Espacios. Vol. 37 (Nº 13) Año 2016. Pág. 23

Party ideology and elections: an analysis of political-economic cycles in Brazilian states

Ideología de partidos y elecciones: un análisis de los ciclos político-económicos en los estados brasileños

Jevuks Matheus de ARAÚJO 1; Paulo Amilton M. LEITE Filho 2; Felipe Alves REISJ 3

Recibido: 27/01/16 • Aprobado: 04/03/2016


1. Introduction

2. Theoretical elements

3. Methodology - Dynamic Panel Data Model

4. Results of Estimations

5. Final Considerations



The objective of this study is to analyze the political-economic cycles in the Brazilian states from 1995 to 2008. The theory was originally developed in the work of Downs (1957), Nordhaus (1975) and Hibbs (1977) who established the assumptions of the model of political-economic cycles. In the 80s of last century, the work of Alesina (1987) and Roggof and Sibert (1988) incorporated the concepts of rational expectations and information asymmetry to the original models. The econometric model was used in dynamic panel data GMM system estimator. The estimation results show an increase of state spending in election years and show that this spending is adjusted in post-election years, showing an opportunistic behavior by governments. The results also showed that the spending of governments, that have different party ideologies, have different behaviors; left-wing parties promote further expansion of public spending.
Keywords: political-economic cycles, public expenditure; dynamic panel data; democracy.


El objetivo de este estudio es analizar los ciclos de política económica en los Estados brasileños de 1995 a 2008. La teoría fue desarrollada originalmente en el trabajo de Downs (1957), Nordhaus (1975) y Hibbs (1977) quienes establecieron los supuestos del modelo de los ciclos políticos económicos. En los años 80 del siglo pasado, el trabajo de Alesina (1987) y Roggof y Sibert (1988) incorpora los conceptos de expectativas racionales y la asimetría de información a los modelos originales. Se utilizó el modelo econométrico en estimador de sistema GMM de datos de panel dinámico. Los resultados de la estimación muestran un aumento del gasto en años de elección estatal y demostración que este gasto es ajustado en los años posteriores a las elecciones, mostrando un comportamiento oportunista por los gobiernos. Los resultados también mostraron que el gasto de los gobiernos, que tienen ideologías diferentes de partido, tienen diferentes comportamientos y son partidos de izquierda, promueven una mayor expansión del gasto público.
Palabras clave: ciclos políticos económicos, gastos públicos; datos de panel dinámico; democracia.

1. Introduction

The analysis of the relationship between economy and politics over time resulted in a greater understanding of how the political design affects the behavior of economic variables. Although many of the results do not converge, it is an act foreign to economics to deny the similarities between economics and politics.

The form of how a government is elected and takes its decisions has become a recurring theme in the studies of political economy. In Downs's work (1957), we found a formalization of how the interests of political agents (parties and candidates) motivate their choices. In the same vein, Nordhaus (1975) shows that the existence of periodic elections determines the pattern of political decisions. The results of the works of  Downs (1957) and Nordhaus (1975) show the idea of ​​an opportunistic political behavior, which refers only to the maintenance of power. According to the theorists of the political opportunism, the choices of the administrators will converge. Regardless of partisan aspect, the government's actions will aim at maximizing votes.

This purely opportunistic behavior, in which the policy choices have no ideological bias, has been questioned and other studies have shown that the ideological component has a strong influence on policy decisions.

Hibbs (1977) argues that governments of different ideological orientation are committed to different interest groups, which leads them to choose different patterns of political decisions. He understands that governments direct their actions to meet the demands of the social class that gives them greater political support.

The goal of governments, from a party standpoint, would not merely be winning elections, but also meeting the needs of a portion of society that is somehow attached to the ideological essence of the ruling party.

Although the core components are divergent in both theses, it is elucidated the idea of ​​the existence of an economic cycle driven solely by political interests to win elections (first thesis), or the partisan interests of favoring a specific portion of the population (second thesis). In both theories, it is clear the strong entanglement between economics and politics.

In later years, the work of Alesina (1987) and Roggof and Sibert (1988) added respectively to the models of Hibbs and Nordhaus, the perception of rational expectations and the problem of asymmetric information.

In Brazil, the discussions on the existence of political-economic cycles have grown in recent years. The return to democracy, the arrangement of the party system and the reliability of the data enabled the advancement of research. Studies have been conducted in all administrative levels (Federal, State and Municipal) and, in most of them, empirical proof for the existence of these cycles has been found.

The focus of this work is the public spending carried out by state governments, which represent on average 30% of government expenditure in Brazil. Among the contributions of this work, we highlight the restricted analysis of the implications of political variables (elections and political parties) on the behavior of state governments' spending.

The research is based on the presuppositions of opportunistic and partisan models that respectively point to the existence of growth in public spending in times of election and different spending patterns in governments of different ideological wings. The working hypothesis is that the electoral calendar and party ideology influence on the performance of public spending.

Within the theoretical framework of political-economic cycles, this work aims to analyze the political-economic cycles in the Brazilian states from 1995 to 2008.

The paper is organized as follows: in the next section we discuss the theoretical approaches to the political-economic cycles, as well as some empirical studies conducted in Brazil and, then, it is presented the methodology and data used in the research. At the end, we discuss the results with the estimates and bring the work's conclusions.

