1In the social sciences, it is common to make propositions which have apparently equivalent plausibility. Maurice Halbwachs formulated the following proposition based on this particular quality of sociological discourse:
Because sociology often appears to discover truisms, justice is not always rendered it… However, in the domain of the plausible, one proposition is generally opposed by another which seems equally obvious. There is, therefore, just as much scientific merit in determining which of two plausible opinions corresponds to reality as there is in bringing to light an entirely new truth. This is exactly what crossing the boundary separating scientific knowledge from common knowledge is about. 
3When it comes to commentary on routine public opinion surveys, this may indeed be said to be the case. Several factors may be marshaled to explain or elucidate fluctuating popularity curves, and we will return to these factors over the course of the present article. Thus, among the plausible propositions relating to presidential popularity, we may say that popularity is strongly dependent on circumstances – notably economic circumstances – or rather on information provided by media coverage of these circumstances. This article aims to identify the underlying “reality”. Does presidential popularity depend more on the state of the economy than on the economy as the media presents it? We are faced here with two causal propositions whose plausibility makes the alternate statement appear to be a truism. And yet each hypothesis is accompanied by heavy consequences, whether these concern the authenticity of citizens’ individual preferences, the strategic direction taken by the president to steer public affairs, or the responsibility of the media in the political game of representative democracy.
4What exactly does “popularity” mean in this context? An October 1990 conference held by the Association française de science politique (AFSP) attempted an interdisciplinary approach, long and laborious, to broach the various perspectives surrounding the notion of popularity. The results were hardly conclusive; the notion of popularity ended up as airy and difficult to conceptualize. Even today, we are obliged to define “popularity” in the context of the questions which allow us to perceive the notion: that is, in relation to the indicators identified, at least in France, in nine empirical ‘barometer’ surveys. Two of these surveys are carried out by the Ifop (Institut français d’opinion publique, the French Institute of Public Opinion); and one each by the Sofres (Société française d’enquêtes par sondages), BVA, CSA (Conseil sondage analyses), Ipsos, LH2, and, most recently, by Viavoice since May 2007 and by Opinion Way since September 2007. These surveys are distinguished from one another not only by their field dates but also by the nature of the questions asked. Each survey defines popularity in a different manner, which they designate as satisfaction, confidence, approval, favorable/unfavorable judgment, and positive/negative opinion. They thus provide information on certain differing aspects of popularity: the Sofres’ survey has a clearly prospective orientation (that is, it measures respondents’ confidence in the president’s ability to solve France’s problems); while the Ifop’s surveys are more clearly retrospective – that is, one can only approve, or be satisfied or dissatisfied with a phenomenon or fact that, as linguists might have it, already appears complete. Similarly, the Sofres’ survey has a definite sociotropic orientation: its questions refer back to the collective.
5The problem we address in this article is clearly delineated by the relations between presidential popularity, economic circumstances, and the media coverage of these circumstances. It gestures towards a whole series of underlying questions regarding how popularity is constituted, and how the economic situation and the media impact this process.
Factors contributing to the construction of presidential popularity
6Not only does the pertinence of the question of presidential popularity provoke controversy,  but over the past two decades, the approach to this question has evolved. Regarding the controversy over opinion polls, we acknowledge some of the critiques that have been levelled against them.  One of the goals of this article is to put the results of these surveys into perspective by illustrating precisely how public opinion is a function of a dependent, contextual variable – information – which prefigures more profound questions regarding the authenticity of opinions and preferences.
Research in evolution
7During the AFSP conference mentioned above, one of the authors of the current article presented a review of the explanations current in 1990 of how popularity is constructed.  His review, essentially based on American research, brought to the fore three main theoretical frameworks. According to Shanto Iyengar, these are still pertinent today. 
8The first framework emphasized the role of “time” and insisted on the inevitable decline of presidential popularity in the pre-Reagan period. In his 1973 book War, Presidents, and Public Opinion, John Mueller had kickstarted a series of studies on presidential popularity after publishing a seminal 1970 article in the American Political Science Review.  For Mueller, this linear decline was a result of the “coalition of minorities” that presidential action never failed to bring out as it distanced itself from its initial supporters. This was the idea of power erosion, described as an almost mechanical effect of time, which James Stinson  describes in more psychological terms, using the idea of expectations created during the electoral campaign which eventually give rise to disillusionment. Popularity is high during the post-electoral “honeymoon” period and sinks during the second and third years, when presidential action “disillusions”. At the end of a term, popularity increases slightly as re-election campaigns revive supporter groups. We thus move from the honeymoon period, during which the president benefits from what George C. Edwards III calls a “positivity bias” and the public in a sense extends credit to him, to a period of ineluctable disillusionment. This reference to the negative effects of time has created the theory of the “No Win Presidency”.
9The second explanatory model was based on conjuncture, or, in Iyengar’s terms, on the course of events. Three types of conjuncture were principally examined: economic conjuncture, international crises, and, in the cases of Nixon and Clinton, scandals. The main economic variables under examination were unemployment and inflation. When these factors increased, presidential popularity invariably sank.  On the international level, John Mueller was the first to point out the phenomenon of “rallies round the flag”, which are specific, dramatic, and sharply focused on international events with clear presidential involvement. Two contrasting theories are used to explain the effects of these rallies behind the president. The first, set forth by Mueller, envisions the rallies as expressions of patriotic reflexes that conflate political identity with national identity. The second theory, from Richard Brody,  suggests that the rallies be seen as expressions of elite partisan leadership. In his model, when the opposition party remains silent, the rallying process may begin, and it comes to an end only when “domestic” critiques are once again voiced. In the case of political scandals – Watergate or the Lewinsky affair – it has been shown that public opinion reactions were much more dependent on evaluations of the performance of public policy (particularly economic policy) than on private conduct.  The third explanatory model recalls the hypothesis of effects deriving from information and communication – especially when these are manipulated by the president. Communication strategies ( “going public”) may be implemented in order to boost sinking curves;  media information may alter the public’s perception of a particular political situation.  The notion of “going public” was recently reformulated by Jeffrey Cohen as “going local”. 
10In 2000, at the Congress of the American Political Science Association (APSA), Paul Gronke and Brian Newman presented a review of research dealing with presidential approval.  According to the authors, research on the topic can be divided into three phases. The first covers Mueller’s work and its extensions into the 1970s. The second phase begins in the early 1980s with work by MacKuen, Hibbs, Ostrom and Simon on the relations between economy and popularity, extending the notion of event in order to explore the rallying phenomenon. The third phase appears in the late 1980s and early 1990s, and is characterized foremost by the development of individual models by MacKuen, Erikson and Stimson, or Peffley, Langley and Goidel, who use individual perceptions of the economic situation as well as opinions on economic crises. This third phase is also characterized by its interest in the role of the media. Brody in particular sees this as the central element of the construction of presidential popularity, and reinterprets via media coverage recurring events such as the honeymoon period or rallying events. For Brody, the frequency and the manner in which the media handle information is decisive. The same is true for John Zaller,  who ascribes to elite opinions and media interpretation a key role in his theory of opinon change via mechanisms of agenda-setting and priming. Various research projects, notably experimental ones, have supported this theory. The third characteristic of the research in this phase concerns the attention paid to the diversity of public opinion reactions, which vary over time and between social groups. Attention is also paid to the growing public expectations of the presidency, as well as to the changes in salience of different issues which may help to explain changes in evaluation criteria over time.  In ascribing more salience to certain issues, the media influence the public’s changing evaluation criteria. The diversity of public opinion reactions is also a function of social groups, distinguished according to their origin, gender, level of knowledge and political sophistication, and particularly by their political allegiance, as numerous studies demonstrate.
