series on monitoring research networks no. 01 report... · also graham thiele, gordon prain, graham...

Series on Monitoring Research Networks No. 01

Acknowledgements

The authors wish to thank Sophie Alvarez, Claudia Oriolo and Joslin Isaacson for invaluable research support, and also Graham Thiele, Gordon Prain, Graham Durant-Law, Douglas Horton and Rick Davies for their useful comments. This project was part of RTB’s research portfolio on partnerships (Theme 7). It was jointly financed by RTB and ILAC with grants from the Ministry of Foreign Affairs of the Netherlands and the International Fund for Agricultural Development (IFAD).

iiiMonitoring the composition and evolution of the research networks of the CGIAR Research Program on Roots, Tubers and Bananas

Contents

Acronyms and abbreviations v

Glossary vi

Executive summary viii

1. Introduction 1

2. Literature review 2

3. Methods 3

3.1 Limitations of the study 4

4. Types of research collaborations reported by RTB researchers 6

4.1 Differences between RTB-induced and non-RTB-induced collaborations 6

4.2 Research models 15

5. The structure of RTB’s research networks 17

5.1 Analysis of the whole data set 17

5.2 Analysis of the network of 92 survey respondents 29

5.3 Centre-based analysis 32

6. Conclusions 36

References 39

Annex 1. Description of the questionnaire 41

Annex 2. Definitions of the terms used in the questionnaire 43

iv Monitoring the composition and evolution of the research networks of the CGIAR Research Program on Roots, Tubers and Bananas

List of figures

Figure 1. Regional distribution of researchers who completed the questionnaire 4Figure 2. Regional focus of all collaborations 13Figure 3. Distribution of in- and out-degrees (whole network of 624 collaborators) 18Figure 4. A tree-like network 19Figure 5. Map of all 15 components in the whole network 20Figure 6. Facilitation of communication by discipline of the researchers in the main component of RTB’s network 21Figure 7. Women-centred sub-network and components 23Figure 8. RTB induced sub-network and components 24Figure 9. Global research sub-network and components 26Figure 10. African research sub-network and components 26Figure 11. Latin American research sub-network 27Figure 12. Asian research sub-network 28Figure 13. Collaborations between researchers and non-research actors 28Figure 14. Distribution of in- and out-degrees (sub-network of 92 respondents) 30Figure 15. Sub-network of 92 respondents – main component, reciprocity of links, and isolates 30Figure 16. Intermediary power of nodes (betweenness) by disciplines 31Figure 17. Intermediation (betweenness) of CGIAR centres in the full RTB network 33Figure 18. Interactions among centres and RTB-induced links (92 respondents) 34Figure 19. Two-mode network: research areas shared by CGIAR centres 35

List of tables

Table 1. Types of organizations mentioned, by type of collaboration and relative importance 6Table 2. Collaborations by type of organization and research area (as a percentage of collaborations by research area) 8Table 3. Collaborations by CGIAR centre 9Table 4. Research areas of CGIAR centres mentioned by RTB researchers 10Table 5. Location of the collaborations 11Table 6. Research areas of all RTB collaborations 11Table 7. Purpose of capacity-building collaborations 12Table 8. Purpose of advocacy collaborations 12Table 9. Location of collaborative activities (by gender of collaborator) 13Table 10. Geographic focus of RTB-induced and other collaborations 13Table 11. RTB’s interactions by region and type of organization 14Table 12. RTB’s research and non-research collaborations by region 14Table 13. RTB’s interactions with different types of non-research organizations by region 15Table 14. Characterization of collaborative research activities 15Table 15. Number of respondents that reported a specific number of collaborators and were mentioned as collaborators by others 18Table 16. Distribution of the components by size 19Table 17. Proportion of female collaborators per component in the women-centred sub-network 23Table 18. Links between and within RTB’s four member centres 25Table 19. Number of researchers with a specific number of links 30Table 20. Intermediation of CGIAR centres (betweenness) 32Table 21. Density matrix – intensity of links within and between centres 33Table 22. External–Internal (E–I) index by centre 34Table 23. Regional distribution of links by centre 35

vMonitoring the composition and evolution of the research networks of the CGIAR Research Program on Roots, Tubers and Bananas

Acronyms and abbreviations

ICRAF: International Centre for Research in Agroforestry (also known as the World Agroforestry Centre)

ICRISAT: International Crops Research Institute for the Semi-Arid Tropics

ICT: information and communication technology

IFAD: International Fund for Agricultural Development

IFPRI: International Food Policy Research Institute

IITA: International Institute of Tropical Agriculture

ILAC: Institutional Learning and Change Initiative

ILRI: International Livestock Research Institute

IRRI: International Rice Research Institute

IWMI: International Water Management Institute

MDS: multidimensional scaling

NGO: non-governmental organization

RTB: Roots, Tubers and Bananas

SNA: social network analysis

SPIA: Standing Panel on Impact Assessment

AfricaRice: Africa Rice Center

Bioversity: Bioversity International

CBO: community-based organization

CGIAR: A global research partnership for a food-secure future (formerly Consultative Group on International Agricultural Research)

CIAT: Centro Internacional de Agricultura Tropical (International Center for Tropical Agriculture)

CIMMYT: Centro Internacional de Mejoramiento de Maíz y Trigo (International Maize and Wheat Improvement Center)

CIP: Centro Internacional de la Papa (International Potato Center)

CIRAD: Centre de coopération Internationale en Recherche Agronomique pour le Développement (Agricultural Research for Development), France

FAO: Food and Agriculture Organization of the United Nations

GFAR: Global Forum on Agricultural Research

GIS: geographic information system

GIZ: Deutsche Gesellschaft für Internationale Zusammenarbeit GmbH

vi Monitoring the composition and evolution of the research networks of the CGIAR Research Program on Roots, Tubers and Bananas

This glossary is intended to help non-specialists understand this report; technical terms have been avoided as much as possible in these definitions.

Betweenness centrality: For node i, this is the number of shortest paths between other nodes that run through i. It can also be expressed as a percentage of the maximum possible betweenness that a node could have.

Centrality or global centrality: A node is globally central if it lies at short distances from many other nodes (see definition of ‘distance’).

Centralization: This measures how central a node is in a network by looking at the differences between the centrality scores of the most central node and those of all other nodes. The most common centrality scores are degree and betweenness.

Clustering coefficient: For node i, this is calculated by dividing the number of links between the nodes close to i by the number of links that could possibly exist between them. The clustering coefficient for the whole network is the average of the clustering coefficients for all the nodes in the network.

Component: This refers to a set of nodes that are connected through one or more paths, but that have no connections outside this group.

Cut-point: A node whose removal would increase the number of components by dividing the network or any of its components into two or more disjointed subsets is a cut-point.

Degree: This is the total number of nodes to which a particular node is connected.

Density: This is the number of links in a network divided by the maximum possible number of links.

Diameter: This is the largest distance (or geodesic distance) in the largest subset of connected nodes (i.e., the largest component; see definition of ‘component’).

Distance (or geodesic distance): This is the length of the shortest path between two nodes.

Ego network: The ego network for node i contains i and the other nodes that i is connected to (i’s neighbours). It is possible to define as many ego networks as there are nodes in the network. The size of the ego networks is set by the researcher, usually extending two or three steps from the central node.

Eigenvalue centrality: For node i, this is the sum of its connections to other nodes, weighted by their degree centrality.

External–Internal (E–I) index: This index compares the numbers of links within groups and between groups. It is calculated by taking the number of links of group members to outsiders, subtracting the number of links to other group members, and dividing by the total number of links.

Freeman’s graph centralization index: This index is the proportion of existing links to the links that would be present in a star graph of the same dimension (see definition of star graph).

In-degree: This is the total number of links that end in a particular node.

k-core: This is a connected set of nodes in which each node is adjacent to at least k other nodes; all the nodes within the k-core have a degree greater than or equal to k.

Link: A link is a connection between two nodes. In this study it is also known as a collaboration.

Main component: This is the largest component in a network.

Navigability: This refers to the possibility that any particular node can reach any other node in the network. The ‘small world’ property means that short paths exist between any two nodes and the ‘scale-free’ property implies that nodes can find those paths using only local information.

Neighbourhood: For node i, this includes all the nodes that are connected directly or indirectly to i and that are within a predefined distance, usually, two steps.

Glossary

viiMonitoring the composition and evolution of the research networks of the CGIAR Research Program on Roots, Tubers and Bananas

Node: A node is a point (actor) in a network; in this study it refers to each survey respondent and any person named by a respondent as a partner or collaborator.

Out-degree: This is the total number of links that originate in a particular node.

Path: A path is a continuous sequence of alternating nodes and links that starts at a particular node and ends at a different node.

Power law: A distribution where the frequency with which an event occurs varies as a power of some attribute of that event (e.g., its size) is said to follow a ‘power law’ distribution. The number of links (degrees) in many networks follows ‘power law’ distributions. These are networks in which a few nodes have many links while most nodes have very few links. The distribution of links in the World Wide Web is an example of a power law because Google or Amazon have millions of links while the vast majority of nodes have fewer than 20 links.

Scale-free network: This is a network where the distribution of degrees follows a power law. A major feature of scale-free networks is that they do not have a representative parameter (e.g., the mean does not explain the behaviour of the distribution).

Small world: This property applies to a network with low average distances and a high clustering coefficient. All nodes in a network with a small world property are connected to other nodes by short distances and are organized in relatively local tight groups (i.e., they have high clustering coefficients). The best-known example of this property is the theory of ‘six degrees of separation’.

Star graph: A network where all nodes are only connected to one central node is a star graph; there are no links other than those to the central node. A star graph has the highest possible centralization score.

viii Monitoring the composition and evolution of the research networks of the CGIAR Research Program on Roots, Tubers and Bananas

Executive summary

CGIAR is implementing a change process that aims to develop new approaches for research and innovation, focus research activities and increase the collaboration among centres and with other partners. An essential component of the change process is the creation of 16 research programmes, known as CGIAR Research Programs; Roots, Tubers and Bananas (RTB), which started operations in January 2012, is one of them. Given the complexity of CGIAR and the breadth of the reforms, successful implementation of the change process will require substantial adaptation of goals and strategies as the process evolves. This report describes a cost-effective, easy-to-use methodology based on social network analysis (SNA) to monitor changes in a CGIAR Research Program with the goal of providing useful information primarily to RTB’s management and stakeholders, but also to other CGIAR Research Programs and CGIAR as a whole.

This report describes a pilot project that developed a methodology that seeks to contribute to answering important questions about RTB’s impact pathway by identifying the actors RTB is collaborating with. The methodology used a user-friendly, Internet-based survey to collect information about active collaborations between RTB-financed researchers and a wide range of partners. Collecting the information and cleaning the database required almost three months of intensive work by the research team and the active support of RTB’s director.

All researchers mentioned in RTB’s project documents were asked to list the collaborations they engaged in while conducting their research activities along with some attributes of each collaboration, including type and topic of research conducted, gender of the collaborator, geographical focus and location of activities. The re-searchers were asked to report only collaborations that had been active in the 12 months prior to the survey.

Although it was not possible to fully map RTB’s impact pathway, it was possible to sketch them. Analysis of the survey data showed that about 80 percent of RTB’s collaborations involve other researchers. A reduction in the proportion of partnerships with non-research partners was observed in the RTB-induced collaborations. This is to be expected in the initial phase of a complex programme that seeks to build up links between CGIAR partners and to make the existing research portfolio more coherent. With the explicit priority CGIAR places on

partnerships with actors in the agricultural innovation system, new partnerships outside CGIAR are expected to emerge in the near future.

The lack of diversity in RTB’s research network could affect its productivity. Studies of the interactions between research and diffusion networks have shown that often these networks have different structures and that they share different types of information. It has also been shown that researchers who interact with different types of actors are more productive academically and more creative than researchers who only work with other researchers. To address these issues, RTB should explore new approaches to conducting research as part of innovation processes, including sharing different types of information among different types of actors. It should be noted that the process should not involve only sharing scientific information with non-scientists (e.g., extension) but also passing information from non-scientists to researchers so that the latter can better understand the needs of the potential users and the opportunities created by their research.

While the centres and the CGIAR Research Programs are the operative and administrative units, increasing CGIAR’s impact will require focusing strongly on collaborations across the CGIAR Research Programs and on collaborations with other actors in the agricultural innovation system. From the perspective of evaluation, understanding coalitions of CGIAR Research Programs and the changing interactions among research and non-research actors should be an important component of the evaluation of the CGIAR Research Programs.

The interactions with non-research partners follow geographical patterns, since collaborations with extension organizations and private firms dominate in Africa, Asia and Latin American, while interactions with a global focus involve mainly policy-makers and private firms.

RTB has few direct contacts with non-researchers but there may be more indirect links. Mapping the networks of all RTB’s partners could shed light on this issue, but the pilot project showed that this would be an extremely difficult and costly endeavour. A more manageable alternative is to map other CGIAR Research Programs, especially those that collaborate closely with RTB, to assess the nature and evolution of their joint impact pathway.

ixMonitoring the composition and evolution of the research networks of the CGIAR Research Program on Roots, Tubers and Bananas

The CGIAR Research Programs were created as umbrella organizations for pre-existing projects with the expectation that over time they would reshape CGIAR’s research portfolio according to predetermined priorities. As RTB was only recently created, RTB’s research activities do not yet follow a coherent strategy and this is reflected in the network’s sparse connectivity – especially the dearth of reciprocal interactions. This architecture hampers the sharing of information across RTB and with partners essential for integrating the research and innovation processes. However, RTB has already induced important changes in its research portfolio, fostering greater interaction among CGIAR centres, and refocusing partnerships according to the partners’ capabilities and RTB’s research priorities.

The survey was only partially successful in capturing the different types of research conducted by RTB researchers. From the answers it was clear that many researchers were not familiar with different research approaches (e.g., the difference between ‘on-farm research’ and ‘action research’). This lack of understanding constrains their ability to explore alternative research models and, therefore, diminishes their ability to achieve CGIAR’s development goals.

Most researchers engage in multidisciplinary networks, but it is not clear to what extent their research is influenced by other disciplines. Further research should explore how other disciplines and non-research actors influence RTB researchers.

In addition to the information on the structure of RTB, this pilot project has provided useful lessons about the methodology’s possibilities and limitations. On the one hand, the information enabled mapping of the research networks and provided a baseline that will facilitate future monitoring of RTB’s evolution. The baseline contains information on the type and number of research organizations that participate in RTB’s research activities, gender issues, research approaches and locations, the range of disciplines and geographic focus. Repeating this exercise periodically will enable RTB to identify changes in its research portfolio and to link those changes with learning along its impact pathway. On the other hand, the project did not collect information on financial issues or the size of the research projects (in terms of the number of partners and disciplines involved and the geographic area covered) or the expected length of the projects.

1Introduction

1. Introduction

The CGIAR change process aims to develop new approaches for research and innovation, focus research activities and increase the collaboration among centres and with external partners (CGIAR, 2011). An essential component of the change process is the creation of 16 research programmes, known as CGIAR Research Programs. Roots, Tubers and Bananas (RTB), which started operations in January 2012, is one of the CGIAR Research Programs. RTB brings together four CGIAR centres (Bioversity, CIAT, CIP and IITA) and many other partners. It includes research on six sets of crops: bananas (and plantains), cassava, potatoes, sweet potatoes, yams and other roots and tubers.

Given the complexity of CGIAR and the breadth of the reforms, successful implementation will require substantial adaptation of goals and strategies as the process evolves (Christensen, Anthony and Roth, 2004). The execution of such an adaptive strategy requires timely information on the evolution of the change process (Patton, 2010; Davila, Epstein and Shelton, 2006). In the case of the CGIAR Research Programs, the implementation of an adaptive strategy has two main components: defining impact pathways and monitoring whether the actions of the CGIAR Research Programs conform to those pathways. For the impact pathways to provide effective guidance for action, they should first identify the partners the CGIAR Research Programs expect to work with, which may differ from the partners they actually work with. Such discrepancies have been identified in many organizations and are usually explained in terms of the ‘espoused theory’ and the ‘theory-in-use’ (Argyris and Schön, 1974).1 For example, a CGIAR Research Program may define its impact pathway (the espoused theory) as developing research outputs (e.g., basic seed) that are passed on to private firms (e.g., seed companies) who bring them to the farmers. However, the analysis of the actual interactions (the theory-in-use) may indicate that the CGIAR Research Program researchers interact mainly with other researchers and not with the firms. Recognizing differences between the espoused theories and the theories-in-use can help the CGIAR Research Programs to discover emerging problems and opportunities.

