Karthik Muralidharan and Paul Niehaus of University of California, San Diego, argue that when we test things at a small scale, they might not be predictive of how they perform at a larger scale.

Find out more at VoxDev.org