Skip to main content
SpringerLink
Log in

Voting and unemployment in French national elections

  • Data, Measures and Methods
  • Published:
French Politics Aims and scope

Abstract

The purpose of this article is to study and to compare the influence of local unemployment on the vote in French elections for the European Parliament, in French legislative elections and in French presidential elections. The results of our paper showed that if we study the vote in the second round of French presidential elections over the 1988–2007 period (without 2002), the influence of the annual change in the local unemployment on the vote for the Left is much greater than for the vote for the Left in the first round of legislative elections (1986–2007), and French elections for the European Parliament (1984–2009). For French presidential elections, we build two variables taking into account the responsibility of the incumbent President for the economic situation after a cohabitation period. We make estimates for the second-round vote of French presidential elections (without the 2002 French presidential election or with an estimated vote for this election). We show that over the 1988–2007 and 1988–2012 periods (without 2002), it is not necessary to take into account the influence of cohabitation periods on the responsibility of the government in relation to the economic situation.

This is a preview of subscription content, log in via an institution to check access.

Access this article

Log in via an institution

Price excludes VAT (USA)
Tax calculation will be finalised during checkout.

Instant access to the full article PDF.

Institutional subscriptions

Similar content being viewed by others

Notes

  1. The national elections in France are: the presidential election, the legislative elections and the French elections for the European Parliament (since 2004, the French elections for the European Parliament are held by electoral district (big region) but some chief candidates (or candidates) do not have any local mandate or are only little known at the local level; thus, we think that the nationwide campaign is more significant than the local grassroot one.

  2. We chose to present the article of Fair (2009) that enabled us to compare the influence of the economic situation between the US presidential elections and the US House elections.

  3. We find a symposium: Forecasting the 2012 American national elections with a recap in PS: Political Science & Politics, 2012, p. 45 and in PS: Political Science & Politics, 2013, p. 46; and Forecasts of the 2010 mid-term elections in PS: Political Science & Politics, 2010, p. 43.

  4. In Auberger (2013a) for French presidential elections and in Auberger (2013b) for French legislative elections, we used a corrected popularity variable (estimated by removing the influence of the economic situation).

  5. For example, in Lewis-Beck and Tien (2012), the dependent variable is the two-party popular presidential vote share in the American presidential elections, and in Lewis-Beck and Tien (2010), the dependent variable is the House Seat Change in the American mid-term elections.

  6. Number of quarters in the first 15 quarters of the administration in which the growth rate of real per capita GDP is greater than 3.2 per cent at an annual rate except for 1920, 1944 and 1948, where the values are 0 (1916–2004 for US presidential elections and US on-term House elections and 1918–2006 for US mid-term elections).

  7. The results can be nuanced if we take into account that in both equations for the vote in the US House elections, there is a variable for the vote in the previous elections (2 years before); then, past economic situation before this previous election is taken into account by this variable.

  8. The comparison was possible but was more difficult because the dependent variables and the number of quarters were not the same.

  9. Precise definition of the variables and sources of the data are shown in Appendix.

  10. That is, two quarters before the elections for French elections for the European Parliament and one quarter before the elections for French presidential elections.

  11. For French elections for the European Parliament, we obtained more satisfactory statistical indicators with a local unemployment variable calculated with data of two quarters before the elections (December) than with a local unemployment variable calculated with data of one quarter before the elections (March).

  12. The random effects reflect the influence of other independent variables not incorporated into the model. In a model with FE, we can have a correlation between the individual effects (f i ) and the independent variables of the model, whereas in a model with RE, we suppose that there is an absence of correlation between the individual effects (u i ) and the independent variables of the model.

  13. For the ML1(Pc), L2(Pa), L2(Pc), L2(Pf) models the results of the Hausman test were favorable to the model with RE at the statistical level of 5 per cent; for the ML(Ea), ML1(La), ML1(Pb), L2(Pb), L2(Pd), L2(Pe) models the results of the Hausman test were favorable to the model with FE at the statistical level of 5 per cent; for the ML1(Pa) model the results of the tests were favorable to the model WE at the statistical level of 5 per cent.

