1The French presidential election that took place in April and May 2012 was associated with major financial crisis. [1] With economic growth at zero and unemployment hitting a symbolic 10% during the first quarter of 2012, outgoing president Nicolas Sarkozy appeared to be fighting a losing election battle. Far from downplaying the historic nature of the situation, Sarkozy largely sensationalised it. In the months leading up to the election, he made a series of announcements that sought to increase taxes and reduce social gains and public services – despite having in fact reduced compulsory levies at the very beginning of his mandate. [2] Although such volte-faces are not rare in politics, they usually follow an inverse logic. The traditional political and economic cycle generally entails an initial drop in the public deficit at the beginning of a president’s term, followed by a rise in public spending at the end of the cycle, designed to boost a candidate’s chances of being re-elected. [3] Given the situation, Sarkozy’s defeat in the presidential election remained moderate, albeit undeniable. Despite an unprecedented high score for the Front National candidate Marine Le Pen in the first round (17.9%), Sarkozy managed to qualify for the second round of the election, comfortably in the lead by more than 3 million votes (27.2%). In the second round, with 48.4% of the vote, Sarkozy obtained a marginally better score than Valéry Giscard d’Estaing did in 1981 (48.2%), even though the latter benefited from a slightly more favourable context.

2This analysis is supported both by a European comparison and by considering certain specific elements of the 2012 French presidential election. From a European point of view, alternating changeovers in government between two major parties has become the norm since 2010. Between 2010 and 2012, sixteen out of nineteen member countries of the European Union which held general elections witnessed a power shift. [4] From the French perspective, Nicolas Sarkozy’s chances at re-election seemed unlikely even as early as 2008. The first polls taken that year already showed that a majority of French voters were dissatisfied with his performance. The president’s approval rate dropped to 33% in July and did not significantly climb back up thereafter, [5] despite the fact that the full effects of the financial crisis had not then really been felt in France. In short, Sarkozy managed to lose by only a small margin even though all factors were pointing to a massive defeat. How can this relative performance be explained?

3The goal of this article is to offer a partial answer to the question above, by assessing the electoral impact of the economic and financial crisis in 2012. Was the crisis’s impact ultimately only marginal? Did it benefit the parties in government or the more fringe minority candidates? To answer these questions, the primary empirical source used is the 2012 FES survey, conducted via questionnaire after the second round of the presidential election on a representative sample of 2,000 French citizens living in mainland France and registered to vote. [6]

4Using this data, this article will test two main hypotheses which stem from recent developments in scholarly research on economic voting. The first hypothesis is that the deterioration of economic conditions had a direct impact on voting decisions. Our second, more complex, hypothesis is that the potential effect of the financial crisis was in reality mediatised by the debate on the direction of economic policy. In other words, we shall observe the respective impact of economic valence voting and economic positional voting on electoral behaviour in 2012. These two hypotheses are described in depth in the following section before being put to the test.

Economics and voting in crisis situations

5Over the past few decades, numerous studies on the impact of economic conditions on voting patterns have been conducted in the field of political science. Although many questions still remained unanswered, one conclusion has consistently emerged emerged when tested in different contexts and according to different specifications: voters “punish” an outgoing government when the national economic context has significantly deteriorated. Many voting prediction models have thus been proposed, most of them identifying economic growth or fluctuating unemployment rates as the crucial elements informing anticipated electoral results. Arising from the analysis of electoral behaviour which flourished during the 1990s and 2000s (in 2007, Lewis Beck and Stegmaier referenced more than 400 recent works on the question), [7] analysing “economic voting” has become indispensable ever since the crisis that occurred at the end of the 2000s.

6Although literature on economic voting generally adopts an instrumental view of electoral behaviour, it is nevertheless divided as regards both approaches and conclusions. [8] Its baseline remains the study, at the aggregate level, of the relationship between economic conditions, the popularity of the government, and election results. The dominant theory [9] paints the picture of an uninformed voter, largely unaware of macroeconomic issues and, more importantly, short-sighted – unable to remember the past or anticipate the future more than a few months in either direction. Voters thus choose to reward the outgoing majority party, or sanction it (by voting for its main opponent), according to the party’s perceived ability to ensure a certain level of economic prosperity, measured in terms of GDP growth and limited levels of inflation. In other words, economic voting is first and foremost a sociotropic and retrospective phenomenon that bases an electoral choice for the future on an evaluation of the recent economic performance of the outgoing government. Based on this general model, a number of different variations have been elaborated. In particular, research has focused on the choice of sociotropic and egotropic approaches (to what extent is it actually changes in personal economic circumstances which are being taken into account?); on the connections between retrospective and prospective dimensions (are voters capable of assessing a government’s ability to ensure future economic prosperity, independent of past experiences?). More generally, research on economic voting has largely remained split between an approach built around aggregate data (how do economic circumstances influence election results?) and an approach which concentrates on polling data (how do perceptions of the economic situation influence voting choices?), even if certain contemporary studies using multi-level analyses have tried to combine both approaches. [10] This structural difference explains both the volume of research produced as well as the – sometimes vehement – critiques that appear within the discipline. The articles penned by Martin Paldam in 1991 [11] and Christopher J. Anderson in 2007 [12] are telling examples, as they reveal how an internal technical and epistemological critique has been inseparable from the development of this approach.

