I extend my gratitude to my colleagues at GSGAS and DZHW for providing both critical and constructive feedback. Likewise, I thank the two anonymous reviewers for their input and recommendations.
1Transparency has been at the heart of policy and empirical debates since the turn to New Public Management (NPM) processes. This turn – focused mostly on the efficiency of resource allocation, the clear distribution of tasks, and the fulfillment of indicator-based objectives – has led many public institutions on a path of stringent budget reconsideration. Cost-efficiency and cost-effectiveness measurements and, most pressingly, corruption (Abreu & Gomes, 2021; De Simone et al., 2017) were at the heart of public administrations under the NPM paradigm. The fight against corruption was accompanied by campaigns aimed at increasing citizen trust in governments and with efforts to make public an array of records and data about the dealings of different levels of administration (supranational and international, national and local) (Ott et al., 2019; Rossi et al., 2018; Zhang & Wang, 2021). In order to reduce corruption and increase citizen trust in institutions, public administrations were driven to maximise their efforts towards openness and accountability. For over 30 years, this maximisation has been the established principle for public offices as well as the preferred slogan for political actors.
2New indices have since been developed to assess how each polity (macro-level) or institution (micro-level) could handle the task of becoming more transparent and trustworthy (Albassam, 2015; Hollyer et al., 2014; Williams, 2015). A myriad of concepts have thus been operationalised as mechanisms aimed at curbing corruption and other governance deficits. Among these, transparency stands out as the most discernible high-level concept meant to aggregate the different dynamics that underpin political and administrative mandates. Transparency then becomes a catch-all term without clear boundaries – transparency represents fiscal control, administrative checks, or even political acts wholly dependent on the empirical application in which it is specified.
3Efforts have been made to operationalise more concrete measurements of transparency in specific realms that more directly affect the public domain, such as fiscal governance. Indices such as the Open Budget Index (OBI) (cf. de Renzio & Masud, 2011; Seifert et al., 2013) stand out for their systematisation of different sources of data that, unlike perception measurements, are aggregated according to normative prescriptions, i.e., in scales that deviate from an ideal type. Such an exercise allows for direct comparability and more detailed case analyses in that it reflects the idea of openness multidimensionally. In this sense, fiscal governance structures can indeed benefit from transparency, mainly as a positive contribution to making budgetary and planning policies more open and democratic. However, the OBI coexists with other measurements that do not necessarily complement or build upon each other. This is further confounded by different foci and research designs that broaden and diversify the methodological landscape.
4The goal of this paper is therefore to present and discuss the diversity and implications of the many approaches to measuring transparency, specifically in the field of fiscal governance. A systematic review of (35) selected empirical studies in this area is presented and discussed based on the analytical dimensions they operationalise, and contrasted with a conceptual framework I propose. Thus, attention is paid to the particular traits and dimensions of transparency measured and utilised in these studies, highlighting both methodological and analytical elements. To provide further empirical evidence of the unharmonised nature of the measurements operationalised in these studies, the findings are summarised in a meta-analytic approach. From these, key empirical elements are discussed and assessed from the perspective of governance outcomes. The paper concludes with some remarks on the validity of these findings and a critical approach to the study of transparency.
5One of the most famous examples regarding the measurement of perceptions is the Corruption Perception Index (CPI) developed by Transparency International (TI). This index compiles a number of country expert reports and interviews that evaluate the respective national context on a yearly basis (see de Renzio & Masud, 2011). Despite the potential value of acknowledging the context-specific insights of experts and civil society members, the downside of these types of measurements has to do with the issue of representativeness of the problems framed relative to the problems described (Moloney & Stoycheva, 2018). For example, expert scholars or involved stakeholders will necessarily possess a more informed perspective than that of laypeople. This is due to the level of exposure to direct sources of information and a more comprehensive understanding of the functioning of accountability channels. In addition, the perceptions of groups, laypeople, and experts may differ significantly in terms of perceived and experienced corruption (Vadlamannati & Cooray, 2017), while the core of the problem – actual corruption – remains under-described at best.
6As this article explores a particular aspect of transparency, namely fiscal transparency, the application of the literal/phenomenological framework provides a more defined perspective on the elements that make up the sample of publications analysed later. To this end, I focus on the governance specificities that determine the openness of fiscal policies and their implementation. Naturally, I then look for these characteristics and their operationalised structures in the empirical settings analysed. Hence, for the sake of precision – and based on a variety of contributions that discuss conceptual approaches (e.g., Birchall, 2014; Cruz Romero, 2023; Islam, 2006; Pozen, 2020; Stasavage, 2003) – I take fiscal transparency as the intersection of information availability and accounting channels for matters of resource allocation, taxation structures, and oversight mechanisms. Thus, I also look for these elements in the empirical designs and results of the studies analysed.
- 1 The defining characteristic being that an objective state of transparency cannot be quantified. In (...)
- 2 Rowe and Slutzky’s text (“Transparency: Literal and Phenomenal”) was written for the fields of arch (...)
7As noted above, there are two major strands of measurement in the study of transparency (or transparency-based phenomena); namely subjective and objective (as in Chen & Ganapati, 2021). Now, this terminology has been further developed in corruption studies, where the contrast between perceived and actual corrupt practices has long been problematised as a source of conflict with tangible policy and analytical repercussions (Gebel, 2012)1. Despite the motivating and comprehensive debates around the validity or suitability of these labels, I will not engage in the discussion, as it is tangential to the main point of the research. For the sake of coherence and conceptual precision, transparency measures are categorised into a typology in terms of subjects’ relationship with transparency. These distinctions are consistent with the terminology and analyses presented and with a broader strand of the transparency literature (Birchall, 2014; Heald, 2003; Pozen, 2020). Moreover, its expansion into fiscal transparency (see Table 2) addresses specific analytical elements. Hence, I will work with two categories; a) literal or normative, and b) phenomenological or experiential (see Rowe & Slutzky, 1963)2.
