Hostname: page-component-cd9895bd7-fscjk Total loading time: 0 Render date: 2024-12-28T21:12:10.919Z Has data issue: false hasContentIssue false
  • English
  • Français

Question Design, Response Set and the Measurement of Left/Right Thinking in Survey Research*

Published online by Cambridge University Press:  10 November 2009

Ronald D. Lambert
Ronald D. Lambert
Affiliation:
University of Waterloo
Get access Rights & Permissions [Opens in a new window]

Abstract

This note considers the possibility that the format of the interview schedules in the 1965 and 1968 national election surveys may have influenced respondents' ratings of the major political parties on the seven-point left-wing/right-wing scale. It argues that the use of 12 additional scales, defined by more widely understood adjectives such as good/bad and powerful/weak, reduced the willingness of respondents to admit their ignorance and increased their use of the midpoint on this scale. The instructions in 1968 also urged respondents to use the scales even if they were unsure of their meanings. Some idea of the magnitude of these effects is provided by the 1979 survey in which only the left/right scale was employed. Eliminating these artifacts in 1979 contributed to a picture of a party system that was more clearly differentiated and polarized along the left/right continuum.

Résumé

Dans cette note l'auteur se demande si le format de protocole d'interview dans les enquêtes sur les élections nationales de 1965 et 1968 n'a pu influencer la position des répondants sur les principaux partis politiques dans l'échelle gauche/droite à sept points. L'auteur soutient que l'utilisation de 12 échelles additionnelles, définies par des adjectifs plus faciles à comprendre, tels que bon/mauvais et fort/faible, réduit le désir des répondants d'admettre leur ignorance et fait augmenter leur utilisation de la catégorie médiane. Les instructions de 1968 poussaient les personnes enquêetées à utiliser les échelles même si elles n'étaient pas sûres de leur sens. L'enquête de 1979 où on a seulement utilisé l'échelle gauche/droite, donne une idée de l'ampleur de ces effets. L'élimination de ces artifices en 1979 a contribué à l'image d'un systéme partisan plus clairement différencié et polarisé tout au long du continuum gauche/droite.

Type
Note
Information
Canadian Journal of Political Science/Revue canadienne de science politique , Volume 16 , Issue 1 , March 1983 , pp. 135 - 144
Copyright
Copyright © Canadian Political Science Association (l'Association canadienne de science politique) and/et la Société québécoise de science politique 1983

Access options

Get access to the full version of this content by using one of the access options below. (Log in options will check for institutional or personal access. Content may require purchase if you do not have access.)

References

1 See, for example, Kay, Barry J., “An Examination of Class and Left-Right Party Images in Canadian Voting,” this JOURNAL 10 (1977), 127-43;Google ScholarLambert, Ronald D and Hunter, Alfred A., “Social Stratification, Voting Behaviour and the Images of Canadian Federal Political Parties,” Canadian Review of Sociology and Anthropology 16 (1979), 287–304;CrossRefGoogle ScholarLaponce, Jean A., “Note on the Use of the Left-Right Dimension,” Comparative Political Studies 2 (1970), 481–502;CrossRefGoogle Scholar“In Search of the Stable Elements of the Left-Right Landscape,” Comparative Politics 4 (1972), 455-75;CrossRefGoogle Scholar and Left and Right: The Topography of Political Perceptions (Toronto: University of Toronto Press, 1981);Google Scholar and Meisel, John, “Party Images in Canada: A Report on Work in Progress,” Working Papers on Canadian Politics (2nd ed.; Montreal: McGill-Queen's University Press, 1975), 63–126.Google Scholar The use of the semantic differential in the British case is described in Butler, David and Stokes, Donald, Political Change in Britain (2nd ed.; London: Macmillan, 1974)CrossRefGoogle Scholar, 323ff., and Lawrence LeDuc, “Semantic Differential Measures of British Party Images,” British Journal of Political Science 6 (1976), 115-28.CrossRefGoogle Scholar Researchers in the United States have tended to focus on the liberal/conservative rather than the left/right distinction. See Converse, Philip E., “Public Opinion and Voting Behavior,” in Greenstein, F. I. and Polsby, N. W. (eds.), Handbook of Political Science, vol. 4Google Scholar, Nongovernmental Politics (Reading, Mass.: Addison-Wesley, 1975), 109-11Google Scholar, and Sullivan, John L., Piereson, James and Marcus, George E., Political Tolerance and American Democracy (Chicago: University of Chicago Press, 1982), 70–76.Google Scholar

2 Lambert and Hunter, “Social Stratification,” 302.

3 Ogmundson, Rick, “A Note on the Ambiguous Meanings of Survey Research Measures which Use the Words ‘Left’ and ‘Right, ’ ” this JOURNAL 12 (1979), 800.Google Scholar Ogmundson was disturbed by the abnormally high refusal rates on the left/right scale in the 1965 and 1968 surveys, compared to the other scales. The possible multi-dimensional nature of the left/right distinction was the subject of an earlier work by Ogmundson, “Mass-Elite Linkages and Class Issues in Canada,” Canadian Review of Sociology and Anthropology 13 (1976), 1–12.CrossRefGoogle Scholar

4 Respondents were also asked about the Social Credit/Créditiste parties in 1965 and 1968. Only Quebec respondents were asked about the Créditistes in 1979. This report is confined to the three national parties described in all three surveys.

