This question was posed on behalf of an ACE user.

What factors influence the performance of Biometric Voter Registration (BVR) systems globally?

Summary of Responses

Practitioners’ Network members suggested a range of considerations for the performance of Biometric Voter Registration (BVR) systems. The most common suggestions are consolidated in the themes below.

Support

A member from IFES suggested that while specifications of biometric kits are important, it is also vital to consider the capacity of the implementing body, or the abilities of the implementing staff to operate the equipment and the process. As he noted, the performance of BVR systems is intrinsically linked to the support that the equipment has. Several other members also emphasized the importance of system maintenance, and institutionalizing the process so the equipment delivers the intended results. In line with this, another member suggested that technical partners are crucial for installing and maintaining biometric systems, and EMBs must identify appropriately experienced contractors.  

Testing

Additionally, another consideration several members suggested was testing. Piloting BVR systems is integral to their successful implementation. As a member noted, the equipment must undergo technology tests (to understanding error rates, response time, etc.) and scenario testing to understand environmental effects on the system (e.g. operational temperature, load effects). As one member suggested, a significant lesson learned from Kenya’s 2013 general election is allowing sufficient time to test and pilot such systems before they are deployed.

Buy-In

A few members suggested that while there are technological and system-related factors, acceptance of BVR systems is also important. For example, one member suggested that very rarely are populations asked whether they feel comfortable with sharing their biometric information.  As another member noted, in order for biometric systems to function “citizens [must] understand the need of using biometrics and [they have to] accept it.”  In line with this, a member shared that in Kenya, rural communities stayed away from BVR due to rumors that it caused cancer, barrenness, and erectile dysfunction.  Furthermore, one member suggested that implementers must also have buy in from the government and other stakeholders to avoid system failure (e.g. the government may refuse to release funds, contest the EMB, or de-campaign the registration system).

Other Considerations

Lastly, a member with extensive technical and management experience in the Information Technology field shared a set of considerations for every step of implementing a BVR system. 

The 4-phases of the process the member laid out include:

1. Live capture of finger prints (FP) from the field
2. Building and running the central voters’ database
3. Running the Automatic Fingerprint Identification System (AFIS)
4. Actual use during the election

Across these stages, the member outlined a series of important considerations (see member’s contribution in the original discussion thread below).

Contributing Members

• Staffan Darnolf
• Ahmed Gedel
• Ernesto R. del Rosario
• Henry Makabayi
• Jersa Kide Barsaba
• Francisco Barrera
• Víctor Hugo Ajila
• Ronan McDermott
• Vincent de Paul Taty
• Roger Thord-Gray
• Jesús Antonio Castellanos Vásquez
• Mahouton Marius Elvis Djossou
• Michel Chajes
• Woda Jago
• Idi Boina
• Abdul Aziz Mbond
• Amin S. Wasike Yusuf
• Abdiwahidi Hussein
• Manuel Wally
• Hadija Miiro

In addition to the specs of the biometric kits, how well the staff can actually operate the equipment and the registrants understanding of the process are additional important factors. Performance of the BVR system is, of course, also intimately related to the actual support/services to the equipment itself. Do keep in mind that the actual processing time for a registrants is often very different to what you achieve under ideal conditions in the capital city.    

The factors that can influence the performance of Biometric systems may vary from country to country.Some of the factors are 1.Biometric Equipment  kits Selection.This is very important aspect that can influence the Performnce. The failure of a country to conduct appropriate selection affects the performance.2.Acquisition Methodology 3. Installation/Commissioning 4. Testing.There are three testing methods.i.Technology testing which involves ascertaining the error rates,the sensitivity of the sensor to provide the accepted image quality ie PPI agreement in the RFP document,Response time,etc.ii.Scenario testing:This is to consider environmental effects which include operational temperature,Loading effects,etc iii.Piloting:This involve taking a random samples of the equipment and allowed it to take measurement/registration.In this case you  an evaluation report is then written for all the three tests before the equipment is then opened for public usage.

The next is to established a maintenance programme that ensures that the equipment provided the needed results expected and accepted by all Stakeholdrers etc  etc

Part One

These are my thoughts right off the top of my mind based on my experience on setting up and managing end-to-end a 50-million-voter fingerprint-based BMR system. It will take four  parts to do this.