Presidential popularity and the economy
11Our intention here is not to take stock of the innumerable research projects that have dealt with the relation between economic conditions and political popularity, but simply to highlight several landmark projects in the field, both in France and abroad. As early as 1983, when George Edwards III published his book on presidential popularity,  American research on the subject was already considerable. The tendency to evaluate the economic situation in “sociotropic” terms – a more conclusive approach than the ‘egocentric’ one, according to Kinder and Kiewiet  – was already visible: in other words, evaluation on the collective level as opposed to on the level of the individual citizen and his pocketbook. Whether it is a question of unemployment or inflation, neither circumstance as experienced on a personal level affects presidential support. In the same way, the distinction between retrospective and prospective evaluation does not majorly influence popularity. However, there is a much greater correlation between presidential popularity and the perception of government action. In other words, citizens evaluate the president’s economic role more on the basis of his performance than on their own representation of the economic situation. In the following years, controversy raged over the question of the rationality of homo politicus, and, in 1992, MacKuen et al. wrote that “after decades of attention, we are little beyond introspection in understanding the processes by which citizens come to perceive economic movements”. 
12In France, in 1981, Pommerehne, Schneider and Lafay published a long article, presented as a summary of past analyses, on the interactions between economics and politics.  In a section devoted to the influence of economic and political conditions on the policy sector, functions of popularity are particularly seen as “the principal link between the economy, voter reactions, and government action”. The authors compare the effects of three economic variables (unemployment, inflation, and growth in real disposable income) for six democracies with representative government. The authors conclude that each of these variables has a significant effect on the government’s popularity, though unemployment always has the strongest (negative) impact. “Excepting France and Germany, an average 1% increase in unemployment corresponds to a popularity decrease of approximately 2%… 1% increased inflation corresponds to 1% decreased popularity.” In 1995, Christopher Anderson confirmed the lesser impact of economic variables in France as opposed to other countries.  André Bernard  notes that for studies dealing with conjuncture and voting, “in the majority of equations, it is popularity, not the vote, which is the dependent variable”, which underlines the relevance of this type of research for our own questions on the links between popularity and the economic climate, or the mediatized version of the latter. From the multitude of studies he compiles, Bernard concludes that “a given climate does not have a decisive impact on most voters’ choices”, and that the electoral impact of specific economic situations tends to vary over time and across countries, as Pacek and Radcliffe observed. 
13The widest-ranging study on presidential popularity under the Fifth Republic (from 1960-2003) was the work of an American scholar.  First and foremost, Richard S. Conley observes that the rare systematic studies of presidential popularity in France have mainly been focused on macroeconomic factors. Lecaillon, Lewis-Beck, Hibbs and Lafay  are early examples of this. Conley does a good job of highlighting the variety of factors in play in presidential popularity: the timing and effects of electoral cycles, economic conditions, cohabitations, strikes and rallying events. Indeed, two of the three initial factors find their way into this group: erosion of power and political, economic, social, or even international conjuncture combined with rallying events; although he does not consider the effects of the media. The general model is specified by the following equation:
14Popularity = cste +b1(months in office) + b2(transition) + b3(campaign period) + b4(legislative election) + b5(unemployment) + b6(inflation/change in CPI)+ b7(strikes) + b8(rally)+ b9(terrorism)+ b10(cohabitation) + e
16Economic variables appear to have had significant and powerful effects on the popularity of Valéry Giscard d’Estaing: inflation at its peak cost him nine points in the ratings. During the first fifteen months of his term of office, François Mitterrand dropped 7-8.5 points for the same reason. A one-point unemployment increase generally translates to a 1.4% drop in popularity. During the 1980s, unemployment became a structural problem and cost these two presidents a total of 14 points in the ratings. Because of unemployment rates, Mitterrand and Chirac were 10-12% less popular than De Gaulle and Pompidou. Structural unemployment is thus fundamental in determining the basis on which opinions of the Fifth Republic’s first presidents and their successors were formed. But between 1992 and Richard S. Conley’s 2006 assertions, much progress has been made on a third approach opened up by our question concerning how economic information affects the construction of presidential popularity: that is, the impact of the media.
Media information and popularity
17This research field developed rapidly from the 1990s, particularly in the United States under the influence of Richard Brody,  Shanto Iyengar,  and Benjamin Page.  Experimental research revealed the priming effects created when media attention focuses on a specific issue, leading the public to afford particular weight to that issue when evaluating presidential performance.  Initially applied to rallies, as in Krosnick and Kinder’s study of Irangate  or Krosnick and Brannon’s Gulf War study,  the notion of priming allowed Goidel et al.  to show that media coverage of economic phenomena lent these factors greater influence in G. H. Bush’s popularity decline, thus confirming Hetherington’s 1996 results (see Conclusion). Clearly the media emphasis on certain issues serves to weigh upon presidential evaluation criteria,  thus opening up important strategic perspectives for presidential control of the media.  This kind of control appears more likely for those issues “mastered” by the president in an “issue ownership” sense; in other words for which he has a political reputation for superior competence, or sectors where he enjoys higher credibility. George C. Edwards III  however remains generally skeptical of the American president’s capacity to influence public support for his own policies; Canes-Wrone,  on the other hand, theorizes that influencing power remains conditional.
18In the British context, from 1987 onwards, David Sanders’ research has been revealing. Sanders first examined the impact of the Falklands War on government popularity, later focusing on the relationship between support for a political party in power and its economic competency as perceived by the public. In this vein, he and Neil T. Gavin began to work together, studying the impact of economic information broadcast by British television between 1993 and 1994.  This particular research has considerably influenced the current article. Sanders and Gavin compare the respective impacts of the mediatized economy, personal retrospective perceptions of the economy, and the “objective” economy on perceptions of the government’s competency – a determining factor of support for the party in power. The results of Sanders and Gavin’s work reveal a weak correspondence between the “objective” and mediatized economies. However, the mediatized economy has an impact on (perceived) competence and, as a result, on government popularity. In 2004, Sanders and Gavin published an article about televised information, economic perceptions and British political preferences during the 1997-2001 period. Their work confirmed the lesser importance of the objective economy compared to the mediatized economy in terms of personal prospective evaluations. A link is thus established between the perceptions of a government’s economic competence, the personal evaluations of voters, and televised economic information.
19Iyengar and McGrady came to an appreciably different conclusion.  For them, televised information concerning unemployment reflects “objective” economic indicators, and both “objective” and “televisual” unemployment influence presidential popularity.
20What about in France, during Nicolas Sarkozy’s first three years as president of the Republic? There have, certainly, been analyses of the effect of circumstance on his popularity ratings.  However, these studies fail to answer the following question: is presidential popularity – based on confidence –more sensitive to the “true” economic climate or the economic climate as represented by the media?
The model and the observation mechanism
21To answer this question, we have developed a model to measure the economy’s influence on the French president’s popularity. We have chosen a certain number of indicators corresponding to the different variables used in order to devise a logistic regression model.
Modeling the economy’s influence on the popularity of the French president
22Drawing on earlier work, notably that by Sanders and Gavin, we have constructed an empirical model to observe the relationship between popularity, economic information, and the “objective” economy in the case of the presidency of Nicolas Sarkozy. Our hypothesis is that presidential popularity depends on the evaluation of governmental economic policy, which in itself is largely swayed by both the elected candidate’s electoral program and by the decisions made as president. The evaluation must be considered in terms of citizen concerns as well as of the economic perceptions of this same group. Taking into account Sanders, Gavin, and Iyengar’s results, we too hypothesize that these concerns are largely influenced by perceptions of the economy, by mediatized economic information, and by the “objective” economy. The model we have used may be represented by Figure 1 below and is a classic recursive (or interlocking) model. Each element influences the others, located further down the causal chain. For example, media treatment of the economy influences economic perceptions and concerns, but also support for economic policy and, ultimately, presidential popularity. Economic perceptions then influence economic concerns, then support for economic policy, and finally presidential popularity. The cycle goes on. It should be noted that the real economy is logically outside this model.