8The literal-phenomenological typology attempts to provide a clearer and more concise distinction from the controversial claims of truth and reason (for example, as objectivity in opposition to subjectivity). As discussed in the literature (cf. Chen & Ganapati, 2021), the objective-subjective perspective works insofar as the contextual dependencies and measurement specificities (e.g., roles and experience of the experts surveyed) are assumed to be uniform. The problem then becomes the objective transformation of subjective appraisals, which is exactly what this framework seeks to highlight and problematise. Thus, my intention is to argue in favour of a literal sense of transparency vis-a-vis an experienced sense of transparency, focusing on the distance between the norm and the perceived effects of the norm. For instance, I argue that there are only literal implementations of norms and policies depending on the level of governance and the degree of influence involved. Likewise, these implementations are experienced more or less directly, also depending on various factors. For example, time, as policies may still be perceived as opaque or unchanged “over time because of deterrence effects” (Chen & Ganapati, 2021, p. 267).
9As shown in Table 1, the main characteristics of the proposed typology are based on the unit of analysis that serves as the basis for constructing transparency indices, i.e., their proxies. In the literal-normative definition these units can be found in official documents or statistics, and they serve as a basis for social and economic policies. In other words, they are ideal types of transparency objects meant to guide the debates on the state of different transparency dimensions or mechanisms. On the other hand, phenomenological transparency is anchored in perceptions and focuses on the experiential nature of similar sources of compiled data (documents or statistics). This approach underscores the defining distance between the normative and the experienced. In contrast, the objective-subjective approach equates both sides as the same measurement construct, artificially erasing this distance (and thus concealing the respective measurement nuances of, for example, surveys versus oversight reports). The objective-subjective column portrays this agglutination of elements. Most importantly, the weakness row of this last column shows precisely this problem: it is (mostly) not generalisable and depends on data collection factors (e.g., survey) prone to biases that are not acknowledged.
Table 1. Proposed Typology of Transparency Measurements
|
Literal/Normative
|
Phenomenological/Experiential
|
Objective/Subjective
|
Unit of Analysis
|
Documents
Policies
Laws
|
Perceptions
Compiled information
|
Documents
Policies
Perceptions
|
Applications
|
Qualitative studies
Comparative analyses
Quantitative assessments
|
Comparative analyses
Quantitative assessments
|
Qualitative studies
Comparative analyses
Quantitative assessments
|
Sources
|
Official documentation
|
Expert opinions
Official data sources
|
Official documentation
Expert opinions
|
Strength
|
Focalised
In-depth
Specific
|
Broad
Multi-level
Generalisable
|
Focalised/Broad
In-depth/Multi-level
Specific/Generalisable
|
Weakness
|
Non-generalisable
Less quantifiable
|
Measurement device
Lack of specificity
|
Non-generalisable
Measurement device
|
Source: the Author
10The framework therefore allows for a focus on the discursive and policy implications of transparency initiatives in different settings and against different backdrops of socioeconomic and political consolidation. Moreover, as can be seen in Table 2, the approach summarises three key categories relating to the institutional and contextual determinants of fiscal transparency, namely institutional dimensions, measured effects, and characteristics of indices. These categories are observed in the sample of studies considered and include various other elements that permit a much narrower analysis. Hence, the analytical framework is further subdivided according to the defining characteristics found in the studies on the three main dimensions identified.
11It must be said, however, that this approach is not meant to replace the objective-subjective logic with an empirical one. Under the logic of a systematic review (and a meta-analysis), the literal-phenomenological classification is meant to be a guiding typology for understanding the literature and not only assessing results but also the analytical underpinnings that lead to them. As will be discussed, I argue that this framework is better suited to making sense of empirical settings in which little attention is paid to the underlying biases in different data sources. In doing so, I hope that this approach will then be adopted as a more critical reflection on different data compositions.
Table 2. Dimensions of Literal and Phenomenological Fiscal Transparency
Fiscal Transparency
|
Literal/Normative
|
Phenomenological/Experiential
|
Institutional Dimensions
|
Index Characterisationa
|
Measured Effects
|
Efficiency
|
Efficiency
|
Spending
|
Good Governance
|
Scores
|
Sustainability
|
Processes
|
Institutional Dynamics
|
Reputation
|
Corruption Perceptionb
|
|
|
-
In line with the systematisation of the reviewed papers, the index characterisation refers primarily to the OBI.
-
Corruption perception is included as a subcategory because this dimension is used as a common measure of institutional quality.
Source: the Author
12The literal-phenomenological approach also represents a dynamic approach that does not constitute an exclusive classification. As per the division in Table 2, the literal-phenomenological framework is cross-cutting and allows for overlap. Indices, for example, fall into the intersection between literal and experiential frameworks because, as will be discussed below, the characterisation of indices revolves around the design and construction of perceptual items. Hence, the reported efficiency levels or the institutional dynamics at play stem from the reported perceptions of their occurrence or success factors. I address the effects category as a constitutive element of the phenomenological dimension, given that it deals (almost) exclusively with (self-)reporting, e.g., in the case of perceptions of reputations. Noticeably, I place perceptions of corruption under the literal/normative dimension. This is due to the consistent use of corruption perception measures for purposes of assessing institutional quality, which makes it impossible to disentangle them empirically in the studies analysed. With this multi-category outlook, I characterise all studies in the sample and offer a broader perspective for further studies.
13As will be seen in the meta-analytical part of this article, the overlapping nature of this typology also plays a role in the coding process. I argue that several studies analyse fiscal transparency settings from an allegedly objective perspective (what I understand in an expanded form as literal-normative), while overlooking the subjective elements therein (again, from my perspective, relating to a wider array of analytical determinants that constitute a phenomenological-experiential dimension). I therefore take single studies and place them under a multi-category systematisation of the operationalised issues and variables.
14The following section is concerned with developing the dimensions listed in Table 2 in a structured manner alongside the major takeaways from the papers reviewed and analysed. That is, the literal-phenomenological distinction is emphasised, while the three main dimensions guiding the empirical research identified (i.e., institutional dimensions, index characterisation, and measured effects) are characterised.