2. Theoretical elements

Studies of political-economic cycles are based on the models of opportunist politics and partisanship. They refer, respectively, to the influence of the electoral calendar and the party ideology on the policy choices of government. The existence of elections and political parties is a characteristic of a democratic political system.

The starting point of analysis is to clearly display the theoretical arguments that make it possible to understand what is being sought, within the limits of this work, on understanding the concept of democracy. Schumpeter (1975, p.321) states that "the democratic method is an institutional framework for policy making, in which the individual acquires the power to decide upon a competitive struggle for the votes of the voter"

The theory of Schumpeter democracy refers to a minimal definition of democracy, although even this minimal definition is subject to certain conditions. Schumpeter puts the existence of such conditions: a guide to leadership as collective action, collective wills emerged by the leaders, competition and free vote, the existence of individual freedom in which "anyone can run for political leadership" and the primary function of the electorate, which is to elect the government and its ability to remove it. These peculiar conditions of democratic government, explained by the author, are distinguished or opposed to the core of other possible forms of government.

By analyzing the democratic theory of Schumpeter and others authors who are described as minimalist, O'Donnell (1999) points out that, although their definitions revolve around rules and election procedures, these settings are realistic, that is, it is possible to verify them empirically, he adds two other important elements assumed by these theories: the existence of elections that do not violate the constitution and also the non violation of the rights of freedom.

Thus, accepting the conditions aforementioned, we can accept or understand democracy as a set of rules and procedures that enable the individual to make their choices, without running the risk of being complacent with the definition of democracy or unrealistic as to the empiricism of this understanding. Bobbio (1997, p. 18) characterizes democracy as "a set of rules (primary or key) that establishes who is authorized to make collective decisions and with what procedures".

The organization of the democratic political system allows governments to behave strategically, making decisions that favor their personal interests or the interests of a small group that supports them. In Economic Theory, these choices originate the concept of political-economic cycles.

2.1 Political-Economic Cycles

The relationship between the behavior of economic variables and political variables was denominated by Nordhaus (1975) as political business cycles. The author described the relation of the political choices with electoral period, noting an opportunistic behavior of the ruler seeking to maximize votes.

Hibbs (1977) questioned the results of the work of Nordhaus and verified, empirically, that, for some countries, the choices of the government do not have a purely opportunistic character, but a strong partisan bias, i. e., the choices are directed to the group that supports government policy.

Nordhaus and Hibbs's original models were expanded in the eighties, incorporating new elements of the economic theory such as rational expectations and asymmetric information. In the following section, we describe each of the original models and their extensions.

The basic assumption of Downs's (1957) model is the existence of a democratic system in which governing parties, opposition parties and voters coexist. The model seeks to structure the behavior of these agents based on the tenets of traditional economic theory, applying these same rules of rational behavior imposed on firms and consumers.

According to Downs, the objective of political parties is to win the election; therefore, they seek to maximize votes. Thus, policy-making functions as a means to increase the chances of victory in the electoral process. According to Downs (1957, p.52) "the party that runs the government handles their policies and actions the way they think it will bring more votes without violating constitutional principles."

The work of Nordhaus (1975) analyzes the macroeconomic variables (unemployment and inflation) in nine countries, based on the trade-off of the Phillips curve and demonstrates a political behavior towards controlling and managing these variables in electoral periods. According to the author, there is a reduction in unemployment in periods before elections and also high inflation rates.

In this perspective, the proximity of elections compels the government to adopt policies aimed at meeting the majority of the electorate preferences, extending, right before an election period, policies that encourage economic growth and reduce unemployment levels. The policies used are commonly guided by the expansion of public spending and monetary issuing (Borsani, 2003).

The logic of the model is plausible, especially if we consider the process of structuring the votes submitted by Downs (1957) in which voters vote by comparing the utility income flow from government activity.

The voter chooses, based on past information, the government that will bring greater perceptible benefits (there are a number of benefits generated by government action which are not easily visible to the voter, for example, checking the quality of water that is supplied to homes.) This behavior coming from voters incentivizes the government to increase actions that will increase such perceptible benefits during election periods.

Rogoff and Sibert (1988) reformulated the original opportunistic model, expanding it to incorporate and replace some assumptions. They replace the adaptive expectations with rational expectations, incorporating to the model asymmetric information and relevant importance is given to the level of government competence of providing public services.

According to the authors, the most competent government is the one that requires less revenue to provide a given amount of services and voters prefer more competent governments. However, information is not perfect, the government has a bigger set of current information in comparison to the one that voters have. This information asymmetry allows electioneering maneuvers as it signals to the government the possibility to omit information that might jeopardize the current level of competence.

Voters evaluate a government's competence based on information acquired in a certain period of time, however, this information is outdated, and so the voter's assessment is influenced by information not consistent with the situation of the moment.

The government has incentives to generate, on the eve of the election period, signals to indicate their level of competence. The shock of powers, given by the government, directly influences the election results. These shocks can be represented by the implementation of policies which aim at increasing production and reducing unemployment.

Rogoff (1990) presents refinements to the model and expands the analysis of the responsibilities of the government in other variables, especially the expansion of government spending in the most tangible benefits to society. Thus, the government has the stimulus to increase spending on consumption and transfer in pre-election periods.

The expansion of public spending may represent a clash of powers clearly visible and quickly assimilated by the voters, directly influencing the elections.

In Brazil, a classic case of a clash of jurisdiction is the inauguration of works on the eve of the election. A ruler who plans a round of inaugurations next to the election period is clearly trying to signal to voters a presumable administrative competence.

