Food security and agriculture value-added: Do they asymmetrically matter for Korean environmental sustainability?

Abstract

Limited research has previously probed the interplay between Korea’s food security, agricultural sector, and environmental ramifications. This study delves into the asymmetric influence of food security and agricultural value-added on Korea’s carbon dioxide emissions (an environmental sustainability proxy) from 1970 to 2020, incorporating non-renewable energy consumption and foreign direct investment. Employing a nonlinear auto-regressive distributed lag methodology, our findings unveil long-term asymmetric effects of food security and agricultural value-added on carbon dioxide emissions. Specifically, carbon dioxide emissions rise with positive shocks to food security and agricultural value-added but decline with negative shocks. Moreover, our analysis reveals that non-renewable energy consumption positively impacts carbon dioxide emissions, whereas foreign direct investment exerts a negative effect. These insights may underpin policy recommendations for the Korean government to bolster environmental sustainability.

Keywords:

• Food security
• agriculture value-added
• environmental sustainability
• nonrenewable energy consumption
• foreign direct investment
• nonlinear auto-regressive distributed lag methodology

1. Introduction

Korea’s economy is burgeoning, with escalating industrialization and urbanization driving substantial growth in energy consumption and carbon dioxide emissions, posing challenges to environmental sustainability. As per International Energy Agency data, South Korea ranks among the top global carbon dioxide emitters, generating 586 million metric tons in 2019. Though emissions decreased in 2018 and 2019, Korea remains among the top 10 emitters worldwide. Several factors contribute to Korea’s rising carbon dioxide emissions. Primarily, the nation’s heavy reliance on fossil fuels exacerbates its carbon emissions footprint (Pata & Kartal, 2023). Additionally, Korea’s transportation sector contributes significantly to overall emissions, with a marked increase in private vehicles and a dense road infrastructure (Lu et al., 2007). Industrialization also plays a role, as major steel, chemical, and power industries emit considerable amounts of carbon dioxide (Song et al., 2022). Lastly, insufficient government initiatives contribute to growing emissions (N. Zhang et al., 2013). While the Korean government has implemented a carbon emissions trading system and increased renewable energy usage, the effects remain unclear. To curb emissions, the government must enforce stricter regulations, improve carbon trading efficacy, and incentivize businesses and individuals to adopt emission reduction measures.

Building upon the aforementioned research context, this study aims to investigate the impact of food security and agricultural value-added on carbon dioxide emissions (a proxy for environmental sustainability) in Korea, factoring in non-renewable energy consumption and foreign direct investment. Utilizing a nonlinear autoregressive distributed lag approach, we conducted an empirical analysis based on time series data spanning 1970–2020. Our findings reveal a long-term asymmetric relationship between food security, agricultural value-added, and carbon dioxide emissions. Specifically, positive shocks in food security and agricultural value-added correspond to increased emissions, while negative shocks result in decreased emissions. Additionally, our analysis indicates that non-renewable energy consumption positively affects carbon dioxide emissions, whereas foreign direct investment exerts a negative influence.

Diverging from existing literature, this study emphasizes environmental sustainability and asymmetry, shedding light on both the positive and negative ramifications of food security and agriculture on carbon dioxide emissions, thus enriching the knowledge base in these domains. Our examination of positive and negative shocks to food security and agricultural value-added already presents a valuable contribution to environmental research. Moreover, employing a nonlinear autoregressive distributed lag methodology, we scrutinize the long- and short-term associations and asymmetries between agriculture, fossil fuel usage, food security, and carbon dioxide emissions in Korea from 1970 to 2020. To the best of our knowledge, this is the first instance in Korean literature where a single variable in a specific period has been considered in this context. Prior studies predominantly investigated shocks to all independent variables, potentially obfuscating the performance of individual variables.

The remainder of this article is broken down into the following sections: In Section 2, a literature review is offered. Section 3 provides a description of the variables as well as an econometric technique. Section 4 presents the findings along with the discussion. Section 5 lays out the conclusion that was reached.

2. Literature review

As global food demand surges, agricultural carbon emissions present a formidable challenge in combating climate change. However, there is a paucity of methodical and insightful empirical literature on food security and carbon emissions in Korea, as well as a gap in examining optimal trajectories for balancing food security and emission reduction. Edoja et al (Edoja et al., 2016), utilizing annual time series data from 1961 to 2010, probed the dynamic relationship between carbon dioxide emissions and food security in Nigeria, finding no long-term association but identifying a significant, negative short-term linkage using vector auto-regressive estimates and impulse response functions. Naseem et al (Naseem, Guang Ji, et al., 2020) analyzed the asymmetric impact of energy consumption and food security on carbon emissions in Pakistan from 1970 to 2019, employing an asymmetric autoregressive distributed lag approach. Employing population data as a proxy for food security, they observed environmental degradation driven by food security and energy consumption. Ntiamoah et al (Ntiamoah et al., 2023) investigated the connection between carbon dioxide emissions and food security in East Africa from 1990 to 2020. Applying fully modified ordinary least squares and dynamic ordinary least squares methods, they revealed a long-term equilibrium relationship, with increased food security leading to elevated carbon dioxide emissions. This finding is supported by Surahman et al (Surahman et al., 2018), Lee et al (S. -H. Lee et al., 2018), Ziska et al (Ziska et al., 2012), Zhang et al (T. Zhang et al., 2023), and Ziska and Bunce (Ziska & Bunce, 2007).

