Real time re­port­ing is more trans­par­ent since we have ac­tual num­bers, not statis­tics


The East African - - OPINION -

Iwant to make some ob­ser­va­tions about a nar­row but im­por­tant is­sue that arose dur­ing the re­cently an­nulled Kenyan elec­tion and that is still rel­e­vant as we move to the fresh elec­tion in Oc­to­ber. The mat­ter I have in mind is how elec­tion re­sults are re­ported to the public. I am not go­ing to say any­thing about the count­ing or trans­mis­sion pro­cesses, nor will I make any claims about the re­sults of the Au­gust elec­tion.

What I will claim is that the man­ner in which re­sults were shared with the public on the In­de­pen­dent Elec­toral and Bound­aries Com­mis­sion (IEBC) web­site and on tele­vi­sion over sev­eral days was de­fi­cient from the per­spec­tive of trans­parency and le­git­i­macy.

Af­ter the elec­tion, some pun­dits claimed that the fact that there was a con­sis­tent gap be­tween the two can­di­dates in the re­sults as re­ported sug­gested foul play.

Oth­ers re­sponded that this is ex­actly what one would ex­pect. This was a typ­i­cal fight about some­thing seem­ingly ob­jec­tive — what sta­tis­ti­cal prob­a­bil­ity would tell us about what to ex­pect from elec­tion re­sults, re­gard­less of who was win­ning — where peo­ple’s views have been cor­rupted by bias (whether ex­plicit or im­plicit).

The ba­sic facts are th­ese. The con­di­tion un­der which we would ex­pect that the re­sults from, say, 100 polling sta­tions out of the 40,000-plus used in the elec­tion, would re­flect the true re­sult of the elec­tion is the fol­low­ing: th­ese polling sta­tions must be a ran­dom sam­ple drawn from the to­tal. Ran­dom­ness en­sures that there is no re­la­tion­ship be­tween the fac­tors that led to se­lect­ing the 100 polling sta­tions and any fac­tors that af­fect a per­son’s vote.

If the rea­son why a polling sta­tion is in the sam­ple is re­lated to vote choice, then the sam­ple will give us a bi­ased es­ti­mate of the true re­sult.

If the IEBC was re­port­ing to us ev­ery few hours with up­dates from scores of polling sta­tions that were ran­domly se­lected, we would ex­pect re­sults to come in close to the ac­tual re­sult through­out the pe­riod of re­port­ing. Be­cause of ran­dom­ness, there would be no rea­son to ex­pect any bias in the sam­ples of polling sta­tions se­lected.

The ques­tion is whether the re­sults we were see­ing in Au­gust were the re­sult of ran­domly sam­pled polling sta­tions.

The­o­ret­i­cally, this is cer­tainly plau­si­ble: all polling sta­tions were to open at the same time, close at the same time, and trans­mit re­sults around the same time.

While plau­si­ble, I have no in­for­ma­tion about which polling sta­tions were in­cluded in IEBC’S up­dates to test whether they were in fact ran­dom.

Prac­ti­cally, it seems un­likely that they were. The main fac­tor af­fect­ing when a polling sta­tion would have been in­cluded in IEBC’S sam­ple would be time: when a sta­tion trans­mit­ted re­sults. What fac­tors would de­ter­mine time of trans­mis­sion?

We know that polling sta­tions that are in more re­mote ar­eas or that were af­fected by bad weather would be un­likely to start or fin­ish polling on time.

A sub­stan­tial share of the to­tal polling sta­tions, and dis­pro­por­tion­ately those in more re­mote ar­eas, did not have net­work cover­age, and of­fi­cers were in­structed to leave the sta­tion and move to an­other with bet­ter cover­age to re­port. They would have in­evitably re­ported their re­sults later than other sta­tions.

So it is ex­tremely un­likely that the se­lec­tion of polling sta­tion re­sults re­ported by IEBC was ran­dom. What we do not know is whether those non­ran­dom fac­tors, such as the weather or re­mote­ness of the sta­tion, were re­lated to how peo­ple voted. If more re­mote or marginalised ar­eas of the coun­try leaned to­ward one party, while less marginalised ar­eas leaned to­ward an­other, then we would ex­pect that the votes com­ing in ear­lier would be skewed to­ward the pre­ferred can­di­dates of the less marginalised ar­eas, and the re­verse would be true for re­sults re­ported later.

There may have been other non­ran­dom fac­tors af­fect­ing the time when re­sults came in.

For ex­am­ple, sup­pose that sta­tions with more polling agents present were more likely to fight about the re­sults be­fore cer­ti­fy­ing them, and sup­pose that there were more polling agents in more hotly con­tested ar­eas.

This would mean that re­sults would come in faster from sta­tions with ex­treme votes for ei­ther party, and slower for sta­tions where can­di­dates were more evenly matched.

I make no claims about whether th­ese things are true, only that they are plau­si­ble. If they were true, they would mean that a con­stant gap be­tween can­di­dates was less likely.

The best way to re­port re­sults to the public is con­tin­u­ously, as they are re­ceived from polling sta­tions, rather than via pe­ri­odic sam­ples.

Real time re­port­ing is more trans­par­ent, and al­lows peo­ple to quickly spot anom­alies where a par­tic­u­lar polling sta­tion seems to have a vote in­con­sis­tent with past ex­pe­ri­ence of how peo­ple vote in that area. We don’t need statis­tics when we have the ac­tual num­bers.

The ques­tion is whether the re­sults were of ran­domly sam­pled polling sta­tions.” What I will claim is that the man­ner in which re­sults were shared with the public on the IEBC web­site and on tele­vi­sion over sev­eral days was de­fi­cient from the per­spec­tive of trans­parency and le­git­i­macy

Ja­son Lakin is head of re­search for the In­ter­na­tional Bud­get Part­ner­ship. E-mail: ja­

Newspapers in English

Newspapers from Kenya

© PressReader. All rights reserved.