Under any form of divided democratic authority, voters wishing to hold government to account for the results of government policy must judge how responsible was each authority for a given policy outcome. Judgments of responsibility intervene between evaluation of policy outcomes and voting behaviour. “Responsibility judgments are the principal mechanism through which citizens hold representatives accountable for their conduct” (Rudolph, Reference Rudolph2006: 99). Since multiple governments—sometimes all three levels—are involved in Canadian policy-making, voters face a daunting challenge in attributing responsibility to their governments. One may reasonably wonder if there is any useful information that can help voters assign responsibility. Cairns, in fact, went so far as to suggest that elites build federal institutions deliberately to foster confusion and blunt accountability (Reference Cairns1977: 708). Ultimately, this challenge may threaten electoral accountability in Canada and other federations (Richards, Reference Richards and Lazar1998; Smiley, Reference Smiley1987), yet little is known about the seriousness of this threat. To assess the quality of democratic accountability in Canada, therefore, we must understand how well voters are coping with this additional burden of federal citizenship.
Even before assessing the impact of responsibility judgments on voting, we must understand voters' judgments of responsibility. At one extreme, voters might be motivated to assess responsibility, be capable of doing so, and have good information enabling them to make such judgments. These voters would assign responsibility accurately and discriminate among the actions of multiple governments, ideally inducing those governments to make and implement good public policy. At the other extreme, however, voters might not want to go to the trouble of assessing responsibility, might not be capable of dealing with the inherent complexity, or might not have any decent information on which to base their responsibility judgments. If the latter seems more likely, one might ask whether shortcomings in citizens' responsibility judgments are due mainly to the inaccessibility of reliable information or to limitations in citizens' motivation and capacity to process political information.
Canadians could hardly be blamed if they were confused about responsibility. Canada is one of the most decentralized federations in the world (Watts, Reference Watts2005), the tax field is shared in complicated ways, the most significant social services are delivered by provinces but with significant block funding from the federal government, and there is a very complex revenue equalization formula. For a half-century, most Canadian policy has involved negotiations, co-operation, and conflict between the two senior levels of government (Cameron and Simeon, Reference Cameron and Simeon2002; Rocher and Smith, Reference Rocher and Smith2003; Simeon, Reference Simeon1973). Making the challenge even tougher, politicians can take advantage of intergovernmental entanglement and play the blame game (Hood, Reference Hood2002). For instance, First Nations leaders have long accused the federal and provincial governments of passing the buck, arguing about responsibility while living conditions on reserve remain poor (Prince and Abele, Reference Prince, Abele and Lazar2000). Governments in Canada even go so far as to mount advertising campaigns to heap responsibility on the other level of government. A recent example is the Premiers' Council on Canadian Health Awareness, which ran newspaper ads telling citizens that the federal government was not contributing enough money to provincial health care budgets.Footnote 1 Given this context, it is no surprise that surveys find that three-quarters of Canadians agree with the statement: ‘it is often difficult to figure out which government is responsible for what’.
This study presents the first comprehensive diagnosis of this situation, in Canada or beyond. Using original survey research designed for the purpose, it provides a nuanced account of the contours of Canadians' judgments of their governments' responsibility for policy outputs. The following section argues that judgments of responsibility are an integral part of the mechanism of electoral accountability and then reviews the rather sparse literature on citizens' allocation of responsibility. The first empirical section of the paper describes the distribution of responsibility judgments across fifteen policy domains in three electoral contexts. The remainder of the paper asks which citizens make better judgments than others in order to determine whether there is any relevant information that attentive citizens might use to make these judgments.
Divided Power and Accountability
Theoretical attention to the tension between divided power and clear lines of accountability dates back to Hamilton's Federalist Paper, number 70 (Rossiter, ed., Reference Rossiter2003; more recently, Alesina and Rosenthal, Reference Alesina and Rosenthal1995; Anderson, Reference Anderson2006; Downs, Reference Downs1999; Hood, Reference Hood2002; McKinnon and Nechyba, Reference McKinnon, Nechyba, Ferejohn and Weingast1997; Richards, Reference Richards and Lazar1998; Smiley, Reference Smiley1987; Weaver, Reference Weaver1986). If voters do not hold a government responsible for a policy outcome, then that government will not suffer for poor results or benefit from good ones. Citizens' judgments of governmental responsibility are therefore the linchpin of accountability in federal states. Failures of accountability are possible if confusion about responsibility distorts some voters' attempts to punish governments for policy failures. A contemporary recognition of this threat to accountability is to be found in Germany, where federal reforms in 2006—the most significant in that country's history—were aimed at disentangling levels of government and were principally motivated by concerns about voters' ability to point the finger for policy outcomes (Heckel, Reference Heckel2006).
