The Environmental Approach: The Ecological Tradition

16The first research tradition, generally called “ecological” or geographical, is based on a simple fact: the observable relationships between the geographical, social, or political characteristics of an aggregated unit—here an environment as defined by Huckfeldt—and the electoral behavior that occurs within that unit represent more than just the mere accumulation of individual characteristics. The basic hypothesis is that individual behavior is defined more by the opportunities for interaction available within an individual’s environment than by the individual’s own characteristics. [11] The goal of this analytical tradition, therefore, is twofold: first, to empirically demonstrate, using aggregated data, that the form of an environment influences the electoral behavior of individuals within it; second, to formulate hypotheses capable of explaining the “contextual” foundations of this relationship.

17The American Kevin R. Cox developed the most well-known ecological hypothesis during a research carried out in Great Britain. According to Cox, if the decision to vote for a candidate or a party depends at least as much on the sociodemographic composition of the area where a voter lives as on the voter’s individual characteristics, that is because an area’s composition sustains a “culture” of norms that are transmitted from voter to voter. He called this the “neighborhood effect”. [12] Following Cox, William L. Miller applied this hypothesis to his own analysis of the results of legislative elections in England from 1964 to 1974. Miller’s theory is that the social environment determines opportunities for interaction, and that those interactions are largely responsible for creating consensus effects. On that basis, he claims that “people who talk together vote together”. [13] Miller proposed a total of four other hypotheses regarding the impact of the social environment on voting:

18– The “no environmental effect model”: no difference in electoral behavior between the different types of neighborhood studied; interactions between neighbors have no effect on voting.

19– The “reactive environmental effect model”: voters’ electoral preferences are reinforced in reaction to interactions with people from a different social class.

20– The “consensual environmental effect model”: voters will conform to the opinions of the people they interact with.

21– The “‘Przeworski’ environmental effect model”: environmental effects differ depending on social class. [14]

22Most authors today assimilate the neighborhood effect hypothesis to the third model above, which is undoubtedly the one that best corresponds to Miller’s well-known phrase, but also the one that has been most criticized, and rightly: Miller’s argument was based on aggregated data, and he never empirically verified that “people who talk together vote together”. [15]

23The neighborhood effect hypothesis has been re-examined by various British geographers since the beginning of the 1990s, the most prominent being Ron Johnston and Charles Pattie. The two authors first clarified Miller’s theoretical scenarios. Johnston then identified five hypotheses capable of explaining the consensus effect. He summarizes them as follows: [16]

24– “I talk with them and vote as they do” (local social interaction): situation that corresponds directly to Miller’s consensus model.

25– “I want to be like them so I live with them” (environmental selection): here, consensus is based on the shared values or desires for identification that cause people to choose to live in the same place.

26– “I live among them and want to be like them” (emulation): this refers to conformist behavior, which can occur even when people do not actually engage in conversation.

27– “What I observe around me makes me vote with them” (environmental observation): a similar perception of similar life circumstances leads to similar voting habits.

28– “They want me to vote for them here” (local pressure): here, Johnston directly addresses the question of the social pressure that can be exerted by political parties. Similarities can be drawn between this idea and another hypothesis put forward by Valdimer Key regarding the “friends and neighbors effect”, [17] whereby candidates are more successful in the areas they come from and/or where they have networks of supporters.

29Based on these new hypotheses, the British geographers developed explanatory models adapted to the statistical relationships that can be observed with the help of modern technologies. Their current analyses are carried out at very detailed levels of aggregation, drawing on intracommunal survey data, which reduces the risk of ecological fallacy. The use of multilevel modeling, or sophisticated models for calculating the “predicted” votes for each aggregated unit, which can then be compared with the actual results (a method that was already being used by Miller), has allowed them to confirm, in large part, the conclusions drawn by their predecessors: taking the characteristics of areas into account reduces or alters the effect of individual variables. But statistical sophistication does not change the fact that, in the geographic tradition, the mechanisms that actually cause context-related statistical variation have still only ever been expressed in the form of hypotheses. It comes down to a question of approach: the use of aggregated data precludes direct investigation of the forms and effects of social interaction. To remedy this “empirical gap”, Johnston and Pattie have conducted several surveys in which they abandon the analysis of aggregated data in favor of a method borrowed from another tradition of contextual analysis.