In the contemporary context of global warming and the imperative to regulate greenhouse gas emissions, fostering low-carbon agriculture is essential for achieving ecologically harmonious development, economic growth, and sustainable agricultural advancement. Agricultural production efficiency is the principal determinant of agricultural carbon emission intensity. Employing Chinese provincial panel data from 2010–2019, Zhu and Huo (Zhu & Huo, 2022) devised an indicator set and utilized the super-efficient slacks-based measure model to appraise agricultural output efficiency. Estimating regional agricultural carbon emissions through agricultural production activities, they identified a U-shaped relationship between agricultural production efficiency and carbon emission intensity. Specifically, in regions with high efficiency, enhanced production efficiency curtailed emission intensity, while in low-efficiency regions, it intensified emissions. Balsalobre-Lorente et al (Balsalobre-Lorente et al., 2019) examined the environmental Kuznets curve hypothesis in BRICS nations (Brazil, Russia, India, China, South Africa) from 1990 to 2014, revealing agriculture’s detrimental impact on environmental sustainability via dynamic constant least squares and fully modified ordinary least squares techniques for long-term regressions. The G7 countries’ elevated greenhouse gas emissions pose a significant global environmental threat. Wang et al (L. Wang et al., 2020) scrutinized the effect of value-added agriculture on the relationship between carbon dioxide emissions and agriculture using 1996–2017 data. Employing innovative econometric methods, such as the cross-sectional augmented autoregressive distributed lag approach, they ascertained that increased value-added agriculture led to reduced carbon emissions. Comparable findings were reported by Wang et al (R. Wang et al., 2022), Yasmeen et al (Yasmeen et al., 2022), Zhang et al (L. Zhang et al., 2019), Guo et al (Guo et al., 2022), and Koondhar et al (Koondhar et al., 2021).

Furthermore, non-renewable energy consumption and foreign direct investment represent substantial influences on carbon dioxide emissions. Amin and Song (Amin & Song, 2022) employed a comparative approach to assess the connection between non-renewable energy usage and carbon dioxide emissions in South and East Asian countries from 2000 to 2018. Utilizing the dynamic co-integration technique, they investigated short- and long-term linkages and identified that non-renewable energy consumption contributed to increased long-term carbon dioxide emissions in South Asia. Zhang et al (Y. Zhang et al., 2023) examined the relationship between non-renewable energy consumption and carbon dioxide emissions in China, India, Bangladesh, Japan, South Korea, and Singapore from 1975 to 2020, uncovering a considerable increase in long-term carbon emissions due to non-renewable energy utilization. Apergis et al (Apergis et al., 2023) explored the correlation between carbon dioxide emissions and both renewable and non-renewable energy consumption in Uzbekistan from 1985 to 2020. Employing an autoregressive distributed lag model, they revealed that gas and oil energy consumption positively affected per capita carbon dioxide emissions in short and long terms, while coal usage showed a short-term positive and long-term negative impact. Regarding foreign direct investment’s effect on carbon dioxide emissions, Wang and He (Y. Wang & He, 2023) assessed its influence on China’s energy transition and environmental quality using 2000–2021 annual data and an autoregressive distributed lag model, discovering that foreign direct investment increased carbon dioxide emissions. Boamah et al (Boamah et al., 2023). gathered data from 41 African nations between 2005 and 2019 to determine the impact of foreign direct investment and other factors on carbon dioxide emissions. Applying pooled least squares, fixed and random effects models, and the generalized method of moments, they identified that foreign direct investment inflows led to increased carbon dioxide emissions in the studied African countries. These findings are supported by Lee et al (C. -C. Lee et al., 2023), Zhang et al (Z. Zhang et al., 2023), Kheir (Kheir, 2023), Duan et al (Duan et al., 2022), and Boohene and Darkwah (Boohene & Darkwah, 2023).

3. Variable description and econometric approach

3.1. Variable description

This study endeavors to explore the consequences of food security and agriculture on environmental degradation. As indicated by Adeleye et al (Adeleye et al., 2021). and Alhassan (Alhassan, 2021), carbon dioxide emissions serve as a proxy for environmental sustainability. Presently, various scholars utilize diverse proxies for food security. For instance, Subramaniam and Masron (Subramaniam & Masron, 2021), and Subramaniam et al (Subramaniam et al., 2020) employed the food security index as a proxy. Qadir et al (Qadir et al., 2018) used a comprehensive index encompassing socioeconomic status, while Russell et al (Russell et al., 2018), Nicholson et al (Nicholson et al., 2021), and Molotoks et al (Molotoks et al., 2021) utilized population as a proxy. This is due to the fact that food security encompasses the economic and physical accessibility of a nutritious diet meeting an individual’s dietary needs for a prolonged, healthy life. Moreover, in line with Ridzuan et al (Ridzuan et al., 2020), Cai et al (Cai et al., 2022), and Faling (Faling, 2020), agriculture value-added serves as a proxy for agriculture. This article also incorporates several control variables. Following Ren et al (Ren et al., 2020), Zhukovskiy et al (Zhukovskiy et al., 2021), and Chen et al (Chen et al., 2021), fossil fuel energy consumption is included. Additionally, Sabir et al (Sabir et al., 2020), Abdo et al (Abdo et al., 2020), and Baskurt et al (Baskurt et al., 2022) inform the inclusion of foreign direct investment. Table 1 encapsulates the details of all variables examined within this article.

Table 1. Results of variable description

Variable	Abbreviation	Definition	Source
Carbon dioxide emissions	car	Emissions of carbon dioxide (unit: Mt) in log	UN environment programme
Food security^1	foo	Total population (unit: million) in log	World Bank Data
Agriculture	agr	Agriculture value-added (unit: billion US$)	Food and Agriculture Organization of the United Nations
Non-renewable energy consumption	non	Fossil fuel energy consumption (% of total)	World Bank Data
Foreign direct investment	fdi	Foreign direct investment, net inflows (% of GDP)	World Bank Data

Note: 1 the reason is that food security is the economic and physical availability of a nutritious meal that satisfies a person’s nutritional requirements for a long and healthy life.

3.2. Econometric approach

Drawing inspiration from Ozatac et al (Ozatac et al., 2017), Amri (Amri, 2017), and Lin and Moubarak (Lin & Moubarak, 2014), Equation (1) has been formulated to examine the long-term impacts of food security and agriculture on carbon dioxide emissions, which serve as a proxy for environmental sustainability. The expression for Equation (1) is presented as follows:

