The impact of climate change on agricultural productivity in Asian countries: a heterogeneous panel data approach

While climate change is having serious impacts on agriculture and may require ongoing adaptation, short-run threats to global food security are also crucial for developing countries. We use dynamic and asymmetric panel autoregressive distributed lag estimators to investigate how the effects of climate change on agricultural productivity vary depending upon the short run and long run in Asia over the period of 1980–2016. The results confirmed that there is a long-run relationship between agricultural productivity and climate change variables; however, only CO2 emissions could be linked to agricultural productivity in the short run. Moreover, while the direction of this effect is positive for the short run, it turns into negative in the long run confirming that carbon fertilization in the atmosphere can to some extent have a positive effect on agricultural productivity.


Introduction
Changes in the world's climate will bring major shifts in food security. While the supply of food was increasing in Asia and the Pacific, rising incomes and emerging middle class continue to drive demand for food and agricultural commodities and resources (ESCAP 2009). However, the shrinkage of agricultural areas day by day due to rapid urbanization, construction, climate, and environmental factors, such as energy consumption, mining, and urbanization, stands as an obstacle to a rapid growth in agricultural production. Climate change caused by emissions of greenhouse gases affects agricultural productivity cycles directly or indirectly through the temperature, amount of precipitation, and sunshine duration. These changes have affected the productivity pattern of agricultural products and have become a leading source of worsened food insecurity, especially in developing countries. Climate change is to have a negative effect on agricultural productivity especially in least developed countries (UNCTAD 2015). For instance, Cline (2007) examined three key factors on modeling changes in yields due to climate change for country-level impact estimates: carbon fertilization, trade and irrigation, and demonstrated strong negative impacts that climate change brings on most developing countries. According to this, by the 2080s, global agricultural productivity may fall by 15.9% if global warming progresses at its current rate emphasizing that the developing countries are most at risk. In order to ensure food security and self-sufficiency in the agricultural sector, as well as to meet the increasing demand for food despite land degradation, a lot can be done in terms of incorporating numerous strategies for sustainable development.
Climate change also leads to multiple irregular agricultural production patterns. That is, addressing the threats from current climate risks is key to understand agricultural productivity, particularly in parts of Sub-Saharan Africa, south-eastern Asia, and Western Asia (Bruinsma 2003). Since developing countries are much more vulnerable to climate change than the developed countries, the importance of agricultural sustainable development is particularly crucial particularly in South Asia and Southeast Asia. Agricultural output increased as a result of the Green Revolution, started in Asia in the 1960s under the leadership of the International Rice Research Institute (IRRI). Since then, the Asia and the Pacific region have experienced a rapid expansion and extensive structural changes with accompanying a remarkable progress in reducing food insecurity and malnutrition (FAO 2018 Asia region has made remarkable progress toward reducing food insecurity over the past quarter of a century. According to UNCTAD's Agricultural Productivity in the Least Developed Countries report, agricultural productivity in Asian least developed countries has surpassed that of African least developed countries since 2006 (UNCTAD 2015). Some authors point out that the driving forces behind agricultural productivity growth in many Asian countries are technological progress that heavily rely on research and development (R&D) in food and agriculture and improvements in human capital (Chang and Zepeda 2001;Thirtle et al. 2003). However, in comparison to the other major Asian economies, as can be seen from Fig. 1, Japan's process of agricultural productivity change appears to be a bit of an outlier. According to the OECD's report on agricultural policy developments in all countries, Japan has reduced its support to agriculture, but more recently the change in support levels has been moderate (OECD 2020). Additionally, Yamashita (2008) argues that the decline in agricultural productivity in recent years is due to the increased structural problems in the Japanese agriculture sector, stated earlier by the 1957 White Paper on Agriculture. It is also mentioned in the literature that more climate extremes tend to be associated with more resilient agricultural productivity reducing economic growth in Asia and Africa (Stern 2007;Biemans et al. 2006;FAO 2017). At the same time, since rapid economic development and heavy demand for environmental goods hinder the sustainability of natural resources, the threats to agricultural development posed by pollution and other forms of climate change and hereby the association of agricultural productivity to climate change should be addressed specifically on the basis of short and long run. So this raises crucial question and that is, to what extent are the climate change factors can affect agricultural productivity in the short run and to what extent is it due to long-run patterns in Asia. Gornall et al. (2010) point out that long-run effects of climate variability include global food production and food security as well as changes in gene expression and enzyme activity in the shorter term. Besides this, rising carbon dioxide levels cause an increase of photosynthetic rates, crop yields, and agriculture efficiency. On the other hand, the rise in global temperature due mainly to the increasing concentrations of greenhouse gases in the atmosphere is likely to reduce yields in many areas (Allen 1991;FAO 2013;Sperry et al. 2019). Consequently, since the impact of elevated CO 2 levels on agriculture is complicated, in order to determine its net effect on agricultural productivity, a number of indicators related to climate change to be examined by considering the short and long-run distinction.
Empirical studies on the impact of climate variables on value-added agriculture are limited, but essential for policymakers to adopt the policies that reduce poor farmers' vulnerability to climate change. In this context, examining the consequences of climate change on agricultural productivity under short-run and long-run distinction, especially for certain Asian countries, will contribute to the active discussion in the recent literature on climate change and its implications for agriculture. For this purpose, this study aims to analyze the effects of climate change on agricultural productivity for selected Asian countries by applying advanced panel data techniques including three different tests, namely, mean group (MG) introduced by Pesaran and Smith (1995), pooled mean group (PMG) developed by Pesaran et al. (1999), and dynamic fixed effect (DFE) estimators. The novelty of this approach is that it focuses on the short and long-run effect of three climate variables as CO 2 level, average annual temperature, and rainfall on agricultural productivity in Asian countries. Primary energy consumption per capita and total fertilizers by nutrients used in agricultural sector were also used as control variables. Furthermore, for robustness purposes, we also report the results of common correlated effect pooled mean group (CCEPMG) and common correlated effect mean group (CCEMG) estimators.
The rest of this paper is structured as follows: The "Literature review" section provides a literature review, the "Methodology, model specification, and data" section is about the methodology, model specification, and data, the "Empirical results" section provides the empirical results, the "Discussion and policy recommendations" section contains a discussion of the results and some remarks on policy recommendations, and finally, we draw some general conclusions in the "Conclusion" section.

