This paper analyzes the debates surrounding the privatization of health services financing in Quebec. The objective is to clarify policy-making processes with regard to this important issue and, more generally, to provide a realistic understanding of health-related policy processes in Canada. The analysis is based on a large and continuous sample of mass media and National Assembly debates on the question during the four-and-a-half years following the Chaoulli ruling of the Supreme Court of Canada. These data are used to test four hypotheses about relationships among the types of political actors involved, their policy preferences, the rhetoric they use and the anticipated policy effects they assert. The results are applied to a discussion of questions about the factors that influence the effectiveness of political communication.

Political debate about healthcare financing is hardly new. Over the past 20 years, healthcare financing has become a central feature of political debates across Canada (Evans 2002; Hurley and Guindon 2008), especially in Alberta (Government of Alberta 2006; Mazankowski 2001; Rachlis 2000) and Quebec (Béland et al. 2008; Castonguay 2008; Contandriopoulos and Bilodeau 2008). In June 2005, in a very close four-against-three majority, the Supreme Court of Canada (2005) ruled in Chaoulli v. Quebec that the prohibition of substitutive private insurance for medically necessary services available under medicare in Quebec violated Quebec's charter of rights (Flood et al. 2005; Maioni and Manfredi 2005). However, it soon became clear that the effects of that ruling were not so much legal as political. The ruling opened a policy window (Kingdon 1984) prompted by a strong surge in the overall salience of the issue, resulting in related legislative activities (Castonguay 2008; Government of Quebec 2006a,b, 2008). In this paper, we use the post-Chaoulli healthcare financing policy debate as a revelatory case study to understand health-related policy processes in Quebec and Canada.

Our analysis is based on a large and continuous sample of the publicly observable components of policy-making processes during a four-and-a-half-year period. The data were used to test four hypotheses regarding the interdependence of groups' preferences, policy options supported, rhetoric used and anticipated effects. We then compared our results against the evolution of relevant healthcare financing policies and practices in Quebec during the same period, to discuss factors that appear to influence the effectiveness of policy-making efforts.

Conceptual Framework

Contemporary models of policy processes have generally broadened the definition of policy making beyond the boundaries of governmental institutions, beyond the core set of decision-making legislators and beyond specific decisions (Heaney 2006; Heinz et al. 1993; Jones and Baumgartner 2005; Kingdon 1984; Sabatier 1999b; Sabatier and Jenkins-Smith 1993; Stone 2002). There is a growing consensus that the most realistic lens for understanding policy making is one that focuses on policy subsystems, where numerous participants of various kinds, all interested in a given policy issue, interact over the long term to further their interests or preferences in the elaboration of public policies (Jordan and Maloney 1997; Kingdon 1984; Nakamura and Smallwood 1980; Rhodes 1990; Sabatier 1999b; Sabatier and Jenkins-Smith 1993). Participants obviously include legislators, civil servants and interest groups, but also lay citizens, community organizations, journalists, academics and others (Abelson 2002; Baumgartner and Leech 1998; Contandriopoulos 2011; Dearing and Rogers 1996; Jordan and Maloney 1997; Kingdon 1984; Rhodes 1990; Sabatier and Jenkins-Smith 1999). Participants tend to cluster in more or less formalized coalitions according to their preferences regarding the issue at stake (Heaney 2006; Heinz et al. 1993; Jordan and Maloney 1997; Sabatier 1999a; Sabatier and Brasher 1993). Our conceptualization of the policy-making process is based on this broad view of policy making as an open phenomenon occurring, to a significant degree, in public arenas.

Another core component of our conceptual framework is the definition of policy activity. The conceptualization of policy activity we have adopted is directly influenced by the literature on lobbying, which argues that the single most important way to influence policy making is through strategic communication of information (de Figueiredo 2002; Kollman 1998; Milbrath 1960; Smith 1999). We conceive political communications as discourses where actors use rhetoric to confer as much legitimacy and plausibility as they can on the link between a given situation (problem), a given intervention (policy option) and given effects (Elder and Cobb 1983; Hardy et al. 2000; Majone 1989; Stone 2002; Terry 2001). The link between interventions and effects can either be positive ("we should do X because it would produce positive effect Y") or negative ("we shouldn't do X because it would produce undesirable effect Y"). This conception of policy making as a communication process and the underlying typology of message content are core components of the coding and analysis presented in the next section.

