There are statistical fluctuations even with expert election judges. This post about the 2008 MN Sen recount goes into more detail.
Why not count every state every time several times then? It was very stupid to do.
There were actual issues that justified a recount. In that given case, I can't see how anyone wouldn't want to know who actually got x amount of votes, and this election was what first got me thinking about politics, even if it did take many years to become a significant interest.
Just couldn't understand how they could halt a recount in a very close election where there were obvious issues during voting. It's not like we were electing a city council member. This was for POTUS damn it.
Depends on however that State resolves ties. I think most would make it be a matter of random chance, but frankly I doubt any tie after the initial result of a statewide election would still be a tie after a recount.
I agree, and the 2008 MN Sen race is a good example of the amount of swing that might be expected. MN has very good audit procedures for their elections, but even so there can be ballots in dispute. The initial count was Coleman leading Franken by 215 votes out of 2,885,555 cast. After the recount Franken led by 225 votes out of 2,887,337 cast. After the final court challenge the margin favored Franken by 312 votes out of 2,887,646 cast.
There are two factors to consider here. First is the change in the ballots cast. MN ended up with and additional 2091 ballots found to be valid. Missing votes and unreported or partial precincts can happen just due to human or technical errors. In this case it resulted in an additional 0.072% ballots cast which make up less than one in a thousand. However, when the margin is also less than one in a thousand, that matters.
The second is that the recovered ballots are unlikely to exactly mirror the statewide vote. That was true in MN where Franken gained 1,254 votes and Coleman gained 727. Statistical fluctuations can seem large because the added ballots are not uniformly distributed across the state. So in the case of a tie, it's very likely that a recount will find more valid votes, and it is highly unlikely that they will split exactly evenly when those added ballots total in the hundreds or thousands as they would for presidential electors.
Statistically this is not an exact number that can be known in any vote system. There will always be inconsistencies that can't be eliminated - think of voting as polling with a very small margin of error.
What makes my example above interesting is that MN requires an election audit every cycle and it involves a recount of a large number of precincts around the state by a group of experts. Many of those same experts were involved in the Franken recount, so it was possible to compare the same recounted precincts by two different trained experts. The audit results were mostly the same but there were slight differences, enough that the expected margin of error measuring the total statewide vote was greater than the swing from the recount. The race was a statistical tie, and another recount from scratch could just as likely to swing the race back as to keep the official recount result. The effect of the court picking the count that it did was effectively a coin flip from a statistician's viewpoint.
