2% of APIP are at level 2, and 7.06% of APIP are still at level 1.The comparison of these conditions can be seen in Figure 1.According to BPKP (2021), the APIP at level 3 has been able to carry out supervisory activities in accordance with professional standards and practices, and the results of APIP supervision are already of good quality and provide reasonable assurance of compliance and 3E (economy, efficiency, and effectiveness), early warning and improvement of the effectiveness of risk management and improvement of governance for K/L/D organizations.Based on the data, it can be concluded that most APIPs today (54.74%) have had a fairly effective contribution to the organization.
Nonetheless, this capability level focuses more on the effectiveness aspect of the APIP role and does not analyze further whether the APIP has operated efficiently in achieving the effectiveness of its role.In fact, efficiency, alongside effectiveness, has been the main objective of state finance reform that has been mandated in the law.It is suggested that efficiency has been a core component of government reform.Implementing a reform often puts an advanced bureaucracy as a focus to 0.00% 10.00% 20.00% 30.00% 40.00% 50.00% 60.00% 70.00% 80.00% 90.00% 100.00% improve service delivery, which requires the government to be more efficient (Liou, 2017).Furthermore, higher government efficiency is necessary to give fiscal space to fulfill the increasing need for achieving sustainability (Baum et al., 2017).Numerous studies discuss the role of APIP in supporting efficiency.In the context of listed public companies in Indonesia, Ardianto et al. (2023) found that the internal audit function only has a minor role in giving value-added service regarding investment decisions.Meanwhile, in the Polish public sector, Postula et al., (2020) found that the efficiency of internal audits plays a significant role in streamlining operations and adding value to the functioning of public administration in Poland regarding efficient public spending and public entities' completion of public service tasks.However, to the authors' best knowledge, the systematic empirical study that discusses the efficiency of APIP in Indonesia has never been concluded.It is paradoxical since APIP measures the effectiveness and efficiency of other units while APIP's efficiency itself is not much explored.For this reason, this study will analyze the efficiency of the internal control function by APIP K/L/D in order to determine which APIP has operated efficiently.
The study of efficiency measurement for APIP is arguably important.The highest level of internal audit capability model inspires APIP to be a key change agent and catalyst for the organization (Internal Audit Foundation, 2017).In other words, APIP acts as a role model for other units to conduct effective and efficient business.In addition, according to Presidential Regulation Number 54 of 2018 concerning the National Anti-Corruption Strategy, one of the challenges faced in the aspects of law enforcement and bureaucratic reform is the persistent lack of independence, oversight, and internal control within government institutions, including ministries, agencies, and regional and local governments (K/L/D).Thus, the study of APIP's efficiency could bring valuable insight for understanding the efficiency level among APIPs and use the most efficient APIP as a benchmark for other APIPs.Considering the importance of APIPs in preventing and detecting corruption, understanding the efficiency level is critical for the government to formulate a better policy for optimizing the resources of APIPs to enhance their capabilities in combating corruption in Indonesia.

