#ACTvotes – but why does it take so long?

For those of you from outside of the ACT, you may not be aware that we’ve just had our ACT Election for our next term of government. The election coverage can be followed in a few places:

Of course, there are a myriad of other places for information too, but these tend to be the ones I use each time there is an election on I’m interested in following. This post hasn’t happened as a result of me wanting to talk about the results (although another minority government in the ACT isn’t anything new), but about the length of time it takes to get a result after each poll.

You see, the ACT has only one house of parliament (most States and the Federal government in Australia are bicameral and have a lower and upper house), and the Legislative Assembly is made up of 17 members spread across 3 electorates. As such, the election uses a quota-based system known as Hare-Clark, combined with a Robson Rotation for listing candidates on the ballot paper. As Antony points out on his blog and in the coverage on the ABC last night, the results can take a long time to finalise because:

  1. A single transferable vote method means that until all votes are distributed to lower preferences, candidates may not meet the required quota and therefore it cannot be determined who has won a seat;
  2. Robson rotation means there are a LOT of different ballot papers (since candidates are in a different order) so the scrutineering process takes longer since you cannot use the position of the preference on the ballot as an indication of who the vote is for; and
  3. Historically, the ACT always ends up with minority governments, so the cross bench needs time to negotiate with the major parties to determine who will be supported to form government.

It amazes me that we still rely on a manual count of votes to determine our winner given what computers are capable of today. Security is often raised as the reason why voting cannot occur over the Internet or via electronic means, and although the ACT has electronic voting facilities available, these are only installed in about 6 or so polling places and are used predominantly for pre-polling. So, although about 20% of the vote is entered electronically, the greater majority of votes are done by filling out ballot papers.

We’ll put the costs of printing and the environmental impact of the campaigns aside for the time being – I want to focus on the actual counting process. I’ve thought about this a lot, and one of the most frustrating things about the whole thing is that any voting system, including Hare-Clark, can be easily represented algorithmically. So, this would mean it would be trivial to write a computer program that could use the voting data to automatically determine the winner of the election in a small amount of time. In fact, one of the tasks I’m setting for some students involves writing a program to do just this, and I’m pretty confident that these Year 10 students with about a year of programming experience will be able to do just that. So, that clearly isn’t a deal breaker in terms of improving the efficiency of the process.

The biggest hurdle, then, would appear to be taking the votes themselves and converting them into an electronic format that could be used by the computer program to generate the result. If electronic voting and/or Internet voting are still a way off, then with the advances in OCR I can’t see why the paper ballots couldn’t be processed using a workflow like:

  1. Ballot papers are scanned into a computer;
  2. The OCR reads the vote to determine:
    1. The position of each candidate on the ballot paper; and
    2. The preference, if any, awarded to each candidate on that particular vote;
  3. The OCR-generated data is then displayed on a screen next to the scanned copy of the ballot paper, and is checked by 2 or more scrutineers for accuracy (this step currently takes place with paper ballots, but this approach could allow the workload to be distributed much more easily):
    1. If accurate, the scrutineers approve the result;
    2. If OCR has generated an error, the scrutineers manually correct the error, then approve the result;
  4. The accurate data is then formatted and stored in a database so that it can be read by the ccounting program/algorithm;
  5. At any stage, the votes that have been entered can be analysed using the algorithm, and this could generate not just first preference votes, but the final result of the Hare-Clark allocation based on the votes that have been collected.

Given the manual process still required to transfer the votes from paper form into an appropriate digital format, it would probably still take an evening to have all of the votes entered into the computer and processed. However, at any time during the evening, a full analysis of the outcome of the vote could be determined (based on a partial vote count) and this would make it easier for analysts like Antony Green to predict not just the parties that would win the seats, but the candidates as well. By 11pm that night, I dare say that everyone would know which candidates had achieved a quota, and by Monday morning negotiations could begin between the cross-benches and major parties with the result of the election known and confirmed.

It seems absurd to me that this kind of process isn’t already in place – not only does it make sense economically, but it’d also mean that our pollies could get on with running the place rather than being forced to wait for two weeks to determine what the composition of the parliament would actually look like.

I’d be interested to hear from people who have either been involved in the scrutineering or counting process, or from one of the analysts like Antony Green himself, to try to determine why we haven’t got a system like the one I’ve described in place already. Given it could be used for any voting system currently in use in Australia (since they all need to have defined algorithms that can easily be programmed into a computer), there has to be an incentive for the AEC to implement such a solution.

3 thoughts on “#ACTvotes – but why does it take so long?”

  1. Antony Green

    The system already is in place. All ballot papers will be scanned this week and the result determined by an electronic count. If you check the ACT Electoral Commission’s website you will find the code that carries out the distribution of preferences, and an actual distribution was published on election night for all electronic votes. The electronic votes do not appear to be a good sample of the final vote.

    1. Bruce Post Author

      That’s great – but why isn’t it done that way on election night then? I mean, I know it is done for the electronic ones (that makes sense), but surely the paper ballots could have just been done as they came in from the booths, so you would have had access to 73% of the votes for your analysis by the time 11pm came around?

    2. Bruce Post Author

      And this is the page Antony is referring to in his reply above – sure enough, the system does exist, but I’d still be interested to know why it isn’t utilised for the count in the first instance.

      http://www.elections.act.gov.au/elections_and_voting/electronic_voting_and_counting/development_of_the_system

      The only reason I can think of is that it is preferable to use the electronic (faster) method for the verification of results obtained the paper (slow) way, because doing it the other way would mean that any fault in the electronic system could declare an incorrect winner and checking the result would take a long time.

      Given that the voting system has been checked for accuracy already though, I still think it seems to be a very conservative approach.

Leave a Reply