“The Rational Public? A Canadian Replication of the Page and Shapiro Argument.” With Éric...

International Journal of Public Opinion Research Vol. No. © World Association for Public Opinion Research; all rights reserveddoi:./ijpor/edh, available online at www.ijpor.oupjournals.org

THE RATIONAL PUBLIC? A CANADIAN TEST OF THE PAGE AND SHAPIRO

ARGUMENT

Éric Bélanger and François Pétry

ABSTRACT

Over a decade ago, Page and Shapiro challenged the classical view of public opinion byarguing that Americans’ collective policy preferences, instead of being volatile and mean-ingless, were ‘rational’ in the sense that they were reasonable and stable over time. Wepropose the first non-American test of Page and Shapiro’s argument. We analyze

years of trends in Canadian public opinion on policy issues (–). Our resultsreveal that Canadian public opinion appears to be as stable as American public opinion.However, since government acts in advance of public opinion more often in Canada thanin the United States, we argue that stability must be viewed as a reflection of collectivecomplacency as much as it is a sign of a rational public opinion. Multivariate regressionanalyses of changes in public opinion further indicate that, other things being the same,the stability of public opinion is positively affected by issue salience and by the numberof times a survey question is repeated. Although the positive impact of these variables onstability is consistent with the collective rationality model, we show that it can also beexplained without assuming that opinion is collectively rational in the sense implied byPage and Shapiro.

Over a decade ago, Page and Shapiro () challenged the classical view ofpublic opinion by arguing that collective opinion, instead of being volatile andmeaningless, was ‘rational’. By rational public opinion, they meant two things:() that the collective policy preferences of the American public are stable overtime, tending towards gradual rather than rapid change, and only followingdevelopments that are relevant to the policy issues; and () that stable publicopinion is meaningful and reasonable in the sense of providing a clearer pictureof the public’s underlying interests than could be inferred from individual opin-ions. The thrust of their rationality argument rests upon the processes of statisticalaggregation, which cancels out the distorting effects of both random errors in the

An earlier version of this paper was presented at the 2002 meetings of the Midwest Political ScienceAssociation in Chicago. The article was first submitted to IJPOR January 23, 2004. The final version wasreceived July 6, 2004.

edh050.fm Page 1 Monday, November 22, 2004 5:55 PM

I N T E R N A T I O N A L J O U R N A L O F P U B L I C O P I N I O N R E S E A R C H

measurement of individuals’ opinions and random responses given by surveyedindividuals, and of collective deliberation, which enables citizens to arrive atreasonable opinions without an extensive informational base. Aggregate policypreferences, then, are not capricious but are trustworthy.

Consensus on this view is, however, still premature. The claim that publicopinion is rational is itself very controversial. But, even leaving aside this contro-versial claim, the lack of comparative evidence of opinion stability also explainsat least in part why some authors are still not convinced that a compelling casehas been made for a ‘rational public’. Page and Shapiro’s test of the stability ofpublic opinion has been replicated recently by Isernia, Juhász, and Rattinger() with data from France, Germany, and Italy, but the replication onlyinvolves foreign policy issues. The opinion stability argument has yet to be fullytested (with data on mass opinion towards a wider range of issues, not just foreignpolicy) outside of the United States.

We propose to test the Page and Shapiro stability argument with Canadiandata. We analyze years of trends in Canadian public opinion on a variety ofpolicy issues. Our time-series analysis relies on aggregate poll results covering policy questions asked repeatedly by Gallup Canada over the period–. The first part of the article is a comparative description of the stabilityand movement of Canadian and American public opinion over time. Cross-national comparative research on this question is essential because the nature ofpublic opinion and its role in policy making might differ in important respectsbetween presidential and parliamentary systems. The second part of the essayuses multivariate regression techniques to test whether the stability of Canadianopinion is positively correlated with issue salience and with the number of timesa survey question is repeated. In conclusion, we discuss the implications of ourresults for Canadian democracy in general and public policy in particular.

IS PUBLIC OPINION ‘RATIONAL’?

The classical view of public opinion holds that people’s attitudes on public policymatters are highly unstable, volatile, and mood-driven. In an influential study,Converse () found major instability over time in individuals’ opinions aboutpolitical issues. Using panel data from the , , and Americannational elections, he observed that for a substantial part of the surveyed popula-tion, responses to public policy questions shifted from one wave to the next. ForConverse, this inconsistency over time in individual preferences indicated thatmost citizens did not have ‘true’ attitudes about politics and thus simplyanswered survey questions at random. Converse’s empirical results significantlycontributed to the conventional view according to which public opinion is fickleand meaningless, and cannot serve as a foundation for sound public policy (e.g.Lippmann, ).

A C A N A D I A N T E S T O F T H E P A G E A N D S H A P I R O A R G U M E N T

More recently, theoretical research and empirical research have sought torehabilitate mass opinion. Scholars have demonstrated that a substantial propor-tion of apparently random fluctuations in public opinion were the product ofmeasurement error (Achen, ; Erikson, ; Judd & Milburn, ;Inglehart, ) and often reflected reasonable responses to differing cues in thesurvey instrument (Zaller & Feldman, ). Other revisionist scholars haveargued that shallow and unstable individual preferences may be consistent withstable and coherent collective opinion (Miller, ; Feld & Grofman, ). Thelatter argument stems from Condorcet’s jury theorem. Based entirely on statist-ical probabilities, the theorem shows that, under certain conditions, groups tend toprovide better-informed decisions than individuals. This is due to random individ-ual errors canceling out in the aggregation process, thus generating a collectivedecision that has a higher likelihood of being ‘right’ than a single individual onewould.

Taking this argument one step further, Page and Shapiro () proposedthat, in spite of the presence of instability and inconsistency in individual atti-tudes, aggregate public opinion is ‘rational’. They argue that, as the jury theoremsuggests, there are emergent properties in the collective phenomenon of publicopinion that make the latter coherent and meaningful, and that result mainlyfrom the statistical process of aggregation. Since random elements of publicopinion, which can be attributed as much to measurement errors as to the lack ofsophistication of some citizens, tend to cancel each other out over a large sample,aggregate public opinion can express at any given moment the true distributionof the collectivity’s preferences. This means that collective opinion can be sensi-ble and predictable even if individual opinions are not. This also means thatwhen aggregate opinion changes, it does so in reasonable response to significantsocial and value change.

