CAIRN-INT.INFO : International Edition

1Employing the methodology of focus groups, this research takes as its point of departure a clearly observable contradiction that resides within the general public: that between the strength of disapproval following the revelation of breaches of integrity and abuses of power by political leaders, on the one hand, and the weakness of the political effect of such accusations, on the other hand, particularly in electoral dynamics. By occasionally re-electing politicians suspected of breaches of integrity, voters show that the value they attribute to norms supposedly defining the right political behavior is at very least ambiguous. At present, the thesis most often employed to explain this democratic paradox points to the existence of normative conflicts, defined as the coexistence of different criteria that are used by citizens to judge politics. Citizens bring moral or even legal principles to bear on a host of other normative referents related to politics. [1] Hence, in their appraisal of elected officials and leaders, citizens attribute considerable importance to pragmatic factors (the effectiveness of local implementation) or symbolic factors (the ability to represent the collectivity). Such factors determine citizens’ moral principles, and their combination translates into diverse forms of tolerance and pardon. Consequently, in the context of political judgments, how should we seek to comprehend this ambiguity towards breaches of conduct?

2The research model for the present study was implemented through a Cevipof group project, [2] and contains three components combining qualitative methods (position papers and focus groups) and quantitative methods. Our focus group approach thus dovetails with other “mixedmethod” research. [3] This approach provided three main advantages: it allowed for the “triangulation” of results, whereby the results of one method helped to test and validate a portion of the results of another; it ensured “complementarity” between the results, whereby those resulting from one approach built upon the results of another; and finally, this approach allowed for “development”, whereby the general research question evolved in accordance with the output of results, the research having been undertaken in sequences composed of several methods.

3This article is centered on the second component of this research, based on the use of focus groups. In the sociological endeavor to better understand the complexity of forms of political judgment, the contribution of qualitative research is undeniable, particularly when conducted using structured focus groups. Such research allows for the exploration of areas of judgment in which heterogeneous norms and values – often poorly connected to one another and sometimes contradictory – clash together. Tellingly, this methodology has often been employed to explore and to clarify the normative conflicts related to the ethics of healthcare or in the domain of sexuality. [4] During group discussions, the problems raised put personal choices and the autonomous reasoning underlying them, into contrast with issues of a public nature that concern collective morals. Adopting this framework, this article works toward three goals. It aims to demonstrate the utility of a focus group approach to understanding normative, and particularly moral, dimensions of the way people relate to politics. It supports the idea that focus groups constitute an “apparatus” [dispositif] which – by stimulating agreement and disagreement within discussion – brings forth the arguments constitutive of the very normative grammar that individuals employ in their relationship to the political. Finally, this article promotes a new quantitative methodology of focus group analysis, which allows us to shed light upon forms of political judgment and their sociopolitical groundings, in a way that is more elegant than traditional data decomposition methods.

4To demonstrate the full relevance of a focus group approach to the study of perceptions and judgments of political transgressions, we proceed in four stages. In the first part, we show the relevance of this approach with regard to more traditional quantitative analyses. We clarify the design of our study and the originality of our approach to data treatment, adopting firstly qualitative methods and then quantitative ones. Situating the study within a sociology of judgment approach, which emphasizes the diversity of perceptions and argument justifications, we bring to light the central contributions of our use of focus groups. In a second phase, we analyze the judgments put forth by participants, looking not only at their content and the order in which they are used, but also at how the discussion dynamics alter the severity of participants’ evaluations. The data thus collected enables us, on the one hand, to constitute a relatively stable system of measurement (tables for qualifying and classifying judgments) based on the commonalities between groups, and on the other hand, to identify certain mechanisms that generally moderate the collective discussion. A third phase takes into consideration the modulations and variations in the way judgments are expressed, and relates these to the impact of the groups’ different social characteristics. Finally, we analyze the reasoning and justifications developed by the focus groups. We show that it is possible to identify argumentative structures that are common to all groups (“argumentation structures”), [5] whose differing combinations constitute so-called “argumentative repertoires”, some of which are more idiosyncratic to one social group’s forms of judgment than to those of another social group.

The renewed contribution of focus groups in understanding the moral dimension of political judgment

The limits of questionnaire-based studies

5The general purpose of our research is to describe and analyze the normative and moral dimensions of the way we relate to politics, i.e. the form of judgments held by ordinary citizens regarding the behavior and integrity of elected officials, but also, by extension, regarding that of their fellow citizens. Concerning the political activity and behavior of elected officials, individuals often formulate normative expectations and criticisms by insisting on notions of integrity, duties and obligations. Such expectations and criticisms are then expressed through polarized categories of “good/bad”, “just/unjust”, “honest/dishonest”. To construct such judgments, ordinary citizens make use of diverse resources – generalized representations of politics, personal principles, legal categories that are more or less accurate, and personal experiences. The attempt, using inductive methods, to lay bare the principles of classification, the categories and the arguments employed by citizens when judging politics presents obvious methodological problems. Such an attempt constitutes nothing less than the identification and analysis of the normative dimensions of our relation to politics. How can we isolate the moral principles applied to politics in light of the fact that, as François Dubet has put it, “if social agents are embedded within moral frameworks, the latter still do not totally determine their judgments”? [6] In other words, the way we relate to politics is shaped through a great many more parameters than the simple requirement that elected officials be moral. The judgments held by citizens pertaining to the integrity of their elected officials belong within the much vaster framework of the relation that exists between those who govern and those who are governed.

6On the basis of quantitative questionnaire-based studies, American and Canadian research has focused on ordinary citizens’ perceptions of integrity and corruption in elected officials. These studies emphasize the complexity and ambivalence in citizens’ judgments of what they consider “good” or “bad” political behavior. [7] They also show the importance of different kinds of breaches of integrity in the harshness of judgments. [8] And yet, as revelatory as such approaches are, they nevertheless present real limits. First of all, the empirical data provided by survey techniques is insufficient if the aim is to understand the way in which respondents construct and mobilize normative systems. The use of scenarios in questionnaires makes it possible to record appraisals of severity (or tolerance) in reference to different kinds of deeds, but it does not aid in the detection of the justifications underlying these judgments. Studies using questionnaires do, however, point to the existence of judgmental heuristics, linked to the way in which perception of scenarios varies according to the political practices under consideration. [9]

7Nevertheless, while the combinatorial analysis of questionnaire items allows one to identify regularities and coherencies that give shape to stable attitude systems, this type of study does not give access to the systems of argumentation developed by individuals. What’s more, in the short time it takes to fill out the questionnaire, judgment is “instantaneous” and “compressed”, making it impossible for the participants to explicate, to nuance, or to rethink their initial categories of judgment. In that questionnaires are based on declarative statements, they are also likely to privilege the expression of more normative judgments, which heighten the denunciatory posture. [10] This procedure rarely manages to shed light on potential discords between attitudes, on the proximity of disapproving judgments to more tolerant approaches, or on the distinction between perception and severity of judgment; these can only be explained through a contextualization of the judgment and by taking into account an assortment of variables. [11] From this point of view, quantitative questionnaire-based approaches can only link opinions with explicative variables measured by closed questions regarding gender, age, socioprofessional milieu, and political orientation, etc.

8Thus, the results of questionnaire-based studies – those undertaken in the United States, Canada, or in Japan, [12] such as the Cevipof survey of 2007 – show that, while the relation of such judgments to social conditions or political orientations is not one of total independence, it is nevertheless a link made tenuous and complex by the broader meaning that individuals vest in their relation to the political; it is a link constructed through a series of experiences and processes of socialization (be they political, professional, social, economic, or other). In one sense, it is hardly surprising that questionnaire-based studies primarily focus on the way in which the properties specific to the judged acts (which are different according to the scenarios presented) influence the variation of individuals’ perception and judgment: the model encourages (in a heuristic way) the detection of these first-level judgments. The contextualized sociopolitical groundings of the “ethics maps” brought to light by these studies are trickier to identify and to analyze.

The contributions of a focus group approach

9Our focus group approach aims to record ordinary discourse induced by group interaction using a simple stimulus: a scenario submitted to the group for evaluation. [13] The advantage of this methodological tool is that it significantly reduces the risk of the researcher imposing his or her own preconceptions, both when writing the questionnaire and when conducting the interviews. [14] First and foremost, it also enables one to pay special attention to the structures of argumentation and to the “systems of reasoning” that aid individuals to bring meaning to [15] and to judge political behavior. Such assorted modes of reasoning, reflective of social actors’ norms, are far from constituting a rational and perfectly ordered system that operationalizes, in some way, overarching principles of political morality. On the contrary, individuals develop schemas of perception and political judgment that superimpose different and often contradictory registers. As Dubet has described the situation regarding injustices at work, “each person behaves as a theorist, not because he is capable of condemning injustice, but because his criticism is always associated with an argumentation structure, composed of principles perceived as more or less universal, and subject to the constraint of coherence and reciprocity”. [16] When individuals judge politics, they produce arguments and lines of reasoning that are more or less developed, and which they use to bring meaning to their relationship to those who govern, be they relatively far-removed figures (executive or legislative) or local officials that are more easily discernible. In so doing, they create and use “commonplace theories” of representative government in order tomake sense of their relationship to the political. Such constructions compile social, professional, and political experiences, as well as previously assimilated normative principles, bringing these to bear on the situations under judgment.

10To gather such material, we organized eleven different discussion groups, homogenous in socioprofessional status, [17] in two waves (before and after the quantitative study). These groups, which brought together five to eight people previously unknown to one another, were organized around a collective “exercise” and moderated in a very focused manner. The exercise involved discussing a series of brief scenarios in which the principal actors were at times elected officials, and at times citizens. Such scenarios are employed in different types of studies as a way of inciting respondents’ capacity for reasoning by placing them “in a virtual deliberative situation”. [18] Before asking people to formulate judgments (here in terms of the seriousness of the situation; in other studies as a preference ranking exercise), the scenario proffers a condensed form of information that enables one to establish the different aspects of the issue under investigation.

11The actions presented in the scenarios oscillate between extreme cases that, in principle, touch the limits of the range of proposed choices: they go from the most common behaviors (political namedropping in order to obtain a place in a daycare center) to legally defined forms of corruption. [19] The other situations proposed in the focus groups deliberately contain information that is likely, depending on the recipient, to lead to an intensification/attenuation of the prevailing judgments (the type of perpetrator, the level of the advantages obtained, extenuating circumstances). In contrast to American or Canadian questionnaire-based studies, which favor situations involving national-level officials (the President, members of the Ministerial cabinet, Senators, etc.), the cases treated in the scenarios of our study arose from examples of “neighborhood politics” [politique de proximité] predominantly involving local officials.

