Mobile Politicians: Opportunistic Career Moves

Abstract

Using information on more than 1,500 elected Members of the Parliament (MPs) and the votes received by political parties in five consecutive elections in Turkey (1991–2011) and employing a controlling strategy, we show that elected MPs are more likely to switch parties after an election if they faced greater electoral uncertainty and experienced a narrowly-won victory. Politicians switch parties after an election to improve their ex-ante re-election probability in the following election, and party-switching MPs are more likely to get elected in the next election. MPs switch parties to move towards the median voter. These results point to forward-looking opportunistic behavior of politicians regarding their strategy to win future elections to have a longer tenure in the Parliament.

Appendices

Appendix for Mobile Politicians: Opportunistic Career Moves

Appendix 1: d’Hondt Example

In this section, we first explain how seats are allocated to political parties in the d’Hondt method. We then provide evidence that shows small changes in the distribution of votes can impact the number of seats won by each party. We additionally show that MPs with smaller Margins of Victory (MVs) are at the risk of losing their seats if vote distribution in the district changes even slightly. Finally, we demonstrate how we compute the MVs.

In the d’Hondt method, the number of votes obtained by each party in the district is divided by consecutive positive integers up to, and including, the total number of seats in that district. Parties with the largest quotients win seats. Consider the hypothetical example of a district with 100 voters. Assume that there are seven seats to be allocated and five competing parties: A, B, C, D, and E. Suppose that in the election, these parties have obtained 10, 13, 19, 25, and 33 votes, respectively. d’Hondt method divides each party’s votes by 1, 2, 3, 4, 5, 6, and 7. Parties with the seven largest quotients win seats. This is illustrated in Panel 1 of Appendix 1 Table 9. The seven biggest quotients (highlighted in bold) are 33, 25, 19, 16.5, 13, 12.5 and 11. The quotients 33, 16.5, and 11 are associated with party E. Therefore, Party E wins 3 of the seven available seats. The quotients 25 and 12.5 belong to Party D, so Party D wins two seats, and Parties B and C win one seat each. As a result, party B’s and party C’s top-ranked candidates are elected as MPs. Party D sends its top 2 candidates to the Parliament, and party E’s top 3 candidates are elected as MPs. Party A does not win a seat.

Table 9 Hypothetical d’Hondt example of votes shares and seat allocations

To see how small changes in the distribution of votes can impact who is elected from a given party, compare the vote distributions depicted in Panels 1 and 2 of Appendix 1 Table 9. In both cases, there are 100 voters, and in both cases, parties B, C, and E receive the same number of votes. The only difference between the election results shown in Panels 1 and 2 is that in Panel 2, Party D received 23 votes (instead of 25), and these two voters moved from Party D to Party A. The resultant d’Hondt calculation is presented in Panel 2, which shows that the 2% loss suffered by Party D did not impact the number of seats of Party D (party D still qualified for two seats). On the other hand, this 2% vote loss of Party D produced a seat in the Parliament for Party A at the expense of a seat for Party E. More specifically, this slight shift of votes from Party D to Party A cost one MP to Party E, while Party A gained one seat. As a result of these changes, the 3rd-ranked candidate of Party E is not elected, while the 1st ranked candidate of Party A is elected. As shown in Panel 3 of Appendix 1 Table 9, if two more individuals had voted for party A instead of D, the election outcome would have been more different yet: In that case, parties A, B, C, and D would qualify for one seat each, while party E would have 3 MPs.

The examples demonstrated above suggest that some candidates, especially those who won by a small margin of votes, were at the risk of losing their seats if the vote distribution in the district was slightly different. To identify the MPs who were at high risk of losing the election, we compute the Margin of Victory (MV) for each elected MP for in vote distribution depicted in Panel 1 of Appendix 1 Table 9. Panel 4 of Appendix 1 Table 9 displays these calculations. We define MV as the difference between the quotients of each elected candidate and the candidate who just missed qualifying for the last seat from that district. For example, the person who won the 7th and the last seat in this district is candidate E3, and her margin of victory is 1 (the difference between the E3’s and A1’s quotients). Candidate E1 is the first-ranked candidate of party E; her quotient is 33, and therefore, her MV is 23. Notice that the seats won by MPs with small MVs (D2 and E3) could have been easily lost to another candidate if the vote distribution changed even slightly (as demonstrated in Panels 2 and 3). In our analyses, in addition to using the whole sample of MPs, we employ the subsample of MPs who won by small margins in order to identify the impact of a candidate’s rank on the probability of party switching. In addition, in some specifications, we use the Margin of Victory as an (inverse) measure of electoral uncertainty faced by politicians.