Recent public opinion scholarship has been rather critical of Page and Shapiro’sargument. The principal objection is against their second claim, that mass opinionis collectively ‘rational’. Mass preferences as measured by public opinion surveysmay appear stable and coherent, but that does not mean it is rational in the sensethat it reflects faithfully the true beliefs and preferences of individual citizens.Indeed, some scholars contend that the rational public argument is not convinc-ing at the micro level. Important differences in levels of information among thecitizenry and in the distribution of political knowledge in the mass public arebelieved to affect the ‘quality’ and representativeness of aggregate public opinion,and to induce systematic bias in majority opinion on policy issues (see Kuklinski &Quirk, ; Converse, ; Alvarez & Brehm, ; Althaus, ). The quest-ions raised by these criticisms are important and, while we cannot test this issuedirectly because our study does not deal with public opinion at the individuallevel, we must bear in mind that the evidence of stability of collective opinion inno way implies a ‘rational’ public.

Our essay focuses primarily on Page and Shapiro’s first claim, that mass opin-ion is coherent and stable. Examining years of aggregate public opinion dataon the policy preferences of Americans, Page and Shapiro showed that collectiveopinion was generally stable and predictable, and that it moves gradually andmeaningfully as a result of changes in the demographic and value makeup ofsociety, important domestic or foreign events, and transmission of new informa-tion within society. Recent studies have provided some empirical support toPage and Shapiro’s stability argument in America (Stimson, ; Mayer, ;Smith, ; Knopf, ; Erikson, MacKuen, & Stimson, ) and in Europe(Isernia et al., ). There is also partial evidence in Canada of coherence incollective preferences on political institutions (Johnston, ), on peacekeeping(Martin & Fortmann, ) and on some policy issues (Johnston, ; Pétry,). But more complete comparative evidence would certainly go some waytowards strengthening Page and Shapiro’s first claim.

This study thus presents a test of the opinion stability argument using aggre-gate poll data on Canadians’ policy preferences over the last years. We areinterested to know whether Page and Shapiro’s conclusions regarding the stabil-ity of American mass opinion also apply to the Canadian case. ComparingCanada and the United States is important because, aside of obvious institutionaldifferences, they are very much alike with respect to the mix of issues in theirrespective public agendas and to the role of the mass media. Americans’ andCanadians’ public attitudes towards policy and their government also share manysimilarities (Mauser, ; Mauser & Margolis, ; Perlin, ; Simpson,; Nesbitt-Larking, ).1 Because it is so similar to the U.S. in theserespects, Canada constitutes an ideal laboratory for conducting a controlled com-parison of the stability of mass opinion.

Even though Canadians appear to have become more like Americans in theirvalues and public attitudes, their Westminster-style parliamentary institutionalsystem has remained very different from the American presidential system. Theabsence of checks-and-balances and the strong party discipline in the Canadiansystem have contributed to concentrating powers in the hands of the prime min-ister to an extraordinary degree (Savoie, ). As a consequence, the Canadiangovernment is able to control the public and the media agendas more often thanthe American government can (Soroka, ).

It is against this background that our analysis of the stability of public opinionis set. There are powerful economic and social forces that appear to compel pub-lic opinion in the two countries towards similar sets of concerns and attitudes.But at the same time, institutional differences would lead us to expect differ-ences in their opinion environment. The implications of these differences on the

1 A contrary view emphasizing the differences in the policy attitudes of Canadians and Americans is offeredby Lipset () and, more recently, Adams ().

stability/volatility of public opinion are not entirely clear, however. On onehand, close involvement of public opinion in policy making in America mayrender mass opinion more rather than less volatile, especially if the issue underconsideration is divisive. By contrast, the more deferential attitude of the public inCanada may coexist with a more complacent and hence more stable mass opinion.On the other hand, more direct involvement of mass opinion in issue resolutionmight promote more intense public debate, thereby increasing the quality and,ultimately, the reasonableness of mass opinion.

DATA AND METHODOLOGY

Available published national surveys conducted by Gallup Canada between and were inspected for repeated questions on policy preferences. Theaggregate survey data—the marginal frequencies of responses—were drawnfrom the Gallup Report and, for a limited number of cases, from Fletcher andDrummond () and Sanders ().2 Only those polls dealing with publicpreferences on an identifiable question of national policy were used in the ana-lysis. These policy questions also had to have been asked with identical wordingat least twice during the -year period to be included in our dataset. It wasimportant that question wordings be identical because, as Monroe (, p. )points out, ‘different question wordings produce different levels of support for apolicy, but comparisons of the same wording across time . . . provide a consistentmeasure of change.’

For each poll, the reported opinion was recorded as either favoring or oppos-ing the policy change raised in the survey question.3 A total of repeated pollsover survey items were found, resulting in instances of potential opinionchanges—an average of . repeated polls per item. These policy questions thusprovide a substantial database for analyzing opinion change. To our knowledge,they constitute the most extensive collection of trends about Canadians’ policypreferences yet assembled. They address a large variety of policy issues and theycover virtually every existing policy domain. Domestic policy areas include civilrights ( percent of all items), government administration ( percent), welfare( percent), and the economy ( percent). Sixteen items ( percent) concernforeign affairs or national defense.

This dataset is not as large as Page and Shapiro’s, who accumulated repeatedpolls on , policy items. At least two reasons can explain this difference.

2 This is an extension of the public opinion data used in Pétry’s () study of the opinion–policyrelationship.

Answers to questions that had more than two response categories were merged so as to dichotomize thepreferences into favoring either ‘change’ or ‘status quo’ in policy. Polls that gave respondents a choice ofincreasing, maintaining, or decreasing levels of government intervention or spending were excluded becausethey cannot be operationalized within our dichotomous framework.

First, Page and Shapiro gathered all available political questions from fivepolling organizations, often ending up using survey questions asking for rathervague policy preferences and concerning executive, legislative, and judicialpolicies at all levels of government (, pp. –). We wanted to be morerestrictive in our selection process, and to only use survey questions clearlyreferring to identifiable national policies and to governmental actions at thefederal level. Second, Canadian researchers do not have access to the sameamount of opinion surveys as their American colleagues, especially when it comesto surveys dealing with citizens’ policy preferences. Gallup Canada usually asksaround – policy questions a year (sometimes as many as –), and thisrate was even lower before the mid-s.

