Party organization, discipline and switchers in the Congress of Peru

 Milagros Campos, Eduardo Moron, Jean Paul Rabanal

Abstract

Using voting records on the motions approved for the period 2006-2011 in the Congress of Peru, we compute party cohesiveness and legislators’ ideal points. We find that the incumbent party (Partido Aprista Peruano) exhibits the highest score on cohesiveness. We also show that incumbent party gained votes from alliances with other political groups and from the switchers that abandon their party affiliation. The ideal points estimation shows a clear distinction between parties that are located in the right and left. The ideal points of the switchers are closer to the members of the incumbent party.

Keywords: Roll Call Voting Record, Ideal Point Estimation.

 

1. Introduction

The peruvian parliament was exposed to high levels of corruption during Alberto Fujimori mandate and in the last ten years has gained legislative transparency and strengthened the accountability of individual legislators to voters. Our objective in this paper is to measure party discipline and compute ideal points following the traditional political literature.

Party discipline is reflected in voting patterns. We use the standard measure of rice to compute party cohesiveness. The party discipline is also affected by the number of congressmen that switches their party affiliation. This phenomenon is common in developed countries as well as developing countries. In Peru, there is a permanent debate about how to change the legislative rules to mitigate it. However, there has not been any studies that identify the switchers and analyze their voting behavior. Collecting voting records for the period 2006-2011, we measure group cohesiveness and estimate ideal points.

The estimation of ideal points has been applied to many countries except Peru. The spatial model assumes that policy alternatives can be represented as points in a geometric space. We estimate these points following the bayesian approach of Clinton, Jackman and Rivers (2004).

Using voting records on the motions approved for the period 2006-2011 we compute party cohesiveness and legislators’ ideal points. We find that the incumbent party (PAP) exhibits the highest score on cohesiveness. We also show that incumbent party gained votes from alliances with other political groups (in particular AF) and from the switchers coming from different groups.

The ideal points estimation shows that Unidad Nacional (UN), Alianza Fujimorista (AF) and Partido Aprista Peruano (PAP) are on one side (right) and Nacionalista (NAC) are on the left. Union por el Peru (UPP), Alianza Parlamentaria (AP) and the switchers are somewhere in between.

The organization of the paper is as follows. Section 2 describes the political parties and groups in the period 2006- 2011, and the data we collect from the congress. Section 3 includes our main analysis. We start with the standard index of party discipline and then compute the ideal points of individual legislators. Finally we conclude in section 4.

 

2. Political parties and congress groups

The Congress is conformed by groups that could be either a political party, congressmen that decide to abandon their party and conform a new group, or an alliance between political parties form in order to gain higher representation in the Congress.

The groups in the Congress 2006-2011 are constituted by the parties UPP (Union por el Peru), Nacionalista (NAC), Partido Aprista Peruano (PAP), the alliances Alianza por el Futuro (AF), Unidad Nacional (UN), and the groups Alianza Parlamentaria (AP), Alianza Nacional (AN), Grupo Especial Democratico (GE) and Bloque Popular (BP). There is a category that classifies as No group (NG) the congressmen that do not have any official group affiliation. Table 2 shows the group affiliation.

At the beginning of the period, UPP and NAC decided to split. Notice that they would be the biggest group in the congress if they maintained the unity. The incumbent party is PAP with around 36 members. The alliance AF resembles the initials of the ex president Alberto Fujimori and had about 13 members. The alliance UN also had an important number of congressmen. However, the alliance broke apart and the members of Solidaridad Nacional and Renovacion Nacional moved away. AP is conformed by small parties that need to merge in order to have a higher representation in the parliament such as Peru Posible and Accion Popular.

Congressmen that abandon their parties –mainly from UPP and UN– also formed new groups: Alianza Nacional (AN), Grupo Especial Democratico (GE) and Bloque Popular (BP).

Notice that we incorporate a row with our definition of Switchers. They are congressmen that decide to abandon their initial party to either form a new group or be part of the no group category. At the end of the period, we have a total of 21 switchers (18 percent of the total congressmen).

Table 2 shows the composition at the end of period of the congressmen by region. We define the region variable Lima and Callao that gets the value of 1 if the congressmen is from that region and 0 otherwise. Of the total 120 legislators, 37 legislators are from Lima and Callao (31 percent). Notice that parties NAC and UPP are mainly conformed by members outside the Lima and Callao region.

2.1 Data Description

Records of legislative votes, indicating how each legislator voted, began to be made available on the Peruvian Congress’s website in 1999. Our database consider the law proposals that were approved from July 2006 to January 2011. In total we work with 1121 votes.

Each legislator’s vote (aye, nay, abstain, noresponse, absent, or suspended) is recorded on each motion, allowing us to construct cohesiveness scores and legislator’s ideal points . It is important to mention that we consider the nay, abstain, noreponse and absent responses as a single nay in our analysis. We also classify the motions as economic, social, government or international issues accordingly to the proposal’s summary .

 

3. Measuring party organization

We analyze the roll data with two standard concepts: the rice index and ideal points. Rice (1927) suggested an index that measures the group cohesiveness |%ayei j −%nayi j| for group i on vote j. Notice that according to our definition, nay, abstain, no response and absent responses are represented as a single nay.

The score ranges from zero (evenly split between aye and nay votes) to one (for a group fully mobilized and unified). Since we are working with law proposals that were approved we expect an upward biased. Table 3 shows the results for the period 2006-2011 classified by issues. We follow Carey (2003) in the classification of issues.

In every issue, PAP (incumbent party) exhibits the highest cohesiveness. Interestingly, the economic issues have relative low cohesiveness for every party except PAP and UN. The social issues shows a high cohesiveness and the UN party shows a low cohesiveness for government issues. The Nacionalista group (NAC) shows a higher index for international issues.