Data and Methods

As stated, our aim was to document and analyze the healthcare financing policy debate in Quebec following the Chaoulli ruling of the Supreme Court. Data were collected over a continuous four-and-a-half-year period between June 2005 (Chaoulli v. Quebec ruling) and January 1, 2010. Data sources can be divided into two groups. The first consists of data pertaining to formal political debates, specifically the transcripts of all discussions in Quebec's National Assembly as well as transcripts of all discussions within the Social Affairs Committee, the parliamentary committee in charge of all health-related legislative and regulatory activity. The second source comprises a broad sample of mass media in print, radio and television. This sample includes all daily newspapers in Quebec (excluding tabloids, owing to difficulties in accessing these data), two major pan-Canadian daily newspapers and all news-related radio and television shows indexed in the Eureka.cc media content database.

Each data set includes hundreds of thousands of pages of transcripts. The first step was thus to design, test, refine and use a set of search phrases (keywords with Boolean operators) to perform a preliminary skimming of the data. The goal at this stage was to be as sensitive as possible while retaining sufficient specificity to achieve the purpose. In all, 53 different search phrases were used and their results combined. In a second step, the documents were manually sorted on the basis of their relevance to the issue at hand. We retained only documents dealing with issues related to our definition of healthcare financing (see discussion of coding below). Ultimately, 1,330 documents were selected for analysis.

All data were then entered into an QSR NVivo 8 database and manually coded using a tree-based node system derived from the conceptual framework presented above. The tree structure of the nodes started with four branches: "groups" (who is speaking); "policy proposals" (what policy options are put forward); "rhetoric" (how the argument is framed and supported) and "effects" (the asserted effect of the proposal, according to the speaker). Inside this tree structure, nodes were inductively developed, discussed, adjusted and merged throughout the coding and analysis phases. Figure 1 offers a basic summary of the final coding structure.

The data analyzed included statements from 1,625 different persons, for many of whom (especially lay citizens) we had only one statement. At the other extreme, we had 589 different statements from a single individual, who was Quebec's health minister during most of the period studied. All individuals in the data set were sorted into 11 subnode groups which, in turn, were grouped into three larger groups: Policy Makers, Bystanders and Interest Groups. The logic behind the group taxonomy rests on multiple theory-based criteria related to control over the policy (policy makers versus others), the nature of the speaker (institutional spokespersons versus individuals), the nature of their stake in the debates (direct financial stake versus value-based preferences) and the reason for their involvement (professional or personal) (Abelson 2002; Baumgartner and Leech 1998; Contandriopoulos 2011; Jones and Baumgartner 2005; Nakamura and Smallwood 1980; Sabatier and Brasher 1993). The research team's knowledge about this policy debate was used to attribute each speaker to the appropriate group. When doubts were raised, web search engine results were used to gather the necessary information.

The second main coding node is the nature of each participant's policy proposal. All statements in the data set that could be related to either defining or supporting/opposing an identifiable policy proposal were coded. As discussed earlier, we adopted a broad definition of policy proposals as explicit pleas supporting or opposing a given path of action. The main branches of this node are, predictably, divided according to whether the policy proposals favour private or public financing of healthcare services. We define privatization of healthcare financing as any measure that can foreseeably increase the proportion of healthcare expenditures coming from sources outside the public sector, whether insurance-based or out-of-pocket. However, because it is possible to manage private-like insurance schemes in the public sector, we also consider akin to privatization any transfer of income-related funding towards use-based or risk-based funding (Colombo and Tapay 2004; OECD 2004a,b; Tuohy et al. 2004). Concretely, some statements in the data set were limited to simple affirmations of a preference for or against privatization and were therefore coded generically to the pro-private or pro-public main nodes. Other statements provided clear, operationalizable propositions (e.g., ER patients should incur a co-payment of $20 per visit). The structure of the tree node was defined inductively through team discussions to merge similar proposals. The sub-branches, as listed in Figure 1, are idiosyncratic both to Quebec's existing regulatory and legislative structures for healthcare delivery and financing and to long-standing policy options in the debate (Deber 2002; Evans 1997a,b; Evans et al. 1993).