The Network Approach: The Interactionist Tradition

30In an article published in 2001, [18] Pattie and Johnston present a survey carried out in order to evaluate the impact of conversations on voting, and more specifically on voting trajectories. Their goal was to empirically test the hypothesis that voting intentions can be altered by conversations, with the expectation that voters would alter their vote to be more in line with the opinions they most often encountered. The remarkable thing about the authors’ discourse here is not their sources—electoral panel surveys conducted in Britain from 1992 to 1997 that asked more than 2,000 voters about the three people with whom they most often discussed politics; or their results, which are interesting but of limited relevance for the initial hypothesis—the observed statistical effects of conversations on voting intentions are sometimes significant, but nevertheless weak and indirect (conversations seem to cause changes in attitude rather than changes in voting intentions), and moreover only based on conversations reported in 1992, five years before the 1997 election that was used to track changes in voting decisions. The really remarkable thing is the novel theoretical approach the study represents. The authors talk of “conversational contexts” and “networks” rather than “neighborhoods”. The demonstrative logic has changed: it is no longer a matter of observing collective units in order to deduce the existence of relationships that affect voting decisions, but rather isolating and directly measuring those relationships starting from the individuals involved. [19] Although its overall goal is quite similar, this way of working belongs to a quite different tradition of contextual analysis, with its own theoretical and methodological peculiarities: the “interactionist” approach.

31Paul Lazarsfeld is unquestionably the pioneer of this approach in the field of electoral analysis. In everything he wrote, interpersonal relationships are always at the heart of his vision of the social and political world, theorized in well-known concepts like the “two-step flow of communication” and “opinion leadership”, and tackled using a constantly evolving range of methods that testify to his ongoing desire to adapt the empirical approach to contextual demonstration. In that respect, Lazarsfeld draws heavily on the work and thought of Gabriel Tarde, as Elihu Katz points out in the preface to the new edition of Personal Influence. The Austrian author’s work was deeply marked by the idea of a social world made up of networks and chains of influence, where an individual’s place in society is shaped by the ebb and flow of the psychological processes of imitation and opposition. Although this type of approach has been featured in numerous surveys and theoretical reflections by sociologists, [20] economists, [21] and social psychology researchers, [22] it only became widely popular at the beginning of the 1980s, when Huckfeldt first started to publish his works. A rather eclectic tradition, [23] interactionism nonetheless relies on one fundamental theoretical and methodological perspective: that context and its effects can be defined through the study of conversational networks—formal, informal, or digital. [24] Here, again, the main heuristic benefit of the interactionist tradition is its capacity for developing theoretical models that can generate powerful hypotheses regarding electoral analysis. To demonstrate, this, we will now examine three of the most significant of these models.

32The first is that of “social cohesion”: that the degree of cohesion and homogeneity within a network determines the network’s ability to influence behavior, particularly through the mechanisms of conformism and social pressure. In this theoretical model, the most cohesive and homogeneous networks—sociologically as well as politically—are the most likely to produce stable opinions and encourage participation; vice versa, heterogeneous networks with low levels of cohesion produce indecision and political exit strategies. This model, which draws on social psychology, is perfectly illustrated in Lazarsfeld’s work. According to the findings presented in The People’s Choice, the most socially and politically structured groups exert a fundamentally conformist influence on their members. Social integration leads to political integration: the development of a group consciousness encourages individuals to orient themselves within the public sphere, translate their own social interests into political choices, and participate in elections. Heterogeneous environments, on the other hand, have the opposite effect. Lazarsfeld and his colleagues emphasize the notion of cross pressures, or in other words the mutually incompatible forces that individuals are subjected to both internally (psychologically) and externally (socio-psychologically, in the form of the conflicting influences exerted on individuals by those around them), and that motivate them to develop exit strategies.