A simple normative theory of vote choice in a federation is that a voter should weigh policy areas according to her concern about them. But if voters are not up to the challenge of attributing responsibility, one consequence might be that policy domains where responsibility is clearest would disproportionately dominate voters' decisions. Health care might get short shrift in voters' decisions because responsibility is much clearer in other policy areas—say, foreign affairs in federal elections and labour relations in provincial ones. A second distortion might be if voters treat both governments as necessary contributors to the policy outcomes, giving both levels full responsibility. This would lead to a form of exaggerated accountability, where both governments respond as if fully responsible and thus collectively overreact—for example, overcorrecting a policy that has upset voters. Harrison (Reference Harrison1996) argues that this may characterize governments' behaviour when public interest in the environment is at its peak. And it may account for deep budget and program cuts by both federal and Ontario governments in the mid-1990s, interpreted by many as an overreaction to high levels of government debt. Finally, a third problem might be that confusion, particularly if fostered by an intergovernmental blame game, might turn voters off. Voters might punish governments that induce uncertainty about responsibility. Or, worse, they might throw up their hands, saying, “Why bother voting in this election when I can't tell whether this government is responsible for any of the things I care about?”
It is not my aim in this paper to test for these distortions in Canadians' voting behaviour. They serve only to emphasize the importance of responsibility judgments in the formula of electoral accountability. This study takes the prior step required before evaluating these possibilities by describing voters' responsibility judgments.
Responsibility as Mediator
What, then, is known about citizens' attributions of responsibility? Empirical work demonstrates that attribution of responsibility mediates the effect of policy judgments on voting behaviour (Alesina and Rosenthal, Reference Alesina and Rosenthal1995; Anderson, Reference Anderson1995a; Anderson, Reference Anderson1995b; Anderson, Reference Anderson2000; Anderson, Reference Anderson2006; Atkeson and Partin, Reference Atkeson and Partin1995; Fiorina, Reference Fiorina1981; Kenney, Reference Kenney1983; Lau and Sears, Reference Lau and Sears1981; Lewis-Beck and Nadeau, Reference Lewis-Beck and Nadeau2000; Lewis-Beck and Paldam, Reference Lewis-Beck and Paldam2000; Leyden and Borelli, Reference Leyden and Borelli1985; Lowry, Alt, and Ferree, Reference Lowry, Alt and Ferree1998; McGraw, Reference McGraw1991; Nadeau and Lewis-Beck, Reference Nadeau and Lewis-Beck2001; Niemi, Stanley, and Vogel, Reference Niemi, Stanley and Vogel1995; Partin, Reference Partin1995; Peffley, Reference Peffley1984; Peffley and Williams, Reference Peffley and Williams1985; Powell and Whitten, Reference Powell and Whitten1993; Simon, Reference Simon1989). The hypothesis underlying these studies is that voters' electoral response to policy outcomes should be stronger as the institutional context promotes clarity of responsibility and as the voter's own judgments about the lines of responsibility are clearer. The obvious implication is that mechanisms of electoral accountability depend on clear judgments of responsibility by citizens.
Only a few existing studies take a step back to ask how responsibility judgments are formed. Iyengar differentiated causal (retrospective) and treatment (prospective) responsibility, arguing that citizens use the two types of judgment separately and independently in their political cognition. He also demonstrated that attributions of responsibility, like other political attitudes, are sensitive to framing effects (Reference Iyengar1989). Rudolph used survey research to show that attributions of responsibility are heavily coloured by partisanship, but also sensitive to features of the institutional context, such as divided government and the relative power of the executive and legislative branches (Reference Rudolph2003a; Reference Rudolph2003b). Separate experimental research confirmed this conclusion, showing that partisans devalue information relevant to responsibility attributions (Rudolph, Reference Rudolph2006). Relatedly, Arseneaux demonstrated that American voters do differentiate among the functional responsibilities of the three levels of government in coming to their attributions of responsibility (Arceneaux, Reference Arceneaux2006).
One limitation of nearly all of the empirical work on this topic, however, is that it stems from the economic voting literature.Footnote 2 This has meant a rather straightforward conceptualization of responsibility and its role as a mediator of judgments of performance. The economy is usually the most important policy domain and citizens and governments are assumed to want maximum economic growth. A more complete picture of responsibility judgments requires a view to other policy domains where policy goals are less clear and salience varies across voters and time.
A second limitation of existing attempts to understand the role of responsibility judgments in voting behaviour is crude measurement of responsibility. The few surveys that have asked explicit responsibility questions about more than one political authority ask which one is responsible or more responsible than the others (for example, Arceneaux, Reference Arceneaux2006; Johnston, Reference Johnston1986; Rudolph, Reference Rudolph2003b). But responsibility judgments are more nuanced than this, especially when policy is the result of conflict between, and co-operation among, centres of power. Responsibility should therefore be measured for each government separately. Existing questions have also assumed that respondents recognize that responsibility is zero-sum. As this paper will show, voters may attribute maximum responsibility to all relevant institutions if they believe that all were necessary causes of conditions that resulted from government policy.
Third, considerations of the role of responsibility judgments have also by and large neglected the fact that these judgments have two components: the level of responsibility and the certainty of that attribution. These two components may interact in complicated ways as a result of the political context. For instance, a voter who hears conflicting accounts of responsibility might not change the level of responsibility she attributes to political actors but she may become less certain that her attributions are accurate. One consequence might be that voters react to uncertainty by hedging their bets and maximally blaming all potentially responsible actors.
The research reported here addresses all of these shortcomings by measuring responsibility for federal and provincial governments separately, across a range of policy domains, with follow-up questions on the certainty of voters' attributions of responsibility.
