Recounting votes - isn't a recount just as prone to error as the initial count?

Question

In elections, whenever there is a too close margin between the two winning candidates, there is the legal opportunity to initiate a recount in order to minimize possible errors. However isn't the following scenario likely?

Assume the first count lets Candidate A win by 10 votes. There is a recount, resulting in Candidate B by 10 votes. Thus, Candidate B is declared the winner.

Is this

1. even how the system works (is the recount automatically accepted) and

2. isn't this widely inaccurate because the recount is just as subject to error as the initial count? (ie. it is just as likely that the initial count was correct)

Answer 1

Recounts tend to be heavily contested. So rather than just doing a new count and going home, they tend to be the source of various arguments.

For example, in the Florida recount of 2000, each ballot was available for examination by both the Al Gore and George W. Bush campaigns. Both campaigns could argue their assignment to various categories. For example, they could include:

1. Overvote including candidate (i.e. two or more spots removed).
2. Overvote where one candidate is entirely removed and second candidate is partially attached.
3. Undervote, where candidate's opening is partially detached.

The original counters would have counted these differently depending on how each interpreted the rules. In the recount, when it was determined that these distinctions really mattered, both sides have to agree on the interpretations. If they can't agree, they go to court and the court interprets for them.

There are two main reasons why recounts are more accurate:

1. There's less of a hurry. The original count is done at the end of a long day, and everyone is in a hurry to get the count out that night. Recounts are expected to take days and can be done strictly in normal business hours.

2. Both sides oversee the counts directly. If one side thinks that an actual physical count is wrong, they'll insist that it be redone in their presence.

The recount isn't finished until either both sides agree or a court tells them to stop wasting time. It's not as simple as just picking up the ballots and counting them a second time.

Answer 2

No. A recount, while almost never perfect, is going to be more accurate. That's because, while there are going to be human errors, a recount is a process where an error would have to get by multiple counters and observers who are verifying the same information. Also, there are a lot of machine-reading issues where many votes are not counted at all that should have been ("undercounts") where examination of the ballot can show clear intent, or, if there was a machine tabulation error, that machine tabulation error would not affect a human being.

That, of course, assumes a hand recount. In many states, their "recount" process consists of feeding the same ballots through the same machine readers, or, with electronic voting machines, looking at the tapes and verifying totals. Both of those, but especially the second, are a complete farce and will only repeat any problems, and should not technically be considered a "recount."

This is why there was such an outcry for some kind of paper trail on electronic machines when they were first implemented without them, especially after demonstrations of how insecure they were came to light - so they could be hand-verified, if necessary. However, many laws are built so they actually discourage the most accurate means of recounting, ostensibly for considerations of time and expense (not necessarily malice or shenanigans).

Some recount laws are so antiquated (Michigan comes to mind), that, if there is a major discrepancy between ballots cast and the final total, those ballots are EXCLUDED from recounts.

Detroit News: Half of Detroit's Ballots May Be Ineligible For Recount

Answer 3

In the UK, a recount can be requested by the candidates or their agents once they are told the provisional result by the returning officer, and before this result is announced. Once a recount is performed, the opposing candidate is also at liberty to request a further recount. The returning officer acts as arbiter as to whether such a recount is reasonable to perform or not, although if she refuses, then candidates can attempt to take action via an election petition to get the decision overturned. Previous decisions have suggested that two counts in a row generating the same result is sufficient to justify halting.

Since the UK uses paper ballot, the possibility for human error in the recount is obviously still there, however sorting errors should on average be reduced, since this will be a further check on resorted piles, rather than the initial sorting. Further, it will be clearer to everybody as to which doubtful or disputed ballots are significant, and which cannot meaningfully change the result. It should be noted that it's possible for candidates other than the winner or runner-up to request a recount, since parliamentary candidates must deposit a sum of money in order to stand, which is returned to them if they receive at least 5% of the valid votes cast.

Answer 4

You ask a very general question, so it is only possible to give a very general answer. It is less likely that there will be clerical errors in a recount, though one could argue that there has been more opportunity for intentional manipulation.

Recounts are typically not just asked for willy-nilly. It's rare that someone says "I want a recount just because I think there were statistical errors." Typically that individual has some information about what they think was wrong about the previous count. Perhaps they think individuals were not counted properly, or they think the previous counting mechanism misrepresented votes. That information can then be used to tune the recount to properly address these issues.

I suppose that, if you didn't have the foggiest idea why the first count was wrong, but you asked for a recount anyways and it changed the outcome, you might do a third recount. That would at least let you do statistical analyses to see how confident you were in the vote.