Accountability, Corruption, and Opposition Groups: Evidence from Latin America

Abstract

Under democratic regimes, accountability discourages the existence of corruption, because constituents typically constantly observe public servants. Although every citizen can drive accountability, only a few sectors actively engage in it, such as opposition groups, to a greater extent and for different reasons. Through a comprehensive measure of accountability developed by Varieties of Democracy, which includes vertical, horizontal, and diagonal accountability, this study aims to explore the relationship among the type of opposition in government, level of accountability, and level of corruption, to unravel the relevance of the characteristics of opposition groups when holding a government accountable. Are governments less corrupt when opposition groups from civil society exercise accountability? This study employs latent Markov class analysis applied to Latin America (1980–2020) and concludes that high levels of accountability are achieved and low levels of corruption are evidenced with the existence of opposition groups from civil society.

1. Introduction

The concept of accountability, which refers to the mechanisms by which individuals or groups of people are held responsible for their acts, has become an important and debatable issue within democracy studies (Bovens et al. 2014). Although accountability and democracy are closely related concepts, they are not interchangeable. Formally, democracy is defined by entitlement of individuals to proportional influence over collective decisions that affect them. In this case, accountability supports this democracy-defining norm by generally connecting citizens to agents who make and organize these decisions on their behalf (Lindberg 2009). Thus, accountability enables the consolidation of democracy, because it enables a sense of responsiveness from elected officials and public servants to constituents in general.

Under a democracy and, consequentially, under accountability, public servants should respond to the demands of citizens and hold transparency and honesty as inherent values of the public chairs they hold. Although this study does not discuss the intrinsic nature of public service, accountability would guarantee the appropriate functions of a democracy in terms of preventing the existence of negative practices, such as corruption and embezzlement, because conducting these practices is severely punishable by the relevant institutions and the citizenry (Bovens et al. 2014). Despite increased worldwide efforts to reinforce the appropriate mechanisms to hold politicians accountable to citizens, malfunctioning governments remain a concern for democratic consolidation. One of the possibilities for better understanding this includes the exploration of who, when, and how accountability practices develop.

Oppositional forces play a crucial role in terms of holding the government accountable and preventing corruption. Although one can intuitively predict that groups aligned with the status quo would keep the government less accountable than oppositional forces from civil society, this relationship lacks exploration. Naturally, people in power (in terms of political or economic influence) employ a strategic interest in undermining the de facto effectiveness of institutions of accountability. In contrast, oppositional forces intend to obtain stronger means for exposing the illicit practices and failing decisions of ruling elites (Mechkova et al. 2019). Many contemporary leaders decide this struggle in their favor by introducing de jure institutions but manipulate them to limit the constraints on their rule (Schedler 2013).

Similar to studies on the relationship between accountability and corruption, the academic literature on the negative consequences of corruption for human and democratic development is vast (Drury et al. 2006; Shen and Williamson 2005; Chatterjee 2022). However, the understanding of the causes of a functional (non-corrupt and accountable) government remains limited. In this sense, this study contributes to the literature by exploring the relationship not only between accountability and corruption but also of groups that drive it; by doing so, the study explores the role of oppositional forces in keeping governments accountable. Furthermore, the study aims to elucidate the possible transitions from one “state of accountability” to another, such as the plausibility of migrating from a low accountability and high corruption state to a high accountability and low corruption state.

Exploring this issue is relevant for Latin America, as it remains a region often struggling with the consolidation of democracy. According to Munck and Luna (2022) one of the main problematics in Latin America is the lack of state capacity under which institutions often fail to guarantee impartial check and balance mechanisms. Furthermore, it is relevant to analyze the role that opposition groups play in a politically heterogenous region: Latin America includes governments with different levels of openness to citizens’ participation, posing an interesting framework to evaluate the relationship among the variables of interest considered in this study.

The remainder of the paper is structured as follows. The first section explores the relationship between accountability and corruption and presents the Varieties of Democracy (V-Dem) conceptualization of accountability. The second section explores the role of oppositional forces when fostering accountability and demonstrates the permeation of this relation according to the type of oppositional group. The methodological section then applies an empirical strategy to examine the effect of the type of regime opposition on the probability of reporting particular levels of accountability and public corruption. Finally, the concluding section discusses the main findings and implications of the study. In summary, the article concludes and demonstrates that accountability increases, and corruption decreases, when oppositional groups emerge from the sphere of civil society.

1.1. What Is Accountability and What Is Its Relationship with Democracy?

To elucidate the relationship between accountability and democracy, we first need to explore what democracy entails; nonetheless, defining and characterizing democracy are difficult tasks (Treier and Jackman 2008). A parting point is Schumpeter’s (1942) minimalist definition, which captures the institutional basic principle of democracy: a regime is democratic when it holds contested elections through free and open participation. The latter principles are complementary, because we cannot ensure that a country is democratic if full participation exists but not contestation and vice versa.