Comparing the espoused theory with the theory-in-use requires describing the impact pathway with a certain

1 The ‘espoused theory’ is the theory of change that an organization explicitly develops, which guides its decisions; the ‘theory-in-use’ is the actual strategy that emerges from the many individual decisions made by the members of an organization. The gap between the espoused theory and the theory-in-use differs from the discrepancies that arise between actual and planned activities, in the sense that the former is not usually conscious and organizations are often unaware of its existence.

degree of detail and making value judgements about observed discrepancies. For example, how many non-research actors should a CGIAR Research Program interact with? And how large should the difference between the expected and actual pathway be to trigger a managerial intervention? Appropriate answers to these questions cannot be provided in the abstract; they require an understanding of the different types of research conducted by the CGIAR Research Programs and of the role of research in pro-poor agricultural innovation processes.

Previous research has found that research and innovation networks change as the research process matures (Kratzer, Gemuenden and Lettl, 2008; Gay and Dousset, 2005). However, no practical guidelines were found on how to use social network analysis (SNA) and the impact pathways for monitoring and evaluation of research for development activities. This report describes a pilot project that developed a methodology that seeks to contribute to answering important questions about RTB’s impact pathway by identifying the actors RTB is collaborating with; future research will need to address the comparison of actual collaborations with those posited by the impact pathway. This project was commissioned by the management of RTB with the explicit goal of supporting organizational change – not as an accountability tool. Monitoring changes in research and diffusion processes as they are implemented complements the management system being developed by the CGIAR Consortium Office.

The project had three objectives:• To help RTB understand how its research products are

developed, how other actors in the agricultural innovation system influence (and are influenced by) the research portfolio and how the research outputs are diffused;

• To provide a baseline for monitoring the evolution of the RTB research networks, by identifying the main partners, disciplinary areas of research, geographical focus and how the research is being conducted;

• To develop a methodology that can be used to monitor RTB’s learning along its impact pathway.

Section 2 reviews the approaches taken by previous studies of research networks and Section 3 describes the research methods of the current study. Section 4 uses basic tables to analyse important features of the collaborations of RTB researchers. Section 5 applies social network analysis (SNA) methods to the structure of the researchers’ networks. Section 6 presents the conclusions of the study.

2 Monitoring the composition and evolution of the research networks of the CGIAR Research Program on Roots, Tubers and Bananas

2. Literature review

Three reasons are behind the recent upsurge of analysis of research networks. First, research is increasingly implemented by multidisciplinary, multi-institutional teams that form networks with both local and global connections, and researchers and policy-makers are increasingly interested in understanding the structure and dynamics of these networks (Cassi et al., 2008; Powell and Grodal, 2005; Bennett, Gadlin and Levine-Finley, 2010; Wagner, 2008). Second, programmes to foster interdisciplinary, inter-institutional collaborations between researchers and other actors in innovation systems have been implemented in several countries and policy-makers are asking about the impact of these programmes on research activities. Third, it has been found that scientific collaborations are an important influence on researchers’ productivity (Klenk, Hickey and MacLellan, 2010; Wagner, 2008), and collaborations with non-researchers help researchers to better contribute to innovation processes and to become more productive and creative (Rivera-Huerta et al., 2011). When researchers learn how to participate in innovation processes, they change the way they conduct research (Ekboir et al., 2009).

Several methodologies have been used to analyse research networks. For this project, only publications that reported on SNA approaches were reviewed. Published analyses of the structure and dynamics of research networks have focused on networks classified in three different ways: networks identified from joint publications (co-authorship relationships), networks identified from project documents (project partnerships) and networks based on data from direct surveys of researchers

Studies based on co-authorship relationships explore databases of scientific articles to identify patterns of interaction. Two authors are considered to have a direct link if they have co-authored at least one paper. When a large number of co-authorships are identified, a map of interactions emerges. If several years of data are available, it is possible to analyse the evolution of these networks. Due to the public availability of data, the structure and evolution of co-authorship networks have been well studied; recent papers include Klenk, Hickey and MacLellan (2010), Palla, Barabási and Vicsek (2007), Newman (2004), and Lyrette (2002). These data only

document collaborations that resulted in a scientific publication while they ignore connections based on capacity-building and advocacy activities. Also, they do not include informal collaborations and exchanges of information, which have been recognized as important components of scientific work even though they may not result in scientific publications (Varga and Parag, 2009; Wagner, 2008; Kratzer, Gemuenden and Lettl, 2008). Another problem with this approach is that publication patterns vary among scientific disciplines (Okubo, 1997); some researchers may be included as co-authors for social reasons (La Follette, 1992) or because they provide data or equipment (Stokes and Hartley, 1989). Finally, scientific works often take several months to be published, by which time the collaboration may no longer be active.

Studies of project partnerships analyse project documents, especially joint proposals and publications, to identify research collaborations. The main difference with co-authorship relationships is that project partnerships are more likely to include non-scientific collaborations. In recent years, many countries have implemented projects promoting research networks. Among these, the European Framework Programmes and the Canadian Network of Centres of Excellence were particularly important because of their continuity and level of investment. Some of these programmes have been analysed by combining SNA with qualitative study methods (Klenk and Hickey, 2012; Protogerou, Caloghirou and Siokas, 2010; Cassi et al., 2008). One drawback is that these data only capture interactions that are relevant for documentation of the project, which may not have resulted in effective collaborations. These data also overlook informal interactions.

Approaches based on co-authorship relationships and project partnerships cannot be used in research areas with low propensity to publish (such as development of agricultural equipment or ‘action research’ projects) or where informal interactions are common. Additionally, they may lead to the inclusion of collaborations that exist only on paper. Following Bozeman and Corley (2004), we instead adopted a strategy based on self-reported information, which captures active formal and informal collaborations as well as non-research interactions. This approach is described in the next section.

3Methods

3. Methods

The mapping of RTB’s networks was conducted in three stages: data collection, data cleaning and data analysis.

The data for this project were collected by asking RTB researchers to complete a questionnaire about all the collaborations they have engaged in as part of their research activities during the 12 months prior to the survey. A research collaboration was defined as a relationship where researchers effectively cooperate with other actors in the innovation system (including researchers, extension agents, research managers, input suppliers, output buyers and policy-makers) and that includes both formal and informal links. This definition is broader than that of a partnership, which is usually defined as a formal interaction (Horton, Prain and Thiele, 2009).

Potential study respondents were identified from a variety of sources. RTB provided four lists of professionals working with RTB-funded projects, including researchers, communication specialists and support professionals. The lists were collated and checked for inconsistencies, missing information was completed, and non-researchers and those no longer working for RTB were excluded. Some researchers forwarded the names of colleagues who were not included in the original lists, and they too were added to the list of potential respondents.

In all, 126 researchers were invited to complete the questionnaire, including 33 women and 93 men. Only seven professionals in the list were not affiliated with CGIAR. The final database included each researcher’s full name, gender, job title, institutional affiliation, the country where she or he was based and contact information. This updated list of researchers was the first output generated by the project.

An Internet-based, user-friendly questionnaire was designed so that completion would take half an hour at most (full details are provided in Annex 1). In the questionnaire, researchers were asked to identify their collaborators and describe the nature of each of these relationships. The only information provided to the researchers was the definition of a collaboration (see Annex 2), leaving to them to decide which relationships to report and how to categorize them. In other words, their responses indicated how they perceived the interactions, which may be different from what those interactions actually involved.

An important limitation of self-reported information is that researchers may forget to report important collaborations because they value their interactions differently. One alternative is to present respondents with a list of potential contacts and let them choose which ones to report as collaborators (Marschall, 2012). This alternative was not feasible in this study because a complete list of all RTB partners was not available.

Different approaches to agricultural research have been developed in the last half-century, including traditional research, on-farm research, participatory methods and action research. However, the names of the different approaches were not used in this survey because it was expected that many researchers would not be familiar with them. To describe the nature of each reported collaboration, the researchers were instead asked to choose among six statements (see Annex 2).

To avoid the problem of memory lapses with regard to past collaborations and to capture active relationships instead, researchers were asked to report only collaborations that had been active in the 12 months prior to the survey.

The questionnaire was pretested before being provided to respondents. However, a problem with the definition of the type of research conducted by the respondents and another with the lack of variability found for two variables (frequency and importance of the collaboration) only became evident after all the information had been collected.

There were two stages of data collection. In the first stage, the questionnaire was sent to the 126 researchers on the compiled list, 90 of whom responded, including 61 who provided complete and valid answers. The survey was opened for a second time after three problems were identified. First, the effective response rate of the first round (49 percent) was considered insufficient for a valid analysis of the RTB network. Second, 22 respondents reported exactly 10 collaborations. This was the limit mentioned on the first page of the questionnaire, but respondents could report more collaborations by requesting a new link. The decision to include only 10 pages on which to report collaborations was based on the assumption that many researchers would be discouraged if they were asked to report more contacts. It was also expected that few researchers would report


more than 10 collaborators. These two assumptions were found to be incorrect. The large number of respondents reporting exactly 10 collaborations suggested that some researchers had more collaborators but did not know how to report them. In the second stage of data collection, the maximum number of collaborations that could be reported on the questionnaire was increased to 40 and no limit was mentioned to the respondents. The largest number of collaborations reported was 21. The third problem during the first stage of data collection was that 11 respondents reported only one collaboration. It was suspected that in fact these respondents had more collaborators but did not report them for various reasons (for example, perhaps they forgot to complete the questionnaire). In the second stage, respondents were invited to revise the answers they had provided in the first stage, or to fill out the questionnaire if they had not yet done so. At the end of the second stage, the number of useful responses had increased to 92 out 126, giving an effective response rate of 73 percent (see Figure 1).

The data were checked for errors (e.g., misspelling of collaborators’ names) and analysed with tables and simple statistical tools, including tests for differences in distributions, as well as with specialized SNA programs (UCINET 6 and NETDRAW).

This report includes analysis of many network param-eters. Although some of these parameters do not have clear relevance to current RTB activities, they are included because the literature on SNA has shown their relevance for most networks, and it is not currently known whether they may in future help to explain changes in the patterns of collaboration. The most effective set of parameters will be identified after several successive studies of RTB’s networks and/or those of other CGIAR Research Programs.

3.1 Limitations of the studyThis project succeeded in exploring several dimensions of RTB’s research portfolio; however, there were important limitations. First, RTB’s networks were not fully characterized because information was collected only from researchers funded by RTB, but not from their other collaborators. Since the latter are likely to also collaborate among themselves as well as with RTB researchers, the full network is likely to be more connected than the one mapped. Furthermore, most named collaborators did not complete the survey. Therefore, two separate analyses were conducted, one with the full data set (i.e., respondents and collaborators mentioned by the respondents) and another containing information only on the 92 respondents. The larger data set was used to study RTB’s range of contacts, geographic reach, main research areas and some patterns of interaction. The smaller data set was used to explore the network’s structure, connectivity and interactions.

A second limitation is that the questionnaire asked researchers to name their collaborators but it did not explicitly state that non-research partners should be included. Thus, it is possible that non-research collaborations are underrepresented, since respondents who are focused on research may not be fully aware of the other dimensions of their work, such as capacity building or advocacy. Anecdotal evidence indicates that this is quite a common problem. Third, information from other CGIAR Research Programs with which RTB researchers collaborate (whether formally or informally) is lacking. These CGIAR Research Programs may be intermediaries between RTB and non-research actors providing additional impact pathways. Fourth, the survey did not collect information on the impacts of RTB on collaborations that were not induced by it, for example, existing partnerships that were strengthened by additional funding. Future research is expected to address these issues.

As with all empirical methods, the self-reporting approach adopted for this study has several shortcomings; we identified three potential problems. First, it lacks operational precision as each respondent decides which collaborations she or he will report. An analysis of the researchers’ motivations for reporting or not reporting certain partnerships is beyond the scope of this report. Second, there is great variation among researchers in their willingness to provide information. Taken together, these problems mean that some collaborations may not be reported at all. Differences in the propensity to provide information may result from personal characteristics and from different perceptions of the value of the exercise. With regard to the perceived value of the exercise, the willingness to provide

Figure 1. Regional distribution of researchers who completed the questionnaire (n=92)

Africa

Asia

Europe

Latin America

North America

228

3

12

47

5Methods

information should increase as researchers and managers come to see the value of the analysis. Repeating this exercise over time and for other CGIAR Research Programs will help to identify systematic changes in researchers’ willingness to provide information and could help to identify biases caused by the self-reporting method. Finally, if researchers are not aware of non-research outputs resulting from their activities, such as capacity building and advocacy, they will fail to report non-traditional collaborations. These problems have been recognized in the specialized literature (Marschall, 2012; Newman, 2003) but are no different from those found in any survey-based study. For example, all statistics and econometric textbooks indicate that the distributions of estimates are conditional on the information used in the estimation.

One of the major limitations of self-reporting is that respondents tend to forget some collaborations, such that the estimates of some parameters (e.g., density) are biased downwards due to the exclusion of some existing links from the calculations (Marschall, 2012). This means

that the estimates obtained with self-reported data are the lower bounds of the true parameters. Some solutions for correcting these biases have been proposed in the literature (Marschall, 2012), but they require the imposition of strong assumptions about the true (unknown) distribution of links. Thus, the correction of the bias comes at the cost of using more prior information, which may cause additional biases. Due to the lack of clear indications on how to deal with these problems, the estimates of the network parameters were not corrected.

The above-mentioned limitations of self-reported data should be less important when comparing changes in the structure of networks over time, as long as the data collection methods are repeated consistently; any systematic effects would be cancelled out in the comparison.

Despite these limitations, the study provides important insights into the nature and structure of the RTB activities and partnerships.


Table 1. Types of organizations mentioned, by type of collaboration, and relative importance

Type of organization1

Not RTB-induced RTB-induced

Not RTB-induced (%)

RTB-induced (%)

Difference in %

(1) (2) (3) (4) (4)–(3)3

International research institutes (mainly CGIAR centres)2 124 68 22 51 29

National NGOs 10 6 2 4 2

Extension agencies 0 1 0 1 1

National private firms 10 3 2 2 0

Independent consultants 7 2 1 1 0

Ministries or other public offices (not public research organizations)

12 3 2 2 0

International NGOs 12 2 2 1 –1

Farmer organizations and CBOs 11 1 2 1 –1

Multilateral organizations (e.g., FAO, GFAR or World Bank) 14 2 3 1 –2

Multinational firms 10 0 2 0 –2

International cooperation agencies (e.g., CIRAD or GIZ) 23 0 4 0 –4

National research organizations or national universities 182 36 33 27 –6

Advanced research institutes (including universities from developed countries)

138 10 25 7 –18

Other 2 0 0 0 0

Total 555 134 100 100

1 For 13 collaborations, the type of organization involved was not specified.2 This includes two international research centres that do not belong to CGIAR, which were mentioned six times.3 A x2 test indicated that the probability that columns 3 and 4 were derived from the same distribution was almost 0 (exactly 1.28E–8).

The influence of RTB is shown to be associated with important differences in the patterns of interactions and collaborations, especially within CGIAR. This reflects the need to bring together a diverse group of previously independent projects and researchers. It should be noted that a major goal of the CGIAR change process is to develop new partnerships with a broad range of partners, including non-research organizations. Future surveys will enable measurement of the progress towards this goal.

4. Types of research collaborations reported by RTB researchers

In its short life, RTB has already made an important impact, fostering greater interaction among CGIAR centres and inducing changes in research topics. The survey identified 702 links, including 134 (19 percent) that were induced by RTB. The survey did not, however, ask how RTB influenced other collaborations. RTB’s research network was found to be quite diverse, including 624 individual collaborators from 302 distinct organizations. Some collaborators were mentioned by more than one survey respondent.

