| This dissertation examines intercountry differences in agricultural efficiency and productivity. In one approach, traditional Cobb-Douglas production functions are econometrically estimated with fixed and random effects models using panel data. These models are estimated with traditional, quality-adjusted inputs with annual data (N = 117, T = 31), including the former centrally planned economies (CPEs); with non-traditional inputs (agricultural R&D and human capital), the sample size is reduced to quinquennial data (N = 89, T = 4) (excludes CPEs). Nearly all inputs were statistically significant, but the random effects estimates are generally more efficient. The aggregate models indicate high correlation of the output levels with the productivity levels. Therefore, the models are also estimated on a per hectare basis. This procedure produces more reliable coefficients and productivity rankings. The United States had the highest productivity level in the aggregate model, but fell to the 21st rank in the per hectare model.;Another approach, the deterministic nonparametric methodology, is also used with the same database (although the data were smoothed to minimize weather shocks). This approach shows that 27 countries were technically efficient in every period with the variable returns to scale assumption, while only 9 countries were efficient in every period with the constant returns to scale assumption. The Malmquist productivity index is also estimated, which decomposes productivity growth into changes in efficiency change and technical change. The United States' productivity grew at 2.01 percent over the 1962-1990 period. The most consistent innovator countries were the United States, Germany, The Netherlands, Israel, and Singapore.;A final chapter compares the two methodologies and results. The two methodologies, each with strengths and weaknesses, provide complementary information to different research questions. The econometric approach assumes efficiency, estimates coefficients of a functional form that may be arbitrarily selected, and estimates period-wide productivity levels. The nonparametric productivity approach measures inefficiency, doesn't impose any functional form, estimates year-to-year productivity growth rates, and identifies the technological innovators. One can estimate the statistical significance of inputs with the econometric approach, but the nonparametric approach assumes good measurement and may confuse measurement error with efficiency. |
