Exploring differences in level of construction innovation: an empirical analysis in China

Innovations can overcome constraints posed by resource depletion, increasing environmental and ecological protection concerns. There is considerable amount of innovation that occurs in the construction industry. Accordingly, construction innovation is receiving increasing attention in China. However, the provincial development level of construction innovation remains unclear. To address this gap, the present study employs data-driven measurement for the level of construction innovation. A total of 25 alternative indexes were selected based on the innovation ecosystem theory. Then, text preprocessing, statistical methods, and search statistics were employed to acquire index data. The indexes weights were determined through expert scoring and a cloud model. The quantitative measurement of the level of construction innovation was finally performed. Additionally, the exploratory results were revealed with the analyses of the overall, regional, and cluster. The results revealed that overall level of construction innovation in China is not high and regional distribution is uneven. Simultaneously, the level of construction innovation is consistent with local economic strength, and it is most sensitive to innovation output in regional level. Moreover, the spatial distribution of the level of construction innovation in China was showed, which has similarity with the characteristics of geographical location. The measurement system this study represents breakthroughs over traditional methods that rely on statistics, cases, or questionnaires, which can be applied to other research fields.


Introduction
The construction industry, one of the largest economic sectors, is critical to the functioning of a domestic economy in China. The construction industry has a wide range of responsibilities, including improving the living standards of people, boosting employment, and protecting the environment (Manseau 1998). However, buildings consume 46.5% of Chinese energy, produce more than 1.8 billion tons of waste, and contribute 51.3% of CO 2 emissions. It is widely accepted that the productive activities of the construction industry consume significant amounts of resources and negatively impact the ecological balance (Peter et al. 2012). Innovation promotes sustainable development and benefits, such as energy consumption reductions, environment protections, and productivity advancements (Sexton and Barrett 2003). Construction innovation, which has drawn wide attention within the world, is a determinant of the sustainable development of this industry and national growth (Panuwatwanich and Stewart 2012).
Construction innovation, which includes product, process, marketing, and organizational innovation, represents the changes through something new and can bring benefits or improvements in cost, time, quality, safety, and environment (Maria and Victor 1995;Pierce and Delbecq 1977;Bygballe and Ingemansson 2014;Bassioni et al. 2004;). The innovation in the construction industry is inactive, and governments and related firms lack the correct cognization of the level of construction innovation in China, because it is difficult to collect the empirical data in the construction sector (Meng and Brown 2018). The construction industry in China has serious uneven development due to economic condition, ecological environment, and differentiated policies (Wang et al. 2021;Cole 1999). Accordingly, there are great differences in the level of construction innovation in China. Research on the differences in level of construction innovation is a topic of great importance for governments and has been investigated with positive results by many scholars (Yan and Liu 2010). The current research on the differences in different regions focuses on three main areas: the differentiated influences of factors, the change trend of development level, and the spatial distribution in level (Liu et al. 2016;Chen et al. 2017). Although being influenced by the same factor, the level of construction innovation in different regions is significantly different. Scholars had explored the regional differences in level with being influenced by factors of productive efficiency, environmental protects, and construction cost and policy (Chau 1993). The difference directly led to the change trend of development level. Several models including the two-hierarchical analysis framework and the stochastic frontier analysis (SFA) model had been used to measure the change trend (Chen et al. 2017;Bai 2013). Different regions can present a hierarchical spatial distribution in terms of the similar characteristics. The assessment system is the most common method to explore the differences in level of construction innovation, because this method can be used to explore the differences under the influence of multiple factors (Ramirez et al. 2004;Xu et al. 2019).
Previous research always focuses on the differences in level of construction innovation in projects, firms, and nations (Yang and Huang 2016;Bossink 2004). Many factors including technology, management, and environment have been considered in the difference exploration (Xu et al. 2019;Yang and Huang 2016). However, limited by data and analysis methods, there is not a comprehensive and broadly applicable assessment process to explore the differences in level of construction innovation from a national level (Bygballe and Ingemansson 2014;Yitmen 2007). To address this gap, the present study employs data-driven measurement for the development level of construction innovation. A measure system is built based on innovation ecosystem theory. The indicators are comprehensive and systemic, which broadens the perspective of the measurement for the level of construction innovation. Simultaneously, the measurement index quantification driven by data in this study represents breakthroughs over traditional methods that rely on statistics, cases, or questionnaires. Subsequently, an empirical case study on 30 provinces in China is conducted to analyze the causes for the differences of regional construction innovation. This research attempts to address the following questions: • How to develop a suitable method for measuring the level of construction innovation in China? • What causes the difference in the level of construction innovation in China? • How to improve the level of regional construction innovation in China?
In this research, the index weight is calculated with the cloud model, and the objective data is obtained using the method of text preprocessing. The cloud model is a new approach to study the uncertain transformation between a qualitative concept and its quantitative instantiations and has been widely used for evaluating development level (Li et al. 2009). Additionally, the text preprocessing can obtain valuable information from a large amount of text data and is an efficient method to obtain objective data (Dou et al. 2019). The research methods provide a theoretical support for exploring the differences in level of construction innovation in China. This paper is structured as follows: "Literature review" reviews past literatures; this is followed by a discussion on the research framework, the construction of measurement system, the acquisition and quantification of indicator data, and the calculation of indicator weight in "Methodology" "Results analysis" elaborates on the empirical analysis results. The last section summarizes the main findings.

