China’s leadership in the hydropower sector: identifying green windows of opportunity for technological catch-up

Abstract

From the sectoral systems of innovation perspective, the windows of opportunity (hereafter referred to as WoOs) for industrial latecomers to catch-up could be opened up through abrupt changes in the technological, market, and institutional dimensions. Existing literature discusses different dimensional changes in isolation. Nevertheless, for green industries, the systemic interplay of these dimensions is of key importance; yet few studies have probed into this. These limitations in the literature are largely rooted in the lack of novel methods to detect and specify these abrupt changes, especially in a quantitative way. This paper, therefore, proposes a framework combining natural language processing methods with experts’ knowledge to detect these abrupt changes—named turbulences—by using multi-source heterogeneous data, in order to better identify the co-occurrences and interactions of turbulences across the technological, market, and institutional dimensions that have a high probability to open up WoOs. We apply this framework to analyze China’s hydropower sector as a case study. The hydropower sector is considered a “green” energy sector, in which China, as this study finds, has recently gained technological leadership. By analyzing the interactions between these multiple dimensions of WoOs, we discover that institutional turbulences proactively intertwine with other turbulences, and collectively form Green WoOs for the successful catch-up of China’s hydropower sector.

1. Introduction

China’s hydropower sector may be considered a latecomer on the global stage; yet various studies propose that it has successfully caught up with leading countries in terms of technology development and gaining global market shares (Li et al., 2018). From a sectoral systems of innovation (SSI) perspective, latecomer countries could catch-up and even forge ahead in certain sectors by seizing windows of opportunity (WoOs) that are opened by abrupt changes or discontinuities in technologies (Perez and Soete, 1988; Geels, 2002 ), markets, and institutions (Lee and Malerba, 2017) however, these literature discuss different dimensions separately.

Traditional WoOs literature mainly discusses technological changes or discontinuities, such as the emergence of new technological paradigms or radical innovations, and some research has broadened the concept to market dimensions in relation to the emergence of new demand or abrupt market changes (Morrison and Rabellotti, 2017). Recent studies have extended the notion to the institutional dimension (Kim and Lee, 2008; Lee and Malerba, 2017), such as in developmental state studies (Wade, 2018) and the national systems of innovation literature (Freeman, 1988;,Lundvall, 1992;,Nelson, 1993) . However, how WoOs could be opened up through policy support and institutional changes is still under-studied (Xu et al., 2018b; Yap and Truffer, 2019). Specifically from a sectoral system’s perspective, WoOs could be associated with abrupt changes in all three dimensions and thus should be recognized as a three-dimensional concept. This needs further enquiry.

In addition, existing studies fail to discuss the co-occurrences and interactions between technological, market, and institutional changes (Yap and Truffer, 2019). As observed in catch-up sectoral cases, like steel (Lee and Ki, 2017), wine (Morrison and Rabellotti, 2017), and semiconductors (Cho and Lee, 2003) , catch-up often involves interactions of important changes across different SSI dimensions. For instance, technological WOs are often opened by the emergence of new technologies, while the new technologies might be triggered by rising demand in bourgeoning markets such as niche or bottom-of-pyramid markets—the latter could even lead to disruptive innovations (Christensen, 1993). Few studies have investigated the interactions of multi-dimensional changes that may synergically lead to WoOs for latecomers.

Lastly, WoOs are still mainly considered as exogenous and thus unpredictable for firms and countries (Yap and Truffer, 2019). However, they could be endogenized if changes across different dimensions are intertwined, especially when they involve institutional changes that are endogenous in nature. For example, governments use policy tools to intervene in technology or market development, and latecomer firms influence policymaking through lobbying (Yap and Truffer, 2019), especially in highly regulated industries such as green energy sectors in which industrial policies contain endogenous strategic planning by governments that also receive feedback from industries, focal firms, and relevant stakeholders. This is neither unpredictable nor exogenous.

These limitations in existing WoOs literature may be caused by the lack of novel methods and data to identify and analyze multi-dimensional WoOs, which involve institutional, technology, and market changes that may be intertwined. There are two main weaknesses in existing WoOs studies. First, existing WoOs literature is mainly qualitative and conceptual, and the notion of abrupt changes that may open up WoOs still remains vague and has yet to be specified in a quantitative way. Instead, the qualitative interpretations rely heavily on the individual knowledge and inconsistent intuitions of industrial experts (Li et al., 2019). Significant changes may be neglected if domain experts fail to differentiate these from gradual changes, which is something that often happens. Second, the co-occurrences and interactions between institutional, technology, and market changes are overlooked in existing WoOs studies, both qualitative and quantitative ones. The analysis of multi-dimensional interactions requires multi-source data—data which are generally heterogeneous and thus need novel methods to deal with—together with experts’ interpretations that involve cross-disciplinary knowledge (e.g. policy/technology/business).

To address these research gaps, this article will develop a framework that combines natural language processing (NLP) methods with experts’ knowledge using multi-source heterogeneous data, in order to detect those abrupt changes—called turbulences—that may lead to WoOs. Specifically, this article applies this framework to identify and analyze the co-occurrences or interactions of turbulences across the technological, market, and institutional dimensions that have a high probability to open up cross-dimensional WoOs in a sectoral system. First, the Latent Dirichlet Allocation (LDA) topic model and turbulence detection algorithm are used to detect technological, market, and institutional turbulences (IT) by analyzing multi-source data including patents, market news and reports, and policy documents, respectively, considering the criterion of the intensity and scale of topic variations. Second, by combining the data results and experts’ opinions, the co-occurrences and interactions of turbulences in three dimensions are identified and analyzed, which are recognized as cross-dimensional WoOs.

The proposed framework is applied to analyze China’s hydropower sector as a case study. For many decades, Western firms were the largest players in the global hydropower market. This changed recently when Chinese hydropower firms not only caught up but also became leaders in the global hydropower market and technology. This study aims to analyze this phenomenon both quantitatively and qualitatively. This article addresses three main research questions. (i) What is the progress of the catch-up in China’s hydropower sector? (ii) What are the institutional, market, and technological turbulences (TT) along the catch-up pathway of China’s hydropower sector? (iii) How do these turbulences co-occur and interact, thus triggering cross-dimensional WoOs for the catch-up of China’s hydropower sectors?

This article is structured as follows: Section 2 introduces the literature review and framework, Section 3 discusses the methodology, Section 4 presents the case of China’s catch-up in the hydropower sector, Section 5 shows the empirical results and analysis, and Section 6 provides a short discussion and concludes this article.