Literature review
The importance of seeking relevance between global warming and agricultural productivity is concerned not just with the uncertainty in climate projections but also with the shifts in environmental conditions that can lead to potential risks causing further jeopardized food security in developing countries (Rosegrant et al. 2008;Khor 2009;Dudu and Cakmak 2018). More importantly, the impact of climate change on agricultural productivity indirectly causes significant changes in consumption trends through prices, such as higher animal feed costs due to drought result in higher meat prices and consequently, lower meat consumption. Hence, it is crucial for policy makers in the agriculture sector to assess the impacts that climate change will have on agricultural productivity. Essentially, the effects of climate change on agriculture can be analyzed from different perspectives within different contexts, notably "Ricardian approach" and "time series/panel data approach" (Mendelsohn 2008).
Ricardian approach focuses on the estimates of the cost of climate changes analyzing associations between land value and agro-climatic variables using net revenue climate response function under the assumption that land rent would reflect the long-run net productivity of farmland on the basis of survey or country-level data (Mendelsohn et al. 1994(Mendelsohn et al. , 1996Mendelsohn andDinar 1999, 2003;Liu et al. 2004;Gbetibouo and Hassan 2005;Schlenker et al. 2005;Seo et al. 2005;Mano and Nhemachena 2007;Deressa and Hassan 2009;Lippert et al. 2009;De Salvo et al. 2013;Closset et al. 2015;Mishra and Sahu 2014;Van Passel et al. 2017;Trinh 2018;Sadiq et al. 2019;DePaula 2020;De Siano et al. 2020;Jawid 2020;Nicita et al. 2020;Ortiz-Bobea 2020). Some of these studies based on Ricardian approach are reported in Table 1.
Time series/panel data models have become popular in recent years, as more data is available. This approach has been used to examine the association between weather and net income (Chang 2002;Deschenes and Greenstone 2007;Gay et al. 2006;Gupta et al. 2014;Sarker et al. 2012;Barnwal and Kotani 2013;Guntukula 2020;Guntukula and Goyari 2020). However, while these studies examine the issue with food production and net revenues focusing on farm net revenue, net agricultural revenue, land value, net agricultural income, yields of grain or cereal yield, and etc., only a few studies have addressed short-run and long-run effects of climate change on agricultural productivity based on advanced panel data techniques. For instance, Zaied and Cheikh (2015) analyzed the short-run and long-run association between agriculture production and climate change in Tunisia from 1979 to 2011. Using the panel data cointegration method, the study concluded that while the long-run effect of temperature on the crop production is generally negative, the effect of precipitation is positive. Besides, they concluded that an increased annual temperature decreases both cereal and date productions. Zhai et al. (2017) have assessed the wheat productivity response to climate change and technological progress on the wheat yield per unit area during 1970 to 2014 in China using ARDL model. The findings showed a long-run relationship among climate change, technical progress, and the wheat yield per unit area and positive land size impact on the per unit area wheat yield in the short run. Another empirical study has been conducted in Tunisia from the period of 1975-2014 by Attiaoui and Boufateh (2019), which investigated the long-run and short-run effects of climate change on cereal farming using pooled mean group (PMG) estimation method. Findings revealed that climate change can negatively affect cereal production, mostly due to the shortage of rainfall, whereas current temperature level has a positive impact on cereal production in Tunisia. More recently,   Chandio et al. (2020a) have analyzed the short-run and longrun impacts of climate change on agricultural output in China over the period of 1982-2014. They used annual climate change and other control variables by using the ARDL model. Findings showed that while CO 2 emissions, land area under cereal crops, fertilizer consumption, and energy consumption have a positive impact on the agricultural output in both the short run and long run, temperature and rainfall have a negative effect on agricultural output in the long run but positive in the short run. Finally, Abbas (2020) applied an autoregressive distributed lag (ARDL) model bounds testing to investigate the short-run and long-run relationships between climate change, the area under cultivation, fertilizer consumption, and cotton production in Pakistan from 1980 to 2018. According to the results, there is no evidence on increasing cotton yield through increased temperatures in the long run in Pakistan. Some other and most recent studies used panel data are reported in Table 2. This brief review of the nexus between climate change and agricultural output suggests that there is no doubt that climate change might have a possible impact on agricultural productivity. However, they do not sufficiently address or comprehensively explain climate change-related agricultural productivity on a short-run basis as well as on a long-run basis. This research aims to investigate the long-run and short-run association between climate change and agricultural productivity in the context of selected Asian countries.

