International comparison of economic aggregates requires the conversion of such aggregates, expressed in domestic currencies, into a common numeraire currency, such as the U.S. dollar or the SDR. As market exchange rates are often subject to fluctuations that do not fully reflect fundamentals and can result in inconsistent cross-country comparisons of real economic activity, it is often recommended to use purchasing power parity (PFP) rates instead. In this context, PPP refers to the purchasing power of a country’s currency. This paper reviews the data and methodology underlying the construction of PPP indices (based on extensive price surveys conducted by the International Comparison Programme in selected countries) and examines some of the issues associated with the potential use of PPP-based estimates of GDP for the Fund’s operational purposes. The paper concludes that, in general, because of unresolved data and methodological issues, the use of PFP-adjusted estimates would seem inappropriate for the Fund’s operational purposes at this time.
Although it is generally agreed that PPP rates are appropriate conversion factors from a conceptual viewpoint, their practical implementation has been hampered by the uneven quality of the PPP indices currently available. Perhaps the most critical problem with the ICP data base is its incomplete coverage of countries and even entire regions. Other data-related problems include extrapolations of current price data from constant price growth rates and difficulties in comparing baskets of goods and services across widely diverse countries (a particularly acute problem with nontradables).
Apart from data deficiencies, a number of methodological issues remain unresolved. The method most commonly used by the ICP has been criticized for imparting an upward bias in GDP estimation for those countries whose price structures differ significantly from price structures in the high-income industrial countries. However, alternative methodologies have also been subject to criticism on other grounds. The various methodological choices produce substantially different results, and no consensus has emerged as to the appropriate estimation procedure.