Despite the usefulness of this minimalist definition, one may overlook the other relevant defining components of democracy by acknowledging only the electoral dimension. Complimentarily, Dahl (1971), who goes beyond the electoral dimension of democracy, returns to the origins of the definition: democracy embodies popular control over collective decision making as well as political equality. In his theorization, Dahl recognizes that a few key “guarantees” must be met for a society to be democratic. In other words, to exert popular control over collective decision making, holding politicians, who literally represent the people, accountable, is necessary. In summary, democracy evidently builds on the notions of contestation and participation; additionally, the existence of certain types of mechanisms of checks and balances is crucial.

The concept of accountability becomes relevant at this point. Accountability is a relationship between two actors, where “A is accountable to B when A is obliged to inform B about A’s actions and decisions, to justify them, and to suffer punishment in the case of eventual misconduct” (Plattner et al. 1999, p. 17). Based on this definition, this study infers that accountability is necessary to ensure democracy in the sense that it enables the sharing of power and public control over the use of public resources (control over collective decision making). In this sense, accountability contributes to the reduction in the risk of the abuse of power and corruptive practices, which, in turn, is essential for ensuring the fulfillment of the basic rights of the people (Tsai 2007). Lastly, well-functioning accountability relationships largely contribute to building trust in state institutions.

In summary, assuming that accountability acts as a gatekeeper of democracy to ensure that politicians, in this case, are responsive to constituents is possible. However, how can accountability be observed in action? The extant literature responds mostly based on the electoral component of democracy, that is, it frequently focuses on how citizens use elections to hold politicians accountable (Cheibub and Przeworksi 1999; Thomassen 2014). However, empirically, a potential substantial variation exists in the degree to which accountability actors, apart than voters, constrain governments (Lindberg 2013). Although certain actors display strong and systematic repertoires of collective action, such as protests and strikes, other actors from civil society are more sporadic. An example of the former is an indigenous movement in Ecuador, which featured mass mobilizations that concluded with the overthrow of several presidential mandates considering unmet social demands. Such episodes illustrate that accountability does not function only through elections, although they remain a crucial mechanism for holding governments accountable.

1.2. Types of Accountability

The academic scholarship in political science emphasizes the importance of accountability in preventing the abuse of political power. Similarly, a concern emerges for checks and oversight in states to guarantee democracy at large (Schedler 1999, p. 13). In this sense, although an agreement exists on the importance of accountability within a democracy, a great divide is observed in its empirical conceptualization and eventual measurement, which renders difficult the development of an in-depth understanding of the mechanisms related to the concept and its impact on other dimensions. Similarly, research on the impact of opposition groups on accountability remains in its infancy, that is, knowledge of whether or not certain characteristics of opposition groups play a role in the achievement of certain levels of accountability remains scarce (Voltmer 2010; Voorn et al. 2019).

Concretely, two challenges exist in relation to the concept of accountability, namely, conceptual ambiguity and limited data availability. The existing measures, such as the Voice and Accountability Index of the World Bank, are criticized for their conceptual inconsistency, lack of transparency in construction, and limited coverage (Thomas 2009; Apaza 2009). Consequently, researchers resort to measures of democracy to proxy for accountability, which poses a problem in comparability, coverage, and theoretical soundness (Gerring et al. 2012; Harding and Wantchekon 2010).

Within the most recent developments in the measurement of accountability, the study refers to the scholarly work developed by Lührmann et al. (2020), who organized accountability into subtypes on the basis of the spatial direction between its actors (Lindberg 2013). The authors operationalize accountability into three categories, namely, vertical, horizontal, and diagonal. Vertical accountability is a relationship between unequal institutions (the government and citizens), and horizontal accountability is a relationship between more or less equal institutions (different branches of government; O’Donnell 2001). Diagonal accountability represents the extent to which actors outside of formal political institutions (the media and civil society) hold the government accountable.

Empirical research (Laebens and Lührmann 2021) illustrates that diagonal accountability, in particular pressure from civil society, is a vital doorkeeper to democracy. Although protests and civil society pressures, by themselves, are insufficient for pressuring politicians to heed the demands of the people, fulfill their electoral promises, or assure their own transparency, they may be crucial for the effectiveness of other mechanisms of accountability, which influence levels of accountability overall.

In general, diagonal accountability features opposition groups within a government; hence, examining the characteristics of opposition groups within a government is key to understanding accountability and its relationship with other variables of democracy. Procedurally, to obtain an incentive to check on the incumbent and to successfully sanction incumbents through elections (vertical accountability) or through institutional checks and balances (horizontal accountability), civil society must frequently pressure elites (diagonal accountability).