Firstly I am bounding the term performance of a Biometrics Voter Registration system within four phases: 1) live capture of fingerprints (FP) from the field, 2) building and running the central voters' database, 3) running the Automated Fingerprint Identification System (AFIS) 4) actual use during an election.

For the live capture stage, what are critical to good performance are the ease of use of the FP capture station (which normally captures also the picture and demographics of the registrant) for this will be used by field people and non-techies; the efficiency of the capture session (a completion turn-around of about 3 minutes or better would be a good target to minimize the queue during crunch times), very important as well is the one-time correct  capture of the FP images (an automatic FP quality accept/reject routine should be built-in in the capture software to avoid the need for a re-capture later which can be messy and costly); lastly is a good quality data transmission of the field-captured voters' information to the central processing and voters' database build-up which can be in physically transmitted batches on media or online.

Follow Part Two....

Part Two (of 4 parts)

For the central voters' data base build up, the important performance drivers are are: 1) an efficient staging  system for the field-captured data whether in physical batches on media or online transmitted prior processing and build-up in the central database; 2) a very good Data Base Management System (I used Oracle) as the core software for the central database (the voters' database once built is a permanent and continuously growing database of the country so a very good DBMS with a stable company behind it is very critical - meaning avoid cheap, here-now-gone-tomorrow DBMS products); 3) a high-performance, high-throughput, and high-quality (laser) printing system (avoid desk-top/office printers which are inefficient and will surely conk out given the volume of the printing that will be done - remember elections cannot be postponed) with adequate capacity buffer.  Printing of the election day field voters list with picture and voter sign-off will be done here.  Of course, an online voter authentication option can be used but this will need an enormous bandwidth on election day and election day ONLY - expensive and wasteful during the intervening years between elections.

Follow part Three....

Part Three (of 4 parts)

For the AFIS Facility or database cleansing using FP-matching, the critical performance determinant is: A highly accurate, high performance and stable AFIS product (NEVER from a fly-by-night company) for this will be the core functionality of the entire BMR system -  correctly SIZED and pilot-tested (to near perfection !).  This will require an adequately sized  AFIS at the needed accuracy which means it will deliver the prescribed FP match detection/discriminating accuracy (much higher than 99.99% TAR perhaps) depending on what is needed by the biggest election of the country (national usually) addressing the closest possible margin (of 1 in a simple majority-win type of elections) between the just-winning and just-losing candidates in the smallest race in the same election. Delivery of the cleansed database should be within a prescribed window prior an election.  Here the EMB can be a bit creative (highly suggested really !) for the first cleansing effort will most likely need a very large-capacity and expensive AFIS to finish within the window for use in the first targeted election.  The needed capacity then tapers off tremendously (depending on the annual voter population growth) between succeeding elections (in my case a 10 to 1 drop).  So the EMB might want to outsource the first massive cleansing effort and then buy a less-capacity facility after from the same vendor preferably to tide it over all succeeding elections.   Of course, the AFIS selection is a very technically-rigid process (too long to include here) - this must not be done in a rush-to-disaster mode.

Follow Part Four...

Part Four (of 4 parts)

For the actual use during an election phase, the sole critical performance parameter is an efficient and cost effective voter authentication system using the cleansed central voters' database which can be of two basic types:  1) centrally batch printed election-day voters list (off the cleansed central database) with voter picture and voter sign-off feature disseminated shortly before election day   or 2) on-line real-time authentication.  As mentioned in Part Two, online voter authentication can be unnecessarily expensive due to the very high bandwidth requirement (except maybe if there is a massive and pervasive volume of flying voters expected on election day) for its utility is on election day ONLY and remain highly under-utilized in between elections.

End Notes

I wrote this long comment from the practical, running-an-actual-BVR perspective for I believe it can be of some use to some colleagues.  A BVR selection process is normally covered extensively in identification and elections technologies literature.

Of course there can be other biometrics technologies that can be used in a BVR system like voice, iris, palm print, facial, odor, gait, and even DNA (maybe ?) but the fingerprint-based one is still the dominant one used at this time.  And also, there might not even be a need for a BVR at all depending on the "honesty" of the conduct of elections in a country.

