Environmental performance of Malaysia’s air pollutants based on data envelopment analysis with slack-based measure and Malmquist productivity index

Air pollutants have a significant impact on humans and the environment, making their reduction and mitigation crucial and requiring attention from policymakers. In this study, the impact of air pollutant concentrations, namely particulate matter with an aerodynamic diameter less than 10 μm (PM10), particulate matter with an aerodynamic diameter less than 2.5 μm (PM2.5), sulphur dioxide (SO2), nitrogen dioxide (NO2), ozone (O3) and carbon monoxide (CO) on the environmental performance of 15 states in Malaysia was examined using available data from 2018 to 2021. The analysis was performed using data envelopment analysis (DEA) with slack-based measure (SBM) and the Malmquist productivity index (MPI). The efficiency values were used for principal component analysis (PCA) to infer the influencing factors that are highly redundant or dependent on each other. Results from SBM-DEA suggested an important aspect of gross domestic product toward efficiency score, where high efficiency values were observed for Selangor, Kuala Lumpur, and a small state like Perlis (efficiency value of 1.000). The MPI score indicated that the performance of each state was relatively low for the years 2019–2020, which suggested a regression in performance productivity due to the haze episode. PCA analysis showed that there were two factors, where the highest contribution for Factor 1 was Melaka and Johor with an average contribution of 8.15% and for Factor 2 was Perlis and Perak with an average contribution of 23%. This study’s findings showed that air pollutants play significant roles in achieving good environmental performance in order to tackle global issues such as global warming and climate change.


Introduction
Environmental problems are becoming major issues, and reduction of pollution is one of the mainstream agendas.From year to year, many industries and sectors have recognised and adapted to the need for environmental protection, one of which is the evaluation of performance.On the other hand, measuring and reviewing environmental performance is vital for developing environmental policy and plans (Matsumoto et al 2020).There are several available approaches to determining the environmental impact and performance of a system or product.One of the methods for assessing environmental performance is through a life cycle assessment approach, which could resolve the influences of society, environment, and economy within a sustainable development system, while evaluating various environmental impacts (Liu et al 2019, Böckin et al 2022).This approach can be complemented by other performance evaluations such as data envelopment analysis (DEA).
The DEA methodology is a type of linear programming used to assess the efficiency of multiple decision-making units (DMUs) in situations where the production process involves multiple inputs and outputs (Jin et al 2014, Zhou et al 2019, Li et al 2020).DEA has been applied for evaluating environmental efficiency in various fields, such as air pollution (Li et al 2020, Moutinho et al 2020, Zhao et al 2022), CO 2 emissions (Bian et  Several key benefits of DEA: ability to handle multiple inputs and outputs; it does not make any inherent assumptions about the underlying function that connects inputs and outputs, making it applicable in various situations; it is also modifiable; and it can also be applied to big data for precise estimations (Liu et al 2019, Matsumoto et al 2020).
The DEA approach, combined with other effective models has been explored for the past 20 years.One of the models that has the most representative and usually performed for DEA studies is the slackbased measure (SBM) (Zhang et al 2016).SBM is applied to investigate the efficiency based on the slack values, where the SBM is usually used to evaluate the context that produces an appropriate stratification of the DMU performance levels (Liu et al 2017, Tian et al 2020).SBM-DEA is particularly fitted for dealing with undesirable outputs in environmental efficiency evaluation, and this approach can be applied in many related fields that have undesirable outputs (Wang et al 2020, Zhang et al 2021).Additionally, DEA studies also considered the evaluation of productivity changes, for which the Malmquist productivity index (MPI) approach is prevailing.MPI measures efficiency performance and productivity growth combined with DEA, where this approach is more comprehensive and assesses the productivity change based on the changes in the technical performance and efficiency frontiers over time (Huong et al 2021, Liu et al 2021, Khoshroo et al 2022).Hence, the measurement of efficiency that is related to time evaluation with targeted outputs such as air pollution was also the main interest of DEA.
The performance of cities in reducing air pollutants plays a significant role, as reducing the environmental impact of cities is one of the targets in the sustainable development goals.Based on the severity of air pollution and its harmful effects on human health, it is important to have air pollution reduction strategies, especially for urban and city development.(2020), after assessing the environmental performance of cities or local governments with regard to air pollution, adjustments to local government policy can be suggested, such as improving and enhancing industrial structures based on geographical and climatic conditions, using more advanced management techniques, and optimising their environmental governance resources to suit local conditions.
Due to the impact of various factors on the level of air pollutants, a detailed evaluation with strategic mitigation needs to be performed not only at the scale of a country but also at the state and city levels.The evaluation of air pollutants with input parameters that have a significant impact on the environmental equilibrium can be part of the DEA analysis.Thus, this study aimed to investigate the efficiency of each state in Malaysia in terms of air pollution levels, using vehicle numbers, population, and meteorological parameters as input parameters from 2018 to 2021.As air pollutant variability can be affected by various pollution source, thus this study investigates the selection of input and output parameters, focusing on air pollutant levels as the primary output.This paper also evaluated the relationship between input and output parameters in evaluating the environmental efficiency of air pollutants in each state in Malaysia.