4.1 Differences between RTB-induced and non-RTB-induced collaborationsThe influence of RTB is shown to be associated with important differences in the patterns of interactions and collaborations. Table 1 shows the types of organizations identified in the survey and the number and proportion of collaborations that were RTB-induced versus

non-RTB-induced. The column on the far right in Table 1 shows the difference between the percentages of all collaborations of each type (RTB-induced minus non-RTB-induced), per type of organization. More than half of collaborations induced by RTB (51 percent) were established among international research institutes

7Types of reseach collaborations reported by RTB researchers

(mainly CGIAR centres), compared with just 22 percent of non-RTB-induced collaborations. Meanwhile, the proportion of interactions with non-research partners was slightly lower among RTB-induced collaborations, and the share of interactions with advanced research institutes and national research organizations was substantially lower as compared with non-RTB-induced collaborations. The different focus among RTB-induced collaborations reflects the recently prioritized need to bring together a diverse group of projects and researchers that previously operated under a different organizational structure. It should be noted, though, that a major goal of the CGIAR change process is to develop new partnerships with a broad range of partners, including non-research organizations. Future surveys will enable measurement of the progress towards this goal.

The changes in the patterns of interaction can also be tracked by analyzing the collaborations by research area and type of organization (Table 2). The findings indicated that during its first year of existence RTB has induced a shift in partnerships according to the partners’ strengths. Collaborations among CGIAR scientists formed a larger share of the collaborations in areas where they have traditionally had strong capabilities (e.g., breeding and germplasm conservation), new areas that are critical in the change process (e.g., research management, impact assessment and gender issues) and ‘emerging’ areas that do not require major investments (e.g., GIS and climate change). On the other hand, there was a comparatively higher proportion of collaborations with other research institutions in areas where the latter have stronger capabilities, such as biotechnology (in the case of advanced research institutes), and innovation platforms, seed systems and post-harvest (in the case of national research organizations). At the same time, RTB had a

The survey identified 702 collaborations, of which 134 (19 percent) were induced by RTB.

smaller share of collaborations with these partners in areas where they lack a clear advantage, for example, collaborations on breeding with advanced research institutes. Most researchers engage in interdisciplinary collaborations (see Section 5.1).

Only three collaborations with donors were reported, indicating either a rather distant relationship between donors and researchers or that researchers do not see their interactions with donors as part of their work, despite the fact that they build the capacity of donors when they pass on the latest information and insights on their areas of expertise. This finding has great importance for the future of RTB. Since the establishment of the CGIAR Research Programs is a complex process, their management teams must react to emerging needs and opportunities. But this can only be done if the donors and the Consortium Office allow the CGIAR Research Programs to change. From the donors’ perspective, such limited interactions prevent them from learning first-hand how CGIAR is evolving on the ground, potentially leading to biased perceptions of the change process.

Collaborations between CGIAR centres were mentioned 270 times, representing 38 percent of the total 702 reported collaborations. Collaborations between CGIAR centres are strongly concentrated among RTB’s four initial partners (Bioversity, CIAT, CIP and IITA), accounting for 91 percent of the 270 collaborations. Moreover, all except two of the 81 RTB-induced collaborations with CGIAR centres were between researchers from these four centres. The distributions of all collaborations and of RTB-induced collaborations by CGIAR centre are shown in Table 3.

RTB is helping the CGIAR centres to refocus their research portfolios. As shown in Table 4, new collaborations between centres were focused on a few traditional areas (e.g., breeding) and topics whose importance is growing (i.e., germplasm conservation, impact assessment, gender and large environmental issues, such as climate change). The relatively large number of collaborations in impact assessment is possibly the result of a large exercise on priority setting that was implemented by RTB in 2012, as a condition for the approval of the formation of RTB. To simplify the data presentation, Table 4 does not distinguish between RTB-induced collaborations and other collaborations.

Survey respondents could indicate more than one objective per collaboration, and indeed most of the 702 reported collaborations had multiple purposes: 650 had research objectives, while 444 included capacity-building activities and 200 incorporated advocacy goals. However,

RTB has induced a shift in partnerships according to the partners’ strengths. Collaborations among CGIAR scientists formed a larger share of collaborations in areas where they have traditionally had strong capabilities, new areas that are critical in the change process and ‘emerging’ areas that do not require major investments. On the other hand, there was a comparatively higher proportion of collaborations with other research institutions in areas where the latter have stronger capabilities and a smaller share of collaborations in areas where the partners lack a clear advantage.


Table 2. Collaborations by type of organization and research area (as a percentage of collaborations by research area)

Type of organization

Research area ARI CBO IC ICA INGO IRI M-firm Min M-org NARO N-firm NNGO Other

Biotechnology non-RTB-induced (N-R)

26 0 0 6 3 18 6 0 0 35 6 0 0

RTB-induced (R)

67 0 0 0 0 0 0 0 0 33 0 0 0

Breeding (plant, animal, fish)

N-R 18 0 0 5 0 18 4 1 0 49 3 0 1

R 0 0 0 0 0 53 0 0 0 33 0 13 0

Crop production

N-R 5 0 9 0 0 36 0 0 0 50 0 0 0

R 0 0 0 0 0 33 0 0 0 33 33 0 0

Gender issues N-R 0 0 0 0 0 0 0 0 0 0 0 0 0

R 0 0 0 0 0 100 0 0 0 0 0 0 0

Germplasm conservation

N-R 28 2 0 9 0 15 0 4 4 37 0 0 0

R 18 0 0 0 0 53 0 0 0 24 0 6 0

GIS N-R 33 0 0 0 0 33 0 0 0 33 0 0 0

R 0 0 0 0 0 100 0 0 0 0 0 0 0

Human health and nutrition

N-R 33 0 0 4 8 21 0 4 0 29 0 0 0

R 20 0 0 0 0 0 0 0 0 60 0 20 0

ICT N-R 20 0 0 0 20 20 20 0 20 0 0 0 0

R 0 50 0 0 0 50 0 0 0 0 0 0 0

Impact assessment

N-R 10 0 3 3 0 77 0 0 0 3 3 0 0

R 0 0 0 0 0 91 0 0 0 0 0 9 0

Innovation platforms

N-R 16 13 3 6 0 13 0 0 6 29 6 3 3

R 0 0 0 0 0 0 0 0 50 50 0 0 0

‘Large’ natural systems (e.g., climate change)

N-R 69 0 0 0 0 19 0 0 0 13 0 0 0

R 0 0 17 0 0 50 0 0 0 33 0 0 0

‘Local’ natural systems (e.g., agronomy, agroforestry)

N-R 22 0 0 7 4 15 0 0 0 44 0 7 0

R 33 0 0 0 0 33 0 0 0 33 0 0 0

Pest and disease management

N-R 27 1 1 0 1 21 2 2 2 42 0 0 0

R 18 0 0 0 0 24 0 6 0 41 6 0 6

Policies N-R 23 0 0 0 8 46 0 0 8 8 8 0 0

R 0 0 33 0 0 67 0 0 0 0 0 0 0

Policies N-R 35 0 0 12 0 12 0 0 0 38 4 0 0

R 0 0 0 0 0 17 0 17 0 58 0 8 0

Research management

N-R 6 0 13 0 6 44 0 6 13 6 0 6 0

R 0 0 0 0 0 88 0 0 0 13 0 0 0

Seed systems N-R 8 11 0 5 8 16 0 3 3 32 3 13 0

R 0 0 0 0 40 0 0 0 20 40 0 0 0

Key: ARI, advanced research institute; ICA, international cooperation agency; INGO, international NGO; IRI, international research institute (including CGIAR centres); NARO, national agricultural research institute; NNGO, national NGO; M-firm, multinational firm; N-firm, national firm; Min, ministry or public organization; IC, independent consultant; CBO, community-based organization; M-org, multilateral organization; N-R, non-RTB-induced; R, RTB-induced; GIS, geographic information systems; ICT, information and communication technology.


RTB’s research activities are concentrated on the traditional CGIAR topics: pest and disease management, breeding, germplasm conservation and biotechnology. However, during its first year of existence, RTB has had an influence on the nature of research activities. Comparing the collaborations not induced by RTB with those induced by it, germplasm conservation and gender issues were a larger proportion of the RTB-induced collaborations, while biotechnology, value chains, breeding, and pest and disease management had a smaller share.

RTB has influenced the type of research CGIAR centres perform, increasing what is usually known as ‘downstream’ research and decreasing ‘upstream’ research.

Table 3. Collaborations by CGIAR centre

Name*All

collaborationsRTB-induced

collaborations

CIP 97 28

Bioversity International 57 22

CIAT 50 15

IITA 43 14

IFPRI 8 0

IRRI 3 0

ICRISAT 2 1

ILRI 2 0

WorldFish 2 0

CGIAR Consortium Office 1 1

CIMMYT 1 0

ICRAF 1 0

IWMI 1 0

SPIA 1 0

AfricaRice 1 0

HarvestPlus 0 0

Total 270 81

* Please refer to list of acronyms and abbreviations as needed (page v).

RTB has a clear research focus: 558 collaborations (79 percent of the total) included national research organizations, CGIAR centres and advanced research institutions, while only 75 collaborations (11 percent) were established with disseminators of technical information, such as non-governmental organizations (NGOs), community-based organizations (CBOs) and private firms. RTB’s research focus is confirmed by the fact that 508 collaborations (72 percent) occurred in research environments (desks, experimental stations and advanced and regular laboratories).

A larger share of RTB-induced collaborations among CGIAR centres can be characterized as what is usually known as ‘downstream’ research while a smaller share involves ‘upstream’ research.2 As shown in Table 5, a higher proportion of RTB-induced collaborations documented by the survey involved desk work and research in farmers’ fields, while a lower proportion took place in regular and advanced laboratories, compared with non-RTB-induced collaborations.

RTB’s research activities are concentrated on the traditional CGIAR topics: pest and disease management, breeding, germplasm conservation and biotechnology. However, as indicated by the last column of Table 6, during its first year of existence, RTB has had an influence on the nature of research activities. Comparing the collaborations not induced by RTB with those induced by it, germplasm conservation and gender issues were a larger proportion in the second group, while biotechnology, value chains, breeding and pest and disease management had a smaller share. Impact assessment accounted for a large share of the researchers’ efforts, since 7 percent of all collaborations were reportedly related to this activity, as were 9 percent of the RTB-induced collaborations. The large share of collaborations on impact assessment could be a result of the implementation of a major priority-setting exercise in 2012 that used ex ante assessments. Tables 4 and 6 show that differences in collaboration patterns within CGIAR are similar to those found in the full network.

Capacity building is an important dimension of RTB’s activities, involved in 444 (63 percent) of the documented collaborations. Table 7 shows a varied pattern of collaborations for capacity building, with facilitation, extension and exchanges of information being more common in RTB-induced collaborations than in non-RTB-induced collaborations, while mentoring and thesis supervision accounted for a comparatively low share of RTB-induced collaborations. These differences could be related to the stronger downstream focus of RTB-induced collaborations. However, interactions with non-research partners (i.e., facilitation, extension and innovation brokers) made up only a small share of all collaborations (18 percent).

2 Downstream research is expected to be used by non-researchers shortly after the release of the research results/outputs while there is no such expectation for upstream research. Upstream research is similar to (but not exactly the same as) basic research, while downstream research includes applied research and development.


Table 4. Research areas of CGIAR centres mentioned by RTB researchers

Biov

ersi

ty

CIAT

CIM

MYT

CIP

Cons

orti

um O

ffic

e

Har

vest

Plus

ICRA

F

ICRI

SAT

IFPR

I

IITA

ILRI

IRRI

IWM

I

SPIA

Wor

ldFi

sh

Tota

l

Biotechnology 0 1 0 4 0 0 0 0 0 1 0 0 0 0 0 6


0 2 0 14 0 0 0 0 0 6 0 0 0 0 0 22

Crop production 0 5 0 2 0 0 0 0 0 2 0 0 0 0 0 9

Gender issues 2 1 0 1 1 0 0 0 0 1 0 0 0 0 0 6


6 1 0 7 0 0 0 0 0 1 0 0 0 0 0 15

GIS 0 4 0 2 0 0 0 0 0 0 0 0 0 0 0 6

Human health and nutrition

0 0 0 5 0 0 0 0 0 0 0 0 0 0 0 5

ICT 1 1 0 0 0 0 0 0 0 0 0 0 0 0 0 2

Impact assessment

6 7 1 3 0 0 0 2 5 4 1 1 0 1 2 33


1 0 0 1 0 0 1 0 0 0 1 0 0 0 0 4


1 0 0 4 0 0 0 0 1 0 0 0 0 0 0 6

‘Local’ natural systems (e.g., agronomy, agroforestry)

0 0 0 1 0 0 0 0 0 2 0 0 1 0 0 4

Pest and disease management

2 3 0 11 0 0 0 0 0 3 0 1 0 0 0 20

Policies 0 0 0 1 0 0 0 0 3 2 0 0 0 0 0 6

Post-harvest 0 0 0 3 0 0 0 0 0 1 0 0 0 0 0 4

Research management

5 1 0 6 0 0 0 0 0 1 0 0 0 0 0 13

Seed systems 1 1 0 3 0 1 0 0 0 2 0 0 0 0 0 8

Total 25 27 1 68 1 1 1 2 9 26 2 2 1 1 2 169

Note: Some collaborations were not counted because the relevant information was not available.

Finally, Table 8 shows that RTB-induced advocacy collaborations were also comparatively more likely to involve participation in global and regional institutions (e.g., FAO), but less likely than non-RTB-induced collaborations to involve policy advice or participation in global and regional forums (e.g., GFAR).

Most of the reported collaborations were quite intense, in terms of the frequency of interaction. Among all the

collaborations, 498 (71 percent) involved interacting at least monthly while only 60 collaborations (9 percent) involved interactions every six months (the remainder were somewhere in between these extremes).

Several studies have analysed the importance of formality in research activities. On the one hand, since the 1980s governments and donors have increasingly demanded that research activities be formally contracted (Vera-Cruz


Table 6. Research areas of all RTB collaborations

Research area

Not RTB-induced RTB-inducedNot RTB-induced

(%)RTB-induced

(%) Difference in %

(1) (2) (3) (4) (4)–(3)

Biotechnology 54 3 10 2 –8

Breeding (plant, animal, fish) 77 15 15 12 –3

Crop production 22 3 4 2 –2

Gender issues 0 6 0 5 5

Germplasm conservation 46 17 9 14 5

GIS 3 5 1 4 3

Human health and nutrition 24 5 5 4 –1

ICT 5 2 1 2 1

Impact assessment 35 11 7 9 2

Innovation platforms and value chains 31 2 6 2 –4


16 6 3 5 2

‘Local’ natural systems (e.g., agronomy, agroforestry, conservation agriculture)

27 3 5 2 –3

Pest and disease management 90 18 17 14 –3

Policies 14 3 3 2 –1

Post-harvest 27 9 5 7 2

Research management 16 8 3 6 3

Seed systems 38 9 7 7 0

Total1 525 125 100 100

1 For 52 collaborations, information about their purpose was not available.

Table 5. Location of the collaborations

Location of collaboration

Not RTB-induced RTB-induced Not RTB-induced

(%) RTB-induced

(%) Difference in %

(1) (2) (3) (4) (4)–(3)2

Desk 118 47 21 35 14

Farmers’ fields 117 40 21 30 9

Partner’s location (e.g., market or ministry)

24 4 4 3 –1

Experimental station 99 21 18 16 –2

Regular laboratory 45 7 8 5 –3

Advanced laboratory 150 14 27 11 –16

Total1 553 133 100 100

1 For 16 collaborations, the location was not specified.2 A x2 test indicated that the probability that columns 3 and 4 were derived from the same distribution was almost 0.