Innovation in the construction industry
Innovation is a critical contributor to the growth in the construction industry (Manseau 1998). The concept of construction innovation is initially used to describe the first use of new technology within construction firms (Tatum 1987). Now, the most acknowledged definition of construction innovation is that a process to develop, disseminate, and apply a new or improved product and service (Na et al. 2006;Steele and Murray 2004). Given this definition, scholars conducted relevant researches including the drivers of innovation and the influencing factors which can foster or hinder innovation for improving innovation within the construction industry.
A number of previous researches have discussed various drivers of the development and adoption of innovation in construction firms (Slaughter 2000). The initial objective of implementing innovations in firms is to increase the competitive advantage in construction markets with new technologies (Pelliceret al. 2014). With the complication of engineering construction projects, client requirements and project performance improvement provide new incentives for innovation (Ozorhon 2013). Recently, in order to ensure the sustainability of construction projects, changes toward sustainability have been regarded as a key driver of innovation (Dewick and Miozzo 2010). The changes include concerns to environment from managers and regulatory pressures from the sectors of environmental protection (Qi et al. 2010). Firms have a good innovation environment with these drivers and begin to innovate actively.
A solid body of work centered on the factors to foster or hinder innovation has been achieved by researchers. Tatum (1989) and Aronson and Lechler (2010) first proposed that the philosophy and organizational culture in construction firms influence innovation behavior. The preference of upper management for innovation promotes firms to make business strategies that can improve the innovation performance. Employees' innovation thinking can be triggered through fostering an innovation-supportive culture. A good material foundation is another advantage factor for innovation. A large number of technical personnel and sophisticated equipment can increase the success rate of innovation (Wen et al. 2020). The firms increasingly innovate with these fostering factors.
The construction industry lacks innovation (Brockmann et al. 2016). The factors hindering innovation include the structure of project organization, innovation value, and the complexity of the construction process (Winch 1998). A project involves many organizations with different demands for interests. The separated responsibility among organizations reduces the possibility of innovation (Slaughter 2000). The lack of long-term cooperation relationships further increases the difficulty of innovation diffusion and adoption. Innovation is difficult to be boosted within a short term (Weerd-Nederhof 2007). Firms have little momentum to innovate with the pressure of high costs and high risk from innovation. With the increases of major engineering, the firms prefer to refuse innovation due to the more complex construction technology and higher uncertainty of innovation.

Assessment of the level of innovation
The assessment of the level of innovation is crucially considered in the context of firms, industry, or regions . The innovation in firms emphasizes technology innovation, the capacity of innovation management, and the influence of innovation Cáceres and Rekowski (2011). An assessment system based on the dimensions of inputs and outputs was used to measure the level of technology innovation in firms. However, the system ignored the role of managers (Flor and Oltra 2004)). Lobo and Samaranayake (2020) combined the stage-gate model to build the assessment system of innovation management and proved that effective innovation management can improve the level of innovation. The level of innovation is also influenced by other factors, such as investor experience and size of firms (Boh et al. 2020). Firms achieved better innovation through assessing the impacts of these factors.
The assessment of the level of industrial innovation involved many perspectives including patent protection, industry structure, innovation environment, and technology (Sun 2010;Bi et al. 2016). Sun (2010) measured the level of industrial innovation in China based on various indicators, including patent grants, new product sales, and R&D spending with the data of statistical yearbook. The industry structure is the bottleneck that hampers the level of innovation of the construction industry (Sha et al. 2008). In order to improve the negative effect of industry structure, Chen et al. (2013) assessed the innovation environment of the industry using data envelopment analysis (DEA) and found that technology is a good way to improve the negative effect.
Since the interpretation of innovation in different industry is totally diverse, the assessment of the level of regional innovation is complicated (Daugeliene and Juocepyte 2012). Previous researches suggested that the level of regional innovation was measured by innovation organizations, innovation policies, and patent. Cooke (2004) achieved assessment of regional innovation from a dual perspective with institutional and organizational and emphasized the importance of a national innovation system. Indians accurately assessed technological advances and innovations with the analysis of patent data from the Indian patent office (Abraham and Moitra 2001). In addition, the concept of sustainable development provides the assessment with a new perspective (Smith et al. 2010). An eco-sustainable framework has been built to measure the level of innovation in Australia. After summarizing indicators of predecessors, Daugeliene and Juocepyte (2012) built an assessment system from various dimensions including environment, innovators, economic effect, and scientific output and realized systematic criteria for the assessment.
The assessment of the level of construction innovation can be conducted from two dimensions of technological and management (Manley 2008). The assessment of new technology aims to enhance the economic and environmental values (Kajikawa et al. 2011). There are many models with the assessment of the technological impacts to cost and societal benefit (Chang et al. 1988). These assessment systems include many dimensions of resource input, product output, innovation environment, innovation diffusion, and market structure. According to project data, Carrillo et al. (2019) assessed the reliability of project technology and encouraged participants to improve their construction technology to increase productivity. With increasing attention paid to the sustainable construction, an assessment system of green technology has been proposed, which incorporated the concept of sustainability. Since the influence of society, economy, and ecological environment is widely considered, the assessment system of green technology is more advantageous than traditional technology assessment.
Managers, stakeholders, and innovation network receive extensive attention when measuring the level of construction innovation from the management perspective (Rothschild and Darr 2005). The level of innovation in construction firms can be measured through quantifying the preference degree of manager with the methods of questionnaires and focus categories. Noktehdan et al. (2019) researched the management behavior of innovation activities using practical project case and argued that the key indicators of measuring innovation are different at various stages of project. The assessment of the level of innovation at various stages of project can respectively provide the basis of decisions for stakeholders, including investors, policy-makers, and contractors (Shahpari et al. 2019). Moreover, innovation networks consisting of many stakeholders have been used to assess the innovation performance that is closely related to the level of innovation (Zhang et al. 2021).

