When I read pop-sci interpretations of it (which is basically all I've ever read) my impression was that the top percentile estimate themselves as worse than th

When I read pop-sci interpretations of it (which is basically all I've ever read) my impression was that the top percentile estimate themselves as worse than the lower-percentile does. The graphs cited in the article show that the top still estimate themselves better than the lower percentile, which is new to me.