Study area
In this study, all states in Malaysia, covering Peninsular Malaysia and East Malaysia, were chosen.The 15 states were Perlis, Kedah, Pulau Pinang, Perak, Selangor, Negeri Sembilan, Melaka, Johor, Pahang, Terengganu, Kelantan, Sarawak, Sabah, Putrajaya, and Kuala Lumpur; each state has its own capital city.Overall, the total population of Malaysia is 32.7 million, with an annual growth rate of 0.2% (DOSM 2021).Malaysia is a developing country with a mix of industries, including agriculture, mining, manufacturing, services, and other economic activities.The location of Malaysia in a tropical area gives the country a climate of uniform temperature, high humidity, and copious rainfall (METMalaysia 2022).Malaysia also has monsoon weather conditions where the northeast monsoon season is the main rainy season while the southwest monsoon is the dry period that occurs between May and September (Othman et al 2022).During the southwest monsoon, especially during dry and hazy conditions, the southern and coastal areas of Peninsular Malaysia are facing low levels of air quality, particularly due to biomass burning activity.Among all states, Kuala Lumpur had the highest population with the greatest numbers of motor vehicles.This is attributed to the impact of economic development, particularly the predominant economic activities are the service sector that relate to wholesale and retail trade; finance and insurance; and information and communication (DOSM 2022).According to MOT (2020), Kuala Lumpur had the highest number of vehicle number registrations, especially for motorcycles and motor vehicles, compared to other states.The rapid development in Kuala Lumpur also affects the neighbouring states of Selangor and Negeri Sembilan, where development of residential areas and economic activities are active.Other states have their own economic activities such as mining and quarrying, which are dominated by Sarawak and Sabah, while states such as Pulau Pinang, Melaka, Negeri Sembilan, and Selangor are active in manufacturing (DOSM 2022).

Data collection
Hourly data of air pollutant concentrations (PM 10 , PM 2.5 , SO 2 , NO 2 , O 3 , and CO) and meteorological parameters (wind direction, wind speed, relative humidity, solar radiation, and temperature) from 1 January 2018 until 31 December 2021 were obtained from the Department of Environment, Ministry of Natural Resources, Environment and Climate Change.The selection of air monitoring stations was based on the criterion that each station must be located within a city in order to represent the ambient air concentration for each state.In total, there are 65 air quality monitoring stations distributed across Malaysia, categorised as urban, suburban, or rural stations.For this study, we chose 15 monitoring stations in each state.All selected stations were urban except for Perlis, Kedah, Pahang, Kelantan, Sabah, and Putrajaya, which were classified as sub-urban stations.The description of air monitoring station in each state in Malaysia that was applied in this study is detailed in supplementary text 1 and table S1.Each state was also known as DMU in this study.
Several social-economic data were accessible online or via annual reports for each department and ministry in Malaysia.However, most of the time, the data was tabulated for Malaysia as a whole and was unavailable for individual states.Thus, this study only gathered data on total population and gross domestic product (GDP) by states that was obtained from yearly reports by the Department of Statistics Malaysia (DOSM 2021(DOSM , 2022)).While data on the total number of vehicles on the road for each state was collected from the Ministry of Transport Malaysia (MOT 2020).All the data were analysed on an annual basis from 2018 to 2021.

