[EM] working paper on delegable proxy voting
jarmyta at antioch-college.edu
Sun Nov 13 21:01:30 PST 2005
I thought that I posted this newer version of my proxy paper when I wrote
it, but now I'm not finding it in the archives, so I'll post it now. The
url is http://fc.antioch.edu/~james_green-armytage/vm/proxy.htm
Direct Democracy by Delegable Proxy
by James Green-Armytage
It seems axiomatic that in a democracy, all citizens should have the
opportunity to vote directly on the key decisions of the collective.
However, when such public decisions reach a certain level of complexity,
it becomes impractical for every citizen to become fully informed on every
In a traditional direct democracy system, every citizen has two options
with respect to a given issue: vote or abstain. If most citizens do not
take the time to become fully educated on the issues, but nearly all of
them choose to vote, then public decisions are likely to be somewhat
arbitrary, and easily manipulable by public relations campaigns. If most
citizens do not take the time to be fully educated on the issues, and many
of them do not vote, then the system may become a discriminatory one,
excluding the values held by people in particular sectors of society. If
nearly all citizens do take the time to become fully educated on the
issues, this may excessively remove some peoples attention from other
valuable endeavors. One could argue that there would be a certain amount
of wastefully redundant effort in this last scenario.
Surprisingly, there is a satisfying resolution to this ancient paradox of
democracy: the delegable proxy system.
THE FUNDAMENTAL ELEMENTS
Direct democracy / delegable proxy system
Assume that I am a voter, and there is a certain issue to be decided via
Direct vote option: I have the option of voting directly on the issue
(or formally abstaining).
Proxy option: I also have the option of deferring my vote to a proxy of
my choosing. If my proxy votes directly on the issue, then the weight of
my vote is added to his. There is no minimum threshold of votes needed for
anyone to serve as a proxy.
Delegable proxy: If my proxy doesnt vote directly on the issue, but
rather names another proxy in turn, then the weight of my vote and the
weight of my proxys vote are both carried by this second proxy. There is
no limit to the number of times that a vote can be transferred along a
proxy chain in this way.
Issue independence: Even when there are multiple issues on the same
ballot, I should have the option of indicating separate proxies for
separate issues, while still voting directly on other issues, if I choose.
Comments: The proxy system combines some of the best features of
direct-issue voting and representative voting. People who have a definite
opinion on an issue retain the option of a direct vote. People who do not
have time to become fully educated on the issue may defer their vote to
someone who does, someone with whom they share common values and beliefs.
Proxies differ from traditional representatives in that voters always get
their first choice of proxy. To a much greater extent than traditional
representatives, it is reasonable to expect that my proxy will vote the
way I would vote if I had time to become fully informed on the issue.
The importance of the delegable proxy feature should not be
underestimated. This allows me to choose as a proxy someone whom I know
and trust directly, without having to worry that my vote will be wasted if
he doesnt have time to become educated on that particular issue. This
feature facilitates the accumulation of votes by people who are trusted by
people who are in turn trusted by others.
Issue independence allows voters to indicate as proxies people who are
knowledgeable in the field that a specific issue relates to. For example,
if one issue is relevant to ecology, I might indicate an ecologist as
their proxy for that issue, a staff member at an NGO that deals with the
environment, or perhaps just someone who I know to have researched all of
the different options for that issue in particular.
Ranked ballot, pairwise tally
Assume that there are more than two options for a given issue...
Ranked ballots should be used, such that voters have the option of
assigning a distinct ranking to every option (although they should also be
able to indicate indifference between two or more options).
When majority rule is appropriate (which is likely to be most of the
time), a pairwise comparison method should be used to tally the votes. A
pairwise comparison uses ranked ballots to simulate head-to-head contests
between different options. Given two candidates A and B, A has a pairwise
defeat over B if and only if A is ranked above B on more ballots than B is
ranked above A. (All candidates who are not ranked on a given ballot are
considered to be tied for last place on that ballot.)
The majority rule method should always choose a winner from the minimal
dominant set. This is the smallest set of candidates such that every
candidate within the set has a pairwise victory over every candidate
outside the set. When the minimal dominant set has only one member, this
candidate is a Condorcet winner. A method that always chooses from the
minimal dominant set is by definition Condorcet-efficient. The majority
rule method should also be relatively resistant to strategic manipulation.
When proportional representation is appropriate, some type of single
transferable vote tally method should be used.
