1. Introduction
China’s economy has been reformed and opened up for more than 40 years and has made brilliant achievements in the economic field. However, since the reform and opening up, China’s economy has been in a rough economic development model for many years, making China’s environmental pollution problem increasingly serious. According to the World Health Organization (WHO), air pollution causes nearly 7 million deaths worldwide each year. As one of the most polluted countries globally, China has endured haze pollution, acid rain, and excessive greenhouse gas emissions that continue to plague the country in the fall and winter of recent years [
1,
2]. According to the Cost of Pollution in China 2007, a study completed by the World Bank in collaboration with the Chinese government, severe air pollution in China causes 350,000–400,000 deaths per year due to indoor air pollution [
3]. Currently, China is among the world leaders in PM
2.5, SO
2, NO
2, and greenhouse gas emissions. More seriously, in 2011 and 2015, several Chinese provinces experienced PM
2.5 explosions with pollution levels reaching up to 1155 μg/m
3, and serious environmental pollution problems will pose a significant threat to the health of residents [
4,
5]. Data in the 2017 China Ecological Environment Status Bulletin also show that the percentage of cities with excessive environmental pollution in China is as high as 70.7%. The rate of days with PM
2.5 and PM10 as the primary pollutants was 74.2% and 20.4%, respectively [
6]. Numerous pieces of evidence show that the increasing environmental pollution has induced a continuous increase in the rate of various chronic diseases among Chinese residents. The environmental pollution problem poses a significant threat to the quality of China’s economic development [
7,
8,
9].
In the face of severe environmental problems and the need to transform the economic development model, China has continued to transition to a greener economy in recent years. The key is to promote technological innovation to achieve pollution reduction and sustainable economic development [
10]. However, the answer to whether the promotion of technological progress can effectively reduce environmental pollution is inconclusive. By studying technological progress and environmental pollution in five developing countries—Brazil, Russia, India, China, and Mexico—scholars found that technological progress leads to increased pollution in developing countries when their GVCs are below a threshold. Otherwise such technological progress can reduce emissions [
11]. Some scholars have also analyzed Chinese industrial data and found that technological progress can reduce pollution [
12]. Some studies have shown that technological progress from different sources has different effects on pollution, with independent innovation having a significant inhibitory effect on haze pollution, while technology introduction exacerbates haze pollution [
13]. Neutral technological advances and labor-saving technological advances are beneficial to haze emission reduction. In contrast, capital-saving technological advances have insignificant effects on haze pollution, and energy-saving technological advances cannot effectively reduce haze pollution due to the energy rebound effect [
14].
The invention of the steam engine set off a wave of the industrial revolution. It also brought about quite serious pollution events such as the London toxic fog that caused thousands of deaths [
15,
16]. Some studies have shown that the economy’s structure based on fossil energy sources has also created certain carbon lock-in and energy rebound effects [
17,
18,
19,
20]. Some scholars have argued that the carbon lock-in effect is one of the major obstacles in the transition to a low-carbon economy and that carbon lock-in can lead to greater welfare losses in the absence of regulatory conditions [
18]. This was also found when analyzing the energy consumption behavior of urban households [
19]. This is because the path of industrial countries when building industrial systems is locked in the fossil energy-based energy system. Thus, technological progress will rely more on that path, and the scale effect of technological progress will make the production efficiency and production scale further expand. The rise in demand for energy leads to more pollution problems, so even if technological progress occurs, it is not always possible to reduce environmental pollution [
14,
20]. Technological progress pushing enterprises to expand their production scale, leading to more energy consumption, can bring more serious environmental pollution problems. The existing fossil energy-based technology system also forms a certain obstacle to the promotion and application of green technological progress. For example, studies of urban residential electricity consumption in China show that there is indeed an energy rebound effect [
21]. An analysis of the rebound effect of energy consumption in China shows that this phenomenon does exist in China, where technological advances have improved energy efficiency, but the rebound effect has made energy savings less effective than expected [
22]. Some scholars have shown by estimation that the rebound effect in China during 1981–2011 was between 30% and 40% [
23]. However, some scholars’ studies have shown that technological progress can significantly improve environmental quality and reduce environmental pollution. Unlike previous industrial revolutions, technological progress in recent years has been about efficiency improvements. The application of green energy sources such as photovoltaics and wind power has led to technological progress toward cleaner and energy-efficient technologies. China has made great progress in clean energy technology [
24], and studies have shown that the development and application of clean energy technology can effectively reduce pollution [
25,
26]. It has been found that through the analysis of the impact of China’s trade on technological progress, imports from developed countries such as Europe and the United States increase the rate of green technological progress. In contrast, imports from developing countries decrease the rate of green technological progress [
27]. Moreover, some scholars have studied the impact of environmental regulation, fiscal decentralization, and foreign investment on green technological progress in order to achieve a reduction in pollution emissions [
28,
29,
30,
31,
32].
