Portal de Periódicos da CAPES

What Are (Semi)Presidential Elections About? A Case Study of the Portuguese 2006 Elections

2007; Taylor & Francis; Volume: 17; Issue: 3 Linguagem: Inglês

10.1080/17457280701617094

1745-7297

Pedro C. Magalhães,

Political Systems and Governance

Abstract Abstract This article examines what has driven voters’ choices in the 2006 presidential election in Portugal. Electing a semipresidential head of state has often been treated either as a “popularity contest” or as a full‐fledged “first‐order” election, depending on the particular national political system in which those elections were studied. Using data from a panel survey conducted following the 2005 legislative and 2006 presidential elections, this article suggests that, in a “premier‐presidentialist” system such as Portugal – where presidents are neither the heads of the executive nor mere figureheads – voters are unlikely to be oblivious to the conventional partisan and ideological cues provided by campaigns, but also unlikely to see these elections as a mechanism with which to hold government accountable. Instead, patterns of defection from the government party seem to conform to theoretical expectations derived from the notion that presidential elections in such cases can be conceived as “less important”, but where parties and voters remain aware of the connections between what is at stake in different electoral arenas. Acknowledgements Early versions of this article were presented at the Lisbon Discussion Group on Institutions and Public Policy of the ICS‐UL and the New University’s Faculty of Economics, at the University of Notre Dame’s Kellogg Institute’s Lecture Series, and at the CONNEX‐Conference on the Multilevel Electoral System of the European Union that took place in Cadenabbia, Italy, 2007. I am grateful to the audience in those venues for all the comments I received, as well as to three anonymous referees for their useful suggestions. Notes 1. For other analyses of these amendments and their consequences, see, for example, Canotilho and Moreira (1991 Canotilho, J. J. G. and Moreira, V. 1991. Fundamentos da constituição, Coimbra: Coimbra Editora. [Google Scholar]) and Sousa (1992 Sousa, M. R. de. 1992. O sistema de governo português, Lisbon: AAFDL. [Google Scholar]). Some discussion persists about the extent to which the 1982 constitutional amendments really eliminated the accountability of the executive before parliament. By stating, since 1982, that the president can only dismiss the executive “when that is necessary to ensure the regular functioning of the democratic institutions” (Portuguese Constitution, article 195), constitutional law obviously leaves open what this “institutional accountability” – rather than “political accountability” – really means, allowing the president the last word in the interpretation of what is required to ensure such “regular functioning”. In any case, the fact is that no prime minister has been dismissed by the president since 1982 on these grounds, and at least one president has, in public statements, openly expressed his interpretation of the Constitution as precluding the political accountability of the executive vis‐à‐vis the presidency (see Araújo, 2003 Araújo, A. 2003. “El presidente de la república en la evolución del sistema politico de Portugal”. In Portugal: democracia y politico, Edited by: Barreto, A., Fortes, B. G. and Magalhães, P. Madrid: Siglo XXI. [Google Scholar]). 2. One visible implication of this is the fact that turnout is higher in French presidential than in parliamentary elections. On average, the level of turnout in the first round of the eight French presidential elections conducted from 1965 to 2007 was 81.1%, against 73.8% in the first round of the ten legislative elections conducted since 1967. 3. Such power (not shared by all premier‐presidential systems) could be treated, at first sight, as a “functional equivalent” of the ability to dismiss the cabinet. However, a crucial difference is that, unlike what occurs with the presidential power of cabinet dismissal, the composition of the following cabinet is a decision that requires the electorate to be heard in elections, whose results, as we have seen previously, place crucial constraints on the president’s nomination of a prime minister. This is not to say that the ability to dissolve the assembly is not an important prerogative, bargaining chip or power‐enhancing threat for the president in his or her relation with the parliamentary majority, but “it is the electorate that ultimately determines whether the president’s bargaining authority is enhanced or diminished” (Shugart, 2005 Shugart, M. S. 2005. Semi‐presidential systems: dual executive and mixed authority patterns. French Politics, 3: 323–351. [Crossref] , [Google Scholar]). 