In African elections, the period between polling and announcement can be
protracted and tense. In the best cases, this intermission is marked by hopeful
candidates urging tense supporters to stay calm. In the worst cases, such
periods are used by politicians to hurl accusations of fraud back and forth to
work up partisanship and devalue electoral institutions. The days between
an election and its results are stressful because incomplete information about
this constituency or that trickles out, but partisans have few systematic
ways to compare these data with past results or exit polling, and worry
that the missing data are somehow being tampered with. This paper shows
how OLS regression using past results to fill in partial results can not only
reduce uncertainty in the short term, but may also point out whether or
not withheld results seem plausible. What began as a simple social media
experiment is presented here as an elegant formula that accurately predicts
outcomes across Ghana’s Fourth Republic and in Nigeria’s 2015 presidential
election. This accuracy was achieved with as little as 10% of the results in,
and extremely biased samples.