Economic Indicators And Climate Change For BRICS Economies In The Post COVID-19 World: Neural Network Approach

Climate Change has emerged as one of the challenges of the global economy. Climate Change economics has focused on the economic aspects of climate trade-off. Studies have been conducted on the causal association of economic indicators and climate change indicators. However, for the sample of BRICS countries, that are important participants of global climate change, no study has attempted to identify whether causal connections apply to them. The study is an endeavor to identify the underlying causal connections between economic indicators and carbon emissions for BRICS economies. Six economic indicators, Current Account Balance, Inflation, Foreign Direct Investment Inflows, Gross Domestic Product, Real Effective Exchange Rate, and Trade Openness are selected for the sample period 2005-2019. Neural Network analysis as a method of computational economics is applied for the superior methodology over standard statistical techniques. The outcome suggests that for BRICS economies economic indicators have a significant relationship with Carbon Emissions, differing in intensity as per node strength of the neural network.


Figure 1. Climate Change Theories
Source: Climate Change issues in BRICS countries All these theories attempt to identify the cause and suggests controlling the cause may reduce climate change. The anthropogenic theory is the most important for our study. This theory states that human causes are the primary causes responsible for altering climate change. Other theories are more subtle and complex. Due to the popularity of anthropogenic theory, in general, climate change is understood as the changes in human factors including the carbon emissions due to production activities of humans. COVID-19 as a pandemic entered the world creating havoc and chaos, shutting down the physical activities, countries went for the nationwide lockdown. BRICS economies also responded to global pandemic and implemented nationwide lockdown in 2020. Brazil was one of the countries to implement the lockdown with delay. The first case of COVID-19 was reported on 25 th February 2020. However, due to the reluctance of the President, the lockdown was implemented on May 7 in several states when the cases started rising rapidly. The first case of COVID-19 was reported on 31 st January after which preventive measures were imposed. However, as the cases started to rise in end February and March, lockdown was imposed on April 17 2020. However, in the end of March it was already implemented on the Russian borders. On 24 th March 2020, a nationwide lockdown was imposed in India for 21 days which was later extended. He cases were already rising and this created a total shutdown of economic and industrial activities. China was the destination for the origination of the COVID-19 virus. It all started from Wuhan and a lockdown was implemented in Wuhan on 23 rd January 2020. Lockdown was followed by quarantine measures, follow up checks, and travel history analysis for possible carrier of virus. On 5 th March 2020 officially it was reported in South Africa that cases have started spreading rapidly. As a breakdown strategy, nationwide lockdown was imposed on 27 th March 2020. BRICS economies responded quickly to the COVID-19 pandemic except Brazil that was late. This lockdown resulted in almost total lockdown of industrial and economic activities. This has resulted in reduction in the carbon emissions for BRICS countries. However, no official statistics has been released by any of the multilateral agencies. In absence of any official statistics, it can only remain conjecture on the carbon emissions. In figure 2, the study conceptualizes Covid10 lockdown and Climate Change with the help of quadrant. The first quadrant (+, +) denotes the positive impact of Climate Change and COVID-19 lockdown resulting in reduced carbon emissions and a protected ecosystem. Due to lockdown, the movement of humans were at minimum in the BRICS economies for the first time in history. This reduced the carbon emissions as well as gave time to ecosystem to rejuvenate. The second quadrant (+, -) captures the positive role of COVID-19 lockdown but still a negative impact on Climate Change. Energy consumption was still high due to home consumption, rather the home consumption of energy went up as people were locked in their homes. Also due to the pressure on digital economy, it is expected that digital waste has gone up which further leads to degradation of environment in the form of toxic waste. The third quadrant (-, -) presents the negative role of COVID-19 as well as a negative impact on Climate Change. Again, the increased consumption at home of food items including meat consumption (it leads to carbon content) will in the long-term negative impact the climate change. Finally, the fourth quadrant (-, +) indicated a negative impact of COVID-19 lockdown but a positive impact on Climate Change. It has been reported that the number of home disputes have gone up as people were locked in their homes. The economic cost in the form job loss cannot be ignored which will have a long-term consequence for the people. In this quadrant, no negative impact on climate change is to be identified. It is imperative to highlight the stringency index of COVID-19 lock for the initial months to assess the intensity of the shutdown of the BRICS economies. Stringency Index is a comprehensive index of 17 indicators of COVID-19 lockdown including containment at schools, workplace, public events, commute services et cetera. Figure 2 shows the Stringency Index for BRICS economies.

