Abstract
Few cross-country studies examine the effects of auditing quality on public sector corruption. We present a definition of good auditing consisting of three principles: independence, professionalism, and recognizing the people as the principal. Using novel data from an original expert survey covering more than 100 countries, the concept is then operationalized and tested empirically. The results demonstrate that good auditing has a positive effect on national levels of public sector corruption. This lends reason to believe that auditing which is organized according to certain principles has potential to contribute to well-functioning public administrations with a low degree of corruption.
Introduction
Understanding how to reduce public sector corruption—generally defined as “the abuse of public power for private gain” (Transparency International [TI], 2011)—is crucial for policymakers that seek to get rid of such behavior, and it has been described as the “million dollar question” facing public administration scholars (Rothstein, 2011). The comparative research on public sector corruption prevention has examined a range of institutional as well as cultural explanatory factors across countries (see Treisman, 2007). Seminal work on how to deter public agents to engage in corrupt behavior propose that the risk of getting caught is of crucial importance (cf. Becker & Stigler, 1974), and in modern societies, auditing activities is the main tool for detecting such behavior in public administrations. Yet, the departure for this article is that comparative research seldom has focused on the quality of auditing of the public sector as a mean to reduce corrupt behavior among bureaucratic agents. We therefore ask the following: Can well-functioning national auditing of the public sector reduce such behavior?
Scholars have studied the shift from central regulation and detailed instructions to decentralization, market solutions, and ex-post controls within the New Public Management (NPM) reform package. This change increased the demand for oversight and inspection of the public sector to such an extent that Power (1999) claims that we were facing an “audit society.” Power argued that this would have undesirable consequences, and in line with his view, much of the literature has focused on the harmful impact of public sector audit activities (Lonsdale, Wilkins, & Ling, 2011; Pollitt et al., 1999). 1 However, not all scholarly contributions adapt a pessimistic view on auditing and few have conducted empirical studies to demonstrate the impact of audit activities on public sector outcomes (Andrews et al., 2008; Boyne, 2003; Cabral & Lazzarini, 2015; Reichborn-Kjennerud & Johnsen, 2015; Walker, Boyne, & Brewer, 2010).
Notably, a feature of the literature studying the impact from auditing on public sector outcomes is that very few studies focus on corruption per se. Among the exceptions are two intervention studies by Duflo, Hanna, and Ryan (2012; Duflo, Greenstone, Pande, & Ryan, 2013) that show through field experiments that auditing and monitoring of third-party agents can reduce misbehavior related to corruption by public agents. More commonly, this vein of research often study outcomes that operationalize the concept of public sector performance (Walker et al., 2010). These studies depict a mixed picture of how auditing activities impact the performance of public sector organizations (Bevan & Hood, 2006; Boyne, 2003; Carlson, Cowen, & Fleming, 2014; Rutherford, 2014). Yet, the findings regarding this relationship have further nuances. For instance, Rutherford (2014) finds that audit interventions increased schools’ performance on some indicators while there were no improvements on other performance indicators. 2 These patterns relate to the importance of how public sector performance is defined and measured. As Andersen, Boesen, and Pedersen (2016) discusses in a recent conceptual article, the concept of performance is so broad that researchers should be as specific as possible when using this terminology. To be clear, we relate our inquiry to insights from research on auditing and public sector performance. However, while we see the degree of corruption in the public sector as a feature that may have further ramifications for different types of performance (as bribery often induce implementation failures and hampers bureaucratic efficiency), we do not view corruption as a type of public sector performance as such.
Almost all countries around the world have an established supreme audit institution (SAI), defined as “national audit agencies responsible for auditing government revenue and spending” (World Bank, 2001). However, the organization of these SAI’s vary in several aspects, such as their structure, degree of professionalism, degree of independence from the government, and whether they communicate their results to the public. Previous research emphasizes some of these aspects as important in relation to efforts to improve different performance measures in the public sector. Yet research lack an overall theoretical understanding of how auditing should be organized to generate lower degrees of public sector corruption and has not extensively subjected such reasoning to systematic empirical testing (Boyne, Day, & Walker, 2002; Cabral & Lazzarini, 2015; Walker et al., 2010). 3 The specific focus of this study is on SAI’s in a comparative perspective and national levels of public sector corruption. This study aims to expand this field of research through examining how “good auditing” impacts public sector corruption in a cross-national comparative study.