4. Although presidential vetoes of government decrees are absolute and final (Amorim Neto, 2003 Amorim Neto, O. 2003. “Portugal: changing patterns of delegation and accountability under the president’s watchful eyes”. In Delegation and Accountability in Parliamentary Democracies, Edited by: Strøm, K., Müller, W. C. and Bergman, T. Oxford: Oxford University Press. [Crossref] , [Google Scholar]), governments supported by an absolute majority can introduce and approve those decrees as parliamentary bills. Nevertheless, the considerations below about political costs also apply here. 5. See Araújo (2003 Araújo, A. 2003. “El presidente de la república en la evolución del sistema politico de Portugal”. In Portugal: democracia y politico, Edited by: Barreto, A., Fortes, B. G. and Magalhães, P. Madrid: Siglo XXI. [Google Scholar]: 87–89), for a brief summary. 6. For a detailed analysis of these elections, their antecedents and short‐term consequences, see Almeida and Freire (2005 Almeida, P. T. de and Freire, A. 2005. The 2004 European election and the 2005 legislative election in Portugal. South European Society and Politics, 10: 451–464. [Taylor & Francis Online] , [Google Scholar]). 7. On the political context of the 2006 election, see Salgado (2007 Salgado, S. 2007. Parliamentary and presidential elections in Portugal 2005 and 2006. Electoral Studies, 26: 506–510. [Crossref], [Web of Science ®] , [Google Scholar]). 8. The 2005 and 2006 post‐electoral surveys were coordinated by the Social Sciences Institute of the University of Lisbon (ICS‐UL). Fieldwork was conducted by the Centre for Public Opinion Polls and Studies of the Portuguese Catholic University. These surveys were conducted under the auspices of the research programme “Comportamento Eleitoral dos Portugueses”, coordinated by António Barreto, at ICS‐UL. Financial support was received from ICS‐UL, the Fundação para a Ciência e Tecnologia, the Tinker Foundation, the Fundação Calouste Gulbenkian, the Fundação Luso‐Americana para o Desenvolvimento, the Comissão Nacional the Eleições and the Secretariado Técnico dos Assuntos para o Processo Eleitoral. For the 2005 survey, a multistage area sampling method was employed, with localities in Continental Portugal (64) randomly selected, with probability proportional to size, within strata defined by region and locality size. Households were selected by random route and respondents were selected randomly within each household, provided they were 18 years of age or older. In the absence of the respondent, two additional visits to each household were made. The survey’s fieldwork took place between 5 March and 8 May. Sample size is 2,801. The 2006 survey was applied, by phone, to a sub‐sample of the original 2005 sample. Of the 1,200 respondents selected from the original sample, preserving the original stratification by region and size of locality, 812 responded to the 2006 survey, meaning that 812 respondents were surveyed in the two waves. Fieldwork of the second wave took place between 25 January and 7 February. The 2005 survey was conducted in connection to the Comparative Study of Electoral Systems network. 9. For the analysis of voting behaviour in the 2005 elections, vote for the PSD or the CDS‐PP, as expressed in the vote recall question was coded as 1, PS as 2, PCP, BE and others as 3, and abstention as 4. For the 2006 presidential elections, vote for Soares or Alegre was coded as 1, Cavaco Silva as 2, Jerónimo, Louçã or Garcia as 3, and abstention as 4. 10. Although party identification used to be seen as a mere proxy for actual voting behaviour in the European context (Butler and Stokes, 1969 Butler, D. and Stokes, D. E. 1969. Political Change in Britain, London: Macmillan. [Google Scholar]), the decline of party identification in Western democracies in the last decades and the factors found to influence it, together with the fact that the impact of party identification on the vote seems to vary across countries according to predictable institutional factors (Dalton, 2000 Dalton, R. J. 2000. “The decline of party identification”. In Parties without Partisans: Political Change in Advanced Industrial Democracies, Edited by: Dalton, R. J. and Wattenberg, M. Oxford: Oxford University Press. [Google Scholar]; Norris, 2004 Norris, P. 2004. Electoral Engineering: Voting Rules and Electoral Behavior, Cambridge: Cambridge University Press. [Crossref] , [Google Scholar]), is scarcely compatible with the notion that party identification, as measured in surveys, is a mere proxy for the vote. Party identification can “be usefully applied in most democratic systems” (Dalton, 2000 Dalton, R. J. 2000. “The decline of party identification”. In Parties without Partisans: Political Change in Advanced Industrial Democracies, Edited by: Dalton, R. J. and Wattenberg, M. Oxford: Oxford University Press. [Google Scholar]: 20), particularly those where party systems that are weakly anchored in social structures (Marsh, 2006 Marsh, M. 2006. Party identification in Ireland: an insecure anchor for a floating party system. Electoral Studies, 25: 489–508. [Crossref], [Web of Science ®] , [Google Scholar]). Besides, around half of the Portuguese voters declare themselves not to be close to any party and, as we shall see in the following analyses, Portuguese partisans – particularly when strength of partisanship is taken into account – are, albeit more likely, not certain to vote for a party or candidate. 