1 Introduction

The global warming and resource restriction have raised widespread concern in the world. As the country with the largest CO2 emission globally (Liu et al. 2014), Chinese government has faced enormous pressure of trimming down its emission. To ease the pressure of emission reduction, Chinese government has pledged that by 2030, the CO2 emissions per unit of GDP would be reduced by 60–65% compared to the level of 2005. In the conference of China–US joint statement on climate change, China also proclaimed that its peak for CO2 emissions would be reached and the proportion of non-fossil energy in primary energy consumption would increase to 20% by 2030. However, economic development is still the main priority in China. It is widely recognized that controlling CO2 emissions is especially difficult with the premise of ensuring economic growth (Wang and Li 2016; Zhu et al. 2015; Noailly and Smeets 2015). Under such circumstance, relying on technological advancement is always the pillar of China’s climate change mitigation strategy (Yuan et al. 2016). Technological advancement is a primary concern in curbing CO2 emissions and also plays a crucial role in transforming the Chinese economy toward a low-carbon pathway.

In light of this, there are increasing studies concerning about the effect of technological advancement on CO2 emissions. After scrutinizing the literature, this study figures out that the role of technological advancement in reducing emissions remains controversial. Some studies reported that the technological advancement played an important role in reducing CO2 emissions (Wei and Yang 2010; Wang et al. 2012; Yang et al. 2014). Researchers also found that increasing CO2 emissions are positively associated with technological advancement, which mainly resulted from the existence of rebound effects (Li and Lin 2015; Wang and Lu 2014). However, other researcher stated technological advancement had no significant impact on CO2 emissions (Hu and Huang 2008; Teng 2012). The influences of technological advancement on environmental issues are complicated and inconsistent because of different factors, models and research samples.

Most studies in this area examined separately the impacts of various technological channels on CO2 emissions and failed to comprehensively investigate the integrated effect of a variety of channels which together may influence the performance of CO2 emissions reduction. Thus, as technological advancement impose complicated and changeable influences on CO2 emissions, a further and detailed exploration of various technological channels for CO2 emissions reduction in a unified framework is urgently required.

To fill in this research gap, this paper pays special attention to different technological channels influencing CO2 emissions and considers different technological channels for technology progress in one united framework, which can further compare their relative advantages in abating the CO2 emissions and avoid the estimation bias caused by omitted variable. In empirical studies, technological advancement can be measured by input indicators [R&D expenditure (Yang et al. 2014; Boeing et al. 2016)] or output indicators [patents (Wang et al. 2012; Wu 2016), literature (Wong et al. 2014)]. But there are some limitations in measuring technological advancement by using output indicators. Taking patents counts, for example, not all innovations are patentable (Albino et al. 2014), and patents counts have no clear interpretation (Pakes 1985). In this sense, this study employs input indicators to measure technological advancement, mainly including indigenous R&D and foreign technology import (Wei and Yang 2010; Teng 2012). Moreover, technical renovation is regarded as another important source of technological advancement (Xu et al. 2013). Considering its significant role in improving environmental performance, it is expected that technical renovation can effectively trim down the CO2 emissions. However, the impact of technical renovation on CO2 emissions is still ignored so far. In summary, this study takes into account the role of indigenous R&D, foreign technology import and technical renovation, the main channels of technological advancement, in mitigating the CO2 emissions.