(1) $\mathrm{c}\mathrm{a}{\mathrm{r}}_{\mathrm{t}}={\alpha }_0+{\alpha }_1\mathrm{f}\mathrm{o}{\mathrm{o}}_{\mathrm{t}}+{\alpha }_2\mathrm{a}\mathrm{g}{\mathrm{r}}_{\mathrm{t}}+{\alpha }_3\mathrm{n}\mathrm{o}{\mathrm{n}}_{\mathrm{t}}+{\alpha }_4\mathrm{f}\mathrm{d}{\mathrm{i}}_{\mathrm{t}}+{ϵ}_{\mathrm{t}},$(1)

where the time variable is denoted by $\mathrm{t}$, while ${\alpha }_0$ represents the constant and [${\alpha }_1$, ${\alpha }_4$] signify the coefficients to be estimated; ${ϵ}_{\mathrm{t}}$ stands for the white noise. As suggested by Sam et al (Sam et al., 2019), Panopoulou and Pittis (Panopoulou & Pittis, 2004), and Kapetanios et al (Kapetanios et al., 2006), the asymmetric autoregressive distributed lag approach is employed to discern the long- and short-term effects of food security and agriculture on carbon dioxide emissions, utilizing the decomposition of negative and positive partial sums. The nonlinear autoregressive distributed lag methodology boasts several advantages over conventional cointegration approaches. According to Kapetanios et al (Kapetanios et al., 2006) and Kumar et al (Kumar et al., 2020), the first advantage is its efficacy even with relatively small sample sizes. Alsamara et al (Alsamara et al., 2020) and Kouton (Kouton, 2019) highlight the second advantage, which is the absence of a requirement for a stationarity test in nonlinear autoregressive distributed lag estimation. Ali et al (Ali et al., 2020) and Zafar et al (Zafar et al., 2023) emphasize the third advantage, which is the effectiveness of the nonlinear autoregressive distributed lag estimation when cointegration is present under the conditions of I(0), I(1), or a mixture of I(0) and I(1). However, He et al (He et al., 2022). and He (He, 2022) noted that the nonlinear autoregressive distributed lag estimation would be ineffective if any variables exhibit stationarity at I(2). Consequently, this study utilizes the nonlinear autoregressive distributed lag approach due to its ability to examine both the short- and long-term asymmetric impacts of food security and agriculture on carbon dioxide emissions. The nonlinear autoregressive distributed lag bound test, as proposed by Shin et al (Shin et al., 2014), is used to explore both long- and short-term relationships, while Equation (1) solely generates long-term implications for a specified model. To incorporate both error correction terms and short-term relationships, Equation (1) must be modified. Following Pesaran et al (Pesaran et al., 2001), the model encompassing error correction terms and short-term connections is presented as follows:

(2) $\mathrm{\Delta }{\mathrm{c}\mathrm{a}\mathrm{r}}_{\mathrm{t}}={\beta }_0+{\beta }_1{\mathrm{c}\mathrm{a}\mathrm{r}}_{\mathrm{t}-1}+{\beta }_2{\mathrm{f}\mathrm{o}\mathrm{o}}_{\mathrm{t}-1}+{\beta }_3{\mathrm{a}\mathrm{g}\mathrm{r}}_{\mathrm{t}-1}+{\beta }_4{\mathrm{n}\mathrm{o}\mathrm{n}}_{\mathrm{t}-1}+{\beta }_5{\mathrm{f}\mathrm{d}\mathrm{i}}_{\mathrm{t}-1}+{\beta }_6{\mathrm{d}\mathrm{u}\mathrm{m}}_{\mathrm{t}-1}+\lambda {\mathrm{e}\mathrm{c}\mathrm{m}}_{\mathrm{t}-\mathrm{i}}+{\beta }_{1\mathrm{i}}\sum _{\mathrm{k}=\mathrm{i}}^{\mathrm{n}1}{\mathrm{\Delta }\mathrm{c}\mathrm{a}\mathrm{r}}_{\mathrm{t}-\mathrm{i}}+{\beta }_{2\mathrm{i}}\sum _{\mathrm{k}=1}^{\mathrm{i}}{\mathrm{\Delta }\mathrm{f}\mathrm{o}\mathrm{o}}_{\mathrm{t}-\mathrm{i}}+{\beta }_{3\mathrm{i}}\sum _{\mathrm{k}=1}^{\mathrm{i}}{\mathrm{\Delta }\mathrm{a}\mathrm{g}\mathrm{r}}_{\mathrm{t}-\mathrm{i}}+{\beta }_{4\mathrm{i}}\sum _{\mathrm{k}=1}^{\mathrm{i}}{\mathrm{\Delta }\mathrm{n}\mathrm{o}\mathrm{n}}_{\mathrm{t}-\mathrm{i}}+{\beta }_{5\mathrm{i}}\sum _{\mathrm{k}=1}^{\mathrm{i}}{\mathrm{\Delta }\mathrm{f}\mathrm{d}\mathrm{i}}_{\mathrm{t}-\mathrm{i}}+{\beta }_{6\mathrm{i}}\sum _{\mathrm{k}=1}^{\mathrm{i}}{\mathrm{d}\mathrm{u}\mathrm{m}}_{\mathrm{t}-\mathrm{i}}+{ϵ}_{\mathrm{t}},$(2)

where ${\beta }_0$ represents the constant, while [${\beta }_1$, ${\beta }_{6\mathrm{i}}$] signify the coefficients to be estimated; $\mathrm{\Delta }$ stands for the first-order difference, ${ϵ}_{\mathrm{t}}$ denotes the white noise, $\mathrm{e}\mathrm{c}\mathrm{m}$ symbolizes the error correction term, and $\mathrm{d}\mathrm{u}\mathrm{m}$ is a dummy variable. The approach developed by Engle and Granger (Engle & Granger, 1987) aligns with Equation (2), achieved by substituting a proxy for the lag of the error correction term in Equation (1) with a linear combination of these variables’ lags. Notably, Equation (2) enables the examination of both long- and short-term impacts, an enhancement over Engle and Granger’s (Engle & Granger, 1987) presentation. In Equation (2), [${\beta }_{1\mathrm{i}}$, ${\beta }_{6\mathrm{i}}$] imply the long-term relationship among the investigated variables, while [${\beta }_{1\mathrm{i}}$, ${\beta }_{6\mathrm{i}}$] indicate the short-term relationship. Furthermore, to validate the long-term relationship between these variables, it is crucial to establish long-run causality. Following Pesaran et al (Pesaran et al., 2001), the F-statistics bound test can be employed to confirm the cointegration between these investigated variables. Although Equation (2) presumes symmetrical impacts of the investigated independent variables on carbon dioxide emissions, this study aims to explore the asymmetric influences of food security and agriculture on carbon dioxide emissions in Korea. To assess the asymmetric effects on carbon dioxide emissions, the investigated variables are decomposed into positive and negative components. The asymmetric model is presented as follows:

