THE STRUCTURAL AND TECHNOLOGICAL CHANGES IN TAIWAN'S AGRICULTURAL SECTOR-1951 THROUGH 1976' A TIME-VARYING RARAMETER APPROACH

The assumption of non-stochastic parameters has long been recognized as restrictive to the solution of many production problems arid to economic modeling in general. Parameter variation historically has been treated with the use of non-stochastic adjustments through interaction variables and the use of proxy dummy and trend variables (Solow, 1956, 1957; Amana; Ferguson; Uzaga; Fei and Ranis). Though these empirical techniques in many cases give reasonable results, the presuppose that the researcher can specify the nature of the parameter change. In fact, it may not be obvious that random parameters are part of the estimation problem. Furthermore, specification of structural shifts in parameters is usually difficult. Comparison of parameter changes through techniques such as grouping of data and using various F-tests is most often dependent on the criteria for grouping (Maddala, pp. 390-404).