Data and Hypotheses
The raw material for this investigation is survey data collected during the provincial election campaigns in Ontario and Saskatchewan in the fall of 2003 and then re-interviews of the same voters during and after the federal election campaign in June 2004.Footnote 3 I will refer to these as the provincial study or “wave” and the federal study or “wave” of this panel survey. The surveys were similar to national election studies but with a focus on retrospective evaluation of the state of the world and then the addition of responsibility judgments. A federal-provincial panel study provides crucial leverage for understanding responsibility, but, obviously, the cost of conducting separate provincial election studies for all provinces is prohibitive. While Saskatchewan and Ontario are not fully representative of Canadian provincial electoral contexts, the 2003 elections in these provinces are by no means atypical. More importantly, the impact of federalism on the policy context in these two elections was quite typical of Canadian elections, both federal and provincial. The policy issues and areas of government performance salient in these campaigns varied in terms of de facto and de jure jurisdiction—some mainly provincial, some mainly federal.
The Ontario survey asked specific responsibility questions in five areas: the state of health care, changes to health care, the economy, electricity, and taxes. In Saskatchewan, voters were asked about the economy, health care, and the farm crisis. In the provincial wave, respondents were also asked about the certainty of their responsibility judgments. In the federal election wave, voters in both provinces were asked about value for money in health care, social services, and the economy; and a fourth issue, different in each province, was the farm crisis in Saskatchewan and the budget deficit in Ontario.
Obviously, responsibility has many meanings and will be interpreted in various ways by survey respondents. Since the focus of the study was electoral accountability for policy outcomes, respondents were presented with “responsibility” in the context of questions about the policy results. In most cases, the responsibility question asked “How much responsibility does the government have for the way things are going?” in a given policy area. Responsibility was not elaborated or specified in the questionnaire so I assume that most respondents interpreted it to mean causal responsibility: how much had the government in question done to bring about the current conditions in that policy area. Undoubtedly, some respondents will combine this with treatment responsibility, a judgment about how much the government can or should do to improve the situation. In this study I make no effort to separate the two, assuming that the evaluation of the sources and quality of these judgments applies to responsibility in general as interpreted by citizens themselves.
In the provincial surveys, the responsibility scale provided to respondents was “zero to ten where zero means that the government is not at all responsible and ten means that the government is fully responsible. In the middle, five means that the government is partly responsible for what is going on.” The federal wave used a four-point verbal scale comprising “no responsibility, a little responsibility, a lot of responsibility, or full responsibility.” Part of the objective was to investigate the measurement properties of the two scales to evaluate the trade-off between useable variation and simplicity.
The following question sequence was used on the economy in Saskatchewan:
Now we would like to ask you some questions about how much responsibility the federal government and the Saskatchewan government have for some of the things going on in Saskatchewan at the moment. On all of these questions we will use a scale from 0 to 10 where zero means that the government is not at all responsible and 10 means that the government is fully responsible. In the middle, five means that the government is partly responsible for what is going on. If you're not sure, please say so.
Now I'd like to talk about the economy. What about Saskatchewan's economy? Over the PAST FEW YEARS, has SASKATCHEWAN's economy GOTTEN BETTER, GOTTEN WORSE, or STAYED ABOUT THE SAME?
Now, thinking about the economy. How responsible is the [random rotation: Saskatchewan/Federal] government for the way the economy is going in the province?
And how responsible is the [fill other government] government for the way the economy is going in this province?
How certain are you about how much responsibility each level of government has for the way the economy is going here? Are you very certain, somewhat certain, or not very certain?
It is important to note that all questions in both waves of the study identified the geographic locus of the policy outcomes as within the province; for example, “The overall state of the health care system in Saskatchewan” or “Overall, have taxes been going up or down for Ontarians in the last couple of years?”
The first set of propositions to be evaluated in this study concerns the expectations for absolute judgments of responsibility and variation across issues. If voters were fully informed and making expert responsibility attributions they should apportion responsibility in line with expert opinion. Table 1 provides this expert opinion using the 0 to 10 scale. The entries are averages taken from a web-administered survey of 33 Canadian political scientists who specialize in federalism or provincial politics. The experts were asked questions identical to those used in the public surveys.Footnote 4
Table 1. Experts' and Citizens' Mean Responsibility Judgments by Issue
In considering Table 1 and the results that follow it, it is important to keep in mind that responsibility pertains to the facts on the ground for a given issue in these two provinces at the time of the surveys, not to some general features of the issues. For instance, on electricity in Ontario in the year before the election both the supply and predictable pricing were threatened. Provincial, not federal, governments have long been responsible for ensuring the supply of electricity at fair prices, and section 92a of the Constitution Act, 1982 formalized this power. In practice, the Ontario government, like the Alberta government a few years earlier, had initiated well-publicized deregulation which was followed by a sharp rise in average prices and the threat of shortages. This led to a policy reversal on deregulation and an imposed cap on prices, all of which was coincidentally punctuated by the major blackout in northeastern North America one month before the election call.Footnote 5 By contrast, the federal government had not done anything to precipitate the problems in the industry. Not surprisingly, then, the expert opinion given in Table 1 has electricity in Ontario as the most differentiated issue, with a mean responsibility of 1.8 for the federal government and 8.6 for the province. Note, finally, that even experts are willing to attribute responsibility for some issues to both governments equally, with total responsibility adding up to more than the logical maximum of 10: an expert average of 11.4 total responsibility for health care, for instance.