Hope this helps.

End

The law, equipment and software, cost, environment, security, perception, power source, details captured, time and personnel are among the things that affect biometric voter registration.

The enabling laws may detail confidentiality of the biometrics collected. 

Equipment and software affect the quality of biometrics such as photographs and fingerprints that will be collected.  People involved in daily manual work like farmers using hand hoes and mechanics who do not use gloves normally have worn-out ridges of their fingerprints and a fingerprint scanner may even 'hang' when such fingerprints are being captured.

Software for duplicate analysis affects the rate and extent of identifying and eliminating multiple registration by some individuals.

Photographs taken on very bright days in an environment where a shade may not be available can cause poor images. where a face may not be clearly identifiable.  Heating of equipment may cause poor performance.

If persons deem that the collection of their biometric particulars may be for other purposes, they may decline to present themselves for registration which would reduce the percentage of persons eligible to register as voters.

An insecure environment will lead to lack of safety for the equipment and low turn-up at registration centres.

Biometric registration, especially in rural settings with poor or no electricity supply, calls for backup batteries or even solar panels to supply power for the registration equipment.

Details captured include the number of fingerprints.  If more fingerprints are captured, it improves the chances of weeding out people who would have registered more than once.  The more the details, the more the storage capacity that will be required for the data.  If only one fingerprint is captured per person, it would require less storage space than if ten fingerprints per voter are captured.

The time taken to register an individual has a bearing on how many will be registered.  It also has a bearing on how patient persons turning up to register will be and this in turn affects the overall number of voters that would be registered.

A long time frame allotted for the registration exercise normally leads to laxity in turn-up because people assume they have a lot of time.  Most people turn up when the deadline is approaching which increases pressure on the registration officials which can affect the quality of their work.

The skill, temperament and level of training given to officials handling the biometric registration exercise also has a bearing on the outcome.

South Sudan as a new country has not yet tried the BVR systems but still as experience got from other countries during elections e.g. Kenya, the factors that influence the performance of biometric voter registration systems can be

1. Experience of the staff in handling the equipment.  Are they well trained and can handle the system very well without messing which will affect the whole process.

2. In a country like South Sudan where most of the people are illiterate can influence the performance of the BVR Systems.

3. In a country whereby electricity is not there all the time or only covers certain parts of the country is also another factor.

Factores que influyen de forma contrarios al registro manual o registro parcial en medio magnético:

Costos de tecnología, como equipos de adquisición de huellas (imágenes y minucias o características), medios de almacenamiento y de procesamiento de imágenes y de datos, equipos para comprobación local o remota de datos y de imágenes (triada de huella, foto y firma), sistemas que garanticen la tolerabilidad de los equipos, como respaldo de potencia, servicios de comunicaciones, espacios adecuados para las labores de investigación, principalmente seguridades a todo nivel en los sistemas y los datos.

Factores que favorecen la tenencia de un registro biométrico:

Se hace una sola vez la adquisición de imágenes y de datos y para el futuro se hace actualización de la información en las bases de datos (Mantenimiento por adiciones, cambios o eliminación en el registro de votantes).

Se garantiza que no hay suplantación, de las personas que inicialmente se involucraron en el registro, siempre hay disponibilidad y facilidad de tener la información disponible.

Se pueden hacer controles locales, nacionales, en línea, en impresos o en un documento de identidad que puede ser verificable en cualquier parte del proceso de votación.

Hay ahorros en tiempo relativamente corto, pues las labores de adquisición, de impresión, de almacenamiento, se hacen una sola vez

Hay economía de los presupuestos posteriores pues la gran inversión se hace la primera vez

Considero que los factores que influyen en el rendimiento del Registro Biométrico de Votantes son:

- Características de la tecnología empleada

- Nivel de conocimiento del personal encargado de operar el sistema

- Capacitación a la ciudadanía sobre el funcionamiento del registro.

