How Econometrics Can Help Us Understand the Effects …
31
Dealing with Collinearities in a Multivariate Framework
As it has been previously stated, in order to econometrically model the effect of
climate change on crop yields, a multivariate framework should be used to control
for and also to integrate different groups of determinants (climatic, technological,
or economic) that may affect crop yields (see, for example, Huang and Khanna
2010). However, collinearity is often present in time-series data that show trending
behavior. Moreover, Auffhammer et al. (2013) indicate that many empirical studies
do not take into account all relevant climate dimensions, and many of them are
difficult to measure. Therefore, estimating a model that includes different climate
variables implies dealing with collinearities.
Some works focus on studying the effect of a single climate variable (such as
temperature or rainfall). Auffhammer and Schlenker (2014) warn that if a single
climate variable is used as a regressor, this measure will be subject to confounding
variation of other climate measures that are correlated with it and also affect the
variable of interest. This approach can lead to a classic problem of bias due to
omitted variables.
In order to consistently estimate the long-run coefficient, a partial system approach
can be followed as discussed in Juselius (2006). To take into account different potential determinants and to avoid collinearities, sub-systems of long-run relationships
due to climate, technological, and economic factors can be estimated and then evaluated by encompassing of the different ECM representations (as in Ahumada and
Cornejo 2019). The aim of testing encompassing is to address if crop yields adjust
to one or several long-run relationships. We should recall the invariance of the cointegration property to the extension of the information set (Juselius 2006). This property implies that once co-integration is found in a partial system, the co-integration
results will remain valid if more variables are added. Therefore, there would be no
omitted variable effects present for co-integration when adopting this partial system
approach.
Final Remarks
Effective adaptation strategies of crop production require long-run time-series analysis and a multivariate framework. In this line, the aim of this chapter has been to
illustrate how econometrics can help understand the effects of climate change on
the time behavior of crop yields at a country-level scale for the main producers and
exporter of soybeans. We have focused on this crop as a particular case of adaptation
and mitigation, with emphasis on the Argentine experience.
Climate econometrics provides an approach to give a rigorous basis for many
hypotheses related to climate change. In this line, we have discussed different empirical issues that an accurate econometric strategy should address in order to model
the effects of climate change on crop yields: the non-stationarity nature of climate
31
Dealing with Collinearities in a Multivariate Framework
As it has been previously stated, in order to econometrically model the effect of
climate change on crop yields, a multivariate framework should be used to control
for and also to integrate different groups of determinants (climatic, technological,
or economic) that may affect crop yields (see, for example, Huang and Khanna
2010). However, collinearity is often present in time-series data that show trending
behavior. Moreover, Auffhammer et al. (2013) indicate that many empirical studies
do not take into account all relevant climate dimensions, and many of them are
difficult to measure. Therefore, estimating a model that includes different climate
variables implies dealing with collinearities.
Some works focus on studying the effect of a single climate variable (such as
temperature or rainfall). Auffhammer and Schlenker (2014) warn that if a single
climate variable is used as a regressor, this measure will be subject to confounding
variation of other climate measures that are correlated with it and also affect the
variable of interest. This approach can lead to a classic problem of bias due to
omitted variables.
In order to consistently estimate the long-run coefficient, a partial system approach
can be followed as discussed in Juselius (2006). To take into account different potential determinants and to avoid collinearities, sub-systems of long-run relationships
due to climate, technological, and economic factors can be estimated and then evaluated by encompassing of the different ECM representations (as in Ahumada and
Cornejo 2019). The aim of testing encompassing is to address if crop yields adjust
to one or several long-run relationships. We should recall the invariance of the cointegration property to the extension of the information set (Juselius 2006). This property implies that once co-integration is found in a partial system, the co-integration
results will remain valid if more variables are added. Therefore, there would be no
omitted variable effects present for co-integration when adopting this partial system
approach.
Final Remarks
Effective adaptation strategies of crop production require long-run time-series analysis and a multivariate framework. In this line, the aim of this chapter has been to
illustrate how econometrics can help understand the effects of climate change on
the time behavior of crop yields at a country-level scale for the main producers and
exporter of soybeans. We have focused on this crop as a particular case of adaptation
and mitigation, with emphasis on the Argentine experience.
Climate econometrics provides an approach to give a rigorous basis for many
hypotheses related to climate change. In this line, we have discussed different empirical issues that an accurate econometric strategy should address in order to model
the effects of climate change on crop yields: the non-stationarity nature of climate