Asymmetric information leads to an opportunistic behavior on the part of the government, since the government has a greater knowledge about their competence, as well as their limitations. In the pre-election, the government tends to exacerbate actions that create broad impact of their competence. Similarly, they might as well postpone to the post-election actions that create or increase the limitations of their government (Rogoff and Sibert, 1988).

The model with rational expectations assumes that voters associate the economic results with government competence. Thus, positive economic results reverberate favorably to the government at the polls. The reformulations of the opportunistic model consider that the voter chooses their candidate based on past information and on the level of benefits they expect to receive in the future. Voters use retrospective behavior as a basis for prospective choices.

The current governor keeps a vote maximizing behavior. Rogoff (1990) states that any governor, regardless of party ideology, in electoral periods, seeks to convince voters of their administrative skills increasing the chances of reelection.

The work of Rogoff and Sibert (1988) and Rogoff (1990) open space for a political-economic cycles analysis in any level of government under a wide range of fiscal variables. They state that, during elections, the leaders often engage in a series of actions that improve the visibility of their competence. To do so, they distort fiscal policies by cutting taxes and increasing the direct transfers, this way, making it more visible to the voters that they are investing more. The logic of the opportunistic model disregards party ideology as an important element in directing the actions of the government.

The approach taken by Hibbs (1977) considers the existence of political decisions driven by ideological interests and presents a causal relationship between these decisions and economic cycles. In this model, parties with different ideologies behave differently, trying to meet the interests of those groups to which they are closer ideologically.

Within an ideological spectrum formed by left and right-wing parties, political decisions tend to have a very divergent path when there are changes in the party or in the ruling coalition. The work of Hibbs (1977) analyzed the pattern of macroeconomic policies of 12 countries including the United States and England.

Following the same reasoning used by Nordhaus (1975), the Phillips Curve, Hibbs (1977) notes that, one the one hand, lower income groups have a strong aversion to unemployment and, on the other hand, higher income groups have a strong aversion to inflation. This piece of information is important because policy choices will have different impacts on the groups that make up a country's electorate.

The results showed that, in the U.S., a Democratic government combats unemployment, while a Republican government combats inflation as its main goal, and in England, Democrats and Republicans are represented respectively by the Labor and the Conservative parties.

Thus, the author concludes:

The general conclusion of the study is that the macroeconomic policies pursued by left and right-wing governments are broadly in accordance with the subjective preferences of their class defined core political constituencies. (Hibbs, 1977)

The analysis of Hibbs (1977) suggests the construction of a new theoretical model considering the role of party ideology in the choice of the ruler. The constituencies of each party would have an important role in the policy measures taken by the government. The partisan model structure presented in the next section differs from the opportunistic model for posing the ideological stance of the government as a cause of the cyclical behavior.

Based on the existence of a partisan bias in the behavior of policymakers, Alesina (1987 and 1988b) incorporates to the original model of Hibbs (1977) the existence of a voter behavior guided by rational expectations.

In this new perspective, elections are considered sources of uncertainty about the future. Voters do not know which party will win the elections, and they do not know what the policy choices will be either, since parties from distinct wings (left or right) choose different policies.

Alesina and Sachs (1988) tested the hypothesis of partisan political cycles with rational expectations for U.S. data from 1945 to 1984. Their results confirm the hypothesis of political bias in the choice of economic policies, noting that there are differences between the goals of each party. Nonetheless, the differences are more evident at the beginning of the administration. There is a period dominated by a certain political neutrality and both administrations (left or right) behave similarly.

Alesina and Roubini (1992) advance in the discussion and test the hypotheses of the existence of political economic cycles for 18 European countries, rejecting the hypothesis of electoral opportunism put forward by Nordhaus (1975). However, their study shows some consistency for the hypothesis developed by Rogoff and Sibert's (1988) expanded model and proves the existence of political bias in the choice of economic policies; such results confirm the model of Alesina (1987) and corroborate the results of Alesina and Sachs (1988).

Alesina and Roubini (1992) find that even a government that acts strictly according to its ideology prefers to remain in power than to leave it, leading to a possible opportunistic behavior during election periods. Thus, we can consider that the opportunistic and partisan models are not mutually exclusive and it is possible to empirically verify a typical electioneering behavior and an ideological bias for the same government.

The ruling party seeks to maximize its objective function in accordance with their ideological bias. The application of the model should result in a different behavior for left and right-wing governments. It is expected that right-wing governments have contractionary policies which are more concerned about controlling inflation, and left-wing government should have expansionary policies which are more concerned with reducing unemployment.

As to Brazil, the left-wing parties have an electoral link to social movements and a strong association with lower social classes. In this paper, we consider that this electoral relationship implies, according to the theory, in the case of a leftist government, in an expansion of public spending to meet the social needs of the poor.

2.2 Evidence in Brazil

A number of empirical studies were carried out to verify the existence of political-economic cycles in developed countries. Recently, some researchers have proposed to study the hypothesis of political-economic cycles in developing countries.

Empirical studies on political-economic cycles and economic policy in Brazil are recent; however, there has been a growth in the publications in this area over the last ten years. Most studies show results that demonstrate the existence of political cycles for a number of economic variables using different methodologies.