Voters' attributions will be judged against the standard provided by Table 1. The first set of propositions to be tested involves the distribution of judgments of responsibility. Even if the 0-to-10 scale is subject to varying interpretations, the relative responsibility attributed across governments and issues should conform at least roughly to the pattern in Table 1. If Canadians are coping with the burdens of voting in a federal context, we should expect the following:
• The average of citizens' attributions of responsibility should vary across issues and governments in a pattern similar to the experts.
• The federal government should bear nearly no responsibility for electricity in Ontario, while the provincial government should bear significant responsibility.
• The share of total responsibility attributed to the province should be significantly higher than to the federal government on all issues except the economy and taxes.
• Responsibility should be closest to half and half on the economy and taxes.
• Total governmental responsibility (federal plus provincial) in a given issue area should add up to the logical limit, 10 or “full responsibility.”
○ If responsibility is truly zero-sum, as it must logically be, then regression of federal responsibility on provincial responsibility should produce negative coefficients.
• Total responsibility (federal plus provincial) should be lower on the economy and the farm crisis because on these two issues much is beyond the control of domestic governments.
The second set of analyses involves the determinants of responsibility judgments at the individual level. I examine the effect of political awareness, education, and certainty about responsibility on the accuracy, clarity and the over-time stability of responsibility judgments. This analysis provides an answer to the question of whether shortcomings in citizens' responsibility attributions are a function of citizens' political-cognitive limitations or a lack of available, reliable information about responsibility.
To do so I require an assumption now uncontroversial in the public opinion literature (for example, Zaller, Reference Zaller1992): respondents who know more about politics are more attentive to political information and have therefore taken in more information relevant to responsibility attributions. If this information is useful, these attentive citizens should be making qualitatively better responsibility judgments. Accordingly, a steep gradient in accuracy, clarity, and stability according to political awareness and education will indicate that the quality of responsibility judgments is, like most other objects of citizens' political cognition, characterized by a “low mean and high variance” (Converse, Reference Converse and Apter1964) and that this variance is driven by consumption of, and skill using, political information. If this is true, we can characterize any weakness in this responsibility link in the chain of electoral accountability as entirely general, in line with the now familiar picture of poorly informed democratic citizens (Delli Carpini and Keeter, Reference Delli Carpini and Keeter1997; Fournier, Reference Fournier, Everitt and O'Neill2002). If not, and politically sophisticated, better educated, well-informed citizens are not much more accurate, clearer, or stable, then our conclusion will be that the challenges for voters are presented by the institutions of federalism, particularly the lack of useful information about responsibility.
The accuracy of responsibility judgments is, for a given issue and government, simply the difference between the expert average given in Table 1 and the respondent's own response.
The stability of responsibility judgments over time is measured by the regression of responsibility in the federal wave (t2) on responsibility in the provincial wave (t1), for the same government on the same issue. Interacting the provincial wave measure with indicators of political information, education, and certainty allows us to gauge whether or not some citizens are making more stable judgments than others.
The clarity of responsibility judgments is measured by the total responsibility attributed to both governments, which, logically, and given the question wording, should equal 10.
With these measures in hand, the second set of propositions to be evaluated follows.
• If coping with the challenge of federalism is a skill like other political skills and dependent on attention to, and skill using, the political information available in the mass media, we should expect voters who are:
○ more attentive to political information (measured by factual knowledge of politics), and
○ better educated, and
○ more certain about responsibility in the provincial wave to:
• be closer to the expert judgments given in Table 1
• have more stable judgments of responsibility over time; that is, interactions of these three variables with t1 responsibility will take on positive coefficients, and
• give total responsibility closer to the logical limit of 10
The final empirical analysis evaluates the success of the blame game in diverting responsibility. Though there is a probably a balanced intergovernmental blame game over the long haul, Canada's non-concurrent federal and provincial elections present a quasi-experiment that can indicate whether or not voters are federal dupes, as Cairns once suggested, tacking back and forth between provincial and federal responsibility in response to politicians' attempts to shift blame (Reference Cairns1977: 708). With non-concurrent elections, the campaigning government can attempt to shift blame to the other level without much immediate, direct response from that other level. During the campaign, reporters are unlikely to make the effort to ask for a response from the other level of government. Moreover, leaders of governments at one level usually want to avoid the appearance of meddling in the other-level election.
But even this is asymmetrical: provincial governments can shift blame relatively unproblematically, while federal ones face important obstacles to shifting blame. Federal governments campaign across the country and so have to blame all provincial governments, diffusing the effectiveness of the claim, or blame one specific province when speaking directly to voters in that province. So we should expect provincial governments to be more successful at shifting blame and stealing credit, if either level is.
The upshot is that if a blame game (and credit-claiming) is being played on some of these issues, and if the game were successful, then:
• voters will attribute more responsibility to the government up for election for positive judgements of the status quo than for negative judgements, and
• this effect will be stronger—or only visible—at the provincial level.
