There are few studies analysing whether different types of environmental regulation have differential impacts on the efficiency of the construction industry. Using 2012–2016 panel data from 30 provinces in China, the green total factor productivity (GTFP) of the construction industry is measured with a global Malmquist–Luenberger productivity index based on the epsilon measure model. Thereafter, a panel tobit regression model is proposed to explore the relationship between three types of environmental regulation and the GTFP of the construction industry. The results show that (a) from 2012 to 2016, the GTFP of the Chinese construction industry grew slowly at an average annual rate of 0.14%; (b) both one-phase lagged command-and-control and current phase market-based environmental regulation had a positive linear relationship with GTFP, while one-phase lagged voluntary environmental regulation, on the other hand, had an inverted U-shaped relationship with GTFP; (c) the three types of environmental regulation can be combined to establish a suitable environmental regulation system. The findings of this study provide guidance for the sustainable development of the construction industry by combining the actions of different types of environmental regulation.
Notation
1 Introduction
Economic globalisation not only has promoted the development of the world economy but has also caused environmental problems in many countries. Indeed, from the perspective of long-term development, economic growth at the cost of environmental pollution is unsustainable (Beck et al., 2011; Millimet and Roy, 2016). The coordination and integration of industrial economic development and environmental protection through environmental regulation (ER) has attracted the attention of governments across the world (Gouldson et al., 2009). ER provides important means for reducing the impact of economic activities on the ecological environment. Governments can use different types of ER to guide enterprises in order to make technological innovations for improving energy efficiency, reducing pollution emissions, increasing green total factor productivity (GTFP) and ultimately realising sustainable development (Liu et al., 2018). In this context, GTFP is defined as the ratio of the total output of a production system to the actual input of all production factors when considering undesirable output and energy consumption (Chen et al., 2018; Wang et al., 2018).
However, most studies focus on the influence of ER on the economy and industry efficiency (Zhang et al., 2012, 2018). Very few consider the effect of different types of ER on construction industry performance (Shen et al., 2019; Xie et al., 2017). Additionally, while previous studies point out that different ERs have different impacts on efficiency (Liu et al., 2018; Ren et al., 2018; Shen et al., 2019), just one has considered the effects on construction industry performance. This study, though, focused on certain EU regions only (Testa et al., 2011).
Furthermore, since GTFP represents the contribution to industrial growth, its change over time can directly reflect construction industry performance. Different types of ER have either a positive (i.e. innovation compensation) or negative impact (i.e. crowding-out effect) on GTFP (Chen et al., 2018). Thus, it is important to explore whether there is a relationship and time lag effect between the implementation of different types of ER and the GTFP.
This study analyses these two factors in the context of the Chinese construction industry. First, as the world’s largest developing country and the second largest economy, China has severe environmental problems from continued economic growth. According to the BP Statistical Review of World Energy 2018 (BP, 2018), China is still the world’s largest energy consumer and carbon dioxide (CO2) emitter. However, the rapid development of its construction industry not only has consumed vast resources and energy but has also caused considerable damage to the environment. This has created multiple air, water, solid waste and noise pollution problems (Li et al., 2019; Zhang et al., 2016, 2019).
Second, to address these issues, the Chinese government has applied different types of ER to regulate the development of the construction industry and promote GTFP – namely, it has made full use of three types of ER: (a) command-and-control ER (CER), which requires the industry to adopt green technologies to avoid environmental administrative penalties issued by governmental authorities; (b) market-based ER (MER), which exerts economic pressure to make the industry improve its production efficiency and reduce pollutant emissions for cost reasons – pollution-charging taxes and environmental protection taxes are some MER examples; and (c) voluntary ER (VER), which relies on the environmental awareness of citizens to supervise and influence the way that the industry operates – for example, from environmental letters and visits (Feng and Chen, 2018; Li and Ramanathan, 2018; Ren et al., 2018). The government requires the construction industry to adopt building materials with low carbon dioxide emissions, advanced energy saving and emission-reduction technologies. These aim to reduce energy consumption and environmental pollution in the manufacturing process and promote GTFP (Mohurd, 2017).