The remainder of the article is structured as follows: The “Theory: The Organization of Auditing and Its Impact on Public Sector Corruption” section discusses prior research, conceptualizes good auditing, and clarifies our theoretical expectations. The “Study Design and Data” section presents the data and methods. The “Results” section reports the results, and the “Concluding Discussion” section concludes.
Theory: The Organization of Auditing and Its Impact on Public Sector Corruption
The Importance of Auditing of the Public Sector
Oversight of the public sector is commonly described theoretically as following the logic of the principal-agent problem (Bendor, Glazer, & Hammond, 2001; McCubbins, Noll, & Weingast, 1987; Weingast, 1984). As the principals (the political representatives and, ultimately, the people) delegate the execution of public affairs to the agents (the public officials), the need to control the public officials’ actions and performance increases. If there were no oversight of public officials, there would be an impending risk of policy drift, shirking (i.e., the avoidance of duties), or corruption in public authorities (McCubbins et al., 1987). This need for control is enforced by the asymmetric relationship between the principals and agents in terms of knowledge. Particularly in modern societies with highly complex and specialized administrations, political representatives and ordinary citizens may have a difficult time constituting sufficient control merely through access and insight into the public sector organizations. This has given rise to the need for specific government agencies to control other parts of the public sector, such as auditing or inspection agencies (O’Donnell, 1999; Scott, 2000).
Research on auditing and inspection of organizations has demonstrated mixed results, with some studies showing a positive impact and others a negative impact or no impact at all. These studies have mainly been centered on how the communication of audit results can play an important role. Audit results have proved to have a positive impact on organizations’ performance (measured in a diverse array of outcome variables) when they have been made public to citizens (Carlson et al., 2014; Bevan & Hood, 2006; Ferraz & Finan, 2008). In particular, such information has proved to be significant in decentralized public sectors with market solutions, where citizens are allowed to choose among various service providers. Citizens need information on which to base their choices, and public service providers need to perform well in published reports to attract citizens as “customers” (Carlson et al., 2014). The source of performance information has also proved to be important. Research holds that citizens are more likely to trust information produced by independent auditing bodies than they are to trust information produced by other actors (James, 2011). Furthermore, the capacity of the agency, in terms of the auditors’ expertise and skills, has been demonstrated to affect its ability to fulfill its assignment. In developing countries in particular, capacity constraints in terms of lack of education and expertise among auditors place strict limitations on how auditing agencies in these countries operate (Gustavson, 2014; Isaksson & Bigsten, 2012). A reason for the mixed results could be the lack of an overall theoretical framework establishing what may constitute “good auditing” of the public sector. Just as it is not certain that “governance” alone will lead to any positive outcomes in a country, neither does “auditing” alone ensure that an administration is monitored in a way that will improve its operations. A similar discussion on what constitutes good auditing has been lacking.
Apart from the overarching understanding of the need for and purpose of auditing agencies in democracies, there are few established comprehensive theoretical frameworks for understanding how such agencies should be organized to achieve better outcomes in general and reduce the degree of corruption in the administrations being audited in particular (Boyne et al., 2002). In the following section, the features of good auditing will be reviewed and discussed in the light of historical political theory to outline a definition of good auditing and anchor it to democratic accountability. We then explicate why we expect these features of good auditing to reduce public sector corruption.
