Discussion
of Instant Runoff Voting Systems
& Fraud Resistant Voting
by Michael
Andrews
There are at least two remedies necessary to reinstate a semblance of democracy in the USA. The first is finance reform, that is the absolute removal of wealth in any form from the electoral process. This includes the bankable currency of fame, notoriety and celebrity. Along with the separation of church and state a democratic voter residing any place on the planet should have the right to a well regulated economy, the separation of state and markets, and the complete disassociation of wealth from determining election results.
True democracy will only
when the poorest and least known citizen has an equal chance of winning
electoral office as the wealthiest and most famous candidate.
This paper is focused solely
on the second remedy which is the establishment of more democratic voting
systems, and includes the issue of fraud resistant electronic voting.
The following discussion is
broken into the following six sections:
Standard
IRV
Distributed
IRV
Implementing
IRV On Paper Ballots
Implementing
Electronic IRV
Minimizing
Fraud Potential In Electronic Voting
Comparing
Standard IRV With Distributed IRV
Standard
IRV (Instant Runoff Voting)
The only way to democratize
a voting system in which more than two candidates compete for a single off ice
is to utilize runoff voting.
Runoff voting allows the
voter to cast his first vote for his first choice. But, in the event his
candidate finishes last, the voter does not waste his vote. A runoff vote is
held for the remaining candidates and so, the voter then gets to choose between
them. This cycle of runoff votes is repeated until some candidate gets more
than 50% of the total votes.
This is democratically fair,
but expensive and inconvenient to hold separate multiple voting events on
separate voting days. Also, each cycle increases the opportunity for error and
fraud.
Instant runoff voting
removes the expense and inconvenience by capturing every voter’s second, third,
or more choices in one, single voting event, and then computing the winner the
same way as if multiple runoff elections took place.
This is done by ranking the
candidates.
The first rank of votes are
totaled. If any candidate gets more than 50% he is the winner and process
stops.
If not, then the candidate
with the least votes is eliminated and the ranks reflecting votes that were
initially allocated to him are then reallocated to the rank of the following
choice. The first ranks are then totaled again. This process is repeated until
some candidate gets more than 50% of the total vote.
As an example we will use
the 2000 presidential candidates; Gore, Bush, Nader and Buchanan.
In and IRV election the
voter simply ranks the candidates in the order of choice; for example a voter
could rank:
1.
Nader
2.
Gore
3.
Bush
4.
Buchanan
This means that if the
voter’s first choice, Nader, is low man, then Gore gets his vote on the runoff.
If Gore is the new low man, then Bush gets his vote on the second runoff, which
would be the last since in this case only two candidates remain.
For the sake of simplicity,
using a sample of 10 voters the 10 IRV ballots could be as follows:
1.
Nader
2.
Gore
3.
Bush
4.
Buchanan
1.
Nader
2.
Gore
3.
Buchanan
4.
Bush
1.
Gore
2.
Nader
3.
Bush
4.
Buchanan
1.
Gore
2.
Nader
3.
Bush
4.
Buchanan
1.
Gore
2.
Bush
3.
Nader
4.
Buchanan
1.
Gore
2.
Buchanan
3.
Bush
4.
Nader
1.
Bush
2.
Buchanan
3.
Gore
4.
Nader
1.
Bush
2.
Gore
3.
Buchanan
4.
Nader
1.
Bush
2.
Buchanan
3.
Nader
4.
Gore
1.
Buchanan
2.
Bush
3.
Gore
4.
Nader
The total of the first ranks
then becomes: Nader, 2 votes; Gore, 4 votes; Bush, 3 votes and Buchanan, 1
vote.
No candidate has more than
50% so the low candidate, Buchanan is eliminated and his vote is redistributed
by that ballot to Bush. After eliminating Buchanan the ranks now look as
follows:
1.
Nader
2.
Gore
3.
Bush
1.
Nader
2.
Gore
3.
Bush
1.
Gore
2.
Nader
3.
Bush
1.
Gore
2.
Nader
3.
Bush
1.
Gore
2.
Bush
3.
Nader
1.
Gore
2.
Bush
3.
Nader
1.
Bush
2.
Gore
3.
Nader
1.
Bush
2.
Gore
3.
Nader
1.
Bush
2.
Nader
3.
Gore
1.
Bush
2.
Gore
3.
