Introduction by Professor Sir Charles Bean, London School of Economics
Using Machine Learning to Quality-Adjust Prices at Scale
Matthew D. Shapiro, University of Michigan
This paper presents findings from the Re-Engineering Statistics using Economic Transactions (RESET) project using machine learning (ML) to estimate hedonic price indices at scale from item-level transactions and product characteristics. The procedure uses state-of-the-art approaches from hedonic econometrics and implements them with a neural network ML approach. Applying the methodology to Nielsen Retail Scanner data leads to a large hedonic adjustment to the Tornqvist index for food product groups: Cumulative food inflation over the period from 2007 through 2015 is reduced by half—from 5.9% to 2.8%—owing to quality adjustment. These results suggest that quality improvement via product turnover is important even in product groups that are not normally considered to feature rapid technological progress. These findings demonstrate the feasibility and importance of implementing hedonic adjustment at scale, especially when using item-level transactions data for measuring inflation and real expenditures.
Matthew D. Shapiro is the Lawrence R. Klein Collegiate Professor of Economics and Research Professor, Survey Research Center, Institute for Social Research at the University of Michigan. He is a Research Associate of the National Bureau of Economic Research and received B.A. and M.A. degrees from Yale in 1979 and a Ph.D. from M.I.T. in 1984.
Matthew’s general area of expertise is macroeconomics. He has carried out research on investment and capital utilization, business-cycle fluctuations, consumption and saving, financial markets, fiscal policy, monetary policy, time-series econometrics, economics of aging, economic measurement, and survey methodology. Among his current research interests are use of big data in economics; modeling saving, labor supply, retirement, health, insurance, and portfolio choices of older Americans; using surveys and administrative data to address questions in macroeconomics and individual decision making; improving the quality of national economic statistics; and using naturally-occurring data such as account records, retail transactions, and social media to measure and understand economic activity.