2. Literature review and conceptual framework

2.1 Woos: a three-dimensional concept

Traditionally, the concept of “windows of opportunity (WoOs)” describes the emergence of a new technological paradigm (Perez and Soete, 1988) or the abrupt changes in market demand (Mathews, 2005). Latecomer countries can catch-up using technological leapfrogging. Following this, a line of literature views WoOs as technological windows of opportunity that are triggered by the emergence of radical innovations or new generations of technologies, during which many incumbent firms are complacent with their prior success and thus fail to have effective responses to the technological discontinuities.

Adding to this, some argue that WoOs have a broader scope than mere technological concerns. WoOs can also be opened by the creation of new demand, the emergence of new markets, or abrupt changes in market demand (Mathews, 2005). For example, incumbents may be locked in their success in existing mainstream demand, while neglecting the newly emerged demand. This gives new entrants chances to dethrone incumbents with their fast and effective responses to the market changes (Christensen et al., 2015).

Some recent research has started to realize the significance of the institutional aspects of WoOs, especially when these are concerned with industrial-level and systemic factors. From a sectoral system of innovation perspective, this research (Malerba, 2002; Lee and Malerba, 2017) indicates that institutional factors such as public policies and other institutional changes are very important to the opening of WoOs. Specifically, governments may intervene in the industrial catch-up by supporting R&D programs (Markard et al., 2012), creating asymmetric environments for competitions between latecomers and incumbents (Kim and Lee, 2008), and changing regulations and legislation systems to better adapt to specific industries (Guennif and Ramani, 2012).

Based on the aforementioned literature, recent literature summarizes that WoOs have three dimensions: technological, market, and institutional aspects (Lee and Malerba, 2017). Technological change refers to the emergence of innovation, both radical and incremental, that changes the technological trajectories of specific technologies in specific sectors and that eventually changes the technological paradigms (Dosi, 1982,Christensen and Rosenbloom, 1995) (). In the case of green windows of opportunity, the technological paradigm is changing due to environmental and social pressures, such as the need to mitigate and adapt to climate change, reduce energy poverty, and increase environmental sustainability. Institutional change refers to changes in policies, regulations, legislation, and institutions, for instance by raising industrial standards, controlling harmful substances such as greenhouse gas emissions or toxic chemicals, setting up commissions to protect the environment, etc. Market change refers to shifts in supply and demand of goods and services, such as consumer demand, which opens new windows of opportunity to latecomers who are capable of meeting new demand conditions. This is, for example, the case with Chinese hydropower dams overseas, which have opened up under-explored markets in Asia and Africa. Each of those three changes provides windows of opportunities for latecomers to catc- up with leading countries in certain sectors; for example, Korea’s catch-up with Japan in the steel sector and China’s catch-up with OECD countries in the hydropower sector.

Specifically in a sectoral system, the WoOs appear when the industry experiences abrupt changes in the above three dimensions—these changes open up WoOs for the latecomer to catch-up or even forge ahead of incumbents. A limited number of studies have discussed the interactions between these three dimensions that may have a high probability to open up WoOs. As one of the few attempts, Lee et al. (2014) observe that the emergence of new technology followed by institutional and regulation changes have formed WoOs for Indian firms in the IT service sector. In addition, some argue that the opening-up of WoOs is associated with demand change intertwined with institutional support from governments (Giachetti and Marchi, 2017). However, none have extended the discussion to the interactions of the three-dimensional changes to open up WoOs in a sectoral system, especially as regards highly regulated industries such as green energy industries (Zhou et al., 2019b). We, therefore, argue that, in these sectors, WoOs may be formed through the interplays of three-dimensional changes rather than single-dimensional discontinuities. We define this as cross-dimensional WoOs in this study.

The multi-dimensional changes may co-occur and interact in three forms. First, institutional changes may co-occur with market ones. For example, the WoO that facilitated the catch-up of the Japanese camera industry after World War II was opened by the government launching industrial policies to expand the industry (Donzé, 2014), and rapidly growing demand for cameras in Japan and the United States back then (Kang and Song, 2017).

Second, institutional changes may co-occur with technological ones. For instance, WoO that built the leadership of Europe in GSM (Global System for Mobile Communications) technologies was opened by governments’ decisions on launching a unified GSM technical standard in Europe which facilitated the rapid growth of technological innovation, also resulting in strengthened university-industry linkages (Fuentelsaz et al., 2008; Giachetti and Marchi, 2017).

Third, changes at all three dimensions may co-occur within a certain period of time. For example, the WoO that facilitated Korea’s catch-up to Japan in the steel sector was triggered by changes in all three dimensions around the 1980s. The interaction regarding industrial changes was initiated by the Korean government’s large amount of financial investment, and favorable polices provided to a state-owned steel firm to develop its initial technical capability in the sector, which paved its way to explore low-end market neglected by incumbents (D'Costa, 1999). Meanwhile, the steel sector in developed countries suffers from overcapacity issues so that leading-edge steel technologies were sold to the Korean firms at low price. These advanced technologies allowed Korea to gain leadership in the steelmaking sector in the 1980s (Lee and Ki, 2017). Drawing on historical experience from the aforementioned catch-up cases, WoOs opened by multi-dimensional turbulences may provide latecomers with higher probability for technological catch-up.

In this article, the cross-dimensional WoOs are likely to open endogenously due to the proactive roles of public policy in stimulating market demand and prioritizing innovation efforts in specific technological fields (Yap and Truffer, 2019). Particularly in highly regulated sectors, WoOs may be endogenous in nature because both developing and developed countries tend to launch mission-oriented innovation policies in strategic sectors (Mazzucato, 2018) to promote certain technological developments (Weber and Rohracher, 2012; Yap and Truffer, 2019). Thus, cross-dimensional WoOs may be endogenous and predictable.

2.2 Turbulences and WoOs

As previously mentioned, WoOs can be opened up by a variety of changes in terms of technological discontinuities (Lee et al., 2005; Park and Lee, 2006; Wang et al., 2014; Wu and Zhang, 2010), market alternations (Morrison and Rabellotti, 2017), and institutional transformations (Lee and Malerba, 2017). However, most of these discussions remain at the concept level (Guo et al., 2018); while many argue that changes may open up WoOs, few have discussed what size of changes may cause the latter. The size of changes, especially quantitatively (e.g. intensity and scale), will help to distinguish critical changes that are possibly more significant to the opening-up of WoOs.