Methodology, model specification, and data
This study adopts a panel data approach covering eleven Asian countries over the period 1980-2016, to examine the dynamic relationship between agricultural productivity and some primary climate change indicators. This section can be divided into two subsections. The first subsection gives a brief summary of the method and some comment on its appropriateness. The second subsection is used to describe the model specification and data-related issues 1 .

Preliminary tests
Before panel unit root tests, we conduct a cross-sectional dependence (CSD) test, introduced by Pesaran (2015), to check systematic residual correlation across different units in the panel. To investigate whether the variables are non-stationary, we then employ panel unit root tests before performing the main estimations. In the panel unit root test framework, two generations of tests have been employed: first-generation tests as the Im et al. (2003) panel unit root test (hereafter IPS) and the Levin et al. (2002) test (hereafter LLC) and the secondgeneration test of IPS test as the cross-sectional augmented IPS (hereafter CIPS) developed by Pesaran (2007). The Im-Pesaran-Shin (IPS) test is not as restrictive as the Levin-Lin-Chu test, since it allows for heterogeneous coefficients. The LLC test has a null hypothesis of the common unit root process presence, while null hypothesis for the IPS test is the presence of individual unit root process in series. If the results are statistically significant under the LLC and IPS test, that is to say that all our series are non-stationary. However, unlike LLC and IPS panel unit root tests, the CIPS panel unit root test accounts for cross-sectional dependence (Asteriou et al. 2021).

