SOUTHERN JOURNAL OF AGRICULTURAL ECONOMICS DECEMBER, 1982 RELATIONSHIP BETWEEN FARM SIZE AND TECHNICAL EFFICIENCY IN WEST TENNESSEE AGRICULTURE

There has been persistent interest in the issues fixed frontiers of these studies are not estimated relating to farm size and survival of family farms in any statistical sense, but are simply "com-since the beginning of agricultural economics as a puted " via mathematical programming methods discipline. Since the early 1950s, the number of (Schmidt and Lovell, p. 344). In order to over-farms has been decreasing and the average farm come these shortcomings, in this study, the com-size has been increasing. Some poverty has al- parative average technical efficiency of small and ways existed in the rural areas, but the difference large farms has been estimated relative to the in the level of incomes between small and large stochastic frontier production functions, using farms has been widening (Singh and Bagi). Con- the maximum likelihood method. sequently, some persons are concerned with the implications of concentration of farm production in the hands of fewer farms and the potential for STOCHASTIC PRODUCTION FRONTIER industrialized organization of agriculture (Stan-ton, p. 727). Serious concern has been expressed Recently, Aigner et al., and Meeusen and van about America's very large farms and what their den Broeck have specified and estimated a growth signals for the rest of agriculture stochastic production frontier which can be writ-(Breimyer; Nikolitch, 1964, 1970, 1972). The ten as changes in the structure of agriculture are sus-pected to be accelerated by government policies (1) Yt = F(Xt,B)e t t =,---,N(Balobaum; Bravo-Ureta and Helmers; Carman; Coffman; Gardner and Pope; Jensen; Penn and Boehm; Raup, 1969, 1978; Upchurch). where Yt is the output of t-th farm, Xt is a vector In brief, agricultural economists are interested of inputs, B is a vector of coefficients, and Et is a more than ever in the issues of equity and effi- random disturbance. This error term is further ciency (Humphries, p. 879; Schuh; Schultz, decomposed into two error components as fol-p. 876; Stanton, p. 727; Tweeten, p. 863). There- lows