DEA modelling 2.3.1. Input and output selection
A DEA model is known to be used to assess the efficiency of several DMUs, where the model's inputs and outputs must be identified.The application of the DEA model in this study was anticipated to provide the environmental efficiency value of each state/DMU based on the selection of input and output parameters.The input parameters used in this study were the number of populations, the total number of vehicles on the road, meteorological parameter, GDP and air pollutants (PM 10 , PM 2.5 , SO 2 , NO 2 , O 3 and CO) (see table S2).To ensure the accuracy of the data required for DEA analysis, both input and output parameters were considered as options (Jia and Yuan 2017) where most of the studies have applied input and output parameters such as energy, population, capital and GDP when evaluating the environmental efficiency of air pollutants (Lee et al 2014, Li et al 2020, Matsumoto et al 2020, Wu et al 2021).The selection of output parameters also took undesirable outputs into account, such as various environmental pollutants and air emissions, which would be generated as by-products and were unavoidable while GDP was the desired output (Zhang et al 2016, Khoshroo et al 2022).The importance of having socio-economic data, such as GDP, lies in its ability to reflect the economic development level of a selected area.GDP is also commonly used as a determinant of environmental efficiency (Wang and Wang 2019, Wang and Chen 2020).Hence, the undesirable output for this study was chosen to be all studied air pollutants.
In order to evaluate the environmental efficiency of each state/DMUs based on different inputs and outputs, scenario analyses were performed, and three scenarios were suggested.The computation of the SBM-DEA did not directly look at the level of air pollutant concentration, but this analysis considered all possible impacts, such as inputs and desired outputs.The scenarios were as follows: 1. Scenario 1 was aimed at evaluating the effect of input parameters that were the number of populations, the total number of vehicles and meteorological parameters (wind speed, wind direction, solar radiation, temperature, and relative humidity) to air pollutants and GDP.Thus, the output was GDP and undesirable outputs were all air pollutant parameters (PM 10 , PM 2.5 , SO 2 , NO 2 , O 3 and CO). 2. Scenario 2 was aimed at evaluating the effect of population and the total number of vehicles as inputs.The output was selected to be GDP, and the undesirable outputs were SO 2 , NO 2 and CO. 3. Scenario 3 was aimed at evaluating the effect of several meteorological parameters (temperature, relative humidity, and solar radiation) on GDP which the undesirable outputs were PM 10 and PM 2.5 .Because of year 2019 was a significant year with increased PM 10 and PM 2.5 concentrations, the effect of meteorological factors was investigated.

SBM-DEA model
The DEA model analysis emerged with a new approach to suit its application in certain fields of study.The SBM-DEA incorporated slack variables, addressed scheduling issues, and was more effective in handling undesirable outputs (Zhang et al 2021).
The SBM-DEA approached applied in this study were described in supplementary text 2. After calculating the efficiency of each DMU, we employed sensitivity analysis to assess the robustness and validity of the efficient scores obtained from the SBM-DEA analysis.The technique used in the sensitivity analysis involved eliminating efficient DMUs from the reference set.The detailed process of the sensitivity analysis is presented in supplementary text 3.

Malmquist productivity index (MPI)
The MPI was used to investigate the productivity growth and technical progress in a production system (Färe et al 1994, Zhang et al 2021).Moreover, as MPI is suitable for determining the performance of a system over time, the application of this method helps in evaluating the effects of air pollutants from year to year.This study had applied MPI calculation to investigate each state/DMUs performance, efficiency change (EC) and technological change (TC) for different scenarios (see supplementary text 4).