Model of the effects of the economy on presidential popularity
Model of the effects of the economy on presidential popularity
23However, we have introduced an important distinction into our model. In effect, within the economic factors (concerns, media treatment, and the real economy) we distinguish between purchasing power on the one hand and unemployment on the other. This distinction allows us to take into account the key question for the 2007 presidential electoral campaign, and the impact of the economic crisis in the second half of 2008. Moreover, this is an example of a classic macroeconomics arbitration mechanism (a Phillips curve) between inflation and unemployment.  This curve is the basis for all models of politico-economic cycles in political economy. Empirical studies  dealing with voting and popularity showed that the two macroeconomic evolutions most likely to influence these two functions were the general level of prices and unemployment.
Presidential popularity and support for economic policy
24To evaluate the empirical pertinence of our model, we use separate measures of presidential popularity, popularity of economic policy, of economic concerns, perceptions of the economy, and information on economic changes. 
27The equation gives the ratio of positive opinions (confident responders) to negative opinions (distrustful responders) among the interviewees. The percentage of non-responders is very low, varying between 2% and 5% for the president, whereas the same figure varies between 4% and 10% for the prime minister. (The latter generally enters into office with an unknown reputation and projects which are as yet unclear.) This particular popularity measurement neutralizes the “don’t know” (dk) category, which can be relatively significant at the beginning of a term but decreases gradually as the term progresses. Indirectly, the “dk” figures are a way to measure dwindling “honeymoon” effects.  The pivot value for this measurement is 1. Below 1, popularity is considered negative (there are more negative than positive opinions) and, symmetrically, popularity is considered positive over the value of 1. This remains the case regardless of changes in the “dk” category. Figure 2 illustrates the evolution in popularity measures for the president and the prime minister.
Evolution of popularity indices for the president and prime minister (June 2007-May 2010)*
Evolution of popularity indices for the president and prime minister (June 2007-May 2010)** Note: The vertical axis shows the ratio of positive and negative opinions for both executives.
28A graphic reading of the comparative evolution in popularity for the two heads of the executive clearly shows that overall their popularity remains at similar levels. This is the case despite the prime minister’s domination, which remains unflagging from the beginning of 2008 and contradicts the traditionally observed tendencies under the Fifth Republic.  This is confirmed by the fact that the coefficient of linear correlation between the two popularities is as high as 0.92%.  In other words, because the two series are highly correlative, analyzing either one will not introduce significant differences. We will thus concentrate on the president’s popularity; though this does not mean we shall negate the significance of gaps and reversals in the dominant position.
30Figure 3 provides an initial graphic representation of the link between presidential popularity and the popularity of the latter’s economic policy. The popularity of economic policy is additionally measured by the ratio of positive to negative opinions. The portion of “dk” responders is as such neutralized.
Presidential popularity and the popularity of economic policy (June 2007-May 2010)*
Presidential popularity and the popularity of economic policy (June 2007-May 2010)** Note: The vertical axis shows the ratio of positive and negative opinions for the president and for his economic policy.
31Once again, it appears from a graphic point of view that changes in both levels of popularity are closely linked. The two series have a correlation coefficient of 0.98.  In other words, the evaluation of the president’s economic policy almost perfectly matches the popularity of the president himself: the two evolutions are almost exactly linked.
Economic concerns, perceptions of the economy, media treatment and “real” economic variables
32The economic concerns of the French people are measured through TNS Sofres’ monthly survey, “Barometer of French concerns”, which provides information on the specific concerns – largely economic  – of the interviewees. We have isolated economic concerns dealing with jobs and unemployment on the one hand, and purchasing power on the other. Figure 4 illustrates the evolution of these two concerns.
Concerns regarding jobs and purchasing power (June 2007-May 2010)
Concerns regarding jobs and purchasing power (June 2007-May 2010)
33The levels of the two areas of concern do not show the same evolution. Whilst concerns about unemployment are initially stable and even decrease, from June 2008 they increase steadily and significantly, peaking in April 2009. From April 2009 on, concerns about unemployment remain high at around 75%. Concerns regarding purchasing power initially increase, peaking in June 2008, and subsequently decrease, stabilizing at below 50%. In September 2008, concerns regarding unemployment overtook concerns regarding purchasing power.
34In order to measure voters’ perceptions of the economy, we developed a composite indicator based on the Insee (French national statistics office) monthly household confidence barometer. Four different perceptions of the economic situation are aggregated: a retrospective egotropic perception, a prospective egotropic perception, a retrospective sociotropic perception, and, finally, a prospective sociotropic perception. The two evaluation criteria we have isolated are in keeping with the traditional criteria used in the literature on popularity and voting functions.  Each is obtained using a specific question, and using the difference between the proportion of households with positive perceptions to those with negative perceptions. 
35The two dimensions are highly linked: the egotropic is linked with the sociotropic, and the prospective is linked with the retrospective. We thus decided to combine them for a general perception of the economy.
36Economic perceptions improved during the six first months of the new president’s term (until November 2007), but then declined significantly, from –3 to –120, over the course of November 2007 to September 2009. Afterwards, they improved slightly (Figure 5).
Household perceptions of the economy (June 2007-May 2010)
Household perceptions of the economy (June 2007-May 2010)
37On the question of economic information or mediatized economy, we have used UBM measures (Unité de bruit médiatique, or Unit of Media Reach) produced by TNS Media Intelligence.  The UBM measures are composite indices built on a corpus provided by the principal media organs. They measure the editorial space devoted to a subject, the target audience of 15 or older, and the tone of the discourse used. This third variable is not used in our analysis. The UBM score thus corresponds to the media product, measured in terms of the number of pages or minutes per the potential audience of 15 years or older. The UBM analysis examines 43 written press documents (twelve daily newspapers, fifteen regional papers, ten weekly papers, and six economy or finance magazines). Forty-two audiovisual items are examined (fourteen news broadcasts, fifteen national radio broadcasts, six morning interviews and seven morning radio programs). We have tallied the items strictly related to unemployment, and not those dealing with jobs more generally, which would have expanded the semantic research field to other, connected themes (Contrat nouvelles embauches [New Employment Contracts], Contrat de travail unique [Single Employment Contracts], Hiring, Employment, Employment of Older Workers, Downsizing, Relocation, Professional Training – internships and work/study arrangements – Interim Staffing, Dismissals, the Job Market, Redundancy Plans, Workforce Reductions, and the Wage System). Evolutions in UBM for unemployment and purchasing power show contrasting patterns (Figure 6).
Media reach based on theme (June 2007-May 2010)
Media reach based on theme (June 2007-May 2010)
38Although the UBM score for unemployment progressively increases throughout the entire period, over and beyond marked monthly changes like those seen during April 2008 or February 2009, the UBM score for purchasing power decreases continuously. From March 2009 on it is weak despite an enormous increase in November 2007.
39The “objective” economic data collected comes from the Insee and is made up of monthly variations in unemployment (using the ILO definition of the percentage of the active population) and monthly inflation (using the Harmonised Index of Consumer Prices). Figure 7 shows the evolution of these two indicators.
Monthly evolutions in unemployment and prices (June 2007-May 2010)
Monthly evolutions in unemployment and prices (June 2007-May 2010)
40We observe that unemployment fluctuates more than inflation; the former increases dramatically from July 2008 to April 2009 despite previously being in decline; subsequent evolutions are erratic. During this same period, monthly price variations tend to be weak, even negative (that is, in deflation).
