World Scientific
  • Search
  •   
Skip main navigation

Cookies Notification

We use cookies on this site to enhance your user experience. By continuing to browse the site, you consent to the use of our cookies. Learn More
×

System Upgrade on Tue, May 28th, 2024 at 2am (EDT)

Existing users will be able to log into the site and access content. However, E-commerce and registration of new users may not be available for up to 12 hours.
For online purchase, please visit us again. Contact us at [email protected] for any enquiries.

Public-Sector Accounting Reforms and Governmental Efficiency: A Two-Stage Approach

    https://doi.org/10.1142/S1094406022500172Cited by:2 (Source: Crossref)

    Abstract

    Synopsis

    The research problem

    We investigated the association between public-sector accounting reforms and efficiency at a central-government level, assessing efficiency through a global perspective.

    Motivation

    The effects of financial management reforms on efficiency have rarely been investigated. We contribute to the academic debate concerning both public-sector accounting reforms (implementation of international accounting standards and accrual-accounting systems) and governmental efficiency, merging two streams of literature that have not been investigated thoroughly.

    The test hypotheses

    H1: Countries that have implemented IPSAS are more efficient.

    H2: Countries that have implemented accrual-accounting systems are more efficient.

    Target population

    We used a sample of 22 European countries in the period 2010–2018.

    Adopted methodology

    We adopted a two-step approach: first, we created several efficiency indicators using different techniques; second, we implemented a model to test our hypotheses.

    Analyses

    Data Envelopment Analysis (DEA) technique, DEA with application of the bootstrap technique, and Order-m model

    Findings

    Accrual accounting was positively associated with efficiency while findings did not totally support an association with IPSAS.

    1. Introduction

    Achieving public spending efficiency is a cornerstone in public management, especially because of budget cuts due to the global financial crisis (Papi et al., 2018). Previous studies investigated efficiency mainly at a local-government level (Narbón-Perpiñá & De Witte, 2018a,b) while few have analyzed government policy implementation and its effects on efficiency in different areas (administration, education, health, income distribution, and economic stability) at a central-government level. In this context, public-spending efficiency has been interpreted as the ability of the government to maximize the policy outcomes given a level of spending or the ability to minimize its spending given a level of economic activity (Chan & Karim, 2012).

    The financial crisis also emphasized the relevance and usefulness of management tools that support and improve decision-making in the public sector (Caruana et al., 2019). Consequently, many countries have implemented financial management reforms to use public resources more efficiently and promote citizens’ engagement.

    Within this “reform movement,” the language of accountancy has acquired a new significance (Lapsley, 1999), and accounting plays a central role (OECD/IFAC, 2017). Many countries innovated accounting systems implementing accrual-based accounting and International Public Sector Accounting Standards (IPSAS). These reforms are expected to ensure high-quality information (Bastida & Benito, 2007; Kopits & Craig, 1998; Wang, 2002) that can better support decision-making processes (Sutcliffe, 2003), thereby improving the management of public resources (Bergmann et al., 2019). Accordingly, this study aimed to investigate the relationship between public-sector accounting innovations (IPSAS and accrual-based systems) and public-spending efficiency.

    A major motivation for this study was that the impact of public-sector accounting reforms on governmental performance (particularly efficiency) had not been investigated thoroughly. Indeed, there is a risk of retaining public-sector accounting as a sort of “black box” or bundle of techniques (Steccolini, 2019), as occurred in several studies investigating key public administration concepts such as accountability and transparency (e.g., Brandsma & Schillemans, 2013; Romzek & Johnston, 2005; Schillemans, 2015). This could lead to a functionalist view (Steccolini, 2019) in which the relationship between accounting innovations and efficiency could be taken for granted.

    Following Micheli and Pavlov (2020), this study avoided considering accounting merely as the “technical lifeblood” (Olson et al., 1998) of public financial management reforms. Conversely, it referred to the concept of “publicness” (Steccolini, 2019) to investigate the effects of accounting innovations at a central government level on efficiency within a decision-making context (Nutt & Backoff, 1993). Publicness has been considered as a way to conceptualize public performance (Alford & O’Flynn, 2009), supporting the idea that the search for greater efficiency and a better performance can be blended with the aspirations for the common good (Dahl & Soss, 2014), considering both the citizen’s and the public administration’s perspective (Papi et al., 2018). Consequently, the hypothesized relationship does not necessarily imply a causal link. Instead, it should be interpreted as an association, meaning that a greater level of efficiency is expected to occur in countries that have also implemented accounting reforms.

    We investigated a sample of 22 European countries in the period 2010–2018, adopting a two-step approach. First, different government spending-efficiency indicators were created using nonparametric techniques. Second, a model was proposed to show the effect of the IPSAS/accrual reform(s) in Europe on these efficiency indicators. The European context was selected because most European countries have adopted accrual accounting. Furthermore, a European task force was created in 2014 to issue European Public Sector Accounting Standards (EPSAS), which took IPSAS as a reference, and some European countries were doing the same when issuing or updating their national standards (OECD/IFAC, 2017).

    Our findings showed that accrual-accounting systems were related to spending efficiency, although the evidence was insufficient to support an association with IPSAS implementation. The economic explanation of these findings is that IPSAS implementation is mainly oriented towards external stakeholders while accrual accounting could also impact internal decision-making processes, contributing to improving efficiency.