Furthermore, several econometric energy models have been adopted to estimate the impact of technological advancement on CO2 emissions, which generally involve building the STIRPAT (stochastic impacts by regression on population, affluence and technology) model (Lin and Xie 2014; Wang and Zhao 2015), the Kaya equation (Wang and Li 2016) and Logarithmic Mean Divisia Index (Ang 2004), etc. Besides the aforementioned models, the environmental learning curve (ELC) model is increasingly applied to estimate the role of technological advancement in curbing CO2 emissions (Fehr 2003; Sun et al. 2008, 2011; Yu et al. 2011, 2015; Guo et al. 2016). Moreover, ELC model integrates the concept of conventional learning curves and depicts the environmental protection progress along with the development of technology (Yu et al. 2015). This model is thereby more plausible and internally consistent than the intuition of the modeler (Grübler et al. 1999; Yu et al. 2011). More importantly, the strong CO2 emissions learning ability of technological advancement estimated by ELC model indicates the technologies would effectively curb the emissions, which is crucial for China to achieve the ambitious CO2 emissions reduction target. Therefore, the ELC model is applied in this paper to address the impacts of different technological channels on CO2 emissions.

Besides, China has a vast territory and exhibits significant regional differences. Previous regional research mainly bases on the classification of geographical location (Wang et al. 2012; Zhang and Zhou 2016). However, there are significant differences among regions in China in terms of CO2 emissions performance (Wang and Zhao 2015). Hence, to analyze the characteristics of regions with various CO2 emissions and emission efficiency, this paper divides China into four regions according to the “CO2 emission-efficiency” features.

The rest sections of this paper are organized as follows. Section 2 presents a literature review; model specification, methodology and data sources are introduced in Sect. 3. Section 4 presents the empirical results and discussions of CO2 emissions learning curve through technological advancement; main conclusions and policy recommendations of the current research are summarized in Sect. 5.

2 Literature review

This paper constructs the environmental learning curve to identify how different technological channels can influence the regional CO2 emissions, and further makes a distinction among the four regions. In this section, based on relevant literatures, we analyze the role of indigenous R&D, foreign technology import and technical renovation in CO2 emissions reduction.

2.1 The role of indigenous R&D in CO2 emission reduction

Indigenous R&D is recognized as the engine of sustained economic growth in endogenous growth theory (Romer 1990) and is also found to be a vital channel to promote the technology innovation and accumulate absorptive capacity. These capacities are crucial for increasing energy efficiency and curbing CO2 emissions for a long run (IPCC 2006). Many related papers have been conducted, including those by Ang (2009), Wei and Yang (2010) and Yang et al. (2014). Moreover, previous studies have also reported that finite resources and stringent environmental goals in developing country have facilitated the indigenous R&D investment for energy-saving technologies, which in turn reduced the CO2 emissions (Teng 2012, Fisher et al. 2004). Additionally, indigenous R&D, which determines the pace and direction of the technology transformation, is used to measure the input of technological innovation. Li and Lin (2016) also confirmed that higher R&D investment in China has resulted in more energy technology patents and thus accelerated the development of energy conservation and low-carbon technology. Consistent with these previous studies, we expect that indigenous R&D could help to reduce CO2 emissions in China. Thus, we hypothesize:

Hypothesis 1

The indigenous R&D is positively associated with the reduction of CO2 emissions in China.

2.2 The role of foreign technology import in CO2 emission reduction

Technological advancement is costly and risky (Boeing et al. 2016). Compared to indigenous R&D, foreign technology import is a relatively low-cost R&D activity (Wei and Yang 2010). Hence, foreign technology import is another key channel for technological advancement. Importing technologies from abroad causes fierce competition to domestic firms and promotes domestic firms’ impetus to invest in R&D activities in order to enhance their competitiveness (Boeing et al. 2016). Therefore, importing foreign technology not only accelerates the technological advancement, but also generates a series of beneficial effect, such as vertical linkages (Wei and Yang 2010) and international cooperation (Boeing et al. 2016). Yet, the empirical evidences concerning whether foreign technology import is beneficial to the reduction of CO2 emission or not are mixed. Wei and Yang (2010) implied that foreign technology import played a crucial role in reducing CO2 emissions according to a sample of 29 provinces in China, whereas Teng (2012) suggested that foreign technology import has a significant negative influence on energy intensity of the 31 industrial sectors, which in turn may raise CO2 emission. Besides, Teng (2009) found that the impact of foreign technology import on energy intensity is highly dependent on the technology level of the provinces in China. Following the intuitions derived from the above literatures, we develop the research hypotheses:

Hypothesis 2

The learning ability of foreign technology import in CO2 emission reduction is modified by regional technological capacity in such a way that greater technological capacity is associated with a higher CO2 learning ability.