Some innovation literature uses the concept of “turbulence” to indicate abrupt technological changes or discontinuities (Christensen, 1993; Calantone et al., 2003; Lichtenthaler, 2012) although these discussions remain conceptual. Meanwhile, in the computer science domain, recent research uses machine-learning methods to specify the size of sudden and violent changes (Akcora et al., 2010; Hu et al., 2011). Studies have, for example, used NLP techniques to detect abrupt changes in technological (Chen et al., 2019) or market semantic topics (Hu et al., 2011) when analyzing text-based data such as patents and web-news. In addition, some literature examines the intensities of text topics and visualizes the changes by using text-flow diagrams (Cui et al., 2011; Li et al., 2020), and some detect those violent changes when considering the intensity and scale (Zhou et al., 2018b). Some authors also discuss sudden spikes in public controversies around a technology reflect deeper and transformational changes within the sector (Hajer, 1995; Rosenbloom et al., 2016; Rosenbloom, 2018).These studies use different terms to name these abrupt changes, such as breakpoints (Hu et al., 2011), topic variations (Miao et al., 2020) and sudden spikes.

Combining expert insight with these NLP techniques, we have been able to extend the notion of turbulence to quantitatively specify the size of abrupt changes by observing variations in topic intensities and scales; in addition, these changes can be detected in different dimensions of a sectoral system. Specifically, those changes with a high-degree variation in topic intensities and scales can be called turbulences (see Section 3.3.2), which can be distinguished from gradual changes and thus help to identify WoOs better—the former would have much higher probabilities to result in WoOs than the latter.

As discussed in Section 2.1, WoOs could open up when multi-dimensional abrupt changes (called turbulences in this study) occur simultaneously or even interact. These co-occurrences or interactions generate a variety of opportunities for latecomers in a sectoral system that evolves over time (Lee and Malerba, 2017). Research on the co-occurrence and interactions of turbulences remains sparse, which is largely due to the lack of novel techniques to detect these turbulences in technologies, markets, and institutions. In green sectors such as hydropower, the co-occurrences and interactions of these turbulences may result in endogenous WoOs, while the government may be more proactive in these interactions of turbulences. This needs further exploration.

2.3 Conceptual framework

Based on the above literature, this article proposes a multi-layered conceptual framework to identify WoOs that are opened by the co-occurrences or interactions of turbulences (Figure 1). In the proposed framework, turbulences in all three SSI dimensions (technology, market, and institution) can be detected using patents, business news, and reports, as well as policy documents (see Section 3.2.2). On top of this, analyzing the co-occurrences or interactions of turbulences will help to identify the multi-dimensional WoOs, during which Chinese latecomers may catch-up with incumbents.

3. Methodology and data

3.1 Research design and case selection

This article uses a single longitudinal case study (Yin, 2017) of China’s large hydropower sector to explain how China successfully caught up in the hydropower sector by using green windows of opportunities. The hydropower sector is a special case as it was China’s first green energy sector. This is also the case for many other countries such as those in the Nordic region and North America, as well as many developing nations across Asia, Africa, and Latin America, where the hydropower sector was the first green energy sector that managed to operate commercially and at large scales (see Xin et al., 2019). The origins of the large hydropower sector are rooted in energy security concerns and in being able to provide large-scale electricity generation at low cost, over long periods of time, and from indigenous natural resources. The relevance of hydropower for climate change mitigation only became important much later, following the Kyoto Protocol and particularly the Paris Agreement. In China, innovation within the hydropower sector was initially meant for the domestic market and it was not developed for export, unlike solar photovoltaic (PV). It is only since the government’s 2000 “Going Out” strategy that hydropower technology has been exported overseas on a large scale (Urban et al., 2013).

The large hydropower sector is characterized by very sophisticated, large-scale operations, a heavy industry that involves the creation of a reservoir (which may involve large-scale logging of forest areas), the construction of a dam wall (often made of large amounts of concrete) and the equipment, most importantly the turbines and other heavy machinery. It is also a service-intensive sector that involves a large number of contractors, suppliers, and service providers along the value chain to, for example, conduct feasibility studies, Environmental and Social Impact Assessments (ESIA) and other consultancy services, to supply equipment (e.g. for the turbines, blades, generator, transformer), and to provide engineering services for building the dam wall, water diversion tunnels, power lines, etc.

For many decades, there has been a discussion about how “green” the hydropower sector really is. The proponents of the hydropower sector argue that it is a renewable, low carbon energy technology that is crucial for mitigating climate change. This is also the official stance of important international organizations such as the Intergovernmental Panel on Climate Change (IPCC) and the International Energy Agency (IEA). Hydropower opponents, however, argue that large hydropower has large-scale and irreversible environmental impacts including ecosystem destruction, geomorphological changes, hydrological changes, impacts on aquatic species, habitat and biodiversity loss, etc. It has also been analyzed how reservoirs contribute to the release in greenhouse gases, most importantly methane and nitrous oxide (Fearnside, 2015). Many analysts come to the conclusion that large hydropower can have more devastating environmental impacts than small hydropower, although a cascade of small-hydro plants can also be damaging for the riverine environment. Moran et al., (2018)make a series of suggestions on how hydropower can become more sustainable.

Since about 1999/2000, China has started to catch-up with leading countries in the hydropower sector. According to Bosshard (2009) and Jia (2016), about half of the world’s large dams are based in China. China clearly leads the field as the country with the largest installed capacity and the largest generating capacity. The International Hydropower Association (IHA) estimates the current installed hydropower capacity in China to be over 340,000 MW, providing a generation capacity of about 1,195,000 GWh (IHA 2019). Much research has been conducted on large hydropower dams in China (e.g. Magee, 2006; McNally et al., 2009; Chang et al. 2010).

In the past two decades, a large number of large hydropower dams have been built by Chinese firms internationally, including countries in Southeast Asia and Africa. Research on Chinese-linked overseas hydropower projects has therefore increased (e.g. Yu, 2003; McDonald et al., 2009; Urban et al., 2013, 2016, 2018; Tan-Mullins et al., 2017; Brautigam and Hwang, 2019) examined the role of technology and knowledge transfer from China to other countries in the hydropower sector.

With the increasing attention placed on environmental protection and social welfare of affected communities (Tilt et al., 2009), the focus of the hydropower sector has shifted from dam design and project quality control to ecology and migration management (Jiang et al., 2016a). This requires more accountability and governance of the dams’ impacts by local governments. Due to the high involvement of government agencies in dealing with some aspect(s) of hydropower developments, the window of opportunity opened by institutional turbulence will be very important for latecomer countries.

The hydropower industry case is being investigated for the period 1999–2017. The year 1999 was selected as the starting year because in that year China’s hydropower sector started to grow rapidly due to the construction of the Three Gorges Dam; however, China was far from being a global leader either in market or in technology at that time. The turning points occur afterwards: China’s installation capacity (market) became the world’s No.1 in 2004, and its technology (patent counts) also overrode other countries in 2008 (Section 4). Our case study ends in 2017 due to data constraints. This study uses basic patent applications in the priority year for the analysis, but as it generally takes 18 months for the patent applications to be collected comprehensively into the patent database, the patent application data for 2018–2019 is non-exhaustive. In order to avoid data bias, we, therefore, set the end date at December 2017.