Nader
Now the total of the first
ranks then becomes: Nader, 2 votes; Gore, 4 votes; Bush, (3+1) or 4 votes.
Again, no candidate has more
than 50% so the low candidate, Nader is eliminated and his vote is
redistributed by those two ballots as 2
votes to Gore. Now the ranks look like a typical 2 candidate eletion:
1.
Gore
2.
Bush
1.
Gore
2.
Bush
1.
Gore
2.
Bush
1.
Gore
2.
Bush
1.
Gore
2.
Bush
1.
Gore
2.
Bush
1.
Bush
2.
Gore
1.
Bush
2.
Gore
1.
Bush
2.
Gore
1.
Bush
2.
Gore
Now the total of the first
ranks then becomes: Gore, (4+2) or 6 votes; and Bush, 4 votes.
Since Gore has more than 50%
he is declared the winner.
This process is iterative
and for the sake of practicality in any election of even a modest number of
candidates and voters it would be typically determined by a computer program.
It is also worth noting that
the totaling of the ballots must happen in a single computer which has the data
for all ballots.
Distributed
Instant Runoff Voting
The main distinction between
standard IRV and distributed IRV is that standard IRV requires an iterative
process to complete it calculation, where a distributed IRV achieves the same
results by totaling only once. This benefit means that standard IRV is reliant
on a single computer using black box software and the distributed IRV spreads
the computation among many computers all of which can cross validate one
another using open, simple, cheap and inspectable software. All of this, in
turn, makes the distributed process more complicated and more difficult to
understand.
The concept that the
arithmetically challenged voter will not be able to fathom is that any one
singe vote is broken into values. The sum of these values represent one, entire
vote. These values are distributed among the candidates as ranked by the voter.
Distributed IRV is also
based on simple ranking. There are two methods of distributed IRV described
here. If there are many candidates in an election a voter may very well decide
to not rank the candidates in which he has no interest. In fact, a voter need
only vote for one candidate. The two forms of distributed IRV reflect two ways
to handle the remaining candidates that were not ranked. The first method, the
Even Distribution Method Of Distributed IRV, evenly distributes the vote values
among the remaining candidates. The second method, the Zero Distribution Method
Of Distributed IRV, allocates zero values among the remaining candidates and
distributes all the values to only those candidates that have been ranked by
the voter
Even Distribution Method Of Distributed IRV
Take for example, this
year’s presidential elections; John, W and Ralph. For the sake of simplicity
the example is limited to three candidates.
The voter simply ranks them
in order. A single vote is then broken down and distributed among the
candidates based on the total number of candidates. In this case, with 3
candidates the single vote is broken into the distribution of 6 values. The top
pick gets a value of 3, the second 2 and the last 1. In this case there is, by
combination, 6 possible rankings. Each possible ranking of the six distributes
a total of 6 values based on the voter’s choice of rank as follows:
3
Ralph
2
John
1 W
or
3
Ralph
2 W
1
John
or
3
John
2
Ralph
1 W
or
3
John
2 W
1
Ralph
or
3 W
2
John
1
Ralph
or
3 W
2
Ralph
1
John
When voting is halted all
the values are totaled. The candidate with the highest values wins. Even if all
those who vote their conscience for Ralph lose, their next pick John would
still beat W. This entirely eliminates the standard IRV method of employing an
iterative process of redistributing secondary choices and re-totaling until a
candidate with more than 50% of the total votes emerges. It automatically
accomplishes the same goals in one totaling calculation.
Any of the IRV methods will
work for any number of candidates, including a couple of hundred in a
California Gubernatorial recall vote.
For the sake of clarity
compare a vote of 2, 3, 4 and 5 candidates.
First, the number of
possible combinations of rankings:
2
candidates has 2 combinations.
3
candidates has 6 combinations.
4
candidates has 24 combinations.
5
candidates has 120 combinations.
It is obvious that the
combinations quickly escalate. But it is not significant because the method of
choosing is easily handled and hugely simplified since the voter simply ranks
the candidates according to preference. In the first example above
3
Ralph
2
John
1 W
the voter is basically
saying he wants Ralph to win, but in the event Ralph cannot, then he wants his
vote to go to John. This is exactly what would happen in a standard Instant
Runoff Vote.
The breakdown of a single
vote into distributable values is based on the number of candidates, which in
turn determines the number of ranks. The more candidates there are the more
values are needed to represent the hierarchy of ranking. But no matter what,
the total values always represent one single vote.