However, one must remember that not all civil society mobilizations are equally effective in triggering accountability mechanisms. In fact, when opposition groups are perceived as illegitimate (oftentimes, when they are close to the government), incumbents may succeed in marginalizing and criminalizing anti-government protests or developing other mechanisms of resource mobilization such as coopting groups or leaders (Walker 2002; Wrede 2006). Conversely, when opposition groups emerge from civil society, a different process may occur as these groups could be perceived as highly legitimate, which renders ignoring or coopting them more difficult.

Whether or not an accumulation of grievances and related mobilizations trigger accountability is partially attributable to the strategy of opposition groups and their composition. Previous studies, such as Pinckney et al. (2022), suggest that quotidian civil society organizations (QCSOs) could drive successful democratic transitions. QCSOs are more likely to have stable preferences for democracy and durable mobilization structures; thus, high levels of accountability for new elites could be achieved.

An exemplification of the abovementioned concept could be, again, the indigenous movement in Ecuador, which was a strong movement that displayed a capacity for overthrowing the government. In this case, the pressure exerted by the movement in the context of the October 2019 protests in Ecuador led to a televised dialogue between the government and civil society in which several demands were addressed. In addition, Ecuadorian society, such as students and transport-sector syndicates, widely supported the movement. This scenario indicated that important changes can be achieved especially through diagonal accountability. Hence, exploring the role of civil society within broad accountability mechanisms is crucial for better understanding its impact.

1.3. Who Holds Governments Accountable?

In a broad sense, citizens hold the government accountable, and this accountability can be exercised vertically, horizontally, or diagonally, as previously mentioned. O’Donnell (2001) argues that “various social agents and demands” exercise a type of vertical accountability that links the state to society. Conversely, Schmitter (2004) conceives of them originally in the realm of horizontal accountability but later proposes a third type of accountability mechanism (oblique). To reflect that non-state actors play an important intermediary role and support voters (vertical) and legislators (horizontal) through the provision of information (media) and sanctions (civil society protest), Lührmann et al. (2020) label such forces ‘diagonal’ accountability. This distinction is helpful, because the actions of civil society are highly relevant for holding governments accountable.

Building on this framework, the current study argues that diagonal accountability, which refers to the ability of civil society actors and the media to constrain governments, is key to preventing the occurrence of corruption, because they can use a broad range of actions to challenge the existence of corrupt practices. For example, independent journalists may uncover corrupt practices through a state bureaucratic apparatus, and civil society actors may mobilize against such action through mass protests or other forms of citizen engagement.

After briefly revising the mechanisms that citizens could utilize to hold governments accountable, revising how frequently these groups are observed in practice is important. Incumbents who intend to pursue undemocratic and corrupt practices in the government are likely to face opposition from various agents. However, these practices are not necessarily motivated only by a commitment to democracy and its institutions. In fact, certain people who hold positions from which they can constrain the incumbent may also be concerned with their personal or their group’s interests. For example, judges may consider the rule of law important but put their personal careers first. Similarly, elected politicians may not only seek to defend their ideological principles or deploy the mandated policies of voters. They may also want to ensure reelection and the continuity of their political careers. In turn, voters may prioritize partisan loyalties and economic outcomes and, hence, tolerate power abuses or corrupt practices.

This multiplicity of interests and goals in practice complicates the realization of accountability pressures because it provides the incumbent with the possibility of dividing and coopting potential opponents. Is this cooptation (and, in turn, less realization of accountability) contingent on the type of opposition that demands it? To answer this question in an analysis, including the composition of opposition groups within a country, could shed light on the role of opposition groups in ensuring accountability by considering the sector from which they emerge. Specifically, the case may be that opposition groups close to the establishment could hinder accountability, whereas those more aligned to civil society could foster accountability and, in turn, eliminate corruption and improve governance in general.

In places such as Latin America, repression has played an important and pernicious role, which contributed to the suppression of opposition groups (from civil society and military groups), especially those coming from civil society. In this sense, by combining cooptation and repression, politicians may successfully evade and undermine accountability, such as through oversight and sanctions, and, consequentially, diminish democratic standards. Similarly, members of a ruling party or actors from economic and political elites may be less likely to oppose the incumbent, although he or she has incurred corrupt practices, when they expect him or her to stay in office.

Apart from boosting public support for the incumbent, a positive economic performance may also render coopting elites, who are likely to disproportionately benefit from economic expansion, easy for the government.

1.4. Accountability and Corruption

The problem of corruption in the public sphere is nearly a direct consequence of the nature of government interventions. Transactions within the government always imply a certain asymmetry of information between citizens and politicians and, at the same time, governments intervene precisely in situations where market failures exist, such that private provision is not regarded as a viable alternative (Banerjee 1997). In this context, corruption emerges spontaneously as a consequence of the existence of rents and monitored failures.