Comments: There are many political issues that are too complex to be
reduced to a single up-or-down vote; in these cases voters should be able
to choose between multiple options. The minimal dominant set provides the
best definition of majority rule between multiple options. Furthermore,
Condorcet winners are the most likely options to serve as compromise
For example, imagine that there is a vote where the options are as
follows: (1) drastic change, (2) moderate change, (3) status quo. (The
moderate change is in the same direction as the drastic change, but it is
not as extreme.) The votes are cast as follows:
33%: drastic > moderate > status quo
16%: moderate > drastic > status quo
16%: moderate > status quo > drastic
35%: status quo > moderate > drastic
~ In light of these votes, we can predict that a single up-down vote on
the drastic change would fail. Use of plurality or IRV to tally votes
would also result in the status quo. A Condorcet tally, however, picks the
compromise solution when there is one, in this case the moderate option.
A Condorcet-efficient method is likely to pick the median even with a
large number of options, as long as they are arrayed across a single
spectrum. A Condorcet-efficient method is essential to identifying
centrist, non-polarized choices, provided of course that a sufficient
compromise option exists on the ballot.
VARIABLE ELEMENTS AND INTERESTING QUESTIONS
Delegable proxy systems can be used for a very wide variety of voting
purposes. In this paper, I will focus mostly on their potential use for
direct-issue voting on government policy.
Several degrees of bindingness are possible. One option would be to have
a non-binding system that allowed the citizens of a nation to express
their opinion without carrying any legal weight in itself. The government
would still fund the system and ensure its integrity, and there should be
some general pressure put on government officials to act in accordance
with its results.
Another option would be to subject the public decision to veto by other
branches of government, e.g. the executive, a majority of the legislature,
a supreme court. A third option would be to allow other branches of the
government to postpone implementation of the public decision, but not to
reverse it. A fourth option would be to give the public decision
preeminent legal status.
Frequency of votes
The number of issues decided by this system per year is a matter of
preference. It might even be possible to decide all issues that would pass
through the legislature this way, thus effectively supplanting the
legislative assembly. However, I suggest that this system should coexist
with, rather than supplant, the elected legislators. For example, one
could have three public votes per year, with ten issues on each ballot.
(Any other combination of numbers is possible here.) Thus, the proxy
system could serve as a guide for the legislative process rather than its
A question: Should the votes should take place over the internet, or only
at controlled polling stations? The internet poses problems of security
and problems of equal access, so I suggest that official polling stations
are a preferable venue. The voter interface should be electronic (paper
ballots would probably just be too clumsy for this system), and every
effort should be made to assure that the votes are being counted
Voters should be able to choose their proxies from a complete list of
people registered as proxies. This list should be kept on a secure master
file that is also a matter of public record, available on the internet,
etc. Aside from just their name, there should be a few distinct keywords
and some basic information about them so that people who want to choose
them as a proxy can distinguish them from others with the same name. Proxy
registrations should be filed some time in advance of the vote, to give
couriers a chance to bring secure copies of the proxy list from the
Ranked proxy lists
One may allow the voters to indicate a ranked list of proxies, rather
than a single proxy.
One may allow the voters to register for a standing proxy, so that if a
voter doesnt show up to vote on a given issue, his vote is delegated
automatically to his standing proxy. Or, one could require voters to show
up in person, if only to indicate one proxy for all of the issues on the
ballot. If one does not allow standing proxies, then the ranked proxy
lists could be used to assure that votes are not wasted. That is, if my
first-ranked proxy doesnt show up to vote, my vote could be deferred to
my second-ranked proxy instead.
Resolving proxy loops
It is possible that a proxy loop might arise, if for example voter A
indicates B as his first proxy, B indicates C as his first proxy, and C
indicates A as his first proxy. Ranked proxy lists can be used to resolve
loops. One possible rule is as follows: "A vote shouldn't travel the same
proxy path twice." Given the above case, A's vote has traveled the path
A-->B, then the path B-->C, and then the path C-->A. Therefore, according
to this rule, once A's vote returns to A, it should not once again travel
the path from A to B. Instead, it should travel to the next proxy as
ranked on A's proxy list. The proxy path rule is not very important, since
such loops are not an especially daunting problem. Other rules are
possible, for example "a vote shouldn't be assigned to the same person
twice," in which case A's vote would be transferred to C's second proxy
rather than being assigned to A once again.
In some circumstances it might be possible to resolve proxy loops without
the use of ranked proxy lists, for example by informing the people
involved in the loop and requesting that they indicate another proxy.
A question: Should someone who serves as a proxy for other voters still
be entitled to keep their votes secret? It is theoretically possible to
keep their votes secret from the general public, but one cannot avoid the
fact that information will be stored in the computers responsible for
tallying the result. Also, it seems likely that most voters would like to
know for sure how their proxy voted.