Conducting focus groups

The groups were always led by the same two people and took place in three phases. First, after a presentation of the study, each person classified the proposed scenarios individually into four categories according to a crescendo of increasing seriousness: from “not very serious” to “very serious”. The scores were then compiled scenario-by-scenario on a board (without indicating which participant gave which score). Finally, collective discussion was brought to bear on the final result, scenario-by-scenario, without fixating on individual choices. The idea was to make explicit the arguments undergirding agreements and disagreements regarding the final scores. Initially, four exploratory groups were developed. They helped to refine the approach and served to prepare the questionnaire used during the quantitative phase. Seven other groups were subsequently developed, differentiated through their socioprofessional composition. [20] We thus observed, videotaped, and analyzed the different phases of the group interviews: the work of individually classifying (in four categories [21]) the sixteen scenarios figuring on the cards, [22] followed by the collective discussion of results for each of the scenarios, in which participants were asked to justify and explain their choices. The verbal exchanges first focused on situations featuring higher levels of consensus regarding the degree of seriousness, then on those in which there was dissension. Because the participants had to justify the degrees of seriousness accorded to each situation, it was possible to explore which criteria of perception and judgment were employed to demarcate acts likely to be considered corrupt from the others. Responses were first revisited with the aim of discussing mitigating factors that might justify the acts described in the scenarios, then, in the cases of a strong consensus, counter-arguments were discussed. More general questions about people’s experience of and relation to politics were always asked following certain scenarios in order to broaden the discussion. Each collective interview lasted approximately three hours.

12What are the primary virtues of this method of inquiry and what do we expect it to achieve? Starting from concrete cases potentially similar to an experience that the participants may have had, we forced them to avoid the usual stereotypes by not only making them differentiate and hierarchize situations, but also by making them justify their judgments. The collective interview here reveals itself to be doubly useful. On the one hand, it encourages individuals to express a set of judgments and reiterated justifications, given shape through the pronouncement of categorizations, principles, arguments, systems of reasoning, and good reasons. On the other hand, the collective situation brings forth a deliberative process in which arguments are exchanged in order to make clearer a decision (in this case, an appraisal of seriousness). The focus group gives interviewees the time to develop their points of view, rather than addressing the subject abruptly. Group exchanges create a dynamic conducive to discussions that often become quite lively. This dynamic encourages participants to justify their judgments, makes possible the involvement of each person, and can even help to reveal practices considered deviant (attempts at personal gain, petty fraud, inaction in the face of rule-breaking). Without going as far as to force it out of them, people are sometimes led to reveal their true opinions. The collective interview thereby allows for the observation of social interactions between individuals discussing a variety of topics at length: the meaning of scenarios; whether or not the types of behavior described are “commonplace” in the context of everyday political activity; what can be deduced from this, and the different judgments that this behavior evokes. The collective interview bears evidence of how firm agreement or disagreement is, and lets us test “resistance” against the attenuating effects of discussion. It also makes it possible for judgments to evolve between the initial classification of the scenario and the phase in which the latter is justified and defended through argument. By facilitating participants’ reflection on sensitive subjects by means of several concrete cases, the collective interview can go beyond the expression of dominant and univocal normative points of view. The artificially constructed situation of judgment and justification incites participants to eschew stereotypes – i.e. “they’re all corrupt” – by obliging them to sustain arguments throughout the duration of the study, and by presenting a range of situations differentiated by the facts, the actor and issues concerned, context, and so on. Individuals thereby establish differences and nuances in their judgment. Confronted with different positions and also with their own contradictions brought up in the course of discussion, they seek to bring some sort of coherence to their arguments.

13With respect to existing empirical studies using group interviews, our approach seeks neither to draw out shared meanings common to all groups, which would be the case of a cultural model proper to society at large, nor to analyze the dynamics of interaction between participants as such, even though such dynamics and their self-compounding effects do come under observation. We seek to understand how, over the course of the discussion, individuals belonging to similar socioprofessional milieus reason, through a process that combines different principles, rules, symbols, images and subjective experiences with their interpretation of the situations proposed in the study. Following the approach adopted by François Dubet, [23] we attempt to draw out the basic elements of a syntax – a vocabulary, a grammar – that composes individuals’ normative stance toward the political. We therefore concentrate on each of the groups in its entirety, centering the analysis on arguments and systems of argumentation expressed collectively within the group. In so doing, we are careful to characterize any eventual specificity in the content, comparing groups with one another. In this approach, closer to the works of William Gamson [24] than to those of Elizabeth Frazer [25] or of Michael Billig, [26] it is a question of identifying the “casuistry” through which members of groups judge politics. The latter assures the harmony between a handful of principles (see below) and a diverse set of judgments. This approach does not posit that the structures of arguments exchanged within a group are homogenous, but helps rather to identify regularities internal to the arguments. The occurrence of such regularities can be explained by participants’ shared experiences (social, political, or professional) that link back to comparable stock arguments and to similar “structural conditions” such as their social position, career path, socialization, etc.

14Nevertheless, the fact remains that this methodology proffers the same scenarios for discussion by all of the groups, which is to say that it obliges participants to judge and argue along the same continuum of more or less controversial political practices. By comparing the different argumentative systems, this methodology allows for the identification of inter-group regularities between opinions and judgments (notably with regard to practices that are unanimously denounced or tolerated), and between the arguments and argumentative systems that are employed. We thus speak of “points of agreement” or of minimal “normative consensuses” [27] pertaining to condemned or accepted practices. In this vein, it was even possible, to an extent, to confront the results of the qualitative focus group study with those of the quantitative study, [28] as the latter involved the evaluation of a number of scenarios common to both studies. Such regularities can be explained by the fact that, while it is true that the normative activity of individuals is based on a casuistry that draws on each person’s specific social, political and professional experiences, it is nevertheless possible to identify a number of recurrent arguments, as well as structuring principles that are stable and that all of the individuals and groups employ, even if they combine them differently.

New quantitative methods for the analysis of qualitative data

15The first use of focus groups was based on standardized transcriptions of the discussions, which featured all of the spoken enunciations as well as the structuring elements of the conversation (laughing, silences, hesitations and so forth). These focus groups, carried out in an iterative manner, gave rise to the first effort of codification and to the identification of principles, arguments and repertoires. [29] Though it may be heuristic, this qualitative method of proceeding has a certain number of limitations: the degree of judgmental severity is insufficiently measured; arguments are not apprehended systematically, and the temporal dimension of arguments is left out. Through recourse to a complementary method of innovative quantitative treatment, such limitations are surpassed and the analysis can be considerably enriched.

16Three objectives guided the conception of the quantitative methodology of analysis employed throughout the study. Firstly, we sought to compile indicators that would allow for an evaluation of participants’ judgmental severity. Secondly, we sought to assess the proportional frequency with which participants used particular arguments to back up such judgments. Finally, using these indicators, the third objective of this study attempted to weigh up judgmental severity and argumentative strategy according to different external variables (such as the socioprofessional status) or variables relative to the proposed scenarios (such as the seriousness of the situation, or the nature of the protagonist – a politician as opposed to a citizen, etc.).

17Bearing these three imperatives in mind, our strategy was to put in place two data treatment procedures (one focusing on judgment, the other on argumentation), which would both be suited to the analysis of semantic content and to quantitative methods. The first procedure provided an index for judgmental severity based on a measure of the time devoted to argumentation during the discussion. It was created by calculating the time spent denouncing kinds of behavior relative to the total time spent on argumentation (which is to say the time spent either denouncing or justifying conducts). The second procedure deals with argumentative analysis and is presented in the fourth section of the article. It is based on a process of identification (codification, categorization, classification) and the tallying of arguments. This procedure then uses factorial analysis to extract the associations between arguments, as spontaneously produced by the participants during the discussion. It thereby enables the calculation of the number of arguments relative to the types of arguments thus constituted. [30]

The creation of an analytical procedure for quantifying judgmental severity

The transcription of the total corpus identified 3,321 enunciations (405,542 characters). To facilitate the analysis, the first phase of our procedure involved defining a “unit of meaning”. [31] By “unit of meaning” we mean a word or an enunciation in which one kind of judgment is expressed (either justifying or denouncing the practices presented by the scenario) measuring the seriousness of the situation. Subsequently, our task involved categorizing each enunciation according to the orientation of the judgment expressed therein. Enunciations or segments of enunciations that did not contain any kind of judgment were eschewed from analysis, as were those of the discussion moderators (cf. Annex 2). Having analyzed enunciations according to this procedure, we identified some 1,887 units of meaning, which were subsequently classed into one of two categories, “serious” or “not serious”.
The second phase of our analysis involved the measurement of time spent by participants upon arguing for or against certain kinds of behavior. In order to do so, we measured the units of meaning previously identified. In a preliminary study, we illustrated that the duration (in seconds) closely correlated to the number of characters in the transcription: r (141) = .97, p .0001. [32] This led us to adopt the latter unit of measurement, much better suited to the task at hand. We thereby calculated the number of characters of all of the units of meaning from our transcription (cf. Annex 2), having carefully removed all commentary and explanation relating to the events. The total number of characters of these 1,887 units of meaning was 289,357 (71.4% of the total number of characters; A = 153 characters per unit of meaning, SD = 199). [33]
For different contexts (a given scenario, group, participant, and so forth) we could therefore provide the number of characters associated with the justification or denouncement of the behavior under discussion. Finally, our indicator of the discussion’s severity was obtained by calculating the percentage of severe enunciations (denouncing the behavior of the situation’s protagonists) in relation to the total number of characters of all argumentative enunciations (either justifying or denouncing practices). A value of 50% corresponds to a time spent denouncing equal to the time spent justifying: in other words, to a balanced argument. A value greater than 50% corresponds to a greater time spent denouncing than spent justifying, or to a generally severe argumentation. Symmetrically, a value less than 50% corresponds to less time spent denouncing than spent justifying, or to a generally tolerant argumentation.

18One of the advantages of our focus group study is that because it is situated within a broader research agenda, it lends itself to the comparison of results with other methodologies. So for example, the quantitative study [34] and the focus groups logically share the same material (six scenarios in common). During this study, participants were put into face-to-face encounters and led to evaluate the seriousness of scenarios by situating them on a scale from 1 (“not serious”) to 4 (“very serious”). In addition, at the beginning of the focus groups, we decided to submit all participants to an identical task: they were asked to individually evaluate the seriousness of the scenarios (subsequently discussed by the group) by situating them on the very same scale as that used in the quantitative study.

19We therefore disposed of two further indicators of judgmental severity: an initial individual index of judgmental severity (from the beginning of the focus groups) and the index of judgmental severity from the quantitative study. In order to facilitate the comparison of these indexes with that of the discussion, we transformed them into percentages. [35] The responses ranged from 0% (“not serious”) to 100% (very serious), whereby a value of 50% corresponds to the center of the scale, situated between “not very serious” (33%) and “serious” (67%). Bringing together these three indicators of severity (by triangulating data) notably enabled us to guarantee the reliability of our model. [36]

Consistency of judgments and dynamics of deliberation

20Schematism too often prevails in analyses of how breaches of integrity and “corruption” are perceived, wherein attention is brought to bear primarily on very generalizing proclamations such as “They’re all corrupt” [37], despite the fact that such statements properly belong only to stereotypes and soapboxes. Analysis of the data that comes firstly from the eleven focus groups, and then from the eight groups in which we carried out in-depth enquiries, led to the fleshing-out of nuanced arguments and enabled us to specify exactly which factors varied. Three preliminary findings were made clear. First of all, regardless of the moment in which judgments were made in relation to the given scenario, [38] the latter were ranked into an order ranging from attributions of significant severity to significant tolerance, and this order of rank is stable. All of the participants thereby demonstrated forms of political and moral competency, [39] in other words, of the capacity to differentiate between situations and rank them in a hierarchy. Second, we empirically demonstrate that participants’ judgments were consistent; the elaboration of arguments during discussion is correlated with the initial ranking. Finally, by comparing, for a given scenario, the initial rankings with the results of the quantitative study and the analysis of argumentative exchange during discussion, it is clearly apparent that the discussion contributes to the evolution of the initial judgments, be they disapproval or expressions of tolerance.