The literature on political science and economics extensively discusses the role of political accountability in generating good governance practices and, particularly, in reducing corruption; see, for example, Fackler and Lin (1995), Linz and Stepan (1996), and Laffont and Meleu (2001). The central argument is that accountability enables the punishment of politicians that adopt “bad policies,” which, thus, aligns their preferences with those of their citizens—that is, they play the role of watchmen. In turn, the specific features of the political system determine the degree of accountability in the system. In other words, one can expect to observe less corruption under a democracy, because politicians respond to the demands of their constituents. Despite the wide exploration of this relationship, studies that explore the type of participation (group) that renders accountability possible or facilitates it are lacking. For instance, if a country lacks an engaged civil society, then making the government accountable would be more difficult compared with a country whose citizens are highly involved in its affairs. In summary, assuming that less corruption exists in the presence of high levels of accountability is possible.

In this sense, although one can intuitively predict that the use of accountability mechanisms by civil society exerts an impact on (decreases) the level of corruption in any given government, we lack further information regarding the composition of such a civil society (in terms of whether or not it is close to the government). A possibility exists that the characteristics of opposition groups should be considered when observing the interaction between accountability and corruption. Simply put, if an opposition group is close to the government, then observing high levels of accountability and low levels of corruption would be more difficult. This study precisely addresses this relationship.

2. Dataset

The study conducts an empirical analysis on data extracted from V-Dem. We consider the questions concerning the measure of accountability, public sector corruption, equal distribution of resources, level of mass mobilization, and type of opposition group as follows:

(a)

Accountability index (“v2x_accountability_osp”). This variable responds to the question: To what extent is the ideal of government accountability achieved? This variable scales from low to high in the interval (0,1).

(b)

Regime most important opposition group (“v2regimpoppgroup”). This variable responds to the question: Which (one) group constitutes the greatest threat to the current regime? It is a categorical variable of 14 possible outcomes.

(c)

Public sector corruption index (“v2x_pubcorr”). This index responds to the question: To what extent do public sector employees grant favors in exchange for bribes, kickbacks, or other material inducements, and how often do they steal, embezzle, or misappropriate public funds or other state resources for personal or family use? It is a continuous variable in the interval (0,1).

(d)

Equal distribution of resources index (“v2xeg_eqdr”). This index measures the distribution of tangible and intangible resources in society. It responds to the question: How equal is the distribution of resources?

(e)

Mass mobilization (“v2cagenmob”). This index responds to the question: In this year, how frequent and large have events of mass mobilization been?

(f)

CSO repression (“v2csreprss”). This index responds to the question: Does the government attempt to repress civil society organizations (CSOs).

The study opted to employ variables c–f, because they will enable us to perform an initial characterization of countries and then identify a set of latent states that characterize different features of the data. Given the nature of the research question, we characterize the initial states according to the levels of repression (to capture whether organizations and civil society are subject to repression mechanisms), mass mobilization (to capture whether citizens realize diagonal accountability), and indicators such as public corruption and distribution of resources (to capture whether corruption and inequality permeate accountability).

The data correspond to a subset of $n=19$ individuals followed for $T=40$ years from 1980 to 2020. Given our interest in the dynamics experienced in the Latin American region, we consider data for the following countries: Mexico, Colombia, Brazil, El Salvador, Bolivia, Haiti, Honduras, Peru, Argentina, Venezuela, Chile, Costa Rica, Ecuador, Guatemala, Panama, Uruguay, Dominican Republic, Cuba, and Paraguay. The time frame considered by the study generally captures the return to democracy experienced in Latin America after 1978.

3. Econometric Methodology

The study implements an application of the latent Markov class model introduced by Wiggins (1973).

Let ${\mathit{Y}}^t=\left(Y_1^{\left(t\right)},\dots ,Y_r^{\left(t\right)}\right)$ be the vector of response variables $Y_j$ for $j=1,2,\dots ,r$ observed at time $t=1980,\dots ,2020$ for each country. The number of individuals is denoted by $n=19$. The response variables are continuous and $r=5$. Let $\tilde{\mathit{Y}}$ be the stacked vector (for $t=1,\dots ,T)$of dimension $\left(40\times 5\right)$.

In the empirical analysis:

$$\tilde{\mathit{Y}}=\left[\begin{array}{c}{\left(Accountability index\right)}^t\\ {\left(Public sector corruption index\right)}^t\\ {\left(Equal distribution of resources index\right)}^t\\ {\left(Mass mobilization index\right)}^t\\ {\left(Repression index\right)}^t\end{array}\right]$$

(1)

The response variables will be used to reduce the dimensionality of the dataset. The goal is to identify a set of latent states that characterizes different features of the data.

We assume that the latent state at time $t$, $W^t,$ lies on a state space $\left\{1,\dots ,k\right\},$ where $k$ is the number of latent non-observable classes or states.

Let $\mathit{W}=\left(W^1,\dots ,W^T\right)$ be a vector that describes the latent process. This process is assumed to influence the distribution of the response variables. This vector follows a first-order Markov chain with $k$ latent states. Alternatively, we consider ${\mathit{X}}^{\left(t\right)}$ to be the vector of individual covariates. This vector is assumed to influence the latent process of the model.

Furthermore, we obtain the following in the empirical analysis:

$${\mathit{X}}^{\left(t\right)}=\left[\begin{array}{c}{\left(Type of Regime Oposition \right)}^t\end{array}\right]$$

(2)

The type of regime opposition is the only covariate that the study uses in the empirical specification. Notably, in this case, the stacked vector of covariates $\tilde{\mathit{X}}$ is of dimension $\left(1\times 40\right).$

This variable is central to the study, because we propose that the type of regime opposition influences the distribution of the latent process of the model. This variable is categorical with 14 possible outcomes, namely, the aristocracy, agrarian elites, party elites, business elites, civil servants, the military, an ethnic or racial group(s), a religious group(s), local elites, urban working classes, rural working classes, rural middle classes, and a foreign government (or colonial power).

We perform a transformation of this categorical variable by constructing a dummy variable equal to 1 for civil opposition and 0 for elite opposition groups.

We denote ${\phi }_u$ as the initial probability of belonging to a particular latent state:

$${\phi }_{w|\mathit{x}}=p\left(W^{\left(1\right)}=w|{\mathit{X}}^{\left(1\right)}=\mathit{x}\right);w=1,\dots ,k$$

(3)

The initial probabilities are the parameters of the latent process and indicate the initial probabilities of belonging to a particular class given the type of regime opposition. Alternatively, the conditional response means are represented as follows:

$${\mu }_{jy|w\mathit{x}}^t=E\left(Y_j^{\left(t\right)}=y|W^t=w,{\mathit{X}}^{\left(t\right)}=\mathit{x}\right) j=1,\dots ,19;t=1980,\dots ,2020; w=1,\dots ,k$$

(4)

and

$${\mu }_{y|w}^t={{\displaystyle \prod }}_{j=1}^{19}{\mu }_{jy|w}^t=p\left(Y_1^{\left(t\right)}=y_1,\dots ,Y_{19}^{\left(t\right)}=y_r|W^t=w\right)$$

(5)

We describe the transition probabilities (from one latent state to another) using the following equation:

$${\phi }_{w|vx}^t=p\left(W^t=w|W^{t-1}=v,{\mathit{X}}^{\left(t\right)}=\mathit{x}\right) t=1980,\dots ,2020; w,v=1,\dots ,k$$

(6)

The probability distributions of the latent states and response variables are characterized by the following equations:

$$p\left(\mathit{W}=\mathit{w}\mathbf{|}\tilde{\mathit{X}}=\tilde{\mathit{x}}\right)={\phi }_{w^1|{\mathit{x}}^{\mathbf{1}}}{{\displaystyle \prod }}_{t=1980}^{2020}{\phi }_{w|vx}^t={\phi }_{w^1|{\mathit{x}}^{\mathbf{1}}}\left({\phi }_{w_{1981}|w_{1980},x_{1981}}^{\left(1981\right)}\dots {\phi }_{w_{2020}|w_{2019},x_{2020}}^{\left(2020\right)}\right)$$

(7)

$$p\left(\mathit{Y}=\mathit{y}\mathbf{|}\mathit{W}=\mathit{w},\tilde{\mathit{X}}=\tilde{\mathit{x}}\right)={{\displaystyle \prod }}_{t=1980}^{2020}{\mu }_{\mathit{y}\mathbf{|}\mathit{w}\mathbf{,}\mathit{x}}^{\left(t\right)}={\mu }_{\mathit{y}\mathbf{|}\mathit{w}\mathbf{,}\mathit{x}}^{\left(1980\right)}·{\mu }_{\mathit{y}\mathbf{|}\mathit{w}\mathbf{,}\mathit{x}}^{\left(1981\right)}\dots {\mu }_{\mathit{y}\mathbf{|}\mathit{w}\mathbf{,}\mathit{x}}^{\left(2020\right)}$$

(8)