3.2 Methods

3.2.1 Methods for analyzing the catch-up and growth

This article, in Section 4, uses market statistics and patent counts/citation networks to analyze the catch-up and growth of China's hydropower sector. First, this article investigates the market share of the top 10 countries (including China) in the world by using the statistical reports. Second, it examines the technology growth of the top 10 countries by using patent counts. Additionally, this article also uses patent citation and network analysis to study the positions of Chinese firms in the global knowledge network. Following the methodological success in prior studies (Wang et al., 2018; Xu et al., 2018; Zhou et al., 2016, 2018a), this article constructs the patent citation network of the global top 20 hydropower firms (see details in Appendix C). It uses degree centrality, which considers the number of links a node has in a network (Zhou et al., 2016), to measure the positions of firms in the network. In addition, expert interviews helped to triangulate the outcomes using a mixed-methods approach, relying both on quantitative and qualitative methods and sources of information (also see Section 3.2.3).

3.2.2 Methods for detecting multi-dimensional turbulences

This article, in Section 5, will use an NLP-based turbulence detection model to detect the institutional, technological, and market turbulences (MT) (Step I in Figure 1), which can be differentiated from mild changes considering the size (i.e. intensity and scale) of changes. Following existing NLP literature (Hu et al., 2011; Wei and Croft, 2006; Zhou et al., 2018b), this turbulence detection model consists of two procedures: (i) LDA topic model to extract the semantic topics from datasets and (ii) turbulence detection algorithm to measure the intensity and scale of topic changes.

Procedure I: LDA for topic extraction. LDA is used to extract semantic topics from text-based datasets, which is an unsupervised machine-learning technique for NLP processing that can be used to identify latent topic information in large-scale document collections (Blei et al., 2003, 2012; Dong et al., 2017; Jiang et al., 2016a,b).¹ Following prior literature (Wang and Xu, 2018; Zhou et al., 2019), this study uses LDA analysis in two steps: First, we determine the number of semantic topics (denoted as K) from a text-document dataset (also called a corpus) by using a perplexity test (K is set 20 in this study).² Second, using the optimal K, we can extract K semantic topics at different time-phases in a specific period (Wang and Xu, 2018; Zhou et al., 2019) (see technical details in Appendix A). In turn, we can observe or even measure the topic changes (the weights of K topics) over time—this will be conducted in the next procedure.

Procedure II: turbulence detection algorithm to detect abrupt topic changes. The turbulence detection algorithm is used to quantitatively specify the size of topic changes over time, while there are K topics at n time-phases obtained from previous LDA procedures. Following existing literature (Akcora et al., 2010; Hu et al., 2011), this study uses the turbulence detection algorithm in three steps. First, we calculate the topic variations by differencing the weights of K topics (or called topic strength)³ in neighboring time-phases such as t_i and t_{i+ 1}, and we define this (t_{i+ 1} − t_i) as the difference in topic strengths. Second, we measure the intensity of topic changes by comparing these differences in topic strengths over time. Following existing literature (Miao et al., 2020), we use a parameter C to indicate the threshold of the differences in topics strengths—the higher the parameter C, the more abrupt the topic changes—and we recognize the above-threshold changes as turbulences. Third, we measure the scale of these topic changes by counting the number of critical topics that contribute to at least 80% of the total topic variations. Combining the intensity (step 2) and scale (step 3), we can detect the most abruptly varied topics as turbulences.

3.2.3 Methods for identifying cross-dimensional WoOs from turbulences

Based on turbulence detection results, we will identify the WoOs (Step II in Figure 1) that may be opened up by the co-occurrences and interactions of these turbulences, in two procedures. First, we map these turbulences in China's hydropower along with growth in SSI dimensions (see Figure 8), based on which we recognize the co-occurred turbulences and suggest possible cross-dimensional WoOs. Second, through interviews and discussions with experts (see Appendix B), we investigate the interactions between these co-occurred turbulences, based on which we identify those highly interacted turbulences as cross-dimensional WoOs and we maintain that these WoOs opened up for China’s hydropower sector to catch-up and ultimately led to the global leadership change in this industry.

The analytical framework proposed in this article has three advantages. First, WoOs identified using this framework would be more accurate comparing to existing qualitative studies. This framework is primarily data-driven thus less influenced by subjectivity of interviewees. Second, the interactions among turbulences could be identified. This framework combines NLP method with interviews of industry experts to identify multi-dimensional turbulences and analyze their interactions. This is absent in the conventional qualitative studies of WoOs. Third, this framework allows researchers to analyze large-scale multi-source heterogeneous data efficiently when studying WoOs, which would be challenging to do manually.

This framework could be employed in cross-dimensional WoOs studies of other highly regulated sectors that are likely to experience multi-dimensional turbulences. However, it is worth to note that this framework may not be an effective alternative to qualitative analytical framework focusing on single-dimensional WoOs. This is due to the heavy workload resulted from the complexity of turbulence detection model and large amount of data required beforehand.

3.3 Data

This study develops three datasets: technology, market, and institution (Figure 1).⁴

3.3.1 Technological dataset

This article uses the patent dataset (1999–2017) retrieved from the Derwent World Patents Index (DWPI) and Derwent Patents Citation Index (DPCI) databases.⁵ These two databases contain patent information and citation data from 61 patent-issuing authorities worldwide.⁶ Using the list of keywords derived from existing literature (Jiang et al., 2016b) and suggestions from experts, we retrieved 14,435 worldwide patents related to the hydropower industry. Patent information includes title, abstract, claims, and priority dates. For patent citation analysis, we selected the patents affiliated with key firms in patent citation networks (see Appendix C). For the identification of turbulences in technology, the abstracts and claims of all patents in this dataset were used as the corpus (i.e. text documents) for LDA analysis.

3.3.2 Market dataset

Following prior literature (Sternitzke, 2010; George et al., 2016; Boudoukh et al., 2019), this article develops the market dataset (1999–2017) that incorporates market news released by the IHA, the China Society for Hydropower Engineering (CSHE), as well as all company news and reports released on the website of key hydropower firms. For the analysis of China’s market, a total of 8109 text documents were collected from websites of Chinese firms and CSHE using a web-crawler service.⁷ For the global market, 7578 text documents were collected from websites of key firms overseas and the IHA website.