However, in the face of the multiple pollutants embedded in the environmental pollution brought about by the rough economic development, can the promotion of technology effectively reduce the emissions of all pollutants? Is the relationship between the action of air pollutants represented by PM
2.5, SO
2, and CO
2 levels and technological progress consistent with the inverted U-shaped relationship of the environmental Kuznets curve [
33,
34]? Although there are studies related to the social causes of pollutants such as ozone and PM
2.5 [
35,
36], is there a bias of technological progress towards the reduction of different kinds of pollutants? There are few relevant studies on these questions, especially on whether there is a certain bias of technological progress towards the reduction targets of multiple pollutants. Related studies have mostly focused on the relationship between technological progress in input bias and individual pollutants [
29,
36]. Biased technological progress is currently a hot research topic in the academic community [
37,
38,
39,
40]. Most of the literature has focused on the bias of technological progress on input-based technological progress, with the aim of verifying whether the preference of input factors contributes to technological progress, without considering the issue of multiple outputs [
41,
42,
43]. Most of the available studies on output-based technological progress bias have focused on undesirable outputs represented by CO
2 and have not paid attention to the bias of technological progress for multiple types of undesirable outputs [
29,
44,
45,
46].
In this paper, based on the existing studies, we measure the green total factor productivity of Chinese provincial panel data and decompose the green technological progress from it; secondly, we verify the possible inverted U-shaped relationship between multiple air pollutants and technological progress. Then, this paper analyzes the bias of technological progress for emission reduction among different pollutants in detail to investigate whether there is a preference for different types of pollutants and whether there is heterogeneity in the bias of technological progress among different provinces in China. Finally, the selective bias of technological progress on pollution emissions is empirically analyzed by means of an econometric model to verify the analysis made in the previous paper.
Compared with previous studies, the marginal contributions of this paper are as follows. (1) Verifying whether the inverted U-shaped relationship between multiple air pollutants and technological progress holds. (2) Decompose technological progress and analyze the bias of output-based technological progress for different kinds of air pollutants. (3) Analysis of the possible bias of pollutant emission reduction in different provinces in China and attempt to explain the bias that causes different biases among different provinces. (4) To verify the bias of output-based technological progress for different types of air pollutants by using econometric models. By solving the above problems, this paper hopes to provide suggestions for China to adjust its environmental protection policies.
2. Methods and Data
This section presents the econometric models involved in this paper, where the DEA-SBM (Slack-Based Measure, SBM) model is used to measure green total factor productivity and describes the process of decomposing it into technological progress and technological progress bias. The regression model is a validation of the inverted U-shaped relationship between multiple pollutants and technological progress and empirical analysis of socioeconomic factors.
2.1. The DEA-SBM Model
Among the methods for measuring total factor productivity, the DEA model approach can be used for multiple inputs and outputs and does not require constructing a production function to estimate the parameters, thus avoiding the errors associated with an artificially set production framework. Therefore, the DEA model is selected in this paper to estimate the required total factor productivity. The traditional DEA model fails to solve the problem of slack variables better when measuring efficiency evaluation. Therefore, in this paper, we adopt the SBM model proposed by Tone [
47], with undesired outputs, which can consider the relationship between inputs, outputs, and undesirable outputs in an integrated way and can better solve the slack problem in efficiency evaluation.
