The database employed in this study was compiled by consultants in the countries involved, working in coordination with FAO Representatives. At present the database includes data for major food staples in 16 countries. In more than one country, however, data for cash crops and agricultural exportables which are important in the local context are not considered. For some countries data is very scarce: e.g. for Ethiopia and Argentina, for which only few annual data points are available. Monthly prices series are also available for seven countries (Costa Rica, Egypt, Ethiopia, Ghana, Indonesia, Senegal, Thailand, and Turkey), with a variable number of observations, ranging in most cases from 100 up to more than 300.
All prices have been reported in US dollars. Annual data for five countries (Brazil, Costa Rica, India and Indonesia) and monthly data for all countries were converted from local currencies to US dollars using the current average period exchange rates reported by the IMF (2003) International Financial Statistics database. Annual data were supplied in US dollars by the local consultants for the remaining ten countries (Chile, Egypt, Ghana, Mexico, Pakistan, Senegal, Thailand, Turkey, Uganda, and Uruguay). For few countries data are available at all the three stages of the food chain (producer, wholesale, retail).[2]
For the analysis presented in this paper it was necessary to select: (i) annual price series including at least 30 observations; this excluded the annual prices reported for Ethiopia, and all data for Argentina; (ii) monthly price series showing an acceptable degree of continuity. Most price series in fact showed considerable gaps. Missing observations were replaced through interpolation, and the consistency of the new data points was checked on the basis of the parameter of the underlying AR (1) model.[3]
The very nature of the prices included in the database is variable. Firstly, included border prices from FAOSTAT are indeed import unit values. Secondly, wholesale, retail and producer prices may inevitably refer to a different price in each market, depending on their specific characteristics. The local consultants have chosen those reported as being representative prices at the three levels. Their level of accuracy, therefore, depends on the variability of market features, and on the size of the countries. The same inevitably applies also to product definitions.[4]
[2] Particularly
[3] In practice, data were
first replaced through interpolation, and then replaced with the fitted
data of the underlying AR(1) model until the parameters of the model stabilized.
|