The manifest distribution of $\mathit{Y}$ is given by the following equation:

$$P\left(\tilde{\mathit{y}}\mathbf{|}\tilde{\mathit{x}}\right)=P\left(\tilde{\mathit{Y}}=\tilde{\mathit{y}}\mathbf{|}\tilde{\mathit{X}}=\tilde{\mathit{x}}\right)$$

$$={{\displaystyle \sum }}_w{\phi }_{w^{1980}|{\mathit{x}}^{\mathbf{1980}}}·{\mu }_{\mathit{y}|w,\mathit{x}}^{\left(1980\right)}\times {\phi }_{w_{1981}|w_{1981}}^{\left(1981\right)}·{\mu }_{\mathit{y}\mathrm{|}\mathit{w}\mathbf{,}\mathit{x}}^{\left(1981\right)}\times \dots \times {\phi }_{w_{2020}|w_{2019},x_{2020}}^{\left(2020\right)}{\mu }_{\mathit{y}|w,\mathit{x}}^{\left(2020\right)}$$

(9)

$$={{\displaystyle \sum }}_w·{{\displaystyle \sum }}_{w_{1981}}·\dots ·{{\displaystyle \sum }}_{w_{2020}}{\phi }_w{{\displaystyle \prod }}_{t=1981}^{2020}{\phi }_{w|v}^{\left(t\right)}{{\displaystyle \prod }}_{t=1981}^{\left(t\right)}{\mu }_{\mathit{y}|w}^{\left(t\right)}$$

(10)

The estimation of the model relies on the maximization of the following log-likelihood function:

$$\ell \left(\mathsf{\theta }\right)={{\displaystyle \sum }}_{i=1}^{19}\log P\left(\widetilde{{\mathit{y}}_{\mathit{i}}}|\widetilde{{\mathit{x}}_{\mathit{i}}}\right)$$

(11)

The latter formula is maximized using the expected maximization algorithm.

In this model, we assume that the response variables measure the unobserved individual characteristics of interest. We are interested in the effect of certain covariates on the latent distribution.

If we want to allow that a set of covariates influences the distribution of the initial probabilities, we adopt a multinomial logit parametrization as follows:

$$\log \frac{P\left(W^1=w|{\mathit{X}}^{\mathbf{1}}=\mathit{x}\right)}{P\left(W^1=1|{\mathit{X}}^1=\mathit{x}\right)}=\log \frac{{\phi }_{w|\mathit{x}}}{{\phi }_{1|\mathit{x}}}={\beta }_{0w}+{\mathit{x}}^{\mathbf{\prime }}{\beta }_{1w},w=2,\dots ,k$$

(12)

$$\log \frac{P\left(W^t=w|W^{t-1}=\overline{w},{\mathit{X}}^t=\mathit{x}\right)}{P\left(W^t=\overline{w}|W^{t-1}=\overline{w},{\mathit{X}}^t=\mathit{x}\right)}=\log \frac{{\phi }_{w|\overline{w}x}^t}{{\phi }_{\overline{w}|\overline{w}x}^t}={\gamma }_{0\overline{w}w}+{\mathit{x}}^{\mathbf{\prime }}{\mathsf{\gamma }}_{\mathbf{1}\overline{\mathit{w}}\mathit{w}}$$

(13)

For $t=2,\dots ,T;$ $\overline{w},w=1,\dots ,k$ with $\overline{w}\ne w$.

The objective is to estimate ${\mathsf{\beta }}_u={\left({\beta }_{0,w},{\beta }^T{}_{1,w}\right)}^T$ and ${\gamma }_{\overline{w}w}={\left({\gamma }_{0\overline{w}w},{\gamma }^T{}_{1\overline{w}w}\right)}^T.$ For this reason, we use the statistical R package LMest (Bartolucci and Pandolfi 2017); available at https://cran.r-project.org/package=LMest (accessed on 10 July 2022), which provides a set of functions that can be used to estimate the LM models for longitudinal continuous data. This package is based on (Bartolucci et al. 2017).

Based on Equations (12) and (13), the objective of the empirical methodology is to estimate the impact of the type of regime opposition of a country on the probability of belonging to a particular class relative to the probability of belonging to a reference group (in this case, the first one). Alternatively, the study intends to determine the impact of the covariate on the transition probability of belonging to a particular class given that the individual belonged to another one in the last period.

4. Results

For the data described in the previous section, we estimate a multivariate LM model with covariates in the latent process.

Table 1. Autocorrelation matrix of response variables.

	Accountability	Public Corruption	Distribution Equality	Mass Mobilization	CSO’s Repression
Accountability	1	−0.6870	0.5233	0.0079	0.7909
Public Corruption	−0.6870	1	−0.8742	−0.1642	−0.5739
Distribution Equality	0.5233	−0.8742	1	0.1072	0.4653
Mass Mobilization	0.0079	−0.1642	0.1072	1	−0.1589
CSO’ Repression	0.7909	−0.5739	0.4653	−0.1589	1

describes the autocorrelation matrix of the response variables.

We observe high autocorrelations among the response variables. We endeavor to classify the individuals of the dataset in a small number of groups according to the abovementioned response variables. The first step in the empirical methodology is selecting the number of latent states. We perform a test based on the AIC and BIC information criteria with consideration of a maximum tolerance of four latent states.

