There are several issues starting with the regional comparisons. The world is divided into geographic regions on the assumption there is more homogeneity in the economies being compared including purchasing patterns and economic structure. The regional structure is not completely followed because some countries such as Japan, Korea, and Mexico are part of the OECD comparison instead of in their respective geographic regions. Because of political considerations, some countries want to be considered for more than one comparison. (e.g., Russia in Eurostat-OECD and the CIS; Sudan and Egypt in the Western. Asia and Africa Comparisons).
In 2005, the regions (and the Eurostat-OECD comparison) were linked using prices provided by 18 countries from the ring list. The linking was done in two stages. First, the ring prices for each of the ring countries were deflated to a regional price using PPPs coming from the respective regional comparisons. This provided a matrix of regional prices which via the CPD produced a set of between region PPPs which when multiplied times the each country’s regional PPP calibrated it to the global level. The aggregation was also done in two stages with linking factors computed at each level of aggregation. The aggregation could have been done using a matrix of basic heading global PPPs as a column and the 146 countries across the rows and aggregating to the GDP in one step.
Different methods produce different results. The properties of the single vs. two stage linking need to be examined. The use of the core list which is to be priced by all countries provides different opportunities for aggregation. The affect of the changing mixture of countries by region raises the question whether countries should be linked by measures of similarity rather than by geographic region.
The use of a core list may result in the prices for some countries appear as “outliers” in the Dikhanov and Quaranta table diagnostics. Criteria need to be established for when a country’s prices for the core are or are not to be used.
A group comprised of Y. Dikhanov, E. Diewert, A. Deaton, S. Sergeev, and P. Rao collaborated on a paper examining these issues which provided recommendations for TAG consideration.