- Percepción de los sujetos políticos y generadores de opinión pública

- Transparencia y rendición de cuentas

Hard to answer because the word "performance" is vague. It could refer to any or all of the following:

1. Speed of a 1:1 match
2. Speed of a 1:N match
3. Accuracy of matching - False Match and False Non-match rates, etc.
4. Percentage of Failure to Enrol
5. Battery life (for mobile kits)
6. ...
   

and so on. If the person who asked the question could elaborate on what they mean by performance, more appropriate responses are likely to flow.

In many places, because of highly compressed timelines, the AFIS performance is hugely (and expensively) over-specified only for the procured systems to sit largely idle between registration or election cycles. Ernesto mentioned this in Part 3 of his excellent response. 10 to 1 is being kind, I suspect!

There is a very strong relationship between the following:

• The training of the staff who will operate the equipment used to enroll voters.
• The percentage of voters whose fingerprints are captured
• The quality of those captured fingerprints
• The resulting accuracy of the AFIS used to match those fingerprints.

Great question (if vague!), good and useful discussion. Looking forward to other responses.

Pour se faire une idée effective sur l'influence de la biométrie sur le vote, il faille avoir une vue générale sur les faiblesse de la biométrie. Chez nous au Congo, nous venons d'opter pour la Biométrie, je ne peux donc pas vous fournir un argumentaire fiable sur ce que cela représente quant à la fiabilité et la qualité des donnés collectées. Ce qui pour moi semble être préoccupant c'est la situation des frontières poreuses en Afrique et que l'on n'est pas certain de prendre rien que les fils du pays. Les gouvernants s'arrangent à enrouler les électeurs venant des pays tiers et le mieux serait de lier l'état civil de chaque pays et la biométrie. C'est vrai que la bio a pour objectif de réduire la fraude pendant le vote, mais rien ne garantie que celui qui se fait enrôler est un vrai citoyen surtout que les cartes d'identités et les passeports se donnent à coups de billets de banque.             

The question is a bit open ended - there are obviously technological and system related factors, and cultural and acceptance factors. A tremendous amount of work usually goes into determining the best technology for a given environment, identifying the most suitable equipment, training the staff .. even media campaigns telling the population how good it is for them. What we DON'T see often are the opinion polls ASKING the population whether they want to give up their biometric information. Success of electronic voter pocessing, whether it is biometric registration or eVoting, is largely based on trust of the electrorate for the EMB. Many people will resister electronically if told to do so. Some will only do so if the trust the system. And these are the ones we need to find ..

Proper campaigns describing the proposed system, followed by acceptance polls, prior to contract negotiation for acquisition, should help to develop  this trust.

En Venezuela se vienen utilizando sistemas de identificación biométricos desde el año 2004. El primer sistema biométrico,denominado captahuellas (2004-2010),fue instalado en un número reducido de estados (unidad subnacional) principalmente aquellos con mayor densidad poblacional. Dicho sistema se colocó en los centros de votación con mayor número de electores. Vale decir que el mismo estaba colocado fuera del recinto de votación, constituyendo una etapa previa al acto efectivo del sufragio y fue considerado por los partidos de oposición como un mecanismo de demora de la votación por los retrasos ocasionados en ocasión de su uso.

(Para mayor información técnica del sistema biométrico  recomendamos ver  Manual Funcionamiento para Miembros de Mesa Electoral y Secretarios. Elecciones Parlamentarias 2010).

http://www.cne.gob.ve/web/normativa_electoral/elecciones/2010/parlamentarias/documentos/MANUAL_OPERATIVO_PARLAMENTARIAS_parte1.pdf

http://www.cne.gob.ve/web/normativa_electoral/elecciones/2010/parlamentarias/documentos/MANUAL_OPERATIVO_PARLAMENTARIAS_parte2.pdf

En el año 2012 se modifica la identificación biométrica electoral, implementando a nivel nacional el uso de un sistema integrado biométrico y de votación llamado "Sistema de Autenticación Integral SAI". El mismo se incorporaba en el acto de votación, en la medida que a través de él se permitía el ejercicio efectivo del sufragio. Tal modificación generó dudas en los partidos opositores en la medida que se temia que al estar integrados los sistemas de identificación biométrica-sistema de votación se podía lesionar el secreto del voto. (Entre otras razones por la secuencialidad de la identificación del elector el dia de las elecciones y el voto como tal)