2.3 The role of technical renovation in CO2 emission reduction

Technical revolution is different from indigenous R&D or foreign technology import. The term “technical revolution” in this paper means that the enterprises implement effective measures to refurbish and retrofit the existing facilities, technological conditions or production services, etc. (SCC 2012). For example, replacing low-efficient coal combustion in industrial and residential heat supply and installing metering equipment for heating energy use (Xu et al. 2013). Thus, the purposes of technical revolution are to eliminate backward production capacity, improve production efficiency, reduce pollutions emission, and ultimately achieve the technological advancement. It implied that technical revolution plays an important role in improving environmental performance. In case of China, technical revolution has become the major initiatives of China’s national energy and environment policy. During the 11th FYP period, pushing clean coal power development in both newly built and existing fleets mainly relies on the technical revolution (Yuan et al. 2016). In 2014, the Action Plan on Energy Saving and Emissions Reduction Upgrading & Retrofitting in Coal Power Plants (2014–2020) was issued by National Development and Reform Commission (Yuan et al. 2016). The Action Plan aims to build a clean and high efficient coal power sector through technical revolution in China. Thus, it is necessary to know whether technical revolution can cut down the CO2 emissions in China. However, only a few environmental literatures have taken into account the role of technical revolution (Rodney and Phil 2009; Xu et al. 2013). Thus, this study will place particular emphasis on the role of technical revolution in mitigating CO2 emissions.

Hypothesis 3

Boosting the investment of the Technical revolution is beneficial to curb the CO2 emissions in China.

3 Data and methodology

3.1 Multivariable ELC model

The original learning curve model was firstly proposed by Wright (1936) to depict the time savings (or cost reduction) accompanying the increasing output of aircraft manufacturing. The learning curve model reveals the gradual manufacturing cost decline with repeating production process or increasing output in the enterprises. It also reflects that enterprises succeeded in reducing the manufacturing cost with the experience accumulation or technological advancement. Later, learning curve has been extended to not only manufacturing industry, but also engineering technology innovation, products research and development and other industries (Fehr 2003; Sun et al. 2011). Recently, borrowing from learning curve, “environmental learning curve” has been proposed. Previous related studies have employed the ELC to explore the energy, water consumption and SO2 emissions (Sun et al. 2008, 2011; Yu et al. 2011). The ELC model introduces a concept of “environmental learning” that reflects environmental improvements (i.e., CO2 emissions abatement) through the economic development and technological advancement (Fehr 2003). However, previous studies in this area only considered the per capita GDP factor (Sun et al. 2008, 2011) and technological factors were ignored. Therefore, the environmental learning by technological advancement needs to be further explored.

The ELC model is commonly estimated by Alchian multilog model, Cobb–Douglas multiplicative exponential model and Womer’s variable production rate model (Badiru 1992). Due to the feature of flexibility and concise structure, Cobb–Douglas model has been multiply utilized (Rubin et al. 2004). The standard Cobb–Douglas model can be expressed as Eq. (1).

$$C = A\mathop \prod \limits_{i = 1}^{ n} (x_{i} )^{{ - b_{i} }} \varepsilon$$
(1)

where \(C\) denotes the environmental cost; \(A\) indicates the initial environmental cost with fixed value which is determined by the development of society and economy. \(b_{i}\) represents the learning coefficient of the \(i\)th factor, which can be estimated by regression; \(x_{i}\) stands for the \(i\)th learning factor; and \(\varepsilon\) is the error term.