One of the pioneering works of analysis of the political-economic cycles in Brazil was Fialho's (1996), which analyzes the influence of elections on the behavior of macroeconomic indicators from 1953 to 1995. The estimation was carried out using econometric time series with a dummy variable for election years, considering all elections (local, state and federal). According to the author "the concepts of political cycles may indicate an additional explanation for fluctuations of the Brazilian economy." The main conclusion of the work is to evidence the opportunistic cycles in macroeconomic variables.

It is important to highlight the temporal extension of Fialho's (1996) work, for the text itself highlights the possible lack of reliability on the data. Another aspect is the military dictatorship period, when there are no electoral pressures.

Sakurai and Gremaud (2007) studied the behavior of public spending in São Paulo's small towns from 1989 to 2001. Using panel data, they clearly observed the influence of the electoral period on government spending. However, the results do not show the influence of political parties, highlighting the lack of ideological consistency in political parties that can differentiate the behavior of public expenditure in the towns analyzed.

Lucinda, and Arvate Avelino (2008) found, using panel data, the role of party ideology in the behavior of public spending for the Brazilian states from 1987 to 2000. They observed that the results vary depending on the classification adopted to define a party as left or right-wing, using the definitions established by Figueiredo and Limongi (1999) and Coppedge (1997). The results of their study suggest the influence of party ideology, in which left-wing governments increase the total expenditure and also increase the share of social spending in the total spending.

The party classification used by the authors differs essentially by the position of the PSDB, considered left-wing following the classification of Coppedge (1997) and as center following the classification of Figueiredo and Limongi (1999).

Lucinda and Arvate Avelino (2008) open an important discussion about how to classify a party, which is a barrier for the analysis of partisan cycles in Brazil. The classification, in most cases, is associated with experts' opinions. Thus, it has a subjective character, which may bring different results, especially if the classifications differ around a party of major political expression, such as the PSDB.

Borsani (2003) studied political-economic cycles in Latin American countries from 1979 to 1998 using a model of static panel data with fixed effect. In the analysis of macroeconomic variables, the results show little evidence of opportunistic cycles and check the influence of party ideology on variations of employment levels.

Fenólio (2007) tested, having a Taylor rule as a starting point, the existence of influence of the electoral calendar over the Selic rate. The results showed statistical insignificance for the political variables of the model, indicating the absence of opportunistic behavior in the Brazilian interest rate policy from July 1999 to December 2006.

Shikida et ali (2007) used an ARIMA model to verify the existence of electoral cycles in certain macroeconomic variables (GDP growth rate, inflation rate, unemployment rate) from 1985 to 2006. The results for all variables showed a cyclic behavior with an unconscious growth in the pre-election period and a drop in post-election period. They emphasize that the goal of reelection overlaps the collective interests, strongly corroborating the opportunistic model.

Empirical studies in Brazil have, like in the international literature, divergent results. However, these differences arise primarily from the variables studied, the period and the methods used. The existence of political-economic cycles in Brazil is not fully accepted, nor completely rejected. Studies are needed to broaden the discussion about the phenomenon of political-economic cycles in all levels of government.

In the present work, new elements are incorporated into the discussion such as: a period of time that has not been studied yet, a restriction in the analysis that checks only the influence of both party ideology and elections on the spending behavior of state governments. Moreover, dynamic panel data estimates are used, which have been little explored by the studies on political cycles carried out in Brazil.

3. Methodology - Dynamic Panel Data Model

The method of econometric panel data consists of a set of information on a number 'n' of individuals (families, companies, countries, etc.) over a period of time (Pindyck, Rubinfeld, 2004). Hsiao (2003) cited in Baltagi (2005) presents a list of benefits of using panel data, among them we highlight: the control of individual heterogeneity, a bigger amount of information on the sample, a greater variability, a reduction in the collinearity among variables, a greater degree of freedom and more efficiency.

Another important advantage of using panel data is the ability to identify and measure some affects that are not verifiable using time series or cross section data (Baltagi, 2005). Thus, the panel data model picks up the effect of variables that would normally be omitted in time series or cross section models.

The empirical strategy of estimation consists of applying the method of dynamic panel data. To get the results, we used the software STATA 12. The analysis was the following: organization of data in panel form, application of unit root test in ​​dependent variables; estimation of dynamic model with the Blundell / Bond estimator. In order to verify the robustness and consistency of the models, the autocorrelation tests (Bond testing) and the specification or validation of instruments (Sargan test) were applied, both suggested by Arellano and Bond (1991).

The implementation of the specification and autocorrelation tests allowed us to evaluate the best models. Based on the results of the tests, it was possible to compare models and find the best specification and the most appropriate number of lags. The Sargan test's goal was to verify the validity of the instruments. The acceptance of the null hypothesis implies the validity of the instruments. The autocorrelation test aimed to check the serial correlation of first and second order residues. The acceptance of the null hypothesis of the inexistence second order serial correlation is expected.

The dependent variables were estimated with logarithmic specification, so the coefficients refer to the percentage changes. The estimated econometric models can be represented by the following equations:

The opportunistic behavior will be captured by dummy variables: year of election (elei) pre-election year and post-election years (pos).  The ideological bias of the government is captured by the dummy government ideology (esq). The dummies take value 1 if the variable is affirmative and 0 otherwise.

