A Constrained State-Space Approach to the Prediction of Comparable Real Income Across Countries

Studies on growth performance and catch-up and convergence of countries require and make extensive use of internationally and temporally comparable data on real gross domestic product (GDP) expressed in a common currency unit. The International Comparison Program (ICP), a project supported by the World Bank, OECD and a host of other international bodies, provides estimates of purchasing power parities (PPPs) as the most robust and appropriate converter of nominal GDP into a common currency unit reflecting differences in levels of prices of goods and services in different countries. The coverage of the ICP, however, has been limited to a few benchmark years, roughly every five years since 1970, and to only those countries participating in benchmark comparisons. Over the last two decades, the Penn World Tables (PWT) filled this gap by providing extensive tabulations of real GDP data for a large number of countries and for a 50-year period covering both participating and non-participating countries and benchmark and non-benchmark years. The PWT figures are essentially extrapolations based on results from benchmark years and country-specific growth rates with particular emphasis on comparisons from specific benchmark years. The main purpose of the paper is to show how a constrained state-space formulation of the problem can be used to generate PPPs, and real GDP data, that is consistent with data generated by the ICP for the benchmark years. Treating the ICP data as an unbalanced panel, the paper presents a suitable spatially autocorrelated econometric model for PPPs which is reformulated as a constrained state-space form to complete the panel. The empirical illustration focuses on 24 OECD countries from 1971 to 2000 and generates a complete set of predicted PPPs for all the countries and years. Unlike the PWTs figures, it is possible to compute standard errors associated with the predicted figures