Similar to the ELC, the CO2 emissions learning curve reflects that enterprises (i.e., industries or regions) succeed in reducing CO2 emissions through experience accumulation and technological advancement. The CO2 emissions learning by technolocal advancement implies the technology transformation from carbon-intensive technologies towa low-carbon ones, which is important to curb the CO2 emissions in China for a long run.

Above all, this paper explores the CO2 emissions learning by technological advancement in China, and three different technological channels are adopted to construct the process of technological advancement, including indigenous R&D activity, foreign technology import and technical renovation. To avoid omitted variable bias, GDP per capita is also included in the present study due to its important role in influencing CO2 emissions and technological advancement. By applying the natural logarithms of Eq. (1) to both sides, an extended ELC was proposed.

$${\text{lnCE}}_{i ,t} = A_{0} - b_{1i} { \ln }Y_{i ,t} - b_{2i} {\text{lnRD}}_{{i ,t{ - 1}}} - b_{3i} {\text{lnFT}}_{{i ,t{ - 1}}} - b_{4i} {\text{lnTR}}_{{i ,t{ - 1}}} + \varepsilon_{i ,t}$$
(2)

where \(i\) is the \(i\)th region, and \(t\) is the \(t\)th year; \({\text{CE}}\) represents CO2 emissions, \(A_{0}\) denotes the initial CO2 emissions which are determined by the development of society and economy. \(Y\) stands for GDP per capita computed as GDP divided by population at the end of the year \(t\). \({\text{RD}}\), \({\text{FT}}\) and \({\text{TR}}\) indicate the knowledge stock of indigenous R&D, foreign technology import and technical renovation, respectively. Further,\({\text{RD}}\), \({\text{FT}}\) and \({\text{TR}}\) variables are lagged by one year to allow their impacts on CO2 emissions to be sluggish. \(b_{1i} , b_{2i} , b_{3i}\) and \(b_{4i}\) indicate the learning coefficient of the \({\text{RD}}\), \({\text{FT}}\) and \({\text{TR}}\), respectively. Taking the learning coefficient of indigenous R&D, for example, the larger value of \(b_{2i}\) implies the stronger CO2 learning ability of indigenous R&D in the \(i\)th region, while the impact on CO2 emissions reduction associated with indigenous R&D is also greater. Additionally, if \(b_{2i} { > 0}\), it denotes that there exists reverse changes in the relationship between indigenous R&D and CO2 emissions. That is, if indigenous R&D increases, then CO2 emissions decrease. If \(b_{2i} { < 0}\), the CO2 emissions change in the same direction as indigenous R&D. The specific descriptions of the variables are shown in Table 1.

Table 1 Description of variables used in the analysis for the period 1997–2014
Full size table

4 Methodology

To avoid the pseudo-regression issue caused by nonstationary panel data, four panel unit root tests are employed, including Levin, Lin and Chu (LLC) (Levin et al. 2002), Im, Pesaran and Shin (IPS) (Im et al. 2003) Fisher-ADF and Fisher-PP tests (Maddala and Wu 1999). Table 2 summarizes the statistics at the national and regional levels. As shown, the series are nonstationary in levels but stationary in first differences, which indicates that the series are first-order integrated series. Further, Pedroni (1999) and Kao (1999) cointegration test are adopted to test the long-run equilibrium among variables. In accordance with the descriptive evidence in Table 3, the null of no cointegration is rejected in our sample.