(3) ${\mathrm{x}}_{\mathrm{t}}={\varphi }^{+}{\mathrm{y}}_{\mathrm{t}}^{+}+{\varphi }^{-}{\mathrm{y}}_{\mathrm{t}}^{-}+{ϵ}_{\mathrm{t}}$(3)

where $\mathrm{y}$ denotes $\mathrm{f}\mathrm{o}\mathrm{o}$, $\mathrm{a}\mathrm{g}\mathrm{r}$, $\mathrm{n}\mathrm{o}\mathrm{n}$, and $\mathrm{f}\mathrm{d}\mathrm{i}$; ${\varphi }^{+}$ and ${\varphi }^{-}$ denote the long-run coefficients. Therefore, $\mathrm{y}$ can be shown as follows:

(4) ${\mathrm{y}}_{\mathrm{t}}={\mathrm{y}}_0+{\mathrm{y}}_{\mathrm{t}}^{+}+{\mathrm{y}}_{\mathrm{t}}^{-}$(4)

where ${\mathrm{y}}_{\mathrm{t}}^{+}$ denotes the positive change in partial sums; ${\mathrm{y}}_{\mathrm{t}}^{-}$ denotes the negative change in partial sums. Consequently, the positive and negative shifts in partial sums for both food security and agriculture are illustrated as follows:

(5) $\mathrm{f}\mathrm{o}{o}^{+}={\sum }_{\mathrm{i}=1}^{\mathrm{t}}{\mathrm{\Delta }\mathrm{f}\mathrm{o}\mathrm{o}}_{\mathrm{i}}^{+}={\sum }_{\mathrm{i}=1}^{\mathrm{t}}\mathrm{m}\mathrm{a}\mathrm{x}\left(\mathrm{\Delta }\mathrm{f}\mathrm{o}{\mathrm{o}}_{\mathrm{i}},0\right)$(5)

(6) $\mathrm{f}\mathrm{o}{o}^{-}={\sum }_{\mathrm{i}=1}^{\mathrm{t}}{\mathrm{\Delta }\mathrm{f}\mathrm{o}\mathrm{o}}_{\mathrm{i}}^{-}={\sum }_{\mathrm{i}=1}^{\mathrm{t}}\mathrm{m}\mathrm{i}\mathrm{n}\left(\mathrm{\Delta }\mathrm{f}\mathrm{o}{\mathrm{o}}_{\mathrm{i}},0\right)$(6)

(7) $\mathrm{a}\mathrm{g}{r}^{+}={\sum }_{\mathrm{i}=1}^{\mathrm{t}}{\mathrm{\Delta }\mathrm{a}\mathrm{g}\mathrm{r}}_{\mathrm{i}}^{+}={\sum }_{\mathrm{i}=1}^{\mathrm{t}}\mathrm{m}\mathrm{a}\mathrm{x}\left(\mathrm{\Delta }\mathrm{a}\mathrm{g}{\mathrm{r}}_{\mathrm{i}},0\right)$(7)

(8) $\mathrm{a}\mathrm{g}{r}^{-}={\sum }_{\mathrm{i}=1}^{\mathrm{t}}{\mathrm{\Delta }\mathrm{a}\mathrm{g}\mathrm{r}}_{\mathrm{i}}^{-}={\sum }_{\mathrm{i}=1}^{\mathrm{t}}\mathrm{m}\mathrm{i}\mathrm{n}\left(\mathrm{\Delta }\mathrm{a}\mathrm{g}{\mathrm{r}}_{\mathrm{i}},0\right)$(8)

Integrating equations (2), (5), (6), (7), and (8), we present the nonlinear autoregressive distributed lag model as follows:

(9) $\mathrm{\Delta }{\mathrm{c}\mathrm{a}\mathrm{r}}_{\mathrm{t}}={\gamma }_0+{\gamma }_1{\mathrm{c}\mathrm{a}\mathrm{r}}_{\mathrm{t}-1}+{\gamma }_2{\mathrm{f}\mathrm{o}\mathrm{o}}_{\mathrm{t}-1}^{+}+{\gamma }_3{\mathrm{f}\mathrm{o}\mathrm{o}}_{\mathrm{t}-1}^{-}+{\gamma }_4{\mathrm{a}\mathrm{g}\mathrm{r}}_{\mathrm{t}-1}^{+}+{\gamma }_5{\mathrm{a}\mathrm{g}\mathrm{r}}_{\mathrm{t}-1}^{-}+{\gamma }_6{\mathrm{n}\mathrm{o}\mathrm{n}}_{\mathrm{t}-1}+{\gamma }_7{\mathrm{f}\mathrm{d}\mathrm{i}}_{\mathrm{t}-1}+{\gamma }_8{\mathrm{d}\mathrm{u}\mathrm{m}}_{\mathrm{t}-1}+\lambda {\mathrm{e}\mathrm{c}\mathrm{m}}_{\mathrm{t}-\mathrm{i}}+{\gamma }_{1\mathrm{i}}\sum _{\mathrm{k}=\mathrm{i}}^{\mathrm{n}1}{\mathrm{\Delta }\mathrm{c}\mathrm{a}\mathrm{r}}_{\mathrm{t}-\mathrm{i}}+{\gamma }_{2\mathrm{i}}\sum _{\mathrm{k}=1}^{\mathrm{i}}\mathrm{\Delta }{\mathrm{f}\mathrm{o}\mathrm{o}}_{\mathrm{t}-\mathrm{i}}^{+}+{\gamma }_{3\mathrm{i}}\sum _{\mathrm{k}=1}^{\mathrm{i}}\mathrm{\Delta }{\mathrm{f}\mathrm{o}\mathrm{o}}_{\mathrm{t}-\mathrm{i}}^{-}+{\gamma }_{4\mathrm{i}}\sum _{\mathrm{k}=1}^{\mathrm{i}}\mathrm{\Delta }{\mathrm{a}\mathrm{g}\mathrm{r}}_{\mathrm{t}-\mathrm{i}}^{+}+{\gamma }_{5\mathrm{i}}\sum _{\mathrm{k}=1}^{\mathrm{i}}\mathrm{\Delta }{\mathrm{a}\mathrm{g}\mathrm{r}}_{\mathrm{t}-\mathrm{i}}^{-}+{\gamma }_{6\mathrm{i}}\sum _{\mathrm{k}=1}^{\mathrm{i}}{\mathrm{\Delta }\mathrm{n}\mathrm{o}\mathrm{n}}_{\mathrm{t}-\mathrm{i}}+{\gamma }_{7\mathrm{i}}\sum _{\mathrm{k}=1}^{\mathrm{i}}{\mathrm{\Delta }\mathrm{f}\mathrm{d}\mathrm{i}}_{\mathrm{t}-\mathrm{i}}+{\gamma }_{8\mathrm{i}}\sum _{\mathrm{k}=1}^{\mathrm{i}}{\mathrm{d}\mathrm{u}\mathrm{m}}_{\mathrm{t}-\mathrm{i}}+{ϵ}_{\mathrm{t}}$(9)

