Alex,
I'm usually pretty good at manipulating data in SPSS, but I'm stumped on this one. It's an issue that Dr. St. Clair and I discussed last year when we conceptualized this thing, and maybe there's no easy way to do it.
So the unit of analysis is the candidate, right? So if a candidate runs for office 10 times, he or she will have ten rows of data -- with a consistent candidate id -- in the dataset. Similarly, there will be 20 lines of data for each seat (one for each contest in each of the 20 election cycles -- 1968 to 2006 -- that are included). My question is: How can I examine trends in competition for seats, rather than individual candidates.
Here's an example. Say I want to look to see how well Republicans did each year over the period and map that out with a table or figure such like:
1968: Republicans won 45% of contests
1970: Republicans won 43% of contests
Etc.
If I try to crosstabulate plurality party with year, I'm getting an N of like 800 each year because the row includes all of the candidates, not just all of the contests. So the value for "plurality party" for each candidate in the race will be the same. Does that make sense? So it would be like this, if Joe Smith, the Republican, wins:
Joe Smith, Republican, plurality party = Republican
Jane Doe, Democrat, plurality party = Republican
So in 1968, for instance, the data show Republican Party Plurality as 379 and Democratic Party Plurality as 440, but there were only 435 or so elections that year.
This will be more important as we try to examine, for instance, how often African Americans win against Whites, how often African American women win against Latinos, etc. This isn't necessarily something that needs to be fixed for the paper that's due next week, but I wanted to see if you had any ideas.
Thanks for all of your help,
Dr. C.