Generation of issues
I propose that some issues to be decided by the proxy system could be
generated by the legislature, while others are generated by the public
process itself.~I suggest that we might want to set the agenda for a few
direct votes at a time. For example, let's say that there were three
direct votes per year, each with about ten issues on the ballot. At the
end of one year, we could decide which issues to vote on over the course
of the next year. Each time we went to the polls for a direct vote, there
could be a combination of congressionally generated and publicly generated
issues on the ballot.
For issue-generation inside the legislature, I suggest a system of single
transferable vote (STV) proportional representation. For example, the
legislators could take an STV vote to fill a certain fixed number of slots
for issues in an upcoming direct vote. Perhaps the legislature should also
have the ability to add extra issues in case of emergency.
For issue-generation outside the legislature, I propose that issues
should first be nominated via a public process, and that nominated issues
should then be placed on a ballot for a direct agenda-setting vote. (Of
course, we do not need to drag voters to the polls just to do an
agenda-setting vote; it can be put on the ballot with other issues.) The
public agenda-setting vote can also be based on STV, filling a fixed
number of slots.~
Question: How would issues qualify for the agenda-setting vote itself? A
certain number of petition signatures?
Question: Should the legislature (or some other entity) maintain the
power to assert that two or more issues should be combined into one to
avoid redundancy and possible contradiction? What should the boundaries of
this power be?
To sum up, we could do an agenda-setting vote about once a year, with the
legislature filling a preset number of issue slots, and the public filling
a preset number of issue slots. Once the issues-to-be-voted-on had been
decided, we would divvy them up onto different ballotings (e.g. one
balloting in February, one in June, etc.), and we would get to work on
generating the options for the issues.~
Generation of options
Once it has been decided that there will be a direct vote on a given
issue, the next step is to generate the different options that voters will
choose from when voting that issue. The goal here is to make sure that a
less-than-ideal option doesn't win because a more effective compromise
option didn't make it onto the ballot.
Option-generation inside the legislature: Again, an STV vote is logical,
but in this case, an option called "no additional option" should be in
competition with the other options that have been proposed. A certain
number of maximum slots would be available to be filled with options, but
if many legislators are satisfied with less than the maximum, some of the
extra slots could be filled with "no additional option", that is to say,
Option-generation outside the legislature: There should also be a public
process for generating options. Question: should there be a separate
public STV vote for option selection as well as issue selection? This
might prove too cumbersome to be worthwhile, in which case any proposed
options meeting the given requirements (e.g. acquiring a certain number of
petition signatures) could be placed directly on the ballot without
further filtering or a hard limit as to the number of possibilities.
How technical should the options be?
Should options be written in the sort of dense legalese that
characterizes the handiwork of most professional legislators, or should
they be worded in a more general, approachable way? Where can the balance
be found between legal precision and common understandability?
People who serve as proxies would be given a score, calculated yearly,
which would be based on how many votes they actually cast on specific
issues. That is, if there was a vote where many people named me as a
proxy, but I deferred the votes to someone else, it wouldn't count toward
my proxy score. However, if I applied the weight of those votes toward a
specific option, it would count. Proxy scores should be averaged /
normalized so that a score of 300 basically means that on average during
the previous year, I executed direct votes on each issue for 300 people
including myself. Hence it would be meaningful to say that a single
proxy's score was equivalent to a specific fraction of the total number of
Proxy scores could potentially be used to facilitate the process of issue
generation and option generation, in that the signatures of proxies could
carry the full weight of their normalized proxy score. Hence, instead of
having to get thousands of individual signatures, one could achieve the
same effect by getting the support of a few high-scoring proxies.
Question: Should those who serve as proxies for many other people be
given monetary compensation? If we do decide to compensate super-proxies,
what money-allocation formulas might we want to use? Here is one
possibility: There is a total preset budget for proxy remuneration, which
is divided among those who receive compensation. Each super-proxy's share
of this overall budget is proportional to a modified proxy score, which is
calculated as follows: The modification comes as a result of a minimum
threshold, such that those whose proxy scores are below the minimum will
not receive remuneration. Instead, their scores are added to those whom
they list as a standing proxy, and they continue up the proxy chain until
they coalesce into an above-threshold score somewhere along the line.