Estimation model choice and estimation strategies
41Given the strong correlation between the popularity of the president and the popularity of economic policy, it would seem unnecessary to distinguish between the two in an estimation model. A distinction would, moreover, introduce redundancy when attempting to determine influences of the different explanatory factors. Thus we will focus exclusively on the popularity of economic policy. Our model is made of four interlocking levels (Figure 1): the popularity of economic policy, which is largely explained by economic concerns. Economic concerns are, for their part, explained by perceptions of the economy and by information produced about the economy, both of which are endogenous to the model. The causal model we have chosen is very close to the one proposed and empirically tested by Sanders and Gavin (2004) and in which the “objective” economy influences the televised economy, which then influences economic evaluations (our “perceptions of the economy”). In turn, economic evaluations influence perceptions of economic competence (our “popularity of economic policy”), which, finally, influence support for political parties. The economic “evaluations” introduced by Sanders and Gavin have been refined in our model to distinguish between perceptions and concerns relative to unemployment and to prices. Unlike Sanders and Gavin, though, we have limited our model to these two economic indicators.
42As a result of the interlocking nature of the model and the temporal nature of the data, we face two difficulties when attempting to estimate the model’s parameters. Each difficulty is linked to these two characteristics. First, it is possible that the series are not stationary;  and second, the explained variables are determined simultaneously and concomitantly through a system of equations. 
43We can overcome the first difficulty by examining the stationarity of the series (see Annex 1 for the results of this examination). All of the series appear stationary  and thus it is possible to estimate them directly with the exception of the series dealing with unemployment and purchasing power concerns. Taken separately, these series are not stationary. However, logically, they are stationary when combined  and thus we will examine them in aggregated mode. To overcome the second difficulty, we may use a specific estimation technique  which will eliminate the problem by correcting the correlation between the error terms of the different equations.
44When it comes to statistical inferences, the restricted number of observations (36) presents another problem. Thus, in order to maintain proper degrees of freedom, we cannot introduce too many explanatory variables at the same time.
45To estimate the popularity of economic policy (Pop) using our model, we group together concerns about unemployment and purchasing power (PUnem & PP), perceptions of the economy (Percep), and media treatment of these two themes (UBMUnem and UBMPP respectively). According to traditional results obtained from popularity functions, we can expect that a combination of these elements will have a negative influence on the popularity of economic policy.
46To estimate economic concerns, we introduce perceptions of the economy (Percep), media treatment (UBMUnem and UBMPP) and the corresponding evolutions in economic variables (?TxUnem and ?IPC). To explain perceptions of the economy, we use media treatment (UBMUnem and UBMPP) as well as “objective” economic variables, in other words evolutions in inflation and the unemployment rate. Finally, for a separate explanation of the media treatment of both economic themes (UBMUnem and UBMPP), we use real or “objective” economic evolutions. Each explanatory variable is explained by its delayed value (t-1). As such, our model is similar to VAR models,  and we are able to account for structural phenomena as well as the accumulated effects of the other explanatory variables. 
47In the end, the estimating model includes five explained variables and seven explanatory values: it is represented below.
49We needed to estimate the popularity of economic policy, economic concerns, the perception of the economy, and media coverage of the economic situation during the month t. These four variables are endogenous to the model; and we add to it the exogenous variables tracing changes in the “objective” economy – that is, inflation and unemployment rates.
Results and findings
50Annex 2 provides the detailed results of the simultaneous estimations of the five equations, and figure 8 presents a synthesis. Generally speaking, the model works well and provides interesting results. Before dwelling on this at greater length, we should specify here that the results are very stable at the chosen specification. 
The estimated effects of the economy on presidential popularity 
The estimated effects of the economy on presidential popularity 
51The following commentary will be divided into four sections. First we will elaborate on the factors that influence the three principal variables: popularity of economic policy, economic concerns and, finally, economic perceptions. We will then be able to observe which factors influence (or not) these dependent variables in our model. Then we will discuss the effects of these factors: that is, our reasoning will be based on the explanatory variables, and not the explained (or dependent) variables. This allows us to compare the effects of the same explanatory variable on the main explained value contained within the model. This comparison and the model’s general results led us to modulate the model in two ways. Firstly, we differentiated between the popularity of economic policy according to the political affiliations of the interviewees (left- or right-leaning). Secondly, we applied the model during two periods: before and after the financial crisis.
Popularity of economic policy, economic concerns, perceptions of the economy: which explanatory factors?
52First let us concentrate separately on the results dealing with the three main explained variables for our model: that is, popularity of economic policy, economic concerns, and perceptions of the economy.
53The popularity of economic policy is a factor with a strong memory effect because the coefficient of its delayed value is highly significant and elevated. In other words, it is marked by strong inertia. In addition, only two explanatory factors directly influence the popularity of economic policy: economic concerns and changes in prices. As a result, perceptions of the economy and the way it is treated in the media do not have a direct impact on its popularity.
54Increased economic concerns do have a negative and statistically significant effect. The more citizens claim to be concerned about unemployment and/or purchasing power, the less a given economic policy is popular. The extent of this effect is relatively large, because elasticity  is higher than 1. A 1% increase in economic concerns can be translated into a 1.3% decrease in the popularity index for economic policy. We observe that perceptions of the economy and its treatment by the media have no direct influence on the popularity of economic policy. In the same way, only monthly inflation, and not unemployment, impacts popularity. The influence remains rather limited because a 1% increase in the monthly inflation rate implies a decrease of 0.02% in the popularity of economic policy.
55These results are quite surprising in that they contradict the typical results found in other literature on the subject. It seems therefore that the popularity of a given economic policy is not affected by economic information, and is only partially and minimally affected by actual economic evolutions.
56As for economic concerns, they are very clearly altered by perceptions of the economy. The more negative the households’ perceptions, the more concerns about unemployment and purchasing power increase. The response elasticity for economic concerns following a 1% increase in economic perceptions is -0.2. These concerns are influenced by media coverage of unemployment, but not purchasing power, and by changes in unemployment levels but not in prices. Both effects are positive, as predicted: an increase in unemployment or in the media treatment of the same translates to an increase in concerns (of 0.06% and 0.003% respectively). And so it seems as though economic concerns in terms of unemployment and purchasing power can essentially be explained by perceptions of the economy and by unemployment (either as represented by the media or objectively). There is, on the other hand, no structural effect for economic concerns: the delayed variable in question has no significant coefficient. In other words, concerns during past periods do not influence current economic concerns.
57Insofar as perceptions of the economy are concerned, the most remarkable element is that media coverage of economic changes has a significant negative influence on perceptions. Elasticity figures are close: -0.03% for media coverage of unemployment and -0.04% for coverage of inflation. That is to say, as themes of unemployment and inflation are dealt with more and more in the media, perceptions of the economy by French households become more negative. Conversely, “objective” economic variables do not influence perceptions of the economy.
58Aside from the “information effects” to which we will return later on, this difference between media effects and macroeconomic effects may have two possible origins. The first concerns the evolution in prices, which does not influence perceptions of the economy. Here we are reminded of the debate surrounding measures of price evolution. Since the changeover to the Euro, the price index is often criticized for its failure to chart households’ perceptions of evolutions in prices. It is thus possible that our results – which indicate the absence of an effect of price indices on economic perceptions – reflect this gap between measurements and perceptions. This possible explanation seems all the more plausible given the absence of an effect of price variations on economic concerns.
59The second possible explanation relates to the discrepancies between media coverage of unemployment and its evolution. Media coverage of unemployment increased significantly during the first half of 2008 (Figure 6a) with a first peak in April of 2008, although the unemployment rate itself did not begin to increase until July-August 2008.  French households may have anticipated the increase in unemployment, thus explaining the temporal discrepancy between the real effects of unemployment and the effect of media coverage of unemployment on perceptions of the economy. The monthly UBM estimation for unemployment corroborates this hypothesis. This third variable for economic perceptions cannot be explained by variations in the unemployment rate, whose estimated coefficient is not statistically different from zero. Media coverage of unemployment thus appears to remain disconnected from macroeconomic change in this area, at least for the period covered by our study. On the other hand, and in the expected direction, media coverage of changes in purchasing power is influenced by price variations: as prices increase, the theme of inflation is dealt with more often by the media.