Based on the abovementioned test, the optimal number of latent states is four (Figure 1). This classification enables a relatively clear characterization of the different latent states in terms of the levels of the response variables.

Based on the conditional response means, we characterize such states in terms of the levels of accountability, public corruption, mass mobilization, equality in the distribution of resources in society and the repression of CSOs.

Table 2. Conditional response means (${\mu }_{\mathit{y}|w}^t)$.

	State One	State Two	State Three	State Four
Accountability	0.1652 (0.0310)	0.7754 (0.0084)	0.8058 (0.0075)	0.2318 (0.0228)
Public Corruption	0.4758 (0.0316)	0.4069 (0.0100)	0.5862 (0.0105)	0.8088 (0.0284)
Distribution Equality	0.3030 (0.0383)	0.4167 (0.0141)	0.4370 (0.0115)	0.7657 (0.0338)
Mass Mobilization	−0.0236 (0.2132)	0.3214 (0.0675)	0.4223 (0.0501)	1.8063 (0.1393)
Repression of Civil Society Organizations	−2.0790 (0.1866)	1.1957 (0.0573)	1.2264 (0.0494)	−0.8827 (0.1527)

illustrates the estimated conditional response means ${\mu }_{\mathit{y}|w}^t$ for the model.

The results in Table 2 define the estimated conditional means of levels of the five response variables given that a subject belongs to a particular latent state. Such estimations enable the characterization of the latent states: countries in the first state have the highest probability of reporting high levels of public corruption relative to levels of accountability. In this class of countries, we also observe high levels of repression of civil organizations. State four presents high expected values of public corruption and low expected values of accountability. In addition, high levels of social mobilization and repression characterize this class.

States two and three contain countries that display a mixture of levels of the five response variables. The two latent classes exhibit higher levels of accountability relative to the levels of corruption. Furthermore, the expected values of repression are low in both states. The major difference between states two and three is that state 3, high levels of accountability coexist with high expected values of public corruption.

Figure 2 presents the average marginal distributions of the four latent states. We find that a country from the dataset belonging to class three followed by state two as of 2020 is more probable. The probability of belonging to states one and four decreases over time. At the beginning of the survey, the probability of belonging to state one was the highest among all states, followed by that of state two.

Table 3. Parameters influencing the logit for the initial probabilities.

	State Two	State Three	State Four
Intercept	−0.8986 *** (0.2198)	−3.6944 *** (1.5937)	13.6746 *** (0.4012)
Type of opposition	0.5470 ** (0.2830)	2.0938 *** (1.0429)	−14.6399 *** (0.4012)

*** Statistical significance at $\alpha =5\%$; ** Statistical significance at $\alpha =10\%$.

presents the estimated regression coefficients that influence the distribution of the initial probabilities. The estimated negative intercepts in states two and three indicate a general tendency to report low levels of accountability and high levels of public corruption at the beginning of the survey.

From the estimation, we observe that the covariate exerts a positive impact on the probability of belonging to latent class two relative to that of belonging to class one. In other words, if the most important opposition regime group comes from civil society and not from elites, the probability of belonging to class two is higher at the beginning of the survey. Latent class two is characterized by higher probabilities of reporting better accountability outcomes relative to the levels of corruption. We find that the impact is also positive when examining latent class three.

On the contrary, the impact of the covariate on the probability of belonging to class four is negative, which indicates the negative relationship that emerges between regime opposition groups coming from elite groups and levels of accountability and corruption. At the beginning of the survey, we find that the existence of civil regime opposition groups exerts a negative impact on the probability of belonging to class four.

The major interest of this study is the analysis of the transition probabilities of the model.

Table 4. Parameters influencing the logit for the transition probabilities.

LOGIT 1	Transition 1–2	Transition 1–3	Transition 1–4
Intercept	−2.2395 *** (0.1541)	−3.1479 *** (0.0147)	−15.2764 *** ($\approx 0)$
Regime opposition	0.6943 *** (0.1997)	−1.4282 *** (0.0151)	−4.1538 *** ($\approx 0)$
LOGIT 2	Transition 2–1	Transition 2–3	Transition 2–4
Intercept	−26.9348 *** ($\approx 0)$	−4.1842 *** (0.1265)	−8.6351 *** ($\approx 0)$
Regime opposition	6.2682 *** ($\approx 0)$	0.6160 *** (0.1734)	−8.8160 *** ($\approx 0)$
LOGIT 3	Transition 3–1	Transition 3–2	Transition 3–4
Intercept	−22.2451 *** ($\approx 0)$	−0.4674 *** (0.4759$)$	7.9271 *** ($\approx 0.4884)$
Regime opposition	−1.4486 *** ($\approx 0)$	−4.4872 *** $\left(0.4798\right)$	−13.3283 *** ($\approx 0.4884)$
LOGIT 4	Transition 4–1	Transition 4–2	Transition 4–3
Intercept	−20.3134 *** ($\approx 0)$	1.1493 *** ($\approx 0)$	10.6325 *** (0.3851)
Regime opposition	−1.4404 *** ($\approx 0)$	−12.6972 *** ($\approx 0)$	−13.6928 *** (0.3851)

*** Statistical significance at $\alpha =5\%$.

provides the estimated parameters that influence the distribution of the transition probabilities. Logit $i=1,\dots ,4$ presents the impact of the covariate in the model on the transition probabilities from one latent state to another.