Table 8. Purpose of advocacy collaborations

Type of advocacy activity


(%) RTB-induced (%) Difference in %

(1) (2) (3) (4) (4)–(3)

Participation in global and regional institutions

58 24 38 51 13

Participation in global and regional forums

61 17 40 36 –4

Policy advice 34 6 22 13 –9

Total 153 47 100 100

Table 7. Purpose of capacity-building collaborations

Type of capacity-building activity


(%) RTB-induced (%) Difference in %

(1) (2) (3) (4) (4)–(3)

Facilitation 19 17 6 16 11

Extension (advisory services) 21 13 6 12 6

Exchange information 75 28 22 27 5

Innovation broker 10 0 3 0 –3

Technical 136 38 40 36 –4

Mentoring 27 3 8 3 –5

Thesis supervision 51 6 15 6 –9

Total 339 105 100 100

et al., 2008; Kraemer, 2006). On the other hand, informal relationships play a major role in the exploration of new research areas and the development of research and innovation partnerships (Heinze et al., 2009; Wagner, 2008).

For this study, the importance of informal collaborations was analysed, where formality was explicitly defined by the existence of an agreement between organizations. Informal interactions were found to account for about one third (237, or 34 percent) of all collaborations. However, informality seems to be important only for the most connected researchers; just 31 researchers reported informal collaborations and all but 3 of them had more than 6 partners. There was no evidence of gender differences in the share of formal collaborations or of RTB’s influence on the pattern of informal collaborations. It should be noted that the distinction between formal and informal collaborations may not capture all the nuances of research partnerships. For example, two researchers may have a long history of informal collaboration and yet their current work may involve a formal contract. However, the responses make it clear

that informal collaborations are an important component of the researchers’ activities.

The overall gender distribution of the named collaborators was 150 females and 474 males (24 percent and 76 percent, respectively). The proportion of women represented among the RTB-induced partnerships was higher than among the non-RTB-induced collaborations, but the difference was not significant (25 percent and 23 percent, respectively). As shown in Table 9, a slightly higher proportion of collaborations with women occur in advanced laboratories (29 percent) and at desks (26 percent), compared with the proportion of collaborations with men. Meanwhile, collaborations with men are more likely to occur in experimental stations (19 percent) and farmers’ fields (24 percent) than are collaborations with women.

While collaborations with a global focus predominate, RTB clearly has an especially strong presence in Africa and Latin America, while RTB activities in Asia are limited. The geographic distribution of the collaborations is shown in Figure 2. Based on the survey findings, female


Table 10. Geographic focus of RTB-induced and other collaborations (percentage)

Region Not RTB-induced RTB-induced

Africa 33 32

Asia 10 10

Europe 1 0

Global 34 40

Latin America 20 17

North America 1 0

Region not reported 1 0

Total 100 100

Table 9. Location of collaborative activities (by gender of collaborator)

Location

Female Male

n % n %

Advanced laboratory 46 28.6 120 23.2

Desk 42 26.1 125 24.1

Experimental station 24 14.9 98 18.9

Farmers’ fields 32 19.9 125 24.1

Partner’s location (e.g., market, ministry) 8 5.0 20 3.9

Regular laboratory 9 5.6 30 5.8

Total1 161 100.0 518 100.0

1 For 23 collaborations, information on both gender and location was not available.

Figure 2. Regional focus of all collaborations (n=702)

0 50 100 150 200 250

Not available

North America

Latin America

Global

Europe

Asia

Africa

5

5

141

238

10

72

231

researchers were more likely to be involved in collaborations with a global and Latin American focus (in both cases, 28 percent of these collaborations involved women compared with 25 percent of all collaborations) and less likely to work on collaborations focused on Africa (21 percent) and Asia (22 percent).

The RTB-induced collaborations vary slightly in their geographic focus, with a larger proportion having a global focus (40 percent) compared with one third (34 percent) of non-RTB-induced collaborations (Table 10).

RTB collaborations with a global focus mainly involve CGIAR centres and advanced research institutes (see Table 11). Among the other three regions (not including Europe, which only involved 10 collaborations), collaborations were most likely to be with national agricultural research organizations (42 to 49 percent of collaborations), followed by collaborations with CGIAR centres (17 to 21 percent) and advanced research institutes (7 to 14 percent).

While collaborations with a research focus dominate in all regions among both RTB-induced and other collaborations, there are clear differences in their relative importance. As shown in Table 12, all RTB-induced global collaborations and 94 percent of non-RTB-induced global collaborations were research-oriented. RTB-induced


Table 11. RTB’s interactions by region and type of organization (percentage)

Africa Asia Europe Global Latin America

Advanced research institute 14 7 70 35 13

CBO 2 7 0 0 5

Extension 0 0 0 0 0

International cooperation agency 2 0 10 5 1

International NGO 4 0 0 0 2

Independent consultant 1 1 0 2 1

International research institute (mainly CGIAR centres)

17 21 0 47 20

Ministry or public office 2 6 0 0 2

Multinational firm 2 0 0 1 2

Multilateral organization 3 1 0 2 1

National agricultural research organization 48 49 20 6 42

National firm 1 4 0 0 4

National NGO 2 3 0 0 6

Total 100 100 100 100 100

collaborations with non-research organizations were most prevalent in Asia (29 percent), followed by Africa (26 percent) and Latin America (13 percent of all collaborations). This differs from the pattern in non-RTB-induced collaborations, among which non-research-

oriented collaborations were most common in Latin America (25 percent). All 10 collaborations in Europe were research-oriented and none were RTB-induced.

In order to analyse the collaborations with non-research organizations, they were categorized into three groups: extension actors (extension agencies, community based organizations, and national and international NGOs); policy-makers (ministries or public offices and multilateral organizations); and private sector (national and multinational firms). Reported collaborations with non-research organizations were mainly focused on Africa, Asia or Latin America, indicating that RTB’s

Table 12. RTB’s research and non-research collaborations by region


Not induced by RTB

Research 153 (84%) 45 (80%) 10 (100%) 166 (94%) 85 (75%)

Non-research 30 (16%) 11 (20%) 0 (0%) 11 (6%) 29 (25%)

Total 183 (100%) 56 (100%) 10 (100%) 177 (100%) 114 (100%)

Induced by RTB

Research 32 (74%) 10 (71%) 0 (NA) 54 (100%) 20 (87%)

Non-research 11 (26%) 4 (29%) 0 (NA) 0 (0%) 3 (13%)

Total 43 (100%) 14 (100%) 0 (NA) 54 (100%) 23 (100%)

NA, not applicable.

RTB-induced collaborations are more likely to have a global reach and are less likely than other collaborations to focus on Latin America. Most collaborations with a global focus involve CGIAR centres and advanced research institutes; most collaborations with a regional focus (except Europe) involve national and local partners.


Table 13. RTB’s interactions with different types of non-research organizations by region


Not RTB-

inducedRTB-

induced

Not RTB-

inducedRTB-

induced

Not RTB-

inducedRTB-

induced

Not RTB-

inducedRTB-

induced

Not RTB-

inducedRTB-

induced

Extension 14 (47%) 7 (64%) 6 (55%) 1 (25%) 0 (NA) 0 (NA) 1 (9%) 0 (NA) 16 (55%) 2 (67%)

Policy-makers

10 (33%) 3 (27%) 4 (36%) 1 (24%) 0 (NA) 0 (NA) 6 (55%) 0 (NA) 4 (14%) 1 (33%)

Private sector

6 (20%) 1 (9%) 1 (9%) 2 (50%) 0 (NA) 0 (NA) 4 (36%) 0 (NA) 9 (31%) 0 (0%)

Total 30 (100%) 11 (100%) 11 (100%) 4 (100%) 0 (NA) 0 (NA) 11 (100%) 0 (NA) 29 (100%) 3 (100%)

NA, not applicable.

portfolio goes beyond pure research. As shown in Table 13, the non-research collaborations in these three regions were most likely to be with extension actors and most were not induced by RTB. However, the number of collaborations with non-research organizations is too small for further meaningful analysis.

The survey data do not shed light on whether RTB is developing an effective impact pathway because it is not known (a) what the optimal combination of research and non-research partnerships should be for a research-for-development organization, or (b) what types of non-

research partners are necessary for effective impact on the ground. Future surveys should help to clarify these issues.

4.2 Research modelsTo explore the use of six research models, the survey required respondents to choose among six statements the one that best described each of their reported collaborative research activities. These statements (see Table 14; more details in Annex 1) describe different models of research that have been used in agricultural research for the last 50 years.

Table 14. Characterization of collaborative research activities

Statement Research model Number of cases

1. You work mainly with him/her in labs or experimental stations, and occasionally with extension agents.

Example: Finding molecular markers for particular traits, or providing occasional policy advice to policy-makers

Research conducted with limited interaction with non-researchers

253

2. You work with him/her in farmers’ fields replicating the farmers’ conditions. Farmers may be hired to do field work or to manage experiments, but they do not participate in the design and evaluation of experiments.

Example: A farmer is hired to plant a crop in a particular way.

Research conducted in farmers’ fields with limited interaction with non-researchers

50

3. You work with him/her (a farmer or other type of stakeholder) as equal partners in the design and implementation of the research activities, but you do not explicitly seek to influence his/her behaviour.

Example: Participatory breeding or experimenting on conservation tillage.

Participatory research 108

4. You interact with him/her to conduct systematic inquiries that directly help improve his/her practices.

Example: Innovation brokers work with market actors to improve value chains.

Action research 113

5. You facilitate his/her access to scientific information or to a research network, but you do not have joint research activities; you may also participate in communities of practice.

Gatekeeper: linking local researchers with research networks

57

6. You contract him/her to perform specific components of a research project. Contract research 118

No information provided (no statement selected) 3


In the 1960s and 1970s, the dominant public research paradigm was the linear model; most activities were conducted in laboratories and experimental stations and the research outputs were transferred to national research organizations, which in turn modified them and passed them to extension agents, who transferred them to farmers. A number of researchers recognized that this research model often did not respond to the farmers’ needs and they developed on-farm research approaches, in which researchers conduct experiments under conditions in farmers’ fields. In the 1980s, on-farm research gained acceptance, but in many cases it did not represent a major departure from the linear model because researchers did not involve farmers in the planning and interpretation of the experiments.

Participatory research methods were developed in the 1990s to address the shortcomings of on-farm research. In this new approach, farmers participated in the planning, design and interpretation of the research activities, but the researchers positioned themselves outside the social and productive processes that were the subject of the research. In the 2000s, again several researchers observed that farmers and other actors often did not make use of research outputs, even those obtained with participatory approaches. In response to this, ‘action research’ methods were developed, in which researchers participated in innovation processes as change agents. Action research methods have not been mainstreamed and are still unknown to many researchers. This classification of research activities does

not have a value connotation; some activities are best conducted within the linear model (e.g., biotechnology), while others require strong stakeholder involvement (e.g., value chain development). Table 9 presents the distribution of collaborations by research type, as reported by survey respondents using the six statements that represent the six research models.

The category ‘research conducted in farmers’ fields’ (statement 2) was selected to describe surprisingly few collaborations while ‘action research’ (statement 4) was mentioned more often than expected. Respondents listed the activity location as ‘in farmers’ fields’ 157 times (Table 5), which seems to contradict the selection of statement 2 only 50 times (Table 9). On the other hand, ‘in the partner’s location’ was mentioned only 28 times (Table 5) while statement 4 indicating ‘action research’ was mentioned 113 times (Table 9). These numbers suggest that the statements did not succeed in capturing the difference between action research and other research approaches. Nevertheless, research activities where researchers had little interaction with non-researchers clearly predominated among the reported collaborations (i.e., statements 1, 2 and 6 combined, accounting for 421 collaborations, or 60 percent).

Based on the survey results, the predominant types of research activities are those in which researchers have little interaction with non-researchers.

17The structure of RTB’s research networks

RTB’s network is weakly connected, with no clear intermediaries. Most researchers were engaged in multidisciplinary networks, and no clusters defined by discipline or geographic focus were identified.

5. The structure of RTB’s research networks

As a basis for the analysis of RTB’s research network, the boundaries of the network first had to be defined. Some networks have clearly defined limits; for instance, co-authorship relationships in a specific discipline, or telephone calls in a particular region (Palla, Barabási and Vicsek, 2007). In other cases, the limits must be set by the analyst. For example, researchers interact with other researchers and actors in the innovation system, who in turn interact with other actors, and so on. Therefore, the whole network could potentially include a very large number of contacts. Collecting information from all these contacts would require a massive amount of resources. In the case of the present project, even contacting the respondents’ immediate collaborators was beyond the scope and would have exceeded the available resources. Thus, it was decided to limit the survey only to the researchers funded by RTB (see Section 3 for a detailed description of data collection methods). Because most of the people mentioned by the respondents were not asked to provide information, some network parameters (e.g., density) could not be calculated for the whole data set. Therefore, separate analyses were conducted for (1) the whole data set of 624 nodes and 702 links/collaborations, and (2) the data set of the 92 survey respondents and 91 links. Finally, the network was analysed in terms of interactions among centres, rather than individual researchers.

5.1 Analysis of the whole data setThe analysis of the whole data set reveals a network of weakly connected actors, with no individuals playing a clear intermediary role. This structure reflects in part the genesis of the CGIAR Research Programs (including RTB), which originated as umbrella organizations for a large number of pre-existing projects that didn’t have a unifying strategy. Future surveys will help to determine whether RTB evolves to become a more coherent research programme. Having relatively sparse connections is also a consequence of the network’s size; because humans have limited time to interact among themselves, as nodes are added to the network, the number of actual links each person has grows more

slowly than the maximum number of possible interactions. The analysis also revealed that most researchers were engaged in multidisciplinary networks, and no clusters defined by discipline or geographic focus were identified. On the other hand, there was some grouping based on the specific location of the research activities (e.g., a laboratory or farmers’ fields).

The complete data set had 624 nodes (individual names) and 702 links. The links are directed because the originator of the link is known. Although most people in the data set did not complete the survey (i.e., only 92 of the 624 nodes were survey respondents, while the others were named collaborators), the in- and out-degrees provide some information about the network’s structure. For this analysis, it should be kept in mind that the data reflect only the point of view of the respondents and not of the named collaborators.

The out-degrees (i.e., the number of contacts a person sends information to) ranged from 0 to 21, while the in-degrees (i.e., the number of contacts a person receives information from) ranged from 0 to 5 (Table 15). There are various reasons for the differences in the distributions of the in- and out-degrees. One consideration is that 532 nodes (the collaborators named by the respondents) were not asked to complete the survey; had they provided information, the pattern of interaction would likely be different. Also, people have different commitments to relationships and value the interactions differently; what is important for someone may not be important for someone else. For example, a researcher that allocates 80 percent of her time to a project values the collaboration differently than a researcher that allocates only 5 percent of his time. Similar asymmetries arise in mentor and mentee relationships, in hierarchical relationships and in situations where a researcher provides crucial information for a project but receives little in return. In such cases, the links are not valued equally and the in- and out-degrees differ.

The distributions of the in- and out- degrees seem to have ‘fat tails’, a result commonly found in research networks (Figure 3). However, the number of nodes is relatively small and does not support definitive conclusions about the shape of the distribution.

In the SNA literature, nodes with out-degrees are considered to be sources of information, while those with in-degrees are receivers. In the whole network of RTB


No central actors were identified in the network; the researchers with higher in- and out-degrees were not directly linked among themselves.