    The paper is structured as follows. Section 2 reviews the literature on efficiency and public-sector accounting. Section 3 proposes the research hypotheses. Section 4 describes the research methodology while Section 5 presents and discusses the results of the analysis. Section 6 concludes and provides suggestions for further research.

    2. Background

    2.1. Efficiency of public-sector entities

    Previous studies investigated a wide range of efficiency determinants. Chan and Karim (2012) found that political stability and financial freedom positively affected public-spending efficiency while governance indicators had a negative effect. Rayp and Van De Sijpe (2007) found that government efficiency was determined by structural country variables and governance indicators while economic policy determinants apparently counted less. Hauner and Kyobe (2010) showed that richer countries tend to exhibit better public-sector performance and efficiency with accountability and demographic factors playing a significant role. Rahmayanti and Horn (2010) noted that a critical level of spending efficiency had a positive effect on growth in developing countries. Adam et al. (2014) and Ubago Martínez et al. (2018) analyzed decentralization as a determinant of governmental efficiency. Montes et al. (2019) noted that fiscal transparency is relevant to improving government effectiveness and spending efficiency; they found that management of public resources is part of the principal–agent problem, in which the principal (the public) has less information on the intentions and actions of the agent (government). Information is then essential to reduce asymmetries between the two actors, which leads to a suboptimal outcome since the government will act not in the best interest of the public but, rather, in its best self-interest.

    A further stream of literature concentrated on the dichotomy between the private and public management of public services, investigating whether use of the private sector, through outsourcing strategies and public–private partnerships, led to greater efficiency. However, the concept of efficiency in the public-sector realm is different (Button, 1979) since decision-making must consider different—sometimes less tangible—aspects. Consequently, should governments refer uncritically to the approaches generally used in the private sector, they may end up rationally maximizing utility in ways that lead to inefficiencies. To avoid this risk, Rainey et al. (2010) suggested considering a contingency perspective (Thompson, 1967), in which economic rationality is based on goal agreement and technical knowledge, which rely on complete and reliable information. Considering that information is mainly provided by accounting systems, an improved accounting system is then supposed to enhance knowledge of the economic and financial health of a government (Hyndman & Connolly, 2011). Following Voghouei and Jamali (2018), the use of more complete information not only can enable governments to better understand citizens’ needs, helping them to prioritize decisions to enhance the quality of public services, but can also help detect inefficient and illegitimate spending of resources.

    2.2. Accrual accounting and IPSAS implementation in the public sector

    Because of the central role of the budget in the public-sector context, budgetary accounting and cash-based information have been the mainstream accounting methods in this arena for many years (Oulasvirta, 2019). However, cash-based accounting does not provide the required information to support the decision-making processes useful to plan and manage resources related to service provisions (Bergmann et al., 2019; FEE, 2007). The movement towards accrual accounting denotes a shift from a “strict situation of cashed incomes and paid expenses to a situation where the emphasis is on achievements in performance conditions”; consequently, “the emphasis is on efficiency” (Tiron-Tudor, 2017, p. 3).

    Although scholars documented several difficulties related to the implementation of an accrual-based accounting system, mainly due to public-sector specificities, an increasing trend towards accounting innovations has been observed (Brusca et al., 2015).

    Accrual-based accounting provides more complete and comprehensive data, making it possible to assess financial sustainability better than cash accounting, evaluate the costs of services and political programs, better identify liabilities, and evaluate outsourcing strategies (Anessi-Pessina & Steccolini, 2007). Accrual accounting is believed to satisfy the information needs of investors (Caperchione & Salvatore, 2012), providing valuable information on solvency (Pina & Torres, 2003). Furthermore, the adoption of accrual-accounting systems can lead to the implementation of cost accounting techniques (Liguori et al., 2012) to support decision-making processes.

    This debate has frequently gone together with the implementation of IPSAS, which are considered a useful support for strategic management decision-making processes (Sutcliffe, 2003). IPSAS provide reliable and comprehensive information (Bastida & Benito, 2007; Kopits & Craig, 1998) and positively influence transparency and accountability, reducing opportunities for corruption (Cuadrado-Ballesteros et al., 2020). IPSAS implementation facilitates the comparability of financial information (Wang, 2002) because of the international harmonization of public-sector accounting (Benito et al., 2007; Christiaens et al., 2015).

    Scholars have claimed that IPSAS are close to their private-sector counterpart, the International Financial Reporting Standards (IFRS) issued by the International Accounting Standards Board (IASB) based on an external “decision-usefulness” approach (Ellwood & Newberry, 2016). Indeed, Brusca et al. (2016) argued that, in certain contexts (they investigated Latin American countries), IPSAS implementation can be interpreted as a response to international pressures where main actors are external stakeholders such as multinational agencies. However, it should be noted that IPSASB’s current projects deal with public-sector-specific topics, reducing the IFRS’s “dependence.”

    Mussari (2014) also claimed that, while the implementation of accrual-accounting systems does not seem to be an issue anymore, reaching a harmonized status is a key point at both an international and a regional level. In this last respect, the European Union launched the EPSAS project to harmonize European public-sector accounting.

