Developing network models of development projects: An introduction > Network Examples > Networks changing over time
Water research in West Africa: Identifying possible impact pathways

The first model
The network diagram below shows the linkages between eight research projects being undertaken in 10 countries, with 5 participating institutions, who themselves are spread over 4 continents. Square nodes are "the projects" and round nodes are the institutions participating in those projects. Green round nodes are participants located within West Africa, while red nodes are located elsewhere elsewhere in Africa, or on other contintents. These eight projects form part of a larger global Water and Food program, of more than 50 projects in three continents.


To what extent does the network presented represent what is expected to happen?  One program concern is that research findings have an influence beyond the immediate sorroundings of each project. And that some progress is made with identifying potential impact pathways, whereby this wider impact might be realised. One set of pathways that might be available, and utilised relatively early on, are the linkages between the participants, across different projects. In the network diagram above there are eight organisations which link participants in the different projects, four of which are in Ghana and four of which  are outside West Africa.

The four Ghanain organisations could be: (a) the subject of program interventions aimed at enhancing cross-project influences, and (b) they could also be subject of program monitoring activities, aimed at tracking whether cross-project research influencing is happening as expected, or not. But before then, a more detailed network analysis might help to focus those efforts. Some of the bridging organisations are better connected than others. CSIR-WRI has a higher "closeness centrality" and higher "betweenness centrality" than the other participants. Next in order on these same measures are IWMI, then ZEF and INERA.

This analysis may still be too simple because it has not taken into account any of the attributes of the participants in the network, all it has looked at is the structure of the relationships. But some participants may be more capable than others, of acquiring information and sharing it with others. Or, there may be important status differences that are likely to hinder or enable the flow of information between any two participants. Or, geographical location might have a big effect on information flows. A decision needs to be made about which of these attributes is likely to have an influence. Then it could be built into the network model, by weighting the relationship each participant has with the others, in the way suggested earlier in this network document. (If multiple factors are likely to be important, then they could be aggregated into a single attribute score, weighted by relative importance if necessary)

As well as looking at the linking organisations it is also worth looking at those participants on the "margins". Project 34 is conspicuous in that it is linked to other projects by only one participant (CSIR-WRI), whereas the other projects have between two and four such links. The sole participants in Project 34 are at greater risk of not hearing about the other projects and not being able to influence them. Their sole intermediary (CSIR-WRI) might become an unintentional or intentional bottleneck in the flow of information between that project and the others.  This potential problem could be managed by linking some of the sole particpants in Project 34 with participants in the other projects, via periodic workshops, or participation in standing committees.

The second model

The eight projects are linked in other ways. One important potential linkage is the overlap in the contents of the projects - what they are trying to do. This could be a major motivating factor influencing who talks to whom.  Or, it could influence the views of program management about who should be talking to to whom.

A partial analysis has already been made of commonalities of research interests across the eight projects, and is available on the program website. Project Managers of 31 projects were given a list of 21 key words, and asked to allocated 100 points across these key words, according to the extent that their project addressed those topics.  Four of these project belong to the set above (and possibly others whose identities could not be established via the information available). The network diagram below show how these four projects are linked by their reported interests.  The red nodes are the key word description of topics, and their size varies according to the number of links projects made to these topics.



In the first network diagram above these three projects had either 1 to 2 links with each other (average 1.6) But because this second (but partial) network provides more details (i.e. more linkages per participant) we can see more visible differences between the projects. In this  model project 47 is the most inter-connected of all the projects (it has 13 links with other projects), versus the others which have between 6 to 8. Amongts those projects with similar numbers of linkages, we can see the differences in the contents of their conenctions with others. Projects 40 and 6 have the same number of connections, but they have some non-shared interests in common with other actors (see the bottom right and bottom left key word nodes).

This model could be made more sophisticated by taking account of the strength of each participants interests in each of the key word topics. This information was provided by the 100 point system they were asked to use, but has not been shown in the diagram above. A topic in which two participants both have a strong interest might, or should be, expected to provide a stronger link than one where their interests are relatively weak. 

The relevance of the detail provided by this second network is that it can help set more detailed expectations about the types of information that might, or should, be shared between these participants. It also suggests which projects might, or should, be more successfull in sharing information (because of their greater commonality of interest with others (i.e. more links).

Caveat: All the above analysis was made on the basis of information publicly available on the CGIAR Challenge Program on Food and Water website at www.waterandfood.org The analysis has not been sanctioned by CGIAR and any errors in the above data may not be their responsibility.