Why Experiments are Misleading
By Bryan Caplan
Tyler Cowen discusses an interview with Harvard’s Mullainathan, in which he talks about a bank’s randomized marketing experiment:
“What we found stunned me,” he says. “We found that any one of these things had an effect equal to one to five percentage points of interest! A woman’s photo instead of a man’s increased demand among men by as much as dropping the interest rate five points! These things are not small. And this is very much an economic problem. We are talking about big loans here; customers would end up with monthly loan payments of around 10 percent of their annual income. You’d think that if you really needed the money enough to pay this interest rate, you’re not going to be affected by a photo. The photo, cell phone lottery, simple or complicated table, and deadline all had effects on loan applications comparable to interest. Interest rate may not even be the third most important factor. As an economist, even when you think psychology is important, you don’t think it’s this important. And changing interest rates is expensive, but these psychological elements cost nothing.”
But if that’s true, then wouldn’t every bank already be maxing out on psychological elements? In actual markets, every glass has already been labelled “half full.” If you want to loan more money, then, you have to do something costly – like cut interest rates.