33The second model is that of “structural equivalence”. The important thing here is not so much the structure of networks as the position of individuals within them. In this model, individuals who occupy the same position within a network act in similar ways because they have relationships with similar people, fulfil the same social roles, and therefore have the same interests, knowledge, and perceptions. This model is backed up by socio-psychological theories: individuals adopt certain behaviors as a result of a process of imitation. [25] But it also finds support in rational choice theory: for example, assuming that voters use their social capital rationally when it comes to politics, they may find it beneficial to talk to their “structural equivalents” in order to obtain information cheaply. In their analysis of extrafamilial networks, Huckfeldt and John Sprague argued that interpersonal influence is determined by structural and socio-psychological factors—the ability to identify the preferences of discussion partners; opportunities for interacting with people with similar preferences—but also by the frequency of discussions, which is more important than the closeness of relationships: in their sample, discussions between acquaintances (“regular contacts and friends”) had a more significant statistical effect than discussions between close friends. [26] The influence that social relationships have on individuals, therefore, does not necessarily operate through processes of persuasion. [27]

34Finally, the third model can be described as that of “deliberation”. It concerns the positive impact that heterogeneous networks with low levels of cohesion have on democratic participation: by familiarizing individuals with conflicting opinions and behavior, such networks help accustom individuals to the practice of debating ideas and make them more tolerant, and so have a beneficial effect on citizen engagement. [28] This model is particularly relevant to weak interpersonal ties, which function as sources of diverse information and introduce variety into cohesive, stable networks. Huckfeldt and Sprague have developed it in their recent work. [29] The social cohesion model, which deals with how homogeneous, cohesive groups relate to politics, cannot in their view explain why political alternation occurs between one election and the next. Nevertheless, it would be unlikely for such alternation to be totally independent of the mechanisms of everyday interpersonal influence. After having empirically verified the heterogeneity of networks of political discussion in the United States, the two authors then investigate the possibility of exchanges between individuals with politically opposed views, before drawing more general theoretical conclusions from their observations. The “strength of weak ties” comes into play here in that weak ties can sustain disagreements between individuals because, according to the authors, they do not involve persuasive communication, in which discussion partners try to change each other’s mind or condemn each other’s behavior. Communication in such relationships is simply “effective”: both partners are content to understand the messages expressed by the other. Their underlying difference of opinion remains unaltered because each has their own circle of close friends and family, unknown to the other: people who share their opinions, serve as points of reference for them, and help them relativize what their discussion partner says. [30] On that basis, the authors conclude that heterogeneous networks create ambivalence and instability among citizens, but do not lead to a reduction in electoral participation. [31]

35This review of the various approaches and theoretical models used in the contextual analysis of electoral behavior demonstrates the diversity and heuristic potential of such analysis. This is not the place to discuss the advantages and disadvantages of each model. Instead, our next question is this: to what extent have these theoretical models been imported into France, particularly in periods following appeals for a return to the contextual approach in electoral analysis? After a brief contextualization of this “renaissance” of the contextual approach in France, a comparison of the theoretical models we have just described and the French scientific works that claim to adopt the same approach should help us scrutinize the revitalizing ambitions of the latter group.

Contextualization of Variables Versus Contextual Demonstration

43Our first reproach concerns studies in the ecological tradition and comes primarily from a comparative perspective. Our assessment is quantitative as well as qualitative. Besides the fact that relatively few studies of this nature have been carried out in France compared to in Anglophone countries, the major difference between the two research traditions lies in the type of demonstration they use. While the British and American studies essentially use contextual demonstration to evaluate the effects of social context on behavior, the French studies primarily try to contextualize the analysis. Semantic subtleties aside, we believe this difference is the more important one, because it ultimately stems from two fundamentally opposed theoretical orientations.

44The French contextualization of analysis should be understood as the desire to resituate behavior in its geographical or historical contexts of production in order to re-examine and refine “decontextualized” theoretical frameworks, i.e. frameworks developed within macrosociological research. The main innovation here is not theoretical. Many ecological studies of voting in France have developed and investigated hypotheses borrowed wholesale from the paradigm of rational choice [47] or the analysis of heavy variables. [48] The key analytical factor in such studies is still the effect of individual characteristics or choices on voting as seen in its spatial dimension. This type of analysis is not properly speaking contextual, because the phenomena studied are not presented as being determined by social interaction.