Innovation ecosystem
The concept of innovation ecosystem was first put forward by Adner (2006). It is defined as a system that can meet the need of consumer with the recombination of advantageous resources among firms. Afterwards, scholars defined innovation ecosystem from different perspectives (Asheim and Gertler 2005). The most widely used definition is that a complex system includes the evolving set of actors, activities, artifacts, and the institutions and relations that are important for the innovative performance of actors. The definition emphasizes the complementary, cooperative, substitute, and competitive relations.
The concept of innovation ecosystem has been widely applied in various fields to solve different problems focused on a firm, industrial, and national level (Oh et al. 2016;Philip 1992). Previous researches have built the innovation ecosystem at firm level so that firms can increase the competitive advantages and identify risks quickly (Adner 2006). The most serious risk of applying innovation ecosystem is to decrease the firms' enthusiasm of innovation. Due to this deficiency, Song (2016) built an interaction model based on innovation ecosystem to study the impact of the interactive activities among members on cooperative innovation performance. Then, the US Council on Competitiveness proposed the national innovation ecosystem for sustaining national innovation vitality (Fukuda and Watanabe 2008). The national policy has been shifted from technology policy to the innovation policy based on the ecosystem concept.
The concept of innovation ecosystem has been used to build the assessment system of the level of innovation (Tolstykh et al. 2020). According to the concept of industrial innovation ecosystem, Yan et al. (2020) provides a comprehensive set of technology indicators including technology, policy, economic, new business, and industrial clusters to measure the level of technology innovation in science parks. The innovation ecosystem is also applied to measure the level of innovation in the relatively low-tech forestry sector, and the assessment system was built from the three dimensions of the actors' landscape, knowledge and information provision, and project actions and interventions (Minang et al. 2019). Based on the above indicators, Cai and Huang (2018) built an assessment system including innovation resources, resource flow, basic environment, and policy environment and found the uneven level of innovation in 30 provinces of China. Moreover, since digital technology pervades many industries, the indicators of strategic management and key actors have been incorporated to the assessment system that was built based on the concept of business innovation ecosystem (Mohelska and Sokolova 2016).

Research framework
A systems approach in studies of complex phenomena has a long tradition in a broad range of disciplines. Innovation research paradigm has experienced the development process of linear innovation, innovation system, and innovation ecosystem (Amitrano et al. 2018). Innovation ecosystem is viewed as a situational analysis perspective argued to be very beneficial for innovation research (Jin et al. 2022). Innovation ecosystem introduces the concept of "ecology" into the innovation system and observes the economy and society with the natural ecological law. Construction innovation was originally regarded as the application of new technology to construction. With the updating of the concept of innovation, construction innovation has been given more significance, including management innovation, institutional innovation, and organizational innovation. Construction innovation has been greatly enriched. The research basis of this paper is to deduce the measurement system of construction innovation level from the general theory of innovation ecosystem.
Accordingly, the influence of construction innovation can be considered by integrating four elements, innovation environment, innovation subject, innovation input, and innovation output, from the perspective of innovation ecosystem. This research explores the differences in level of construction innovation in 30 provinces of China using a comprehensive and broadly applicable assessment system. Firstly, the system of indicators was constructed based on the innovation ecosystem theory. The second-level indicators of the assessment system were determined with building-related literature. Secondly, the initial data of indicators were obtained with three methods of search statistics, statistical data, and text preprocessing and quantified using the linear dimensionless method. Thirdly, considering expert scoring, the indicator weights were calculated by applying a cloud model. Finally, the assessment results of the level of construction innovation can be obtained based on indicator data and their weights. The causes leading to the differences are analyzed considering overall and region and the spatial distribution according to the level categories. The framework of this research is described in Fig. 1.