Conceptualizing Good Auditing
We argue that there are three main principles that can be regarded as essential elements in a definition of good auditing, based on a democratic perspective. The first is the principle of independence (Flint, 1988; Hollingsworth, White, & Harden, 1998; Mautz & Sharaf, 1961; Normanton, 1966; Power, 1999, 2005). If we regard auditing as a mechanism for the principal to control the agent, the separation between the agent and the auditing agency becomes central. If there is no separation between the oversight mechanism and the public administration that is subjected to control, the oversight mechanism would instead work primarily as a self-evaluation function (Wildavsky, 1979). Internal auditing serves as such a mechanism, operating as an internal control function for management to evaluate and control the organization. Although internal auditing may be a valuable instrument for management in the public sector, for politicians and the public to enhance efficiency and prevent policy drift, fraud, and corruption in the administration relying on self-evaluative mechanisms has obvious limitations. For example, there is a risk that self-estimations will overvalue performance and downplay underperformance and other problems in the organization due to management’s desire to present a picture of success (see Meier, Winter, O’Toole, Favero, & Andersen, 2015). The history of auditing in democratic theory also illustrates how independence has been an essential element in the relationship between auditors and public administrations. In The Politics, Aristotle (1996) states that officials handling public money must be controlled by other officials who are separate from them and have no other function: “But since some, not to say all, of these offices handle the public money, there must of necessity be another office that examines and audits them, and has no other function” (1322b5-15). Similar to Aristotle, John Stuart Mill (1861/2001) clarifies the need to create systems to control the administration and to separate “these offices of control and criticism” from the administration whose work they are to examine (p. 70). As independence is well established as a cornerstone in auditing, the issue discussed in the literature is not whether or not independence is necessary, but rather how various dimensions of independence can be guaranteed, such as financial independence; protection from interference in planning, selecting, and executing audits; and individual ethics among auditors (Antle, 1984; Bayou, Reinstein, & Williams, 2011; Cullinan & Sutton, 2002; Gendron, Cooper, & Townely, 2000; Preston, Cooper, Scarbrough, & Chilton, 1995; Sikka & Willmott, 1995; International Organization of Supreme Audit Institutions [INTOSAI], 1998).
The second principle of good auditing, this article argues, is professionalism. The creation of specific auditing agencies builds on the notion that officials working for these agencies are better able than ordinary citizens and politicians to control financial transactions and the ethical conduct of agents in very complex organizations such as contemporary public administrations. Scholars also argue that professionalism among officials conducting oversight activities generates better outcomes in terms of improving performance of the auditees (Boyne et al., 2002; Gustavson, 2014; Isaksson & Bigsten, 2012). In addition to controlling the public sector, auditing also builds on recommendations to the auditee with the aim of improving its performance (Reichborn-Kjennerud, 2013). For their recommendations to actually lead to improvements, the auditors’ understanding of these organizations must be significant (Boyne et al., 2002, p. 1199). In comparison with other oversight agencies, constituting a specific profession also has implications for how the expertise among auditors is governed (for a recent discussion drawing on experiences from Nordic countries, see Jeppesen et al., 2017). A profession is defined by its claim to abstract knowledge and the application of such knowledge to particular cases (Abbott, 1988, p. 8). To limit who can claim legitimate membership in the profession, professional groups exercise control and authority in several ways. Entrance to the profession is generally regulated through demands for specific education and exams; furthermore, apart from national legal frameworks, members’ work is internally regulated through requirements that they follow professional norms such as specific standards and guidelines (Bédard, 1989; Byington, Sutton, & Munter, 1990; Preston et al., 1995). Hence, a part of the trust in auditors’ competence is also based on whether they follow the norms established by the profession for their work (Gustavson & Rothstein, 2013). In a democratic perspective, the relationship between democracy and professionalism is not entirely clear. The historical political theories discussing a function such as audit are not fully consistent in their approach to professionalism. The democratic system promoted by Aristotle was an amateur system based on a rotation of positions between the public officials, in order to be truly democratic. 4 More recent democracy and public administration theorists such as Max Weber (1922/1978) nonetheless argue that the need for professionalism in an audit function is essential due to the asymmetric relationship in information and knowledge between a specialized administration and the people and politicians. Mill also emphasized the importance of using highly skilled people in the administration. In addition, he argued that they should carry out the work in an impersonal manner, according to specific procedures (Urbinati, 2002, pp. 54-55).
The third and last principle that we argue constitutes the foundation of good auditing is recognizing the people as the principal. The organization of public administrations in democratic societies can be described as a chain of delegation. The people delegate authority to elected politicians who in turn delegate power to various public agencies, their management, and the individual public officials. Who should be considered the principal in this vertical accountability framework then depends on where we focus in the chain of delegation (Brandsma & Schillemans, 2013). Although delegation builds on a shift in thinking about who should be considered the principal, from a democratic perspective, it can be argued that the main principal, whose will and power are delegated, is always ultimately the people. The question then becomes how auditing can be organized to enable the people to become a true principal, for whom auditing agencies hold the public sector organizations to account. The literature pinpoints that communication of the findings from audits is a crucial aspect of this accountability process. Audit reports and the information produced by independent bodies like auditing agencies can be used directly by the citizens, creating a direct vertical accountability link between the auditing agency and the people. People are affected by, and may act upon, information on public sector performance generated by external audit agencies. For instance, they might change service provider if the audit reports present negative reviews of their current service provider’s performance or they may voice concerns for political change (Carlson et al., 2014; James, 2011). Communication of audit results directly to the people may hence lead to higher levels of political accountability.
