Columns
|
Data
Quality and the Irish Abortion Referendum
On the 6th of March this year, the
Irish people went to the polls to determine the legislative response to
a thorny moral, religious and philosophical issue Ð namely the protection
of and balancing of the right to life of a mother and an unborn child.
This article is not about that issue. The people have spoken in regard
to that amendment to the Irish Constitution and I am not about to retread
old arguments.
This article is about the fundamental issues of data quality which aspects
of the coverage and analysis of the referendum decision throw up. These
are issues which largely have been dropped to the way-side in the aftermath
of the referendum. However, given the low electoral turn out many of these
issues are of extreme concern to me as a Data Quality/Data Management
professional.
As voter turn out decreases (irrespective of the cause), the remaining
active electorate will be increasingly polarised on either end of the
spectrum of Yes/No decisions. This is a factor of both the smaller statistical
population and also of the fundamental fact that as voter activity decreases
those that Òmake the effortÓ increasingly represent a growing proportion
of extreme opinion either way. As such, society (much like the free market)
will achieve an ideological equilibrium whenever faced with a simple Yes/No
decision on a question such as that posed on March 6th. This is borne
out by the narrow margins in both the Divorce referendum (narrowly carried)
and the recent referendum on abortion (narrowly defeated).
Voter turn out is, as a general trend, in decline in Ireland. Therefore
small errors in the way in which the ballot and the voter data gathering/validation
exercises that precede any vote by the electorate will have an increasingly
significant impact on the outcome of any election or referendum vote.
These errors will, should the current trend in voter turnout continue,
have an increasingly powerful affect on the outcome of any poll. Given
the recent tight margins between the Yes and No votes in Òbinary choiceÓ
referenda this could have the affect of misrepresenting the will of the
people and turning a pro-life vote into a pro-choice vote or electing
one candidate over another in a general election.
These issues need to be addressed with the seriousness that they deserve
given that they have the potential to overturn the democratic will of
the people. What are these issues?
1. On the day of the poll, a caller to a national radio station complained
that they had been unable to vote. When they turned up with their polling
card (which has a numeric identifier associated with it which I assume
should be unique within any constituency) they were told that they had
already voted. As a result they did not have a ballot paper issued to
them and were unable to vote.
2. In the days preceding the poll, many national newspapers carried light
relief stories of family pets having valid polling issued to them Ð giving
them, technically, the right to vote and express a preference on fundamental
matters of national importance.
One key point to make at this stage is that I have no means of quantifying
the incidences of either of these scenarios. Therefore I am not in a position
to put forward any empirical data as to the impact of these two scenarios
on the overall referendum outcome. In any event, the poll has been concluded
now and, irrespective of any Supreme Court challenges that might be in
the offing, the outcome can only be overturned by a new vote.
If we examine these scenarios from a pure Total Data Quality Management
perspective it is apparent that there are significant flaws in the way
in which the data, which control the way in which the will of the people
is manifest, are gathered and managed. I would hypothesise that this is
a result of the historic culture and attitudes that pervade in many organisations
(particularly Civil Service strongholds) about data management and the
auditing and control of data quality. Furthermore, I suggest that unless
the voters recognise their role in validating and auditing this data,
the situation will continue to be difficult to manage.
The Irish electoral system contains a number of, for want of a better
expression, business rules that govern the conduct of elections and referenda
in the Irish Republic. The fundamental rules (ignoring issues such as
proportional representation for a while) are:
á All voters must be registered to vote in an electoral region
á Only those voters registered to vote in an electoral region may vote
in that region
á All those eligible to vote in an electoral region will receive evidence
of this (a polling card with their name, address and a numeric identifier).
á Voters register to vote in an electoral area by applying to be included
on the register of electors (the voter Òmaster-fileÓ in IT terms).
á To be registered to vote, voters must be over 18 prior to any application
to be included on the register of electors.
á To be registered to vote, voters must be citizens of Ireland or citizens
of another EU country normally resident in Ireland.
If we apply these rules to the scenarios identified above, it is apparent
that the system is not structured to fully ensure purity of data.
Scenario 1
In this scenario the voter (letÕs call them W) registered to vote by submitting
their name and details to the draft register of electors. W is issued
a polling card, and on the day of the referendum, W goes to the local
polling station to cast his/her vote.
To WÕs surprise, they are told that they have already voted.
Root Cause:
Either duplicate polling cards were issued, giving an unknown person the
ability to impersonate W and vote or (more likely) the polling clerks
simply crossed off the wrong name on their copy of the Electoral Register.
Problem: What is the problem here? W has the right to receive a ballot
paper once they can prove who they are. Therefore, they should be able
to cast their vote, eliminating the problem. This is the legalistic answer
to the problem.