Development a system of indicators
Innovation ecosystem, which is characterized by diversity, dynamism, network, and openness, is an open complex system where different interest subjects and system elements adapt to each other, coordinate and integrate, and evolve symbiotically. System elements include information, capital, technology, and human resources (Yin et al. 2020). Construction innovation is the process of creating value through the comprehensive action of capital, information, talents, and other innovative resources under the innovative environment. It can be found that they are similar in structure and function by analyzing construction innovation activities from a systematic perspective. As is shown in Table 1, construction innovation activities are the exchange of material, energy, information, and capital between the ecological community and the ecological environment in a certain space and industrial field. The ecological community is formed by enterprises, government institutions, and scientific research institutions. The ecological environment includes the economic, political, technological, and market aspects. Under the interaction of various elements, a comprehensive system of competition, symbiosis, and collaborative evolution is formed.
Based on the innovation ecosystem theory, the system of indicators for the evaluation of the level of construction innovation is built from four dimensions: innovation environment, innovation subject, innovation input, and innovation output as is shown on Fig. 2.

Innovation environment dimension
The quality of the innovation environment directly affects the output of industrial innovation activities (Cole 1999). The innovation environment can be considered from both the political environment and the social environment. The policy support of governments can effectively improve the level of construction innovation. The most powerful methods to improve the level of construction innovation for the local governments are introducing technical personnel, encouraging innovation boldly, and strengthening the protection of innovation achievements (Wang and Zou 2018). The more social resources are invested, the more benefits the construction innovation activities. Innovation activities are inseparable from the investment of talent, technology, and capital resources and other social resources (Russo 1997). In view of the considerations mentioned above, the indicators of innovation environment have been selected.

Innovation subject dimension
The innovation subject is the main carrier of innovation activities and plays a leading role in innovation behavior. The impact of innovation subject on the level of construction innovation is from two perspectives, both the subject and the cooperative relationship among subjects. Universities and research institutes are creators of new ideas, and firms play an important role in the conversion and promotion of the innovation achievements (Latham and Bas 2006). The cohesion among subjects decides the innovation efficient of the system. The impacts of subjects on the level of innovation are measured with the number of universities, research institutions, construction firms, and industrial parks.

Innovation input dimension
Innovation input can be examined from three different angles: policy intervention, market regulation, and the efforts of innovation subject (Tatum 1986). From a policy perspective, the influence of local governments on construction innovation can be analyzed from the numbers, sources, and types of the construction technology policies (Dou et al. 2019). From a market perspective, the market share of local firms and the care for buildings from people promote indirectly the level of construction innovation. The innovation activities of firms are inseparable from the support of human, material, and financial resources. In view of the above analysis, we can get relevant indicators of innovation input.

Innovation output dimension
Innovation output, which embodies the efficiency of innovation activity, is an important dimension to measure the level of regional innovation (Fritsch and Slavtchev 2010). Innovation output can be measured from three dimensions of the knowledge and technology output, the economy output, and the society output. The number of building technology patents and building-related papers represents the knowledge and technology output (Murovec and Prodan 2009). The knowledge spillover and diffusion depends on the collaboration of innovation subjects. Firms, who gain benefits for themselves and bring a lot of social benefits, are the main contributors of economic and social output (Russo et al. 1997). The determination of economy and society output indicators is closely related to firms.

Data acquisition and processing
All of the indicators in Table 2

Innovation environment indicator data
The indicators included in this subgroup reflect political direction and social support in the context of innovation environment. The content of political indicators includes innovation, talent, and the protection of intellectual property. The data is crawled with the data crawling software of octopus and is classified by region after cleaning and structuring the data. The investment of talent, technology, and capital indicate local social support. The data of this resource is from China Statistical Yearbook. The dimensionless process of E1 is listed in Table 3. The dimensionless process of others' indicators is similar to E1.

Innovation subject indicator data
The driver of construction innovation can be attributed to the innovation subjects or their collaboration. This research focuses mainly on construction firms, universities, and related research institution. The selection of these subjects is simply because they can generate innovative knowledge and apply innovation. The number of construction firms and universities can be queried on the Statistical Yearbook, and the data of research institutions is from the open resources of information firms. Industrial parks consisting of various firms have the greater innovation potential, and the data is from a public database. The dimensionless process of these indicators is similar to E1.

Innovation input indicator data
The political support of the government, the resource investment of the construction industry, and social promotion have been the driving force behind construction innovation. Governments provide the construction industry with many opportunities to innovate and have issued a large number of incentive policies involving technology, material, and equipment. The initial data of policies is obtained with the method of text preprocessing. The implementation intensity of policies is quantified according to different sectors of issuing policies. A 5 to 1 Likert scale is assigned to the corresponding policies from high to low, respectively, according to the administrative levels of the provincial people's congress, provincial government office, bureaus, municipal government offices, and construction bureaus. This study takes the average of policies' values as the quantized data. Meanwhile, the degree of encouragement of various policy types is different for innovation. Three policy types of technical work processing, technical work document issuance, and technical standard document issuance take respectively 1, 3, and 5 points. The construction industry also invests large resources including talent, technology, and capital that can promote innovation. The data of related indicators can be obtained from Statistical Yearbook. The related indicators of social promotion embody the support of local society. The data of market share is from Statistical Yearbook. The degree of public concern for construction innovation is reflected by the search frequency of keywords in Baidu Index. The dimensionless processes of I2 and I3 are similar. The dimensionless processes of others' indicators are similar to E1.

