22
H. Ahumada and M. Cornejo
grown GM soybeans are concentrated in a few countries, mainly USA, Brazil, and
Argentina.
Therefore, given the technological advances experienced by this sector, empirical
studies focused on studying the effect of climate change on crop yields should also
control for mitigation and adaptation processes that are actually taking place in the
agricultural sector.
On the Econometric Modeling of Soybean Yields
In this section, we discuss different empirical issues that an econometric model
of soybean yield determination should consider in order to measure the effects of
climate change in the agricultural sector. Hsiang (2016) offers a wide revision of the
different econometric methods that can be used to study the effect of climate change
on social and economic outcomes, in general.
Dealing with Trends
Variables may be classified according to their degree of time persistence into nonstationary or stationary. Non-stationarity is associated with the idea of long memory
(high persistence) of past shocks on the behavior of a time series (e.g., crop yields).
Such series could be stationary with a short-time dependence—that is, they could
exhibit a significant tendency to mean reversion—after first differencing. In those
cases, the series under study is said to have a unit root (a stochastic trend) or be
integrated of first order, I(1).
This kind of behavior is generally compared with a typical model of deterministic
trend to approximate the long-run behavior of a series. As stated by Lobell (2009), the
trend in crop yields results largely from improvements in technology and, thus, for
most crops the technology trend can be approximated with a first-order polynomial
(a linear trend).
For the series of our interest, different unit root tests reported in Table 1 show that
soybean yields can be represented as stationary around a deterministic linear trend.
This trending behavior was also observed in Fig. 1. Because of that, many studies
remove deterministic trends before studying the effects of climate factors on yields
(see for example Thomasz et al. (2016) in the Argentine case or Tao et al. (2008) in
the Chinese case).
Furthermore, given the nonlinear and non-stationary nature of crop yields, many
detrending methods have been suggested to model them (see the comparison of
detrending crop yield data techniques in Lu et al. 2017).
Nonetheless, it should be noted that if the aim of an empirical study is to understand which drivers could be behind the observed trending behavior, and the long-run
relationships between crop yields and their potential determinants should be stud-
H. Ahumada and M. Cornejo
grown GM soybeans are concentrated in a few countries, mainly USA, Brazil, and
Argentina.
Therefore, given the technological advances experienced by this sector, empirical
studies focused on studying the effect of climate change on crop yields should also
control for mitigation and adaptation processes that are actually taking place in the
agricultural sector.
On the Econometric Modeling of Soybean Yields
In this section, we discuss different empirical issues that an econometric model
of soybean yield determination should consider in order to measure the effects of
climate change in the agricultural sector. Hsiang (2016) offers a wide revision of the
different econometric methods that can be used to study the effect of climate change
on social and economic outcomes, in general.
Dealing with Trends
Variables may be classified according to their degree of time persistence into nonstationary or stationary. Non-stationarity is associated with the idea of long memory
(high persistence) of past shocks on the behavior of a time series (e.g., crop yields).
Such series could be stationary with a short-time dependence—that is, they could
exhibit a significant tendency to mean reversion—after first differencing. In those
cases, the series under study is said to have a unit root (a stochastic trend) or be
integrated of first order, I(1).
This kind of behavior is generally compared with a typical model of deterministic
trend to approximate the long-run behavior of a series. As stated by Lobell (2009), the
trend in crop yields results largely from improvements in technology and, thus, for
most crops the technology trend can be approximated with a first-order polynomial
(a linear trend).
For the series of our interest, different unit root tests reported in Table 1 show that
soybean yields can be represented as stationary around a deterministic linear trend.
This trending behavior was also observed in Fig. 1. Because of that, many studies
remove deterministic trends before studying the effects of climate factors on yields
(see for example Thomasz et al. (2016) in the Argentine case or Tao et al. (2008) in
the Chinese case).
Furthermore, given the nonlinear and non-stationary nature of crop yields, many
detrending methods have been suggested to model them (see the comparison of
detrending crop yield data techniques in Lu et al. 2017).
Nonetheless, it should be noted that if the aim of an empirical study is to understand which drivers could be behind the observed trending behavior, and the long-run
relationships between crop yields and their potential determinants should be stud-