Differentiated and ordered judgments

21As described above, the collection of eight scenarios retained for quantitative analysis was put together using two criteria. On the one hand, we marked out the extreme limits of the potentially deviant situations by retaining situations that ran from simple everyday favoritism to qualified corruption. Having consulted with colleagues expert in the matter, we retained two situations theoretically considered as “not very serious”. They are related to local-level clientelism and the financial stakes are paltry. [40] Two other scenarios, however, seemed to us to be theoretically “very serious”, because they involved cases of juridically defined corruption (the existence of a conspiracy to commit corruption) and included very significant personal gain. [41] On the other hand, for the remaining four scenarios we did not hold any a priori hypotheses pertaining to the potential degree of seriousness they might be attributed. We intentionally included elements that helped vary both the type of protagonist (citizen / elected official) and the justification of the transgressions (finding work, housing).

22The data thus gathered makes possible the observation of regularities in forms of judgment. The diversity of judgments and their ordering into three coherent groups indicates first and foremost that the criteria of variation that we have retained were pertinent in that they allowed for the differentiation of reactions. Accordingly, our study also validates participants’ capacity to order situations in a stable manner. Indeed, the initial graduation of judgments (formed through individual votes without prior discussion) is confirmed in the analysis of arguments exchanged through interaction. Given that our material only relates to eight different scenarios, it would be somewhat rash to extrapolate any sort of general hierarchy of cases from this. That being so, we can nevertheless validate the existence of three kinds of situations, regardless of the group concerned: a first type in which judgments of very slight severity prevail (two cases in which the initial average severity index is only 16%); a second in which the controversial cases are to be found (three cases in which the initial average severity index is 47%); and a third, in which severe judgments predominate (three cases whose initial average severity index is 82%).

Figure 1

Index of initial severity by socioprofessional status and the seriousness of the scenario

Figure 1

Index of initial severity by socioprofessional status and the seriousness of the scenario

23These results are confirmed by comparison with results from the quantitative study relative to the judgment of the six common scenarios. [42] Here once more we find equivalent differences between the three groups, with, however, elevated severity in the case of the quantitative study: the global index of severity for all three scenarios is 58% for the quantitative, 52% for the focus groups (cf. Table 1 below). For the type 1 scenario (not or not very serious) the average index is 35%; for type 2 (serious and very serious), the average index is 81%; finally, for the third type (controversial), the average index is 57%. Between the two studies, while the difference between the scenarios unanimously judged to be not very serious or to be controversial is relatively negligible, that between the scenarios unanimously judged to be very serious is significant, representing some 17%. The elevated degree of seriousness attributed in the quantitative study can be explained through three factors. First of all, the evaluation of scenarios took place after the questionnaire had run its course, and in particular directly after a series of questions on opinions about political corruption (potentially provoking a cumulative effect). Secondly, attributing gravity to breaches of integrity corresponds to the socially expected judgment (an expression of political correctness). Finally, the context of the enunciation differs: in one, an instantaneous individual response face-to-face to an interviewer; in the other, an individual response in a collective situation without any immediate justification of the choice.

Table 1

Comparison of types of judgment according to the quantitative and qualitative studies and to the degree of seriousness of the scenarios

Table 1
Degree of seriousness attributed to scenarios Consensus: “not very serious” Disagreement on seriousness Consensus: “very serious” Quantitative study 2,012 people 35% 57% 81% Qualitative study 8 groups 32% 53% 64%

Comparison of types of judgment according to the quantitative and qualitative studies and to the degree of seriousness of the scenarios

24Our results thus confirm Heidenheimer’s hypothesis on the existence of three categories of corruption: “white”, “grey”, and “black”. [43] His analytical model is culturalist, and shows the difference between the perceptions held by the political elite and those held by ordinary citizens. “White corruption” corresponds to situations wherein opinion between the two groups is united to mitigate certain transgressions (financing of politics); “black corruption”, on the other hand, designates situations wherein the opinion is united in its denunciation of practices (personal enrichment); finally, “grey corruption” designates situations that generate disagreement (fringe benefits, conflicts of interest). In our study, the existence of a common ranking is related to a social and cognitive phenomenon. The frameworks employed to qualify and classify acts are relatively homogenous (from the most serious to the most tolerable), regardless of the social or professional group concerned. Such groups cannot be differentiated through their recourse to different scales, but rather via the variation of the scales that demarcate for each the intensity of disapproval or of tolerance. [44] The study also validates the findings of Peters and Welch and those of Jackson and Smith on the differences in judgment regarding elected officials and citizens. [45] As these authors have already shown, for equivalent matters, elected officials are always judged more severely: their index of initial severity is generally 55% (across all cases and groups) and only 48% for citizens. The divide is greater for the scenarios where the “not very serious” consensus prevails (48% for elected officials, 30% for citizens). It is also higher for “serious” scenarios (68% for elected officials, 57% for citizens). The difference is very slight for those scenarios that are controversial (52% for elected officials, 55% for citizens).

Consistency of judgments

25The measurements and calculations that we have effected allow us to highlight a significant linear correlation between the initial evaluation of the scenario made by each of the members of the group before any interaction takes place, and the time spent on denouncing this scenario during the discussion. Two findings were made. On the one hand, the more severe the initial judgment, so too the time spent upon arguing in favor of this position will be greater. [46] On the other hand, Figure 2 shows that, the more severe the initial judgment so too, during the discussion, the ratio of time spent on disapproval relative to the total time of the exchange (our index of discussion severity) will be greater: r (40) = .68, p .0001. [47]

Figure 2

Array of points denoting the linear relations between the initial severity index and the discussion severity index

Figure 2

Array of points denoting the linear relations between the initial severity index and the discussion severity index

Note: the higher the index, the more severely the scenario is judged. The dotted line corresponds to a perfect correlation that is also perfectly linear. The other line (solid) corresponds to the regression slope (line of least squares) produced by the data. The individual points correspond to the focus groups of the same social group evaluating the same scenario, with the x axis corresponding to their average initial severity, the y axis to their average severity during the discussion.

Deliberation: a moment for the evolution of judgments and for the expression of ambivalence

26Another result relates to the dynamics of discussion. Figure 3 shows that the average judgment (across all groups and all scenarios) is situated more or less at the centre of the scale. Judgments of approval and disapproval are quantitatively equivalent. This validates our calibration of the scenarios’ different levels of seriousness, carried out during the selection phase. The most interesting result, however, concerns the specific effects of group discussion on the participants’ judgment. Thus, in previous work, effects of a polarizing nature are most commonly observed. [48] Here, on the other hand, regardless of the relative severity or tolerance of the initial judgment, the discussion appears to moderate this judgment. This result is observable across all groups (in the following sections, we shall bring further nuance to this point). Situations that were initially judged on average to be less serious were subject to harsher appraisals. Inversely, those for which a severe judgment was initially expressed were subsequently judged with greater tolerance. Following Cass R. Sunstein, [49] we hypothesize that the moderation (in both directions) evidenced in discussion is attributable to the expression of commonly shared and persuasive arguments, which lead in opposite directions. [50] This result is consistent with those arrived at using alternative methods that have demonstrated, in the case of judgments relating to breaches of integrity, the coexistence and tension between divergent principles and arguments. [51] Here, the discussion setting appears to allow for the expression of such ambivalence. All the same, one may also note that interaction only very slightly modifies the situations of disagreement, which are often reinforced through discourse.

Figure 3

Severity of the discussion versus that of the initial judgment according to the scenario and its degree of seriousness

Figure 3

Severity of the discussion versus that of the initial judgment according to the scenario and its degree of seriousness

27The most significant example is that of the “box of chocolates” (scenario #2). Overwhelmingly considered as “not very serious” during the individual votes (the initial severity index averaging at only 12%), the collective confrontation of arguments during discussion produced a clear hardening of participants’ judgments (the discussion severity index averaged 48%). This act, initially perceived as inoffensive or even as having a positive connotation (a Mother’s Day gift), is quickly reconsidered through the lens of clientelism. The elected official’s gesture is made worthy of criticism not only due to its suspected political purpose, but also due to its cost (the taxpayer pays for the official’s self-promotion). This act was also sometimes delegitimized because it could be seen as discriminatory (in relation to women without children); or due to the historical links between la fête des mères [mother’s day] and the Vichy regime. Finally, one can also note a snowball effect, whereby one person’s objections to the tolerance of a given act are generally enough to incite the rest of the group to follow their opinion. An equivalent dynamic is observable for the scenario that is always the most harshly condemned (the mayor who lies under oath in order to save the political career of a friend, in scenario #7). Initially, the scenario is judged very serious at 95%, but the discussion attenuates this condemnation by some 30% (judgment reaches 68%). The initial judgment is based on three essential elements: its illegal nature (lying under oath), the compromised dignitas of the elected official (lying is improper to his role) and the dubious purpose (saving a political career). Over the course of the interaction, other aspects of the situation are brought to light and help moderate the judgment: specifically, the importance of friendship and the banality of this lie given the context of everyday politics.

The effect of social status on variations in the severity of judgments

28Yet another advantage of the focus groups, as compared with the quantitative studies, resides in the differences they make apparent between groups of diverse socioprofessional status. The comparison of the data collected amongst different groups is demonstrative of just how fruitful the approach can be. In this light, bringing these findings to bear upon those of the quantitative study is instructive. In the quantitative study, the effects of social statuses and inequalities are not immediately apparent. [52] Generally speaking, the results suggest the inexistence of any direct correlation between social status and judgments made about public integrity. They support the hypothesis according to which opinions about public integrity and corruption are multidimensional.

The contribution of focus groups in exploring the social dimension of judgment

29Seen in this light, focus groups provide a methodology that, in line with the construction of the model, is more conducive to making felt (during discussion) and making apparent (during analysis) the social grounding of judgments about politics and integrity. During the interview with the group composed of young manual laborers and low-paid workers, one of the participants, a mason, adopted an ambivalent judgment, which he defended in a nuanced manner, regarding the scenario in which a mayor allocates a five-bedroom apartment in a social housing block to one of his cousins, a bachelor. While the behavior of the elected official may be condemned, the result for the beneficiary is nevertheless positively defended in the broader context of restricted access to housing.


“On the part of the mayor, it’s unacceptable; but again, the tenant is totally in the right. Someone’s just handing him a new apartment, he’s right to accept it. Still, it’s not acceptable that he should receive it and not a large family […] Such situations are incredibly common, I’ve personally seen it myself. Not so flagrant as this bachelor in his five-bedroom apartment, but I’ve seen similar such things. Especially in the smaller villages in which the mayor has the applications sitting right next to his desk; it’s incredibly common you know. My employer was once the mayor and he found me an apartment that was about 100 meters from work, in some social housing in the village, even though some people had been waiting, like, two weeks or god knows how long. As for me, I had an apartment in three weeks. I was single, I had two bedrooms with a garden, a garage; I was well set up, you know. The mayor was the one acting like a fool, and I was totally right to take advantage.”

31Disapproval, which is explicitly articulated, is linked to the realist observation according to which “the ends justify the means”. During the entire collective interview with the young manual laborers and low-paid workers, the social grounding of judgment is palpable within the speech of all the participants, linked to personal economic difficulties that lead to the justification of a kind of disillusioned pragmatism, which consciously valorizes the efficiency of action over the intrinsic worth of moral principles. In contrast to this view, the focus groups undertaken with traders, artisans and small business-owners are striking in the homogeneity evidenced in the attitudes and behaviors (the manners of speaking). Everyone seems to adhere to a realist and avowedly cynical perception of political practices.


