The Question
This question is posted on behalf of Granville Abrahams, member of the Practitioners' Network
 
To what extent should inaccuracies of voter rolls be tolerated in a First Past The Post constituency-based electoral system without compromising on the freeness and fairness of an election?

Summary of responses
There are differing views as to how significant inaccuracies in the electoral register are and to what degree they should be tolerated. Several experts posit that inaccuracies are in themselves not a problem if they do not disenfranchise or disqualify legitimate voters from voting, so the degree that inaccuracies should be tolerated depends to a large extent on their nature.

Several replies point out that there is bound to be a certain amount of error in any electoral register. The real issue, therefore, is determining what margin of error is acceptable. There is general agreement that the credibility of an election comes under threat “as the margin of victory approaches the margin of error.” In other words, the closer an election is anticipated to be, the greater the need for accuracy of the voters’ roll.  Another danger is that inaccuracies can be cited and used in the event of disputes over election results. It is also noted, however, that there needs to be an element of pragmatism regarding what level of accuracy is realistically attainable by a given electoral management body (EMB).

It is also key to ascertain whether errors are distributed randomly or if there are patterns which raise suspicion of intentional inaccuracies.

One expert advises that EMBs must be proactive: it is not a matter of deciding what level of error is acceptable; but rather defining first what is meant by inaccuracies, then measuring their scope and extent before finding ways to address and prevent them in the future. The degree to which different types of errors impinge on an election’s credibility will vary depending on the context. With the above approach, it is argued that EMBs will be better prepared to avert a post-electoral crisis and to determine the extent to which any remaining inaccuracies will threaten the credibility of an election.

On a more detailed note, one practitioner advocates for ‘provisional’ ballots to be issued to voters who do not appear on the list on election day. These ballots would then be reviewed to see if they should be counted or not.

There is disagreement as to whether inaccuracies are of more significance in a FPTP system compared with say a Proportional Representation system.

Examples of related ACE Articles and Resources
Encyclopaedia:
• Quality standards of voter lists 

Consolidated replies
• Voter registration standards and country case studies 

External Resources
• Foley, E., (2006) The Legitimacy of Imperfect Elections: Optimality, Not Perfections,
Should be the Goal of Election Administration, [Online] Princeton, New Jersey, Policy
Research Institute for the Region 

Names of contributors
1. Staffan Darnolf
2. Francisco Cobos Flores
3. Stanley Collins Onsembe
4. Ronan McDermott
5. Monte McMurchy
6. John Maphephe
7. Charles Winfree
8. Granville Abrahams
9. Chris Kyriakides
10. Juma Thomas Otieno
11. Atem Oben Henry Ekpeni
12. Hadija Miiro
13. Zekria Barakzai
14. Louis Lohlé-Tart

Inaccuracies are strictly speaking not necessarily an issue. If voters can still cast their ballots in spite of spelling errors, or dead people are not able to vote from the grave the roll is technically speaking still ok. Having said that, these kind of things can effectively erode the trust and confidence of the register and then you definitely have a challenge on your hands.

The first issue that we would need to know is what type of accuracies we are speaking about. Is the voters' roll not inclusive (leaving groups -i.e. political party sympathizers outside), simply outdated (details of some random voters are not accurate) or not comprehensive (all voters in the right age are not there)? In any case, FPTP is more critical with accuracy as one single vote can change the result, in proportional representation systems you have more margin. So, extremely important to get the rolls right, or if not let people vote as much as possible with extra procedures in place to make sure no major fallout occurs after the elections in tight races. 

Inaccuracies in the voters roll are of different nature and there is need to specify the kind of inaccuracies that you are referring to. As Staffan and Francisco have rightly put it some inaccuracies may be hindrances that will cause the voters to be denied access to their right to vote. For example where the voters number on the roll and that on the card are different or appear more than once on the roll and hence being deleted from the main roll. 

Inaccuracies in the voters roll can be used as a source of disputing the final election results of an election especially if its a tight race and the EMB is also not trusted by the various stakeholders for one reason or the other. Inaccuracies of one or two letters misspelled or the gender and or date of birth missing would be acceptable at a 10-15% level. This data categories on the voters roll are not very important on the election day but aer important when running tests to determine if the roll has been manipulated in any way to inflate the number of voters and or if there are dead and or over and underage voters in the roll. 