Responsibility Judgments in Federal and Provincial Elections
The first task is to characterize experts' and ordinary Canadians' responsibility judgments. Table 1 shows the mean public responsibility beside the mean expert responsibility. Experts show much more variation across issues and levels of government. The public averages are so minimally differentiated that it is immediately clear that citizens are not responsibility experts. Citizens appear to get close to expert judgments only when the mean expert judgment happens to be around the 6–8 citizen average that obtains across all issues on the numerical scale. That they are sometimes close to expert judgment appears to be just luck. When they are, it is for one government only; in the aggregate they never share the experts' judgment that the two governments bear significantly different levels of responsibility. The economy is a good example: though the citizen average is close to the expert judgment on federal responsibility, citizens give the province far too much responsibility for the provincial economy. Note, finally, that citizens appear to get closer to expert judgments when using the verbal scale rather than the numerical one, but the difference is not significant when proportional differences are compared.
Figures 1a and 1b present smoothed histogramsFootnote 6 showing the distribution of citizens' responsibility attributions to the provincial governments in the context of the provincial elections in Ontario (1a) and Saskatchewan (1b). First, and most importantly, variation across issues is strikingly minimal and respondents are making little use of the lower end of the scale. The means are between six and eight on the 0–10 scale, indicating that voters saw the provincial government, on average, as more than partly responsible. On no issue in Ontario did more than 10 per cent of respondents give the province less than five. In Saskatchewan a few more did so, with the “farm crisis” standing out as the only issue where the provincial government escaped substantial responsibility among one-quarter of its citizens. At the other end of the scale, on each issue we find 20 per cent in Ontario and 10 per cent in Saskatchewan saying that the provincial government is “fully responsible”!
Figure 1a–1c Responsibility by Policy Area, Provincial and Federal Studies; 1a Provincial Responsibility – Ontario, Oct. 2003; 1b Provincial Responsibility – Saskatchewan, Nov. 2003; 1c Federal Responsibility – Ontario & Saskatchewan, June 2004
Does this just reflect low levels of information about provincial politics? Or perhaps the 10-point measurement scale is too demanding, generating a lot of random measurement error? Neither is true. A similar pattern is evident in Figure 1c, showing federal responsibility as measured on the semantic “no responsibility” to “full responsibility” scale during the federal campaign. The modal response is “a lot of responsibility” on all issues except the Saskatchewan economy and the Ontario deficit, where slightly more respondents gave the federal government only “a little” responsibility. Across a range of issues, then, voters hold both governments jointly responsible for current conditions. Few voters are willing to use the ends of the responsibility scale, declaring either government not responsible or fully responsible, that is, unless they are willing to do so for both governments. And the pattern holds if we look at federal responsibility in the context of the provincial campaign, and vice-versa.
These graphs present responsibility in isolation, but there are two governments involved in each of these policy areas. Just as relevant, therefore, is the total and relative allocation of responsibility to the two governments. The evidence shreds any remnants of a view of Canadian federalism of “watertight compartments.”Footnote 7 On all eight issues in the provincial campaigns, more than three-quarters of voters have a total federal-plus-provincial responsibility score more than 10 on the 0 to 20 scale (tabular results not shown). That is, both governments are more than half responsible. In the federal election context the story is the same. Total responsibility with the four-point scale can run from zero (both governments not responsible) to six (both governments fully responsible), with a logical maximum of three. The means for federal responsibility range from 2.9 (SK economy) to 3.8 (health care, both provinces). The majority of respondents, on most issues, see no need to impose some kind of logical limit to responsibility such that the two governments' responsibilities add up to full responsibility.
The absolute measure of responsibility, however, is confounded by interpersonal variation in understanding of the responsibility scale. A citizen's assessment of relative responsibility is a cleaner measure; so I divide the responsibility of the government in question by the total provincial-plus-federal responsibility for each respondent. Figures 2a–2c present the evidence. On all issues, the relative share of responsibility is tightly clustered around a judgment of equally shared responsibility, though remember that for many respondents this involves both governments judged more than partly responsible. All issues at the provincial level except electricity in Ontario have mean provincial responsibility shares between 53 per cent and 57 per cent. Few voters judge relative responsibility outside of the 40/60 to 60/40 range. The same goes for the federal share of responsibility in the federal study (Figure 2c). The economy, health care value for money, and the farm crisis are all clustered around 50 per cent, while only on social services and the Ontario deficit do we see responsibility on average tilting toward the province. Given that social services are clearly a provincial responsibility and the Ontario deficit was widely attributed to aggressive provincial tax cuts, we can only be somewhat encouraged by the public's ability to lean very slightly toward provincial responsibility on these matters.
Figure 2a–2c Share of Responsibility by Policy Area, Provincial and Federal Studies; 2a Provincial Share of Responsibility, Ont. Oct. 2003; 2b Provincial Share of Responsibility, Sask. Nov. 2003; 2c Federal Share of Responsibility, Ont. & Sask. 2004.
On this evidence, we should therefore dismiss the notion that citizens take a zero-sum approach to responsibility attributions. To see this clearly, it is straightforward to summarize the relationship between the two judgments with a regression of provincial on federal responsibility for a given issue as measured within one interview (results not shown). A zero-sum approach to responsibility would imply strongly negative coefficients—the more responsibility for one government, the less for another. But using the provincial election wave, we find no significant relationship between federal and provincial responsibility for taxes, changes to health care, and electricity in Ontario. The relationship is positive and significant for health care and the economy in Ontario and for all three issues in Saskatchewan. In the federal election wave, four issues have positive relationships while only one is zero.Footnote 8 Putting it simply, those who attribute above-average responsibility to one level of government also attribute above-average responsibility to the other level.