L. (taxi driver): All politicians are dishonest, essentially. To succeed in such professions, you have to be dishonest.
D. (clothing merchant): Yeah, essentially that’s it: if you want to become a politician, you have to be a good liar already. It shocks me because he lied once or twice under oath to save his career. He must have had to go before a jury, swear on his honor, etc. That’s what shocks me. Not the fact that he lied. If you’re honest, there’s no way you can be in politics, it’s just not possible.
L.: Yeah, I reckon that in politics, the first thing you learn – even at the ENA[53] – is to lie.

33The functioning of the “system” is unanimously presented as based upon individual plots and clannish schemes, and on networks. Other group interviews shed light upon the professional grounding of judgment. During the focus group carried out with young executives and those working in “liberal” (highly-qualified, self-employed) professions, one of the participants, an architect, aptly evoked the professional context and the ambiguity of a situation in which the mayor of a city required a restaurant-owner to cater for his son’s wedding banquet in order to obtain the contract for the municipality’s school canteens.


“We had this same problem with a communist municipality; the situation really got pretty unsavoury. There was no transparency whatsoever. Even people in the neighborhood were saying, ‘this whole thing’s rotten’. We even had different restaurant owners who talked to us of the situation, independently of one another, for a good two hours. There was also a real pressure coming from the mayor. The mayor allowed things to happen that we could have sent to the Canard enchaîné. [54] For us architects, that sort of call for tenders, with all kinds of pressures from the mayor, who might promise things without sticking to it, that can really upset the dynamics when you’re competing for a contract, you never know how to go about it. You always say to yourself that you represent a good level of quality, that the architecture has been conceived of at a reasonably high level, that therefore, there is no reason not to have the same chances as everyone else; and yet, it’s pretty clear how it really goes down. You have to go kiss up to the mayor, you have to go…”.

35In this group, the means of identification with the scenario and, more broadly, of judging it, are premised less on the identification with personal situations than on the analytical apprehension of dilemmas that can arise in professional situations.

36The systematically quantitative treatment of the judgments elaborated during the discussion of eight scenarios provides further material for identifying the impact of social status, and, by refining the results of the qualitative study, works to further confirm them. The quantitative analysis also makes it possible to establish differences between groups in ways that were not possible using the questionnaire-based study.

The social grounding of judgment

37In the quantitative study, [55] the different socioprofessional statuses do not greatly diverge regarding the general level of judgmental severity brought to bear on the scenarios. This level varies very little, going from 57% (minimum) to 58% (maximum). [56] Crunching the focus group data establishes interesting variations: initial judgmental severity varies from 40% to 63% [57] according to status (cf. Figure 4). Generally speaking, we observe, not surprisingly, that the most controversial scenarios (grey zones of judgment) are the ones for which we find the greatest judgmental variability between socioprofessional statuses (cf. Figure 1).

Figure 4

Initial judgmental severity by group socioprofessional status

Figure 4

Initial judgmental severity by group socioprofessional status

38Nevertheless, it is possible to identify certain specificities that are attributable to status. We will firstly present the two groups most highly contrasted in terms of judgment (the most tolerant – traders, artisans and small business-owners – and the most disapproving – young manual laborers and low-skill employees), and then the four intermediary situations.

39The group of traders, artisans and small business-owners distinguish themselves from other statuses because they are the only group (generally) to express tolerance with regard to the political conduct presented in the scenarios. Across all scenarios, this group has an initial severity index of 40% (average 2.2), whereas, for the group of manual laborers and employees over 40 years of age, the initial severity index is 51% (2.53), but also 53% (2.59) for business people and 55% (2.65) for bureaucrats. Taking into account the index of severity of the discussion confirms these trends. Regardless of the scenario, the judgments of the group of traders, artisans, and small business-owners are the most tolerant of all of the groups, including for the practices agreed upon as “serious” (“black corruption”), such as lying under oath and buying favors (cf. Figure 5).

Figure 5

Initial severity index (solid lines) versus discussion severity index (dotted lines) according to the degree of gravity of the scenario and the socioprofessional status of the group

Figure 5

Initial severity index (solid lines) versus discussion severity index (dotted lines) according to the degree of gravity of the scenario and the socioprofessional status of the group

40On the other hand, the group of young manual laborers and employees is the one to express the harshest judgments, both during initial evaluations (63% on average) and during discussion (59% on average). Economically precarious situations, linked to the uncertainty of low-skilled employment, tend to favor the expression of a general sense of disappointment, resignation and also of harshness with regard to politics. Analyzing the variations in severity from one scenario to another can refine our understanding of this elevated attribution of seriousness. The group of young blue-collar workers and low-level employees revealed itself to be clearly more intransigent regarding scenarios linked to obtaining economic and social goods that are hard to find (jobs, housing) by means of self-interested practices (making a donation, joining the mayor’s party, getting a letter of recommendation from the wife of the elected official). Their initial level of severity is thus 71% (average on a scale of 3.13), whereas it is only 44% for businessmen (2.32), or even 30% for the group of traders, artisans and small business-owners (1.9). Such situations are condemned all the more harshly for the fact that they correspond to the lived experience of this group’s participants. Acts are not condemned without some degree of ambivalence, however, because certain participants admit to having benefited from such practices. These kinds of behavior seem to be all the more harshly judged for the fact that they highlight compromises that result from economic difficulty and precariousness.

41Turning now to the way in which judgmental severity during discussion varies according to the scenario’s protagonist, one cannot help but notice the uniqueness of the group of young low-skill workers and manual laborers. This group has by far the highest severity index – at 68% (3.04) – when an elected official carries out the action. In such situations, criticism of political transgression is the harshest. In contrast to this phenomenon, the group of traders, artisans and small business-owners (generally more lax) judge actions carried out by citizens and those carried out by elected officials in an equal manner (39% and 35% respectively). This judgment reflects their “realist” view of the social and political spheres. They value, on the one hand, the capacity of elected officials to produce visible results and, on the other hand, the personal benefits that social individuals are able to derive from such actions, regardless of the measures that had to be taken to realize these benefits. In this world, the boundaries between elected officials and citizens are blurred. Elected officials are no more responsible for corruption than citizens. It is because society as a whole functions according to power relations and favoritism that the elected official, who is merely the product of this system, engages in collusion, lying and abuses of power.

42Sitting in the middle of the two groups that occupy opposing ends of the judgmental continuum, the differences between the groups of senior bureaucrats, of business people, and even of low-skill employees and manual laborers over the age of 40, are less immediately perceptible, especially if we focus solely on judgmental severity. By comparing initial severity indexes on a scenario-by-scenario basis, however, we can identify a number of telling differences. The groups of bureaucrats thus belong to the most tolerant (along with low-skilled adults and the group of traders, artisans and small business-owners, always more “laxiste” [relaxed]) concerning scenarios unanimously deemed “not serious” (the box of chocolates, the nursery). They are also the harshest of all the groups regarding scenarios unanimously condemned or considered serious (lying under oath, the renovation or taxi scenarios). Their level of severity is, on the other hand, close to the two other categories concerning scenarios that provoke the most dissension. A possible explanation is the “proximity” of the bureaucrats to the two sorts of behaviors, which their professional positions regularly allow them to observe. Their judgments are based on forms of insider’s knowledge of the limits of acceptability. They trace boundaries between practices that belong to “politics as usual” (tolerated due to their normality) and those that clearly breach the legal framework (condemned without hesitation due to their illegality). A discussion about the “chocolate box” scenario, taken from one of the groups of bureaucrats, adequately illustrates this knowledge of the “do’s and don’ts” of politics.


Group leader: Why do you think that it’s not serious?
Participant 1: It’s done out of friendliness.
Participant 2: Everyone knows about it. And everyone benefits; there’s no discrimination.
Participant 3: In this case mothers are all put on equal footing; there is no preferential treatment or favoritism.
Leader: Do you think this sort of thing is common? Can it mask electoral ambitions?
Participant 4: Obviously! But since it’s non-discriminatory… it’s a nice gesture. All of the measures that are undertaken are always in favor of the population. So it’s also a gesture in favor of the population, like any other. And given that it’s not discriminatory, it works for the common good. It’s for the mothers. If you build a sports center, that only concerns those who like sport. At the end of the day, it’s not going to concern anyone else; and yet, it’s for everyone. The common good is common to everyone, that’s what it means. In this case, it’s for all the mothers. […] But it’s obvious that there are ulterior motives, but all the same, that’s not reprehensible.
Leader: And if it comes with a personalized note from the mayor?
Participant 4: But of course it’ll come with a note! When you give a gift, you always say…otherwise it’s not a gift. When you give a gift as a private citizen, you always say: “Here, this is for your 20th birthday, hugs, your name.” Here, it’s the same thing, except it’s not signed “hugs”, but it’s the same. It obviously has a note. And no one is fooled about its purpose. Yet all the same, to make a nice gesture should not be reprehensible; if that were the case, we couldn’t shake hands anymore, because, effectively, that would then constitute making a nice gesture to a potential voter.

44Other kinds of differences come to light between the group of high-level bureaucrats and the businessmen when we take into account the effects of the discussion. Of all the socioprofessional groups, the bureaucrats held initial judgments that evolved to the greatest extent during discussion, notably because they significantly tempered their initially tolerant judgments of scenarios that they did not consider to be very serious (the gift of the box of chocolates by the mayor on mother’s day or the request for a place in the nursery). In contrast to such tolerance, the business people constituted the group who “averaged out” their initial judgments to the least extent during discussion, and, notably, who subsequently applied the least moderation to their harshest initial judgments. The manual laborers / older employees, for their part, appeared more severe during the discussion than during the initial evaluation concerning scenarios that gave rise to disagreement and that centered on coveted goods (housing, employment). Explaining such differences, which combine social status and discussion effects, clearly is no simple task. The concept of the socioprofessional grounding of judgment is important in understanding such differences. In the bureaucrats’ case, the collective discussion led to the unveiling of hidden (and implicit) issues embedded in these practices. Even though they know them to be tolerated, bureaucrats are also perfectly conscious of the political machinations and clientelism that they disguise. That business people did not, as a result of their discussions, moderate their judgments regarding scenarios that are unanimously condemned is more difficult to interpret. It would almost appear as if they applied general principles and legal rules in order to ground their judgments, without wanting to consider – or in any case, less than other groups (namely the low-skilled employees and manual laborers over 40) – the sorts of mitigating factors (economic context, difficult access to coveted goods, etc.) that might moderate the harshness of their judgment. Despite the fact that the situation is judged from a professional point of view, the interview extract cited above clearly illustrates that, to the contrary, it is the general analytical principles that are determinant in the final count.

Identifying and analyzing argumentative repertoires

45Our model for quantitative database analysis not only enables us to evaluate the severity of the judgments made about political transgressions, it also makes possible an exhaustive analysis of the justifications that people produce in support of their judgments. Our initial exploration of the collected material, based on an iterative approach, made it possible to identify principles concurrent with political judgments: three moral principles emphasizing impartiality, the exemplary nature of the behavior, and selflessness; three principles of action rooted in a consequentialist approach, putting high value on the efficacy of public policy, on actions that truly further the common good, and on the elected official’s availability to citizens. [58] The way in which, according to the group under analysis, these principles are variously combined, structures what we call the different “repertoires” (legalism, contractualism, pragmatism and realism), which define both the general relation to the political and the contrasting normative expectations built into this relationship. [59] Quantitative analysis of discussions considerably enriches this treatment of the data. In this way, arguments can be analyzed more systematically, and it is even possible to construct more precise “argumentative repertoires” and to measure the frequency with which they are employed.