7The relative instability of economic voting estimates has been key within such critiques. One solution to this problem was found by paying greater attention to the nuances of institutional and contextual factors. The clarity of responsibility was the first aspect to be examined. [13] At this level, it is a matter of understanding to what extent specific actors can be attributed a particular political responsibility, in the case of a coalition government for instance, or in the context of a divided majority. We should note that from this perspective, multiple factors exist that affect the clarity of responsibility and which can at times contradict one other. We shall nevertheless try to differentiate between institutional aspects (shared responsibilities between different levels of government, for example) and more specifically political aspects (government cohesion, for instance [14]). Context itself can also have an impact on the exercise of political responsibility. But institutional and contextual effects are especially complex, as parties and candidates also help to create the arena where political judgements are made. Strategies used to blur lines, put issues on the political agenda, or frame the discussion also greatly disturb the notion of a retrospective vote as a rational judgement on an objective performance.

8This article situates itself in the margins of this classical approach to economic voting. It is evidently difficult to produce a reliable estimate of the impact of economic circumstances when using data from only one election, and especially without a panel dimension. More substantially, the choice of a case study is justified both by the desire to account for a more “complete” approach to economic voting, and the impossibility of measuring today’s economic crisis against other recent episodes of recession.

9Michael Lewis-Beck and his colleagues have recently called for the elaboration of a “complete theory” of economic voting. [15] Their main argument is that the conventional theory of economic voting completely overlooks two fundamental aspects of the economy’s impact on voting behaviour: preferences in terms of economic policy (what these authors call the positional aspect) and the economic status of individuals, especially in terms of property and assets. Preferences in terms of economic policy have in fact been observed since the very beginning of studies on economic voting. As Gerald H. Kramer has argued, [16] economic voting is a type of voting on public policies to the extent that voters react to macroeconomic consequences that were originally determined by macroeconomic policies. These preferences are excluded from analyses because of the allegedly limited information that individuals possess. But such a hypothesis then leads to a major obstacle for a theory of economic voting. If voters really do possess very limited information, how can their actual perception of national economic conditions be explained? [17] The reason behind this can likely be found in the media’s role in defining this perception, thus leading us to discuss as much the effect of the media, observed in few other areas, as the truly economic nature of economic voting. Hence the positional aspect should not in principle be excluded from a study of economic voting. The idea is not inherently new: D. Roderick Kiewiet mentioned it in 1983 in the field of economic voting, [18] and the values established around “economic liberalism” have long been associated with the study of electoral behaviour.

10The inclusion of assets and property in the definition of economic voting appears more questionable. This dimension does not initially appear to concern exactly the same level as the traditional definition of economic voting since it does not directly concern economic policies and their consequences – unless, ultimately, we consider that it is only a matter of re-evaluating the role of egotropic reasons for voting. But even if we concede the legitimacy of this conceptual expansion of economic voting, the role of personal property remains debatable. Even in terms of Lewis-Beck and his colleagues’ own logic, assets in fact seem to include three different aspects that are hard to untangle: assets define social status (by wealth) and economic status (independent status automatically increases the assets held by an individual), inheritance (and thus a form of socialisation), and a certain attitude to risk. Conversely, focusing only on assets when defining voter economic status seems equally prejudicial. Only the empirical observation of a closer correlation with electoral behaviour would justify it. In short, considering the economic status of voters seems to constitute more a contribution to the understanding of economic preferences than function as a direct component of the economic dimension of voting itself. The empirical results produced by Lewis-Beck and his colleagues appear to corroborate this hypothesis. [19]

11In recent studies on economic voting, a second important aspect has been examined: the effects of the contemporary economic crisis. The major economic crisis prompted by 2008’s financial crisis in fact represents a crucial contextual change. [20] Research has identified at least three major changes provoked (or at least emphasised) by this crisis.

12The first is the attention paid to the effects of economic circumstances over and above support for parties in government. Although the problems associated with applying a model conceived in continental European majoritarian democracies have been long recognised, the rise of “protest” or far-right parties – notably in Greece and Italy – has led researchers more systematically to include the study of economic fallout on party systems as a whole. Following the same logic, Hans-Peter Kriesi has proposed extending the logic of economic voting to the whole of political behaviour, in order to better emphasise the diversity of protest reactions to the crisis. [21]

13In general, economic crises are also supposed to strengthen the salience of economic questions on the political agenda, [22] as, according to the thermostatic approach, the importance of a change is exponentially linked to its scope. But the economic crisis that began in 2008 may have affected not only the hierarchy of issues on the agenda, but also the very nature of those issues. The current economic crisis cannot in reality be interpreted as simply the classic coupling of a recession and a rise in unemployment. It has also helped to make two related questions more concrete: the sovereign debt crisis and the interdependence of economic policy at the European level. The sovereign debt crisis and the related question of bank rescue operations run by individual states have attracted an unprecedented amount of attention in recent times, although this is not entirely a new phenomenon. The conflict between budgetary orthodoxy and pro-growth policy was one of the cornerstones of the French presidential election in 2012, with an outgoing president who blasted the Socialist Party’s refusal to approve the constitutionalisation of a budgetary “golden rule” when the socialist candidate was making the renegotiation of European agreements to incorporate a “growth policy” one of the central planks of his campaign. We can thus see how, from this perspective, the economic crisis may not only have affected the importance within electoral decisions of judgements regarding the performance of economic policies, but also helped to reframe the very meaning of economic questions within the campaign.