  14. For vote models with RE, we applied the OLS to the transformed data as indicated in Wooldridge (2002, pp. 286–287); and we used the residuals (Wooldridge, 2002, p. 288) to make these two tests. We showed that the null hypothesis of homoscedasticity is rejected at the statistical level of 5 per cent for ML(Ea), ML1(Pa), ML1(Pb), ML1(Pc), L2(Pc) estimates; the null hypothesis of homoscedasticity is accepted at the statistical level of 5 per cent for ML1(La), L2(Pa), L2(Pb), L2(Pd), L2(Pe), L2(Pf) estimates; and the null hypothesis of a first-order autocorrelation of the errors is rejected at the statistical level of 5 per cent for all the estimates except for ML(Ea), L2(Pd). For all the estimates, we used the Newey–West (or the White) method.

  15. A vote function of the form: VOTEML1=α0+α1DUNEMNL t +α2DUNEMNL × COHAB t (COHABA or COHABB)+u i +ɛ it was estimated. We obtain (expected sign) but (this is not the expected result but this result is consistent with the estimates of Table 6).

  16. The decrease of the local unemployment rate in 44 departments did not lead to an increase in the first-round vote for the moderate Left (a decrease in the local unemployment rate in 37 departments and stability in 15 departments).

  17. The average of the first-round vote received by the extreme Left is equal to 6.06 per cent for French presidential elections, 3.39 per cent for the EMP elections and 2.09 per cent for French legislative elections.

  18. In the EMP elections, there is only one round; in French legislative elections, there are two rounds but the first round is very important because some candidates are elected in the first round.

  19. When we study the first-round vote for the moderate Left over the 1988–2002 period, we notice that the estimated coefficient of the DUNEMLCOA variable is negative (this is the expected sign) but is not significantly different from 0 at the level of 10 per cent; over the 1988–1995 period, the estimated coefficient of the DUNEMLCOA variable is negative and significantly different from 0 at the level of 5 per cent.

  20. Victory of the Left (incumbent President) with a little increase over 1 year in the national unemployment rate.

  21. If we add the INFLL-independent variable (calculated in December of the year preceding the elections) in the estimations, its estimated coefficient is equal to 0.22 (this is not the expected sign) and the estimated coefficient of the DUNEML-independent variable is equal to −2.87.

  22. If we add the INFLL-independent variable (calculated with the data of the quarter preceding the elections) in the estimations, its estimated coefficient is equal to −0.50 (this is the expected sign) and the estimated coefficient of the DUNEML-independent variable is equal to −2.37.

  23. The unemployment rate was calculated with a methodology used until July 2007 for the published data and until March 2009 for the calculated data.

  24. A part of the voters for the National Front are former voters of the moderate Left.

  25. We find the ML1(a) and ML1(b) estimates in Auberger (2012).

  26. If we add the INFLL-independent variable (calculated in December of the year preceding the elections) in the estimations, its estimated coefficient is equal to 2.65 (this is not the expected sign) and the estimated coefficient of the DUNEML-independent variable is equal to 2.92 (this is not the expected sign).

  27. At the beginning of this section, we showed the instability of the estimations obtained for the first-round vote of French presidential elections, and, in particular, for the estimates with the DUNEMLCO variable.

  28. A vote function of the form: VOTEL2=α0+α1DUNEMNL t +α2DUNEMNL × COHAB t (COHABA or COHABB)+u i +ɛ it was estimated. Over the period 1988–2007 (without 2002), we obtain (this is the expected sign) but is not different from 0 at the statistical level of 10 per cent (this is not the expected result but this result is consistent with the estimates of Table 7).

  29. We give the priority to the results of Table 7, because for the results of Table 8, we used an estimate for the result of the second-round vote of the 2002 French presidential election (there was no second round in the 2002 French presidential election between the moderate Left and the moderate Right but a duel between the moderate Right and the extreme Right).

  30. Over the 1988–2007 (without 2002), if we add the INFLL-independent variable (calculated in December of the year preceding the elections) in the estimations, its estimated coefficient is equal to 1.39 (this is not the expected sign) and the estimated coefficient of the DUNEML-independent variable is equal to −0.99. Over the 1988–2007 (with 2002), if we add the INFLL-independent variable (calculated in December of the year preceding the elections) in the estimations, its estimated coefficient is equal to 1.44 (this is not the expected sign) and the estimated coefficient of the DUNEML-independent variable is equal to 1.40 (this is not the expected sign).

  31. It seems preferable to use the DUNEML or DUNEMLCOA-independent variable because over the 1988–1995 period, the estimated coefficients obtained for the DUNEML and DUNEMLCOA variables are negative (expected sign) and significantly different from 0 at the level of 1 per cent whereas that of the DUNEMLCO variable is positive and significantly different from 0 at the level of 1 per cent.