    It is worth observing that cash and (full) accrual should be considered the extremes of a continuum of approaches with several intermediate solutions, such as modified-cash or modified-accrual (Manes-Rossi, 2016). Similarly, IPSAS implementation should not be considered a dichotomic variable, also considering that a country could adopt IPSAS indirectly, implementing national accounting practices that are essentially based on IPSAS. Accordingly, intermediate situations need to be considered that seek to assess the level of accounting maturity of each country (PwC, 2014).

    3. Research Hypotheses Development

    The idea that implicitly underlines the studies mentioned above is that information could play a relevant role in improving efficiency thanks to a more effective decision-making process. Many studies that focused principally on private-sector entities documented the positive impact of high-quality information (and IT infrastructures as well) on performance. In such studies, information is treated as a factor that allows organizations to monitor internal and external issues during the decision-making process regardless of the organizational decision-making model (optimizing, satisfying, incremental, and so on).

    Management information has frequently been considered as a production factor that is readily available (Citroen, 2011); therefore, its role has been regularly taken for granted and seldom discussed and analyzed as such. In other words, information has rarely been viewed as a determining factor. Scholars label this incongruity the “illusion of control” paradox (Davis & Kottemann, 1994; Raghunathan, 1999), meaning that having better information would (automatically) imply an improved performance. Consequently, the sources or the quality of information are taken for granted during the decision-making process (Citroen, 2011), although scholars emphasized that the quality of the decisions made and their effects on performance depend on the quality of information (Raghunathan, 1999). Further research had mixed results: while the bulk of previous studies evidenced a positive effect of information and decision comprehensiveness on performance, especially in unstable environments, others documented opposite results (for a review, Miller, 2008; see also Forbes, 2007). Meissner and Wulf (2014) called for further research to investigate the role of information quality in decision quality and then in organizational performance.

    Few studies have investigated the effects of higher information quality on the performance of public-sector entities (Heintze & Bretschneider, 2000; Rainey et al., 2010). However, the broad adoption of e-government initiatives has stimulated an intense debate as they facilitate the active participation of citizens in, and public control over, activities performed by public-sector entities (Cuadrado-Ballesteros et al., 2022; Guillamón-López et al., 2011). Governments are required to change their consolidated decision-making processes to improve the quality of services and the relationships with citizens (Aiello et al., 2018). Accordingly, a “central issue in organizational performance is the quality of decision making [which] is a particularly acceptable proxy for public organizations because [they] generally require decisions to be more timely and correct than do private firms” (Heintze & Bretschneider, 2000, p. 807).

    We investigated the critical issue of the quality of information from accrual-accounting systems and IPSAS implementation. Information quality would mean reliability, completeness, and timeliness (Citroen, 2011), which are important requisites for information to be retained as a supportive base for decision-making processes, leading to better evaluations for strategic choices. However, information could be used in a purposeful or passive way with the main aim to improve efficiency and effectiveness or to comply with certain requirements, respectively (Micheli & Pavlov, 2020; Moynihan, 2009). Furthermore, implementing a new accounting system (accrual- or IPSAS-based) is not only a technical issue but shapes a renewed managerial culture. Some studies investigating the use and usefulness of accrual-accounting information documented a limited use of it while others had more positive findings (see van Helden, 2016 for a review). This means that investigating the effects of better information quality on efficiency would require additional (external) factors, such as the emergence perspective and the organizational imperative (Markus & Robey, 1988).

    The emergence perspective considers the inherent dynamicity of the implementation and use of a different accounting system (accrual- or IPSAS-based), which could affect the decision-making style and culture of management. Accordingly, the interaction of conflicting objectives and preferences should be heeded. Organizational imperative relates to the rationality of decision-making processes and orienting behavior towards a specific outcome (e.g., improving efficiency); however, decision-making processes are not retained as certain, and contextual variables are deemed as contingencies. Improving the knowledge from accounting systems concerning the overall financial condition of the entity helps to detect inefficient spending of resources (Voghouei & Jamali, 2018) and to understand how contingencies can be addressed (Rainey et al., 2010; Thompson, 1967). Consistently with the concept of publicness (Steccolini, 2019), accounting is not a mere technical tool (Ellwood & Newberry, 2016) but is investigated to understand if decision processes, based on better-quality information (Nutt & Backoff, 1993), contribute to improve the overall governmental efficiency.

    According to all these arguments, more complete and reliable information from accounting innovations is expected to increase the quality of decision-making processes, which should bring about greater governmental efficiency (and, more generally, its performance). However—bearing in mind that accounting is not a “black box” (Steccolini, 2019)—these relationships should not be taken for granted as several contingencies might play a role, and several barriers may impede the achievement of positive effects of better-quality information. Consequently, although a direct, causal link between accounting reforms and efficiency could be hypothesized, it seems rational to claim that an association exists, meaning that greater efficiency is expected to occur in countries that have also implemented accrual-accounting systems and/or IPSAS. Therefore, the following hypotheses are tested:

    H1: Countries that have implemented IPSAS are more efficient.

    H2: Countries that have implemented accrual-accounting systems are more efficient.

    4. Methodology

    4.1. Sample

    To investigate these hypotheses, a sample of 22 European countries was selected: Austria, Belgium, Czech Republic, Denmark, Estonia, Finland, France, Germany, Greece, Hungary, Ireland, Italy, Latvia, Lithuania, Netherlands, Poland, Portugal, Slovak Republic, Slovenia, Spain, Sweden, and the UK. The European context is highly appropriate since most of these countries have adopted accrual accounting; furthermore, although the level of IPSAS adoption is low, the EPSAS project (and national standard-setters as well) used IPSAS as a primary reference (OECD/IFAC, 2017). The analysis covers the period 2010–2018.

