Original question:

We’re facing a challenge in first-run local elections in Aceh.

The Aceh Independent Election Commission (KIP) needs to quickly come up with a system for verifying nominations for independent candidates at provincial and district levels. At the provincial level independent candidates require more than 120,000 signatures (>3% pop).

My questions are:

• Is there software that can scrutinize and cross-check lists?
• Do you know how this problem has been tackled in other places?

Responses were received, with thanks, from:

• Jérôme Leyraud

• Alan Wall

• Michael Maley
• Domenico Tuccinardi
• Horacio Boneo
• Michael Meadowcroft

Links to the ACE Encyclopaedia and other related resources:

• ACE Encyclopaedia: Candidate Nomination
• ACE Encyclopaedia: Technology for Regulation of Party and Candidates
• International IDEA: Ballot access for political parties
  and candidates

Quote from the ACE Encyclopaedia on signature collection and verification:

"Once an initiative has been filed and the title agreed, its proponents must begin the process of collecting the relevant number of signatures required to get the initiative on the ballot. Signatures and identifying information such as voter registration number and/or address are often gathered on official forms provided by the electoral management body which include the full text or a summary of the proposal.

The number of signatures required can be expressed as a percentage of the electorate or as a fixed number of signatures from registered voters; these thresholds can vary from about one or two percent of the electorate up to 10 or even 20 percent. In addition, some geographical distribution of signatures, in a number or percentage of voters from all or most of the relevant geographical subdivisions, may be required for qualified support.

Signature verification is undertaken by the department responsible for administering direct democracy and will normally be undertaken on a random sampling basis. Opponents are sometimes also given the opportunity to inspect the petitions and to file objections to individual signatures or to the entire submission on the grounds that it does not fulfil the legal requirements.

Usually, a proportion of the signatures collected will be found to be invalid, so proponents generally seek to collect a number of signatures in excess of the actual threshold. If and when the relevant number of signatures is collected, the initiative is included on the ballot (usually in the next election in the appropriate jurisdiction)".

Individual responses in full:

Jérôme Leyraud:

I have a few comments to offer to the question:

First, if you have a computerized voter roll you can indeed intend cross-check names with a name recognition software although you will likely end up getting a lot of names misspelled (in particular because the list provided by the candidates is likely to be hand-typed), hence having a high number of rejections. So if technically it is possible, realistically it’s another matter!  

At some point in the Democratic Republic of Congo (DRC) there were discussions to introduce such a software to both cross-check the voter list and comply with the nomination procedures. It was, however, abandoned for not been adequate. I have not even mentioned the need for a powerful data processing system to cross-check hundred thousand of names against a multi-million FVR.  

Second, although electoral laws may include such provisions, they are in practice not truly implemented, electoral management bodies not necessary having the capacity to process such high volume data. That was the case in Afghanistan where were received candidate nominations with copies of voter registration cards attached to them. The Joint Electoral Management Body (JEMB) never had a chance to go through them and endorsed nominations without scrutinizing the data (the FVR was not even ready!).

In DRC again, the Independent Electoral Commission (IEC) endorsed nominations assuming that lists provided by candidates were accurate. Usually such provisions are implemented through a "good faith" approach.  Random checks can always be made to assess the accuracy of the list provided.

Once again, the issue is less about the law, than how legal requirements fit with a given context and can be effectively implemented. In the context of Aceh, the provisions of the law may just not be implementable! Cross-checking 120,000 signatures for each candidate would be very cumbersome and time consuming for the electoral authorities.

Alan Wall:

On the software question. There are two issues relevant to checking the validity of lists of supporters of candidates' nominations. 

The first is establishing that the person whose name appears on a list of supporters is likely to be a real person. This can possibly be done by matching name data  from the list of supporters (which would need to be first compiled in a database) with another database which is  reliable and comprehensive - such as voter registration, as suggested by Jerome.

However this assumes a number of things:

• a reliable list to match against will be available - the census for voter registration and national ID card reissue in Aceh still does not appear to be finalised, and there is no other comprehensive, relatively reliable database;
• there is accurate conversion of, in Aceh, the more likely handwritten than typed lists of supporters to a database - legibility of the lists that may have been tramped around villages for some time  and accuracy of the data entry staff are both required; and
• each name that appears on the supporters list appears in the same or very close spelling and order as that person's name appears on the reference listing (voters register etc).

At the very least some sophisticated fuzzy matching routines and some pretty impressive computing power will be required. And this sort of matching can still require significant manual follow up/verification - it's not as though a foolproof verified list of names suddenly appears out of a printer.

This test may be enough to knock out some prospective candidates, but passing this test need not necessarily mean acceptance as a candidate as it  does not deal at all with the second  major issue of verifying the list of supporters  - that the signature of each supporter is the real signature for the person whose name is provided.

Aceh law requires that the name and signature of each supporter (no provisions for thumb mark)  be recorded in the list of people supporting a candidate team's nomination, and this entry be supported by a photocopy of the person's KTP - national identity card.  There are no digitally recorded signatures for voters in Aceh.  Any signature comparison would thus have to be done visually - against the copy of the KTP - so why not check both name and signature against the KTP? (this assumes. of course, that all in Aceh have an ID card - which are currently being rather slowly re issued).

I'm not at all convinced that it is worth trying to  half computerise this checking system under immense time pressure (election likely in early  December, with end September/most of October a no work period for Ramadan) in this difficult environment. 

As Jerome notes, burdensome signature support provisions are in many countries noted for not being implemented seriously, or at all, by EMBs, rather than for any innovations in effectively implementing them.  I'm becoming more convinced that if administrations really want to deter frivolous candidates, then monetary deposits are both more equitable (the funding/effort required of an independent candidate to amass 130,000 signatures in Aceh would cost far more than any feasible monetary deposit) and much more administratively effective.