  32. A vote function of the form: VOTEL2=α0+α1DUNEMNL t +α2DUNEMNL × COHAB t (COHABA or COHABB)+u i +ɛ it was estimated. Over the period 1988–2007 (with 2002), we obtain (this is the expected sign) but (this is not the expected result but this result is consistent with the estimates of Table 8).

  33. The unemployment rate was calculated with a methodology used since March 2008.

  34. We remind you that we give priority to the results obtained for the second-round vote for the Left without taking into account the 2002 French presidential election. Over the 1988–2002 period, the estimated coefficient of the DUNEML variable is positive but is not significantly different from 0 at the level of 10 per cent because there was a decrease in the local unemployment rate in numerous departments but that do not lead to a victory of the Left; and the estimated coefficients of the DUNEMLCO and DUNEMLCOA variables are negative (equal to −1.48 and −4.39, respectively) and significantly different from 0 at the level of 5 per cent.

  35. For all the models, the results of the Hausman test were favorable to the model with RE at the statistical level of 5 per cent except for ML(Eb) and ML1(Lb).

  36. We showed that the null hypothesis of homoscedasticity is rejected at the statistical level of 5 per cent for ML1(Pd), ML1(Pe), ML1(Pf), L2(Pi), L2(Pj), L2(Pl) estimates, and the null hypothesis of homoscedasticity is accepted at the statistical level of 5 per cent for ML(Eb), ML1(Lb), L2(Pg), L2(Ph), L2(Pk) estimates; and the null hypothesis of a first-order autocorrelation of the errors is rejected at the statistical level of 5 per cent for ML1(Lb), ML1(Pd), ML1(Pe), ML1(Pf) estimates, and the null hypothesis of a first-order autocorrelation of the errors is rejected at the statistical level of 5 per cent for ML1(Lb), L2(Pg), L2(Ph), L2(Pk), L2(Pj), L2(Pk), L2(Pl) estimates. For all the estimates, we used the Newey–West (or the White) method except for ML(Eb), L2(Ph), L2(Pi), L2(Pk).

  37. In that case, if we do not take into account the influence of the cohabitation periods, the influence of unemployment on the vote for the Left is smaller for French presidential elections than for French legislative elections and is comparable to that of the French elections for the European Parliament for the 1988–2007 period; but, when we include the 2012 French presidential election, the influence of unemployment on the vote for the Left for French presidential elections is comparable to that of French legislative elections and is greater to that of the French elections for the European Parliament (1988–2012 period).

  38. With always the same reserve for the estimates obtained for the second-round vote over the 1988–2007 and 1988–2012 periods (with 2002).

References

  • Auberger, A. (2005) Forecasts of the 2004 French European election. Swiss Political Science Review 11 (3): 61–78.

    Article  Google Scholar 

  • Auberger, A. (2012) Voting and economic factors in French elections for the European parliament. Public Choice 153 (3–4): 329–340.

    Article  Google Scholar 

  • Auberger, A. (2013a) Vote, popularité, chômage et élections présidentielles françaises. Paris, University of Paris 2 (mimeo).

  • Auberger, A. (2013b) Vote, Popularity, Unemployment and French Legislative Elections. Paris, University of Paris 2 (mimeo).

  • Auberger, A. and Dubois, E. (2005) The influence of local and national economic conditions on French legislative elections. Public Choice 125 (3–4): 363–383.

    Article  Google Scholar 

  • Bélanger, E., Lewis-Beck, M.S. and Nadeau, R. (2012) Choosing the dependent variable: Sarkozy’s forecasting lesson. French Politics 10 (4): 364–368.

    Article  Google Scholar 

  • Fair, R.C. (2009) Presidential and congressional vote-share equations. American Journal of Political Science 53 (1): 55–72.

    Article  Google Scholar 

  • Fauvelle-Aymar, C. and Lewis-Beck, M.S. (1997) L’Iowa donne l’opposition gagnante. Libération 23 May: p. 15.

  • Foucault, M. (2012a) Forecasting the 2012 French legislative election. French Politics 10 (1): 68–83.

    Article  Google Scholar 

  • Foucault, M. (2012b) Looking back on the forecasting of the French legislative election. French Politics 10 (4): 383–388.

    Article  Google Scholar 

  • Foucault, M. and Nadeau, R. (2012) Forecasting the 2012 French presidential election. Political Science & Politics 45 (2): 218–222.