Good Auditing and Reduced Public Sector Corruption: Theoretical Expectations
In a wide sense, it is quite plausible to assume that actors in public authorities will have the incentives to limit corrupt practices if the risk of being exposed is a real threat (cf. Becker & Stigler, 1974). If auditing is truly effective, it will hence have a deterring effect on the behavior of agents in such authorities. As discussed above, three main principles of what may be considered good auditing, from a democratic perspective, have been outlined and the following definition is proposed: Good auditing is characterized by independence from the auditee, professionalism in the exercise of the audit practice, and recognizing the people as the principal. The overall theoretical understanding is thus that these principles together organize auditing in a way that generates a public sector with less corruption; that is, countries where national auditing agencies is organized according to these principles also will tend to have little public sector corruption. This leads to the following hypothesis:
Although the overall theoretical understanding of what characterizes good auditing builds on the three principles in combination, it is reasonable to assume that each one of the principles has an independent effect on how well auditing can generate a public sector without corruption. There are specific reasons to believe that the components of this concept have an effect on reduced public sector corruption. To start with, previous discussions lend reason to believe that independence matters. Two main systems for organizing SAI’s are discussed in the literature. In some countries, for instance, France and the Netherlands, the SAI serve as courts wherein the public entities may face penalties if irregularities are detected (Bundt, 2000). In such accountability arrangements, the auditors acquire a great deal of discretion and have the final say in how the administrations perform, with limited participation from the people. In other countries, such as Sweden and Great Britain, auditing instead builds on recommendations from the auditors to elected politicians and the administrative management in the audited entity, which then are responsible for implementing the auditors’ recommendations (Pollitt & Summa, 1997). In such an accountability arrangement, it mainly becomes the responsibility of the political leaders and the administrative management to act on the audit results on behalf of the citizens. If the contemporary juridical audit system has limited democratic influence, the audit system building on recommendations to administrative management and political leadership has also proven to be inadequate from a democratic perspective. Scholars have argued that auditors are frequently too closely aligned with administrative management to report sufficiently on maladministration (Sikka & Willmott, 1995). In particular, it has been pointed out that the advisory role of auditors may conflict with their role as reviewers of the organization (Cullinan & Sutton, 2002; Sikka, 2009). For government audit agencies, independence is a cornerstone, and if they become “too cozy” in their relationship with the audited entities, there is a risk that they will no longer be perceived to be independent (Pollitt & Summa, 1997, p. 334). To illustrate, independence has been characterized by Barzelay (1997) as “an essential attribute of performance auditing” (p. 241). In addition, studies have demonstrated that when auditors do report on mismanagement and recommend changes, reports and recommendations are not always taken into account by the administrative management and the political leadership (Freides, 1992; Hanberger, 2009). As noted in our conceptual discussion, an auditing which is not independent may run the risk of becoming a self-evaluation of public officials and therefore may downplay corrupt behavior. Building on this argument, we conclude that it is reasonable to assume that if the auditing agency is independent, it will be better able to monitor the public sector in a manner that will reduce corruption. This leads to the following hypothesis:
Likewise, we find reason to believe that if the auditors who work in the SAI have the appropriate skills and education, they will be able to conduct the audits in a way that reduces public sector corruption. As noted above, theory suggests that professionalism should characterize such organizations. Weber (1922/1978) argues that officials need to be held to account; moreover, he argues that it should be experts who are tasked to hold other experts to account, as leaders generally lack the knowledge needed to be able to check the experts’ work (p. 236). We find it likely that this also should hold true for the staff in audit organization in regard to anticorruption investigations. To illustrate; if a SAI do not have the expertise to identify acts of misbehavior for personal gain in complex organizations—acts which most often is hidden in paper trails so that outsiders do not easily find out about them—its work would be quite toothless as a tool to reduce public sector corruption. Moreover, it can be noted that a lack of professional expertise creates more dependency on the auditee and it limits possibilities to take a critical stance toward the information obtained in the audit process (Isaksson & Bigsten, 2012); it also has a negative impact on the trustworthiness of the auditors’ assessments, opinions, and reports (Flint, 1988). It is therefore plausible to assume that SAI’s need to be staffed with skilled employees given the appropriate resources to conduct meaningful investigations of fraud and bribery in public authorities. In other words, a SAI which is characterized by professionalism should be much more likely to expose corruption than one that is not, everything else alike. To conclude, this leads to the following hypothesis:
