Spatial Effects of Economic Activity and Environmental Carrying Capacity on Air Quality in Java and Bali Islands

The purpose of this research is to determine the direct effects and spillover effects of environmental carrying capacity, urbanization, and economic activities on air quality in the provinces of Java and Bali Islands. The research contributes academically and practically by providing a spatial-based model design and policy recommendations for improving air quality in the provinces of Java and Bali Islands. The research method used is quantitative, utilizing secondary data from seven provinces in Java and Bali Islands during 2013-2019, obtained from publications and official documents of the Indonesian Central Statistics Agency, Ministry of Environment and Forestry, and Ministry of Finance. The research variables include air quality, urbanization, per capita income, exports, imports, foreign direct investment, and environmental carrying capacity. The analysis technique employed is Spatial Econometrics. The results show that spillover effects occur due to economic activities and the environmental carrying capacity of each region. Policy synergies between regions oriented towards air quality improvement are necessary.


Introduction
Air quality is an important factor for human health.All regions require clean air as the most crucial element for the environment and health [1].Efforts to improve air quality are part of Indonesia's commitment at the 26th UN Climate Change Conference of the Parties (COP 26) in Glasgow, Scotland, from October 31 to November 12, 2021.According to the Ministry of Environment and Forestry of the Republic of Indonesia, Indonesia is committed to climate change mitigation.At COP 26, Indonesia has set a target for Net Sink Carbon in the land and forest sector by no later than 2030 (FOLU Net Sink 2030) and aims to achieve "Net Zero" by 2060 or earlier [2].Climate change mitigation involves identifying the determinants of CO2 emissions, which serve as an indicator of air quality.Economic 1248 (2023) 012025 IOP Publishing doi:10.1088/1755-1315/1248/1/012025 2 activities, such as high-tech investments from foreign direct investment (FDI) and international trade (exports and imports), are currently being extensively studied.The goal is to determine whether FDI and international trade align with the pollution haven hypothesis (PHH) or the pollution halo hypothesis (PH).Many studies are also conducted to examine the Environmental Kuznet Curve (EKC) with varying results.
The population in Indonesia is unevenly distributed across the archipelago.The population distribution can be observed through population density, measured as the number of inhabitants per square kilometer (Km2).The distribution of population density exhibits significant inequality (Table 1).Table 1 shows that nationally, the population density has been increasing at a rate of 1.24% per year.The population is concentrated in Java and Bali Islands, with a density of 74% in Java and 15.25% in Bali out of the total Indonesian population.Higher population density is associated with increased economic activity.This is also evident in the per capita income of each province.Based on data from the Central Statistics Agency in 2022, it is found that Java and Bali Islands had an above-average per capita gross regional domestic product (GRDP) growth rate (3.68%) during the period 2014-2019.Java had a growth rate of 4.61%, while Bali had a growth rate of 4.92%.This situation will have implications for increased environmental exploration, leading to a decline in air quality.. Data from the Ministry of Environment and Forestry of the Republic of Indonesia for the years 2013-2020 show that Java has an average air quality of 74.28, making it the island with the lowest air quality compared to other islands in Indonesia.On the other hand, Bali, with the second largest population density after Java, has good air quality, with a value of 88.61, which is above the national average (86.70).
In reality, each location has different characteristics, availability, and resilience of resources.[3] states that the environment should have the ability to support human life by providing resources and resilience to maintain its stability.Environmental carrying capacity provides benefits (ecosystem services).The higher the ecosystem services of an area, the greater the environmental carrying capacity to absorb substances, energy, and/or other components that enter or are introduced into it [4].The Land Cover Quality Index can be used as a parameter to measure the area of vegetation that functions to absorb substances, energy, and/or other pollutant-like components.When looking at the land cover area in Indonesia, there is a gap between Java and Bali compared to other islands.Based on data from the Ministry of Environment and Forestry of the Republic of Indonesia for the period 2013-2020, the national average Land Cover Quality Index is 59.74 (in the range of 50 < LCQI ≤ 60, which falls under the category of less good).Java and Bali are islands with land cover averages below the national average, at 39.43 and 39.84, respectively.