Innovation output indicators data
Innovation output presents in the form of knowledge, technology, economy, and social benefit. This research takes scientific papers and patents as the knowledge and technology output of construction innovation. The method to obtain data of patents is similar to the indicator of E1. The data of scientific papers is from Chinese National Knowledge Infrastructure and can be quantified according to different journal levels. The journal levels have three types; the first level is SCI, EI, and Peking University Indexed core; the second level is CSSCI and CSCD, and others are the third level. The paper can obtain ); j, r = 1, 2, …, n C 1j indicates the number of innovation policies issued by local government in region j E 1j indicates the value of this indicator in region j ); j, r = 1, 2, …, n F 8j indicates the total carbon emissions of the construction industry in region j that year P 8j indicates the value of this indicator in region j 3 points, 2 points, or 1 point depending on the journal level. The final data of scientific papers is from the accumulation of scoring results. The dimensionless process of P2 is listed in Table 2. The related indicators of economy extracted from the statistics include total output value, labor productivity, and total profit and tax in the construction industry. The data of these indicators is from China Statistical Yearbook, and the dimensionless processes are similar to E1. The contribution of construction innovation to society is reflected by the increased employees, the decreased pollution, and the obtained honors. The dimensionless process of P8 is listed in Table 2. The dimensionless processes of P6 and P7 are similar to E1.

Indicator weight calculation process
Cloud model, proposed by Li Deyi, is an uncertain transformation model of handling qualitative concepts and quantitative descriptions (Li et al. 2009). The determining of weights with cloud model is in line with the laws of human understanding of the objective world (Wang et al. 2015). Cloud model is used to determine indicator weight in this research. The application process of cloud model is shown in Fig. 3. Three eigenvalues of EX (expectation), EN (entropy), and HE (hyper entropy) can be obtained by running the reverse cloud model after obtaining the scoring results of n experts on an indicator. After determining cloud drop number, a cloud graph can be obtained through running the forward cloud model. It indicates that there is no problem with the data when the graph is clear and shows a normal distribution. The indicator weight can be indicated with EX. We can request the experts to re-value the indicator and re-run the cloud model if the graphics are not clear.

The method of expert scoring
Expert scoring results are the data basis of cloud model operation, and the weight calculation with cloud model is regarded as the visual optimization of expert scoring results. The method of expert scoring is used to obtain expert scoring results, which can be used to solve problems with the expert knowledge and experience. Generally, this method adopts questionnaire survey to collect data, and the final data is the result of expert consensus after collection, feedback, induction, and modification of expert opinions for many times. In order to ensure the reliability of data, the selected experts should be familiar with the status quo of innovation in the construction industry and have high authority and representativeness, and the number of experts should be appropriate.
The product output of the construction industry is based on the project. Many construction innovations take place at the level of project, mainly involving general contractors, construction units, design units, scientific research institutions, and many other participants. Considering the comprehensiveness of innovation subjects, experts from universities, scientific research institutions, real estate, construction enterprises, and relevant government departments will be invited respectively. The selection criteria of experts are indicated in Table 4. In addition, relevant studies prove that when the number of cloud drops is greater than 10, the error of EX is less than 0.01 (Lv et al. 2003), that is, the number of selected experts should be greater than 10.
In order to ensure the reliability of weights, 20 experts have been invited to score the importance of indicators, and 15 experts retained the scoring results finally. Sample screening steps of expert scoring results are as follows: 1. Invite 20 experts in different work units who meet the selection criteria to judge the influence of various indicators in the measurement system on the level of construction innovation according to their experience and give appropriate scores 2. Collect 20 pieces of scoring data; EX (expectation) can be calculated with the reverse cloud model. The data with the biggest gap with EX will be removed. The reverse cloud model can be used to calculate the EX of the remaining 19 pieces of data, and the data with the biggest gap with EX will be removed. In the end, 15 pieces of data are retained 3. Run the forward cloud model based on the existing 15 data, and output the cloud image. If the cloud image Fig. 3 The running process of cloud model is dispersive, 15 experts will get the feedback and be invited to give a new score. Through many attempts, the cloud image is agglomerate 4. The EX of the final scores of 15 experts will be used to calculate weight

Weight calculation
Fifteen experts from universities, research institutions, and firms were invited to value for each indicator, and the total value of each indicator is 100 points. Take the indicator of P1 as an example, the scoring result of P1 from 15 experts was < 95, 90,90,90,90,98,85,90,80,85,80,92,85,90,95 > . The results of EX, EN, and HE are respectively 88.45, 5.43, and 1.01 with running the reverse cloud model. A cloud drop represents the value of P1 from an expert. We simulated 500 cloud drops in the forward cloud model, and the cloud graph is indicated in Fig. 4. The cloud graph is dispersive through observing Fig. 4, and there is a discrepancy of the experts' opinions. A cloud graph with good cohesion can be obtained after revaluing by the experts. The cloud graph is shown in Fig. 5. The verification of the indicator of P1 was passed when the experts' opinions are unanimous. The expected value of P1 is 89, which are the data of the next step of calculating indicator weight. Calculating weights of all indicators, the weights are shown in Table 5.