Comparing the effects of explanatory variables
60If we now compare the effects of the different explanatory variables, we should focus on three elements. The first is the effect of the delayed value of the explained variables. Except for those related to economic concerns, the coefficients associated with these variables are always positive and significantly different from zero. For the four variables with a statistically significant effect, this means that there must be a memory effect of at least one month. In other words, the other explanatory factors for the preceding month also have an influencing power on the current month’s explained variables. Economic concerns, however, appear to have zero memory effect: the delayed value has no influence.
61Next, we observe the fluctuating impact of the “objective” economy, measured by the evolutions in Insee indices for unemployment and prices. The objective economy logically influences media treatment and unexpectedly influences popularity by population (left/ right). We will return to these results. Only variations in unemployment influence concerns and neither measure (unemployment or prices) influences perceptions. In this way, we are able to partly confirm Gavin and Sanders’ results on the absence of a link between the “objective” economy and the perception of the British government’s competency in matters of economic management. At the same time, though, we undermine their findings on the question of the link, in France, between the “objective” economy and the mediatized economy (though only for the measure of inflation), which is not found by Gavin and Sanders for the case of Great Britain.
62As far as media treatment of the economy is concerned, its effects are also strongly differentiated. We observe symmetrical, opposed effects for the left- and right-leaning populations for popularity and a similar effect for perceptions of the economy. However, only media treatment of unemployment influences concerns. These differences are not due to delayed effects over time, because introducing t-1 and/or t-2 values does not affect the results. The media influence thus seems clearly differentiated and more or less instantaneous. These contrasting results, as well as the results for the macroeconomic variables, make it possible to nuance Iyengar and McGrady’s results, according to which, at least in appearance, “presidential popularity was more responsive to economic events than to news coverage of these events” (p. 281) and “as unemployment rises, the president’s approval level falls” (p. 283). This first diagnostic is significantly tempered later on when they affirm that “in summary, popularity is not driven entirely by media management or by history, but by a combination of the two”. In the case of the economy, circumstances are determining factors when it comes to presidential popularity. However, in the case of security, for example, events are less important factors than the media representations of these events. The French presidential election of 2002 and its high-pressure security agenda clearly illustrate this fact.  In matters of security, the American president exerts considerable power over information content and can thus more easily maintain his popularity regardless of the events that arise.
Different economic policy evaluations according to party affiliation
63Given the surprising results for the popularity of economic policy, we were curious to know if these results were not perhaps the product of having combined several elements in order to measure it. One possible answer to this question is that this measurement considers evaluations both from those politically close to the president as well as evaluations from those who are more politically distant. In other words, in combining the evaluations of both populations in the same measurement, we have made the implicit hypothesis that the reactions of both these populations are identical in relation to economic evolutions and media coverage of these evolutions, as well as in terms of perceptions of the economy and economic concerns.
64This hypothesis may be pursued using the idea that “partisan identification or simply political proximity pre-orient perceptions and anticipations in different economic circumstances, and these end up being no more than mere reflections, changing with each situation”.  Some scholars, like Evans and Andersen,  hold that too much weight is given to retrospective economic considerations. In their research on dynamic relations between support for the (Conservative) party and retrospective economic perceptions in the United Kingdom from 1992 to 1997, they find that sociotropic perceptions are highly conditioned by political opinions of the party in power, and have little effect on the party’s popularity. In any case, political support for the party in power more strongly influences sociotropic perceptions than these perceptions influence support for the Conservative party. This hypothesis is further pursued by distinguishing the popularity of economic policy for two separate populations: politically left-leaning and right-leaning. Figure 9 shows that although the profiles for these two measures of popularity are similar, they are not identical, especially not after the general drop in popularity which reached its nadir in May 2008.
Popularity of economic policy (June 2007-May 2010)
Popularity of economic policy (June 2007-May 2010)
65We have thus revised our estimation (Figure 10, details in Annex 3) to include the distinction between the two populations. The model remains the same and includes the explanatory variables as well as the other explained variables in the equation system.
Determining support variables for economic policy according to respondent political affiliation 
Determining support variables for economic policy according to respondent political affiliation 
66Distinguishing between the two populations allows us to see that the evolutions in popularity are not the same. This may explain the surprising results obtained when the two populations are combined. First, the delayed effect is similar for both the left- and right-leaning respondents, producing a quasi-identical structural effect on the popularity of economic policy and conforming to what we know about the profound inertia of popularity mechanisms. Second, economic concerns and economic perceptions have different effects for the two populations. Right-leaning respondents’ evaluations of economic policy are negatively influenced by economic concerns, whereas evaluations by left-leaning respondents are positively influenced by economic perceptions. Thus both variables respond as expected: when economic concerns or perceptions become more negative, popularity falls, but for different sub-populations. Hence the usefulness of distinguishing between perceptions of the economy and economic concerns, which Gavin and Sanders fail to do in their study.
67Media coverage and economic variables have symmetrical and opposite effects. Among the left-leaning respondents, media treatment of the themes of unemployment or purchasing power has no significant effect on popularity. However, real evolutions in inflation have a significant negative influence. Inversely and symmetrically, among right-leaning respondents, popularity evaluations are positively influenced by media treatment of unemployment (but not purchasing power). Likewise, monthly variations in unemployment figures (and not prices) positively influence the popularity of economic policy among right-leaning respondents.
68In other words, left-leaning voters negatively evaluate economic policy when purchasing power is weak (in terms of macroeconomic evolutions only). As inflation grows, the government’s economic policy is evaluated more negatively. Inversely, right-leaning voters positively evaluate economic policy when unemployment increases and becomes an important topic in the eyes of the media. Our breakdown illustrates symmetry in the effects on popularity based on political proximity, both for macroeconomic evolutions and for media coverage. But there is another interesting aspect of these results. Theories of politico-economic cycles  have more or less verified the hypothesis that in choosing between unemployment and inflation, left-leaning politicians and voters are more concerned with fighting unemployment than inflation. On the other hand, right-leaning politicians and voters more often fight inflation than unemployment. Our results, however, show that left-leaning voters tended to evaluate the right-leaning government on the basis of inflation, and right-leaning voters evaluated the same government on the basis of its fight against unemployment.
69Another interesting finding from these results concerns the positive effect of unemployment and unemployment media coverage on the right-leaning respondents’ evaluations of economic policy. According to the economic theory of vote sanctions, these are unexpected effects:  negative macroeconomic results should mean a decrease in presidential popularity. One possible explanation for these surprising results is that during a time of global crisis, the president is not seen as accountable for macroeconomic evolutions such as increases in unemployment levels. Among right-leaning voters, economic policy is then not evaluated on the basis of actual results but on the basis of their own political position in relation to the president.
Before and after the economic crisis
70Another element we must take into consideration when evaluating our results is the unchanging nature of behaviors over time. One might assume that voters evaluating economic policy would react to explanatory variables – concerns, perceptions, media treatment and macroeconomic evolutions – in the same manner throughout the entire study period.
71However, during Nicolas Sarkozy’s term of office a major economic crisis began, triggered by Lehman Brothers bank filing for bankruptcy in the third trimester of 2008. September 2008 thus marks a break in some of the series we have studied. Perceptions of the economy, for example, significantly deteriorated during this month (Figure 5). The UBM for unemployment increased strongly (Figure 6a) and, on average, the figure for the following months was much higher than before; and the reverse is true for the UBM for purchasing power (Figure 6b).  We might therefore question the stability of reactions between September 2008 and June 2010.
72We have, in order to address this question, used the estimation model for two periods: June 2007 to August 2008 and September 2008 to June 2010. The results have been summarized in Table 1 and are detailed in Annex 4. Significant changes can be observed between the two periods.