    4.2. Creating efficiency indicators

    Efficiency refers to the maximum potential output obtained from a given input or the minimum potential input required to produce a given output (Knox Lovell, 1993). Extrapolating this concept to the central government level, outputs would be the outcomes of public policies while inputs would be financial resources used to fund these policies. Accordingly, a government is considered efficient if it can provide the maximum level outcomes (outputs) utilizing the minimum level of spending (inputs). This is referred to as “government spending efficiency.”

    Operationalizing this concept was a difficult task because identifying inputs and outputs is a thorny issue (Afonso et al., 2010). We followed the methodology of Adam et al. (2011), Chan and Karim (2012), Chan et al. (2017), and Montes et al. (2019), grouping inputs and outputs into two categories, heeding Afonso et al. (2005, 2010):

    Opportunity indicators reflect the interaction between fiscal policies and the market process and the impact on individual opportunities. A well-functioning public administration and a healthy and well-educated population who enjoy high-quality public infrastructure could be considered essential for well-functioning markets and secure property rights with plenty of opportunities for all (Afonso et al., 2005). Then, four dimensions should be represented by different inputs and outputs: administrative, education, health, and infrastructure dimensions.

    Musgravian indicators represent the outcomes of the interaction with and reactions to the market process by the government. They refer to income distribution, economic stability, and allocative efficiency (i.e., general economic performance).

    Tables A.1 and A.2 in Appendix describe input and output used to research each dimension and also provide some descriptive statistics.

    Having selected these variables, frontier functions were used to represent the best possible combination input–output. Researchers revealed a wide range of statistical techniques, the most frequently used being the nonparametric approach. This method determines the best frontier as a linear envelopment of the data, which is created with the most efficient decision-making units (here, general governments); thus, it is necessary to find the maximum ratio of the linear combination of outputs compared to that of inputs, contemporarily selecting the optimal weights of inputs and outputs. The main advantage of nonparametric techniques is that they do not require a specific functional form as parametric techniques do. This proves to be a shortcoming because they have a deterministic nature; however, it is also an advantage because it avoids specification and estimation problems in the definition of a specific functional form.

    We used three specific nonparametric techniques to represent the efficiency of the general governments in the sample: (i) Data Envelopment Analysis (DEA), (ii) DEA with application of the bootstrap technique, and (iii) the Order-m model.

    4.2.1. Data envelopment analysis (DEA)

    The DEA model (Charnes et al., 1978) is based on linear programming techniques to define an empirical frontier that creates an envelope of the most efficient decision-making units, trying to get the maximum level of output with the minimum input. In the specific case of the public sector, outputs (public services) are totally or partially regulated externally by the law; therefore, it is more appropriate to evaluate efficiency in terms of the minimization of inputs (budgetary variables) while retaining variable returns to scale. The minimum is found by annually selecting the optimal weights associated with inputs and outputs through the resolution of the following program :

    Minθ,λθs.t.yrii=1nλiyri,r=1,,pθxjii=1nλixji,j=1,,qλi0,i=1,,ni=1nλi=1
    where i represents each government (i=1,,n), yr refers to each output (r=1,,p), and xj refers to each input (j=1,,q). The restriction i=1nλi=1 implies the assumption of variable returns to scale, which ensures that each government is compared only to others of similar sizes. For each government, the value of θ is obtained—that is, the efficiency score (DEA): if θ=1, it is defined as efficient; if θ<1, it is inefficient.

    However, the DEA technique has some shortcomings due to its deterministic nature (De Witte & Marques, 2010). First, it is highly sensitive to extreme values and outliers as it creates a frontier that envelops all data. Second, DEA assumes the absence of statistical noise, so it is sensitive to measurement errors. Accordingly, other methodologies may overcome such shortcomings, such as bias-corrected DEA via bootstrapping techniques and the Order-m methodology.

    4.2.2. DEA with application of bootstrap technique

    The bootstrap technique is a way to analyze the sensitivity of efficiency to the sampling variations, simulating the efficiency for different subsamples (Simar & Wilson, 1998). We used the Simar and Wilson (1998) algorithm that applies the smoothed bootstrapping procedure to generate θi (i=1,,n) with replacement from (θ̂1,,θ̂n), producing (θ1b,θ2b,,θnb), where b is the b-th iteration of the resampling process (Assaf & Matawie, 2010).

    The bootstrap inputs were then obtained as xib=(θ̂iθit)xi; these bootstrap inputs were used to obtain the new estimates of efficiency, namely θ̂ib. These steps were repeated B times, producing a set of θ̂ib where b=1,,B. Finally, the mean of the bootstrap estimator, being DEA_bst=1Bb=1Bθ̂ib, was used as the bootstrap DEA estimate. Therefore, the difference between the original DEA estimates and these newly created scores are usually called bias (biasî=1Bb=1Bθ̂ibθ̂in).

    Moreover, confidence intervals can be obtained via (θ̃inα,θ̃in1α) where θ̃inα is the 100α percentile of the distribution of θin, and shifting the bounds of the interval by the factors (2biasî) will ensure that the bootstrap distribution centers on the bias-corrected estimate θ̃in=θ̂inbiasî (Assaf & Matawie, 2010).