Table 2 Results of panel unit root tests
Full size table
Table 3 Results of Pedroni and Kao panel cointegration tests
Full size table

Five regression methods are used after building the national-level and regional-level CO2 learning curve. These methods contribute to gain more reliable regression results, including fixed effects (FE), linear regression with Newey–West standard errors (N–W), feasible generalized least squares (FGLS), linear regression with panel corrected standard errors (PCSE) and linear regression with Driscoll–Kraay standard errors (DK) estimation. To select the appropriate regression methods for different regions, four statistic tests are carried out in this study. The robust Hausman test is firstly adopted to choose between the fixed and random effects models. The results are summarized in Table 4. As we can see, all the statistics reject the null hypothesis, suggesting that the fixed effects model is more suitable in both national and regional levels. In addition, modified Wald test (Greene 2000), Wooldridge test (Wooldridge 2002) and cross-sectional dependence test (Pesaran 2004) are conducted to detect the heteroskedasticity, autocorrelation and cross-sectional dependence in FE models. As shown in Table 5, the three statistics are all significant at national-level. There exist heteroskedasticity, autocorrelation and cross-sectional dependence among H–H region, H–L region and L–H region, whereas heteroskedasticity and autocorrelation in L–H region are significant in spite of insignificant cross-sectional dependence. From the above analysis, the estimation results of FE models could possibly be biased. To solve these issues, N–W, FGLS and PCSE are used to obtain the more reliable results. N–W estimation could deal with the heteroskedasticity, autocorrelation without consideration of cross-sectional dependence (Newey and West 1987). Additionally, FGLS and PCSE estimation are carried out to overcome cross-sectional correlation. The FGLS estimation could be against the disadvantage of inaccurate standards errors and will be suitable when the cross-sectional dimension \(N\) is smaller than the time dimension \(T\) (Beck and Katz 1995). Nevertheless, in the finite sample, PCSE estimation is poor when the \(N\) is rather large compared to the \(T\) (Hoechle 2007). Hence, DK estimation is implemented to overcome the bias results by modifying the standard nonparametric time series covariance matrix estimator (Driscoll and Kraay 1998).

Table 4 Results of Robust Hausman test
Full size table
Table 5 Results of group-wise heteroskedasticity, autocorrelation and cross-sectional dependence test
Full size table

In summary, this paper estimates the empirical models (model 4, 6 and 10) by DK estimation to solve heteroskedasticity, autocorrelation and cross-sectional dependence issues, and focuses on N–W estimation for L–H regions (model 8) to address the heteroskedasticity, autocorrelation problems in this region.

4.1 The division of the provinces in China

To further investigate the CO2 emissions learning curve in regions with different CO2 emissions and efficiency, this paper selects the average amount of CO2 emissions as emission targets and CO2 intensity as emission efficiency indicators, and the 29 provinces in China are divided into four regions using the method proposed by Shao et al. (2014). The four regions are high emission–high efficiency region (H–H), high emission–low efficiency region (H–L), low emission–high efficiency region (L–H) and low emission–low efficiency region (L–L). The results are shown in Table 6 and Fig. 1.

Table 6 “Emission-efficiency” type of provinces in China
Full size table
Fig. 1
figure 1

Histogram of average CO2 emission and emission efficiency in China from 1997 to 2014

Full size image

4.2 Data source and description

This paper includes annual data of China’s 29 provinces covering the period from 1997 to 2014. Hong Kong, Macao, Taiwan and Tibet are not included because of lack of data, and Chongqing is merged in Sichuan province to ensure the consistency of data. The provincial data of CO2 emissions are computed according to the Intergovernmental Panel on Climate Change (IPCC) (IPCC 2006). The calculation is based on the final energy consumption provided by the China Energy Statistical Yearbook (NBSC 1997–2014b). Three types of fossil fuels (raw coal, crude oil and natural gas) are included.

GDP per capita is computed as GDP divided by population at the end of year, the panel data of GDP (in 1997 constant price, Yuan) and population are collected from the China Statistical Yearbook (NBSC 1997–2014a)

The R&D expenditure has invariably influences on CO2 emissions, so the knowledge stock of R&D expenditures is more appropriate chosen as a factor in our paper than R&D expenditure flow. The knowledge stock of R&D expenditures is computed using the perpetual inventory method as \(K_{i,t} = I_{i,t} + (1 - \delta )K_{i,t - 1}\) where \(\delta\) represents the knowledge stocks depreciation rate and is assumed by 15% as same as the previous literature (Hall and Mairesse 1995). Expenditures of indigenous R&D, foreign technology import and technical renovation are collected from China Statistical Yearbook on Science and Technology (NBSC 1997–2014a).