Methods
The concept of efficiency permeates various disciplines since its measurement is crucial for understanding performance and identifying areas for improvement.Data Envelopment Analysis (DEA) has emerged as a powerful tool for quantifying efficiency, particularly in situations with multiple inputs and outputs.Emrouznejad & Yang (2018) list 9,881 DEA articles and research papers highlighting the use of DEA analysis from 1978 until the end of 2016.The top five application fields of DEA with the most journal articles in 2015 and 2016 include agriculture, banking, supply chain, transportation, and public policy.Charnes et al. (1978) introduced DEA as a non-parametric frontier-based approach.It constructs an efficient frontier based on the observed data of decision-making units (DMUs), representing best practices.Each DMU is evaluated against this frontier, revealing its relative efficiency score, in which a score of 1 indicates full efficiency and scores below 1 imply inefficiency.This model was then known as DEA-CCR.It holds the assumption of constant return to scale, which means that if all inputs are proportionally increased, outputs will also increase proportionally.Later, Banker et al. (1984) developed the CCR model into the BCC model.It holds variable returns to scale assumption, which allows for different efficiency patterns at different output levels.The CCR model is preferred when DMUs are likely homogenous in terms of scale, while the BCC model is preferred when DMUs operate at different scales.Furthermore, Pradeep et al. (2020) concluded situations where CCR and BCC are applicable.The basic DEA approach using the CCR model is applied to the unitary evaluation of homogeneous units (rather than organizations).They also strongly recommend applying the BCC model for evaluating the efficiency of the public sector and private sector services businesses.This study will use the BCC model to analyze the efficiency level.
The main difficulty in any application of DEA is in the selection of inputs and outputs.The criteria for the selection of these inputs and outputs are quite subjective.There is no specific rule In the input categories, the number of employees and goods expenditure budget ratios are used to handle the difference in the size of operations between APIPs.It follows Benazić (2012), who utilizes the number of employees and the organization's cost as an input to measure the efficiency of customs service in Croatia.It also follows Kruger et al. (2002), who used labor and monetary value (costs) as input variables to analyze determinants of internal audit efficiency.It also aligns with the Ministry of Home Affairs' regulations (2019)(2020)(2021)(2022)(2023), which use a ratio from total region expenditures in the local government budget that ranges between 0.3%-0.9% to calculate the supervision budget.
In the output categories, the variables are selected following Article 11 PP Number 60 of 2008 that states the following: the realization of the effective role of APIP must at least (1) provide adequate assurance of compliance, economy, efficiency, and effectiveness in achieving the objectives of the implementation of the tasks and functions of government agencies; (2) provide early warning and improve the effectiveness of risk management in the implementation of the tasks and functions of government agencies; and (3) maintain and improve the quality of governance in the implementation of the tasks and functions of government agencies.Thus, we use the bureaucratic reform, risk management, and integrity perception scores of the K/L/D to measure the effectiveness of APIPs.Those measures follow the expected contribution of APIPs mentioned in the law but also represent the outcome of anti-corruption efforts of K/L/D supported by APIP.
The bureaucracy reform is chosen as the output since it enhances the anti-corruption agenda due to the improvement of services from the government (Yustia & Arifin, 2023).The level of bureaucracy reforms is measured using the bureaucratic reform index initiated by Kemenpan-RB.The index is obtained from an assessment model based on Total Quality Management principles and is used as a method for conducting a comprehensive assessment and analysis of the performance of government agencies.The index varies on a scale of 1 to 100 and is obtained through self-assessment by K/L/D and then evaluated by the Ministry of State Apparatus Empowerment and Bureaucratic Reform (Kemenpan-RB, 2020).
Risk management is used as the output since it substantially influences the effort on fraud prevention and detection (Tarjo et al., 2022).The implementation of effective risk management may diagnose the potential of fraud and improve the control to prevent it (Bento et al., 2018).In this study, risk management is measured by the score of the risk management index that describes the quality of risk management implementation within the K/L/D scope obtained from calculating risk management assessment parameters.In the MRI assessment model, assessment parameters are grouped into three main components, namely planning, capability, and results.The index is obtained through self-assessment by management, quality assurance by APIP, and then evaluated by BPKP (BPKP, 2021).
The K/L/D integrity perception survey is used to assess corruption risk in K/L/D.The assessment comes from the perceptions and experiences of stakeholders in K/L/D, namely employees, service users and collaboration partners, and experts from various groups.The assessment includes transparency, integrity in carrying out duties, trading in influence, budget management, procurement, human resource management, and anti-corruption outreach in each agency.The survey results are in the form of numbers, which indicate the level of agency integrity on a scale of 1 to 100.The higher the integrity score of an agency, the better the system in place to detect corruption risks and handle corruption crimes in the K/L/D.(Yudha et al., 2021) After the efficiency analysis using DEA, we extend our analysis by mapping out the level of efficiency and capability of APIPs using quadrant analysis.The analysis will divide each APIP into four quadrants based on the capabilities and efficiency score.Quadrant 1 reflects APIPs who are efficient and effective.Quadrant 2 means that APIPs achieve their effectiveness but are not quite efficient.Quadrant 3 means that APIPs are quite efficient but not effective, and Quadrant 4 places APIPs that are not efficient or effective.