Table 15. Number of respondents that reported a specific number of collaborators and were mentioned as collaborators by others

Number of collaborations

Out-degrees (mentioned others)

In-degrees (mentioned by others)

21 1 0

20 7 0

19 2 0

18 1 0

17 0 0

16 1 0

15 0 0

14 0 0

13 0 0

12 0 0

11 2 0

10 23 0

9 1 0

8 3 0

7 5 0

6 8 0

5 6 2

4 8 6

3 4 20

2 7 61

1 13 486

0 532 49

3 The in-degree centralization is meaningless in a network like this where a large proportion of nodes did not provide information.

researchers, a small number of nodes sent information to many people (11 people mentioned more than 16 collaborations), but there were no central receivers (no individual was named as a collaborator more than five times and the vast majority of nodes had fewer than three incoming links). In addition, no node was found to have a strong influence as measured by the average number of degrees (i.e., a node’s number of connections as a proportion of the total number of nodes) because all nodes had relatively few connections compared with the large size of the network. The maximum average out-degree was 0.034, meaning that even the best-connected node sent information to just 3.4 percent of the nodes in the network. The maximum average in-degree was 0.008, meaning that the best-connected node received information from just 0.8 percent of the nodes in the network. In other words, information within RTB does

Figure 3. Distribution of in- and out-degrees (whole network of 624 collaborators)

Out-degrees In-degrees

0 100 200 300 400 500 600

25

20

15

10

5

0

not flow easily through personal relationships. This observation is confirmed by other parameters discussed below and the implications for RTB are discussed in the conclusions (Section 6).

The researchers with relatively high degrees (both in- and out-degrees) interacted with researchers from other disciplines. Additional analysis is necessary to explore the depth of these exchanges across disciplines.

We analysed the connectivity of the network with two parameters. The first was Freeman’s graph centralization index, which measures the number of existing links as a proportion of the links that would be present in a star graph of the same dimension. The out-degree centralization index is 3.20 percent; the small value indicates that no researcher had a positional advantage in the network.3 The most central actors, i.e., the researchers with higher in- and out-degrees, were not directly connected among themselves.

Freeman’s graph centralization index can be misleading when the map of the network resembles a tree with many branches (Figure 4) because such a network is highly centralized, i.e., removal of the central node would split the network into separate components.


Freeman’s index looks at the differences in degrees between the most central actor and all others, so in a network where all nodes have low degrees, a degree-based centralization index would result in a low index.

To take these considerations into account, the eigenvalue centrality was proposed. It is defined as the sum of each node’s connections to other nodes, weighted by their degree centrality. In this calculation, connections to actors who are themselves well connected are more influential than connections to poorly connected actors.

The eigenvalue centrality index for the whole network in this study is 46 percent, which would indicate that a large number of paths converge on a few nodes; however, the standard deviation of the eigenvalues is 60 percent higher than their mean, indicating high variability of the data. Additionally, the eigenvalue centrality index contradicts other parameters of this network that point to a decentralized structure (especially the low Freeman centrality index and the fact that removing the cut-points has little effect on the network structure, which is discussed further below).

Women were found to be less important than men as sources of information, but equally important as recipients of information. This is evidenced by the fact that although the proportion of female researchers in the whole network was 24 percent, 15 percent of the 20 top out-degrees and 25 percent of the 20 top in-degrees were women.

The connectivity of the network depends on the existence (or lack) of paths that link a particular node with all other nodes. A component is defined as a subset of nodes that are connected through one or more paths,

but that have no connections outside this group. RTB’s network is disconnected; some researchers formed their own small networks that were not connected to the larger group. As shown in Figure 5, the whole network comprises one large (main) component with 561 nodes (90 percent of the total) and 14 smaller components that range in size from 2 to 11 nodes. The distribution of the components by size is shown in Table 16. All the collaborations with a global focus were part of the main component, indicating that information of global importance can circulate to most nodes in the network. The small components were largely made up of partners working at experimental stations or in farmers’ fields and with a regional focus, indicating that these components may be relatively local, resulting from isolated projects. About half of the interactions within the smaller components were established as a result of RTB, and most of the interactions were formal. Examining what roles these projects play in RTB’s research portfolio is beyond the scope of this document, but future studies will be able to assess their integration into the main component.

Figure 4. A tree-like networkSource: Marschall, 2012

RTB’s network comprises one large group and several small sub-networks. The smaller networks have a local focus and result from isolated projects.

Table 16. Distribution of the components by size

Component No. Number of nodesProportion of all

nodes (%)

1 561 89.9

2 11 1.8

3 9 1.4

4 8 1.3

5 7 1.1

6 5 0.8

7 4 0.6

8 3 0.5

9 3 0.5

10 3 0.5

11 2 0.3

12 2 0.3

13 2 0.3

14 2 0.3

15 2 0.3


A few individuals in the network facilitate communication between nodes, although alternative longer paths bypassing these individuals also exist. As RTB researchers become more integrated, the facilitation role played by these individuals should become less important.

Figure 5. Map of all 15 components in the whole network

Note: Each component has been given a different colour simply to facilitate visualization.

Because the network is disconnected, the distance-based measures are only calculated within the main component. The network density is defined as the number of links in the graph divided by the maximum possible number of links (Scott, 2000). The average density (0.002) is quite low, reflecting the facts that it is a relatively large network and that most nodes did not provide information about their connections. The main component’s cohesion measures (density, diameter, average degree and average diameter) are very similar to those of the network as a whole; this was expected given that most nodes of the network are in the main component.

The main component is relatively well connected; the average distance (the average length of the shortest path between two points) is 4.093 and the diameter (the largest geodesic distance in the connected network) is 10. Geodesic distance is defined as the number of links in the shortest possible path from one actor to another. In other words, in the main component in this study, a message sent by a node can reach any other node in an average of 4 steps and no more than 10 steps.

The betweenness centrality of a node is defined as the number of shortest paths between other pairs of nodes that pass through the node; it indicates the node’s intermediary power. The strength of a node’s intermediary power, as measured by this parameter, is not absolute. The

betweenness centrality indicates that several most efficient (i.e., shortest) paths connecting any two nodes pass through a small number of individuals. If these individuals were removed from the network, the new shortest paths would be longer than the original ones, but the individuals at both ends of the paths would still be connected. Betweenness centrality for the whole network is 73 percent, indicating that a few intermediaries not only had many connections but were also linked to people who were well connected. Some of the nodes with high betweenness centrality have low degrees, meaning that although these individuals did not have many connections, they could help different groups of researchers to communicate. As shown in Figure 6, the nodes with highest intermediary power (betweenness centrality) included breeders (brown) and agronomists (blue). With a couple of exceptions, social scientists (red) and communications specialists (yellow) did not play a central role in the network. The most connected researchers interacted with collaborators from different disciplines.


Figure 6. Facilitation of communication by discipline of the researchers in the main component of RTB’s network (nodes colour-coded by discipline)

Note: The size of the node indicates its betweenness. The nodes are colour-coded to represent disciplines and affiliations, as follows: brown = plant breeding, blue = agronomy, red = social science, black = research management, grey = nutrition/health, pink = plant pathology/plant biology/entomology, dark green = post-harvest, light green = agriculture system research, light blue = genetics/molecular biology/genomics/biotechnology, yellow = communications/ICT/bioinformatics, light purple = government, dark blue = biology/biochemistry, dark purple = geography/GIS/meteorology, olive green = donors, opaque green = livestock, orange = private sector/NGO, white = forestry. Some of the nodes are too small for the colour to show.

RTB researchers do not tend to interact intensively with small groups of collaborators. The removal of a few key researchers eliminates only a few nodes from the main component, corroborating the observation that no researchers have a particularly central role. This implies that the departure of particular researchers, even the most connected, would not have major consequences in terms of network connectivity, even though it may significantly affect the efficiency of information transmittal (lower betweenness centrality) and research capacity (i.e., if researchers with highly specialized knowledge were lost).

However, the researchers working on plant breeding and plant genetics had less diverse networks. Few of the collaborations of the central nodes resulted from RTB – a result that was anticipated since the well-connected researchers have been in the system for several years.

The betweenness centrality depends greatly on the density of the network. As the number of links grows, the number of possible shortest paths increases, reducing the centrality of any particular node (Marschall, 2012). Reductions of the betweenness centrality would indicate a greater integration of RTB as a research programme.

Another indication of the structure of the network is the organization of local ‘neighbourhoods,’ which can be measured by the clustering coefficient. This coefficient for a particular node is the proportion that is calculated by dividing the number of links between the nearby nodes (in its neighbourhood) by the number of links that could possibly exist between them. The clustering coefficient for the network is the average of the clustering coefficients for all the individual nodes. The clustering coefficient was found to be 0.07. While this is

significantly larger than the average density, it is still very small, indicating that RTB researchers do not tend to interact intensively with small groups of collaborators, as illustrated in Figure 5. The same result is obtained when the analysis is performed by geographic region, type of relationship and main subject of research. These small values reflect both the large size of the network and the fact that only 15 percent of the nodes completed the survey (i.e., information was not sought from the collaborators named by the respondents).


The main component is quite robust. When a network has low density it can be fragile; the removal of an important node may break the network into separate parts. Such nodes are known as ‘cut-points’ and the nodes that would become separated are the ‘blocks’. The main component in this study has six blocks and four cut-points. The removal of the cut-points has little effect on the network structure because only the collaborators of the removed nodes would be lost. While the cut-points are not directly connected to each other, they can be linked in two steps at most. The four networks originated by the cut-points are multidisciplinary, but laboratory work predominates.

The small effect that the removal of a cut-point has on the structure of the network implies that the departure of particular researchers, even the most connected, would not have major consequences in terms of network connectivity, even though it may significantly affect the efficiency of information transmittal (lower betweenness centrality) and research capacity (i.e., if researchers with highly specialized knowledge were lost).

The six blocks are quite different in size and composition. The block sizes are 73, 31, 27, 13, 10 and 7 nodes. The largest three blocks and the smallest one are multidisciplinary. The fourth block includes six geneticists, four plant pathologists and one breeder. The fifth block is composed of eight social scientists, one agronomist and a post-harvest specialist. In terms of regional focus, the largest block is mainly focused on Africa and global research; the second has a Latin American focus and, to a lesser degree, an African focus; the third-largest block focuses on Africa; and the last three blocks have a global focus.

The connectivity of the network is also measured by the ‘k-core’, a set of connected points in which each point is adjacent to at least k other points. All the points within the k-core have a degree greater than or equal to k (Scott, 2000). A k-core is an area of relatively high connectivity. In this study, the 2-core has 110 nodes and the 3-core has 12 nodes – 18 percent and 2 percent, respectively, of the total number of nodes. There are no higher-order k-cores. As with the other parameters, the relatively small number of nodes in the k-cores indicates that the network is sparsely connected.

The presence of a k-core would indicate the existence of a ‘community of practice’ or at least, of a set of researchers that interact assiduously. Such communities have been identified in specialized areas of research, such as the European space industry (Kratzer, Gemuneden and Lettl, 2008). No k-cores of order higher

than 3 were found in RTB’s network, despite the fact that work on the central crops has been ongoing for many years.

Women-centred sub-network and componentsThe role of women in RTB’s network was analysed by identifying the components centred on female researchers – the women-centred ‘ego networks’.

The survey was answered by 20 women researchers, and their combined collaborations include 163 nodes and 154 links (Figure 7). This sub-network of 163 nodes comprises 12 separate components, the largest of which includes 22 percent of the nodes, while three components include between 11 and 13 percent of the nodes and the rest have smaller shares.

The largest component has 37 nodes, about half of whom are geneticists. The other members of the component show no pattern in terms of disciplines. The component has a largely global focus, with some work in Latin America. Only six links were created as a result of RTB. The second largest component has 22 nodes and most of its members are research managers, but it also includes a few communication specialists and a social scientist. All links in that component have a global focus (most were not induced by RTB) and all the work, except for one link, is desk work. The third largest component has 21 nodes and is composed exclusively of plant pathologists and GIS specialists; almost all the research (90 percent) is conducted in advanced laboratories.

The findings indicate that women tend to interact more intensively with other women. While the proportion of female collaborators (nodes) in the whole network is 24 percent, their share never falls below 30 percent within the women-centred components, as shown in Table 17.

Sub-network of RTB-induced collaborationsRTB induced the creation of 134 links, involving a total of 131 nodes, and this sub-network is split into 16 components of at least three people each (Figure 8). The largest components all have a clear geographical focus. The largest component of this sub-network has 42 nodes (31 percent of the RTB network) and is highly multidisciplinary, as represented by the diversity of colours in Figure 8: 11 social scientists, 10 plant pathologists, 7 research managers, 5 ICT specialists, 3 geneticists, 2 geographers, 1 agricultural systems researcher and 1 plant breeder. However, most social scientists (red nodes) are connected to the component by one central node. Plant pathologists, entomologists and plant biologists (pink) are also mostly connected by one node. Reportedly, half of the collaborations


Table 17. Proportion of female collaborators per component in the women-centred sub-network

Component No.Number of female

collaborators (nodes)

Proportion of nodes in the component

that are women (%)

1 13 36

2 8 38

3 7 41

4 6 30

5 6 40

6 4 40

7 4 57

8 3 30

9 3 60

10 2 33

11 1 100

12 1 100

Whole network 130 24

Figure 7. Women-centred sub-network and components (nodes colour-coded by discipline)

Note: The nodes are colour-coded to represent disciplines and affiliations, as follows: brown = plant breeding, blue = agronomy, red = social science, black = research management, grey = nutrition/health, pink = plant pathology/plant biology/entomology, dark green = post-harvest, light green = agriculture system research, light blue = genetics/molecular biology/genomics/biotechnology, yellow = communications/ICT/bioinformatics, light purple = government, dark blue = biology/biochemistry, dark purple = geography/GIS/meteorology, olive green = donors, opaque green = livestock, orange = private sector/NGO, white = forestry.

occurred on a monthly basis, about 70 percent were based on desk work and 65 percent of the links had a global focus.

The second-largest component has 32 nodes, with a less diverse mix of disciplines: 8 plant breeders, 6 agronomists, 6 geneticists and 12 nodes from other disciplines. As expected, there is a strong connection with Latin America (the focus of about half of the collaborations), since CIP and CIAT headquarters are in that region. The third-largest component has an African focus. The work is conducted mainly in farmers’ fields, and about half of it is on a contract basis, involving partners from several disciplines. The component has two distinct sub-groups formed around individual nodes and linked by one single node, such that this component is like two of the other small components that are each formed around one node, but linked in this case by one mutual collaborator (top right of Figure 8). The relatively dispersed structure of the RTB-induced sub-network

Female researchers tend to collaborate more intensively with other women than with men.


Figure 8. RTB-induced sub-network and components (nodes colour-coded by discipline)


If RTB succeeds in promoting a holistic view of the mandate crops, then collaborations among the centres should increase. This effect has already been observed in the relationships between Bioversity, CIAT and CIP; about half of the interactions among these centres originated in RTB.

reflects RTB’s origins as a combination of pre-existing projects. As RTB reorients research according to its priorities, the smaller components should become more inter-connected.

One goal of the CGIAR change process is to help the 15 CGIAR centres to better integrate by forming partnerships with a diverse set of collaborators. At first, the focus was to increase interactions among CGIAR centres in order to increase the coherence of the joint research portfolio. For example, often the mandate to conduct research on a particular crop is shared between two or more centres that specialize in different aspects of that crop. In the case of bananas, the responsibility is shared between Bioversity and IITA, since Bioversity specializes in diseases and germplasm characterization, post-harvest and nutrition, while IITA specializes in breeding, integrated pest management and agronomy. If RTB succeeds in promoting a more holistic view of the crop, then collaborations among the centres should increase. Moreover, in the future it is expected that collaborations with non-CGIAR partners will increase. Thus, the evolution of the CGIAR centres’ patterns of interaction will be an important indicator of the influence of RTB.

This effect has already been observed in the relationships between some CGIAR centres. RTB has been highly instrumental in fostering interactions among Bioversity, CIAT and CIP; about half of these interactions originated in the CGIAR Research Program (Table 18). The RTB-induced collaborations between IITA and the other three initial partners represent a relative small proportion of all collaborations. On the other hand, with the exception of Bioversity, RTB induced few collaborations among scientists working for the same centre. Unfortunately, it is not possible to analyse the interactions with non-CGIAR partners because they were not included in the survey.

Geographical sub-networksWhen analysed by geographical focus, the RTB network has clear distinguishing features. Research with a global focus was conducted mainly at desks (where social


All the geographic sub-networks have one large component held together by a few nodes whose removal would break the component into smaller blocks. This feature indicates that while information can circulate within the component, the central role played by a few nodes limits the volume and type of information that can be shared.

scientists and research managers predominated) and in advanced laboratories (by geneticists and plant pathologists). Additionally, global collaborations were less multidisciplinary than the collaborations in the regional sub-networks. On the other hand, research with a regional focus (Africa, Latin America and Asia) was conducted primarily in farmers’ fields and at experi mental stations, while most of the research with a focus on developed countries was conducted in advanced laboratories.