The Electoral Roll should thus be at least 85% accurate with the endorsement of key stakeholders (Political parties and Local domestic observation groups) having confidence in it.

I'm not sure that FPTP is any more susceptible to voter roll error than PR - in tightly-fought races in Ireland (where I'm from) parliamentary seats can be decided on a handful of votes late in the count. 

Let me quote my dissertation (on VR database quality) directly:

"Setting standards for voter registration databases is a difficult task indeed. Foley* offers a useful reminder of the "problem of perfectionist expectations" when he explains  why some errors are inevitable in a VRDB. 'There is some inevitable risk of inaccuracy in the content of the voter registration database' - Foley continues - '[w]e should, therefore, determine an acceptable margin of error for an electoral system and then hold that system accountable to that standard'. While he advises the setting of standards - he does not offer any guidelines on what those standards should be. Should the threshold for completeness of a VRDB be that it contains 80% or 90% or 95% of eligible citizens? And what exactly is meant by accuracy - is it the correct spelling of names or can it refer to the proper assignment of voters to polling stations? Usefully, Foley stresses the importance 'that errors be randomly distributed' (Foley*) - the implications for a data-driven evaluation are fairly clear - errors that are geographically or demographically focussed are more suspicious as to source and intent, while those that are consistent across the database less so. "

*Foley, E., (2006) The Legitimacy of Imperfect Elections: Optimality, Not Perfections,
Should be the Goal of Election Administration, [Online] Princeton, New Jersey, Policy
Research Institute for the Region. Available at:
http://www.princeton.edu/prior/events/conferences/archives/april_06_2006/PRIOR_9
_Foley.pdf

There are two facets to this. One is quantitative the other political. Or, to put that another way, one is objective, the other subjective. My rule is this - as the margin of victory approaches the margin of error, you have the potential for a crisis of credibility in the given election.

What constitutes error in an election? And how do you measure it? Is a low turnout because of voter apathy erroneous? Is very high turnout in one demographic (say, for example, youth voters) because of a successful GTVO (Get The Vote Out) campaign an error? Inaccurate voter rolls are just one source of error in an election.

As previous contributors have pointed out - there are inaccuracies in the voter rolls that disenfranchise and those that do not. Even this can be subjective - in some countries, overzealous (or poorly trained or partisan) polling station officials will deny a ballot to someone where any discrepancy exists between spelling on the ID offered and on the voter roll entry.

So first define what is meant by inaccuracies in the voter roll. Second, measure these inaccuracies. Third, put in place measures to reduce and eliminate these inaccuracies (and measures on election day to inhibit the exploitation of inaccuracies for electoral gain). These first three must be proactively undertaken by the EMB - do not wait until you have a post-electoral crisis. Now you may be in a position to determine the extent to which residual errors impact the credibility of an election.

Then we get political. If there is any evidence of inaccuracy in the voters roll and the scale of this approaches the winning (or losing) margin, stakeholders can make the claim that the election is not credible. Where an EMB has proactively addressed voter roll accuracy, they are better able to defend the credibility of an election. Where an EMB has ignored voter roll accuracy or VR database quality issues, or is dismissive of unsubstantiated allegations, the complaints may gain more traction. 

As in war, the first casualty is the truth - where there are few facts, the truth is an easy target. 

EMBs must be proactive and must define, measure and mitigate voter list inaccuracies well in advance of any election. It should be a routine activity for the EMB. It should also be transparent and inclusive. This way, stakeholder trust and confidence is built up and inaccuracies that do remain will not threaten the credibility of a given election.

The "freeness and fairness" of an election is dependant on a multiplicity of other factors notwithstanding innacuracies relating to voting polling records. The issue of salience is what the EMB will do once this element concerning poll discrepancy which is indeed an important component of the electoral process is directed to their attention. EMB performance will be crucial as to how this electoral process is to be considered in terms of probative conduct in ensuring that all those who are eligible to vote are indeed able to vote.