Finally, if the SAI in a country communicate the results from audits to the public regularly, resulting public pressure will tend to lead to reduced public sector corruption. To illustrate, Andersson and Bergman (2009) show in their study that the significance the administrative management attaches to audits can influence its ability to protect the administration from mismanagement. Weak audits and limited attention to audit reports constituted part of their explanation for the higher level of corruption found in one out of two otherwise similar regions in Sweden (Andersson & Bergman, 2009). A closely related and important issue regards communication of audits and accountability structures. Ferraz and Finan (2008) illustrate how audit reports can significantly impact the way citizens hold political leaders accountable in elections. As part of an anticorruption strategy, the Brazilian state conducted an extended audit of randomly selected municipalities. The results from the audit were published on the Internet and handed over to the local media. The audit reports on the prevalence of corruption in some municipalities greatly influenced whether or not the local politicians were reelected. Politicians who were reported as corrupt in the audits were much less likely to be reappointed, whereas those reported as “clean” were much more likely to be reelected (Ferraz & Finan, 2008). As we can see, when audit results are communicated directly to the people, accountability increases and the chances of maladministration proceeding undetected diminish. Of course it is not possible to inform the general public of all audit findings, but if the people are to be recognized as the principal, auditing agencies need to seriously consider how best to handle audit findings with a view to serving the public interest. This entails communicating audit results to the public, including those that are inconvenient for the government. This reasoning informs us to formulate the final theoretical expectation in this article and leads to the following hypothesis:
Study Design and Data
To assess the impact from good auditing on public sector corruption, this study conducts an empirical investigation with a cross-sectional design focusing on a global sample of countries. The following sections will discuss the operationalizations, methods, and data used in this study.
Operationalizing the Dependent Variable
The dependent variable in this study is the degree of corruption in the public sector of a country. To measure corruption, we use figures from the TI’s Corruption Perceptions Index (CPI). The CPI is a perception-based expert assessment, which combines investigations performed by a number of established institutions, such as the World Bank, the Economist Intelligence Unit, Freedom House, and the Bertelsmann Foundation. The index combines these investigations to a single indicator of the extent to which bribery and corruption is pervasive in the public sector of a country. The CPI is reported as an annual figure for a global sample of countries and a nation is assigned a value ranked from 0 (high corrupt) to 100 (low corrupt). We make use of the CPI from the year 2015. As will be discussed in the “Results” section, we also perform the analyses in this article by using a different measure of corruption. As a mean of robustness test, we run the models with Control of Corruption data from the World Bank Governance Indicators as an alternative dependent variable (see Kaufmann et al., 2009). Both variables are taken from the Quality of Government Institute’s cross-sectional database (Teorell et al., 2016).
Operationalizing the Independent Variable
To measure the main independent variable in focus, good auditing, we use original data from the Quality of Government (QoG) Institute Expert Survey II. The survey was launched by researchers at the QoG Institute at the University of Gothenburg, Sweden, in 2014, and is based on responses from 1,294 country experts, covering 159 countries. These experts were purposively recruited in four rounds during 2014 through the network of the organizing researchers and their involvement in associations such as the United Nations Public Administration Network. The survey was administered per email and from a pool of 7,096 experts; 1,784 started the survey; and 1,294 finished it. The survey was available in English, French, Spanish as well as Russian, and its items were pretested on a sample of experts that participated in an earlier round of the survey. The participants are more often than not of male sex, with a PhD degree, employed at a public university and are born or live in the country that they are assessing. 5 Experts participated in the survey on a voluntary basis with no remuneration for their involvement (Dahlström et al., 2015b). 6 The 122 countries that are represented by three or more experts in the survey are used in this dataset. While a reduced number of countries may slightly skew the sample geographically, we are informed by the writings of Lee and colleagues (2012) and deem it important not to rely on information from single experts when analyzing the answers pertaining to a country. On average, roughly eight experts assess a country and the responses have been subject to tests for experts’ perception bias (see also Dahlström et al., 2015a). The dataset has been validated in previous publications and earlier rounds of this data collection—the QoG Institute Expert Survey I—have been reported in a number of articles in political science and public administration journals (e.g., Dahlström, Lapuente, & Teorell, 2012; Sundell, 2014).