In general, our methodology follows that of Page and Shapiro (, p. ).For each of the polls included in the dataset, we first recomputed the dichot-omous marginal frequencies of responses so as to exclude ‘don’t know’ and ‘noopinion’ responses. This was done in order to track the balance of opinion amongthose people actually holding opinions, and to provide a definition of opinionchange as large as possible (see Page & Shapiro, , p. , n. , and Iserniaet al., , p. ). We then proceeded to ascertain whether or not any signifi-cant change in opinion occurred, taking sampling error into account. Since thesamples used in most of the surveys considered have a percent margin of errorwith a confidence level of percent, we defined a statistically significantinstance of opinion change as a shift in response frequencies of percentagepoints or more between two consecutive repetitions of a survey question. When-ever significant changes occurred—whether only once, or more than once if anitem were repeated several times—we classified each as a separate instance ofchange, to be used as a unit of analysis for investigating the magnitude and rapidityof opinion changes that did occur.

FINDINGS

STABILITY AND CHANGE IN COLLECTIVE OPINION

The first question of interest is just how much change and how much stability arefound in the Canadian data as compared to the American data. Table presentscomparative evidence of significant changes in collective policy preferences. Of the pairs of Canadian survey questions, about percent reveal a shift of per-centage points or more. This figure is close to the one presented in Page andShapiro (, p. ), except that their figure of percent is based on policy itemsas a whole rather than on successive question pairs. That is, they considered a policyitem as revealing change the moment they found at least one instance of significantopinion change within the item. Using this method, we find instead that percentof Canadian items show at least one significant shift in opinion.

One reason for the difference in results could be that, as Mayer (, p. ,n. ) argued, Page and Shapiro’s dataset is somewhat biased in favor of non-change—something the authors themselves acknowledge.4 Their surveyquestions classified as ‘stable’ were repeated . times, while their itemsshowing change were repeated . times (on average). By comparison, our stableitems were asked . times and our items revealing at least one significant opin-ion change were repeated . times (on average). We were forced to classify as‘revealing change’ a large number of frequently repeated items that were gener-ally stable and that covered many decades, but that often showed only one or twoinstances of significant change over time.5 We believe this is the main factor thataccounts for the differing figures in Table with regards to the frequency ofchange in items. We would also argue that the frequency of change in successivequestion pairs is a more reliable indicator in assessing the degree of stability andchange in collective opinion since this measure is based on all possible instancesof opinion change observed within each policy item. It thus provides a morestringent test of Page and Shapiro’s argument. Our finding that up to percentof question pairs reveal no significant opinion change should be appropriatelycompared with Page and Shapiro’s finding of percent item stability.6

Canadian public opinion, then, appears strikingly stable over time. It is notimpervious to change either, but as Table shows, most of the changes thatdid occur were rather small. More than half of them (, or percent) wereless than percentage points. Most of those involved preference shifts of or percentage points—statistically significant but hardly impressive. Moreover, ifwe compare with the American results reported in Page and Shapiro (see first

4 ‘Many (, or percent) questions with significant changes were asked at only two time points, so theycould not reveal any fluctuations in opinion; some were repeated with long time intervals between surveys, sothat any abrupt changes would not be likely to show up either’ (Page & Shapiro, , p. ).

5 For example, one question asking if immigration in Canada should be increased was repeated times overa period of years and showed only two small significant shifts in opinion ( percentage points between and , and percentage points between and ).

6 Page and Shapiro do not present results based on pairs of repeated questions, an odd decision consideringthat the rest of their discussion of opinion change rests upon figures based on question pairs.

TABLE Frequency of significant changes in collective opinion on repeatedpolicy questions

Source: For the U.S.: Page & Shapiro (, p. ).

No change Change Total

% (N) % (N) % (N)

United States (by items) () () (,)Canada: by pairs () () ()

by items () () ()

row of Table ), we can see that opinion changes in Canada appear smaller thanthose observed in the United States. Only instances of significant change( percent) revealed opinion shifts of percentage points or more. Among themost noteworthy cases, opinion became much more favorable towards civilservice strikes, the mandatory use of seat belts in cars, the privatization ofgovernment-owned PetroCanada, a maximum four-year term for government,and the participation of Canadian armed forces in the Gulf War. Therewere also huge drops in support for wage and price controls to fight inflationduring the s, and for the possibility of amnesty to illegal immigrants (thelatter being the largest opinion shift we observed: percentage points between and ).

To further verify the extent to which Canadian public opinion moved duringthe -year period, we classified the instances of opinion change according towhether they were gradual, abrupt, or fluctuating. Following Page and Shapiro,a change is considered ‘abrupt’ if it occurred at a rate of percentage points ormore per year.7 We also included as abrupt any significant ( percentage pointsor more) change that occurred within seven months. A ‘fluctuation’ was definedby the number of successive reversals in direction of significant change within agiven time interval. We took two or more significant changes in opposite direc-tions within two years, or three or more within four years as constituting a fluc-tuation. Gradual changes are shifts in opinion that are neither abrupt nor part ofa fluctuation.

Overall, many of the changes we found in policy preferences involved gradualtrends. We found cases of abrupt changes ( percent of the instances ofsignificant change) in our dataset, and percent of our items showed atleast one abrupt change in opinion. By comparison, percent of significantchanges found in the United States by Page and Shapiro (, p. ) wereabrupt. Figure presents an interesting case of abrupt changes in Canadian publicopinion. Up to , general attitudes towards free trade with the United States

7 The rate of change was calculated by dividing the magnitude of change by the time interval over which itoccurred.

TABLE Magnitude of significant changes in collective policy preferences (byquestion pairs)

Source: For the U.S.: Page and Shapiro (, p. ).

–% –% %+ Total

% (N) % (N) % (N) % (N)

United States () () () ()Canada () () () ()

were either stable or moved gradually, even during the – period notshown in the figure. Since however, we can observe six abrupt changes onthe question pertaining to the Free Trade Agreement (FTA) and two moreabrupt shifts related to the North American Free Trade Agreement issue(NAFTA). This example suggests that Canadian attitudes are more volatilewhen faced with actual trade agreements than on the more general question oftrade liberalization (see also Clarke, LeDuc, Jenson, & Pammett, , p. ).

Figure also offers examples of fluctuations in public opinion. Two fluctua-tions can be observed on the FTA issue between and , while two morecan be found on the NAFTA question between and . Such fluctua-tions in Canadians’ policy preferences, however, appear to be uncommon. Of the items repeated frequently enough to detect fluctuations, percent showed atleast two successive opinion changes in opposite directions. This result is nearlyidentical to Page and Shapiro’s (, p. ); they detected fluctuations in percent of questions asked at three or more points in time. Apart from the twofree trade agreement items, we found fluctuations in opinions regarding thelegalization of abortion, wage and price controls, the ‘Buying Back Canada’ pro-gram, the Meech Lake constitutional accord, and the recognition of Quebec as a‘distinct society’ in the Canadian constitution.