15This section presents the systematic review of the 35 studies collected, focusing primarily on the main effects and outcomes reported. Thus, using key examples from the sample, I discuss the conceptual foundations of each dimension and how they relate to the operational and narrative levels of analysis of fiscal transparency. This section therefore offers a broad discussion of the empirical approaches of the highlighted studies. It is not only a descriptive approach, but also a critical approach to the main findings and methods. I seek to contextualise the findings of the studies in relation to my framework and their estimated governance and development outcomes. An introductory sub-section introduces the data collection and coding steps followed, underlining the broad scope of the study in relation to its bibliographic foundation. This is followed by a brief introduction to the review dimensions (from Table 2) and an overview of how the studies fit into each dimension (section 3.2). Finally, I present a more in-depth approach to each of the analytical dimensions in separate sub-headings. An overview of the main bibliographic information is provided in the appendix.
16Nearly four thousand studies in the field of transparency were initially identified, and their metadata was downloaded and checked for completeness. Since the purpose of this analysis is to assess the distance between empirical measurements and theorised groundings, only studies assessing a quantitative aspect of transparency were selected. Thus, after filtering by the topic of fiscal transparency, an overview of relevant scholarship (Berliner, 2014; Cicatiello et al., 2021; Heald, 2006) was followed by a snowballing approach within the identified sample. These steps follow the PRISMA protocol (Liberati et al., 2009) as found in similar bibliometric approaches (e.g., Cruz Romero, 2022). This larger set of records included numerous publishing formats, such as papers, reports, books and book chapters on the topic of transparency. The studies were retrieved from Web of Science and PlosOne. Additional studies were searched and downloaded from Google Scholar and Elsevier’s Scopus.
17The 35 studies were selected after several rounds of eligibility checks, excluding all other records that did not meet the corresponding characteristics. The eligibility procedure was carried out using a semi-supervised procedure that checked for keyword matches in titles and abstracts. The studies were then manually screened to pre-code general bibliographic and analytical information (e.g., authors, year and type of publication, as well as variables and approach). The pre-coding step is also the foundation for the systematic review, in that it helped to identify and develop the literal-phenomenological framework introduced previously (see Figure A1 in the appendix). Thus, from this structured filtering, I gained insights into the explicit and latent determinants that I discuss – both earlier and in more detail below. Sections 3.3, 3.4, and 3.5 present and discuss the analytical dimensions shown in Table 2 using key studies as examples for each (institutional dimensions, index characterisation, and measured effects, respectively).
- 3 Not all effects from each study were coded, but only those whose (independent) variables relate to (...)
18The coding for the meta-analysis focused on the individual effects reported by each study. As mentioned above, 350 effects were coded and standardised using a Fischer z scale transformation (see appendix). The distribution of effect sizes between studies is rather large, with some studies reporting as much as 36 effects and others only one (the mean number of effects reported for the entire sample is ten).3 Again, this introduces a significant amount of variance, which is reflected in the estimated confidence interval for the overall effect size (see Figure 1). The latter introduces an implicit bias into the estimation which, alongside other pre-existing biases (e.g., in sample selection, testing, or estimation within each study), may confound the estimated effects with omitted variables or inherent measurement errors that are not even problematised in the primary source (see Higgins et al., 2020). These issues are discussed further in sections 4.1 and 4.2.
19From the scope of analysis identified in the 35 studies, I find that the empirical framing of fiscal transparency relates not to one monolithic concept but to a variety of elements that can be classified into three specific sub-dimensions. Each of these sub-dimensions deals with specific nuances of fiscal transparency, as well as with the analytical (or, as some studies investigate, causal) relationship with governance and development goals, the latter also assessed through a variety of key indicators. As presented in Table 2, I divide these analytical dimensions into a) those focusing (predominantly) on the institutional dimensions of fiscal transparency, b) those assessing the characteristics of the indices in terms of methodological consistency, and c) those focused on the empirical specificities of the estimated effects. Moreover, each sub-dimension is made up of three (or four) topics that shape not only the design and results of each study associated to them, but also the results and outcomes reported (see the rows for each sub-dimension in Table 2).
- 4 Each dimension is a cluster, where each sub-dimension is a categorical variable computed in the met (...)
20It is also paramount to underline the intersection of these three sub-dimensions. As stated above, the literal-normative dimensions of fiscal transparency relate (almost exclusively) to the institutional determinants measured in the scholarship. The sub-dimensions related to the specificities of the indices also fall (partially) under both the literal and experiential typologies. Inversely, the same applies for the sub-dimensions focused on the effects measured in relation to the phenomenological typology. For this reason, I do not define a clear distinction between the sub-dimensions; my coding approach is based on similarities in approaches and findings, and thus classified accordingly. These sub-dimensions are defined as such for two operational reasons: firstly, to guide and structure the review of the studies, where classification and differentiation are essential tasks and secondly, to facilitate the coding of the studies with regard to the meta-analytic dimension of this review4. The classification then transposes these analytical categories into the analysis of each sub-dimension, in particular, as already mentioned, due to the emphasis on the main findings and methods.
21Then, of the 35 studies reviewed, the majority (21) focus on the a) institutional dimensions of fiscal transparency, whether on the efficiency of fiscal policies, on the governance factors that determine them, or on the processes that define them. Slightly fewer (13) focus on the b) characteristics of the indices analysed, specifically assessing the scores, the measured efficiency, or the dynamics that such indices measure. Finally, others (13) focus on c) the effects of fiscal transparency policies looking at outcomes in terms of spending, (financial) sustainability, or reputation gains. Again, the classification is non-exclusive, so studies are coded according to their primary objective; where one study may focus on reported effects, it may do so mainly by assessing institutional dynamics, thus leading to a common classification.
22So, for example, the majority of studies focus on the elements of budgeting as their independent variable, specifically taking the aforementioned OBI as the basis for empirical analysis. As it has been advanced, open budgeting and budget processes represent a key component of the understanding of transparent fiscal governance. These elements bridge to other relevant avenues of governance, in particular through their relationship with deliberation and participation or, from the language promoted by the open government movement, co-creation (e.g., OGP, 2019). Additionally, these studies focus primarily on the systematic effect of open budgeting practices on normative and practical governance factors, i.e., perception of corruption, efficiency, and investment expectations (e.g., Cicatiello et al., 2021; Jung, 2022).