From the above description, it can be seen that the high or low economic growth does not necessarily indicate the good or poor air quality of an area.This contradicts studies [5], [6], and [7] which state that as the population increases, the CO2 emissions also increase, leading to a decline in air quality.The low land cover does not necessarily mean low air quality [4].This research is conducted to analyze the spatial effects of urbanization, per capita income, economic activities, and environmental carrying capacity on air quality in Java and Bali.It aims to synergize with national goals to achieve sustainable development (SDGs, Sustainable Development Goals).Spatial studies are crucial to implementing specific policies in each region.[8].Research considering the role of regions or regional factors has been suggested by [9], [10], and [11].Such conditions can occur because the sources of environmental loads may be spatially distributed, so it is advisable to analyze them at the city or regional level.It is expected that if the appropriate policies are implemented in one area to improve air quality, the air quality in the surrounding areas will also improve.

Methods
The objective of this research is to determine the spatial effects of urbanization, per capita income, economic activities, and environmental carrying capacity on air quality in Java and Bali.This research utilizes a quantitative method.Spatial Econometric Analysis is applied to panel data, which is a combination of cross-sectional data across regions in Java and Bali Islands (7 provinces) with time series data spanning from 2013 to 2019.The research variables, measurement units, and data sources are presented in Table 2. To address heteroskedasticity, all variables are transformed into the logarithmic form [12], except for air quality, which is in the level form [13], and environmental carrying capacity.The environmental carrying capacity variable is modified based on Klassen's typology and classified into quadrants based on economic and environmental indicators.The determination of Quadrants 1 to 4 utilizes the Classification of Environmental Carrying Capacity Matrix Applying Klassen Typology [14].
In this research, two econometric models will be compared: the non-panel model (OLS) and the spatial panel model.Firstly, the general panel OLS model can be written as Equation 1: Remark   is AQ province i in year t;  is the measured parameter; x is the set of variables (urbanization, per capita income, exports, imports, and foreign direct investment) that affect y;  0 and  are constants and error terms, respectively.Secondly, the spatial panel model.This model uses two types of spatial weighting matrices: Contiguity Matrix-based Weight (Wcon) and Inverse Distance Matrixbased Weight (WDis).The determination of the proximity weighting matrix is done to measure the proximity between observations/regions.In other words, whether one province shares a border or adjacency with another province.Meanwhile, the determination of the distance-based weighting matrix requires identifying each location using its longitude and latitude coordinates.
Meanwhile, the specifications for the spatial model that will be built follow [15], namely the Spatial Lag X Model (SLXM).In the SLX model, the dependent variable at one location is related to the independent variable at a neighboring or other local unit.
is the variable under study, which is air quality.Independent variables have been shown from   include urbanization, income per capita, exports, imports, and FDI.  is a spatial weights matrix.While   ,   and  are spatial-specific effects, error terms, and parameter coefficients, respectively. is province 1,…, N (7 provinces) and  represents the period (2013-2019).Just like the value of ,  s are also the value of the parameter coefficient used to determine the effect of spatial spillover [16].
The best model is selected based on the likelihood value.This refers to the Maximum Likelihood Estimation (MLE) method proposed by [17].[18] explains that MLE is an estimation method that maximizes the likelihood function, which also maximizes the log-likelihood function.Therefore, the best model is the one that yields the highest log-likelihood.Meanwhile, the AIC is a measure indicating the relative quality of a model compared to others, where a smaller AIC value indicates a better fit compared to other models [17].

Results and Discussion
The population density in each region of Indonesia is mostly in the range of moderate population density, with a range of 125-263 people/km2.This condition is very different from the population density outside of Java and Bali Islands.The population density in provinces other than Java and Bali Islands falls into the low category.However, their land area is larger, and the availability of natural resources is greater compared to Java and Bali Islands.This becomes an important consideration in this dissertation research to conduct spatial panel regression for Java and Bali Islands.From the regression results, an overview of the air quality model in Java and Bali Islands will be obtained, providing additional research findings on the influence of population concentration in these two islands on the air quality model in Indonesia.
