I'm surprised this was published now, given that I saw a talk on this at a math conference in either 2008 or 2009. The memorable anecdote was that they filmed children who worked with cash in the market, brought them to the classroom, and given pen and paper in the classroom they would be unable to duplicate the calculations they had already done in the market. The speaker was promoting VR to simulate the market context in the classroom.

I guess what this paper adds is a higher N and the reverse case, that classroom skills don't transfer to the market.