11. See descriptive statistics for all variables in the Appendix. 12. In cases where these values are outside the range of the scales, we use the appropriate endpoints. 13. All calculations made with the help of XPost: Post‐Estimation Interpretation Using Excel, by Simon Cheng and Scott Long, available at http://www.indiana.edu/~jslsoc/xpost.htm. 14. Several alternative models were run to check the robustness of these results. A model with voting for Soares and Alegre coded as different values of the dependent variable shows that voters for the two candidates emanating from the Socialist area are predictably similar in terms of their level of partisan attachment to the Socialist Party and, in fact, in all other characteristics except their separate evaluation of the qualities of the two candidates. All remaining results were confirmed. In order to address potential problems of endogeneity, the model presented in Table 3 was estimated while measures of partisan attachment to the PS in the second wave of the survey were replaced by similar measures obtained in the first wave of the survey. In all cases, coefficients for partisanship preserved the same sign and level of statistical significance, albeit their size was slightly reduced in the case of voting for Cavaco and slightly increased in the case of voting for Others and Abstention. All results are available with the author upon request. 15. Alternative specifications of the model were tested in order to confirm this finding. Replacing voters’ evaluation of government performance with the sociotropic retrospective evaluation of the economy, in order to test a more specific economic accountability hypothesis, produced statistically insignificant results across the board. Keeping government performance in the model and replacing partisanship in the second wave of the survey by partisanship attachment measured in the first wave makes the coefficient for Cavaco Silva significant, but only marginally so (p < .04). In fact, coefficients for evaluation of government performance were only statistically significant at the levels and directions found for the legislative voting model in Table 2 when both partisanship and leader effects were excluded altogether from the presidential model. This suggests that, in presidential elections, voters with better evaluations of the government may have been more likely to vote for Socialist than opposition candidates, but only insofar as those evaluations contributed to shape their evaluations of candidates’ qualities or to update their partisan affiliations. In legislative elections, those effects on the vote were strong and direct, resisting the introduction of controls for partisanship and leader effects. One potential objection to generalization of these findings beyond the context of the 2006 election would be to point out that, by the time of that election, the government was still enjoying a “honeymoon” period, thus depressing whatever effects government performance might have on the vote. However, the data fail to support this notion. In the 2006 survey, about 40% of respondents rated the performance of the Sócrates government as “bad” or “very bad”. Furthermore, the prime minister’s approval rating, as measured by the monthly Marktest barometer, shows a sharp decline throughout the first year of government: in April 2005, one month after the legislative election, the difference between positive and negative evaluations of Sócrates’ performance amounted to 43 percentage points; by February 2006, net support had dropped to mere 7 points. By the time of the presidential elections, then, whatever “honeymoon” may have occurred was most certainly over. 16. Another potential explanation of defections among former Socialist voters could be “regime voting” (Gschwend and Leuffen, 2005 Gschwend, T. and Leuffen, D. 2005. Divided we stand – unified we govern? Cohabitation and regime voting in the 2002 French elections. British Journal of Political Science, 35: 691–712. [Crossref], [Web of Science ®] , [Google Scholar]), particularly the possibility that such voters might have a preference for cohabitation regardless of their ideological views. In our survey, however, we lacked measures of such preferences, and were unable to test this interesting hypothesis. 17. It could be argued that the size of this “surge” turned the 2005 election into such an exceptional event in Portugal that no generalization can be made on the basis of the 2005 case. As it happens, however, we have reasons to believe this “surge” was not “exceptional” at all. Of all eight legislative elections conducted from 1983 to 2005, only in the 1983 and 1985 cases has the winning party (invariably, the PS or the PSD) obtained less than 40% of the valid vote.