45Some might see all this as simply the exercise of due caution in the interpretation of available data: as we discussed above, because analysis in the ecological tradition uses aggregated data, the effects of social interaction can only ever form the basis of hypotheses that lack empirical support. But although the contextual hypothesis can be rightly described as empirically unfounded, the same criticism applies to the theory that spatial variation in correlations between demographic characteristics and votes is caused primarily by social status or by strictly individual evaluative criteria. This is not to say we should dismiss the theory altogether: it may be true, for example, that if a manual worker in northern France votes differently from a manual worker in the east or south of the country, that is simply because the category of manual worker in each region refers to different legal or social realities, even if the interactional contexts in each case are not significantly different, or at least not decisive in terms of voting. [49] But because the methodology is the same, we cannot claim that it rests on stronger empirical foundations than the contextual hypothesis. The fact that the contextual hypothesis has not been properly tested in France is not due to methodological or empirical factors. Rather, it is the result of a theoretical choice that, although legitimate, hampers the revitalizing ambitions of the “return to context” because it ultimately produces results that are “orthodox”, to say the least. This is the first part of our paradox: for reasons of theoretical choice, the return to context in France does not generally result in true contextual demonstration, by which is meant the development and/or testing of hypotheses regarding the effect of social interaction on voting.

46The goal set by French political scientists in the early 2000s appears all the more appropriate when we consider that, for them, the theoretical choice seemed clear-cut. That choice was to rely on the work already done in areas studied by the geographical tradition in order to observe, in situ, the interactions that cause the neighborhood effect. Nevertheless, once again the results obtained have several limitations. However—and this is the second part of the paradox—this time it is because of the methodological choices made by the advocates of the contextual approach.

Ethnography and Contextual Analysis

47Qualitative methods, and in particular ethnographic ones, have long been used in the contextual approach to voting, including in American studies. They have been deployed in the service of two fundamental goals: to reflect on which social and interactional conditions are conducive to the politicization of individual relationships and exchanges, [50] and to investigate how collective political preferences form and, above all, are sustained. [51] However, such methods have not yet been used to demonstrate that context has an effect on electoral behavior, or to explicitly test the theoretical models developed by the ecological and interactionist schools. The ambition to do so is, therefore, the most novel aspect of the plan to bring back the contextual approach in France. It is also the aspect that we will attempt to critique in the following section. In what follows, we by no means intend to question the value of ethnographic analysis itself; rather, we would like to discuss certain limitations that arise when it is used to evaluate contextual theoretical models, including the hypothesis that context affects electoral behavior. In our view, this use of the ethnographic approach poses two major problems.

48The first is methodological. The difficulties facing researchers engaged in the ethnography of ordinary political practices have already received plenty of attention and discussion. [52] But what is true at the general level is also true more specifically of contextual analysis: how can we even observe the interactions that constitute contextual effects and isolate them from those that do not, let alone quantify their impact on electoral practices? As the challenges encountered by American ethnographers can attest, there is no simple answer to this question. [53] And there is another difficulty associated with the use of ethnography in contextual analysis in France, where it is used to empirically investigate the neighborhood effects theorized by the ecological tradition: many studies have demonstrated a long-standing and significant decline in neighborhood sociability in both France and in the United States, at least in urban contexts. [54]

49The second problem, which is in part caused by this methodological limitation, is theoretical. Citizens who choose to respond to ethnographic surveys have such specific relationships to politics that is it hard to imagine that they are representative of anyone other than those at the politicized fringes of the areas studied. But besides the question of the statistical representativeness of case studies—a practical problem that can be countered by the claim that analysis never needs to be representative of anything other than the group being analyzed—the major stumbling block for these studies is a theoretical one. The mechanisms of interpersonal influence that can be observed within studied groups do not go beyond a narrow understanding of the model of social cohesion. There is no conversion through conversation here: individuals share the same preferences, deviations from the norm are rare, and conflict is quickly resolved; group life reinforces already firmly-rooted individual opinions. At the opposite end of the scale, ethnographic surveys reveal collective units where the norm is to reject or avoid politics, but little else. It is as if all the possible forms of political action were covered by these two ideal-typical case studies, which hardly seems like a satisfying conclusion.

