Tsang Shu-ki (1990) is quite right in stating that aggregation under conditions typical of centrally planned economies (CPEs) is problematic and that results obtained to date are by no means definitive. While commending his effort to "identify the key problems that remain unsolved" (p. 1), I would like to offer another perspective on two of the problems: smoothing by aggregation using a hyperbolic relationship and uncertainty concerning the specification of demand and supply functions. A third problem-identification in a discrete switching model-has already been clarified by Portes and Quandt (1989) and need not be discussed here.
With regard to smoothing by aggregation, several authors have proposed models making use of various microeconomic assumptions and data. In contrast to some others, the hyperbolic model proposed in Burkett (1988) makes no use of information on the proportion of submarkets in excess demand. Whether that feature of the model is regrettable or not depends on whether such information is available. When econometricians have access to such information, they should, of course, adopt a model that makes efficient use of it. When the information is unavailable -as is often the case for CPEs -the hyperbolic model may be useful. Unavailable information clearly cannot be "hidden or lost in the process of smoothing" (Tsang 1990, p. 20).
We should distinguish between the general hyperbolic model and specific applications. The general model allows shifts of the hyperbola as well as movements along it. Variables that shift the hyperbola "might include indicators of mobility of goods and buyers, storage capacity, and flexibility in production and consumption" (Burkett, 1988, p. 494). In application of the model of East European economies, I have found no observed variables that significantly shift the hyperbola. However, such variables may be uncovered by future research. Much of Tsang's paper is devoted to a discussion of factors that might in principle shift the hyperbola. It would be useful to measure these factors and estimate their quantitative impact.
With regard to specification uncertainty, Tsang concentrates on consumer demand functions. It is undoubtedly true that these functions can be further refined. As Kornai (1971) stresses, the optimizing models