The distribution of CO2 emission and emission efficiency in different regions during 1997–2014 is illustrated in Fig. 2. As shown, all emission regions show an ascending trend in CO2 emissions with different change rates. The average annual growth rate of CO2 emission in L–L region (7.56%) is the highest, following by H–H (7.46%), H–L (6.86%), L–H region (6.03%). In 2014, CO2 emission of L–L, H–H, H–L and L–H region has been raised by 237.24, 230.52, 202.63 and 163.99%, respectively. Nevertheless, there is an obvious upward trend in emission efficiency of four regions during 1997 and 2014 except L–L region. The emission efficiency in L–L region slightly increases with an average growth rate of 2.24% and aggregate growth rate of 43.17%. Conversely, the emission efficiency of the H–H region is 4.40% with a total rapid increase of 104.37%, while that in H–L and L–H region are 4.27 and 4.31%, respectively, growing by 101.04 and 101.61% in all.

Fig. 2
figure 2

CO2 emissions and emission efficiency of different regions in China from 1997 to 2014

Full size image

5 Empirical results and discussion

In current study, CO2 emissions learning curve is adopted to explore the role of three different technological factors in CO2 emissions reduction with a consideration of regional differences. The results are shown in Table 7. Additional results based on alternative regression methods (i.e., FGLS/ PCSE method) are present in supplementary materials Tables S1-S5. For the ease of interpretation, the analysis of technological factors is discussed separately.

Table 7 Panel estimation results for CO2 emissions of different regions in China during 1997–2014
Full size table

5.1 Impact analysis of indigenous R&D

The estimation results for the whole country are clearly shown in Table 7. From model 2, it can be seen that indigenous R&D has no significant impact on CO2 emissions, indicating the poor CO2 learning ability for indigenous R&D and insufficient environmental conscious across the country, which is inconsistent with Hypothesis 1.

Table 7 shows the results of the H–H, H–L, L–H and L–L region, respectively. As shown, the learning coefficients of indigenous R&D in the four regions are apparently different. Indigenous R&D in H–H and H–L region has negatively influences on the CO2 emissions. On the contrary, the elasticity of indigenous R&D in L–L region is −0.346. This means that the CO2 learning ability in L–L region is rather weak. Furthermore, the learning coefficient of the indigenous R&D in L–H region is insignificant at the 5% level, indicating that L–H region cannot benefit from indigenous R&D. In sum, this study finds an interesting phenomenon that the CO2 learning ability in regions with higher emissions is much better than the lower ones.

The possible reason is that high emission pressure may motivate indigenous R&D to invest in the low-carbon technology, and then may promote the technological progress away from carbon-intensive technologies toward carbon-free technologies. For example, renewable energy technologies are better developed in the regions with higher CO2 emissions. According to the data from Collection of China Electric Power Statistics, the cumulative installed capacity of photovoltaic power in H–H region accounts for 48% of the total capacity in China by the end of 2013. Moreover, the wind power capacity in H–L region increased much more quickly, such as Inner Mongolia; the cumulative installation of wind power sharply increased from 157 to 22,312 MW during the period of 2006 to 2014. However, due to the increasing demand of energy consumption in China, only regions with high emissions can benefit from the indigenous R&D, which raises the concern that the CO2 leaning abilities in L–L and L–H region need to be further improved by increasing emissions reduction pressure. Particularly, the L–L region is rich in renewable energy resources, such as solar PV power, and the development potential of renewables technology in this area is tremendous. But the weak learning capacity of indigenous R&D has severely impeded the development of renewables technologies. Meanwhile, most of the provinces in L–L region are inland provinces and in the advanced situation for foreign trade with Central Asia energy-rich countries. Thus, importing foreign technology from abroad and enhancing the international cooperation may make up the poor indigenous R&D capacity.