Asymmetric panel ARDL tests
Referring to the technique introduced by Pesaran et al. (1999), the dynamic heterogeneous panel regression model can be incorporated into an error correction modeling format using panel autoregressive distributed lag (ARDL) model (Samargandi et al. 2015). As Pesaran and Shin (1999) argued, panel ARDL model is applicable in a condition that independent variables are integrated of order zero or one, I(0) or I(1), respectively or a combination of both; however, the dependent variable has to be I(1). Furthermore, the potential endogeneity bias and small sample problem tends to be irrelevant and very small. Therefore, the dynamic panel model in error correction model based on panel ARDL (p,q) approach, with p as the lag of the dependent variable and q as the lag of the independent variables, is formulated accordingly as follows (Pesaran et al. 1999;Loayza and Rancière 2006): where y is the real agricultural value added in agriculture, forestry, and fishing, X is the vector set of explanatory variables including carbon dioxide (CO 2 ) levels, average annual temperature and rainfall primary energy consumption per capita, and total fertilizers by nutrients used in agricultural sector, γ and δ are the short-run dynamic coefficients related to dependent variable and its determinants respectively, β represents the long-run coefficients, and φ shows the speed of adjustment in the long run. Lastly, i and t represent the country and time respectively. The bracketed values represent the long-run regression as follows: There are different types of methods to estimate the above model. According to Pesaran et al. (1999), the autoregressive distributed lag (ARDL, p,q) model includes the mean group (MG) and pooled mean group (PMG) estimators as well as the dynamic fixed effect (DFE) model. The MG estimation includes separate regressions for each country with all the coefficients to vary and being heterogeneous in the short and long run. That is, the MG model does not impose any restrictions on the cross-sectional parameters and hence, ignores any possible homogeneity of some parameters across countries. The PMG model allows the short-run coefficients, error variances, and the regression intercept to be heterogeneous, but constraints the long-run estimates to vary across cross sections, which is a prominent difference between the PMG and MG technique. Finally, the dynamic fixed effect (DFE) model presumes all slope coefficients (both short run and long run) to be homogeneous across countries allowing panel-specific intercepts except the constant term (intercept). However, Baltagi et al. (2000) clarify that this model is subject to potentially inconsistent and misleading estimates caused by the endogeneity existing between the lagged dependent variable and error term. To find the efficient model to provide reliable results, the Hausman h-test based on panel ARDL approach that measures the efficiency and consistency has been used to analyze whether there are significant differences among the PMG, MG, and DFE. Under the null hypothesis of log-run homogeneity, the difference between PMG and MG or PMG and DFE estimation is not significant; that is, the efficient estimator under the null hypothesis is PMG. Otherwise, if the null hypothesis is rejected, then the efficient estimator, MG or DFE, is preferred respectively. As a test for robustness, the dynamic common correlated effects mean group (CCEMG) and the dynamic common correlated effects pooled mean group (CCEPMG) estimators, introduced by Pesaran (2006), would be conducted to affirm the findings of PMG estimator. Both estimators are robust to CSD as they consider heterogeneous panels with cross-sectional dependence by using a common factor structure in the error terms. The dynamic CCEMG and CCEPMG estimators account for heterogeneous time effects and deal with crosssectional dependencies.

Model specification and data description
This research is intended to estimate short-run and long-run relationship between agricultural productivity and climate change variables without being able to observe the short-run and long-run relevant components of variables employed. Agriculture, forestry, and fishing value added per worker series were used as the proxy for agricultural productivity, the ratio between value added in agriculture and number of workers employed in agricultural sector, as identified in the "World Bank Data Catalog". Based on Climate Change Indicators in the United States Report by the US Environmental Protection Agency (2016), average annual temperature, average annual rainfall, and CO 2 emission levels were used in the model as a proxy for climate change indicators. Besides, primary energy consumption per capita and total fertilizers by nutrients used in agricultural sector were also added as control variables. To obtain evidence on the potential impact of climate change on agricultural productivity, the current study incorporates an interaction between agriculture, forestry, and fishing value added per unit of input and relevant climate change indicators and control variables as follows: where AGR is the agriculture, forestry, and fishing, value added (constant US$) as a proxy for agricultural productivity; CO 2 is the CO 2 emissions (metric tons per capita); Temp is the temperature as measured°C; RF is the average annual rainfall (mm per year); EC represents the primary energy consumption (gigajoules per capita); and finally, FE represents the total fertilizers by nutrients used in the agricultural sector measured as tones.

