Explanation and causality: Some suggestions from econometrics