Is the apparent vagueness of Canadians' attributions of responsibility reflected in uncertainty, by their own admission? It might be that Canadians admit to guessing on the responsibility questions, in which case we can hardly expect them to be using these judgments consciously in their voting decisions. Despite question wording making it easy to do so, few respondents responded with “don't know” on the original responsibility questions—below 3 per cent on most issues. Most Canadians are willing to at least think about government responsibility and thus, presumably, understand the concept.
On the follow-up question asking about their certainty, in both provinces across the eight issues the distribution is similar: 30 per cent per cent say they are “not very certain,” 50 per cent “somewhat certain,” and 20 per cent “very certain” about responsibility. The only notable variation from this pattern, sensibly, is greater certainty on taxes and electricity in Ontario, where 36 per cent and 40 per cent respectively were “very certain.” Certainty, however, is no guarantee of accuracy. Being certain does increase the use of the ends of the responsibility scale (see the analysis of accuracy and discrimination in the next section). But this is hardly reassuring. Although we find one-third of those who are certain about responsibility on electricity giving the federal government responsibility of two or less, just as many who are certain give responsibility of eight or more! Given this, it is unlikely that any conscious feeling of certainty makes these responsibility judgments more useful in meeting the challenge of federal voting behaviour. Uncertainty would surely be a reasonable position for many citizens and might even improve voting decisions if it dulled the influence of misguided responsibility judgments. But evidence presented below shows no relationship between certainty and accuracy.
What about variation in responsibility across issues? Fully informed judgments about responsibility would surely show significant variation across issues and governments. Looking at the broad pattern of results in Figures 1a–c and 2a–c suggests pessimism about voters' ability to cope with attributing governmental responsibility. There is only minimal variation across issues in the level of responsibility attributed to either government. What little variation appears is as perverse as sensible: for example, the federal government gets more responsibility for health care, an area of formal provincial jurisdiction. Reassuringly, the highest mean provincial responsibility, and highest provincial share, is found on electricity in Ontario, as we would hope. But we do not find the lion's share of responsibility given to the province on value for money in health care, changes to health care, or the Ontario budget deficit, as experts do. Nor is provincial responsibility appreciably lower for the farm crisis or the economy in Saskatchewan. The variation across issues given by experts in Table 1 are only hinted at by the public. Canadians are no responsibility experts.
These are aggregate results, though. It is possible that they conceal a good deal of variation in responsibility across issues, within respondents. How different are the levels of responsibility attributed to a government across issues by the typical respondent? While most respondents show some variation across issues, the mean intra-respondent range is quite low: under 3 for both levels of government in Saskatchewan, 3.6 for the provincial government in Ontario, and just over 4 for the federal government in Ontario. In Saskatchewan about 80 per cent of respondents have a range below 5; in Ontario approximately 60 per cent are in this 0 to 4 range. These results suggest two things: First, because there is some non-random range across issues, responsibility judgments are not meaningless “non-attitudes.” By and large, citizens understand what is being asked. And second, Canadians have at least a vague understanding of the high level of intergovernmentalism in Canada; almost all variation in responsibility across issues is between partial (5) and full (10) responsibility.
Two possible mechanisms might produce the patterns observed so far. The first, and more flattering, is that voters recognize that in a truly intergovernmental policy world, both governments are necessary conditions for the results of government policy. Even experts' judgments bear something of this quality. The second, more pessimistic, is that voters cannot tell which government is responsible, so they guess that both are and, accordingly, give them similar values on all but the most obvious issues.
One issue in particular points us toward the latter account. The issue is: “problems with the supply of electricity in this province”. A well-informed observer would give the federal government almost no responsibility for this situation—experts gave it an average of 1.8 on the 0–10 scale in Table 1. And on this issue, there was no blame game. Provincial attempts to blame the federal government would have lacked credibility. But just 43 per cent of Ontario voters gave the federal government less than a score of five on the responsibility scale, while 15 per cent said the federal government was “fully responsible,” giving a score of 10! And of this latter group, more than two-thirds attributed full responsibility to the province as well. With this evidence from separate questions asking about both levels' responsibility, our conclusion has to be distinctly more pessimistic than the findings from questions in other studies that have asked which government is “more responsible.” If responsibility is a mediator with respect to the voting decision, the high level of responsibility given to the federal government for electricity implies that there were some Ontarians whose estimation of the federal Liberals declined in the wake of the problems with electricity in Ontario.
To be sure, aggregate variation across issues corresponds roughly to expert judgment of the shares of responsibility but the variation is so small that the real story is the muddled nature of responsibility judgments. The bottom line is that the great bulk of voters see both governments as at least partly responsible in most domestic policy areas. Canadians tend to give more responsibility to both governments than the experts. And on all of these issues the public differentiates the two governments' roles far less clearly. The public may have it right in giving roughly 50/50 responsibility on some issues like taxes, but the results on the other issues make it more likely that they are guessing, employing a default assumption that both governments are more than partly responsible. In areas where the provincial government is unequivocally more responsible, like changes to health care, value for money in health care, and problems with electricity in Ontario, voters' attributions are, unlike the experts, hardly different than in the more equally shared policy domains.