3.1 Specification of Variables

The period under study goes from 1995 to 2008 covering three elections (1998, 2002 and 2006). It is important to point out that Brazilian elections have a fixed schedule, allowing a possible adjustment of public spending policies with the election period. Secondary sources were used to collect the data analyzed in this study. The data on public spending come from the National Treasury Secretariat publications, the information referring to electoral data were searched in the publications of the Superior Electoral Court.

This study aims at analyzing the behavior of public expenditure in Brazilian states in order to verify possible changes in these variables that might have occurred in election years, as well as confirming the distinction of behavior among governments with different ideological arrangements; thus, investigating the existence of political-economic cycles in Brazilian states.

This way, the variables considered in the model can be classified as dependent or political variables.

The dependent variables estimated refer to the expenditure in the categories costs and capital, health and education. The goal is to identify changes in the behavior of these expenditures over the period studied. The education and health variables reflect social spending, while capital characterizes government investment and cost is the administrative state cost.

The political variables are intended to capture the effects of elections and party ideology on the behavior of public spending. When it comes to election, the year before and the year after it are considered. The logarithm of population is used as a control variable.

For political variables, public spending is expected to grow in election years (elec), confirming the existence of political opportunism. It is also expected that the left-wing parties (left) increase spending, confirming the ideological influence on political decisions.

3.2 Public Expenditure and Elections

As to public spending, cost and capital aggregated expenditures and specific expenditures on education and health were considered. The graphs below demonstrate the behavior of these variables in electoral periods.

Figure 1 shows the variation of the sum of state expenditure from 1997 to 1999. In all variables, it is possible to notice an increase in election year (1998) and a significant downturn in 1999. In Figure 2, from 2001 to 2003, shows that the variable cost is the only one that shows growth in an election year, there is a reduction in the others. In figure 3, from 2005 to 2007, the variables were higher in election year and lower in the following year.

Figure 1 - Behavior of expenses in the period 1997 to 1999.

Source: Own elaboration from the database of the National Treasury (STN).

Figure 2 - Behavior of expenses in the period 2001 to 2003.

Source: Own elaboration from the database of the National Treasury (STN).         

Figure 3 - Behavior of expenses in the period 2005 to 2007.

Source: Own elaboration from the database of the National Treasury (STN).

The party ideology is polarized into left and right, but there are also the subdivisions; center, center-left and center-right. For methodological reasons, the subdivisions center-left and center-right will not be considered, belonging to the classification of pole, that is, the party classified as center-left will be now classified as left.

The lack of relationship between electoral calendar and the expansion of spending, in certain election periods, as it is the apparent case of the 2002 elections, does not hurt the fullness of the theoretical basis of opportunistic cycles. Rogoff and Sibert's(1987)  work, after expanding the model, points to a less frequent political and economic cycles than those held in Nordhaus's (1977) article.

The party system established by the Constitution of 1988 is a multiparty system, which grants the right of establishment, merger and dissolution of political parties. In Brazil, according to the Superior Electoral Court (TSE), there were in 2009 27 (twenty seven) parties, 21 (twenty one) of which had representatives in the House of Representatives and only 11 (eleven) in the Senate. In 2006, governors from only 8 (eight) different parties were elected.

The large number of parties and alliances hinder the ideological classification. However, we will concentrate on seeking a classification for only the twelve (12) parties which have owned or currently own elective seats in the period studied.

The parties are the Brazilian Democratic Movement Party (PMDB), Brazilian Socialist Party (PSB), the Workers' Party (PT), Democratic Labor Party (PDT), the Brazilian Social Democracy Party (PSDB), the Liberal Front Party (PFL) , currently called the Democrats (DEM), Brazilian Progressive Party (PPB), Progressive Reform Party (PPR) Progressive Party (PP), Popular Socialist Party (PPS), the Brazilian Labor Party (PTB), Social Liberal Party (PSL).

Among all parties listed above, the PSDB is the one with larger differences on their classification. Coppedge (1997) and Boschi (2004) classify PSDB as being leftist, while Figueiredo and Limongi (1999) and Kinzo (2005), as center-left. However, Borsani (2003) presents a classification based on the policy measures taken by the administration of President Fernando Henrique Cardoso (PSDB), which ideologically fits the PSDB as a right-wing party.

Box1 presents, based on the authors above mentioned, the classification of the parties, as well as the states where they won the elections. We found that out of 27 (twenty seven) states, only São Paulo, Roraima and Amapá had no ideological changes in the classification of their rulers, the first two had right-wing administrations and the third, left-wing administrations.

The PSDB has four consecutive terms of office in São Paulo, each term lasting four years. Amapá has two by PSB (94 and 98) and two by PDT (02 and 06) and Roraíma was governed by PTB (94), PPB (98), PSL (02) and PSDB (06).

The Figure 4 shows the percentage distribution of the administrations of center, left and right in the elections of 94, 98, 02 and 06. A significant increase in the leftist administrations can be seen, which went from 22% in 94 and 98 to 37% in 2002 and 44% in 2006. In contrast to the expansion of the left-wing, there is a reduction of right-wing governments of 44% in 94 to 22% in 2006.

Box 1 - Classification of Parties and Election Results.



State elections













































































Source: Own elaboration from the database the Superior Electoral Court (TSE).

Figure 4 - Percentage Distribution of State Governments
by rating Ideological Elections 1994, 1998, 2002 and 2006.

Source: Own elaboration from the database the Superior Electoral Court (TSE).