Hence, this study aims to explore the relationship between different types of ER and GTFP in the Chinese construction industry. Panel data from 30 provinces from 2012 to 2016 are used, with the global Malmquist–Luenberger productivity index based on the epsilon measure (EBM-GML) model. This model also considers undesirable outputs when estimating the GTFP. Finally, a tobit regression model is used to analyse the relationship, impacts and possible time lags between different types of ER and GTFP.
2 Literature review and hypothesis development
2.1 ER and the construction industry
According to externality theory, the pollution discharges of a construction enterprise will bring losses to surrounding enterprises or consumers and have a negative effect on society and the environment. These are negative externalities, and in environmental economics, ER is the main way to curb those caused by pollution (Arrow, 1969).
However, ER can make the production process easier or more difficult. Sometimes, due to lower energy utilisation rates, more environmental pollutants are produced by construction enterprises. When ER is strengthened, construction enterprises are forced to invest more capital in pollution control, thereby crowding out the enterprises’ investment in other aspects (Jaffe et al., 1995, 2002). Conversely, an appropriate ER can steer technological innovation, help enterprises adopt new methods to improve energy usage and reduce waste emissions (Alpay et al., 2002; Porter and van der Linde, 1995).
Although previous studies (e.g. Zhang et al., 2012, 2018; Zhong et al., 2017) have considered ER as a sole indicator when exploring its impact on the construction industry, there are different types of ER, and each can have different effects. Hence, simply selecting one indicator may not be fully representative. It is necessary to devise an indicator system that discriminates according to the type of ER to capture fully the intensity of the regulation and potential impact on GTFP.
Additionally, previous studies measured only the efficiency (also termed ‘total factor productivity’ (TFP)) of the construction industry. They did not consider the sustainable development dimension of productivity – that is, they did not measure GTFP. In response, this study considers the different impacts of the three types of ER; takes energy consumption as an input factor and carbon dioxide emissions as undesirable output; and uses other auxiliary indices to calculate GTFP.
2.2 ER and GTFP
The growth rate of TFP and its contribution to output growth is regarded as the main basis for assessing economic development and improvement in economic quality (Lin and Su, 2007). However, TFP does not consider environmental pollution or the loss of resources caused by economic growth. This can bias economic efficiency evaluations and produce misleading policies (Hailu and Veeman, 2000). An improvement proposed by Chung et al. (1997) when measuring the TFP of the Swedish pulp mill is to use a directional distance function that can treat pollution emissions as undesirable output. With the maturity of technology efficiency, other factors such as pollution emissions, resources and energy consumption have started to be included in the TFP analysis framework, now called GTFP.
ER is a conventional tool for addressing the environmental issues arising from economic growth (Wang et al., 2016). The relationship between ER and GTFP is discussed in many studies (e.g. Chen et al., 2018; Li and Wu, 2017; Wang et al., 2018), but there is still common agreement on how both variables are related – another reason why there is a need to drill down different types of ER to find clearer answers.
2.2.1 Command-and-control environmental regulation
Faced with increasing emissions of environmental pollutants, many governments formulate emission standards and production technology standards to reduce or eliminate such emissions. This series of policies is often referred to as command-and-control ER (CER) (Blackman, 2010). When pollutant emissions exceed these standard limits, enterprises face heavy fines or administrative penalties such as forced closure (Yang et al., 2012).
Therefore, some environmental emission standards can prompt the industry to find its own low production efficiency. This may lead the industry to improve continuously its production processes to reduce pollution and ultimately improve GTFP, suggesting the following hypotheses.
H1a: CER has a positive linear relationship with the construction industry GTFP.
H1b: CER has a non-linear relationship with the construction industry GTFP.
2.2.2 Market-based environmental regulation
Market-based ER (MER) tools for control pollution are currently widely accepted by countries worldwide. Similar to CER, MER also has corresponding emission standards, but they are more dependent on other economic tools (Camison, 2010). Under the control of MER, the production options of the industry are more flexible, as appropriate investment methods can be chosen to reduce the negative effects of production on the environment (Jacobs et al., 2010).