So,
with 2 candidates you have:
1st
rank gets 2 values
2nd rank gets 1 value
for
a total of 3 distributable values.
3
candidates:
1st
rank gets 3 values
2nd rank gets 2 values
3rd
rank gets 1 value
6
total distributable values.
4
candidates:
1st
rank gets 4 values
2nd rank gets 3 values
3rd
rank gets 2 values
4th
rank gets 1 value
10
total distributable values.
5
candidates:
1st
rank gets 5 values
2nd rank gets 4 values
3rd
rank gets 3 values
4th
rank gets 2 values
5th
rank gets 1 value
15
total distributable values.
The next issue is one of
clarity. In the event there are many candidates, say 10 or 20 in a Democratic
primary, or worse, hundreds in a California Gubernatorial recall, then no voter
need be concerned with trivial choices. For example the voter only need vote
for 1 out of hundred, or rank the first 3 candidates and can safely ignore the
rest.
The method of evenly
distributing the remaining values to the remaining candidates is easily
accomplished by the computer using a simple algorithm:
(Total Distributable Values
- Total Values Assigned) / (Total Number of Candidates - Total of Ranks
Assigned)
For example in a 2 candidate
election:
there
are 2 candidates, and 3 distributable values
If
the voter only assigns his first pick, or stops after the 1st rank
then
1st
rank receives 2 values assigned, then (2-1) / (3-2) = 1 which is the remaining
value distributed to each remaining candidate.
This is exactly the same as
if the voter assigned both ranks, 2 values for the 1st rank and 1
for the last rank. And this is exactly the same as if the voter voted for 1
candidate in a 2 candidate race in a conventional election.
For
3 candidates to be assigned in 3 ranks & 6 distributable values to be
assigned:
0
ranks assigned = (6-0) / (3-0) = 2 values are evenly distributed among all 3
candidates
1st rank only assigned = (6-3) / (3-1) = 1.5
values are evenly distributed among remaining 2 candidates
2nd rank only assigned = (6-5) / (3-2) = 1
values are evenly distributed among remaining 1 candidate
3rd rank assigned = ranking completed, no values
left to distribute
For
4 candidates to be assigned in 4 ranks & 10 distributable values to be
assigned:
0
ranks assigned = (10-0) / (4-0) = 2.5 values are evenly distributed among all 4
candidates
1st rank only assigned = (10-3) / (4-1) = 2
values are evenly distributed among remaining 3 candidates
2nd rank only assigned = (10-3) / (4-1) = 1.5
values are evenly distributed among remaining 2 candidates
3rd rank only assigned = (10-5) / (4-2) = 1
values are evenly distributed among remaining 1 candidate
4th rank assigned = ranking completed, no values
left to distribute
For
5 candidates to be assigned in 5 ranks & 15 distributable values to be
assigned:
0
ranks assigned = (15-0) / (3-0) = 3 values are evenly distributed among all 5
candidates
1st rank only assigned = (15-3) / (5-1) = 2.5
values are evenly distributed among remaining 4 candidates
2nd rank only assigned = (15-3) / (53-1) = 2
values are evenly distributed among remaining 3 candidates
3rd rank only assigned = (15-3) / (5-1) = 1.5
values are evenly distributed among remaining 2 candidates
4th rank only assigned = (15-5) / (5-2) = 1
values are evenly distributed among remaining 1 candidate
5th
rank assigned = ranking completed, no values left to distribute
Zero Distribution Method Of Distributed IRV
The second method is to
distribute all the vote values among only those candidates that the voter has
ranked. If the voter chooses only one candidate out of 20 then all the values of his vote are assigned
to a single candidate and all remaining candidates are given zero. It he ranks
only 2 or 3 candidates, then the values are assigned in a descending scale to
just those 2 or 3, and the remaining candidates are assigned zero.
This is more democratic than
the first, simpler method in that a voter who wishes either Ralph or John to
win and to give no support to W, would rank Ralph with 2 values, John with 1 value and W would get 0.
In a 2 candidate race this
looks exactly like conventional voting. The candidate chosen gets the total vote
of 1 represented by a total of vote values of 1, and the rejected candidate
gets 0.
