INNOVA Research Journal, ISSN 2477-9024  
Septiembre, 2018). Vol. 3, No.9 pp. 157-171  
(
DOI: https://doi.org/10.33890/innova.v3.n9.2018.645  
URL: http://revistas.uide.edu.ec/index.php/innova/index  
Correo: innova@uide.edu.ec  
Methane emissions, economic growth and agriculture: evidence of  
environmental kuznets curve for Argentina  
Emisiones de metano, crecimiento económico y agricultura: evidencia de la  
curva ambiental de kuznets para Argentina  
Jessica Lissette Sánchez Cruz  
Luis Eduardo Solis Granda  
Manuel Leonardo Perrazo Viteri  
Universidad Estatal de Milagro-UNEMI, Ecuador  
Autor para correspondencia: lsolisg@unemi.edu.ec; mperrazov@unemi.edu.ec;  
jsanchezc5@unemi.edu.ec  
Fecha de recepción: 10 de abril de 2018 - Fecha de aceptación: 15 de septiembre de 2018  
Abstract: This paper provides evidence of the existence of an Environmental Kuznets Curve  
(EKC) in the long-run for Argentina from 1970 to 2012 which is the country with most production  
of meat in the region. There is a dynamic relationship between methane emissions, economic  
growth and agriculture activities. The autoregressive distributed lag methodology was used to test  
for cointegration in the long-run. Furthermore, we used the vector error correction model to test  
for causality and to verify the predictive value of independent variables. In fact, a quadratic  
relationship was found between methane emissions and economic growth. The effect of agriculture  
was the only unexpected, and that is because the reduction of methane emissions thanks to suitable  
policies related to the use of technology in agriculture activities.  
Key words: environmental kuznets curve; ardl; Argentina; methane emissions, gdp; agriculture  
JEL Code: C32, Q01, Q50, Q51, Q56  
Resumen: Este documento proporciona evidencia de la existencia de una curva ambiental de  
Kuznets (EKC) a largo plazo para Argentina desde 1970 hasta 2012, que es el país con mayor  
producción de carne en la región. Existe una relación dinámica entre las emisiones de metano, el  
crecimiento económico y las actividades agrícolas. La metodología de retraso distribuido  
autorregresivo se utilizó para probar la cointegración a largo plazo. Además, utilizamos el modelo  
de corrección de errores vectoriales para probar la causalidad y verificar el valor predictivo de las  
variables independientes. De hecho, se encontró una relación cuadrática entre las emisiones de  
metano y el crecimiento económico. El efecto de la agricultura fue el único inesperado, y esto se  
debe a la reducción de las emisiones de metano gracias a políticas adecuadas relacionadas con el  
uso de la tecnología en las actividades agrícolas.  
Palabras clave: curva de kuznets ambiental; ardl; argentina; emisiones de metano; pib; agricultura  
Revista de la Universidad Internacional del Ecuador. URL: https://www.uide.edu.ec/  
157  
INNOVA Research Journal 2018, Vol 3, No. 9, pp. 157-171  
Introduction  
The Kuznets Curve was proposed by Simon Kuznets in 1955. He found that there was an  
existence of a quadratic relationship between economic growth and income inequality. Inequality  
rises up along with economic growth until a turning point in which the trend inverts (Kuznets,  
1
955). The same explanation was used to describe the environmental degradation relating a Green  
House Gases (GHG) such as carbon dioxide (CO2), methane (CH4) or Nitrous Oxide (N2O) with  
an economic growth variable such as Gross Domestic Product. As long as a country increases its  
gross domestic product, its GHG emissions will increase as well, until a turning point in which  
technology makes a country more efficient in the way the anthropogenic activities are done and  
the emissions of those gases start to reduce (Kraft and Kraft, 1978).  
Since Kuznets discovery, several different countries were studied to provide empirical  
evidence of the existence of an EKC. Besides GDP, other variables related to environmental  
degradation were included in the models over time, including foreign trade (Hossain, 2011),  
urbanization (Zhang and Cheng, 2009) and energy consumption (Saboori and Sulaiman, 2013).  
There are few studies that demonstrate an EKC with a methane emissions and GDP per  
capita relationship, but in this paper we are going to show that GDP per capita and agriculture have  
an inverted U-shaped relationship.  
As shown in figure 1, methane emissions are the second largest GHG emissions in the  
world (IPCC, 2014),and they are principally generated as a result of agriculture and livestock  
farming activities. Argentina is one of the largest producers of meat in the region (INTA, 2014)  
with around 51 million cattle. Agriculture is the third principal economic activity in Argentina,  
accounting for around 10% of the total gross domestic product (MECON, 2012).  
Those are the reasons that motivate us to study the Argentinian case and found the existence  
of an EKC for the period of 1970 to 2012.  