In summary, MER creates more economic pressure on enterprises by encouraging the transformation of industrial production to a greener one. Thus, two further hypotheses are as follows.
H2a: MER has a positive linear relationship with the construction industry GTFP.
H2b: MER has a non-linear relationship with the construction industry GTFP.
2.2.3 Voluntary environmental regulation
Unlike the two types of ER tools mentioned earlier, voluntary ER (VER) relies on the environmental awareness and capabilities of citizens to influence the way that the industry produces (Borck and Coglianese, 2009). When the citizens’ environmental awareness is enhanced, they are more inclined to a greener and environmentally friendly lifestyle. Due to changes in market preferences, consumer lifestyles and the awareness of environmental protection, the industry should move (directly or indirectly) towards greener production methods increasing its GTFP (Vitiea and Lim, 2019).
This provides the following final two hypotheses.
H3a: VER has a positive linear relationship with the construction industry GTFP.
H3b: VER has a non-linear relationship with the construction industry GTFP.
3 Data and methodology
3.1 Data source
The data for this study cover 30 provinces in China from 2011 to 2016 (with 2011 as the base period). Due to the data from the provinces of Hong Kong, Macau, Taiwan and Tibet not being available, they were not included in the analysis. Data were mainly drawn from China Statistical Yearbook (NBS, 2012a, 2013a, 2014a, 2015a, 2016a, 2017a), China Statistical Yearbook on Construction (NBS, 2012b, 2013b, 2014b, 2015b, 2016b, 2017b), China Energy Statistical Yearbook (NBS, 2012c, 2013c, 2014c, 2015c, 2016c, 2017c), China Environment Yearbook (NBS, 2012d, 2013d, 2014d, 2015d, 2016d, 2017d), China Statistical Yearbook on Environment (NBS, 2012e, 2013e, 2014e, 2015e, 2016e, 2017e) and relevant annual data concerning Chinese provinces from the website of the National Bureau of Statistics of China (NBS, 2020).
3.2 Construction industry GTFP
3.2.1 Selection of indicators
The input and output indicators of GTFP are selected from a literature review as follows (references and measurement units from all indicators are summarised in Table 1).
Input indicators. (a) Labour force: the number of employees is usually selected as the labour force indicator in the construction industry. (b) Capital: denoted here by the total assets of construction enterprises in each region. This indicator is used because of the construction industry’s complex capital structure and the inability to obtain the depreciation rate of the fixed assets in each province. (c) Machinery and equipment: the total annual power of machinery and equipment owned by construction enterprises in each region. (d) Energy: under the constraint of limited resources, the total energy consumption is a frequent measure of energy input. The terminal energy consumption of the industry is selected here to represent energy input.
Output indicators. (a) Desirable output: considering that the products of the construction industry are mostly buildings, structures and facilities, it is difficult to summarise their actual volume. Therefore, commonly used output indicators are used, such as gross output value, total profit, gross added value and area of completed construction. From the perspective of output and profit capacity, the gross output value and total profit of the construction industry are assumed desirable output indicators. (b) Undesirable output: as an unpaid environmental cost, carbon dioxide is the world’s most well-known environmental pollutant, and the carbon dioxide emissions of the construction industry are therefore used as the undesirable output indicator (Zhang et al., 2018). This indicator is estimated using the UN Intergovernmental Panel on Climate Change’s formula (Yin et al., 2015), according to which the conversion coefficient into carbon dioxide emissions from the consumption of various common energies is easily estimated.