Empirical results
Before going into the main estimation of the panel ARDL estimators, slope homogeneity test (Pesaran and Yamagata 2008) and panel unit root tests were conducted to verify the heterogeneity and the stationarity of the variables used as in the literature due to their superiority to time series unit root tests. The slope homogeneity test results showed that the slope coefficients are found to be heterogeneous 2 . On the other hand, although ARDL is applicable for variables with a mixture of integration of I(0) and of I(1), unit root tests still need to be undertaken to ensure that the variables are not I(2) so as to avoid spurious results. The results of the unit root tests are presented in Table 3 showing mix results regarding the existence of unit root in their levels for the variables that are employed in the model. Since all the variables are stationary in their first differences, the common components of variables all turn out to be integrated of order one, or I(1) 3 . The results of the second-generation panel unit root test (CIPS) results are also presented in Table 3 . Accordingly, except lnTEMP lnRF and lnFE, the series are not stationary in level, but stationary at level I(1). Panel cointegration test, developed by Westerlund (2007), was also conducted to test the existence of long-run relationships between integrated series in both time series dimension (T) and cross-sectional dimension (N). Accordingly, four new panel cointegration tests for the null hypothesis of no cointegration developed as G t , G a , P t , and P a based on the error correction model (ECM), allowing for a large degree of heterogeneity between the cross-sectional units and can account for cross-sectional dependence via bootstraps 4 . As suggested by Westerlund (2007), the probability values obtained from the "bootstrap" distribution were obtained to take into account the cross-sectional dependency. Bootstrap resampling procedures are employed at 500 estimations for each Westerlund panel cointegration test and provide us with robust p-values. The results are summarized in Table 4 and in all cases there is no evidence of cointegration. Table 5 presents the results of baseline estimates obtained from MG, PMG, and DFE estimators for a linear specification of the effects of climate change on agricultural productivity. The Hausman test was used to test the null hypothesis of homogeneity restriction on the long-run coefficients based on the comparison between the pooled mean group and the mean group estimators, as the respective Hausman h-test pvalues of 8.61 and 0.34 for MG and DFE are both insignificant. This suggests that the PMG estimators are consistent and more efficient than MG and DFE. Summarizing the findings from Table 5, the PMG results can be interpreted as follows: CO 2 emission The estimated results display that CO 2 emissions have a positive short-run impact on the agricultural productivity. However, this effect turns into negative in the long run; that is, the results indicate that the coefficient of CO 2 emission level is significant and negative in the long run, indicating a negative impact of carbon emissions on agricultural productivity. Specifically, a 1% increase in CO 2 emission level would lead to a reduction in agricultural productivity by 1.94%. Our findings on positive effect of carbon emission on agricultural productivity in the short run but negative in the long run are partially consistent with some studies, which have found same effects for both periods, such as Chandio et al. (2020aChandio et al. ( , 2020bChandio et al. ( , 2021b and Rehman et al. (2021).
Temperature Results from short-run dynamics indicate that 1% increase in temperature would cause to lower productivity by 2.28% and vice versa. Temperature was also found to produce an insignificant positive impact on agricultural productivity in the long run. The results are consistent with Zaied and 4 For more details, please see Westerlund (2007).   Cheikh (2015), Attiaoui and Boufateh (2019), and Chandio et al. (2020aand Chandio et al. ( , 2020c. Annual rainfall The results reveal that the annual rainfall is insignificant in influencing agricultural productivity in the short run, but turns into positive in the long run, with a coefficient of 0.95, which means a 1% increase in annual rainfall will boost the agricultural productivity of about 0.95%. The findings coincide with Attiaoui and Boufateh (2019) and Zaied and Cheikh (2015) for both periods, but coincide with Abbas (2020) for only the short run.
Energy consumption As a control variable, the impact of energy consumption on agricultural productivity is negative in the short run, and these effects turned out to be positive, consistent with the study of Chandio et al. (2020aChandio et al. ( , 2020b. Total fertilizers As a control variable, total fertilizer consumption has a positive impact on agricultural productivity in the long run, similarly with Chandio et al. (2020a).
Overall, the results of the CCEMG and CCEPMG estimations in Table 6 are in line with those obtained in Table X. According to CCEMG and CCEPMG model results, CO 2 emissions positively affect agricultural productivity in the short run, while this effect turns negative in the long run. Similarly, the findings on the negative long-run effect of temperature, positive long-run effect primary energy consumption, and total fertilizers are consistent with the results obtained through using PMG estimator.

