Why 1990 international Geary-Khamis dollars cannot be a foundation for reliable long run comparisons of GDP

Abstract

Using a large, new dataset of agricultural prices and quantities for many countries and regions, we create five new international Geary-Khamis pounds – for 1870, 1845, 1775, 1705, and a superior chained series. We show that estimated levels and changes in output per worker look very different – more extreme – using 1705 international pounds and 1990 international dollars, compared to all other series; growth rates appear substantially higher using 1990 international dollars. In short, out-of-sample baskets and/or prices create extremely unreliable output estimates. We also show that individual country price indices (rather than international indices) can generate substantially different estimated growth rates.