All the geographic sub-networks have one large component held together by a few nodes (cut-points) with strong intermediary roles (high betweenness), the removal of which would break the component into smaller blocks. This feature indicates that while information can circulate within the component, the fact that it has to pass through the central nodes limits the volume and type of information that can be shared.

The global sub-network has 212 nodes with 238 links. It has a major component with 182 nodes and 11 small ones with 6 or fewer (Figure 9). The main component has 7 nodes with high betweenness. The global sub-network has 46 geneticists (21.6 percent), 35 social scientists (16.5 percent), 20 research managers and 20 ICT specialists (9 percent of the network each). This sub-network shows a lower level of interaction between researchers from different disciplines than the regional sub-networks and the different disciplines tend to form clusters connected to the main component by one single link. This is shown in Figure 9 by the concentration of colours in certain areas. Social scientists and research managers tend to dominate at one end of the network

Table 18. Links between and within RTB’s four member centres

Total linksRTB-induced

links %

CIP–Bioversity 39 18 46

CIP–CIAT 24 12 50

CIP–IITA 26 8 31

CIAT–Bioversity 23 12 52

CIAT–IITA 9 1 11

Bioversity–IITA 13 7 54

CIP–CIP 20 4 25

CIAT–CIAT 4 0 0

Bioversity–Bioversity 9 6 67

IITA–IITA 1 0 0

Research with a global focus was conducted mainly at desks (where social scientists and research managers predominated) and in advanced laboratories (by geneticists and plant pathologists. Additionally, global collaborations were less multidisciplinary than the collaborations in the regional sub-networks. On the other hand, research with a regional focus (Africa, Latin America and Asia) was conducted primarily in farmers’ fields and at experimental stations, while most of the research with a focus on developed countries was conducted in advanced laboratories.

(right) while geneticists prevail at the other end (left). Half of the research was carried out in laboratories, especially by geneticists and plant pathologists, while social scientists tended to relate in the context of contractual relationships (22 percent) and desk work. These features reflect discipline-related differences in the way research is conducted.

The largest and most connected regional sub-network is located in Africa. It has 241 nodes with 231 links. The network has 18 components with the largest accounting for 59 percent of the nodes (Figure 10). About 17 percent of nodes in the largest component are plant pathologists, 15 percent research managers, 13 percent plant breeders, 12 percent geneticists and 11 percent social scientists. While the composition of the largest component suggests multidisciplinary interactions, further analysis indicates a stronger presence of collaborations within disciplines. The two nodes with the largest networks collaborate mostly with plant breeders and research managers. The next most connected node interacts mainly with plant breeders and the third most connected node is linked mainly with geneticists. The second-largest component has 35 nodes, and is centred around one key node (black/research management) whose immediate network includes mostly plant pathologists (pink).

The Latin American sub-network has 163 nodes with 141 links. It is less connected than the African network because it is split into 26 components, with the largest one incorporating 33 percent of the nodes (Figure 11).


Figure 9. Global research sub-network and components (nodes coloured-coded by discipline)


Figure 10. African research sub-network and components (nodes colour-coded by discipline)



Figure 11. Latin American research sub-network (nodes colour-coded by discipline)


The most common disciplines in this region are plant pathology (18 percent), genetics (16 percent), agronomy (15 percent) and social science (14 percent). In the main component, a third of the nodes are agronomists and the next largest group are social scientists. Genetic analysis and plant systems are the main types of research (33 per-cent each). No discipline-based clusters are discernible; most researchers have multidisciplinary networks. One third of the links involved laboratory work and another third involved action research projects, reflecting the strong programmes CIAT and CIP have developed on partnerships, learning alliances and value chains. More than 30 percent of research in this region was reportedly conducted on farmers’ fields and 20 percent at experi-mental stations. Half of the interactions occurred on a monthly basis. About 65 percent of the links were formal and 82 percent did not result from RTB.

The Asian sub-network is the smallest and least connected. It includes 83 nodes with 72 links and is split into 11 components (Figure 12). There is a predominance of research managers (19 percent), plant pathologists (18 percent), plant breeders and geneticists (13 percent each discipline). The main component has 33 nodes, is highly multidisciplinary, and is held together by a few

cut-points. The proportion of nodes in the main component is 40 percent. Some of the researchers linking the components together were not actually based in the region. About 37 percent of collaborations involved farmers and other actors (e.g., participatory breeding or conservation agriculture) followed by laboratory work and participatory research projects (about 22 percent each). Also, 30 percent of the work was carried out on farmers’ fields and 26 percent at experimental stations. Collaboration occurred on a monthly (41 percent) to quarterly basis (33 percent). A little over 60 percent of the links were formal and 19 percent resulted from RTB. Genetic analysis and plant systems were the most common subjects of research (75 percent combined).

Sub-network of collaborations linking CGIAR researchers with non-research partnersCollaborations with non-research actors show two clear features: they have a strong geographic focus (mainly Latin America and Africa) and they are mainly research-based. The sub-network that links CGIAR researchers with non-research actors has 114 nodes and 80 links; the nodes include 40 researchers, 35 extension agents, 23 public organizations and policy-makers and 16 people from the private sector (see Figure 13).


Figure 12. Asian research sub-network (nodes colour-coded by discipline)


Figure 13. Collaborations between researchers and non-research actors

Note: Nodes are colour-coded by the type of institution/occupation they represent, as follows: red = CGIAR centres, green = extension agents, brown = public organizations and policy-makers, orange = private firms.

29

Collaborations in the network of the survey respondents are quite sparse, with many isolated researchers and very few reciprocal links. This structure indicates that RTB still does not have an integrated research portfolio.

Collaborations with extension agents: As expected, 30 out of the 35 collaborations (86 percent) took place in farmers’ fields. Furthermore, 17 of these collaborations (49 percent) had a Latin American focus, 15 (40 percent) occurred in Africa and only 6 were induced by RTB. The disciplines of the researchers show a strong concentration: there were seven geneticists, six agronomists, six social scientists and six research managers; the rest of the collaborations involved actors from several disciplines.

Collaborations with public organizations and policy-makers: These 23 collaborations are more variably distributed by location, with six occurring at desks, four at experimental stations, nine in farmers’ fields and three in ministries. Most collaborations (11, or 40 percent) had an African focus, while five (20 percent) had a Latin American focus, another five had a global focus and three (12 percent) were focused on Asia. Only two collaborations originated in RTB. The disciplines of the researchers showed greater variability compared to those working with extension agents: three conducted research on agricultural systems, two were agronomists, two were plant breeders and the rest represented an assortment of disciplines.

Collaborations with private firms: These 16 collaborations have a clear research focus: five were based in advanced laboratories, five were at experimental stations, five involved desk work and the rest were a diverse group.

The map of this sub-network shows very few connections (Figure 13), indicating that few researchers worked with the same collaborators, as would be expected if their research projects followed a more integrated strategy.

Only 11 researchers (28 percent) collaborated with more than one type of non-research partner. Additionally, most components are very small involving two or three collaborators. This lack of diversity is to be expected when the collaborations are not part of multi-actor innovation processes. Further research is needed to clarify whether the interactions with innovation actors occur indirectly through other CGIAR Research Programs.

5.2 Analysis of the network of 92 survey respondentsThe analysis of the full network in Section 5.1 did not allow for the exploration of reciprocal links and redundant paths because most nodes did not complete the survey. To overcome this problem, a sub-network that only included the researchers who provided information was analysed. This restricted data set has 92 nodes and 91 links.

Collaborations in the network are quite sparse. The in- and out-degrees range from 0 to 5 and their distributions are quite similar. As shown in Table 19, five researchers had five out-links (i.e., each mentioned five respondents) and only one researcher had five in-links (mentioned by five people). Meanwhile, 53 respondents did not mention any collaborations with other survey respondents, thus they had 0 degrees. Almost 53 percent (49) of the researchers reported not receiving any information from other respondents and about 58 percent (53) did not send any information to other nodes in this sub-network – an indication of asymmetric collaborations (Table 19).

As in other research networks, the distributions of the in- and out-degrees seem to have ‘fat tails’. However, the small size of the network means that a definitive conclusion cannot be drawn (Figure 14).

The sub-network of survey respondents is quite disjointed, having a main component formed by 55 nodes (59 percent of the network) and 37 isolates that did not interact even among themselves (Figure 15). The main component includes 91 links, of which only 11 (13 percent) are reciprocal. The small number of reciprocal links is likely the result of three factors. First, RTB started as a collection of many existing disjointed projects, such that the researchers had originally developed their networks independently. Second, there is great variation in the level of effort devoted by researchers to particular collaborations, and they may value their links differently. Third, the researcher’s role in the network defines a de facto hierarchy; for example, a research manager has more power than a researcher because the former determines the allocation of resources and may even have a say over the researcher’s

Collaborations with non-research actors show three clear features: they have a strong geographic focus (mainly Latin America and Africa), they tend to be research-based, and few researchers work with the same collaborators as would be expected if their research projects followed a more integrated strategy.


Table 19. Number of researchers with a specific number of links

Number of collaborations


5 5 1

4 3 4

3 6 12

2 11 8

1 14 18

0 53 49

employment. Asymmetric power relationships may influence how a person views a collaboration. Further research is needed to investigate the causes of the lack of reciprocity. All reciprocal interactions but one involve researchers from different institutions and all but three

Figure 15. Sub-network of 92 respondents – main component, reciprocity of links, and isolates

Note: The 55 blue nodes belong to the main component while the 37 red nodes (the stack on the left) are isolates not connected to other nodes. Reciprocal links are shown in red, non-reciprocal links in blue.

of the nodes involved are senior researchers. Seven of the 11 reciprocal collaborations involve the social sciences (i.e., impact assessment, policies, gender issues or research management).

Figure 14. Distribution of in- and out-degrees (sub-network of 92 respondents)

6

5

4

3

2

1

00 20 40 60

Deg

ree

Frequency



Figure 16. Intermediary power of nodes (betweenness) by disciplines (nodes colour-coded by discipline)


Even though all the links in the network are concentrated in the main component, connections there are sparse. The average degree is 1.6. The average distance is 3.6 and the diameter is 9, both slightly shorter than in the full network of 624 nodes. The 2-core has 39 nodes and there are no cores of higher degree. The density of the main component is relatively high (0.11), especially when compared with the density of the whole network (0.002). However, a large number of researchers (37 of the 92) were isolated and therefore were not included in the calculation. This result indicates that at the core of RTB there is a group of researchers who have established a sparsely connected network, while 40 percent of the researchers are isolated from the main group.

The isolates in Figure 15 tended to work more with non-RTB partners. The geographical focus of the isolates’ collaborations was mainly Africa (47 percent), followed by a global focus (16 percent), Latin America (14 percent) and Asia (6 percent). About 58 percent of the collaborators they named on the survey were from advanced or national research institutes (that share is 53 percent for the whole network), 19 percent work at international research organizations (compared with 28 percent for the whole network), 10 percent work for ministries or international cooperation agencies (6 percent for the whole network) and 10 percent work for private firms or NGOs (8 percent for the whole

network). About 18 percent of these links stemmed from RTB. A third of the activities with the isolates’ collaborators were carried out on farmers’ fields, 23 percent in advanced laboratories and 16 percent at experimental stations.

No individual is significantly influential in the sub-network of the 92 respondents. The centralization of the component is 6.4 percent both for the in- and out-degrees. The component centralization index is 6.7 percent, a relatively low value, explained by the fact that most nodes can be reached by more than one path (Figure 16). Research managers are shown in black, social scientists in red and agronomists in blue.

The cluster analysis identified two clusters within the main component. The first one has 27 nodes with 34 links (collaborations). While it includes researchers from many disciplines, plant scientists predominate. With regard to research approaches, about half (53 percent) of the links involved laboratory work followed by participatory research (18 percent). In terms of location of activities, 44 percent of the links involved work in advanced laboratories, 26 percent at experimental stations and 20 percent at a desk. Two thirds of the links had a global focus, 15 percent had a Latin American perspective and 18 percent focused on Africa. RTB spurred 62 percent of the collaborations.


The second cluster has 28 nodes with 48 links. While it also has a multidisciplinary composition, it is more oriented towards the social sciences; 56 percent of the collaborations involved social scientists. Only 23 percent of the links involved research teams, 11 percent were contracted research and 23 percent facilitated the access of local partners to international scientific networks. As much as 85 percent (41 links) of the collaborations in this cluster involved desk work, 80 percent had a global focus and 15 percent had an African perspective. RTB induced 56 percent of the interactions.

The large proportion of RTB-induced interactions in both clusters indicates that, in just one year of existence, RTB has fostered a large increase in collaborations among a core group of researchers. An expansion of this core group in the future would provide evidence that RTB is reshaping research patterns.

Structural equivalence defines a situation in which two nodes occupy similar places in a network and can exchange their positions without affecting the network’s structure. Several measures of structural equivalence were analysed and no evidence of this phenomenon was found. This result was expected, given that the network is relatively large.

Many research networks have a structure known as ‘small world’, i.e., networks that have both local clustering and, on average, short distances between nodes (White and Houseman, 2002). The small world structure is important because it enables easy interaction among nodes. A related property of many networks is a ‘scale-free’ distribution of degrees, i.e., networks in which a few nodes have a large number of links while most nodes have a few links. Networks with a scale-free distribution have high centralization. The small world structure implies that short paths exist, while the scale-free distribution implies that those paths can be found using only local information (Clauset and Moore, 2003). The combination of these two properties defines the navigability of the network, i.e., the possibility that a member can find what he or she is looking for in the network.

No evidence of the small world structure was found within RTB’s networks. In particular, we did not find evidence of clustering, nor did we find evidence of a scale-free distribution, possibly because the network was too small to allow fitting the data to such distribution. These findings indicate that information does not circulate easily within RTB’s networks, especially to the peripheral nodes. On the other hand, the departure of a few scientists would have little effect on the network’s connectivity (as shown by the analysis of the cut-points).

5.3 Centre-based analysisFurther insights on the structure of RTB can be gained by viewing the nodes not as individuals but as affiliates of one of RTB’s member centres: Bioversity, CIAT, CIP and IITA. The analysis was performed with the 624 nodes and 702 links. As expected, there is a very high level of centralization (0.81) because almost all nodes belong to one of these four centres (Table 20).

The normalized betweenness is the ratio of each node’s betweenness coefficient to the maximum coefficient this network could obtain.

Figure 17 illustrates the central roles played by RTB’s four member centres. CIP (pink) has the highest level of intermediation (the most central role), followed by Bioversity (yellow), CIAT (black) and IITA (red). All nodes that do not belong to the four centres are shown in grey. Only one node that does not belong to one of these four centres is an important intermediary: this researcher is linked to CIP and brings to the network 16 people belonging to CIP who would have otherwise been isolated (large grey node in Figure 17, see also Table 20).

The difference between IITA’s betweenness rank in Figure 17 and Table 20 stems from differences in the way UCINET and NETDRAW (the two computer programs used to analyse the data) calculate the coefficients; this issue will be addressed in future analysis. The important intermediary role played by IITA contrasts with the relatively small number of collaborations reported in Table 3.

A better understanding of the intermediary roles of the centres is important because in Section 4.1 it was pointed out that (a) RTB’s impact pathway could not be identified from the reported collaborations since most of them were with research institutions; and (b) RTB induced the interactions among CGIAR centres. During the CGIAR change process, the CGIAR Research Programs were

Table 20. Intermediation of CGIAR centres (betweenness)

BetweennessNormalized

betweenness

CIP 1.685.833 0.813

IITA 1.341.000 0.646

Bioversity 1.237.333 0.596

CIAT 627.500 0.302

Large grey node1 128.000 0.062

Other 77.667 0.037

1 See Figure 17.


Figure 17. Intermediation (betweenness) of CGIAR centres in the full RTB network (nodes colour-coded by centre affiliation)

Note: The size of the node indicates its betweenness. yellow = Bioversity, black = CIAT, pink = CIP, red = IITA, grey = other centres.