23Coincidentally, it is the element of normative governance that stands out as a cross-cutting paradigm, as it links factors of fiscal responsibility with notions of accountability and, crucially, corruption management (De Simone et al., 2017; Montes & Luna, 2021). The latter emphasises a sequential logic of increased transparency leading to a decrease in perceived corruption. This sequential idea is followed by a number of studies that focus on the constituting elements of fiscal oversight and fiscal transparency – whether through a normative framework or through quantitative approaches, e.g., a fiscal transparency index. These types of studies can be further divided into two research strands that analyse the relationships between more normative effects and more empirical effects: corruption and indices’ construction on the one hand (e.g., Abreu & Gomes, 2021), and spending efficiency and economic performance on the other (e.g., Jung, 2022). Again, this highlights the overarching normative dimension of transparency, by which composite measurements can be effectively incorporated into many different types of analytical frameworks, addressing a wide range of questions.
24However, there is little discussion of the broader structure of transparency in terms of the information-accountability dynamic; instead, studies that focus on information as a proxy for transparency assess an alleged direct relationship from information to budget efficiency. The usual controls are applied and briefly discussed, but rather than reaffirming the empirical estimates, they cast greater doubt on the most appropriate determinants and intervening factors shaping the fiscal impact of transparency. Regarding governance and development outcomes, fiscal transparency acts as a powerful medium to make declarative claims operational and promote effective service delivery. In other words, “[t]ransparency becomes more important than a rigid interpretation of bureaucratic regulations” (Spahn, 2006, p. 97) in that it represents a covenant for public-private interactions in search of more active participation.
25One of the key areas of research on the impact of fiscal transparency on governance and development outcomes has to do with the institutional dynamics affecting transparency policies and initiatives. Within this framework, three major topics are addressed in the literature, namely transparency in terms of a) efficiency, b) good governance, and c) institutional processes. A fourth, i.e., corruption perception, is also a factor in this sub-dimension, as it is often associated with institutional quality. Many studies touch on these three (or four) topics in parallel, but my classification in this regard refers to which of them is specifically problematised and framed as predominantly explanatory in its scope. These topics relate primarily to the explanatory mechanisms presented in the empirical designs, and thus exemplify a rich body of scholarship that focuses on the effects of fiscal transparency on a set of highly intertwined factors related to governance dynamics and structures.
26An important study carried out within the IMF framework correlates fiscal transparency measures (a general index focused on distinct aspects of openness and oversight) with economic outcomes in factors such as credit ratings (Hameed, 2005). Its results are marginally positive for the effect of transparency in the economy (broadly measured), but this broadness, coupled with a large variance, shows how sensitive such results are. However, an important methodological consideration shown in this paper is the assessment of different samples along the predefined transparency proxies in order to provide different baseline results according to distinct groups of countries classified under economic growth parameters. The main takeaway is summarised then in the institutional diversity of context-dependent processes, highlighting an underlying limitation of the multi-regional, macro approach.
27Working with a similar dataset on a subnational index of fiscal transparency, two studies focus on similar designs and attempt to link this index to broad dimensions of budget administration (Krause & Melusky, 2012; Li & Feiock, 2020). On the one hand, Krause and Melusky analyse the impact of fiscal transparency on expenditure growth, i.e., how open budgeting structures lead to more effective budget planning (highlighting the relevance of efficient planning and governance processes). On the other hand, Li and Feiock offer an updated take on the (mostly identical) dataset, assessing how the fiscal transparency index correlates with fiscal stability, or “punctuation” as the authors put it (2020, p.2). The results for both analyses are as expected, highlighting the intuitive directionality of the effects. For example, in the first study, the authors find a negative effect of fiscal transparency on expenditure growth; this is relevant for two reasons, firstly because it shows that participatory spaces have an institutional impact, and secondly because this same civic impact is translated into otherwise desirable outcomes (more efficient budget allocation). In the same vein, the second study finds a positive correlation between transparency policies and budget stability (interpreted inversely, a negative correlation with budget punctuations). Transparency, but also a host of political-administrative intervening variables, play different roles in the relationship, highlighting the role of contextual and structural determinants of policy outcomes at the institutional level.
28Furthermore, as mentioned above, the OBI stands out as a recognised analytical dimension used to assess fiscal transparency from a normative perspective. Some studies use this index and propose different methodological approaches regarding the respective dependent variables. Key examples of this approach focus on distinct elements of fiscal governance and governance in general (Cicatiello et al., 2021; Cuadrado-Ballesteros & Bisogno, 2022; Montes & Luna, 2021). From expected FDI inflows to corruption perception, research on the effects of transparency through institutional channels offers many avenues to pursue. Again, these institutional channels are co-determined by elements surrounding the good governance framework, but also by oversight and efficiency.
29Another perspective from which the institutional sub-dimension is evident relates to econometric approaches to the impact of transparency on financial stability. The results presented by Cuadrado-Ballesteros and Bisogno (2022) provide a methodologically sound perspective on the various mechanisms that influence the adoption of transparency policies in the pursuit of greater financial stability. The results indicate a positive relationship between transparency in budgetary processes and the capacity to spend and generate debt. As in previous studies, the findings presented are intuitive and help to consolidate some background evidence on the role of transparency in policy actions (e.g., Guillamón et al., 2011; Ríos et al., 2016). In particular, the authors point to the relevance of the study on both theoretical and practical levels, arguing that making such institutional processes transparent is not only a normative but also a practical tool of public administration. Nevertheless, as is the case with the subject of this review, the scholars note that the estimated effects are neither causal nor direct, as the operationalised concepts are complex and only partially observed (Cuadrado-Ballesteros & Bisogno, 2022 p. 226).
30Similarly, the work of Cicatiello et al. (2021) identifies the mechanisms through which transparency policies impact investment flows in a broad sample of countries. The research also highlights the econometric approach as a key methodological tool to untangle diverse relationships (causal or associative) between the identified variables. In the case of this study, economic controls are included to argue in line with disciplinary standards, but crucially, political-institutional variables are added to precisely pinpoint the intervening mechanisms in the transparency-investment relationship. The scholars also find intuitive results that support the idea that greater transparency in fiscal processes (especially budgeting) is positively correlated with greater (foreign direct) investment. In addition, the authors find that political dimensions have mixed effects (of varying significance) regarding the hypothesised effects. Overall, the authors argue that the “main channels through which fiscal transparency attracts FDI are by making government preferences explicit, and by increasing government commitment to fiscal evaluation” (p. 5). Moreover, the results also highlight how different framings of a single concept can alter an estimation; in this case, the effect of open budgeting is particularly strong when assessing its execution is assessed rather than just its existence. This is in line with good governance stipulations and therefore highlights the institutional processes as a key analytical factor.