Summary of input and output indicators of all variables
| Category | Indicator/variable name | Indicator/variable description | Unit | Source | |
|---|---|---|---|---|---|
| Dependent variable (GTFP) | Input indicator | Labour force | Number of employees | 10 000 CNY | Hu and Liu (2015), Wang et al. (2013), Xu et al. (2016), Gao and Wang (2013), Chen et al. (2016), Nazarko and Chodakowska (2015), Chau et al. (2005), Li and Liu (2010), Zhong et al. (2010) |
| Capital | Total assets of construction enterprises | 10 000 CNY | Wang et al. (2013), Wei and Niu (2013), Xu et al. (2016), Gao and Wang (2013), Zhong et al. (2010) | ||
| Machinery and equipment | Total power of machinery and equipment owned | 10 000 kW | Wang et al. (2013), Wei and Niu (2013), Xu et al. (2016), Gao and Wang (2013), Chen et al. (2016) | ||
| Energy | Energy consumption of construction | 10 000 t | Hu and Liu (2015), Chen et al. (2016) | ||
| Output indicator | Desirable output indicator | Gross output value of the construction industry | 10 000 CNY | Wang et al. (2013), Wei and Niu (2013), Xu et al. (2016), Gao and Wang (2013), Chen et al. (2016) | |
| Total profits of the construction industry | 10 000 CNY | Liu et al. (2016), Xu et al. (2016), Chen et al. (2016), Nazarko and Chodakowska (2015) | |||
| Undesirable output indicator | Carbon dioxide emissions | t | Zhang et al. (2018) | ||
| Independent variable | CER | Number of environmental administrative penalty cases | Piece | Li and Ramanathan (2018), Liu et al. (2018) | |
| MER | Pollution discharge fee | 10 000 CNY | Feng and Chen (2018), Li and Ramanathan (2018), Ren et al. (2018) | ||
| VER | Total number of environmental letters and visits | Pieces | Li and Ramanathan (2018), Feng and Chen (2018) | ||
| Control variables | GDP | Regional GDP | 100 million CNY | Li and Ramanathan (2018) | |
| OSS | Proportion of state-owned assets in total assets of regional construction enterprises | % | Feng and Chen (2018), Chen et al. (2018) | ||
| IDP | Proportion of the gross output value of the regional construction industry in regional GDP | % | Ren et al. (2018), Li and Ramanathan (2018) | ||
1 CNY = US$0.143; GDP, gross domestic product; IDP, industrial development degree; OSS, ownership structure
3.2.2 Calculation model
The EBM-GML model is used to estimate GTFP. Tone and Tsutsui (2010) propose the epsilon-based measure (EBM) model, which integrates the advantages of radial and non-radial models and provides a more accurate measure of decision making unit (DMU) efficiency (Qin et al., 2017; Yang et al., 2018). Oh (2010) constructs the global Malmquist–Luenberger (GML) productivity index, which overcomes the defects caused by using the geometric average and ensures that linear programming can provide feasible solutions. Thus, the directional distance function is defined based on an EBM model, and the GTFP of the construction industry is calculated by using the EBM-GML model.
According to Tone and Tsutsui (2010), it is assumed that there are n DMUs with p inputs and q outputs. The EBM model results in
where γ* represents the optimal efficiency value satisfying 0 ≤ γ* ≤ 1; θ denotes the radial efficiency value; wi is the weight of each input indicator satisfying ; si is a slack variable corresponding to the ith input indicator; ϵx is the integrated parameter of the radial efficiency value and the non-radial slack variable; and λ represents the relative importance of the reference DMU. X = {xik} ∈ Rp×n is the input vector, i = 1, … p; Y = {yjk} ∈ Rq×n is the output vector, j = 1, … q, with X > 0 and Y > 0 always.
After building the directional distance function based on the EBM model and according to the GML productivity index constructed by Oh (2010), the simultaneous production technology aims to provide a reference technology set in period t for each observed DMU. These can be defined as Pt(xt) = {(yt,bt) | xt can produce (yt,bt)}, t = 1, …, T, where bt represents the undesirable output of the DMU in period t. The union of all the simultaneous production technology sets constitutes the global production technology set PG(x) = P1(x1) ∪ P2(x2) ∪ … ∪ PT(xT). Combining the global production possible sets in t and t + 1 periods, the GML productivity index is
where the directional distance function is , s = t, t + 1.