5 The left-wing/right-wing scale was preceded by scales defined by the following paired adjectives in 1965: out of date/modern, competent/incompetent, powerful/weak, foolish/wise, for the middle class/for the working class, united/split and good/bad; in 1968: modern/out of date, competent/incompetent, powerful/weak, for the working class/for the middle class, united/split and good/bad. Scales labelled as follows appeared after the left-wing/right-wing scale in 1965: strong-minded/weak-minded, honest/dishonest, dull/exciting, young/old and slow/fast; in 1968: honest/dishonest, exciting/dull, young/old, concerned/unconcerned for people's welfare, good/bad for Canadian unity and reiigious/non-religious.

6 Laponce was also sensitive to the problem of methodological artifact. The concepts to be rated as left or right were aligned in the centre of the questionnaire page and respondents were instructed to draw arrows to the left or to the right of each concept. This meant that respondents had to return to the centre of the page to start their arrows, thus encouraging them to think independently about each concept and scale and discouraging the kind of serial effects suspected in the 1965 and 1968 surveys. Positioning each concept in the centre of the page also discouraged central responses. A nine-point scale running from extreme left to extreme right was superimposed on each arrow by the researcher in order to assign a score (see Laponce, “In Search,” 501).

7 See Wells, William D. and Smith, Georgianna, “Four Semantic Rating Scales Compared,” Journal of Applied Psychology 44 (1960), 396.CrossRefGoogle Scholar Ogmundson was also troubled by the meaning of the midpoint of the scale whose ends were defined as “for the middle class” versus “for the working class.” In one set of analyses, “4” was defined as a “middle class” or status quo response, while in parallel analyses it was declared missing. His results in the case of this scale were unaffected by the procedure followed. (“Party Class Images and the Class Vote in Canada,” American Sociological Review 40 [1975], 509). Lambert and Hunter adopted this convention in the case of the left-wing/right-wing scale by “assigning the score ‘4’ to the status quo” (“Social Stratification,” 295, 298).

8 The 1965 survey is described in John Meisel and Richard Van Loon, “Committee on Election Expenses 1965–66,” in Committee on Election Expenses, Studies in Canadian Party Finance (Ottawa: Queen's Printer, 1966), 32–41, 143–45.Google ScholarFor the 1968 survey, see John Meisel, “Party Images in Canada.” The 1974 and 1979 panel survey is described in Lawrence LeDucGoogle Scholar, Clarke, Harold D., Jenson, Jane and Pammett, Jon H., “A National Sample Design,” this JOURNAL 7 (1974), 701-08:Google Scholar and Clarke, Harold D. et al., Political Choice in Canada (Toronto: McGraw-Hill Ryerson, 1979), 397.Google Scholar

9 It might be argued that only the seven scales in 1965 and the six scales in 1968 that immediately preceded the left/right scale applied to a given party would affect people's ratings of the party in question. The results calculated in this way closely resembled the findings based on the 12 scales, as reported here.

10 I qualify the evidence presented here as circumstantial because a full test of my working hypothesis would require a sample of surveys differing in terms of the presence or absence of additional scales preceding the left/right scale. An example of this kind of quasi-experimental analysis using individual surveys as the unit of analysis is reported in John Goyder and Peter Pineo, “Social Class Self-Identification,” in Curtis, J. and Scott, W. (eds.), Social Stratification: Canada (2nd ed.; Toronto: Prentice-Hall, 1979), 431-47.Google Scholar

11 Lambert and Hunter, “Social Stratification,” 293.

12 Lambert and Hunter subjected the full set of semantic differential scales in each of the 1965 and 1968 surveys to factor analysis. The popularity of the midpoint in these surveys helps to explain the failure of these researchers to find evidence of a class or left/right factor (“Social Stratification,” 299–300). It was not possible to perform a comparable analysis with the 1979 data because only the left/right scale was employed. If the response set interpretation has any merit, however, one might expect the left/right data in 1979 to be more meaningfully patterned than in 1965 and 1968. One place to look for this kind of patterning is in the correlations among the left/right ratings assigned to the different parties. I therefore calculated product-moment correlations among the three parties taking into account the strength of respondents‘ party identification. Using the 1979 data, the correlations between respondents’ preferred parties and each of their nonpreferred parties tended to be negative, while the correlations between the two nonpreferred parties tended to be positive. This pattern was most pronounced among respondents who identified strongly with their preferred parties and diminished in strength or even changed signs the weaker the identification. As expected, the patterns in 1965 and 1968 were much less consistent.