Statistical analysis and data interpolation
A descriptive analysis was performed to explore the state/DMUs-specific characteristics of air pollutants, where trend analysis of air pollutant concentration was plotted against time using the openair package of R software (Carslaw and Ropkins 2012).Correlation matrices were performed using the corrplot package, and principal component analysis (PCA) was analysed using the FactoMineR package, both packages utilising R software.Modelling of SBM-DEA and MPI was carried out using the deaR package of R software.Spatial distributions of SBM-DEA efficiency results were also mapped using ArcMap 10.8 (ESRI Inc., Redlands, CA, USA).

Variations in air pollutants
The Comparing the yearly average of air pollutants with the New Malaysia Ambient Air Quality Standard for Interim Target 2018 and 2020, both PM 10 and PM 2.5 were below the yearly standard, while the highest concentration was recorded in 2019 (table S2).The trend concentration for the daily average of SO 2 showed that this pollutant's concentration was always below the 24 h average of the standard, while certain days in Selangor and Johor exceeded the standard for NO 2 (figure 1(b)).The high concentration of NO 2 is believed to be primarily contributed by vehicle emissions, especially during rush hours in the morning and late afternoon, particularly in economically active areas such as Kuala Lumpur (Azhari et al 2021).For O 3 and CO, most of the time the concentrations were lower than the standard, but there was a peak concentration that exceeded the standard, while for all studied years, the highest yearly concentrations were 2019 (O 3 ) and 2018 (CO).Sarawak had a peak concentration for CO in second of 2019, which could be due to some local pollution in the surrounding area, while a peak concentration of all sizes of PM was observed in all states in August to early September 2019.
Overall, the trend of daily air pollutants in Malaysia was uniform and below the standard, except for PM 10 and PM 2.5 .The concentrations of PM 10 and PM 2.5 were highly influenced by local sources and contributions from long-range transport, such as haze episodes from biomass burning in Sumatra and Kalimantan (Sulong et al 2017, Latif et al 2018).High concentrations of PM between August and September 2019, particularly in dry weather conditions in Malaysia, which were contributed by biomass burning events that were also observed in a previous study by Othman et al (2022).While pollutant such as O 3 was observed to have vary daily concentration