In Indonesia in 2003, when for national party registration around 20-30,000 party member names/signatures had to be verified in some provinces for each new party attempting to register, checks could not be carried out effectively. 

An Afghan/DRC style approach - if a supporter's name has a signature and a current KTP photocopy attached it passes muster - may be more likely to be adopted as the most practical course in the circumstances in Aceh, particularly as the Aceh Independent Election Commissions at provincial and local levels will have more than enough other issues to fully occupy them at the same time. But if it is, in Aceh I wouldn't be broadcasting that beforehand.

If Aceh Independent Election Commission (KIP) does want to undertake serious checking, my current and not fully considered inclination would be to do a random sample check (minimum 5%). If a supporters list fails this check only by a relatively small, defined margin, a more extensive check could be undertaken for final determination. KIP could also usefully first screen out all obvious invalid entries on the supporters list (obvious multiples, fake names, entries unsupported by a signature or copy of KTP/ID card) by eyeball check and make the relevant adjustment to total signatures provided (maybe knocking out some candidates at this stage).

And in any support checking process, KIP needs to consider the protection of the rights of nominated prospective candidates against KIP errors or unfounded decisions.  

Michael Maley:

I had some correspondence yesterday with our friend who subsequently lodged this question with ACE Practitioners' Network, so I won't reiterate it.

But there may be more general interest in the formulae for using a random sampling process - probably the only remotely feasible option - to assess a nomination which requires signature support.

The attached note (enclosed at the end of this consolidated reply) was worked up a few years ago in a different context (checking the required number of signatures on a citizen initiated referendum petition), but I've updated it to refer to nomination signature checking.

The maths looks ugly, but it's logical, and can be set up in an Excel spreadsheet with only minor effort.

Domenico Tuccinardi:

In my view the problem does not lie in the capacity of available software applications to performs such checks (the list of names supporting a candidate against the voter register and the eventual cross-check of the signatures against a digitalised list of signatures); such applications are indeed widely available on the market and can perform the mentioned operations much faster than it is normally assumed.

The challenge is rather the creation of reliable lists of supporting signatures in digitalised form and the reliability of the voter register or census database that can actually be used for such verification as mentioned by Alan. Both elements seem rather difficult to achieve in the current conditions of the process.

I am also of the opinion that such signature verification exercise is not the most important aspect of the entire candidates’ requirement verification process, and might not be worth a significant investment in time, human and financial resources at this stage. A random signature check in the measure of anything between 3 and 5% as described by Michael in his attachment would more than suffice to the case.

Given the current time constraints, there is another issue that must be taken into account. Any candidate who is knocked off the electoral lists should have guaranteed the right to appeal the KIP decision. The entire process might in fact take about a month between notification of the decision to the candidate, time afforded to the stricken off candidate to submit an appeal and time needed to reach a decision from the KIP or whoever else has such competence in Aceh.

Any eventual desire to perform a more thorough cross-check of signatures by KIP might be in reality mask the intent of delaying further the electoral date and would be better dealt with by the measures described by Alan.

Horacio Boneo:

I fully agree with the various points made by other respondents. I just want to add that I can recall a similar experience in Nicaragua. 

On the basis of that experience, which reached the stage of comparing signatures, I can add a few practical issues that might also happen there:

1. When there is such a high requirement of signatures (3% of the population should probably be around 7 to 8 % of the people on the voters’ roll) the intention is not to verify the existence of support to an independent candidate, but rather to exclude them.  That intention implicit in the regulation will also be present in subsequent stages, so all sort of problems can be expected.

2. In Nicaragua it was not possible to conduct the verification exclusively within the Electoral Commission.  The parties insisted in sending agents to verify that the process was rightly done, and some of them really created hell.  Those against the registration of the independent candidate – really, all of them – would challenge every decision, so the whole process took a lot of time.

3. The actual comparison was a nightmare.  Even if at the end it was decided that a sample could be used for verification, it was necessary for the Commission to put all the info of the supporters on a database to check that no one had supported twice a candidate – a common occurrence.

4. The comparison of signatures was another nightmare.  There are differences between signatures collected on the field in difficult conditions, and those included in the voter registration cards done under more adequate conditions.  There were cases of people who presently signed differently – it is quite common for young people to have evolving signatures.  When the mark is just an X, how can you be sure that it is not counterfeited?

5. The comparison of fingerprints was even worst.  All you could get from the field were blots of ink, so that at the end the Commission had to decide to accept them all.

6. The people who collected signatures in areas of low literacy, would usually write the name and numbers of those signing the form, because the supporters couldn’t do it.  Therefore, you have whole pages with the same hand-writing, and all there was from the real supporters was just lots of ink blots and X in the column for signatures.  Parties would usually challenge those pages.  What can be done in those cases except filed checking, which is a nightmare?

7. One last, not from Nicaragua, but from New York.  Years ago, Senator Mc Cain was trying to raise signatures to be inscribed in the primaries of the State.  About 200 signatures from the small town where my daughter lives were declared null because at the bottom of the forms the volunteer had make the mistake of putting Mamaroneck as a “village”, when it is really a “town”.  That was enough for declaring invalid some 6 pages of supporter signatures !!!!

So, if you are not accepting good faith on the part of the people presenting the signatures, you are in deep trouble.

Michael Meadowcroft:

It's ill-advised to have too high a qualifying figure. The problem is that local civic representatives etc tend to want to include in the qualifications for nomination the kind of conditions they would look for in a candidate to vote for.

If the figure is pitched very high a trade will grow up in signatures! I recall that in the Russian presidential elections in 1996 it was $100 per 1,000 signatures!