Energy , CO 2 , and AQI Efficiency and Improvement of the Yangtze River Economic Belt

With the rapid development of its economy, environmental governance is becoming more important in China. The Yangtze River Economic Belt (YREB), as the world’s largest inland shipping channel, can lead the country’s regional green economy development. As most research on China’s environmental efficiency focuses on provinces or the east and west regions, this paper examines its energy input and output and environmental effects from the aspects of YREB and non-YREB, breaking through the limitations of previous studies that only used cross-section or panel data for environmental assessment. This paper employs the meta-frontier dynamic SBM model, selects fixed assets as carry-over indicators, and considers the interrelationships between the dynamics variables during 2014–2016. The results are as follows: The overall energy efficiency and CO2 emission efficiency of YREB are higher than those of non-YREB. The difference in energy consumption, CO2, and AQI efficiency is large, but the performance of YREB is generally better than that of non-YREB. After setting the meta-frontier, non-YREB is better than YREB, for the main reason that the technology gap values of YREB are smaller than those of non-YREB. Our findings thus suggest that YREB should strengthen technical exchanges and promotion within its region, thereby decreasing regional technology differences, while non-YREB should address environment protection and CO2 emissions and advocate a low-carbon production mode.


Introduction
China's government and its citizens have gradually become more and more concerned about the environment in recent years.However, the results of China's efforts at environmental governance are not significant.According to the Ministry of Environment and Ecology, in 2017, the Air Quality Index (AQI) of 239 cities was greater than 100, with 70.7% of cities having air pollution exceeding the standard, while 99 cities' AQI was between 0-100.It is clear that China needs to put forth much greater efforts to improve domestic air quality.
The Yangtze River Economic Belt (YREB) covers a broad economic area that borders China's largest river.In 2016, the Yangtze River's shipping volume was four times the amount on the Mississippi River and six times that on the Rhine.YREB has rich resources, convenient water and land transportation, and dense urban areas, and plays an important role in promoting economic development and green development.From 2014 to 2016, the China government issued the "Guiding Opinions on Promoting Energies 2019, 12, 647 3 of 17 taking energy consumption, employees, and government expenditures as inputs; GDP (desirable) and AQI (undesirable) as outputs; and assets as the carry-over variable to assess intertemporal efficiency from one period to the next period.This paper offers further innovation in the literature by embedding AQI and technology gap variables into the meta-frontier dynamic SBM model to conduct an efficiency evaluation.
For the assessment of energy efficiency, most research studies utilize data envelopment analysis (DEA) models (such as DEA, Malmquist index, and Tobit model (Lv et al. [3]); SBM-DEA (Choi et al. [20]); DEA: Multi-directional Efficiency Analysis (MEA) (Wang et al. [21]); DEA-VRS (complete and window analyses) and Tobit regression (Zhang et al. [22]); and DEA-Luenberger productivity index (Chang and Hu [23])).For static models, Wang et al. [24]) used small cities' data in 2008 in non-radial DEA to assess energy performance, by considering the differences among Chinese cities. Discussing the potential for energy conservation and emission reduction, they provided a theory for achieving China's 2020 energy conservation and emission reduction targets.Yao et al. [25] offered detailed information on the energy efficiency and carbon emissions performance in 2011, helping to overcome bias in energy and carbon performance evaluation.Yan et al. [26] used the non-radial Malmquist productivity index, which was adapted to model the changes in 3E productivity during 2011-2013 for the 30 Chinese administrative regions.Miao et al. [27] employed the slack-based measure method and an extended Luenberger productivity indicator to estimate and decompose the atmospheric environmental performance.
For dynamic analysis, Wang et al. [28] took the improved DEA model and considered desirable and undesirable outputs and energy and non-energy inputs in order to assess regional energy efficiency in China in 2013.Wang et al. [29] designed a new all factor energy efficiency indicator to explain the heterogeneity of production technologies in various provinces of China, showing that the difference in energy efficiency and production technology between 29 provinces from 2001 to 2010 is very significant.Guo et al. [12] addressed the efficiency assessment of regional energy conservation and emission reduction and constructed an input and output selection evaluation index system to measure energy savings and emission reduction in 30 provinces of China from 2011 to 2012.However, most of these works in the literature have the following two problems: (1) they simply consider the trend of the data, but fail to describe the interrelationships between the years; and (2) their comparability between different indicators is weak, along with a lack of overall indicators to evaluate various factors.Although Wang et al. [30] proposed the concept of comprehensive indicators, it was still limited to static models.
In terms of research samples in the literature, Apergis [31] selected OECD member countries, Chang [32] employed 27 EU members, Song et al. [4] researched energy efficiency and carbon emissions in BRICS, Cui [33] selected nine countries, Zhang et al. [22] looked at 23 developing countries, Choi et al. [20] utilized panel data from 30 provinces in China, and Honma and Hu [34] used data on 47 regions in Japan.For research on China's resource utilization efficiency and environmental efficiency, most scholars use provincial panel data, such as Wei et al. [35], Jia and Liu [15], Wang et al. [21], Wu et al. [6], Shen et al. [36], Li and Lin [37], Zou et al. [38], and Guo et al. [12].For research on regions, Tang [1] divided China into eastern, central, and western regions to study the input-output efficiency of resources from 2000 to 2015.As one can see, the literature has not considered samples separated into YREB and non-YREB, and thus we employ these two regions as the research object.

