Building a more flexible model for firm productivityMay 31, 2018
Story by Rob Rombouts
It was 11 years in the making, challenged long-held assumptions in the study of economics and has resulted in interest from governments and agencies around the world.
Salvador Navarro and David Rivers, professors in the Department of Economics, have worked to develop new empirical methods for studying firm-level data, which can be used to employ richer models of firm behaviour, while reducing or avoiding restrictive assumptions.
Firm-level productivity is a key input to models in several areas in economics, including International Trade, Macroeconomics, Labor, as well as Industrial Organization. The pair, working with Amit Gandhi, Professor of Economics at the University of Pennsylvania, showed that the standard approach relied on flawed assumptions around productivity.
“A common approach to measuring firm productivity is to look at value added, taking the output of a firm and subtracting the value of intermediate inputs,” said Rivers. “We wanted to point out that, by just doing that, we are missing something important. We can now look at the output of a firm and all the inputs of a firm, and how they interact.”
“There was a widely used method that everyone had agreed upon being used, but red flags were being raised, and we came along and determined what was wrong, and how to fix it,” said Navarro.
When economists estimate productivity, they “impose convenient restrictions on how they think a firm works, but in many cases we do not know if firms really satisfy these criteria,” said Navarro.
“Our approach allows the model to be more flexible,” said Rivers.
The methods “allow researchers to study questions that were previously unexplored or to revisit existing ones armed with better tools,” said Rivers.
“Imagine if it’s the Olympics,” said Navarro. “You can talk about who is the fastest man alive. You can talk about that; or you can talk about precise measurement. We have developed a more accurate ‘clock’ if you will.”
The resulting model and code has allowed for the creation of an estimation routine that has been distributed widely, generating interest from groups that measure productivity across economies.
Navarro has had meetings with representatives from the central banks in Mexico and Colombia to talk about the methods.
Requests have come in from around the world, including from Belgium, Bolivia, Canada, China, Chile, Denmark, Italy, the Netherlands, Romania, Singapore, Sweden, Switzerland, Turkey, the United Kingdom and the United States.
The World Bank, Microsoft Research, the Federal Reserve Board, the United Nations and the OECD have all requested the code.