In addition to assessing whether public opinion at the margins was stable, wealso wanted to verify the extent to which there were contradictory majorities inCanadian policy preferences (see Johnston, , p. ). Contradictory majori-ties are defined as cases where majority opinion can be found to support eitherside of a policy issue (change or status quo in policy) depending upon themoment it is measured. Sampling error was again taken into account in opera-tionalizing contradictory majorities. Each policy item showing at least one signi-ficant shift ( percentage points or more) in majority opinion was considered as

1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996

Encourage free trade with USA

Support FTA

Support NAFTA

FIGURE Attitudes Towards free trade, -

presenting a case of contradictory majorities. In other words, the movement inpolicy support had to go from percent (or less) to percent (or more)—orvice versa—to be considered as representing a statistically significant switch inthe direction of majority opinion.

All policy items were examined in order to detect changes in the directionof majority preferences. Nearly one fifth of all items (, or percent) revealedat least one instance of contradictory majorities in opinion. This number is compa-rable to the proportion of abrupt changes ( percent) or fluctuations ( percent)we previously observed in these same items. Only of those items revealedmore than one change over time in the direction of majority opinion. One ofthese cases, support towards the FTA, appears in Figure . Majority opinion onthis particular item jumped back and forth between and , opposing theagreement at one point, then supporting it at another before opposing it again.Other noteworthy cases of contradictory majorities are observed on questionsrelated to the breaking-off of continuing ties with the British monarchy, the ban onstrikes in the civil service, regional development spending, and financial assistanceto unemployed people in the country’s poorer regions. Contradictory majoritiesappear to be more frequent on questions of welfare and foreign policy; nearly athird of all items in each of these two policy areas present contradictory majorities.

We do not want to downplay the fact that Canadian public opinion displaysabrupt changes, fluctuations, and contradictory majorities. Nonetheless, theoverall story, as in the case of the United States, is one of a generally stableCanadian public opinion. Two examples of stability and gradual change in opin-ion appear in Figure . Favorable attitudes towards an increase in immigrationremained low throughout the – period, with very few significant upsand downs. And public opinion gradually moved towards higher levels ofapproval for mercy killing, with the majority opinion increasing by about per-centage points over a period of nearly years.

1966 1969 1972 1975 1978 1981 1984 1987 1990 1993 1996 1999 2002

Mercy killing permitted for incurable disease

Increase immigration

FIGURE Examples of Stability and gradual change in Canadian

The data provide little evidence, therefore, that Canadian public preferences arevolatile, or tend to move wildly over time.8 It also appears that the Canadian publicis slightly more gradual in its opinion movements than American opinion (Page &Shapiro, , pp. –). Among the instances of significant opinion change, percent were gradual—neither abrupt nor part of a fluctuation ( percent in theUnited States). And percent of our policy items repeated at three or moretime points involved gradual changes or no change at all, while this proportiondrops to percent in the American case. In most cases, changes in Canadians’ pol-icy preferences involved gradual trends or small incremental movements of opinion.

EXPLAINING AGGREGATE OPINION CHANGE IN CANADA

To better account for opinion change in Canada, we propose several explanatoryhypotheses of opinion change. These hypotheses will be tested with a multivariateanalysis that will first explore the impact of four independent variables on opin-ion movement. These variables are the time interval between polls, the timeperiod, policy areas, and issue salience. We also present and discuss resultsrelated to a fifth explanatory variable: the number of times a survey is repeated.Other variables (such as the socio-economic status of individual survey respond-ents) certainly have an important effect on the volatility of public opinion, buttesting their impact would necessitate a different research design than theaggregate-level one used in this paper.9

First, we logically assume that the bigger the time interval between twosuccessive polls, the greater the probability of detecting a significant change inopinion. If public opinion moves in response to new environmental stimuli (orshocks), the larger the time interval between two polls, the larger the number ofpotential shocks, and, therefore, the larger the amplitude of the change. Putdifferently, if collective opinion is completely random, we should observe nosystematic difference between opinion shifts occurring over short periods andthose recorded over long periods of time. We use the number of months separat-ing two successive repeated polls as an indicator of the time interval between polls.

A second explanatory variable is the time period in which a poll is taken. Con-sistent with the collective rationality model, Page and Shapiro argue that citizens

8 A simple multivariate regression analysis of opinion change confirms the strong element of inertia inCanadians’ policy preferences. Since collective opinion appears generally stable, and opinion changes are moregradual than abrupt, we would expect that the level of support for a particular policy in one survey is a verystrong predictor of the level of support for that same policy in a subsequent survey. One way to verify this is toregress the level of opinion at t upon the level at t– using ordinary least squares and controlling for policyareas. For our question pairs, we obtain an unstandardized regression coefficient of . for opinion at t–,and a R of .. Even if we restrict the sample to the instances of significant opinion change, opinion at t–remains a strong predictor of opinion at t (regression coefficient of . and R of .). These figures reveal avery high degree of stability in public opinion.

9 We had access only to the marginal frequencies as published by Gallup, not to their original survey data.

are better educated and more informed about public policy today than they were years ago.10 The collective rationality model also assumes that public opinionmade out of ignorant citizens tends to be more volatile, whereas informed citizensmake up a more stable collective opinion. If ignorant opinion is more volatilethan knowledgeable opinion, and if there are less ignorant citizens (in propor-tion) today than years ago, then we should expect to find more importantmovements of opinion at the beginning of the period under study because citi-zens were less informed about policy issues on average and hence more likely tochange opinion on those issues. The problem with the expectation of increasedopinion stability over time is that the collective rationality assumptions on whichit is based do not appear supported by more recent empirical evidence. First, theclaim that the collective opinion of ignorant citizens should be more volatile thanthe collective opinion of knowledgeable citizens is strongly contested by Althaus(), who demonstrates that ill-informed opinion is often more stable thanwell-informed opinion.11 Second, even if well-informed opinion were more stablethan ill-informed opinion, there is no reason to believe that citizens are betterinformed about policy issues today than they were in the past (see Delli Carpini &Keeter, , for the United States, and Johnston, Blais, Gidengil, & Nevitte,, and Fournier, , for Canada).