According to Chung et al. (1997), the GML index can be divided into two parts to analyse the causes of productivity changes – namely, the efficiency change index (GECH) and the technological progress index (GTCH). A GML index bigger/lower than unity means that productivity has increased/decreased, while GECH and GTCH bigger/lower than unity mean that efficiency is higher/lower and technology has improved/regressed, respectively. Since the GML index is not the GTFP itself but represents its growth rate, assuming that GTFP in period t is unity, the GTFP in period t + 1 corresponds to the GTFP in period t multiplied by the GML productivity index in period t – that is
3.3 Environmental regulation
3.3.1 Selection of variables
The variables are selected from the ER tools that have regulatory effects on the production behaviours of the construction industry that cause environmental pollution. Then, the following are selected:
the number of environmental administrative penalty cases for CER
the pollution discharge fees levied for MER in China
the total number of environmental letters and visits for VER in each region.
Also, the following control variables are used: (a) economic development level as measured by the regional gross domestic product (GDP), (b) ownership structure (OSS) as the proportion of state-owned assets of the total assets of regional construction enterprises and (c) industrial development degree (IDP) as the proportion of construction industry gross output in regional GDP.
3.3.2 Regression model
A panel tobit regression model is used to explore the relationship between different types of ER and GTFP in the Chinese construction industry. The productivity values calculated by the EBM-GML method always have non-negative values, which require some restricted variables. For the estimation of these variables, ordinary least squares methods are not suitable, as they are likely to produce biased estimation results (Otero et al., 2012). Therefore, a tobit model (Tobin, 1958) is used instead. As the data of 30 provinces in China from 2012 to 2016 are used for empirical analysis, the sample data are short in time span but contain many cross-sectional units. Therefore, a random-effects panel tobit model is used to explore the relationship between the three types of ER and the GTFP. Model 1 is defined as
where i represents the provinces and t is the year. The specific value of GTFP is calculated by using EBM-GML with a one-phase lag (1 year). ERj denotes CER, MER and VER, j = 1, 2, 3. GDP, OSS and IDP represent the level of economic development, ownership structure and industrial development degree, respectively. α0 is the constant term, and ϵi,t is the disturbance term.
In order to examine the non-linear relationships between the three types of ER and GTFP, a quadratic ER term is introduced into model 1 to produce model 2 as
Both models represent the effects of ER on GTFP over the period of analysis. However, considering that the effect of ER on the construction industry GTFP will take some time to materialise, there will be an inevitable time lag. Consequently, further linear and non-linear models (models 3 and 4) are established with one-phase lags in their independent variables. Their control variables are also lagged one phase (year) to avoid a two-way causal relationship with productivity (Rubashkina et al., 2015; Xie et al., 2017).
The selected input and output indicators/variables are summarised in Table 1. The original data are rendered dimensionless by standardisation to eliminate dimensional influence.
4 Results
4.1 GTFP of the construction industry
Each input indicator and each output indicator are positively correlated at the 1% significance level when subjected to the Pearson correlation test, thereby satisfying the monotonic hypothesis – that is, when the input increases, the output cannot decrease.
Assuming that the construction industry GTFP was 1 in 2011, its values from 2012 to 2016 are calculated with the EBM-GML model. The average values of provincial GTFP, GML and their composition for the construction industry in 2012–2016 are shown in Table 2, with the dynamic changes (trend) shown in Figure 1.
