The effect of within-candidate variation in appearance on vote share - pre-analysis plan

We are examining whether within-candidate variation in appearance in Oregon voting pamphlets predicts candidates' actual vote share. We measured appearance by having Mechanical Turk workers rate the competence of pictures of state senate and state representative candidates. We have already conducted the analysis with pictures from 2004-2012, but the estimates were not sufficiently precise. So, we have now collected a large number of missing pictures from 2004 and 2014 and are about to have them rated. Since we have already written the code, we thought we would register the code and analysis before we collect these new ratings.