Dynamic DEA
Data envelopment analysis is based on the concept of the Pareto optimal solution, using linear programming to assess the relative efficiency of a decision making unit (DMU).Charnes et al. [39] provided the CCR model in 1978 that extended Farrell's theory to multiple inputs and outputs, naming it data envelopment analysis (DEA).Banker et al [40] proposed the BCC model in 1984 based on variable returns to scale (VRS).Tone [41] introduced the slack-based measure (SBM) and used non-radial models to deal with slacks in each input/output.The SBM model adds non-radial slacks and can avoid the question of radial proportionality.
To evaluate the efficiency for multiple carry-over periods, Färe and Grosskopf [42], Bogetoft et al. [43], Chen [44], Nemoto and Goto [45,46], and Park and Park [47] proposed putting the effect of internal linkages into Dynamic DEA.Tone and Tsutsui [48] extended the model to the Dynamic Slack-Based Measures (SBM) model and next proposed the weighted Dynamic SBM model.The Dynamic SBM model is divided into input-oriented, output-oriented, and non-oriented types.
Tone and Tsutsui [48] did not consider undesirable output and regional differences; thus, this paper uses a modified meta-frontier dynamics slack-based measure model.The model is introduced in the following section.

The Modified Meta-frontier Dynamic Slack-Based Measures (SBM) Model
This study utilizes panel data collected from 30 provinces in China, which represent the most developed cities within the Yangtze River Economic Belt (YREB) (11 provinces in total) and non-Yangtze River Economic Belt (non-YREB) (19 cities in total).Among them, we employ six variables: two are input variables, three are outputs variables, and one is a carry-over variable.The input variables are employees and energy consumed; the output variables are GDP (desirable), AQI (undesirable), and CO 2 (undesirable); and the carry-over variable is fixed assets.The structure of dynamic DEA is in the Figure 1.To evaluate the efficiency for multiple carry-over periods, Färe and Grosskopf [42], Bogetoft et al. [43], Chen [44], Nemoto and Goto [45,46], and Park and Park [47] proposed putting the effect of internal linkages into Dynamic DEA.Tone and Tsutsui [48] extended the model to the Dynamic Slack-Based Measures (SBM) model and next proposed the weighted Dynamic SBM model.The Dynamic SBM model is divided into input-oriented, output-oriented, and non-oriented types.
Tone and Tsutsui [48] did not consider undesirable output and regional differences; thus, this paper uses a modified meta-frontier dynamics slack-based measure model.The model is introduced in the following section.