However, from a Data Quality perspective, if W now casts a vote, the data
is corrupted. In the case of a simple Yes/No decision WÕs vote will either
á cancel out the ballot which W allegedly cast earlier
á compound the ballot which W allegedly cast earlier.
The problem with this solution is simply that it overlooks the underlying
data quality issue and potentially ignores the possibility of electoral
fraud. While from a legal perspective it is a tidy solution, from a data
quality perspective it merely compounds one error with another. Two wrongs
should not be interpreted as a right. Furthermore, this solution assumes
that W is sufficiently aware of their rights to demand a ballot paper
and has sufficient identification to satisfy the authorities that they
are in fact who they say they are.
In the case reported to the radio programme, the voter was not aware of
their rights in this regard and left the polling station without casting
a vote. They were effectively disenfranchised by poor attention to data
quality on at least two grounds:
1. If the root cause was a duplicate polling card ID number, then the
process by which the voter Ids are assigned to polling cards was flawed
as it allowed a duplicate number to be generated.
2. If the root cause was polling clerks mis-marking the electoral roll,
then the disenfranchisement is as a result of poor enforcement of manual
checks to ensure that the correct record was marked off.
3. The polling clerks and/or the local returning officer should have been
aware that the voter had the right to a ballot paper once they were able
to identify themselves. The authorities should then have been in able
to explain these rights to the voter.
In effect, the voter failed to exercise their right due to a breakdown
in the data management process governing the poll and a failure of the
stewards of that process to know or communicate important rules governing
the process. However, given that the data quality interpretation of legal
rule is effectively Òtwo wrongs make a rightÓ, neither option is acceptable
in the current environment.
Scenario 2
In this scenario, a family pet receives a valid polling card evidencing
their eligibility to vote (this is a favourite of the tabloid and some
broad-sheet press prior to any poll). Therefore, the pet (letÕs call it
ÒRoverÓ) has apparently fulfilled the key criteria for voting rights in
the Irish Republic:
á Rover must be over 18
á Rover must be living in an electoral constituency
á Rover must be registered on the register of electors.
The electoral data systems are unaware of one key fact about Rover which
has a direct bearing on RoverÕs status as a voter, namely that Rover is
a dog.
Root Cause:
The root cause of this scenario is simple. RoverÕs details found their
way into the electoral register via the normal data gathering processes.
In addition, RoverÕs details met the essential requirements for a ÒvoterÓ
in that they had a name, an address, and a date of birth, all of which
indicated that Rover was
á Resident in an electoral constituency
á Over 18 at the date of registration
This data could have entered the system by either of two channels.
1) Rover was included on the Census of Population for that household.
(The electoral authorities in preparing the ÒDraft Register of ElectorsÓ
use census data. If Rover were included in the Census as a family member
then Rover would have been entered into the base data used for the Register
of Electors. ) The source of Census data is the public who provide this
data to the government for a number of purposes, of which the Register
of Electors is only one (others include planning of government spending
on health and education services).
2) Rover was included in an addition to the Draft Register of Electors
(Before any referendum or election, the public is again asked to update
the data on the Register of Electors to take account of children leaving
home, changes address etc. Filling out a simple form with any removals
from or additions to the Register does this. ) Again the source of this
data is the public.
As the Register of Electors is the ÒVoter Master-fileÓ which is used to
issue polling cards, it is imperative that this data be correct. To this
end, the electorate are urged before each election or referendum to check
this register to ensure that they are included and that it is correct.
Therefore, the existence of Rover as a voter could be caught at that stage
and be corrected. However in the case in hand it wasnÕt spotted until
Rover received a polling card.
Problem:
There are a number of distinct problems because of this scenario. Once
Rover was issued a valid polling card, Rover technically could vote as
on the basis of the data provided, Rover was an eligible voter living
at an address in an electoral constituency (letÕs say Ò122 Labrador Place,
Dublin 22Ó)
Therefore, any person who could adequately identify himself as ÒMr. R.
DogÓ of Ò122 Labrador Place, Dublin 22Ó could exercise that vote without
question or hindrance. At a push, ÒRoger DogÓ could claim a typo in the
electoral register if challenged (ÒGÓ is directly above ÒVÓ on a computer
keyboard). Among the valid forms of identification accepted under the
Irish electoral system are credit cards. If you apply to any of the credit
card companies for a card, you can request whatever name you want on the
card, with very few restrictions. Therefore, it would be a simple matter
to use the existence of Rover to cast a fraudulent vote and distort the
outcome of an election or referendum. Should a special interest group
or political party exploit this weakness in the system, it would really
be a case of the dogs in the street knowing the outcome before the rest
of us.
