The global adoption and diffusion of conservation tillage has made considerable progress over the last 20 years. No-till and zero-tillage could be seen as representing the current technological end-point of the conservation tillage movement. This thesis uses descriptive statistics and both logit and duration regressions to analyse the influence of cross-sectional and time-dependent factors on the probability of no-till adoption by growers in Australia’s southern grain growing regions. Cross-section and time-series data on individual adoption decisions was gathered through interviews and employed in conjunction with generic time series data from various government agencies in a duration analysis modelling framework. Descriptive statistics suggest that weed management and herbicide resistance are important considerations for growers in their tillage decisions, predominantly due to the substitution of herbicides for the physical weed control provided by cultivation. Logit and duration regressions identify a number of significant factors influencing growers’ adoption decisions. These include growers’ perceptions of herbicide efficacy and sowing timeliness in no-till systems; the declining price of glyphosate relative to diesel; average annual rainfall and growers’ proximity to other adopters and opportunities to observe the beneficial effects of no-till. The results suggest that research and development of integrated weed management practices that are compatible with no-till systems is highly important if no-till systems are to be sustained in Australia’s southern wheatbelt. Such research and development should acknowledge the high value which growers place on locally generated information and the channels used to acquire such information, namely local extension events and consulting services. This thesis shows how duration analysis, with its ability to take account of both cross sectional and time-varying factors, can provide a statistical modelling framework better suited to the study of adoption decisions than traditional cross sectional methods based on logit and tobit analyses.
|Publication status||Unpublished - 2006|