The Modified Meta-frontier Dynamic Slack-Based Measures (SBM) Model
This study utilizes panel data collected from 30 provinces in China, which represent the most developed cities within the Yangtze River Economic Belt (YREB) (11 provinces in total) and non-Yangtze River Economic Belt (non-YREB) (19 cities in total).Among them, we employ six variables: two are input variables, three are outputs variables, and one is a carry-over variable.The input variables are employees and energy consumed; the output variables are GDP (desirable), AQI (undesirable), and CO2 (undesirable); and the carry-over variable is fixed assets.The structure of dynamic DEA is in the Figure 1.Since this study considers undesirable output and regional differences in the dynamic SBM model, we can modify Tone and Tsutsui's [48] dynamic SBM model (See Appendix A) to make it the undesirable output in the meta-frontier dynamic SBM model.Battese and Rao [49] and Battese et al. [50] proposed a meta-frontier model that can be compared with the technical efficiency of different groups.O'Donnell et al. [51] set up a meta-frontier model that can accurately calculate the metafrontier efficiencies and group efficiencies.However, most of DEA usually assumes that all DMUs have the same level of technology, but the assessed DMUs are often in different geographical locations or under different national policies, resulting in different technology levels.Therefore, based on the Tone and Tsutsui [48] Dynamic SBM, the O'Donnell et al. [51] meta-frontier model, and undesirable output, we set up the model as follows: (  Since this study considers undesirable output and regional differences in the dynamic SBM model, we can modify Tone and Tsutsui's [48] dynamic SBM model (See Appendix A) to make it the undesirable output in the meta-frontier dynamic SBM model.Battese and Rao [49] and Battese et al. [50] proposed a meta-frontier model that can be compared with the technical efficiency of different groups.O'Donnell et al. [51] set up a meta-frontier model that can accurately calculate the meta-frontier efficiencies and group efficiencies.However, most of DEA usually assumes that all DMUs have the same level of technology, but the assessed DMUs are often in different geographical locations or under different national policies, resulting in different technology levels.Therefore, based on the Tone and Tsutsui [48] Dynamic SBM, the O'Donnell et al. [51] meta-frontier model, and undesirable output, we set up the model as follows: Energies 2019, 12, 647 5 of 17 (1) Meta-frontier (MF) We assume all units (N) are composed of DMUs in g groups (N = N 1 + N 2 + . . .+ N G ); y sj and x ij indicate output item s (s = 1, 2, . . ., S) of item j (j = 1, 2, . . ., N) and input item i (i = 1, 2, . . ., m); Z indicates desirable carryover (i = 1, 2, . . ., n); g is desirable output; and b is undesirable output of item j (j = 1, 2, . . ., N) under the meta-frontier.The meta-frontier k of DMU efficiency can be solved by the following linear programming (LP): Using Equations ( 1) and ( 2), we can find the overall technical efficiency (OTE) value of all DMUs under the meta-frontier model.
(2) Group-frontier (GF) Each DMU under the group frontier chooses the most favorable final output weighted, so that the efficiency of the DMUs under the group frontier can be solved by the following equation: (3) Technology gap ratio (TGR) Since the meta-frontier model contains g groups, the technical efficiency of the meta-frontier (MFE) will be less than the technical efficiency of the group frontier (GFE).The ratio value, called the technology gap ratio (TGR), is shown as:

Energy, AQI, and CO 2 Efficiencies
This paper follows the Hu and Wang [52] total-factor energy efficiency index to overcome any possible bias in the traditional energy efficiency indicators.There are three key features of this study: energy efficiency, AQI efficiency, and CO 2 efficiency.In this study, we note that i represents area and t represents time.
By definition, energy efficiency is the ratio of target energy input to actual energy input.AQI or CO 2 efficiency is the ratio of target AQI or CO 2 undesirable output to actual AQI or CO 2 (i, t) .
If the target energy input is equal to the actual input level, then the energy efficiency equals 1, representing efficiency.If the target energy input is less than the actual input level, then the energy efficiency is less than 1, representing inefficiency.
If target AQI or CO 2 undesirable output is equal to the actual AQI or CO 2 undesirable output level, then the AQI or CO 2 efficiency equals 1, representing efficiency.If target AQI or CO 2 undesirable output is less than the actual AQI or CO 2 undesirable output level, then the AQI or CO 2 efficiency is less than 1, representing inefficiency.

Data and Variables
This study uses panel data for 30 provinces that represent most provinces of China (except for Xizang).We collected data for 2014-2016 from the Statistical Yearbook of China [53], Demographics and Employment Statistical Yearbook of China, and Statistical Yearbooks from all cities [54].We took air pollutants data from China Environmental and Protection Bureau reports [55].Input, output, and carry-over variables are as follows.

Input variables:
Labor input (em): number of employees in each city at year-end.Unit: persons.Energy consumed (com): total energy consumed in each province.Unit: 100 million.

Output variable:
Desired output (GDP): each province's GDP is taken as its output.Unit: 100 million RMB.Undesired output: AQI: Air Pollution Index; the pollutants in the evaluation are SO 2 , NO 2 , PM 10 , PM 2.5 , O 3 , CO, and six other items.Unit: µg/m 3 .CO 2 : data on CO 2 emissions for each city are estimated from the energy consumption breakdown by each fuel category.