Figure 1: Total annual anthropogenic GHG emissions by gases 1970–2010  
Source: IPCC, 2014  
The methodology used is an autoregressive distributive lag (ARDL) bounds testing  
approach to cointegration with a series time analysis from 1970 to 2012. Results show an EKC for  
the short-run as well as for the long-run. As expected, agriculture is statistically significant;  
Revista de la Universidad Internacional del Ecuador. URL: https://www.uide.edu.ec/  
158  
INNOVA Research Journal 2018, Vol 3, No. 9, pp. 157-171  
however, in the long run has a negative impact on emissions, due to Argentinian environmental  
policies and the incremental technology used for those activities.  
The rest of the paper is organized as follows: Section 2 explains the environmental and  
economic situation of Argentina. Section 3 defines the theoretical and modeling framework.  
Section 4 presents the methodology to be used. Section 5 shows the empirical results. Section 6  
concludes.  
Argentine Context.  
Argentina is a South American country with a population of 41.45 million people, of which  
almost 10% are rural population (WBG, 2013). Argentina is the third largest economy in South  
America, just behind Brazil and Chile. Its average growth rate of real gross domestic product  
(GDP) per capita was 2.55% between 1980 and 2012, as shown in figure 2 (MECON, 2012).  
Argentina has a Gini coefficient of 0.423 (WBG, 2013) and a Score of 0.836 in human  
development (United Nations Development Program, 2014). Both scores are improving over time,  
illustrating the improvement in Argentinian living conditions as a result of economic growth.  
Figure 2: Graphic representations of GDP pc, Gini Index, HDI and Methane emissions for Argentina  
In the environmental context, Carbon Dioxide (Co2) is the GHG most produced because  
of the human activity, but in this paper we’re going to focus on the second GHG most emitted,  
methane (CH4). Methane, in general, is generated as a result of anthropogenic activities,  
principally agriculture. According to the World Bank Group data, Argentina and Brazil produce  
the most meat in the region, and also emit the most methane from livestock farming activities.  
Agriculture and livestock farming contribute 44% of total GHG emissions in the country,  
just behind the energy sector (Berra, 2000) with 48% of total methane emissions. Of the total GHG  
emissions, 30% comes from livestock farming, and 95% of that comes from cattle (IICA, 2015).  
Livestock farming contributes to the methane emissions from enteric fermentation and excretions  
of animals. These last two are also a source of nitrous oxide, just as nitrogen-fixing fodder. In  
Revista de la Universidad Internacional del Ecuador. URL: https://www.uide.edu.ec/  
159  
INNOVA Research Journal 2018, Vol 3, No. 9, pp. 157-171  
agricultural activities, these emissions occur as a result of nitrogen-fixing crops, including  
soybeans and Stubble. Commercial fertilizers also contribute to the emission of nitrous oxide,  
while rice cultivation generates methane emissions. Finally, burning Stubble produces nitrous  
oxide emissions, other nitrogen oxides, carbon monoxide and methane.  
Even though methane is the second largest GHG in Argentina, its emissions have been  
reduced as a result of an increase in the technology used in agriculture and livestock farming, as  
shown in figure 2.Argentina had to apply Innovative processes and change macroeconomics  
policies in order to let the rural population to begin a process of economic and productive recovery.  
This growth is accompanied by sustainable development policies for agricultural and rural sector.  
The Agri-Food and Agribusiness Strategic Plan (PEA2) and the Smart Agriculture Plan (AI) were  
created in order to generate a more efficient, competitive and sustainable production. Promoting  
smart agriculture involves developing active policies in the agricultural sector to harmonize  
production and environmental systems, while at the same time representing the Argentine  
government's response to the challenge of food security in the context of climate change.  
Theoretical and modeling framework  
The EKC hypothesis indicates that the relationship between economic growth and  
environmental degradation has an inverted U shape.In the short-run, the economic growth of a  
country has a negative impact on the environment seen in the rising part of the curve; but in the  
long-run, when the economy reaches its highest point of income, known as the turning point, the  
curve descends, illustrating the positive impact of economic growth on the environment.  
The model is structured as follows  
Ln (Et)=β_(0 )+ β_1 Ln(Yt)+β_2 〖(Ln (Yt))〗^2+ β_3 〖(Ln (Yt))〗^3+β_4 Ln 〖(Z〗_t)+μt  
Where the dependent variable is an indicator of environmental contamination measured in  
logarithms, β0, β1, β2, β3 are the parameters to be estimated, Y is the per capita income in  
logarithms, Z is the vector of additional variables, also measured in logarithms and finally μ is the  
error term.  
This work suggest in the equation 1 that methane emissions (CH4) depend on GDP, square  
of GDP (GDP^2) and the agriculture (AGRI) for the period 1970-2012 in the case of Argentina.  
