Since last year, I’ve been doing some work with Zimbabwe Peace Project (ZPP) on human rights monitoring stuff. In the course the work, one man Perl strikeforce Sam Smith coded a script which claws its way through ZPP’s Human Rights Monthly Monitoring Reports (MMRs) and makes the content more accessible.
The MMR is an information-rich rundown of politically-motivated acts of violence which its 240 field monitors investigated that month (here‘s a sample, for July 2007). As you’ll see, the documents are structured along geographical and chronological lines: Region > Province > Constituency > Incidents ordered in date order. There are easily 500 incidents in each report, which has been produced monthly since late 2002: that’s more public domain information about Zimbabwe than you could shake a stick at, but its format makes it very hard to get at, even for the authors. Sam’s script takes this content and puts it into an Excel sheet, allowing a better measure of re-use and quick analysis than is possible from the document itself.
What’s striking in seeing this vast amount of retooled information (or spreadsheet of horror, as a colleague named it) is the absence of a stable, detailed geography underpinning the recording of incident information. Constituencies can be changed, so it’s probably short-sighted to use them as the main locational value when recording or processing data from incident reports.
Just such a thing has happened on quite a grand scale at least twice in the last five years in Zimbabwe. I have just seen the list of freshly updated constituencies for the forthcoming 29 March 2008 election in Zimbabwe. There are now roughly double the number of parliamentary seats up for grabs, but how have the boundaries changed? In the absence of accurate geographical data, though, it’s not clear to me how the constituencies differ precisely and which areas would now find themselves in different constituencies.
For any monitoring organisation, boundary changes are a nightmare. Obviously, an incident happens in a place irrespective of the constituency it’s in; if this precise location isn’t captured in a database or somesuch system first time, the original records – most likely paper monitoring forms – will have to be hauled out and re-processed. This radically increases the cost of making useful comparisons between patterns of violence currently being experienced and those observed in previous elections.
To avoid this problem, future monitoring efforts should make sure that precise locations are recorded first time. So, here are two questions for our five or so readers: what’s working well on this issue in the real world; and, what’s the most practical way to manage information about electoral boundaries?
Update – 25/02/2008:
On Disruptive Proactivity, Sam has responded in more detail about his part in this work, with some smart comments about how to resolve the geography issue.
