Online survey of MandE NEWS email list members

Subject: Their membership of other email lists

Date of the survey 3-11 Aug 2006
Number of members of MandE NEWS email list invited to participate 954
Number of  respondents to the email survey 114
As a percentage of all members
Not all respondents to the (individualised) invitations to the online survey answered the survey questions. 96 (84%) answered at least one question. This is equivalent to 10.06% of the total membership.
Survey results:
Percentage of all respondents who belong to the other 29  listed (M&E)  email lists
Equal to 82% of all those who answered at least one question
Median number of other (M&E) email lists they belonged to.
But see the distribution as shown in Figure 1 below
Median number of MandE NEWS members belonging to each of the other M&E email lists.
But see the distribution as shown in Figure 2 below.
Links between M&E email lists, arising from overlapping membership (of MandE NEWS members). Diagram 1
Profile of the respondents Figure 3
Send your comments to Emails go via, so be patient. You may have to answer a "challenge" email, to ensure that your email is not treated as spam.

Figure 1

Figure 2

Diagram 1

Red nodes = email lists that people were asked about
Grey lines = links between email lists, in the form of overlapping memberships
Thicker lines = bigger overlap in membership
Network analysis and diagrams created with UCINET and NetDraw, available from


Email lists can be linked by having some members in common. Those people may chose to pass messages they hear from within one email list, on to another email list that they also belong to. Strictly speaking, common membership only creates potential links, until people choose to make use of their membership of two or more email lists.

The network diagram above shows all the links between the email lists mentioned in question 1 of the email survey. It looks like every list is connected to every other list, almost. With the exception of  three email lists:
EGAD List, Evaluation Feedback, and MONEV_NGO

But many of these links are very tenuous.  Twenty seven of the email lists were connected to each other by one-person links.   On the other hand, in such a dense network there is clearly more than one way of receiving a message from another group. It could go via multiple other email lists acting as intermediaries.

The part of the network that is most likely to be effectively sharing information is that where there are dense interconnections, and where those  dense interconnections are made up of strong links (i.e. multiple people). About half the email lists have links to others made of 4 or more people . These are shown in Diagram 2 below. Within this network, Xceval, EvalChat and MostSignificantChange, have the most number of direct connections with the others, and they are also the closest to the others, when all other intermediary routes are taken into account.

This analysis has implications for how one might use scarce time. Subscribing to all three of these email lists may produce quite a lot of redundant information, whereas subscribing to others that are less densly connected might provide a wider range of information. But a complicating factor here could be the size of those other groups. If they are very small they may not produce much information at all.

This interpretation assumes people do pass information from one email lists to another. Perhaps that doies not happen much at all. What might be more common is that a person who belongs to multiple email lists will "inject" their own new information into each of these email lists around the same time.

Further analysis is needed of this survey data and other related data at hand. For instance, how does size of an email list relate to its connectedness with other email lists? A positive correlation might be expected, with bigger email lists having more connections, because there are more members who each can afford time to have connections with other email lists.  I need to edit the network Diagram 2 so that node size reflects membership size.

And time may be on their side, bigger email lists may be bigger  because they have had more time to grow. The longer an email list has been around, the more time there would have been for others to hear about it and join up, even though they already belong to another email list.
There is data available out there available to be analysed, on the Yahoo website at  Any search for groups on almost any topic will bring up a list, and each one has its membership size on display. Clicking through to the home page of any group will then show when that group was formed (in the Message History) section. Late Note: See some initial analysis via the Footnote below. 

What may be more interesting are the exceptions: the small and new email lists with lots of activity and old and big email lists that are not very active. Connecting up to the active new and/or small networks might help bring new ideas into the wider network.

Caveat?:  The connections shown in Diagrams 1 and 2 are a partial picture. If we also surveyed members of Xceval and EvalInfo they might report different types of overlapping memberships with the other 28 emails lists covered in this survey.  On the other hand, MandE NEWS is one of the biggest M&E email lists, so its membership's survey responses may still be significant indication of the overall structure.

Diagram 2

Figure 3

Footnote: Well, there goes a not very good theory. I looked at Yahoo groups: Top > Science > Social Sciences > International Relations . I copied the details of the top 39 groups. Both their size and age. Here below are two graphs showing the size distribution of these groups, and the relationship between group size and group age.  Most groups are small, and very few are large. Including MandE NEWS. That was not too much of a surprise. But what was a surprise, and should not have been, was that there does not seem to be a relationship between age and size. Possibly because unlike firms (economic entities) that cost money to keep running, small and non-functional Yahoo groups do not tend to die off, because there is no maintenance cost to an inactive group.

Some of you may recognise this as being very near to a power law distribution. These can be found all over the place. See Power Laws, Weblogs, and Inequality  for more this subject.