Bad Economics: You should know better than this, Priceonomics
Today we have Priceonomics using data from one of their customers to drive home the point that if omitted variable bias were to show up to their house uninvited wearing a neon suit and started hitting on their wife, they probably would still not recognize it.[footnote]It has come to my attention that Redbooth wrote the original material. I would still expect Priceonomics’ blog peer review standards to rise above this, so I’m holding them accountable.[/footnote]
Some context: the customer is a kanban-style task tracking application. They use the time to completion and completion rate by country of tasks on the application to “see if there was an difference in task productivity across countries”. They find that Peru finishes tasks the fastest, Finland and the Czech Republic finish the most tasks, while Russians can’t seem to be able to finish tasks. [footnote]Thus of course confirming the null hypothesis of [alcohol consumption on work ethic](https://en.wikipedia.org/wiki/List_of_countries_by_alcohol_consumption_per_capita).[/footnote]
Now one might say that we should let them have fun with their data. Of course country correlates with things like industry size, education, average required skill for a task or even cultural tendency to break tasks into smaller or larger subtasks, but that doesn’t prevent us learning some fun facts about countries’ work styles. One would be wrong in this case. This analysis is pure noise – we do not know anything more after reading this than before because the omitted variable bias is so crippling. We don’t even know what latent variable correlating with country we’re measuring because the research stopped so short. Had the blog post brought a single controlling variable, I would be tempted to agree, but as it is there is nothing at all to be gained from reading this.