Overall, it really does look like the badges help, not just with increasing sharing rates but with making sure that shared data is helpful to the research community. Of all the 2,478 articles used in the study, those without badges were very weak about sharing: “Just six of 37 articles from journals without badges and two of 10 articles from [Psychological Science] before badges that reported available data had accessible, correct, usable data,” write the authors. By contrast, of the articles with badges, “actual sharing was very similar to reported sharing.”
This is both amazing and inspiring. Just putting badges on papers if they have open data dramatically increases the papers including open data. It's not perfect, but it is clearly an incentive system that helps a lot.