If time has a positive impact on opinion stability, this has probably more to dowith technical factors associated with sampling methods than with collectiverationality considerations. Gallup’s sampling methodology has been improvedand refined since the s. Because they rely on larger samples and more relia-ble instruments, the repeated surveys of today lead to smaller inter-sample varia-bility (Whalen, ), thereby providing a more stable measure of public opinionthan was the case years ago. To test for the possible impact of the time periodon opinion stability, we use the year of polling as an indicator, that is the year ofthe second survey for each pair of successive polls.

The frequency of change in public preferences might also be contingent uponthe policy area involved. A well-known argument in the literature (the ‘moodtheory’) suggests that public opinion on foreign policy issues is ill-informed andhighly volatile—more so than on domestic policy issues (see Knopf, , for areview). Recent studies in Canada (Martin & Fortmann, ), the United States(Knopf, ) and Europe (Isernia et al., ) have shown, however, that collec-tive opinion on foreign policy was neither random nor unpredictable, but could becharacterized as rational in Page and Shapiro’s sense of the term. We test whetheropinions on foreign policy issues (including national defence) are more prone to

10 They state on p. that ‘the rational public revealed by recent survey data may be partly a creation of thetwentieth-century system of public and private formal education together with the modern media and masscommunications.’

11 Althaus (, p. ) states that ‘contrary to the predictions of collective rationality models, ill-informedopinion is usually less evenly dispersed than well-informed opinion.’

change than attitudes towards domestic policy issues with a dummy coded forforeign policy issues and for domestic issues. Within domestic policy issues, wealso check whether collective opinion moves differently on questions of civil rights,government administration, welfare, and the economy using similar dummies.

A fourth explanatory variable is issue salience. Policy issues that are particu-larly salient to the public should exhibit less opinion volatility. Salient issuesusually attract more interest and attention, and are more thoroughly debatedamong the public. Hence, people should possess more information about theseissues, resulting in stronger, more intensely held, and therefore more stable,opinions (Erikson & Luttbeg, , pp. –; Althaus, , p. ). Issue sali-ence is measured for each poll by the proportion of respondents answering ‘don’tknow’ or ‘no opinion’ to the question: the higher the percentage of non-response,the less salient the issue.12 It is, therefore, hypothesized that opinion change willcorrelate positively with the measure.

We test the effect of these four variables on opinion movement using multi-variate regression methodology (either ordinary least squares or logistic) withrobust standard errors, using the Huber–White sandwich estimator of varianceto address potential problems of heteroskedasticity. Because of the cross-sectional nature of the dataset, residuals cannot be assumed to be independentwithin survey items. For this reason, a cluster specification identifying the policyquestion to which each observation (i.e. each survey) belongs is introduced inevery regression analysis.

Ordinary least squares (OLS) estimates from a first multivariate regressionmodel appear in Table . The model is tested on all question pairs. Thedependent variable is the actual percentage-points absolute change in opinionobserved for each pair (the variable’s range goes from to ). The model showsthat the most important explanatory variable of opinion change is time. Thestatistically significant positive impact of the interval of time between two repeti-tions of a survey question indicates that public opinion tends to change more overthe long term than the short term. This would not be the case if public opinionwere moving entirely at random. It also appears that the magnitude of change inopinion has slightly, but significantly, diminished over the period, suggesting thatthe variability between successive Gallup surveys has been reduced as the accu-racy and reliability of the survey instruments have improved over time. The datasuggest that opinion changes occur significantly more often on economic issuesthan on welfare. However, attitudes towards foreign policy issues do not appearmore volatile than opinions on domestic policy matters. This result can possiblybe attributable in part to the increasing ‘domestication’ of Canadian foreign

12 This measure of issue salience is used in studies of opinion-policy relationship (e.g. Page & Shapiro, ;Brooks, ; Pétry, ). It is not a perfectly valid measure: a high proportion of ‘don’t know’ responses maybe a sign of low salience, but in some cases it may also be a sign of indecision on salient and contentious issues.

policy, which makes foreign policy issues undistinguishable from domestic ones(Stairs, ; Thérien & Noël, ). Finally, the impact of issue salience onopinion movement is statistically significant and of the expected positive sign. Inother words, the less salient an issue, the more policy preferences are volatile.

Table presents three logistic regression models that use as dependent varia-bles the binary codings identifying the presence of non-gradual opinion changesas discussed in the preceding section. The first model is based on the instances of significant opinion change, and tests whether issue salience andpolicy areas can account for abrupt changes.13 The results show that, comparedto welfare issues, the odds of opinion changes on economic issues being abruptare multiplied by four. Issue salience also appears to have a significant effect onabrupt changes: the less salient an issue, the more abrupt the opinion change.This result suggests that, while opinions on salient policy issues are not impervi-ous to change, they are more stable than policy preferences on less salient issuesbecause when they do change, they appear significantly less prone to moveabruptly. Cases selected for the second logistic regression model are furtherrestricted to instances of significant opinion change for items in which it waspossible to determine whether opinions fluctuated (N = ). Policy domains andthe salience of issues do not, however, seem to significantly affect the presence offluctuations in Canadian public opinion.

Table also presents a third logistic regression model in which the dependentvariable takes the value of when a policy item shows contradictory majorities,

13 For the first and second regression models of Table , time was left out of the equations since units ofanalysis in both cases are opinion changes in the first place, and time was already taken into account whencoding abrupt changes and fluctuations.

TABLE Determinants of general opinion change

Note: Entries are unstandardized OLS regression coefficients with robust standard errors in parentheses. Theitalicized numbers are the standardized coefficients. The excluded policy domain is Welfare. * p < .; ** p < .; *** p < . (two-tailed).

Independent variables B S.E. β

Foreign policy . (.) .Civil rights . (.) .Government administration . (.) .Economy .* (.) .Issue salience (non-response rate) .* (.) .Time interval between polls .*** (.) .Year of poll −.*** (.) −.Constant .*** (.) Adjusted R . N (pairs)

and otherwise. The same independent variables as in Table (although indifferent form in the case of time14 and issue salience15) are included on the right-hand side of the equation. The results show that time does not have a significanteffect on the presence of contradictory majorities in survey items. They alsoindicate that contradictory majorities occur significantly less in civil rights and theeconomy than in welfare. As in the case of abrupt changes in opinion, issue sali-ence has a significantly negative effect on the presence of contradictory majoritiesin Canadian policy preferences. Everything else being equal, a one-unit increasein our salience indicator (which really means a one-unit decrease in salience)results in an percent increase in the odds of finding contradictory majorities.There are a few exceptions to this general rule, however, as salient issues like theadoption of the FTA show (see Figure ). But, in general, the finding is robust(confirmed in three out of four regression equations): majority preferences appearsignificantly less stable and clear-cut on less salient policy issues.

