Cataloging Growth: A Re-Evaluation of 1900–1990 [pdf] (veronicabp.github.io)

A new paper by Veronica Bäcker-Peral and Benjamin Wittenbrink proposes a significant re-evaluation of U.S. real GDP growth from 1900 to 1990 by constructing a novel quality-adjusted price index for consumer goods using 5.1 million product listings from Sears catalogs. Utilizing large language models to extract product information and assess hedonic price schedules, the researchers found that real goods consumption during the century grew by a factor of 39, starkly contrasting with the conventional estimate of 10.3. This adjustment suggests that inflation was substantially lower than previously thought, and highlights that the period before World War II experienced more robust economic growth than acknowledged by traditional metrics. This work is particularly significant for the AI/ML community as it showcases the innovative application of machine learning techniques in economic research, particularly through the use of high-dimensional text embeddings for data extraction. The findings indicate that quality adjustments to traditional indices can yield insights into consumer goods inflation and economic growth that significantly alter historical perspectives, making it crucial for economic historians and policymakers to consider quality improvements in their evaluations of economic performance. The study also demonstrates the power of empirical methods in adapting historical data for modern analytical paradigms, potentially paving the way for future research that combines AI methodologies with economic analysis.