In this study, we use Chinese province panel data, wherein each of the 30 provinces is made a production decision unit DMU to construct the optimal production frontier of China in each period. Each province uses M kinds of inputs to obtain S kinds of desired outputs and I kinds of non-desired outputs, then the DEA-SBM model can speak about the production process of Chinese provinces as:
According to Equation (1), the DEA-SBM model for undesired output can be written as:
where
,
,
represent the slack values of the input, good output, and bad output, respectively.
,
,
represent the
mth input of the
kth DMU, the
pth desired output, and the
qth undesired output, respectively.
is a variable between 0 and 1 representing the efficiency of the
kth DMU environment, where less than 1 means that the
kth DMU is inefficient.
This study constructs the Malmquist index distance function in conjunction with the SBM model dealing with undesirable output. According to the Malmquist exponential decomposition method, the total factor productivity (TFP) growth rate is decomposed into technological change (TC) and efficiency change (EC). We further decompose technological change into output-biased technological progress (OBTC) and input-biased technological progress (IBTC) and magnitude of technological change (MATC).
First, assuming that
and
are the efficiency of the
kth DMU in period
t to
t + 1, China’s green Malmquist productivity index is defined as follows:
> 1 indicates that the green
TFP is growing from period
t to period
t + 1, and
< 1 indicates that the green
TFP is reduced from period
t to period
t + 1. According to the Malmquist exponential decomposition method by Fare [
48], the
TFP growth rate is decomposed into technological change and efficiency change as follows:
denotes the shift of the kth DMU in the period t to t + 1 of the technological change, i.e., the technological frontier. indicates a change in relative efficiency.
After decomposed
TFP Fare [
49], decomposing
TC into an index of the magnitude of technological change and an index of technological bias (
BTC), the technology bias index can be decomposed into input-biased technological progress and output-biased technological progress indices as follows:
i.e.,
MATC represents the scale effect of technological progress, refers to the neutral transfer of the technological frontier, while BTC means the bias of technological progress refers to the “non-neutral” transfer of the technological frontier. IBTC and OBTC reflect the impact of input and output changes on technological progress. If IBTC (OBTC) > 1 (<1), indicates progress (regression) in input-biased technology. When IBTC and OBTC = 1, it means that the technology change is Hicks-neutral.
It needs to be pointed out that output-based technological progress bias refers to the ability of technological progress to produce more output when the input factors are unchanged. We draw on the ideas of Weber and Domazlicky [
50] and Li et al. [
41] on the discriminative approach to the relationship between the direction of technological change and the elements. When
,
OBTC > 1 indicates a
-producing biased technological change and
OBTC < 1 indicates a
-producing biased technological change. When
OBTC = 1, the output-biased technological change is Hicks-neutral. When
,
OBTC > 1 indicates a
-producing biased technological change and
OBTC < 1 indicates a
-producing biased technological change. The
-producing biased technological change means that output-biased technology tends to produce more
relative to
, while the
-producing biased technological change tends to produce more
relative to
.
This paper will focus on analyzing the output bias of technological progress. The specific descriptions of technical bias relationships are listed in
Table 1.
represents desired output;
represents the undesired output.
2.2. The Regression Model
In general, the relationship between the impact of technological progress on environmental pollution has not yet reached a unified conclusion. One view is that the abatement effect of technological progress on pollutants cannot offset the environmental pollution caused by the increase in production efficiency brought about by technological progress. There is also a view that the reduction effect of technological progress and pollutants is in line with the inverted Kuznets U-curve for the environment. When a country’s technological progress is on the left side of the inflection point, and the technological progress itself is based on the fossil energy-based industrial system, the essence of its technological innovation is with pollution attributes. Its pollution reduction effect is necessarily limited. When the level of technological progress is on the right side of the inflection point, then the technological progress is biased toward the use of clean energy, with environmental attributes. The progress of technology can promote pollution reduction. According to the above idea, to further explore the relationship between technological progress and pollutants, the scatter fit of the relationship between technological progress and pollutants required for the analysis of this paper is shown in
Figure 1. From
Figure 1, an obvious non-linear relationship is seen between technological progress and CO
2, SO
2, and PM
2.5. Therefore, the quadratic term of technological progress is introduced when constructing the mechanism of the effect of technological progress on pollutants to draw relevant conclusions in favor of reducing environmental pollution.
