In a perfect world, an economist could run a controlled experiment just as a physicist or a biologist does: setting up two samples, randomly manipulating one of them, and measuring the effect. But an economist rarely has the luxury of such pure experimentation.