〖
CH〗_4=f(GDP,〖GDP〗^2, AGRI)  
(1)  
The model would be as follows:  
Ln (〖CH〗_4) =β_ (0) + β_1 Ln (GDP) +β_2 〖 (Ln (GDP)) 〗^2+β_3 Ln (AGRI) +μt  
2)  
(
The theory suggests that in order to get an EKC, this should have the following relationship:  
〖
β〗_1> 0, β_2<0 which have the shape of an inverted U. It is expected thatβ_GDP>0 andβ_  
Revista de la Universidad Internacional del Ecuador. URL: https://www.uide.edu.ec/  
160  
INNOVA Research Journal 2018, Vol 3, No. 9, pp. 157-171  
(GDP^2) <0; the sign of β_AGRI>0 if we assume that activities concerning agriculture are handled  
without any significant technological improvement during the period analyzed.  
The data on all variables come from World Development Indicators (WDI). The methane  
emissions (〖CH〗_4) is proxied by Methane emissions (kt of CO2 equivalent), the GDP per  
capita (constant 2005 US$) and AGRI is for Agriculture, value added (% of GDP).  
Methodology  
ARDL Bounds Testing of Cointegration  
The application of the ARDL bounds testing approach to cointegration developed by  
Pesaran and Pesaran (1997), Pesaran et al. (2000, 2001) allows us to examine the long-run  
relationship between methane emission, economic growth and the agricultural.  
The methodology used is the Auto regressive model with distributed lags which was  
proposed by Pesaran et al. (2001), which provides better results for small samples as proposed by  
Haug (2002), less than 50 data samples, such as the proposed case. This model also can be applied  
without investigation the order of integration, but a requirement is that they should not be at second  
difference I(2). The ARDL model provides better results with this type of samples, compared to  
traditional approaches to cointegration, like Engle and Granger Granger (1987), Johansen and  
Juselius (1990) and Phillips and Hansen (1990). Laurenceson and Chai (2003) affirm that another  
advantage of ARDL limit testing is that the model is not restricted model error correction (ECM),  
and has sufficient flexibility to accommodate lags that capture the data generating process in a  
general framework of specification.  
The unrestricted model is indicated as follows:  
∆
ln 퐶퐻 = 훼 + 훽퐺퐷푃푙푛ꢀꢁꢂ푡−1 + 훽  
2
푙푛ꢀꢁꢂꢃ  
+ 훽  
푙푛ꢄꢀꢅꢆ푡−1  
퐴퐺푅퐼  
푚
4푡  
0
퐺퐷푃  
푡−1  
푝
푞
+
∑ 훼 ∆푙푛퐶퐻  
+ ∑ 훼 ∆푙푛ꢀꢁꢂ + ∑ 훼 ∆푙푛ꢀꢁꢂꢃ  
4푡−푖 푗 푡−푗 푘  
푡−푘  
푖
푖ꢇ1  
ꢉ
푗ꢇ0  
푘ꢇ0  
+
∑ 훼 ∆푙푛ꢄꢀꢅꢆ + 휇푡  
(3)  
ꢈ
푡−ꢈ  
ꢈ
In order to determine whether there is cointegration of the variables, it is necessary to use  
the critical values tabulated by Pesaran et al. (2001), where the null hypothesis of no cointegration  
is β_GDP=β_(GDP^2 )=β_AGRI=0 and the alternative hypothesis that represents cointegration of  
the variables is β_GDP≠β_(GDP^2 )≠β_AGRI≠0. With this, we can obtain the F-calculated which  
is compared with the upper and lower critical bound values from Pesaran et al. (2001). Another  
option is to use the critical values proposed by Narayan (2005) as these are more appropriate for  
small samples, as in our case. If the value of F-calculated exceeds the critical value, then we have  
evidence that the variables are cointegrated. On the other hand, if the F-statistic is less than the  
critical value, we cannot reject the null hypothesis of no cointegration. Finally, if the calculated F-  
statistic is between lower and upper critical bounds, the cointegration decision is not conclusive.  
If the null hypothesis of no cointegration is rejected, the behavior of the variables in the short-run  
Revista de la Universidad Internacional del Ecuador. URL: https://www.uide.edu.ec/  
161  
INNOVA Research Journal 2018, Vol 3, No. 9, pp. 157-171  
will be captured by the error correction term 〖 (ECT〗_ (t-1)) incorporated in equation 3as  
follows:  
푝
푞
푚
∆
ln 퐶퐻 = 훿 + ∑ 훿 ∆푙푛퐶퐻  
+ ∑ 훿 ∆푙푛ꢀꢁꢂ + ∑ 훿 ∆푙푛ꢀꢁꢂꢃ  
4푡−푖 ꢃ푗 푡−푗 ꢊ푘  
푡−푘  
4푡  
1
1푖  
푖ꢇ1  
ꢉ
푗ꢇ0  
푘ꢇ0  
+
+
∑ 훿 ∆푙푛ꢄꢀꢅꢆ  
4ꢈ 푡−ꢈ  
ꢈ
훾퐸퐶푇푡−1+휇푡  
(ꢋ)  
The coefficient of ECT (γ) indicates the speed of adjustment and shows how quickly the  
variables return to the long-run equilibrium (Masih and Masih, 1997), that coefficient should be  
negative and significant.  