In order to identify the optimal nonlinear autoregressive distributed lag parameters, we evaluated equation (9)(9) $\mathrm{\Delta }{\mathrm{c}\mathrm{a}\mathrm{r}}_{\mathrm{t}}={\gamma }_0+{\gamma }_1{\mathrm{c}\mathrm{a}\mathrm{r}}_{\mathrm{t}-1}+{\gamma }_2{\mathrm{f}\mathrm{o}\mathrm{o}}_{\mathrm{t}-1}^{+}+{\gamma }_3{\mathrm{f}\mathrm{o}\mathrm{o}}_{\mathrm{t}-1}^{-}+{\gamma }_4{\mathrm{a}\mathrm{g}\mathrm{r}}_{\mathrm{t}-1}^{+}+{\gamma }_5{\mathrm{a}\mathrm{g}\mathrm{r}}_{\mathrm{t}-1}^{-}+{\gamma }_6{\mathrm{n}\mathrm{o}\mathrm{n}}_{\mathrm{t}-1}+{\gamma }_7{\mathrm{f}\mathrm{d}\mathrm{i}}_{\mathrm{t}-1}+{\gamma }_8{\mathrm{d}\mathrm{u}\mathrm{m}}_{\mathrm{t}-1}+\lambda {\mathrm{e}\mathrm{c}\mathrm{m}}_{\mathrm{t}-\mathrm{i}}+{\gamma }_{1\mathrm{i}}\sum _{\mathrm{k}=\mathrm{i}}^{\mathrm{n}1}{\mathrm{\Delta }\mathrm{c}\mathrm{a}\mathrm{r}}_{\mathrm{t}-\mathrm{i}}+{\gamma }_{2\mathrm{i}}\sum _{\mathrm{k}=1}^{\mathrm{i}}\mathrm{\Delta }{\mathrm{f}\mathrm{o}\mathrm{o}}_{\mathrm{t}-\mathrm{i}}^{+}+{\gamma }_{3\mathrm{i}}\sum _{\mathrm{k}=1}^{\mathrm{i}}\mathrm{\Delta }{\mathrm{f}\mathrm{o}\mathrm{o}}_{\mathrm{t}-\mathrm{i}}^{-}+{\gamma }_{4\mathrm{i}}\sum _{\mathrm{k}=1}^{\mathrm{i}}\mathrm{\Delta }{\mathrm{a}\mathrm{g}\mathrm{r}}_{\mathrm{t}-\mathrm{i}}^{+}+{\gamma }_{5\mathrm{i}}\sum _{\mathrm{k}=1}^{\mathrm{i}}\mathrm{\Delta }{\mathrm{a}\mathrm{g}\mathrm{r}}_{\mathrm{t}-\mathrm{i}}^{-}+{\gamma }_{6\mathrm{i}}\sum _{\mathrm{k}=1}^{\mathrm{i}}{\mathrm{\Delta }\mathrm{n}\mathrm{o}\mathrm{n}}_{\mathrm{t}-\mathrm{i}}+{\gamma }_{7\mathrm{i}}\sum _{\mathrm{k}=1}^{\mathrm{i}}{\mathrm{\Delta }\mathrm{f}\mathrm{d}\mathrm{i}}_{\mathrm{t}-\mathrm{i}}+{\gamma }_{8\mathrm{i}}\sum _{\mathrm{k}=1}^{\mathrm{i}}{\mathrm{d}\mathrm{u}\mathrm{m}}_{\mathrm{t}-\mathrm{i}}+{ϵ}_{\mathrm{t}}$(9) utilizing the general-to-specific methodology, which has been corroborated by studies such as Arize et al (Arize et al., 2017), Karamelikli and Karimi (Karamelikli & Karimi, 2022), and Hussain et al (Hussain et al., 2019). To circumvent potential disruptions that may impede the cointegration test from dismissing the null hypothesis of long-run association, we employed the structural break test as suggested by Bai and Perron (Bai & Perron, 2003).

4. Results and discussion

4.1. Basic statistical tests

The aim of this subsection is to conduct essential statistical analyses, establishing a foundation for exploring the influence of food security and agriculture on carbon dioxide emissions utilizing the nonlinear autoregressive distributed lag approach. The analysis encompasses five distinct facets, including the unit root test, the Zivot-Andrews test, descriptive statistical assessment, correlation examination, and the bounds test. The nonlinear autoregressive distributed lag approach proves effective when the investigated variables are stationary at I(0), I(1), or a combination of I(0) and I(1). However, the presence of any I(2) variable complicates the application of this method, as noted by Sek (Sek, 2017) and Cheng and Cao (Cheng & Cao, 2019). Consequently, verifying the absence of I(2) stationary variables and confirming the integration of the investigated variables was essential. To achieve this, the conventional unit root tests, ADF and PP, were employed. Panel A of Table 2 presents a summary of the unit root test results. Likewise, the outcomes of the Zivot-Andrews structural break unit root test are displayed in Panel B of Table 2. Panel C of Table 2 showcases the findings of the descriptive statistical assessment, while the correlation test results are documented in Panel D. Lastly, Panel E of Table 2 features the outcomes of the bound test.