Finally, we look at a fifth variable that might further help explain opinionchange at the aggregate level, that is the number of times a survey is repeated. Itis plausible to expect the magnitude of opinion change to drop as a poll isrepeated over time. With repeated surveys, the public can learn to recognize andunderstand the issue, in part because of education or mobilization by the elites.As a consequence, public opinion may become more crystallized as the issueprogresses over the policy cycle. As V. O. Key (, p. ) once suggested:‘from time to time questions arise that command general attention and set offwide discussion, which eventually crystallizes an opinion not readily explicableas an unthinking reaction.’ Page and Shapiro’s (, pp. –) and Iserniaet al.’s (, p. ) studies both allude to this possibility without providingsubstantial evidence. But the proposition that citizens acquire new knowledge assimilar polls get repeated is questionable theoretically because it contradicts theassumption of uninformed individual opinion. A less demanding hypothesis isthat repeated polls do not affect the level of knowledge of most citizens (whichremains low), but they constitute cues that facilitate the use of cognitive short-cuts that help citizens overcome their lack of political knowledge and act as ifthey were fully informed. For example, the increase in opinion stability as a pollis repeated identically over time could result from on-line processing (Fiske &Taylor, ). Citizens may process the information from repeated polls ‘on-line’, at the time they are exposed to it, update their opinion based on the newinformation, and then quickly forget the information itself while retaining the

14 Since the unit of analysis in the third regression model of Table are items, we drop the ‘year of poll’variable, and we add up the total number of months covered by each item as an indicator of time, with theexpectation that the longer the time span of an item, the higher the probability of detecting change in thedirection of majority opinion on the issue.

15 We use an indicator of item salience in the third regression of Table , which corresponds to the averageproportion of ‘don’t know’ or ‘no answer’ responses given to each repeated polls asked on a given item.

updated summary judgment. Thus, citizens may express informed opiniondespite being unable to recall the actual information used to shape their opinion.Such information shortcut processes might have two observable effects oncollective opinion. First, it might diminish non-response and hence increase thesalience of the policy issue in the public mind. Second, it might gradually dimin-ish the gap between informed and uninformed citizens, resulting in a more stablepublic opinion.

Some scholars question the empirical validity of the claim that the use of cog-nitive shortcuts helps uninformed citizens act as if they were fully informed.Althaus () demonstrates that the use of shortcutting strategies rarely helpsAmerican citizens to express opinions similar to those they would give if theywere better informed. In Canada, Johnston et al. (, p. ) show thatinformation shortcuts during the constitutional referendum ‘did not closethe gap . . . between the poorly and the well informed.’ Based on the contraryevidence presented by Althaus and by Johnston et al., we should expect thatrepeated polls have no impact on opinion stability.

These contradictory expectations are tested in Table . The repeated surveyvariable is operationalized as a counter starting with for the first poll on a givenissue, for the second poll, and so on. If Page and Shapiro are correct, the variableshould correlate positively with issue salience and with the stability of public opin-ion by hypotheses. According to Althaus’ null hypothesis, the variables should not

TABLE Impact of poll repetition on collective opinion

Note: Entries are unstandardized OLS regression coefficients with robust standard errors in parentheses. Theitalicized numbers are the standardized β coefficients. The excluded policy domain is Welfare. * p < .; ** p < .; *** p < . (two-tailed).

Independent variables Dependent variables

Salience (Non-response) Opinion change

Foreign policy . (.) . . (.) .Civil rights −. (.) −. . (.) .Government

administration .*** (.) . . (.) .

Economy .*** (.) . .** (.) .Issue salience .* (.) .Time interval

between polls .*** (.) .

Year of poll −.** (.) −.Number of times

a poll is repeated−.* (.) −. −.** (.) −.

Constant .*** (.) .** (.) Adjusted R . . N (polls) N (pairs)

be correlated at all. The table indicates that the variable for the repeated surveyshas a positive and statistically significant effect on both issue salience and opinionstability. In general, with each new poll, the non-response rate on the same issuesignificantly diminishes by . percentage point. The effect of repeated polls onthe magnitude of opinion change is also substantial: each time the Canadianpublic is questioned again about a given policy, the aggregate movement of opin-ion observed since the last time the question was asked is reduced by about .percentage point. This effect is significant notwithstanding the time intervalbetween the two polls, the time period in which the poll was taken, as well as thepolicy area and the salience of the issue. The variable’s significant impact suggeststhat, as an issue is repeated in the polls, it becomes more salient to the public, andopinion on this issue stabilizes in the process.16

CONCLUSIONS

A systematic examination of repeated Canadian public opinion polls on a widerange of policy issues over a -year period reveals considerable stability inCanadians’ policy preferences. When collective opinion did change, opinionmovements were usually gradual instead of being wild and unpredictable. Asmall proportion of significant opinion changes were abrupt or part of a fluctua-tion. Our study also revealed the presence of a few cases of contradictory majori-ties in Canadian public opinion. However, as with Page and Shapiro’s study ofAmerican policy preferences, the more general pattern appears to be one of grad-ual change or no change at all.

Looking for possible determinants of opinion stability at the aggregate level,we find that stability is positively affected by the time period, by issue salience,and by the number of successive polls on the same issue. The R-squared in theregressions of Tables through may be small, but this is perfectly understand-able. A much larger proportion of the variance in opinion volatility could beexplained if we had data on individual levels of knowledge about policy issues.The positive correlations between issue salience, the number of successive polls,the time period, and the stability of opinion could be due to collective rationalityin the sense implied by Page and Shapiro.17 However, we have shown that these

16 We tested for the possibility of an ‘issue-attention cycle’ whereby policy issues gradually become salient tothe public before fading into the public mind after some time (see Downs, ). We proceeded by squaringthe number of times each survey was repeated, and then by adding this squared variable to the equation inTable , Column , with the expected coefficient signs being negative for ‘repetitions’ and positive for‘repetitions-squared’ (remember that our indicator of issue salience is the percentage of non-response). Theresults were not significant and the signs of both coefficients went contrary to expectations.

17 In their book, Page and Shapiro () suggest that the stability of public opinion is a result of collectiverationality. They write (p. ): ‘The reason for these [stability] findings may have to do with the offsettingeffects of random opinion movements by individuals, and with collective deliberation, so that collective publicopinion can remain stable or respond coherently, even while many individuals’ responses shift randomly.

positive correlations can also be explained without assuming that Canadianpublic opinion is collectively rational.