(Evrensel, 2010:1) argues that [a] sound voter registration process is crucial to a successful election. Voter registration is not just the technical implementation of an activity; it is a holistic political, administrative and practical process. The role of voter registration is especially important when it comes to emerging democracies: it can make or break an election. The quality of the process and the product – that is, the voters’ roll – can determine the outcome of an election and consequently the stability of the democratic institutions in a country. Trust in democracy is promoted when the voter registration process is open and transparent and allows for the participation of all electoral stakeholders –namely, political parties, civil society organizations (CSOs), the media, security forces, the international community and all potential voters in a country. Importantly, the electoral process should support a culture of dialogue and shared responsibility.

Inline with the above paragraph statement, one will argue that voter registration must be approved by the set international standards, characteristics of being voter register and again some sort of creditability in the entire electoral stakeholders, it should strive to get a rating of 95-98% data accuracy, inclusiveness openness’s and allow registered citizens to vote during polling day.

Clerical errors are inevitable.  In the event that the voters who appear at the precinct do not appear on the voter list for that precinct, the voter should be offered a "provisional" ballot.  This ballot would be sealed, relevant registration information would be recorded on the exterior of the envelope, and the relevant authorities can determine, before a final canvass is conducted, whether the voter is eligible to cast a ballot in that election.   Wrongful inclusion would dilute properly cast ballots, wrongful exclusion would disenfranchise the voter in question.

There must also be a safeguard against the voter presenting themselves at multiple precincts on election day.  Provisional ballots considered by a central authority can also prevent that, assuming bona fide identities are used.

My apologies for coming a little late into this discussion... every time I went over my input I added a little more... and now its too long   I generally agree with most of the points raised thus far; however, I feel that on some points of our brush strokes may be a little too broad.

As was rightly pointed out by earlier respondents, confidence in the electoral outcome being truly representative of the will of the people can be seriously eroded as the percentage of an election’s winning margin approaches the percentage of the voter roll inaccuracy.

Sometimes we need to approach the analysis of the electoral cycle in reverse – comparable to how a reviewer approaches an event after it has occurred.

If we knew how close the winning margin was going to be, we would know the maximum level of errors (both, in the voters roll and the results management process) that could be tolerated by the electoral system.  There in lies the crux – the more closely we anticipate a contested election to be, the more accurate the voters roll must be.

In electoral support projects by necessity we are often required to assume roles wearing different hats – we adopt the roles of: technical advisor, optimist, pessimist and sometimes even the role of a detective. We must perpetually seek out in all critical electoral matters the why, what, how, what if and by whom. As rightly pointed out, as far as the Voter Roll is concerned, inaccuracies in any operation of this magnitude are inevitable. Nonetheless, we must always consider how such inaccuracies can be (intentionally or unintentionally) introduced and possibly exploited for various purposes – sometimes innocently and other times not. It is not a matter of suspicion on our part but rather due diligence.

The Voters Roll accuracy must always be seen within a more holistic context that takes into account:

·      The specific type of inaccuracies; as well as,

·      The mechanisms employed by that specific EMB to ensure accuracy of the voters roll.

·      Another aspect that has bearing on voter roll inaccuracies is the actual polling day procedures and the conduct of polling station staff on Election Day. (The latter bears a degree of variance whose impact cannot accurately be quantified in advance).

We must bear in mind that the possibility exists for errors to be intentionally introduced into the VR data capturing process specifically so as to circumvent the very checks and balances that the EMB put in place to ensure the accuracy of the voters roll. (For the sake of completeness, I’ll put forth a flip side argument to this later on).

 Hence depending on the checks and balances employed by an EMB … not all types of voter registration inaccuracies are of equivalent importance in every EMB scenario!

Voter Roll inaccuracies that do not in themselves (or by actions of polling station officials) disenfranchise legitimate voters may in most cases be inconsequential. A percentage figure of such inaccuracies will in most cases prove of purely academic value and bears little pragmatic impact on the outcome of an election. A word of caution however! The very same type of inaccuracy under another EMB scenario may prove to be more damaging.

Let us say a computer-based system is used by an EMB to detect duplicate voters – in such a scenario, the actual mechanism used to detect duplicates will determine what type of inaccuracies we must consider as being acceptable or unacceptable (also see holistic perspective below). In other words, if say the computer-based methodology relied on certain criteria to be accurate then there may be legitimate reason to reject such a voter based on these specific types of inaccuracy. Basically, any inaccuracies that serve to circumvent the safety checks put in place by the EMB cannot be ignored.