The rice index was only calculated with the congressmen that decided to stay at their originally group. A first step to analyze the switching behavior of the congressmen is to compute individual agreement scores. The agreement score tell us how many times (in terms of number of votes) legislator k agrees with legislator n. We can then identify who decides to switch her more preferred option by computing how many times the legislator agrees with the majority of the relevant group. Table 4 shows the legislator that did switch her most preferred option and we compare whether the “empirical" measure match our switchers definition of Table 2.

Comparing the agreement score and the definition of switchers, we find that half of the legislators that switch their most preferred option also decided to abandon their parties. Furthermore, looking at the initial group affiliation and the most preferred group (information given in parenthesis) there are evidence of: i) a strong alliance between PAP and A; ii) switching is not in favor of NAC and UN and iii) switching does not originate in the incumbent party if we exclude the strategic alliance between AF and PAP.

Before we conclude our analysis of agreement scores, we depict individual agreements according with the main three parties than run for presidential election in 2006 in a triangle (NAC, PAP and UN).1 Each edge represents the majority of each group and each dot is the legislator agreement score with respect to the three parties. Therefore, dots closer to the north agree most with PAP with respect to the other parties. Dots closer to the south west agree most with NAC and dots closer to the south east with UN. Notice that a higher cohesiveness between the three parties implies that agreement scores are located in the middle of the triangle. We plot the agreement scores with respect to the majority relevant parties for the years 2006, 2008 and 2010 in Figure 1.

Figure 1: Legislator agreement with respect to PAP (south west), UN (north) and NAC (south east) parties (the years are depicted clockwise 2006, 2008 and 2010.)

We observe that UPP (brown), AP (yellow) and switchers (black) are somewhere in between the NAC (red) and PAP (blue) parties. The AF (orange) party is closer to the PAP party. Other important feature is that agreement tends to get closer with respect to time. Probably this is line with the common view that the congress become more “populist” and gets closer to the median voter. However, we need a more formal analysis of how groups evolve and alternative measures of polarization in the congress.

3.1 Ideal points

The model of roll call voting we employ is standard in the political science literature: the Euclidean spatial (Enelow and Hinich, 1984). Roll call j ∈ {1,...,m} presents legislator i ∈ {1,...,n} with a choice between an aye position ζj and a nay position ψj , location is R d where d denotes the dimension of the policy space (d = 1 in our case). Let  yi j = 1 if legislator i votes aye on the jth roll call and yi j = 0 otherwise. Legislators are assumed to have quadratic utility functions over the proposals and status quos; i.e. Ui(ζj) = − k ξi −ζj k 2 +ηi j and Ui(ψj) = − k ξi −ψj k 2 +vi j where ξi ∈ R d is the ideal point of legislator i and ηi j and vi j are the errors or stochastic elements of utility, and k . k represents the euclidean norm. Utility maximization implies yi j = 1 if Ui(ζj) > Ui(ψj) and yi j = 0 otherwise. By assumption, the errors ηi j and vi j have a joint normal distribution with E(ηi j) = E(vi j), var(ηi j −vi j) = σ 2 j and the errors are independent across both legislators and roll calls. The probit model that we use is then the following, where πi j is the probability of legislator i voting aye on the jth roll call.

where βj = 2(ζj −ψj)/σj and αj = (ψ ′ jψj −ζ ′ j ζj)/σj and Φ(.) denotes the standard normal distribution.

Figure 2: Posterior means and 95% credible intervals for ideal points (ξi), using 1 121 votes for 2006-2011. Colors represent parties: Orange: AF; Blue: PAP, brown: UPP, red: Nacionalista, yellow: AP and black are the recorded switchers in 2011.

We will estimate the parameters, including the ideal points, following the item-response theory (IRT) model applied to roll call data a la Clinton, Jackman and Rivers (2004). The parameters in the probit model (βj ,ξi ,αj) are estimated via a data-augmented Gibb sampler, see Jackman (2009) pp. 457 for further detail. Our estimation includes 1 121 votes. We observe on the right side the parties UN, AF and PAP. Switchers (black) appear closer to the incumbent party (blue). On the left hand side we locate AP, UPP and NAC. It is worth noting the difference location between the NAC (red) and the incumbent party (blue); and the NAC and the UPP (brown) party, which is consistent with the early division between both parties. One remark is important to mention before concluding. Here the ideal points reflect a number of different influences including personal ideology, the ideology of the legislator’s party, lobbying by interest groups among others.

 

4. Conclusions

Using voting records on the motions approved for the period 2006-2011 we compute party cohesiveness and legislators’ ideal points. We find that the incumbent party (PAP) exhibits the highest score on cohesiveness. We also show that incumbent party gained votes from alliances with other political parties (in particular AF) and from the switchers coming from different groups. The ideal points estimation shows that the parties UN, AF and PAP are on one side and the NAC are on the other side. UPP, AP and the switchers are somewhere in between these sides. There is still work to be done. Our research agenda consists on measuring the legislative polarization. Using agreement scores, it seems that the polarization decreased with time. However, a more formal analysis is still required.

 

References

[1] Carey, J.M. “Transparency versus Collective Action: Fujimori’s Legacy and the Peruvian Congress.” Comparative Political Studies, 36 (9) 983-1006, 2003.

[2] Clinton, J., S. Jackman and D. Rivers. “The Statistical Analysis of Roll Call Data.” The American Political Science Review, 98(2):355-370. 2004.

[3] Rice, S.. “The identification of blocs in small political bodies.” The American Political Science Review, 21(3):619Ð627, 1927.

[4] Jackman, S. Bayesian Analysis for the Social Sciences., John Wiley & Sons, Ltd, Chichester, UK, 2009.