With elections coming up in the US, I thought it might be interesting to have another go at discussing the problems of fraud, tampering and assorted irregularities. These subjects come up every single time there is an election in the US and it would be nice to be able to go through the most important facts and arguments before the storm hits. As we know, the greatest predictor of whether or not a person will believe an election was fair is whether or not that person's preferred candidate won. Nevertheless, the topic is objectively important and can be discussed without being partisan.

I'd like to propose an open-minded but not conspiratorial or partisan discussion of some of the following questions and subjects (not all at once, you can pick and choose and we have plenty of time):

- what kinds of fraud, tampering and irregularities are there?

- how can we detect them?

- how often do they occur?

- where do they occur?

- what impact do they have on elections at various levels?

- how can we prevent them?


I think that, this time around, we should avoid wasting time on the possible vote-suppressing effects of various voter-ID laws, gerrymandering, voter-roll purges etc because they remain legal and agreement on those issues is unlikely.


Some familiar topics are:

- dead people casting absentee ballots
- ballot-box stuffing
- security-issues with certain electronic vote-counting systems (known hackable machines, systems where there is no auditable paper trail etc).
- glitches or more severe problems with registration, addresses etc.
- ridiculously incompetent election-boards
- statistical tests based on Benford's law
- discrepancies between exit polls and final results



Afaict, although fraud and irregularities seem to occur from time to time in various places at various levels of govt., there is currently little persuasive evidence supporting a significant impact of fraud and various irregularities on the outcomes of major US elections.

In spite of this, the potential threat to the integrity of the democratic process presented by the use of vulnerable electronic vote-counting systems may be worth discussing and I hope that those of you interested in those matters--security, hacking, software design etc--can weigh in.


Re. arguments based on discrepancies between exit polls and final results, I think these are problematic because they rest on the assumptions about the validity of exit polls that are not entirely supported by the available evidence. For example, exit polls don't necessarily give you a random sample because not everyone is equally likely to vote in person, not everyone is equally likely to participate, the polls are conducted at specific locations etc. Some more arguments: http://fivethirtyeight.com/features/...d-ignore-exit/


Evidence based on analyses using Benford's law also seems unpersuasive. The validity of using such methods, for detecting electoral fraud or irregularities in US elections, has not been demonstrated. In some cases, there is evidence suggesting that such methods are in fact useless. Although such methods are still used to analyse elections in countries where fraud and tampering are known or strongly suspected to occur at significant levels, I have yet to see evidence supporting their general usefulness in the context of elections. I would expect these methods to be more likely to return false positives and false negatives than to return true positives.


Given the weaknesses of these arguments, how can we improve our ability to reliably detect electoral fraud, tampering etc?

Talk amongst yourselves