Looking at the distribution of preference of the government services and spending scale seemed to be logical place to start to see what empirical data exists before I construct my own scales. There actually is a little natural experiment in the data. The NES asked respondents to place two candidates, Kennedy and Jackson, who were not actually in the race. No data were collected for a candidate who was in the race, Perot, in 1992, but there are data for Anderson and Perot in 1996.
Examining the placements of the respondents and their perceptions of the Presidential candidates from 1980-1996, I am struck by the following observations (please see the Excel File which did not properly convert to HTML: http://www.duke.edu/~rml5/dis/ANES-Presidents.xls ).
Respondents: As John promised, the self-placement of the respondents varies little over time. Respondents were more liberal in 1980, and more likely to be unsure of how to answer the question. Other than that first year, the distribution of preferences varied only slightly.
Republican Nominees: The placement of the Republican nominees did not vary widely, although there was more variation among Republican nominees than the Democrats. Reagan was seen as very conservative in 1984- the modal placement was "2" on the seven point scale. All other years, the modal placement of the Republican nominee was at "3." Bush in 1988 was placed by nearly as many people at the midpoint of "4," and had the highest mean placement of any Republican nominee.
Question: Was George H.W. Bush seen as a moderate because Dukakis was far to the left and he used Reagan as a foil to demonstrate that he was not so far to the right?
One surprising finding was that Dole was actually perceived as being more moderate than Perot in 1996. This suggested to me that many respondents had little problem placing more than one candidate at the same location on the seven-point scale. According to my analysis, in 1980, 409 respondents placed two (or more) of the five candidates at the same location on the scale.
Question: What implications does this have for the measurement methodology that I employ? Should I modify the NES’ format or are there good reasons to preserve their methodology?
Democrats: There was very little variation in how respondents placed the Democratic nominees for President. Dukakis in 1988 was seen as the most liberal Democrat nominee. Duke’s modal placement was a "6" on the seven point scale, and Dukakis was located at "7" more than any other Democratic nominee. The mean location of all the nominees was about 5. In 1980, respondents were also asked to locate Kennedy and in 1988, respondents placed Jesse Jackson. These candidates were seen as much more liberal. The modal placement by respondents of these Democrats was at the extreme "7."
Question: My theory would suggest that both Dukakis and Carter should appear more moderate with an extreme candidate in the race. Why didn’t this happen?
One explanation is that neither Kennedy nor Jackson was actually in the race at the time of the questions. A second explanation might be history: While Dukakis was painted as a liberal and linked negatively with Jackson, while Carter took a middle ground and differentiated himself from Ted Kennedy?
Partisan Differentiation.
I looked at how these placements corresponded to levels of affect using ANES 1980. My preliminary analysis suggests that partisans tend to place their candidate closer to the middle of the scale, while opponents tend to place the same candidate closer to the furthest extreme. (Tables can be found as a Word document at: http://www.duke.edu/~rml5/dis/2-16-partisan-affect.doc )
If a respondent thinks a candidate is expected to win, he or she tends to have a higher mean level of affect for the candidate (no regression analysis run yet). The respondents also tend to see the winning candidate as being closer to the middle of the government spending/services scale. (See Table 1)
Compared to the opponents’ partisans, partisans of one party tend to view their candidate as more moderate. Republicans thought Carter was more liberal than Democrats did. Democrats thought that Reagan was more conservative than Republicans. Independents tended to locate all three candidates close to the middle of the issue scale. Anderson was seen as center-right by Democrats and center-left by Republicans (talk about being the wrong thing to all people!). (see Table 2)
The same pattern was visible when I broke the data down by whom the respondent voted for. Carter voters thought their man was more conservative than Reagan voters did. Reagan voters thought Reagan was closer to moderation than Carter voter did. (see Table 3)
Those respondents who placed themselves at the conservative end of the scale saw Carter as more liberal. Perceptions of Reagan were more consistent across the placements, except among the same conservative respondents who saw Reagan as closer to their own position. Anderson’s woes continued as those who located themselves at places 1-3 on the scale placed Anderson higher than "4." The reverse was true for the more liberal respondents. (see Table 4)
The mean level of affect for Carter was highest for those who saw him as more conservative than themselves (compared to those who saw him as equally more liberal). The mean level of affect for Reagan was highest for those who saw him as more liberal. Anderson was liked better by those who thought he was more Conservative than they were. My hunch is that these tabular results are largely driven by partisans who locate the candidate closer to the center than they locate themselves. (see Tables 5a-c).