3.3.3 Institutional dataset

This article compiles a set of policy documents for use as an institutional dataset. In this article, the word “policy” is defined broadly to include policy tools, regulations, and laws. Changes in these policies play crucial roles in driving institutional changes since policy could influence other types of institutional aspects like values, norms, and industrial standards. Therefore, this article chose to capture institutional dynamics by studying the topic changes in policy documents. Policy documents from China were retrieved from the PKULAW database.⁸ Using “hydropower” (in Chinese) as the keyword, a total of 2435 text documents were retrieved from this database. Policy documents from other countries were retrieved from the WESTLAW database.⁹ A total of 1797 policy documents from countries worldwide (excluding China) have been retrieved; the collected policies come from conventional leading countries in the hydropower sector.

3.3.4 Expert interviews

The expert interviews were based on semi-structured, open-ended key informant interviews with representatives from firms and business associations, policymaking authorities, non-governmental organizations, and experts from academia and think tanks working in the hydropower sector. These qualitative interviews were critical to verify the outcomes that resulted from the quantitative data analysis.

4. Catch-up and growth of China's hydropower sector

Following the methods described in Section 3.2.1, this section studies the catch-up and growth of China’s hydropower sector in both market and technology aspects.

4.1 Market catch-up

As a latecomer in the global hydropower market, China has developed rapidly in the past two decades. From Figure 2, China clearly leads the field as the country with the largest generating capacity since 2004. The IHA estimates the current installed hydropower capacity in China to be over 340,000 MW, providing a generation capacity of about 1,195,000 GWh (IHA, 2019). Figure 2 also shows that the years 2001, 2004, 2008, and 2012 stand out for having significantly higher growth rates in terms of hydropower-based electricity generation. We argue that these may indicate critical tipping-points in China’s catch-up, but some may also be explained by climatic variations, e.g. years with higher precipitation than usual that result in higher water flow rates and therefore higher electricity generation capacity. This needs further investigation.

4.2 Technology catch-up

We use the number of patent applications in the hydropower sector to examine the technology development of global top-10 countries. Figure 3 illustrates that China became No.1 in the number of hydropower patent applications in 2008, and China accounted for more than 50% of global patent counts in 2008 and 94% in 2017. We argue, at least from a patent count perspective, that China has caught up with leading countries in the hydropower sector in terms of technology scale (though not quality) since 2008.

We also examine the technology quality of China’s hydropower sector by using patent citation and network analysis (depicted in Section 3.2.1). Leading firms are generating high-quality patents that are highly cited by followers, and this makes them move to central positions of the patent citation network. Following existing methods (Section 3.2.1), this study builds two patent citation networks that involve global top-20 firms (considering patent citations) in 1999–2008 (Figure 4a) and 2009–2017 (Figure 4b), respectively. 2008 is selected as a cutoff year because in this year China’s hydropower patent application reached over 50% in this year, and this denotes a tipping point (Figure 3).

• Firm’s citation network, 1999–2008.

• Firm’s citation network, 2009–2017.

• Number of citations of top-20 firms–

Figure 4a shows that during 1999–2008, none of the Chinese firms (red nodes) have high patent citations or are located in the central position, but instead are found at the periphery of the network.¹⁰ By contrast, during 2009–2017 (Figure 4b), Chinese firms move to the central (core) positions in the patent citation network. This indicates that Chinese firms have started to catch-up with incumbents in the global knowledge network. Figure 4c shows the comparison of patent citations of top-20 firms as a proxy for the quality of technology, and illustrates that by 2017, some Chinese hydropower companies had more patent citations than their international competitors, the former including State Grid, Southern Power Grid, PowerChina (parent company of Sinohydro), Three Gorges Corporation, and Huadian.

Based on the above analysis, we argue that China has successfully caught up in the hydropower sector in both market and technological aspects.

5. Empirical results: identifying WoOs in China’s hydropower sector

This section identifies and analyzes the WoOs for China’s hydropower sector in two steps. In Section 5.1, by using NLP methods (Section 3.2), we detect technological, market, and IT based on three datasets. In Section 5.2, we identify the WoOs by mapping the co-occurrences of these turbulences (Figure 8) and analyzing the interactions between them. Following this, we discuss the responses of China’s hydro-sector to these WoOs.

5.1 Detecting multi-dimensional turbulences

5.1.1 Detecting technological turbulences

Following the methods in Section 3.2, we use the patent database (3.3) and successfully detect four technological turbulences (TT) during 1999–2017 (see Figure 5). These TTs are coded as TT1 to TT4, as described below. Figure 5 also illustrates the number of patents related to specific topics that account for the most abrupt changes in each technological turbulence (see details in Appendix D). This helps us to better interpret which topic variations constitute these turbulences.

TT1 (2001)—new technologies in water protection and treatment. TT1 comprises five major topics that change abruptly in the year 2001, including “water protection,” “generator,” “water treatment,” “waste control,” and “cooling system.” These changes mainly happen in China. Out of these five topics, three topics are highly related to water resources/protections (Figure 5). We argue that this may be triggered by the Three Gorges Project (second phase) in China during 1997–2002, which induced new technologies to cope with water-resource protection issues that significantly threatened the Three Gorges and Yangtze River at that time.

TT2 (2004)—emerging domestic technologies of hydro-turbines and dam construction. TT2 has four additional topics in comparison with TT1, including “water pump,” “sand removal,” “fish passage,” and “dam construction,” which are related to hydro-turbines and ecological concerns (Figure 5). This turbulence may be associated with the first-ever domestic manufacturing of large hydro-turbines (700 MW) in China as import substitutions, and these turbines begin operation in the third phase of the Three Gorges project, which started in 2003.

TT3 (2008)—the emergence of distributed hydropower and smart grid. TT3 comprises three additional topics, including “pumped storage,” “gate panel,” and “electricity converter,” which are highly related to the emergence of distributed hydropower and the smart-grid technological paradigm in the world (see Figure 5). According to experts, smart-grid technologies are revolutionary, enabling distributed power generation and upgraded power-grid management, both in China and the rest of the world.

TT4 (2013)—drastic technological growth in dam construction. TT4 has one additional topic, “monitoring system.” Other relevant topics are highly related to the construction in hydropower projects (see Figure 5). We argue that this may be associated with the booming of hydropower constructions, including China’s second-largest hydropower project, Baihetan station. According to the experts we interviewed, this project faces many engineering challenges when dealing with tough geographical conditions like loosely structured rock and columnar jointed basalt. In response to these challenges, many leading-edge technologies have been developed by Chinese firms, such as real-time monitoring systems, new water pump technologies, tunnel-digging technologies for dam construction, etc.

5.1.2 Detecting market turbulences

Following Sections 3.2 and 3.3, three market turbulences (MT) have been detected during 1999–2017 (see details in AppendixD). These MTs are coded as MT1 to MT3 (see Figure 6), as described below. Additionally, Figure 6 illustrates the specific topics that account for the most abrupt changes in each market turbulence.