The overall level of construction innovation in China
The values and rankings of the level of construction innovation in 30 provinces are obtained with the comprehensive calculation based on the dimensionless data of "Data acquisition and processing" and the indicator weight of "Indicator weight calculation", as shown in Fig. 6. The average value of 30 provinces is 3.4 while a perfect value is 10. The level of construction innovation in China is not high. The figure also demonstrates the average values of the level of construction innovation of four major economic regions in China. As can be seen from the information presented in the figure, the  . 6 The values and rankings of the level of construction innovation in 30 provinces average values of eastern, central, northeastern, and western are 4.8, 3.6, 2.6, and 2.1, respectively. Meanwhile, more than half of the provinces valued below the average value, with a difference of 7.2 between the highest and the lowest. The huge gap demonstrates the significant differences in level of construction innovation in various regions of China.
As shown in Fig. 7, besides individual provinces, the rankings of the level of economy and the level of construction innovation are similar. The impact of economy on the construction innovation is significant. The level of construction innovation in the four major economic regions has unique regional characteristics.
The high-level economy attracts many labors and highquality talents. It is clearly visible that these provinces in eastern can accept faster the newest information that can develop innovative thinking and facilitate the communication of domestic and foreign countries. The level of construction innovation in the eastern region is far ahead of other regions. The level of innovation of the central region is a little higher than the national average level. Around the Yangtze River Economic Belt, the information can flow from the east to the central region directly. The central region, which is the important hub of the railway, plays an important role of carrying on the industrial transfer of the east and driving the development of the west. The northeast region with the unique economic characteristics is consisted of three provinces of Liaoning, Heilongjiang, and Jilin. Shenyang, the capital of Liaoning province, is in line with Beijing to help the development of Beijing-Tianjin-Hebei. The construction industry in the northeast has great potential. The western region has the largest number of provinces and the largest area. However, limited with the factors of location, economy, and population, the level of construction innovation is low. The level of construction innovation in the western can be further improved through promoting the development of economic circles, which are centered on Chengdu, Chongqing, and Xian.

The regional level of construction innovation of 30 provinces
The different values of four dimensions of innovation environment, subject, and input and output in each province are listed in Table 6. Through comparing the ranking of four dimensions and the comprehensive ranking, the level of construction innovation is most sensitive to innovation output. The results also show that the bad innovation atmosphere and low innovation efficiency lead to the low level of innovation.
Deeply analyze the impact of different factors on the level of construction innovation according to the values of different dimensions in each province. As shown in Fig. 8, four dimensions respectively account for about 2.5 points. The values of innovation input are generally greater than that of innovation environment and innovation output; there is not a good innovation atmosphere, and the overall innovation efficiency is not high. The ranking of innovation output is closest to the comprehensive result. It is the most feasible to measure the level of construction innovation with innovation output while not considering other dimensions. The value fluctuation of innovation subject is the biggest, and the impact of innovation subject on the level of construction innovation is relatively small.
Verify the consistency of the ranking of each province under different dimensions; the differences are demonstrated in Fig. 9. The deviation values of Liaoning, Yunnan, and Inner Mongolia are greater than 13 and the sub-item deviation value of Beijing is greater than 7.
The sub-item deviation of innovation subject in Beijing is 10. Beijing is a developed city with good innovation environment and a lot of construction innovation output. However, the value of innovation subject of Beijing is poorer than its comprehensive value. As a municipality, Beijing is limited by the geographical factors and is small compared with other provinces. The number of construction firms and universities Fig. 7 The ranking difference between economy and construction in China is restricted to a certain extent. Beijing is not suitable for large-scale industrial agglomeration. In general, the number of innovation subject in Beijing is limited.
The deviation of four dimensions in Liaoning amounts to 21. Liaoning Province is one of the three northeast provinces. The relevant policies issued by the government of Liaoning Province in 2016 have effectively promoted the construction of innovation environment. However, the ranking of the Liaoning in the 2018 provincial GDP growth statistics is low, and a large number of highly educated talents in the province are lost. The population of the province has been decreasing year by year since 2014 with the influence of the one-child policy. The lack of innovation motivation has a bad impact on the development of the construction industry.
The deviation of four dimensions in Yunnan amounts to 18. The values of innovation environment and innovation output of Yunnan are both lower than the comprehensive value. The number of innovation policies is low due to the weak innovation sense of government. The government has not played its due leading role in construction on innovation. The total output value of the construction industry has an average annual growth rate of more than 15% based on the increasing economic strength. Nevertheless, the number of building-related papers is few, and the number of new employees in the construction industry has grown negatively. The overall innovation output of the construction industry is weak.
The deviation of four dimensions in Inner Mongolia amounts to 18. Inner Mongolia has a good innovation environment that has greatly improved the numbering of Mongolia across the country. This is because Inner Mongolia has been advocating optimizing the innovation environment and creating a good innovation atmosphere. The objects of carrying out the strategic deployment of building an innovative Inner Mongolia issued by the central government and the district government have been insisted by Inner Mongolia. However, Inner Mongolia lacks the innovation subject and has insufficient innovation motivation due to regional restrictions, the lack of education, and economic backwardness.