It is expected that the growth in the left-wing governments is associated with the increase in public spending.  However, the Figure 5 shows the average change in spending, showing that from 1995 to 2008 there are no significant differences between left or right-wing governments. However, the center governments stand out for having the highest average growth rate in spending on capital and health, which grew respectively 54% and 63%.

Another relevant piece of information to the description of the comparative data is the composition, in relation to the total spending, of what represents spending on costing, capital, education and health (Figure 6), this way, measuring the percentage of spending in the total spending. The focus is to verify if there are differences in the composition of total spending in different administrations.

The Figure 6 shows the average representation, in percentage terms, of the spending analyzed in this study, in relation to the total spending. It is observed in each column the average percentage according to the type of expenditure and according to the classification of government. 

Figure 5 - Variation of Average Expenditures According to the
rating of the Government in the period 1995 to 2008. In%.

Source: Own elaboration from the database of the STN and TSE.


Figure 6 - Composition of Average Spending According to the rating of the Government in the period 1995 to 2008. In%.

Source: Own elaboration from the database of the STN and TSE.

On average composition, it is possible to see that cost represents the largest share of total expenditures. Moreover, it is possible to verify that the left-wing governments have a greater spending on cost, which represents 67% of their total spending.  Right-wing governments are the ones that spend more on capital, although the differences are not that significant (16% right, 14% center and 13.5% left).

Spending on education represents approximately 16% of total spending for all types of government. Health expenditures represent about 9% of total spending for the governments of right and center and approximately 13.5% for the left-wing governments.

The average composition of spending provides a better understanding of the relationship between parties and public expenditure than the average variation of costs, since it allows to identify choices or preferences of each administration.

The initial observation allows us to assess, however inaccurately, that the behavior of the data has a clear association with the electoral calendar and a less evident relationship with the ideology of the ruling party. A more precise investigation of the data is developed in the next section, in which the analysis of the results of econometric estimation is given.

4. Results of estimations

The results for the estimation of the equations 1, 2, 3 and 4 are reported in Table 4. One can observe in the (2), (3), (4) and (5) columns the results of the Blundell / Bond estimator for the four variables studied. The levels of lag for the adjustment of the model are represented by the terms L1, L2 and L3 in the first column of the table. The lagged variables are significant from a statistical point of view, which indicate a correct adjustment to a dynamic behavior pattern of the estimated variables. 

The results for the best dynamic adjustment for the model are in table 4, whose choice was based on the autocorrelation and Sargan tests. The models chosen are those that, in both tests, accept the null hypothesis, there is no second order serial correlation (M2) and validation of instruments, respectively.

Column (2) presents the result for estimating the expenses on costing. The best model was adjusted with three lags where all coefficients are relevant and the signs are as expected.

The costs of funding represents on average 65% of the total expense of the states and are directly related to personnel expenses and government consumption. The model results [4] show that these costs have a small increase (5%) in pre-election years, and  strong growth in election years (86%) followed by a drastic drop in post-election years (40%).

Table 4. - Results of Estimations.

( 2 )

( 3 )

( 4 )

( 5 )
































Election year









Before election













































 Autocorrelation test







 Sargan test







Prob > chi2





Source: Prepared by the authors.
Notes: The p-values ​​are in parentheses. ***, ** And * represent significance level,
Respectively 1%, 5% and 10%. The regression presented in bold represents that best fit

The estimates show a cyclic behavior of expansion of spending during electoral period accompanied by a contraction or adjustment in the post-election period. For analysis of partisan positioning, the coefficient indicates that the left-wing parties increase the expenses on costing by 62%, which also implies a cyclical performance influenced by the ideology of the ruler.

In column (3), we have the estimates for investments. We have that the best adjusted dynamic model was, again, with three lags. The results point to a behavior similar to that of spending on cost, both with respect to the election period and the government classification. However, for capital expenditures, they reduce in post-election years by 47%, i.e., the adjustment in capital spending is more intense.

The capital expenditures consist of financial investments, debt repayment and investment. However, investment accounts for approximately 78% of this type of expense. Thus, the result indicates an increase of planning and execution of works, at election time. It is a typical behavior of the Brazilian rulers to accelerate works at election time, aiming to show society tangible results of their administrations.

Not long ago, these results, besides of being tangible, were also personified with names that led to the figure of the ruler. There are numerous housing projects, schools and many other public buildings that are named after rulers' relatives or the rulers themselves.

Column (4) gives the estimates for the spending on education. As in previous estimates, the best model is the one with three lags. The coefficient of the variable pre-election year showed statistical insignificance; however, the coefficient of the election year is significant from a statistical point of view and indicates a 62% increase in spending on education, implying the existence of opportunistic behavior.

The party behavior was also significant for spending on education, where left-wing parties increased the expenses by 54%. One intriguing result was the negative sign of the coefficient of the population logarithm, indicating that an increase in population reduces the spending on education, what seems contradictory. However, education reaches only some population groups, perhaps, for this reason; the variable has provided a negative impact on education spending.

Column (5) presents the estimating results of health. One can observe that the model with one lag showed the best adjustment. The coefficients that capture the influence of the electoral calendar and the parties have statistical significance.

Although health care costs begin to increase in pre-election periods, this growth is very small; only 1%. However, in an election year, there is an increase of 42% of health spending which, as well as the other variables, has a very strong growth peak in the election year. The coefficient of party ideology indicates a 51% increase in spending when the ruler is a leftist. 