Results and Discussion
This study uses five variables, including two inputs and three outputs, with 50 APIPs as the number of DMUs.Ramanathan (2003) suggests that the number of DMUs should be at least two or three times larger than the number of inputs and outputs.This is because DEA is a non-parametric technique that relies on the number of DMUs to estimate the efficient frontier.The larger the number of DMUs, the more accurate the estimated frontier will be.Avkiran (2001) goes a step further and suggests that the number of DMUs should be larger than the product of the number of inputs and outputs.This is because it is possible for two DMUs to have the same number of inputs and outputs but have different levels of efficiency.By having more DMUs than the product of the number of inputs and outputs, it is more likely that the efficient frontier will be able to distinguish between these two DMUs.In this study, the number of DMUs ( 50) is more than the minimum suggested by Ramanathan (>18), and it is also larger than the product of the number of inputs and outputs (>6).The descriptive statistics for the input and output data can be seen in Table 3.The results of the analysis of the efficiency of 50 APIPs are shown in Table 4. Based on the test results, either with input or output oriented, there are 7 of 50 DMUs (14%) that get efficient results (efficiency score =1), and the remaining 43 DMUs are relatively inefficient.The average efficiency of entire DMUs is 0.42747 with input-oriented and 1.147 with output-oriented.This score indicates that there is still room for improvement.Reference groups are naturally generated as outputs of DEA algorithms and represent efficient DMUs.If a DMU is efficient, it does not have a reference group.If a DMU is inefficient, then it has a reference group including at least one efficient DMU.For inefficient DMUs to become efficient DMUs, reference groups must be benchmarked to reduce or increase their input or output (Park & Kim, 2019).Table 5 shows the lambda (λ) value of the input-based BCC model.The lambda (λ) value is the weight of the reference DMU at which the efficiency frontier of the inefficient focal DMU is represented by the ratio of the reference DMU.Reference group DMUs are relatively close to the efficiency frontier of the focal DMU, and therefore, the reference DMUs have similar characteristics to the focal DMU.For example, DMU1 (Ministry of Religious Affairs), DMU3 (Ministry of Finance), DMU26 (Jakarta Province), and DMU28 (Central Java Province) are the reference groups for DMU2 (Ministry of Law and Human Rights.).The reference frequency of a DMU is the number of times it is used for the evaluation of other DMUs.After obtaining the efficiency test results and based on APIP capability score data in Appendix A, we conduct a quadrant analysis to classify the DMUs into four quadrants.As shown in Figure 2, the samples fall into three categories.The DMUs in category I are divided into two subcategories where seven DMUs have high capabilities and could perform their duty relatively efficiently with an efficiency score = 1 (category I.1) and another seven DMUs with an efficiency score above 0.5 (category I.2).In Category II, DMUs have high capabilities with low-efficiency scores (< 0.5) relative to other DMUs.In the third category, DMUs have low capability and low efficiency scores (< 0.5) relative to other DMUs.The detailed results and list of DMUs' categories are shown in Appendix B. This result indicates that even though APIPs are at the same capability level, they vary greatly in the use of resources in carrying out their roles.Category I.I is filled by seven relatively efficient APIPs (efficiency score = 1), namely DMU 1,3,4,13,26,28,and 48.The data shows that the input utilization is relatively low, and output attainment in these DMUs is relatively higher compared to the average of others.This comparison can be seen in Table 6.
As shown in Table 6, the mean of all the input and output for the category II and III DMUs are below the data mean.It shows that the use of inputs is relatively high while the attainment of the outputs in these DMUs is relatively lower compared to the DMUs in categories I.1 and I.2.It is indicated that the difference lies in variations in work processes and methodologies of APIPs which can vary across all stages of the supervision process, starting from the planning stage to determin-