The third main coding node is the anticipated effect used to justify the proposal in the statement coded. Again, the level of specificity varied greatly. Some statements included explicit anticipated effects (e.g., "we should do X because doing so will increase accessibility to service Y"), while others remained extremely vague and referred implicitly to undefined positive or negative effects. The definition of the tree structure was inductive during the coding, but at the end of the process there were a lot of subnodes with low occurrences, which we merged into broader, conceptually driven nodes (e.g., accessibility, costs, equity, etc.). Generic positive- and negative-effect subnodes were used to ensure that all statements were coded to the effect main node.

The fourth main coding node is the nature of the rhetoric used. The process of defining tree node structure for the rhetoric node was very similar to that for the effect node. An inductive process was initially used to identify different types of rhetoric, after which subnodes were inductively grouped into five conceptually driven ones. Arguments in the "subjectivity" subnode are framed according to a very basic "just trust me" rhetoric. Those in the "juridical" subnode are framed around the need to respect the laws. In the "comparative" subnode are arguments referring to what occurs in other countries or outside the health sector. The "evidentiary" subnode includes all arguments based on factual data. Finally, in the "realist" sub-node are arguments justifying a course of action using such logics as "it already exists" or "we can't modify this reality."

With few exceptions, all statements were simultaneously coded to the four main nodes presented above. Some statements were coded to more than one subnode where appropriate, hence the variation in the totals presented in the tables in the appendix. This data coding framework allowed the use of NVivo matrix coding query functions to build contingency across tables for the main node sets.

The data were analyzed in the context of a larger project and examined both longitudinally over time, to analyze the processes related to the evolution of the policy issue, and transversely, to analyze the interrelations between each node treated as a variable. The present paper focuses on the transverse analysis. It offers robust descriptive evidence on the nature and content of the healthcare financing debate in Quebec. We also used the data to test four main hypotheses, which were that (1) groups differ in terms of policy preferences; (2) they show preferences for specific rhetorics; and (3) the rhetorics they use and (4) the anticipated effects they assert are both linked to their preferred policy options.

Results and Discussion

The four hypotheses presented above were tested through simple chi-square tests of the contingency tables produced from NVivo matrix coding queries. Hypotheses were tested using both main nodes and subnodes. In each case, chi-square probabilities were below 0.001 for all hypotheses. The data analyzed thus strongly support our four hypotheses. Real and expected contingency tables for the four main hypotheses are presented in the appendix, and many more such tables were used to test hypotheses at the subnode level. Complementary tests were also conducted for each data source (media versus National Assembly), and their results are similar to those obtained with the aggregated data presented here.

Some of the hypotheses tested are intuitive (e.g., it is not surprising that owners of private clinics or the banking and insurance industries would favour more private financing), but the full picture emerging from detailed analysis of the contingency tables provides relevant insights into the content of policy-related discourses at the core of the healthcare financing debate. In the following sections we discuss these insights. Then, we conclude by comparing the analysis of our results with the evolution of healthcare financing policies and practices in Quebec to discuss some factors that could influence the effectiveness of policy-making activities.

Hypothesis 1: Group affiliation is linked to policy preferences

(See Table 1 in appendix at https://www.longwoods.com/content/23005)

An interesting characteristic of the post-Chaoulli policy debate on healthcare financing is that it was almost perfectly balanced. Overall, 48% of the statements were coded at pro-private nodes and 52% at pro-public ones. Such a pattern could be explained in part by the nature of mass media reporting, where issues tend to be framed dialectically and journalists seek out opponents to any idea put forward in reports. However, because the same patterns characterize the data from both sub-arenas, mass media reporting alone cannot explain the findings. Likewise, available poll results offer a similar picture of evenly divided opinion on the issue (Contandriopoulos and Bilodeau 2008).