Finally, diagnostic tests are performed to check the suitability of the model, including the  
Jarque-Bera normality test, Breusch-Godfrey serial correlation LM test, ARCH hetoscedasticity  
test, Ramsey RESET test and cumulative sum/- squared (CUSUM/CUSUMSQ) test.  
Causality Analysis  
The presence of cointegration between variables implies that the causal relation must exist  
at least in one direction; the ARDL model does not show what the causality direction is. In order  
to explain the causality in the short run and long run of the variables it is necessary to apply a  
vector error correction model (VECM) to examine for cointegrated variables.  
VECM permits to analyze two forms of causality. One of them is the short-run causal  
relationship and the other one is the long-run causal relationship. It is necessary, in order to get a  
short-run granger-causal relationship, for the lagged differenced explanatory variables to be  
significant. To get a long-run granger causal relationship, it is necessary for the lagged ECT to be  
significant, as well. (Masih and Masih, 1996).  
In this case, Estimate the residuals of the long-run model as a proxy of the ECT is the first  
step. Then, as a second step, we need to estimate the VECM as follows:  
∆
∆
ln 퐶퐻4푡  
ꢀꢁꢂ푡  
휇1  
휏11,1 휏1ꢃ,1  
휏 휏  
1ꢊ,1 14,1  
ꢌ
ꢌ
ꢌ
⌊
ꢍ
ꢍ
휇
ꢃ
휏
휏ꢃꢃ,1  
휏
휏
휏
=
[ ] + ꢎ ꢃ1,1  
ꢃꢊ,1 ꢃ4,1ꢏ  
ꢃ
휇
ꢊ
∆
ꢀꢁꢂ  
푡
ꢍ  
휏ꢊ1,1 휏ꢊꢃ,1  
휏
ꢊꢊ,1 ꢊ4,1  
휇4  
∆ꢄꢀꢅꢆ ⌋  
푡
∆
∆
ln 퐶퐻4푡−1  
휎1  
휗1  
ꢌ
ꢌ
ꢌ
⌊
ꢍ
ꢍ
ꢀꢁꢂ푡−1  
ꢀꢁꢂꢃ  
휎
휗
ꢃ
ꢃ
+
[ ] 퐸퐶푇 + ꢐ ꢑ  
휎
푡−1  
∆
ꢊ
휗ꢊ  
푡−1
ꢍ  
휎4  
∆ꢄꢀꢅꢆ푡−1  
⌋
휗4  
Where the vector of ϑ t's is white noise.The σkare interpreted as the speed of adjustment  
which represent the response of the dependent variable to deviations from the long-run  
equilibrium.  
Revista de la Universidad Internacional del Ecuador. URL: https://www.uide.edu.ec/  
162  
INNOVA Research Journal 2018, Vol 3, No. 9, pp. 157-171  
Empirical Results  
According to Ouattara (2004), if a variable is integrated into I (2), then the F-statistic for  
the cointegrating is inconclusive.It is necessary that the variables become stationary until I(1). We  
use the ADF unit root test to check the stationarity of our variables. The results show that there are  
not unit root problems. Table 1 indicates that all variables are non-stationary at level, but these  
become stationary at the first difference.  
The selection of the maximum lag length for each variable has been determined using the  
SIC (Schwarz information criteria) in which the minimum value is taken. Table 2 presents several  
of the combination sets of lags, including the one chosen for the model (1, 0, 0, 1).  
The next step is the calculation of F-statistic cointegration, as shown in Table 3. The results  
indicate that the calculated value is above the upper bound of 3.454 obtained through critical values  
proposed by Narayan (2005). It is concluded thata cointegration relationship exists between  
methane emissions, GDP, 〖GDP〗_2 and AGRI, when methane emissions is the dependent  
variable. Table 3 also shows the results of the respective diagnostic tests.  
The long-run estimates are reported in Table 4. The results show that the coefficients of the  
variables are significant. While the values for GDP and GDP_2are the expected, the AGRI  
coefficient is contrary to the expected. In this case heteroskedasticity and serial autocorrelation  
was detected; we will work with robust errors of white and residues laggards one period to deal  
with both problems respectively. All the coefficients are significant at 1%. The estimations in the  
long-run show the existence of an EKC in Argentina. The methane emissions increase when  
income does, until a turning point and then the emissions start to decrease while income continues  
to rise.The long-run elasticity between methane missions and agriculture is -0.058%. This means  
that a 1% rises in agriculture, the methane emissions decrease by -0.058%.  
The short-run model is shown in Table 5. The AGRI variable is significant at 1% as the  
error correction term, wherein the coefficient of the latter is shown negative; this confirms the  
existence of the cointegrating equation. Moreover, the coefficient of ECT means that the  
deviations from equilibrium methane are corrected by 28.18% within a year.  
