Productivity growth, technical progress, and efficiency

摘要: In their comment, Subhash C. Ray and Evangelia Desli (1997) (hereafter RD) point out that the specification of decomposition Malmquist productivity index used by Fare et al. (1994) FGNZ) is not unique, propose compute an alternative decomposition. We will discuss additional decompositions at end this note, but proceed here comparing RD with FGNZ based on both conceptual computational grounds. provide a discussion overall index, including important issue when equivalent to traditional notion total factor (TFP) -namely under condition reference technology be consistent constant returns scale (CRS). As they out, yield measure TFP even if "true" underlying CRS, for example. Both use CRS productivity. One key issues raised role properties benchmark technologies define its components. particular, two "reference" are employed in FGNZ: what we refer as variable (VRS) technologies.' By construction, these nested: "contains" VRS technology, Figure 1 RD. This nestedness provides logical basis our At very intuitive level, would argue benchmarks can bounds true-but unknown-technology.2 Intuitively see providing type convex "inner approximation," whereas "outer approximation." Thus benchmarks; do require data satisfy either or VRS. Another possible interpretation captures (perhaps hypothetical) "long run" approximates short run. has some useful features; example, it maximal