The PhD economists on our team looked at it. Economists from other teams looked at it and signed off. There was manual testing — ie, trying some “choice value

The PhD economists on our team looked at it. Economists from other teams looked at it and signed off. There was manual testing — ie, trying some “choice values”. We discovered their error in convexity when our test suite allowed us to randomly sample the models at scale. (Actually, I had questions before that — but unsurprisingly, when it was just me questioning a PhD economist, the lowly SDE was ignored.) Good intentions aren’t enough; you need mechanisms.