The causality based on VECM is reported in Table 6. There are two portion of this table.  
The first portion is showing the short- run causality (F-statistic). The second portion presenting  
the long-run causality indicated through significance of ECT (t-statistic). The short run causal  
effects revealed that the agriculture is the only variable that has effect on methane emissions, The  
short run causal effects revealed that the agriculture is the only variable that has effect on methane  
emissions, while in the long run the results indicate that there is a bidirectional causality in all  
variables.  
To verify this, the variance decomposition was implemented, Table 7 shows the results  
which indicates that; a change in one standard deviation in GDP, GDP2 and AGRI represents a  
shock of the 26.23%, 22.10% and 30.71% respectively in CH4 emissions. Given that these shocks  
are higher if it was contrary (0.23%, 0.22%, 10.52%), then there is an unidirectional Granger  
causality of the variables to CH4.  
Revista de la Universidad Internacional del Ecuador. URL: https://www.uide.edu.ec/  
163  
INNOVA Research Journal 2018, Vol 3, No. 9, pp. 157-171  
The CUSUM and CUSUMQ are used to check the stability of the coefficients for the short  
and long-run. Figures 2 and 3 show that the coefficients are stable with a significance level of 5%.  
The results suggest that the model can be used for policy proposal.  
Table 1: Unitroot test  
Table 1: Unitroot test  
Variable  
T-Statistics  
P value*  
ADF test at level with intercept and without trend  
Ln CH4t  
LnGDPt  
Ln GDPt2  
LnAGRIt  
-2.0837  
-0.6068  
-0.5588  
-2.8324  
0.252  
0.8580  
0.8687  
0.0624  
ADF test at first difference with intercept and without trend  
Δ Ln CH4t  
Δ LnGDPt  
Δ Ln GDPt2  
Δ LnAGRIt  
-4.9075  
-5.0591  
-5.0375  
-6.6400  
0.0002  
0.0002  
0.0002  
0.0000  
*MacKinnon (1996) considered P values  
Table 2: Lag Length selection criteria  
Table 2: Lag Length selection criteria  
Lagcombination  
SIC  
F-statistic  
2.1388  
2.170226  
2.785578  
2.97092  
2.475947  
3.529861  
2.555225  
3.005556  
4.021315**  
2.960585  
P value  
(
(
(
(
(
(
(
(
(
(
2.2.2.2)  
2.0.1.0)  
2.0.0.2)  
2.0.0.1)  
2.0.0.0)  
1.1.1.1)  
1.0.1.0)  
1.0.0.2)  
1.0.0.1)  
1.0.0.0)  
-4.085603  
-4.256634  
-4.344973  
-4.40326  
-4.345665  
-4.415757  
-4.302162  
-4.409142  
-4.520785  
-4.391765  
0.0468  
0.0507  
0.014012  
0.010657  
0.030139  
0.003156  
0.025157  
0.00998  
0.001729  
0.013467  
SIC: Schwarz information criteria, **indicates statistical significance at 5% level  
Table 3: Cointegrationtestsresults  
Table 3: Cointegrationtestsresults  
Boundstesting to cointegration  
Estimatedequation  
Optimallagstructure  
F-statistics  
CH4=f(GDP, GDP2, AGRI)  
SIC: Schwarzinformationcriteria  
3.532992***  
Diagnosticcheck  
R2  
Adjusted-R2  
F-statistics (P)  
J-B Normality test  
Breusch-Godfrey LM test [2]  
0.5386  
0.4047  
4.0213  
0.6934  
1.5870  
Revista de la Universidad Internacional del Ecuador. URL: https://www.uide.edu.ec/  
164  
INNOVA Research Journal 2018, Vol 3, No. 9, pp. 157-171  
ARCH LM test [2]  
Ramsey RESET  
CUSUM  
0.6309  
0.8098  
Stable  
Stable  
CUSUMSQ  
*
** The significant at 10% level. The optimal structure is determined by SIC  
SIC: Schwarzinformationcriteria  
Table4: Long run estimates  
Table4: Long run estimates  
Dependent variable=Ln CH4t  
Variable  
Coefficient  
Standard error  
6.6444  
T-statistic  
-1.089042  
2.933441*  
-3.03114*  
-3.702024*  
Constant  
-7.2360  
4.5249  
LnGDPt  
1.5425  
Ln GDP2t  
-0.27157  
-0.058341  
0.089593  
0.015759  
LnAGRIt  
Diagnosticcheck  
R-squared  
0.816254  
-4.6340  
-4.4272  
41.0913  
1.61607  
1.2463  
Akaikeinfocriterion  
Schwarzcriterion  
F-statistic  
Durbin–Watson  
Serial correlation LM [2]  
ARCH test [2]  
Normality test  
Ramsey RESET test  
0.0846  
1.1439  
0.4262  
*1% level of significance  
Table 5: Short run estimates  
Table 5: Short run estimates  
Dependent variable=Δ Ln CH4t  
Variable  
Coefficient  
Standard error  
0.003091  
3.390862  
0.199407  
0.017582  
0.085189  
T-statistic  
0.49142  
0.379037  
-0.363845  
-1.530678  
-3.308902*  
Constant  
Δ LnGDPt  
Δ Ln GDPt2  
Δ LnAGRIt  
ECT(-1)  
0.001519  
1.285262  
-0.072553  
-0.026913  
-0.281883  
Diagnosticchecks  
R-squared  
0.491924  
-4.988632  
-4.696071  
5.486531  
1.903009  
0.4643  
Akaikeinfocriterion  
Schwarzcriterion  
F-statistic  
Durbin–Watson  
Serial correlation LM [2]  
ARCH test [2]  
Normality test  
0.0222  
0.3433  
Revista de la Universidad Internacional del Ecuador. URL: https://www.uide.edu.ec/  
165  
INNOVA Research Journal 2018, Vol 3, No. 9, pp. 157-171  
Ramsey RESET test  
1% level of significance  
0.5945  
*
Table 6: Causality results based on VECM.  