MT1 (2003)—growing demand induced by large dams including the Three Gorges. MT1 comprises six major topics that changed abruptly in 2003, including “disaster management,” “transportation,” “resettlement,” “natural resource protection,” “quality evaluation,” and “electricity generation” (Figure 6). We argue that these topics may be associated with the Three Gorges project (Phase 3, started in 2003) in China, which was concerned with new demands including social issues such as resettlement and disaster management, economic issues like transportation on Yangtze River, and protecting the environment during construction.

MT2 (2009)—new demand for distributed power generation. MT2 contains four new topics in comparison with MT1, including “research & development,” “emission reduction,” “international collaboration,” and “investment,” and changes occurred both in China and worldwide (Figure 6). We argue that this may be related to the rising demand for small hydropower and pumped-storage hydropower in the world. According to the expert opinions of our interviewees, these new hydropower stations are distributed-based and more ecological.

MT3 (2011)—rising demand for hydropower construction. MT3 encompasses two new topics, including “electricity transmission” and “workplace safety,” while changes primarily occurred in China. This turbulence may be triggered by the fast-growing hydro-constructions in China, including both large and small dams such as Baihetan station (started in 2011). Additionally, the demand for small hydropower projects also grows substantially to form a more decentralized national power grid. The diversity among these hydropower projects requires tremendous efforts in terms of electricity transmission and project management on safety.

5.1.3 Detecting institutional turbulences

Following Sections 3.2 and 3.3, we successfully detect three institutional turbulences (IT) during 1999–2017 (see Figure 7). These ITs are coded as IT1 to IT3. Additionally, major topics that account for the most rapid changes in each institutional turbulence are illustrated in Figure 7 (see details in Appendix D).

IT1 (2002)—new social and environmental policies regulating large dams like the Three Gorges. IT1 comprises eight major topics in the year 2002. Three of the eight topics are highly related to the social policies that involve “land expropriation,” “resettlement,” and “health and safety of migrants”; the other three are related to the environmental policies including “natural resource management,” “water standards,” and “water protection.” According to our expert interviewees, a bundle of policies concerning the large dams, including the Three Gorges project (third phase), constitutes this institutional turbulence, such as the “Technical code of safety & protection facility for hydropower & water conservancy construction engineering (2002)” and “Regulations on the construction and resettlement of the Three Gorges Project of the Yangtze River (2001)”.

IT2 (2009)—new regulations on distributed power generation and smart grid. IT2 has two new topics in comparison with IT1, including “water conservancy” and “training & education.” These mainly occurred in China, and are probably linked to a series of new policies that are related to distributed hydropower stations, such as small hydro-stations that are dispersed in a rural area in China, e.g. “Notice on strengthening the construction and management of small hydropower replacing fuel and rural electrification (2009)” and “Notice on strengthening safety supervision of small hydropower stations (2009).”

IT3 (2013)—new policies regulating and promoting hydropower constructions. IT3 contains three new topics, including “construction supervision,” “engineer qualification,” and “renewable energy.” This happened primarily in China. We argue that these may be triggered by the new regulation policies related to the rapid expansion of both large and small hydropower constructions in China. Since 2011, the government has started to increase the level of supervision on hydropower projects by raising environmental and technological standards when evaluating hydropower projects.

5.2 Identifying and analyzing WoOs: co-occurrence and interactions of multi-dimensional turbulences

Following Section 3.2.3, we map the above turbulences along with the growth of China’s hydropower in three SSI dimensions (see Figure 8). From this, we can discover the co-occurrences of these multi-dimensional turbulences in 2002–2004, 2008–2009, and 2011–2013, respectively. We bundle these three sets of co-occurred turbulences and recognize them as possible WoOs—designated WoO1 to WoO3. According to our discussions with experts, we further analyze the interactions of these turbulences in each WoO as follows.

5.2.1 Woo1: market catch-up

WoO1 (opened around 2002)—domestic level and international level.Figure 8 shows that there are three turbulences (IT1, MT1, and TT2) that co-occurred in 2002–2004. Based on this observation and our discussions with experts, we argue that institutional turbulence IT1 inside China (i.e. policies on large hydro-dams such as the Three Gorges) firstly initiated this WoO1 and then interacted with the market (MT1) and technological turbulence (TT2) that opened up this WoO1 around 2002–2004. Specifically, in 2002, the construction of the Three Gorges Dam was transitioning from its second phase (construction) to the third phase (operation), a time during which China’s government had increased attention on environment protection and social welfare of affected communities (Oud, 2002; Yuksek et al., 2006; Tilt et al., 2009), and IT1 thereby includes a series of social and environmental policies. These policies propel the emergence of new demands (MT1) on transportation, resettlement, and natural resource protection, and, combined with the emergence of new relevant technologies (TT2), these aspects may also reinforce each other.

Additionally, at the international level, experts we interviewed also argue that WoO1 may be co-opened by the global vacuum that was created by the downsizing of OECD large dams’ activity around 2000 following the World Commission on Dams (WCD) recommendations (Gagnon et al., 2002). This was a time when the World Bank and other major OECD dam-builders reduced their activities due to the large-scale irreversible ecological and social impacts of large dams (Kucukali and Baris, 2009). This market void was filled by Chinese companies that started operating in emerging markets such as in Africa and Asia. From a policy perspective, this was also promoted by the Chinese “Going Out” policy in 2000, which encouraged Chinese firms to invest and operate in new markets overseas (Urban et al., 2013).

At the domestic level, Chinese firms strived to respond to WoO1. Specifically, in 2003, a Chinese-led firm (Three Gorges Corporation) activated a batch of six 700 MW-based generators and connected to power grids in China’s Three Gorges project. This effort partially helped China become the global leader in hydropower capacity in 2004 (see Section 4). Technology-wise, Chinese leading hydropower firms like Three Gorges Corporation and PowerChina worked together with other firms through international technology cooperation to build up core technology capacities (Urban et al., 2016) and domestic R&D efforts. In addition, these firms accumulated immense technical and engineering skills that were required for building large hydro-dams—this echoes the TT2 we find in 5.1.1.

5.2.2 Woo2: the technological catch-up

WoO2 (opened around 2008)—the emergence of distributed hydropower paradigm globally. Three turbulences (TT3, MT2, and IT2) co-occurred between 2008 and 2009. From Figure 8 and our discussions with experts, we argue that WoO2 was firstly triggered by TT3. The experts maintain that around 2008, the technological paradigm of distributed hydropower (i.e. integrations of smart-grid and small/pumped-storage hydropower) emerged and diffused in the world, and this new trend also influenced China. In 2006, IBM developed solutions to build smart grid collectively with other firms and research institutes. This technological turbulence stimulates the rising market demand (MT2) for the diffusion of small hydropower and pumped-storage hydropower in China, and this requires more accountability and governance of the dams’ impacts by local governments—governments follow-up to regulate the risks by launching new policies (IT2) on small hydropower regulations in terms of water conservancy and education. These three turbulences synergically opened up WoO2 in China.

