Robustness of geography as an instrument to assess impact of climate change on agriculture

Purpose - The empirical literature on climate change and agriculture does not adequately address the issue of potential endogeneity between climatic variables and agriculture, which makes their estimates unreliable. This paper aims to investigate the relationships between climate change and agriculture and test the potential reverse causality and endogeneity of climatic variables to agriculture. Design/methodology/approach - This study introduces a geographical instrument, longitude and latitude, for temperature to assess the impact of climate change on agriculture by estimating regression using IV-two-stage least squares method over annual panel data for 60 countries for the period of 1999-2011. The identification and F-statistic tests are used to choose and exclude the instrument. The inclusion of some control variables is supposed to reduce the omitted variable bias. Findings - The study finds a negative relationship between temperature and agriculture. Surprisingly, the magnitude of the coefficient on temperature is mild, at least 20 per cent, as compared to previous studies, which may be because of the use of the instrumental variable (IV), which is also supported by an alternative robust measure when estimated across different regions. Practical implications - The study provides strong implications for policymakers to confront climate change, which is an impending danger to agriculture. In designing effective policies and strategies, policymakers should focus not only on crop production but also on other agricultural activities such as livestock production and fisheries, in addition to national and international socio-economic and geopolitical dynamics. Originality/value - This paper contributes to the growing literature in at least four aspects. First, empirical settings introduce an innovative geographical instrument, Second, it includes a wider set of control variables in the analysis. Third, it extends previous studies by involving agriculture value addition. Finally, the effects of temperature and precipitation on a single aggregate measure, agriculture value addition, are separately investigated.