Table 6: Causality results based on VECM.  
Variable  
Short Run (F-stat.)  
Long run (t-stat.)  
ECT  
Δ ln(CH4)  
Δ ln(GDP)  
0.851638  
4.075853)  
Δ ln(GDP2)  
0.870510  
(0.240178)  
2.766037  
(0.639166)  
-
Δ ln(AGRI)  
15.41688***  
(0.018777)  
0.024518  
(0.049969)  
0.025557  
(0.847712)  
-
Δ ln(CH4)  
Δ ln(GDP)  
Δ ln(GDP2)  
Δ ln(AGRI)  
-
-2.132135**  
(0.079872)  
3.077004***  
0.252556  
(
3.794231  
-
(
0.388378)  
3.956387  
2.752860  
(184.0129)  
0.000897  
3.094094***  
3.605982  
(
6.588783)  
0.939578  
0.003157  
-2.422722**  
0.660550  
(
1.206945)  
* Significance at 5% level.  
** Significance at 1% level.  
(33.70782)  
(1.986306)  
*
*
Table 7: Error VarianceDecomposition  
Table 7: Error VarianceDecomposition  
Variance Decomposition of Δln(CH4):  
Period  
S.E.  
Δ ln(CH4)  
100  
Δ ln(GDP)  
0
Δ ln(GDP2)  
0
Δ ln(AGRI)  
1
2
3
4
5
6
7
8
9
0.017924  
0.025177  
0.027707  
0.030826  
0.034713  
0.039347  
0.044972  
0.051341  
0.058277  
0.065763  
0
88.81764  
85.75045  
77.24769  
65.94547  
53.78145  
42.47137  
33.36988  
26.35869  
20.96539  
0.014629  
4.236748  
10.8478  
15.15791  
18.36229  
21.01229  
23.16881  
24.9032  
26.22884  
7.425548  
6.785603  
6.644078  
8.530882  
11.75429  
15.28986  
18.19393  
20.39408  
22.10349  
3.742183  
3.227198  
5.260431  
10.36573  
16.10197  
21.22648  
25.26738  
28.34403  
30.70228  
1
0
Variance Decomposition ofΔln(GDP):  
Period  
S.E.  
Δ ln(CH4)  
1.563922  
1.027026  
0.714228  
0.655445  
0.554025  
0.44595  
Δ ln(GDP)  
98.43608  
92.07062  
81.57238  
67.73927  
56.40919  
49.53755  
45.51527  
42.92245  
Δ ln(GDP2)  
0
Δ ln(AGRI)  
0
1
2
3
4
5
6
7
8
0.052967  
0.08043  
0.097611  
0.115068  
0.132993  
0.149031  
0.163801  
0.178447  
0.589216  
2.705943  
8.614837  
14.96274  
19.17593  
21.70486  
23.28829  
6.313138  
15.00745  
22.99045  
28.07405  
30.84058  
32.41067  
33.47794  
0.369199  
0.311324  
Revista de la Universidad Internacional del Ecuador. URL: https://www.uide.edu.ec/  
166  
INNOVA Research Journal 2018, Vol 3, No. 9, pp. 157-171  
9
1
0.193417  
0.20885  
0.265764  
0.228767  
41.00211  
39.42088  
24.37428  
25.21497  
34.35784  
35.13538  
0
Variance Decomposition of Δln(GDP2):  
Period  
S.E.  