46We begin with an explanation of the analytical procedure that enabled us to identify a fixed number of argumentative models that have a very general scope and range, in terms of both the social groups and the material presented. The constitution of these models, that we term argumentative repertoires, is the final result of a quantitative, “bottom-up” analysis in which we analyze semantic content by using patterns in the data itself as a guide. We then go on to present these repertoires and the portions of discussion that they each represent, relative to one another. We shall see how social groups can be distinguished according to the relative significance that they accord to these repertoires in order to justify or denounce deviant behavior.

How can the relative significance of argumentative content be quantified? Implementing a new analytical procedure

47The analytical procedure used to identify and analyze the argumentative content of the corpus is implemented through three main steps.

Categorization and classification of arguments

48The first involves coding, categorizing and classifying arguments (into units of meaning), taking into account each time the context of one scenario at a time (into units of context, across all groups). The categorization of arguments was informed by the advice of René L’Ecuyer regarding the “open model”. [60] This model is said to be open “insofar as there are no default categories given from the outset; categories are formed from the material under analysis by successive groupings of enunciations, according to their relatedness or the similarity in their meaning”. [61] We reused the same breakdown of data undertaken during the first analysis (of judgmental severity). Thus for each of the enunciations in which a judgment was expressed (either approving or disapproving of the presented conduct), we identified and then briefly summarized the argument(s) presented. After this coding phase, undertaken for each scenario (across all groups), we proceeded with the first groupings of summarized arguments according to the similarity of their meaning, and then named these preliminary categories. We then reorganized the categories (by subdividing, combining and removing redundant categories), making sure at every step that they accurately reflected the enunciations classified within them. We gave a precise name to the categories of arguments thus created and drew up an analytic grid for each scenario (cf. Annex 3 for an example of a grid). These eight grids, composed of categories justifying or denouncing the conduct (from 11 to 24 categories of “arguments” depending on the scenario), [62] enabled us to proceed to the classification of each of the corpus’s 1,187 argument-bearing enunciations. In the final count, some 2,586 arguments were identified and sorted into our categories.

Extracting arguments

49Using the appropriate method of statistical analysis (Principal Component Analysis), the following step involved the extraction of argumentative links intuitively produced by participants during discussion.

Building data matrices

50Firstly, we divided the corpus of each scenario (for all groups) into successive 500 character intervals. In each interval, we grouped together the enunciations per participant. Subsequently, we counted and classified the arguments expressed in each of these groupings. Looking at individual speakers, the idea was to create a data matrix that enabled us to make clear which arguments were systematically contiguous during the discussion (cf. Table 2). The rows of this matrix refer to participants who, during a given temporal interval, made comments or segments of comments supporting their arguments; the columns contain the lists of argument categories produced for each scenario. In each line, the arguments contained in the enunciations were counted and distributed throughout the categories.

Table 2

Example of a data matrix in preparation for factorial analysis

Table 2
Arguments (Scenario A) Group # Line # interval # Participant Arg. 1 Arg. 2 Arg. 3 … Arg. 20 1 1 1 Philippe 3 0 2 … 0 1 2 1 Alain 0 2 0 … 1 1 3 2 Alain 2 0 1 … 0 1 4 2 Jean-François 0 1 0 … 2 1 5 3 Aline 0 2 0 … 3 … … … … … … … … … 8 189 6 Sergio 1 0 1 … 0

Example of a data matrix in preparation for factorial analysis

51In the example presented in Table 2, Philippe, a participant in focus group 1, makes three arguments from category 1 and two arguments from category 3 in the first temporal interval of the discussion of scenario A.

Exploratory factorial analyses

52For each of these eight data matrices (one per scenario) we carried out a Principal Component Analysis (PCA). [63] By examining the matrix of factorial weights we were able to determine which arguments contributed most significantly to a single component (cf. Annex 4 in order to examine the detailed results of a PCA). [64] These associations between arguments constitute, in fact, the very argumentation structures produced during the discussion of each scenario. Finally, we attempted to label these broad argumentation structures identified within the data by looking at the relative contributions of the different sub-arguments. Over the whole corpus, 53 argumentation structures were identified (between 5 and 8 per scenario). Among these argumentation structures, 31 belong to arguments justifying conducts and 22 to those denouncing them. Consequently, we were able to quantify the number of arguments constitutive of argumentation structures, and thereby to quantify the relative weight of these in the discussion of a given scenario.

From argumentation structures to the constitution of argumentative repertoires

53Recall that the argumentation structures identified in the preceding step are relative to a single scenario. One of our objectives, however, was to be able to compare the argumentative resources that people made use of according to the type of scenario or their status, which is to say, across all scenarios. We therefore proceeded with a final stage in which similarities of meaning between the 53 argumentation structures were identified. We quickly realized that it was indeed possible to draw out categories from the data that were both illuminating and common to all scenarios; we have called these “argumentative repertoires”. Having completed a categorization process following the recommendations of R. L’Écuyer, [65] we were finally able to classify the argumentation structures into eight repertoires (three to ten structures per repertoire). [66] Annex 5 features an example of the manner in which several argumentation structures were classed into a single repertoire. The eight repertoires constituted in this manner are much more precise and far less schematic than those identified during the first qualitative analysis of the corpus. [67] Each is based on a principle or on the absence of one (availability of elected officials, morality, justice, efficiency, and so on), depending on whether the principle is intended to denounce or justify conduct.

54By employing this framework, we can calculate an index of the relative use of these repertoires during discourse. We do so by determining the portion of arguments belonging to the repertoire in question (its constitutive arguments) relative to all arguments put forth (belonging to all repertoires), according to variations in context (for example, participant, scenario, group, status).

Relative use of argumentative repertoires and social grounding

55Because the repertoires are limited in number (eight) and because they correspond to principles commonly employed in the judgment of the political, it is not surprising to find them used across all groups. The ways in which they are put together appears, however, much more specific and often linked to one social group in particular.

Relative weight of repertoires in the corpus

56Examination of the relative portion occupied by each of the eight argumentative repertoires (across all groups and scenarios) is instructive. In the group of repertoires, five relate to the justification of conducts (representing 46% of the arguments) and three relate to their denunciation (54% of the arguments). By quantifying the arguments constitutive of repertoires in this manner, we arrive at the following result: the general judgment of deviant conduct is rather severe, despite the fact that participants use a more diverse range of repertoires in order to justify them (five against three, see Figure 6).

Figure 6

Relative use frequency of the eight argumentative repertoires (across all groups and scenarios)

Figure 6

Relative use frequency of the eight argumentative repertoires (across all groups and scenarios)

Note: N corresponds to the number of arguments belonging to the argumentative repertoire (across all groups and scenarios).

57When they need to provide support for a disapproving judgment, the participants employ primarily (in 21% of the arguments) the repertoire denouncing misdemeanors and a lack of morality (“No morality”). We find here one or several of the moral principles identified in the first treatment of the focus groups: impartiality, exemplary nature of the conduct, and selflessness. The second place (19%) is occupied by the repertoire emphasizing a legal approach to the situation, and recalling formal rules (“Legalism – serious – corruption or attempted corruption – unfair advantages”).

58When participants make arguments to support tolerant judgments (cf. Figure 6), they do so on the basis of three categories that organize the five repertoires in use. The first is constituted of a single repertoire, without question the most commonly employed (18% of arguments), and which is distinct from the other four. This approach develops a pragmatic point of view in which rules can be subject to “arrangements”, and in which favors are viewed with considerable indulgence (“Pragmatism – giving it a shot – saving a situation”). In having recourse to this repertoire, the participants place high value in the potential efficiency of deviant conduct as presented in the scenarios, and they emphasize the legitimacy of actors’ motivations. The second category is composed of two repertoires that clearly exhibit the same use frequency (11% and 10% of arguments). The first belongs to a group of justifications emphasizing the attenuating circumstances of the actors’ conduct (“Not serious – attenuating circumstances”). The second is consequentialist in inspiration; it either consists of denying the efficacy of deviant conduct, or stresses the uncertain results of transgression (“Without consequence or uncertain”). The third category (least frequently employed, 6%) is composed of a first repertoire (4%) according to which, in a political and social universe that defies all attempts at regulation, the ends justify the means (“Realism – Benefit from the corrupt system”). The second repertoire (only 2% of arguments) justifies transgressive conduct by placing value in the relations of intimacy maintained by elected officials with their elector. It considers that the role played by the elected official involves the satisfaction of the demands of the citizen and the giving of favors, even if this is in conflict with certain principles such as that of equal treatment (“Availability of the elected official”).

Differences in the employment of repertoires according to status

59When we examine the use of repertoires, comparison between groups once again reveals itself to be very effective. While each group employs all of the repertoires, thereby demonstrating the stability of these argumentation structures and their connection to a limited number of general principles, certain groups, because they privilege certain repertoires, create assemblages [Translator’s note: combinatoires] that are specific to them. Two socioprofessional groups can be distinguished from the others in their employment of repertoires (cf. Figure 7): the group of bureaucrats and the group of small business-owners (traders, artisans and small business-owners).

60The group of bureaucrats is set apart by its overwhelming use (31%) of the repertoire disapproving of the protagonists’ lack of morality; this repertoire is most intensely used when compared with other statuses (the average use across statuses is 21%). The bureaucrats also employ references to legalism (16% of all arguments put forth) and, to a lesser extent, to the principles of just equality (11%) in their denunciation of deviant conduct. When justifying transgressions, the bureaucrats use the consequentialist repertoire more than any other group or emphasize the uncertainty of the result (16%; the average across all statuses is 10%). The employment of these specific repertoires can be explained through reference to the internal familiarity of these actors with political procedures. The bureaucrats are familiar with the legal rules but also – better than other groups – they know that the existence of significant leeway requires a level of behavioral integrity that they attribute, specifically, to a heightened conception of role and function (morality).

61The group of small-scale independent business-owners is characterized by its infrequent recourse to disapproving repertoires, whose use frequency is always below average (across statuses). Taking into account the five repertoires justifying deviant conducts, tolerant arguments comprise 62% of their argumentation. One also observes in this group a disproportionate presence of pragmatic arguments (27%, while the average is only 18%). They also invoke attenuating circumstances that minimize actors’ responsibility (15%; the average is 11%) and realist arguments (9%; the average is 4%). This group makes the highest use of this last repertoire.

62When we examine the use that the other three statuses make of the repertoires, one notices a clear pattern of resemblance (cf. Figure 7). This is most striking for the two groups of low-skilled workers (young and over 40). This final result seems to validate the hypothesis according to which the variation observed in the use of arguments is indeed correlated to status effects. Generally, these three statuses (the business people, the two groups of manual laborers/low-skilled workers) use disapproving repertoires with an average level of frequency. The two groups of manual laborers / low-skilled workers, however, more frequently have recourse to pragmatism (20% for the younger group and 22% for the elder group) compared to the business people (16%).

