Is top 5% of patents really a good measure of scientist productivity as opposed to published articles or research awards?
Is outmigration more relevant than absolute residency numbers? If "top scientists" choose to start their career in a high tax state and don't move, they will not show up in this study.
What percent of this rather arbitrary group are professors, working at large corporations, or some other affiliation?
What is the effect of outliers? No sample size is given and the individual tax rate effect doesn't look very robust in the scatterplots. Does that fact that a large amount of patent law gets set in Eastern Texas influence this more than taxes?
Even for the corporate tax rate effect that does look like a positive correlation, the effect could be due to one or a few large organizations moving their R&D department.
This just seems like some half baked regressions that aren't seriously trying to understand the issue.