Section 3: Review of Literature
Studies have started pouring in on the subject of COVID-19 and climate change. A survey of New York residents identified that electricity consumption has become much higher due to work from home. The perception of the respondents towards climate change has remained unchanged during COVID-19 (Chen, de Rubens, Xu, and Li, 2020). A study conducted on the province of Ontario identified that the electricity consumption in the month of April 2020 declined by 14%. The pattern of highest electricity demand has changed after the COVID-19. In pre-covid1, Mon-Fri were the days with highest electricity demand. However, in the post COVID-19, Mon-Tue became the days with the highest electricity demand. The hourly electricity demand curve also flattened during the lockdown period (Abu-Rayash and Dincer, 2020). The wave of Coronavirus has given an opportunity to think about the wat of life in the context of climate change. There are lessons to be drawn on sustainability for the future world (Mair, 2020). The COVID-19 pandemic poses five set of issues with respect to climate change such as impacts on emissions, environmental policy, investment in green deals, deglobalizing climate change policies and intergenerational environmental impacts. The study argued that the positive consequences of the COVID-19 on environment will be short term and the world should not dump the environmental concerns (Helm, 2020). Thus, for future environmental policies it is important to put clean energy at the front of the stimulus package (Birol, 2020). As we have conceptualized the rejuvenation of ecosystem, evidence suggests that a strong association of biodiversity conservation with coronavirus (Corlett, et al. 2020). Using a multi-regional macro-economic model, researchers have attempted to identify the spill over effects of lockdown and other containments during the COVID-19. It was identified that global atmospheric emissions are reduced by 2.5Gt of greenhouse gases, 0.6Mt of PM2.5, and 5.1Mt of SO2 and NOx. This is remarkable in absence of any specific endeavor. However, the study also suggests socio-economic challenges for the global economy including unsustainable global patterns (Lenzen, et al. 2020). The post pandemic world requires stringent planning to tackle climate change issues. The public policy for climate change must incorporate the lessons drawn from the COVID-19 crisis (Pinner, Rogers, and Samandari, 2020). With respect to the relationship between Climate Change and Economic Indicators, most of the studies focus on carbon emissions and economic growth. Evidence is available with dynamic effects model on saving and capital accumulation for economic impact of climate change. There will be lower output when savings rate is constant in the presence of climate change, this has a spill over effect on the investment (Fankhauser, and Tol, 2005). SMEs also are influenced by government policies pertaining to climate change (Iqbal and Rahman, 2015). Evidence from last 50 years on the relationship between climate change and economic growth suggest that higher temperature substantially 0 20 40 60 80 100 120 01jan2020 06jan2020 11jan2020 16jan2020 21jan2020 26jan2020 31jan2020 05feb2020 10feb2020 15feb2020 20feb2020 25feb2020 01mar2020 06mar2020 11mar2020 16mar2020 21mar2020 26mar2020 31mar2020 05apr2020 10apr2020 15apr2020 20apr2020 25apr2020 30apr2020 Brazil Russia India China South Africa reduces economic growth in poor countries reducing agricultural and industrial output. Thus, poor countries feel more burn of the climate change on their economic growth (Dell, Jones, and Olken, 2008). With the help of an integrated assessment model for economic growth and climate change, it was found that economic growth has substantial effect on climate change and vice-versa particularly for the developing countries (Roson, and Van der Mensbrugghe, 2012).