The algorithm needed to
effect this is more complicated than the first, simpler system. The numbers
that the algorithm needs to calculate are functions of the total number of
candidates/ranks and on the last rank assigned by the voter. These numbers are the
lowest beginning number, the increment and the first rank. This allows the
total values to be distributed only to the candidates ranked. This requires
three separate algorithms, for which the pseudo code would look like:
Lowest
Beginning Number
LN
= (TR - LR) + INT( ABS(TR-LR-1) / INT(TR / 2))
Increment
I =
1 + INT( ABS(TR-LR) / INT(TR/2))
and
First Rank
FR
= V - TVU
where
TR
is the total number of ranks, or candidates
LR
is the Last Rank Assigned
V
is the total vote Values to be distributed for a single vote
LN
is the Lowest beginning Number
I
is the Increment ascending from rank to rank beginning with the LN
FR
is the First Rank - a special case
TVU
is the total number of values used beginning with the LN and ascending up
through the second rank
In a 6 candidate race the
total values for a single vote is 15 and is distributed as some combination of
integers from 0 to 15. The simple case is when all candidates are ranked, then
the distribution of 15 becomes 0, 1, 2, 3, 4 and 5. The other simple case is
when the voter ranks only one candidate, in which case that candidate receives
all 15 values.
A: 5 values
C: 4 values
E: 3 values
B: 2
D: 1
F: 0
The in between options are
more complex. For example, in a vote for 6 candidates A, B, C, D, E and F,
where the voter only wants A, C or E to win and is against B, D and F, if the
voter ranks all 6 candidates then the value distribution would be:
A: 5 values
C: 4 values
E: 3 values
B: 2
D: 1
F: 0
If the voter ranks only his
first choice candidate then the value distribution would be:
A: 15 values
C: 0 values
E: 0 values
B: 0
D: 0
F: 0
If the voter ranks only his
first three choices and ignores the candidates he is against then the value
distribution would be:
A: 7 values
C: 5 values
E: 3 values
B: 0
D: 0
F: 0
These are the two distinct
methods for accounting for the remaining values and remaining candidates. The
first, and least complicated method is that the many remaining distributable
vote values are evenly distributed to the remaining candidates. The second, and
more democratic method, is that all the values are evenly distributed to only
those candidates that the voter chooses to rank and the remaining candidates
get 0.
Either of these two
distributed IRV methods is scalable in that the same system of ranking,
assigning vote values and distribution to unassigned candidates remains the
same for 1 candidate or for a thousand.
In most races there are
probably an average of 5 or 6 candidates for a voter to consider. But no matter
how few or how many candidates are in a race, since the total distributed votes
is always the same for each single vote, then the grand total of all values
divided by the total distributable values for a single vote will equal the
total number of voters.
Both methods accomplish two
things; they reflect the intention of the voter who is saying that the
remaining candidates are of no concern to him, and it keeps the total number of
the distributable values always the same for every single vote. That is an important
benefit because after the polls close and all the values are totaled, then the
overall total of points divided by the distributable values for one vote will
equal the total number of voters and, by definition, the total number of single
votes. This is a crucial parity check for fraud prevention.
In passing, it should be
noted, that in the event a voter ranks no candidate at all, then either the
vote should not be finalized or counted, or optionally, the vote may be counted
anyway by assigning every candidate an even fraction of the total values which
would be equivalent to a null vote and would still maintain the integrity of
matching total values with total voters.
Implementing
IRV On Paper Ballots
Both Standard IRV and
Distributed IRV can be implemented on paper ballots. Practically speaking, it
becomes confusing, lengthy and costly if the number of candidates exceeds 10 or
20.
Both methods could be
manually tabulated, but once again, in terms of real world practicality the
computation of the final tabulations would take place in a computer.
However, the paper ballots
would have be to accounted for, packaged, shipped, and processed manually for
entry into the computer, including the difficulties related to fraud, human
error and damaged ballots.
Implementing
Electronic IRV
Both Standard IRV and
Distributed IRV can be implemented by employing a computer based electronic
system. The issue here is one of practicality, which translates into the
democratic benefit of an increase in the voter turnout, and the avoidance of
error inherent in the paper ballot system.
There are several ways that an electronic system may be employed. The first is by confining the data processing to a single computer, utilizing a compiled black box form of software. This is the method currently being developed by Bush’s HAVA project and using Diebold as the computer software developer. It is rife with potentials for fraud and manipulation. It is not verifiable or inspectable. And it is hugely more expensive, and once installed, holds the American electorate hostage to that system for an indefinite time. No doubt, Mr. Bush will receive some pecuniary benefit from Diebold, since it will come into a virtual monopoly on voting software for a very long time.