50For one thing, while individuals’ relationships with politics do seem to conform to the norms and views perceived to be dominant within their groups, we do not know whether this is true for electoral behavior itself. This is especially the case given that we do not know whether we are dealing with genuine conformism or with elective affinities. In fact, this is one of the questions that ethnographic studies fail to answer: is ambient conformism in case study groups the result of social pressure exerted by the group or its leaders over other members, or is it the result of a similarity-attraction effect that would, for example, drive more politicized citizens to form lasting relationships with others who think like them? The distinction is important, because in the second hypothesis the processes observed are not strictly speaking the effects of context: rather, they are the result of similar social statuses and individual psychological mechanisms (for example, the tendency to avoid “conflict”).

51Moreover, ethnographic studies have a persistent blind spot: they never address the fact that the individuals they study inevitably also associate with other groups about which we know nothing—the nature of the relationships individuals have with them, for example, or whether such groups might represent conflicting influences that could counteract the conformism observed in the study area. Social psychology draws an important distinction here: the conformism individuals display within a group may be caused by real changes of attitude or opinion or by the reinforcement of existing attitudes or opinions—in short, some sort of effect on an individual mental state—, but alternatively it may be the manifestation of a strategy adopted by individuals in order to present (in Erving Goffman’s dramaturgical sense) their own convictions to the world. In other words, conformism may be nothing more than a show put on in order to avoid coming into conflict with the perceived majority view and unsettling an individual’s more general relationships with the group or a part of the group. Thus, while most empirical case studies correspond in an almost ideal-typical way to the primary cohesive groups theorized by Lazarsfeld, it is striking that the opposite hypothesis of politically heterogeneous groups that have a destabilizing effect on the individual—which is also represented in Lazarsfeld’s work by the concept of cross pressures—has never been empirically investigated. This is the basis of our second reproach: if we take into account what has been observed and theorized up to now using other survey methods, and if we accept that the influence of context on individuals must necessarily be complex and variable, we must then admit that data obtained through ethnographic methods cannot be used to make general statements about all individuals or all groups.

52Direct observation of the interactions underlying contextual effects is undeniably useful. French studies have made at least two major contributions in this respect: they describe the content of the interactions that might underlie contextual effects, and they try to define contextual effects in terms of the social resources of the individuals and groups actually observed. In contrast, British and American analyses using quantitative surveys tend to generalize the electoral impact of interactions with regard to all social groups. Nevertheless, French studies are hampered by empirical and theoretical difficulties all their own, and it is impossible to justify the neglect within French academia—the third paradox we would like to discuss—of other methods or conceptual tools that could open up new avenues of research into the phenomenon of context.

An Alternative Conceptual Framework to Generate New Perspectives?

53The failure to adopt the concept of social networks seems both questionable and damaging. Questionable, because the rejection is ultimately rooted in a theoretical—structuralist—bias. In fact, use of the concept of networks implies a methodological position—emphasizing interpersonal (sometimes “transgroup”) relationships over well-defined groups—more than a specific theoretical one. Networks are, therefore, equally applicable to structuralist and individualist models, and their underlying assumptions can justify the use of a wide variety of methodological tools, both quantitative (snowball sampling, main respondent surveys, econometrics) and qualitative (ethnographic analysis, [55] snowball interviews).

54And damaging, because in practice the concept of social networks covers such a broad understanding of democracy and voters that it can reveal aspects of contextual analysis that have scarcely been touched on in France. Using a few examples, we would like to demonstrate the heuristic potential of the concept of networks when used in the contextual approach to voting.

55Studies of the degree of heterogeneity within networks, or in other words the possibilities for disagreement available within them, aim to develop theoretical and methodological tools capable of re-examining the old ideal of a deliberative democracy from a contextual perspective by looking at the most mundane areas of social activity. [56] They also try to address important phenomena such as the rise in voting abstention and electoral volatility, but they do not provide definitive or unambiguous answers to such questions. They paint a rather more nuanced picture of democratic activity than studies of groups or areas with clearly defined borders, which are generally politicized, socially and politically homogeneous, and overwhelmingly conformist. The latter seem out of their depth when it comes to explaining the phenomena mentioned above. If individuals can belong to several types of groups connected through the networks they associate with, then the potentially varied impacts of those multiple forms of belonging must be properly investigated.