Table 2. Results of basic statistical tests

Panel A: Unit root test
Variable		ADF test					PP test
		Level		First difference			Level			First difference
car		−0.349		−7.309***			0.182			−9.403***
foo		−1.235		−3.295*			−2.687			−3.317*
agr		−3.043		−9.742***			−2.883			−12.136***
non		−0.315		−5.478***			−0.884			−5.506***
fdi		−6.317***		−10.801***			−5.614***			−37.320***
Panel B: Zivot-Andrews test
Variable		Level		Break-year			First difference			Break-year
car		−3.395		1982			−8.338***			1998
foo		−2.057		1997			−19.602***			1973
agr		−3.510		1980			−10.353***			1978
non		−1.535		1982			−6.318***			2011
fdi		−1.475		1998			−13.638***			1997
Panel C: Descriptive statistical test
Variable		Mean		Maximum			Minimum			Standard error
car		2.550		2.880			1.997			0.283
foo		1.639		1.712			1.508			0.060
agr		1.317		1.469			1.040			0.119
non		0.882		0.995			0.805			0.067
fdi		0.722		4.566			0.010			0.733
Panel D: Correlation test
Variable	car		foo		agr			non			fdi
car	1.000
foo	0.995***		1.000
agr	−0.975***		0.982***		1.000
non	0.913***		−0.917***		−0.903***			1.000
fdi	−0.154		0.146***		0.107			−0.223			1.000
Panel E: Bounds test
Test statistic	Value		Significant level			I(0)			I(1)
F-statistic	4.934***		10%			2.20			3.09
k	4		5%			2.56			3.49
			1%			3.29			4.37

Note: *1% significant level; **5% significant level; ***10% significant level.

The unit root test findings in Panel A of Table 2 reveal that carbon dioxide emissions, food security, agriculture value-added, and non-renewable energy consumption are non-stationary at their respective levels, but stationary at their first-order differences. Meanwhile, foreign direct investment is stationary both at its level and first-order difference. Consequently, the investigated variables comprise a mix of I(0) and I(1), which meets the prerequisite for implementing the nonlinear autoregressive distributed lag approach. As per the Zivot-Andrews test results in Panel B, these variables exhibit non-stationarity or unit root issues at the level of structural breaks occurring in 1982 (carbon dioxide emissions), 1997 (food security), 1980 (agriculture value-added), 1982 (non-renewable energy consumption), and 1998 (foreign direct investment). Despite these structural breaks, the variables are found to be stationary at the intended level. The descriptive statistical test outcomes in Panel C of Table 2 indicate that food security and non-renewable energy consumption exhibit less variability than carbon dioxide emissions, agriculture value-added, and foreign direct investment. Furthermore, the positive mean values of the five investigated variables suggest a rising trend over time. The correlation test results in Panel D reveal that food security and non-renewable energy consumption share a positive correlation with carbon dioxide emissions, while agriculture value-added and foreign direct investment are negatively correlated. Lastly, the bounds test results in Panel E show that the calculated F-statistic value surpasses the critical value at a 1% significance level, confirming the cointegration between food security, agriculture value-added, non-renewable energy consumption, foreign direct investment, and carbon dioxide emissions.

4.2. Asymmetric effects of food security and agriculture on environmental deterioration

The purpose of this segment is to explore the asymmetric effects of food security and agriculture on carbon dioxide emissions. The results of this analysis are presented in Table 3.

Table 3. Results of effects of food security and agriculture on environmental deterioration

Long-run effect
$\mathrm{f}\mathrm{o}{\mathrm{o}}^{+}$	0.042* (1.958)
$\mathrm{f}\mathrm{o}{\mathrm{o}}^{-}$	0.090* (−1.931)
$\mathrm{a}\mathrm{g}{\mathrm{r}}^{+}$	0.748*** (3.814)
$\mathrm{a}\mathrm{g}{\mathrm{r}}^{-}$	−0.688*** (−3.529)
$\mathrm{n}\mathrm{o}\mathrm{n}$	0.202** (2.092)
$\mathrm{f}\mathrm{d}\mathrm{i}$	−0.348*** (−11.032)
Short-run effect
${\mathrm{\Delta }\mathrm{f}\mathrm{o}\mathrm{o}}_{-1}^{+}$	0.019** (2.032)
${\mathrm{\Delta }\mathrm{f}\mathrm{o}\mathrm{o}}_{-1}^{-}$	−0.049*** (−4.223)
${\mathrm{\Delta }\mathrm{f}\mathrm{o}\mathrm{o}}_{-2}^{-}$	−0.029 (−1.071)
${\mathrm{\Delta }\mathrm{a}\mathrm{g}\mathrm{r}}_{-1}^{+}$	0.294*** (2.699)
${\mathrm{\Delta }\mathrm{a}\mathrm{g}\mathrm{r}}_{-1}^{-}$	−0.532** (−2.155)
$\mathrm{\Delta }\mathrm{n}\mathrm{o}{\mathrm{n}}_{-1}$	0.168** (2.223)
$\mathrm{\Delta }\mathrm{n}\mathrm{o}{\mathrm{n}}_{-2}$	0.034*** (2.378)
$\mathrm{\Delta }\mathrm{f}\mathrm{d}{\mathrm{i}}_{-1}$	−0.146*** (−5.178)
${\mathrm{d}}_{1997}$	−0.023** (−2.297)
${\mathrm{d}}_{2011}$	0.034*** (4.295)
$\mathrm{c}$	0.032** (2.159)
$\mathrm{e}\mathrm{c}{\mathrm{m}}_{-1}$	−0.087*** (−6.490)

Note: + positive component; - negative component; $\mathrm{\Delta }$ difference operator; ${\mathrm{d}}_{1997}$ dummy variable; ${\mathrm{d}}_{2011}$ dummy variable; $\mathrm{c}$ constant; $\mathrm{e}\mathrm{c}\mathrm{m}$ error correction term; * 10% significant level; ** 5% significant level; * 1% significant level.