Comparing models before and after September 2008 
Comparing models before and after September 2008 
73After September, price variations negatively impacted the popularity of economic policy, but before September, real economic changes had no impact. On the other hand, we find that before September 2008, economic concerns negatively influenced the popularity of economic policy. And, inversely, media coverage of unemployment positively influenced this popularity, but only after the start of the economic crisis. This last result can be explained by the respondents’ differences in party affiliation. It is only after the Lehman Brothers collapse that we observe this “rallying around” economic policy. Before the collapse, the phenomenon did not exist at all. This observation would tend to confirm the explanation of popularity differences for the left- and right-leaning respondents.
74The changes in economic concerns and economic perceptions over the first two periods are particularly interesting. While perceptions of the economy negatively influence economic concerns throughout the two periods, media coverage of unemployment and unemployment’s true monthly variations only have a (positive) impact after September 2008. In other words, neither unemployment as a true phenomenon nor unemployment as a mediatized phenomenon affect economic concerns until after the beginning of the crisis. Between the two periods, the effect of media representations of purchasing power is inversed: before September, it is positive, and turns negative afterwards.
75The determining factors for perceptions of the economy vary from one period to another. Real economic changes do not affect measurements, but media treatment of purchasing power does weaken perceptions before September 2008. After this month, media treatment of unemployment begins to affect perceptions. It is thus clear that reactions vary from one period to another, and that purchasing power plays a more important role before the economic crisis. After the collapse of Lehman Brothers, however, it is unemployment that weighs more heavily.
76Four main conclusions may be drawn from this study of Nicolas Sarkozy’s popularity over the period ranging from June 2007 to July 2010. First, it appears that the president’s popularity is strongly linked to the popularity of economic policy. However, assimilating presidential popularity and the popularity of his economic policy may be purely conjunctural. We need only reflect on how George W. Bush’s popularity was positively affected by his reaction to the 11 September 2001 attacks on the World Trade Center and its aftermath in Iraq to understand the historical, and therefore situated, nature of such an assimilation. In framing the attacks as a national security problem and a war on terror, G. W. Bush saw his popularity increase by 35 points between 10 and 15 September (Gallup survey). This typical “rally around the flag” strategy meant that he largely surpassed the 18-point popularity boost enjoyed by his father, G. H. Bush, in January 1991 when he launched the Desert Storm operation against Iraq.  Nevertheless, assimilation of this type undoubtedly took place in the French situation between 2007 and 2010. This study’s goal is thus not to show that economic variables precede political variables, nor that they structure all political opinions always and everywhere.
77Second, we conclude that economic concerns are primarily driven by real variations and media coverage of unemployment. Perceptions of the economy, however, are altered only by media coverage of broad-brush economic phenomena, and not by the phenomena themselves. This is an important conclusion given the initial question posed in the article’s introduction: what are the respective impacts of real economic results – or the media depictions of these results – on representations of popularity? We found that the mediatized economy conditions perceptions of the economy more so than actual economic conditions. Here we recognize the phenomenon of “impersonal influence”, identified by Diana Mutz as the depersonalization of political experience. Media groups thus become the vectors of this impersonal influence.  This was demonstrated in the US when the American public, overwhelmed with media frenzy during the 1991 recession, perceived the economy as declining even after the end of the economic downturn. Clinton took advantage of this situation, deliberately drawing upon it for his electoral strategy.  These results are a partial confirmation of W. Thomas’ theorem: “If men define situations as real, then they are real in their consequences.” Once again, our composite indicator for economic perceptions combines four types of perceptions: egocentric and sociotropic, and prospective and retrospective. The indicator thus provides a synoptic view of a given individual’s perception of the economy.
78Our third conclusion concerns the significant gap between unemployment growth and the effects of this growth on both the popularity of the president and the popularity of his economic policy. Premature media coverage of the economic crisis’s effects, and especially of the job market, meant that real unemployment figures influenced neither media coverage of unemployment nor perceptions of the economy.
79The fourth conclusion we come to is that real economic changes and the mediatized economy affect presidential popularity in two ways: thematically and according to political beliefs. The popularity of economic policy varies depending on an individual’s political beliefs, and political identification is a particularly sensitive variable when it comes to changes in popularity as a result of real economic changes and media representations of these changes. In the same way, the popularity of economic policy, and thus of the president, are not the same before and after the onset of the financial crisis.
80The first 36 months of Sarkozy’s presidency can thus be divided into two clear phases. The first, from June 2007 to the summer of 2008, was characterized by a roughly year-long decline in popularity with an increase in concerns regarding purchasing power. The theme had been an important one in Sarkozy’s electoral campaign; the slogan “Travailler plus pour gagner plus” ( “Work more to earn more”) was emblematic of the then-candidate’s eventual conquest of the presidency. During this phase, pre-Lehman Brothers collapse with its ensuing global crisis, there was significant media coverage of purchasing power – more so than coverage of unemployment. As such, the public was thereby encouraged to use purchasing power as the main criterion for evaluating the political situation. The second phase began in September 2008 with the Lehman Brothers collapse. Public perceptions, particularly of the state of the job market, were modified as a result, and media coverage of unemployment within the economic context accordingly increased significantly, although there was no reduction in the coverage of purchasing power. Presidential popularity remained relatively stable during this period (confidence levels fluctuated between 33% and 39%). It appears as if the financial crisis, strongly emphasized as a global problem, ended up freeing the president from all responsibility in regard to the economic situation in general and the employment question in particular, this latter issue having become a major preoccupation for the French people.
81The problem was recognized at the highest levels of government, but it was a “difficult” question and reactions to it were formulated belatedly. Through the Service d’information du gouvernement (SIG, Governmental Information Service), a governmental campaign was launched in 2008 from 23 June to 11 July. With a total budget of 4,300,000 Euros, and entitled the “purchasing power campaign”, the program’s objective was to “raise awareness of government action to support purchasing power, in particular with regard to measures that would allow increases in work-related revenues, and decreases in taxes and current expenditures for households”. The campaign message indicated that “the government is aware of French anxiety in regards to purchasing power and distribution of wealth, and concrete responses will be provided for various different life-situations” and concluded on the following note: “We will win the purchasing power battle month by month”. The message was widely diffused on terrestrial, cable and satellite television, and in the daily regional, national, and free press as well as on the Internet. However, the campaign did not have the chance to take effect: the global financial crisis hit at almost exactly the same time. As a general point, it is thus crucial to carefully distinguish between the effects of purchasing power and unemployment: whether this is in relation to media coverage, structural concerns, or evaluations of economic policy as a whole.
82Finally, the main contribution of these results is to reveal – or confirm, depending on the level of knowledge – that popularity largely depends on the media coverage given to presidential policy action. The results reinforce the theory of “the persuasive effects of information” according to which information does not only possess a cognitive value but a persuasive value as well.  In focusing collective attention on certain problems (unemployment or purchasing power, in our cases), the media spurs its audiences to structure their individual perceptions and concerns around and through them. Cognitively and via the media, citizens are encouraged to convert their perceptions and/or concerns into criteria for evaluating policy action, and economic policy in particular. This explains why we obtain a 0.98 correlation coefficient between the two chronological series relating to presidential popularity and the popularity of economic policy. As Iyengar and McGrady affirm, “in summary, events do not speak for themselves but require interpretation and explanation” (p. 275). The media, indeed, holds this particular power to contextualize, and it is thus largely to the media that we owe the changing popularity of Nicolas Sarkozy or, for that matter, of any other president.
83* * *
Series stationarity test (augmented Dickey-Fuller test)
Series stationarity test (augmented Dickey-Fuller test)
Results of the estimation model on the popularity of economic policyNote: ***, **, and * indicate coefficients statistically different from zero at a level of 1%, 5% and 10% respectively.