Who Makes More Accurate, Stable, Clearer Responsibility Judgments?
Canadian federal voters appear to be confused, vague, and imprecise when attributing responsibility to their governments. But is this due to the challenging context provided by Canadian federalism or can it be laid at the feet of citizens themselves, a simple consequence of widespread ignorance in all matters political? Relatedly, does the information available through the mass media clarify or confuse matters of responsibility? To find out, we assess the determinants of the accuracy, stability, and clarity of these judgments.
Accuracy
Which voters are getting closer to expert responsibility judgments than others? The degree to which better-informed, more attentive citizens are significantly more accurate in their judgments will indicate how much information relevant to these judgments is available in the mass media. To judge accuracy we take the respondent's distance from the mean expert judgment on the 0–10 scale on each issue and average across issues.Footnote 9 Lower values are more accurate judgments. The variable has a range from 0.5 to 6.3 with a mean of 2.3.
Table 2 shows a negative binomial regression of accuracy on education and political information.Footnote 10 It indicates that better educated and more attentive citizens get substantially closer to the expert judgment. The difference between the most and least educated is expected to be .42 (.25) and the best and worst-informed .66 (.20).Footnote 11 Combined, the difference is 1.11 (.29), which is roughly the space between the 25th and 75th percentiles of the distribution of accuracy. This is a substantial effect, indicating that responsibility is a political cognition like others in that education and attention make these judgments more accurate. The implication is that there is some information in the mass media or in formal education that helps the more attentive get closer to the expert attribution of responsibility. A more far-reaching consequence is that to the extent accurate responsibility judgments are necessary for reasoned voting decisions, a federal system may amplify the gap in democratic decision making between more- and less-sophisticated citizens (Bartels, Reference Bartels1996; Delli Carpini and Keeter, Reference Delli Carpini and Keeter1997; Fournier, Reference Fournier, Everitt and O'Neill2002).
Table 2. Accuracy
Negative binomial regression coefficients with standard errors in parentheses.
Coefficients in bold are more than 1.64 times their standard errors.
Stability
Stability is a feature of more sophisticated belief systems (Converse, Reference Converse1962). More specifically, judgments about responsibility should change only marginally over the eight-month period represented by this study. As with accuracy, if attention to politics improves stability, this is evidence that there is at least some information available that is relevant to responsibility judgments. Questions about responsibility for health care, the economy, and the farm crisis were asked in both provincial and federal election waves and they were asked about both levels of government. Table 3 presents only pooled two-province analysis of the economy and health care, showing regressions of the 2004 federal wave response on the 2003 provincial wave response and on interactions of the provincial response with education, political information, and the respondent's professed certainty about her responsibility judgment in the provincial wave. If these characteristics enhance stability, the coefficients on the interaction terms should be positive.Footnote 12
Table 3. Stability: Provincial to Federal Surveys
OLS regression coefficients with standard errors in parentheses.
Coefficients in bold are more than 1.64 times their standard errors.
Only a few results stand out. First, judgments about responsibility are unstable over time. A perfect correspondence would have required a coefficient of .36 on the provincial response variable to get from the 0-to-10 provincial wave scale to the four-point federal wave scale so that respondents one point higher than average in the provincial wave's 0–10 scale would be .36 higher on the federal wave's four point scale. The coefficients in the top row (provincial response) are not even one third this size. Even on the most stable issue in the disaggregated analysis, provincial responsibility for the economy in Saskatchewan (disaggregated results not shown), respondents five points apart on the provincial wave 0–10 scale are predicted to be only .55 apart on the four-point federal wave responsibility scale.
What enhances stability? The short answer is not much. Education, if anything, weakens it. Being certain about the judgment does not promote stability. Table 3 does give some evidence that political information makes respondents a bit more stable in their assignment of responsibility. On health care, the best informed voters are about twice as stable as the least attentive ones, though this does not apply to responsibility for the economy. Given the finding that attentiveness bolsters accuracy, it is surprising that it does not also make the judgments more stable. Although the more attentive tend to make better responsibility judgments, they are no more likely to carry around these judgments fully formed and ready for the survey interview or use in a voting decision. This instability suggests that there may indeed be an opportunity for politicians to play the blame game, as Cairns suggested (Reference Cairns1977: 708).
Clarity
Another measure of the quality of these judgments is the total responsibility allocated to the two governments, given that 10 is logically defined as “full responsibility” in the provincial wave. Table 4 presents total responsibility regressed on education, certainty, and political awareness from the provincial surveys. The better educated attribute lower total responsibility, closer to the logical maximum of 10, except on Health Care and on the Saskatchewan economy, where they are no different. On the others, a university-educated voter gave on average one point lower total responsibility than one who had not completed high school. The real impact comes from attentiveness to politics, where the best-informed voters gave roughly three points lower total responsibility than the least-well-informed voters. Those who said they were most certain, by contrast, tended to give higher responsibility. So again, like most political judgments, those about responsibility are clearer in the minds of the more educated, more attentive, more sophisticated citizens; the less attentive tend to give high levels of responsibility to both governments on all issues, perhaps hedging their bets in the absence of information.