The spatial regression for Java and Bali Islands is carried out in several stages, starting with nonspatial panel regression, followed by a test of regional dependence (spatial dependence test).The results of the non-spatial regression estimation models, such as Ordinary Least Square (OLS), Fixed Effect (FE), and Random Effect (RE) for Java and Bali Islands, are shown in Table 3.The appropriate estimation model can be obtained by conducting several tests.First, the Chow Test is used to determine whether the ordinary least square (OLS) model or the fixed effect model is appropriate for panel data analysis.The testing criteria are as follows: null hypothesis (H0): the OLS model is the same as the fixed effect model, alternative hypothesis (Ha): the OLS model is not the same as the fixed effect model.The decision rule is to reject H0 if the p-value is less than the significance level of 0.05, indicating that the fixed effect model is used.Based on the Chow test results (Table 3), the significant p-value is less than 0.05.The fixed effect model is better than the OLS model.Second, the Hausman test is conducted to choose between the fixed effect model or the random effect model in panel data estimation.The testing criteria are as follows: null hypothesis (H0): the fixed effect model is the same as the random effect model, alternative hypothesis (Ha): the fixed effect model is not the same as the random effect model.The decision rule is to accept H0 if the p-value is greater than the significance level of 0.05 (5%), indicating that the random effect model is used.Based on the Hausman test results (Table 3), the p-value is 0.00, which is greater than 0.05 (>5%).The Ho is accepted, and the random effect model is used for estimating and analysing the spatial air quality in Java and Bali Islands.The next step is to test the dependence among regions in Java and Bali Islands using the Pesaran test.The criteria for the Pesaran test are as follows: null hypothesis (H0), no spatial dependence, the alternative hypothesis (Ha), spatial dependence exists in the model.The decision rule is to reject H0 if the p-value is less than the significance level of 0.05 (5%).This indicates the presence of spatial dependence in the model.Conversely, if the p-value is greater than 0.05 (>5%), H0 is accepted, indicating no spatial dependence in the model.The results of the Pesaran test are presented in Table 4.  4 shows that the p-value is less than 5%, indicating the rejection of H0.This suggests that there is spatial dependence in the non-spatial panel model.
Spatial regression for Java and Bali Islands is conducted after non-spatial panel regression and the test of spatial dependence.Spatial regression is performed using the Spatial Lag X (SLXM) panel model.The results of the spatial regression based on the proximity weighting matrix and distance weighting matrix allow for model comparison.The summary of the model comparison is shown in Table 5.In Table 5, Model 1 has the lowest AIC value and a higher likelihood value compared to Model 2. The Pseudo R2 value and the Wald test of spatial in Model 1 are also higher than those in Model 2. Based on these results, Model 1 is the most appropriate for the analysis of air quality in Java and Bali Islands.

TABLE 5. Comparison of Panel Spatial Lag X Models (SLXM) Based on Weighting Matrices in Java and Bali Islands
Model 1 showed that urbanization has a negative effect on air quality in Java and Bali Islands.According to [13], the interpretation of the results for a level-log model suggests that a 1% increase in urbanization in Java and Bali Islands leads to a 0.2269% decrease in air quality, ceteris paribus.This is consistent with the research conducted by [19], [20], [21], [22], and [23].The high population density in Java and Bali Islands falls into the category of very high, which results in a decrease in air quality with an increase in urbanization.These findings align with the negative impulse response of CO2 emissions at the national level to urbanization [24].This indicates that the population density in Java and Bali Islands is in the counter-urbanization phase in the short term.This phase is characterized by a decrease in the environmental, physical, and socio-economic carrying capacity of the population in larger metropolitan areas [25].The high rate of urbanization can lead to rapid population growth, resulting in agglomeration and subsequent efforts by the community to meet their needs.The speed and scale of escalation in these cities can put significant pressure on the environment and pose threats to sustainable development [26].Consequently, the air quality in Java and Bali Islands deteriorates.
The hypothesis of the Environmental Kuznets Curve (EKC) is validated in Java and Bali Islands.The findings indicate that the null hypothesis (Ho) is rejected because the coefficient of per capita income is negative (<0 and significant <0.1%), and the coefficient of squared per capita income is positive (>0 and significant <0.05).This suggests that the process of increasing per capita income initially leads to a decrease in air quality.However, in the long run, air quality improves, but it requires time and a process.Similar findings have been reported by [27] in 42 countries, [28] in the United States (USA), [29] in 8 ASEAN countries from 1965 to 2010, and [30] in countries within the North American Free Trade Agreement (NAFTA) from 1971 to 2014.
Positive spatial effects are observed in Java and Bali Islands for the variables of per capita income and imports from neighboring provinces (Table 5).If the per capita income and imports of neighboring provinces increase by 1%, the air quality in their respective regions improves by the corresponding coefficient, ceteris paribus.On the other hand, the variable of neighboring province exports has a negative impact on air quality in its region.If the exports of neighboring provinces increase by 1%, the air quality in their respective regions decreases by the corresponding coefficient, ceteris paribus.
