Patterns in the Measurement Errors of PPPs and Their Growth Rates

The construction of purchasing-power-parity exchange rates (PPPs) is based on a solid theoretical framework guiding real GDP (and its components) comparisons across countries and over time. In practice this involves a harmonization between prices collected (sporadically) across countries at a given time period (as part of the International Comparisons Program - ICP), known as benchmarks, and the annual movement of prices within domestic economies which is measured by the GDP deflators. Harmonization can only be achieved by making assumptions about measurement errors in these two sources. The new generation PWT reconciles these by assuming ICP measurements are exact when observed and deflator movements are exact when ICP measurements are not observed. This study treats both sources as measured with errors at every period and evaluates a number of alternative measurement error (ME) scenarios for ICP PPPs and their growth rates. The econometric framework used is general and encompasses the extrapolations methodologies of the World Bank, old and new generations of PWT as special cases. Results are evaluated using the mean absolute deviation of estimated PPPs from the observed benchmarks and that of the implied growth rates from observed relative deflators movement. The best performing specifications show measurement errors are inversely proportional to each countries’ group income level and show the regional design of the last two global rounds of the ICP influences the patterns of measurement errors in the constructed ICP PPPs. A possible use of this modeling approach might be to aid the final compilation of the ICP PPPs by tempering the effect of the regional design prior to their release by the World Bank