Daily Archives: December 28, 2008


Ghana Election Projection: Too Close To Call (again)

Based on 37 constituencies for which I have full results, the current average swing between the second and first rounds is 2.9% from the NDC to the NPP. That would result in a overall majority for John Atta Mills of 28,000 votes. That is definitely well within the margin of error at this comparaively early stage.

Turnout is higher everywhere, except in Mion where a strangely low turnout resulted in a big swing to the NPP. That may bear investigation. On the other side, turnout in Anlo, where the NDC further extended its lead, reached suspiciously high levels and may also bear investigation.

But so far the patterns of voter behaviour appear consistent and explicable and the indications are that the election is broadly fair, despite both parties positioning themselves to cry foul if they lose.

The problem is that, if the result is as close as it looks at this early stage it will be, then small disputes become critically important. I pray that Ghanaians maintain their hard-won tradition of peace and democracy.

STOP PRESS

My calculator is suffering the strain and I have bits of paper covered in figures strewn all over the kitchen table, but with 53 constituency swings now worked out I am projecting a win for Atta Mills by just 8,000 votes. Again, that is so close as to be statistically meaningless.

00.39 I have now calculated the swings from 82 different full constituency results, calculated the swing in each, calculated an average swing, and projected this on to the first round results nationally.

The projection now shows a majoriy for Atta Mills of 20,000. Again still too small to be decisive, but there has been a real consistency to the projection results which reduces the margin of error. I am confident that, whoever wins, they will not obtain more than 50.2% and probably less than 50.1%.

01.31 With swings fully calculated for exactly 100 constituencies, the projection is for a majority for Atta Mills of just 5,000 votes. Again the projections are remarkably stable, but again the result is so close it could easily still go either way.

View with comments