Another, more democratic
method is by utilizing a distributed system based on interpreted languages
which means that they are open to inspection, cheap, easily modified, offer
huge benefits for fraud prevention and cannot easily become anyone’s monopoly.
Finally, electronic voting
means that every ballot is complete and coherent. The distributed totals of
votes or vote values acts as a parity check to detect fraud or system failure.
Ultimately every vote counts and democracy is better served.
Minimizing
Fraud Potential In Electronic Voting
In order to discuss security
against fraud it is necessary to first consider the technology and the
description of the software.
The most important software
consideration is openness. That means it must be written in an interpreted
language and not in compiled code. Interpreted languages are open to
inspection, where compiled languages are difficult and require arcane technical
expertise. Web based sites on the internet use commonly understood interpreted
languages which are in common usage today such as VB Script and Java Script.
Additionally, the open code should not call compiled, third party components
written specifically for voting software, such as touchscreen systems that were
not written and compiled for general software purposes.
In addition, the voter
should never be required to use only that software presented by the central
voting organization, or by some government sponsored program such as HAVA and
its associated commercial software developers such as Diebold.
In order to guarantee
security against fraud, the voter must be allowed to log to, via the internet,
any one of many Independent Polling Web
Sites. Such Independent Polling Web Sites can be created, implemented and
maintained by anyone, but most probably by the parties and by independent
candidates, but also by special interest groups, or coalitions of concerned
voters. Citizens have the potential to monitor their own voting system. Also
special consideration for the blind or for foreign language voters can be
handled by the relevant special interest groups directly. Furthermore, the
expense of developing such special polling sites need not be born by the
government alone, or with raids on public funds, or borne by the general voter,
but can be defrayed to the special interest group.
If the voter is restricted
to one centralized polling site and to one voting program, it is open to
fraudulent manipulation. Therefore, touchscreen voting is open to fraud so long
as it is controlled only by a single, closed polling program. If touchscreen
voting occurs by passing control to a remote Independent Polling Web Site then
it can be inspected and regulated.
A distributed electronic
voting system means that the computers, web sites and their software are
distributed openly to anyone willing to create and maintain such a site. It is
not restricted to a single black box program on a single computer, controlled
by a single polling authority.
A distributed electronic
voting system protects against the fraud caused by the artificial generation of
power-outages, black-outs, machine failures, or simply unplugging the computer.
It does so because voting can occur on any machine with Web access. It also
prevents fraud based on discarding votes, having a paper trail, and data
redundancy.
The real danger in
electronic voting is the same as for paper voting; who is responsible for
maintaining and authenticating registered voters, and black-box, compiled
voting software.
This means that in a
distributed electronic voting system there are two types of polling software
and two types of polling web site. The Central Polling Web Site which runs on
the polling place computer and the Independent Polling Web Sites which are operated
remotely by independent parties. When a voter enters a polling place he logs to
some authorized site of his choice. If he is a registered Democrat he will
probably chose an Independent Polling Web Site operated by the Democratic
Party. If he is a Chinese speaker, he may choose a site operated by a Chinese
Language group. His voting screen is then controlled by that remote site and
not by the Central Polling Web Site. When cast, that is the ballot is filed to
the data base, his vote is counted by both the Independent Polling Web Site and
by the Central Polling Web Site.
This double counting is
crucial to fraud and system error prevention. When the polls are closed a
“committee” is convened consisting of all the authorized and contributing
Independent Polling Web Site and the Central Polling Web Site. The total votes,
or vote values of the Independent Polling Web Sites must equal the totals
reported by the Central Polling Web Site. That means the Central Polling Web
Site and all the Independent Polling Web Site are cross validating one another.
Any discrepancy is either
fraud or system failure. In either case the problem is immediately defined by
cross referencing the Central Polling Web Site and the Independent Polling Web
Sites. If the discrepancy is a system failure the redundancy of the parallel
data will, in nearly all cases, resolve the issue. If fraud is discovered then
parallel data may resolve the issue. If the vote has been found to be totally
invalidated by fraud, then recourse to a Mail-In Runoff Vote automatically
occurs. This is in no way different from conventional voting and/or using paper
ballots with the most important distinction of not open the polls for a second
voting event..