Δ ln(CH4)  
1.522484  
1.051577  
0.736277  
0.651917  
0.541912  
0.433508  
0.357789  
0.301136  
0.256776  
0.220851  
Δ ln(GDP)  
98.45155  
91.77875  
81.07585  
67.13369  
55.7987  
Δ ln(GDP2)  
0.02597  
Δ ln(AGRI)  
0
1
2
3
4
5
6
7
8
9
0.89965  
1.37085  
1.668548  
1.972737  
2.2856  
0.792375  
3.071479  
9.100331  
15.45998  
19.63881  
22.12314  
23.66076  
24.70324  
25.5043  
6.377295  
15.11639  
23.11406  
28.19941  
30.9681  
32.53871  
33.60297  
34.47663  
35.24587  
2.566035  
2.824543  
3.080823  
3.34263  
3.6124  
48.95958  
44.98036  
42.43513  
40.56335  
39.02898  
1
0
Variance Decomposition of Δln(AGRI):  
Period  
S.E.  
Δ ln(CH4)  
3.651318  
6.481976  
6.827817  
7.68437  
Δ ln(GDP)  
0.68262  
Δ ln(GDP2)  
17.82866  
12.27059  
13.41201  
15.83099  
18.44538  
19.85527  
19.93336  
19.67769  
19.56565  
19.63535  
Δ ln(AGRI)  
77.8374  
1
2
3
4
5
6
7
8
9
0.163402  
0.197369  
0.203531  
0.208048  
0.215282  
0.220193  
0.222258  
0.223702  
0.225526  
0.228005  
9.538751  
11.42669  
11.0265  
71.70868  
68.33348  
65.45814  
61.41675  
58.8075  
8.958847  
9.68185  
11.17903  
11.65538  
12.23311  
12.65993  
12.97685  
13.33153  
10.08876  
10.36469  
10.50875  
10.51623  
57.74477  
57.29769  
56.94875  
56.51688  
1
0
Figure3: Plot of cumulative sum of recursive residuals  
Revista de la Universidad Internacional del Ecuador. URL: https://www.uide.edu.ec/  
167  
INNOVA Research Journal 2018, Vol 3, No. 9, pp. 157-171  
1
1
1
0
0
0
0
0
.4  
.2  
.0  
.8  
.6  
.4  
.2  
.0  
-
0.2  
0.4  
-
1
980  
1985  
1990  
1995  
2000  
2005  
2010  
CUSUM of Squares  
5% Significance  
Figure 4:Plot of cumulative cum of squares of recursive residuals  
Conclusion and policy implications  
The objective of this paper was to empirically examine the short-run and long-run  
relationships of methane emissions, GDP per capita, and agriculture in Argentina for the period of  
1
970-2012. Using the ARDL model proposed by Pesaran et al, (2001) we observed that the  
coefficients of the variables GDP and 〖GDP〗^2 were positive and negative respectively,  
suggesting the existence of an inverted curve U-shape. Assuming that there is indeed an  
Environmental Kuznets Curve (EKC) in Argentina, and considering the publication of Stern et al.  
(1996), we cannot conclude that economic growth can improve the environment, but we should  
consider policies that have been established in Argentina to achieve sustainable development.  
Contrary to expectations, the coefficient of agriculture variable was negative; this can be  
justified with the technological innovations employed in the agricultural sector in this country, the  
Agri-Food and Agribusiness Strategic Plan (PEA2) and the Smart Agriculture Plan (AI).  
Countries whose economic policies induce a rapid expansion of income and employment may  
experience serious environmental damage unless appropriate environmental regulations are taken  
Dasgupta (2002). Martin (2002) came to the same conclusion, that the Environmental Kuznets  
Curve can only be expected when the respective measures are taken.  
The existence of ECK in Argentina shows changes of a growing economy in which  
appropriate technologies have been implemented to reduce the environmental impact.  
Bibliography  
Basso, L. R., C. Pascale Medina, E. S. de Obschatko, J. Preciado Patiño, 2013. Agricultura  
Inteligente: la iniciativa de la Argentina para la sustentabilidad en la producción de  
alimentos y energía.  
Ministerio de Agricultura, Ganadería y Pesca de la Nación, Instituto Interamericano de  
Cooperación para la Agricultura. Buenos Aires, 124 pp.  
Berra Guillermo. (2000), Influencia de la ganadería argentina: Emisión de Gases de Efecto  
Invernadero.  
Revista de la Universidad Internacional del Ecuador. URL: https://www.uide.edu.ec/  
168  
INNOVA Research Journal 2018, Vol 3, No. 9, pp. 157-171  
INTA. Castelar, 1-4.  
CEPAL, FAO, IICA(2014) Perspectivas de la agricultura y del desarrollo rural en las Américas:  
una mirada hacia América Latina y el Caribe. San José, C.R.: IICA, 2013. 230 p.  
Dasgupta, S., Laplante, B., Wang, H., and Wheeler, D. (2002).Confronting the Environmental  
Kuznets Curve. The Journal of Economic Perspectives, Vol. 16, No. 1.Edited by Winkler,  
H., Davidson, O.,  
Kenny, A., Prasad, G., Nkomo, J., Sparks, D., Howells, M., and Alfstad, T. (2006).  