14The final identified impact of the economic crisis challenges the very relationship between macro-economic policies and economic fluctuations. From this point of view, the crisis had two contradictory effects. First of all, it helped to emphasise the limits of state economic governance; it underscored these limits of state intervention both as a result of financial constraints and also because of the limits imposed by European or international agreements. Second, it also proved the limited effectiveness of state action, which is seemingly unable to steer the economy beyond the partial bailout of certain sectors about to collapse. Diminishing state power had already been observed before the crisis, however, brought about as much by the end of an era of post-war economic expansion as by the dynamics of integration and interdependence produced by European integration and globalisation. Considered in this longer perspective, the crisis may have also been seen as a renewal of state intervention, either directly through a return to the nationalisation of businesses or, more indirectly, through new instruments implemented at the European level. The 2012 presidential campaigns were thus marked by the omnipresence of Europe, whether it was Nicolas Sarkozy’s strengthening of the Franco-German partnership as a solution to the crisis, or François Hollande’s reorienting European integration as an equally effective solution. We should also note that the European Union – cited as the reason for the economic crisis – was likewise omnipresent in the rhetoric of Marine Le Pen and Jean-Luc Mélenchon, despite their radically opposing positions.

15Based on these different suggestions outlined by the literature on economic voting, the rest of this article will be devoted to analysing the impact of the economic crisis on the 2012 French presidential election, first by examining the economy as a valence issue, before then treating it as a positional issue.

Context, macro-economic policy performance, and the vote in 2012

16The impact of the economic crisis on the French electorate in 2012 is easily discernable. When voters answered an open-ended question on the most important problems in France, for 78% of voters, their first response focused on economic issues: [23]among and 6.9% about purchasing power. [24] As a corollary, during the twelve months leading up to the election, belief that the French economy had improved was shared by only 3.4% of the sample population: on the contrary, 80.9% believed that it had suffered. For reference: the second quarter of 2012 witnessed an effective 0.2% decrease in GDP (0.1% GDP growth during the previous twelve months), while unemployment reached 10.2% at the end of this period, having increased 0.6% over the previous twelve months. [25] We should also note here that these circumstantial elements do not single-handedly, however, explain the specific nature of the economic crisis which began in 2008, as these indicators were even worse at the end of the 1990s.

17The perception of the evolution of the economic situation serves here as a preliminary estimate of the impact of the crisis. Two dependent variables are used: the first-round vote between the five major candidates (from left to right: Jean-Luc Mélenchon, François Hollande, François Bayrou, Nicolas Sarkozy, Marine Le Pen) and the second-round vote between Hollande and Sarkozy. [26]

18As in the majority of empirical analyses of individual voting decisions, [27] the empirical approach used here does not rely on a traditional regression model with an ordinary least square estimation method. Here we shall consider the voting variable as a discrete variable; [28] in other words, all the candidates cannot be ranked in a homogenous fashion for all voters along a single dimension. In this case, French political science usually recommends using logistical estimation models: that is to say, models based on the estimation of the relative probability of belonging to one group rather than another. Instead of a logistical model, we have here opted for a probit model, very similar to logistical models save for hypotheses regarding the shape of the distribution of probabilities for the dependent variable. [29] This choice was based less on a real difference in the quality of estimation than on the subsequent development of this analysis, which relies on the new specifications of probit models. Interpreting the results of a probit regression is similar to interpreting the results of an ordinary least squares regression, but the coefficients must be viewed as representing the marginal effect on the probability of a positive result of a unit’s increase on the cumulative scale of a normal distribution of probabilities. In simpler terms: a negative coefficient is associated with a negative effect of the explicative variable, a positive coefficient with a positive effect and a zero coefficient with no effect.

19Since the dependent variable has more than two modalities, a multinomial probit regression model was used for votes cast in the first round, while a simple probit regression model was used for the second round. Traditional control variables were added to this model: age, gender, education level (9 points), professional occupation, income (10 points); and for socio-demographic traits, religion and religious practices, political awareness (based on a series of four political questions [30]) and, finally, position on the left/right political spectrum. This article focuses on ideological positioning rather than political affiliation, since it appeared more stable than political preference and was thus less likely to maintain a reverse causal relationship with electoral choice.