Furthermore, even if Rover does not exercise the vote available, the fact
that a dog is counted amongst the valid electorate distorts the count
of potential voters. This has an impact on the reporting of voter turnout
against potential voters. This in turn has a direct effect on government
spending on voter awareness campaigns. It also affects the way in which
political parties shape policy given that they are always trying to attract
more voters in those areas where turnout is low.
Finally, this problem points to a distinct flippancy, lack of concern,
or lack of understanding towards Census and Electoral Register data on
the part of the public. The fact that a family pet is entered on a Census
form as a person speaks volumes for the ability of the common person to
follow the instructions set out in the Census forms. People apparently
do not care if Census data is correct or the forms and/or process are
too complicated for people to understand. This appears to carry through
to the electoral register checks that we are all supposed to carry out.
None of these problems is helped by the fact that the media would rather
treat this as a Òdumb governmentÓ story and write it up as a light-relief
piece. Rather it should be treated as a Òlazy and flippant electorateÓ
story and write it up as a serious question as to who is ultimately responsible
for making sure that family pets arenÕt counted among the electorate.
It is a stark reality that governments are unlikely to provide significant
or sufficient investment in performing a data cleanup process given the
other demands on the public purse. However, unless the citizens wake up
to their responsibility for poor data quality in the electoral system
then the data cleanup overhead will continue to be there and will continue
to grow.
Conclusion
Voter activity in Ireland is in decline. Concern amongst the electorate
for the issues of the day is steadily decreasing. We are now at the point
where referenda on fundamental issues of moral direction or nationhood
are decided by margins of 1% or less. It is in this sort of environment
that the quality of the source data used to identify those who can and
cannot vote is of prime importance. It is in this sort of environment
that every effort is essential to ensure that every vote counts and that
every eligible voter should be able to vote.
In the polling stations, immediate action is required to ensure that a
slip of a pencil or the misplacement of a ruler does not disenfranchise
voters. The inherent data quality flaw in the current legal regulation
is that it effectively supports the notion that in matters of morality
or nationhood, two wrongs DO make a right. This needs action, as it is
a disturbing principle of law in such an important process. In relation
to the generation of polling cards, steps are urgently needed to ensure
that a polling card registration number is globally unique within the
electoral population. ÒOne card Ð one voteÓ should be the guiding principle.
This should help prevent disenfranchisement by Òdouble dippingÓ where
duplicate polling card numbers mean someone loses his or her vote. Finally,
in relation to the data gathering exercise that builds the electoral register,
this needs immediate attention as it is blatantly open to abuse and corruption.
1) Personal data gathered from the census data need to be matched against
standard libraries of names and name constructions. _ Distinguish ÒDaragh
O BrienÓ as a name from ÒRover T. DogÓ as a question mark name.
2) Any question mark names gathered from this process should be addressed
back to the head of household for validation _ ÒDear Mr. O Brien, is Rover
T. Dog a person who meets the criteria to be a voterÉ.Ó This response
should be treated as a statutory declaration and should have penalties
associated with it for making a false or misleading declaration
3) Likewise, any data gathered from the forms to update the electoral
register needs to be matched against the same standard libraries in the
same way.
4) The standards of identification acceptable at a polling station must
be raised. Credit cards were fine when very few had them. Now their use
is so common and it has become so easy to get a card (without any form
of rigorous checking of data) that the quality of this as a form of identification
is now extremely poor. As such they should be removed from the list of
valid forms of identification. Ideally, supporting identification should
be of a form that requires the use of an RSI number and/or birth certificate
to obtain (driverÕs licence/passport) and should have a photograph of
the person.
5) More efforts should be made to engage the electorate in the process,
not just in the vote itself but in the full gamut of what it means to
be a voter. This MUST include education on the responsibilities that come
with being a citizen in a democratic republic. It MUST also include education
on the penalties for including false or misleading information on Census
forms or on voter registration forms. It MUST also include ACTUAL ENFORCEMENT
of these penalties. The owner of Rover should receive the appropriate
penalty if they register him as a voter and then fail to take the necessary
steps to deregister him. INCREASING VOTER TURNOUT should also be a key
component of any such programme as it will reduce the percentage impact
of any data issues on the overall result. Democracy brings with it many
rights, amongst them the right to vote and express a choice in the make
up of the government and fundamental laws of the nation. However, these
rights carry with them responsibilities.
Citizens in a democracy should recognise their role in ensuring that the
underlying data which drives the electoral process is as accurate as possible.
Until pets are recognised as citizens, Rover should not have the Vote.
by
Daragh O'Brien
22nd April 2002
|
Topics
|