56The theory of deliberation generates potentially interesting hypotheses that are completely unrelated to the figure of the rational voter. Investigating the political and social structure of networks raises topics such as cognitive dissonance, a question that is almost exclusive to social psychology but that we know is important when it comes to electoral behavior, or the cross pressures identified a long time ago by Lazarsfeld but still never examined by French political scientists. This is despite the fact that individuals being exposed to conflicting influences from the various groups they belong to was presented in The People’s Choice as one of the drivers of indecision, instability, and electoral abstention. [57] The same theory appears in the work of Diana Mutz, which has no equivalent in France. In Hearing the Other Side, she does not challenge the idea that networks of sociability might be more heterogeneous than the theory of social cohesion suggests. Instead, she draws attention to an important theoretical and practical dilemma: although democracies based on ideologically closed groups have citizens who are less good at exchanging conflicting opinions, they are also more participatory than democracies where citizens have been made tolerant by regular encounters with political alternatives. This is because, while social pressure tends to push citizens towards stable, deeply entrenched preferences in the former, the proliferation of cross pressures makes them more indecisive and fickle in the latter. [58]

57Another question raised by network analysis—that of how individuals use context—is also deserving of more attention from French researchers. Although the effects on electoral behavior of socioeconomic status or education have been extensively evaluated in France since the early days of electoral science, the same is not true of social capital. Networks are essentially sources of information and resources for individuals. [59] Whether or not individuals use them in a rational and utilitarian way, it does not seem unrealistic to suggest that such information and resources would influence their behavior, including electoral behavior. Is the figure of the “expert” that appears in several American studies [60] as someone to whom individuals in a social circle go for information not at least partly related to Lazarsfeld’s concept of opinion leaders, which is regularly cited but rarely empirically investigated in French studies?

58But the concept of networks also has applications in other frameworks, including the study of internal relations within “closed” groups. In France, the most convincing work on the contextual analysis of voting—and more specifically of the influence of familial ties on electoral participation—has not come from ethnographic analysis, but rather from studies of quantitatively identifiable interpersonal links, for example using electoral registers. [61] Another topic that can be approached via the concept of networks is online relationships, in that their influence on electoral behavior has rarely been looked at from a contextual perspective. Finally, the last twenty years have seen a growing number of studies in the United States looking at how effective canvassing by activist networks is in terms of increasing electoral participation. [62] Such studies, which put experimental approaches into practice in the field, have so far yet to be repeated in France, with the notable exception of the experiment on the topic of voter registration by Braconnier, Dormagen, and Vincent Pons, [63] although their focus was specifically on schemes that encourage registration rather than on the influence of social context on electoral behavior. These studies operationalizing the concept of networks offer new research perspectives that French political scientists have still not begun to investigate.

59* * *

60In Lazarsfeld’s work, the concept of context was mobilized in the service of two distinct goals.

61First, to clarify, or rather flesh out, the concept of class-based voting: taking context into account shows that the ways in which individual interests translate into votes are not just determined by objectively and subjectively perceived social status; they are also influenced by everyday interactions with the mostly homogeneous (and perceived as such) reference groups to which individuals belong and/or against which they position and define themselves.

62Second, to show how context may counteract these processes: interactions sometimes take place within non-homogeneous groups or between individuals from opposing groups, without necessarily provoking positive reactions (the reinforcement of existing opinions or identities).

63In France, while the representatives of the ecological tradition basically treat context as a scale variable, the successors of the interactionist tradition seem to have forgotten about the second idea just as it starts to receive more attention among their British and American counterparts. In that sense, the French return to the contextual approach to electoral behavior restricts its own initial ambitions in two ways. It neglects a whole range of theoretical, conceptual, and methodological tools that deserve to be tested and discussed—especially because they have not been investigated before now. It also deprives itself of a basis for dialog with other research traditions. After more than half a century of research into voting, the revitalization of electoral studies and its main paradigms can only be brought about by the systematic opening up of this sort of dialog.