20The results of these regression analyses are presented in Table 1. Unsurprisingly, and in keeping with traditional analyses of electoral behaviour in France, position on the political spectrum and religion remained the most important variables for understanding voting choices in 2012. Those who voted for François Bayrou, Nicolas Sarkozy or Marine Le Pen largely came from the right end of the political spectrum. Those who voted for Bayrou or Sarkozy also tended to be practising Catholics and independent professionals. Nicolas Sarkozy’s electorate was mostly female, while Mélenchon and Le Pen’s electorates were both significantly younger. The important point of these analyses nevertheless remains the notable impact of sociotropic and retrospective judgement on the state of the economy. Judging the economy more positively than the average voter significantly increased an individual’s chances of voting for the outgoing Sarkozy. More precisely, assessing the economy more positively on a 5-point scale increased an individual’s chances of voting for Sarkozy by 40%. We should note that the coefficients associated with this variable are only significant for votes cast in favour of Sarkozy. This in fact means that economic voting, defined from a sociotropic and retrospective point of view, primarily affects support or disapproval for the outgoing incumbent. From this point of view, outlook on the economy did not influence votes between the outgoing president’s different adversaries.

Table 1

Outlook on the French economy and voting in 2012 (probit models)^*,**,***

First round (ref: Hollande) Second round (ref: Hollande) Mélenchon Bayrou Sarkozy Le Pen Sarkozy Socio-retro judgement Age Gender Education Profession: Farmer/self-employed Senior exec./Liberal prof. Mid-level occupation Clerical workers Manual workers Income Religion: Pract. Cath. Non-pract. Cath. Other religion No religion Left/right Political awareness -0.05 (0.08) -0.01 (0.00) *** 0.14 (0.14) -0.08 (0.03) ** -0.82 (0.41) ** 0.17 (0.19) -0.10 (0.22) -0.15 (0.17) - 0.00 (0.03) -0.88 (0.27) *** -0.48 (0.15) *** -0.75 (0.25) *** - -0.06 (0.03) * 0.04 (0.04) 0.13 (0.09) -0.01 (0.00) * -0.14 (0.16) -0.02 (0.04) 0.46 (0.29) -0.02 (0.23) 0.20 (0.24) 0.22 (0.19) - 0.04 (0.03) 0.68 (0.24) *** 0.10 (0.19) -0.05 (0.29) - 0.41 (0.04) *** -0.01 (0.04) 0.27 (0.09) *** 0.01 (0.00) * -0.26 (0.15) * -0.06 (0.04) 0.52 (0.27) * 0.17 (0.22) 0.14 (0.25) -0.11 (0.19) - 0.06 (0.03) * 0.68 (0.24) *** 0.41 (0.20) ** -0.37 (0.32) - 0.83 (0.04) *** -0.05 (0.04) -0.09 (0.09) -0.01 (0.00) ** 0.15 (0.15) -0.18 (0.04) *** -0.19 (0.29) -0.33 (0.24) -0.21 (0.23) -0.04 (0.17) - 0.02 (0.03) -0.08 (0.23) -0.05 (0.17) *** -1.35 (0.32) - 0.63 (0.04) *** -0.17 (0.03) *** 0.18 (0.06) *** 0.00 (0.00) -0.16 (0.10) -0.04 (0.02) 0.27 (0.18) 0.10 (0.15) 0.19 (0.16) -0.00 (0.12) - 0.03 (0.02) 0.36 (0.15) ** 0.24 (0.12) * -0.24 (0.19) - 0.61 (0.03) *** -0.01 (0.03) R2 0.302 0.511 N 1433 1455

Outlook on the French economy and voting in 2012 (probit models)^*,**,***

Note: standard deviations are noted in parentheses.
* p < 0.1;
** p < 0.05;
*** p < 0.01.

21These results thus seem to confirm standard economic voting theory, illustrating the importance of the perception of economic circumstances with regard to deciding to support (or not) the incumbent’s re-election. This outcome is also confirmed if we use the egotropic approach, where an individual assesses his or her quality of life during the last twelve months. This variable is likewise significant with regard to second-round choices (to a threshold of 5%), but the value of the coefficient is lower and the standard error is higher. Moreover, this variable is not significant for the first round of the election. However, two limits to these results emerge as equally important. The explanatory power of an individual’s perception of the economic situation remains limited, while the assessment of its impact remains subject to the risk of endogeneity.

22Perception of the economic landscape in fact only marginally helps to explain the results: although this variable has an excellent level of statistical significance in our models, it in fact does not provide much in the way of a justification for the existence of different electoral behaviours. For example, omitting this variable in one model or another only marginally affects R² (that is, the degree of variance explained by the model), with only 0.75 points less for the first round and 0.6 points less for the second round. One important reason for this can be found in the limited variability of this indicator. [31] More than 80% of the sample population selected two out of the five modalities for this variable. One solution to this problem could be found by adding an alternative variable: the respondent’s satisfaction with how Nicolas Sarkozy dealt with the economic crisis. 38% of the sample population stated that they were satisfied (out of two modalities) and 57% stated that they were dissatisfied (also out of two modalities), thus illustrating a more even distribution. If we integrate this variable into the previous model, only three variables remain significant: gender, position on the political spectrum, and satisfaction with regard to Sarkozy’s handling of the economic crisis. The quality of the model’s estimation consequently increases significantly, with an R² augmented by almost eight points (to 0.605). But adding in this measure in fact reveals that the main problem is the endogeneity linked to such an approach.