The use of open, interpreted
code is also valuable for boiler-plating the software. That means relatively
simple software is distributed to the Independent Polling Web Site for
implementation. It is easy to inspect and to modify by that Independent Polling
Web Site. Modifications largely reflect the identity of the Independent Polling
Web Site and the identities of the particular election and of the numerous
polling places. A functional demonstration program can be created in a matter
of a few weeks, and the implementation and modification of such software can be
done by the Independent Polling Web Sites in a matter of a few hours.
All the software implemented
by all the Independent Polling Web Sites that are authorized for any given
election can be inspected by all the other Independent Polling Web Sites. It is
an open system with easy inspection.
The same applies to the data
base that collects the voting results. A simple, commonly used DB such as
Access allows for easy maintenance, easy boilerplate distribution and allows
for the common reporting of results; although any DB may be used by any
Independent Polling Web Site that can transmit its data to another DB.
The data base caries the
following information for every vote:
The
Election ID; for example the Presidential elections of 2004
The
polling place ID
The
ID of the Independent Polling Web Site, for example, the Democratic Party’s
Site
The
Sequential Ballot ID
The
ranking of candidates as chosen by the voter
The
time stamp is the same on the two duplicates databases as well as the printed
form
Each ballot is recorded with
a sequential ballot number beginning with 1. This number is not identified with
the voter, only the sequence of valid, unique ballots. It is sequential from 1
to the last ballot for any particular Independent Polling Web Site.
In addition, there should be
duplicated data bases of verified and registered voters. Voters can be asked to
create their own passwords (which seems less open to the criticism of
big-brother intrusions of privacy).
Both data bases will reflect
that a voter has voted and register the time stamp.
Any discrepancies between
time stamps either for the ballot or the registered voter represent either
fraud of system failure. This is particularly expedient in preventing the
insertion of additional fabricated votes.
Every polling place has its
own identity and a printer. When the ballot is cast, the voter receives 2
copies of a printed record of his vote. This is printed on some form with a
water mark or other type of counterfeiting prevention. The printed form contains
the following critical information:
The
Election ID; for example the Presidential elections of 2004
The
polling place ID
The
ID of the Independent Polling Web Site
The
Sequential Ballot ID
The
ranking of candidates as chosen by the voter
The
time stamp is the same on the two duplicates databases as well as the printed
form
It does not identify the
voter.
This is the paper trail.
In the event of catastrophic
fraud or system failure that cannot be corrected by cross validation and data
redundancy an automatic Mail-In Runoff Vote can be implemented without the need
of executing a second voting event. This consists of every voter mailing in a
copy of his ballot. There is no need to hold another voting event.
The results are tabulated
once for each unique ballot identified with a Sequential Ballot ID, an
Independent Polling Web Site ID, the Polling Place ID and the Election ID, thus minimizing the effects of multiple mail-in
types of fraud. The results are easily compared to the data base maintained by
the Central Polling Web Site and cross validated with the data maintained by
the Independent Polling Web Sites and the results are corrected.
In the extreme and unlikely
case that this does not resolve the issue, then holding another voting event as
a conventional Runoff Vote always remains possible.
A distributed polling system
utilizing redundant data bases is the best protection against hacking. Too be
successful a hacking effort would have to hack every one of the distributed
data bases and then modify the data the same in each. At best, a hacking type
of fraud could only trigger a Mail-In Runoff.
Additional safeguards
against hacking is that the polling sites remain off-line, except for system testing
shortly before the actual election, and only on-line during the election
itself. This minimizes exposure to hacking efforts which would require more
time than would provided in order to crack even modest anti-hacking precautions.
Furthermore, a voter is free
to not choose an Independent Polling Web Site and use only the Central Polling
Web Site. This is in keeping with the spirit of free choice and acts as a
further check against fraud. The Central Polling Web Site in this case would
act as both the Central Polling Web Site and as one Independent Polling Web
Site of many. Since it is the Central Polling Web Site that is the most likely
source for fraud, this is a risky choice. But if the voter is Republican and
knows that fraud will be in his favor he may feel comfortable using the Central
Polling Web Site alone.
Incidentally, the Central
Polling Web Site can use any black box, compiled voting software it chooses,
including a touchscreen interface in order to accommodate voters with special
needs or who are too computer phobic to utilize any other type of interface.
