Marriage of love or marriage of convenience? The determinants of pre-electoral coalition formation during the French Fifth Republic 1962–2012

ABSTRACT

In multiparty countries, parties often form strategic alliances to contest elections together. Pre-electoral coalitions can offer a clear alternative to the government, and as such may get the chance to govern. However, the parties have to convince their voters to support the alliance as a unit. Hence the parties may face a dilemma: whether to choose pre-electoral coalition partners that they have the potential to win the most seats with, or ideologically close allies. Pre-electoral coalitions form before an election; thus, normally, we can only infer the parties’ motivations to form them from the results of their cooperation. I circumvent this problem by using manifesto and electoral data from the first round of the French legislative elections from 1962 to 2012 to predict which alliances form in the second round. I find that ideological closeness is an important factor in PEC formation but only if parties can gain electorally from the alliance.

KEYWORDS:

• France
• pre-electoral coalition
• joint competitiveness
• two-round electoral system
• Comparative Manifestos Project

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

1 Parties may write their manifestos after they form a pre-electoral coalition. This means that the ideological distance that we observe between pre-electoral coalition partners may be endogenous to the cooperation itself (Eder, Jenny, and Müller 2017).

2 In this paper, I follow the definition of pre-electoral coalitions of Golder (2006, 12). “A pre-electoral coalition is a collection of parties that do not compete independently in an election, either because they agree to coordinate their campaigns, run join candidates or joint lists, or enter government together following the election.”

3 The threshold increased throughout the years covered by this study. First it was 5% of all the votes cast, in 1966 this increased to 10% of the registered voters and again in 1976 to 12.5% (Clift 2003). As the 12.5% threshold is calculated as the percentage of the registered voters, in reality around 17% of the votes cast were needed to advance to the second round of the elections (Golder 2006).

4 Parties form pre-electoral coalitions before the first electoral round to surpass the threshold necessary to enter to the second electoral round. These coalitions often form on the local level between smaller parties (Blais and Indridason 2007). While it would be very interesting to study these coalitions, they are outside of the scope of this study. Local pre-electoral coalitions do not necessarily lead to a joint government and may have a local policy impact. In this paper, I examine the influence of the potential seat gains within the entire country on the formation of nationwide pre-electoral coalitions. See a more detailed analysis in Blais and Indridason (2007).

5 These are the elections for which district level legislative data are available, which are necessary to conduct this study. The elections covered in this analysis are the French legislative elections of 1962, 1967, 1968, 1973, 1978, 1981, 1988, 1993, 1997, 2002, 2007 and 2012. Data: BDSP/CDSP (1962, 1967, 1968, 1973, 1978, 1981), Ministère de l’Intérieur/CDSP (1988, 1993, 1997, 2002, 2007, 2012). I exclude 1986 because that election was run under Proportional Representation (PR) electoral rules. See Appendix A for a list of successful party dyads.

6 This is not a big problem for this study, since smaller parties join a pre-election coalition before the first round of the elections and may run joint candidates on a local level with a bigger party.

7 If a pre-electoral coalition of three parties forms, they enter into the dataset as three separate party dyads (AB, AC, and BC).

8 In reality the vote share of a realized coalition can be lower or even higher than the votes that the individual parties realized depending on the voter turnout (Tillman 2013).

9 The average vote share of the second candidates were: 0.38 in 1962, 0.42 in 1967, 0.42 in 1968, 0.43 in 1973, 0.44 in 1978, 0.43 in 1981, 0.45 in 1988, 0.42 in 1993, 0.42 in 1997, 0.43 in 2002, 0.44 in 2007 and 0.42 in 2012.

10 In all cases I use the joint vote share of the parties. For instance, let us imagine a country with five electoral districts where in each district there are 100 voters. Let’s suppose further that in the second electoral round the second candidate got 40% of the votes on average, this is the voteshare that the two parties want to exceed. Let us suppose in a given election, Party A gets {15,16,17,14,11} votes, while Party B gets {30, 31, 44, 11, 60} votes, in the electoral districts. To get the vote share we have to divide these numbers with 100. Thus, the joint vote shares of Party A and Party B in the five districts are: {0.45, 0.47, 0.61, 0.25, 0.71}. So now we can calculate how far the two parties together are in each district from the 50% threshold, to the positive or to the negative direction: {|0.45 − 0.4|, |0.47 − 0.4|, |0.61 − 0.4|, |0.25 −0.4|, |0.71 − 0.4|}. Next, we divide each value with 5 which is the number of districts and the probability of the observation in the distribution: {0.05/5, 0.014/5, 0.042/5, 0.03/5, 0.062/5}. Finally, we add the numbers together, which results in 0.79/5, and we multiply this number with 2.5 since we want the measure to have a range between 0 and 1. As multiplication is a linear transformation this transformation will not change the estimates. At the end we find that the Vote Share Spread of Party A and Party B is 0.395.

11 A value of 1 indicates that in any district the two parties together would either get 0% of the votes or 100% of the votes.

12 Duverger predicts that under a majoritarian electoral system in a Single Member District competition only two parties remain. One of the reasons for this is psychological: the voters will not vote for third party candidates (Riker 1982). Knowing this, big parties could hope that voters of small parties vote strategically even if their candidate stays in the competition.

13 The issue categories cover a variety of economic and social topics that could be important for the parties.

14 As the thresholds in each year are calculated from the registered voters after 1962 (it was 0.05 of the turnout in 1962), I adjust this measure with the turnout (Blais and Indridason 2007). The variable is calculated similarly to the Vote Share Spread; the formula is $\mathrm{T}\mathrm{u}\mathrm{r}\mathrm{n}\mathrm{o}\mathrm{u}\mathrm{t}\ast {\sum }_{k=1}^n\left|x_{k\left(i+j\right)}-{\displaystyle \frac{\mathrm{T}\mathrm{h}\mathrm{r}\mathrm{e}\mathrm{s}\mathrm{h}\mathrm{o}\mathrm{l}\mathrm{d}}{\mathrm{T}\mathrm{u}\mathrm{r}\mathrm{n}\mathrm{o}\mathrm{u}\mathrm{t}}}\right|p\left(x_{k\left(i+j\right)}\right),$ The thresholds were 0.05 in 1962, 0.123 in 1967, 0.125 in 1968, 0.123 in 1973, 0.15 in 1978, 0.176 in 1981, 0.19 in 1988, 0.18 in 1993, 0.18 in 1997, 0.19 in 2002, 0.21 in 2007 and 0.21 in 2012.

15 I use this model as the results are sufficiently similar to the model with year dummies, and I do not want to represent any particular year in the predictive plots.

16 I do not argue here that parties cannot signal in any other way that they will govern together, but instead that pre-electoral coalitions are not formed solely for that purpose in France.

17 The control variables gain significance in certain models (Model 5 and Model 6) for certain years. However, the results are not robust to alternative specifications. The log likelihood and the AIC values show that the inclusion of the control variables does not improve the model fit.

18 For this analysis I chose the biggest party on the left and the biggest party on the right party block.

19 I use this approach, as opposed to having fixed effects by party, since in France parties often change names. While the Comparative Manifestos Project accounts for some of these changes within their coding scheme this is not always the case.

Additional information

Notes on contributors

Zsuzsanna Magyar

Zsuzsanna Magyar is a Postdoctoral Fellow at the Institutions and Political Economy Research Group at the University of Barcelona. She has a PhD in Political Science and an MS in Statistics from the University of California, Los Angeles. Her research focuses on party alliances, legislative oppositions, the relationship between party system structure and policy outcomes, and the emergence of party systems in Europe and in Asia.