Results of the estimation model on the popularity of economic policy relative to political affiliation of interviewees
Results of the estimation model on the popularity of economic policy relative to political affiliation of intervieweesNote: ***, **, and * indicate coefficients statistically different from zero at a level of 1%, 5% and 10% respectively. The remainder of the system of equations is identical to the previous model. The results are similar to those obtained previously and are thus not reproduced here but may be obtained from the authors on request.
Results of the estimation model depending on time periodNote: ***, **, and * indicate coefficients statistically different from zero at a level of 1%, 5% and 10% respectively.
Maurice Halbwachs, The Causes of Suicide, trans. Harold Goldblatt (London: Routledge & Kegan Paul, 1978).
Philippe Juhem, “Les ‘favoris des sondages’. Contribution à une analyse structurale de la popularité mesurée par les baromètres d’opinion”, paper given at the doctoral conference at the École doctorale de science politique at the Université Paris I-Panthéon Sorbonne, “Des discours aux effets de la communication en démocratie”, 21 September 2007.
Daniel Gaxie, Nicolas Hubé (eds), L’Europe des européens (Paris: Economica, 2011).
Jacques Gerstlé, “Évènement, communication et effets de popularité”, conference contribution at “Popularité politique”, Association française de science politique, 16-18 October 1990.
Shanto Iyengar, Jennifer A. McGrady, Media Politics. A Citizen’s Guide (New York: Norton & Company, 2007). On page 271, we read that “Research suggests that three principal factors influence presidential popularity: length of time in office, the course of events and public relations.”
John Mueller, War, Presidents and Public Opinion (Lanham: University Press of America, 1973), and “Presidential popularity from Truman to Johnson”, American Political Science Review, 64(1), 1970, 18-34.
James A. Stimson, “Public support for American presidents: a cyclical model”, Public Opinion Quarterly, 40, 1976, 1-21.Online
Michael B. MacKuen, “Political drama, economic conditions and the dynamics of presidential popularity”, American Journal of Political Science, 27, 1983, 165-92; Kristen R. Monroe, Presidential Popularity and the Economy (New York: Praeger, 1984); Helmut Norpoth, Economics, Politics and the Cycle of Presidential Popularity. Economic Conditions and Electoral Outcomes: The United States and Western Europe (New York: Agathon Press, 1985).
Richard A. Brody, Assessing the President. The Media, Elite Opinion and Public Support (Stanford: Stanford University Press, 1991).
John R. Zaller, “Monica Lewinsky’s contribution to political science”, Political Science and Politics, 33, 1998, 182-9.
Samuel Kernell, Going Public: New Strategies of Presidential Leadership (Washington, D.C.: Congressional Quarterly Press, 1986).
Shanto Iyengar, Donald R. Kinder, News that Matters. Television and American Opinion (Chicago: The University of Chicago Press), 1987.
Jeffrey E. Cohen, Going Local. Presidential Leadership in the Post-Broadcast Age (Cambridge: Cambridge University Press, 2010).
Paul Gronke, Brian Newman, “FDR to Clinton, Mueller to ? A ‘state of the discipline’ review of presidential approval”, report presented at the annual Congress of the American Political Science Association, Washington D.C., 31 August to 3 September 2000.
John Zaller, The Nature and Origins of Mass Opinion (Cambridge, Cambridge University Press, 1992).Online
George C. Edwards III, William Mitchell, Reed Welch, “Explaining presidential approval: the significance of issue salience”, American Journal of Political Science, 39, 1995, 108-34.Online
George C. Edwards III, The Public Presidency. The Pursuit of Popular Support (New York: Saint Martin’s Press, 1983).
Donald R. Kinder, Roderick Kiewiet, “Sociotropic politics: the American case”, British Journal of Political Science, 11, 1981, 129-61.Online
Michael B. MacKuen, Robert S. Erikson, James S. Stimson, “Peasants or bankers? The American electorate and the US economy”, American Political Science Review, 86(3), 1992, 597-611.Online
Werner W. Pommerehne, Friedrich Schneider, Jean-Dominique Lafay, “Les interactions entre économie et politique: synthèse des analyses théoriques et empiriques”, Revue économique, 32(1), 1981, 110-62.Online
Christopher Anderson, Blaming the Government. Citizens and the Economy in Five European Democracies (Armonk: M.E. Sharpe, 1995).
André Bernard, “La conjoncture économique et le vote: une relation ambiguë”, Revue française de sociologie, 38(2), 1997, 245-64.Online
Alexander C. Pacek, Benjamin Radcliff, “Economic voting and the welfare state: a cross-national analysis”, Journal of Politics, 57(1), 1995, 44-61.Online
Richard S. Conley, “From Elysian fields to the guillotine? The dynamics of presidential and prime ministerial approval in Fifth Republic France”, Comparative Political Studies, 39(5), 2006, 570-98.Online
Jacques Lecaillon, “Salaires, chômage, et situation politique”, Revue d’économie politique, 5, 1980, 615-27; Michael S. Lewis-Beck, “Economic conditions and executive popularity: the French experience”, American Journal of Political Science, 24, 1980, 306-23; Douglas A. Hibbs, “Economics and politics in France: economic performance and political support for presidents Pompidou and Giscard d’Estaing”, European Journal of Political Research, 9, 1981, 133-45; Jean-Dominique Lafay, “The impact of economic variables on political behavior in France”, in Douglas A. Hibbs, Heiko Fassbender (eds), Contemporary Political Economy (New York: North-Holland, 1981).Online
The ARIMA method (Autoregressive Integrated Moving Average) is a way of estimating time series. The numbers in parentheses correspond respectively to the number of autoregressive terms, the differentiation number, and the number of terms in moving average.
R. A. Brody, Assessing the President…
Shanto Iyengar, Donald R. Kinder, Mark D. Peters, Jon A. Krosnick, “The evening news and presidential evaluations”, Journal of Personality and Social Psychology, 46, 1984, 778-87.Online
Benjamin I. Page, Robert Shapiro, Glenn R. Dempsey, “What moves public opinion?”, American Political Science Review, 81(1), 1984, 23-43.Online
S. Iyengar, D. R. Kinder, News that Matters…
Jon A. Krosnick, Donald R. Kinder, “Altering the foundations of support for the president through priming”, American Political Science Review, 84, 1990, 497-512.Online
Jon A. Krosnick, Laura A. Brannon, “The impact of the Gulf War on the ingredients of presidential evaluations: multidimensional effects of political involvement”, American Political Science Review, 87, 1993, 963-75.Online
Robert K. Goidel, Todd G. Shields, Mark Peffley, “Priming theory and RAS model: toward an integrated perspective of media influence”, American Politics Quarterly, 35, 1997, 287-318.
G. C. Edwards III, W. Mitchell, R. Welch, “Explaining presidential approval…”
George C. Edwards III, Dan B. Wood, “Who influences whom? The president and the public agenda”, American Political Science Review, 93, 1999, 327-44.Online
George C. Edwards III, On Deaf Ears. The Limits of the Bully Pulpit (New Haven: Yale University Press, 2003), and The Strategic President. Persuasion and Opportunity in Presidential Leadership (Princeton: Princeton University Press, 2009).
Brandice Canes-Wrone, Who Leads Whom? Presidents, Policy and the Public (Chicago: University of Chicago Press, 2006).
Neil T. Gavin, David Sanders, “Television, economy and the public’s political attitudes”, in Neil T. Gavin (ed.), The Economy, Media and Public Knowledge (London: Leicester University Press, 1998), 90-111.