Good auditing is measured with three specific items in the expert survey, capturing each one of the three principles outlined in the theory section. The three items follows after the question “To what extent would you say the following applies today to the country for which you have chosen to submit your answers?” Respondents are then asked to assess the following three dimensions according to this question: (a) the National Audit Office is independent of the government; (b) auditors at the National Audit Office have the appropriate education and qualifications; and (c) the National Audit Office regularly communicates their results, including results that may be inconvenient for the government, to the general public. Apart from a “no answer” option, the answer options range from 1 to 7, where 1 represents “not at all” and 7 “to a very large extent.” The answers were aggregated to the country level by taking the mean of all experts per country. The good auditing index was then created by taking the mean of the values for the three questions (Cronbach’s α = .939).
Operationalizing Control Variables
To ensure that the relationship between good auditing and public sector corruption is robust to other features that may explain variance in the degree of corruption across countries, we make use of a range of control variables. These variables have been shown in cross-country comparative studies to be factors contributing to the degree of corruption in the public administration of a nation.
Several studies pinpoint that the economic development in a country is a strong predictor for corruption in a country (see an overview in Treisman, 2007). To control for economic development, the gross domestic product (GDP) per capita in current prices by Gleditsch is used (year 2011). International integration has also proved to be a significant factor leading to lower levels of corruption in the public sector of a country (Sandholtz & Gray, 2003). In the models, international integration is specified using an indicator of trade openness, from the World Bank Development Indicators (2012), which is measured by the sum of the country’s exports and imports divided by GDP (World Bank, 2012). A dummy variable for whether or not the country has been a British colony is also included, as scholars have argued that the British left norms and legal structures in their colonies that, after independence, made these countries more likely than other former colonies to have administrations based on the rule of law (La Porta, Lopez-de-Silanes, Shleifer, & Vishny, 1999; Treisman, 2000). For this variable, the article uses the measure of colonial origin created by Hadenius and Teorell (2007). A part of good auditing is the communication of audit findings to the general public. To ensure that the impact from good auditing exists also when taking levels of press freedom in the country in to account, Freedom House’s measure of freedom of the press was added into the models as a control variable. This scale assesses countries on whether their press is considered “free,” “partly free,” or “not free.” The variable was reversed so that high scores indicate high levels of press freedom. Finally, we control for the level of democracy in a country as this factor as been shown to be a strong predictor of corruption in a country (e.g., Bäck & Hadenius, 2008; Charron & Lapuente, 2010). We use the imputed polity scores to measure this concept, a standard measure in this literature. This indicator ranges from 0 to 10, the upper score indicating the highest level of democracy in a country. All control variables are taken from the Quality of Government Institute’s cross-sectional database (Teorell et al., 2016).
Empirical Strategy
Using our cross-sectional dataset, the empirical strategy proceeds in two steps: We first investigate the bivariate relationship between good auditing and public sector corruption. We here utilize ordinary least squares (OLS) regression methods as the dependent variable is expressed as an interval scale. As a range of factors may explain levels of public sector corruption—some of which may be rival explanations to the impact from auditing—we then proceed to investigate the impact from good auditing in a multivariate framework in which we include the control variables discussed above. In the final multivariate models, we have data for all variables available for 104 countries, distributed across the globe. The summary statistics of the variables used in these models are reported in Table 1. It should be noted we made sure to investigate if there is multicollinearity in the models we report. Collinearity diagnostics indicate that the variables are not problematic in terms of tolerance and that variance inflation factors are within secure boundaries.
Descriptive Statistics.
Note. GDP = gross domestic product.
Results
Focusing on the relationship between good auditing and the levels of public sector corruption in a country, we first report the bivariate patterns in Figure 1. In this scatterplot, it is evident that countries with higher scores on the good auditing index also tend to have low degrees of corruption. As expected, the Nordic countries are found at the top right corner, together with countries such as New Zealand, the Netherlands, and France, with high levels of good auditing and well-functioning public administrations. The fact that countries representing different kinds of audit systems—that is, the parliamentarian system (the Nordic countries, New Zealand and Canada) and the juridical system (France and the Netherlands)—are found at the top indicates that it does not matter what kind of audit system is established in the country as long as it is an independent and professional auditing agency that communicates with the general public. At the bottom left corner, we find countries such as Afghanistan, Venezuela, and Bangladesh, countries where corruption in the public sector is perceived as being very high; these countries are also distinguished by not having national public auditing agencies that are organized according to principles of good auditing.
