18
Endnotes
1
The split sample design also varied question order for items on partisan identification and voting behavior. The split
sample rating the five policy areas identified partisanship before the vote. Half of respondents, those rating three
policies, identified the 1992 party vote prior to personal party affiliation.
2
The equation transforming a value from the seven point scale (x7) to a corresponding point on the eleven-point
scale (x11) is: x11=[(x7-1)/6]*10+1 .
3
This construction carries the primitive assumption that preferences on issues are unique and separable, since the
model tests for independent effects of relative distance on each scale. On the other hand, if issue preferences were
completely nonseparable or if private opinions correspond to party affiliation or information source, then we would
fail to find statistical significance for rejecting a hypothesis of dependence. A factor analysis of placements on the
policy scales suggests multidimensionality in party- and self-placements.
4
In the full sample, the Tories receive 45.6% of the vote, Labour gets 34.3%. The remainder of votes are distributed
among the smaller parties: Liberal Democrat 17.1%, Scottish National 2.2%, Green 0.3%, Plaid Cymru 0.2%, and
other 0.3%. The Conservatives win 57.1% of the two-party vote in the 1992 BES.
5
The BES prompts for party affiliation and, if none is supplied, presses for the party closer to the respondent. If
Labour or Conservative is answered at either stage, the variables measure party affiliation.
6
No important characteristics distinguish non-readers and readers of independent papers in the analysis here and
below.
Convention