When analyzing the data at the group level, unsurprisingly, group affiliation explained a significant portion of policy preferences. Private clinic owners and managers voiced the most unambiguous opinion on the desirability of privatization, but the publicly expressed opinions of justice court members were also skewed in favour of pro-private policy options. This finding is likely explained by the fact that most statements from actors in this latter group were defending the Chaoulli v. Quebec ruling. Equally unsurprising was academic experts' strong skew in favour of pro-public financing (because their two preferred options were maintaining the status quo and increasing the efficiency of public systems). This rationale is likely due to this group's awareness of the strong body of evidence on the programmatic effects of private healthcare financing. Most other groups (public servants, politicians, lay citizens, journalists, think tanks and similar institutions, as well as unions and union-like groups) overall produced a relatively balanced number of statements. For unions, there was in fact a clear divide between vocal physicians' unions and associations (such as the FMSQ [Quebec Federation of Specialist Physicians], the Canadian Medical Association, the AMQ [Quebec Medical Association], and others) on the pro-private side and all other unions (Confédération des syndicats nationaux [CSN], Fédération des travailleurs du Québec [FTQ], etc.) on the pro-public side. Physician-based organizations produced 107 pro-private and 76 pro-public statements; other health professional organizations produced 17 pro-private and 53 pro-public statements, while all other unions combined produced 5 pro-private statements and 105 pro-public ones.

Hypothesis 2: Groups display preferences for specific rhetorics

(See Table 2 in appendix at https://www.longwoods.com/content/23005)

In terms of rhetoric used, more than half of the statements (51%) employed little more than a "just trust me" subjective rhetoric. Evidence was used as a rhetorical argument in only 11% of the statements. Notwithstanding its low overall prevalence, academic experts, think tanks and public servants were the ones who relied most on an evidence-based rhetoric. Again, as expected, justice court members tended to use a juridical rhetoric, and journalists tended to use more of a comparative rhetoric. The "just trust me" subjective rhetoric was especially favoured by politicians, who showed a particularly low recourse to evidence-based rhetoric.

The portrait presented by hypotheses 1 and 2 is convergent with previous studies showing that the healthcare financing policy debate in Canada is far from being a rational, evidence-based examination of the programmatic impact of given policy options (Béland et al. 2008; Deber 2002; Evans 1997b, 2005; Flood et al. 2005; Quesnel-Vallée et al. 2006). Rather, as any political science–inspired framework would suggest, the debate looks like a struggle by structured groups to impose their own preferences and opinions by legitimizing categories, ideas and goals (Stone 2002), something we will discuss in conclusion. Factual presentation of policy options and objective examination of their likely programmatic effects, described by Stone (2002) as the "production model" of policy making, were an insignificant part of the data analyzed here.

Hypothesis 3: Reliance on a given rhetoric is linked to policy option preferences

(See Table 3 in appendix at https://www.longwoods.com/content/23005)

Using the same data, we also analyzed the link between rhetorical choices and policy options. Subjectivity remained the most prevalent kind of rhetoric and was almost perfectly shared by both the pro-private and pro-public sides. The most striking difference in the rhetoric used by the two policy sides is their differential reliance on an evidence-based rhetoric. When pro-public and pro-private policy options are disaggregated into subnodes, it is striking to note that real occurrences were below expected ones for the evidence-based rhetoric for all of the eight pro-private nodes, and above expected occurrences for all but one of the eight pro-public ones. The only pro-public node not associated with an evidence-based rhetoric consisted of appeals for the Quebec government to use the notwithstanding clause to ignore the Chaoulli v. Quebec ruling; these appeals, as might be predicted, were defended through a juridical rhetoric. Although we did not expect the pro-private statements to have a natural affinity with an evidence-based rhetoric, the unambiguous link between an evidence-based rhetoric and the public–private policy divide was much clearer than expected.