Table 3 displays the long-run and short-run effects of the investigated variables on carbon dioxide emissions. In terms of long-term impacts, it is evident that a positive shock in food security results in increased carbon dioxide emissions, whereas a negative shock leads to reduced emissions. This observation aligns with the findings of Mahdavian et al (Mahdavian et al., 2022) and Naseem et al (Naseem, Guangji, et al., 2020). A possible explanation is that maintaining food security necessitates the involvement of supporting industries such as processing and transportation, which contribute to emissions. Additionally, the analysis reveals that a positive shock in agriculture increases carbon dioxide emissions, while a negative shock decreases them. This supports the conclusions of Asumadu-Sarkodie and Owusu (Asumadu-Sarkodie & Owusu, 2016), Dogan (Dogan, 2016), and Ramzan et al (Ramzan et al., 2022). One potential reason for a positive shock in agriculture is Korea’s need to enhance agricultural production within its limited land resources, which may involve increased use of agricultural technology and chemicals, leading to higher emissions. Conversely, a negative shock in agriculture can be attributed to Korea’s strong focus on export and import, with a significant portion of import demand consisting of agricultural and sideline products. This suggests that a decrease in domestic agricultural production could result in lower carbon dioxide emissions. Furthermore, the study finds that nonrenewable energy consumption positively impacts carbon dioxide emissions, consistent with He and Zhang (He & Zhang, 2022) and He et al (He et al., 2022). This can be attributed to the exploitation and use of fossil fuels, the primary source of carbon dioxide emissions. Simultaneously, foreign direct investment is found to reduce carbon dioxide emissions, corroborating the findings of Abid et al (Abid et al., 2022), Jafri et al (Jafri et al., 2022), and Wang et al (Z. Wang et al., 2022). This may be due to Korea relocating highly polluting industries to other countries and implementing advanced pollution treatment technology, both of which contribute to reduced carbon dioxide emissions.

In terms of short-run impacts on carbon dioxide emissions, it is observed that a positive shock in food security leads to increased emissions, while a negative shock results in a decrease. Similarly, a positive shock in agriculture causes a rise in emissions, whereas a negative shock leads to a reduction. Meanwhile, nonrenewable energy consumption is found to have a positive effect on carbon dioxide emissions, while foreign direct investment negatively influences them. Furthermore, the coefficients of both dummy variables (${\mathrm{d}}_{1997}$ and ${\mathrm{d}}_{2011}$) are statistically significant. Two potential explanations underlie these results: Firstly, South Korea’s energy structure overhaul in 2011 is a plausible reason. Reuters (Seoul) reported on January 25 that South Korean companies were expected to invest 4.5 trillion won (4.03 billion US dollars) in renewable energy in 2011, a 23% increase compared to 2010. This suggests that, in line with the findings of Yu et al (Yu et al., 2022), Saboori et al (Saboori et al., 2017), and Zhang (S. Zhang, 2018), South Korea’s dependence on nonrenewable energy has been gradually decreasing since 2011. Secondly, the 1997 financial crisis had a significant impact on the Korean economy, prompting the country to adopt advanced technology through foreign direct investment for economic growth. Studies by Bulus and Koc (Bulus & Koc, 2021), Park et al (Park et al., 2019), and Kim and Seok (Kim & Seok, 2023) indicate that advanced technology has increased energy efficiency while reducing its adverse environmental effects. Additionally, the speed at which the system returns to equilibrium following a short-run shock to the investigated variable is 0.087% in the subsequent period.

To ensure the reliability and validity of the results presented in Table 3, a series of diagnostic tests are conducted, as outlined below. The findings from these diagnostic assessments can be found in Table 4.

Table 4. Results of diagnostic tests

Panel F: Brock, Dechert, Scheinkman and LeBaron (1996) Independence Test
Dimension	car	foo	agr	non	fdi
2	0.207 (0.000)	0.081 (0.000)	0.178 (0.000)	0.183 (0.000)	0.207 (0.000)
3	0.351 (0.000)	0.121 (0.000)	0.308 (0.000)	0.313 (0.000)	0.352 (0.000)
4	0.451 (0.000)	0.159 (0.000)	0.399 (0.000)	0.400 (0.000)	0.455 (0.000)
5	0.523 (0.000)	0.182 (0.000)	0.461 (0.000)	0.456 (0.000)	0.528 (0.000)
6	0.574 (0.000)	0.203 (0.000)	0.502 (0.000)	0.492 (0.000)	0.581 (0.000)
Panel G: asymmetric test
Long-run asymmetric test			Short-run asymmetric test
Variable	F-statistic	Outcome	Variable	F-statistic	Outcome
foo	8.526 (0.006)	yes	foo	1.322 (0.257)	no
agr	26.850 (0.000)	yes	agr	0.542 (0.66)	no
Panel H: stability test
Method		F-statistic		P-value
Normality		1.371		0.339
${\chi }^2$-serial		1.217 (Nicholson et al., 2021; Ziska et al., 2012)		0.312
${\chi }^2$-white		0.083 (Pata & Kartal, 2023; Sabir et al., 2020)		0.774
${\chi }^2$-Ramsey		1.544 (Lu et al., 2007; Qadir et al., 2018)		0.258
CUSUM test		Stable
CUSUM of Squares		Stable

Note: Standard error in the parenthesis; yes, having asymmetric effect; no, having no asymmetric effect.

The correlation integral, a concept utilized in this test, measures frequency and is designed to identify any unobserved logical or predictable non-stationary characteristics within a time series. Concurrently, this test aids in distinguishing chaotic processes from nonlinear ones. The test is structured to provide high efficiency compared to linear chaos, revealing its potential to evaluate various forms of nonlinearity. Panel F of Table 4 presents the results of the independence test conducted by Brock, Dechert, Scheinkman, and LeBaron (Broock et al., 1996). These findings do not support the null hypothesis of linear correlations, indicating nonlinearity exists across all investigated variables. This is demonstrated by the significant Brock, Dechert, Scheinkman, and LeBaron statistics for each of the embedding dimensions. As the embedding dimensions’ increase, so do the Brock, Dechert, Scheinkman, and LeBaron statistics, confirming the presence of substantial nonlinearity for larger dimensions. Panel G of Table 4 displays the results of the Wald test, which evaluates the presence of asymmetrical impacts in the testing hypothesis. The findings reveal that food security, agriculture value-added, and carbon dioxide emissions exhibit asymmetrical effects in the long run. However, the asymmetric short-run effects for these variables—food security, agriculture value-added, and carbon dioxide emissions—have not yet manifested themselves. Panel H of Table 4 presents the results of the stability test. The normality test outcome indicates that the residual follows a normal distribution. The serial correlation test result reveals an absence of serial correlation, as evidenced by the ${\chi }^2$-serial value. Additionally, the heteroscedasticity test outcome suggests no heteroscedasticity, based on the ${\chi }^2$-white value. Furthermore, in line with McLeod and Li (McLeod & Li, 1983), Broock et al (Broock et al., 1996), and Baek and Brock (Baek & Brock, 1992), the stability diagnostics results (${\chi }^2$-Ramsey, CUSUM test, and CUSUM of Squares) indicate that the model employed in this study is robust, valid, and reliable for making inferences and predictions.