Table 4. Total Responsibility–OLS Regression
OLS regression coefficients with standard errors in parentheses.
Coefficients in bold are more than 1.64 times their standard errors.
Election Campaigns and Responsibility
Finally, I evaluate the success of the blame game. If it works, citizens would attribute more credit to the government up for election for a positive situation than it would blame for a negative one. Table 5 shows no evidence of success by either federal or provincial governments. It presents mean responsibility scores for the government facing election broken down by respondents' judgment of the status quo for health care and the economy (the issue areas constant across the provincial and federal waves). Results on other issues are very similar (not shown). There is no difference in either the government's level of responsibility or its share of responsibility across the range of positive and negative evaluations of the status quo. In three of the four cases, there is simply no significant variation in responsibility share for the different judgments. In the bottom right table (5d), there are differences, but not in a consistent direction. The provincial government does get a greater share of responsibility for a positive economy, but the difference is only 5 per cent and this likely reflects a fundamental attribution error whereby greater responsibility is attributed to positive events that occur to objects closer to the self, in this case the province (Madsen, Reference Madsen1987). As for absolute levels of responsibility, there is no difference for the federal responsibility on health value during the federal election or for provincial responsibility on the economy during the provincial election. In the other two cases (5b and 5c), it looks more like blame avoidance would backfire: voters who saw negative conditions gave more, not less, responsibility to the government up for election than to the other government.Footnote 13
Table 5. Mean Responsibility Shares by Status Quo Judgments
One plausible explanation for this is that the logic of electoral competition prevents these “presentational strategies” (Hood, Reference Hood2002) from succeeding. Despite the silence of the other level of government, any attempts at shifting blame are balanced by opposition parties' attempts to pin as much blame as possible on the government. Opinion about responsibility does not shift because there are balanced messages of equal intensity cancelling each other out (Zaller, Reference Zaller1992). As long as the opposition parties have an incentive to correct the government's claims about responsibility, the result may be, at worst, greater confusion about responsibility.
Avoidance of blame in general probably meets with limited success, first, because voters are so vague already about responsibility in the federal system and, second, because increasing the blame of another level of government may not decrease responsibility attributed to the blame avoider's own. Furthermore, the voters who will hear the message are the ones least likely to change their assessment of responsibility, perhaps because the better educated are more aware of the de jure division of powers in the federation.
Conclusion
All of this evidence, taken together, indicates that responsibility judgments are not meaningless non-attitudes in Canada, nor are they highly rational, well-informed, or centrally processed by most voters. If they were, we would expect them to be more accurate. Some voters, at least, would be differentiating the two governments' roles more clearly. We would also expect more variation across issues and more stability over time. Instead, this paper's central finding is that many Canadian voters appear to defer to the “muddle” of federalism (Richards, Reference Richards and Lazar1998) by assuming that both governments have contributed to current conditions and both should therefore be credited or blamed to roughly the same degree.
Just as evidence has piled up showing the gap between real voters and an idealized version of a democratic theory of electoral accountability, the evidence here points to a gap between real federal voters and assumptions about how electoral accountability should work under multi-level government. Some of voters' vagueness is undoubtedly a reflection of the well-documented rational ignorance of citizens in modern democracies, but some of it comes from the context. No expert would contest the claim that attributing responsibility to federal and provincial governments in Canada is exceedingly difficult. Even the experts canvassed in this study show significant variation in attributions of responsibility to a government for a given issue. So the gradual entanglement of provincial and federal governments over the last half-century (Rocher and Smith, Reference Rocher and Smith2003) may be administratively efficient, but it flies in the face of the assumptions about electoral accountability underlying theories of federalism (Richards, Reference Richards and Lazar1998). If the vast majority of voters believe on the basis of little information that both governments bear a lot of responsibility on all issues, then many of federalism's vaunted advantages—government “closer to the people,” more accurate communication of citizens' preferences to goverment, etc.—are rendered dubious (Cairns, Reference Cairns1977). Worse, if politicians think voters lack motivation or the tools to accurately attribute responsibility, they may be able shirk in their role as citizens' agents. And this may be the case in some policy domains and not others, determined by the particular context of federalism in each domain.
To find out how serious is the threat to accountability outlined early in this paper, another study is required to take these responsibility attributions, such as they are, and interact them with policy judgments in a model of retrospective voting. It seems unlikely that accountability will not be threatened by the challenge of attributing responsibility to multiple governments. However, the consequences are not obvious. It may be that Canadians hedge their bets, blame both governments, and try to get a response from either or both. It is equally possible that Canadians ignore policy areas where they have trouble pointing the finger. These distortions of electoral accountability are likely endemic to federalism. Canadians, and Canadian scholars, should be mindful of the consequences.
Acknowledgments
The author appreciates contributions from Greg Lyle, Patrick Fournier, Stuart Soroka, Cristine de Clercy, Peter Ferguson, Amanda Bittner, Marc-André Bodet, Richard Johnston, and the participants of seminars at McGill University's Social Statistics Program, the Australian National University, the UBC Centre for Health Services and Policy Research, and the European Consortium for Political Research Joint Sessions (2005). My gratitude, also, to 33 generous colleagues who took time to complete the expert survey on responsibility.
