Machine Learning, Big Data, and Causation: #TWET
Perhaps you’ve grown accustomed to EconTalk host Russ Roberts bemoaning the limits of statistical analysis…Well fasten your seat belts and get ready for a data-driven ride in this week’s episode with Stanford University’s Susan Athey.
Athey’s research centers on industrial organization, micro theory, and applied econometrics. She’s also consulted with tech firms, and in this week’s conversation, she discusses some of the work she’s done for Bing. Athey encourages listeners to take a new view of “data mining,” using examples as diverse as FDA drug trials, minimum wage legislation, and Google algorithms. While statisticians and social scientists working in the tech landscape have made tremendous strides in statistical analysis, as Athey lucidly points out, Roberts wonders why the same sort of reliability and replicability isn’t found in economics and social science academic literature. Is it a question of different incentive structures? Inferior or outdated statistical methods? Why can’t the same data-driven acumen drive public policy as well as it can influence our online habits? Has Athey been successful in taking the “con” out of econometrics, or will we always have trouble dealing with counterfactuals in a statistically rigorous fashion?
Have a listen, and be on the lookout for additional reflections in our forthcoming EconTalk Extra this week. Thanks for listening!