This paper contributes to the productivity literature by tying together work from the firm-level productivity studies with macro-level efforts to forecast productivity growth. While the paper spends considerable effort to measure productivity at the firm-level, the main question is how to use information available at the firm level to improve macro-level productivity forecasts. To our knowledge, our study is a novel attempt to connect micro and macro level analysis whereby micro-level productivity estimates and decompositions of aggregate productivity provide additional information to be used in making aggregate forecasts.
Our forecasting experiments show that, although there are some remaining questions, micro-aggregated decompositions improve aggregate total factor productivity forecasts. This result is corroborated by our Bayesian analysis where we found that Bayesian (BMA) forecasts of these decompositions were always better than the aggregate alternative.