Figure 7

Relative use frequency of argumentative repertoires according to socioprofessional status

Figure 7

Relative use frequency of argumentative repertoires according to socioprofessional status

Note: For each status, the value associated with a given repertoire corresponds to the percentage of arguments (produced by the groups of that status) that belongs to the repertoire, relative to the arguments associated with all repertoires.

63* * *

64The focus group study is part of a research program proposing three different and complementary methodologies – quantitative questionnaire studies, individual monographs and focus groups. By way of conclusion, we can begin by placing these approaches alongside one another and comparing their respective contributions. With regard to the focus groups, three distinct contributions should be made clear. First of all, initial analysis of the discussions made it possible to focus the majority of the quantitative study questions on neighborhood political relations (local scale). In effect, it is in this domain that argumentation and reasoning are most immediate, and, moreover, most nuanced and least stereotyped (contrary to the national level). Second, compared to the interviews carried out in the context of in-depth studies of local administrative districts [monographies de communes], the focus groups made it possible to discern nuances and ambiguities in the judgments made of elected officials’ integrity. By making clear a variety of argumentative repertoires, in addition to the effects of the discussion dynamics on the initial judgments, we have been able to shed light on the variability and the importance of contextual factors in formulating judgments. Finally, a systematic analysis of focus groups makes tangible the effects of social status, whereas they were much less evident in the two other approaches.

65The tool used to help with the analysis of specific content, conceived of to quantify the data produced by focus groups, can also be further explored. It comprises a database that, using statistical and spreadsheet software, makes possible: a) the compilation and codification of the enunciations in the corpus; b) statistical treatment; and c) exploration of the corpus using sophisticated commands. As this tool is flexible and capable of development, it can be extended further to deepen our analysis. Let us suggest three possible avenues. First of all, regarding the level of judgmental severity during discussion, the results would appear to lend significant weight to the validity of our temporal index. Yet, this could be lent more weight if we were to better calibrate the measure of severity by taking into account the force of the argumentation associated with it. We might also refine our analysis by evaluating certain complex interactive effects (for example, effects related to status, to the seriousness of the scenario and to the protagonist of the action). Secondly, with regard to argumentative repertoires, we observed their utility when used across groups. But it would be interesting to explore the relations that they do or do not engender amongst themselves (for example, associations, oppositions), and in relation to severity indexes. [68] It would appear that the type of association differs according to the context of the judgment. It would also be judicious to examine the cumulative effect of certain variables on the recourse to these repertoires: for example, if the bureaucrats denounce the absence of morality more than others, do they do so differently in accordance with the seriousness of the scenario or with the nature of the protagonist? Finally, in a more general sense, we might optimize the significance of results concerning the comparisons between social statuses by increasing the number of groups. If we had to compare our database to the existing discursive analysis software, whether the analysis is lexical (such as Alceste) or thematic (such as Prospéro), we would say that our database seems more adapted to the semantic coding of judgments and arguments. In effect, the advantage of such specialized software is limited because it does not allow one to automatically break down the corpus into units of meaning (which is to say, into segments that vary in length with the same meaning); it can only be done manually. [69] Moreover, using our database in conjunction with professional statistical software, allows us to overcome the constraint of the limited choices offered by such software.

66To conclude, we believe that the analysis of focus groups helped us to progress in answering our initial research question. The support that citizens accord to unscrupulous elected officials within democratic regimes becomes less enigmatic when one makes clearer the various criteria employed in judging politics and policy. Generally speaking, the repertoire analysis illustrates a very strong tension between two types of principles that confront one another with relatively equivalent weight (54%/46% – cf. Figure 6). On the one hand, judgments rely on specific expectations and principles regarding elected officials and manners of exercising power. The disapproving repertoires most frequently employed, regardless of the group, insist on the moral aspects of civic service and on the respect of both formal rules and just principles (54%). Normative expectations regarding political conduct do exist. Yet, on the other hand, analysis of argumentation structures puts into stark light the predominance of pragmatic and realist considerations in politics (46%). Justifications valorize the social, economic, and personal outcomes of elected officials’ actions. They also reveal the constraints weighing on certain social categories, which lead them to seek out favors, or even to act cynically. In such cases the political space is predominantly perceived of as (relatively) autonomous in a normative sense. This tension, between two strong structurations of political judgment, goes a long way towards explaining the ambiguities and contradictions of French citizens regarding politics. This tension should be studied further, using an increasingly systematic analysis of argumentative repertoires and of the principles that are employed and by further refining the understanding of their social grounding. Recourse to focus group methodologies in comparative studies would provide us with valuable data. [70]

Annex 1 - List of scenarios

Not very serious scenarios

67“Nursery” (To obtain a place in a nursery, one of your friends requests a meeting with the wife of the mayor.)

68“Box of chocolates” (The mayor of your town sends married women in his constituency a large box of chocolates for Mother’s Day each year.)

Scenarios of average seriousness

69“Job for son” (To help his son get a job in an association with links to the municipality, one of your friends gives a gift of 2,000 euros to the political party to which the mayor of your commune belongs.)

70“Municipal employee” (An elected official suggests to one of your friends that he join the city mayor’s political party in order to get a job as a municipal employee.)

71“Social housing” (To get subsidized housing, a city representative suggests to one of your neighbors that he get a letter of support from the wife of the mayor.)

Very serious scenarios

72“Lying under oath” (To save the political career of a friend, the mayor of your commune lies under oath.)

73“Renovations” (One of your friends is the director of a construction company with 30 employees but is having economic difficulties. In order to obtain a contract from the regional board, he offers to renovate a local representative’s country house free of charge.)

74“Taxi” (To get his driver’s license back after having it suspended for drunk driving, a taxi driver offers 5,000 euros to an elected official to intervene in his favor.)

Final corpus: Inclusion criteria for scenarios used in analysis

75Among all of the scenarios presented, eight were retained on the basis of two criteria. The first consisted in excluding those scenarios for which the presentation differed among groups. More precisely, as the focus groups were conducted in two waves, some presentations were adjusted slightly during the second wave, which then made comparison between groups impossible. The second criteria arose from a process of categorization that we carried out from the outset of the study. Effectively, we wanted the scenarios to represent differing levels of seriousness, all with a certain kind of balance. This therefore led us to choose scenarios evaluated from the outset as not serious, moderately serious and very serious. Moreover, the data processing made it possible to verify whether the participants’ own judgments fit with our initial evaluation. The transcription of the entire final corpus was made up of 3,321 enunciations (405,542 characters).

Annex 2

Table 3

Example of a breakdown of comments, of classification of units of meaning, and of characters counted

Table 3
Comment number Participant Transcription Classification Unit of meaning number No of characters in French 1 Mireille That costs the commune money. SERIOUS 1 31 But on the other hand, everyone is treated the same: Mother’s Day is celebrated in France. NOT SERIOUS 2 91 2 Marilyne It is celebrated by children, but by the mayor, I think that’s bad form. I would not want to be bought like that. SERIOUS 3 114 3 Leader Oh, you have the impression he’s looking for something? – – – 4 Marilyne Yes, it’s an electoral message. That would bother me. SERIOUS 4 50 It’s true that it’s very small, NOT SERIOUS 5 31 but me, that would bother me. SERIOUS 6 26 So, time for break? – – –

Example of a breakdown of comments, of classification of units of meaning, and of characters counted

Comments 1 and 4 were broken down. Comment 3 and the last segment of comment 4 were removed from the analysis. The six units of meaning were eventually classified in two categories according to the orientation of the judgment (“serious” versus “not serious”).

Annex 3

Table 4

Argumentative categories denouncing behaviors in the “box of chocolate” scenario versus those justifying them

Table 4
Argumentative categories.. … denouncing the behaviors 1. Other ways of improving the lives of constituents – not for the common good – social actions would be more useful 2. Immoral undertakings, unethical – shocking acts 3. Misappropriation of public revenues – direct or indirect costs 4. Serious – more serious than the other cases cited 5. The elected official takes advantage of the weakness of a population – does not respect his constituents, talks down to them 6. The elected official is not there to make his constituents or personnel happy 7. Breaking with equality 8. Attempted corruption 9. Electoral goals of the official – hypocrisy, hidden agendas 10. Gaudy, ridiculous, worthless..justifying the behaviors 11. Decision legitimized by a vote 12. An undertaking that has no effect on constituents’ voting – free choice of the decision-maker 13. Elected official wants to make his constituents happy 14. A friendly gesture, special attention – creating social ties, working against isolation 15. Does not cost the region money, little to no money at stake 16. No unequal treatment, others receive gifts as well 17. No electoral goals 18. Not at all or not very serious – less serious than the other cases cited – not in the same register as corruption or attempted corruption 19. Paternalism by elected official 20. Normal, banal, common practice – it’s traditional 21. Elected official’s availability

Argumentative categories denouncing behaviors in the “box of chocolate” scenario versus those justifying them

Annex 4 - Results of the Principal Component Analysis (PCA) for scenario 5 (“job for son”)

Table 5

Determinant for the correlation matrix, KMO index and Bartlett Test

Table 5
Determinant 0.093 Measure of adequacy of the Kaiser-Olkin sampling 0.634 Bartlett’s test of sphericity Approx. Chi-2 (ddl = 136) = 432, p = 1,526E-32

Determinant for the correlation matrix, KMO index and Bartlett Test

Table 6

Extraction of components: Percentage of total variance explained before and after orthogonal rotation (Method: Varimax with Kaiser normalization)

Table 6
Initial eigenvalues Extraction sums of squared loadings Rotation sums of squared loadings Component Total % of variance % combined Total % of variance % combined Total % of variance % combined 1 2.82 16.58 16.58 2.82 16.58 16.58 2.17 12.75 12.75 2 1.56 9.20 25.79 1.56 9.20 25.79 1.82 10.72 23.47 3 1.54 9.03 34.82 1.54 9.03 34.82 1.54 9.05 32.51 4 1.30 7.66 42.47 1.30 7.66 42.47 1.34 7.90 40.42 5 1.22 7.21 49.68 1.22 7.21 49.68 1.31 7.71 48.13 6 1.18 6.92 56.59 1.18 6.92 56.59 1.29 7.62 55.75 7 1.06 6.24 62.84 1.06 6.24 62.84 1.21 7.09 62.84 8 0.91 5.38 68.22 9 0.86 5.03 73.25 10 0.77 4.51 77.76 11 0.73 4.29 82.05 12 0.66 3.89 85.94 13 0.59 3.49 89.43 14 0.54 3.15 92.58 15 0.46 2.72 95.31 16 0.41 2.41 97.72 17 0.39 2.28 100.00

Extraction of components: Percentage of total variance explained before and after orthogonal rotation (Method: Varimax with Kaiser normalization)

Figure 8

Horn’s parallel analysis: observed versus random eigenvalues

Figure 8

Horn’s parallel analysis: observed versus random eigenvalues

Table 7

Correlation matrix of components after oblique rotation (Method: Oblimin with Kaiser normalization)

Table 7
1 2 3 4 5 6 7 1 1.00 -0.06 0.25 -0.09 -0.14 -0.16 -0.16 2 -0.06 1.00 -0.04 -0.02 0.02 0.02 -0.01 3 0.25 -0.04 1.00 -0.06 -0.07 -0.11 -0.10 4 -0.09 -0.02 -0.06 1.00 0.04 0.07 0.03 5 -0.14 0.02 -0.07 0.04 1.00 0.08 0.07 6 -0.16 0.02 -0.11 0.07 0.08 1.00 0.09 7 -0.16 -0.01 -0.10 0.03 0.07 0.09 1.00