The estimation results demonstrate the existence of partisan cycles in the behavior of public expenditures made by state governments. The left-wing parties have a greater intervention in economic outcomes through the increase of public spending. In Brazil, the expenses of costs and capital correspond in average to 17% of the state GDP. When measuring this relationship considering left-wing governments, the percentage rises to 22% of state GDP.

The increase in spending on education and health, in left-wing governments, shows the ideological bias in the choice of public policies. The expansion of this spending represents an effort to meet the interests of a specific group of voters. An allusion to the level of welfare of an individual can be made by adding up the amount of private goods and public goods consumed by them. The poorest people, obviously with lower incomes, have a strong restriction on the consumption of private goods. For this population, the consumption of public goods is of primary importance, for as the supply of these goods increases, so will the level of welfare of this group of people. The left-wing parties are politically linked to the less affluent social strata, thus increasing social spending is related to the interests of social groups that support the government policy.

The election period also causes fluctuations in the performance of public spending. In an election year, there is a sharp increase in all variables analyzed, confirming the assumption of political opportunism. The spending increase represents the creation of shocks of competence by governments.

The shocks of competence are made by exceeding the expenditures without properly expanding the revenue. Spending on capital and cost represents, on average, for the entire period, 143% of tax revenues of states. In election years, this average rises to 167% of tax revenues, which represents an even bigger bottleneck to the public accounts. The adjustment of the accounts is done through the reduction in spending in post-election years, in which the relationship of expenses with the revenue drops to 118%.The results show the coexistence of opportunistic and partisan cycles in the Brazilian states. In summary we can say that:

  1. The maintenance of power is among the objectives of state governments and government spending is increased as a way to achieve these goals;
  2. The ruler simulates to society a false level of competence through increased government spending;
  3. There is an ideological bias in the choice of government actions; the government seeks to meet the demands of the social group that supports them.
  4. Left-wing parties adopt more expansionary policies in public spending;

5. Final Considerations

The starting point of this research was to understand the role that the electoral process has on the performance of government spending. It should be noted, within the democratic system, as essential components, the existence of political parties representing the interests of different groups that make up society and the free elections that allow the community to choose their representatives.

The theory of political-economic cycle studies, within a democratic political system, how the policy choices of governments affect the behavior of economic variables. There are two strands: the first one preaches political opportunism, in which policy choices are merely electioneering, whose single aim is to stay in power; the second asserts the existence of an ideological partisan bias, where political decisions are taken in order to meet the desires of the ruler's supporting group.

The assumptions of the theory of political-economic cycles endorsed the objective of this work which was to analyze the influence of the electoral calendar and party ideology on the behavior of public expenditures made by state governments in Brazil.

Two features of the Brazilian democratic environment should be highlighted: the Brazilian elections are pre-fixed, prior knowledge allows leaders to plan their spending, so they are extended in the period around the election date, and the multiparty political system that somehow complicates the ideological classification of governments within the continuum left / right.

The difficulties of classifying parties were overcome through the use of the categorizations established by experts, which have been accepted and used in other studies with analysis similar to those applied in this research.

The public expenditure analysis was performed on costing, capital, health and education. The objective was to capture the influence of the electoral calendar and party ideology on consumption (cost), investment (capital) and in social areas (education and health).

In observing the behavior of public spending, the existence of electoral cycles was evident in the elections of 1998 and 2006. It was not possible, based on data examination, to find a cyclic behavior in the series during the 2002 elections.

Regarding partisan influence, the series are apparently indifferent, i.e., the mere consideration of data showed no significant differences between the spending of a left or a right-wing government. However, when the examination was performed based on the composition of spending, it was found that governments of different ideological wings choose different arrangements for public spending.

The econometric estimation showed that the series have a correlated behavior, i.e., the present results depend on past results. The results show, for all series, a cyclic behavior determined by the election period, in which the expenses expand in the election year and contract the in year following the election. The results also show that policy choices are driven by the ideological bias of the ruling party and that left-wing parties have a bigger expansion of spending.

The theory of political-economic cycles proved valid when contrasted with data from the governments of the Brazilian states. The assumptions of political opportunism and partisanship were corroborated by the results of the analysis. Thus, it confirms the initial hypothesis that election and the ideological bias of the government influence on the behavior of public spending in Brazil.

Although the research objectives have been achieved, the approach suggests new developments. For example, in this research, it was considered that the government ideology is represented by the ruling party. However, there is a coalition that elected them and possibly an even broader coalition to maintain governability. Thus we have the following problem: what is the impact of political alliances in the behavior of public spending?

Another extension of the approach would be the evaluation of efficiency: left and right-wing governments have different patterns in resource allocation, but is this allocation efficient? What outcomes are more desirable from the standpoint of economic and social development, the left or right-wing governments'?

There are many questions to be answered in the relationship between economics and politics. However, we limited ourselves to achieve our goals. Thus, we restricted ourselves to understanding how political variables, elections and political parties influence on economic choices of allocation of public resources.

  Finally, we conclude that the electoral calendar generates a cyclic behavior in public spending and that governments from different ideologies choose different levels of resource allocation, where left-wing governments increase spending and right-wing governments reduce them.


ALESINA, Alberto (1987). Macroeconomic policy in a two-party system as repeated game. Quarterly of economics. n0 102, páginas 651-678.

ALESINA, Alberto (1988a). Macroeconomics and Politics. NBER Macroeconomics Annual, Cambridge: M.I.T. Press, p. 13-52.