Bivariate relationship between good auditing and lack of public sector corruption.
Proceeding to the results from the OLS regression model, Table 2 reports the findings from this analysis in five different models. Model 1, reporting the bivariate relationship, clearly shows that there is a strong association between the two variables good auditing index and corruption. The b value suggests that there is a substantially large and statistically significant effect from the auditing variable on national levels of public sector corruption. When introducing the set of control variables in Model 2, the magnitude of this effect decreases but is still substantial. The size of the b value in this full model indicates that a 1-point increase in good auditing would decrease corruption by about 4 points on this scale (p < .001). Using the full index as the main independent variable, this lends support for our first hypothesis posed earlier in this article, that good auditing leads to a public administration with less corruption.
Good Auditing and Lack of Public Sector Corruption, Results From OLS Regression (b Values).
Note. Standard errors are reported in parentheses. The dependent variables in all models are the CPI country scores from 2015 (Higher values on this scale indicate less corruption). OLS = ordinary least squares; GDP = gross domestic product; CPI = Corruption Perception Index.
p < .010.
We then proceed to examine the three dimensions that constitute this aggregate index. In accordance with the theory, the three principles of good auditing are also expected to have independent effects on public sector corruption. Hence, in Model 3, the second hypothesis is tested: If the national auditing agency is independent of the government, it generates less public sector corruption. Similarly, Model 4 investigates Hypothesis 3: If the national auditing agency is professional (i.e., has the appropriate skills and education), it generates less public sector corruption. Model 5 finally tests Hypothesis 4: If the national auditing agency regularly communicates its results to the public, it generates less public sector corruption. Running the full regression model with each of the components of the good auditing index also makes it possible to see if any part in the index is more important than the others and is perhaps driving the results for the whole index.
Models 3 to 5 illustrate how each of the principles of good auditing has a clear impact on corruption in the public administration, significant at the 99% level. Yet, the size of the effects in these three models differs slightly. The strongest effect found is that of professionalism (a b value of 5), reported in Model 4 (the variable Component 2: Audit professionalism). This is also the model with the highest level of explained variance. The principles of independence, as reported in Model 3 (the variable Component 1: Audit independence), and communication (the variable Component 3: Audit communication, reported in Model 5) demonstrate b values of around 2.6. The effect from professionalism therefore seems to be somewhat driving the results of the index.
Notably, the above results are reported when including a number of control variables. These factors mainly behave as predicted from theory. As expected, GDP per capita has a positive significant effect. Trade openness does not have any significant effect in the analysis, nor does being a former British colony. More surprisingly, the democracy levels of a country exhibit a negative impact. Pertaining to press freedom, this effect is in line with prior research positive and statistically significant throughout all four models.
It should be noted that we also reran all these five models with an alternative dependent variable that operationalize public sector corruption. By using the Control of Corruption data from the World Bank Governance Indicators as the variable to be explained, we obtain basically identical results: The effect from the good auditing index is positive and when disaggregating the three components of this index, it is the professionalism feature that exhibit the largest substantive effect. These results may be obtained from the authors upon request.
Concluding Discussion
Most countries have a SAI to monitor finances in public authorities and the behavior of officials to ensure that public resources are used without corrupt practices. Yet, from current comparative research, it is unclear whether or not auditing actually leads to a public administration without corruption. Among existing work showing the potential of audits to reduce corrupt behavior in the public sector (e.g., Andersson & Bergman, 2009), few studies adapt a cross-country perspective. In this article, three core principles of good auditing have been suggested as well as operationalized and tested empirically using original data from a novel expert survey with a global scope.
The results from the investigation reported in this article suggest that having authorities of national auditing that is organized according to these principles is associated with low degrees of public sector corruption. Moreover, the data suggest that of the different features of good auditing, the indicator gauging audit professionalism demonstrates the largest effect on corruption levels. These results resonate findings from previous research that stress the importance of skills among the officials conducting controls for the controls to have a positive impact. In developing countries, often facing problems with corruption, it is particularly important to have auditors with appropriate skills and education for the auditing agency to fulfill its assignment (Gustavson, 2014; Isaksson & Bigsten, 2012).
