SC vote counts and exit polls - how odd is this?

[seaclyr]

Exit polls in South Carolina showed Bush winning by 8% but final vote counts had a much higher margin of 18%. Vote count percentages, however, varied tremendously over election night and did not reach 18% until well into the evening. Taking a look at the numbers as they first came in, early percentages reported by cnn (first image) were strongly pro-Bush but they dropped off quickly so that by 8:16pm EST the Bush margin(%B-%K)was down to 8%, matching the exit polls at this point. Shortly after this, however, there was another big shift to Bush that continued through the evening, bringing the final margin up to 18%.




Differences in actual votes, rather than percentages, provide another take on this. Vote differences (B-K), i.e. votes for Bush minus votes for Kerry, were calculated from the cnn numbers reported over the evening (2nd image; the y-axis shows the difference in votes measured in 100,000s). It's clear that vote differences early on favored Bush but the increase was relatively slow and linear up until 8:16 - 8:24pm. If the slope of the line up until this point had been maintained, the final Bush margin would have been 119,000 votes (the number obtained by least-squares analysis of all points up to and including 8:24pm EST), or 7.9% of the total, which again would match the exit polls. Instead, the slope abruptly increased beyond this point as the vote difference climbed to 276,000 (i.e 18% of the total).




Does anyone know if this pattern is indeed odd - on the surface it seems a curious, quite abrupt shift from numbers reflecting a 8% Bush advantage to a 18% Bush advantage (the difference between exit polls and final vote count) - or might it simply reflect some idiosyncracy in SC regional patterns and voter preferences?