We propose estimating production functions using firms' subjective expectations of future output and inputs; data which are becoming increasingly available in surveys. This note compares their proposed estimator to traditional dynamic panel data (e.g., Blundell and Bond, 2000) and proxy variable methods (e.g., Olley and Pakes, 1996). While NPR allows for nonlinear productivity processes, we discuss commonalities with the former when those processes are linear. We note that NPR may be more robust to oligopolistic competition than the latter since it does not employ input demand relations to proxy for
productivity.
Stephen Bond, Agnes Norris Keiller, Tim Obermeier, John Van Reenen and Aureo de Paula
4 February 2026 Paper Number POIDWP138
Download PDF - Leveraging subjective expectations for production functions