Figure 2 demonstrates that the proportion of countries in latent class one decreased over time. On the contrary, the number of countries characterized by latent states two and three increased. Transition probabilities are a measure of the effect of the presence of civil regime opposition on the levels of accountability and corruption of a country over time.

The study finds that the impact of the type of regime opposition on the transition from state one to the second one is positive, which indicates that oppositions from civil society increase the probability of moving from environments characterized by low levels of accountability and high levels of corruption to states with higher levels of such response variables. The study also observes a significant negative impact on the transition from states one to four, which indicates that civil opposition groups avoid moving to worse states in terms of accountability and corruption.

The impact of the covariate on the probability of transition from states two to four is significant and negative, as is the coefficient related to the transition from states three to four. In other words, regime opposition groups coming from elites are related to states with low levels of accountability and high levels of corruption.

Finally, the results illustrate that the impact of the covariate on the probability of transit from state four to states three and two is negative. States four and one are characterized by high levels of repression, which prevents the formation of regime opposition groups coming from the civil sphere. This scenario can explain the negative impact of the covariate on the probability of moving from state four to better states (two or three).

5. Conclusions

This study intended to elucidate the relationship between different states of corruption and accountability using the characteristics of opposition groups within these states for a subset of countries in Latin America. Furthermore, it explored the possibilities for transition from one state to another. Toward this end, we first classified the countries according to their levels of corruption, accountability, mass mobilization, inequality and repression of civil organizations.

We enabled this classification to be influenced by the inclusion of a covariate, that is, the most important type of regime opposition in the country. This produced a latent effect, which influenced the initial probability of a country belonging to a specific “class or state” and to the transition probabilities between states.

In the initial classification, we identified four classes of countries, in which the first and fourth classes depicted countries with low levels of accountability relative to the levels of public corruption. In addition, these states are associated with high levels of repression and mass mobilization. On the contrary, states two and three are characterized by countries with high levels of accountability relative to the levels of corruption. Furthermore, they depict low levels of repression of civil organizations.

Table 5. Classification of countries based on accountability, corruption, repression and mass mobilization.

Class	Characteristics	Category
1	Low levels of accountability with high levels of corruption Very high levels of repression and low levels of mobilization	Bad state
2	High levels of accountability, moderate to high levels of corruption, moderate mobilization, low levels of repression	Good state
3
4	Low levels of accountability and very high levels of corruption High levels of repression and high levels of mobilization	Bad state

depicts the resultant classification.

Given this classification, we estimated the initial and transition probabilities of belonging to a particular state relative to that of belonging to a bad state. After applying latent Markov class analysis to 19 countries in Latin America from 1980 to 2020, the results demonstrate that the type of opposition exerts an effect on the levels of accountability and corruption at the beginning of the survey. Concretely, the initial probability of belonging to the group of countries with high accountability and low corruption increases when the type of opposition is sustained in groups closely related to civil society.

Regarding the probability of migrating from one state to another, two results are relevant. First, the presence of civil opposition groups prevents a country from migrating from a good state to a bad one. Second, when a country belongs to a bad state, which is characterized by high levels of repression, the probability of migrating to a good state is negatively influenced by the presence of civil opposition groups. This finding indicates that the effects of corruption, low accountability and repression are very pervasive and difficult to escape (an environment that can be checked by the persistence behavior of the average transition probabilities).

These results are very relevant to a region such as Latin America, in which the capacity of the state remains underdeveloped and issues such as public corruption and lack of accountability deeply influence the development and consolidation of democratic systems. On a positive note, however, if the opposition in a country stems from civil society, the probability of building highly accountable states is higher, which depicts the relevance of civil society when fostering accountability. This finding, indeed, is the most relevant one for this study and is aligned with the existent literature, which correlates mass mobilization and social change (Kadivar 2018; Hellmeier and Bernhard 2022).

In addition, the results reveal new venues for future research. An interesting avenue consists of exploring the reasons behind the existence of states with high levels of accountability and high levels of corruption, and second, exploring the factors that contributed to the migration of certain countries from bad to good latent states, including the role of repression mechanisms employed by the government.