23The endogeneity of independent variables in a regression model is a common problem for these types of estimations. For example, it occurs in cases of reverse causality, when the dependent variable determines the independent variable. Technically, endogeneity is defined by the fact that the explanatory variable being considered is correlated with the error term of the regression model. In our case, the endogeneity of an individual’s perception of the economic situation could stem from the fact that political preferences in fact explain this perception, or at least the statement made regarding the matter. Advocates for “objective” measures and believers in perception measures regularly debate this subject in research on economic voting. [32] In this case, we can observe significant correlation, for example between satisfaction expressed regarding Sarkozy’s economic performance and being favourably disposed towards this candidate; likewise, the preceding results clearly illustrate the correlation between perception of the economic situation and judgement vis-à-vis Sarkozy’s performance.

24Despite the frequent occurrence of endogeneity, there are few easy ways to address the problem. In this case, the simplest solution would be to base estimates on objective data about the economic situation. However, I did not adopt this approach, for both methodological and substantive reasons. Substantively, the theory of economic voting presupposes that it is indeed perception of the economy and not the reality of the situation that influences voting. Postulating the homogeneity of this perception seems to involve as much potential bias as the direct use of perceptions. Whatever the truth of this, using situational data assumes a comparative design, whether temporal or spatial, which also poses certain difficulties for the theory of homogeneity of effects throughout different contexts. As we are interested in a specific context, the direct use of situational data appears to lack pertinence.

25The solution that I therefore propose to address the issue of endogeneity is a technical one, derived from econometrics, and which is commonly called regression with instrumental variables. The principle is relatively simple: it consists in swapping out the endogenous variable for this variable’s prediction score, based on variables considered to be exogenous and defined as “instruments”. This strategy used to treat the problem of endogeneity faces two obstacles: identifying the instruments and, more importantly, the difficulty of treating categorical or ordinal variables.

26Identifying the instruments to use presupposes two conditions: having access to variables that are sufficiently exogenous with regard to our dependent variable (the vote), although these variables must also be linked to our variable of interest (perception of the economic situation). The solution offered here is to use local economic circumstances to this end. [33] The economic situation is indeed exogenous with regard to voting. We can likewise prove that it is statistically linked to perceptions of economic circumstances. To do so, we used two indicators of economic situation, available at the level of the département, in order to describe economic circumstances during the twelve months leading up to the election: fluctuation in departmental unemployment rates, as measured by Insee (variation measured by percentage point) and variation in corporate insolvencies between 2011 and 2012, as measured by the Banque de France (variation measured by percentage). These two variables are only moderately correlated (0.27). Table 2 shows that they did in fact have a significant statistical impact, despite the inclusion of classic control variables (excluding the non-pertinent variables [34]). The probit model of analysis was retained, although this time it was ordered (since it can reasonably be argued that the importance of the economy’s improvement or its deterioration refers to a hierarchy understood in similar fashion by the whole of the sample). The statistical significance of the estimations is corrected for the existence of multiple observations in each département.

27The second difficulty with using instrumental variables stems from the complexity of integrating ordinal or categorical variables into this kind of model, at the level of both the dependent variable and the independent exogenous variables. For the second round of the presidential election, an intermediate solution consists of applying the instrumental variables technique to a probit model, since the choice in question is a binary one.

Table 2

The impact of local economic conditions on the subjective evaluation of the economy^*,**,***

Ordered probit model Change in unemployment rate Change in corporate insolvency rate Age Gender Education Occupation Farmer/self-employed Senior exec./Liberal professions Mid-level occupation Clerical workers Labourers Income Left/right Political awareness 0.45 (0.10) 0.06 (0.02) -0.00 (0.00) -0.13 (0.06) -0.01 (0.01) 0.05 (0.10) -0.04 (0.08) 0.01 (0.09) -0.03 (0.07) - -0.01 (0.01) -0.08 (0.01) 0.02 (0.01) *** ** ** *** R2 0.25 N 1753

The impact of local economic conditions on the subjective evaluation of the economy^*,**,***

Note: The standard deviations are indicated in parentheses.
* p < 0.01;
** p < 0.05;
*** p < 0.01.

28Although the first-round decision was taken into consideration, no simple solution exists for voting variables. One option would be to follow Wouter van der Brug’s solution, [35] in other words to understand the impact of the economic conditions not on votes cast but on the propensity to vote for the different candidates in competition. This solution is not employed here because of the numerous technical difficulties that it raises and, moreover, because it complicates any comparisons with previous results. In a slightly less elegant fashion, I thus propose only to approximate the results of a multinomial regression by first contrasting the decision to vote for Sarkozy with the decision to vote for Hollande in the first round; and second, with the decision to vote for any other candidate. The results of these analyses are presented in Table 3.