31In a similar vein, but from a perspective more focused on the intersection of political and institutional dynamics, the work by Montes and Luna (2021) assesses the impact of fiscal transparency on corruption control, specifically in contexts where discretionary mandates are in place. It links to the study by Cicatiello et al. (2021), as corruption is one of the key control variables that has been shown to have little effect on how investment is driven (attracted or diverted) by a transparency outlook. In their study, however, Montes and Luna find an indirect effect of transparency on the control of corruption (perception) through a rule of law composite indicator (Kaufmann et al., 2011) in which the strength of the judiciary institutional system is systematised. Additional co-variates are included to control for perceptions of the independence, impartiality, and integrity of legal entities. In terms of the literal-phenomenological framework, this study presents a paradigmatic case in which, when dealing with questions of efficiency, inherent problems of measurement precision are taken for granted.
32Regarding the characterisation of different measures, similar to the previous sub-dimension, three topics constitute the main analytical approach. Namely, I identified the role of transparency indices in terms of a) efficiency, b) scoring schemes, and c) institutional dynamics. The studies classified in this sub-dimension thus -focus on the role of the policy dynamics captured in the quantifiable dimensions. For example, information policies are one of the key factors assessed in the sample. Such policies are based on basic characteristics of information provision. Hence, the studies identified and included in this review category are characterised by their focus from information to governance goals. Three key studies stand out due to their emphasis on the characteristic dynamics surrounding the different measurements employed.
33For example, the first of the highlighted studies points to the broad dimension of bureaucratic efficiency as measured by the International Country Risk Guide (ICRG) indicators (Vadlamannati & Cooray, 2016). The authors provide insights at a micro-level, assessing the role of information availability regarding the efficiency of government institutions, controlling for the usual dimensions of economic and political characteristics (pp. 7-8). Aware of the risks of unobserved heterogeneity and endogenous biases, the researchers turn to an instrumental variable (IV) estimation in two stages. Regional and cultural dimensions are used in this regard to attempt to control for the influence of similar polities in contexts where geopolitical interaction is high. Another dimension observed by the authors has to do with revenue from natural resources, as extensive literature (in transparency studies) stresses the influence of these (Kolstad & Wiig, 2009; Sovacool & Andrews, 2015). The authors report estimates of positive correlations for the level of bureaucratic efficiency, meaning that the availability of an information and accountability framework tends to be a determinant of more efficient institutions. Contrary to some perspectives on the matter, the existence of such normative and regulatory frameworks does not “stifle bureaucratic autonomy and thereby reduce efficiency” (Vadlamannati & Cooray, 2016, p. 18). The authors seek to extend these findings to the interaction of key dimensions associated with civic freedom (free media, political competition, presence of non-government organisations) and report more positive correlations. This strengthens the intuition that the more information is needed by social organisations, the more government institutions need to be involved in the provision of information thus, making such provision channels open and efficient. Additionally, the approach highlights the extent to which potential ‘subjective’ biases are cause for concern in terms of achieving statistical robustness.
34Similarly, in both the micro-level analysis and the focus on information (as a key determinant of governance and development outcomes), the research by Honig et al. (2022) focuses on the availability of information on specific projects to analyse implementation success. It uses a large cross-sectional dataset of thousands of donor agency projects with their respective scores and assessments. It has a specific focus on financial reporting, since this is a key component of aid projects that determines their sustainability and, arguably, their success. The methodological design is highly relevant as the authors highlight the mechanisms driven by information availability alone, and the critical role of accountability channels, which they operationalise as “appeals mechanisms” (pp. 9-10). The authors also test geopolitical variables to control for potential unobserved but latent biases (endogeneity and autocorrelation) in the data.
35Their argument for controlling for such mechanisms is similar to that in the paper described earlier; on the one hand, spatial and political proximity may determine the policy expectations of a government and/or its citizens. On the other hand, since the unit of analysis comes from specific organisations (donor agencies), these are also correlated with the national interests of the funding parties (for a broader methodological discussion, see Berliner, 2014) and, as argued here, are often interpreted as true information (objective) from perception reports (subjective). Beyond the methodological nuances, the authors find a reassuring effect that supports the idea introduced previously; namely, that the availability of information alone does not determine positive governance outcomes. In order to “help to repair the broken feedback loop between public institutions and beneficiaries by reducing information asymmetries within the multiple principal–agent chains connecting them” (p. 18), information policies need effective accountability mechanisms, especially when financial information is at the centre.
36Finally, and doubling down on the micro-level dimension that characterises most of the work in this category, the research by Abdellatif et al. (2016) presents an analysis of policy debates on fiscal governance at the level of legislators. Their research is relevant in two respects: firstly, the breadth and scope of the empirical analysis provides a panoramic outlook of policy-making in the parliamentary setting, assessing expectations and incentives (i.e., the particular institutional dynamics of parliamentary politics). Secondly, the research points to the reiterated factor of the information-accountability linkage; information alone is not a panacea, especially in the case of financial data where “[i]f unchecked, these could be used as tools to achieve political or bureaucratic interests in case of asymmetric information” (p. 1). The researchers double down on the role of political motivations as a determinant of the extent and intensity with which politicians argue for or against fiscal openness. Expectations of political outcomes may undermine the critical acceptance or development of specific fiscal transparency policies. Although not explicitly problematised, these ideas are directly related to my presentation of the phenomenological dimension of transparency, emphasising the experiential component. The baseline of transparency in the research context is low, but the lesson stands as it shows that the need for enforcement is paramount in order to check the potential negatives of fiscal governance.