Overall, our data showed that, on one side of the debate, the pro-private arguments were mostly based on comparisons and a juridical rhetoric, while on the other, pro-public arguments were more often based on evidence and realist rhetoric. Some parts of the debate, notably the parallel "pro–user fees" and "anti–user fees" arguments, showed a higher-than-expected reliance on a subjective "just trust me" rhetoric. The arguments in favour of increasing healthcare human resources, opposing private insurance, and improving the efficiency of the public system showed a higher-than-expected reliance on an evidence-based rhetoric.

Hypothesis 4: The nature of asserted anticipated effects is linked to policy option preferences

(See Table 4 in appendix at https://www.longwoods.com/content/23005)

The last hypothesis focuses on the asserted links between policy options and their anticipated effects. Our conceptual framework suggests that statements to support a given policy option will use rhetoric to legitimize a potential link between this option and a set of anticipated effects. Results of the chi-square tests showed there is indeed a clear link between policy options and anticipated effects. In terms of specific effects, the pro-private arguments largely suggested a positive link between more private financing and greater accessibility, combined with reduced costs. Given that the available scientific evidence suggests the opposite, it is understandable that, as seen earlier, the "evidence-based" rhetoric was not favoured in those statements. Conversely, the pro-public arguments suggested more often than expected a link between public financing and better distribution of healthcare human resources and greater equity, which is more in line with the scientific evidence on the topic.

In addition to the specific associations described above, an important conclusion of our analysis is the very strong association between pro-private arguments and generic positive effects and, conversely, the very strong association between pro-public arguments and generic negative effects. We coded statements to generic positive effects when the speaker suggested that a given path of action was desirable because its unspecified or idiosyncratic effects were desirable. Likewise, statements were coded to generic negative effects when the speaker suggested that a given path of action was undesirable because its unspecified or idiosyncratic effects were detrimental. Our results show that the pro-public policy arguments were mostly defensive (e.g., "we should keep a publicly financed system because doing otherwise would be detrimental"), while the pro-private arguments were generally presented positively (e.g., "we should develop private financing because doing so would bring desirable effects"). This held true for every one of the eight pro-private and eight pro-public individual policy options when the analysis was done at the subnode level. In fact, increased equity in healthcare services access was the only positive effect that exhibited a higher-than-expected association with all the pro-public policy options, while all pro-private policy options were less associated than expected with increased equity.

The defensive nature of the pro-public discourse is probably explained in part by the nature of the policy window in our case, as the Supreme Court ruling had just set the stage for more private financing, leaving groups opposing the idea in a defensive position. As we discuss in our conclusion, this stance could also influence the effectiveness of pro-public political communication.

Conclusions

During the four-and-a-half years covered by the data analyzed, Quebec adopted many policies related to healthcare financing. In 2006, Bill 33, which was the government's direct answer to the Chaoulli ruling, modified existing laws to allow substitutive private insurance for hip and knee surgery and cataract removal (Government of Quebec 2006a). It also included legislative modifications to strengthen private-sector delivery of ambulatory care, and was further enhanced in 2008 by a list of interventions to be offered in the private sector (Government of Quebec 2008) that included dozens of others besides the three targeted in 2006. In 2010 the provincial budget included a fixed per capita contribution to healthcare financing that corresponded with our definition of private financing (Government of Quebec 2010). During the same period, media reports of illegal billing practices at many private clinics were followed by discussions between medical associations and the government, with a resultant broadening of permissible complementary fees for medicare-covered services offered in the private sector (FMSQ 2007).

Overall, although our data show that the policy debate was remarkably balanced between the pro-public and pro-private sides, both policies and actual practices relating to healthcare services financing evolved in a clearly pro-private direction over the period studied here. This observation raises the question of how to explain this one-sided evolution of policies. One explanatory factor is related to the convergence between the ideological preferences of the elected party during that period and those of most pro-private groups. Politics are deeply structured by ideologically based coalitions (Heinz et al. 1993; Sabatier 1999b), and thus statements made by political foes are unlikely to be as effective as those made by political friends.