SBM-DEA Efficiency for different scenarios
The SBM-DEA approach was applied for input and output data according to 15 states/DMUs in Malaysia, and the efficiency values were derived for three scenarios as listed in table 1.Based on the result, it showed that there was a difference in efficiency in each state/DMU for the years 2018-2021.With the input data of population, total number of vehicles, and meteorological parameters for scenario 1, it seemed that these input data had not much effect on air pollutants, where the efficiency values were clearly not showing much difference from year to year.There were several states/DMUs had the highest efficiency values that remained consistent from 2018 to 2021 such as Perlis, Selangor, Negeri Sembilan, Terengganu, Putrajaya, and Kuala Lumpur, which showed stable efficiency.Several state/DMU had reduced efficiency in 2019 and 2020, including Johor, Pahang, Sabah, and Sarawak, while Kelantan had reduced efficiency in 2020 with a value of 0.697 and then increased back in 2021 with an efficiency value of 1.000.It can be said that the efficiency value of 1.000 indicated the best use of input to maximise GDP and minimise air pollutant as an output.It showed that all input parameters applied in this study influenced the efficiency value of each state/DMU, especially GDP.The relationship of all input parameters in this study was also explored for its correlation using Pearson correlation matrix that shown that strong correlation of CO and NO 2 with GDP (see supplementary text 5 and figure S1).
For the second scenario (scenario 2), which focused on the input of population and total vehicle with outputs of SO 2 , NO 2 , and CO, the efficiency was quite different from year to year except for Perlis and Putrajaya, which had values of 1.000 for every Scenario 3, which modelled the effect of input data for meteorological parameters such as solar radiation, temperature, and relative humidity on the PM 10 and PM 2.5 was observed to be different for in each state/DMU and year, where only Selangor had an efficiency value of 1.000 in all of the years.Reduced efficiency values were recorded for certain state/DMU for 2019.In comparison to 2018, Perlis had a reduction of 17%, Negeri Sembilan had a reduction of 64%, Johor had a reduction of 10%, Pahang had a reduction of 52%, Terengganu had a reduction of 9.6%, Sarawak had a reduction of 10%, and Putrajaya had a reduction of 10%.Positive increase in efficiency values from 2019 to 2021 were observed; only five states/DMUs had reductions in efficiency values in 2021 compared to 2020 but, the reductions were still small (ranging from 6% to 31%).
Figure 2 depicts the overall average efficiency for the first, second, and third scenarios, spatially.It is clear that each state/DMU had its own efficiency, which was affected by the computation of input and output parameters, as shown in figure 2(a), where the range of the average efficiency value was large, indicating that most of the state/DMU had efficiency values greater than 0.500.However, when looking at different input parameters such as those in scenario 2 (figure 2(b)), which specifically evaluated the performance of each states/DMU on the emission of gases (SO 2 , NO 2 , and CO), it showed that the efficiency values had a large gap, with only two states/DMUs having efficiency = 1.000.Scenario 3 (figure 2(c)) which intended to show the effects of meteorological parameters on the efficiency of PM 10 and PM 2.5 , provided the information that high efficiency values were recorded for states/DMUs such as Sabah and Sarawak and was the same for smaller sizes of state/DMU like Kuala Lumpur, Putrajaya, and Selangor.There was no clear indication to relate the haze episode in 2019 to the efficiency value of all state/DMU, where in some state/DMU, the efficiency was still higher in year 2019 compared to the previous year.Overall, the SBM-DEA values obtained in all three scenarios are statistically different (p < 0.001), indicating that input and output parameters have a significant impact on the DEA analysis.Therefore, the selection of both input and output parameters needs to be made carefully for each scenario.
As the SBM-DEA approach used input and output parameters in the computation of efficiency, where GDP was the desired output, GDP played a significant role in indicating the best efficiency value.For example, the state/DMU that had high values of GDP, such as Kuala Lumpur and Selangor, had high efficiency as both state/DMU had the higher high rate of economic activity, especially in scenario 1.As reported by Matsumoto et al (2020), Zhang et al (2016), the output parameter, such as GDP, played a significant role in determining the efficiency score; nevertheless, the selection of environmental variables was also a critical aspect for the DEA related study.Thus, the increment of undesirable output could also be linked with the increase of input (Tian et al 2020).A high efficiency value was also suggested to indicate that the particular DMU had performed environmental mitigation, while a low efficiency suggested that there was inefficiency in the economic and social dimensions (Lee et al 2014, Tian et al 2020).

MPI for different scenarios
The changes in the MPI for 15 states/DMUs for scenario 1, 2, and 3 are shown in figure 3  in MPI values, with seven state/DMU had MPI >1.000 (figure 4), indicating productivity improvement, while other state/DMU were catching up very well from the previous year, which recorded MPI slightly lower than 1.000.In scenario 2, improved performance from year 2018-2019 was observed for Perlis, Pulau Pinang, Negeri Sembilan, Melaka, Johor, Kelantan, Sabah, and Sarawak, with the highest improved performance of about 31% recorded for Pulau Pinang, where it was clear that the lowest MPI value was recorded for 2019-2020, except for Perak, where Perak showed declining performance for all studied years.A big jump in improvement was recorded for 2020-2021 compared to 2019-2020, when almost all state/DMU had MPI >0.800 except for Perak.The low value of MPI for Perak was clearly shown in figure 4, with a statically below 2.000 MPI value from 2018-2019 until 2020-2021.
Scenario 3 showed that almost all state/DMU had improved performance in 2018-2019, except for Kedah and Pahang, where both state/DMU recessed about 4% and 11%, respectively.Moreover, the trend of MPI clearly showed that the recessed performance was from 2019-2020 for all state/DMU that had MPI <0.100.While from 2020-2021, there were increased performances of all state/DMU, except for Pulau Pinang, Sarawak, and Putrajaya, with MPI values of 0.985, 0.929, and 0.975, respectively.
Overall performance of each state/DMU for each scenario is shown in figure S2, where a higher MPI value was recorded for scenario 3.This result can be attributed to the input and output parameters used in scenario 3, which focused on only two outputs.Both Kelantan and Terengganu had MPI values more than 1.000, indicating increase productivity change for Scenario 3. The overall average MPI values for other states/DMUs were discussed in supplementary text 6.