Table 21. Density matrix – intensity of links within and between centres

Bioversity CIAT CIP IITA Other

Bioversity 0.030 0.058 0.061 0.042 0.017

CIAT 0.058 0.067 0.050 0.033 0.040

CIP 0.061 0.050 0.078 0.045 0.018

IITA 0.042 0.033 0.045 0.067 0.067

Other 0.017 0.040 0.018 0.067 0

Note: The matrix is symmetrical, so it is only necessary to analyse the lower half.

organized into several categories, including CGIAR Research Programs that conduct research on particular crops, fish and livestock, and CGIAR Research Programs that focus on improving the productivity, profitability, sustainability and resilience of entire farming systems. The first category has a mandate to generate scientific information on specific crops and animals, while the second group must integrate that and other information to improve the performance of agricultural systems, defined in a broad sense (including ecological dimensions). Since IITA and CIAT are lead centres for non-commodity CGIAR Research Programs, they could provide the link between RTB and other actors in the innovation system. This hypothesis is supported by the

fact that IITA has a higher density of links with non-CGIAR collaborators than with the three other centres (the blue cell in Table 21). This evidence, however, is just an indication of a possible behavioural pattern and must be analysed in conjunction with additional information before definitive conclusions can be made.

The density matrix of links (Table 21) also shows the strength of the interactions within and between centres and with other actors. Because the matrix is symmetrical, the upper and lower halves contain the same information. The main diagonal (yellow cells) reflects interactions within each centre. CIP is the centre that most interacts within itself, while Bioversity has the fewest internal


Table 22. External–Internal (E–I) index by centre

Centre Internal External Total E–I

Bioversity 4 27 31 0.742

CIAT 6 22 28 0.571

CIP 36 35 71 –0.014

IITA 2 13 15 0.733

All four 48 104 152 0.368

interactions (indicated by the highest and lowest entries in the diagonal, respectively). In particular, Bioversity interacts strongly with CIAT and CIP (green cells). The strength of this behavioural pattern is also measured by the External–Internal index (E–I index), which takes the value of 1 when all links are external to a group (a CGIAR centre in this case) and –1 when all links are internal. For the 55 connected researchers who completed the survey (see Figure 15, main component), the E–I index has a value of 0.368, indicating a predominance of external links. The E–I indices by centre are given in Table 22. Bioversity and IITA are the most externally oriented centres, while CIP is relatively more inwardly oriented.

This pattern is also evident in Figure 18, where CIP’s nodes form a grey cluster at the right side, and CIAT’s blue nodes form a line at the centre, while Bioversity’s black nodes at the far left interact mostly with nodes from other centres. IITA’s pink nodes lie mostly at the outer limits of the map.

Figure 18 also shows that most interactions established as a result of RTB (thin lines) involve partners from different centres. Also, all reciprocal links but one (red lines) involve different organizations; an indication that these partners value their interactions similarly.

Further insights can be obtained by a two-way analysis of the interactions among centres and academic

Figure 18. Interactions among centres and RTB-induced links (92 respondents)

Notes: Links: red lines = reciprocal links, blue lines = non-reciprocal links, thin lines = links induced by RTB, thick lines = links not induced by RTB. Nodes: black = Bioversity, blue = CIAT, grey = CIP, pink = IITA, red = all other organizations.

disciplines. In Figure 19, the centres (red nodes) are connected through shared research areas (blue nodes). This ‘two-mode network’ is mapped with a method known as multidimensional scaling (MDS) in which the centres appear closer to each other the more research areas they share. As expected, RTB’s four member centres have many research areas in common. The most striking feature of the map, however, is that almost all interactions with other centres involved social sciences, especially impact assessment. The only relationships that could involve disciplines other than social sciences are ‘large natural systems’ with IFPRI (modelling adaptation to and impacts of climate change), seed systems with HarvestPlus, and ‘local natural systems’ with IWMI.


Bioversity

CIAT

CIP

IITA

IFPRI

ICRISAT

ILRI

IRRI

WorldFish

Consortium OfficeCIMMYT

ICRAF

IWMI

SPIA

HarvestPlus

Large natural systems (e.g., climate change)

Local natural systems (e.g., agronomy, agroforestry)

Biotechnology


Crop production

Gender issues


GIS

Human Health & nutrition

ICT

Impact assessment


Pest and Disease management

Policies

Postharvest

Research management

Seed systems

Figure 19. Two-mode network: research areas shared by CGIAR centres

Note: Centres are shown as red circles and research areas as blue squares. The map was produced using multidimensional scaling (MDS).

Table 23. Regional distribution of links by centre

Region Bioversity CIAT CIP IITA

Africa 25 (20%) 9 (7%) 82 (28%) 95 (81%)

Asia 15 (12%) 26 (20%) 30 (10%) 9 (8%)

Europe 4 (3%) 1 (1%) 1 (0%) 0 (0%)

Global 62 (49%) 53 (41%) 94 (33%) 12 (10%)

Latin America 17 (13%) 40 (31%) 81 (28%) 0 (0%)

North America 1 (1%) 0 (0%) 1 (0%) 0 (0%)

Region not reported 2 (2%) 0 (0%) 0 (0%) 1 (1%)

Total 126 (100%) 130 (100%) 289 (100%) 117 (100%)

All centres except IITA have a strong global focus, while IITA has established most of its collaborations in Africa (Table 23). Bioversity’s secondary focus has also been on

Africa, while CIAT has focused mostly on Latin America and Asia. Finally, CIP has developed an equal number of interactions with Latin America and Africa.


The analysis of RTB’s research networks reveals a weakly connected set of actors with few reciprocal collaborations. This structure reflects in part the genesis of the CGIAR Research Programs (including RTB), which originated as umbrella organizations for a large number of pre-existing projects that didn’t have a unifying strategy. Before the start of the current change process, there were few incentives for collaboration among centres. During 2010–2011, the design process for RTB forced researchers to interact intensively, creating opportunities for new partnerships and this trend was reinforced during RTB’s first year (2012). Therefore, RTB’s current structure reflects both prior engagements and new partnerships. Future surveys will help to determine whether RTB evolves to become a more coherent research programme. A second reason for the observed relative lack of collaborations is the large size of RTB’s network. As a human network expands, the number of potential links increases faster than the actual number of interactions because people can collaborate only with a limited number of partners; thus, the ratio of actual links to potential links is usually higher in small networks than in large ones.

The analysis not only provides a picture of how RTB’s activities are currently distributed across geographic regions, disciplines and institutional landscapes, but, by repeating the exercise every couple of years, it can also be used to monitor how they change over relatively short periods of time. This information can help RTB to assess how it is moving along its impact pathway; in particular, whether (a) new partnerships are created, (b) existing collaborations are closed and (c) interactions between the RTB and external partners are strengthened. This knowledge is important for the implementation of adaptive management approaches that take corrective actions as needs and opportunities emerge.

The CGIAR Research Programs were created with the expectation that over time they would reshape CGIAR’s research portfolio, according to predetermined priorities. Our analysis indicated that, while RTB is not yet a consolidated research programme, it has already induced important changes in research activities, fostering greater interaction among CGIAR centres and refocusing partnerships according to the partners’ capabilities and RTB’s research priorities. However, we also found that the research portfolio still lacks coherence, as reflected in the network’s sparse connectivity and a dearth of reciprocal

interactions. This structure hampers the sharing of information across RTB and with partners who are essential for integrating the research into innovation processes.

The survey requested information on all types of collaborations researchers participate in while doing their research. To this end, it distinguished between partnerships (i.e., formalized collaborations) and more informal collaborations. Informality was important for a relatively large number of researchers, especially the senior ones. This contrasts with the trend in many research systems, including CGIAR, to formalize research activities and clearly define ex ante expected outputs and deliverables. It has been shown that when outputs cannot be clearly predefined, contracts are not optimal (MacLeod, 2007; Huffman and Just, 2000). RTB should explore new organizational settings that balance accountability with informality in order to foster creativity and the search for new research methods.

The survey was only partially successful in capturing the different types of research conducted by RTB researchers. From the answers, it was clear that many researchers were not familiar with different research approaches (e.g., the difference between on-farm research and action research). This lack of understanding constrains their ability to explore alternative research models and, therefore, diminishes their ability to achieve CGIAR’s development goals. Developing a cost-effective strategy to explore alternative research approaches and to build the researchers’ understanding of the multiple dimensions of research exceeds the scope of this study, but it should be mentioned that the exploration is a complex process for which no simple recipes exist. Several CGIAR projects have explored new research organizational models (see, for example, Ekboir et al., 2013) but their lessons have not been systematically analysed.

Although it was not possible to fully map RTB’s impact pathway, it was possible to sketch them and to identify several important issues that should be further investigated. First, it is not known what an effective impact pathway for research for development organizations should look like. While the specialized literature and best practices indicate that the pathway should include both research and non-research partners, there are no guidelines on how many collaborations should be included or what the right proportions are for

6. Conclusions

37Conclusions

the two groups. For example, about 80 percent of RTB’s collaborations involve researchers and the proportion of collaborations with non-research partners in the RTB-induced collaborations is smaller than the proportion in the set that were not induced by RTB. This means that RTB has relatively few direct contacts with non-researchers, but it does not necessarily mean that there are not enough of these contacts or that many indirect links do not exist. Additionally, there are indications that effective interventions do not depend only on the quantity of the partnerships, but also on their quality and the research approaches used (see, for example, Ekboir, 2002). The method of analysis presented in this report cannot answer these questions, and complementary studies to explore the complex relationship between agricultural research and development impacts should be implemented.

Second, to fully monitor changes along RTB’s impact pathway it would be necessary to map not only RTB’s networks, but also those of their partners. Mapping all RTB partners’ networks could shed light on this issue, but the pilot project showed that this would be an extremely difficult and costly endeavour. A more manageable alternative is to map other CGIAR Research Programs, especially those that collaborate closely with RTB, to assess the nature and evolution of their joint impact pathway.

Third, most researchers engage in multidisciplinary networks. It is not clear, though, to what extent their research is influenced by other disciplines. Further research should explore how other disciplines and non-research actors influence RTB researchers.

Fourth, the lack of diversity in RTB’s research networks could affect its productivity. Studies of the interactions between research and diffusion networks (e.g., Cassi et al., 2008) have shown that often these networks have different structures and that they share different types of information. It has also been shown that researchers who interact with different types of actors are more productive and more creative than researchers who only talk to other researchers (Rivera-Huerta et al., 2011). To address these issues, RTB should explore new approaches to conducting research as part of innovation processes and to sharing different types of information among more types of actors. It should be noted that the process should not only involve sharing scientific information with non-scientists (e.g., extension agents), but also passing information from non-scientists to researchers so that the latter can better understand the needs of the potential users and the opportunities created by their research.

Fifth, due to the complexity of the processes of research for development and the novelty of the CGIAR Research Program structure, it is not possible at this stage to definitively identify the most effective set of parameters to monitor the movement of the CGIAR Research Programs along their impact pathways. However, it is possible to assert that this set of parameters should include: size of the network, proportions of different types of collaborations (especially disciplines and types of non-research partners and geographic focus), gender dimensions, degree distribution, connectivity (density, betweenness), analysis of components, cut-points and blocks, reciprocity, composition per discipline and whether a small world and/or scale-free network emerges.

Future research should also include mapping several CGIAR Research Programs and exploring their links to shed light on a number of important issues regarding the structure and management of the CGIAR Research Programs. For example, are there substantial differences in the structure and composition of the networks of commodity CGIAR Research Programs and systems CGIAR Research Programs? Do the systems CGIAR Research Programs provide effective pathways for the commodity CGIAR Research Programs? How does CGIAR actually participate in research and innovation activities? Also, repeating the mapping over time will help to understand not only the dynamics of individual CGIAR Research Programs, but also of their interactions, of CGIAR’s research portfolio and of the modes in which it engages partners. Such information would provide important input in a flexible management process, enabling CGIAR to respond rapidly to emerging problems, needs and opportunities.

These observations have important consequences for the management and evaluation of research programmes and CGIAR. While the centres and the CGIAR Research Programs are the operative and administrative units, increasing CGIAR’s impact will require focusing strongly on cross-CGIAR Research Program collaborations. From the perspective of evaluation, our research indicates that looking only into isolated CGIAR Research Programs may overlook important issues that determine their effectiveness; in other words, understanding coalitions of CGIAR Research Programs and the changing interactions between research and non-research actors should be an important component of the evaluation of the CGIAR Research Programs.

In addition to the information on the structure of RTB, the pilot project provides useful lessons about the methodology’s possibilities and limitations. On the one


hand, the information enabled mapping of the networks of researchers and construction of a baseline that will facilitate monitoring RTB’s evolution. On the other hand, the project did not collect financial information or record the magnitude of the research projects, data which could be used to weigh the different collaborations. The use of this information should be explored in future studies of CGIAR research networks.

Most of the questions asked in the questionnaire provided useful information. If the method is used for other CGIAR Research Programs or again for RTB, it would only be necessary to revise a couple of questions, especially the ones that explore different research approaches.

Two actions can be implemented to improve the collection of information. First, a strong awareness campaign should be targeted at potential respondents to explain the importance of the exercise and the lessons the analysis will yield. Second, information for all CGIAR Research Programs should be collected at the same time. Most CGIAR researchers work in more than one CGIAR Research Program; asking them to complete the same questionnaire several times would discourage them from participating. Also, for the information to be used for monitoring the evolution of networks, it should reflect the state of all parts of the network with reference to the same time period.

39References

Argyris, M. and Schön, D. (1974) Theory in Practice: Increasing Professional Effectiveness. Jossey-Bass, San Francisco, CA, USA.

Bennett, L.M., Gadlin, H. and Levine-Finley, S. (2010) Collaboration and Team Science: a Field Guide. National Institutes of Health, Bethesda, MD, USA. Available at: https://ccrod.cancer.gov/confluence/download/attachments/47284665/TeamScience_FieldGuide.pdf

Bozeman, B. and Corley, E. (2004) Scientists’ collaboration strategies: implications for scientific and technical human capital. Research Policy, 33: 599–616.

Cassi, L., Corrocher, N., Malerba, F. and Vonortas, N. (2008) Research networks as infrastructure for knowledge diffusion in European regions. Economics of Innovation and New Technologies, 17(7&8): 663–676.

CGIAR (2011) A Strategy and Results Framework for the CGIAR. CGIAR Consortium Board, Montpellier, France. Available at: http://library.cgiar.org/bitstream/handle/10947/2608/Strategy_and_Results_Framework.pdf?sequence=4

Christensen, C.M., Anthony, S.D. and Roth, E.A. (2004) Seeing What’s Next: Using the Theories of Innovation to Predict Industry Change. Harvard Business School Press, Boston, MA, USA.

Clauset, A. and Moore, C. (2003) How do networks become navigable? Available at: http://arxiv.org/pdf/cond-mat/0309415.pdf

Davila, T., Epstein, M.J. and Shelton, R. (2006) Making Innovation Work: How to Manage It, Measure It, and Profit from It. Wharton School Publishing, Upper Saddle River, NJ, USA.

Ekboir, J., Douthwaite, B., Sette, C. and Alvarez, S. (2013) Dealing with complexity, adaptability and continuity in agricultural research for development organizations. ILAC Working Paper 15, Rome, Italy: Institutional Learning and Change Initiative. Available at: http://www.cgiar-ilac.org/files/publications/working_papers/ilac_working_paper_no._15_cpwf.pdf

Ekboir, J.M., Dutrénit, G., Martínez, G., Torres Vargas, V.A. and Vera-Cruz, A. (2009) Successful Organizational

Learning in the Management of Agricultural Research and Innovation: the Mexican Produce Foundations. International Food Policy Research Institute (IFPRI) Research Report 162. IFPRI, Washington, DC, USA.