37Along the same lines, the studies that primarily focus on the specificities of effects and methods also have a cross-cutting dimension defined by how corruption can be incorporated (e.g., Chen & Neshkova, 2020). Intuitively, the relationship is positive; the more transparent governments are, the better the perception that corruption is under control. A number of mediating variables exist and are also present in the fiscal transparency sub-field, as they highlight the role of specific institutional designs or the mere consolidation stage of liberal democratic organisations. The studies classified in this category are also co-determined by three main topics dealing with the effects of transparency on a) spending, b) financial sustainability, and c) perceptions of reputation. In my sample, some studies stand out for their focus on thematic and methodological dimensions of transparency. These studies link corruption issues to notions such as opacity, which has been identified as a determinant of poor governance and low economic performance (Heald, 2015), i.e., more (unchecked) spending, more fiscal pressure, and less institutional credibility.
38More precisely, the research by Brusca et al. (2018) for instance, shows that the relationship between fiscal transparency and development outcomes is cross-cut by a number of (intertwined) co-determining factors. The researchers focus on specific dimensions of the OBI as a way of operationalising fiscal transparency. As such, they stress the role of accountability as an overarching dimension and estimate the effect of the quality of accounting mechanisms. It is relevant to highlight this idea, since transparency research tends to focus only on the role of information, whereas the authors of this study identify the underlying link between information and audit systems. They find, as expected, a positive relationship: the better designed and empowered the audit entities are, the better the perception of governments’ handling of corrupt practices.
39On a similar but broader scale, a study conducted by Abreu and Gomes (2021) shows that the correlation between budget openness (taken as a general score from the OBI) and democracy (as a broad schematic measurement of polyarchy dimensions compiled by the Economist Intelligence Unit) shows a positive sign. The authors estimate an interaction term between these two variables across a balanced panel to confirm their expected results regarding the perception of corruption (p. 9). However, the authors caution against the over-generalisation of limited quantitative results, particularly with regard to the role of institutional dynamics. They note that “[c]are must be taken not to jump to conclusions too quickly about the results of democratic countries that have imperfections in their institutions” (p. 11), as the paradigmatic case of Brazil (the authors’ home country) reflects a normative and empirical dissociation in terms of fiscal transparency and corruption perception. In other words, it highlights the fuzziness of quantitative measurements vis-à-vis specific contextual determinants or, again, the ‘objective transformation of subjective appraisals’.
40Focusing on such regional and national specificities, the study by De Simone et al. (2017) takes on corruption perception but correlates it with another measure of fiscal transparency – this one provided by the IMF (Wang et al., 2015). The researchers control for a number of standard econometric dimensions, such as wealth, growth, and taxation income, but also focus on specific determinants of a socio-political nature. For example, the study assesses whether and to what extent a country is considered to be linguistically and ethnically fragmented, as well as by religious denomination. These controls are meant to identify cultural differences affecting the ways in which both transparency and corruption are framed. Similar to previous studies, the authors estimate a dynamic panel regression and find a significant and negative relationship between fiscal transparency and corruption – the caveat being that, in this study, the authors operationalise corruption as a hard measure rather than its perception. Thus, an increase in transparency (according to the index used) is correlated with a decrease in corruption (as per the data sources). Indirectly, the authors recognise the limitations of the underlying data that I have pointed out, but further discussion seems to be dismissed.
41A common case of methodological clarity in these studies (and many others in the sample) has to do, as I argue in this review, with the fact that different indices give different estimates. Most strikingly, however, is the diverse use of the same or similar indices that, even with similarly sized samples, give different estimates (whether in its size or statistical significance). Of course, covariates and controls vary from study to study, but the problem of different indices remains, whether one is analysing transparency broadly or in a more limited sense. The same issue arises when considering dependent variables such as corruption measurements (from its reported perception to its supposed real quantification). Again, this issue highlights the literal-phenomenological framework I proposed, based on the premise that seemingly objective indicators depend on a myriad of contextual features. Furthermore, as I have argued, a subjective dimension that reflects researchers’ paradigmatic preferences is deeply embedded in the specific data sources and methods.
- 5 It is important to note that testing my hypotheses is not central to my approach. Rather, I expect (...)
42The meta-analytic approach requires a working hypothesis to be defined prior to the meta-estimation of the overall effect size. In this case, as I have argued as the core problem with measurements of transparency, I start from the assumption that the expected governance and development outcomes are not entirely dependent on the operationalised independent variables5. This has to do with different elements: on the one hand, there are research design factors such as time-series analyses or large N studies. On the other hand, this expectation is dependent on other methodological elements, such as the cases studied, or the type of transparency conceptualised. The main issue lies not in the direction or size of the expected effect, but in the determining factors that correlate with the estimation itself, in other words, in the methodological aspects that define the statistical robustness of the estimated effects. In this respect, the effect size transformation for the meta-analytic approach considers the sample sizes as well as the reported errors and variances (see Harrer et al., 2021). For this reason, and because of the different levels of sample sizes and reported error terms, it is expected that a conclusive effect size may not be found.
43Other meta-analytic factors (moderator variables) were included in the estimation of the overall effect shown in Figure 1, as they “systematically influence the magnitude of the effect size in a study” (Chen & Ganapati, 2021, p. 262). For the purposes of meta-analysis, they take the form of dichotomous values indicating their presence or absence in each study. In this case, the set of moderators is divided into three broad categories: a) those related to the specific dimensions of fiscal transparency (the institutional dynamics, the index characterisation, or the measured effects; broadly speaking, they relate to the response variables operationalised), b) those related to the observed relationship between transparency and governance and development outcomes (through the independent variable classification – see above), and c) those related to the methodological design (e.g., has an international scope, measures many time periods, or uses more comprehensive panel data). Furthermore, an additional factor was considered in the meta-equation as a moderating variable, namely the type of transparency studied, whether it is framed as literal or phenomenological, as per the analytical differentiation shown above.
44The estimated overall effect (0.06 [0.02, 0,10], p = 0.01) is positive and significant, providing evidence that, despite specific methodological omissions and biases, fiscal transparency is correlated with higher development outcomes. The overall effect represents the estimated change in governance and development outcomes from a total change in fiscal transparency. For example, a 100% increase in fiscal transparency policies would, on average across its different values, lead to a 6% increase in reported governance and development outcomes. This type of result would strengthen previously reported but broadly consistent results in the literature (de Renzio & Masud, 2011; de Renzio & Wehner, 2017). The results are broadly consistent due to the aforementioned differences in the meanings and contexts in which fiscal transparency is observed and measured. Furthermore, this very interpretation is an example of how abstract such estimates can be; moreover, what would a double increase of a policy imply? What would be the baseline? And, more importantly, how does this translate into effective policy planning? The dependence on context specific dimensions (e.g., political system, institutional independence, or even democratic consolidation) is thus an unobserved factor that generates bias in this regard. The lack of uniformity in the literature permits only a macro focus on the relationship between these dimensions.