Taking a hypothetical case of a database system relying purely on comparing textual information to detect duplicates – such a detection mechanism will fail when faced with numerous textual data capturing errors (e.g., spelling errors in names). A similar argument would hold for a system relying purely on a National ID field for de-duplication.

Biometric systems can prove of enormous value to the voter de-duplication process when implemented correctly. Unfortunately the Achilles heel of such biometric systems can often be that they are undermined by poor implementation (e.g., data capturing staff with insufficient skills, or even buggy VR kit software). However, when it comes to voter roll accuracy one needs to understand how such biometric systems operate and precisely how even such systems can be circumvented.

·      When it comes to central data processing of biometric systems, there is sometimes a trade-off between the computer processing power required and the time available to complete the biometric de-duplication process. When adequate time or processing power is in short supply, biometric systems may often resort to buckets to reduce the processing workload. Such reliance on bucketing becomes more prominent as the number of voters that need to be processed increases and the time available to complete all such processing decreases! In simple terms, when buckets are used, a person’s biometrics are only checked against others in the bucket he/she matches.

·      Purely for arguments sake, let us assume that a given system uses only two buckets -- one for males and another for females. If a person is registered twice, once as a male and once as a female then he/she may not be detected under such a biometric system configuration. What should happen if such a bucketing system was used and a person turned up to vote that appears registered under the wrong sex? Without adopting a holistic perspective of the electoral system we actually cannot answer this question. For example, we must consider if indelible ink is being used to mark a voter’s finger in this election? If so, then if that persons finger is not marked then most likely he/she should be allowed to vote. If no indelible ink is being used for this election however then the case is not so clear cut… the person may have already voted under a “correct sex registration”.

·      If bucketing is used within any biometric system then it must be clear from the very onset what type of buckets are used and what type of data capturing errors may be tolerated and which ones must be rejected.

The field of biometrics is rather extensive and I’ll leave the discussion there.

Whatever protocol is proposed to accept or reject voters it must be examined for possible exploitation,  bias or ambiguity and, if it passes this check, then it must be applied uniformly through all polling stations.

Whenever we review any electoral process we must adopt both a pragmatic perspective and a holistic perspective.

·      A pragmatic perspective must take into account what is realistically achievable by an EMB within that specific country scenario.

o   As far as percentage of accuracy goes what is pragmatically achievable by one EMB may not be possible for another EMB under different prevailing country scenarios.

§  I also mentioned in earlier replies that we should be cognizant of the cyclic fluctuations in capacities, funding and available skills level within an EMB over an entire electoral cycle. These factors have a bearing on what the EMB can reasonably be expected to achieve in the given timeframe.

o   Thus, in my opinion, EMBs being expected to reach universal accuracy in their voters roll may prove impractical.

·      A holistic perspective must take into account that an Electoral System by design contains multiple checks and balances specifically to ensure that limited shortcomings in the implementation of one part of the electoral cycle does not necessarily derail the entire process.

o   For example, say a person managed somehow to register and appear more than once on the voters roll. However, as mentioned, this does not necessarily mean he/she will be able to vote multiple times…

§  If his/her finger is marked with indelible ink; and

§  The person presenting the voter card is confirmed by visual inspection of the voters photograph, as well as, his/her appearance on the polling station list; and

§  At the moment of voting each voter should be marked off the list.

If all these checks are in place and correctly performed then consider how would such a person manage to vote twice?

o   Proper implementation of the checks and balances can often serve to restrict the impact of earlier activity imperfections.

o   By contrast, improper implementation can often undermine even flawlessly executed prior activities.

Earlier we presented an argument where an individual would have tried to get around the checks and balances used by the EMB. However, for the sake of completeness, the flip side to this coin where the EMB is placed under the spotlight, is also worth consideration. By knowing precisely what types of inaccuracies can justifiably be argued as grounds for rejecting a voter then, “in theory”, such inaccuracies during voter registration could intentionally be introduced with the nefarious intent to disenfranchise specific voters. Hence, after classification of inaccuracies, we must also observe whether we have an even geographic distribution of inaccuracies that may prevent (or have prevented) a person from voting.

One aspect not touched on yet by anybody deals with the post-electoral review. If we do not know:

a) how many voters actually exist; and

b) how many did turn up to vote -- but were turned away owing to voter roll inaccuracies

then can we accurately assess figures for voter turnout or voter apathy?