    4.2.3. Order-m methodology

    The Order-m frontier (Cazals et al., 2002) may overcome DEA shortcomings since it does not require enveloping all data. We also took an input orientation because, as indicated previously, the outputs are required externally. In this case, the Order-m estimator is used as a benchmark for the expected minimum level of inputs given a fixed number of m governments producing at least an output level y (Narbón-Perpiñá et al., 2017). Therefore, following a similar notation, efficiency (Order-m) is defined as

    θ̂m(x,y)=E[(θ̂m(x,y)|Yy)].

    This means that, for a given level of input–output, the estimation defines the expected maximum of m random variables drawn from the conditional distribution of the output matrix Y observing the condition Yy. A value greater than 1 indicates super-efficiency, suggesting that the general government that operates at the level (x, y) is more efficient than the average of m peers randomly drawn from the rest of the population producing more output than y (Narbón-Perpiñá et al., 2017).

    4.3. Stage 2: Testing the association between public-sector accounting and efficiency

    Having obtained the efficiency indicators, the following model was proposed to test our hypotheses :

    Es,it=α+βjAj,it+γkCk,it+ηi+νit.(1)

    In model (1), subindexes i and t refer to country and year, respectively; Es is the vector of the s dependent variables representing the efficiency of the investigated governments; Aj is the vector of the j independent variables representing the public-sector accounting situation; Ck is the vector of the k control variables; α, β, and γ are the parameters to be estimated; νit is the classical disturbance term; and ηi is the unobservable heterogeneity that represents the characteristics of each country, which are different from other countries but are invariant over time.

    The dependent variables (Es) are the efficiency indicators previously described: (i) DEA refers to efficiency as obtained by the DEA technique (Charnes et al., 1978), (ii) DEA_bst was obtained using bootstrap methods based on subsampling (Simar & Wilson, 1998), and (iii) Order-m refers to the partial frontier proposed by Cazals et al. (2002). The two former variables take values between 0 and 1 from the worst to the best level of efficiency, but Order-m may also take values greater than 1, which indicates super-efficiency (the government that operates using such an input–output combination is more efficient than the average of the peers randomly drawn from the rest of the population producing a greater output level).

    The independent variables (Aj) represent the accounting situation through two indicators, called IPSAS and Accruals. The IPSAS variable represents the different levels of IPSAS implementation at a central government level by taking three values:

    IPSAS=1: No actions were undertaken to adopt IPSAS up to now.

    IPSAS=2: The legislative process was undertaken, and IPSAS are being adopted or partially applied.

    IPSAS=3: IPSAS are adopted, or national standards are broadly consistent with IPSAS.

    The Accruals variable refers to the status of accruals reform(s), also taking three values:

    Accruals=1: Public-sector accounting standards are cash-based.

    Accruals=2: Public-sector accounting standards are in transition to accrual, or they require modified-accrual or modified-cash systems.

    Accruals=3: Public-sector accounting standards are accrual-based.

    In both cases, a value from 1 to 3 is assigned to each country in each year. Therefore, a country can have different values over the sample period (2010–2018) if its status changes during it. Both IPSAS and Accruals are ordinal variables because the values are assigned by observing the “development” or “improvement” in the implementation of IPSAS or accrual-based accounting (from the minimum to the maximum level of adoption). This approach is more appropriate than the alternative one based on dummy variables; in fact, IPSAS/Accruals take different values (as continuous) so that the association with governmental efficiency due to an improvement in accounting maturity (in terms of IPSAS implementation or accrual-accounting adoption) can be observed.

    The data were obtained from two sources. The first is the OECD/IFAC (2017) report, which is based on information gathered from a survey sent to the Ministries of Finance and the equivalent bodies of all OECD countries. The second is the IFAC website,1 which collects information from member organizations and publicly available sources listed in each jurisdiction profile. The IFAC triangulates various sources within and outside the jurisdiction (country), such as relevant organizations (e.g., institutes of auditors and accountants, financial supervisory authorities, accountants’ regulatory boards, and so on), legislation, and publications of international organizations (e.g., Deloitte, World Bank, European Commission, and so on).

    Finally, control variables (Ck) are socioeconomic and political factors whose effects on efficiency have been previously evidenced in the literature (Adam et al., 2011, 2014; Hauner & Kyobe, 2010; Chan & Karim, 2012; Ubago Martínez et al., 2018; Montes et al., 2019):

    Population density (Density) is the midyear population divided by land area in square kilometers.

    Dependent population (Dependency) is the sum of the share of the population over 65 and below 15 years old.

    Political ideology is represented by a categorical variable (Ideology) that takes the value 1 for right-wing governments, 2 for center, and 3 for left-wing governments.

    Political competition is represented by the fraction of seats held by the governmental party (Govseats).

    Trade openness (Openness) is represented by the sum of exports and imports of goods and services.

    The level of fiscal balance (Deficit) is represented by the net lending (+)/net borrowing () of the central government (% of GDP).

    Data regarding control variables were obtained from the World Bank and Eurostat, except political ideology and competition, which were retrieved from the Database of Political Institutions (Cruz et al., 2016).