45As for the estimated heterogeneity of the overall effect, the reported I2 (p = 0.001) of 97.45% indicates that almost all of the variation between the reported effects is due to actual variation in the effects. This is consistent with my efforts to include diverse studies that represent a rich source of empirical work through academic articles and grey literature. This heterogeneity is also evidenced in the results of asymmetry tests – Begg’s test (0.08, p = 0.535) and Egger’s test (0.56, p = 0.580). The non-significance of the tests indicates a failure to reject the null hypothesis (i.e., that there is publication bias in the sample). In other words, the sample seems to be balanced in terms of the inclusion of academic and professional literature, as shown in the funnel plots included in the appendix.
46Regarding the significance of specific moderators, I stress that the estimations show specific effects with respect to the three categories, namely the grouping of independent variables, the types of transparency, and the analytical dimensions. Several factors show some significance in the meta-regression specifications, indicating a broad relevance of such variables in the studies. As part of the baseline estimation, I also use a set of moderators related to distinct publication and methodological elements. These moderating variables help to capture any variance in the breadth and scope of publications in the sample, while also pointing to specific analytical nuances, such as estimation based on (one or multiple) time periods or panel-type data.
47Further, the signs of some estimates are noteworthy as they highlight the relevance of the variable conceptualisation to the overall methodological design. For example, this is seen in the grouping of independent variables focusing on information availability, where the estimate is significant but negative (both in the baseline and full models). This highlights a problematic premise, namely that approaching information availability normatively only evidences and underlines its deficiency (e.g., Vadlamannati & Cooray, 2016; Yu et al., 2021). Similar to the overall effect, this outlook is only indicative of the relevance of properly assessing contextual factors in the methodological design when assessing (fiscal) transparency (see Andreula & Chong, 2016).
- 6 Again, I highlight the coding and conceptual background for this classification in order to problem (...)
48Similarly, as argued earlier, the effect of a phenomenological framing of transparency (understood as the experienced distance to the measured type of transparency variables) seems to be a relevant factor, especially in direct comparison with the effect shown in studies using a literal framing of transparency. In other words, studies reporting on perceptions of transparency tend to report more optimistic results than those based on quantitative and index-based data (as in the case of corruption; see also Montes & Luna, 2021)6. Interestingly, however, the inclusion of time periods (as one or multiple cross-sections – but not panel data) is associated with negative effects in the correlation studied. This phenomenon may be driven by a) the fact that overall scores deteriorate, from one period to the next; b) differences in samples in different time periods, or c) unaccounted for measurement differences that generate implicit biases. The latter two seem more credible, although the first is plausible as it would signify a general deterioration that would call into question the whole object of the study and its effects.
49It is important to note that the meta-regression estimation of the moderators’ effects was carried out using a Weighted Least Squares (WLS) random effects approach (models 1 – 5 in appendix) (see Viechtbauer, 2010 for details), as well as a Generalised Method of Moments (GMM) approach (model 6) (see Bakbergenuly & Kulinskaya, 2018).
Figure . Forest Plot for Effect Sizes for Fiscal Transparency
Source: the Author
50Regarding the main arguments discussed in the first part, these results are indicative of the underlying phenomena surrounding the measurement, the framing of the data, and the conceptual laxity of (fiscal) transparency. Further, I show the wide variance with which studies have reported their findings within the field. These differences provide a consistent indication of how researchers design their methodological approaches, particularly regarding case selection, sample size, and time dimensions in their empirical studies. As these latter categories are also taken into account in the sub-group analysis, the estimated overall effect provides tentative evidence of the importance of such key methodological features. More broadly, these elements also call into question the contested nature of transparency research agendas. In particular, in the field of fiscal transparency, it could be argued that there is no consistent framework and very little follow-up research. Whether this is an issue of data availability, replicability, or even research agendas needs to be investigated further.
51The estimated positive overall effect is then hardly cause to celebrate the virtuous impact of fiscal openness on development. It could be questioned whether sample selection induces bias into the respective estimates, or whether the operationalised variables are treated with the same robustness across the sample studies. However, as noted above, these elements overlap in most cases and the empirical designs are based on common statistical approaches; my classification into different categories serves the function of establishing typologies that can be empirically estimated. Thus, despite the atomisation of studies into different fields and sub-fields, the broad dispersion of results (and the variance within them) call for further study, potentially into a unifying framework that could serve as a basis for future reference.
52A wide diversity of approaches and analytical foci serve as a basis for studying the effects (positive or negative) of transparency policies on governance and development outcomes. Correspondingly, a variety of measurements have been operationalised to reflect the conceptual foundation of transparency for organisations and administrations. This heterogeneity offers both positive and negative perspectives for the study of transparency, particularly regarding the fiscal dimension assessed here. On the one hand, it allows for a multidimensional approach, stimulating different research designs that take into account specific analytical aspects in particular governance contexts. On the other hand, it creates the situation highlighted earlier: an inconsistent research agenda, lacking a conceptual basis and offering contrasting results. These characteristics atomise the field of study and impede general research, thereby accentuating the under-conceptualisation.
53It is not surprising, then, that studies focusing on distinct levels of governance produce conflicting results on issues that are, at least conceptually, closely related. For example, national case studies using either cross-sectional or time-series data (or both) regularly diverge from the findings of locally focused studies (e.g., cf. Galli et al., 2020; Hameed, 2005), regardless of the time dimensions used in the latter. Similarly, studies with large N dataset are useful for uncovering patterns of dependency and policy development (De Simone et al., 2017), highlighting overarching policy topics. Studies in this category offer analytical perspectives in many directions of national and supra-national policy-making, which can be critically reflected in terms of world systems and globalisation dynamics. In contrast, the analysis of single (or reduced) case studies is helpful in delving deeper into governance specificities that only possible with greater contextualisation and detailed policy background (Jung, 2022). Such studies offer rich data sets, including administrative records that would otherwise not be considered outside economic or accounting realms. Few studies combine both perspectives, as the methodological hurdles make such efforts particularly challenging (Alt et al., 2006 offer a longitudinal analysis of US State governments, which are by most accounts divergent cases concerning fiscal governance and transparency). Hence the inconsistent results when using one or another measure to operationalise fiscal transparency.
