Nonattitude

How the Problem was Uncovered

The nonattitude problem became prominent when the National Election Survey (NES) reinterviewed panels of Americans at 2-year intervals in the 1950s. Political scientist Philip Converse observed the low stability of individuals' answers given 2 years apart on issues that had been widely discussed by political leaders and the media. Question reliabilities ranged from .23 to .46. He also noted a lack of constraint or structure in responses to different policies: Most people did not consistently choose liberal or conservative policies within a single survey. The mean correlation between domestic policy responses was .23. A survey of congressional candidates of the two parties, interviewed with the same questions, found a mean inter-item correlation of .53. Later NES panel studies of political elites showed that their responses had much more reliability over time as well as much greater interitem correlation. These findings were confirmed by panel studies in the 1970s and surveys since in many countries.

Converse concluded that a great many people had no opinions on major issues of the day and were concealing this by randomly choosing responses rather than answer “Don't know,” “Undecided,” or “No opinion” even when these alternatives were offered in a nonjudgmental manner. The observed (low) correlations over time and between issues could be produced by one stratum holding real opinions, which were highly stable and consistent, and another stratum of covert nonopinion-holders expressing pseudo-opinions. Assuming no real attitude change over the 2-year periods, he estimated the percentage of covert non-opinion-holders on each question from the number of changed answers, added in the overt nonopinions, and argued that from 20% to 80% of the public had nonattitudes on a wide range of policy questions. This cast doubt on the meaning of most reported opinion survey results, and on the ability of much of the public to form meaningful opinions on the political issues of the day and influence elite decision making. It also led to a major methodological, theoretical, and ideological controversy.

Alternative Models with Latent Attitudes

Critics analyzing the same data rejected the idea that a large part of the public had nonattitudes on leading public issues. Alternative theories to explain the observed instability and incoherence of responses include the following:

• 1.
  Measurement error produced by vague and ambiguous questions, concealing real attitudes, which could be revealed by better questions
• 2.
  The influence of temporary stimuli—events in the news or in personal life—leading to wide variations in momentary feelings around underlying attitudes
• 3.
  The possibility that each object has a variety of elements or considerations about which the individual has positive or negative feelings, but “samples” unsystematically in answering the questions—perhaps randomly, perhaps in response to recent events or cues given by question wording or sequence
• 4.
  Those who more systematically inventory the considerations they hold in mind may have a near balance of positive and negative feelings—an ambivalence making their answers unstable from time to time or under different question wordings, although they have strong feelings about the issue.

Critics of the nonattitude hypothesis have used structural equation models to show that the pattern of observed (low) correlations could be the result of most people having stable underlying attitudes, albeit very weakly connected to their responses to the particular questions. According to some estimates, these latent attitudes were quite stable, with correlations over 2-year periods ranging from .8 to .9. Instead of a public made up of people with attitudes and people without, public responses in a particular issue area might come from a latent continuum of attitude holding, ranging from those with highly reliable and interrelated opinions (such as those found in elites), through those with general pre-dispositions producing only modest degrees of reliability and structure, to a residue with total absence of attitudes, admitted or concealed. Another model uses the idea of issue publics—that there are small groups of people with stable, organized ideas in particular issue areas but with only loose underlying attitudes, or none at all, toward policies in other areas. The rest of the public may have poorly structured attitudes in all of the areas, or nonattitudes in some or all. Because political elites have to inform themselves, discuss, and take stands on a wide range of issues, they develop consistent attitudes, based on a general ideology or party loyalty linking many issues.

...