    The truncated estimator (with fixed effects) was used to test model (1) as it offers more robust results than the traditional OLS and Tobit methods (Da Cruz & Marques, 2014; Simar & Wilson, 2007) and avoids correlation problems between the first (creating the efficiency indicators) and second (testing the determinants of efficiency) stages.

    5. Findings and Discussion

    5.1. Descriptive analysis

    Table 1 shows the descriptive statistics of variables in Model (1). The level of efficiency was quite large in terms of the three indicators. More concretely, Figure 1 illustrates the mean value of efficiency during the period 2010–2018 by country. In general, DEA and DEA_bst showed similar results, but they differed slightly from that given by the Order-m variable, suggesting the need to use different indicators to represent efficiency.

    Table 1. Descriptive Statistics

    VariableObsMeanStd. Dev.MinMax
    DEA1980.99950.00170.99111
    DEA_bst1980.99870.00150.99080.9996
    Orderm1980.97480.04630.63731
    IPSAS1981.49490.585313
    Accruals1982.31820.763913
    Density198137.8056114.990817.6484511.4759
    Dependency19833.90052.087727.787437.9911
    Ideology1471.81630.936513
    Govseats17656.52117.957836.076079.8732
    Openness198113.459844.482552.0062226.0414
    Deficit1982.99753.487332.11.6
    Distribution of Observations of Ordinal Variables
    Freq.PercentCum.
    IPSAS=110955.0555.05
    IPSAS=28040.495.45
    IPSAS=394.55100
    Accruals=13618.1818.18
    Accruals=26331.8250
    Accruals=39950100
    Ideology=18054.4254.42
    Ideology=2149.5263.95
    Ideology=35336.05100
    Fig. 1.

    Fig. 1. Distribution of efficiency indicators by each country.

    Regarding the independent variables, the mean values of IPSAS and Accruals were 1.49 and 2.32, respectively, with a range between 1 and 3 in both cases. These results indicate that, whereas most of the sample countries implemented accrual-based systems, IPSAS implementation was still low. The bottom of Table 1 shows the distribution of observations of the two variables while Figure 2 represents the situation graphically, considering the mean value during the period 2010–2018. Indeed, the only country in the sample that can be considered IPSAS-compliant is Estonia while, in the remaining countries, no actions were undertaken (IPSAS=1) or IPSAS were only partially adopted (IPSAS=2). Although IPSAS compliance is still low, this figure shows that the accrual-based system is widely implemented in Europe.

    Fig. 2.

    Fig. 2. Distribution of accounting indicators by each country.

    Turning to Table 1, the statistics of control variables indicate the mean population density (137.8 inhabitants per km2) and the share of population over 65 and under 15 (33.9%). Dependency ratios were more similar among the sample. The mean value of Ideology does not have economic meaning because it is an ordinal variable, only controlling for government ideology in the model; the mean value of Govseats indicates that 56.52% of seats were held by the governmental party. The mean value of Openness was 113.46, which is the sum of the mean value of imports and exports (% of GDP). Finally, the mean value of Deficit was 2.99, representing the poor financial situation that Europe suffered during the period 2010–2018.

    Table 2 shows the bivariate correlations. There were high and relevant correlations between DEA and DEA_bst (0.9976) because both were obtained through the DEA technique. Nevertheless, this does not introduce multicollinearity problems because efficiency indicators were considered individually as dependent variables. Correlation coefficients between these two variables and Order-m were statistically relevant, although smaller. This supports the idea of using different efficiency indicators in the testing model (1). IPSAS and Accruals were also highly correlated (0.452), so they will be entered individually into the model. The remaining correlations between control and independent variables were not so high (lower than 0.5).

    Table 2. Bivariate Correlations

    DEADEA_bstOrdermIPSASAccrualsDensityDependencyIdeologyGovseatsOpenness
    DEA_bst0.9976***1
    Orderm0.12310.12781
    IPSAS0.2077**0.2064**0.05011
    Accruals0.0230.01930.07220.452***1
    Density0.04860.05020.01920.1691*0.3225***1
    Dependency0.1428*0.13740.1627*0.08960.2516***0.07951
    Ideology0.2461**0.246**0.0690.10220.2198**0.2084*0.14571
    Govseats0.06790.06670.04130.14240.01010.12060.03260.266**1
    Openness0.03170.03740.1527*0.2686***0.268***0.05920.414***0.03770.2279**1
    Deficit0.08310.08240.2241**0.1906**0.2063**0.0650.2381***0.01760.10760.0799

    Notes: , *, **, ***significant at 10, 5, 1, and 0.1 percent level, respectively.

    5.2. Explanatory analysis

    Panels A and B of Table 3 exhibit the effects on the dependent variables of IPSAS and Accruals, respectively. While the Accruals variable impacted the three efficiency indicators positively and was statistically relevant, the IPSAS variable was significant only in the first equation and impacted the indicators negatively. These results support H2; not enough evidence was found to support H1. This means that accrual-accounting data satisfies the information needs and better supports public management decision-making regarding planning and management of public resources (Bergmann et al., 2019), confirming how important it is to focus on efficiency (Tiron-Tudor, 2017). Conversely, IPSAS implementation does not seem to be relevant in this respect.