Table 3

Voting models for the outgoing president in 2012, with economic perception instrumentalised through local economic circumstances (probit models)^*,**,***

First Round Second Round Sarkozy vs. Hollande Sarkozy vs. others Ref: Hollande Sarkozy Socio-retro judgement Age Gender Education Occupation: Farmer/self-employed Senior exec./Liberal professions Mid-level occupation Clerical workers Labourers Income Religion: Practising Catholic Non-pract. Catholic Other Religion No religion Left/right Political awareness 0.38 (0.61) -0.00 (0.00) 0.15 (0.17) -0.04 (0.04) 0.43 (0.24) -0.05 (0.20) 0.11 (0.17) -0.10 (0.19) - 0.06 (0.04) 0.47 (0.20) 0.21 (0.18) -0.44 (0.32) - 0.76 (0.16) -0.03 (0.03) * ** *** 0.43 (0.39) -0.01 (0.00) 0.14 (0.11) 0.01 (0.02) 0.39 (0.15) 0.23 (0.13) 0.17 (0.15) -0.02 (0.13) - 0.04 (0.02) 0.59 (0.15) 0.40 (0.14) 0.26 (0.18) - 0.36 (0.04) -0.01 (0.02) *** *** * * *** *** *** 0.60 (0.31) -0.00 (0.00) 0.06 (0.13) -0.02 (0.02) 0.18 (0.18) 0.08 (0.13) 0.16 (0.11) 0.01 (0.15) - 0.03 (0.02) 0.38 (0.12) 0.25 (0.10) -0.11 (0.19) - 0.54 (0.08) -0.02 (0.01) ** ** ** *** LogP. Lik N -1202.7 863 -2155.8 1433 2137.2 1455

Voting models for the outgoing president in 2012, with economic perception instrumentalised through local economic circumstances (probit models)^*,**,***

Note: Standard deviations are indicated in parentheses. They are robust and corrected for the departmental clusters.
* p < 0.1;
** p < 0.05;
*** p < 0.01.

29We can draw several contrasting conclusions from Table 3. For the second round of the presidential election, the instrumental variables technique allowed us to confirm the weight of perceptions of economic conditions with regard to voting choices. The coefficient associated with this variable remains significant and positive. On the other hand, for the first round this variable loses all significance, regardless of whether votes cast for Sarkozy are compared with the totality of other votes or only with those cast for Hollande. Going further, I compared candidates in pairs to verify the results. The perception of the economy was significant in only one case, albeit majorly: when comparing votes for Sarkozy with votes for Bayrou. The associated coefficient then reached -1.14 (with a standard deviation of 0.35), the variable being significant at 0.001% despite a smaller sample size for this analysis (445). The votes cast against Sarkozy in favour of Bayrou nonetheless remain limited in number. Let us recall that, unlike in 2007 when Bayrou managed an important electoral break-through, this was not the case in 2012, when he remained at the average level for the centrist electorate in recent years. [36]

30Together, these analyses of the impact of the economic situation on voting behaviour in 2012 thus largely confirm the findings of literature on economic voting. The perception of a deteriorating economic situation leads to a form of electoral sanction for the outgoing president; the effect appears robust for all the different statistical specifications. This electoral sanction ultimately remains limited in scope, however. In the first round, perception of the economy primarily affected the voting decisions of individuals on the centre or centre-right, who were hesitating between Sarkozy and Bayrou; this nuances one of the traditional postulates of existing research, which presumes an impact principally focused on the outgoing candidate and his or her main opponent. In the second round, the impact of economic perception was less pronounced in its scope but still significant, this time in the choice made between Sarkozy and Hollande. It appears evident that the latter owes his victory at least in part to fringe members of the centrist electorate who, having preferred Bayrou to Sarkozy in the first round, ultimately voted for Hollande in the second – less because of any real affinity with the left than because of dissatisfaction with current economic circumstances. Paradoxically, however, the crisis probably limited the effects of economic perception, as ultimately, the belief in a deteriorating situation was shared almost unanimously by the whole of the electorate.

Economic positions and voting in 2012

31To complement our understanding of the effects of the economic crisis in 2012, I propose exploring the positional dimension of economic voting. Several recent studies [37] have illustrated not only the existence, but also the continuity, of French voters’ attitudes with regard to economic matters. The main conclusions of these studies fall into three camps. Attitudes regarding economic issues are structured along two major axes: interventionism (or liberalism) and egalitarianism (or a preference for redistribution). Intervention appears to be closely linked to the left-right dimension, while egalitarianism is only moderately linked. [38] The structure of these axes remains relatively stable over time, but the levels recorded along both of these axes have increased over time towards demands for more intervention and redistribution. The 2008 economic crisis did not in fact signal a marked rupture from these mid-term tendencies. The objective here is to observe how, for the data regarding 2012, these relationships remained stable and to what extent another major economic issue on the political agenda – public debt and deficit – contributed to these axes or represented a specific issue for this election.

32The study used here combines a limited number of economic attitude indicators and economic public policy preference indicators. Given the possibility of an endogeneity bias, I also chose not to use those questions which referred specifically to one candidate or another. Eight indicators remained: two directly pertaining to the issue of redistribution, four to economic liberalism, one to public deficit and one to the desired levels of taxation, which remained perpendicular to these axes.