Appendix 2: Distribution of Party Switchers by Party Over Time

Tables 10 and 11

Table 10 The change in parties’ seats in the parliament by parliamentary term

Table 11 The number of MPs and party switchers by parliamentary term in the Grand National Assembly of Turkey

Appendix 3: Discussion of the Estimates of the Control Variables in Table 2 of the Main Text

Personal Attributes of the MP

The estimated impact of MP characteristics on the propensity to switch parties is listed in panel B of Table 2 in the main text. The variable, titled “Relative Salary,” is the ratio of real MP salary to the per capita income in the province from which the MP was elected. MP salaries are periodically adjusted upwards by the Parliament, and there is variation in per capita income between cities. Thus, Relative Salary varies both over time and between cities. Table 2 of the main text shows that an increase in MP salaries in comparison to per capita income has a positive impact on the propensity to switch parties, likely because an increase in salary makes the MP’s post more attractive.

Members of the Parliament who are also members of the cabinet are less likely to switch to another party. Arguably, this is because Cabinet Members have access to government resources, and defecting from their current party may reduce opportunities to obtain governmental pork (Desposato 2006). Those who are elected multiple times from the same party (and have not switched parties) are more likely to be loyal to the party. The coefficient of Elected from the Same Party Before is negative and significant in all specifications, but omitting this variable, which captures party loyalty, had no impact on other coefficients. Freshman parliamentarians are less likely to switch parties. These MPs are elected to the parliament for the first time, and presumably because of this, they did not have enough time to network and make connections with competing parties to be able to switch their party affiliation.

In Table 2, the coefficient of the variable Influence of the MP is about -0.016. This result suggests that a one standard deviation increase in an MP’s influence reduces his/her probability of switching parties by 1.7 percentage points. This is probably because for those MPs with strong influence on other members of their party, party switching is more costly, in that defection from their party involves giving up their sway over other members. The coefficient of Effort of the MP is positive but insignificant in Table 2.

Members of the Parliament who were 50 years of age or older when they were elected are less likely to switch parties. This could be because it may not pay off to switch parties and start off anew at a different party when the time left until retirement is shorter. This finding is consistent with Alesina et al. (2015). In Appendix 3 Table 12, we present a more detailed investigation of the effect of MP’s age. Our results suggest that the effect of age on the probability of party switching is monotonic. In addition, we find that the effect of election uncertainty has a statistically significant impact on party switching in the sample of younger MPs, while older MPs do not respond to election uncertainty. Female MPs are more likely to switch to another party. Those MPs who have received a college or graduate degree outside Turkey (School Abroad) are less likely to switch parties. To receive a college degree abroad could be an indicator of wealth or superior academic performance in high school. MPs who obtained a degree abroad are more likely to be a member of a wealthy families compared to MPs who went to college in Turkey because attending school abroad is more expensive compared to obtaining a degree in the country. Alternatively, an MP could obtain a degree abroad with the support of a scholarship thanks to their superior academic performance in school. In either case, wealthy or academically successful MPs have better outside options in the labor market. Having a graduate degree (MA/PhD) does not significantly influence the probability of party switching. However, our analysis, presented in Appendix 3 Table 13, suggests that MPs who don’t hold graduate degrees respond more strongly to election uncertainty. Specifically, election uncertainty faced by the MP has a statistically significant effect on party switching in the sample of MPs who do not have graduate degrees but not in the sample of MPs who have MA or Ph.D. degrees. Conditioning on the MPs’ area of study or their universities’ quality does not change this finding. The quality of MP’s university, as measured by the selectiveness of the colleges, reduces the probability of party switching in the sample of MPs who do not have a graduate degree. These results are presented in Appendix 3 Table 13. In addition, Appendix 3 Fig. 2 shows that the impact of Ranked 2nd or Lower on the probability of party switching is greater when the university quality of the MP is smaller.

Table 12 The effect of electoral uncertainty on the probability of party switching by MP’s age

Table 13 The effect of electoral uncertainty on the probability of party switching by MP’s education conditional on MP’s area of study and university quality

An interesting result in Table 2 of the main text is the impact of Member of the Government Party. This variable takes the value of one if the MP’s party is the governing party in the Parliament, either as the majority party or as part of a coalition government. The results show that if MP is a member of the majority (the governing party), then his/her probability of switching parties is seven percentage points larger.¹⁸ This is likely because it is a more attractive proposition for the MP to transfer to another party if he/she is a member of the majority party because such a move weakens the government and strengthens the opposition. As an extension, we investigated whether the impact is larger when the margin of the majority of the government is smaller. Results presented in Appendix 3 Fig. 3 support this idea. Specifically, compared to their counterparts in the opposition parties, members of the parties that formed the government are more likely to switch parties when the majority of the government in the Parliament is smaller.

The Impact of Party and District Attributes

Panel C of Table 2 in the main text shows that, all else the same, MPs are more likely to switch parties after an election if they are members of a party that has a longer existence in politics. On the other hand, if the leader of the party has held that position of leadership for at least ten consecutive years (Single Leader Party = 1), this reduces the propensity to switch parties. This could be because in such parties, the entire leadership team may be more stable, and therefore, they may be more attuned to the rank and file of the party, more effective in dealing with grievances, and reacting more promptly to any attempts to defection from the party.

Panel C of Table 2 of the main text shows that the higher the number of Contested Seats in a District, the lower the likelihood of party switching of an MP who was elected from that district. This result, which indicates that the MPs who are elected from smaller districts are more likely to switch parties after an election, makes sense to the extent that those seats are more valuable to transfer to another party (Figs. 2 and 3).