5.2 Impact analysis of the foreign technology import

The estimated results of the impacts of foreign technology import on CO2 emissions for the whole country are presented in Table 7. The learning coefficient of foreign technology import for the whole country is greater than zero. Boosting the investment in foreign technology import will lead to CO2 emissions reduction over the period, which enhances the confidence of the government in mitigating the emissions through foreign technology import. This finding is consistent with that of Wei and Yang (2010).

The learning abilities of foreign technology import on emissions abatement also differ among the four regions. With the exception of H–H region, all influences of foreign technology import on emissions are statistically significant at 1% level. Learning coefficient of foreign technology import in H–L regional is less than zero (\(b_{31} =\) −0.522), which means raising the investment in foreign technology will lead to the increase in the CO2 emissions. Conversely, foreign technology import for the L–H and L–L region makes an achievement about declining the CO2 emissions during the research period. A 1% increase in foreign technology import will decrease the emissions of L–H and L–L region by 0.169 and 0.356%, respectively. From the discussions above, we can conclude that the CO2 learning abilities in L–H and L–L region are stronger than that in H–H and H–L region.

This finding does not conform to Hypothesis 2, as L–L region with poor technological capacity has a relatively stronger CO2 learning ability. Although unexpectedly, the possible reason for such finding may be that the expenditures of foreign technology import in H–L and L–L region are spatially more concentrated. Data show that approximately 74.0% expenditure for importing the foreign technology is concentrated on 50% provinces of the L–L region. Similarly, 75.3% of that is concentrated on 44% provinces of L–H region. Nevertheless, H–H and H–L regions are more evenly distributed across provinces. A higher degree of spatial concentration of expenditure will have better technology capabilities, which benefit the technology transfer and absorbing knowledge from abroad (Zhang and Zhou 2016), and this is essential for China to improve energy efficiency and reduce the emissions. This raises the concern that the CO2 emissions gap between low emissions regions (L–H and L–L region) with high CO2 learning ability and high emissions regions (H–H and H–L region) with low CO2 learning ability may be broadened. Meanwhile, the H–H and H–L region still face limitations with regard to low-carbon technology transfer and absorbing knowledge from abroad. Hence, there is a great need to strengthen technology capabilities in these regions through concentrated for investment of foreign technology import. Finally, these regions should seek to import low-carbon and energy-saving technologies due to its energy-guzzling industrial structure.

5.3 Impact analysis of the technical revolution

The results from Table 7 strongly suggest that technical renovation is the effective measure for abating CO2 emissions both in the national and regional levels, which is consistent with Hypothesis 3. The technical revolution is the dominant contributor to the CO2 emissions reduction in the national level. Furthermore, technical renovation in H–H region has theoretical maximum CO2 learning ability (\(b_{41} =\) 0.285), followed by L–H region (\(b_{43} =\) 0.191), L–L region (\(b_{44} =\) 0.186) and H–L region (\(b_{42} =\) 0.125). This indicates that throughout the entire period of 1997–2014, emission reduction in H–H region is most effective in the technical revolution.

Our finding implies that the more investment of technical renovation will result in lower CO2 emissions, and the reasons mainly lie in the following aspects. Firstly, the main purpose of technical renovation is to improve the energy efficiency by refurbishing and retrofitting low-efficient equipment, which can directly cut down the CO2 emissions. Secondly, the renovation relies less on the technology capabilities, i.e., high human capital and economics base, to adopt and absorb the new technologies (Xu et al. 2013). Thus, technical renovation can reduce investment risk and increase the efficiency in a short term that enables the enterprises to more effectively assimilate low-carbon technology. Finally, due to indispensable role of coal power, clean and efficient coal power development through technical renovation has been the priority of Chinese government. The government has implemented a series of measures concerning energy efficiency retrofitting in traditional coal-fired generating units (Yuan et al. 2016). Nevertheless, technical renovation tends to achieve high energy efficient by implementing the mature and efficient technologies, but cannot promote the development of new renewables technologies. Hence, apart from the technical renovation, it is crucial to emphasis on the investment of indigenous R&D and foreign technology import to accelerate the renewables technologies and achieve the CO2 emissions reduction target for a long run.

