Some observations on the European Count and talk of electronic counting

The vacuum of incomplete European election results will be filled today by discussion of the speed of the count and desires to introduce electronic counting.

The verification of the ballot papers had already been completed in the local government count centres, allowing European ballots placed in local government ballot boxes to be retrieved. [The local government and Euro elections will collide again in 2019, and the super councils can look forward to verifying both ballots again.]

countingThe sorting of first preference votes seemed to get off to a good start. Working through the new super councils in alphabetical order, ballot boxes were emptied into huge grey tubs from which sorters picked up bundles of ballots. I counted 108 sorting stations – custom constructions with a 3×4 grid that allow sorters to stack up the ballots for the ten candidates on one side, while other staff ran along the back and gathered up sorted ballots for a particular candidate and fed them through to the checking team and then on for counting.

Staffing levels seemed to be based on the 2009 Euro count with some changes made in the balance between different teams and disciplines. There was a 30% increase in the number of Euro ballot papers between 2009 and 2014 – up from 489k to 636k.

There was a great variation in the speed of individual sorters. Some lacked urgency (though the Chief Electoral Officer is keen to stress accuracy over speed). Sorters seemed to be sorting 1st preferences between 10 and 25 ballots per minute.

Count centres are a great source of information, misinformation and disinformation. Metal shelving along one wall of the hall turns into a giant bar chart as bundles of 1000 counted votes are stacked up under each candidate’s name.

Announcements throughout the morning noted the start and finish of opening ballot boxes from each council area. However, there seemed to be a very long lag between the boxes being opened and emptied into the grey tubs and bundles of counted ballots arriving on the shelving.

Indeed, the stage two exclusion of Mark Brotherston – who’s vote was dwarfed by the number of rejected ballots (the majority of which seemed blank rather than spoiled) – was sorted, checked and counted in a matter of minutes … yet there was a very long delay at the end of the chain before an announcement could be made.

You’d predict that the time taken to complete a stage would be proportional to the number of ballots handled (4,114 in the case of Mark Brotherson’s exclusion, 159,813 when Martina Anderson’s votes needed to be resorted to transfer her surplus) plus a constant administrative time at the end. My observation from last night watching sorters and counters sitting idle for long periods (one was busy revising for an exam this morning!) was that the final part of each stage – double checking the figures, making sure the candidate totals were accurate and labelling ballots was taking much longer than observers could fathom.

Comparisons with counts in other European regions need to come with a health warning. How many other countries are using a multi-stage voting system like STV? Labour rates and staff availability will all have major implications on the end-to-end timescales. But I’d love to see a time and motion analysis of the stages of this major five-yearly STV count.

city hall demoBack at the end of May 2012, Derry-based otp2vote demonstrated their electronic counting solution in Belfast City Hall.

The brochure from the day illustrates the layout they could use in one of the partitioned rooms in the City Hall that are used for council counts.

Observers could watch the opening of boxes and registration of papers; multiple scanning machines reading the ballot papers; adjudication stations where illegible marks could be accepted or rejected; and the final results.

opt2vote software was used by ”all 32 councils in the Scottish Local Government elections” in May 2012. Many other e-counting solutions are also available. e-counting may be faster, but it is expensive to rent the hardware and software, along with the supplier’s staff to be on site to rectify any problems.

Manual counting solutions are very flexible and can work around many unforeseen issues. Hairdryers have been deployed at previous Northern Ireland counts when ballot boxes were left out in the rain. I wouldn’t fancy scanning those ballot papers electronically.

[Note the difference between electronic voting (touch screen computers in polling stations) and electronic counting (paper ballots scanned in rather than sorted and counted by hand).]

And as Comrade Stalin blogged over on Lapsed Communist, the already-complicated STV process will be even less well understood and appreciated by a system that takes a couple of hours to scan in ballot papers and then spits out the results of every stage of the election – exclusions, surplus redistribution and final numbers – in an instant at the end. The pace of the current election system allows candidates to come to terms with potential disappointing results during the day.

Graham Shields, the Chief Electoral Officer for Northern Ireland, is keen to introduce an electronic-counting solution. Earlier today on Good Morning Ulster he noted that a change in legislation would be needed to allow e-counting in Northern Ireland, and that it requires the agreement of local parties to lobby the Secretary of State to proceed.

Elections are not cheap to run, and I don’t believe that the Electoral Office have infinite budget. There would have been the space to nearly double the number of staff in the Kings Hall Pavilion … but would fiscally prudent politicians and tax payers want to bear the burden of the extra wages to shave a few hours off the count?

I’m not at the second day of the count, but with fewer staff available to sort and count, the next few rounds of exclusions and vote transfers may still be painfully slow.