The spatial distribution in the level of construction innovation
The software of Statistical Product and Service Solutions (SPSS) can be used to achieve a hierarchical cluster analysis. This software provides a simple and useful method of the Q-type clustering to achieve the regional classification of 30 provinces in China. The block method is finally adopted after trying many methods of squared Euclidean distance, cosine, Pearson's correlation, Chebyshev, block based on the inter-group connectivity. The result of classification is indicted in Fig. 10 where 30 provinces are divided into six categories. In order to verify the credibility of the results, 30 provinces were classified again using the partially ordered set (Kelly 1981). According to the ranking of indicator weight and the original data, the partial order relationship among assessment objects is built with the method of partially ordered set, and the structural The objective classification result is shown in Fig. 11. The final classification result is listed in Table 7 after considering comprehensively two classification results that are roughly consistent. Figure 12 shows the spatial distribution in the level of construction innovation among the 30 provinces with an intuitive way. As can be seen from the information presented in this figure, the result of spatial distribution is similar with the characteristics of geographical location. There are obvious regional differences of the level of construction innovation of China due to physical, economic, and cultural environments varying across regions. The difference is most obvious among the east coast, central provinces, and western areas. The construction innovation in China is highly concentrated in the eastern coastal areas.
Jiangsu, who owned the highest assessment value, is a separate category. As coastal areas with highly developed economy, this province contributes 13% of China's total output value of the construction industry and owns the most universities and construction firms that can provide a steady stream of power for innovation. Its output of knowledge including scientific papers and patents also is the most in China. The level of construction innovation of Jiangsu is far beyond other's provinces.
The second category comprises Beijing, Guangdong, and Shandong. Guangdong and Shandong, which locate in eastern coastal areas, have absolute economic strength to build innovation environment. Guangdong and Shandong have a large demand for buildings and attract many labors due to being the most populous in China. There are also a large number of universities that provide many high-quality talents. The unique advantages of geography, economy, and education have created a high level of construction innovation in these provinces.
Sichuan, Henan, Hebei and Hubei, which are populous provinces, provide a lot of labors for China. There is also Fig. 10 Provincial clustering process with SPSS a large demand for buildings due to the high population density of these areas. These provinces locate in the geographical middle of China and own the strong power in policy, economy, and culture. Zhejiang and Fujian, which are the coastal areas with relatively developed economies, have unique geographical and economic advantages and can provide strong resource support for construction innovation.
The next category comprises Shanghai, Tianjin, and Liaoning. Shanghai and Tianjin, two of the four municipalities, are key cities for national development. They are supported by the government and owned the strong power of economy.
However, the total output value of the construction industry of Shanghai and Tianjin is far lower than the developed provinces that include Jiangsu, Zhejiang, and Beijing. Despite a large amount of research fund, the level of technology in Tianjin is not high, and the comprehensive power of local construction firms is weak. Meanwhile, the number of employees in the construction industry of Shanghai has grown negatively. The government pays significant attention to Liaoning, which is close to Beijing. It also provides the great convenience to develop local construction technology for Liaoning.
The fifth category is formed by eight provinces including Guangxi, Shaanxi Chongqing, Jiangxi, Shanxi, Hunan,  Anhui, and Yunnan. The level of construction innovation capability of these provinces is below the national middle level. Although a lot of resources have been invested in the construction industry of these provinces, the innovation situation is not ideal due to the bad innovation environment. Now, many construction firms are attracted to promote the development of the local construction industry. The remaining provinces are classified to the last category. This category is defined as an area characterized by worse socio-economic situations and lower level of construction innovation. Most of them are located in the remote area with poor natural conditions. For example, Jilin and Heilongjiang have the coldest weather, and Gansu and Guizhou have many mountainous areas. The local construction industry is in a period of slow development, and there is a large room for development.