High Capabilities
Category II 35 DMUs

Category I 14 DMUs
Category III 1 DMU Category IV ing the objects, the formulation of work programs, the implementation, and the reporting of oversight results.In determining oversight objects, it should ideally be based on risk, considering the human resources and budgets (Le et al., 2022) and the implementation of oversight that can utilize information technology to optimize the supervision process and minimize the input (Keskinen & Tarwireyi, n.d.;Shaikh et al., 2018).These results convinced the necessity for inclusivity in the execution of oversight roles by APIPs, particularly those situated within Quadrant I.The role of professional organizations of APIPs (AAIPI) and BPKP as their overseeing body is also anticipated to accelerate the enhancement of APIPs.

Conclusion
The important role of APIP is acknowledged by the government by putting APIPs as the focal point in maintaining internal control as well as in preventing and detecting corruption.Unlike any other agencies, APIP in Indonesia's governmental system exists in all agencies from the central to the regional level, which indicates the fundamental role of APIP in the organization.However, this study concludes that the efficiency level of APIPs in Indonesia is still low.Using Data Envelopment Analysis (DEA) and purposive sampling of 50 APIPs at the central and regional levels, this study finds only seven APIPs (14%) achieved a maximum efficiency level with a score of 1.Those are APIPs in the Ministry of Religious Affairs, Ministry of Finance, Ministry of Education and Culture, National Family Planning Coordinating Board, the Jakarta Province, the Central Java Province, and the Boyolali Regency.Furthermore, this study also set out those APIPs as a reference group for other APIPs to carry out a benchmark based on the similarity of inputs and outputs.For example, the APIP of the Ministry of Law and Human Rights, whose efficiency score is not optimal, could learn from the APIP of the Ministry of Religious Affairs, Ministry of Finance, Jakarta Province, and Central Java Province.Moreover, this study also analyzes the efficiency level of APIPs and their capability.It finds that 14 APIPs are relatively efficient and capable, while 35 APIPs are capable but not relatively efficient, and one APIP is not capable or efficient.It is suggested that APIP's approach to planning, executing, and reporting processes may influence the level of its efficiency.
One limitation of this study is the result only covers the year 2021 and 50 APIPs due to the availability of data.Some data related to the number of APIP employees and the realization of goods expenditure are not publicly available.The absence of this information may indicate that the level of transparency, one important element for good governance, is still inadequate among government institutions.It causes the comparison of efficiency between APIPs at the central and regional levels to not be reliably concluded.Furthermore, the input used in the analysis only depicts the number but does not define its quality, such as the quality of government spending and the quality of human resources.The measurement of output is also taken directly from APIP's mandate in PP 60/2008, which may be biased due to external factors besides APIP's role.Thus, future research could be improved by extending the sample of APIPs and periods covered, as well as considering the quality of inputs and the measurement of outputs that can be isolated from other external factors.Despite its limitations, it could be argued that this study will enrich the theoretical framework regarding APIP's efficiency, which is still relatively unexplored.This study also gives a practical implication for the government to put serious effort into improving the efficiency level of APIPs to make the available resources allocated for APIPs optimally utilized to support good governance and clean government in Indonesia.

Figure 2 .
Figure 2. Quadrant Analysis for Capability and Efficiency of APIP

Table 1 .
List of DMUs

Table 2 .
Data Types and Sources The Corruption Eradication Commission (jaga.id)

Table 3 .
Descriptive statistics for the input and output data

Table 4 .
Efficiency Analysis Results and Summary

Table 5 .
The reference DMUs and lambda values of the BCC model

Table 6 .
Input and Output Data in Each Categories