Average 2012–2016 values of provincial GTFP, GML and their composition
| Province | GML | GECH | GTCH | GTFP | Up (U)/down (D) trend |
|---|---|---|---|---|---|
| Beijing | 1.0076 | 1.0000 | 1.0076 | 1.0381 | U |
| Tianjin | 0.9839 | 1.0009 | 0.9818 | 0.9398 | D |
| Hebei | 1.0040 | 1.0105 | 0.9940 | 1.0629 | U |
| Shanxi | 1.0004 | 1.0122 | 0.9878 | 1.0349 | U |
| Inner Mongolia | 0.9205 | 0.9323 | 0.9858 | 0.6859 | D |
| Liaoning | 0.9037 | 0.9133 | 0.9869 | 0.8524 | D |
| Jilin | 0.9706 | 0.9697 | 0.9938 | 0.8490 | D |
| Heilongjiang | 0.9782 | 0.9845 | 0.9887 | 0.9714 | D |
| Shanghai | 1.0291 | 1.0152 | 1.0148 | 1.0882 | U |
| Jiangsu | 1.0227 | 1.0000 | 1.0227 | 1.1071 | U |
| Zhejiang | 1.0097 | 1.0000 | 1.0097 | 1.0393 | U |
| Anhui | 1.0247 | 1.0273 | 1.0011 | 1.0732 | U |
| Fujian | 1.0091 | 1.0239 | 0.9882 | 1.0301 | U |
| Jiangxi | 1.0211 | 1.0000 | 1.0211 | 1.1023 | U |
| Shandong | 1.0209 | 1.0172 | 1.0153 | 1.1195 | U |
| Henan | 1.0275 | 1.0296 | 0.9998 | 1.0724 | U |
| Hubei | 1.0489 | 1.0494 | 1.0072 | 1.1825 | U |
| Hunan | 0.9724 | 0.9771 | 0.9941 | 0.9036 | D |
| Guangdong | 1.0166 | 0.9739 | 1.0533 | 1.0784 | U |
| Guangxi | 1.0436 | 1.0000 | 1.0436 | 1.1931 | U |
| Hainan | 0.9588 | 0.9836 | 0.9772 | 0.8702 | D |
| Chongqing | 1.0315 | 1.0000 | 1.0315 | 1.0827 | U |
| Sichuan | 1.0526 | 1.0219 | 1.0308 | 1.2412 | U |
| Guizhou | 1.0077 | 1.0109 | 0.9967 | 1.0991 | U |
| Yunnan | 1.0183 | 1.0223 | 0.9989 | 1.0967 | U |
| Shaanxi | 0.9804 | 0.9898 | 0.9931 | 0.9007 | D |
| Gansu | 1.0138 | 1.0275 | 0.9892 | 1.0542 | U |
| Qinghai | 0.9720 | 0.9939 | 0.9793 | 0.8943 | D |
| Ningxia | 0.9845 | 1.0078 | 0.9749 | 0.9709 | D |
| Xinjiang | 1.0079 | 1.0166 | 0.9916 | 1.0828 | U |
| Eastern average | 1.0008 | 0.9949 | 1.0079 | 1.0349 | U |
| Central average | 0.9960 | 0.9980 | 0.9977 | 0.9861 | D |
| Western average | 1.0076 | 1.0101 | 0.9984 | 1.0470 | U |
| Total average | 1.0014 | 1.0004 | 1.0020 | 1.0239 | U |
Dynamic changes of the provincial GTFP of the construction industry in 2012–2016
Dynamic changes of the provincial GTFP of the construction industry in 2012–2016
Overall, the average GTFP of the national construction industry in 2012–2016 was greater than 1, indicating that, considering environmental factors, productivity was rising. However, the growth rate of GTFP in 2012–2016 was relatively low (average annual rate of 0.14%). Two-thirds of the provinces’ construction industry GTFP was increasing. Among these, Sichuan had the highest annual average GTFP and GML index (its GTFP was 1.2412, with an average annual growth rate of 5.26%). However, one-third of the provinces showed a downward GTFP trend, indicating that the industry was in need of extensive development. For example, Inner Mongolia had an average annual GTFP of only 0.6859, the lowest in the whole country, with an average annual reduction of 7.95%. The national GECH was 1.0004, with an increase of 0.04%, and the GTCH was 1.0020, with an increase of 0.2%. This indicates that the improvement in GTFP was mainly due to technological advancement.
The dynamic changes of the national GTFP, GML and composition of the construction industry in 2012–2016 are shown in Figure 2. From the national perspective, the average GTFP was strongly volatile in 2012–2016, with a slight overall increase. The GECH and GTCH were also volatile. The efficiency of the construction industry increased at first and then decreased. Except for 2012 and 2013, the efficiency changes were all below 1, which hindered the construction industry’s economic growth. Except for 2012 and 2015, the technological progress indices were all increasing.