Another point of departure from Page and Shapiro concerns our diagnosis ofthe consequences of stability for policy and, more generally, for representativedemocracy. Page and Shapiro view the stability of public opinion in a favorablelight because they implicitly assume that opinion leads and precedes policy. Intheir interpretation, evidence of stability means that opinion constitutes a relia-ble guide to policy decisions, whereas the presence of unstable public opinionraises doubt about the capability (and the morality) of mass opinion to influencepolicy. Opinion stability is important because of the fear that the public policiesof a democratically responsive government will be irrational or unreasonable ifthey reflect too closely an unstable and volatile public opinion. A positive corre-lation between stability and issue salience would suggest that such fear isunfounded because instability is mostly circumscribed to low salience issues.Rational politicians have little incentive to follow public opinion on low salienceissues. Instead, they will try to be responsive to public opinion on highly salientissues because these issues have a major impact on individual voting decisions. Ifopinion leads policy, it would indeed be a good thing that opinion stabilityincreases with issue saliency.

A similar argument could be made about our finding that opinion becomesmore stable as a survey is repeated over time. Assuming that opinion precedespolicy (so that government decision is the final stage in the process) the findingthat public opinion stabilizes towards the end of the process, as we get nearer thedecision stage, would be a good sign. By contrast, if opinion volatility increasedas we get nearer decision time, that would be bad news from a democraticresponsiveness point of view.

However, a number of scholars, especially outside of the United States, havepointed out the possibility that policy leads opinion, in which case mass opinionmust be seen as endogenous, not exogenous (Bourdieu, ; Ginsberg, ;Margolis & Mauser, ; Champagne, ; Blondiaux, ; Maravall, ;Jacobs & Shapiro, ). While we cannot directly address this debate, it isimportant to emphasize that the government acts in advance of the electoratemore often in Canada than in the United States. When Canadian opinion iscongruent with policy, it is far more likely to be because opinion defers togovernment authority than because decision makers respond to public opinion(Mauser & Margolis, ; Howlett & Ramesh, ; Pétry & Mendelsohn,). If policy leads opinion (a serious possibility in Canada) then the evidenceof stability of opinion must be looked at in a different, less optimistic light. Thepositive correlations that we observe between opinion stability, issue salience,and the number of repeated polls could be the result of spurious relationsprompted by the effect of elite influence. A well-known example of the eliteinfluence scenario is issue-priming, where repeated polls are used to increase the

relative salience of an issue in the public mind, thereby increasing opinion stability,but only as a side effect. More generally, if policy leads opinion, stability may notbe as reasonable or functional as the collective rationality model assumes. Opinionstability may reflect public complacency whereas fluctuations in public opinionand contradictory majorities may indicate that democratic debates and argumentshave a chance. Far from being dysfunctional, fluctuations in public opinion mightbe an element that contributes to the good operation of liberal democracies.

REFERENCES

Achen, C. H. (). Mass political attitudes and the survey response. American PoliticalScience Review, , –.

Adams, M. (). Fire and ice: The United States, Canada and the myth of convergingvalues. Toronto: Penguin Canada.

Althaus, S. L. (). Collective preferences in democratic politics: Opinion surveys and thewill of the people. Cambridge: Cambridge University Press.

Alvarez, R. M., & Brehm, J. (). Hard choices, easy answers: Values, information, andAmerican public opinion, Princeton: Princeton University Press.

Blondiaux, L. (). La fabrique de l’opinion: une histoire sociale des sondages. Paris: Seuil. Bourdieu, P. (). L’opinion publique n’existe pas. Les Temps Modernes, , –. Brooks, J. E. (). The opinion–policy nexus in Germany. Public Opinion Quarterly,

, –. Champagne, P. (). Faire l’opinion: le nouveau jeu politique. Paris: Éditions de Minuit. Clarke, H. D., LeDuc, L., Jenson, J., & Pammett, J. H. (). Absent mandate: Inter-

preting change in Canadian elections. Toronto: Gage. Converse, P. E. (). The nature of belief systems in mass publics. In D. E. Apter

(Ed.), Ideology and discontent (pp. –). New York: Free Press. Converse, P. E. (). Assessing the capacity of mass electorates. Annual Review of

Political Science, , –. Delli Carpini, M. X., & Keeter, S. (). What Americans know about politics and why it

matters. New Haven: Yale University Press. Downs, A. (). Up and down with ecology: The ‘issue-attention cycle’. The Public

Interest, , –. Erikson, R. S. (). The SRC panel data and mass political attitudes. British Journal of

Political Science, , –. Erikson, R. S., & Luttbeg, N. R. (). American public opinion: Its origins, content, and

impact. New York: Wiley. Erikson, R. S., MacKuen, M. B., & Stimson, J. A. (). The macro polity. Cambridge:

Cambridge University Press. Feld, S. L., & Grofman, B. (). Ideological consistency as a collective phenomenon.

American Political Science Review, , –. Fiske, S. T., & Taylor, S. E. (). Social cognition (nd ed.). New York: McGraw-Hill. Fletcher, F. J., & Drummond, R. J. (). Canadian attitude trends (–) (Work-

ing Paper No. ). Montreal: Institute for Research on Public Policy.

Fournier, P. (). The uninformed Canadian voter. In J. Everitt & B. O’Neill (Eds.),Citizen politics: Research and theory in Canadian political behaviour (pp. –). DonMills: Oxford University Press.

Gallup Report. (–). Toronto: Gallup Canada, Inc. Ginsberg, B. (). The captive public: How mass opinion promotes state power. New

York: Basic Books. Howlett, M., & Ramesh, M. (). Studying public policy (nd ed.). Don Mills: Oxford

University Press. Inglehart, R. (). Aggregate stability and individual-level flux in mass belief systems:

The level of analysis paradox. American Political Science Review, , –. Isernia, P., Juhász, Z., & Rattinger, H. (). Foreign policy and the rational public in

comparative perspective. Journal of Conflict Resolution, , –. Jacobs, L. R., & Shapiro, R. Y. (). Politicians don’t pander: Political manipulation and

the loss of democratic responsiveness. Chicago: University of Chicago Press. Johnston, R. (). Public opinion and public policy in Canada: Questions of confidence.

Toronto: University of Toronto Press. Johnston, R., Blais, A., Gidengil, E., & Nevitte, N. (). The challenge of direct

democracy: The Canadian referendum. Montreal: McGill-Queen’s UnversityPress.