6 Conclusion and policy implications

This paper establishes an environmental learning curve model to explore the impact of technological advancement on CO2 emission in China from 1997 to 2014 by using a panel data of 29 provinces, with a particular emphasis on different technological channels and regional differences. There are three technological channels including indigenous R&D activity, foreign technology import and technical renovation. Furthermore, China’s 29 provinces are decomposed into high emission–high efficiency region (H–H), high emission–low efficiency region (H–L), low emission–high efficiency region (L–H) and low emission–low efficiency region (L–L). The empirical results of this study are as follows:

  1. (1)

    Technological progress accompanied by the indigenous R&D has environmental learning effect on CO2 emission in H–H and H–L region, while the increase in indigenous R&D leads to the raise of CO2 emission in L–L region. In terms of L–H region, it has no significant impact on CO2 emissions.

  2. (2)

    The environmental learning effects by the foreign technology import are different among the different regions. Only L–H region and L–L region have a negative learning effect on reducing CO2 emission. However, the learning effect of foreign technological import on CO2 emission is not obvious in H–H region, while the learning ability in H–L region is weak.

  3. (3)

    The technical renovation in all regions has positive impact on curbing the CO2 emissions. The environmental learning ability in H–H region is the largest, followed by L–H region, L–L region and H–L region.

These results not only contribute to enrich the technology innovation theory, but also proposing specific policies for improving the environmental performance in China, which are summarized as follows:

Firstly, the results suggest that different regions should take differentiated investment strategy and policy measures to curb CO2 emissions according to local conditions. Firstly, CO2 emission learning abilities of indigenous R&D in high emissions regions are stronger than the low ones, which imply that high pressure of emissions reduction may arouse the environmental consciousness and facilitate the indigenous R&D in low-carbon technologies. However, high emissions regions still remain high emissions and it suggests that the indigenous R&D’s learning abilities in these regions should continually be improved in order to transform their emissions-efficiency performance to L–H regions. Indigenous R&D is the source of the creation of new clean technology and mainly relies on the enterprises. Therefore, in order to encourage them to invest in low-carbon technologies, China should promulgate more stringent CO2 emissions and efficiency policies to provide incentives for enterprises to be involved in low-carbon technology innovation. In addition, designing proper subsidies for firms will be another effective incentive policy and can ensure compensation for their investment in clean technologies.

Secondly, the CO2 emission learning effects of foreign technology import depend on the concentrated degree of the local expenditure. The results suggest that H–H and H–L region should import foreign technology more spatially concentrative, and focus on low-carbon and energy-saving technologies due to its energy-intensive industrial structure. Meanwhile, L–H and L–L region should seek to import foreign technology to further reduce emissions. Furthermore, the government should insist on strengthening competitiveness in energy-intensive industries and, emphasizing on cultivating technology capabilities, especially for L–L region, to adopt and absorb advanced foreign technologies.

Thirdly, the excellent learning ability of technical renovation implies that it will effectively reduce the CO2 emission and improve energy efficiency in China. Thus, it calls for striving to lessen CO2 emissions through technical renovation. In particular, the resource-dependent provinces are located in H–L regions, such as Shanxi and Inner Mongolia, which are rich in resources of coal and major energy exporter through coal power generation. Thus, the Chinese government should particularly emphasis on the technical revolution in coal power of H–L region. Besides, the haze smog frequently occurred in China during recently years. The government should also be cautious of the strict pollution control standard in the heavy industry and implement the electricity substitution projects.