EC and TC for different scenarios
The EC and TC values was obtained from MPI modelling where EC component measures how the efficiency of a system or entity has changed over time while TC component measures how technology or knowledge has progressed over time.The cutting point of EC was 1.000, where EC > 1 indicates relatively improved efficiency, EC < 1.000 indicates efficiency recession, TC > 1.000 indicates technological progression, and TC < 1.000 indicates technological regression (Wang et al 2020).Both EC and TC equalling 1.000 showed no efficiency and TC .
Values for EC and TC variation over time are shown in figure 4, where the EC values for each state/DMU did not correspond to the TC values.The highest value of EC was obviously indicated by Perak for scenario 2, where the TC and MPI values for Perak were relatively low.As Perak had low efficiency values for TC and MPI in 2019-2020, the EC would be higher, which showed the variability of the efficiency of Perak.Scenario 3 also recorded high variability of EC value, where several state/DMU had increased EC values except for Sarawak, while Pahang had tremendously an increased EC value, which was about 12% for 2019-2020 and 25% for 2020-2021.While for TC, all state/DMU had similar trends from year to year.Comparing between scenarios, Melaka, Perlis, and Selangor had recorded EC ⩾ 1.000, which showed efficiency in managing air pollutants.Comparison of average EC and TC values between scenarios was described in supplementary text 7 and table S4.The environmental performance of these state/DMU could be investigated with other suitable input parameters, such as energy consumption.The indication of EC would help in clearly knowing the efficiency of each DMU that was separated by innovation development, which provided information on efficiency performance individually (Firsova and Chernyshova 2020).High efficiency progress necessitates technical change in areas where improvement is required (Wang et al 2020).

PCA, k-mean clustering, and correlation matrix
PCA is an analysis that aimed to modify the original data into a few groups of principal component/factors, where the first component/factor has the greatest variance, followed by those with the second and third largest variances, and so on (Gunawardana et al 2012).Thus, this study modelled the efficiency values for SBM-DEA and MPI values for all the scenarios.Results of the PCA analysis that provided two factors with k-means clustering of the PCA values are shown in figure 5. Results for Factor 1 showed the contributions of Melaka, Johor, Negeri Sembilan, Pahang, Selangor, Kedah, and Pulau Pinang (8.20%, 8.19%, 8.05%, 7.91%, 7.71%, 7.62%, and 7.29%) while for Factor 2, the contributions were for Perlis, Perak, Putrajaya, Sarawak, and Pulau Pinang (26.6%, 20.3%, 13.3%, 12.6%, and 7.4%).Factor 1 primarily explains the variation in state/DMU that has high efficiency in SBM-DEA.Factor 1 had a clear indication that states such as Melaka and Johor, which had the highest contribution to this factor, were most efficient in dealing with air pollution.Moreover, because Perak had relatively low efficiency values in scenarios 1 and 2, it can be stated that Factor 2 is dominated by the state/DMU with minimum and moderate efficiency.
The cluster analysis, which was performed using the k-mean method, showed that the efficiency value of each state/DMU was divided into three clusters (figure 5(c)).State/DMUs that were grouped in cluster 1 were Perlis, Sarawak, Putrajaya, and Kelantan, while those in cluster 2 were Sabah, Selangor, Johor, Pahang, and Kuala Lumpur, and those cluster 3 were Perak, Pulau Pinang, Negeri Sembilan, Kedah, Terengganu, and Melaka.Overall, the state/DMUs that were grouped in clusters 1 and 2 were suggested to have almost similar efficiency values, whereas cluster 3 had state/DMU with a higher value of efficiency.However, this clustering method was applied for all scenario values, thus the state/DMUs did not have the same sequence as the SBM-DEA and MPI analyses.A study by Firsova and Chernyshova (2020) identified three clusters based on the DEA Malmquist Index analysis, where the author analysed 80 Russian Federation, and the k-mean clustering showed that the indication of each cluster was subject to economic and innovative development.