Judd, C. M., & Milburn, M. A. (). The structure of attitude systems in the generalpublic: Comparisons of a structural equation model. American Sociological Review, ,–.

Key, V. O., Jr. (). Public opinion and American democracy. New York: Knopf. Knopf, J. W. (). How rational is ‘The Rational Public’? Evidence from U.S. public

opinion on military spending. Journal of Conflict Resolution, , –. Kuklinski, J. H., & Quirk, P. J. (). Reconsidering the rational public: Cognition,

heuristics, and mass opinion. In A. Lupia, M. D. McCubbins, & S. L. Popkin (Eds.),Elements of reason: Cognition, choice, and the bounds of rationality (pp. –).Cambridge: Cambridge University Press.

Lippmann, W. (). The phantom public. New York: Harcourt Brace. Lipset, S. M. (). Continental divide: The values and institutions of the United States

and Canada. London: Routledge. Maravall, J. M. (). Accountability and manipulation. In A. Przeworski, S. C. Stokes, &

B. Manin (Eds.), Democracy, accountability, and representation (pp. –).Cambridge: Cambridge University Press.

Margolis, M., & Mauser, G. A. (Eds.). (). Manipulating public opinion: Essays onpublic opinion as a dependent variable. Pacific Grove, CA: Brooks/Cole.

Martin, P., & Fortmann, M. (). Canadian public opinion and peacekeeping in aturbulent world. International Journal, , –.

Mauser, G. A. (). A comparison of Canadian and American attitudes towardsfirearms. Canadian Journal of Criminology, , –.

Mauser, G. A., & Margolis, M. (). The politics of gun control: Comparing Canadianand American patterns. Government and Policy, , –.

Mayer, W. G. (). The changing American mind: How and why American public opinionchanged between and . Ann Arbor: University of Michigan Press.

Miller, N. R. (). Information, electorates, and democracy: Some extensions andinterpretations of the Condorcet jury theorem. In B. Grofman & G. Owen (Eds.),Information pooling and group decision making (pp. –). Greenwich, CT: JAI.

Monroe, A. D. (). Public opinion and public policy, –. Public OpinionQuarterly, , –.

Nesbitt-Larking, P. (). Canadian political culture: The problem of Americanization.In M. Charlton & P. Barker (Eds.), Crosscurrents: Contremporary political issues (thed., pp. –). Toronto: ITP Nelson,

Page, B. I., & Shapiro, R. Y. (). Effects of public opinion on policy. American PoliticalScience Review, , –.

Page, B. I., & Shapiro, R. Y. (). The rational public: Fifty years of trends in Americans’policy preferences. Chicago: University of Chicago Press.

Perlin, G. (). The constraints of public opinion: Diverging or converging paths? InK. Banting, G. Hoberg, & R. Simeon (Eds.), Degrees of freedom: Canada and the UnitedStates in a changing world (pp. –). Montreal: McGill-Queen’s University Press.

Pétry, F. (). The opinion–policy relationship in Canada. Journal of Politics, ,–.

Pétry, F., & Mendelsohn, M. (). Public opinion and policy making in Canada–. Canadian Journal of Political Science, ().

Sanders, W. (). Jack et Jacques: l’opinion publique au Canada pendant la DeuxièmeGuerre mondiale. Montréal: Comeau & Nadeau.

Savoie, D. J. (). Governing from the centre: The concentration of power in Canadianpolitics. Toronto: University of Toronto Press.

Simpson, J. (). Star-spangled Canadians. Toronto: HarperCollins. Smith, T. W. (). Is there real opinion change? International Journal of Public Opinion

Research, , –. Soroka, S. N. (). Agenda-setting dynamics in Canada. Vancouver: University of

British Columbia Press. Stairs, D. (). Public opinion and external affairs: Reflections on the domestication of

Canadian foreign policy. International Journal, , –. Stimson, J. A. (). Public opinion in America: Moods, cycles, and swings. Boulder, CO:

Westview Press. Thérien, J.-P., & Noël, A. (). Welfare institutions and foreign aid: Domestic foun-

dations of Canadian foreign policy. Canadian Journal of Political Science, , –. Whalen, H. (). The perils of polling. In P. W. Fox (Ed.), Politics: Canada (th ed.,

pp. –). Toronto: McGraw-Hill Ryerson. Zaller, J., & Feldman, S. (). A simple theory of the survey response: Answering

questions versus revealing preferences. American Journal of Political Science, ,–.

BIOGRAPHICAL NOTES

Éric Bélanger is Assistant Professor in the Department of Political Science at McGillUniversity. His research interests include political parties, public opinion, and votingbehavior. He has written or co-written articles published in Comparative Political Studies,

Electoral Studies, French Politics, the European Journal of Political Research, and theCanadian Journal of Political Science.

François Pétry is Professor in the Department of Political Science at Université Laval.His main areas of interest are public policy, comparative politics, and research method-ology. His most recent publication is Politics, Institutions, and Fiscal Policy: Deficits andSurpluses in Federated States (co-edited with Louis Imbeau, ).

Address correspondence to: Éric Bélanger, Department of Political Science, McGillUniversity, Sherbrooke Street West, Montreal, Canada, HA T, email:eric.belanger@mcgill.ca

“The Rational Public? A Canadian Replication of the Page and Shapiro Argument.” With Éric...

Documents

Transcript of “The Rational Public? A Canadian Replication of the Page and Shapiro Argument.” With Éric...

Making Argument Systems for Outsourced Computation ...

Theta Roles and Argument Structure in Manipuri

Art, Argument, and Advocacy

Advancing an Idea for Material Argument

KONSISTENSI STRUKTUR ARGUMENT BY COMMITMENT

Argument from Appearance: A New Argumentation Scheme

Directives: argument and engagement in academic writing

CRITICAL THINKING AND ARGUMENT ANALYSIS - INDONESIA NATIONAL EXAMINATION

How George Bush silenced Anita Hill: A Derridian view of the third persona in public argument

Persuasion and Argument in the Perkins Library

Neural Correlates of Verb Argument Structure Processing

Teaching Argument Writing, Grades 6-12

Amato wins over Shapiro 3-2 in township Confused start mars ...

GMAT Argument Writing Samples

Regarding the "Hole Argument"

On an Argument for Humility

AIF+: Dialogue in the Argument Interchange Format

The core argument for veganism

Friedman. 2017. Reference, Argument, and Evidence.

Argument mapping with MindMup - Olivier Lemeire