Correlation matrix of components after oblique rotation (Method: Oblimin with Kaiser normalization)

Note: Since the components of the oblique solution (Oblimin rotation) are not correlated, the orthogonal solution (Varimax rotation) was chosen.
Table 8

Matrix of components (“Factor Matrix”) after orthogonal rotation (Method: Varimax with Kaiser normalization)

Table 8
SERIOUS NOT SERIOUS COMPONENTS (Argumentation structures) 1 2 3 4 5 6 7 Legal 0.81 0.08 0.00 0.00 0.17 0.00 0.01 Free use of one’s money 0.79 0.07 -0.02 -0.03 -0.04 -0.05 -0.05 Giving out of belief 0.65 0.07 0.02 0.02 0.04 0.04 0.07 It’s human 0.57 0.13 -0.06 -0.07 -0.07 -0.12 -0.17 It happens 0.14 0.83 0.02 0.02 -0.02 0.05 0.07 Not serious 0.11 0.74 -0.06 -0.08 -0.02 -0.15 -0.16 Pragmatism 0.08 0.70 -0.03 -0.04 0.18 -0.07 -0.08 Difficult judgment -0.13 0.01 -0.01 -0.01 0.83 -0.01 -0.01 Uncertain outcome 0.26 0.14 -0.04 -0.06 0.73 -0.09 -0.09 Lack of parental responsibility 0.00 0.00 0.87 0.01 0.00 -0.04 0.00 Poor transmission of values -0.03 -0.06 0.87 -0.05 -0.04 0.02 -0.04 Poor use of money 0.00 0.02 0.01 0.82 0.02 0.05 -0.06 Poor image of politics -0.05 -0.10 -0.05 0.80 -0.07 -0.01 0.04 Lack of morals -0.01 0.00 0.02 -0.01 0.01 0.81 -0.01 Serious -0.07 -0.14 -0.05 0.05 -0.10 0.75 0.02 Personal advantage -0.02 0.00 -0.01 0.06 0.01 -0.05 0.71 Unequal treatment -0.06 -0.12 -0.04 -0.08 -0.10 0.05 0.79 Explained variance 2.17 1.82 1.54 1.34 1.31 1.29 1.21 % of total variance explained 12.75 10.72 9.05 7.90 7.71 7.62 7.09 % of combined total variance explained 12.75 23.47 32.51 40.42 48.13 55.75 62.84

Matrix of components (“Factor Matrix”) after orthogonal rotation (Method: Varimax with Kaiser normalization)

Note: The values of the matrix correspond to the correlations (“factor loadings”) between the observed variables (arguments) and the principal components (argumentation structures). Correlations higher than 0.57 are in bold. The rotation converged in 5 iterations.

Annex 5

Table 9

Argumentation structures constitutive of the “NO JUSTICE – NO EQUALITY” repertoire

Table 9
ARGUMENTATION STRUCTURES Scenario REPERTOIRE “no justice – no equality” Box of chocolates Unequal treatment/Other actions are more socially useful Nursery Unequal treatment/Serious Job for son Unequal treatment/Personal gain Municipal employee Unfair undertaking/Instrumentalization of the party – bad image for the elected representative Social housing Unjust – favoritism/Wife is not legitimate Lying under oath – Renovations – Taxi Inequality of treatment/Driver must take responsibility/The elected official should refuse

Argumentation structures constitutive of the “NO JUSTICE – NO EQUALITY” repertoire


  • [1]
    Philippe Bezes and Pierre Lascoumes, “Percevoir et juger la ‘corruption politique’: enjeux et usages des enquêtes sur les représentations des atteintes à la probité publique”, Revue française de science politique, 55(5-6), 2005, 757-86; Pierre Lascoumes and Philippe Bezes, “Les formes de jugement du politique: principes moraux, principes d’action et registre légal”, L’Année sociologique, 59(1), 2009, 109-47.
  • [2]
    Pierre Lascoumes (ed.), Favoritisme et corruption à la française: Petits arrangements avec la probité (Paris: Presses de Sciences Po, 2010).
  • [3]
    Jennifer C. Greene and Valerie J. Caracelli, “Defining and describing the paradigm issue in mixed-method evaluation”, New Directions for Program Evaluation, 74, 1997, 5-18; R. Burke Johnson et al., “Toward a definition of mixed methods research”, Journal of Mixed Methods Research, 1, 2007, 112-33. The following article combines focus groups and questionnaires in order to study political socialization: Janette Habaschi and Jody Worley, “Child geopolitical agency, a mixed methods case study”, Journal of Mixed Methods Research, 3(1), 2009: 42-64.Online
  • [4]
    Joshua Gamson, Freaks Talk Back: Tabloid Talk Show and Sexual Non-Conformity (Chicago: Chicago University Press, 1998); Rosaline S. Barbour and Jenny Kitzinger, Developing Focus Group Research: Politics, Theory and Practice (London: Sage, 1999).Online
  • [5]
    Translator’s note: “Argumentation structure” is used throughout to translate the French “argumentaire”, which describes a combination of several arguments into a broader “line of argument” used to back up a given viewpoint.
  • [6]
    François Dubet, Valérie Caillet, Régis Cortéséro, David Mélo and Françoise Rault, Injustices: L’expérience des inégalités au travail (Paris: Seuil, 2006), 36-7.Online
  • [7]
    See in particular John G. Peters and Susan Welch, “Political corruption in America: a search for definition and a theory; or if political corruption is in the mainstream of American politics, why is it not in the mainstream of American politics research”, American Political Science Review, 78, September 1978, 974-84; Michael Johnston, “Right and wrong in American politics: popular conceptions of corruption”, Polity, 18(3), 1986, 367-91; Michael W. Jackson and Rodney Smith, “Inside moves and outside views: an Australian case study of elite and public perceptions of political corruption”, Governance, 9(1), 1996, 23-42.
  • [8]
    See our presentation of these works: P. Bezes and P. Lascoumes, “Percevoir et juger la ‘corruption politique’…”.
  • [9]
    Maureen Mancuso, Michael M. Atkinson, André Blais, Ian Greene, and Neil Nevitte, A Question of Ethics: Canadians Speak Out (Toronto: Oxford University Press, 1998).
  • [10]
    Later we will show that the same scenarios are generally judged more severely in answers to the questionnaire than during the focus groups.
  • [11]
    Pierre Lascoumes (ed.), “Les Français et la corruption politique: des représentations ambiguës”, Report Phase I, Cevipof, Rapport ANR, March 2007; P. Lascoumes (ed.), Favoritisme et corruption à la française…
  • [12]
    Cf. P. Bezes and P. Lascoumes, “Percevoir et juger la ‘corruption politique’…”, 775-8.
  • [13]
    We use the term focus group because of the organization and properties of our study model. Our collective interviews are centered around a common exercise and task – the classification and the discussion of scenarios describing more or less ordinary political behaviors – which freely admits the constructed and experimental nature of the interactions thus produced (see below). In addition, the groups are composed on the basis of homogenous professional and social status. On these points, cf. Sophie Duchesne and Florence Haegel, L’entretien collectif (Paris: Nathan, 2004), 8-25.
  • [14]
    David L. Morgan, “Focus groups as Qualitative Research”, Qualitative Research Method Series, 16 (London: Sage, 1997); Michael Bloor et al., Focus Groups in Social Research (London: Sage, 2001); Jonathan White, “Thematization and collective positioning in everyday political talk”, British Journal of Political Science, 39, 2009, 699-709; R. S. Barbour and J. Kitzinger, Developing Focus Group Research…; S. Duchesne and F. Haegel, L’entretien collectif.
  • [15]
    Raymond Boudon, Le juste et le vrai: Études sur l’objectivité des valeurs et de la connaissance (Paris: Fayard, 1995), 337.
  • [16]
    F. Dubet et al., Injustices…, 9.
  • [17]
    Two groups of manual laborers and low-level employees: one aged 19-25 (mason, cook, iron worker, hairstylists, manual laborer, clothing workers), the other aged 40 and over (production technician, warehouse worker, secretary, supervisor, nurse, temporary mason, shipbuilding welder); one group of white-collar workers and of mid-level business people (business manager, administrative manager, prison manager, nutrition and diet store manager, commercial manager); one group of traders, artisans, and small business-owners (printer, telephone and automobile dealers, taxi, tile layer); one group of students (first through third years at the Université de Créteil); three groups of managers and liberal professionals, one young (around 30 years old – engineer, architects, IBM managers), the other two older (40-55 years – market research managers); finally, three groups working in public administration (two groups of civil servants and one group of local government workers, working at the higher end of the scale – management director and administrator of a science lab, member of the Paris city council, head of apprenticeship projects, assistant director at the Ministry of Infrastructure, manager of a center for legal information, office manager at the Ministry of the Interior, director of a hospital, etc.). Seven focus groups were undertaken in Paris, four outside (Toulouse, Nantes, Lille).
  • [18]
    Danielle Bütschi, Le raisonnement dans les processus démocratiques, le questionnaire de choix (Paris: L’Harmattan, 1999), 27-34.
  • [19]
    Supposed to be “not very serious”: asking for a recommendation letter from the mayor’s wife to obtain a place in a nursery. Supposed to be “very serious”: a mayor grants a water contract in his city to a local company in exchange for financing his campaign report. The range of proposed scenarios is presented in Annex 1.
  • [20]
    Groups were composed in three different ways, always on a voluntary basis: our own ad hoc groupings, those done through continuing education programs for public- and private-sector managers, and groupings by a research agency with in-depth experience of panel composition.
  • [21]
    Very serious, serious, not very serious, not serious at all.
  • [22]
    Examples of scenarios (presented on cards): no. 4 (nursery): “To get a place in a nursery, one of your friends asks for a meeting with the wife of the mayor”; no. 11 (lying): “To get re-elected, a mayor hides the extent of the city’s budget deficit in his electoral campaign by fixing the accounts of the mayor’s office”; no. 8 (fiscal): “Following a fiscal audit, one of your friends proposes to give a sum of 10,000 euros to his elected parliamentary representative so that he will intervene in the friend’s favor with the new fiscal administration.”
  • [23]
    F. Dubet et al., Injustices…
  • [24]
    William Gamson, Talking Politics (New York: Cambridge University Press, 1992).
  • [25]
    Elizabeth Frazer, “Teenage girls talking about class”, Sociology, 22(3), 1988, 343-58.
  • [26]
    Michael Billig, Talking of the Royal Family (London: Routledge, 1992).
  • [27]
    P. Bezes, P. Lascoumes, “Percevoir et juger la ‘corruption politique’…” 778-80.
  • [28]
    Undertaken using a representative sampling of 2,021 people interviewed directly.
  • [29]
    P. Lascoumes and P. Bezes, “Les formes de jugement du politique…”.
  • [30]
    The enunciations from the corpus and the coding resulting from content analysis were brought together in an Access database. This database, associated with many types of software (Excel, Statistica, SPSS), then allowed for statistical treatment of the data.
  • [31]
    René L’Écuyer, Méthodologie de l’analyse développementale de contenu: Méthode GPS et concept de soi (Québec: Presses de l’Université du Québec, 1990).
  • [32]
    r and p correspond respectively to the coefficient of the linear correlation (the duration of enunciation relative to the number of letters it contains) and to the associated degree of statistical significance.
  • [33]
    A and SD correspond respectively to the average and the standard deviation of the number of characters per unit of meaning of the entirety of the corpus.
  • [34]
    P. Lascoumes (ed.), “Les Français et la corruption politique: des représentations ambiguës…”.
  • [35]
    The transformation takes place in the following manner: new index of severity (of initial individual judgment or of the judgment from the quantitative study) = (response on the scale of 4 points -1)/3.
  • [36]
    While the comparison between the index of severity of the discussion and the two other indices of severity cannot be tested statistically (the data is of a different nature), it does nevertheless allow for the detection of certain convergences or divergences that it will be useful to analyze.
  • [37]
    Or, according to the modern version pronounced by Jean-Luc Mélenchon, “If only they’d all just take a hike” (“Qu’ils s’en aillent tous”, Paris: Flammarion, 2010).
  • [38]
    Moment of individual choice in the beginning of the group, moment of exchanges and responses in the quantitative study.
  • [39]
    Loïc Blondiaux, “Faut-il se débarrasser de la notion de compétence politique? Retour critique sur un concept classique de la science politique”, Revue française de science politique, 57(6), 2007, 759-74.
  • [40]
    Scenario no. 2: “Each year the mayor of your city sends a large box of chocolates to the women of his constituency for Mother’s Day.” No. 4: “To obtain a place in a nursery, one of your friends asks for a meeting with the wife of the mayor.”
  • [41]
    No. 7: “To save the political career of a friend, the mayor of your region lies under oath.” No. 15: “To get back his driver’s license that was suspended for drunk driving, a taxi driver offers 5,000 euros to an official so that he will intervene in the driver’s favor.”
  • [42]
    The questionnaire study (carried out on a representative sample of 2,021 people interviewed directly) included ten scenarios that were to be classified by order of seriousness.
  • [43]
    Arnold Heidenheimer, “Behavioral and normative perspectives on the incidence of corruption”, Political Corruption: Readings in Comparative Analysis (New Brunswick: Transaction Publishers, 1970), 18-28.
  • [44]
    P. Bezes and P. Lascoumes, “Percevoir et juger la ‘corruption politique’…”, 779.
  • [45]
    J. G. Peters and S. Welch, “Political corruption in America: a search for definition and a theory,…”; see also M. W. Jackson and R. Smith, “Inside moves and outside views…”.
  • [46]
    This result was obtained by calculating the coefficient of partial correlation between the index of severity and the length of the disapproving argumentation, the overall time of the argumentation having been controlled for [r (40) = .64, p .0001], r corresponding to the coefficient of the partial correlation and p to the degree of significance that is associated with it.
  • [47]
    r corresponds to the coefficient of linear correlation (between the initial severity index and the discussion severity index) and p to the degree of statistical significance with which it is associated. The value of this coefficient of correlation reveals an average linear correlation. One may also note that several points disperse on the right hand side of the regression. This is due to the fact that each of our indexes is subject, in different ways, to the effect of 1) several sources of variation (such as the scenario, the seriousness of the scenario, the social group) and 2) their interactions. We examine these effects further below.
  • [48]
    Cass R. Sunstein, “The law of group polarization”, in John M. Olin, Law & Economics Working Paper no 91 (2D series) (Chicago: The Law School, The University of Chicago, 12 July 1999); and Going to Extremes: How Like Minds Unite and Divide (New York: Oxford University Press, 2009).
  • [49]
    C. R. Sunstein, “The Law of Group Polarization”, 16.
  • [50]
    We have not measured the impact of the discussion on the evolution of individual judgments. To do this it would have been necessary to take a similar vote after the discussion.
  • [51]
    See P. Lascoumes (ed.), Favoritisme et corruption à la française…; for in-depth studies, see “Prologue”, 21-65, and for the questionnaires, see “Les définitions concurrente de la probité publique”, 73-9.
  • [52]
    For all of these results, see Bruno Cautrès and Jean Chiche, “Le rôle des clivages sociaux dans les jugements sur la probité”, in P. Lascoumes (ed.), Favoritisme et corruption à la française…, 107-24.
  • [53]
    Translator’s note: The École Nationale d’Administration (ENA) is France’s central academic institution for the education of the future political elite. Most (if not all) individuals working in ministerial or senior administrative roles have been educated at this postgraduate institution.
  • [54]
    Translator’s note: A French satirical newspaper that often exposes political scandals.
  • [55]
    P. Lascoumes (ed.), “Les Français et la ‘corruption politique’…”.
  • [56]
    An average of 2.74 on the scale of 1 (not serious) to 4 (very serious) in the judgment of scenarios.
  • [57]
    Concerning the analysis of severity according to socioprofessional status, the data from groups sharing the same status were combined. In each case, we first verified that these groups were not different regarding the severity of judgments.
  • [58]
    See P. Lascoumes and P. Bezes, “Les formes de jugement du politique”, 139-42.
  • [59]
    P. Lascoumes and P. Bezes, “Les formes de jugement du politique”.
  • [60]
    R. L’Écuyer, Méthodologie de l’analyse développementale des contenus…
  • [61]
    R. L’Écuyer, Méthodologie de l’analyse développementale des contenus…, 65-6.
  • [62]
    Average = 17.8 arguments per grid. Standard deviation = 4.4 arguments.
  • [63]
    After ensuring the feasibility of this analysis and determining the number of components to draw out, we examined the solution a) for an orthogonal rotation and b) for an oblique rotation of the components. Conforming to the recommendations of many authors (Elazar J. Pedhazur and Liora Pedhazur Schmelkin, Measurement, Design, and Analysis: An Integrated Approach (Hillsdale: LEA, 1991); Kristopher J. Preacher and Robert C. MacCallum, “Repairing Tom Swift’s electric factor analysis machine”, Understanding Statistics, 2, 2003, 13-32), we used the oblique solution when the components were significantly correlated and the orthogonal solution when the components were independent. Since none of the eight oblique solutions revealed a correlation between the components, we chose the orthogonal solution each time (cf. Annex 4 for an example). Additionally, certain arguments were removed from the analysis when their frequency of appearance was too low.
  • [64]
    These factor loadings correspond to the correlations between the variables observed (i.e. the arguments) and the principal components.
  • [65]
    R. L’Écuyer, Méthodologie de l’analyse développementale des contenus…, 65-6.
  • [66]
    Since the argumentation structures drawn out in the preceding stage were a) composite by nature (i.e. composed of diverse arguments that the participants spontaneously associated with one another) and b) nontransversal (i.e. unique to each scenario), we decided to use a method of empirical classification in order to compose our argumentative repertoires. Nevertheless we were aided by previously developed PCA (component matrices). The classification of an argumentation structure was facilitated by an examination of the factorial loading of the arguments that constituted it.
  • [67]
    Four identified repertoires: legalist, contractualist, pragmatist, and realist (cf. above).
  • [68]
    According to a new data analysis technique currently underway (in which we perform a regression on each index of severity for the scores of the eight repertoires), it is possible to show two types of associations. For the initial judgment of severity (before the discussion) the LEGALISM repertoire is strongly linked to severe judgment (beta = +.43) and the repertoire WITHOUT EFFECT – UNCERTAIN is linked to tolerant judgment (beta = -.24). The situation is different for the discussions: NO MORALITY is linked to disapproving judgment (beta = +.27) and PRAGMATISM is linked to tolerant judgment (beta = -.37).
  • [69]
    Our database is closer to software aiding manual annotation of segments of text (CAQDAS, Computer-Assisted Qualitative Data Analysis Software).
  • [70]
    The authors gratefully thank Sophie Duchesne and Florence Haegel as well as the anonymous readers from the RFSP for their comments and very useful suggestions on a first draft of this article.