May I register my appreciation to contributors to this topic of voters' roll inaccuracies to date. Though am submitting my views late, I believe one statement in it will assist a member of the network in their daily experience with electoral management. I may not strictly refer to FPTP but I will make general comments.

In my understanding and as I have stated in some of my contributions to this network before; Elections and its management can be one way of solving inherent conflicts or triggering the conflict drivers. In line with our topic, though the Voters Roll or Register may not be 100% minus error; Yet Efficiency as a one core value of an EMBs will be tested by such inaccuracies.

The existence of such errors may encourage; 1) Legal battles that can be avoided, 2) Saving of costs arising from petitions and by- elections, and to a great extent 3) Animosity and Conflicts. Based on these, it is necessary to minimize errors and if possible not have them for free and fairness to be exercised. As we know, voters disenfranchisement develops out of this, measures should be put in place to avoid it.

Though many EMBs minimize the employment of permanent staff especially for registry, the intended savings of cost may not equate with the repurcurssions of such errors. I suggest that, if Government registry of persons (for Passports and ID Cards) can be executed throughout the year with a lot of commitment and accuracy THEN there is no justification to casual handling of Voters Roll. If the voters roll is inaccurate how will you explain allegations of cheatings.

Tolerance may be exercised to for example name error as the voter may produce a voter's card with all other details captured but this shows much weakness on registration officers. The starting point of transparent elections is preparation of an upto date voters register. I may pose a suppose question:- Suppose coincidentally the errors affect voters "of one candidate" what picture would it portray? A collusion for rigging? Fellow practitioners, from experience, it is possible to have a clean register with all proper planning, logistics put in place, recruitment of right personnel, and supervision. Voters Roll build up is tied to other functions of an EMBs so we should not just discuss the issue of errors as an island.

Inaccuracies are inevitable and it become a major problem if the EMB is unable to react timely to the satisfaction of key stakeholders. We must not only look at the results of the election as a result of inaccuracies on the roll but also the rights of citizens who are eligible to vote and are unable to do so. Inaccuracies that prevent a voter from exercising his or her civic duty is a violation of international instruments.

Whether FPTP or PR or TRS, the degree of inaccuracies in any electoral roll should be clerical and acceptable by all stakeholders. In Cameroon, we had several persons who registered and had registration receipts but didn't find their names on the roll and consequently they had no voter cards. At the local level EMB staff admitted that these citizens actually registered. The EMB had to react immediate by announcing this shortcoming and passed a decision allowing these citizens with registration receipts to vote and the names provisionally inserted in the electoral roll of the polling station and the receipts withdrawn to prevent multiple voting practices. The attribution of polling station to voters was another problem. In all, decisions were taken timely and those that could not be resolved were noted for improvements. Some of these inaccuracies are also caused by voters, especially multiple registration which result to the voter being attributed a polling station far from place of residence.

In all cases, a handful of votes can change election results. In any case large margins in the seemingly winning candidate can be due to inaccuracies in the voter roll. Since there are now mechanisms like use of biometric technology to minimize multiple registrations, the focus for electoral Management Bodies should be to institute all measures, to produce voter rolls that are up to international standards. Use of photo bearing registers also help to enfranchise voters with inaccuracies, in voters' particulars.

Our experience in Afghanistan where we are using SNTV for parlaimentary and local elections and two round system for presidential election, is that VR is not affecting the fraud as the number of people who succeeds to wipe out the indelible ink and use another VR card is very limited and our problem is turnout because of security and in our situation for future elections is that voters with any kind of valid document showing that they citizens and eligble should allowed to vote as we have app. over 5 million duolicates in 17 million VR data base

In my opinion (an following some observations), what really matters in terms of "unaccuracies" is the number of eligible persons missing. You may find hundred of thousands people "forgotten" -- often apparently for organisational mistakes, lack of something (time, roads, vehicles, registrators, machines, ...) or bugs in data processing. But the impact of these "unaccuracies" is interestingly clear when they are strongly associated (in the statistical sense) with a known trend among underrepresented places/groups of population to be less favourable to the actual government.

I regret to hear very few comments about coverage rates and such indicators.... (with the exception in this discussion of Ronan quoted by Ronan).