(Nota. La principal base de datos biométrica en el país en el 2012 estaba en mano del Poder ejecutivo y la misma tenia serias deficiencias, entre otros motivos por el alto componente manual del levantamiento. En los procesos electorales presidencial 2012, regionales 2012, presidencial 2013 y municipales 2013 el Consejo Nacional Electoral ha logrado importantes avances al respecto)

(Para mayor información recomendamos ver  Manual Funcionamiento para Miembros de Mesa Electoral y Secretarios. Elecciones Municipales 2013).

http://www.cne.gob.ve/web/normativa_electoral/elecciones/2013/municipales/documentos/Manual_Miembro_Mesa.pdf

Vistas las experiencias considero que constituyen factores fundamentales relacionados con la implementación de sistemas de identificación biométricos en materia electoral los siguientes:

1.- Marco normativo, en particular los alcances del sistema biométrico en el derecho al voto.

2.- Tipo de tecnología utilizada. Integración o no al acto de votación

3.- Existencia o no de bases de datos biométricas y acceso real por parte del órgano electoral.  (También es importante es la calidad (rigurosidad y exhaustividad) de esa base de datos)

4.- Recursos económicos y financieros para la implementación de sistemas de identificación biométricos electorales

5.- Capacidad tecnológica para recolectar información biométrica

6.- Credibilidad de la administración electoral.

7.- Transparencia y auditabilidad sel sistema de información biométrico.

8.- Estrategia de aplicación del sistema de información biométrica (progresivo vs. "de golpe")

9.- Nivel de conflictividad política y social.

A mon avis, le succès de l’enregistrement biométrique dépend d’une part du niveau d’instruction des populations et de leurs capacités à cerner les directives qu’elles reçoivent. Mais, elle dépend aussi de la qualité de la sensibilisation et de la précision des informations diffusées par les structures en charge.

Pour moi, il est très important que les personnes en charge de l’enregistrement des données biométriques soient qualifiées et maitrisent les applications qu’il faut  et le bon usage des matériels techniques.  Car, mon expérience m’a révélé que pour ces genres d’opérations, les recrutements ne prennent pas trop en compte la compétence des agents d’enregistrement.  Aussi, le fait que ces agents ne travaillent pas dans de bonnes conditions pourrait être un obstacle à la bonne marche de cette opération. Par exemple, quand ils ne perçoivent pas leurs rémunérations à temps.

I am agree with my dear colleague Ronan,

the question is very vague!

Even if you have the best AFIS or ABIS in the world, if you have during the voter registration phase, a very high 'failure to acquire’ rate at field level  in the system’s attempt to capture the digital fingerprints and face of voters, you will miss maybe 10% of voters, and your de-duplication with be significantly less meaningful,
because of technological registration issues!

The problems can be induced because of  the insufficient training of the operators (it's the bigger  problem on the capture of face photo), because of the profession of the people to register (miners or farmers have very damaged fingerprints), and so on...

The only ways to palliate this problem, are to increase the quality of the training of the operators, to use multi-modal biometrics (fingerprints + face, or better fingerprints+iris), and to reduce the cause of capture problems (i.e. by using two fingers instead of ten)

i will agree with the other collegue who responded refering to South Sudan and other devoloping countries.

The ganeral basic infra structure plat a big role in the success of the BVR eytc power , good place to operate these sensitive machines . cause in most of the African countries people are being registered under the trees or in places like school and so on.

the second thing is the level of traing for the registration officers.

Thirdly in countries which has gone a long civil wars where the numbers of ambutees quite large i think it will be  difficult to take their finger prints.

Thanks

Dr Woda Jeremaih

Bonjour,je n'ai aucune expérience dans ce domaine la biométrie , mais j'ai compris une chose, par rapport aux autres expériences que j'ai au niveau organisation et observation des élections, la performance est d'abord dans le cadre des des textes ,des lois, des personnes qualifiés à les appliquer,les matériels sophistiqués, un environnement très sécurisé et un plan qui reflète à une réalité du terrain . Avec tout ça il faut se fixer des objectifs,des résultats à atteindre et bien sur ce donner les moyens  appropriées.