Engle, R.F., Granger, C.W.J. (1987), Cointegration and error correction representation: estimation  
and testing. Econometrica, 55, 251-276.  
Hossain, S. (2011), Panel estimation for CO2 emissions, energy consumption, economic growth,  
trade openness and urbanization of newly industrialized countries.EnergyPolicy, 39, 6991-  
6
999.  
IICA, (2015). Cambio climático y agricultura en la Argentina: aspectos institucionales y  
herramientas de información para la formulación de políticas. Edith S. de Obschatko.  
Buenos Aires.  
IPCC, 2014: Climate Change 2014: Synthesis Report. Contribution of Working Groups I, II and  
III to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change [Core  
Writing Team, R.K. Pachauri and L.A. Meyer (eds.)]. IPCC, Geneva, Switzerland, 151 pp.  
Johansen, S., Juselius, K. (1990), Maximum likelihood estimation and inference on cointegration  
with applications to the demand for money. Oxford Bulletin of Economics and Statistics,  
5
2, 169-210.  
Kraft, J., Kraft, A. (1978), On the relationship between energy and GNP. Journal of Energy and  
Development, 3, 401-403.  
Kuznets, S. (1955), Economic growth and income inequality. The American Economic Review,  
4
5, 1-28.  
Laurenceson, J., Chai, J.C.H. (2003), Financial Reforms and Economic Development in China.  
Cheltenham, UK: Edward Elgar.  
Masih, A.M.M., Masih, R. (1997), on temporal causal relationship between energy consumption,  
real income and prices: some new evidence from Asian energy dependent NICs based on  
a multivariate cointegration vector error correction approach. Journal of Policy Modeling,  
1
9, 417-440.  
Revista de la Universidad Internacional del Ecuador. URL: https://www.uide.edu.ec/  
169  
INNOVA Research Journal 2018, Vol 3, No. 9, pp. 157-171  
Martin, Christoph. (2002). The Environmental Kuznets Curve: A survey of the empirical evidence  
and possible causes.  
Ministerio de Economía y Finanzas, (2012). La Macroeconomía Argentina: 2003-2012. Buenos  
Aires.  
Narayan, P.K. (2005), the saving and investment nexus for China: evidence from cointegration  
tests.  
Applied Economics, 37, 1979-1990.  
Newey, W.K., West, K.D. (1987), a simple, positive semi-definite heteroskedasticity and  
autocorrelation consistent covariance matrix. Econometrica, 55, 703-708.  
Ouattara, B. (2004), The Impact of Project Aid and Programme Aid on Domestic Savings: a Case  
Study of Cote d’Ivoire.” Centre for the Study of African Economies Conference on Growth,  
Poverty Reduction and Human Development in Africa  
Pesaran, M.H., Pesaran, B. (1997), Working with Microfit 4.0: Interactive Econometric Analysis.  
Oxford: Oxford University Press.  
Pesaran, M.H., Shin, Y., Smith, R.J. (2000), Structural analysis of vector error correction models  
with exogenous I(1) variables. Journal of Econometrics, 97, 293-343.  
Pesaran, M.H., Shin, Y., Smith, R.J. (2001), Bounds testing approaches to the analysis of level  
Relationships. Journal of Applied Econometrics, 16, 289-326.  
Phillips, P.C.B., Hansen, B.E. (1990), Statistical inference in instrumental variables regression  
with I (1) processes. Review of Economic Studies, 57, 99-125.  
Saboori, B., Sulaiman, J. (2013), Environmental degradation, economicgrowth and energy  
consumption: Evidence of the environmental Kuznets curve in Malaysia. Energy Policy,  
6
0, 892-905.  
Shahbaz, M., Dube, S., Ozturk, I., Jalil A. (2015), Testing the Environmental Kuznets Curve  
Hypothesis in Portugal. Energy, 5(2), 2146-4553.  
Stern D.I, Common, M., Barbier, E. (1996), Economic growth and environmental degradation: the  
environmental Kuznets curve and sustainability. World Development, 24, 1151-1160.  
White, H., 1980, Aheteroskedasticity-consistent covariance matrix and a direct test for  
heteroskedasticity, Econometrica 48, 817-838.  
World Bank, Group, (2006). Argentina: Agriculture and Rural Development: Selected Issues,  
Washington D.C. 40-70.  
Revista de la Universidad Internacional del Ecuador. URL: https://www.uide.edu.ec/  
170  
INNOVA Research Journal 2018, Vol 3, No. 9, pp. 157-171  
World Development Indicators (WDI).(2014), World Bank. Available from:  
http://www.data.worldbank.org/data-catalog/world-development-indicators/wdi-2016.  
Zhang, X.P., Cheng, X.M. (2009), Energy consumption, carbon emissions, and economic growth  
in China. Ecological Economics, 68, 2706-2712  
Revista de la Universidad Internacional del Ecuador. URL: https://www.uide.edu.ec/  
171