Dynamic changes of the national GTFP, GML and composition of the construction industry in 2012–2016
Dynamic changes of the national GTFP, GML and composition of the construction industry in 2012–2016
The dynamic changes in the national and regional GTFP of the construction industry in 2012–2016 are shown in Figure 3. This shows that the changes in construction GTFP across the country and in every region were very similar. The eastern region had the highest GTFP in 2012–2013, but the western region surpassed the eastern region since 2014. The central region GTFP was lower than the national average. Its weak economic foundation resulted in the GTFP of the central region to rapidly decline under macroeconomic pressure in 2015, exposing a contradictory relationship between the industry’s economic growth and ecological environment.
Dynamic changes of the national and regional GTFP of the construction industry in 2012–2016
Dynamic changes of the national and regional GTFP of the construction industry in 2012–2016
4.2 Regression results
The values of the Pearson correlation coefficients and the variance inflation factors indicate a low correlation and multicollinearity between the independent variables, respectively.
The authors took the 2012–2016 construction industry GTFP in each province as the dependent variable; the three different ER types as the independent variables; and the GDP, OSS and IDP as the control variables. The random-effects panel tobit model was then used to estimate the relationships between the ERs and GTFP and to analyse their linear and non-linear relationships and the one-phase lags. The results are shown in Table 3. The Wald χ 2 test, the likelihood-ratio (LR) test and the r-values all indicate the suitability of the random-effects panel tobit model. The coefficient of the GTFP lagged one phase is positive and significant at the 99% confidence level. This indicates that GTFP growth had a self-cumulating effect – that is, an increase in the GTFP in the previous year promoted the GTFP growth in the following year.
Regression results of three types of ER combination
| No lag | One-year lag | |||
|---|---|---|---|---|
| Model 1 | Model 2 | Model 3 | Model 4 | |
| GTFP t−1 | 0.4853*** | 0.4907*** | 0.4760*** | 0.4584*** |
| (4.94) | (4.86) | (5.2) | (5.05) | |
| CER | 0.0587 | −0.2059 | 0.1239** | 0.1970 |
| (1.12) | (−0.79) | (2.11) | (0.86) | |
| CER2 | 0.2811 | −0.0797 | ||
| (1.04) | (−0.33) | |||
| MER | 0.1183* | 0.2472 | 0.1063 | 0.1108 |
| (1.7) | (1.07) | (1.53) | (0.5) | |
| MER2 | −0.1292 | −0.0090 | ||
| (−0.55) | (−0.04) | |||
| VER | 0.0155 | −0.1117 | −0.0114 | 0.1874 |
| (0.25) | (−0.74) | (−0.23) | (1.43) | |
| VER2 | 0.1604 | −0.2560* | ||
| (0.86) | (−1.67) | |||
| GDP | 0.0252 | 0.0571 | 0.0601 | 0.0277 |
| (0.35) | (0.55) | (0.92) | (0.33) | |
| OSS | 0.1544** | 0.1583** | 0.2312*** | 0.2301*** |
| (2.36) | (2.43) | (3.65) | (3.60) | |
| IDP | 0.2600*** | 0.2925*** | 0.1429** | 0.1265*** |
| (3.24) | (3.45) | (2.00) | (1.68) | |
| Constant | 0.3632*** | 0.3544*** | 0.3818*** | 0.3877*** |
| (4.03) | (3.91) | (4.38) | (4.54) | |
| σ u | 0.0584*** | 0.0570*** | 0.0620*** | 0.0640*** |
| (4.01) | (3.50) | (4.71) | (4.59) | |
| σ e | 0.0680*** | 0.0678*** | 0.0678*** | 0.0666*** |
| (13.62) | (12.9) | (14.31) | (14.08) | |
| r | 0.4248 | 0.4146 | 0.4555 | 0.4798 |
| LR test | 8.13*** | 6.04*** | 9.99*** | 10.09*** |
| Wald χ 2 | 104.57*** | 107.84*** | 92.95*** | 97.32*** |
| Log likelihood | 167.2559 | 168.1943 | 166.2314 | 167.6217 |
Numbers within parentheses represent the z-values. σ u, error due to differences between units; σ e, error due to differences within units. *, ** and *** represent statistical significance at 10, 5 and 1% levels, respectively
For CER, only the one-phase lagged linear relationship with GTFP is significantly positive (at the 95% confidence level). Hence, H1a is supported when CER is lagged one phase, but H1b is rejected. For MER, only the current phase linear relationship with GTFP is significantly positive (at the 90% confidence level), meaning that H2a is supported, while H2b is rejected. For VER, only the one-phase lagged quadratic term is significant and negative (at the 90% confidence level). Hence, H3a is rejected, but H3b is supported when VER is lagged one phase.