Conclusion and policy implications
In recent years, Malaysia has faced several economic and environmental issues where air pollution can also be affected by both problems.As Malaysia is a developing country, one of the constraints is taking action to reduce air pollution levels without compromising economic activities.In this study, the environmental efficiency of all states in Malaysia towards air pollutants was investigated using SBM-DEA and MPI approaches from year 2018-2021.The state/DMU that had the highest performance with efficiency values of 1.000 for SBM-DEA were Perlis, Selangor Negeri Sembilan, Terengganu, Putrajaya, and Kuala Lumpur for scenario 1, which recorded higher numbers of efficiency values compared to scenario 2 and scenario 3. Therefore, the results suggested that high numbers of input parameters would yield insensitive efficiency values, thus, choosing related input and output variables was a critical aspect of the DEA analysis.Results from MPI had shown fluctuation in productivity, with the lowest productivity value observed in 2019-2020, which clearly related to the contribution of poor air quality levels in Malaysia during that respective year.PCA analysis that combine the efficiency values of SBM-DEA and MPI had provide two factors where Melaka and Johor recorded highest contribution in factor 1; and Perlis and Perak for factor 2. These factors suggested that states/DMUs in factor 1 highly proficiently in managing air quality while moderately efficient for factor 2. In fact, some small state/DMU had low efficiency and performance compared to high-economic development state/DMU, which indicated that air pollution levels were not only controlled by economic activities.In conclusion, assessment of environmental performance is one of the available tools in decision-making and environmental management where air pollution levels are alarming in certain state/DMU that have rapid economic development.Hence, this study's results could provide insight to the policymakers to improve and manage environmental performance, especially air pollutants, at the local government scale, state level, and national level.
Air pollution reduction due to an environmental issue such as the COVID-19 pandemic was widely commented on in previous studies (Agami and Dayan 2021, Latif et al 2021, Othman and Latif 2021, Pal et al 2021), while other studies related to modelling air pollution and climate change mitigation also attracted some interest from several researchers (Zheng et al 2016, Zeng and He 2023).As suggested by Li et al

Figure 3 .
Figure 3. Malmquist productivity index value for each states/DMU based on different scenarios.
−3 was recorded by Sarawak and Sabah in 2019.It also showed that almost all states had peak of PM 10 and PM 2.5 concentrations between August and September 2019.The elevated concentrations of PM 10 and PM 2.5 were due to biomass burning events in Sumatra and Kalimantan, Indonesia, which led to a transboundary haze episode in the region.There was no distinct peak in SO 2 concentration in any of the states, and SO 2 concentrations were consistently lower compared to NO 2 .The concentration of NO 2 exhibited fluctuations, particularly in Johor, Kuala Lumpur, and Selangor, suggesting a significant contribution from high vehicle emissions in these cities.The variability of NO 2 concentration was clearly displayed by Johor, Kuala Lumpur, and Selangor.The daily trend of O 3 in figure 1(c) indicated that there was fluctuation in daily O 3 , with no clear difference in O 3 concentrations from 2018 to 2021 in all states.
trend of air pollutant concentrations from 2018 to 2021 across states/DMU is shown in figures 1(a)-(d).The trend concentration for PM 10 and PM 2.5 was similar through the observed four years, with the highest concentration, exceeding 200 µg m