This can be permitted so long as some independent site is accumulating the
duplicate ballot data. The protection against fraud lies in the distributed
system of numerous independent polling web sites.
Even if the watermark is
forged, fraud is held to a minimum by the capacity of the system to limit the
recount of one vote to a sequential ballot ID.
Security issues related to
the transmission of voting data apply to any and all electronic voting systems
and are resolvable by encryption, authentication, parity checks and system
redundancy.
Comparing
Standard IRV With Distributed IRV
The first comparison to
consider is that between paper ballots and electronic voting.
Paper ballots must be
printed, distributed, manually marked by the voter, packaged, shipped, manually
deciphered, manually input into a computer and responsibly accounted for every
step of the way. If they are lost, damaged or discarded, there is no accountability
and no recourse other than a 2nd runoff election.
They are expensive and time
consuming in terms of man hours. They are all too prone to fraud and to error.
Electronic systems are also
prone to error and fraud, but in different ways that are ultimately better
controlled, easily validated, more easily rectified and more responsibly
accounted for by both a paper trail and the distributed polling sites. In addition,
electronic voting offers easier access to a wider demographic that should be
reflected in increased voter turnout.
An even more profound impact
is that a distributed form of electronic IRV would permit voting from home, using
the same authentication requirements as paper absentee ballots.
Electronic voting means that there is no ballot cast that is
not coherent and complete. Once a ballot is cast there is no ambiguity. By
ranking the votes or vote values for any number of candidates from 1 to some practical
upper limit, say in the thousands, this system will fairly reflect the
intentions of voters who need never vote strategically and who can always vote
their conscience.
Interestingly, electronic
IRV systems could also apply to issue voting. Normally an issue, say the ABC
measure, is a matter of a binary Yes/No vote. An electronic voting system such
as described here would allow for more than simply up/down votes. It could be
Yes, No, Revise and Revote and even list possible revisions as vote options.
Electronic IRV would permit
direct line item voting by the individual voter. This puts democratic power in
the hands of voter and takes it away from corruption prone representatives.
The comparison between using
IRV as opposed to conventional voting is too obvious to dwell on. Electronic
IRV tremendously increases democratic goals by allowing every voter to vote his
conscience without wasting his vote, in races consisting of numerous
candidates, and also validates and make third parties more viable. It will
probably increase voter turnout and electoral participation by firmly and
clearly promoting the understanding that every vote counts and is counted, that
the act of voting has actually regained meaning and power.
The major drawback to a
distributed electronic system is that it would require different voting
software than everyone else's and what is currently being planned, so there
would be considerable resistance to implementing it since there would be
outraged complaints about the need for new software, there would be expense,
very powerful entities would lose the power of fraud to control electoral
outcomes, and powerful commercial entities would lose monopolistic
strangleholds that result is outrageous profits and public extortion.
More contentiously the
voting electorate would have to be educated and retrained. This is an
electorate that is barely able to count up its toes on both hands on any given
day of the week. So the problem is very real indeed.
Since there is expense the
electorate would need to vote on the use of electronic voting and any method of
IRV. Expense is extremely difficult to calculate in any case. If electronic
voting must replace paper ballots sooner rather than later, the expense cannot
be attributed solely to IRV.
Without widespread
electorate understanding, it is assumed that if IRV is presented too soon to it
will be rejected out of hand. This is a serious objection, but an even more
serious consideration is the rapidly diminishing ability of democracy in America
to reinstate itself by democratic means. In eight years democracy may be as
dead as an old boot.
Since our current voting
system is actually the least democratic system possible and since it has been
rendered useless by corruption, it only makes sense that it is of equal
priority with campaign finance if there is any hope of reclaiming some vestige
of democracy in America.
But first it must be
admitted that some form of Instant Runoff Vote that utilizes existing
conventional voting systems is a more immediate and practical benefit to
democratic recovery, and should be implemented wherever possible and as soon as
possible. Instant Runoff Voting will not remove the destructive influence of
wealth on election outcomes, nor stop the power of wealth to buy elections, but
it will greatly improve the viability of third party candidates and make
elections more democratic by allowing the voter to vote his conscience and not
waste the vote. That alone would be a quantum leap of many magnitudes over the
inadequate level of democracy reflected by the current out of date and
cumbersome voting system.
The distinction between
standard IRV and distributed IRV is based on the issue of iterative processes.