33For the redistribution axis, the two indicators were agreement expressed with the following statement: “Government should take measures to reduce income gaps”, as well as a stated preference between four increasingly redistributive tax regimes. [39] Globally, each of these two variables indicates a high level of preference for redistribution (the most redistributive choice was selected by 45.7% and 52.1% of our sample, respectively). The combination of these two variables allows for the creation of a simple scale with more variance than either of the measures used separately. This scale was in turn reduced to six points in order to group together the least redistributive attitudes and conserve sufficient numbers in each category.

34For the economic liberalism axis, I used four classic indicators (positive or negative connotations for the words privatisation, nationalisation and profit, and the choice between favouring economic competitiveness or improving working conditions). Strong statistical links between these indicators allowed the construction of a scale of liberalism, likewise reduced to six points in order to group together the most and least liberal attitudes. [40]

35The last two variables concerned opinions regarding public deficit [41] and taxation levels. [42] With regard to public deficits, 95% of the sample was more or less equally split between the two opinions favourable to balancing public accounts, making this variable essentially dichotomous. For taxation levels, 61.1% of the sample preferred to maintain the current tax levels, 30.6% supported decreasing them and only 8.2% argued for increasing them. The significant point remains that these two variables do not seem to be statistically linked with each other or with the axes mentioned above. We shall thus consider them separately.

36If we now examine the link between these four indicators and position on the left/right political spectrum, it appears that the axis with the closest relationship to this position is economic liberalism, followed by a preference for redistribution. Opinion regarding public deficits is weakly linked statistically to political position, and preference regarding tax levels is not linked at all. In all cases, none of the variables presented a high enough risk of collinearity with this scale. In order to analyse the impact of these different indicators on the vote in 2012, we shall proceed as above. We shall insert these variables into a probit or multinomial probit model depending on whether the focus is on the first or second round of the election. The same control variables will be used. For the sake of simplicity, we shall not include here the elements seen previously on the perception of economic circumstances. We should note that, unsurprisingly, a complete model does not significantly affect the coefficients observed, but slightly decreases statistical significance due to technical and non-substantive reasons. The results of the analysis are presented in Table 4.

Table 4

Model of voting choices in 2012 on the basis of economic variables^*,**,***

First Round (ref. Hollande) Second round (ref. Hollande) Mélenchon Bayrou Sarkozy Le Pen Sarkozy Redistribution Liberalism Deficit Taxes: Taxes - Taxes = Taxes + Age Gender Education Occupation: Farmer/self-employed Senior exec./Liberal professions Mid-level occupation Clerical workers Labourers Income Religion: Practising Catholic Non-pract. Catholic Other religion No religion Left/right Political awareness -0.06 (0.07) -0.21 (0.06) *** -0.03 (0.14) - -0.08 (0.18) 0.13 (0.27) 0.01 (0.00) *** 0.12 (0.17) -0.11 (0.04) ** -1.02 (0.60) ** 0.35 (0.23) -0.23 (0.27) -0.13 (0.21) - 0.03 (0.04) -.075 (0.32) ** -0.34 (0.18) ** -0.60 (0.30) ** - -0.01 (0.04) 0.03 (0.04) -0.29 (0.07) *** -0.01 (0.06) -0.26 (0.16) - 0.30 (0.21) 0.26 (0.34) 0.01 (0.00) * -0.23 (0.09) -0.04 (0.05) 0.47 (0.34) 0.02 (0.29) -0.13 (0.31) 0.18 (0.24) - -0.00 (0.04) 0.85 (0.29) *** 0.24 (0.24) -0.20 (0.38) - 0.37 (0.05) *** 0.02 (0.05) -0.29 (0.07) *** 0.33 (0.07) *** -0.24 (0.16) - -0.23 (0.20) -0.04 (0.36) -0.01 (0.00) ** -0.31 (0.19) -0.08 (0.05) * 0.24 (0.33) 0.04 (0.09) -0.14 (0.30) -0.19 (0.23) - 0.05 (0.04) 0.98 (0.31) *** 0.73 (0.26) *** -0.48 (0.42) 0.79 (0.05) *** -0.11 (0.05) ** -0.23 (0.06) 0.07 (0.06) -0.08 (0.14) - -0.41 (0.18) ** -0.43 (0.34) 0.01 (0.00) ** 0.05 (0.18) -0.20 (0.05) *** -0.39 (0.35) -0.25 (0.29) -0.17 (0.27) -0.10 (0.21) - -0.00 (0.04) -0.13 (0.28) 0.06 (0.21) -1.42 (0.39) *** - 0.59 (0.04) *** -0.15 (0.04) *** -0.16 (0.04) *** 0.23 (0.04) *** -0.18 (0.10) * - -0.04 (0.13) -0.00 (0.00) -0.00 (0.00) -0.31 (0.13) ** -0.05 (0.03) * 0.11 (0.21) 0.02 (0.19) 0.01 (0.20) -0.16 (0.16) - 0.01 (0.03) 0.23 (0.19) 0.27 (0.16) * -0.60 (0.27) ** - 0.60 (0.04) *** -0.02 (0.03) R2 N 0.43 1227 0.56 1239

Model of voting choices in 2012 on the basis of economic variables^*,**,***

Note: The standard deviations are indicated in parentheses.
* p < 0.1;
** p < 0.05;
*** p < 0.01.