In accordance with Lee and Zeng (C. -C. Lee & Zeng, 2011), Hatemi-J et al (Hatemi-J et al., 2017), and Nouira et al (Nouira et al., 2019), this study employs both symmetric and asymmetric causality tests to ascertain the direction and cumulative coefficient of the variables under investigation. The findings of these tests are exhibited in Table 5.

Table 5. Results of asymmetric and symmetric causality tests

Panel I: symmetric causality test
Null hypothesis	F-statistic	P-value
foo $\to$ car	6.511	0.018
agr $\to$ car	7.733	0.013
non $\to$ car	0.273	0.585
fdi $\to$ car	9.970	0.000
car $\to$ foo	0.501	0.618
car $\to$ agr	1.913	0.173
car $\to$ non	2.755	0.104
car$\to$ fdi	4.494	0.039
Panel J: asymmetric causality test
foo+ $\to$ car	5.807	0.048
foo- $\to$ car	3.245	0.057
car $\to$ foo+	1.091	0.542
car $\to$ foo-	2.734	0.141
agr+ $\to$ car	5.514	0.037
agr- $\to$ car	5.589	0.026
car $\to$ agr+	2.329	0.162
car $\to$ agr-	1.952	0.673

Note: + positive component; - negative component.

Panel I of Table 5 presents the outcomes of the symmetric causality test. A unidirectional causality is evident between food security, agriculture value-added, and carbon dioxide emissions. Statistically, no causal link is detected between nonrenewable energy consumption and carbon dioxide emissions. A bidirectional relationship exists between foreign direct investment and carbon dioxide emissions. Based on these insights, one can deduce that food security, agriculture value-added, and foreign direct investment play crucial roles in shaping Korea’s environmental quality. Conversely, it implies that the impact of nonrenewable energy consumption on Korea’s environmental quality has been waning and becoming less significant. Panel J of Table 5 displays the findings of the asymmetric causality test. Unidirectional causality can be established between the positive components of food security and agriculture value-added, and carbon dioxide emissions. Concurrently, unidirectional causality can also be demonstrated between the negative components of food security and agriculture value-added, and carbon dioxide emissions. As a result, considering the asymmetric influences of food security and agriculture value-added on environmental sustainability, the Korean government should work to mitigate its adverse impact on the environment while amplifying its positive contributions to environmental enhancement.

5. Conclusions

In this study, we examine the influence of food security and agriculture value-added on carbon dioxide emissions (a proxy for environmental sustainability) in South Korea, spanning from 1970 to 2020. Employing the nonlinear autoregressive distributed lag approach for our empirical analysis, the findings reveal the presence of long-term asymmetric effects of food security and agriculture value-added on carbon dioxide emissions. We ascertain that positive shocks in food security and agriculture value-added contribute to increased carbon dioxide emissions, whereas negative shocks in food security and agriculture value-added result in decreased emissions. Furthermore, our analysis uncovers that nonrenewable energy consumption positively impacts carbon dioxide emissions, while foreign direct investment exerts a negative influence on emissions. The insights gleaned from this study may serve as a foundation for the formulation of pertinent policies by the Korean government to ensure continued environmental sustainability.

Based on the empirical evidence gathered in this study, we propose the following policy recommendations to address the environmental concerns in South Korea. Firstly, given the asymmetric impact of food security on carbon dioxide emissions, it is crucial for the South Korean government to adopt environmentally friendly food security measures, such as establishing a green grain reserve system, to foster environmental sustainability. Secondly, as the influence of value-added agriculture on carbon dioxide emissions is also asymmetric, it is vital for the government to emphasize eco-friendly farming practices that promote environmental sustainability. Thirdly, considering the negative impact of foreign direct investment on carbon dioxide emissions, the South Korean government should prioritize reducing investments in polluting sectors while promoting investments in environmentally sustainable industries. Fourthly, as nonrenewable energy consumption positively affects carbon dioxide emissions, the government should focus on developing and utilizing renewable energy sources to mitigate emissions. Simultaneously, the government should invest in innovative technologies to enhance energy efficiency and achieve environmental sustainability.

This study, while providing valuable insights, also acknowledges certain limitations, which in turn, present opportunities for future research endeavors. Firstly, the research is focused solely on South Korea, potentially leading to biased findings. Expanding the scope of the study to include countries like China or India could yield more comprehensive results. Secondly, the time-series method is exclusively employed in this investigation, which may result in findings that are not universally applicable. Future researchers could consider utilizing panel data techniques or spatial econometric methodologies to uncover more fascinating insights. Lastly, not all control variables are considered in this study, which could affect the accuracy of the findings. Future investigations might benefit from incorporating additional control factors, leading to more robust and reliable outcomes.

Disclosure statement

No potential conflict of interest was reported by the author.

Data availability statement

The data presented in this study are available from the author upon request.

Additional information

Funding

The author(s) reported there is no funding associated with the work featured in this article.

Notes on contributors

Yugang He

Professor Yugang He, currently an esteemed faculty member at Sejong University in South Korea, boasts an extensive and diverse academic portfolio. His scholastic pursuits traverse the fields of international economics and trade, energy and environmental economics, agricultural and social economics, as well as econometrics. As a testament to his academic standing, Prof. He has assumed pivotal roles as an editor and reviewer for an array of distinguished SSCI, SCI, and A&HCI journals. Prof. He's academic influence extends beyond the realm of scholarly literature, as he is frequently invited to share his expertise at international academic conferences. His research has garnered significant recognition, earning him multiple accolades, including the coveted Best Paper Award on numerous occasions. Furthermore, his pedagogical prowess is often sought after, resulting in his being invited to impart his knowledge to graduate students at various universities. Over the course of his vibrant academic career, Prof. He has been successful in the publication of over 30 scholarly articles in highly respected journals, thereby contributing significantly to the rich tapestry of academic discourse in his fields of interest. His consistent engagement with cutting-edge research and knowledge dissemination solidifies his status as a leading figure in his academic sphere.