Ekboir, J.M. (2002) Developing no-till packages for small-scale farmers. In: World Wheat Overview and Outlook (Ekboir, J., Ed.). CIMMYT, Mexico.

Gay, B. and Dousset, B. (2005) Innovation and network structural dynamics: study of the alliance network of a major sector of the biotechnology industry. Research Policy, 34: 1457–1475.

Heinze, T., Shapira, P., Rogers, J.D. and Senker, J.M. (2009) Organizational and institutional influences on creativity in scientific research. Research Policy, 38: 610–623.

Horton, D., Prain, G. and Thiele, G. (2009) Perspectives on Partnership: a Literature Review. Working Paper 2009-3. International Potato Center (CIP), Lima, Peru.

Huffman, W. and Just, R.E. (2000) Setting efficient incentives for agricultural research: lessons from principal-agent theory. American Journal of Agricultural Economics, 82(4): 828–841.

Klenk, N.L. and Hickey, G.M. (2012) How can formal research networks produce more socially robust forest science? Forest Policy and Economics (in press; available at: http://www.sciencedirect.com/science/article/pii/S1389934112001244 ).

Klenk, N.L., Hickey, G.M. and MacLellan, J.I. (2010) Evaluating the social capital accrued in large research networks: the case of the Sustainable Forest Management Network (1995–2009). Social Studies of Science, 40(6): 931–960.

Kraemer, S. (2006) Science and Technology Policy in the United States. Rutgers University Press, New Brunswick, NJ, USA.

Kratzer, J., Gemuneden, H.G. and Lettl, C. (2008) Revealing dynamics and consequences of fit and misfit between formal and informal networks in multi-institutional product development collaborations. Research Policy, 37: 1356–1370.

References


La Follette, M.C. (1992) Stealing into Print. University of California Press, Berkeley, CA, USA.

Lyrette, J. (2002) Innovations through linkages and networks at the National Research Council: measuring the output. In: Networks, Alliances and Partnerships in the Innovation Process (de la Mothe, J. and Link, A.N., Eds). Kluwer Academic Publisher, Boston, MA, USA.

MacLeod, W.B. (2007) Reputations, relationships and contract enforcement. Journal of Economic Literature, XLV(3): 595–628.

Marschall, M. (2012) Methodological Pitfalls in Social Network Analysis: Why Current Methods Produce Questionable Results. AV Akademikerverlag, Germany.

Newman, M.E.J. (2004) Coauthorship networks and patterns of scientific collaboration. Proceedings of the National Academy of Science of the United States of America, 101(Suppl. 1): 5200–5205.

Newman, M.E.J. (2003) The structure and function of complex networks. SIAM Review, 45: 167–256. Available at: arXiv:cond-mat/0303516

Okubo, Y. (1997) Bibliometric Indicators and Analysis of Research Systems: Methods and Examples. STI Working Paper 1997/1, OECD/GD(97)41. Organisation for Economic Co-operation and Development, Paris, France.

Patton, M.Q. (2010) Developmental Evaluation: Applying Complexity Concepts to Enhance Innovation and Use. Guilford Press, New York, NY, USA.

Palla, G., Barabási, A.L. and Vicsek, T. (2007) Quantifying social group evolution. Nature, 446: 664–667.

Powell, W.W. and Grodal, S. (2005) Networks of innovators. In: The Oxford Handbook of Innovation (Fagerberg, J., Mowery, D.C. and Nelson, R.R., Eds). Oxford University Press, Oxford, UK and New York, NY, USA.

Protogerou, A., Caloghirou, Y. and Siokas, E. (2010) The impact of EU policy-driven research networks on the

diffusion and deployment of innovation at the national level: the case of Greece. Science and Public Policy, 37(4): 283–296.

Rivera-Huerta, R., Dutrénit, G., Ekboir, J.M., Sampedro, J.L. and Vera-Cruz, A.O. (2011) Do linkages between farmers and academic researchers influence researcher productivity? The Mexican case. Research Policy, 40: 932–942.

Scott, J. (2000) Social Network Analysis: A Handbook, Second Edition. Sage Publications Ltd, London, UK.

Stokes, T.D. and Hartley, J.A. (1989) Coauthorship, social structure and influence within specialties. Social Studies of Science, 19: 101–125.

United States of America Government. (2008) Circular No. A–11: Preparation, Submission and Execution of the Budget. Executive Office of the President, Office of Management and Budget, Washington, DC, USA. Available at: http://www.whitehouse.gov/sites/default/files/omb/assets/a11_current_year/a_11_2008.pdf

Varga, A. and Parag, A. (2009) Academic Knowledge Transfers and the Structure of International Research Networks. Working Paper Intangible Assets and Regional Economic Growth (IAREG) 1.3d. Available at: http://www.iareg.org/fileadmin/iareg/media/papers/WP1_03d.pdf

Vera-Cruz, A., Ekboir, J.M., Dutrénit, G., Martínez, G. and Torres-Vargas, A. (2008) Virtues and limits of competitive funds to finance research and innovation: the case of Mexican agriculture. Science and Public Policy, 35(7): 501–513.

Wagner, C.S. (2008) The New Invisible College: Science for Development. Brookings Institution Press, Washington, DC, USA.

White, D.R. and Houseman, M. (2002) The navigability of strong ties: small worlds, tie strength and network topology. Complexity, 8(1): 72–81.

41Annex 1. Description of the questionnaire

Personal information

Respondents were asked to verify information that had been gathered from RTB project documents before administering the survey, as part of the efforts to compile a database of researchers (potential respondents). The information included: name, contact details, country in which he or she is based, title and institutional affiliation, and the specific crop on which he or she works in RTB.

Identification of partners

In the first round of the survey, respondents were asked to name up to 10 of the most important partners with whom he or she collaborated during the 12 months prior to the survey (regardless of whether or not those people receive CGIAR funding). Respondents were also given the option to name additional partners. In the second round, the limit was removed.

For each partner, the following information was requested:

Gender

Male or female.

What type of organization does the partner belong to?

• Advancedresearchinstitute(includinguniversitiesfrom developed countries)

• Internationalresearchinstitute(includingCGIARcentres)

• Nationalresearchorganizationornationaluniversity• Extensionagency• InternationalNGO• NationalNGO• Farmer• Nationalprivatefirm• Multinationalfirm• Farmerorganization,cooperative,CBO,ministryor

other public office (not including public research organizations)

• Multilateralorganization(e.g.,FAO,WFPorWorldBank)

• Independentconsultant• Other.

Interaction models

Six types of research were identified from the literature on management of research: (1) traditional research (in the laboratory or experimental station); (2) on-farm research; (3) participatory research; (4) action research; (5) facilitation of exchange of scientific information; and (6) management of research funds. These labels do not judge the value of the different types of research; each is appropriate for particular objectives and different areas of science. It was anticipated that most researchers would not be able to accurately identify the type of research they were conducting with each partner. Therefore, the survey presented six statements more fully describing the different types of research and asked the researchers to select the statement that best described the collaboration with each named partner.

The actual question included in the survey was: Which of the following statements best describes your interactions with this partner within the project funded by RTB?

1. You work mainly with him/her in labs or experimental stations and occasionally with extension agents. Examples: finding molecular markers for particular traits or providing occasional policy advice to policy-makers.

2. You work with him/her in farmers’ fields replicating the farmers’ conditions. Farmers may be hired to do field work or to manage experiments, but they do not participate in the design and evaluation of experiments. Examples: a farmer is hired to plant a crop in a particular way.

3. You work with him/her (a farmer or other type of stakeholder) as equal partners in the design and implementation of the research activities, but you do not explicitly seek to influence his/her behaviour. Examples: participatory breeding or experimenting on conservation tillage.

4. You interact with him/her to conduct systematic inquiries that directly help improve his/her practices. Example: innovation brokers work with market actors to improve value chains.

5. You facilitate his/her access to scientific information or to a research network, but you do not have joint

Annex 1. Description of the questionnaire


research activities; you may also participate in communities of practice.

6. You contract him/her to perform specific components of a research project.

Physical environment where the research was conducted

The characterization of the type of research was complemented by the identification of the physical environment in which the research was conducted. The options included: advanced lab, regular lab, experimental station, desk, farmers’ fields and partners’ location (e.g., market or ministry).

Intensity of the partnership

Respondents were asked about two dimensions of the relationship: (a) relative importance (i.e., very important, important, not very important, not important at all), and (b) frequency of the interaction in the last 12 months (i.e., daily, weekly, monthly, quarterly and every six months).

Geographic area targeted by the collaboration

The options were: West and Central Africa, East Africa, Southern Africa, North America, Latin America, South Asia, Southeast Asia, Eastern Asia and Pacific, Southwest and Central Asia, Europe, global and national. If the last option was selected, the researcher was asked to specify the country.

Is the partnership formal?

Some collaborations, like communities of practice, are informal, while some become formal after a number of years. The question seeks to understand how important informal collaborations are in the establishment of formal collaborations. A collaboration is defined as formal if it is part of an agreement between organizations.

Did your collaboration with this partner start as a result of the RTB programme?

This question seeks to monitor over the years whether the CGIAR Research Programs are fostering the emergence of research networks. The current survey provides a baseline and future surveys will allow monitoring of any changes in the research activities.

What is the main goal of the collaboration?

Respondents were requested to indicate the main goal of the collaboration with each named partner: research, capacity building or advocacy. Each of the three main goals included the following drop down menus of activities for the respondents to select from:

Research• Breeding(plant,animal,fish)• NewbreedingtoolsforNARSandlocalseed

companies• ‘Large’naturalsystems(e.g.,climatechange)• ‘Local’naturalsystems(e.g.,agronomy,agroforestry,

conservation agriculture) • Post-harvest• Pestanddiseasemanagement• Impactassessment• Genderissues• Valuechainsandinnovationplatforms• Seedsystems• Policies• Germplasmconservation(genebanks,

characterization, in situ conservation)• GIS• Informationandcommunicationtechnologies• Biotechnology• Humanhealthandnutrition• Researchmanagement.

Capacity building• Technical• Mentoring• Facilitation• Exchangeinformation• Thesessupervision• Extension(advisoryservices)• Innovationbroker.

Advocacy• Policyadvice• Participationinfora(globalandregional)• Participationinglobalandregionalinstitutions.

43Annex 2. Definition of the terms used in the questionnair

Annex 2. Definitions of the terms used in the questionnaire

A common understanding of key concepts is necessary to elicit appropriate and consistent information from the researchers who completed the questionnaires. This section describes the definitions that were used in this project. These definitions were provided to the respondents, but no additional activities were implemented to ensure that all respondents understood the definitions in the same way.

Research collaboration

We define a research collaboration as a relationship where researchers effectively cooperate with other actors in the innovation system (including researchers, extension agents, research managers, input supplies, output buyers and policy-makers). Collaborations can be formal or informal; furthermore, partners may have a declared common goal or each collaborator can seek his/her individual goals and still collaborate. In fact, many collaborations start as informal interactions and evolve into formal partnerships. Other collaborations remain informal throughout their whole life (e.g., communities of practice, or decision-makers actively consulting with researchers over several years). Finally, some partnerships are established by formal agreements but never become effective.

What makes a collaboration a research partnership is not its composition or formal goal but the fact that a researcher conducts a research project within it or that it helps the researcher to obtain information and resources for his or her research. For example, an action research project that designs methods to facilitate the participation of small farmers in market chains is research from the point of view of the researcher but a development intervention from the point of view of the market facilitators and farmers. Also, a community of practice is a research network if the researcher obtains information that he or she uses in his or her research activities.

Our definition is broader than the more common concept of partnership. Most of the literature on partnerships in the agricultural sector emphasizes the need to clearly define from the beginning the main parameters of the partnership (e.g., goals, governance, each partner’s contributions). Our definition also includes informal relationships, which are common when collaborations are sustained over time or when collaborators start to explore common interests.

Characterization of research (basic, applied, strategic)

The identification of the main features that define the integration of research into innovation processes can help to monitor how the CGIAR Research Programs move along their impact pathways. Traditional classifications of research are not useful for this purpose. For example, the United States Government defines basic research as a “systematic study to gain more comprehensive knowledge or understanding of the subject under study without specific applications in mind.” Applied research is defined as a “systematic study to gain knowledge or understanding to meet a specific, recognized need (United States of America Government, 2008).” According to these definitions, almost all the research done by CGIAR is applied research, which makes the definition too broad and, therefore, not very useful for classifying CGIAR’s activities.

A more functional classification can be obtained by considering research as a complex process within a larger complex process (innovation). To take into account this complexity, the classification of research activities should simultaneously consider several dimensions, such as the number of networks the researchers participate in; the nature of the partnering organizations; how researchers interact with other stakeholders in innovation processes; the site where the research is conducted (e.g., in an advanced laboratory or on a farm); the disciplinary specialization of the researchers and their partners; the organization in which he or she works; the time devoted to the project and the objectives of the research project. For example, a biotechnologist should have fewer non-research partners than an agronomist working on conservation tillage.

It must be stressed that what defines how research activities are categorized is not the topics that are actually researched but the combination of the several dimensions. In other words, soil dynamics can be studied in the experimental station by an isolated researcher, by a team of diverse partners (including researchers, farmers and private firms) with participatory approaches, or the same team may use action research methods (these types of research are defined below). The differentiation of research on the basis of collaboration patterns does not have a value connotation or a normative content. Some research is best conducted in laboratories while other types can be


performed more effectively in a participatory way. What determines the effectiveness of research activities is that researchers (a) engage different actors in the innovation system, so that the former can understand the needs of the latter and the opportunities opened by their research; and (b) open effective communication channels with other actors in the innovation system so that the scientific information can reach the users (Ekboir et al., 2009; Wagner, 2008).

Using the multidimensional approach, the research activities implemented by CGIAR were classified into six categories:

1. Traditional research includes all activities conducted by groups of researchers, with little or no interaction with other actors in the innovation system, with the possible exception of extension agents. The research is mostly conducted in the lab or experimental station. Examples: finding molecular markers for particular traits, or modelling the adoption of an agronomic practice using survey data (where the only interaction with farmers is the survey).

2. On-farm research is conducted in farmers’ fields seeking to replicate their production conditions but the farmers do not have a say on how the experiment is designed, conducted and analysed. Example: a farmer is hired to plant a crop in a particular way. The original philosophy of on-farm research was to include the farmers in the design and analysis of the experiments, but this principle was not always followed, especially after on-farm research was

mainstreamed into research for development activities. In this study, on-farm research that follows the original philosophy is included in the category of participatory research.

3. Participatory research involves the participation of farmers and/or other stakeholders as equal partners in the design and implementation of the research activities, but the researcher does not explicitly seek to influence the behaviour of his/her partners. Examples: participatory development of IPM practices.

4. Action research requires the researcher to interact with stakeholders in conducting systematic inquiries that directly help to improve their practices. The interactions are jointly monitored to assess how they influence the behaviour of the actors involved and corrections are made as problems or opportunities are identified. Example: innovation brokers work with market actors to improve value chains.

5. Facilitating access to scientific information occurs when the researcher facilitates his/her partners’ access to scientific information, but no joint research activities are conducted; the researcher may also participate in communities of practice.

6. Management of research funds from third parties is one activity of a researcher; in such cases, the relationship with the partners is contractual and restricted to the performance of a specific research activity.

Editing, design and layout: Green Ink (www.greenink.co.uk)

The Institutional Learning and Change (ILAC) Initiative fosters learning from experience and use of the lessons learned to improve the design and implementation of agricultural research and development programmes. The mission of the ILAC Initiative is to develop, field-test and introduce methods and tools that promote organizational learning and institutional change within the CGIAR system, and to extend the contributions of agricultural research towards the achievement of the Millennium Development Goals.

Citation: Ekboir, J., Canto, G.B. and Sette, C. (2013) Monitoring the composition and evolution of the research networks of the CGIAR Research Program on Roots, Tubers and Bananas. Series on Monitoring Research Networks No. 01. Institutional Learning and Change Initiative, Rome, Italy.

series on monitoring research networks no. 01 report... · also graham thiele, gordon prain, graham...

Documents