Responding to this WoO2, China has developed distributed hydro-stations in dispersed rural areas in China; the installed capacity of small hydropower in China increased 4 GW in 2008 (MWR 2009). Specifically, Chinese firms, both state-owned and private firms such as Chongqing Sanxia Hydropower Limited and Yunnan Wenshan Electric Limited, have invested in building small-hydro such as run-of-river hydropower stations in China’s villages (QIRI 2014). Adding to this, the State Grid of China has started to lead the establishment of smart grid in China in order to connect distributed hydropower like small hydro-stations to the national grid. These efforts collectively helped China to accumulate capabilities to develop cutting-edge technologies and keep abreast of global frontiers in the hydropower sector.

5.2.3 Woo3: toward technological-market leadership

WoO3 (opened around 2011)—booming of China’s hydro-projects in both large and small hydropower projects. Three turbulences (MT3, TT4, and IT3) co-occurred in 2011–2013. According to our discussions with experts, we argue that WoO3 is initiated by the MT3 and then opened up interactively with TT4 and IT3. Since 2011, the market demands (Ansar et al., 2014) in both small and large hydropower have proactively risen in China (MT3). This hike in hydro construction in China may be co-driven by the stagnant growth in fossil-fuel-based electricity supply and China’s arising engineering competence in hydropower dams that have accumulated through large-dam constructions such as the Three Gorges project. Specifically, construction began in 2011 on 16 hydro-dams in China, including China's second-largest hydropower project Baihetan; in addition, the investment in small hydropower in rural areas in China also increased about 89.1% from 2010 to 2011 from 2 to 3.9 billion RMB (MWR 2011, 2012). Market booms pulled technological innovation. China’s hydropower technologies achieved a significant growth (TT4), especially as most of these constructions happened in rough geological conditions and thus needed more advanced and environmentally friendly construction technologies (see Section 5.1.1). As evidence of this, the number of Chinese firms’ patent applications rose to the record-high of 973 in 2013 (see Section 4). Market booms also shift the focus of China’s hydropower sector from dam design to project quality control and ecological management (Jiang et al., 2016b), and this creates urgency for China’s governments to launch regulatory policies by increasing the level of supervision on hydropower projects and raising environmental and technological standards (IT3).

According to experts, China’s “Going Out” strategy also helped to boost the market demands. In the last decade, a large number of Chinese-linked overseas hydropower dams have been built by Chinese firms internationally, including in countries in Southeast Asia and Africa (Urban et al., 2013, 2018; Tan-Mullins et al., 2017; Brautigam and Hwang, 2019), echoed by China’s Belt and Road Initiative¹¹ in 2013. Apart from the success in building large hydropower stations, China’s small hydropower also ranked number one globally by 2013, as measured by the total installed capacity (Liu et al., 2013). In response to these fast-growing demands, Chinese-led firms have developed leading-edge innovations. For example, Harbin Electric and Dongfang Electric developed record-breaking hydro-turbines for the Biahetan project (with a capacity of 1000 MW per unit), which are viewed as global-leading hydro-technologies. This echoes China’s leadership position in global networks, as discussed in Section 4. Based on our observations, we argue that China’s hydropower sector successfully seized WoO3 in attaining both market and technological leadership in the global hydropower sector.

6. Discussion and conclusions

This paper analyzed the development of China’s hydropower sector as a case study in order to identify the windows of opportunities that have opened up for China as the latecomer to catch-up in this sector internationally. Specifically, this article developed an integrated framework to detect the market, technological, and institutional turbulences by using NLP methods (Section 5.1), based on which we identified, through our discussions with domain experts, the co-occurrences and interactions of these turbulences that probably lead to cross-dimensional WoOs for China’s catch-up (Section 5.2). The key findings are described below.

First, our research finds that Chinese hydropower firms are not limited to technological catch-up any longer, but instead they are already leading the global hydropower sector in terms of both market share and technological development. The catch-up progress is summarized as follows: (i) In 2004, China became No. 1 in global market share in terms of total installed capacity. (ii) In 2008, China’s hydropower firms became No. 1 in the world in the number of patent applications. In addition, China’s share of global patents in the hydropower sector reached nearly 95% in 2017, as this study shows. This is strong evidence of technological catch-up. However, it also needs to be considered that the traditional hydropower industry based on OECD firms (most importantly in the Northern European countries and North America) developed many of their main technologies and filed core patents a long time ago as some of the firms have been in the sector for many decades, or even a century (e.g. Vattenfall). (iii) Since 2008, China has become technology leaders as they have moved toward the central position of the global knowledge network. We argue that Chinese firms developed an internationally standardized system that acknowledges the role of formal commercialization through patenting. This finding echoes and quantitatively confirms the prior studies on the hydropower sector and China’s domestic and global involvement that were mainly analyzed from a qualitative perspective (e.g. McDonald et al., 2009; Urban, 2018), such as by using interview data. This is where our article adds value.

Second, by using machine-learning (i.e. NLP) methods, we have detected technological, market and institutional turbulences, respectively, along with the catch-up growth of China’s hydropower sector. Technological turbulences (TT) reveal that China’s focus of hydropower technology has evolved from dam design and construction equipment to ecological protection (e.g. fish passages, run-of-river projects, etc.). Adding to existing studies (Jiang et al., 2016b), this article finds that China is also excelling in the development of distributed hydropower (e.g. small-hydro and smart grid) and leading-edge technologies such as real-time monitoring systems. Market turbulences (MT) show that market direction has evolved from large-dam construction to ecological protection, and to export-oriented markets that require higher quality and safety standards in the overseas market, such as in Southeast Asia and Africa. Insititutional turbulences (IT) reveal that China’s government has also changed focus from mere construction of large hydropower projects like the Three Gorges projects to environmental and social policies to distributed and small hydropower to hydro-managerial concerns such as qualification and project supervisions.

Third, we mapped the above turbulences in Figure 8, based on which we bundled these three sets of co-occurred turbulences and recognized them as WoO1 to WoO3. We find that these cross-dimensional WoOs have different interaction patterns, based on which we argue that WoO1 and WoO3 are endogenous while WoO2 is exogenous to China’s hydropower sector. Specifically, WoO1 around 2002 is endogenous and was triggered by a proactive institutional turbulence (IT1) that put forward the large-dam constructions in China (e.g. Three Gorges) and internationally (e.g. China’s “Going Out” strategy particularly in Southeast Asia and Africa), reinforced by its interactions to other co-occurring turbulences in the market (MT1) and technological (TT2) dimensions. WoO2 is rather exogenous, as the new distributed hydropower paradigm (TT3) firstly emerged globally, and influenced China’s growth in small-hydro and smart grid. TT3 is then strengthened by the market creations (MT2) for the diffusion and adoptions of new technologies, followed by the regulations over risks (IT2), which synergically opened up WoO2 in China. WoO3 around 2011 is endogenous, as a proactive market expansion domestically (MT3) and internationally firstly triggered this WoO, and then is collectively opened up by the interplays with indigenous technologies (TT4) and policymaking (IT3). Based on these observations, we argue that these WoOs are not merely policy-driven or market-pulled, or technology-pushed, but are instead initiated and propelled by the highly interacting turbulences of different SSI dimensions. Specifically, government agencies have a high level of involvement in all these WoOs. We argue that institutional turbulences are particularly crucial for latecomer countries in hydropower sectors.

This article contributes to the existing literature in three ways. First, this article extended the notion of the windows of opportunity from a single-dimensional concept to a cross-dimensional one, by defining the multi-dimensional changes as interacting turbulences, which synergically open up the WoOs for China’s hydropower to catch-up . These insights contribute to SSI theory by revealing the interactions between multiple dimensions. On top of this, it integrates the multiple SSI dimensions with the WoOs concept, which helps us to better understand how WoOs could be opened up through policy support and institutional changes (Xu et al., 2018b; Yap and Truffer, 2019). This is where our article contributes insights .

Second, this article proposes a new framework that integrates NLP methods and multiple-source heterogeneous data (i.e. patents, company news and reports, and policy documents) to detect turbulences. When the size of technological, market, and institutional changes are specified, abrupt changes are identified as turbulences. Mapping these detected turbulences along with the growth of China’s hydropower, together with discussions with domain experts, we can identify the co-occurrences of the turbulences and analyze the interactions among them that would lead to cross-dimensional WoOs.

Finally, based on the analysis of these interactions, we argue that sometimes WoOs may be endogenous, and thus do not always open up unexpectedly. This creates contrasts with existing literature and thus brings novelty into WoO theory. Specifically, this article finds that governments could deliberately shape WoOs at the national level and the sectoral level by initiating institutional turbulences through drastic policy changes in green sectors like hydropower. These institutional turbulences may proactively pull other turbulences such as in WoO1, or reinforce endogenously market (expansion) turbulences such as in WoO3. We conclude that these endogenous and intertwined WoOs as green windows of opportunity for sectoral catch-up are rather unique compared to other WoOs.

Further research could go beyond the scope and the limitations of this article. First, the distinction between large and small hydropower has not been made clearly in this research design. Such distinction should be made in future studies in consideration of the heterogeneity in the knowledge based between large and small hydropower technologies. Second, the analytical framework has been validated with only one empirical case. It should be employed in WoOs studies of other green sectors, especially highly regulated sectors, to improve its validity. Third, this framework is complex and primarily data-driven, thus should be complemented by further integration of expert opinions in order to improve the reliability and overall comprehensiveness of the research findings.

To be submitted to a special issue on “Green Windows of Opportunity in Emerging Economies: Towards New Leadership in the Green Transformation?” for Industrial and Corporate Change (ICC).

Footnotes

Acknowledgments

We would like to thank Dr. Xian Ming Qu, national strategic advisor to the “Made in China 2025” strategy; Dr. Xiaoying Yang, acting director of the Engineering Department at the Chinese Academy of Engineering (CAE), and Dr. Jiyuan Zang, researcher at the Engineering Department in Strategic Consultancy Centre at CAE, for their expertise and support for this article.

Funding

This research is supported by the National Natural Science Foundation of China (71974107, 91646102, L1924058, L1824039, L1724034, L1624045, L1524015), the MOE (Ministry of Education in China) Project of Humanities and Social Sciences (Engineering and Technology Talent Cultivation) (16JDGC011), the National Science and Technology Major Project “High-end Numerical Control and Fundamental Manufacturing Equipment” (2016ZX04005002), Beijing Natural Science Foundation Project (9182013), the Chinese Academy of Engineering’s China Knowledge Centre for Engineering Sciences an Technology Project (CKCEST-2020-2-5, CKCEST-2019-2-13, CKCEST-2018-1-13, CKCEST-2017-1-10, CKCEST-2015-4-2), the UK-China Industry Academia Partnership Programme (UK-CIAPP\260), as well as the Volvo-supported Green Economy and Sustainable Development Tsinghua University (20183910020) and the UK Economics and Social Research Council ESRC (ES/J01320X/1).

References

Appendices

A. LDA model perplexity detection

The optimal K values (number of topics) for LDA modeling of the technological, market, and institutional datasets are determined by perplexity test (see Figure A1). K = 20 is selected for LDA modeling of the three datasets since the reduction of perplexity index diminished very slowly after K = 20. Thus, 20 topics will be extracted from each dataset in this article.

B. Participating experts

Expert interviews were held between 2013 and 2019 with key stakeholders in the hydropower sector, including experts from firms and business associations, government agencies, NGOs, and academia.

In 2018–2019, experts made special contributions to this article, especially for the discussion on turbulence detections as well as the identifications of WoOs in Section 5 and insights on the catch-up and growth of China’s hydropower sectors in Section 4.

C. Patent citation analysis

The first step is data acquisition. We download patents through a patent search formula from Derwent Innovation Index, and use the patent citation information to build a network.

The second step is data cleaning. First, we merge the patentees (including the merger of subsidiaries and parent companies, the merger of renamed companies). Then we calculate and rank the citing times and cited times of the patentees based on the patent citation information. Finally, the 20 firms with the largest sum of citing times and cited times are selected. The firms and institutes are shown in the following table.

The third step is network construction. After acquiring 20 companies, we can build the network. Here we use python to construct and analyze the network. We use “network” (version: networkx2.4) for network construction and network attribute analysis, and use “igraph” (version: python-igraph0.8.0) to draw the network graph. The reference matrix between firms and institutes is shown in the following table.

Network is a python package software for creation, manipulation and the study of structure and functions of the complex networks. With this tool, we can load and store networks in a standard data format, can generate multiple types of networks, analyze network structures, build network models, draw networks, etc. (Meghanathan, 2017)

D. Major topics forming each turbulence

At each turbulence, the ranking of the Topics change is shown in Table A2. We organized the Topics that contributed more than 80% to the turbulence (for market and policy dimensions, this value is 90%). Words in bold are newly emerged ones in that turbulence.