    Table 3. Empirical Results

    DEADEA_bstOrderm
    Coef.Std. Err.Coef.Std. Err.Coef.Std. Err.
    Panel A. Effect of IPSAS on Efficiency Indicators
    IPSAS0.0862*0.20080.01360.01020.52630.9096
    Density0.00020.00120.00100.00380.00180.0039
    Dependency0.1288*0.07270.00690.00520.01730.2403
    Ideology0.3361*0.15410.0312*0.01441.11170.9334
    Govseats0.51340.34710.00060.00091.85351.1379
    Openness0.00050.00360.00020.00030.0243*0.0109
    Deficit0.02680.03640.00220.00320.2538*0.1175
    cons0.9956***0.00280.9954***0.00201.0146***0.0918
    /sigma1.5269***0.09260.1408***0.02325.0008***0.3032
    Panel B. Effect of Accruals on Efficiency Indicators
    Accruals0.6984*0.27310.6337*0.25100.0811*0.1844
    Density0.00040.00120.00040.00110.00020.0011
    Dependency0.1460*0.07210.12640.06630.11100.0668
    Ideology0.5499*0.28020.5139*0.25760.3133*0.1415
    Govseats0.57780.34160.52930.31400.47050.3189
    Openness0.00090.00330.00050.00300.00020.0033
    Deficit0.00840.03530.00800.03240.02490.0334
    Cons0.9928***0.00280.9928***0.00250.9954***0.0026
    /sigma1.5012***0.09101.3802***0.08371.4027***0.0851

    Notes:

    Dependent variables: DEA, DEA_bst, and Orderm.

    Independent variables: IPSAS and Accruals.

    , *, **, *** significant at 10, 5, 1, and 0.1 percent level, respectively.

    A first explanation of these results is likely empirical, observing the mean values of IPSAS and Accruals variables, which reflect that accrual accounting is more frequently implemented than IPSAS. A further, more conceptual explanation could rely on the approach used by the IPSASB to issue its standards in the analyzed period. Scholars have claimed that IPSAS are close to IFRS (private-sector standards) based on a “decision-usefulness” approach. Ellwood and Newberry (2016)—having emphasized that this “decision-usefulness is essentially a market-based model” (p. 232)—claimed that IPSAS seemed to devote more attention to external decision-makers. Similarly, Brusca et al. (2016) documented that multinational agencies were among the most important users with the main aim being to address institutional pressures and improve international harmonization (Christiaens et al., 2015). Conversely, accrual accounting data seems to be devoted also to internal decision-makers, fueling management accounting since accrual accounting can imply implementation of cost-accounting techniques (Liguori et al., 2012). Bearing in mind the emergence perspective and the organizational imperative (Markus & Robey, 1988), the rationality of decision-making processes can be affected by contingencies (Rainey et al., 2010), which slacken the process; a progressive increase in knowledge due to accrual accounting data is then essential, supporting the idea that accrual accounting systems can help to understand how to address these contingencies (Rainey et al., 2010; Thompson, 1967) and detect inefficient spending of resources (Voghouei & Jamali, 2018).

    Our results also suggest that accounting innovations cannot be retained as unproblematic “black boxes” or as taken-for-granted elements of a vague administrative capacity (Steccolini, 2019, p. 259). External pressures could represent the main stimulus, especially in the case of IPSAS implementation.

    Regarding control variables, Dependency affects government spending negatively since it represents the segment of the population connected with the provision of educational, health, and welfare services. Therefore, pressure on government spending is increased as the share of dependent population is larger, resulting in a reduction of efficiency (Adam et al., 2014; Ubago Martínez et al., 2018). The negative link of Ideology suggests that efficiency tends to be reduced as the government moves to the left-wing ideology, in line with Borge et al. (2008) and Adam et al. (2011). Govseats shows positive coefficients, being statistically relevant in some equations at 90%. This means that majoritarian governments tend to show larger levels of efficiency since they are stronger and then able to impose hard budget constraints (Roubini & Sachs, 1989), resulting in improvements in government spending efficiency (Adam et al., 2011).

    6. Implications and Concluding Remarks

    Increasing efficiency is one of the main objectives of the reforms implemented in many countries, and researchers have investigated factors that can affect it. We aimed to contribute to the academic debate by considering the benefits due to governmental accounting reforms.

    The results concerning IPSAS suggest that their implementation can be motivated by other reasons—for instance, providing better-quality information to external stakeholders, such as investors or donors (Caperchione & Salvatore, 2012; Ellwood & Newberry, 2016), rather than improving efficiency by supporting the decision-making process of internal stakeholders, like managers and politicians. It could be argued that accrual accounting systems stimulate a purposeful use of information (Moynihan, 2009) more than IPSAS implementation per se since the former can facilitate the use of costing techniques (Liguori et al., 2012) more than the latter.

    In the light of the publicness frame, interpreted as a way to conceptualize public performance (Alford & O’Flynn, 2009) to be blended with the aspirations for the common good (Dahl & Soss, 2014; Papi et al., 2018), these results suggest that information and its purposeful use cannot be conceived as neutral means to improve public-sector performance automatically, reinforcing the idea that accounting is not a “black box” (Steccolini, 2019). The positive role played by information cannot be taken for granted as it occurs under the “illusion of control” paradox (Davis & Kottemann, 1994; Raghunathan, 1999).