Policy implication
According to the above analysis, the overall level of construction innovation in China is low, and the innovation output of a region is the most important indicator to measure the local level of construction innovation. Regional level of construction innovation is affected by regional economic level, innovation environment, and geographical location. In view of this, the following policy implications can be provided from two perspectives between overall and region: • Strengthen macro-policy regulation Local governments can optimize the national innovation environment, improve the innovation atmosphere, and stimulate the vitality of innovation via improving innovation-related encouragement policies, promoting inter-provincial technological exchanges, and strengthening scientific research cooperation among universities across the country. The government optimizes the allocation of regional innovation resources via mobilizing resource among provinces. Innovation output is the result of comprehensive action of all factors, such as resources, information, technology, capital, and talents. Different regions have different advantages in innovation resources; the national government can use macro-policy means to regulate the flow of various resources. The internal innovation vitality of the country can be mobilized with systematical resource mobilization and the improvement of the rationality of resource allocation. In order to improve the level of construction innovation in China, we should first improve the level of construction innovation in each province and improve the differences of regional elements. Local governments should explore the limitations of the development of construction innovation level in local construction industry and put forward corresponding policies to guide the direction and focus of innovation activities of construction enterprises and promote the balanced development of all elements. At the same time, construction enterprises should use regional advantages, seek regional differentiation development, and comprehensively improve the level of national construction innovation.

• Differentiated governance by different regions
The level of construction innovation in China shows obvious regional differences. According to the spatial distribution of construction innovation level, the governments should take corresponding measures in different regions to improve the overall level of construction innovation in China.
The provinces of type I and II have a high level of construction innovation. These provinces with developed economy and abundant innovation resources should focus on the introduction of global construction innovation and provide support for the development of construction innovation in other regions through technology sharing and talent exchange. The provinces of type III and IV are at the middle level of construction innovation. These regions have some advantages in innovation, such as talent, economy, and information. Local governments guide each region to give full play to its own advantages and promote differentiated development. The provinces of type V and VI have a low level of architectural innovation. These provinces are mainly distributed in the western region with the backward economy and the complex geographical environment.
According to the needs of various regions, the government can appropriately increase the investment in innovation research, actively introduce innovative talents, and enhance technical exchanges with strong provinces in construction to solve local technical problems. With the joint efforts of the government and local enterprises, the level of local construction innovation is gradually improved, and the balanced development of national construction innovation will come true.

Conclusion
A system of indicators based on innovation ecosystem theory was proposed to explore the differences in level of construction innovation in China. This research processed the indicator data from statistical yearbooks, databases, and websites using text preprocessing and search statistics and calculated the indicator weight by the cloud model. According to the comprehensive calculation results, the differential analysis was considered from three perspectives of overall, region, and cluster. The results revealed that the level of construction innovation in China is not high and its difference mainly influenced by local economic strength, innovation output, and geographical location. The overall ranking of the level of construction innovation in 30 provinces implicated that the level of construction innovation is consistent with local economic strength. A thorough comparison of innovation environment, subject, input, and output on the level of construction innovation in each province shows that the level of innovation is more sensitive to innovation output; in particular, Beijing, Liaoning, Yunnan, and Inner Mongolia were prominent. The spatial distribution of the level of construction innovation in China was showed using cluster analysis which is similar with the characteristics of the geographical location.
This study has made some contributions to the theory. The findings are conductive to designing effective method for measurement of construction innovation. The system of indicators constructed in this research and the statistical methods were applied in date process granting a holistic view on measurement of construction innovation, which fills the gap in the analysis on differences on construction innovation. The system of indicators, which was built based on the innovation ecosystem theory and systematically considered multiple dimensions, can be used to measure the level of construction innovation after simply adjusting the individual indicators. The statistical methods including text preprocessing and search statistics in this research dug out the more comprehensive and objective data compared with traditional methods, such as statistical data, questionnaires, and project cases.
In practice, the findings of this research offer valuable managerial implications for managers and policymakers in construction innovation in China. Firstly, there is a correct cognization of the differences in the level of construction innovation and the causes leading to the differences in China. Secondly, these findings not only provide the theoretical basis of efficient allocation of resources and regional scientific management for governments in each province, but also indicate the directions of implement innovation for local construction firms, which can narrow the differences in level of construction innovation in China. Thirdly, with the narrowing of the differences, the level of construction innovation and innovation efficiency can further be improved, which can promote the sustainability of the construction industry in China.
Several limitations still exist in this research. First, the exploring difference analysis in the level of construction innovation takes China as the background; in future research, more attention should be paid to the universality and difference of the application of measurement system in different countries. Second, the selection of indicators of measurement system, which is based on the theory of innovation ecosystem, focuses on the basic factors affecting innovation, such as knowledge, technology, and economy without considering the impact of the role of managers in innovation activities. Therefore, a more perfect measurement system must be proposed in future research. Finally, this paper explores the level of construction innovation in China at this time point in 2018, but did not explore and analyze the dynamic development and change of the level of construction innovation in the recent 10 years. A longitudinal study on the level of construction innovation in China should be added in the next research, so that the differences of construction innovation level in different regions of China can be explored deeply. Data availability The datasets used and analyzed during the current study are available from the corresponding author on reasonable request.

Declarations
Ethical approval Written informed consent for publication of this paper was obtained from the Liaoning Technical University of Arts and Science and all authors.

Competing interests
The authors declare no competing interests.