I’ve blogged previously (here and here) about some of the errors which came to light in the London election results. None of these were serious in their own right, but they do cast doubt on how good the checking processes really were. Now a comprehensive review has suggested that up to 41,000 votes went unaccounted for:
Although the glitches are unlikely to have changed the overall result of the mayoral race – Boris Johnson won by almost 140,000 votes after second choices were taken into consideration – today’s report lists a string of potentially serious problems with the counting procedures used in the mayoral and London Assembly elections.
Among the deficiencies highlighted in the report are the counting of blank ballots as valid votes, frequent jams in the scanning machines and a series of bugs and system freezes. ORG’s observers also reported that they were refused access to parts of the process at counts in London Olympia and Alexandra Palace.
In at least two cases, the margin of error was greater than the winning candidate’s margin of victory, leading the group to conclude that there was “insufficient evidence” for it to say that the results were accurate.
That’s not good.
One Comment
I have worked in the audit of technology systems for many years and my advice to anyone is “Don’t let your democracy be subverted by it”. Take the systems that support “Hole in the Wall” cash machines on the High Street. The one key piece of evidence that they are functioning correctly is “How many customers write after checking their statements to say ‘I didn’t take out that £50’. If it is very low, the system is working satisfactorily. However, this check is not available for voting.
The key (and possibly the only worthwhile) check at the counting stage is party members scrutinising the counters. I remember a count in Colne where a pile of 50 ballots were counted as Conservative, where only the top two were, the rest were ours. Our counting agents spotted it and we won by eight votes. I’m sure Mr Greaves will remember the exact details. You can’t check to that level of detail how a machine has worked out its result.