A Theory of Nonseparable Preferences in Survey Responses

Abstract

A person has nonseparable preferences when her preferences for the outcome of one issue or set of issues depend on the outcome of other issues. A model of individual-level responses to issue questions in public opinion surveys implies that when people have nonseparable preferences, their responses will change depending on the order of questions. An individual's responses may also vary over as her perception of the status quo changes. A telephone survey of a random sample of residents of Franklin County, Ohio, reveals that much of the public has nonseparable preferences on a wide range of issues. Results from a survey experiment confirm that aggregate-level question-order effects occur on issues for which people have nonseparable preferences, and order effects do not occur on issues for which most people have separable preferences. At the individual level, people with nonseparable preferences display greater response instability across question orders than people with separable preferences, and a respondent's level of political information has little impact on response instability. ince Converse's The Nature of Belief Systems in Mass Publics (1964), students of public opinion have argued that most people lack well-defined or stable opinions on important political issues. Converse concluded large portions of an electorate simply do not have meaningful beliefs, even on issues that have formed the basis for intense political controversy among elites for substantial periods of time (1964, 245). Several empirical puzzles in opinion research appear to result from the public's lack of meaningful beliefs. First, most people hold opinions on issues independently of their opinions on other issues, without ideological connections across them. Second, individuals' responses to survey questions are often unstable. Converse's original findings of response instability applied to over-time changes in an individual's responses to issue questions, but response instability also results from seemingly trivial changes in the order or wording of questions. Evidence in favor of Converse's thesis comes from surveys in which researchers ask respondents their opinions on one issue after another without explicit links across the issues. For example, the American National Election Studies (ANES) ask respondents their opinions on the level of taxation and on the level of defense spending and on the level of welfare spending; but respondents are rarely, if ever, asked how their opinion on one issue might change given changes in policy on another issue. Although most public policy debates in Congress and in the media involve actors expressing contingent, conditional, or constrained preferences, survey researchers rely on survey instruments that presume the public's preferences on the same issues are isolated, unconditional, and unconstrained. In this article, I argue that many people's preferences on political issues are not isolated, unconditional, and unconstrained; instead, their preferences on multiple issues are often nonseparable. A person has nonseparable

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 109 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.