S. Iyenygar, J. McGrady, Media Politics…Online
Brice Teinturier, “Nicolas Sarkozy et la fonction présidentielle: premier bilan”, in Olivier Duhamel, Brice Teinturier (eds), L’État de l’opinion-Sofres (Paris: Seuil, 2008), 11-31; Emmanuel Rivière, “Nicolas Sarkozy, réformateur paradoxal”, in L’État de l’opinion-Sofres, 31-44.
The Phillips curve (or relationship), from 1968, is based on the empirical observation of an inverse relation between the unemployment rate and inflation: unemployment and inflation increases cannot exist simultaneously. For a long time the curve was used to explain price determination in Keynesian macroeconomic models. Later, it was critiqued: theoretically, by research on inflation anticipation, and empirically, with the arrival of stagflation in the 1970s, when economies did indeed experience high inflation and unemployment rates.
See in particular Michael Lewis-Beck, Mary Stegmaier, “Economic determinants of electoral outcomes”, Annual Review of Political Science, 3, 2000, 183-219, or “Economic models of voting”, in Russel Dalton, Hans-Dieter Klingemann (eds), The Oxford Handbook of Political Behavior (Oxford: Oxford University Press, 2007), 518-37. In French, see Bruno Jérôme, Véronique Jérôme-Speziari, Analyse économique des élections (Paris: Economica, 2010).
Survey data (popularity and concerns) were re-scaled over time. The study dates vary from one survey to the next and from month to month. We have used the following criterion: surveys carried out at the end of the month count for the month in progress, but those performed at the beginning of the month count for the preceding month.Online
This is the political barometer used by Sofres-Le Figaro, based upon the following question: “Are you confident, somewhat confident, not very confident, or not at all confident in the capacities of…to resolve France’s current problems?” Data from August is missing (the survey was not performed during this month). As a result, data was extrapolated linearly, that is, by taking the variation between July and September and splitting it in two.
This question was raised by Éric Dubois in “Économie politique et prévision conjoncturelle: construction d’un modèle macroéconométrique avec prise en compte des facteurs politiques”, doctoral thesis in economic science, Université Paris I-Panthéon Sorbonne, 2005.
Jean-Luc Parodi, “Le Premier ministre sous la Cinquième République: une popularité dominée”, Pouvoirs, 83, 1997, 89-99.
The coefficient is statistically different from zero at a level of 0.0001%.
The popularity of economic policy is measured using a monthly BVA barometer measuring economic policy. The following question is asked: “Would you, personally, say that the economic policy currently pursued by the government is very good, rather good, rather poor or very poor?” The survey is not performed in July or August; the values for these months are linearly extrapolated from the June and September values.
The coefficient is statistically different from zero at a level of 0.0001%.
The exact question is: “Among the following, which of these items concern you, personally, the most?” Among the 16 items, “jobs and unemployment” as well as “changes in purchasing power” are listed. We have used both here. More precisely, we have used the percentage of interviewees who have cited these two responses, regardless of their priority ranking.
On this question, see Dennis Mueller, François Facchini, Martial Foucault, Abel François, Mickaël Melki, Raul Magni-Berton, Choix publics: analyse économique des décisions publiques (Brussels: De Boeck, 2010).
For the retrospective sociotropic dimension, the question is: “In your opinion, over the past twelve months, has the general quality of life in France… significantly increased (+) / slightly increased(+) / remained the same/ slightly decreased (-) / significantly decreased (-)”; for the prospective sociotropic dimension, the question is: “In your opinion, over the next twelve months, will the general quality of life in France…increase significantly (+) / increase slightly (+) / remain the same / decrease slightly (-) / significantly decrease (-)”?; for the retrospective egotropic dimension: “Over the past twelve months, has your household’s financial situation … significantly increased (+) / slightly increased(+) / remained the same/ slightly decreased (-) / significantly decreased (-)”; finally, for the egotropic prospective dimension: “Do you think that over the next twelve months your household’s financial situation will …increase significantly (+) / increase slightly (+) / remain the same / decrease slightly (-) / significantly decrease (-)”?
We graciously thank Sonia Metché, ad hoc Director of Studies at Kantar Media, for having communicated these results to us, and the Service d’information du gouvernement (Governmental Information Service) for having authorized us to analyze them.
When a chronological series is weakly stationary, its average and typical gap do not vary with time, that is, they remain constant. The absence of stationarity make statistical inferences difficult because the quality of the estimated coefficients is such that it is impossible to confirm their statistical significance. This problem is made known via auto-correlation between the errors contained in the model; that is, by the fact that the errors are correlated amongst one another.
The simultaneous determination of explanatory variables also raises problems of statistical inference: the quality of the coefficients is not high enough to confirm their statistical significance. This problem is made known via correlation between the estimation errors for the different explained variables. This kind of correlation considerably contradicts hypotheses on the traditional linear regression model.
The results of the Augmented Dickey-Fuller Unit Root tests are presented in Annex 1.
Granger’s work on series co-integration proved that the sum of two non-stationary time series may itself be stationary.
Here we used the method of “seemingly unrelated regression” (SUR) which allows for a simultaneous estimation of equations while correcting error correlation between them. Two other methods (two-stage least squares and three-stage least squares) were also tested and gave the same results. The authors will provide these results upon demand.
VAR models (Vector Auto Regression) are macroeconomic forecasting models for time series.
If Yt-1 influences Yt, the variables influencing Yt-1 will also influence Yt.
For example, the fact that the value of each explanatory variable is counted in t and t-1 does not fundamentally alter the results, and the estimations obtained do not increase in quality.
The presence of an arrow means that the given variable is statistically significant (at a minimum level of 10%). The plus and minus signs indicate whether the influence is positive or negative.
All elasticity figures are calculated using the averages of the values for the variables concerned.
The first major rise in unemployment happens in August 2008 with a 2.2% increase. Such an increase had not been seen since March of 1993. In October, the increase hits 2.4%, meaning a total of more than two million jobless citizens. In November, unemployment is at a 3.2% increase – the highest rate in fifteen years.
Jacques Gerstlé, “Une fenêtre d’opportunité électorale”, in Pascal Perrineau, Colette Ysmal (eds), Le vote de tous les refus (Paris: Presses de Sciences Po, 2003), 29-52.
Patrick Lehingue, “L’analyse économique des choix électoraux”, in “Les sciences du politique aux États- Unis”, Politix, 41, 1998, 82-122.
Geoffrey Evans, Robert Andersen, “The political conditioning of economic perceptions”, The Journal of Politics, 68, February 2006, 194-207.Online
The presence of an arrow means that the given variable is statistically significant (at a minimum level of 10%). The plus and minus signs indicate whether the influence is positive or negative. See Annex 3 for details.Online
For a presentation of partisan politico-economic cycles, see D. Mueller, F. Facchini, M. Foucault, A. François, R. Magni-Berton, M. Melki, Choix publics. Analyse économique des décisions publiques.
On the other hand, the expected result is obtained for the left-leaning voters. Poor macroeconomic results for inflation mean decreased popularity.
Until August 2008, the UBM for unemployment was stable at around 438, with purchasing power around 1636. From September 2008 to June 2010, these figures settle at 881 and 588 respectively.
Values are given when the estimated parameters are statistically different from zero (at a minimum level of 10%); the plus and minus signs indicate whether the influence is positive or negative. The figure in parentheses represents the effect of a marginal variation of the explanatory variable on the explained variable. For details, see Annex 4.
Marc J. Hetherington, Michael Nelson, “Anatomy of a rally effect: George Bush and the war on terrorism”, Political Science and Politics, 36(1), 2003, 37-42.
Diana Mutz, Impersonal Influence. How Perceptions of Mass Collectives Affect Political Attitudes (Cambridge: Cambridge University Press, 1998).
Marc J. Hetherington, “The media’s role in forming voters’ national economic evaluations in 1992”, American Journal of Political Science, 40, 1996, 372-95.
Jacques Gerstlé, La communication politique (Paris: Armand Colin, 2008).Online