In the meantime, you can listen back to the Stage 1 announcement and hear how the parties reacted to each candidate’s first preference votes being read out.

, , , ,

  • redstar2011

    Usual embarrassment for this backward statelet

    Germany, population 85m had results sorted in 16hrs

  • Just to point out three things. First, turnout was a lot higher than last time, so presumably they were caught short on staffing (and imagine the outrage if they had drafted in too many people).

    Second, the count was just as slow in the South, where they already use STV.

    And third, manual counting represents the ultimate in transparency and the alternative was tried in the South to great acclaim!

  • ctokelly – I’d forgotten to add in the larger turnout. 30% more ballot papers flowing through a more efficiently organised system …

  • Mick Fealty

    [I cannot believe the front page of the BelTel this morning…] This parliamentary count began about 9am, I got there at one pm, and left directly it finished at five the following morning…

    Remember they have been counting in Ireland South since Sunday, and they are also still at it…

  • redstar2011

    The turnout was already expected to be higher as 2elections were on the same say. When this last happened again there was a higher turnout so certainly was not unexpected

  • Actually it doesn’t take very long to count the genuine votes in Tower Hamlets (see Private Eyes for years).

  • The ‘serial’ nature of the STV count is a fundamental problem. *Nothing* happens until a declaration is made. The ballots sit on a shelf, untouched until the opportunity for candidates or agents to challenge the stage result.

    Some time could be saved by paralleling up – pre-sorting the bottom candidate into 2nd preferences while they wait. Still able to gather all those ballots back up onto the shelves if needed.

    But i can see why a more clean room approach is taken to prevent any confusion about where individuals papers are in the process and to prevent larger recounts to get back to a previous state.

  • Reader

    Alan in Belfast: Some time could be saved by paralleling up – pre-sorting the bottom candidate into 2nd preferences while they wait. Still able to gather all those ballots back up onto the shelves if needed.
    Is that a problem with delegation? The person at the top of the tree controlling progress through the steps but refusing to delegate sub-tasks.

  • Brian Walker

    I’m still puzzled why recent counts take so long. in the bad old days of the 1970s when STV was new, results came much quicker on turnouts of around 70%. Honest explanation needed.

  • Worth pointing out that the count is still going on in the South too (as of 13:00).

  • Drumlins Rock

    Went to see the otp2vote counting system last year in City Hall, I’m a fan of keeping a paper trail and transparency of the system, it seemed to tick both boxes, every vote is still visible, allowing for tallies if agents want to keep a check, and although the initial input process will still be quite slow the counting will be immediate. However it should still be conducted in stages and announced accordingly like present.

  • Tochais Síoraí

    Have the counters come back from their lunch yet?

  • Joe_Hoggs

    Painfully slow considering everyone knew the results and order months ago.

  • jagmaster

    They might look into changing the acronym. STV sounds like a nasty rash that will require a trip to your Doctor to get sorted out.

  • Comrade Stalin

    Thank you for the link Alan.

    I see this discussion is continuing in the mainstream media today. The only senior politician who is challenging this situation seems to be Nigel Dodds.

    Talking about electronic voting is premature until we understand why it is that the count is so much slower. I’m trying to wrack my brains and remember, but I am sure in 2009 that the count was complete in full by the early evening with the results being announced. I found this video of Jim Allister’s speech when he conceded defeat at the count in 2009 – the video is dated 8th June 2009 which is a Monday and is the day when the counting started.

    So the question stands. With 20% more votes it took us over twice as long to complete the count. It took us almost as long to complete the first count in 2014 than it took to complete all three counts in 2009. Yes there were 120,000 odd more votes but that doesn’t account for the time difference. This is allied to the situation in 2011 where the STV counts took significantly longer to complete than anyone expected. As I noted in my blog, back then the electoral officer tried to blame STV just like he is doing now.

    The electoral office have changed something in their vote counting procedures since 2010 and we should understand exactly what that is, otherwise we are talking about a solution in search of a problem.

  • Reader

    A measure of the complexity of the count is that there were 7 eliminations (6 were redistributed), and two surpluses to distribute, so more than 70% of ballots (apart from UUP and SDLP) needed to have both the first and second preferences dealt with, and some had several more. I don’t know how that compares with 2009, but if no-one made quota that year things would have been much, much easier.

  • Comrade Stalin

    Reader, indeed, but as I keep saying they took an extremely long time to complete the first count, which is the easiest, requiring no redistributions of any kind.

  • Framer

    But why was the BBC incapable of putting the detailed results up on its website after 48 hours?
    (It still describess turnout as turnip and misspells the province’s name on its council elections results page.)

  • Old Mortality

    I don’t know how the counting staff are paid but might they have an incentive to spin it out as long as possible?