The negative impact of exports on air quality in Java and Bali Islands indicates that a 1% increase in export quantity leads to a 0.131% decrease in air quality, ceteris paribus.This condition is further exacerbated by the validation of interregional dependence in the export variable (significantly below the confidence level of 0.01).If there is an increase in exports in neighboring provinces, it can decrease air quality in those provinces by up to 0.8499%, ceteris paribus.The direct impact of changes in exports on air quality in the province or region itself decreases by 0.131%, ceteris paribus (Table 6).
The spillover effect or indirect impact of neighboring regions or provinces results in an even greater decrease, reaching 0.72%, ceteris paribus, compared to the region itself.In total, it leads to a 0.859% decrease in air quality, ceteris paribus.The summary of the direct impact, spillover impact, and total impact in the spatial lag X model of air quality in Indonesia can be seen in Table 6.The validated spatial influence on the Export variable has a significant negative impact on air quality.This decline is due to the decreasing environmental carrying capacity to absorb the negative externalities arising from high-production activities in densely populated regions.This is supported by the modified Klassen typology study [14], which found that 6 provinces in Java Island, 4 of which are in quadrant 2. Quadrant 2 represents advanced but pressured regions (where the growth of per capita income is increasing, but the growth of environmental carrying capacity is less favorable).The four provinces are West Java, Central Java, Yogyakarta Special Region, and East Java Province.Two other provinces, Banten Province, are in Quadrant 3, and DKI Jakarta is in Quadrant 1. [14] also states that regions in Quadrant 2 and 3 are provinces with poor environmental carrying capacity in absorbing air pollutants.In other words, the majority of provinces in Java Island do not have sufficient environmental resilience to minimize negative externalities from export activities.The negative impact of the export variable on environmental quality aligns with the findings of studies [31] and [21].
The spillover effect from neighboring provinces exacerbates the decline in air quality in the provinces of Java and Bali Islands.The increase in exports, seen from the increase in production in a region, leads to a decrease in air quality.The decline in air quality in neighboring regions is likely to be greater than in their region due to natural factors such as weather and wind.DKI Jakarta Province has poor air quality.In addition to internal factors, the deterioration of air quality is also caused by contributions from Banten Province and West Java Province, which have many industrial centers.
Table 6 also shows that urbanization has a direct impact on the decrease in air quality in its own region by 0.2269%, ceteris paribus.Although neighboring regions do not directly experience spillover effects, it results in a national decrease of 0.219%, ceteris paribus.The direct effect of increasing per capita income can lead to a decrease in air quality in its own region and an improvement in air quality in neighboring regions.The import variable does not have a direct impact in its own region but provides a positive spillover effect in neighboring areas.
Changes in air quality will not be felt by provinces in Quadrant 2 and Quadrant 4. Changes in air quality will occur in provinces in quadrant 3, which will experience a direct impact of a 0.06% decrease in air quality, ceteris paribus, in their own region (Table 6).Meanwhile, the spillover effect is felt by provinces in quadrant 4, with an increase in air quality of 0.085%, ceteris paribus.Provinces in quadrant 3 have low per capita income and poor environmental carrying capacity, resulting in a low ability to absorb air pollution that occurs during development.This causes provinces in Quadrant 3 to have worse air quality compared to provinces in Quadrant 2 or 4 due to the insufficient natural ability to improve air quality.On the other hand, in provinces in quadrant 4, although there is no change in air quality in their own region, neighboring regions will experience spillover effects or indirect impacts.If the environmental carrying capacity in quadrant 4 is good, the air quality in the surrounding areas will also improve.

Conclusion
Indonesia's commitment to improving air quality requires synergy from the regions.Java and Bali Islands play a crucial role in the national economy.The concentration of population in these regions makes air quality improvement a priority.Spatial effects on air quality arise due to economic activities and changes in per capita income in a region.The presence of environmental carrying capacity to support economic growth also determines the quality of air, both in the region itself and in the neighboring areas.
The high level of industrialization in Java and Bali Islands has led to rapid and extensive land conversion, resulting in a decline in air quality.In further research, it is hoped that the role of local governments in improving air quality can be depth explored.