ALESINA, Alberto (1988b) . Credibility and policy convergence in a two-Party system with rational voters. American Economic Review, vol. 78, p. 796-806.

ALESINA, A.; ROUBINI, Nouriel (1992). Political cycles in OECD economies. The Review of Economic Studies, Vol. 59, pp. 663-68.

ALESINA, A.; SACHS, J. (1988). Political parties and the business cycle in the United States, 1948-1984. Journal of Money, Credit and Banking, vol. 20, p. 63-82.

ARRELANO, Manuel; BOND, Stephen (1991). Some tests of specification for panel data: monte carlo evidence and application to employment equations. The Review of Economic Studies. Vol. 58, p. 277-297.

BALTAGI, Badi H. (2005). Econometric Analysis of Panel Data. John Wiley & Sons Ltd 3ºed. England.

BLUNDELL, Richard; BOND, Stephen (1998). Initial conditions and moment restrictions in dynamic panel data models. Journal of Econometrics. Vol.87, p. 115-143.

BOBBIO, Noberto (1997). O futuro da democracia. 6° Ed. Paz e Terra. Rio de Janeiro.

BORSANI, Hugo (2003). Eleições e economia: instituições políticas e resultados macroeconômicos na América Latina (1979-1998). Belo Horizonte: Editora UFMG. Rio  de  Janeiro.

BOSCHI, R. Renato (2004). Instituições políticas, reformas estruturais e cidadania. IUPERJ. Rio de Janeiro.

COPPEDGE, Michael (1997). A classification of Latin American political parties. University of Notre Dame, Working paper, n. 244.

DAHL, Robert; SHAPIRO, Ian & CHEIBUB, José Antônio (2003). The Democracy sourcebook. Massachusetts Institute of Technology.

DOWNS, Anthony (1999). Uma teoria econômica da democracia. Edusp. São Paulo.

FENOLIO, Fernando Roberto (2007). Ciclos eleitorais: evidências para o Brasil. USP, Dissertação de Mestrado. São Paulo.

FIALHO, Tânia Marta (1997). Testando a evidência de ciclos políticos no Brasil. Revista Brasileira de Economia. Vol.51, p. 379-389.

FIGUEIREDO, A.; LIMONGI, F (1999). Executivo e legislativo na nova ordem constitucional. Editora FGV. Rio de Janeiro.

GREENE, William (2000). Econometric Analysis, 4th Edition, Prentice Hall.

KINZO, Maria D'Alva G (2004). Partidos, eleições e democracia no Brasil pós-1985. Revista Brasileira de Ciências Sociais. vol.19 n°54 São Paulo.

HIBBS, D (1977). Political Parties and Macroeconomic Policy. American Political Science Review, Chicago, vol. 71, p. 1467-1487.

LUCINDA, C.; AVELINO, G.; ARVATE, P (2008). Existe influência Ideologia sobre o resultado fiscal dos governos estaduais brasileiros? Estudos Econômicos. Vol.38, São Paulo, p. 789-814.

NORDHAUS, W (1975). The Political Business Cycle. Review of Economic Studies, New York, vol. 42, p. 169-190.

NAKAGUMA, Marco Y (2006). Ciclos políticos e resultados eleitorais. USP, Dissertação de Mestrado. São Paulo, 2006.

O'DONNEL, Guillermo (1999). Teoria democrática e política comparada. Revista Dados. Rio de Janeiro, vol. 42 n° 4.

PEARSON,T; TABELLINI, Guido (1990). Macroeconomics policy, credibility and polities. Harwood Academic Publishers.

PINDYCK, Robert S.; RUBINFELD, Daniel L (2004). Econometria. 3º ed. Elsevier. Rio de Janeiro.

PRZEWORSKI, Adam & LIMONGI, Fernando (1993). Political regimes and economic growth. The Journal of Economic Perspectives, vol 7, n 3, pp 51-69.

PRZEWORSKI, Adam; ALVAREZ, Michael; CHEIBUB, José Antônio & LIMONGI, Fernando (1996). What makes democracies endure? Journal of Democracy, pp. 39-55.

ROGOFF, Kenneth; SIBERT, Anne (1998). Elections and macroeconomic policy cycles. Review of economic studies. no 55, páginas 1-16.

ROGOFF, Kenneth (1990). Equilibrium political budget cycles. The American Economic Review. Vol.80, p.20-36.

SACHS, Jeffrey & ROUBINI, Nouriel (1989). Political and economic determinants of budget deficits in the industrial democracies. European Economic Review. N0 33, paginas 903 – 938.

SALVATO, M. A. et al (2007). Ciclos Políticos. PUC. Minas Gerais.

SAKURAI, Sérgio Naruhiko (2007). Ciclos eleitorais, reeleição e déficit fiscal nos municípios brasileiros: uma análise via dados em painel. USP, Tese de Doutorado. São Paulo.

SHUMPETER, Joseph A (1961). Capitalismo, socialismo e democracia. Fundo de Cultura. Rio de Janeiro.

1. PhD in economics Professor of UFPB. Author Email :
2. PhD in economics Professor of UFPB
3. Master in Economics Professor of UFRPE

4. In the case of dummy variables take the antilog of the coefficients estimated binary and subtract 1 and multiply the result by 100.


Vol. 37 (Nº 13) Año 2016


[En caso de encontrar algún error en este website favor enviar email a webmaster]