Browsing by Subject "Productivity analysis"

As the global population grows and the demand for agricultural commodities increases, empirical measurement of agricultural productivity becomes all the more important. Agricultural productivity is the relationship between yield and inputs and can in simple terms be decomposed into technical change, technical efficiency and scale-effect. Agricultural productivity and its development with time are often studied using price and quantity indices as well as non-parametric and parametric modelling. In this thesis, a mean-value based pseudo-panel created from FADN farm data was used to analyse the productivity development of Finnish cereal farms between 2000-2018. Several functional forms, assumptions of returns to scale and combinations of variables were used to create and test several parametric models. Technical change was observed both at aggregate level for all size classes and for size classes separately. Pooled, fixed effects (FE) and random effects (RE) models were tested with the pseudo-panel. Pooled models were weighted with the number of observations for each mean-value. Results showed little to no technical change in the various size classes. Returns to scale were increasing in all models. Land was a highly significant factor of productivity growth and was found to be notably more significant than the other inputs, labour, capital and materials. FE model were mostly preferred over pooled models and R2-values of all models were over 95% implying over-complexity of models, particularly in the case of the FE models. Very high VIF values were obtained for variables, implying strong multicollinearity between variables. While results regarding technical change are, at a very high level, in line with previous studies on Finnish cereal productivity, the combination of high VIF, fluctuating elasticities between models and limited degrees of freedom indicate that the pseudo-panel method might not provide valuable results in productivity analysis when FADN mean-based data is used. More data-points per year or farm-level data would be required to increase statistical validity of results.