Modelling farmer resilience to rainfall shocks in NSW

This project seeks to understand the the role of weather shocks on livestock farm performance and resilience in the state of New South Wales, Australia. The project is an output of the ARC linkage grant Innovations in Agricultural Greenhouse Gas Management and Policy, where we broadly consider the viability and opportunity costs of carbon farming in NSW. 

The project is utilising the DataLab environment made available from the Australian Bureau of Statistics. In Datalab, we have used BLADE data, which captures business information reported to the Australian Tax Office for all registered ABNs in Australia. We focussed on ABN businesses within NSW with sheep and/or beef operations over a period between 2001 and 2020. 

The modelling of resilience to recognised that:

• The relationship between farm revenue outcomes and weather may be complex and dependent on both current and lagged weather outcomes.

• The relationship of weather and farm revenue may be mediated through expenditure and stocking decisions.

• The relationship may be heterogeneous across different production systems and climates throughout NSW.

The simulation model takes rainfall outcomes for 6 years (4 lags, current, and one expectation), and sequentially models changes in log stock value, expenditure, and revenue (gross margins) per ha over a defined time period. The model uses elasticities from the Arellano Bond Estimators from the econometric analysis to recursively estimate changes in input and output outcomes. The econometric analysis assumed the following decision pathway from the farmer: a) opening stock value decisions are made at the start of the (financial) year, observing revenue, expenditure, stock, rainfall, and terms of trade values in the previous year, while the choice is conditional on asset levels; b) expenditure decisions are then made observing these same past outcomes, and conditional on the stocking decision made earlier; and c) revenue outcomes are conditional on these input decisions, and is also influenced by prevailing rainfall in that year.