How should one then understand these results in a larger perspective? We do not adhere to a view that installing a SAI is a quick fix for corruption in a country: As most countries are equipped with such institutions, this is indicative that having such a body is not sufficient in itself. Instead, we argue in this article that it is the quality of auditing that matters and we have explicated the different cornerstones of this concept. The contribution of our article is hence to advance the discussion about how to design audit institutions to reduce corruption. Theory does not present a straightforward suggestion to understand if one of the three different components identified in the conceptual discussion of good auditing is more important than the others. For instance, is independence more important feature than the other two? Or is a combination of the three necessary for any reduction of public sector corruption from auditing activities? The empirical results from our inquiry, as discussed above, seem to suggest that professionalism is possibly the more important feature of the three components.
Judging from the expert assessments we analyze, it seems that only a minority of countries manages to live up to the principles of good auditing in practice. There is hence room for improvement, and the main policy implication of the article suggests a need to continue to support national auditing agencies in countries plagued by corruption through various capacity building programs. We here wish to stress that the different aspects of what constitutes good auditing should be guiding the discussion on anticorruption work. In other words, policymakers should investigate how the SAI of high-corrupt countries can be given further independence, increased human resources, and educational skills, as well as the institutional support for communicating audit reports to citizens that may in fact paint a grim picture of behavior of actors in the civil service and alike.
It should be noted that we are aware of some limitations of this study. A significant constraint of our empirical analysis is that it builds on a cross-sectional design, which generally poses some challenges in establishing causality between variables. This refers not only to the problem of endogeneity and reverse causation but also the fact that correlations may in fact illustrate spurious relationships. As such, our research design is not perfect. However, being the first global assessment of the impact from good auditing on levels of public sector corruption, we still think that this approach is worthwhile. Moreover, we believe it is plausible that there is a causal relationship on the degree of public sector corruption from having SAI’s characterized by the principles of good auditing. This assumption is grounded both in theory and previous empirical findings. Theoretically, we find the argument convincing that the degree of public sector corruption should decrease when actors face the potential risks of being exposed through effective inspections from the SAI in a country. Empirically, a limited number of intervention studies implement field experiments in which they are able to isolate the causal factor studied on outcome variables related to corruption. Two notable examples relevant for this project are Duflo et al. (2012) and Duflo et al. (2013). The studies implement two interventions that seek to reduce corrupt behavior. The take-home point from Duflo and colleagues from 2012—focusing on teacher absenteeism—is that the threat of being inspected is crucial. When such public agents know that they are being watched, they act according to incentives—to keep their jobs—and corrupt behavior in this setting is reduced. In the study by Duflo et al. (2013), the focus is on the impact from third-party auditors. The results suggest that the unwanted behavior in focus—factory inspectors that take bribes to be lax in emission controls—is reduced in the group that where randomly targeted by auditors. In these field experiments, the risk of reverse causality is significantly reduced because of the use of treatment groups (with auditing) and controls groups (with no auditing). Hence, we believe that these causal effects illustrate the argument we are making and that it is likely that the causal chain runs from good auditing to low corruption in the public sector. A further limitation of our cross-national approach that deserves mention is that we focus on the aggregate level of countries, both in terms of auditing quality (which may vary between audits) and in terms of corruption (which may vary within the public sectors or regions of a country). However, this we believe is a trade-off that is inherent in comparative research across nations, and we hope that our approach will complement the more fine-grained studies mentioned above.
For future research, we anticipate that other researchers could build on our work and further studies would benefit from addressing the topic of this article with additional theoretical and empirical scrutiny. For theoretical development, we would like to see more work on the necessary conditions of good auditing as discussed above. Possible venues in empirical research to explore are comparisons of different sectors within corrupt countries facing different types of auditing reforms. Another area of empirical research that would be worthwhile exploring is, for instance, studies that measure the impact from public sector auditing reforms before and after their implementation: Designing interventions and creating panel studies with participants in public authorities ridden with corrupt behavior that are facing increased auditing would be an interesting area for additional research.
Footnotes
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: The research was funded by the Swedish Research Council (grant number 421-2011-1398).