Hello,

Although the term "performance" can sound a little vague, Indeed many factors can influence the overall performance of a BVR. From what I've learned from the experience we had in Cameroon the factors I think of are:

the stakeholders have to agree to move on to biometrics and prevision should be made for its implementation in the electoral code

citizen or the population have to understand the need of using biometrics and accept it

. This is what makes the difference. Using biometrics for a company is not the same as setting such a system for elections, though the equipements are almost always the same. An emphasis has to be made on the characteristics of the equipements, and on the de-duplication system (AFIS system).

. There are countries where biometrics were implemented but the cards and the rolls were printed in another country, kits were borrowed from other countries...etc. This can affect the performance of the system, and the mastery of the processes by the EMBs. The best thing is to have of the printouts done in the country, as we did in Cameroon and to have a good mastery of all the processes.

During the voter registration and general election, most EMBs rely on temporary staff who are trained for a shorter period because of cost implications. The issue of qualified staff is therefore tricky considering that some lack computer skills and are not give ample time to internalize how the BVR machines work. Consequently, sufficient funding and training duration is key for effective use of the BVRs.

I also find the word 'performance' a bit vague because there are various kinds of performance expected in a BVR e.g. operational performance (e.g.the work of the field staff), the technical performance (e.g. specs, technical aspects like speed.accuracy, quality etc) and strategic performance in terms how has its employment enhanced the overall goals and objectives of the EMB or its key stakeholders. 

I find the explanation provided by Ernesto(in Four parts)  sufficiently detailed and doing justice to the question especially on technical, tactical and operational dimensions.  I also agree with  Ahmed and others  especially on the need for piloting of the technology before roll-out.

In Kenya the greatest lesson we learned from the implementation of BVRin 2012/2013  is never ever deploy technology if you have no enough time to test and pilot it well before deployment for actual registration. And train the end-user thoroughly and in time.

Social,Cultural and political realities/context have significant influence that may provide constraints and opportunities that can derail or aid the 'performance' of the BVR. For example some rural communities in Kenya kept off from the BVR because of rumors that it can cause cancer (both men and women) , barrenness(women) or erectile dysfunction among men. So even if the BVR was perfect in every sense, its performance will be zero if people stayed away from it because of socio-cultural or political consideration. 

Thank you.

If "performance" is meant to produce accuracy and currency of voter registers, then even biometric solutions need to tie into the register of births and deaths (état civil), so that voters who pass away are effectively cleansed off the register. Biometrics can help eliminate double entries, but it cannot in and of itself ensure overall accuracy and currency of the register.

Factors that influence the performance of Biometric Voter Registration (BVR) systems globally.

Some of the important factors not explicitly mentioned above are

·         Corruption and vendor interference: These include Election officials, suppliers, and those who make decisions on behalf of government. Suppliers influence the selection process, officials receive bribes leading to purchase or supply of less efficient or unsuitable systems.

·         Political factors that tend to be ignored for example “Does the government and other people in authority want the system to be implemented?” . This is purely political but may lead to system failure (Government Officials may delay release of funds, fight the EMB or even de-campaign the registration system)

·         Failure to integrate systems within the EMB and with government departments who attempt to solve the same or similar problems. EMBs. passport offices, driving permit/licensing office, policing and national security, banks etc. should integrate their systems and or technologies. This would allow the building and sharing of skills and equipment, reduce costs and improve coordination and efficiency.

·         Failure to coordinate and collaborate regionally that would allow sharing of costs, skills and lessons learned

Other factors mentioned by colleagues but which I consider highly significant are:

·         Failure to pilot the technology and procedures.

·         Poor Preparations and time factors. Systems are always rushed, without sufficient study, timely delivery of equipment and materials, proper training of staff and sensitization of the public

·         Inadequate legal frame works and failure to enforce sanctions for defaulters

·         Environmental factors that include the infrastructure, knowledge and skill levels, system acceptability, organizational politics and the climate

·         Technical Factors: Scalability, Robustness, Ease of Integration, Reliability, Computational Power and Accuracy, Productivity, Speed and Efficiency and Easy of Use of the system.

·         The initial cost, maintenance and sustainability