37The results presented in the table above essentially confirm previous analyses. The major characteristics of each electorate remained stable in relation to the results presented in Table 1. Even the impact of position on the left/right spectrum remained in an equivalent order of magnitude, the only important change being the loss of statistical significance of this variable when comparing voting probabilities between Jean-Luc Mélenchon and François Hollande. For our variables of interest, the results were once again mixed. A preference for redistribution was significantly associated with a vote cast for a right-wing candidate, rather than a left-wing candidate; the impact of a weaker preference for redistribution was identical to the probability of voting for François Bayrou, Nicolas Sarkozy or Marine Le Pen, when compared to that of voting for François Hollande or Jean-Luc Mélenchon. Opinions in terms of economic liberalism were likewise structured along the left/right spectrum, albeit in a significantly different manner than above. Voters for Jean-Luc Mélenchon were much less liberal than those for François Hollande, while Sarkozy supporters were much more economically liberal. On the other hand, however, there was no significant difference between the electorates of Hollande, Bayrou and Le Pen with regard to this matter. Finally, questions regarding public deficit and taxes only allow us to differentiate one candidate at a time. Bayrou’s voters were more likely to believe in the need for limiting public deficits than others; Le Pen’s supporters were more likely to call for a reduction in tax rates. In the second round, redistribution and liberalism played out in the expected fashion: belief in the absolute necessity of limiting public deficits increased the probability of voting for Sarkozy. Opinion regarding the tax regime no longer had a significant impact.

38In terms of explanatory power, then, economic attitude appears to be more heuristic than perception of the economy, with nonetheless some important variations depending on the variable considered. Table 5 thus presents an overview of the evolution of the log-likelihood of the previous models by subtracting our variables of interest, one by one. As a reminder: the log-likelihood for the models that only included the control variables was -1541.78 and -472.40 for the first and second rounds, respectively. We can quickly see that economic liberalism provides additional information, highly significant for our models, while the contribution of opinion regarding taxation levels is much more limited. With regard to this variable, a preference for redistribution and opinion regarding public deficits were slightly more important for explaining the vote.

Table 5

Changes in the log-likelihood of voting explanation models, without economic attitudes

First round Second round General model Model without redistribution Model without liberalism Model without deficit Model without taxes -1030.12 -1079.58 -1310.12 -1080.64 -1066.99 -297.34 -311.65 -403.26 -313.37 -306.02

Changes in the log-likelihood of voting explanation models, without economic attitudes

39The results presented above largely confirm the results of previous studies which showed that economic liberalism remains the most crucial element for explaining political choices. It nevertheless seems that the question of public deficits was, at least in 2012, an important “voting issue” for a section of the electorate. [43] This issue appears to have been favourable to François Bayrou in the first round and Nicolas Sarkozy in the second. Let us recall that Bayrou had expressed his support for Hollande in the second round due to his desire to support honesty rather than budgetary orthodoxy. It seems that part of the centrist electorate made the inverse choice, preferring Sarkozy to Hollande in the hopes that the former would be better able to limit budgetary deficits.

40* * *

41This article has attempted to analyse the relationship between preferences for economic policies, perceptions of the economy, and electoral behaviour during the 2012 French presidential election. It demonstrated that, from this point of view, this election did not represent a dramatic change in the relationship between voting and the economy. Economic liberalism is the primary factor that overwhelmingly explains voting decisions. This factor also remains the dimension closest to the left/right axis, even if it is not sufficient to explain it and cannot replace it. The context of the economic crisis can nevertheless be succinctly measured by the data available. The perception of the changes in the economy over the preceding year had never been so uniformly pessimistic. And yet this perception remained significant for electoral choices – the more pessimist voters being more likely to sanction the outgoing president – even after verifying the robustness of these results with regard to the issue of endogeneity. Moreover, the effect of perception of the economy was compensated by the existence of a voting issue regarding public deficits. In the second round, Nicolas Sarkozy benefited from public opinion which was once again largely consensual, and which expressed itself in favour of balancing public finances.

42The effect of perception of economic crises and the issue of public deficits seems first and foremost to be strongest within the centrist electorate. Perception of a highly deteriorated economic situation favoured votes for Bayrou rather than Sarkozy among centre and centre-right voters during the first round. And these are most certainly the same voters who, having made this choice in the first round, followed the same logic in the second and voted for François Hollande. At the same time, another section of the centrist electorate supported Bayrou in the first round in the hopes of obtaining greater budgetary orthodoxy. This time, however, their choice in the second round leaned towards Sarkozy. We are thus far from a vision of the economic crisis as the standard bearer of a radical transformation of the electoral order, or the vector of a protest movement. Of course, the economic crisis can have an impact through its repercussions on the economic situations of voters themselves. Let us recall, however, that 2012 was not an exceptional situation for France from this point of view in terms of unemployment rates over the course of the past 20 years (since low growth and massive unemployment have ultimately been the medium-term characteristics of the French economic model since 1980). Let us also remember that the perception of individual economic circumstances had no statistically measurable impact on the data. The great economic crisis of 2012 did not thus represent a breaking point for France, as it may have for Iceland, Greece or even Italy.