The second explanatory factor is more directly linked to our conceptual framework and its core idea that what determines the effectiveness of political communication is probably not only the number of statements but also their relative impact, mediated by such elements as the efficacy of the rhetoric used or the credibility/desirability of the asserted causal link between the policy option and specific effects.

Regarding the nature of the asserted anticipated effects of the policy options put forward, an important contribution of our study is to show the deep divergence between the pro-public and pro-private discourses. While the pro-private discourses were centred on assertions of positive effects ("private financing would produce the desirable effect X"), the pro-public discourses were centred on assertions of negative effects ("not relying on public funding would produce the undesirable effect X"). The nature of the data we analyzed does not allow us to determine whether this core characteristic of the visible policy debate on healthcare financing is related to the differential success of the pro-private and pro-public sides in policy implementation. We nevertheless suggest as a hypothesis that positive discourses are potentially more effective than negative ones.

Regarding the rhetoric used, evidence-based rhetoric was much more directly associated with the pro-public than the pro-private side. This finding is due, at least in part, to the fact that quite a few pro-public groups (academic experts being the most obvious example) had a strong affinity with this rhetoric. Regarding the effectiveness of political communication, many well-supported frameworks on policy-making processes suggest that the best way to influence policy process is by framing the basic categories and core elements used to think about the issue at stake (Considine 1998; Dearing and Rogers 1996; Kingdon 1984; Majone 1989; Rogers et al. 1993; Sabatier 1999b; Sabatier and Jenkins-Smith 1993, 1999; Stone 2002). Such framing is achieved by making statements that, explicitly or implicitly, legitimize categories and ideas convergent with the policy options defended. Examples of such statements from the pro-private side would be, "some users abuse the system," "public finances are crumbling," "private companies are efficient by nature," and so on. These examples also highlight that it matters very little whether what is described is true or not; what matters is whether the audience perceives it as plausible. In that regard, many (Edelman 1977; Majone 1989; Stone 2002) suggest that evidence-based rhetoric may not be especially effective to frame policy issues. In fact, a subjective rhetoric ("I'm describing a plain reality that everybody knows about") is probably more effective than an evidence-based rhetoric ("these data show I'm right and things are not as they appear"). This point is well made by Majone in his argument that policy making is about dialectic and rhetoric: "The starting point of a dialectic argument is not abstract assumptions but points of view already present in the community; its conclusion is not a formal proof, but a shared understanding of the issue under discussion; and while scientific disciplines are specialized forms of knowledge, available only to the experts, dialectic can be used by everyone…" (Majone 1989: 6).

La politique visible du débat sur la privatisation au Québec

Résumé

Cet article fait l'analyse du débat autour de la privatisation du financement des services de santé au Québec. L'objectif est de clarifier les processus d'élaboration de politiques quant à cet enjeu d'importance et, de façon plus générale, de mieux comprendre le processus d'élaboration des politiques de santé au Canada. L'analyse se fonde sur un vaste échantillon continu provenant des médias de masse et des débats sur le sujet à l'Assemblée nationale au cours des quatre années et demie qui ont suivi la décision de la Cour suprême du Canada sur l'affaire Chaoulli. Ces données ont servi à tester quatre hypothèses au sujet de la relation entres les types d'acteurs politiques impliqués, leurs préférences politiques, leur discours et les effets qu'ils attribuent aux différentes options politiques. Les résultats sont utilisés pour discuter des facteurs qui influencent l'efficacité des communications politiques.

Appendix

Contingency tables

The totals in the contingency tables vary, as some statements were coded to more than one subnode of the same main node. For example, Table 1 shows a total of 2,067 statements coded both at a given policy option and at a given group, while Table 2 shows a total of 2,186 statements coded at both a given policy option and at a given rhetoric. This difference is explained by the fact that some statements from a single person were coded to two different kinds of rhetoric. This is especially true in longer statements, where the speaker is likely to use more than one kind of rhetoric. Similar arguments explain other differences in the totals.

Comments

Be the first to comment on this!