Qualitative research based on focus groups offers valuable insights in analyzing moral conflicts. This article suggests that it also proves very useful in exploring citizens’ perceptions of political misbehaviour. Moral expectations regarding politicians’ behaviour are ambiguous judgements because of the various criteria that underpin their perceptions of politics. This article is based on the analysis of eight different, socio-culturally homogeneous focus groups. The data collected are analysed through a quantitative method examining discursive contents. The article demonstrates the existence of a common grid for ranking political misbehaviour The moderating effects of collective discussion on judgements are also outlined. The article emphasizes variations in judgments, differentiating groups according to their social and occupational experiences. Finally, different “argumentative repertoires” are identified relying on different principles.

Jean-Michel Lecrique
A social psychologist and a doctor of psychology, Jean-Michel Lecrique is a member of the Laboratoire parisien de psychologie sociale (LAPPS, Parisian research centre for social psychology, Université Paris Ouest). His work focuses primarily on the implicit measurement of attitudes and on the study of psychosocial processes leading to changes of attitude and/or of behavior (persuasion, submission without pressure, cognitive dissonance). He has frequently collaborated on applied research projects in the health field: for example, on quitting smoking for the Interministerial panel against drugs and addiction (Mission interministérielle de lutte contre les drogues et la toxicomanie); and on the psychosocial determinants of health behaviors for the National institute for prevention and health education (Institut national de prévention et d’éducation pour la santé) (
Pierre Lascoumes
Pierre Lascoumes is a research director at the CNRS. He is affiliated to the Center of European Studies (Centre d’études européennes – CEE) at Sciences Po Paris. He gained extensive experience in the field of sociology of law and justice, before specializing in public policy analysis. He has worked on the history and implementation of environmental policies, and has carried out research on the reforms of environmental administration and on “precautionary” policies (REACH regulations). He has also worked on policies aimed at regulating economic and financial misdemeanors. He now leads research projects at the CEE on social representations of corruption and movements against cover-ups. He has recently edited Favoritisme et corruption à la française: Petits arrangements avec la probité (Paris: Presses de Sciences Po, 2010); and published Une démocratie corruptible: Arrangements, favoritisme et conflits d’intérêts (Paris: Le République des idées/Seuil, 2011) (
Philippe Bezes
Philippe Bezes is a researcher at the CNRS, affiliated to the Center for research on administrative and political science (Centre d’études et de recherches de sciences administratives et politiques – CERSA), at the Université de Paris II – Panthéhon Assas, and a teacher at Sciences Po Paris. His work focuses on the history and sociology of state reform policies both in France and in a comparative perspective, and on the transformation of bureaucracies. For many years he has also been interested in the moral dimension of the relationships between citizens and politics within the framework of a collective research program directed by Pierre Lascoumes on representations of public integrity. He has recently published Réinventer l’État: Les réformes de l’administration française (1962-2008) (Paris: PUF, 2009) (Le Lien social series); and (with Odile Join-Lambert) “Comment se font les administrations. Analyser des actes administratives constituants”, Sociologie du travail, 52(2), 2010, 133-50. He is currently working on the transformation of the French administration brought about by the reorganization introduced under the Révision générale des politiques publiques (RGPP) and by management-based logics (
Translated from French by 
Jasper Cooper
This is the latest publication of the author on cairn.
Latest publication on cairn or another partner portal
Uploaded on on 03/03/2014
Distribution électronique pour Presses de Sciences Po © Presses de Sciences Po. Tous droits réservés pour tous pays. Il est interdit, sauf accord préalable et écrit de l’éditeur, de reproduire (notamment par photocopie) partiellement ou totalement le présent article, de le stocker dans une banque de données ou de le communiquer au public sous quelque forme et de quelque manière que ce soit.
Loading... Please wait