54Through a systematic review and a meta-analytic approach, I shed light on analytical (in-)consistencies and propose a novel framework with which to undertake a critical assessment of fiscal transparency. In this way, the empirical results from quantitative studies analysing governance and development implications can be better understood in the context of conceptual and empirical diversity. I therefore emphasise my intention to distil the methodological and analytical diversity into a set of more manageable, albeit general, dimensions.
55By arguing along the lines of an analytical distinction regarding the type of transparency addressed (literal – phenomenological), but also by summarising the different studies according to their framing of both independent and response variables, I have shown that there is still little coherence. This incoherence is evidenced by the wide range of methodological designs and results (although there is an overarching similarity regarding almost all statistical techniques). For example, in terms of the analytical dimensions summarised from the studies, the classification into these categories is non-exclusive and, as problematised, clear parallels are difficult to identify. The same applies to the overarching differences in the operationalised independent variables. The divergent outlook between studies focusing on one or the other aspect of fiscal governance seems to depend on contextual factors that are not entirely unbiased and not fully clarified in the respective discussions of the empirical specificities of each study.
56Thus, even though most of the studies reviewed carry out some form of robustness check, and despite the assumed independence of the samples, a certain amount of variance seems to contaminate the estimated meta-regressions. Furthermore, beyond the nuances of different research designs and empirical settings, some degree of within-study bias also seems to be a source of misdirection in the estimates reported in the different studies. Again, despite many studies using the same explanatory variables (e.g., OBI) with similar control variables and similar (if not the same) response variables (e.g., financial sustainability) (cf. Cuadrado-Ballesteros & Bisogno, 2022; and Sedmihradská, 2019), the magnitude of the (averaged and transformed) reported effects diverge along with the reported significance. Doubts arise about the validity of one or the other result (admittedly, one is a peer-reviewed paper, the other a working paper), but the statistical approaches are not entirely different. This finding emphasises the pluralistic approach to the topic from many disciplinary perspectives which are not necessarily aware of similar studies in other fields.
57I propose a conceptual framework to understand the divergence within the literature through my types of transparency framework. From this, I argue that (at least) some degree of bias can be recognised and better understood in relation to the current pitfalls of an atomised field of study such as fiscal transparency. By identifying each study within the categories of literal (normative) or phenomenological (experiential) transparency, I was able to estimate the difference that this typology entails. As argued in the previous sections, perception-based measurements may, or may not, be entirely reliable due to implicit biases; the results seem to support this claim, as the phenomenological category is more highly correlated with positive effects of fiscal transparency on governance and development outcomes. This finding is hardly unintuitive as the framing of a perceived level of influence may be exaggerated, time-constrained, and malleable in the face of public opinion; this contrasts with the literal framings of transparency, which are limited to actual levels of influence and require a more complex measurement approach.
58For example, as mentioned above, the use of the OBI corresponds to a literal/normative representation of a particular state of (fiscal) affairs. As stated in the core document from which the OBI is developed (International Monetary Fund, 2018), the factors operationalised in the index pertain to “the clarity, reliability, frequency, timeliness, and relevance of public fiscal reporting and the openness of such information” (p.1). Complementing these transparent features, emphasis is also placed on an overview of “the strength of budget oversight institutions such as [...] opportunities for public engagement in the budget process” (de Renzio & Masud, 2011, p. 609). In general, this type of measurement systematises a wealth of information, combining subjective and objective sources to arrive at a synthesis of the current state of policy development. Additionally, and therefore crucially, the focus adopted in many research settings highlights one or more surrounding elements, depending on the response variable observed; this relationship determines the causal and/or confounding mechanisms that scholars have identified in numerous examples. Thus, some studies may compound internal biases and under- or over-estimate the extent to which actual transparency policies are effective.
59Moreover, the field of fiscal transparency is not exactly embedded in a critical perspective that questions the pitfalls of the empirical-analytical paradigm. Rather, it is a field heavily dominated by econometric and economistic approaches, mostly based on the underlying assumptions of efficiency (again, in combination with the NPM characteristics discussed earlier). This trend is observable in the space devoted in each study to a) problematising the data sources selected for the statistical models, and b) describing the specifications, results, and further statistical testing. As shown in the previous sections, such bias is also evident in the categories derived and the main factors discussed in Section 3. Despite its broad disciplinary origins, the sample is methodologically and conceptually skewed towards highly quantitative approaches. Therefore, I attempt to shed light on the empirical foundations behind the literal-phenomenological dimensions and problematise the lack of critical discussion in terms of the ‘objective transformation of subjective appraisals’. In doing so, this study provides a starting point for further discussion of the analytical dimensions within different empirical settings (quantitative or qualitative); as such, it can serve as a nexus for future scholarship in this field.
60However, my study is not free from the same biases or shortcomings that I criticise. Firstly, I acknowledge my strong argument about the nature of quantitative approaches to abstract factors such as transparency. However, this critique is the guiding narrative of my work and motivates the scope of my research; moreover, I argue on the basis of the conceptual and methodological underpinnings of the work reviewed. In this way, I maintain a comprehensive analytical approach as well as my own positioning. For this reason, I strongly emphasise the findings from the systematic review and the insights estimated through the meta-analysis, which serve as evidence of the problem with measurements extensively discussed.
61A further limitation is that the underlying data for the review and analysis is limited to a relatively small sample as discussed, the field of fiscal transparency is rather atomised and not well interconnected (beyond a few key studies). The search and eligibility criteria used for both the review and the meta-analysis allowed us to access only 35 studies within the transparency literature that focused on the effects of fiscal governance (through transparent policies) on development outcomes. Therefore, further research could undertake the task of broadening the sources by searching within more bibliographic databases so as to include more studies from under-represented epistemic spaces.