Section 4: Neural Network Analysis for Economic Indicators and Carbon Emissions
The plausible question in climate change economics has remained the causal linkages between economic indicators and indicators of climate change. Carbon Emissions has remained the single most important indicator of Climate Change due to two factors; the evidence of the significant impact of carbon emissions on climate and the data availability of carbon emissions. The recent techniques of computational economics have the potential to identify causal linkages, neural network approach is one of such endeavors. A neural network is a computational family of models with ample parameter space, independent of hypothesis, flexible structure, developed resembling brain functioning. Neural network analysis has an advantage over traditional regression models. Regression models are based on Ordinary Least Squares (along with finite sample properties) and store judgemental knowledge in the regression coefficients. Regression analysis is just one type of neural network, but neural network analysis is far more than Ordinary Least Squares. Another superiority of a neural network is that it is dynamic rather than static (regression). We use the multilayer perceptron method for a model with one dependent variable (target output) and several predictive variables. The variable description is presented in Annexure-I, and the data description with justification is presented in Annexure-II. The analysis employs three layers, namely output layer, input layer and the unobserved layer.
Step 1 initializes the analysis of the feedforward structure while Step 2 trains the data for revealing the internal dynamics.
Step 3 is the critical step of creating of forward propagation for the data set fed in Step 2.
Step 4 goes with the backward propagation where all reliability and validity are tested, and if now appropriate the model collapses without giving node results. The final step of the cycle is iteration of the data to make it a possible probabilistic model based on the Bayesian mathematics. The power of neural network approach in such a feedforward structure is incredible and has a multitude of benefits over standard statistical procedures. The most prominent being, no application of hypothesis formulation as it is imbibed in the methodology. The normalized importance output becomes the evidence of variables to be associated with the target out, whether due to mediating or moderating role. Step 1

Training
Step 3

Forward Propagation
Step 4

Backward Propagation
Step 5 Iteration Vol. 7, Iss. 4, Article ID 9900058, 2021 Original Research Article The neural analysis is run in a single command for the full dataset to minimize the training time and generate overall results for BRICS economies. All predictors are fed as covariates due to close linkages between economic indicators. Table 1 shows the case processing summary for the multilayer perceptron. The total cases in table 1 are 75, while the total number of observations in the dataset is 510. Cases here means the corresponding year figure as the data is a panel. For Training purposes (Step 2 of the figure), 70% cases are used (357 observations), and for testing purposes (Step 3-5 of the figure 2) 30% cases are used (153 observations). The network, with its internal dynamics, is presented in Table 2. Sum of Squares a. Excluding the bias unit Source: Output generated through Neural Network by the researchers Six input layers denote the predictors (covariates) and one hidden Layer with a hyperbolic tangent for Carbon Emissions (CO2) (target output). The standardized method for covariates and scale dependents is applied to make the output parsimonious. The network structure is selected to make the model robust and reliable. Annexure-III shows the network nodes and the outcome of the analysis. The blue lines in the hidden layer activation function suggest a causal relationship, while the dull lines indicate a weak relationship (meaning thereby insignificant relationship). The significant variables affecting the CO2 for the panel of BRICS Economies are Current Account Balance (CAB) via hidden Layer 1:2, Inflation (CPI) via hidden Layer 1:1 and 1:4; Gross Domestic Product (GDP) via hidden Layer 1:1 and 1:2; Foreign Direct Investment Inflows (FDI) via hidden Layer 1:2, 1:3 and 1:4, Real Effective Exchange Rate (RER) via hidden Layer 1:2, 1:4 and 1:5; and Trade Openness (TOP) via hidden layer 1:1 and 1:2. The hidden layer H(1:1) and H(1:2) have a stronger relationship with the Carbon Emissions of the BRICS Economies. The network suggests that variables having an association with H(1:1) and H(1:2) are having a significant relationship with the Carbon Emissions of BRICS Economies, indicating CPI, FDI, GDP, RER, and TOP. The neural analysis suggest that bias (omitted variables) does play an essential role in impacting the Carbon Emissions of the BRICS economies, which is natural. However, it also suggests that the economic indicators are having a significant impact on the carbon emissions of BRICS economies. The relative importance of the predictors is shown in Figure  5 (For statistics, see Annexure V).
Vol. 7, Iss. 4, Article ID 9900058, 2021 Original Research Article The normalized importance of the predictors suggests the order of predictors in importance with respect to carbon emissions. The most important predictor is the Trade Openness followed by GDP of the country. Next comes the Current Account Balance and Real Effective Exchange Rate of the BRICS economies. However, FDI and Inflation plays a pity role in the climate change economics of BRICS economies. Table 4 captures the summarized results of neural network analysis for the BRICS Economies.