Does ideology influence economists' work? (continued)

12/20/2014 01:49:00 PM
Tweetable
The author of the study I commented on previously has left a comment on my post: the graph I criticized was posted in error, and has been updated. So let me direct readers to the corrected graph:
That makes a bit more sense than the previous graph.
To be clear, this graph is not the only analysis from the paper--they looked at other measures as well and found similar correlations. It's a little easier to see the upward slope in this graph, and maybe that slope is large--since these are constructed measures, I just have no point of reference for whether this is huge or tiny. But, I think Noah Smith's critique still applies here: by any of the measures they looked at, it does not appear that ideology explains that much of the variation in results.

Something that stands out a little more in the revised graph (and which Noah Smith had commented previously) is that the influence of ideology appears to be a bit larger at the extremes. In the middle of the x-axis, there might be a slight correlation but it is miniscule compared to the variation in estimates. But the far right and far left have relatively small ranges, and the far right seems to get slightly higher results on average than the far left. It appears to me--without having any raw data to verify--that these flanks are what's driving most of the slope of the regression line. Weirdly though, both of these flanks are very much lower-end estimates compared to the range of results in the ideological center. Not sure what to make of that.