distinguish between the impact of reform measures and other factors on enterprise productivity. This section provides a brief review of the relevant literature to facilitate discussion of model specification and the estimation of efficiency. Of the many factors that can effect enterprise productivity, only the following are relevant to this study: 1) increasing returns to scale; 2) technical progress; 3) enterprise behaviour; and 4) market structure.
If the production function is such that a proportionate increase in the vector of inputs would result in a more than proportionate increase in output, then the technology is said to have increasing returns to scale. This implies that an improvement in productivity can be realised simply by expanding the scale of the firm, if other factors remain constant. Increasing returns to scale is the technical reason for economies of scale. Productivity gain from larger scales is observable in agriculture (Junanker 1976) and manufacuiring industry (Pitt and Lee 1981; Weiss and Pascoe 1985; Shephard 1972a, 1972b).
In industrial economics, increasing returns to scale is an important source of productivity although, under certain conditions, it may lead to natural monopoly and hence allocative inefficiency (Hollas and Hereen 1982; Gollop and Karlsson 1978). In free market economies, govemment commonly attempts to regulate the electric power industry (Primeau 1977, 1978, 1985, 1986; Stevenson 1982; HoUas and Hereen 1982) and the airline industry (Sickle, Good and Johnson 1986) to exploit increasing returns to scale.
Research on technical progress was pioneered by Solow (1957) and extended by Kendrick and Sato (1963), Diewert (1976), Jorgenson and Nishimizu (1978). The concept of technical progress has been widely applied to the industrial sector in many economies, including Australia (Gregory and James 1973), Tanzania (Shapiro and MuUer 1977) and China (Field 1983; Kuan Chen et al. 1988a; Kalirajan and Cao 1993).
We should note that technical progress differs from change in technical efficiency. Conceptually, the former is defined as the gain in productivity achieved by adopting new technology and the latter as gain in productivity based on a given level of technology. In the production frontier framework, technical progress refers to a shift in the production frontier whereas variations in technical efficiency arise from change in observed output relative to a given frontier.
The assumption of profit maximisation has been controversial since Berle and Means (1932). The most relevant theoretical hypothesis put forward against profit maximisation was Leibenstein's (1966, 1976, 1979) x-efficiency theory. In his theory, technical efficiency was caused by factors that cannot be explained in the neo-classical framework.
37
Non-profit-maximising behaviour also has theoretical implications for technical and allocative efficiency in market economies (Baumol 1959; Marris 1963; Alchian and Demsetz 1972; Jenson and Meckling 1979; Scitovsky 1943). In the empirical literature, economists have suggested that the following factors contribute to non-profit- maximisation behaviour and thus inefficiency: ill-defined ownership- (Berle and Means
1932; Levy 1981), an imperfect market (Alchian and Kessel 1962) and govemment regulation (Joskow 1974; Pustay 1978).
Market structure or competition also has implications for technical and allocative efficiency. The efficiency loss arising from weakening competition may be due to protection for domestic firms-^ (Tyler 1979; Katrak 1980) or a monopolistic market structure associated with a high market concentration^-* (Carlsson 1972; Shephard
1972a. 1972b).
METHODOLOGY: A BRIEF REVIEW OF THE LTTERATURE
Conventionally, economists follow Zellner, Kmenta and Dreze (1966) in using the average production function approach to estimate a firm's productivity. The drawback to this approach is that some observations inevitably lie above the estimated production function. This contradict economic theory, in which production function is defined as the relationship between inputs and maximum output. The attempt to solve this problem led to the development of the production frontier approach, according to which all observations are enveloped under an estimated potential hyperplane.--^ Technical efficiency and allocative efficiency are therefore assessed against this potential hyperplane.
Technical Efficiency
In the application of Farrell's efficiencies, there are two main streams of work. Studies in the first stream, pioneered by .Aigner and Chu (1968) and added to by Afrait (1972), Richmond (1974) and others, tj'pically rely on linear or non-linear programming methods. The second stream, pioneered independentiy by .Aigner, LoveU and Schmidt (1977) and Meeusen and Van den Broeck (1977), uses the stochastic frontier approach. Since the stochastic frontier approach is used in this thesis, only literature in the second stream is reviewed. Strictiy speaJdng, there are no theoretical grounds for choosing this approach over others. All are theoretically valid and empirically each has
--The implications of ov^Tiership issues for firm efficiency is also discussed in Silkman and Young (1982), Shelton (1967), Pitt and Lee (1981), Monsen Oiiu and Cooley (1968). McEachem (1978). Newhouse (1973), Davis (1973). Junanker (1976), Timmer (1971), Bradley and Gelb (1981), Bruggink (1982), Gillis (1982), Page (1980) and Sterner (1990).
--'See also Lecraw (1977,1978 and 1979), White (1979), Bergsman (1974) and McFetridge (1973). -^This is also discussed in Weiss and Pascoe (1985), McFetridge (1973). Katrak (1980) and Primeaux (1977, 1978. 1985. 1986).
38 its own advantages and disadvantages. The major consideration in selecting this approach was its stochastic property. The stochastic frontier approach makes allowances for the effects of factors not under the control of individual firms and as w^ell is consistent with economic theory. Of course, this approach has its own limitations. First, we have to assume a specific production technology, and in doing so it is important to choose the appropriate production function. Second, we have to impose a specific distribution for (technical) inefficiency. The choice of distribution is somewhat arbitrary and again we have to ensure that the most appropriate one is chosen.-®
According to Aigner, Lovell and Schmidt (1977) and Meeusen and Van den Broeck (1977), the realised production function can be written as follows: