Collaborative knowledge-driven governance: Types and mechanisms of collaboration between science, social science, and local knowledge

Collaborative knowledge-driven governance: Types and mechanisms of collaboration between science,... Abstract Knowledge plays an important role in modern public governance characterized by complexity, but collaboration between different types of knowledge in public governance has not been systematically studied. Increasingly, literature has stressed the importance of the application of science, social science, and local knowledge in public governance, whereas it has paid little attention to the types and mechanisms of collaboration between these three fields. The aims of this study were to explore the influence of collaboration between the three types of knowledge on governance performance, the major types of collaboration, and the major institutional design principles for successful collaboration. Based on a combined field study including surveys, interviews, observations, and archive data as well as a meta-analysis study on desertification control in northern China, the largest developing country in the world, this study made the following three key findings: (1) Although natural science was the most widely applied area of knowledge and social science was least applied, the order of the correlation coefficients of the three types of knowledge with governance performance from the highest to the lowest was social science, local knowledge, and natural science. (2) Collaboration between these three types of knowledge influenced governance performance. The types of collaboration with low levels of all three types of knowledge always had low governance performance, and the types of collaboration with high levels of social science and local knowledge often had high performance. (3) Successful collaboration among different types of knowledge shared nine significant institutional design principles. These principles stressed the integration of three types of knowledge, the collaboration among knowledge possessors and other social actors, and reliable and sustainable external support (government, financial, and institutional). These findings shed new light on collaboration between science, social science, and local knowledge in public and environmental governance in China as well as in other countries around the world. 1. Introduction: conflict or collaboration? Although there is an old and protracted debate about whether knowledge is good or bad for human beings (e.g. Chuang 1968), knowledge always plays an important role in our society (Hayek 1945; Gessler 2015) and its public governance (Ostrom 2005; Osborne 2006), especially in modern public governance characterized by complexity (Moynihan et al. 2010). Since Sir Isaac Newton, positivist ways of knowing have dominated modern society (Fazey et al. 2006; Innes and Booher 2010), and people often deem scientific knowledge (natural science and technology) and local knowledge as distinct, and even opposed and conflicting, types of knowledge (Thomas and Twyman 2004) in public governance. Thus, local knowledge and sometimes even many social sciences have been ignored in discourses on public governance. For example, before the Reform and Opening Up Policy in 1978, and especially during the Great Cultural Revolution, local knowledge was often regarded as backward, primitive, ignorant, and even stupid, and many social sciences (e.g. political science, public administration, and sociology) were deemed as useless and bourgeois disciplines and subjects in Mainland China (Zhao 2009). On the contrary, an increasing number of studies have reemphasized the importance of social sciences (e.g. Hayek 1945; Ruttan 1984; Landry et al. 2001; Olmos-Peñuela et al. 2014) and local or traditional knowledge (e.g. Berkes et al. 2000; Fischer 2000; Taylor and de Loë 2012) in various types of public governance and have emphasized the importance of hybrid knowledge (Thomas and Twyman 2004). In addition to the studies on research collaboration (e.g. Katz and Martin 1997; Abbasi et al. 2012), scientists’ or scientific collaboration (e.g. Newman 2001), organizational collaboration (e.g. Powell et al. 1996), and even scientific collaboration between countries (e.g. Wagner and Leydesdorff 2005; Hennemann et al. 2012), rich literature originating from the field of science and technology studies and other related fields explored also emphasized the importance of collaboration between different types of knowledge, in particular, through analyzing co-management or knowledge partnership (Berkes 2009; Zurba et al. 2012; Watson 2013), the co-production of knowledge (Maclean and Cullen 2010; Armitage et al. 2011), co-producing knowledge or joint knowledge production (Edelenbos et al. 2011; Hegger et al. 2012), interactive knowledge production (Pohl 2008; Giebels et al. 2015), knowledge integration (Hill et al. 2012; Ginger 2014), boundary organizations (Guston 2001; Miller 2001; Cash et al. 2006), boundary objects (Star and Griesemer 1989), science–policy interfaces (Heink et al. 2015; Rudd 2015), and bridging tools (Hill 2006) based on the studies mainly done in developed countries such as America, Australia, Canada, and The Netherlands. Within the public administration literature, many studies since the 1960s have emphasized the important roles of knowledge, science, and social sciences in public policy-making, implementation, and other types of public administration. For example, some studied the role of knowledge (e.g. Landry et al. 2003; Meyer et al. 2007; Daviter 2015) in public administration and policy; some focused on the role of science (e.g. Lasswell 1956; Yang et al. 2013); and some explored the role of social sciences (e.g. Bunker 1978; Caplan 1979; Landry et al. 2001) or knowledge about public administration (Behn 1987), local knowledge (e.g. Yang 2015), and even collaboration of social science knowledge and local knowledge or practitioner wisdom (Bardach 1987). Still others even studied the relationships between knowledge (or science and social science) and politics (Torgerson 1986), power (e.g. Albæk 1995), government (e.g. Cairl and Gallagher 1968; Van der Meulen 1998; Lambright 2008), or public policy (e.g. Denny 1967; Bunker 1978) as well as knowledge for development (Borda-Rodriguez and Johnson 2013) and knowledge accumulation (Ko 2013). Furthermore, some studies in the field of environmental governance also have emphasized the importance of collaboration between different types of knowledge in various types of public governance (e.g. Ison et al. 2007; Stringer et al. 2009; Holm et al. 2013; Yang 2015). However, collaboration between different types of knowledge in both public administration and environmental governance has not been systematically studied, especially in many developing countries, such as China (e.g. Yang and Wu 2010; Yang et al. 2013). Thus, in the current study, I examined how collaboration between the three types of knowledge influences the performance of public governance in China. In particular, I explored types of collaboration with the three types of knowledge and the mechanism design (Hurwicz and Reiter 2008) or institutional design principles (Ostrom 1990, 2005) of successful knowledge collaboration in public governance. This investigation can help us not only expand upon the work of many scholars and researchers who have explored the application of knowledge in various types of public governance and the dilemma of knowledge application (Yang 2010) but also understand the influences of knowledge application on public governance performance by analyzing the types and institutional design principles as key mechanisms of knowledge collaboration. Furthermore, if the dilemma caused by the gaps and conflicts between knowledge and practice can be called the first order dilemma of knowledge application (Yang 2010), the dilemma caused by the gaps and conflicts between different types of knowledge can be called the second-order dilemma of knowledge application. Thus, this study can contribute to understanding and resolving both first-order and second-order dilemmas of knowledge application in public governance (Ostrom 1990; Yang 2010), which are linked to the debate about whether knowledge is good or bad for humans, especially in some developing countries (such as China), which are facing multiple and arduous challenges of modernization through the use of modern thoughts and knowledge (Deng 1993; Sun 2011; Hu 2012). World Bank (1998: 3, 6) also pointed out ever: ‘Most of the difficulties that developing countries face involve both knowledge gaps and information problems.’ ‘Development institutions have three roles in reducing knowledge gaps: to provide international public goods, to act as intermediaries in the transfer of knowledge, and to manage the rapidly growing body of knowledge about development.’ Desertification, one of the greatest environmental challenges in our time, refers to land degradation in arid, semi-arid, and dry subhumid areas caused by climate change and human activities (UN 1992; Wu 2001). Desertification often leads to the permanent loss of land productivity, serious dust storms, and various ecological, environmental, economical, and even political consequences (Goudie 2009). Desertified areas account for approximately 41 per cent of the world’s land surface and have influenced the lives of more than 38 per cent of the world’s population (Reynolds et al. 2007). In China, the problem of desertification has also become increasingly serious in the past several decades, and desertified areas have expanded from 137,000 km2 in the 1950s to 385,700 km2 in 2000, with an annual expansion rate of 3,600 km2/year (Wang et al. 2004). Furthermore, different types of knowledge have been applied to combat desertification in China (Yang and Wu 2010). Thus, taking desertification control since 1949, the year of the foundation of new China, as a concrete example of public governance in China can help us not only better understand collaboration between different types of knowledge in public governance but also make a contribution to finally resolving the serious desertification problem in modern China. Based on a collaboration of quantitative and qualitative studies in three adjacent provinces in northern China (Inner Mongolia, Ningxia, and Gansu), I answered the following three research questions: (1) What is the significance of the collaboration between natural science, social science, and local knowledge (or natural scientific, social scientific, and local knowledge) in influencing the performance of desertification control? (2) What are the major types of collaboration between the three types of knowledge? (3) What are the major institutional design principles for successful collaboration between the three types of knowledge in desertification control? Based on the assumption that heterogeneity exists in the levels of three types of knowledge, types, and institutions of knowledge collaboration, and the performance of desertification control, the hypothesis of the study was that the collaboration of the three types of knowledge and their collaboration types and mechanisms have influenced the performance of desertification control. 2. Conceptual background and theory 2.1 Conceptual background In this study, both natural science and social science, or natural and social sciences (Watson 2013), are considered scientific knowledge or science, which had been widely verified using scientific methods and had a high degree of generalizability (Aikenhead and Ogawa 2007; Dickison 2010) (Fig. 1). Concretely, natural science in combating desertification in the study includes agricultural science and technology, agricultural pest control, zoology or animal biology, forestry, knowledge about combating desertification and dust storms, general climatic knowledge, hydraulic engineering, specific knowledge about local desertification, poultry and livestock disease control, etc. (Yang et al. 2013). Social science in desertification control includes political science, public administration, policy analysis, management, economics, law, sociology, general knowledge of environmental governance, knowledge of laws and regulations, understanding local social relations, knowledge of social management, etc. Local knowledge, or indigenous (Smith 1999), traditional (Berkes 1999), and community knowledge (Corburn 2007), is often developed from local experience, rules, wisdom, memories, stories, history, and practices and is only applied or verified in some areas. In this study, local knowledge included all types of ‘culture-specific information, knowledge, skills, norms, taboos, codes of conduct, customs, norms of behavior, conventions, and traditions on desertification control that are based on local experience, wisdom, practices, and histories and are mainly owned by the locals’ (Yang 2015: 617). In order to differentiate the three types of knowledge and make sure people can make distinctions between these categories, natural science and social science were directly measured based on the evaluations of their various concrete branches in the study, while local knowledge was carefully explained when necessary (e.g. in surveys and interviews mentioned later). Figure 1. View largeDownload slide The classification of three types of knowledge. Figure 1. View largeDownload slide The classification of three types of knowledge. Collaboration, as a concept with high valence as defined by Cox and Béland (2013), means two or more parties or stakeholders co-labor or work together (Agranoff and McGuire 2003; O’Leary et al. 2006) to solve the problems which cannot be solved or easily solved by individuals or single organizations (Gray 1985; McGuire 2006), although collaboration does not necessarily lead to increased levels of cooperation (Lubell 2004) and differentiated outcomes (Scott 2015). On the one hand, collaboration between the three types of knowledge means that the three types of knowledge are formally organized (Ansell and Gash 2007) or combined to solve the problems which could not be solved depending on any one of the three and to accomplish the common goal (Katz and Martin 1997) of desertification control. In this stance, ‘collaboration of knowledge’ can be simply changed to ‘combination of knowledge’, and the use of the ‘collaboration’ is different from the usual use of collaboration between multiple actors. On the other hand, collaboration of knowledge also means the collaboration between knowledge possessors because knowledge is often possessed by various social actors, and this use of collaboration is the same as its usual use. That is, in order to avoid the confusion and wordy problem by using two concepts—‘combination of knowledge’ and ‘collaboration of knowledge possessors’, here I use collaboration of knowledge to cover the meaning of both ‘combination of knowledge’ and ‘collaboration of knowledge possessors’. Finally, institutional design principles are essential elements or conditions (Ostrom 1990, 2005) that help to account for the success of knowledge collaboration institutions. 2.2 Theoretical framework In order to study the structural contexts (O'Toole and Meier 1999) of knowledge collaboration, the best method is to develop a taxonomy of knowledge collaboration, because unlike regression, which can only study the relationship between two variables (Davis and Marquis 2005), a taxonomy can not only help researchers investigate the inner structure of knowledge collaboration but also provide a useful foundation to develop theories (Moore and Koontz 2003; Hill et al. 2012). Furthermore, as in the studies on stakeholders’ collaboration in which the level of participation is often used to classify the types of stakeholders’ collaboration (Wellens 1975), the taxonomy of knowledge collaboration can also be classified through analyzing the level of knowledge application in governance. Thus, if we divided the knowledge (natural science, social science, and local knowledge) used in desertification control into high and low levels, then the collaboration types of knowledge could be divided into eight types including High Natural Science, High Social Science, High Local Knowledge and Low Natural Science, Low Social Science, Low Local Knowledge (see Fig. 2). Figure 2. View largeDownload slide Theoretical framework. Figure 2. View largeDownload slide Theoretical framework. Furthermore, a systematical review for the literature over a period of some 20 years found that the application of knowledge and the relationships between different types of applied knowledge were often influenced by nine major factors: (1) levels of knowledge (Thomas 1997; Reynolds et al. 2007; Chasek et al. 2011; Reed et al. 2007; Yang et al. 2013); (2) the application and extension of knowledge (Honadle 1994; Yang and Wu 2010); (3) coordination between different types of knowledge (Thomas and Twyman 2004; Ison et al. 2007; Ansell and Gash 2008; Stringer et al. 2009; Holm et al. 2013); (4) the capabilities and endeavors of knowledge possessors (Yang and Wu 2010; Emerson et al. 2012); (5) communication and collaboration among social actors (Fullen and Mitchell 1994; Papageorgiou 1994; Thomas 1997; Dietz et al. 2003; Tschakert 2007; Stone 2009; Yang and Wu 2010; Borda-Rodriguez and Johnson 2013); (6) other social actors’ support and capabilities (Landry et al. 2003; Yang and Wu 2010; Emerson et al. 2012); (7) government support; (8) legal mechanisms (Hill 2006) and institutional (or cultural) (Cracknell 2001) support; and (9) financial support (Campbell 1992; Yang and Wu 2010; Borda-Rodriguez and Johnson 2013). Previous studies also found that the factors influencing the role of knowledge and science as well as the effectiveness of knowledge collaboration in desertification control could be divided into three aspects: knowledge itself, social actors, and external support (Yang and Li 2015). Thus, in the current study, I integrated the related literature and divided the aforementioned factors into the above three groups and analyzed the mechanisms or nine principles of the collaboration of natural science, social science, and local knowledge. Moreover, I supposed that both the types and the factors shaping the role of knowledge and science influenced governance performance of desertification control. Finally, the theoretical framework of the study was shown in Fig. 2. 3. Research design and methods 3.1 Research design and sites In order to compare knowledge collaboration and their influences on governance performance, a two-step study was conducted. In the first step, the field study was first conducted in twelve typical arid and semi-arid counties from three adjacent provinces with a long history of desertification control and different desertification control results (two counties in Gansu, two in Ningxia, and eight in Inner Mongolia) (Fig. 3). Furthermore, in all of these counties, there were laboratories and field stations of the Chinese Academy of Sciences (CAS) and the Chinese Academy of Forestry. The basic natural characteristics of the twelve counties, such as climate division, total area, and population, are shown in Table 1a. Table 1. Characteristics of the twelve field study counties and 25 meta-analysis cases. Dimensions Counties  Provinces  Total area (km2)  Climate division  Population (10,000)  Annual average temperature (°C)  Annual average precipitation (mm)  Annual average evaporation (mm)  A. Twelve field study casesa                (1) Linze (2001b)  Gansu  3,148  Aridc  15  7.7  115  2212  (2) Minqin (1994)  Gansu  16,016  Arid  27.4  7.8  115  2644  (3) Zhongwei (1995)  Ningxia  5,780  Arid  109.29  9.5  188  1914  (4) Yanchi (2004)  Ningxia  8,661  Arid  15.7  7.7  <300  >2000  (5) Dengkou (1998)  Inner Mongolia  3,554  Arid  9.6  7.6  145  2398  (6) Ejin Horo (2011)  Inner Mongolia  5,600  Semi-arid  15.5  6.7  348  2563  (7) Xinbaerhuzuo (2007)  Inner Mongolia  22,000  Semi-arid  4.2  −0.3  268  1650  (8) Xilinhot (2004)  Inner Mongolia  15,179  Semi-arid  25.2  1.6  250–350  1746  (9) Naiman (2001)  Inner Mongolia  8,159  Semi-arid  41.9  6∼6.5  366  1973–2082  (10) Duolun (2000)  Inner Mongolia  3,773  Semi-arid  10.3  1.9  389  1714  (11) Wengniute (1993)  Inner Mongolia  11,882  Semi-arid  41.6  4.5  370  2106  (12) Aohan (1990)  Inner Mongolia  8,294  Semi-arid  51.3  5–7  310–460  2162  B. Twenty-five meta-analysis cases                (1) Zhengxiangbai (2005)  Inner Mongolia  6,229  Arid  7.3  1.9  268–360  1,932–2,300  (2) Zhenglan (2006)  Inner Mongolia  10,182  Arid  8.3  1.5  365  1,925.5  (3) Siziwang (2005)  Inner Mongolia  25,516  Arid  20.9  1-6  110–350  2,300–2,400  (4) Wulatehou (2001)  Inner Mongolia  25,000  Extremely arid  6.1  3.8  96–105.9  3,000–3,500  (5) Ejina (2006)  Inner Mongolia  114,606  Extremely arid  2  8.6  37  3,841.51  (6) Wushen (2005)  Inner Mongolia  11,645  Arid  10  6.8  350–400  2,200–2,800  (7) Guyang (2002)  Inner Mongolia  5,021  Arid  21  4  300  2,200  (8) Tongliao (2001)  Inner Mongolia  59,535  Arid  309.1  0–6  350–400  2,000  (9) Shangdu (2004)  Inner Mongolia  4,353  Arid  33  3.1  351.5  2,500  (10) Balinyou (2004)  Inner Mongolia  10,256  Semi-arid  18.1  4.9  350  >1,500  (11) Dalate (2004)  Inner Mongolia  8,188  Arid  34  6.1–7.1  240–360  2,600  (12) Wuwei (2000)  Gansu  33,249  Arid  191.8  7.8  60–610  1,400–3,010  (13) Shandan (1998)  Gansu  5,402  Arid  19.2  5.9  195  2,246  (14) Guazhou (2003)  Gansu  23,150  Extremely arid  11.9  8.8  45.7  3,140.6  (15) Dunhuang (2005)  Gansu  31,200  Extremely arid  18.3  9.4  39.9  2,486  (16) Maqu (1995)  Gansu  10,109  Semi-arid  3.71  1.1  616.5  1,353.4  (17) Turpan (2000)  Xinjiang  13,689  Extremely arid  28  13.9  6.9–25.2  3,837  (18) Golmud (1998)  Qinghai  126,220  Extremely arid  12.12  −4.2  41.5  >3,000  2300  (19) Yulin (2005)  Shaanxi  43,578  Arid  335  10.7  290–465    (20) Zhangbei (2001)  Hebei  4,185  Arid  37.2  2.6  300  1,772  (21) Guyuan (2005)  Hebei  3,601  Semi-arid  23  1.4  426  1,787.5  (22) Fengning (1994)  Hebei  8,765  Semi-arid  37.4  3–4  436.7  1,958  (23) Daxing (2006)  Beijing  1,031  Arid  136.5  11.6  556  1,800  (24) Kangping (1995)  Liaoning  2,175  Semi-arid  35  6.9  542.9  2,037.6  (25) Tailai (1998)  Heilongjiang  3,996  Semi-arid  32  4.9  392.6  1,717.1  Dimensions Counties  Provinces  Total area (km2)  Climate division  Population (10,000)  Annual average temperature (°C)  Annual average precipitation (mm)  Annual average evaporation (mm)  A. Twelve field study casesa                (1) Linze (2001b)  Gansu  3,148  Aridc  15  7.7  115  2212  (2) Minqin (1994)  Gansu  16,016  Arid  27.4  7.8  115  2644  (3) Zhongwei (1995)  Ningxia  5,780  Arid  109.29  9.5  188  1914  (4) Yanchi (2004)  Ningxia  8,661  Arid  15.7  7.7  <300  >2000  (5) Dengkou (1998)  Inner Mongolia  3,554  Arid  9.6  7.6  145  2398  (6) Ejin Horo (2011)  Inner Mongolia  5,600  Semi-arid  15.5  6.7  348  2563  (7) Xinbaerhuzuo (2007)  Inner Mongolia  22,000  Semi-arid  4.2  −0.3  268  1650  (8) Xilinhot (2004)  Inner Mongolia  15,179  Semi-arid  25.2  1.6  250–350  1746  (9) Naiman (2001)  Inner Mongolia  8,159  Semi-arid  41.9  6∼6.5  366  1973–2082  (10) Duolun (2000)  Inner Mongolia  3,773  Semi-arid  10.3  1.9  389  1714  (11) Wengniute (1993)  Inner Mongolia  11,882  Semi-arid  41.6  4.5  370  2106  (12) Aohan (1990)  Inner Mongolia  8,294  Semi-arid  51.3  5–7  310–460  2162  B. Twenty-five meta-analysis cases                (1) Zhengxiangbai (2005)  Inner Mongolia  6,229  Arid  7.3  1.9  268–360  1,932–2,300  (2) Zhenglan (2006)  Inner Mongolia  10,182  Arid  8.3  1.5  365  1,925.5  (3) Siziwang (2005)  Inner Mongolia  25,516  Arid  20.9  1-6  110–350  2,300–2,400  (4) Wulatehou (2001)  Inner Mongolia  25,000  Extremely arid  6.1  3.8  96–105.9  3,000–3,500  (5) Ejina (2006)  Inner Mongolia  114,606  Extremely arid  2  8.6  37  3,841.51  (6) Wushen (2005)  Inner Mongolia  11,645  Arid  10  6.8  350–400  2,200–2,800  (7) Guyang (2002)  Inner Mongolia  5,021  Arid  21  4  300  2,200  (8) Tongliao (2001)  Inner Mongolia  59,535  Arid  309.1  0–6  350–400  2,000  (9) Shangdu (2004)  Inner Mongolia  4,353  Arid  33  3.1  351.5  2,500  (10) Balinyou (2004)  Inner Mongolia  10,256  Semi-arid  18.1  4.9  350  >1,500  (11) Dalate (2004)  Inner Mongolia  8,188  Arid  34  6.1–7.1  240–360  2,600  (12) Wuwei (2000)  Gansu  33,249  Arid  191.8  7.8  60–610  1,400–3,010  (13) Shandan (1998)  Gansu  5,402  Arid  19.2  5.9  195  2,246  (14) Guazhou (2003)  Gansu  23,150  Extremely arid  11.9  8.8  45.7  3,140.6  (15) Dunhuang (2005)  Gansu  31,200  Extremely arid  18.3  9.4  39.9  2,486  (16) Maqu (1995)  Gansu  10,109  Semi-arid  3.71  1.1  616.5  1,353.4  (17) Turpan (2000)  Xinjiang  13,689  Extremely arid  28  13.9  6.9–25.2  3,837  (18) Golmud (1998)  Qinghai  126,220  Extremely arid  12.12  −4.2  41.5  >3,000  2300  (19) Yulin (2005)  Shaanxi  43,578  Arid  335  10.7  290–465    (20) Zhangbei (2001)  Hebei  4,185  Arid  37.2  2.6  300  1,772  (21) Guyuan (2005)  Hebei  3,601  Semi-arid  23  1.4  426  1,787.5  (22) Fengning (1994)  Hebei  8,765  Semi-arid  37.4  3–4  436.7  1,958  (23) Daxing (2006)  Beijing  1,031  Arid  136.5  11.6  556  1,800  (24) Kangping (1995)  Liaoning  2,175  Semi-arid  35  6.9  542.9  2,037.6  (25) Tailai (1998)  Heilongjiang  3,996  Semi-arid  32  4.9  392.6  1,717.1  a Adapted from Yang (2015) and Yang et al. (2013). b The year of the source published. c The climate divisions were based on the criteria of United Nations Convention to Combat Desertification. Figure 3. View largeDownload slide The twelve field study counties and twenty-five meta-analysis cases. Source: SFA 2011 and Yang, 2013. Note: The order numbers of the twelve field study counties and the twenty-five meta-analysis cases are consistent with the order shown in Table 1. Figure 3. View largeDownload slide The twelve field study counties and twenty-five meta-analysis cases. Source: SFA 2011 and Yang, 2013. Note: The order numbers of the twelve field study counties and the twenty-five meta-analysis cases are consistent with the order shown in Table 1. In the second step, to test the generalizability of the institutional design principles found in the twelve counties by the field study, I also studied another twenty-five cases in nine provinces in northern China through a meta-analysis (Table 1b). Furthermore, the climate divisions of the twenty-five cases covered not only semi-arid and arid areas but also extremely arid areas. 3.2 Data collection Because there were no other economically scientific methods to collect valid data due to fragmented and incoherent records and non-comparability among different types of records and measures (Sun et al. 2006), the study used four types of methods, including quantitative surveys, qualitative interviews, personal observations, and document analysis, to collect complementary and cross-checked data (Poteete and Ostrom 2008; Poteete et al. 2010) and to form an evidence or data triangle (Patton 1987). Previous studies indicated that a combination of surveys, observations, interviews, and document analysis as a hands-on research technique has proved to be a valid and efficient method for collecting data (e.g. Leach et al. 2002; Poteete et al. 2010; Perry 2012). First, after preinterviews, pre-surveys, archive analysis, and considering the population of each county (Table 2a), in order to guarantee a large sample size to improve the consistency between survey respondents’ perception and actual situations of desertification control, formal surveys were conducted from March to December 2011, with 5,410 copies mailed (the number of sent copies in each county was equal to or over 450) and 4,406 valid responses (Table 2b). Because students in local high schools resided in different regions or townships within the county and could be deemed as a valid and hands-on sample to represent the whole population of the county (Yang et al. 2013), the questionnaires were first randomly distributed to high school students, who were trained to help their relatives and neighbors (including many old farmers and herdsmen, who could not read or did not know how to answer a questionnaire) to complete questionnaires. When respondents forgot or did not know related information, they were also encouraged to get help or talk with other people who had more information or personal experience on desertification control but did not fill in the questionnaires by themselves. Furthermore, because desertification control programs were often conducted in underdeveloped rural villages where people are less influenced by the outside and each decade from 1950s to the 1990s had extremely different programs, movements, and policies, many people could remember (in their own words is ‘would never forget’) their experiences during these decades. The descriptive statistics of the valid responses showed that respondents’ ages ranged from 17 to over 80 years old and their occupations were quite diverse and included fifteen types such as farmers, government officials, researchers, teachers, and businessmen (Table 2b). Table 2. Survey and interview distribution in the twelve counties in northern China (2006–11)a. Counties  Linze  Minqin  Zhongwei  Yanchi  Dengkou  Ejin Horo  Xinbaerhuzuo  Xilinhot  Naiman  Duolun  Wengniute  Aohan  Total  A. Population (Ten thousand)  15.01c  27.47  38.3  16.79  12.34  23.9  4.17  17.37  44.30  10.70  47.70  60.32    B. Survey distribution                            The number of sent copies  450  450  450  450  450  450  450  450  450  450  460  450  5410  Response rates (%)  75.78  100  80.00  99.56  72.00  38.89  86.00  93.56  96.00  100  100  100  86.82  The number of valid copies  328  418  345  439  304  150  387  342  424  449  458  362  4406  Validity rate among received copies (%)  96.19  92.89  95.83  97.99  93.83  85.71  100  81.23  98.15  99.78  99.57  80.44  93.78  C. Types of survey respondents                            Farmers  97  382  130  75  72  53  186  76  70  149  438  256  1984  (29.57)d  (91.39)  (37.68)  (17.08)  (23.68)  (35.33)  (48.06)  (22.22)  (16.51)  (33.18)  (95.63)  (70.72)  (45.03)  Middle schools (teachers and students)  91  8  58  166  99  38  45  99  134  135  11  21  905  (27.74)  (1.91)  (16.81)  (37.81)  (32.57)  (25.33)  (11.63)  (28.95)  (31.60)  (30.07)  (2.40)  (5.80)  (20.54)  General research institutesb  0  1  2  27  2  0  5  3  11  15  0  2  68  (0)  (0.24)  (0.58)  (6.15)  (0.66)  (0)  (1.29)  (0.88)  (2.60)  (3.34)  (0)  (0.55)  (1.54)  Desert control stations  0  0  0  2  2  0  1  0  2  8  0  0  15  (0)  (0)  (0)  (0.46)  (0.66)  (0)  (0.26)  (0)  (0.47)  (1.78)  (0)  (0)  (0.34)  Government  14  1  9  15  13  4  14  24  5  32  1  9  141  (4.27)  (0.24)  (2.61)  (3.42)  (4.28)  (2.67)  (3.62)  (7.02)  (1.18)  (7.13)  (0.22)  (2.49)  (3.20)  Businesses  55  8  48  55  53  10  18  63  14  34  4  28  390  (16.77)  (1.91)  (13.91)  (12.53)  (17.43)  (6.67)  (4.65)  (18.42)  (3.30)  (7.57)  (0.87)  (7.73)  (8.85)  Rural grassroots organizations  7  9  24  15  4  2  2  4  2  41  0  10  120  (2.13)  (2.15)  (6.96)  (3.42)  (1.31)  (1.33)  (0.52)  (1.17)  (0.47)  (9.13)  (0)  (2.76)  (2.72)  Organizations of technology development and promotion in rural areas  4  1  0  3  2  0  4  2  3  1  0  1  21  (1.22)  (0.24)  (0)  (0.68)  (0.66)  (0)  (1.03)  (0.58)  (0.71)  (0.22)  (0)  (0.28)  (0.48)  Universities  1  1  5  18  0  0  4  4  12  0  0  0  45  (0.31)  (0.24)  (1.45)  (4.10)  (0)  (0)  (1.03)  (1.17)  (2.83)  (0)  (0)  (0)  (1.02)  Religious groups  0  0  2  4  0  0  0  2  2  0  0  0  10  (0)  (0)  (0.58)  (0.91)  (0)  0  (0)  (0.58)  (0.47)  (0)  (0)  (0)  (0.23)  Other public institutes  25  1  25  10  20  16  10  40  3  13  3  26  192  (7.62)  (0.24)  (7.25)  (2.28)  (6.58)  (10.67)  (2.58)  (11.70)  (0.71)  (2.90)  (0.66)  (7.18)  (4.36)  Non-governmental organizations  5  0  5  4  3  6  1  3  2  1  0  2  32  (1.52)  (0)  (1.45)  (0.91)  (0.99)  (4)  (0.26)  (0.88)  (0.47)  (0.22)  (0)  (0.55)  (0.73)  News media  1  0  1  2  0  1  0  3  5  0  0  0  13  (0.31)  (0)  (0.29)  (0.46)  (0)  (0.67)  (0)  (0.88)  (1.18)  (0)  (0)  (0)  (0.30)  International organizations  0  0  4  2  0  0  1  0  15  0  0  1  23  (0)  (0)  (1.16)  (0.46)  (0)  (0)  (0.26)  (0)  (3.54)  (0)  (0)  (0.28)  (0.52)  Others  28  6  32  41  34  20  96  19  144  20  1  6  447  (8.54)  (1.44)  (9.27)  (9.33)  (11.18)  (13.33)  (24.81)  (5.56)  (33.96)  (4.46)  (0.22)  (1.66)  (10.15)  D. Interview distribution                            Farmers or residents  4  6  5  1  1  2  2  1  1  1  1  1  26  Scholars, experts & technicians  3  11  4  4  2  3  0  4  5  0  2  4  42  Government officials  1  11  1  3  6  3  3  4  1  3  5  4  45  Businessmen  0  0  0  0  0  0  0  0  2  0  2  0  4  Religious groups or NGOs  0  1  0  0  0  0  0  0  0  0  0  0  1  Total  8  29  10  8  9  8  5  9  9  4  10  9  118  E. Observation distribution                            Numbers  4  11  7  2  9  2  2  2  5  3  2  3  52  Counties  Linze  Minqin  Zhongwei  Yanchi  Dengkou  Ejin Horo  Xinbaerhuzuo  Xilinhot  Naiman  Duolun  Wengniute  Aohan  Total  A. Population (Ten thousand)  15.01c  27.47  38.3  16.79  12.34  23.9  4.17  17.37  44.30  10.70  47.70  60.32    B. Survey distribution                            The number of sent copies  450  450  450  450  450  450  450  450  450  450  460  450  5410  Response rates (%)  75.78  100  80.00  99.56  72.00  38.89  86.00  93.56  96.00  100  100  100  86.82  The number of valid copies  328  418  345  439  304  150  387  342  424  449  458  362  4406  Validity rate among received copies (%)  96.19  92.89  95.83  97.99  93.83  85.71  100  81.23  98.15  99.78  99.57  80.44  93.78  C. Types of survey respondents                            Farmers  97  382  130  75  72  53  186  76  70  149  438  256  1984  (29.57)d  (91.39)  (37.68)  (17.08)  (23.68)  (35.33)  (48.06)  (22.22)  (16.51)  (33.18)  (95.63)  (70.72)  (45.03)  Middle schools (teachers and students)  91  8  58  166  99  38  45  99  134  135  11  21  905  (27.74)  (1.91)  (16.81)  (37.81)  (32.57)  (25.33)  (11.63)  (28.95)  (31.60)  (30.07)  (2.40)  (5.80)  (20.54)  General research institutesb  0  1  2  27  2  0  5  3  11  15  0  2  68  (0)  (0.24)  (0.58)  (6.15)  (0.66)  (0)  (1.29)  (0.88)  (2.60)  (3.34)  (0)  (0.55)  (1.54)  Desert control stations  0  0  0  2  2  0  1  0  2  8  0  0  15  (0)  (0)  (0)  (0.46)  (0.66)  (0)  (0.26)  (0)  (0.47)  (1.78)  (0)  (0)  (0.34)  Government  14  1  9  15  13  4  14  24  5  32  1  9  141  (4.27)  (0.24)  (2.61)  (3.42)  (4.28)  (2.67)  (3.62)  (7.02)  (1.18)  (7.13)  (0.22)  (2.49)  (3.20)  Businesses  55  8  48  55  53  10  18  63  14  34  4  28  390  (16.77)  (1.91)  (13.91)  (12.53)  (17.43)  (6.67)  (4.65)  (18.42)  (3.30)  (7.57)  (0.87)  (7.73)  (8.85)  Rural grassroots organizations  7  9  24  15  4  2  2  4  2  41  0  10  120  (2.13)  (2.15)  (6.96)  (3.42)  (1.31)  (1.33)  (0.52)  (1.17)  (0.47)  (9.13)  (0)  (2.76)  (2.72)  Organizations of technology development and promotion in rural areas  4  1  0  3  2  0  4  2  3  1  0  1  21  (1.22)  (0.24)  (0)  (0.68)  (0.66)  (0)  (1.03)  (0.58)  (0.71)  (0.22)  (0)  (0.28)  (0.48)  Universities  1  1  5  18  0  0  4  4  12  0  0  0  45  (0.31)  (0.24)  (1.45)  (4.10)  (0)  (0)  (1.03)  (1.17)  (2.83)  (0)  (0)  (0)  (1.02)  Religious groups  0  0  2  4  0  0  0  2  2  0  0  0  10  (0)  (0)  (0.58)  (0.91)  (0)  0  (0)  (0.58)  (0.47)  (0)  (0)  (0)  (0.23)  Other public institutes  25  1  25  10  20  16  10  40  3  13  3  26  192  (7.62)  (0.24)  (7.25)  (2.28)  (6.58)  (10.67)  (2.58)  (11.70)  (0.71)  (2.90)  (0.66)  (7.18)  (4.36)  Non-governmental organizations  5  0  5  4  3  6  1  3  2  1  0  2  32  (1.52)  (0)  (1.45)  (0.91)  (0.99)  (4)  (0.26)  (0.88)  (0.47)  (0.22)  (0)  (0.55)  (0.73)  News media  1  0  1  2  0  1  0  3  5  0  0  0  13  (0.31)  (0)  (0.29)  (0.46)  (0)  (0.67)  (0)  (0.88)  (1.18)  (0)  (0)  (0)  (0.30)  International organizations  0  0  4  2  0  0  1  0  15  0  0  1  23  (0)  (0)  (1.16)  (0.46)  (0)  (0)  (0.26)  (0)  (3.54)  (0)  (0)  (0.28)  (0.52)  Others  28  6  32  41  34  20  96  19  144  20  1  6  447  (8.54)  (1.44)  (9.27)  (9.33)  (11.18)  (13.33)  (24.81)  (5.56)  (33.96)  (4.46)  (0.22)  (1.66)  (10.15)  D. Interview distribution                            Farmers or residents  4  6  5  1  1  2  2  1  1  1  1  1  26  Scholars, experts & technicians  3  11  4  4  2  3  0  4  5  0  2  4  42  Government officials  1  11  1  3  6  3  3  4  1  3  5  4  45  Businessmen  0  0  0  0  0  0  0  0  2  0  2  0  4  Religious groups or NGOs  0  1  0  0  0  0  0  0  0  0  0  0  1  Total  8  29  10  8  9  8  5  9  9  4  10  9  118  E. Observation distribution                            Numbers  4  11  7  2  9  2  2  2  5  3  2  3  52  a Adapted from Yang (2015) and Yang et al. (2013). b ‘Types of organizations’ refers to people in these organizations. c The population of the counties was based on data from the Statistical Bulletins of National Economic and Social Development in 2011. d Numbers in brackets are the percentages among the valid copies. Second, in order to cross-check the survey data, semi-structured interviews as well as participatory and nonparticipatory observations in three counties (Minqin, Linze, and Zhongwei) were conducted from June 2006 to February 2008 (the first step) and in the other counties from July to August 2011 (the second step). Although the same questions were asked in both the first and second step interviews, the data from the first step were mainly used to provide basic information for survey questionnaire design. The 118 interviewees included scholars, farmers, citizens, government officials, and researchers in desert control institutes, whose ages ranged from 20 to more than 60 years old (Table 2d). Most interviews lasted approximately 30 minutes to 2 hours, and the open-ended questions corresponded to the survey questions, which provided choices. Furthermore, the 52 observed sites included desert control stations, typical areas of desertification control, famous natural reserves, the Bureau of Forestry, and the Bureau of Environmental Protection (Table 2e). Third, diverse literature and archive data, including county annals, government gazettes, government documents, research reports, news reports, and other published and non-published literature from 1949 to 2011, were collected to form a archive data triangle (Patton 1987) and to complement and cross-check the data from the surveys, interviews, and observations and for a meta-analysis of the twenty-five cases to test the generalizability of the institutional design principles found in the twelve field study counties. 3.3 Variables and measurements According to the theoretical framework, the major research variables in the study were: (1) levels of science, social science, and local knowledge applied in desertification control, (2) performance of desertification control, (3) factors influencing collaboration between the three types of knowledge, and (4) the institutional design principles for successful collaboration between the three types of knowledge. In surveys of the twelve field study counties, questions with a six-point scale (‘very large, large, medium, moderately small, very small, and don’t know’ or ‘strongly agree, agree, neutral, moderately disagree, strongly disagree, and don’t know’) were designed for respondents to directly evaluate the levels of science, social science, and local knowledge applied in desertification control, performance of desertification control, and various factors influencing the application and collaboration between the three types of knowledge of each county. In the six-point scale, the first five points in fact were similar to Likert scale. But a five-point scale as listed above could not cover the possibility when survey respondents did not know actual situations. Thus, in order to reduce the errors caused by survey respondents, I added ‘don’t know’ as the sixth point. However, in the data analysis, any response with ‘don’t know’ selected was excluded to ensure data integrity. The nine factors were evaluated by the average values of the subproblems used to measure the corresponding factors. For example, the factor ‘level of knowledge’ of each county was evaluated by the average values of three subproblems (low levels of science and technology development, low levels of social science development, and low values of local knowledge) as rated by survey respondents of the county. The Harmon one-factor analysis showed that although there was a factor that explained the variance in approximately 41 per cent of the resulting factors based on the sample of all of the valid survey responses (N = 4406), it was not a majority (over 50 per cent). This meant that the common methods bias (Podsakoff and Organ 1986) might not be a serious problem in this study. Furthermore, the study used aggregated data based on calculating the percentages of survey respondents’ answers of each county rather than individual answers to evaluate research variables of the county. This also improved the consistency between respondents’ perception of each county and actual situations of combating desertification of the county (Yang et al. 2013). Moreover, cross-checking and complementing the data from interviews, observations, and the meta-analysis as well as the protection of the respondents’ anonymity and confidentiality in the surveys also increased the validity of the evaluation of the variables. Furthermore, in order to analyze the types of knowledge collaboration in the twelve field study cases, their levels of knowledge application were also divided into two levels (high and low). For example, if the evaluation by one county was higher than the average of the twelve counties, it was deemed ‘high’; if it was less than the average, it was deemed ‘low’. When comparing the average performance of desertification control of the different types of knowledge collaboration in the twelve field study cases with the twenty-five meta-analysis cases, their performance was coded into three groups (successful, semi-successful, and unsuccessful) by trisecting the interval between the maximum and minimum accumulated percentages of the survey respondents and the values of ‘successful’ to ‘unsuccessful’ were assigned 3, 2, and 1. Finally, based on the comprehensive qualitative meta-analysis of diverse literature and archive data, the study also analyzed the types of knowledge collaboration in the twenty-five meta-analysis cases by dividing their levels of knowledge application (science, social science, and local knowledge) into two levels (high and low) and their performance of desertification control into three groups (successful, semi-successful, and unsuccessful), while the satisfaction of the nine institutional design principles corresponding to the nine factors for successful collaboration was coded into four relative levels (high, middle, low, non-satisfaction). For example, as to Principle 8, if a county (or city or district) not only strictly implemented national laws and regulations but also had its local laws and regulations for a long time, it was coded ‘high’; whereas if a county only strictly implemented national laws and regulations but did not have its own local laws and regulations for a long time, it was coded; ‘middle’; and if a county did not implement national laws and regulations very well, it was coded ‘low’. Otherwise, it was coded ‘non-satisfaction’. When calculating the correlation coefficient between the principles and performance, the values of ‘high’ to ‘non-satisfaction’ were assigned 3, 2, 1, and 0, whereas the values of ‘successful’ to ‘unsuccessful’ were also 3, 2, and 1. To avoid personal errors and subjectivity of coding results, the variables were first coded by a research assistant and her ten classmates together, and then were rechecked by the author independently. To avoid the influence by prior knowledge of research hypotheses in particular, all the ten classmates were blind to research purpose, questions, and hypotheses. 4. Results 4.1 The level of knowledge and its influence on the performance of desertification control Among the six choices, on average over 30 per cent of survey respondents indicated that the level of applying natural science and local knowledge was ‘very large’ or ‘large’ in the twelve counties, and 28 per cent said the same for social science. Therefore, the average of the types of knowledge was 30.2 per cent (Table 3a). Furthermore, approximately 30 per cent of survey respondents indicated that the performance of desertification control in the 12 counties was ‘very large’ or ‘large’. Table 3. Average levels of knowledge application and statistical results of the relationship between knowledge application with the performance of desertification control as rated by the survey respondents in the twelve field study counties in northern China (2011) (N = 12). Independent variables  A. Average levels of knowledge application in the twelve cases  B. Method 1: Correlation analysis   C. Method 2: Multivariable liner regression (Enter and Stepwise)   D. Method 3:Multivariable liner regression(Enter)   Coefficients  Sig.  B  Standard error  t  Sig.  B  Standard error  t  Sig.  Constant        −1.775  1.307  −1.359  0.175          Natural science  32.0a  0.335  0.287  −0.248***  0.022  −11.221  0.000          Social science  28.0  0.839***  0.001  0.854***  0.067  12.826  0.000          Local knowledge  30.8  0.789***  0.002  0.565***  .079  7.183  0.000          Average  30.2  0.700**  0.011          0.840***  0.055  15.147  0.000  R        0.894  0.622  R2        0.799  0.386  Adjusted R2        0.797  0.385  Standard error of the estimate        5.73095  9.97516  F(ANOVA)        478.898***  229.437***  (Sig.)        0.000  0.000  Independent variables  A. Average levels of knowledge application in the twelve cases  B. Method 1: Correlation analysis   C. Method 2: Multivariable liner regression (Enter and Stepwise)   D. Method 3:Multivariable liner regression(Enter)   Coefficients  Sig.  B  Standard error  t  Sig.  B  Standard error  t  Sig.  Constant        −1.775  1.307  −1.359  0.175          Natural science  32.0a  0.335  0.287  −0.248***  0.022  −11.221  0.000          Social science  28.0  0.839***  0.001  0.854***  0.067  12.826  0.000          Local knowledge  30.8  0.789***  0.002  0.565***  .079  7.183  0.000          Average  30.2  0.700**  0.011          0.840***  0.055  15.147  0.000  R        0.894  0.622  R2        0.799  0.386  Adjusted R2        0.797  0.385  Standard error of the estimate        5.73095  9.97516  F(ANOVA)        478.898***  229.437***  (Sig.)        0.000  0.000  a The average percentages of ‘very large’ and ‘large’ rated by survey respondents. *** P < 0.01, ** P < 0.05. The results also showed that the correlation coefficients of ‘social science’, ‘local knowledge’, and ‘the average of three types of knowledge’, with the performance of desertification control as rated by the survey respondents, produced through SPSS (Statistical Product and Service Solutions) analysis, were all significant at the 0.05 level. However, the coefficient of natural science was not significant (Table 3b). Furthermore, the results of multivariate linear regression (the data are almost normal) indicated that the coefficient of natural science was even negative, the positive coefficient of social science was larger than the coefficient of local knowledge (Table 3c), the coefficient of the average of the three types of knowledge was positive (Table 3d), and all these coefficients were significant at the 0.01 level. 4.2 Types of knowledge collaboration and their correlation with governance performance Based on the classification of the three types of knowledge (science, social science, and local knowledge) and dividing the levels of knowledge application into two levels (high and low), the types of knowledge collaboration in the twelve counties were divided into seven types. These types were consistent with the eight types shown in Fig. 2 such as (high science, high social science, high local knowledge) and (low natural science, low social science, and low local knowledge), except for (high natural science, low social science, and high local knowledge) (Table 4). Table 4. The types of collaboration among natural science, social science, and local knowledge and the average performance of desertification control in the twelve field study cases as rated by survey respondents (2011) and in the thirty-seven cases (including the twelve field study cases and twenty-five meta-analysis cases) based on the recoded data. TypesItems  (High natural science, high social science, and high local knowledge)  (High natural science, high social science, and low local knowledge)  (High natural science, low social science, and high local knowledge)  (High natural science, low social science, and low local knowledge)  (Low natural science, high social science, and high local knowledge)  (Low natural science, high social science, and low local knowledge)  (Low natural science, low social science, and high local knowledge)  (Low natural science, low social science, and low local knowledge)  Twelve field study cases  Counties falling into the types  Linze, Zhongwei, Dengkou  Duolun  No data  Ejin Horo, Wengniute  Xinbaerhuzuo  Xilinhot  Aohan  Minqin, Yanchi, Naiman  Average performance  39.9a  34.3  No data  22.5  55.2  29.6  18.4  21.4  [Orders]  [2]  [3]  [No data]  [5]  [1]  [4]  [7]  [6]  Thirty-seven casesb  Number of counties  11  3  2  3  4  4  2  8  Average performance  2.82c  2.33  2.00  1.33  2.75  2.00  1.50  1.25  [Orders]  [1]  [3]  [4]  [7]  [2]  [4]  [6]  [8]  TypesItems  (High natural science, high social science, and high local knowledge)  (High natural science, high social science, and low local knowledge)  (High natural science, low social science, and high local knowledge)  (High natural science, low social science, and low local knowledge)  (Low natural science, high social science, and high local knowledge)  (Low natural science, high social science, and low local knowledge)  (Low natural science, low social science, and high local knowledge)  (Low natural science, low social science, and low local knowledge)  Twelve field study cases  Counties falling into the types  Linze, Zhongwei, Dengkou  Duolun  No data  Ejin Horo, Wengniute  Xinbaerhuzuo  Xilinhot  Aohan  Minqin, Yanchi, Naiman  Average performance  39.9a  34.3  No data  22.5  55.2  29.6  18.4  21.4  [Orders]  [2]  [3]  [No data]  [5]  [1]  [4]  [7]  [6]  Thirty-seven casesb  Number of counties  11  3  2  3  4  4  2  8  Average performance  2.82c  2.33  2.00  1.33  2.75  2.00  1.50  1.25  [Orders]  [1]  [3]  [4]  [7]  [2]  [4]  [6]  [8]  a The average percentages of ‘very large’ and ‘large’ rated by survey respondents in the counties fallen into the types of knowledge collaboration. b Based on the recoded data (dived knowledge application into two levels : ‘High’ and ‘Low’ while divided performance into three levels: ‘successful’, ‘semi-successful’, and ‘unsuccessful’). c Given S = 3, Se = 2, and U = 1 in Table 6. Furthermore, the study also analyzed the types of knowledge collaboration in the twenty-five meta-analysis cases, and the results indicated that all the eight types of knowledge collaboration shown in Fig. 2 were found (see Table 6). The average performance of the different types of knowledge collaboration and their orders in the thirty-seven cases in total (twenty-five meta-analysis cases and twelve field study cases based on recorded data) indicated that the type with high levels of all three types of knowledge had the highest performance, while the types with low levels of all three types of knowledge had the lowest performance. Meanwhile, the types with high social science and local knowledge often had high performance, and the types with low social science and local knowledge often had low performance. 4.3 Influencing factors and mechanisms Based on the perceived subproblems influencing the application of the three types of knowledge as rated by the survey respondents, the results indicated that there were nine factors influencing the application and collaboration of knowledge (Appendix A) and they were consistent with the nine factors identified in the theoretical framework (Fig. 2) of the study. The correlation coefficients of the nine factors and the performance of desertification control in the counties were all significant at the 0.01 level (Appendix B). Thus, based on the problems rated by the survey respondents and the nine factors influencing the application and collaboration of knowledge (Appendix A) as well as the nine factors identified in the theoretical framework (Fig. 2), I produced nine design principles for successful collaboration among the three types of knowledge (Table 5). In order to test the generalizability of the findings from the field study as well as the significance of these principles, I used these principles to characterize all the twenty-five meta-analysis cases (Table 6). The table shows that the higher the satisfaction levels of the principles were the more successful performance of desertification was. Table 5. Nine design principles for successful combination between natural science, social science, and local knowledge. Nine design principles  Group 1: Combination among the three types of knowledge  P1. Sufficient developments and continued improvements in the three types of knowledge  P2. The effective refinement, transformation, adaption, application, and extension of the three types of knowledge  P3. High complementation and coordination among the three types of knowledge  Group 2: Collaboration between knowledge possessors and other social actors  P4. High capability and sustained endeavors of knowledge possessors on the application and extension of three types of knowledge  P5. Effective communication and collaboration among knowledge possessors and other social actors  P6. Sufficient support from other social actors and continued improvements on their attention and understanding of the three types of knowledge as well as their capabilities  Group 3: Reliable and sustainable external support  P7. Reliable and sustained support and guidance by governments at different hierarchical levels  P8. Sustained institutional support from laws and regulations  P9. Sufficient financial support  Nine design principles  Group 1: Combination among the three types of knowledge  P1. Sufficient developments and continued improvements in the three types of knowledge  P2. The effective refinement, transformation, adaption, application, and extension of the three types of knowledge  P3. High complementation and coordination among the three types of knowledge  Group 2: Collaboration between knowledge possessors and other social actors  P4. High capability and sustained endeavors of knowledge possessors on the application and extension of three types of knowledge  P5. Effective communication and collaboration among knowledge possessors and other social actors  P6. Sufficient support from other social actors and continued improvements on their attention and understanding of the three types of knowledge as well as their capabilities  Group 3: Reliable and sustainable external support  P7. Reliable and sustained support and guidance by governments at different hierarchical levels  P8. Sustained institutional support from laws and regulations  P9. Sufficient financial support  Note: P1–P9 refer to Principles 1–9. By using the data of Principles 1–9 and the performance of desertification of the twenty-five meta-analysis cases in Table 6, I used five statistical methods to study the relationship between the nine principles and performance of desertification control. The Kruskal–Wallis Test showed that the chi-squares of the nine principles were high and significant at the 0.01 significance level (Table 7a), and this indicated that the nine principles in the twenty-five cases were indeed different. The correlation coefficients between the nine principles and performance of desertification control indicated the correlation coefficients of all the nine principles were also significant at the 0.01 significance level, and the coefficient of Principle 9 was the largest, followed by Principles 2 and 3 (Table 7b). That is, the findings from the twelve field study cases were replicated in the twenty-five meta-analysis cases. Although the Enter multivariable liner regression (the data are normal) indicated that none of the coefficients of the nine principles were significant at the 0.05 level, its values of R and R2 were highest among the three methods of multivariable regression (Table 7c). The Backward multivariable liner regression indicated that the coefficients of Principles 1, 4, 5, 6, and 9 were significant at the 0.1 significant level, while Principles 2, 3, 7, and 8 were excluded from the model. Its values of R and R2 were less than the Enter model but higher than the Stepwise model (Table 7d). The Stepwise multivariable liner regression showed that the coefficients of Principles 3, 6, and 9 were significant at the 0.05 level, while all the other six principles were excluded. Its values of R and R2 were lowest among the three models of multivariable regression (Table 7e). Table 6. Types of knowledge combination, nine design principles, and performance of desertification control of the twelve field study cases and the twenty-five meta-analysis cases in northern China (N = 37). Cases  Types of knowledge combination   Nine design principles   Performance  Natural science  Social science  Local knowledge  P1  P2  P3  P4  P5  P6  P7  P8  P9  a. Twelve field study cases                            (1) Linze  (H  H  H)  –  –  –  –  –  –  –  –  –  S  (2) Minqin  (L  L  L)  –  –  –  –  –  –  –  –  –  U  (3) Zhongwei  (H  H  H)  –  –  –  –  –  –  –  –  –  S  (4) Yanchi  (L  L  L)  –  –  –  –  –  –  –  –  –  Se  (5) Dengkou  (H  H  H)  –  –  –  –  –  –  –  –  –  U  (6) Ejin Horo  (H  L  L)  –  –  –  –  –  –  –  –  –  Se  (7) Xinbaerhuzuo  (L  H  H)  –  –  –  –  –  –  –  –  –  S  (8) Xilinhot  (L  H  L)  –  –  –  –  –  –  –  –  –  Se  (9) Naiman  (L  L  L)  –  –  –  –  –  –  –  –  –  U  (10) Duolun  (H  H  L)  –  –  –  –  –  –  –  –  –  Se  (11) Wengniute  (H  L  L)  –  –  –  –  –  –  –  –  –  U  (12) Aohan  (L  L  H)  –  –  –  –  –  –  –  –  –  U  b. Eighteen document analysis cases                            (1) Zhengxiangbai  (L  H  H)  M  H  M  H  M  H  H  H  H  S  (2) Zhenglan  (H  L  L)  M  M  M  M  L  N  M  L  L  U  (3) Siziwang  (H  H  H)  H  H  H  H  M  H  H  M  M  S  (4) Wulatehou  (L  H  L)  M  L  M  M  H  M  H  H  L  Se  (5) Ejina  (L  L  L)  M  L  L  L  L  N  M  L  L  U  (6) Wushen  (H  H  H)  H  H  M  H  M  H  H  H  H  S  (7) Guyang  (L  L  L)  N  L  ND  L  L  N  L  L  N  U  (8) Tongliao  (H  H  H)  H  H  H  H  M  H  H  H  H  S  (9) Shangdu  (H  H  L)  M  H  M  M  H  M  H  L  M  Se  (10) Balinyou  (H  H  H)  H  H  H  M  H  H  H  H  H  S  (11) Dalate  (H  H  H)  H  H  H  H  H  H  H  H  H  S  (12) Wuwei  (H  L  H)  H  M  M  L  M  H  H  M  M  Se  (13) Shandan  (L  L  L)  L  N  N  L  L  L  H  H  L  U  (14) Guazhou  (L  H  H)  M  H  H  M  H  H  H  H  H  S  (15) Dunhuang  (H  H  H)  H  H  M  M  H  H  H  M  H  S  (16) Maqu  (L  L  L)  N  L  N  L  L  L  M  ND  L  U  (17) Turpan  (H  H  H)  H  H  H  M  H  M  H  H  H  S  (18) Golmud  (L  H  H)  M  H  M  M  M  H  H  M  M  Se  (19) Yulin  (L  L  H)  L  M  L  M  H  M  H  L  H  Se  (20) Zhangbei  (H  L  H)  H  L  L  M  M  M  H  M  L  Se  (21) Guyuan  (L  L  L)  L  L  M  H  M  H  L  M  L  Se  (22) Fengning  (H  H  L)  H  H  M  H  H  H  H  M  H  S  (23) Daxing  (H  H  H)  H  H  H  H  M  M  H  H  H  S  (24) Kangping  (L  H  L)  L  L  M  L  H  L  H  M  M  Se  (25) Tailai  (L  H  L)  L  H  L  M  M  L  H  H  H  Se  Cases  Types of knowledge combination   Nine design principles   Performance  Natural science  Social science  Local knowledge  P1  P2  P3  P4  P5  P6  P7  P8  P9  a. Twelve field study cases                            (1) Linze  (H  H  H)  –  –  –  –  –  –  –  –  –  S  (2) Minqin  (L  L  L)  –  –  –  –  –  –  –  –  –  U  (3) Zhongwei  (H  H  H)  –  –  –  –  –  –  –  –  –  S  (4) Yanchi  (L  L  L)  –  –  –  –  –  –  –  –  –  Se  (5) Dengkou  (H  H  H)  –  –  –  –  –  –  –  –  –  U  (6) Ejin Horo  (H  L  L)  –  –  –  –  –  –  –  –  –  Se  (7) Xinbaerhuzuo  (L  H  H)  –  –  –  –  –  –  –  –  –  S  (8) Xilinhot  (L  H  L)  –  –  –  –  –  –  –  –  –  Se  (9) Naiman  (L  L  L)  –  –  –  –  –  –  –  –  –  U  (10) Duolun  (H  H  L)  –  –  –  –  –  –  –  –  –  Se  (11) Wengniute  (H  L  L)  –  –  –  –  –  –  –  –  –  U  (12) Aohan  (L  L  H)  –  –  –  –  –  –  –  –  –  U  b. Eighteen document analysis cases                            (1) Zhengxiangbai  (L  H  H)  M  H  M  H  M  H  H  H  H  S  (2) Zhenglan  (H  L  L)  M  M  M  M  L  N  M  L  L  U  (3) Siziwang  (H  H  H)  H  H  H  H  M  H  H  M  M  S  (4) Wulatehou  (L  H  L)  M  L  M  M  H  M  H  H  L  Se  (5) Ejina  (L  L  L)  M  L  L  L  L  N  M  L  L  U  (6) Wushen  (H  H  H)  H  H  M  H  M  H  H  H  H  S  (7) Guyang  (L  L  L)  N  L  ND  L  L  N  L  L  N  U  (8) Tongliao  (H  H  H)  H  H  H  H  M  H  H  H  H  S  (9) Shangdu  (H  H  L)  M  H  M  M  H  M  H  L  M  Se  (10) Balinyou  (H  H  H)  H  H  H  M  H  H  H  H  H  S  (11) Dalate  (H  H  H)  H  H  H  H  H  H  H  H  H  S  (12) Wuwei  (H  L  H)  H  M  M  L  M  H  H  M  M  Se  (13) Shandan  (L  L  L)  L  N  N  L  L  L  H  H  L  U  (14) Guazhou  (L  H  H)  M  H  H  M  H  H  H  H  H  S  (15) Dunhuang  (H  H  H)  H  H  M  M  H  H  H  M  H  S  (16) Maqu  (L  L  L)  N  L  N  L  L  L  M  ND  L  U  (17) Turpan  (H  H  H)  H  H  H  M  H  M  H  H  H  S  (18) Golmud  (L  H  H)  M  H  M  M  M  H  H  M  M  Se  (19) Yulin  (L  L  H)  L  M  L  M  H  M  H  L  H  Se  (20) Zhangbei  (H  L  H)  H  L  L  M  M  M  H  M  L  Se  (21) Guyuan  (L  L  L)  L  L  M  H  M  H  L  M  L  Se  (22) Fengning  (H  H  L)  H  H  M  H  H  H  H  M  H  S  (23) Daxing  (H  H  H)  H  H  H  H  M  M  H  H  H  S  (24) Kangping  (L  H  L)  L  L  M  L  H  L  H  M  M  Se  (25) Tailai  (L  H  L)  L  H  L  M  M  L  H  H  H  Se  Notes: P1–P9 refer to Principles 1–9; H = high; M = middle; L = low; N = non-satisfaction; ND = no data; S = successful; Se = semi-successful; and U = unsuccessful. Because the nine principles were produced based on the finding from the twelve field study cases, I did not recoded them in the twelve cases here in order to avoid the problem of self-verification. Table 7. Statistical results of the relationship between the nine design principles (independent variables) and the performance of desertification control (dependent variables) in the twenty-five meta-analysis cases in northern China (N = 25) (2011). Independent variables  A. Method 1: K-W test   B. Method 2: Correlation analysis (Spearman)   C. Method 3: Multivariable liner regression (Enter)   D. Method 3: Multivariable liner regression (Backward)   E. Method 3: Multivariable liner regression (Stepwise)   Chi-square  Coefficients  B  Standard Error  t  Sig.  B  Standard error  t  Sig.  B  Standard error  t  Sig.  Constant      −0.613  0.455  −1.347  0.198  −0.915***  0.283  −3.232  0.004  −0.221  0.207  −1.071  0.296  P1  13.036***  0.735*** [6]  0.145  0.097  1.493  0.156  0.152** [5]  0.070  2.172  0.043          P2  15.846***  0.813*** [2]  0.061  0.117  0.518  0.612                  P3  15.463***  0.803***[3]  0.006  0.100  0.061  0.952          0.195**[3]  .081  2.413  0.025  P4  13.045***  0.736*** [5]  0.158  0.116  1.368  0.191  0.244** [2]  0.097  2.515  0.021          P5  13.441***  0.601*** [9]  0.184  0.107  1.712  0.107  0.171* [3]  0.096  1.790  0.089          P6  15.351***  0.773*** [4]  0.162*  0.081  1.998  0.064  0.162* [4]  0.079  2.043  0.055  0.270***[2]  .077  3.505  0.002  P7  12.749***  0.626*** [8]  −0.151  0.150  −1.005  0.331                  P8  9.467***  0.628*** [7]  0.113  0.074  1.535  0.146                  P9  16.904***  0.838*** [1]  0.266*  0.136  1.958  0.069  0.286***[1]  0.082  3.498  0.002  0.345***[1]  0.085  4.056  0.001  R        0.964        0.957        0.939      R2        0.930        0.915        0.882      Adjusted R2        0.888        0.893        0.865      Standard error of the estimate        0.26102        0.25467        0.28609      D-W        2.114        2.107        2.112      F(AVOVA)        22.078***        14.10***        52.598***      Independent variables  A. Method 1: K-W test   B. Method 2: Correlation analysis (Spearman)   C. Method 3: Multivariable liner regression (Enter)   D. Method 3: Multivariable liner regression (Backward)   E. Method 3: Multivariable liner regression (Stepwise)   Chi-square  Coefficients  B  Standard Error  t  Sig.  B  Standard error  t  Sig.  B  Standard error  t  Sig.  Constant      −0.613  0.455  −1.347  0.198  −0.915***  0.283  −3.232  0.004  −0.221  0.207  −1.071  0.296  P1  13.036***  0.735*** [6]  0.145  0.097  1.493  0.156  0.152** [5]  0.070  2.172  0.043          P2  15.846***  0.813*** [2]  0.061  0.117  0.518  0.612                  P3  15.463***  0.803***[3]  0.006  0.100  0.061  0.952          0.195**[3]  .081  2.413  0.025  P4  13.045***  0.736*** [5]  0.158  0.116  1.368  0.191  0.244** [2]  0.097  2.515  0.021          P5  13.441***  0.601*** [9]  0.184  0.107  1.712  0.107  0.171* [3]  0.096  1.790  0.089          P6  15.351***  0.773*** [4]  0.162*  0.081  1.998  0.064  0.162* [4]  0.079  2.043  0.055  0.270***[2]  .077  3.505  0.002  P7  12.749***  0.626*** [8]  −0.151  0.150  −1.005  0.331                  P8  9.467***  0.628*** [7]  0.113  0.074  1.535  0.146                  P9  16.904***  0.838*** [1]  0.266*  0.136  1.958  0.069  0.286***[1]  0.082  3.498  0.002  0.345***[1]  0.085  4.056  0.001  R        0.964        0.957        0.939      R2        0.930        0.915        0.882      Adjusted R2        0.888        0.893        0.865      Standard error of the estimate        0.26102        0.25467        0.28609      D-W        2.114        2.107        2.112      F(AVOVA)        22.078***        14.10***        52.598***      Notes: P1–P9 refer to Principles 1–9; *** P < 0.01(two-tailed); ** P < 0.05; * P < 0.1; [1]–[9] refers to the order of coefficients from high to low. 5. Discussion 5.1 The important influence of knowledge on governance performance The results indicated that all three types of knowledge were significantly applied in desertification control as rated by survey respondents, but on average natural science was the most widely applied knowledge among the twelve counties, followed by local knowledge, and finally social science (Table 3). However, interestingly the order of the correlation coefficients of the three types of knowledge with the performance of desertification control from the highest to the lowest was social science, local knowledge, and natural science. Furthermore, the results indicated that the coefficient of natural science was not only the lowest but was also the only one that was not significant. That is, although social science was the least applied knowledge in desertification control, it was the most important knowledge among the three influencing the performance of desertification control. On the other hand, although natural science was the most applied knowledge, it was the least important knowledge among the three that influenced performance. In particular, the negative regression coefficient of natural science demonstrated that the mere use of natural science would even have a negative impact on governance performance of desertification control, because without the support of local knowledge and social science or the collaboration among the three types of knowledge, the application of natural science must be blind, and might bring destructive effects to both desertification control and people’s lives. For example, Yang et al. (2010: 159) noted that various natural scientists from universities and research institutes often had negative impact on desertification control. Although they ‘often did not fully understand local condition and also had no concrete and reasonable ideas on how to resolve local problems’, ‘their suggestions were often respected by local officials in their local policy making’. Thus, ‘activities and policies based on their suggestions’ such as ‘planting white poplar to combat desertification’ during the 1980s and 1990s and ‘forcing farmers to grow vegetables in plastic greenhouse’ in the 2000s in Minqin, ‘often deteriorated desertification condition rather than improved land amelioration’. However, the relatively high and significant correlation coefficient and regression coefficient of the average of the three types of knowledge indicated that on average, the three types of knowledge had important and significant influences on the performance of desertification control. Furthermore, it indicated that in order to improve the influence of natural science, it could be combined with social science and local knowledge, because the impacts of natural scientific knowledge on governance performance often can be realized through policy marking, policy implementation, and management, which are the jobs and strength of social science and local knowledge. All of the results showed that knowledge collaboration was very important, and the roles of both social science and local science should be emphasized. These findings were not only consistent with previous literature stressing the importance of local knowledge (Fischer 2000; Yang and Wu 2010; Taylor and de Loë 2012), social science (Hayek 1945; Ruttan 1984), and knowledge collaboration (Thomas and Twyman 2004; Ison et al. 2007; Reynolds et al. 2007; Stringer et al. 2009; Winslow et al. 2011; Holm et al. 2013) but also showed the relative importance of social science compared with local knowledge. However, the fact that social science had the lowest application degree of the three types of knowledge indicated that the importance of social science has not received enough attention. Thus, in the following activities of desertification control and other social affairs, both worldwide researchers and practitioners (including decision makers and implementers) should not only emphasize collaboration between the three types of knowledge but also pay more attention to the development and application of social science, especially in mainland China, where many social sciences are often deemed as useless disciplines and subjects by many people. 5.2 Types of knowledge collaboration and the importance of social science and local knowledge Except for the type of (high natural science, low social science, and high local knowledge), the results indicated that all the other seven types of knowledge collaboration shown in Fig. 2 were found in the twelve field study counties. Furthermore, the analysis of the twenty-five meta-analysis cases showed that all the eight types of knowledge collaboration were found, although some types are more, and some less. All these suggest that the classification of knowledge collaboration shown in Fig. 2 is not only feasible in theory but in line with the actual situation. Thus, the classification can be used not only a useful framework for researchers to studying knowledge collaboration but a simple and convenient tool for practitioners to resolve practical problems. Depending on various traditions, purposes, and practical considerations, researchers and practitioners often divide knowledge into different types. For example, since the time of Sir Isaac Newton, people have also divided knowledge into two types, scientific knowledge (mainly natural science and technology) and unscientific knowledge, and many scientists do not consider knowledge from social science to be scientific knowledge. Thus, when combining social science and local knowledge, the types of knowledge collaboration could be based on the types of knowledge of ‘natural science’ and ‘social and local knowledge’ or ‘science’ and ‘nonscience’ (Guston 2001) and their application levels. Meanwhile, to some scientists, social scientists, other types of researchers, and practitioners, natural science and social science should be put into one group as natural and social science, which is different from local knowledge. In this case, the classification of the types of knowledge application could be based on two types of knowledge (‘natural and social science’ or ‘science’ and ‘local knowledge’) and their application levels as recognized by many social and natural science researchers (e.g. Berkes et al. 2000; Bala and Joseph 2007; Weiss et al. 2013). The classification of the types of knowledge collaboration based on the three types of knowledge (science, social science, and local knowledge) undoubtedly provides a new useful perspective for further theoretical studies on knowledge collaboration and practice. Furthermore, the lowest performance of the type with low levels of all three types of knowledge and the highest performance of the type with high levels of all three types of knowledge once again show the importance of knowledge collaboration. Meanwhile, the low performance of the types with low social science and local knowledge and the high performance of the types with high social science and local knowledge once again emphasize the importance of social science and local knowledge, especially today when people often pay a great deal of attention to natural science rather than to social science and local knowledge (Fazey et al. 2006; Innes and Booher 2010). Meanwhile, these findings suggested that to policy makers and implementers, no matter what their views on the classification of knowledge are, in order to improve governance performance, the types of knowledge collaboration with high social science and local knowledge should be encouraged and get the greatest attention in practice. 5.3 Nine design principles for collaboration of the three types of knowledge The study revealed that collaboration between the three types of knowledge was influenced by the nine factors that were divided into three groups: knowledge itself, social actors (including knowledge possessors and other social actors), and external support. This result was not only consistent with views in the literature about the application of knowledge and relationships between different types of knowledge but also corresponded to the theoretical framework (Fig. 2) of the current study. Furthermore, based on the nine factors, the study proposed nine design principles for successful collaboration among natural science, social science, and local knowledge. Corresponding to the three groups of the nine factors, these nine principles could also be divided into three groups: collaboration among the three types of knowledge, collaboration among knowledge possessors and other social actors, and reliable and sustainable external support. Although many aspects of these principles have been discussed previously in environmental and collaboration literature (Campbell 1992; Fullen and Mitchell 1994; Thomas 1997; Twyman 2004; Ison et al. 2007; Reynolds et al. 2007; Thomas and Tschakert 2007; Stringer et al. 2009; Yang and Wu 2010; Chasek et al. 2011; Reed et al. 2007; Holm et al. 2013), the current study not only reconfirmed the former findings but also found that these findings could be combined to form fundamental design principles for successful collaboration among the three types of knowledge. Furthermore, in addition to collaboration among the three types of knowledge, the nine principles also stressed effective communication and collaboration among knowledge possessors and other social actors. In such collaboration, relying on their knowledge of laws, policies, and management, social scientists could provide suggestions for farmers how to better deal with the relationship between government officials, while seeking more money from the government; while farmers could provide more local knowledge to social scientists to help them better understand the local problems, so as to provide better policy recommendations for the government (Fig. 4a). Natural scientists could help social scientists better understand the problems of desertification control from the perspective of science and technology and provide scientific and technical support for practical governance, meanwhile social scientists could provide law, policy, and management recommendations for the application of science and technology of natural scientists and often played more important roles in law and policy making (Fig. 4b). In particular, in some multi-participant discussion meetings (Fig. 4c and 4d), government officials, farmers, social scientists, and natural scientists could sit together to discuss issues, work out plans, and solve problems. Figure 4. View largeDownload slide Collaboration among farmers, natural scientists, and social scientist. (a) Collaborative discussion between farmers and social scientists in Zhongwei, Ningxia (22 July 2007). (b) Collaborative discussion between natural scientists and social scientists at the Inner Mongolia Grassland Ecosystem Research Station, the CAS, Xilinhot, Inner Mongolia (29 July 2011). (c) Discussion meetings among farmers, natural scientists, social scientists, and government officials in Hongyue village, Jingtai County, Gansu (24 July 2007). (d) Discussion meetings among farmers, natural scientists, social scientists, and government officials in Hongshui Township, Jingtai County, Gansu (14 December 2013). Figure 4. View largeDownload slide Collaboration among farmers, natural scientists, and social scientist. (a) Collaborative discussion between farmers and social scientists in Zhongwei, Ningxia (22 July 2007). (b) Collaborative discussion between natural scientists and social scientists at the Inner Mongolia Grassland Ecosystem Research Station, the CAS, Xilinhot, Inner Mongolia (29 July 2011). (c) Discussion meetings among farmers, natural scientists, social scientists, and government officials in Hongyue village, Jingtai County, Gansu (24 July 2007). (d) Discussion meetings among farmers, natural scientists, social scientists, and government officials in Hongshui Township, Jingtai County, Gansu (14 December 2013). These principles not only provide some useful outlines and references for researchers to further study collaboration between natural science, social science, and local knowledge in public governance but also offer important insights in practice and even a very detailed, operative, and integrative framework for practitioners (including policy makers and implementers) to resolve many practical problems about knowledge collaboration in China and other countries. As concrete guidelines, these principles could be used to design new institutional arrangements, improve existing institutions, transform unsuccessful institutions into successful ones, and diagnose problems for old and new arrangements about knowledge collaboration. However, this does not mean that this integrative framework is a ‘panacea’ (Ostrom 2007) for all the aforementioned problems. The list of design principles in fact is still quite speculative (Ostrom 1990) and more work is needed to further explain the causal mechanisms of each of the principles or to develop and test specific models and hypotheses relating to these principles and their subcomponents. Furthermore, the largest correlation coefficients of Principles 9, 2, and 3 showed that when consider these principles separately, these three were the most important among the nine principles and should be paid more attention. Especially, the extreme importance of financial support was also consistent with many previous findings (e.g. Campbell 1992; Yang and Wu 2010). All these suggested that although all the nine institutional principles are important, practitioners (including policy makers and implementers) should first give more attention to financial support, followed by the application and extension as well as the collaboration of the three types of knowledge. Moreover, the insignificant coefficients of the nine principles and the highest values of R and R2 of the ‘Enter’ regression suggested that although none of the coefficients of the nine principles were significant because of interactive effects among the nine principles, when all the nine principles were considered, the model would have the highest explanatory power. That is, if possible, all the nine principles should be considered together in order to get the highest performance of desertification control. However, if the nine principles could not be considered simultaneously either because of limitations of rationality or cost, the ‘Backward’ regression indicated that when Principles 1, 4, 5, 6, and 9 could be satisfied simultaneously, the model still could get the second highest values of R and R2, because the effects of Principles 2, 3, 7, and 8 on performance had been partly included in the effects of Principles 1, 4, 5, 6, and 9 on performance due to interactive effects between these two groups of principles. If even Principles 1, 4, 5, 6, and 9 could not be satisfied together, the ‘Stepwise’ regression suggested that when Principles 9, 6, and 3 could be satisfied simultaneously, the model still could get relatively high values of R and R2, because the effects of the rest six principles on performance had been partly included in the effects of Principles 9, 6, and 3 on performance due to interactive effects between the three satisfied principles and the rest six principles. Thus, in order to maximize the impact of knowledge collaboration on performance, I suggest practitioners to consider satisfying all the nine principles together as their first choice. If not, I suggest them to deem Principles 1, 4, 5, 6, and 9 as the first subgroup of the nine principles which should be satisfied and consider them from the principles with higher coefficients (e.g. Principles 9 and 4) to the ones with lower coefficients (e.g. Principles 6 and 1), and then also consider the other four principles if possible. When even Principles 1, 4, 5, 6, and 9 could not be satisfied together, they could move to satisfy Principles 9, 6, and 3 as the second subgroup of the nine principles, and then they could also get a relatively high governance performance. If Principles 9, 6, and 3 also could not be satisfied simultaneously, they then could only move to consider the nine principles one by one from the ones with higher correlation with performance to the ones with lower correlation. However, in either case, Principle 9 should be first considered. Certainly, all these should be further studied in the future. Furthermore, at any time, we must bear in mind that the reality is far more complex than the theoretical model. 6. Conclusion Collaboration between science, social science, and local knowledge is a necessary condition of modern collaborative knowledge-driven governance and also a global challenge in the contemporary era of complexity and uncertainty (Coen and Roberts 2012). Although the existing literature stresses the necessity and possibility of the collaboration of the three types of knowledge in public governance, the types and mechanisms of collaboration remained unclear. Based on an empirical study of desertification control in northern China, the current study not only found that knowledge collaboration influenced governance performance, natural science got the most attention but is least essential to success, and the types of collaboration with high levels of social science and local knowledge often led to high governance performance, but also highlighted nine institutional design principles for successful collaboration. In addition to important theoretical and practical principles for both worldwide researchers and practitioners, due to the complexity of modern governance and knowledge application, the findings of the study can also help people understand and resolve first- and second-order dilemmas of knowledge application (Ostrom 1990; Yang 2010) in public governance with a plural and open knowledge system (Popper 1992). That is, through collaboration, the dilemmas and costs of knowledge application could be reduced and the performance of knowledge application and governance could be improved. Together, these findings and reflections shed new light on collaboration between science, social science, and local knowledge not only in desertification control but also in other types of public governance in China and in other countries around the world. Researchers can also attempt to use the types of collaboration listed in Fig. 2 and the nine design principles listed in Table 5 as theoretical references to design their own research to study the collaboration of the three types of knowledge in any type of public governance in any country. However, it should be noted that although the study combined large-sample surveys, interviews, observations, and document analysis, the data analysis was mainly based on survey respondents’ perceptions and qualitative meta-analysis. Thus, further tests should be performed to validate the completeness and consistency of these findings with actual data on desertification and to validate the universal applicability of these findings, especially the statement that the natural science gets the most attention but is least essential to success and the nine design principles for successful collaboration among the three types of knowledge. Acknowledgements The study was supported by State Key Laboratory of Earth Surface Processes and Resource Ecology (2015-KF-14), the National Natural Science Foundation of China (71373016), and the Key Project of National Social Science Fund of China (14ZDB143). The author would like to thank Ms Jiali Yang for her contributions to an earlier version of the article. The author would also like to thank the late Professors Elinor Ostrom and Vincent Ostrom for their comments on and suggestions for this study. In particular, my conversations with Vincent helped me develop the original ideas for this article. Furthermore, special thanks go to the anonymous reviewers and the editors whose suggestions and comments have significantly improved the article. References Abbasi A., Hossain L., Leydesdorff L. ( 2012) ‘ Betweenness Centrality as a Driver of Preferential Attachment in the Evolution of Research Collaboration Networks’, Journal of Informetrics , 6/ 3: 403– 12. Google Scholar CrossRef Search ADS   Agranoff R., McGuire M. ( 2003) Collaborative Public Management: New Strategies for Local Governments . Washington: Georgetown University Press. Aikenhead G., Ogawa M. ( 2007) ‘ Indigenous Knowledge and Science Revisited’, Cultural Studies of Science Education , 2/ 3: 539– 620. Google Scholar CrossRef Search ADS   Albæk E. ( 1995) ‘ Between Knowledge and Power: Utilization of Social Science’, Policy Sciences , 28: 79– 100. Google Scholar CrossRef Search ADS   Ansell C., Gash A. ( 2008) ‘ Collaborative Governance in Theory and Practice’, Journal of Public Administration Research and Theory , 18/ 4: 543– 71. Google Scholar CrossRef Search ADS   Armitage D., Berkes F., Dale A. et al.   ( 2011) ‘ Co-Management and the Co-Production of Knowledge: Learning to Adapt in Canada’s Arctic’, Global Environmental Change , 21: 995– 1004. Google Scholar CrossRef Search ADS   Bala A., Joseph G. G. ( 2007) ‘ Indigenous Knowledge and Western Science: The Possibility of Dialogue’, Race & Class , 49/ 1: 39– 61. Google Scholar CrossRef Search ADS   Bardach E. ( 1987) ‘ From Practitioner Wisdom to Scholarly Knowledge and Back Again’, Journal of Policy Analysis and Management , 7/ 1: 188– 99. Google Scholar CrossRef Search ADS   Behn R. D. ( 1987) ‘ The Nature of Knowledge about Public Management and Teaching from Our Knowledge about Mess and Warfare’, Journal of Policy Analysis and Management , 7/ 1: 200– 12. Google Scholar CrossRef Search ADS   Berkes F. ( 1999) Sacred Ecology: Traditional Ecological Knowledge and Resource Management . London: Taylor and Francis. Berkes F. ( 2009) ‘ Evolution of Co-Management: Role of Knowledge Generation, Bridging Organizations and Social Learning’, Journal of Environmental Management , 90/ 5: 1692– 702. Google Scholar CrossRef Search ADS PubMed  Berkes F., Colding J., Folke C. ( 2000) ‘ Rediscovery of Traditional Ecological Knowledge as Adaptive Management’, Ecological Applications , 10/ 5: 1251– 62. Google Scholar CrossRef Search ADS   Borda-Rodriguez L., Johnson H. ( 2013) ‘ Development on My Terms: Development Consultants and Knowledge for Development’, Public Administration and Development , 33: 343– 56. Google Scholar CrossRef Search ADS   Bunker D. R. ( 1978) ‘ Organizing to Link Social Science with Public Policy Making’, Public Administration Review , 38/ 3: 223– 32. Google Scholar CrossRef Search ADS   Cairl J. G., Gallagher P. R. ( 1968) ‘ Government, Science, and Technology: A Bibliographical Essay’, Public Administration Review , 28/ 4: 373– 81. Google Scholar CrossRef Search ADS   Campbell A. ( 1992) Landcare in Australia: Taking the Long View in Tough Times , Third Annual Report. Canberra: National Soil Conservation Program. Caplan N. ( 1979) ‘ The Two-Communities Theory and Knowledge Utilization’, American Behavioral Scientist , 22/ 3: 459– 70. Google Scholar CrossRef Search ADS   Cash D. H., Borck J. C., Patt A. G. ( 2006) ‘ Countering the Loading-Dock Approach to Linking Science and Decision Making: Comparative Analysis of El Nino/Southern Oscillation (ENSO) Forecasting Systems’, Science, Technology & Human Values , 31/ 4: 465– 94. Google Scholar CrossRef Search ADS   Chasek P., Essahli W., Akhtar-Schuster M. et al.   ( 2011) ‘ Integrated Land Degradation Monitoring and Assessment: Horizontal Knowledge Management at the National and International Levels’, Land Degradation and Development , 22/ 2: 272– 84. Google Scholar CrossRef Search ADS   Chuang T. ( 1968) The Complete Works of Chuang Tzu (translated by Burton Watson).  New York: Columbia University Press. Coen D., Roberts A. ( 2012) ‘ A New Age of Uncertainty’, Governance: An International Journal of Policy, Administration, and Institutions , 25/ 1: 5– 9. Google Scholar CrossRef Search ADS   Corburn J. ( 2007) ‘ Community Knowledge in Environmental Health Science: Co-Producing Policy Expertise’, Environmental Science and Policy , 10/ 2: 150– 61. Google Scholar CrossRef Search ADS   Cox R. H., Béland D. ( 2013) ‘ Valence, Policy Ideas, and the Rise of Sustainability’, Governance: An International Journal of Policy, Administration, and Institutions , 26/ 2: 307– 28. Google Scholar CrossRef Search ADS   Cracknell B. E. ( 2001) ‘ Knowlege is All: Or is It? Some Reflections on Why the Acquisition of Knowledge, Focusing Particularly on Evaluation Activities, Does Not Always Lead To Action’, Public Administration and Development , 21: 371– 9. Google Scholar CrossRef Search ADS   Davis F. D., Marquis C. ( 2005) ‘ Prospects for Organization Theory in the Early Twenty-First Century: Instructional Fields and Mechanisms’, Organization Science , 16/ 4: 332– 43. Google Scholar CrossRef Search ADS   Daviter F. ( 2015) ‘ The Political Use of Knowledge in the Policy Process’, Policy Sciences , 48: 491– 505. Google Scholar CrossRef Search ADS   Deng X. ( 1993) Deng Xiaping Wenxuan . Bejing: Renming Chubanshe. Denny B. C. ( 1967) ‘ Science and Public Policy: A Symposium’, Public Administration Review , 27/ 2: 5– 97. Google Scholar CrossRef Search ADS   Dickison M. ( 2010) ‘ The Asymmetry between Science and Traditional Knowledge’, Journal of the Royal Society of New Zealand , 39/ 4: 171– 2. Google Scholar CrossRef Search ADS   Dietz T., Ostrom E., Stern P. C. ( 2003) ‘ The Struggle to Govern the Commons’, Science , 302/ 5652: 1907– 12. Google Scholar CrossRef Search ADS PubMed  Edelenbos J., van Buuren A., van Schie N. ( 2011) ‘ Co-Producing Knowledge: Joint Knowledge Production between Experts, Bureaucrats and Stakeholders in Dutch Water Management Projects’, Environmental Science & Policy , 14/ 6: 675– 84. Google Scholar CrossRef Search ADS   Emerson K., Nabatchi T., Balogh S. ( 2012) ‘ Integrative Framework for Collaborative Governance’, Journal of Public Administration Research and Theory , 22/ 1: 1– 29. Google Scholar CrossRef Search ADS   Fazey I., Fazey J. A., Salisbury J. G. et al.   ( 2006) ‘ The Nature and Role of Experiential Knowledge for Environmental Conservation’, Environmental Conservation , 33/ 1: 1– 10. Google Scholar CrossRef Search ADS   Fischer F. ( 2000) Citizens, Experts, and the Environment: The Politics of Local Knowledge . Durham and London: Duke University Press. Google Scholar CrossRef Search ADS   Fullen M. A., Mitchell D. J. ( 1994) ‘ Desertification and Reclamation in North-Central China’, Ambio , 23/ 2: 131– 5. Gessler N. ( 2015) ‘ Growing Artificial Societies—Social Science from the Bottom Up’, Artificial Life , 3/ 3: 237– 42. Google Scholar CrossRef Search ADS   Giebels D., van Buuren A., Edelnbos J. ( 2015) ‘ Using Knowledge in a Complex Decision-Making Process-Evidence and Principles from the Danish Houting Project’s Ecosystem-Based Management Approach’, Environmental Science & Policy , 47/ 3: 53– 67. Google Scholar CrossRef Search ADS   Ginger C. ( 2014) ‘ Integrating Knowledge, Interests and Values through Modelling in Participatory Processes: Dimensions of Legitimacy’, Journal of Environmental Planning and Management , 57/ 5: 643– 59. Google Scholar CrossRef Search ADS   Goudie A. S. ( 2009) ‘ Dust Storms: Recent Developments’, Journal of Environmental Management , 90/1: 89– 94. Google Scholar CrossRef Search ADS   Gray B. ( 1985) ‘ Conditions Facilitating Interorganizational Collaboration’, Human Relations , 38/ 10: 911– 36. Google Scholar CrossRef Search ADS   Guston D. H. ( 2001) ‘ Boundary Organizations in Environmental Policy and Science: An Introduction’, Science, Technology, & Human Values , 26/ 4: 399– 408. Google Scholar CrossRef Search ADS   Hayek F. A. ( 1945) ‘ The Use of Knowledge in Society’, The American Economic Review , 35/ 4: 519– 30. Hegger D. L. T., Lamers M., van Zeijl-Rozema A. et al.   ( 2012) ‘ Conceptualising Joint Knowledge Production in Regional Climate Change Adaptation Projects: Success Conditions and Levers for Action’, Environmental Science and Policy , 18/ 18: 52– 65. Google Scholar CrossRef Search ADS   Heink U., Marquard E., Heubach K. et al.   ( 2015) ‘ Conceptualizing Credibility, Relevance and Legitimacy for Evaluating the Effectiveness of Science-Policy Interfaces: Challenges and Opportunities’, Science and Public Policy , 42/ 5: 676– 89. Google Scholar CrossRef Search ADS   Hennemann S., Rybski D., Liefner I. ( 2012) ‘ The Myth of Global Science Collaboration—Collaboration Patterns in Epistemic Communities’, Journal of Informetrics , 6/ 2: 217– 25. Google Scholar CrossRef Search ADS   Hill R. ( 2006) ‘ The Effectiveness of Agreements and Protocols to Bridge between Indigenous and Non-Indigenous Toolboxes for Protected Area Management: A Case Study from the Wet Tropics of Queensland’, Society and Natural Resources , 19/ 7: 577– 90. Google Scholar CrossRef Search ADS   Hill R., Grant C., George M. et al.   ( 2012) ‘ A Typology of Indigenous Engagement in Australian Environmental Management: Implications for Knowledge Integration and Social-Ecological System Sustainability’, Ecology and Society , 17/ 1: 23. Google Scholar CrossRef Search ADS   Holm P., Goodsite M. E., Cloetingh S. et al.   ( 2013) ‘ Collaboration between the Natural, Social and Human Sciences in Global Change Research’, Environmental Science and Policy , 28: 25– 35. Google Scholar CrossRef Search ADS   Honadle B. W. ( 1994) ‘ Making Social Science Research Applicable: The Case of Agricultural Extension Research in Egypt’, Public Administration and Development , 14: 365– 75. Google Scholar CrossRef Search ADS   Hu J. ( 2012). Jiandingbuyi Yanzhe Zhonguo Tese Shehui Zhuyi Daolu Qianjin Wei Quanmian Jian Cheng Xiaokang Shehui Er Fendou . Beijing: People’s Publishing House (Remin Chuban She). Hurwicz L., Reiter S. ( 2008) Designing Economic Mechanisms . New York: Cambridge University Press. Innes J. E., Booher D. E. ( 2010) Planning with complexity: An Introduction To Collaborative Rationality for Public Policy . New York: Routledge. Ison R., Röling N., Watson D. ( 2007) ‘ Challenges to Science and Society in the Sustainable Management and Use of Water: Investigating the Role of Social Learning’, Environmental Science and Policy , 10/ 6: 499– 511. Google Scholar CrossRef Search ADS   Katz J. S., Martin B. R. ( 1997) ‘ What Is Research Collaboration?’, Research Policy , 26/ 1: 1– 18. Google Scholar CrossRef Search ADS   Ko K. ( 2013) ‘ Knowledge Accumulation in Asian Public Administration Research: A Critical Review’, Public Administration and Development , 33: 320– 4. Google Scholar CrossRef Search ADS   Lambright W. H. ( 2008) ‘ Government and Science: A Troubled, Critical Relationship and What Can Be Done about It’, Public Administration Review , 68: 5– 18. Google Scholar CrossRef Search ADS   Landry R., Amara N., Lamari M. ( 2001) ‘ Utilization of Social Science Research Knowledge in Canada’, Research Policy , 30/ 2: 333– 49. Google Scholar CrossRef Search ADS   Landry R., Lamari M., Amara N. ( 2003) ‘ The Extent and Determinants of the Utilization of University Research in Government Agencies’, Public Administration Review , 63/ 2: 192– 205. Google Scholar CrossRef Search ADS   Lasswell H. D. ( 1956) ‘ The Political Science of Science: An Inquiry into the Possible Reconciliation of Mastery and Freedom’, American Political Science Review , 50/ 4: 961– 79. Google Scholar CrossRef Search ADS   Leach W. D., Pelkey N. W., Sabatier P. A. ( 2002) ‘ Stakeholder Partnerships As Collaborative Policymaking: Evaluation Criteria Applied to Watershed Management in California and Washington’, Journal of Policy Analysis and Management , 21/ 4: 645– 70. Google Scholar CrossRef Search ADS   Lubell M. ( 2004) ‘ Collaborative Environmental Institutions: All Talk and No Action?’, Journal of Policy Analysis and Management , 23: 549– 73. Google Scholar CrossRef Search ADS   Maclean K., Cullen L. ( 2010) ‘ Research Methodologies for the Co-Production of Knowledge for Environmental Management in Australia’, Journal of the Royal Society of New Zealand , 39/ 4: 205– 8. Google Scholar CrossRef Search ADS   McGuire M. ( 2006) ‘ Collaborative Public Management: Assessing What We Know and How We Know It’, Public Administration Review , 66: 33– 43. Google Scholar CrossRef Search ADS   Meyer D. R., Cancian M., Nam K. ( 2007) ‘ Welfare and Child Support Program Knowledge Gaps Reduce Program Effectiveness’, Journal of Policy Analysis and Management , 26/ 3: 575– 97. Google Scholar CrossRef Search ADS   Miller C. ( 2001) ‘ Hybrid Management: Boundary Organizations, Science Policy, and Environmental Governance in the Climate Regime’, Science, Technology, and Human Values , 26: 478– 500. Google Scholar CrossRef Search ADS   Moore E. A., Koontz T. M. ( 2003) ‘ A Typology of Collaborative Watershed Groups: Citizen-Based, Agency-Based, and Mixed Partnerships’, Society & Natural Resources , 16/ 5: 451– 60. Google Scholar CrossRef Search ADS   Moynihan D. P., Fernandez S., Kim S. et al.   ( 2010) ‘ Performance Regimes amidst Governance Complexity’, Journal of Public Administration Research and Theory , 21 (Suppl 1): i141– 55. Newman M. E. J. ( 2001) ‘ The Structure of Scientific Collaboration Networks’, Proceedings of the National Academy of Sciences , 98/ 2: 404– 9. Google Scholar CrossRef Search ADS   O’Leary R., Catherine G., Lisa B. B. ( 2006) ‘ Introduction to the Symposium on Collaborative Public Management, Special Issue’, Public Administration Review , 66/ s1: 6– 9. Google Scholar CrossRef Search ADS   Olmos-Peñuela J., Castro-Martínez E., D’Este P. ( 2014) ‘ Knowledge Transfer Activities in Social Sciences and Humanities: Explaining the Interactions of Research Groups with Non-Academic Agents’, Research Policy , 43: 696– 706. Google Scholar CrossRef Search ADS   Osborne S. P. ( 2006) ‘ The New Public Governance?’, Public Management Review , 8/ 3: 377– 87. Google Scholar CrossRef Search ADS   Ostrom E. ( 1990) Governing the Commons: The Evolution of Institutions for Collective Action . Cambridge, NY: Cambridge University Press. Google Scholar CrossRef Search ADS   Ostrom E. ( 2005) Understanding Institutional Diversity . Princeton, NJ: Princeton University Press. Ostrom E. ( 2007) ‘ A Diagnostic Approach for Going Beyond Panaceas’, Proceedings of the National Academy of Sciences , 104/ 39: 15181– 7. Google Scholar CrossRef Search ADS   O'Toole L. J., Meier K. J. ( 1999) ‘ Modeling the Impact of Public Management: The Implications of Structural Context’, Journal of Public Administration Research and Theory , 9/ 4: 505– 26. Google Scholar CrossRef Search ADS   Papageorgiou J. C. ( 1994) ‘ The Role of the Public Sector in Introducing Operations Research/Management Science within Developing Countries’, Public Administration and Development , 14: 293– 312. Google Scholar CrossRef Search ADS   Patton M. Q. ( 1987) How to Use Qualitative Methods in Evaluation . Newbury Park, CA: Sage. Perry J. L. ( 2012) ‘ How Can We Improve Our Science to Generate More Usable Knowledge for Public Professionals?’, Public Administration Review , 72/ 4: 479– 82. Google Scholar CrossRef Search ADS   Podsakoff J. L., Organ D. W. ( 1986) ‘ Self-Reports in Organization Research: Problems and Prospects’, Journal of Management , 12/ 4: 531– 44. Google Scholar CrossRef Search ADS   Pohl C. ( 2008) ‘ From Science to Policy through Transdisciplinary Research’, Environmental Science and Policy , 11/ 1: 46– 53. Google Scholar CrossRef Search ADS   Popper K. R. ( 1992) In Search of A Better World: Lectures and Essays from Thirty Years . London and New York: Routledge. Poteete A. R., Janssen M. A., Ostrom E. ( 2010) Working Together: Collective Action, the Commons, and Multiple Methods in Practice . Princeton, NJ: Princeton University Press. Poteete A. R., Ostrom E. ( 2008) ‘ Fifteen Years of Empirical Research on Collective Action in Natural Resource Management: Struggling to Build Large-N Databases Based on Qualitative Research’, World Development , 36/ 1: 176– 95. Google Scholar CrossRef Search ADS   Powell W. W., Koput K. W., Smith-Doerr L. ( 1996) ‘ Interorganizational Collaboration and the Cocus of innovation: Networks of Learning in Biotechnology’, Administrative Science Quarterly , 41/ 1: 116– 45. Google Scholar CrossRef Search ADS   Reed M. S., Dougill Y. A. J., Taylor M. J. ( 2007) ‘ Integrating Local and Scientific Knowledge for Adaptation to Land Degradation: Kalahari Rangeland Management Options’, Land Degradation and Development , 18/ 3: 249– 68. Google Scholar CrossRef Search ADS   Reynolds J. F., Smith D. M. S., Lambin E. F. et al.   ( 2007) ‘ Global Desertification: Building A Science For Dryland Development’, Science , 316/ 5826: 847– 51. Google Scholar CrossRef Search ADS PubMed  Rudd M. A. ( 2015) ‘ Scientists’ Framing of the Ocean Science–Policy Interface’, Global Environmental Change , 33/ 6: 44– 60. Google Scholar CrossRef Search ADS   Ruttan V. W. ( 1984) ‘ Social Science Knowledge and Institutional Change’, American Journal of Agricultural Economics , 66/ 5: 549– 59. Google Scholar CrossRef Search ADS   Scott T. ( 2015) ‘ Does Collaboration Make Any Difference? Linking Collaborative Governance to Environmental Outcomes’, Journal of Policy Analysis and Management , 34/ 3: 537– 66. Google Scholar CrossRef Search ADS   [SFA] State Forestry Administration, P. R. China. ( 2011) A Bulletin of Status Quo of Desertification and Sandification in China <http://www.forestry.gov.cn/uploadfile/main/2011-1/file/2011-1-5-59315b03587b4d7793d5d9c3aae7ca86.pdf> accessed 24 July 2014. Smith L. T. ( 1999) Decolonizing Methodologies: Research and Indigenous Peoples . London: Zed Books Ltd. Star S. L., Griesemer J. R. ( 1989) ‘ Institutional Ecology, “Translations” and Boundary Objects: Amateurs and Professionals in Berkeley’s Museum of Vertebrate Zoology, 1907–39’, Social Studies of Science , 19/ 3: 387– 420. Google Scholar CrossRef Search ADS   Stone D. ( 2009) ‘ Rapid Knowledge: ‘Bringing Research and Policy’ at the Overseas Development Institute’, Public Administration and Development , 29: 303– 15. Google Scholar CrossRef Search ADS   Stringer L. C., Dyer J. C., Reed M. S. et al.   ( 2009) ‘ Adaptations to Climate Change, Drought and Desertification; Insights to Enhance Policy in Southern Africa’, Environmental Science and Policy , 12: 748– 65. Google Scholar CrossRef Search ADS   Sun Z. ( 2011). Jianguo Fanglue . Beijing: Zhongguo Changan Chubanshe. Sun D., Dawson R., Li B. ( 2006) ‘ Agricultural Causes of Desertification Risk in Minqin, China’, Journal of Environmental Management , 79: 348– 56. Google Scholar PubMed  Taylor B., de Loë R. C. ( 2012) ‘ Conceptualizations of Local Knowledge in Collaborative Environmental Governance’, Geoforum , 43/ 6: 1207– 17. Google Scholar CrossRef Search ADS   Thomas D. S. G. ( 1997) ‘ Science and the Desertification Debate’, Journal of Arid Environments , 37/ 4: 599– 608. Google Scholar CrossRef Search ADS   Thomas D. S. G., Twyman C. ( 2004) ‘ Good or Bad Range Land? Hybrid Knowledge, Science and Local Understandings of Vegetation Dynamics in the Kalahari’, Land Degradation and Development , 15/ 3: 215– 31. Google Scholar CrossRef Search ADS   Torgerson D. ( 1986) ‘ Between knowledge and politics: Three faces of policy analysis’, Policy Sciences , 19: 33– 59. Google Scholar CrossRef Search ADS   Tschakert P. ( 2007) ‘ Views from the Vulnerable: Understanding Climatic and Other Stressors in the Sahel’, Global Environmental Change , 17/ 3-4: 381– 96. Google Scholar CrossRef Search ADS   [ UN] United Nations. ( 1992) Managing Fragile Ecosystems: Combating Desertification and Drought , Chapter 12 of Agenda 2. New York, NY: United Nations. Van der Meulen B. ( 1998) ‘ Science Policies as Principal–Agent Games Institutionalization and Path Dependency in the Relation between Government and Science’, Research Policy , 27: 397– 414. Google Scholar CrossRef Search ADS   Wagner C. S., Leydesdorff L. ( 2005) ‘ Network Structure, Self-Organization, and the Growth of International Collaboration in Science’, Research Policy , 34/ 10: 1608– 18. Google Scholar CrossRef Search ADS   Wang T., Wu W., Xue X. et al.   ( 2004) ‘ Spatial-Temporal Changes of Sandy Desertified Land During Last 5 Decades in Northern China’, Acta Geographica Sinica , 59/ 2: 203– 12. Watson A. ( 2013) ‘ Misunderstanding the “Nature” of Co-Management: A Geography of Regulatory Science and Indigenous Knowledge (IK)’, Environmental Management , 52/ 5: 1085– 102. Google Scholar CrossRef Search ADS PubMed  Weiss J. A., Gruber J. E. ( 1984) ‘ Using Knowledge for Control in Fragmented Policy Arenas’, Journal of Policy Analysis and Management , 3/ 2: 225– 47. Google Scholar CrossRef Search ADS   Weiss K., Hamann M., Marsh H. ( 2013) ‘ Bridging Knowledges: Understanding and Applying Indigenous and Western Scientific Knowledge for Marine Wildlife Management’, Society & Natural Resources , 26/ 3: 285– 302. Google Scholar CrossRef Search ADS   Wellens J. ( 1975) ‘ Levels of Participation’, Industrial and Commercial Training , 73/ 3: 93– 6. Google Scholar CrossRef Search ADS   Winslow M. D., Vogt J. V., Thomas R. J. et al.   ( 2011) ‘ Editorial. Science for Improving the Monitoring and Assessment of Dryland Degradation’, Land Degradation & Development , 22/ 2: 145– 9. Google Scholar CrossRef Search ADS   World Bank ( 1998) World Development Report: Knowledge for Development . New York: Oxford University Press. Wu J. ( 2001). ‘Desertification’, in Robinson R (ed.) Plant Sciences for Students , New York: Macmillan Reference. Yang L. ( 2010) ‘ The Knowledge Dilemma and Its Resolution Methods: Game Theoretical Models of Environmental Governance’, Chinese Public Administration , 10/ 10: 112– 6. Yang J. ( 2013) ‘ Roles and Mechanisms of Multiple Knowledge Collaboration in Institutional Changes of Desertification Control’ , Thesis, Beihang University, Beijing. Yang L. ( 2015) ‘ Local Knowledge, Science, and Institutional Change: The Case of Desertification Control in Northern China’, Environmental Management , 55/ 3: 616– 33. Google Scholar CrossRef Search ADS PubMed  Yang L., Lan Z., Wu J. ( 2010) ‘ Roles of Scholars in the Practice of Combating-Desertification: A Case Study in Northwest China’, Environmental Management , 46/ 2: 154– 66. Google Scholar CrossRef Search ADS PubMed  Yang L., Li C. ( 2015) ‘ Types and Mechanisms of Science-Driven Institutional Change: The case of Desertification Control in Northern China’, Environmental Policy and Governance , 25/ 1: 16– 35. Google Scholar CrossRef Search ADS   Yang L., Wu J. ( 2010) ‘ Seven Design Principles for Promoting Scholars’ Participation In Combating Desertification’, International Journal of Sustainable Development and World Ecology , 17/ 2: 109– 19. Google Scholar CrossRef Search ADS   Yang L., Wu J., Shen P. ( 2013) ‘ Roles of Science in Institutional Changes: The Case of Desertification Control in China’, Environmental Science and Policy , 27: 32– 54. Google Scholar CrossRef Search ADS   Zhao B. ( 2009). Zhengzhixue He Hexie Shehui . Beijing: Beijing University Press. Zurba M., Ross H., Izurieta A. et al.   ( 2012) ‘ Building Co-Management as a Process: Problem Solving through Partnerships in Aboriginal Country, Australia’, Environmental Management , 49/ 6: 1130– 42. Google Scholar CrossRef Search ADS PubMed  Appendix A Perceived factors influencing the application of the three types of knowledge as rated by survey respondents (2011). CountiesNine groups of factors and performance  Linze(%)  Minqin(%)  Zhongwei(%)  Yanchi(%)  Dengkou(%)  Ejin Horo(%)  Xinbaerhuzuo(%)  Xilinhot(%)  Naiman(%)  Duolun(%)  Wengniute(%)  Aohan(%)  Factor 1: Level of knowledge (Average)  36.2a  32.3  38.5  21.9  32.5  24.3  38.8  33.4  11.3  26.1  18.3  32.3  Low levels of science and technology development  31.3b  29.7  37.1  21.4  31.4  22.7  32.1  32.2  12.8  22.8  23.2  26.3  Low levels of social science development  37.5  33.9  40.0  21.9  36.2  26.1  38.8  36.1  10.6  29.7  12.6  35.5  Low values of local knowledge  39.9  33.2  38.5  22.4  30.0  24.1  45.6  31.9  10.6  25.9  19.0  35.3  Factor 2: Knowledge application  37.7  45.4  41.3  25.1  42.2  32.0  40.7  39.2  11  34.7  21.2  38.3  Lack of science suited to local conditions  33.3  48.9  41.0  27.2  44.9  29.3  43.4  41.8  13.1  31.7  19.1  39.7  Lack of effective systems of science and technology transformation, extension and application  36.6  42.2  39.6  22.1  44.4  34.8  41.7  39.1  9.4  38.0  30.0  35.2  Lack of social science suited to local conditions and problems  35.4  49.7  43.8  26.5  37.7  32.9  37.5  37.1  12.4  34.3  19.4  41.0  Lack of effective systems of social science application  45.3  43.2  41.7  24.5  46.6  32.9  44.2  40.1  12.2  37.9  21.7  36.4  Unsuitability of Local knowledge to local conditions  30.4  44.5  38.6  26.2  32.9  31.9  39.4  34.8  8.7  31.0  17.8  38.2  Lack of effective systems of local knowledge refining application, and extension  45.1  43.8  43.0  24.2  46.8  30.3  37.9  42.1  10.1  35.5  19.6  39.1  Factor 3: Relationships among knowledge types  32.1  40.8  38.8  24.8  34.4  34.5  40.2  32.3  11.4  27.4  18.2  36.3  Conflicts between local knowledge and science  32.1  40.8  38.8  24.8  34.4  34.5  40.2  32.3  11.4  27.4  18.2  36.3  Factor 4: Endeavors of knowledge possessors  34.5  42.7  38.9  23.9  37  30.9  38.4  35.1  16  29.2  22.8  36.4  Scientists’ inadequate knowledge of local conditions  36.9  48.4  42.6  25.8  38.1  42.4  38.8  39.8  14.3  34.8  31.7  37.1  Scientists’ sabotage  25.4  36.2  33.7  24.5  28.3  27.7  34.1  24.5  12.2  23.9  19.1  29.1  Social scientists’ inadequate knowledge of local conditions  41.7  51.0  44.0  25.8  44.5  31.9  35.2  39.2  34.6  31.6  22.5  41.2  Social scientists’ sabotage  33.3  44.5  33.9  24.4  35.7  27.7  37.2  30.5  12.4  24.6  15.9  38.2  Low capability of local farmer and herdsmen for applying local knowledge  39.4  38.5  41.8  22.8  39.3  27.5  36.7  40.3  12.8  35.3  27.2  38.1  The sabotage of local people having local knowledge  30.5  37.5  37.3  20.3  36.1  28.4  48.6  36.5  9.6  24.8  20.3  34.8  Factor 5: Relationship among knowledge possessors  32.9  47.8  37.9  26.4  36.9  34.5  40.7  33.9  16.9  30.8  19.2  37.4  Lack of effective communication between scientists and local people, especially farmers and herdsmen  30.9  50.0  42.8  25.9  39.3  34.0  46.7  32.0  40.0  30.0  17.3  35.9  Lack of respect by government and scientists for local people  28.5  42.7  35.3  26.1  28.1  34.3  35.1  28.2  12.8  25.9  20.3  34.4  Lack of effective communication between social scientists and local people, especially farmers and herdsmen  40.4  54.8  40.4  26.4  42.5  35.5  44.4  40.9  14.5  33.2  21.1  39.8  Lack of respect by government and social scientists for local people  32.7  44.9  36.4  25.6  33.9  32.9  34.7  29.4  10.6  26.8  16.3  36.1  Lack of effective communication between other scientific workers and local people, especially farmers and herdsmen  37.0  48.5  40.5  27.3  42.5  35.2  44.7  41.2  11.7  38.1  22.3  39.3  Lack of respect by government and other scientific workers for local people  27.9  45.8  32.2  27.1  35.2  35.2  38.8  31.7  11.9  30.9  17.9  38.8  Factor 6: Support from other social actors  38.1  40.2  38.3  22.6  36.9  31.9  40.1  34.5  12.1  29.2  18.9  34.8  Inadequate attention of society to the function of science and technology in desertification control  36.8  41.6  47.7  21.6  39.7  34.0  35.6  35.9  9.4  30.6  22.6  36.7  Low quality of farmers and pastoralists in applying science  41.6  41.9  40.0  18.8  39.6  33.3  41.8  37.5  12.3  32.0  22.9  33.8  Local people’s noncooperation in applying science  32.3  29.7  33.3  22.0  23.9  30.0  44.2  26.2  14.0  21.1  13.3  31.7  Inadequate attention of society to the development and application of social science  48.0  47.5  38.1  19.5  44.1  34.5  41.7  37.2  12.1  35.0  16.2  36.9  Low quality of farmers and pastoralists in applying social science  37.1  38.4  36.5  24.7  38.2  34.5  47.5  40.9  11.9  32.1  21.9  33.1  Local people’s noncooperation in applying social science  30.1  39.1  35.5  27.0  30.8  30.3  38.5  28.1  11.4  21.5  18.3  36.9  Inadequate attention of society to the function and application of local knowledge  43.3  47.1  41.0  19.4  45.5  29.6  29.6  39.4  13.4  33.9  18.7  33.5  Local people’s noncooperation in applying local knowledge  35.8  35.9  34.6  27.9  33.2  28.9  42.1  30.6  11.9  27.0  17.0  35.5  Factor 7: Local and central government  38  48.8  41  27  43.9  36.8  41  39.3  14.7  32.7  24.3  41.9  Government bureaucracy and corruption in applying science  37.2  56.5  41.0  27.2  46.7  34.0  39.2  46.6  15.9  36.3  31.1  42.2  Inadequate attention of local government to the function of science and technology in desertification control  34.3  47.3  39.9  27.8  41.2  34.0  48.6  34.2  14.3  33.2  25.1  39.2  Inadequate attention of the central government to the function of science and technology in desertification control  33.2  36.5  42.1  30.6  36.7  34.8  35.0  32.0  15.9  25.6  13.2  41.1  Government bureaucracy and corruption in applying social science  37.9  55.9  38.4  26.6  45.2  31.2  33.2  42.4  12.4  33.4  28.9  43.0  Inadequate attention of local government to the function of social science in desertification control  39.1  49.5  37.5  26.3  48.2  40.1  39.8  40.7  15.5  33.9  23.2  39.1  Inadequate attention of the central government to the function of social science in desertification control  38.2  39.6  40.2  25.8  43.2  39.4  39.6  35.0  17.1  27.9  16.3  43.6  Government bureaucracy and corruption in applying local knowledge  38.8  57.8  41.7  31.7  45.0  44.4  45.1  42.0  12.7  35.9  36.3  43.5  Inadequate attention of local government to the function of local knowledge in desertification control  40.2  52.1  42.6  24.1  47.1  38.7  44.7  42.7  12.8  35.7  24.0  40.1  Inadequate attention of the central government to the function of local knowledge in desertification control  43.0  44.0  45.2  23.3  42.1  34.5  43.7  37.7  16.1  32.1  20.4  45.3  Factor 8: Laws and regulations  42  42.1  41.2  21.4  42.4  33.9  36.5  38.7  15  36.8  21.1  36.4  Imperfect laws and regulations of science and technology application  38.4  36.5  42.4  21.9  38.4  30.5  35.5  37.6  17.8  35.7  23.3  34.8  Imperfect laws and regulations of social science application  42.5  45.1  39.9  21.3  43.6  34.5  38.5  38.5  14.9  39.6  19.7  37.2  Imperfect laws and regulations of local knowledge application  45.2  44.7  41.2  21.1  45.2  36.6  35.4  39.9  12.2  35.2  20.3  37.2  Factor 9: Financial support  41.7  43.3  39.1  21.8  43.9  30.2  39.3  37.9  12  33.4  24.6  38.1  Low financial support for science and technology application  42.2  43.2  34.3  16.2  44.9  24.1  35.6  33.5  13.6  29.0  23.4  33.2  Low financial support for social science application  40.7  42.1  38.3  25.0  45.5  34.0  40.6  38.3  11.4  34.1  25.7  40.8  Low financial support for local knowledge application  42.1  44.7  44.8  24.2  41.4  32.4  41.6  41.9  10.9  37.2  24.8  40.3  Performance  46.2  14.3  45.7  29.1  27.7  28  55.2  29.6  20.8  34.3  17.1  18.4  CountiesNine groups of factors and performance  Linze(%)  Minqin(%)  Zhongwei(%)  Yanchi(%)  Dengkou(%)  Ejin Horo(%)  Xinbaerhuzuo(%)  Xilinhot(%)  Naiman(%)  Duolun(%)  Wengniute(%)  Aohan(%)  Factor 1: Level of knowledge (Average)  36.2a  32.3  38.5  21.9  32.5  24.3  38.8  33.4  11.3  26.1  18.3  32.3  Low levels of science and technology development  31.3b  29.7  37.1  21.4  31.4  22.7  32.1  32.2  12.8  22.8  23.2  26.3  Low levels of social science development  37.5  33.9  40.0  21.9  36.2  26.1  38.8  36.1  10.6  29.7  12.6  35.5  Low values of local knowledge  39.9  33.2  38.5  22.4  30.0  24.1  45.6  31.9  10.6  25.9  19.0  35.3  Factor 2: Knowledge application  37.7  45.4  41.3  25.1  42.2  32.0  40.7  39.2  11  34.7  21.2  38.3  Lack of science suited to local conditions  33.3  48.9  41.0  27.2  44.9  29.3  43.4  41.8  13.1  31.7  19.1  39.7  Lack of effective systems of science and technology transformation, extension and application  36.6  42.2  39.6  22.1  44.4  34.8  41.7  39.1  9.4  38.0  30.0  35.2  Lack of social science suited to local conditions and problems  35.4  49.7  43.8  26.5  37.7  32.9  37.5  37.1  12.4  34.3  19.4  41.0  Lack of effective systems of social science application  45.3  43.2  41.7  24.5  46.6  32.9  44.2  40.1  12.2  37.9  21.7  36.4  Unsuitability of Local knowledge to local conditions  30.4  44.5  38.6  26.2  32.9  31.9  39.4  34.8  8.7  31.0  17.8  38.2  Lack of effective systems of local knowledge refining application, and extension  45.1  43.8  43.0  24.2  46.8  30.3  37.9  42.1  10.1  35.5  19.6  39.1  Factor 3: Relationships among knowledge types  32.1  40.8  38.8  24.8  34.4  34.5  40.2  32.3  11.4  27.4  18.2  36.3  Conflicts between local knowledge and science  32.1  40.8  38.8  24.8  34.4  34.5  40.2  32.3  11.4  27.4  18.2  36.3  Factor 4: Endeavors of knowledge possessors  34.5  42.7  38.9  23.9  37  30.9  38.4  35.1  16  29.2  22.8  36.4  Scientists’ inadequate knowledge of local conditions  36.9  48.4  42.6  25.8  38.1  42.4  38.8  39.8  14.3  34.8  31.7  37.1  Scientists’ sabotage  25.4  36.2  33.7  24.5  28.3  27.7  34.1  24.5  12.2  23.9  19.1  29.1  Social scientists’ inadequate knowledge of local conditions  41.7  51.0  44.0  25.8  44.5  31.9  35.2  39.2  34.6  31.6  22.5  41.2  Social scientists’ sabotage  33.3  44.5  33.9  24.4  35.7  27.7  37.2  30.5  12.4  24.6  15.9  38.2  Low capability of local farmer and herdsmen for applying local knowledge  39.4  38.5  41.8  22.8  39.3  27.5  36.7  40.3  12.8  35.3  27.2  38.1  The sabotage of local people having local knowledge  30.5  37.5  37.3  20.3  36.1  28.4  48.6  36.5  9.6  24.8  20.3  34.8  Factor 5: Relationship among knowledge possessors  32.9  47.8  37.9  26.4  36.9  34.5  40.7  33.9  16.9  30.8  19.2  37.4  Lack of effective communication between scientists and local people, especially farmers and herdsmen  30.9  50.0  42.8  25.9  39.3  34.0  46.7  32.0  40.0  30.0  17.3  35.9  Lack of respect by government and scientists for local people  28.5  42.7  35.3  26.1  28.1  34.3  35.1  28.2  12.8  25.9  20.3  34.4  Lack of effective communication between social scientists and local people, especially farmers and herdsmen  40.4  54.8  40.4  26.4  42.5  35.5  44.4  40.9  14.5  33.2  21.1  39.8  Lack of respect by government and social scientists for local people  32.7  44.9  36.4  25.6  33.9  32.9  34.7  29.4  10.6  26.8  16.3  36.1  Lack of effective communication between other scientific workers and local people, especially farmers and herdsmen  37.0  48.5  40.5  27.3  42.5  35.2  44.7  41.2  11.7  38.1  22.3  39.3  Lack of respect by government and other scientific workers for local people  27.9  45.8  32.2  27.1  35.2  35.2  38.8  31.7  11.9  30.9  17.9  38.8  Factor 6: Support from other social actors  38.1  40.2  38.3  22.6  36.9  31.9  40.1  34.5  12.1  29.2  18.9  34.8  Inadequate attention of society to the function of science and technology in desertification control  36.8  41.6  47.7  21.6  39.7  34.0  35.6  35.9  9.4  30.6  22.6  36.7  Low quality of farmers and pastoralists in applying science  41.6  41.9  40.0  18.8  39.6  33.3  41.8  37.5  12.3  32.0  22.9  33.8  Local people’s noncooperation in applying science  32.3  29.7  33.3  22.0  23.9  30.0  44.2  26.2  14.0  21.1  13.3  31.7  Inadequate attention of society to the development and application of social science  48.0  47.5  38.1  19.5  44.1  34.5  41.7  37.2  12.1  35.0  16.2  36.9  Low quality of farmers and pastoralists in applying social science  37.1  38.4  36.5  24.7  38.2  34.5  47.5  40.9  11.9  32.1  21.9  33.1  Local people’s noncooperation in applying social science  30.1  39.1  35.5  27.0  30.8  30.3  38.5  28.1  11.4  21.5  18.3  36.9  Inadequate attention of society to the function and application of local knowledge  43.3  47.1  41.0  19.4  45.5  29.6  29.6  39.4  13.4  33.9  18.7  33.5  Local people’s noncooperation in applying local knowledge  35.8  35.9  34.6  27.9  33.2  28.9  42.1  30.6  11.9  27.0  17.0  35.5  Factor 7: Local and central government  38  48.8  41  27  43.9  36.8  41  39.3  14.7  32.7  24.3  41.9  Government bureaucracy and corruption in applying science  37.2  56.5  41.0  27.2  46.7  34.0  39.2  46.6  15.9  36.3  31.1  42.2  Inadequate attention of local government to the function of science and technology in desertification control  34.3  47.3  39.9  27.8  41.2  34.0  48.6  34.2  14.3  33.2  25.1  39.2  Inadequate attention of the central government to the function of science and technology in desertification control  33.2  36.5  42.1  30.6  36.7  34.8  35.0  32.0  15.9  25.6  13.2  41.1  Government bureaucracy and corruption in applying social science  37.9  55.9  38.4  26.6  45.2  31.2  33.2  42.4  12.4  33.4  28.9  43.0  Inadequate attention of local government to the function of social science in desertification control  39.1  49.5  37.5  26.3  48.2  40.1  39.8  40.7  15.5  33.9  23.2  39.1  Inadequate attention of the central government to the function of social science in desertification control  38.2  39.6  40.2  25.8  43.2  39.4  39.6  35.0  17.1  27.9  16.3  43.6  Government bureaucracy and corruption in applying local knowledge  38.8  57.8  41.7  31.7  45.0  44.4  45.1  42.0  12.7  35.9  36.3  43.5  Inadequate attention of local government to the function of local knowledge in desertification control  40.2  52.1  42.6  24.1  47.1  38.7  44.7  42.7  12.8  35.7  24.0  40.1  Inadequate attention of the central government to the function of local knowledge in desertification control  43.0  44.0  45.2  23.3  42.1  34.5  43.7  37.7  16.1  32.1  20.4  45.3  Factor 8: Laws and regulations  42  42.1  41.2  21.4  42.4  33.9  36.5  38.7  15  36.8  21.1  36.4  Imperfect laws and regulations of science and technology application  38.4  36.5  42.4  21.9  38.4  30.5  35.5  37.6  17.8  35.7  23.3  34.8  Imperfect laws and regulations of social science application  42.5  45.1  39.9  21.3  43.6  34.5  38.5  38.5  14.9  39.6  19.7  37.2  Imperfect laws and regulations of local knowledge application  45.2  44.7  41.2  21.1  45.2  36.6  35.4  39.9  12.2  35.2  20.3  37.2  Factor 9: Financial support  41.7  43.3  39.1  21.8  43.9  30.2  39.3  37.9  12  33.4  24.6  38.1  Low financial support for science and technology application  42.2  43.2  34.3  16.2  44.9  24.1  35.6  33.5  13.6  29.0  23.4  33.2  Low financial support for social science application  40.7  42.1  38.3  25.0  45.5  34.0  40.6  38.3  11.4  34.1  25.7  40.8  Low financial support for local knowledge application  42.1  44.7  44.8  24.2  41.4  32.4  41.6  41.9  10.9  37.2  24.8  40.3  Performance  46.2  14.3  45.7  29.1  27.7  28  55.2  29.6  20.8  34.3  17.1  18.4  a The average percentages of subproblems. b The percentages of ‘very large’ and ‘large’ rated by survey respondents. Appendix B Nine factors for successful collaboration between natural science, social science, and local knowledge and the partial coefficients (controlling for values of local knowledge) of the performance of desertification control in the twelve field study cases in northern China (2011). Nine factors  Coefficients with performance  (significance)  Group 1: Knowledge itself    F1. Level of knowledge  −0.170***  The level of developments and improvements in the three types of knowledge  (0.001)  F2. Knowledge application  −0.569***  The refinement, transformation, adaption, application, and extension of the three types of knowledge  (0.000)  F3. Relationships among knowledge types  −0.528***  Complementation and coordination among three types of knowledge  (0.000)  Group 2: knowledge possessors and others social actors    F4. Endeavors of knowledge possessors  −0.696***  Capability and endeavors of knowledge possessors on the application and extension of the three types of knowledge  (0.000)  F5. Relationship between knowledge possessors  −0.634***  Communication and collaboration among knowledge possessors and other social actors  (0.000)  F6. Supports from other social actors  −0.485***  Support from other social actors and continued improvements on their attention and understanding of the three types of knowledge as well as their capabilities  (0.000)  Group 3: External support    F7. Local and central government  −0.726***  Support and guidance by governments at different hierarchical levels  (0.000)  F8. Laws and regulations  −0.355***  Institutional support from laws and regulations  (0.000)  F9. Financial support  −0.561***  Financial support from government and other social actors  (0.000)  Nine factors  Coefficients with performance  (significance)  Group 1: Knowledge itself    F1. Level of knowledge  −0.170***  The level of developments and improvements in the three types of knowledge  (0.001)  F2. Knowledge application  −0.569***  The refinement, transformation, adaption, application, and extension of the three types of knowledge  (0.000)  F3. Relationships among knowledge types  −0.528***  Complementation and coordination among three types of knowledge  (0.000)  Group 2: knowledge possessors and others social actors    F4. Endeavors of knowledge possessors  −0.696***  Capability and endeavors of knowledge possessors on the application and extension of the three types of knowledge  (0.000)  F5. Relationship between knowledge possessors  −0.634***  Communication and collaboration among knowledge possessors and other social actors  (0.000)  F6. Supports from other social actors  −0.485***  Support from other social actors and continued improvements on their attention and understanding of the three types of knowledge as well as their capabilities  (0.000)  Group 3: External support    F7. Local and central government  −0.726***  Support and guidance by governments at different hierarchical levels  (0.000)  F8. Laws and regulations  −0.355***  Institutional support from laws and regulations  (0.000)  F9. Financial support  −0.561***  Financial support from government and other social actors  (0.000)  Note: F1 to F9 refer to Factors 1 to 9; *** P < 0.01 (two-tailed). © The Author 2017. Published by Oxford University Press. All rights reserved. For Permissions, please email: journals.permissions@oup.com http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Science and Public Policy Oxford University Press

Collaborative knowledge-driven governance: Types and mechanisms of collaboration between science, social science, and local knowledge

Loading next page...
 
/lp/ou_press/collaborative-knowledge-driven-governance-types-and-mechanisms-of-01D6iOJC8L
Publisher
Oxford University Press
Copyright
© The Author 2017. Published by Oxford University Press. All rights reserved. For Permissions, please email: journals.permissions@oup.com
ISSN
0302-3427
eISSN
1471-5430
D.O.I.
10.1093/scipol/scx047
Publisher site
See Article on Publisher Site

Abstract

Abstract Knowledge plays an important role in modern public governance characterized by complexity, but collaboration between different types of knowledge in public governance has not been systematically studied. Increasingly, literature has stressed the importance of the application of science, social science, and local knowledge in public governance, whereas it has paid little attention to the types and mechanisms of collaboration between these three fields. The aims of this study were to explore the influence of collaboration between the three types of knowledge on governance performance, the major types of collaboration, and the major institutional design principles for successful collaboration. Based on a combined field study including surveys, interviews, observations, and archive data as well as a meta-analysis study on desertification control in northern China, the largest developing country in the world, this study made the following three key findings: (1) Although natural science was the most widely applied area of knowledge and social science was least applied, the order of the correlation coefficients of the three types of knowledge with governance performance from the highest to the lowest was social science, local knowledge, and natural science. (2) Collaboration between these three types of knowledge influenced governance performance. The types of collaboration with low levels of all three types of knowledge always had low governance performance, and the types of collaboration with high levels of social science and local knowledge often had high performance. (3) Successful collaboration among different types of knowledge shared nine significant institutional design principles. These principles stressed the integration of three types of knowledge, the collaboration among knowledge possessors and other social actors, and reliable and sustainable external support (government, financial, and institutional). These findings shed new light on collaboration between science, social science, and local knowledge in public and environmental governance in China as well as in other countries around the world. 1. Introduction: conflict or collaboration? Although there is an old and protracted debate about whether knowledge is good or bad for human beings (e.g. Chuang 1968), knowledge always plays an important role in our society (Hayek 1945; Gessler 2015) and its public governance (Ostrom 2005; Osborne 2006), especially in modern public governance characterized by complexity (Moynihan et al. 2010). Since Sir Isaac Newton, positivist ways of knowing have dominated modern society (Fazey et al. 2006; Innes and Booher 2010), and people often deem scientific knowledge (natural science and technology) and local knowledge as distinct, and even opposed and conflicting, types of knowledge (Thomas and Twyman 2004) in public governance. Thus, local knowledge and sometimes even many social sciences have been ignored in discourses on public governance. For example, before the Reform and Opening Up Policy in 1978, and especially during the Great Cultural Revolution, local knowledge was often regarded as backward, primitive, ignorant, and even stupid, and many social sciences (e.g. political science, public administration, and sociology) were deemed as useless and bourgeois disciplines and subjects in Mainland China (Zhao 2009). On the contrary, an increasing number of studies have reemphasized the importance of social sciences (e.g. Hayek 1945; Ruttan 1984; Landry et al. 2001; Olmos-Peñuela et al. 2014) and local or traditional knowledge (e.g. Berkes et al. 2000; Fischer 2000; Taylor and de Loë 2012) in various types of public governance and have emphasized the importance of hybrid knowledge (Thomas and Twyman 2004). In addition to the studies on research collaboration (e.g. Katz and Martin 1997; Abbasi et al. 2012), scientists’ or scientific collaboration (e.g. Newman 2001), organizational collaboration (e.g. Powell et al. 1996), and even scientific collaboration between countries (e.g. Wagner and Leydesdorff 2005; Hennemann et al. 2012), rich literature originating from the field of science and technology studies and other related fields explored also emphasized the importance of collaboration between different types of knowledge, in particular, through analyzing co-management or knowledge partnership (Berkes 2009; Zurba et al. 2012; Watson 2013), the co-production of knowledge (Maclean and Cullen 2010; Armitage et al. 2011), co-producing knowledge or joint knowledge production (Edelenbos et al. 2011; Hegger et al. 2012), interactive knowledge production (Pohl 2008; Giebels et al. 2015), knowledge integration (Hill et al. 2012; Ginger 2014), boundary organizations (Guston 2001; Miller 2001; Cash et al. 2006), boundary objects (Star and Griesemer 1989), science–policy interfaces (Heink et al. 2015; Rudd 2015), and bridging tools (Hill 2006) based on the studies mainly done in developed countries such as America, Australia, Canada, and The Netherlands. Within the public administration literature, many studies since the 1960s have emphasized the important roles of knowledge, science, and social sciences in public policy-making, implementation, and other types of public administration. For example, some studied the role of knowledge (e.g. Landry et al. 2003; Meyer et al. 2007; Daviter 2015) in public administration and policy; some focused on the role of science (e.g. Lasswell 1956; Yang et al. 2013); and some explored the role of social sciences (e.g. Bunker 1978; Caplan 1979; Landry et al. 2001) or knowledge about public administration (Behn 1987), local knowledge (e.g. Yang 2015), and even collaboration of social science knowledge and local knowledge or practitioner wisdom (Bardach 1987). Still others even studied the relationships between knowledge (or science and social science) and politics (Torgerson 1986), power (e.g. Albæk 1995), government (e.g. Cairl and Gallagher 1968; Van der Meulen 1998; Lambright 2008), or public policy (e.g. Denny 1967; Bunker 1978) as well as knowledge for development (Borda-Rodriguez and Johnson 2013) and knowledge accumulation (Ko 2013). Furthermore, some studies in the field of environmental governance also have emphasized the importance of collaboration between different types of knowledge in various types of public governance (e.g. Ison et al. 2007; Stringer et al. 2009; Holm et al. 2013; Yang 2015). However, collaboration between different types of knowledge in both public administration and environmental governance has not been systematically studied, especially in many developing countries, such as China (e.g. Yang and Wu 2010; Yang et al. 2013). Thus, in the current study, I examined how collaboration between the three types of knowledge influences the performance of public governance in China. In particular, I explored types of collaboration with the three types of knowledge and the mechanism design (Hurwicz and Reiter 2008) or institutional design principles (Ostrom 1990, 2005) of successful knowledge collaboration in public governance. This investigation can help us not only expand upon the work of many scholars and researchers who have explored the application of knowledge in various types of public governance and the dilemma of knowledge application (Yang 2010) but also understand the influences of knowledge application on public governance performance by analyzing the types and institutional design principles as key mechanisms of knowledge collaboration. Furthermore, if the dilemma caused by the gaps and conflicts between knowledge and practice can be called the first order dilemma of knowledge application (Yang 2010), the dilemma caused by the gaps and conflicts between different types of knowledge can be called the second-order dilemma of knowledge application. Thus, this study can contribute to understanding and resolving both first-order and second-order dilemmas of knowledge application in public governance (Ostrom 1990; Yang 2010), which are linked to the debate about whether knowledge is good or bad for humans, especially in some developing countries (such as China), which are facing multiple and arduous challenges of modernization through the use of modern thoughts and knowledge (Deng 1993; Sun 2011; Hu 2012). World Bank (1998: 3, 6) also pointed out ever: ‘Most of the difficulties that developing countries face involve both knowledge gaps and information problems.’ ‘Development institutions have three roles in reducing knowledge gaps: to provide international public goods, to act as intermediaries in the transfer of knowledge, and to manage the rapidly growing body of knowledge about development.’ Desertification, one of the greatest environmental challenges in our time, refers to land degradation in arid, semi-arid, and dry subhumid areas caused by climate change and human activities (UN 1992; Wu 2001). Desertification often leads to the permanent loss of land productivity, serious dust storms, and various ecological, environmental, economical, and even political consequences (Goudie 2009). Desertified areas account for approximately 41 per cent of the world’s land surface and have influenced the lives of more than 38 per cent of the world’s population (Reynolds et al. 2007). In China, the problem of desertification has also become increasingly serious in the past several decades, and desertified areas have expanded from 137,000 km2 in the 1950s to 385,700 km2 in 2000, with an annual expansion rate of 3,600 km2/year (Wang et al. 2004). Furthermore, different types of knowledge have been applied to combat desertification in China (Yang and Wu 2010). Thus, taking desertification control since 1949, the year of the foundation of new China, as a concrete example of public governance in China can help us not only better understand collaboration between different types of knowledge in public governance but also make a contribution to finally resolving the serious desertification problem in modern China. Based on a collaboration of quantitative and qualitative studies in three adjacent provinces in northern China (Inner Mongolia, Ningxia, and Gansu), I answered the following three research questions: (1) What is the significance of the collaboration between natural science, social science, and local knowledge (or natural scientific, social scientific, and local knowledge) in influencing the performance of desertification control? (2) What are the major types of collaboration between the three types of knowledge? (3) What are the major institutional design principles for successful collaboration between the three types of knowledge in desertification control? Based on the assumption that heterogeneity exists in the levels of three types of knowledge, types, and institutions of knowledge collaboration, and the performance of desertification control, the hypothesis of the study was that the collaboration of the three types of knowledge and their collaboration types and mechanisms have influenced the performance of desertification control. 2. Conceptual background and theory 2.1 Conceptual background In this study, both natural science and social science, or natural and social sciences (Watson 2013), are considered scientific knowledge or science, which had been widely verified using scientific methods and had a high degree of generalizability (Aikenhead and Ogawa 2007; Dickison 2010) (Fig. 1). Concretely, natural science in combating desertification in the study includes agricultural science and technology, agricultural pest control, zoology or animal biology, forestry, knowledge about combating desertification and dust storms, general climatic knowledge, hydraulic engineering, specific knowledge about local desertification, poultry and livestock disease control, etc. (Yang et al. 2013). Social science in desertification control includes political science, public administration, policy analysis, management, economics, law, sociology, general knowledge of environmental governance, knowledge of laws and regulations, understanding local social relations, knowledge of social management, etc. Local knowledge, or indigenous (Smith 1999), traditional (Berkes 1999), and community knowledge (Corburn 2007), is often developed from local experience, rules, wisdom, memories, stories, history, and practices and is only applied or verified in some areas. In this study, local knowledge included all types of ‘culture-specific information, knowledge, skills, norms, taboos, codes of conduct, customs, norms of behavior, conventions, and traditions on desertification control that are based on local experience, wisdom, practices, and histories and are mainly owned by the locals’ (Yang 2015: 617). In order to differentiate the three types of knowledge and make sure people can make distinctions between these categories, natural science and social science were directly measured based on the evaluations of their various concrete branches in the study, while local knowledge was carefully explained when necessary (e.g. in surveys and interviews mentioned later). Figure 1. View largeDownload slide The classification of three types of knowledge. Figure 1. View largeDownload slide The classification of three types of knowledge. Collaboration, as a concept with high valence as defined by Cox and Béland (2013), means two or more parties or stakeholders co-labor or work together (Agranoff and McGuire 2003; O’Leary et al. 2006) to solve the problems which cannot be solved or easily solved by individuals or single organizations (Gray 1985; McGuire 2006), although collaboration does not necessarily lead to increased levels of cooperation (Lubell 2004) and differentiated outcomes (Scott 2015). On the one hand, collaboration between the three types of knowledge means that the three types of knowledge are formally organized (Ansell and Gash 2007) or combined to solve the problems which could not be solved depending on any one of the three and to accomplish the common goal (Katz and Martin 1997) of desertification control. In this stance, ‘collaboration of knowledge’ can be simply changed to ‘combination of knowledge’, and the use of the ‘collaboration’ is different from the usual use of collaboration between multiple actors. On the other hand, collaboration of knowledge also means the collaboration between knowledge possessors because knowledge is often possessed by various social actors, and this use of collaboration is the same as its usual use. That is, in order to avoid the confusion and wordy problem by using two concepts—‘combination of knowledge’ and ‘collaboration of knowledge possessors’, here I use collaboration of knowledge to cover the meaning of both ‘combination of knowledge’ and ‘collaboration of knowledge possessors’. Finally, institutional design principles are essential elements or conditions (Ostrom 1990, 2005) that help to account for the success of knowledge collaboration institutions. 2.2 Theoretical framework In order to study the structural contexts (O'Toole and Meier 1999) of knowledge collaboration, the best method is to develop a taxonomy of knowledge collaboration, because unlike regression, which can only study the relationship between two variables (Davis and Marquis 2005), a taxonomy can not only help researchers investigate the inner structure of knowledge collaboration but also provide a useful foundation to develop theories (Moore and Koontz 2003; Hill et al. 2012). Furthermore, as in the studies on stakeholders’ collaboration in which the level of participation is often used to classify the types of stakeholders’ collaboration (Wellens 1975), the taxonomy of knowledge collaboration can also be classified through analyzing the level of knowledge application in governance. Thus, if we divided the knowledge (natural science, social science, and local knowledge) used in desertification control into high and low levels, then the collaboration types of knowledge could be divided into eight types including High Natural Science, High Social Science, High Local Knowledge and Low Natural Science, Low Social Science, Low Local Knowledge (see Fig. 2). Figure 2. View largeDownload slide Theoretical framework. Figure 2. View largeDownload slide Theoretical framework. Furthermore, a systematical review for the literature over a period of some 20 years found that the application of knowledge and the relationships between different types of applied knowledge were often influenced by nine major factors: (1) levels of knowledge (Thomas 1997; Reynolds et al. 2007; Chasek et al. 2011; Reed et al. 2007; Yang et al. 2013); (2) the application and extension of knowledge (Honadle 1994; Yang and Wu 2010); (3) coordination between different types of knowledge (Thomas and Twyman 2004; Ison et al. 2007; Ansell and Gash 2008; Stringer et al. 2009; Holm et al. 2013); (4) the capabilities and endeavors of knowledge possessors (Yang and Wu 2010; Emerson et al. 2012); (5) communication and collaboration among social actors (Fullen and Mitchell 1994; Papageorgiou 1994; Thomas 1997; Dietz et al. 2003; Tschakert 2007; Stone 2009; Yang and Wu 2010; Borda-Rodriguez and Johnson 2013); (6) other social actors’ support and capabilities (Landry et al. 2003; Yang and Wu 2010; Emerson et al. 2012); (7) government support; (8) legal mechanisms (Hill 2006) and institutional (or cultural) (Cracknell 2001) support; and (9) financial support (Campbell 1992; Yang and Wu 2010; Borda-Rodriguez and Johnson 2013). Previous studies also found that the factors influencing the role of knowledge and science as well as the effectiveness of knowledge collaboration in desertification control could be divided into three aspects: knowledge itself, social actors, and external support (Yang and Li 2015). Thus, in the current study, I integrated the related literature and divided the aforementioned factors into the above three groups and analyzed the mechanisms or nine principles of the collaboration of natural science, social science, and local knowledge. Moreover, I supposed that both the types and the factors shaping the role of knowledge and science influenced governance performance of desertification control. Finally, the theoretical framework of the study was shown in Fig. 2. 3. Research design and methods 3.1 Research design and sites In order to compare knowledge collaboration and their influences on governance performance, a two-step study was conducted. In the first step, the field study was first conducted in twelve typical arid and semi-arid counties from three adjacent provinces with a long history of desertification control and different desertification control results (two counties in Gansu, two in Ningxia, and eight in Inner Mongolia) (Fig. 3). Furthermore, in all of these counties, there were laboratories and field stations of the Chinese Academy of Sciences (CAS) and the Chinese Academy of Forestry. The basic natural characteristics of the twelve counties, such as climate division, total area, and population, are shown in Table 1a. Table 1. Characteristics of the twelve field study counties and 25 meta-analysis cases. Dimensions Counties  Provinces  Total area (km2)  Climate division  Population (10,000)  Annual average temperature (°C)  Annual average precipitation (mm)  Annual average evaporation (mm)  A. Twelve field study casesa                (1) Linze (2001b)  Gansu  3,148  Aridc  15  7.7  115  2212  (2) Minqin (1994)  Gansu  16,016  Arid  27.4  7.8  115  2644  (3) Zhongwei (1995)  Ningxia  5,780  Arid  109.29  9.5  188  1914  (4) Yanchi (2004)  Ningxia  8,661  Arid  15.7  7.7  <300  >2000  (5) Dengkou (1998)  Inner Mongolia  3,554  Arid  9.6  7.6  145  2398  (6) Ejin Horo (2011)  Inner Mongolia  5,600  Semi-arid  15.5  6.7  348  2563  (7) Xinbaerhuzuo (2007)  Inner Mongolia  22,000  Semi-arid  4.2  −0.3  268  1650  (8) Xilinhot (2004)  Inner Mongolia  15,179  Semi-arid  25.2  1.6  250–350  1746  (9) Naiman (2001)  Inner Mongolia  8,159  Semi-arid  41.9  6∼6.5  366  1973–2082  (10) Duolun (2000)  Inner Mongolia  3,773  Semi-arid  10.3  1.9  389  1714  (11) Wengniute (1993)  Inner Mongolia  11,882  Semi-arid  41.6  4.5  370  2106  (12) Aohan (1990)  Inner Mongolia  8,294  Semi-arid  51.3  5–7  310–460  2162  B. Twenty-five meta-analysis cases                (1) Zhengxiangbai (2005)  Inner Mongolia  6,229  Arid  7.3  1.9  268–360  1,932–2,300  (2) Zhenglan (2006)  Inner Mongolia  10,182  Arid  8.3  1.5  365  1,925.5  (3) Siziwang (2005)  Inner Mongolia  25,516  Arid  20.9  1-6  110–350  2,300–2,400  (4) Wulatehou (2001)  Inner Mongolia  25,000  Extremely arid  6.1  3.8  96–105.9  3,000–3,500  (5) Ejina (2006)  Inner Mongolia  114,606  Extremely arid  2  8.6  37  3,841.51  (6) Wushen (2005)  Inner Mongolia  11,645  Arid  10  6.8  350–400  2,200–2,800  (7) Guyang (2002)  Inner Mongolia  5,021  Arid  21  4  300  2,200  (8) Tongliao (2001)  Inner Mongolia  59,535  Arid  309.1  0–6  350–400  2,000  (9) Shangdu (2004)  Inner Mongolia  4,353  Arid  33  3.1  351.5  2,500  (10) Balinyou (2004)  Inner Mongolia  10,256  Semi-arid  18.1  4.9  350  >1,500  (11) Dalate (2004)  Inner Mongolia  8,188  Arid  34  6.1–7.1  240–360  2,600  (12) Wuwei (2000)  Gansu  33,249  Arid  191.8  7.8  60–610  1,400–3,010  (13) Shandan (1998)  Gansu  5,402  Arid  19.2  5.9  195  2,246  (14) Guazhou (2003)  Gansu  23,150  Extremely arid  11.9  8.8  45.7  3,140.6  (15) Dunhuang (2005)  Gansu  31,200  Extremely arid  18.3  9.4  39.9  2,486  (16) Maqu (1995)  Gansu  10,109  Semi-arid  3.71  1.1  616.5  1,353.4  (17) Turpan (2000)  Xinjiang  13,689  Extremely arid  28  13.9  6.9–25.2  3,837  (18) Golmud (1998)  Qinghai  126,220  Extremely arid  12.12  −4.2  41.5  >3,000  2300  (19) Yulin (2005)  Shaanxi  43,578  Arid  335  10.7  290–465    (20) Zhangbei (2001)  Hebei  4,185  Arid  37.2  2.6  300  1,772  (21) Guyuan (2005)  Hebei  3,601  Semi-arid  23  1.4  426  1,787.5  (22) Fengning (1994)  Hebei  8,765  Semi-arid  37.4  3–4  436.7  1,958  (23) Daxing (2006)  Beijing  1,031  Arid  136.5  11.6  556  1,800  (24) Kangping (1995)  Liaoning  2,175  Semi-arid  35  6.9  542.9  2,037.6  (25) Tailai (1998)  Heilongjiang  3,996  Semi-arid  32  4.9  392.6  1,717.1  Dimensions Counties  Provinces  Total area (km2)  Climate division  Population (10,000)  Annual average temperature (°C)  Annual average precipitation (mm)  Annual average evaporation (mm)  A. Twelve field study casesa                (1) Linze (2001b)  Gansu  3,148  Aridc  15  7.7  115  2212  (2) Minqin (1994)  Gansu  16,016  Arid  27.4  7.8  115  2644  (3) Zhongwei (1995)  Ningxia  5,780  Arid  109.29  9.5  188  1914  (4) Yanchi (2004)  Ningxia  8,661  Arid  15.7  7.7  <300  >2000  (5) Dengkou (1998)  Inner Mongolia  3,554  Arid  9.6  7.6  145  2398  (6) Ejin Horo (2011)  Inner Mongolia  5,600  Semi-arid  15.5  6.7  348  2563  (7) Xinbaerhuzuo (2007)  Inner Mongolia  22,000  Semi-arid  4.2  −0.3  268  1650  (8) Xilinhot (2004)  Inner Mongolia  15,179  Semi-arid  25.2  1.6  250–350  1746  (9) Naiman (2001)  Inner Mongolia  8,159  Semi-arid  41.9  6∼6.5  366  1973–2082  (10) Duolun (2000)  Inner Mongolia  3,773  Semi-arid  10.3  1.9  389  1714  (11) Wengniute (1993)  Inner Mongolia  11,882  Semi-arid  41.6  4.5  370  2106  (12) Aohan (1990)  Inner Mongolia  8,294  Semi-arid  51.3  5–7  310–460  2162  B. Twenty-five meta-analysis cases                (1) Zhengxiangbai (2005)  Inner Mongolia  6,229  Arid  7.3  1.9  268–360  1,932–2,300  (2) Zhenglan (2006)  Inner Mongolia  10,182  Arid  8.3  1.5  365  1,925.5  (3) Siziwang (2005)  Inner Mongolia  25,516  Arid  20.9  1-6  110–350  2,300–2,400  (4) Wulatehou (2001)  Inner Mongolia  25,000  Extremely arid  6.1  3.8  96–105.9  3,000–3,500  (5) Ejina (2006)  Inner Mongolia  114,606  Extremely arid  2  8.6  37  3,841.51  (6) Wushen (2005)  Inner Mongolia  11,645  Arid  10  6.8  350–400  2,200–2,800  (7) Guyang (2002)  Inner Mongolia  5,021  Arid  21  4  300  2,200  (8) Tongliao (2001)  Inner Mongolia  59,535  Arid  309.1  0–6  350–400  2,000  (9) Shangdu (2004)  Inner Mongolia  4,353  Arid  33  3.1  351.5  2,500  (10) Balinyou (2004)  Inner Mongolia  10,256  Semi-arid  18.1  4.9  350  >1,500  (11) Dalate (2004)  Inner Mongolia  8,188  Arid  34  6.1–7.1  240–360  2,600  (12) Wuwei (2000)  Gansu  33,249  Arid  191.8  7.8  60–610  1,400–3,010  (13) Shandan (1998)  Gansu  5,402  Arid  19.2  5.9  195  2,246  (14) Guazhou (2003)  Gansu  23,150  Extremely arid  11.9  8.8  45.7  3,140.6  (15) Dunhuang (2005)  Gansu  31,200  Extremely arid  18.3  9.4  39.9  2,486  (16) Maqu (1995)  Gansu  10,109  Semi-arid  3.71  1.1  616.5  1,353.4  (17) Turpan (2000)  Xinjiang  13,689  Extremely arid  28  13.9  6.9–25.2  3,837  (18) Golmud (1998)  Qinghai  126,220  Extremely arid  12.12  −4.2  41.5  >3,000  2300  (19) Yulin (2005)  Shaanxi  43,578  Arid  335  10.7  290–465    (20) Zhangbei (2001)  Hebei  4,185  Arid  37.2  2.6  300  1,772  (21) Guyuan (2005)  Hebei  3,601  Semi-arid  23  1.4  426  1,787.5  (22) Fengning (1994)  Hebei  8,765  Semi-arid  37.4  3–4  436.7  1,958  (23) Daxing (2006)  Beijing  1,031  Arid  136.5  11.6  556  1,800  (24) Kangping (1995)  Liaoning  2,175  Semi-arid  35  6.9  542.9  2,037.6  (25) Tailai (1998)  Heilongjiang  3,996  Semi-arid  32  4.9  392.6  1,717.1  a Adapted from Yang (2015) and Yang et al. (2013). b The year of the source published. c The climate divisions were based on the criteria of United Nations Convention to Combat Desertification. Figure 3. View largeDownload slide The twelve field study counties and twenty-five meta-analysis cases. Source: SFA 2011 and Yang, 2013. Note: The order numbers of the twelve field study counties and the twenty-five meta-analysis cases are consistent with the order shown in Table 1. Figure 3. View largeDownload slide The twelve field study counties and twenty-five meta-analysis cases. Source: SFA 2011 and Yang, 2013. Note: The order numbers of the twelve field study counties and the twenty-five meta-analysis cases are consistent with the order shown in Table 1. In the second step, to test the generalizability of the institutional design principles found in the twelve counties by the field study, I also studied another twenty-five cases in nine provinces in northern China through a meta-analysis (Table 1b). Furthermore, the climate divisions of the twenty-five cases covered not only semi-arid and arid areas but also extremely arid areas. 3.2 Data collection Because there were no other economically scientific methods to collect valid data due to fragmented and incoherent records and non-comparability among different types of records and measures (Sun et al. 2006), the study used four types of methods, including quantitative surveys, qualitative interviews, personal observations, and document analysis, to collect complementary and cross-checked data (Poteete and Ostrom 2008; Poteete et al. 2010) and to form an evidence or data triangle (Patton 1987). Previous studies indicated that a combination of surveys, observations, interviews, and document analysis as a hands-on research technique has proved to be a valid and efficient method for collecting data (e.g. Leach et al. 2002; Poteete et al. 2010; Perry 2012). First, after preinterviews, pre-surveys, archive analysis, and considering the population of each county (Table 2a), in order to guarantee a large sample size to improve the consistency between survey respondents’ perception and actual situations of desertification control, formal surveys were conducted from March to December 2011, with 5,410 copies mailed (the number of sent copies in each county was equal to or over 450) and 4,406 valid responses (Table 2b). Because students in local high schools resided in different regions or townships within the county and could be deemed as a valid and hands-on sample to represent the whole population of the county (Yang et al. 2013), the questionnaires were first randomly distributed to high school students, who were trained to help their relatives and neighbors (including many old farmers and herdsmen, who could not read or did not know how to answer a questionnaire) to complete questionnaires. When respondents forgot or did not know related information, they were also encouraged to get help or talk with other people who had more information or personal experience on desertification control but did not fill in the questionnaires by themselves. Furthermore, because desertification control programs were often conducted in underdeveloped rural villages where people are less influenced by the outside and each decade from 1950s to the 1990s had extremely different programs, movements, and policies, many people could remember (in their own words is ‘would never forget’) their experiences during these decades. The descriptive statistics of the valid responses showed that respondents’ ages ranged from 17 to over 80 years old and their occupations were quite diverse and included fifteen types such as farmers, government officials, researchers, teachers, and businessmen (Table 2b). Table 2. Survey and interview distribution in the twelve counties in northern China (2006–11)a. Counties  Linze  Minqin  Zhongwei  Yanchi  Dengkou  Ejin Horo  Xinbaerhuzuo  Xilinhot  Naiman  Duolun  Wengniute  Aohan  Total  A. Population (Ten thousand)  15.01c  27.47  38.3  16.79  12.34  23.9  4.17  17.37  44.30  10.70  47.70  60.32    B. Survey distribution                            The number of sent copies  450  450  450  450  450  450  450  450  450  450  460  450  5410  Response rates (%)  75.78  100  80.00  99.56  72.00  38.89  86.00  93.56  96.00  100  100  100  86.82  The number of valid copies  328  418  345  439  304  150  387  342  424  449  458  362  4406  Validity rate among received copies (%)  96.19  92.89  95.83  97.99  93.83  85.71  100  81.23  98.15  99.78  99.57  80.44  93.78  C. Types of survey respondents                            Farmers  97  382  130  75  72  53  186  76  70  149  438  256  1984  (29.57)d  (91.39)  (37.68)  (17.08)  (23.68)  (35.33)  (48.06)  (22.22)  (16.51)  (33.18)  (95.63)  (70.72)  (45.03)  Middle schools (teachers and students)  91  8  58  166  99  38  45  99  134  135  11  21  905  (27.74)  (1.91)  (16.81)  (37.81)  (32.57)  (25.33)  (11.63)  (28.95)  (31.60)  (30.07)  (2.40)  (5.80)  (20.54)  General research institutesb  0  1  2  27  2  0  5  3  11  15  0  2  68  (0)  (0.24)  (0.58)  (6.15)  (0.66)  (0)  (1.29)  (0.88)  (2.60)  (3.34)  (0)  (0.55)  (1.54)  Desert control stations  0  0  0  2  2  0  1  0  2  8  0  0  15  (0)  (0)  (0)  (0.46)  (0.66)  (0)  (0.26)  (0)  (0.47)  (1.78)  (0)  (0)  (0.34)  Government  14  1  9  15  13  4  14  24  5  32  1  9  141  (4.27)  (0.24)  (2.61)  (3.42)  (4.28)  (2.67)  (3.62)  (7.02)  (1.18)  (7.13)  (0.22)  (2.49)  (3.20)  Businesses  55  8  48  55  53  10  18  63  14  34  4  28  390  (16.77)  (1.91)  (13.91)  (12.53)  (17.43)  (6.67)  (4.65)  (18.42)  (3.30)  (7.57)  (0.87)  (7.73)  (8.85)  Rural grassroots organizations  7  9  24  15  4  2  2  4  2  41  0  10  120  (2.13)  (2.15)  (6.96)  (3.42)  (1.31)  (1.33)  (0.52)  (1.17)  (0.47)  (9.13)  (0)  (2.76)  (2.72)  Organizations of technology development and promotion in rural areas  4  1  0  3  2  0  4  2  3  1  0  1  21  (1.22)  (0.24)  (0)  (0.68)  (0.66)  (0)  (1.03)  (0.58)  (0.71)  (0.22)  (0)  (0.28)  (0.48)  Universities  1  1  5  18  0  0  4  4  12  0  0  0  45  (0.31)  (0.24)  (1.45)  (4.10)  (0)  (0)  (1.03)  (1.17)  (2.83)  (0)  (0)  (0)  (1.02)  Religious groups  0  0  2  4  0  0  0  2  2  0  0  0  10  (0)  (0)  (0.58)  (0.91)  (0)  0  (0)  (0.58)  (0.47)  (0)  (0)  (0)  (0.23)  Other public institutes  25  1  25  10  20  16  10  40  3  13  3  26  192  (7.62)  (0.24)  (7.25)  (2.28)  (6.58)  (10.67)  (2.58)  (11.70)  (0.71)  (2.90)  (0.66)  (7.18)  (4.36)  Non-governmental organizations  5  0  5  4  3  6  1  3  2  1  0  2  32  (1.52)  (0)  (1.45)  (0.91)  (0.99)  (4)  (0.26)  (0.88)  (0.47)  (0.22)  (0)  (0.55)  (0.73)  News media  1  0  1  2  0  1  0  3  5  0  0  0  13  (0.31)  (0)  (0.29)  (0.46)  (0)  (0.67)  (0)  (0.88)  (1.18)  (0)  (0)  (0)  (0.30)  International organizations  0  0  4  2  0  0  1  0  15  0  0  1  23  (0)  (0)  (1.16)  (0.46)  (0)  (0)  (0.26)  (0)  (3.54)  (0)  (0)  (0.28)  (0.52)  Others  28  6  32  41  34  20  96  19  144  20  1  6  447  (8.54)  (1.44)  (9.27)  (9.33)  (11.18)  (13.33)  (24.81)  (5.56)  (33.96)  (4.46)  (0.22)  (1.66)  (10.15)  D. Interview distribution                            Farmers or residents  4  6  5  1  1  2  2  1  1  1  1  1  26  Scholars, experts & technicians  3  11  4  4  2  3  0  4  5  0  2  4  42  Government officials  1  11  1  3  6  3  3  4  1  3  5  4  45  Businessmen  0  0  0  0  0  0  0  0  2  0  2  0  4  Religious groups or NGOs  0  1  0  0  0  0  0  0  0  0  0  0  1  Total  8  29  10  8  9  8  5  9  9  4  10  9  118  E. Observation distribution                            Numbers  4  11  7  2  9  2  2  2  5  3  2  3  52  Counties  Linze  Minqin  Zhongwei  Yanchi  Dengkou  Ejin Horo  Xinbaerhuzuo  Xilinhot  Naiman  Duolun  Wengniute  Aohan  Total  A. Population (Ten thousand)  15.01c  27.47  38.3  16.79  12.34  23.9  4.17  17.37  44.30  10.70  47.70  60.32    B. Survey distribution                            The number of sent copies  450  450  450  450  450  450  450  450  450  450  460  450  5410  Response rates (%)  75.78  100  80.00  99.56  72.00  38.89  86.00  93.56  96.00  100  100  100  86.82  The number of valid copies  328  418  345  439  304  150  387  342  424  449  458  362  4406  Validity rate among received copies (%)  96.19  92.89  95.83  97.99  93.83  85.71  100  81.23  98.15  99.78  99.57  80.44  93.78  C. Types of survey respondents                            Farmers  97  382  130  75  72  53  186  76  70  149  438  256  1984  (29.57)d  (91.39)  (37.68)  (17.08)  (23.68)  (35.33)  (48.06)  (22.22)  (16.51)  (33.18)  (95.63)  (70.72)  (45.03)  Middle schools (teachers and students)  91  8  58  166  99  38  45  99  134  135  11  21  905  (27.74)  (1.91)  (16.81)  (37.81)  (32.57)  (25.33)  (11.63)  (28.95)  (31.60)  (30.07)  (2.40)  (5.80)  (20.54)  General research institutesb  0  1  2  27  2  0  5  3  11  15  0  2  68  (0)  (0.24)  (0.58)  (6.15)  (0.66)  (0)  (1.29)  (0.88)  (2.60)  (3.34)  (0)  (0.55)  (1.54)  Desert control stations  0  0  0  2  2  0  1  0  2  8  0  0  15  (0)  (0)  (0)  (0.46)  (0.66)  (0)  (0.26)  (0)  (0.47)  (1.78)  (0)  (0)  (0.34)  Government  14  1  9  15  13  4  14  24  5  32  1  9  141  (4.27)  (0.24)  (2.61)  (3.42)  (4.28)  (2.67)  (3.62)  (7.02)  (1.18)  (7.13)  (0.22)  (2.49)  (3.20)  Businesses  55  8  48  55  53  10  18  63  14  34  4  28  390  (16.77)  (1.91)  (13.91)  (12.53)  (17.43)  (6.67)  (4.65)  (18.42)  (3.30)  (7.57)  (0.87)  (7.73)  (8.85)  Rural grassroots organizations  7  9  24  15  4  2  2  4  2  41  0  10  120  (2.13)  (2.15)  (6.96)  (3.42)  (1.31)  (1.33)  (0.52)  (1.17)  (0.47)  (9.13)  (0)  (2.76)  (2.72)  Organizations of technology development and promotion in rural areas  4  1  0  3  2  0  4  2  3  1  0  1  21  (1.22)  (0.24)  (0)  (0.68)  (0.66)  (0)  (1.03)  (0.58)  (0.71)  (0.22)  (0)  (0.28)  (0.48)  Universities  1  1  5  18  0  0  4  4  12  0  0  0  45  (0.31)  (0.24)  (1.45)  (4.10)  (0)  (0)  (1.03)  (1.17)  (2.83)  (0)  (0)  (0)  (1.02)  Religious groups  0  0  2  4  0  0  0  2  2  0  0  0  10  (0)  (0)  (0.58)  (0.91)  (0)  0  (0)  (0.58)  (0.47)  (0)  (0)  (0)  (0.23)  Other public institutes  25  1  25  10  20  16  10  40  3  13  3  26  192  (7.62)  (0.24)  (7.25)  (2.28)  (6.58)  (10.67)  (2.58)  (11.70)  (0.71)  (2.90)  (0.66)  (7.18)  (4.36)  Non-governmental organizations  5  0  5  4  3  6  1  3  2  1  0  2  32  (1.52)  (0)  (1.45)  (0.91)  (0.99)  (4)  (0.26)  (0.88)  (0.47)  (0.22)  (0)  (0.55)  (0.73)  News media  1  0  1  2  0  1  0  3  5  0  0  0  13  (0.31)  (0)  (0.29)  (0.46)  (0)  (0.67)  (0)  (0.88)  (1.18)  (0)  (0)  (0)  (0.30)  International organizations  0  0  4  2  0  0  1  0  15  0  0  1  23  (0)  (0)  (1.16)  (0.46)  (0)  (0)  (0.26)  (0)  (3.54)  (0)  (0)  (0.28)  (0.52)  Others  28  6  32  41  34  20  96  19  144  20  1  6  447  (8.54)  (1.44)  (9.27)  (9.33)  (11.18)  (13.33)  (24.81)  (5.56)  (33.96)  (4.46)  (0.22)  (1.66)  (10.15)  D. Interview distribution                            Farmers or residents  4  6  5  1  1  2  2  1  1  1  1  1  26  Scholars, experts & technicians  3  11  4  4  2  3  0  4  5  0  2  4  42  Government officials  1  11  1  3  6  3  3  4  1  3  5  4  45  Businessmen  0  0  0  0  0  0  0  0  2  0  2  0  4  Religious groups or NGOs  0  1  0  0  0  0  0  0  0  0  0  0  1  Total  8  29  10  8  9  8  5  9  9  4  10  9  118  E. Observation distribution                            Numbers  4  11  7  2  9  2  2  2  5  3  2  3  52  a Adapted from Yang (2015) and Yang et al. (2013). b ‘Types of organizations’ refers to people in these organizations. c The population of the counties was based on data from the Statistical Bulletins of National Economic and Social Development in 2011. d Numbers in brackets are the percentages among the valid copies. Second, in order to cross-check the survey data, semi-structured interviews as well as participatory and nonparticipatory observations in three counties (Minqin, Linze, and Zhongwei) were conducted from June 2006 to February 2008 (the first step) and in the other counties from July to August 2011 (the second step). Although the same questions were asked in both the first and second step interviews, the data from the first step were mainly used to provide basic information for survey questionnaire design. The 118 interviewees included scholars, farmers, citizens, government officials, and researchers in desert control institutes, whose ages ranged from 20 to more than 60 years old (Table 2d). Most interviews lasted approximately 30 minutes to 2 hours, and the open-ended questions corresponded to the survey questions, which provided choices. Furthermore, the 52 observed sites included desert control stations, typical areas of desertification control, famous natural reserves, the Bureau of Forestry, and the Bureau of Environmental Protection (Table 2e). Third, diverse literature and archive data, including county annals, government gazettes, government documents, research reports, news reports, and other published and non-published literature from 1949 to 2011, were collected to form a archive data triangle (Patton 1987) and to complement and cross-check the data from the surveys, interviews, and observations and for a meta-analysis of the twenty-five cases to test the generalizability of the institutional design principles found in the twelve field study counties. 3.3 Variables and measurements According to the theoretical framework, the major research variables in the study were: (1) levels of science, social science, and local knowledge applied in desertification control, (2) performance of desertification control, (3) factors influencing collaboration between the three types of knowledge, and (4) the institutional design principles for successful collaboration between the three types of knowledge. In surveys of the twelve field study counties, questions with a six-point scale (‘very large, large, medium, moderately small, very small, and don’t know’ or ‘strongly agree, agree, neutral, moderately disagree, strongly disagree, and don’t know’) were designed for respondents to directly evaluate the levels of science, social science, and local knowledge applied in desertification control, performance of desertification control, and various factors influencing the application and collaboration between the three types of knowledge of each county. In the six-point scale, the first five points in fact were similar to Likert scale. But a five-point scale as listed above could not cover the possibility when survey respondents did not know actual situations. Thus, in order to reduce the errors caused by survey respondents, I added ‘don’t know’ as the sixth point. However, in the data analysis, any response with ‘don’t know’ selected was excluded to ensure data integrity. The nine factors were evaluated by the average values of the subproblems used to measure the corresponding factors. For example, the factor ‘level of knowledge’ of each county was evaluated by the average values of three subproblems (low levels of science and technology development, low levels of social science development, and low values of local knowledge) as rated by survey respondents of the county. The Harmon one-factor analysis showed that although there was a factor that explained the variance in approximately 41 per cent of the resulting factors based on the sample of all of the valid survey responses (N = 4406), it was not a majority (over 50 per cent). This meant that the common methods bias (Podsakoff and Organ 1986) might not be a serious problem in this study. Furthermore, the study used aggregated data based on calculating the percentages of survey respondents’ answers of each county rather than individual answers to evaluate research variables of the county. This also improved the consistency between respondents’ perception of each county and actual situations of combating desertification of the county (Yang et al. 2013). Moreover, cross-checking and complementing the data from interviews, observations, and the meta-analysis as well as the protection of the respondents’ anonymity and confidentiality in the surveys also increased the validity of the evaluation of the variables. Furthermore, in order to analyze the types of knowledge collaboration in the twelve field study cases, their levels of knowledge application were also divided into two levels (high and low). For example, if the evaluation by one county was higher than the average of the twelve counties, it was deemed ‘high’; if it was less than the average, it was deemed ‘low’. When comparing the average performance of desertification control of the different types of knowledge collaboration in the twelve field study cases with the twenty-five meta-analysis cases, their performance was coded into three groups (successful, semi-successful, and unsuccessful) by trisecting the interval between the maximum and minimum accumulated percentages of the survey respondents and the values of ‘successful’ to ‘unsuccessful’ were assigned 3, 2, and 1. Finally, based on the comprehensive qualitative meta-analysis of diverse literature and archive data, the study also analyzed the types of knowledge collaboration in the twenty-five meta-analysis cases by dividing their levels of knowledge application (science, social science, and local knowledge) into two levels (high and low) and their performance of desertification control into three groups (successful, semi-successful, and unsuccessful), while the satisfaction of the nine institutional design principles corresponding to the nine factors for successful collaboration was coded into four relative levels (high, middle, low, non-satisfaction). For example, as to Principle 8, if a county (or city or district) not only strictly implemented national laws and regulations but also had its local laws and regulations for a long time, it was coded ‘high’; whereas if a county only strictly implemented national laws and regulations but did not have its own local laws and regulations for a long time, it was coded; ‘middle’; and if a county did not implement national laws and regulations very well, it was coded ‘low’. Otherwise, it was coded ‘non-satisfaction’. When calculating the correlation coefficient between the principles and performance, the values of ‘high’ to ‘non-satisfaction’ were assigned 3, 2, 1, and 0, whereas the values of ‘successful’ to ‘unsuccessful’ were also 3, 2, and 1. To avoid personal errors and subjectivity of coding results, the variables were first coded by a research assistant and her ten classmates together, and then were rechecked by the author independently. To avoid the influence by prior knowledge of research hypotheses in particular, all the ten classmates were blind to research purpose, questions, and hypotheses. 4. Results 4.1 The level of knowledge and its influence on the performance of desertification control Among the six choices, on average over 30 per cent of survey respondents indicated that the level of applying natural science and local knowledge was ‘very large’ or ‘large’ in the twelve counties, and 28 per cent said the same for social science. Therefore, the average of the types of knowledge was 30.2 per cent (Table 3a). Furthermore, approximately 30 per cent of survey respondents indicated that the performance of desertification control in the 12 counties was ‘very large’ or ‘large’. Table 3. Average levels of knowledge application and statistical results of the relationship between knowledge application with the performance of desertification control as rated by the survey respondents in the twelve field study counties in northern China (2011) (N = 12). Independent variables  A. Average levels of knowledge application in the twelve cases  B. Method 1: Correlation analysis   C. Method 2: Multivariable liner regression (Enter and Stepwise)   D. Method 3:Multivariable liner regression(Enter)   Coefficients  Sig.  B  Standard error  t  Sig.  B  Standard error  t  Sig.  Constant        −1.775  1.307  −1.359  0.175          Natural science  32.0a  0.335  0.287  −0.248***  0.022  −11.221  0.000          Social science  28.0  0.839***  0.001  0.854***  0.067  12.826  0.000          Local knowledge  30.8  0.789***  0.002  0.565***  .079  7.183  0.000          Average  30.2  0.700**  0.011          0.840***  0.055  15.147  0.000  R        0.894  0.622  R2        0.799  0.386  Adjusted R2        0.797  0.385  Standard error of the estimate        5.73095  9.97516  F(ANOVA)        478.898***  229.437***  (Sig.)        0.000  0.000  Independent variables  A. Average levels of knowledge application in the twelve cases  B. Method 1: Correlation analysis   C. Method 2: Multivariable liner regression (Enter and Stepwise)   D. Method 3:Multivariable liner regression(Enter)   Coefficients  Sig.  B  Standard error  t  Sig.  B  Standard error  t  Sig.  Constant        −1.775  1.307  −1.359  0.175          Natural science  32.0a  0.335  0.287  −0.248***  0.022  −11.221  0.000          Social science  28.0  0.839***  0.001  0.854***  0.067  12.826  0.000          Local knowledge  30.8  0.789***  0.002  0.565***  .079  7.183  0.000          Average  30.2  0.700**  0.011          0.840***  0.055  15.147  0.000  R        0.894  0.622  R2        0.799  0.386  Adjusted R2        0.797  0.385  Standard error of the estimate        5.73095  9.97516  F(ANOVA)        478.898***  229.437***  (Sig.)        0.000  0.000  a The average percentages of ‘very large’ and ‘large’ rated by survey respondents. *** P < 0.01, ** P < 0.05. The results also showed that the correlation coefficients of ‘social science’, ‘local knowledge’, and ‘the average of three types of knowledge’, with the performance of desertification control as rated by the survey respondents, produced through SPSS (Statistical Product and Service Solutions) analysis, were all significant at the 0.05 level. However, the coefficient of natural science was not significant (Table 3b). Furthermore, the results of multivariate linear regression (the data are almost normal) indicated that the coefficient of natural science was even negative, the positive coefficient of social science was larger than the coefficient of local knowledge (Table 3c), the coefficient of the average of the three types of knowledge was positive (Table 3d), and all these coefficients were significant at the 0.01 level. 4.2 Types of knowledge collaboration and their correlation with governance performance Based on the classification of the three types of knowledge (science, social science, and local knowledge) and dividing the levels of knowledge application into two levels (high and low), the types of knowledge collaboration in the twelve counties were divided into seven types. These types were consistent with the eight types shown in Fig. 2 such as (high science, high social science, high local knowledge) and (low natural science, low social science, and low local knowledge), except for (high natural science, low social science, and high local knowledge) (Table 4). Table 4. The types of collaboration among natural science, social science, and local knowledge and the average performance of desertification control in the twelve field study cases as rated by survey respondents (2011) and in the thirty-seven cases (including the twelve field study cases and twenty-five meta-analysis cases) based on the recoded data. TypesItems  (High natural science, high social science, and high local knowledge)  (High natural science, high social science, and low local knowledge)  (High natural science, low social science, and high local knowledge)  (High natural science, low social science, and low local knowledge)  (Low natural science, high social science, and high local knowledge)  (Low natural science, high social science, and low local knowledge)  (Low natural science, low social science, and high local knowledge)  (Low natural science, low social science, and low local knowledge)  Twelve field study cases  Counties falling into the types  Linze, Zhongwei, Dengkou  Duolun  No data  Ejin Horo, Wengniute  Xinbaerhuzuo  Xilinhot  Aohan  Minqin, Yanchi, Naiman  Average performance  39.9a  34.3  No data  22.5  55.2  29.6  18.4  21.4  [Orders]  [2]  [3]  [No data]  [5]  [1]  [4]  [7]  [6]  Thirty-seven casesb  Number of counties  11  3  2  3  4  4  2  8  Average performance  2.82c  2.33  2.00  1.33  2.75  2.00  1.50  1.25  [Orders]  [1]  [3]  [4]  [7]  [2]  [4]  [6]  [8]  TypesItems  (High natural science, high social science, and high local knowledge)  (High natural science, high social science, and low local knowledge)  (High natural science, low social science, and high local knowledge)  (High natural science, low social science, and low local knowledge)  (Low natural science, high social science, and high local knowledge)  (Low natural science, high social science, and low local knowledge)  (Low natural science, low social science, and high local knowledge)  (Low natural science, low social science, and low local knowledge)  Twelve field study cases  Counties falling into the types  Linze, Zhongwei, Dengkou  Duolun  No data  Ejin Horo, Wengniute  Xinbaerhuzuo  Xilinhot  Aohan  Minqin, Yanchi, Naiman  Average performance  39.9a  34.3  No data  22.5  55.2  29.6  18.4  21.4  [Orders]  [2]  [3]  [No data]  [5]  [1]  [4]  [7]  [6]  Thirty-seven casesb  Number of counties  11  3  2  3  4  4  2  8  Average performance  2.82c  2.33  2.00  1.33  2.75  2.00  1.50  1.25  [Orders]  [1]  [3]  [4]  [7]  [2]  [4]  [6]  [8]  a The average percentages of ‘very large’ and ‘large’ rated by survey respondents in the counties fallen into the types of knowledge collaboration. b Based on the recoded data (dived knowledge application into two levels : ‘High’ and ‘Low’ while divided performance into three levels: ‘successful’, ‘semi-successful’, and ‘unsuccessful’). c Given S = 3, Se = 2, and U = 1 in Table 6. Furthermore, the study also analyzed the types of knowledge collaboration in the twenty-five meta-analysis cases, and the results indicated that all the eight types of knowledge collaboration shown in Fig. 2 were found (see Table 6). The average performance of the different types of knowledge collaboration and their orders in the thirty-seven cases in total (twenty-five meta-analysis cases and twelve field study cases based on recorded data) indicated that the type with high levels of all three types of knowledge had the highest performance, while the types with low levels of all three types of knowledge had the lowest performance. Meanwhile, the types with high social science and local knowledge often had high performance, and the types with low social science and local knowledge often had low performance. 4.3 Influencing factors and mechanisms Based on the perceived subproblems influencing the application of the three types of knowledge as rated by the survey respondents, the results indicated that there were nine factors influencing the application and collaboration of knowledge (Appendix A) and they were consistent with the nine factors identified in the theoretical framework (Fig. 2) of the study. The correlation coefficients of the nine factors and the performance of desertification control in the counties were all significant at the 0.01 level (Appendix B). Thus, based on the problems rated by the survey respondents and the nine factors influencing the application and collaboration of knowledge (Appendix A) as well as the nine factors identified in the theoretical framework (Fig. 2), I produced nine design principles for successful collaboration among the three types of knowledge (Table 5). In order to test the generalizability of the findings from the field study as well as the significance of these principles, I used these principles to characterize all the twenty-five meta-analysis cases (Table 6). The table shows that the higher the satisfaction levels of the principles were the more successful performance of desertification was. Table 5. Nine design principles for successful combination between natural science, social science, and local knowledge. Nine design principles  Group 1: Combination among the three types of knowledge  P1. Sufficient developments and continued improvements in the three types of knowledge  P2. The effective refinement, transformation, adaption, application, and extension of the three types of knowledge  P3. High complementation and coordination among the three types of knowledge  Group 2: Collaboration between knowledge possessors and other social actors  P4. High capability and sustained endeavors of knowledge possessors on the application and extension of three types of knowledge  P5. Effective communication and collaboration among knowledge possessors and other social actors  P6. Sufficient support from other social actors and continued improvements on their attention and understanding of the three types of knowledge as well as their capabilities  Group 3: Reliable and sustainable external support  P7. Reliable and sustained support and guidance by governments at different hierarchical levels  P8. Sustained institutional support from laws and regulations  P9. Sufficient financial support  Nine design principles  Group 1: Combination among the three types of knowledge  P1. Sufficient developments and continued improvements in the three types of knowledge  P2. The effective refinement, transformation, adaption, application, and extension of the three types of knowledge  P3. High complementation and coordination among the three types of knowledge  Group 2: Collaboration between knowledge possessors and other social actors  P4. High capability and sustained endeavors of knowledge possessors on the application and extension of three types of knowledge  P5. Effective communication and collaboration among knowledge possessors and other social actors  P6. Sufficient support from other social actors and continued improvements on their attention and understanding of the three types of knowledge as well as their capabilities  Group 3: Reliable and sustainable external support  P7. Reliable and sustained support and guidance by governments at different hierarchical levels  P8. Sustained institutional support from laws and regulations  P9. Sufficient financial support  Note: P1–P9 refer to Principles 1–9. By using the data of Principles 1–9 and the performance of desertification of the twenty-five meta-analysis cases in Table 6, I used five statistical methods to study the relationship between the nine principles and performance of desertification control. The Kruskal–Wallis Test showed that the chi-squares of the nine principles were high and significant at the 0.01 significance level (Table 7a), and this indicated that the nine principles in the twenty-five cases were indeed different. The correlation coefficients between the nine principles and performance of desertification control indicated the correlation coefficients of all the nine principles were also significant at the 0.01 significance level, and the coefficient of Principle 9 was the largest, followed by Principles 2 and 3 (Table 7b). That is, the findings from the twelve field study cases were replicated in the twenty-five meta-analysis cases. Although the Enter multivariable liner regression (the data are normal) indicated that none of the coefficients of the nine principles were significant at the 0.05 level, its values of R and R2 were highest among the three methods of multivariable regression (Table 7c). The Backward multivariable liner regression indicated that the coefficients of Principles 1, 4, 5, 6, and 9 were significant at the 0.1 significant level, while Principles 2, 3, 7, and 8 were excluded from the model. Its values of R and R2 were less than the Enter model but higher than the Stepwise model (Table 7d). The Stepwise multivariable liner regression showed that the coefficients of Principles 3, 6, and 9 were significant at the 0.05 level, while all the other six principles were excluded. Its values of R and R2 were lowest among the three models of multivariable regression (Table 7e). Table 6. Types of knowledge combination, nine design principles, and performance of desertification control of the twelve field study cases and the twenty-five meta-analysis cases in northern China (N = 37). Cases  Types of knowledge combination   Nine design principles   Performance  Natural science  Social science  Local knowledge  P1  P2  P3  P4  P5  P6  P7  P8  P9  a. Twelve field study cases                            (1) Linze  (H  H  H)  –  –  –  –  –  –  –  –  –  S  (2) Minqin  (L  L  L)  –  –  –  –  –  –  –  –  –  U  (3) Zhongwei  (H  H  H)  –  –  –  –  –  –  –  –  –  S  (4) Yanchi  (L  L  L)  –  –  –  –  –  –  –  –  –  Se  (5) Dengkou  (H  H  H)  –  –  –  –  –  –  –  –  –  U  (6) Ejin Horo  (H  L  L)  –  –  –  –  –  –  –  –  –  Se  (7) Xinbaerhuzuo  (L  H  H)  –  –  –  –  –  –  –  –  –  S  (8) Xilinhot  (L  H  L)  –  –  –  –  –  –  –  –  –  Se  (9) Naiman  (L  L  L)  –  –  –  –  –  –  –  –  –  U  (10) Duolun  (H  H  L)  –  –  –  –  –  –  –  –  –  Se  (11) Wengniute  (H  L  L)  –  –  –  –  –  –  –  –  –  U  (12) Aohan  (L  L  H)  –  –  –  –  –  –  –  –  –  U  b. Eighteen document analysis cases                            (1) Zhengxiangbai  (L  H  H)  M  H  M  H  M  H  H  H  H  S  (2) Zhenglan  (H  L  L)  M  M  M  M  L  N  M  L  L  U  (3) Siziwang  (H  H  H)  H  H  H  H  M  H  H  M  M  S  (4) Wulatehou  (L  H  L)  M  L  M  M  H  M  H  H  L  Se  (5) Ejina  (L  L  L)  M  L  L  L  L  N  M  L  L  U  (6) Wushen  (H  H  H)  H  H  M  H  M  H  H  H  H  S  (7) Guyang  (L  L  L)  N  L  ND  L  L  N  L  L  N  U  (8) Tongliao  (H  H  H)  H  H  H  H  M  H  H  H  H  S  (9) Shangdu  (H  H  L)  M  H  M  M  H  M  H  L  M  Se  (10) Balinyou  (H  H  H)  H  H  H  M  H  H  H  H  H  S  (11) Dalate  (H  H  H)  H  H  H  H  H  H  H  H  H  S  (12) Wuwei  (H  L  H)  H  M  M  L  M  H  H  M  M  Se  (13) Shandan  (L  L  L)  L  N  N  L  L  L  H  H  L  U  (14) Guazhou  (L  H  H)  M  H  H  M  H  H  H  H  H  S  (15) Dunhuang  (H  H  H)  H  H  M  M  H  H  H  M  H  S  (16) Maqu  (L  L  L)  N  L  N  L  L  L  M  ND  L  U  (17) Turpan  (H  H  H)  H  H  H  M  H  M  H  H  H  S  (18) Golmud  (L  H  H)  M  H  M  M  M  H  H  M  M  Se  (19) Yulin  (L  L  H)  L  M  L  M  H  M  H  L  H  Se  (20) Zhangbei  (H  L  H)  H  L  L  M  M  M  H  M  L  Se  (21) Guyuan  (L  L  L)  L  L  M  H  M  H  L  M  L  Se  (22) Fengning  (H  H  L)  H  H  M  H  H  H  H  M  H  S  (23) Daxing  (H  H  H)  H  H  H  H  M  M  H  H  H  S  (24) Kangping  (L  H  L)  L  L  M  L  H  L  H  M  M  Se  (25) Tailai  (L  H  L)  L  H  L  M  M  L  H  H  H  Se  Cases  Types of knowledge combination   Nine design principles   Performance  Natural science  Social science  Local knowledge  P1  P2  P3  P4  P5  P6  P7  P8  P9  a. Twelve field study cases                            (1) Linze  (H  H  H)  –  –  –  –  –  –  –  –  –  S  (2) Minqin  (L  L  L)  –  –  –  –  –  –  –  –  –  U  (3) Zhongwei  (H  H  H)  –  –  –  –  –  –  –  –  –  S  (4) Yanchi  (L  L  L)  –  –  –  –  –  –  –  –  –  Se  (5) Dengkou  (H  H  H)  –  –  –  –  –  –  –  –  –  U  (6) Ejin Horo  (H  L  L)  –  –  –  –  –  –  –  –  –  Se  (7) Xinbaerhuzuo  (L  H  H)  –  –  –  –  –  –  –  –  –  S  (8) Xilinhot  (L  H  L)  –  –  –  –  –  –  –  –  –  Se  (9) Naiman  (L  L  L)  –  –  –  –  –  –  –  –  –  U  (10) Duolun  (H  H  L)  –  –  –  –  –  –  –  –  –  Se  (11) Wengniute  (H  L  L)  –  –  –  –  –  –  –  –  –  U  (12) Aohan  (L  L  H)  –  –  –  –  –  –  –  –  –  U  b. Eighteen document analysis cases                            (1) Zhengxiangbai  (L  H  H)  M  H  M  H  M  H  H  H  H  S  (2) Zhenglan  (H  L  L)  M  M  M  M  L  N  M  L  L  U  (3) Siziwang  (H  H  H)  H  H  H  H  M  H  H  M  M  S  (4) Wulatehou  (L  H  L)  M  L  M  M  H  M  H  H  L  Se  (5) Ejina  (L  L  L)  M  L  L  L  L  N  M  L  L  U  (6) Wushen  (H  H  H)  H  H  M  H  M  H  H  H  H  S  (7) Guyang  (L  L  L)  N  L  ND  L  L  N  L  L  N  U  (8) Tongliao  (H  H  H)  H  H  H  H  M  H  H  H  H  S  (9) Shangdu  (H  H  L)  M  H  M  M  H  M  H  L  M  Se  (10) Balinyou  (H  H  H)  H  H  H  M  H  H  H  H  H  S  (11) Dalate  (H  H  H)  H  H  H  H  H  H  H  H  H  S  (12) Wuwei  (H  L  H)  H  M  M  L  M  H  H  M  M  Se  (13) Shandan  (L  L  L)  L  N  N  L  L  L  H  H  L  U  (14) Guazhou  (L  H  H)  M  H  H  M  H  H  H  H  H  S  (15) Dunhuang  (H  H  H)  H  H  M  M  H  H  H  M  H  S  (16) Maqu  (L  L  L)  N  L  N  L  L  L  M  ND  L  U  (17) Turpan  (H  H  H)  H  H  H  M  H  M  H  H  H  S  (18) Golmud  (L  H  H)  M  H  M  M  M  H  H  M  M  Se  (19) Yulin  (L  L  H)  L  M  L  M  H  M  H  L  H  Se  (20) Zhangbei  (H  L  H)  H  L  L  M  M  M  H  M  L  Se  (21) Guyuan  (L  L  L)  L  L  M  H  M  H  L  M  L  Se  (22) Fengning  (H  H  L)  H  H  M  H  H  H  H  M  H  S  (23) Daxing  (H  H  H)  H  H  H  H  M  M  H  H  H  S  (24) Kangping  (L  H  L)  L  L  M  L  H  L  H  M  M  Se  (25) Tailai  (L  H  L)  L  H  L  M  M  L  H  H  H  Se  Notes: P1–P9 refer to Principles 1–9; H = high; M = middle; L = low; N = non-satisfaction; ND = no data; S = successful; Se = semi-successful; and U = unsuccessful. Because the nine principles were produced based on the finding from the twelve field study cases, I did not recoded them in the twelve cases here in order to avoid the problem of self-verification. Table 7. Statistical results of the relationship between the nine design principles (independent variables) and the performance of desertification control (dependent variables) in the twenty-five meta-analysis cases in northern China (N = 25) (2011). Independent variables  A. Method 1: K-W test   B. Method 2: Correlation analysis (Spearman)   C. Method 3: Multivariable liner regression (Enter)   D. Method 3: Multivariable liner regression (Backward)   E. Method 3: Multivariable liner regression (Stepwise)   Chi-square  Coefficients  B  Standard Error  t  Sig.  B  Standard error  t  Sig.  B  Standard error  t  Sig.  Constant      −0.613  0.455  −1.347  0.198  −0.915***  0.283  −3.232  0.004  −0.221  0.207  −1.071  0.296  P1  13.036***  0.735*** [6]  0.145  0.097  1.493  0.156  0.152** [5]  0.070  2.172  0.043          P2  15.846***  0.813*** [2]  0.061  0.117  0.518  0.612                  P3  15.463***  0.803***[3]  0.006  0.100  0.061  0.952          0.195**[3]  .081  2.413  0.025  P4  13.045***  0.736*** [5]  0.158  0.116  1.368  0.191  0.244** [2]  0.097  2.515  0.021          P5  13.441***  0.601*** [9]  0.184  0.107  1.712  0.107  0.171* [3]  0.096  1.790  0.089          P6  15.351***  0.773*** [4]  0.162*  0.081  1.998  0.064  0.162* [4]  0.079  2.043  0.055  0.270***[2]  .077  3.505  0.002  P7  12.749***  0.626*** [8]  −0.151  0.150  −1.005  0.331                  P8  9.467***  0.628*** [7]  0.113  0.074  1.535  0.146                  P9  16.904***  0.838*** [1]  0.266*  0.136  1.958  0.069  0.286***[1]  0.082  3.498  0.002  0.345***[1]  0.085  4.056  0.001  R        0.964        0.957        0.939      R2        0.930        0.915        0.882      Adjusted R2        0.888        0.893        0.865      Standard error of the estimate        0.26102        0.25467        0.28609      D-W        2.114        2.107        2.112      F(AVOVA)        22.078***        14.10***        52.598***      Independent variables  A. Method 1: K-W test   B. Method 2: Correlation analysis (Spearman)   C. Method 3: Multivariable liner regression (Enter)   D. Method 3: Multivariable liner regression (Backward)   E. Method 3: Multivariable liner regression (Stepwise)   Chi-square  Coefficients  B  Standard Error  t  Sig.  B  Standard error  t  Sig.  B  Standard error  t  Sig.  Constant      −0.613  0.455  −1.347  0.198  −0.915***  0.283  −3.232  0.004  −0.221  0.207  −1.071  0.296  P1  13.036***  0.735*** [6]  0.145  0.097  1.493  0.156  0.152** [5]  0.070  2.172  0.043          P2  15.846***  0.813*** [2]  0.061  0.117  0.518  0.612                  P3  15.463***  0.803***[3]  0.006  0.100  0.061  0.952          0.195**[3]  .081  2.413  0.025  P4  13.045***  0.736*** [5]  0.158  0.116  1.368  0.191  0.244** [2]  0.097  2.515  0.021          P5  13.441***  0.601*** [9]  0.184  0.107  1.712  0.107  0.171* [3]  0.096  1.790  0.089          P6  15.351***  0.773*** [4]  0.162*  0.081  1.998  0.064  0.162* [4]  0.079  2.043  0.055  0.270***[2]  .077  3.505  0.002  P7  12.749***  0.626*** [8]  −0.151  0.150  −1.005  0.331                  P8  9.467***  0.628*** [7]  0.113  0.074  1.535  0.146                  P9  16.904***  0.838*** [1]  0.266*  0.136  1.958  0.069  0.286***[1]  0.082  3.498  0.002  0.345***[1]  0.085  4.056  0.001  R        0.964        0.957        0.939      R2        0.930        0.915        0.882      Adjusted R2        0.888        0.893        0.865      Standard error of the estimate        0.26102        0.25467        0.28609      D-W        2.114        2.107        2.112      F(AVOVA)        22.078***        14.10***        52.598***      Notes: P1–P9 refer to Principles 1–9; *** P < 0.01(two-tailed); ** P < 0.05; * P < 0.1; [1]–[9] refers to the order of coefficients from high to low. 5. Discussion 5.1 The important influence of knowledge on governance performance The results indicated that all three types of knowledge were significantly applied in desertification control as rated by survey respondents, but on average natural science was the most widely applied knowledge among the twelve counties, followed by local knowledge, and finally social science (Table 3). However, interestingly the order of the correlation coefficients of the three types of knowledge with the performance of desertification control from the highest to the lowest was social science, local knowledge, and natural science. Furthermore, the results indicated that the coefficient of natural science was not only the lowest but was also the only one that was not significant. That is, although social science was the least applied knowledge in desertification control, it was the most important knowledge among the three influencing the performance of desertification control. On the other hand, although natural science was the most applied knowledge, it was the least important knowledge among the three that influenced performance. In particular, the negative regression coefficient of natural science demonstrated that the mere use of natural science would even have a negative impact on governance performance of desertification control, because without the support of local knowledge and social science or the collaboration among the three types of knowledge, the application of natural science must be blind, and might bring destructive effects to both desertification control and people’s lives. For example, Yang et al. (2010: 159) noted that various natural scientists from universities and research institutes often had negative impact on desertification control. Although they ‘often did not fully understand local condition and also had no concrete and reasonable ideas on how to resolve local problems’, ‘their suggestions were often respected by local officials in their local policy making’. Thus, ‘activities and policies based on their suggestions’ such as ‘planting white poplar to combat desertification’ during the 1980s and 1990s and ‘forcing farmers to grow vegetables in plastic greenhouse’ in the 2000s in Minqin, ‘often deteriorated desertification condition rather than improved land amelioration’. However, the relatively high and significant correlation coefficient and regression coefficient of the average of the three types of knowledge indicated that on average, the three types of knowledge had important and significant influences on the performance of desertification control. Furthermore, it indicated that in order to improve the influence of natural science, it could be combined with social science and local knowledge, because the impacts of natural scientific knowledge on governance performance often can be realized through policy marking, policy implementation, and management, which are the jobs and strength of social science and local knowledge. All of the results showed that knowledge collaboration was very important, and the roles of both social science and local science should be emphasized. These findings were not only consistent with previous literature stressing the importance of local knowledge (Fischer 2000; Yang and Wu 2010; Taylor and de Loë 2012), social science (Hayek 1945; Ruttan 1984), and knowledge collaboration (Thomas and Twyman 2004; Ison et al. 2007; Reynolds et al. 2007; Stringer et al. 2009; Winslow et al. 2011; Holm et al. 2013) but also showed the relative importance of social science compared with local knowledge. However, the fact that social science had the lowest application degree of the three types of knowledge indicated that the importance of social science has not received enough attention. Thus, in the following activities of desertification control and other social affairs, both worldwide researchers and practitioners (including decision makers and implementers) should not only emphasize collaboration between the three types of knowledge but also pay more attention to the development and application of social science, especially in mainland China, where many social sciences are often deemed as useless disciplines and subjects by many people. 5.2 Types of knowledge collaboration and the importance of social science and local knowledge Except for the type of (high natural science, low social science, and high local knowledge), the results indicated that all the other seven types of knowledge collaboration shown in Fig. 2 were found in the twelve field study counties. Furthermore, the analysis of the twenty-five meta-analysis cases showed that all the eight types of knowledge collaboration were found, although some types are more, and some less. All these suggest that the classification of knowledge collaboration shown in Fig. 2 is not only feasible in theory but in line with the actual situation. Thus, the classification can be used not only a useful framework for researchers to studying knowledge collaboration but a simple and convenient tool for practitioners to resolve practical problems. Depending on various traditions, purposes, and practical considerations, researchers and practitioners often divide knowledge into different types. For example, since the time of Sir Isaac Newton, people have also divided knowledge into two types, scientific knowledge (mainly natural science and technology) and unscientific knowledge, and many scientists do not consider knowledge from social science to be scientific knowledge. Thus, when combining social science and local knowledge, the types of knowledge collaboration could be based on the types of knowledge of ‘natural science’ and ‘social and local knowledge’ or ‘science’ and ‘nonscience’ (Guston 2001) and their application levels. Meanwhile, to some scientists, social scientists, other types of researchers, and practitioners, natural science and social science should be put into one group as natural and social science, which is different from local knowledge. In this case, the classification of the types of knowledge application could be based on two types of knowledge (‘natural and social science’ or ‘science’ and ‘local knowledge’) and their application levels as recognized by many social and natural science researchers (e.g. Berkes et al. 2000; Bala and Joseph 2007; Weiss et al. 2013). The classification of the types of knowledge collaboration based on the three types of knowledge (science, social science, and local knowledge) undoubtedly provides a new useful perspective for further theoretical studies on knowledge collaboration and practice. Furthermore, the lowest performance of the type with low levels of all three types of knowledge and the highest performance of the type with high levels of all three types of knowledge once again show the importance of knowledge collaboration. Meanwhile, the low performance of the types with low social science and local knowledge and the high performance of the types with high social science and local knowledge once again emphasize the importance of social science and local knowledge, especially today when people often pay a great deal of attention to natural science rather than to social science and local knowledge (Fazey et al. 2006; Innes and Booher 2010). Meanwhile, these findings suggested that to policy makers and implementers, no matter what their views on the classification of knowledge are, in order to improve governance performance, the types of knowledge collaboration with high social science and local knowledge should be encouraged and get the greatest attention in practice. 5.3 Nine design principles for collaboration of the three types of knowledge The study revealed that collaboration between the three types of knowledge was influenced by the nine factors that were divided into three groups: knowledge itself, social actors (including knowledge possessors and other social actors), and external support. This result was not only consistent with views in the literature about the application of knowledge and relationships between different types of knowledge but also corresponded to the theoretical framework (Fig. 2) of the current study. Furthermore, based on the nine factors, the study proposed nine design principles for successful collaboration among natural science, social science, and local knowledge. Corresponding to the three groups of the nine factors, these nine principles could also be divided into three groups: collaboration among the three types of knowledge, collaboration among knowledge possessors and other social actors, and reliable and sustainable external support. Although many aspects of these principles have been discussed previously in environmental and collaboration literature (Campbell 1992; Fullen and Mitchell 1994; Thomas 1997; Twyman 2004; Ison et al. 2007; Reynolds et al. 2007; Thomas and Tschakert 2007; Stringer et al. 2009; Yang and Wu 2010; Chasek et al. 2011; Reed et al. 2007; Holm et al. 2013), the current study not only reconfirmed the former findings but also found that these findings could be combined to form fundamental design principles for successful collaboration among the three types of knowledge. Furthermore, in addition to collaboration among the three types of knowledge, the nine principles also stressed effective communication and collaboration among knowledge possessors and other social actors. In such collaboration, relying on their knowledge of laws, policies, and management, social scientists could provide suggestions for farmers how to better deal with the relationship between government officials, while seeking more money from the government; while farmers could provide more local knowledge to social scientists to help them better understand the local problems, so as to provide better policy recommendations for the government (Fig. 4a). Natural scientists could help social scientists better understand the problems of desertification control from the perspective of science and technology and provide scientific and technical support for practical governance, meanwhile social scientists could provide law, policy, and management recommendations for the application of science and technology of natural scientists and often played more important roles in law and policy making (Fig. 4b). In particular, in some multi-participant discussion meetings (Fig. 4c and 4d), government officials, farmers, social scientists, and natural scientists could sit together to discuss issues, work out plans, and solve problems. Figure 4. View largeDownload slide Collaboration among farmers, natural scientists, and social scientist. (a) Collaborative discussion between farmers and social scientists in Zhongwei, Ningxia (22 July 2007). (b) Collaborative discussion between natural scientists and social scientists at the Inner Mongolia Grassland Ecosystem Research Station, the CAS, Xilinhot, Inner Mongolia (29 July 2011). (c) Discussion meetings among farmers, natural scientists, social scientists, and government officials in Hongyue village, Jingtai County, Gansu (24 July 2007). (d) Discussion meetings among farmers, natural scientists, social scientists, and government officials in Hongshui Township, Jingtai County, Gansu (14 December 2013). Figure 4. View largeDownload slide Collaboration among farmers, natural scientists, and social scientist. (a) Collaborative discussion between farmers and social scientists in Zhongwei, Ningxia (22 July 2007). (b) Collaborative discussion between natural scientists and social scientists at the Inner Mongolia Grassland Ecosystem Research Station, the CAS, Xilinhot, Inner Mongolia (29 July 2011). (c) Discussion meetings among farmers, natural scientists, social scientists, and government officials in Hongyue village, Jingtai County, Gansu (24 July 2007). (d) Discussion meetings among farmers, natural scientists, social scientists, and government officials in Hongshui Township, Jingtai County, Gansu (14 December 2013). These principles not only provide some useful outlines and references for researchers to further study collaboration between natural science, social science, and local knowledge in public governance but also offer important insights in practice and even a very detailed, operative, and integrative framework for practitioners (including policy makers and implementers) to resolve many practical problems about knowledge collaboration in China and other countries. As concrete guidelines, these principles could be used to design new institutional arrangements, improve existing institutions, transform unsuccessful institutions into successful ones, and diagnose problems for old and new arrangements about knowledge collaboration. However, this does not mean that this integrative framework is a ‘panacea’ (Ostrom 2007) for all the aforementioned problems. The list of design principles in fact is still quite speculative (Ostrom 1990) and more work is needed to further explain the causal mechanisms of each of the principles or to develop and test specific models and hypotheses relating to these principles and their subcomponents. Furthermore, the largest correlation coefficients of Principles 9, 2, and 3 showed that when consider these principles separately, these three were the most important among the nine principles and should be paid more attention. Especially, the extreme importance of financial support was also consistent with many previous findings (e.g. Campbell 1992; Yang and Wu 2010). All these suggested that although all the nine institutional principles are important, practitioners (including policy makers and implementers) should first give more attention to financial support, followed by the application and extension as well as the collaboration of the three types of knowledge. Moreover, the insignificant coefficients of the nine principles and the highest values of R and R2 of the ‘Enter’ regression suggested that although none of the coefficients of the nine principles were significant because of interactive effects among the nine principles, when all the nine principles were considered, the model would have the highest explanatory power. That is, if possible, all the nine principles should be considered together in order to get the highest performance of desertification control. However, if the nine principles could not be considered simultaneously either because of limitations of rationality or cost, the ‘Backward’ regression indicated that when Principles 1, 4, 5, 6, and 9 could be satisfied simultaneously, the model still could get the second highest values of R and R2, because the effects of Principles 2, 3, 7, and 8 on performance had been partly included in the effects of Principles 1, 4, 5, 6, and 9 on performance due to interactive effects between these two groups of principles. If even Principles 1, 4, 5, 6, and 9 could not be satisfied together, the ‘Stepwise’ regression suggested that when Principles 9, 6, and 3 could be satisfied simultaneously, the model still could get relatively high values of R and R2, because the effects of the rest six principles on performance had been partly included in the effects of Principles 9, 6, and 3 on performance due to interactive effects between the three satisfied principles and the rest six principles. Thus, in order to maximize the impact of knowledge collaboration on performance, I suggest practitioners to consider satisfying all the nine principles together as their first choice. If not, I suggest them to deem Principles 1, 4, 5, 6, and 9 as the first subgroup of the nine principles which should be satisfied and consider them from the principles with higher coefficients (e.g. Principles 9 and 4) to the ones with lower coefficients (e.g. Principles 6 and 1), and then also consider the other four principles if possible. When even Principles 1, 4, 5, 6, and 9 could not be satisfied together, they could move to satisfy Principles 9, 6, and 3 as the second subgroup of the nine principles, and then they could also get a relatively high governance performance. If Principles 9, 6, and 3 also could not be satisfied simultaneously, they then could only move to consider the nine principles one by one from the ones with higher correlation with performance to the ones with lower correlation. However, in either case, Principle 9 should be first considered. Certainly, all these should be further studied in the future. Furthermore, at any time, we must bear in mind that the reality is far more complex than the theoretical model. 6. Conclusion Collaboration between science, social science, and local knowledge is a necessary condition of modern collaborative knowledge-driven governance and also a global challenge in the contemporary era of complexity and uncertainty (Coen and Roberts 2012). Although the existing literature stresses the necessity and possibility of the collaboration of the three types of knowledge in public governance, the types and mechanisms of collaboration remained unclear. Based on an empirical study of desertification control in northern China, the current study not only found that knowledge collaboration influenced governance performance, natural science got the most attention but is least essential to success, and the types of collaboration with high levels of social science and local knowledge often led to high governance performance, but also highlighted nine institutional design principles for successful collaboration. In addition to important theoretical and practical principles for both worldwide researchers and practitioners, due to the complexity of modern governance and knowledge application, the findings of the study can also help people understand and resolve first- and second-order dilemmas of knowledge application (Ostrom 1990; Yang 2010) in public governance with a plural and open knowledge system (Popper 1992). That is, through collaboration, the dilemmas and costs of knowledge application could be reduced and the performance of knowledge application and governance could be improved. Together, these findings and reflections shed new light on collaboration between science, social science, and local knowledge not only in desertification control but also in other types of public governance in China and in other countries around the world. Researchers can also attempt to use the types of collaboration listed in Fig. 2 and the nine design principles listed in Table 5 as theoretical references to design their own research to study the collaboration of the three types of knowledge in any type of public governance in any country. However, it should be noted that although the study combined large-sample surveys, interviews, observations, and document analysis, the data analysis was mainly based on survey respondents’ perceptions and qualitative meta-analysis. Thus, further tests should be performed to validate the completeness and consistency of these findings with actual data on desertification and to validate the universal applicability of these findings, especially the statement that the natural science gets the most attention but is least essential to success and the nine design principles for successful collaboration among the three types of knowledge. Acknowledgements The study was supported by State Key Laboratory of Earth Surface Processes and Resource Ecology (2015-KF-14), the National Natural Science Foundation of China (71373016), and the Key Project of National Social Science Fund of China (14ZDB143). The author would like to thank Ms Jiali Yang for her contributions to an earlier version of the article. The author would also like to thank the late Professors Elinor Ostrom and Vincent Ostrom for their comments on and suggestions for this study. In particular, my conversations with Vincent helped me develop the original ideas for this article. Furthermore, special thanks go to the anonymous reviewers and the editors whose suggestions and comments have significantly improved the article. References Abbasi A., Hossain L., Leydesdorff L. ( 2012) ‘ Betweenness Centrality as a Driver of Preferential Attachment in the Evolution of Research Collaboration Networks’, Journal of Informetrics , 6/ 3: 403– 12. Google Scholar CrossRef Search ADS   Agranoff R., McGuire M. ( 2003) Collaborative Public Management: New Strategies for Local Governments . Washington: Georgetown University Press. Aikenhead G., Ogawa M. ( 2007) ‘ Indigenous Knowledge and Science Revisited’, Cultural Studies of Science Education , 2/ 3: 539– 620. Google Scholar CrossRef Search ADS   Albæk E. ( 1995) ‘ Between Knowledge and Power: Utilization of Social Science’, Policy Sciences , 28: 79– 100. Google Scholar CrossRef Search ADS   Ansell C., Gash A. ( 2008) ‘ Collaborative Governance in Theory and Practice’, Journal of Public Administration Research and Theory , 18/ 4: 543– 71. Google Scholar CrossRef Search ADS   Armitage D., Berkes F., Dale A. et al.   ( 2011) ‘ Co-Management and the Co-Production of Knowledge: Learning to Adapt in Canada’s Arctic’, Global Environmental Change , 21: 995– 1004. Google Scholar CrossRef Search ADS   Bala A., Joseph G. G. ( 2007) ‘ Indigenous Knowledge and Western Science: The Possibility of Dialogue’, Race & Class , 49/ 1: 39– 61. Google Scholar CrossRef Search ADS   Bardach E. ( 1987) ‘ From Practitioner Wisdom to Scholarly Knowledge and Back Again’, Journal of Policy Analysis and Management , 7/ 1: 188– 99. Google Scholar CrossRef Search ADS   Behn R. D. ( 1987) ‘ The Nature of Knowledge about Public Management and Teaching from Our Knowledge about Mess and Warfare’, Journal of Policy Analysis and Management , 7/ 1: 200– 12. Google Scholar CrossRef Search ADS   Berkes F. ( 1999) Sacred Ecology: Traditional Ecological Knowledge and Resource Management . London: Taylor and Francis. Berkes F. ( 2009) ‘ Evolution of Co-Management: Role of Knowledge Generation, Bridging Organizations and Social Learning’, Journal of Environmental Management , 90/ 5: 1692– 702. Google Scholar CrossRef Search ADS PubMed  Berkes F., Colding J., Folke C. ( 2000) ‘ Rediscovery of Traditional Ecological Knowledge as Adaptive Management’, Ecological Applications , 10/ 5: 1251– 62. Google Scholar CrossRef Search ADS   Borda-Rodriguez L., Johnson H. ( 2013) ‘ Development on My Terms: Development Consultants and Knowledge for Development’, Public Administration and Development , 33: 343– 56. Google Scholar CrossRef Search ADS   Bunker D. R. ( 1978) ‘ Organizing to Link Social Science with Public Policy Making’, Public Administration Review , 38/ 3: 223– 32. Google Scholar CrossRef Search ADS   Cairl J. G., Gallagher P. R. ( 1968) ‘ Government, Science, and Technology: A Bibliographical Essay’, Public Administration Review , 28/ 4: 373– 81. Google Scholar CrossRef Search ADS   Campbell A. ( 1992) Landcare in Australia: Taking the Long View in Tough Times , Third Annual Report. Canberra: National Soil Conservation Program. Caplan N. ( 1979) ‘ The Two-Communities Theory and Knowledge Utilization’, American Behavioral Scientist , 22/ 3: 459– 70. Google Scholar CrossRef Search ADS   Cash D. H., Borck J. C., Patt A. G. ( 2006) ‘ Countering the Loading-Dock Approach to Linking Science and Decision Making: Comparative Analysis of El Nino/Southern Oscillation (ENSO) Forecasting Systems’, Science, Technology & Human Values , 31/ 4: 465– 94. Google Scholar CrossRef Search ADS   Chasek P., Essahli W., Akhtar-Schuster M. et al.   ( 2011) ‘ Integrated Land Degradation Monitoring and Assessment: Horizontal Knowledge Management at the National and International Levels’, Land Degradation and Development , 22/ 2: 272– 84. Google Scholar CrossRef Search ADS   Chuang T. ( 1968) The Complete Works of Chuang Tzu (translated by Burton Watson).  New York: Columbia University Press. Coen D., Roberts A. ( 2012) ‘ A New Age of Uncertainty’, Governance: An International Journal of Policy, Administration, and Institutions , 25/ 1: 5– 9. Google Scholar CrossRef Search ADS   Corburn J. ( 2007) ‘ Community Knowledge in Environmental Health Science: Co-Producing Policy Expertise’, Environmental Science and Policy , 10/ 2: 150– 61. Google Scholar CrossRef Search ADS   Cox R. H., Béland D. ( 2013) ‘ Valence, Policy Ideas, and the Rise of Sustainability’, Governance: An International Journal of Policy, Administration, and Institutions , 26/ 2: 307– 28. Google Scholar CrossRef Search ADS   Cracknell B. E. ( 2001) ‘ Knowlege is All: Or is It? Some Reflections on Why the Acquisition of Knowledge, Focusing Particularly on Evaluation Activities, Does Not Always Lead To Action’, Public Administration and Development , 21: 371– 9. Google Scholar CrossRef Search ADS   Davis F. D., Marquis C. ( 2005) ‘ Prospects for Organization Theory in the Early Twenty-First Century: Instructional Fields and Mechanisms’, Organization Science , 16/ 4: 332– 43. Google Scholar CrossRef Search ADS   Daviter F. ( 2015) ‘ The Political Use of Knowledge in the Policy Process’, Policy Sciences , 48: 491– 505. Google Scholar CrossRef Search ADS   Deng X. ( 1993) Deng Xiaping Wenxuan . Bejing: Renming Chubanshe. Denny B. C. ( 1967) ‘ Science and Public Policy: A Symposium’, Public Administration Review , 27/ 2: 5– 97. Google Scholar CrossRef Search ADS   Dickison M. ( 2010) ‘ The Asymmetry between Science and Traditional Knowledge’, Journal of the Royal Society of New Zealand , 39/ 4: 171– 2. Google Scholar CrossRef Search ADS   Dietz T., Ostrom E., Stern P. C. ( 2003) ‘ The Struggle to Govern the Commons’, Science , 302/ 5652: 1907– 12. Google Scholar CrossRef Search ADS PubMed  Edelenbos J., van Buuren A., van Schie N. ( 2011) ‘ Co-Producing Knowledge: Joint Knowledge Production between Experts, Bureaucrats and Stakeholders in Dutch Water Management Projects’, Environmental Science & Policy , 14/ 6: 675– 84. Google Scholar CrossRef Search ADS   Emerson K., Nabatchi T., Balogh S. ( 2012) ‘ Integrative Framework for Collaborative Governance’, Journal of Public Administration Research and Theory , 22/ 1: 1– 29. Google Scholar CrossRef Search ADS   Fazey I., Fazey J. A., Salisbury J. G. et al.   ( 2006) ‘ The Nature and Role of Experiential Knowledge for Environmental Conservation’, Environmental Conservation , 33/ 1: 1– 10. Google Scholar CrossRef Search ADS   Fischer F. ( 2000) Citizens, Experts, and the Environment: The Politics of Local Knowledge . Durham and London: Duke University Press. Google Scholar CrossRef Search ADS   Fullen M. A., Mitchell D. J. ( 1994) ‘ Desertification and Reclamation in North-Central China’, Ambio , 23/ 2: 131– 5. Gessler N. ( 2015) ‘ Growing Artificial Societies—Social Science from the Bottom Up’, Artificial Life , 3/ 3: 237– 42. Google Scholar CrossRef Search ADS   Giebels D., van Buuren A., Edelnbos J. ( 2015) ‘ Using Knowledge in a Complex Decision-Making Process-Evidence and Principles from the Danish Houting Project’s Ecosystem-Based Management Approach’, Environmental Science & Policy , 47/ 3: 53– 67. Google Scholar CrossRef Search ADS   Ginger C. ( 2014) ‘ Integrating Knowledge, Interests and Values through Modelling in Participatory Processes: Dimensions of Legitimacy’, Journal of Environmental Planning and Management , 57/ 5: 643– 59. Google Scholar CrossRef Search ADS   Goudie A. S. ( 2009) ‘ Dust Storms: Recent Developments’, Journal of Environmental Management , 90/1: 89– 94. Google Scholar CrossRef Search ADS   Gray B. ( 1985) ‘ Conditions Facilitating Interorganizational Collaboration’, Human Relations , 38/ 10: 911– 36. Google Scholar CrossRef Search ADS   Guston D. H. ( 2001) ‘ Boundary Organizations in Environmental Policy and Science: An Introduction’, Science, Technology, & Human Values , 26/ 4: 399– 408. Google Scholar CrossRef Search ADS   Hayek F. A. ( 1945) ‘ The Use of Knowledge in Society’, The American Economic Review , 35/ 4: 519– 30. Hegger D. L. T., Lamers M., van Zeijl-Rozema A. et al.   ( 2012) ‘ Conceptualising Joint Knowledge Production in Regional Climate Change Adaptation Projects: Success Conditions and Levers for Action’, Environmental Science and Policy , 18/ 18: 52– 65. Google Scholar CrossRef Search ADS   Heink U., Marquard E., Heubach K. et al.   ( 2015) ‘ Conceptualizing Credibility, Relevance and Legitimacy for Evaluating the Effectiveness of Science-Policy Interfaces: Challenges and Opportunities’, Science and Public Policy , 42/ 5: 676– 89. Google Scholar CrossRef Search ADS   Hennemann S., Rybski D., Liefner I. ( 2012) ‘ The Myth of Global Science Collaboration—Collaboration Patterns in Epistemic Communities’, Journal of Informetrics , 6/ 2: 217– 25. Google Scholar CrossRef Search ADS   Hill R. ( 2006) ‘ The Effectiveness of Agreements and Protocols to Bridge between Indigenous and Non-Indigenous Toolboxes for Protected Area Management: A Case Study from the Wet Tropics of Queensland’, Society and Natural Resources , 19/ 7: 577– 90. Google Scholar CrossRef Search ADS   Hill R., Grant C., George M. et al.   ( 2012) ‘ A Typology of Indigenous Engagement in Australian Environmental Management: Implications for Knowledge Integration and Social-Ecological System Sustainability’, Ecology and Society , 17/ 1: 23. Google Scholar CrossRef Search ADS   Holm P., Goodsite M. E., Cloetingh S. et al.   ( 2013) ‘ Collaboration between the Natural, Social and Human Sciences in Global Change Research’, Environmental Science and Policy , 28: 25– 35. Google Scholar CrossRef Search ADS   Honadle B. W. ( 1994) ‘ Making Social Science Research Applicable: The Case of Agricultural Extension Research in Egypt’, Public Administration and Development , 14: 365– 75. Google Scholar CrossRef Search ADS   Hu J. ( 2012). Jiandingbuyi Yanzhe Zhonguo Tese Shehui Zhuyi Daolu Qianjin Wei Quanmian Jian Cheng Xiaokang Shehui Er Fendou . Beijing: People’s Publishing House (Remin Chuban She). Hurwicz L., Reiter S. ( 2008) Designing Economic Mechanisms . New York: Cambridge University Press. Innes J. E., Booher D. E. ( 2010) Planning with complexity: An Introduction To Collaborative Rationality for Public Policy . New York: Routledge. Ison R., Röling N., Watson D. ( 2007) ‘ Challenges to Science and Society in the Sustainable Management and Use of Water: Investigating the Role of Social Learning’, Environmental Science and Policy , 10/ 6: 499– 511. Google Scholar CrossRef Search ADS   Katz J. S., Martin B. R. ( 1997) ‘ What Is Research Collaboration?’, Research Policy , 26/ 1: 1– 18. Google Scholar CrossRef Search ADS   Ko K. ( 2013) ‘ Knowledge Accumulation in Asian Public Administration Research: A Critical Review’, Public Administration and Development , 33: 320– 4. Google Scholar CrossRef Search ADS   Lambright W. H. ( 2008) ‘ Government and Science: A Troubled, Critical Relationship and What Can Be Done about It’, Public Administration Review , 68: 5– 18. Google Scholar CrossRef Search ADS   Landry R., Amara N., Lamari M. ( 2001) ‘ Utilization of Social Science Research Knowledge in Canada’, Research Policy , 30/ 2: 333– 49. Google Scholar CrossRef Search ADS   Landry R., Lamari M., Amara N. ( 2003) ‘ The Extent and Determinants of the Utilization of University Research in Government Agencies’, Public Administration Review , 63/ 2: 192– 205. Google Scholar CrossRef Search ADS   Lasswell H. D. ( 1956) ‘ The Political Science of Science: An Inquiry into the Possible Reconciliation of Mastery and Freedom’, American Political Science Review , 50/ 4: 961– 79. Google Scholar CrossRef Search ADS   Leach W. D., Pelkey N. W., Sabatier P. A. ( 2002) ‘ Stakeholder Partnerships As Collaborative Policymaking: Evaluation Criteria Applied to Watershed Management in California and Washington’, Journal of Policy Analysis and Management , 21/ 4: 645– 70. Google Scholar CrossRef Search ADS   Lubell M. ( 2004) ‘ Collaborative Environmental Institutions: All Talk and No Action?’, Journal of Policy Analysis and Management , 23: 549– 73. Google Scholar CrossRef Search ADS   Maclean K., Cullen L. ( 2010) ‘ Research Methodologies for the Co-Production of Knowledge for Environmental Management in Australia’, Journal of the Royal Society of New Zealand , 39/ 4: 205– 8. Google Scholar CrossRef Search ADS   McGuire M. ( 2006) ‘ Collaborative Public Management: Assessing What We Know and How We Know It’, Public Administration Review , 66: 33– 43. Google Scholar CrossRef Search ADS   Meyer D. R., Cancian M., Nam K. ( 2007) ‘ Welfare and Child Support Program Knowledge Gaps Reduce Program Effectiveness’, Journal of Policy Analysis and Management , 26/ 3: 575– 97. Google Scholar CrossRef Search ADS   Miller C. ( 2001) ‘ Hybrid Management: Boundary Organizations, Science Policy, and Environmental Governance in the Climate Regime’, Science, Technology, and Human Values , 26: 478– 500. Google Scholar CrossRef Search ADS   Moore E. A., Koontz T. M. ( 2003) ‘ A Typology of Collaborative Watershed Groups: Citizen-Based, Agency-Based, and Mixed Partnerships’, Society & Natural Resources , 16/ 5: 451– 60. Google Scholar CrossRef Search ADS   Moynihan D. P., Fernandez S., Kim S. et al.   ( 2010) ‘ Performance Regimes amidst Governance Complexity’, Journal of Public Administration Research and Theory , 21 (Suppl 1): i141– 55. Newman M. E. J. ( 2001) ‘ The Structure of Scientific Collaboration Networks’, Proceedings of the National Academy of Sciences , 98/ 2: 404– 9. Google Scholar CrossRef Search ADS   O’Leary R., Catherine G., Lisa B. B. ( 2006) ‘ Introduction to the Symposium on Collaborative Public Management, Special Issue’, Public Administration Review , 66/ s1: 6– 9. Google Scholar CrossRef Search ADS   Olmos-Peñuela J., Castro-Martínez E., D’Este P. ( 2014) ‘ Knowledge Transfer Activities in Social Sciences and Humanities: Explaining the Interactions of Research Groups with Non-Academic Agents’, Research Policy , 43: 696– 706. Google Scholar CrossRef Search ADS   Osborne S. P. ( 2006) ‘ The New Public Governance?’, Public Management Review , 8/ 3: 377– 87. Google Scholar CrossRef Search ADS   Ostrom E. ( 1990) Governing the Commons: The Evolution of Institutions for Collective Action . Cambridge, NY: Cambridge University Press. Google Scholar CrossRef Search ADS   Ostrom E. ( 2005) Understanding Institutional Diversity . Princeton, NJ: Princeton University Press. Ostrom E. ( 2007) ‘ A Diagnostic Approach for Going Beyond Panaceas’, Proceedings of the National Academy of Sciences , 104/ 39: 15181– 7. Google Scholar CrossRef Search ADS   O'Toole L. J., Meier K. J. ( 1999) ‘ Modeling the Impact of Public Management: The Implications of Structural Context’, Journal of Public Administration Research and Theory , 9/ 4: 505– 26. Google Scholar CrossRef Search ADS   Papageorgiou J. C. ( 1994) ‘ The Role of the Public Sector in Introducing Operations Research/Management Science within Developing Countries’, Public Administration and Development , 14: 293– 312. Google Scholar CrossRef Search ADS   Patton M. Q. ( 1987) How to Use Qualitative Methods in Evaluation . Newbury Park, CA: Sage. Perry J. L. ( 2012) ‘ How Can We Improve Our Science to Generate More Usable Knowledge for Public Professionals?’, Public Administration Review , 72/ 4: 479– 82. Google Scholar CrossRef Search ADS   Podsakoff J. L., Organ D. W. ( 1986) ‘ Self-Reports in Organization Research: Problems and Prospects’, Journal of Management , 12/ 4: 531– 44. Google Scholar CrossRef Search ADS   Pohl C. ( 2008) ‘ From Science to Policy through Transdisciplinary Research’, Environmental Science and Policy , 11/ 1: 46– 53. Google Scholar CrossRef Search ADS   Popper K. R. ( 1992) In Search of A Better World: Lectures and Essays from Thirty Years . London and New York: Routledge. Poteete A. R., Janssen M. A., Ostrom E. ( 2010) Working Together: Collective Action, the Commons, and Multiple Methods in Practice . Princeton, NJ: Princeton University Press. Poteete A. R., Ostrom E. ( 2008) ‘ Fifteen Years of Empirical Research on Collective Action in Natural Resource Management: Struggling to Build Large-N Databases Based on Qualitative Research’, World Development , 36/ 1: 176– 95. Google Scholar CrossRef Search ADS   Powell W. W., Koput K. W., Smith-Doerr L. ( 1996) ‘ Interorganizational Collaboration and the Cocus of innovation: Networks of Learning in Biotechnology’, Administrative Science Quarterly , 41/ 1: 116– 45. Google Scholar CrossRef Search ADS   Reed M. S., Dougill Y. A. J., Taylor M. J. ( 2007) ‘ Integrating Local and Scientific Knowledge for Adaptation to Land Degradation: Kalahari Rangeland Management Options’, Land Degradation and Development , 18/ 3: 249– 68. Google Scholar CrossRef Search ADS   Reynolds J. F., Smith D. M. S., Lambin E. F. et al.   ( 2007) ‘ Global Desertification: Building A Science For Dryland Development’, Science , 316/ 5826: 847– 51. Google Scholar CrossRef Search ADS PubMed  Rudd M. A. ( 2015) ‘ Scientists’ Framing of the Ocean Science–Policy Interface’, Global Environmental Change , 33/ 6: 44– 60. Google Scholar CrossRef Search ADS   Ruttan V. W. ( 1984) ‘ Social Science Knowledge and Institutional Change’, American Journal of Agricultural Economics , 66/ 5: 549– 59. Google Scholar CrossRef Search ADS   Scott T. ( 2015) ‘ Does Collaboration Make Any Difference? Linking Collaborative Governance to Environmental Outcomes’, Journal of Policy Analysis and Management , 34/ 3: 537– 66. Google Scholar CrossRef Search ADS   [SFA] State Forestry Administration, P. R. China. ( 2011) A Bulletin of Status Quo of Desertification and Sandification in China <http://www.forestry.gov.cn/uploadfile/main/2011-1/file/2011-1-5-59315b03587b4d7793d5d9c3aae7ca86.pdf> accessed 24 July 2014. Smith L. T. ( 1999) Decolonizing Methodologies: Research and Indigenous Peoples . London: Zed Books Ltd. Star S. L., Griesemer J. R. ( 1989) ‘ Institutional Ecology, “Translations” and Boundary Objects: Amateurs and Professionals in Berkeley’s Museum of Vertebrate Zoology, 1907–39’, Social Studies of Science , 19/ 3: 387– 420. Google Scholar CrossRef Search ADS   Stone D. ( 2009) ‘ Rapid Knowledge: ‘Bringing Research and Policy’ at the Overseas Development Institute’, Public Administration and Development , 29: 303– 15. Google Scholar CrossRef Search ADS   Stringer L. C., Dyer J. C., Reed M. S. et al.   ( 2009) ‘ Adaptations to Climate Change, Drought and Desertification; Insights to Enhance Policy in Southern Africa’, Environmental Science and Policy , 12: 748– 65. Google Scholar CrossRef Search ADS   Sun Z. ( 2011). Jianguo Fanglue . Beijing: Zhongguo Changan Chubanshe. Sun D., Dawson R., Li B. ( 2006) ‘ Agricultural Causes of Desertification Risk in Minqin, China’, Journal of Environmental Management , 79: 348– 56. Google Scholar PubMed  Taylor B., de Loë R. C. ( 2012) ‘ Conceptualizations of Local Knowledge in Collaborative Environmental Governance’, Geoforum , 43/ 6: 1207– 17. Google Scholar CrossRef Search ADS   Thomas D. S. G. ( 1997) ‘ Science and the Desertification Debate’, Journal of Arid Environments , 37/ 4: 599– 608. Google Scholar CrossRef Search ADS   Thomas D. S. G., Twyman C. ( 2004) ‘ Good or Bad Range Land? Hybrid Knowledge, Science and Local Understandings of Vegetation Dynamics in the Kalahari’, Land Degradation and Development , 15/ 3: 215– 31. Google Scholar CrossRef Search ADS   Torgerson D. ( 1986) ‘ Between knowledge and politics: Three faces of policy analysis’, Policy Sciences , 19: 33– 59. Google Scholar CrossRef Search ADS   Tschakert P. ( 2007) ‘ Views from the Vulnerable: Understanding Climatic and Other Stressors in the Sahel’, Global Environmental Change , 17/ 3-4: 381– 96. Google Scholar CrossRef Search ADS   [ UN] United Nations. ( 1992) Managing Fragile Ecosystems: Combating Desertification and Drought , Chapter 12 of Agenda 2. New York, NY: United Nations. Van der Meulen B. ( 1998) ‘ Science Policies as Principal–Agent Games Institutionalization and Path Dependency in the Relation between Government and Science’, Research Policy , 27: 397– 414. Google Scholar CrossRef Search ADS   Wagner C. S., Leydesdorff L. ( 2005) ‘ Network Structure, Self-Organization, and the Growth of International Collaboration in Science’, Research Policy , 34/ 10: 1608– 18. Google Scholar CrossRef Search ADS   Wang T., Wu W., Xue X. et al.   ( 2004) ‘ Spatial-Temporal Changes of Sandy Desertified Land During Last 5 Decades in Northern China’, Acta Geographica Sinica , 59/ 2: 203– 12. Watson A. ( 2013) ‘ Misunderstanding the “Nature” of Co-Management: A Geography of Regulatory Science and Indigenous Knowledge (IK)’, Environmental Management , 52/ 5: 1085– 102. Google Scholar CrossRef Search ADS PubMed  Weiss J. A., Gruber J. E. ( 1984) ‘ Using Knowledge for Control in Fragmented Policy Arenas’, Journal of Policy Analysis and Management , 3/ 2: 225– 47. Google Scholar CrossRef Search ADS   Weiss K., Hamann M., Marsh H. ( 2013) ‘ Bridging Knowledges: Understanding and Applying Indigenous and Western Scientific Knowledge for Marine Wildlife Management’, Society & Natural Resources , 26/ 3: 285– 302. Google Scholar CrossRef Search ADS   Wellens J. ( 1975) ‘ Levels of Participation’, Industrial and Commercial Training , 73/ 3: 93– 6. Google Scholar CrossRef Search ADS   Winslow M. D., Vogt J. V., Thomas R. J. et al.   ( 2011) ‘ Editorial. Science for Improving the Monitoring and Assessment of Dryland Degradation’, Land Degradation & Development , 22/ 2: 145– 9. Google Scholar CrossRef Search ADS   World Bank ( 1998) World Development Report: Knowledge for Development . New York: Oxford University Press. Wu J. ( 2001). ‘Desertification’, in Robinson R (ed.) Plant Sciences for Students , New York: Macmillan Reference. Yang L. ( 2010) ‘ The Knowledge Dilemma and Its Resolution Methods: Game Theoretical Models of Environmental Governance’, Chinese Public Administration , 10/ 10: 112– 6. Yang J. ( 2013) ‘ Roles and Mechanisms of Multiple Knowledge Collaboration in Institutional Changes of Desertification Control’ , Thesis, Beihang University, Beijing. Yang L. ( 2015) ‘ Local Knowledge, Science, and Institutional Change: The Case of Desertification Control in Northern China’, Environmental Management , 55/ 3: 616– 33. Google Scholar CrossRef Search ADS PubMed  Yang L., Lan Z., Wu J. ( 2010) ‘ Roles of Scholars in the Practice of Combating-Desertification: A Case Study in Northwest China’, Environmental Management , 46/ 2: 154– 66. Google Scholar CrossRef Search ADS PubMed  Yang L., Li C. ( 2015) ‘ Types and Mechanisms of Science-Driven Institutional Change: The case of Desertification Control in Northern China’, Environmental Policy and Governance , 25/ 1: 16– 35. Google Scholar CrossRef Search ADS   Yang L., Wu J. ( 2010) ‘ Seven Design Principles for Promoting Scholars’ Participation In Combating Desertification’, International Journal of Sustainable Development and World Ecology , 17/ 2: 109– 19. Google Scholar CrossRef Search ADS   Yang L., Wu J., Shen P. ( 2013) ‘ Roles of Science in Institutional Changes: The Case of Desertification Control in China’, Environmental Science and Policy , 27: 32– 54. Google Scholar CrossRef Search ADS   Zhao B. ( 2009). Zhengzhixue He Hexie Shehui . Beijing: Beijing University Press. Zurba M., Ross H., Izurieta A. et al.   ( 2012) ‘ Building Co-Management as a Process: Problem Solving through Partnerships in Aboriginal Country, Australia’, Environmental Management , 49/ 6: 1130– 42. Google Scholar CrossRef Search ADS PubMed  Appendix A Perceived factors influencing the application of the three types of knowledge as rated by survey respondents (2011). CountiesNine groups of factors and performance  Linze(%)  Minqin(%)  Zhongwei(%)  Yanchi(%)  Dengkou(%)  Ejin Horo(%)  Xinbaerhuzuo(%)  Xilinhot(%)  Naiman(%)  Duolun(%)  Wengniute(%)  Aohan(%)  Factor 1: Level of knowledge (Average)  36.2a  32.3  38.5  21.9  32.5  24.3  38.8  33.4  11.3  26.1  18.3  32.3  Low levels of science and technology development  31.3b  29.7  37.1  21.4  31.4  22.7  32.1  32.2  12.8  22.8  23.2  26.3  Low levels of social science development  37.5  33.9  40.0  21.9  36.2  26.1  38.8  36.1  10.6  29.7  12.6  35.5  Low values of local knowledge  39.9  33.2  38.5  22.4  30.0  24.1  45.6  31.9  10.6  25.9  19.0  35.3  Factor 2: Knowledge application  37.7  45.4  41.3  25.1  42.2  32.0  40.7  39.2  11  34.7  21.2  38.3  Lack of science suited to local conditions  33.3  48.9  41.0  27.2  44.9  29.3  43.4  41.8  13.1  31.7  19.1  39.7  Lack of effective systems of science and technology transformation, extension and application  36.6  42.2  39.6  22.1  44.4  34.8  41.7  39.1  9.4  38.0  30.0  35.2  Lack of social science suited to local conditions and problems  35.4  49.7  43.8  26.5  37.7  32.9  37.5  37.1  12.4  34.3  19.4  41.0  Lack of effective systems of social science application  45.3  43.2  41.7  24.5  46.6  32.9  44.2  40.1  12.2  37.9  21.7  36.4  Unsuitability of Local knowledge to local conditions  30.4  44.5  38.6  26.2  32.9  31.9  39.4  34.8  8.7  31.0  17.8  38.2  Lack of effective systems of local knowledge refining application, and extension  45.1  43.8  43.0  24.2  46.8  30.3  37.9  42.1  10.1  35.5  19.6  39.1  Factor 3: Relationships among knowledge types  32.1  40.8  38.8  24.8  34.4  34.5  40.2  32.3  11.4  27.4  18.2  36.3  Conflicts between local knowledge and science  32.1  40.8  38.8  24.8  34.4  34.5  40.2  32.3  11.4  27.4  18.2  36.3  Factor 4: Endeavors of knowledge possessors  34.5  42.7  38.9  23.9  37  30.9  38.4  35.1  16  29.2  22.8  36.4  Scientists’ inadequate knowledge of local conditions  36.9  48.4  42.6  25.8  38.1  42.4  38.8  39.8  14.3  34.8  31.7  37.1  Scientists’ sabotage  25.4  36.2  33.7  24.5  28.3  27.7  34.1  24.5  12.2  23.9  19.1  29.1  Social scientists’ inadequate knowledge of local conditions  41.7  51.0  44.0  25.8  44.5  31.9  35.2  39.2  34.6  31.6  22.5  41.2  Social scientists’ sabotage  33.3  44.5  33.9  24.4  35.7  27.7  37.2  30.5  12.4  24.6  15.9  38.2  Low capability of local farmer and herdsmen for applying local knowledge  39.4  38.5  41.8  22.8  39.3  27.5  36.7  40.3  12.8  35.3  27.2  38.1  The sabotage of local people having local knowledge  30.5  37.5  37.3  20.3  36.1  28.4  48.6  36.5  9.6  24.8  20.3  34.8  Factor 5: Relationship among knowledge possessors  32.9  47.8  37.9  26.4  36.9  34.5  40.7  33.9  16.9  30.8  19.2  37.4  Lack of effective communication between scientists and local people, especially farmers and herdsmen  30.9  50.0  42.8  25.9  39.3  34.0  46.7  32.0  40.0  30.0  17.3  35.9  Lack of respect by government and scientists for local people  28.5  42.7  35.3  26.1  28.1  34.3  35.1  28.2  12.8  25.9  20.3  34.4  Lack of effective communication between social scientists and local people, especially farmers and herdsmen  40.4  54.8  40.4  26.4  42.5  35.5  44.4  40.9  14.5  33.2  21.1  39.8  Lack of respect by government and social scientists for local people  32.7  44.9  36.4  25.6  33.9  32.9  34.7  29.4  10.6  26.8  16.3  36.1  Lack of effective communication between other scientific workers and local people, especially farmers and herdsmen  37.0  48.5  40.5  27.3  42.5  35.2  44.7  41.2  11.7  38.1  22.3  39.3  Lack of respect by government and other scientific workers for local people  27.9  45.8  32.2  27.1  35.2  35.2  38.8  31.7  11.9  30.9  17.9  38.8  Factor 6: Support from other social actors  38.1  40.2  38.3  22.6  36.9  31.9  40.1  34.5  12.1  29.2  18.9  34.8  Inadequate attention of society to the function of science and technology in desertification control  36.8  41.6  47.7  21.6  39.7  34.0  35.6  35.9  9.4  30.6  22.6  36.7  Low quality of farmers and pastoralists in applying science  41.6  41.9  40.0  18.8  39.6  33.3  41.8  37.5  12.3  32.0  22.9  33.8  Local people’s noncooperation in applying science  32.3  29.7  33.3  22.0  23.9  30.0  44.2  26.2  14.0  21.1  13.3  31.7  Inadequate attention of society to the development and application of social science  48.0  47.5  38.1  19.5  44.1  34.5  41.7  37.2  12.1  35.0  16.2  36.9  Low quality of farmers and pastoralists in applying social science  37.1  38.4  36.5  24.7  38.2  34.5  47.5  40.9  11.9  32.1  21.9  33.1  Local people’s noncooperation in applying social science  30.1  39.1  35.5  27.0  30.8  30.3  38.5  28.1  11.4  21.5  18.3  36.9  Inadequate attention of society to the function and application of local knowledge  43.3  47.1  41.0  19.4  45.5  29.6  29.6  39.4  13.4  33.9  18.7  33.5  Local people’s noncooperation in applying local knowledge  35.8  35.9  34.6  27.9  33.2  28.9  42.1  30.6  11.9  27.0  17.0  35.5  Factor 7: Local and central government  38  48.8  41  27  43.9  36.8  41  39.3  14.7  32.7  24.3  41.9  Government bureaucracy and corruption in applying science  37.2  56.5  41.0  27.2  46.7  34.0  39.2  46.6  15.9  36.3  31.1  42.2  Inadequate attention of local government to the function of science and technology in desertification control  34.3  47.3  39.9  27.8  41.2  34.0  48.6  34.2  14.3  33.2  25.1  39.2  Inadequate attention of the central government to the function of science and technology in desertification control  33.2  36.5  42.1  30.6  36.7  34.8  35.0  32.0  15.9  25.6  13.2  41.1  Government bureaucracy and corruption in applying social science  37.9  55.9  38.4  26.6  45.2  31.2  33.2  42.4  12.4  33.4  28.9  43.0  Inadequate attention of local government to the function of social science in desertification control  39.1  49.5  37.5  26.3  48.2  40.1  39.8  40.7  15.5  33.9  23.2  39.1  Inadequate attention of the central government to the function of social science in desertification control  38.2  39.6  40.2  25.8  43.2  39.4  39.6  35.0  17.1  27.9  16.3  43.6  Government bureaucracy and corruption in applying local knowledge  38.8  57.8  41.7  31.7  45.0  44.4  45.1  42.0  12.7  35.9  36.3  43.5  Inadequate attention of local government to the function of local knowledge in desertification control  40.2  52.1  42.6  24.1  47.1  38.7  44.7  42.7  12.8  35.7  24.0  40.1  Inadequate attention of the central government to the function of local knowledge in desertification control  43.0  44.0  45.2  23.3  42.1  34.5  43.7  37.7  16.1  32.1  20.4  45.3  Factor 8: Laws and regulations  42  42.1  41.2  21.4  42.4  33.9  36.5  38.7  15  36.8  21.1  36.4  Imperfect laws and regulations of science and technology application  38.4  36.5  42.4  21.9  38.4  30.5  35.5  37.6  17.8  35.7  23.3  34.8  Imperfect laws and regulations of social science application  42.5  45.1  39.9  21.3  43.6  34.5  38.5  38.5  14.9  39.6  19.7  37.2  Imperfect laws and regulations of local knowledge application  45.2  44.7  41.2  21.1  45.2  36.6  35.4  39.9  12.2  35.2  20.3  37.2  Factor 9: Financial support  41.7  43.3  39.1  21.8  43.9  30.2  39.3  37.9  12  33.4  24.6  38.1  Low financial support for science and technology application  42.2  43.2  34.3  16.2  44.9  24.1  35.6  33.5  13.6  29.0  23.4  33.2  Low financial support for social science application  40.7  42.1  38.3  25.0  45.5  34.0  40.6  38.3  11.4  34.1  25.7  40.8  Low financial support for local knowledge application  42.1  44.7  44.8  24.2  41.4  32.4  41.6  41.9  10.9  37.2  24.8  40.3  Performance  46.2  14.3  45.7  29.1  27.7  28  55.2  29.6  20.8  34.3  17.1  18.4  CountiesNine groups of factors and performance  Linze(%)  Minqin(%)  Zhongwei(%)  Yanchi(%)  Dengkou(%)  Ejin Horo(%)  Xinbaerhuzuo(%)  Xilinhot(%)  Naiman(%)  Duolun(%)  Wengniute(%)  Aohan(%)  Factor 1: Level of knowledge (Average)  36.2a  32.3  38.5  21.9  32.5  24.3  38.8  33.4  11.3  26.1  18.3  32.3  Low levels of science and technology development  31.3b  29.7  37.1  21.4  31.4  22.7  32.1  32.2  12.8  22.8  23.2  26.3  Low levels of social science development  37.5  33.9  40.0  21.9  36.2  26.1  38.8  36.1  10.6  29.7  12.6  35.5  Low values of local knowledge  39.9  33.2  38.5  22.4  30.0  24.1  45.6  31.9  10.6  25.9  19.0  35.3  Factor 2: Knowledge application  37.7  45.4  41.3  25.1  42.2  32.0  40.7  39.2  11  34.7  21.2  38.3  Lack of science suited to local conditions  33.3  48.9  41.0  27.2  44.9  29.3  43.4  41.8  13.1  31.7  19.1  39.7  Lack of effective systems of science and technology transformation, extension and application  36.6  42.2  39.6  22.1  44.4  34.8  41.7  39.1  9.4  38.0  30.0  35.2  Lack of social science suited to local conditions and problems  35.4  49.7  43.8  26.5  37.7  32.9  37.5  37.1  12.4  34.3  19.4  41.0  Lack of effective systems of social science application  45.3  43.2  41.7  24.5  46.6  32.9  44.2  40.1  12.2  37.9  21.7  36.4  Unsuitability of Local knowledge to local conditions  30.4  44.5  38.6  26.2  32.9  31.9  39.4  34.8  8.7  31.0  17.8  38.2  Lack of effective systems of local knowledge refining application, and extension  45.1  43.8  43.0  24.2  46.8  30.3  37.9  42.1  10.1  35.5  19.6  39.1  Factor 3: Relationships among knowledge types  32.1  40.8  38.8  24.8  34.4  34.5  40.2  32.3  11.4  27.4  18.2  36.3  Conflicts between local knowledge and science  32.1  40.8  38.8  24.8  34.4  34.5  40.2  32.3  11.4  27.4  18.2  36.3  Factor 4: Endeavors of knowledge possessors  34.5  42.7  38.9  23.9  37  30.9  38.4  35.1  16  29.2  22.8  36.4  Scientists’ inadequate knowledge of local conditions  36.9  48.4  42.6  25.8  38.1  42.4  38.8  39.8  14.3  34.8  31.7  37.1  Scientists’ sabotage  25.4  36.2  33.7  24.5  28.3  27.7  34.1  24.5  12.2  23.9  19.1  29.1  Social scientists’ inadequate knowledge of local conditions  41.7  51.0  44.0  25.8  44.5  31.9  35.2  39.2  34.6  31.6  22.5  41.2  Social scientists’ sabotage  33.3  44.5  33.9  24.4  35.7  27.7  37.2  30.5  12.4  24.6  15.9  38.2  Low capability of local farmer and herdsmen for applying local knowledge  39.4  38.5  41.8  22.8  39.3  27.5  36.7  40.3  12.8  35.3  27.2  38.1  The sabotage of local people having local knowledge  30.5  37.5  37.3  20.3  36.1  28.4  48.6  36.5  9.6  24.8  20.3  34.8  Factor 5: Relationship among knowledge possessors  32.9  47.8  37.9  26.4  36.9  34.5  40.7  33.9  16.9  30.8  19.2  37.4  Lack of effective communication between scientists and local people, especially farmers and herdsmen  30.9  50.0  42.8  25.9  39.3  34.0  46.7  32.0  40.0  30.0  17.3  35.9  Lack of respect by government and scientists for local people  28.5  42.7  35.3  26.1  28.1  34.3  35.1  28.2  12.8  25.9  20.3  34.4  Lack of effective communication between social scientists and local people, especially farmers and herdsmen  40.4  54.8  40.4  26.4  42.5  35.5  44.4  40.9  14.5  33.2  21.1  39.8  Lack of respect by government and social scientists for local people  32.7  44.9  36.4  25.6  33.9  32.9  34.7  29.4  10.6  26.8  16.3  36.1  Lack of effective communication between other scientific workers and local people, especially farmers and herdsmen  37.0  48.5  40.5  27.3  42.5  35.2  44.7  41.2  11.7  38.1  22.3  39.3  Lack of respect by government and other scientific workers for local people  27.9  45.8  32.2  27.1  35.2  35.2  38.8  31.7  11.9  30.9  17.9  38.8  Factor 6: Support from other social actors  38.1  40.2  38.3  22.6  36.9  31.9  40.1  34.5  12.1  29.2  18.9  34.8  Inadequate attention of society to the function of science and technology in desertification control  36.8  41.6  47.7  21.6  39.7  34.0  35.6  35.9  9.4  30.6  22.6  36.7  Low quality of farmers and pastoralists in applying science  41.6  41.9  40.0  18.8  39.6  33.3  41.8  37.5  12.3  32.0  22.9  33.8  Local people’s noncooperation in applying science  32.3  29.7  33.3  22.0  23.9  30.0  44.2  26.2  14.0  21.1  13.3  31.7  Inadequate attention of society to the development and application of social science  48.0  47.5  38.1  19.5  44.1  34.5  41.7  37.2  12.1  35.0  16.2  36.9  Low quality of farmers and pastoralists in applying social science  37.1  38.4  36.5  24.7  38.2  34.5  47.5  40.9  11.9  32.1  21.9  33.1  Local people’s noncooperation in applying social science  30.1  39.1  35.5  27.0  30.8  30.3  38.5  28.1  11.4  21.5  18.3  36.9  Inadequate attention of society to the function and application of local knowledge  43.3  47.1  41.0  19.4  45.5  29.6  29.6  39.4  13.4  33.9  18.7  33.5  Local people’s noncooperation in applying local knowledge  35.8  35.9  34.6  27.9  33.2  28.9  42.1  30.6  11.9  27.0  17.0  35.5  Factor 7: Local and central government  38  48.8  41  27  43.9  36.8  41  39.3  14.7  32.7  24.3  41.9  Government bureaucracy and corruption in applying science  37.2  56.5  41.0  27.2  46.7  34.0  39.2  46.6  15.9  36.3  31.1  42.2  Inadequate attention of local government to the function of science and technology in desertification control  34.3  47.3  39.9  27.8  41.2  34.0  48.6  34.2  14.3  33.2  25.1  39.2  Inadequate attention of the central government to the function of science and technology in desertification control  33.2  36.5  42.1  30.6  36.7  34.8  35.0  32.0  15.9  25.6  13.2  41.1  Government bureaucracy and corruption in applying social science  37.9  55.9  38.4  26.6  45.2  31.2  33.2  42.4  12.4  33.4  28.9  43.0  Inadequate attention of local government to the function of social science in desertification control  39.1  49.5  37.5  26.3  48.2  40.1  39.8  40.7  15.5  33.9  23.2  39.1  Inadequate attention of the central government to the function of social science in desertification control  38.2  39.6  40.2  25.8  43.2  39.4  39.6  35.0  17.1  27.9  16.3  43.6  Government bureaucracy and corruption in applying local knowledge  38.8  57.8  41.7  31.7  45.0  44.4  45.1  42.0  12.7  35.9  36.3  43.5  Inadequate attention of local government to the function of local knowledge in desertification control  40.2  52.1  42.6  24.1  47.1  38.7  44.7  42.7  12.8  35.7  24.0  40.1  Inadequate attention of the central government to the function of local knowledge in desertification control  43.0  44.0  45.2  23.3  42.1  34.5  43.7  37.7  16.1  32.1  20.4  45.3  Factor 8: Laws and regulations  42  42.1  41.2  21.4  42.4  33.9  36.5  38.7  15  36.8  21.1  36.4  Imperfect laws and regulations of science and technology application  38.4  36.5  42.4  21.9  38.4  30.5  35.5  37.6  17.8  35.7  23.3  34.8  Imperfect laws and regulations of social science application  42.5  45.1  39.9  21.3  43.6  34.5  38.5  38.5  14.9  39.6  19.7  37.2  Imperfect laws and regulations of local knowledge application  45.2  44.7  41.2  21.1  45.2  36.6  35.4  39.9  12.2  35.2  20.3  37.2  Factor 9: Financial support  41.7  43.3  39.1  21.8  43.9  30.2  39.3  37.9  12  33.4  24.6  38.1  Low financial support for science and technology application  42.2  43.2  34.3  16.2  44.9  24.1  35.6  33.5  13.6  29.0  23.4  33.2  Low financial support for social science application  40.7  42.1  38.3  25.0  45.5  34.0  40.6  38.3  11.4  34.1  25.7  40.8  Low financial support for local knowledge application  42.1  44.7  44.8  24.2  41.4  32.4  41.6  41.9  10.9  37.2  24.8  40.3  Performance  46.2  14.3  45.7  29.1  27.7  28  55.2  29.6  20.8  34.3  17.1  18.4  a The average percentages of subproblems. b The percentages of ‘very large’ and ‘large’ rated by survey respondents. Appendix B Nine factors for successful collaboration between natural science, social science, and local knowledge and the partial coefficients (controlling for values of local knowledge) of the performance of desertification control in the twelve field study cases in northern China (2011). Nine factors  Coefficients with performance  (significance)  Group 1: Knowledge itself    F1. Level of knowledge  −0.170***  The level of developments and improvements in the three types of knowledge  (0.001)  F2. Knowledge application  −0.569***  The refinement, transformation, adaption, application, and extension of the three types of knowledge  (0.000)  F3. Relationships among knowledge types  −0.528***  Complementation and coordination among three types of knowledge  (0.000)  Group 2: knowledge possessors and others social actors    F4. Endeavors of knowledge possessors  −0.696***  Capability and endeavors of knowledge possessors on the application and extension of the three types of knowledge  (0.000)  F5. Relationship between knowledge possessors  −0.634***  Communication and collaboration among knowledge possessors and other social actors  (0.000)  F6. Supports from other social actors  −0.485***  Support from other social actors and continued improvements on their attention and understanding of the three types of knowledge as well as their capabilities  (0.000)  Group 3: External support    F7. Local and central government  −0.726***  Support and guidance by governments at different hierarchical levels  (0.000)  F8. Laws and regulations  −0.355***  Institutional support from laws and regulations  (0.000)  F9. Financial support  −0.561***  Financial support from government and other social actors  (0.000)  Nine factors  Coefficients with performance  (significance)  Group 1: Knowledge itself    F1. Level of knowledge  −0.170***  The level of developments and improvements in the three types of knowledge  (0.001)  F2. Knowledge application  −0.569***  The refinement, transformation, adaption, application, and extension of the three types of knowledge  (0.000)  F3. Relationships among knowledge types  −0.528***  Complementation and coordination among three types of knowledge  (0.000)  Group 2: knowledge possessors and others social actors    F4. Endeavors of knowledge possessors  −0.696***  Capability and endeavors of knowledge possessors on the application and extension of the three types of knowledge  (0.000)  F5. Relationship between knowledge possessors  −0.634***  Communication and collaboration among knowledge possessors and other social actors  (0.000)  F6. Supports from other social actors  −0.485***  Support from other social actors and continued improvements on their attention and understanding of the three types of knowledge as well as their capabilities  (0.000)  Group 3: External support    F7. Local and central government  −0.726***  Support and guidance by governments at different hierarchical levels  (0.000)  F8. Laws and regulations  −0.355***  Institutional support from laws and regulations  (0.000)  F9. Financial support  −0.561***  Financial support from government and other social actors  (0.000)  Note: F1 to F9 refer to Factors 1 to 9; *** P < 0.01 (two-tailed). © The Author 2017. Published by Oxford University Press. All rights reserved. For Permissions, please email: journals.permissions@oup.com

Journal

Science and Public PolicyOxford University Press

Published: Feb 1, 2018

There are no references for this article.

You’re reading a free preview. Subscribe to read the entire article.


DeepDyve is your
personal research library

It’s your single place to instantly
discover and read the research
that matters to you.

Enjoy affordable access to
over 12 million articles from more than
10,000 peer-reviewed journals.

All for just $49/month

Explore the DeepDyve Library

Unlimited reading

Read as many articles as you need. Full articles with original layout, charts and figures. Read online, from anywhere.

Stay up to date

Keep up with your field with Personalized Recommendations and Follow Journals to get automatic updates.

Organize your research

It’s easy to organize your research with our built-in tools.

Your journals are on DeepDyve

Read from thousands of the leading scholarly journals from SpringerNature, Elsevier, Wiley-Blackwell, Oxford University Press and more.

All the latest content is available, no embargo periods.

See the journals in your area

DeepDyve Freelancer

DeepDyve Pro

Price
FREE
$49/month

$360/year
Save searches from Google Scholar, PubMed
Create lists to organize your research
Export lists, citations
Access to DeepDyve database
Abstract access only
Unlimited access to over
18 million full-text articles
Print
20 pages/month
PDF Discount
20% off