That is the title from new research in Journal of Regional Science by Nadia Greenhalgh-Stanley (Kent State), Shawn Rohlin (Kent State), and Jeff Thompson (Federal Reserve Bank of Boston). Here is the abstract:

We use panel fixed effects estimation with a border approach creating cross‐border county pairs to identify changes in food sales tax rates on employment, payroll, and hiring. Results suggest food sales taxes have a negligible effect on overall employment but adverse effects in the food and beverage stores industry. We find younger workers, who are more likely to work in the food and beverage industry, are more adversely affected when a neighboring state has preferential tax treatment for food. We also determine that omitting food sales tax rates when studying general sales tax effects on employment does not bias estimates.

That last finding is probably not a throw away result, and might turn out to be the most common reason for being cited over the next few years. A constant source of concern in research general sales taxes is both the definition of the tax base (especially the treatment of food) and the other selective sales taxes. This is some evidence that we can relax a bit on those. (Rohlin and Thompson have another article similarly finding that local sales taxes don’t seem to introduce bias either, also good for state sales tax research).

Another reason to check out the paper is if you are interested in attempting a border-discontinuity regression (discussed in section 4).* The opportunity of state border discontinuities is that you presumably have similar local conditions but different state policies. The new potential bias, as the authors point out, is that cross-border flows (like cross-border shopping) can bias the results. In this case, it seems to me the direction of the bias would be against the authors finding their results, so the estimates are probably conservative.

*For the very very interested, I also have a border discontinuity paper with that lays out the math of what type of bias is eliminated and does some other cool things with regard to border matching on this.