Comments on "Limits of Econometrics" by David Freedman

   David Freedman’s impressive paper reveals well his deep understanding of not only statistical techniques and their uses but also of scientific methodology and its philosophy. On visits to Berkeley over the years, I have had the opportunity to talk with David and to learn much from our friendly and constructive conversations. He was a great member of the intellectual community and his passing is a huge loss to us all.

In particular, I would have enjoyed discussing his paper, “The Limits of Statistical Modeling” with him, especially his final sentence, “The goal of empirical research is---or should be---to increase our understanding of the phenomena, rather than displaying our mastery of technique.” Also, in his wide ranging discussion, he has identified crucial obstacles to progress in several areas of social science research, including econometrics. The overall message conveyed reflects his fruitful and productive life. It questions commonly accepted methodologies and makes a plea for the development of alternatives more solidly grounded in data, appropriate statistical methods and scientific methodology. In particular, Freedman has stressed the difficulties of verifying many assumptions made in regression analyses and the impact that incorrect assumptions can have on the validity of conclusions, one general important theme of his research. In this regard, I can identify deeply with his point of view since I too have been engaged in developing sound alternatives to conventional methods that are often based on weak foundations and lead to unsatisfactory results.