The model is set up as follows.
where
are multiple pollutants such as CO
2, SO
2, and PM
2.5,
is the technological progress,
is the set of control variables for the
period of the
province, and
is the random error term of the econometric model. To avoid possible heteroskedasticity and the effect of differences in different variables, the model is logarithmized.
2.3. Data Sources
In this paper, we use Chinese provincial panel data for green total factor productivity measurement. Based on the availability and reliability of the data, we set the time span of the study as 2004–2018 and apply the DEA-SBM model to measure China’s green total factor productivity based on input-output data of 30 Chinese provinces (except Tibet), and decompose the technological progress and biased technological progress from green total factor productivity. The input-output data and processing are as follows.
- (1)
Labor input. Labor input is measured using the number of employed persons in each province of China, and the data are obtained from the statistical yearbooks of each province of China.
- (2)
Capital input. The amount of completed fixed asset investment in each province of China is used to measure capital input. The data are obtained from the statistical yearbooks of each province of China.
- (3)
Energy input. We use the data of energy consumption of each province in China; the data come from the China Energy Statistical Yearbook.
- (4)
Desired output. Nominal GDP data of each province in China are used and deflated to real GDP using CPI index, data from China Statistical Yearbook.
- (5)
Undesired output. We use pollution data for each province in China, where CO2 is the converted emissions from energy consumption in each province, and the conversion method uses the carbon emission calculation method in the Guidelines for National Greenhouse Gas Inventories prepared by the Intergovernmental Panel on Climate Change (IPCC, 2016).
The calculation formula is:
where CO
2 represents the amount of carbon dioxide emissions to be estimated,
represents different types of energy,
represents the combustion consumption of various energy sources, and
represents the average low calorific value of various energy sources. The value comes from the China Energy Statistical Yearbook.
represents the carbon dioxide emission factor of various energy sources, and the value comes from IPCC (2016).
SO
2 is obtained from the industrial emission data of each province, and the data are obtained from the China Environmental Statistical Yearbook. PM
2.5 is satellite gridded data measured by the Atmospheric Composition Analysis Group at Dalhousie University, Canada (
https://fizz.phys.dal.ca/~atmos/martin/?page_id=140 (accessed on 22 July 2021)).
To determine the relationship between technological progress and pollutants, this paper also includes variables such as Economic Development Level, Energy Structure, Industrial Structure, Human Resources, and Foreign Direct Investment in the econometric model.
Energy Structure (ES): The proportion of coal energy consumption to total energy consumption is used to measure the regional energy consumption structure. The higher the level of this variable, the less green the region is.
Industrial Structure (IS): Measured by the proportion of the output value of the secondary industry in the region’s GDP, the higher the level of this variable, the more the region’s economic development relies on the industry.
Human Capital (HC): Human capital refers to the degree of knowledge, skills, and quantity of labor in the workforce. Regions with higher human capital tend to have more advanced technological standards and higher environmental awareness. Higher human capital is beneficial to local environmental improvement, and this paper uses the number of years of education per employed person to measure the level of human capital.
Foreign Direct Investment (FDI): The current research findings on FDI are not uniform, and some studies suggest that foreign investment will cause the transfer of polluting industries from developed countries to developing countries, leading to environmental pollution in host countries [
51,
52]. Some studies also argue that foreign investment will promote the technological progress of host country enterprises and generate learning effects, which will reduce environmental pollution [
53]. Therefore, the impact of foreign investment on the host country depends on factors such as whether the host country introduces polluting industries and its own technological level, and this paper uses the ratio of FDI amount to local GDP to measure the FDI level.
Economic Level (EL): There is a strong correlation between the level of economic development and the environmental pollution problem. Early economic development often came at the expense of the environment, which also led to serious environmental pollution in China, and with the economic model change in recent years, China’s environmental problems have not been properly solved. In this paper, the GDP per capita of each province in China is selected to measure the economic development level of the region.
The data of the above economic data variables are taken from the China Statistical Yearbook and the statistical yearbooks of each province, and the energy variables are taken from the China Environmental Statistical Yearbook, and the descriptive statistics of each indicator are shown in
Table 2