    From a policy/practical perspective, our findings underline that accounting practices/reforms are not only a “technical” issue (Ellwood & Newberry, 2016). They can lead to organizational changes, technological developments, and the adoption of new managerial approaches (Caruana et al., 2019), going together with a higher level of efficiency. Innovating accounting systems is a dynamic process that helps translate planned activities into decisions and actions (Micheli & Pavlov, 2020; Steccolini, 2019). Accordingly, adopting a wider approach rather than focusing on technicalities could prove useful in assessing the effects on performance of public-sector financial management practices and reforms (Steccolini, 2019).

    This paper is not free of limitations. First, efficiency indicators were obtained through deterministic methodologies. Further studies may replicate our findings by using parametric techniques. Second, this study focused on efficiency as a key component of global performance. Future studies could consider other dimensions (Walker et al., 2010), also including measures of strategy content as explanatory factors of organizational performance (Andrews et al., 2006). Additionally, further studies could consider the implementation of other accounting maturity indicators (PwC, 2014). Finally, the inclusion of non-European countries in the analysis would allow evaluating the role of different cultural and institutional contexts.

    Notes

    1 See https://www.ifac.org/about-ifac/membership/member-organizations-and-country-profiles

    Appendix

    Table A.1. Inputs Description

    Opportunity IndicatorsMeanStd. Dev.Min.Max.
    Administrative dimensionPublic services spendingGovernment expenditure on public services (% of GDP)9.034.693.222.6
    Education dimensionEducation spendingGovernment expenditure on education (% of GDP)3.221.3605.7
    Health dimensionHealth spendingGovernment expenditure on health (% of GDP)2.872.040.17.7
    Infrastructure dimensionInfrastructure spendingGovernment expenditure on economic affairs (% of GDP)3.732.341.124.6
    Musgravian IndicatorsMeanStd. Dev.Min.Max.
    DistributionWelfare spendingGovernment expenditure on social security and welfare (% of GDP)7.363.95117
    StabilityTotal spendingTotal general government expenditure (% of GDP)30.007.4112.562.9
    General economic performance

    Data source: Eurostat

    Table A.2. Outputs Description

    Opportunity IndicatorsMeanStd. Dev.Min.Max.
    Administrative dimension(Data source:(1) Transparency International(2) Worldwide Governance Indicators)Corruption (1)Corruption Perception index assesses how corrupt a country’s public sector is perceived to be by experts and business executives.53.2728.043.492
    Regulatory quality (2)Regulatory quality index captures perceptions of the ability of the government to formulate and implement sound policies and regulations that permit and promote private-sector development.1.240.440.152.05
    Rule of law (2)Rule-of-law index captures perceptions of the extent to which agents have confidence in and abide by the rules of society and, in particular, the quality of contract enforcement, property rights, the police, and the courts as well as the likelihood of crime and violence.1.240.540.082.10
    Government effectiveness (2)Government effectiveness index captures perceptions of the quality of public services, the quality of the civil service and the degree of its independence from political pressures, the quality of policy formulation and implementation, and the credibility of the government’s commitment to such policies.1.220.490.232.24
    Education dimension(Data source: OECD)Secondary school enrollment(% gross)Total enrollment in secondary education regardless of age, expressed as a percentage of the population of official secondary education age113.6317.0990.68163.93
    PISA scores in readingReading performance measures the capacity to understand, use, and reflect on written texts to achieve goals, develop knowledge and potential, and participate in society.494.3216.47453536
    PISA scores in mathMathematical performance measures the capacity to formulate, employ, and interpret mathematics in a variety of contexts to describe, predict, and explain phenomena, recognizing the role that mathematics plays in the world.496.1116.30451541
    PISA scores inScientific performance measures the use of scientific knowledge to identify questions, acquire new knowledge, explain scientific phenomena, and draw evidence-based conclusions about science-related issues.500.7718.86452554
    Health dimension(Data source: World Bank)Life expectancyLife expectancy at birth (years)79.722.6973.2783.5
    Infant mortality rateInfant mortality rate (per 1,000 live births)3.400.881.76.5
    Immunization against measlesImmunization, measles (% of children ages 12–23 months)94.733.368099
    Infrastructure dimension(Data source: World Bank)Electricity powerElectricity generation (gigawatt hours) from fossil fuels, nuclear power plants, hydro-power plants (excluding pumped storage), geothermal systems, solar panels, biofuels, wind, etc.136262.8166843.73314.3619053
    Internet usersIndividuals using the Internet (% of population)78.4910.8444.497.64
    Telephone usersFixed telephone subscriptions (per 100 people)35.7313.705.8565.45
    Musgravian IndicatorsMeanStd. Dev.Min.Max.
    Distribution(Data source: Eurostat)GINI indexGINI index35.533.8424.346.8
    Income share by the highest 10%Share of national equivalized income held by highest 10%23.252.1417.528.8
    StabilityDeviation of the GDP growthStandard deviation of the GDP growth rate(Source: Own elaboration based on World Bank data)1.632.260.00615.20
    Inflation rateInflation, consumer prices (annual %)(Source: World Bank)1.451.361.745.65
    General economic performance(Data source: World Bank)Unemployment rateUnemployment total (% of total labor force)9.655.022.427.47
    GDP per capitaGDP per capita (constant 2010 US$)33827.9315920.5311348.4176662.67
    GDP growth rateGDP growth (annual %)2.012.869.1325.16

    References