Carry-over variables:
Fixed assets: capital stock of each city is calculated by fixed asset investment of each province.Unit: 100 million RMB.

Input-output Indicator Statistics
The column of "Input variables" in Table 1 illustrates numerical analyses of labor, energy, and fixed assets.From 2014 to 2016, the average, maximum, and minimum values of labor do not change a lot.While there is a large difference in labor between provinces, there is no narrowing trend.Average energy consumption presents a decreasing trend from 2014.The maximum value of energy input has increased year by year, the difference in energy consumption between provinces is large, and the standard deviation displays no significant change.The average, maximum, minimum, and standard deviation of fixed assets are all increasing year by year.Capital investment in the provinces is increasing, while the gap among the provinces is gradually widening.The column of "output variables" in Table 1 shows data statistics for the output variables GDP, CO 2 , and AQI.GDP increases year by year from 2014 to 2016.While maximum GDP growth is obvious, the minimum value changes little, and the standard deviation increases year by year.There is a large gap in GDP growth and a clear imbalance in regional development.The mean and minimum value changes of CO 2 emissions are small, and the maximum value and standard deviation increase year by year.The average value of AQI from 2014 to 2016 decreases year by year, with the maximum value being the smallest in 2015.The standard deviation also decreases year by year.The AQI gap among provinces is narrowing.

Results
This study uses MAXDEA 7.0 to evaluate the energy consumption and AQI emission efficiency of 30 provinces in China.First, we calculated the total energy consumption, air pollutant emissions, and dynamic environmental efficiency scores of all provinces.Second, according to the division into YREB and non-YREB, we calculated the dynamic efficiency values of the two regions separately.Finally, we compared the efficiency values of each province in a region with the efficiency values in all provinces.This ratio provides room for improvement in environmental efficiency for each province.
Table 2 shows the overall efficiency scores and ranks for energy and air pollutant emissions in the 30 provinces from 2014 to 2016.The provinces of YREB are marked in red, while the others are non-YREB provinces.Beijing, Guangdong, Hunan, Jiangsu, Neimenggu, Shandong, Shanghai, and Tianjin have energy efficiency values of 1, indicating that these cities have a higher energy output efficiency.Hunan, Shanghai, and Jiangsu are cities of YREB, while Beijing, Guangdong, Neimenggu, Shandong, and Tianjin are non-YREB cities. Gansu, Xinjiang, Ningxia, and Shanxi scores are low, indicating a poor energy consumption efficiency.
The provinces with increased efficiency scores are Fujian, Guizhou, Jilin, Liaoning, Qinghai, Sichuan, and Yunnan.Within YREB, Sichuan has the largest increase, from 0.62 in 2014 to 0.76 in 2016, representing an increase of 22.6%.Guizhou and Yunnan experience a small rise, with an efficiency increase of 0.04.Within non-YREB, the scores of Fujian, Jilin, Liaoning, and Qingha increase slightly.
The overall efficiency of energy consumption in YREB is generally higher than that in non-YREB.Except for Zhejiang, Jiangxi, Hubei, Chongqing, and Anhui, the provinces within YREB all rank better.Within YREB, Shanghai, Hunan, and Jiangsu have the highest overall energy efficiency at 1.The overall efficiencies of Zhejiang, Anhui, Jiangxi, and Yunnan are low and need improvement.
YREB is large.Shaanxi presents the biggest difference, with a CO2 mean value close to 0 and AQI mean value of 0.98.The differences in Shanxi, Liaoning, and Heilongjiang are also large.Conversely, the differences between the AQI scores and CO2 scores in YREB are small.Non-YREB displays a large difference in CO2 efficiency scores, but the difference in AQI scores is relatively small.Compared to YREB, the difference in AQI scores for non-YREB is large.Table 4 shows the technology gap and ranks for each region.The regions with no technology gap from 2014 to 2016 are Beijing, Guangdong, Hunan, Jiangsu, Neimenggu, Shandong, Shanghai, and Tianjin, comprising five regions in non-YREB and three in YREB.The technology gaps of the observed regions are mostly higher than 0.9, and only Zhejiang in YREB is below 0.9.
The technology gaps of Anhui, Fujian, Guangzhou, Henan, Hubei, Jiangxi, Liaoning, Ningxia, and Zhejiang are all decreasing along with their ranks.The technology gaps of Anhui and Hebei drop respectively by 2.44% and 3.37% in 2014, indicating that their actual technology evaluation value is far from the target value.Gansu, Guizhou, Hainan, Hebei, and Jilin have higher technology gaps.In 2016, their technology gaps rose by 1.50%, 5.98%, and 1.39% versus 2014, respectively.
The mean value of the technology gaps in non-YREB is greater than that of the provinces in YREB.After setting the meta-frontier, the technology gap of YREB is not far from the frontier.Moreover, the standard deviation in the technology gap of YREB is 0.9971, or greater than 0.9948 for non-YREB.This shows evidence of greater intra-regional differences.