Fig. 2

Effect of election uncertainty by university quality. We estimate a probit regression of the form $P\left(Party\;Switch\;=\;1\right)=\Phi\;({\beta\;}_{1}Ranked\;2nd\;or\;Lower\;+\;{\beta\;}_{2}Quality+{\beta\;}_{3}Ranked\;2nd\;or\;Lower\times\;Quality\;+\;{{X}}{{{\beta}}}_{4}+\varepsilon\;)$ where $Quality$ is the average university entrance exam score of the MP’s university. Greater $Quality$ implies a more selective university. The solid connected line represents the marginal effect of $Ranked 2nd or Lower$ evaluated at various $Quality$ levels. Dashed lines are two standard error confidence intervals

Fig. 3

The effect of being a member of the government party on party switching (by the strength of the government). The figure presents the effect of Member of the Government Party on the probability of Party Switch = 1. To obtain these marginal effects, we first run a probit regression same as the specifications in Table 2 in the main text with the following exceptions: 1. Among the controls, we include a new variable, No. Seats above Majority, which is constructed as the number of seats the governmental parties have in the Parliament at the beginning of the term minus half of the seats in the Parliament. We also include the interaction of No. Seats above Majority with Member of the Government Party. 2. We exclude Parliamentary term dummies. The connected line represents the marginal effects of Member of the Government Party evaluated at several values of No. Seats above Majority. Dashed lines represent two standard deviation confidence intervals. The confidence intervals start to contain zero after No. Seats above Majority = 50

Appendix 4: Sensitivity Checks

Tables 14, 15, and 16

Table 14 The effect of electoral uncertainty on the probability of party switching and being independent marginal effects from multinomial logit

Table 15 The effect of electoral uncertainty on the probability of remaining independent conditional on leaving the original party after the election

Table 16 Covariate balance in the matched and unmatched samples

Appendix 5: The Impact of Party Switching on the Electoral Success of the Destination Party

The results of Table 7 of the main text indicate that party switching MPs have slightly higher chances of re-election. While this finding explains why MPs switch parties, it seems to go against the prediction that voters would reprimand dishonest politicians by not re-electing them. It should be remembered, however, that in this electoral system, voters cast votes for the entire party ticket. In other words, even if voters are uncomfortable with having a politician on their party’s ticket who was elected to the Parliament in the previous election from another party, voters may still cast their votes for their party because the utility of their party sending more MPs to the Parliament may outweigh the disutility of electing the dishonest party-switcher.

To shed more light on this issue, we also analyze how a party’s election success in a district is impacted by having on the ticket an MP who was transferred from another party. In these regressions, the unit of observation is a province-election-party. Specifically, consider Eq. (6) below:

$${{\%\;Vote\;in\;District}_{pct} = \gamma {\;\%National\;Vote}_{pt} +\theta {Gov{'}tParty}_{pt}+{\lambda }_{c}+{\pi }_{t}+{\xi }_{p}+{\upsilon }_{pct} }$$

(6)

where the proportion of votes received by the party (p) in the province (c) in an election year (t) is depicted by %Vote in District_pct, which is a function of the nation-wide success of the party (% National Vote_pt) and whether the party entered the election as the Governing Party (Gov’t Party_pt). The political sentiment in the province, which impacts the number of votes received by the party, is captured by province fixed effects λ_c. Election year fixed-effects are captured by π_t, and party fixed-effects are represented ξ_p.

Writing Eq. (6) for the next election year t + 1 and taking the first differences, we estimate Eq. (7) below:

$${\Delta\;\%Vote\;in\;District}_{pct}={\gamma\;\Delta\;\% National\;Vote}_{pt} +{\theta\;\Delta\;Gov'tParty}_{pt}+{ \varphi\;MPs\;Gained}_{pct}+{\delta }_{t}+{\mu }_{c}+{\tau }_{p}+{u}_{pct}$$

(7)

where Δ stands for the change in the relevant variable between the two consecutive elections.¹⁹ Although district fixed-effects and party fixed-effects drop out from eq. (7), we nevertheless keep them to absorb any unobserved residual variation. In eq. (7), the variable MPs Gained_pct stands for the number of MPs who switched to party (p) from other parties between the two elections and are now nominated from the party (p). Thus, we analyze how the vote share of the party (p) changes from one election to another and whether this change is impacted by whether or not the party’s ticket in that election in that province contains an MP who was transferred from another party.

The results are reported in Appendix 5 Table 17. The key variable is MPs Gained, which stands for the number of MPs who switched to the party from another party during the parliamentary term just ended. The results show that if the party’s ticket contains an MP who is transferred from another party, then the proportion of votes received by that party in that province goes up by about two percentage points. This means that in a province with 100,000 voters, the party has received 2,000 more votes (in comparison to the previous election) because of the transfer of the MP from another party.

Table 17 The Change in the proportion of votes received by parties (in consecutive elections)