See here for context and here for more discussion on the topic.
I was recently playing with data in the CCES cumulative file, and specifically looking at the news interest variable with the following question in mind: does interest in news and politics increase the closer you get to elections? In the above plot, I show average news interest levels (on a 0-1 scale where higher values indicate more news interest) in every day in the lead-up to the past six elections (and only for days where the weighted survey respondent total was at least 100). Instead of a positive linear trend that would suggest higher interest as the election drew nearer, I find a more parabolic-looking pattern, which holds pretty consistently across the past few elections. Interest is high to start, then declines, then rises back up as the election nears.
My initial interpretation is based on both survey response dynamics and (possibly) election anticipation effects. Perhaps this trend emerges because the people most interested in politics select into taking the CCES during the early period when it’s first starting to get fielded. Then, the less politically interested individuals start to take it. Finally, the closer you get to an election, interest levels start to ramp up as the salience of the election itself grows and it begins to affect individuals. That story is of course speculative, but the pattern and its consistency across election years remain interesting. Moreover, in analysis not shown here, the trend cuts across partisan stripes as well.
Striking evidence of a growing millennial gender gap in partisanship was shown by Pew Research last year. When compared to trends from other available data sources last year, I found that they did not mirror Pew’s results, suggesting Pew’s trend might be more of an outlier rather than indicative of real change. With the recent CCES cumulative file update, I checked this phenomenon again with more recent data; past takeaways hold.