Conclusions
From above analysis, we present the following notable findings: 1.The total efficiency of energy utilization in YREB is higher than that in non-YREB, and regional differences are more obvious; 2. In terms of energy consumption, CO 2 , and AQI efficiency, YREB is better than non-YREB.YREB has fewer provinces with a lower energy consumption, CO 2 , and AQI efficiency.The number of low-score regions in non-YREB is large, and the scores of many provinces there are decreasing year by year; 3. The changes in AQI mean values and CO 2 scores are the same in YREB and non-YREB.However, the changes in CO 2 efficiency scores in YREB are smaller than those in non-YREB.Furthermore, YREB has a better CO 2 efficiency performance than non-YREB; 4.After setting the meta-frontier, non-YREB is better than YREB for two main reasons: the technology gap values of non-YREB are smaller than those of YREB, and the standard deviations of these values for non-YREB are smaller than those of YREB.Therefore, YREB should address the unbalanced development within its region.
This paper offers the following policy suggestions to central and regional governments and related authorities.
The non-Yangtze River Economic Belt should learn from YREB in terms of energy efficiency, emission reduction, and ecological restoration.Non-YREB should also adjust its economic structure and energy structure, focus more on talent recruitment, and develop new technologies to improve the overall energy efficiency.
Non-YREB economies, especially those in the three northeastern provinces, should greatly restrict high-pollution and high-emission industries and change the economic growth mode away from an over-reliance on resources.These three provinces should also cultivate energy-saving and environmental protection industries and look to narrow their gap with YREB.
The non-Yangtze River Economic Belt should also set up better methods for CO 2 emissions reduction and advocate low-carbon production technologies.The government should encourage the development and use of new energy sources such as solar energy, tidal energy, hydro energy, and wind energy through new policies and tax abatement.
There is a large technology gap between the provinces in YREB, especially for Zhejiang, Sichuan, and Guizhou.It indicates that these provinces have low levels of technological development and should comprehensively upgrade the skills and talents of their overall labor force.

Figure 1 .
Figure 1.The structure of dynamic DEA.

Figure 1 .
Figure 1.The structure of dynamic DEA.

Figure 2 .
Figure 2. (a) Average AQI and CO2 scores of the Yangtze River Economic Belt.(b) Average AQI and CO2 scores of the non-Yangtze River Economic Belt.

Figure 2 .
Figure 2. (a) Average AQI and CO 2 scores of the Yangtze River Economic Belt.(b) Average AQI and CO 2 scores of the non-Yangtze River Economic Belt.

Figure 3 Figure 3 .
Figure3shows the changes in efficiency scores and the mean values of AQI for YREB and non-YREB from 2014 to 2016.The changes in the AQI mean values are generally consistent with the changes in the CO2 efficiency scores.The changes in the CO2 efficiency scores of YREB are smaller than those of non-YREB, and the CO2 efficiency scores of most regions rose from 2014 to 2016.Guangxi, Liaoning, Fujian, and Qinghai in non-YREB and Sichuan and Yunnan in YREB all have higher scores.

Figure 3 .
Figure 3. (a) CO 2 efficiency scores and AQI mean values of the Yangtze River Economic Belt.(b) CO 2 efficiency scores and AQI mean values of the non-Yangtze River Economic Belt.
undesirable output.The energy and AQI or CO 2 efficiency models are defined as: Target AQI or CO 2 Undesirable output (i, t) Actual AQI or CO 2 Undesirable output

Table 1 .
Statistical presentation of input-output variables.

Table 2 .
Overall efficiency scores and ranks.

Table 4 .
Regional ranks and technology gaps.

Table 5
Lists the Wilcoxon test scores for the average technology gaps.In 2014-2016, the average gap of YREB and the average gap of non-YREB both pass the significance test.

Table 5 .
Wilcoxon test scores of average technology efficiency gaps.From the two-tailed test, the confidence interval 0.05 is significant.