    Article  Google Scholar 

  • Hibbs, Jr. D.A. (2008) Implications of the ‘bread and peace’ model for the 2008 US presidential election. Public Choice 137 (1): 1–10.

    Article  Google Scholar 

  • Hibbs, Jr. D.A. (2010) The 2010 Midterm Election for the US House of Representatives. Göteborg, Sweden: CEFOS. Working Paper no. 9.

  • Hibbs, Jr. D.A. (2012) Obama’s reelection prospects under ‘bread and peace’ voting in the 2012 US presidential election. Political Science & Politics 45 (4): 635–639.

    Article  Google Scholar 

  • Hibbs, Jr. D.A. (2013) The bread and peace model: 2012 presidential election postmortem. Political Science & Politics 46 (1): 41.

    Article  Google Scholar 

  • Jérôme, B. and Jérôme-Speziari, V. (2012a) Forecasting the 2012 French presidential election: Comparing vote function simulations and vote intention polls. French Politics 10 (1): 22–43.

    Article  Google Scholar 

  • Jérôme, B. and Jérôme-Speziari, V. (2012b) Forecasting the 2012 French presidential election: Lessons from a region-by-region political economy model. French Politics 10 (4): 373–377.

    Article  Google Scholar 

  • Jérôme, B., Jérôme-Speziari, V. and Lewis-Beck, M.S. (2003) Reordering the French election calendar: Forecasting the consequences for 2002. European Journal of Political Research 42 (3): 425–440.

    Article  Google Scholar 

  • Key, Jr. V.O. (1966) The Responsible Electorate. Cambridge: Harvard University Press.

    Book  Google Scholar 

  • Lewis-Beck, M.S. (1986) Une modèle de prévision des élections législatives françaises (avec une application pour 1986). Revue française de science politique 35 (6): 1080–1091.

    Article  Google Scholar 

  • Lewis-Beck, M.S. (1997) Who’s the chef? Economic voting under a dual executive. European Journal of Political Research 31 (3): 315–325.

    Google Scholar 

  • Lewis-Beck, M.S. and Nadeau, R. (2000) French electoral institutions and the economic vote. Electoral Studies 19 (2–3): 171–182.

    Article  Google Scholar 

  • Lewis-Beck, M.S. and Tien, C. (2010) The referendum model: A 2010 congressional forecast. Political Science & Politics 43 (4): 637–638.

    Article  Google Scholar 

  • Lewis-Beck, M.S. and Tien, C. (2012) Election forecasting for turbulent times. Political Science & Politics 45 (4): 625–629.

    Article  Google Scholar 

  • Lewis-Beck, M.S., Bélanger, E. and Fauvelle-Aymar, C. (2008) Forecasting the 2007 French presidential election: Ségolène royal and the Iowa model. French Politics 6 (2): 106–115.

    Article  Google Scholar 

  • Nadeau, R., Lewis-Beck, M.S. and Bélanger, E. (2010) Electoral forecasting in France: A multi-equation solution. French Politics, International Journal of Forecasting 26 (1): 11–18.

    Article  Google Scholar 

  • Nadeau, R., Lewis-Beck, M.S. and Bélanger, E. (2012) Proxy models for election forecasting: The 2012 French test. French Politics 10 (1): 1–10.

    Article  Google Scholar 

  • Paldam, M. (1981) A preliminary survey of the theories and findings on vote and popularity functions. European Journal of Political Research 9 (2): 181–199.

    Article  Google Scholar 

  • Wooldridge, J.M. (2002) Econometric Analysis of Cross Section and Panel Data. Cambridge: MIT Press.

    Google Scholar 

Download references

Author information

Authors and Affiliations

  1. CRED, University of Paris II, 7 rue de la grande chaumière, Paris, 75006, France

    Antoine Auberger

Authors
  1. Antoine Auberger
    View author publications

    You can also search for this author in PubMed Google Scholar

Additional information

This work was partially made when the author was invited researcher in the CES, University of Paris I.

Appendix

Appendix

illustration

figure a

Rights and permissions

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Auberger, A. Voting and unemployment in French national elections. Fr Polit 12, 235–254 (2014). https://doi.org/10.1057/fp.2014.14

Download citation

  • Published:

  • Issue Date:

  • DOI: https://doi.org/10.1057/fp.2014.14

Keywords

Search

Navigation