On the other hand, by comparing the effects of combined and separate ERs, CER lagged one phase still has a significant positive linear relationship with the construction industry GTFP (at the 90% confidence level). MER in the current phase still has a significant positive linear relationship with the construction industry GTFP (at the 90% confidence level). As for VER, the linear and non-linear relationships are insignificant in the current phase or lagged one phase. Finally, by comparing the regression results of the three types of ER combined with the separate ER, it is found that when the ERs work together, the regression coefficient of the CER lagged one phase increases from 0.1145 to 0.1239, and the confidence level increases from 90 to 95%. In addition, the regression coefficient of MER in the current phase increases from 0.1152 to 0.1183.
4.3 Robustness check
In order to validate the models, a robustness test on the estimation results was conducted using a replacement regression model. The panel generalised estimating equation (GEE) model was used to replace the panel tobit model for regression analysis, and the clustering robust standard deviation was also used to estimate the standard deviation. After the panel GEE regression, the coefficients of the core variable and significance remained essentially unchanged, meaning that the model results above were robust.
5 Conclusions
This study uses 2012–2016 panel data from 30 Chinese provinces to determine the national and regional GTFP of the construction industry. It also analyses the relationship between three types of ER and their lag effect on the GTFP. The main findings are as follows.
From 0.999 in 2012 to 1.002 in 2016, the national average construction industry GTFP had a slow upward trend, with an average annual growth rate of only 0.14%. One-third of the provinces had a downward trend, indicating that the construction industry in some provinces was still undergoing extensive development. In particular, Inner Mongolia and other provinces with low GTFP need to strengthen the coordination between construction industry development and environmental protection. The improvement in GTFP was mainly due to technological advancement. In the future, the construction industry needs to improve both its technological efficiency and progress, transform its economic growth mode, rationally adjust and upgrade its industrial structure, fully optimise resource allocation and reach a balanced sustainable development.
There are differences in the impact of different types of ER on GTFP. CER has a positive linear relationship with GTFP, but despite a significant 1-year lag effect, the impact of long-term regulation is even more significant. MER and GTFP have a more immediate positive linear relationship. This is the consequence of a faster adjustment due to the openness and dynamics of the market. VER lagged 1 year and GTFP have an inverted U-shaped relationship, since the public’s response to the government or the market requires some time.
The government should combine the three types of ER – that is, they should implement CER from the perspective of long-term supervision, implement MER from the perspective of short-term supervision and guide the public to watch corporate environmental behaviour in the construction industry (i.e. VER).
The study’s main contribution is to outline the practical implications of implementing different types of ER when promoting sustainable development in the construction industry. The study considers undesirable output and energy consumption in the productivity measurement framework for modelling GTFP. Moreover, the separate and combined effects of three types of ER on the GTFP of China’s construction industry are examined for the first time. This will be conducive for the government to develop a combination of ER policies that support sustainable development in the construction sector.
This study is limited by the constraints of data accessibility, so that the indicators selected to measure different types of ER may not be fully representative. It is also limited to China and other similarly placed countries. Therefore, further research may involve collecting and comparing data from other indices and different countries to compare the impact of different types of ER more generally.
Acknowledgements
This research is supported by the National Social Science Fund post-financing projects (number 19FJYB017), the Humanity and Social Science Program Foundation of the Ministry of Education of China (number 17YJA790091) and the List of Key Science and Technology Projects in China’s Transportation Industry in 2018-International Science and Technology Cooperation Project (numbers 2018-GH-006 and 2019-MS5-100). The sixth author acknowledges the Spanish Ministries of Science, Innovation and Universities for his Ramon y Cajal contract (number RYC-2017-22222) co-funded by the European Social Fund.
