Wednesday, February 10, 2016

The discrepancies in voter counts says more about our data culture than vote theft

First of all, a round of applause to Javie Ssozi and Evelyn Namara, for surfacing the fact that the tallies in the Electoral Commission voter counting are off by 20,000. Yeah, 20,000! You can now start telling trolls, “you missed the point by 20,000.” 

However are the 20,000 ‘ghost voters’? I kind of wish they were. I don’t like vote theft but at least that would be an active and intentional scheme. More likely though, it is more evidence that the people we pay with our taxes just can’t be bothered to show up at office with their heads screwed on. The electoral commission is headed by a mathematician. Badru Kiggundu was the dean of the faculty of technology at Makerere University for a long time. When he wants to stuff the voter register, he will know to run the data through excel and make the tallies match. In this case though, he and his staff just don’t give a hoot what the voter register looks like. They collected wrong tallies from the different polling stations and just threw them onto the page. They didn’t expect anyone to take a second look at. It isn’t the Ugandan thing to do. (Thanks again Evelyn and Javie, for being the exception). 

Why do I think this is clumsiness rather than malicious register stuffing? I work with government generated data all the time for my data journalism classes and writing work. Also, when I worked with Marie Stopes, a reproductive health NGO, I spent a year in close and painful proximity with the national health information system. It really is sad how much the government just doesn’t care about data quality.  

When we look at police crime reports for our trainings at ACME, the first thing we do is create our own totals. The working journalist will tell you there is no story in the discrepancy between our totals and the published crime statistics. The published police numbers are always off. By the thousands. And how could they ever possibly be accurate? Everything; from how crime data is collected, transmitted, tallied and reported takes the attitude that numbers don’t matter. Essentially, this is how the police generates crime statistics. In the best case scenario. 
  • Incidents are written in that large book at the front desk
  • By hand, someone counts the different kinds of incidents and writes them up on a board at the end of day
  • At the end of the week, someone (by hand) tallies the different categories and walks to a bigger police post to submit their tally. 
  • That bigger post tallies the totals submitted by the smaller ones, adds its own manual tally and sends that further up the chain. On it goes, until the records reach police headquarters in Kampala. From one command level to another new tallies are generated, new errors introduced, typos transmitted, etc. 
The only point at which automation or computerization enters the process is when someone is typing the annual crime report. In this day, when for a few million shillings, you could build an SMS based reporting system enabling each police station to submit their day’s record directly to an automated dashboard at police headquarters, police officers are hand delivering handwritten, inaccurately tallied paper records, as if we are some pre-technology civilization somewhere in the Kalahari desert. 

The recording of national health statistics isn’t quite as rudimentary but not much better. The village level clinics (which government preposterously calls health centre 2) fill out a paper form at the end of the month, take the physical copy to the health centre 3 above them, which tallies that area's statistics and submits that tally to the health centre 4, which tallies a larger area and sends it to the district. In some districts, there is a computer loaded with the district health information system which feeds into the national health information system so automation happens at that level. In others, those hand filled forms will later on, be submitted to the ministry of health here in Kampala. If some are lost along the way, rained on as the district health officer makes their way off a bus in Arua park on a rainy morning, well, then, that's tough luck for whoever cares about national statistics. 

Of course, by the time you generate a final statistic out of this kind of data process, the numbers are schizophrenic. Everybody who could have, has had their opportunity to introduce their own variety of error. The guy who can’t count submitted his wrong tally. The district officer who didn’t want her region to look too bad (in crime or ill-health) has done her doctoring. The overworked clerks have missed digits here and there. You wouldn’t even know where the errors came from. You wouldn’t know how to fix any of it. In any case who wants to fix any of it? 

I suspect, the electoral commission’s data collection and transmission is not much more evolved than the police or health information system. Voter numbers and tallies were sent up the electoral administration bureaucracy as if the commission headquarters didn’t have an electronic version of the register itself. Nobody could be bothered to quality check the information, because really, nobody in this country (except the good folks at UBOS) believes that accuracy means anything. When you don't think data quality is even a thing, you have the electoral commission publishing tallies that are sometimes even less than the right total. That's no way to stuff the register but it certainly is a way to show yourself as stunningly clumsy and careless.

I personally get actively suspicious only when I see wrong numbers coming out of UBOS. Those guys don’t make mistakes. If they publish the wrong statistic, they are deliberately doctoring. Everybody else, just deflates me. In my experience, they are just plain clumsy or they don’t give a hoot about what information they are putting out there. That, on some level is an even sadder reality. The fact that Ugandans default to mediocrity in every job we are given, makes me hopeless. 

No comments:

Post a Comment