Discussion and policy recommendations
Since climate change is a primary determinant for agricultural productivity on a sustainable basis, creating and managing sustainability performance in agricultural productivity to satisfy increased demands in agriculture and increase food availabilities is a major task for the global agricultural sector. Climate change might be a major concern to sustainability since it is having a measurable effect on agricultural productivity, especially in rain-fed farming areas. The effects of climate change such as changes in precipitation regime, temperature increase, drought, and natural disasters lead to threats to productivity and growth rates in agriculture. However, conditions such as the level of nutrients in the soil and the amount of water must also be met in order to increase the yield. According to this, higher concentrations of atmospheric carbon dioxide and temperature boost crop yields by increasing the rate of photosynthesis, which in turn contributes to higher growth and productivity (Rosenberg 1982;Kimball and Idso 1983). But some research had found that CO 2 decreases the plant's concentrations of other internal compounds like vitamin B, protein, and micronutrients (Ebi and Ziska 2018). Note: *** , ** , and * show 1%, 5%, and 34% of significant levels Rain-fed agriculture plays a critical role in food production, which covers 80% of the world's cultivated land, and is responsible for about 60% of crop production (UNESCO, UN-Water 2020). The historic records show clear long-run warming trends across the world since the late nineteenth century (Hartmann et al. 2013). Changes in the mean level of temperature and rainfall may lead to stronger droughts, which adversely affect livestock and rain-fed crops (Verner et al. 2018). On the other hand, it is expected that the carbon dioxide concentration accumulated in the atmosphere will have a positive contribution to the growth of certain agricultural products (Warrick 1988). Overall, climate change makes agricultural productivity more vulnerable to climate impacts and could make it more difficult to grow crops and raise animals. It is certain that changes in rainfall levels due to climate change will have effects on agriculture. Soil moisture level should be ideal and agricultural area should be accessible to water.
For this purpose, many studies highlighted the possible threat of negative climate change impacts on agriculture. However, while most of the existing studies on climate change and agricultural productivity nexus are country-specific or region-specific studies, to the best of our knowledge, this study is the first to show the possible association between agriculture value added per worker and the three main climate change indicators as CO 2 emissions, average annual temperature, and rainfall by using pooled cross-sectional data for Asian countries. The main finding we draw from this research, the contradiction of short-run and long-run effects of CO 2 emissions on agricultural productivity, is supported by the study of Goudriaan and Unsworth (1990), suggested that while increasing concentrations of atmospheric CO 2 promote plant growth and agricultural productivity without increasing the water demand for crop transpiration, global warming and climate change vulnerability may tend to reverse positive direct CO 2 effects in the long run. In addition, since agricultural productivity is also affected by long-run trends in average rainfall and temperature, changes in the frequency and intensity of precipitation are also vitally important.
In conclusion, agriculture helps slow climate change by enhancing carbon storage in soils; preserving existing soil carbon; and reducing carbon dioxide, while continuing to remove CO 2 from the air. However, in the long run, it is obvious that climate change is a serious threat to agriculture and to food security. Therefore, since declining agricultural productivity could compound the risk of food insecurity, there is an urgent need to cut risk of long-run effects of climate change. Many options can be carried out at farm level; for instance, the implementation of adaptation and mitigation strategies at farm level, promoting farmers' practices, can be a powerful strategy to minimize the likely dramatic impacts of climate change and also to preserve nature and prevent global warming from becoming worse.

Conclusion
Agricultural vulnerability to climate change depends not only on acceptable temperature ranges and patterns of rainfall, but also on concentration of carbon dioxide in the atmosphere. Indeed, climate change poses a challenge to agricultural productivity and its effects vary according to the regional risks and adaptation and type of production system as well as depending on the dynamic characteristics of climatic indicators. However, the uncertainties about climate impacts appear to have both short-run and long-run components. This research contributes to the agricultural productivity-climate change literature by focusing on a selection of Asian countries which adopted the "Green Revolution" model in the 1960s by introducing high-yielding varieties of food through the intensification of the arable lands through massive investments in irrigation. The aim and novelty of this study is to investigate how the effects of climate change on agricultural productivity vary depending upon the short run and long run in Asia over the period of 1980-2016. Overall, the results confirmed that there is a long-run relationship between agricultural productivity and climate change variables; but more importantly, the short-run positive relationship between carbon emissions and agricultural productivity contradicts with the long-run effect. That is, while the impact of carbon emission on agricultural productivity is positive in the short run, it turns into Note: *** , ** , * shows 1%, 5% and 34% of significant levels negative in the long run, confirming that carbon fertilization in the atmosphere can to some extent have a positive effect on agricultural productivity. In other words, the findings of this study are in line with the argument that rising carbon emission levels benefit agriculture in the short run, stimulating climate change's negative effects on agricultural production in the short run. Nevertheless, it is evident that both carbon emissions and annual temperature have an adverse effect on agricultural productivity in the long run, which has implications on food security for countries; policy interventions that seek to overcome global warming in agriculture sector need to be based on sustainable land management to prevent land degradation and to improve land use management, climate-friendly agricultural practices to increase the capacity to adapt to climate change, and, most importantly, the adaptation and dissemination of low carbon emission technologies within the framework of efficient use of agricultural and forest areas. These findings can form a base for further research such as productivity responses to nitrogen, carbon dioxide, and temperature interactions.
Data availability The datasets used and/or analyzed during the current study are available from the FAOSTAT database, the World Bank Group Climate Change Portal, the BP Statistical Review of World Energy, and the World Bank national accounts data.

Declarations
Ethical approval Not applicable.
Consent to participate Not applicable.

Consent for publication Not applicable.
Competing interests The author declares no competing interests.