Controlling for the effects of climate on total factor productivity: A case study of Australian farms

Abstract

Estimates of agricultural Total Factor Productivity (TFP) can be highly sensitive to both short-run climate variability and long-term climate change. This is particularly true in Australia where drought impacts are responsible for most of the annual volatility in official farm TFP statistics. While climate variability can obscure short-term productivity trends, researchers have typically assumed that long run TFP trends are largely unaffected. However, in the presence of global climate change this assumption becomes problematic. For example, in Australia, shifts to higher temperatures and lower winter season rainfall over the last 20–30 years have had a significant negative impact on agricultural productivity. This study presents a framework to account for the effects of climate variability on TFP estimates. In contrast with previous work, this approach applies a reduced form machine learning based model of farm production to generate synthetic (climate-adjusted) farm-level input and output data sets. It therefore has advantages in terms of flexibility—since the synthetic datasets can be combined with any existing TFP estimation framework. In this study, the approach is applied to estimate climate-adjusted TFP indices (TFP under an assumption of constant long-run average climate conditions) for a range of Australian agricultural sectors.