Regime shifts and technology diffusion in crop yield growth paths with an application to maize yields in Zimbabwe

Abstract

An alternative speci¢cation for the trend component of crop yield growth is developed and applied to maize yield data for Zimbabwe's large-scale farming sector. This accounts for permanent regime shifts as new technologies are discovered but allows gradual absorption as adoption follows a di¡usion path. Econometric methods are used to estimate the timing and importance of innovations, as well as the length of the di¡usion path. Results from an application to Zimbabwe commercial maize yields indicate two major regime shifts that can be associated with the introduction of hybrid seed varieties, and a di¡usion path that extends over a decade.Crop yield time series are often modelled as the sum of two componentsâ trend process representing improvements in seed varieties, increased fertiliser use, better management practices etc. and a stationary process representing transitory deviations around trend caused primarily by weather. Both components are of interest to economists and policy-makers. The trend component can be interpreted broadly as improvements in technology, so that modelling the trend characterises the rate of technological advance. The stationary component accommodates much of the short-term risk inherent in crop yield distributions, and so can be a useful input into farm decision analysis, crop insurance rating and the pricing of yield futures and options.Early models of crop yield growth used deterministic trends in the form of linear or higher-order polynomial functions of time (e.g. Thompson 1969;Menz and Pardey 1983;Gallagher 1986). It soon became apparent, however, that polynomial trends are not adequate for capturing trend yield growth. Polynomial trends assume the trend component of yield grows at a very

Long-run trend in agricultural yield and climatic factors in Europe

“…Surprisingly, applications of more advanced analysis of the trend of crop yield based on structural time series modelling and Kalman filtering are surprisingly sparse and have become even less so in recent years. Kaylen and Koroma (1991), Moss and Shonkwiler (1993) and Myers and Jayne (1997) are three examples of early applications, respectively focusing on the benefits that this methodology brings to the computation of yield distributions, the ability to take into account nonnormal errors and to distinguish between regime shifts with effects diffused over time and more immediate changes influencing the level of the yield. Other applications of the Kalman filtering include Cunha and Richter (2016) in relation to the production of wine.…”

Slow Magic: Agricultural Versus Industrial R&D Lag Models

“…This suspicion is supported to some extent by the evidence in the literature. For example, Myers & Jayne (1997) and Thirtle et al (2008) tried various lag structures and concluded that more positively skewed lag distributions (i.e., with heavier weight on the shorter lags) give rise to higher estimated rates of return. Likewise, in their statistical meta-analysis of studies of returns to agricultural R&D, Alston et al (2000, pp.…”