Iterative processes, as used in standard IRV, tends to constrict the tallying
and distribution of votes to a single, unverifiable computer. This is means a
higher potential for fraud. Fraud can, however be overcome by the ability of a
distributed polling system to cross validate results independently. A
distributed IRV does the tallying automatically for each single vote, for the
local sub totals and for the overall totals in one single count. There is no
iteration of distribution and totaling. As such it is more transparent, less
fraud prone, and less consumptive of human and machine time.
The standard IRV system has a huge advantage over
distributed IRV in that it resembles current voting practices regarding 50%+1
as opposed to the highest number of vote values wins. Although highest vote
values wins achieves the same goals more robustly, it will seem alien to people
who can't figure out the underlying mechanism.
Distributed IRV is a concept
that will be difficult to comprehend by the majority of our undereducated
electorate. It is more complex and it does not appear to be like the good old
system, and even appears to be different than the concept of “one voter, one
vote.” It is not, however, different if principle and is far more democratic.
For these reasons the most
compelling argument in favor of standard IRV is that it is do-able and do-able
now at some small municipality like Santa Monica, Berkeley, or Telluride; at
municipalities small enough and with enough progressive clout to have a chance
of implementing it. Berkeley is the
most likely municipality to implement it since it makes the compelling case for
saving $100,000 to $300,000 by not needing to execute the runoff election that
they are otherwise required to do. It is already being used as UC San Diego,
and at various municipalities as Choice Voting.
The delaying issue seems to
be one of voter acceptance. It is troublesome that enfranchised voters are
deemed incapable of understanding a simple IRV system, but at the same time
they are also deemed capable of making critical decisions concerning the
welfare of an entire nation.
The second and related issue
is one of costs. There are costs incurred with electronic voting whether or not
IRV is implemented. There are costs to implementing IRV either on a paper
ballot or for using electronic voting. It is very difficult to estimate costs
that are dependent on whether or not electronic voting is used, paper ballots
are used, what software is installed, etc, etc.
A sample election that
parallels a real election is a simple way to increase general voter
understanding and acceptance, and to overcome voter resistance to anything new
or more complex than tic-tac-toe. It is very do-able to implement an IRV system
on a web site that would emulate a regular election, say, for example, a city
council or mayoral election. This would only be a practice election that would
parallel a real election and would demonstrate how an IRV system would behave
to both the civic authorities and to the average voter.
It might also be useful to
acquire some financial backing from someone with deep pockets like George Soros
or even Bill Gates, who, despite being a capitalist predator, does like to
throw huge amounts of cash at projects that will give a positive PR spin to
himself or to Microsoft. What better spin than galloping to the rescue of
American democracy. Additionally, big Bill can contribute technical expertise
and already provides all the software tools necessary for creating polling
software with open code.
Because fraud resistant
electronic and democratic voting systems would require national cooperation it
can only practically be initiated first by local election entities such as a
municipalities. From successful city occurrences it can then be applied at the
state level, and finally the national. Consequently, it would probably take at
least 8 years to implement this at the national level. All effort expended to that end is, therefore,
dependent on the optimistic view that democracy, America and even the species
is still extant at that time.
It is understood that it may
be considered a very serious challenge to educate voters to be able to log to
an Independent Polling Web Site rather than using a touchscreen interface. But
that is the price of a reasonable guarantee against fraud and rigged elections.
Our elections are already
rigged by wealth, which determines what candidates can even appear on a ballot.
At the very least IRV would allow a democratic and meaningful choice among
those candidates that we are permitted to vote for.
It may also be quiet a challenge
to train the average voter to learn to drag and drop on a computer screen using
a mouse, or some other method of ranking. These restrictions do not seem to be
any more onerous that those that would normally apply to a blind voter.
If the electorate is not
that mentally capable, then democracy is in principle is unworkable in any
case.
No amount of democratization
of voting systems will overcome the two
other egregious assaults on democracy. No matter how free we are allowed to
democratically choose among candidates on the ballot, so long as money, and
money alone determines what candidates are allowed on the ballot there is no
democracy.
So long as public media can
profit from election advertising and so long as public democratic media in America
is owned by monopolies and profit oriented corporate interests, there is no
democracy in America.
Without addressing these
deeper issues, the democratization of voting systems remains only a relatively
insignificant, albeit necessary, first step.
Michael Andrews, 04/2004