For example, if I am the proxy for about 10 people, and thus have a proxy
score of 10, that might not be enough to justify the paperwork of having
the government send me a check, etc. So the remuneration-weight of my
score should be passed along to my proxies, and perhaps their proxies, and
so on, until it gets to someone who crosses the minimum threshold. If you
like, you could do this in a series of successive rounds, such that first
you eliminate below-threshold people with the very lowest proxy scores,
and transfer their remuneration-weight before doing the next round of
Another question: Should the government place regulations on what proxies
can do with their remuneration? For those who receive only a moderate
amount of money, this might not be worthwhile, but if someones total
remuneration exceeds a certain amount, it might make sense to require them
to invest funds in excess of that amount into policy research ventures
that meet certain standards. I imagine that these research groups could
help to further the democratic goals of the proxy system.
Comments: The argument in favor of proxy remuneration is that the extra
money could help them do good policy research, by allowing them to reduce
the hours they spend at other jobs, and in some cases by allowing them to
hire research staff and acquire research-facilitating capital. The money
may also serve as an incentive to do good research. The argument against
it, aside from the cost, is that it might produce impure incentives for
people to act as proxies, and to over-represent their understanding of
policy in an attempt to advertise themselves.
Virtues of beginning with a non-binding system
(1) It is important for the system to build public participation and
trust before it begins to carry the weight of legal power.
(2) There should be more freedom in the design of a non-binding system,
in that there should be less pressure and more leeway for trial and error.
This will make it easier to incorporate innovative ideas such as the
delegable proxy option and the pairwise tally. It should be much easier to
apply advanced voting rules to an entirely new institution than to an
existing electoral institution, as existing institutions tend to get mired
in entrenched practices and interests.
(3) It should be much easier to achieve than a binding system, and it
should greatly hasten the advent of a binding system.
Already-existing direct-issue voting systems
For those locations that already use direct-issue voting systems of one
kind or another, there are at least two possible goals:
(1) To try to make changes in the system so that it incorporates some of
the essential elements of a good direct-issue voting system, as listed
above. For example, one may try to introduce a delegable proxy option. One
may also try to encourage multiple-option issues with a ranked ballot and
pairwise tally. (Question: which Condorcet completion method should be
used in this situation?)
(2) To create a non-binding system with the essential delegable proxy
elements, as a supplement to the existing system.
Alternative political leaders
The proxy system acts as a perfectly high resolution version of
proportional representation, in that voters always get their first choice
of representative. As such it may provide an increased political role for
third parties, nonprofit organizations, and community leaders.
Expressing and developing the nuance of public opinion
This system attempts to end the state of affairs where most citizens have
only a very diffuse and indirect impact on government policy. In most
plurality elections, the voters essentially have to choose between two
ready-made issue packages. Limiting the voters choices this way totally
glosses over even the slightest bit of nuance that might exist in people's
political beliefs. This oversimplification is so drastic that its
dangerous. Its important to have separate votes on individual issues for
the sake of clarity. There are scores of important and controversial
issues which deserve better than to be lumped together into
winner-take-all packages, or worse yet, to be ignored by all major
parties. It is frustrating to know that there must be several good
solutions to pressing social problems that would be supported by a
majority if they were put to a public vote, but that for some convoluted
reason they are not implemented by the ruling party/parties.
Furthermore, if ordinary people actually have an opportunity to influence
policy by expressing their views on key issues, I suggest that it will
help to clarify their positions on these issues, and to engage far more
people in the political process.
I dont intend to imply that a good direct democracy system eliminates
the need for other electoral reform, campaign finance reform, media
Direct democracy versus privately conducted polls and focus groups
Opinion polls and focus groups have a lot of influence on politics at
present, but they are not an appropriate substitute for direct democracy.
Both are generally controlled by private entities who have a good deal of
power to influence the result, e.g. through suggestive wording and through
the omission of relevant options. Focus groups take place behind closed
doors, routinely have unpublished results, and in general do very little
to further political discourse. Both polls and focus groups are
participated in by only a small segment of the population, leaving
everyone else with their opinions unheard. Also, they reduce the
participants to a relatively passive role that does not foster political
organization, action, or education.
This is why even a non-binding system would be a tremendous improvement
over the current system. It would allow the popular will to come through
as a result of direct political action in broad daylight rather than
through the dubious conduits mentioned above. It engages people to get
involved, to discuss, and to act. It includes all adult citizens who are
willing to participate, rather than just a few people who are randomly
selected by a network news agency, or who strike a focus group research
firm as being representative of swing voters.
Application of the delegable proxy system is by no means limited to
governments. Almost any organization with a sufficiently large membership
can potentially benefit by using it. (E.g. unions, schools, churches,
etc.) The larger the organization, the more important the delegability
feature becomes. One particularly interesting application is to publicly
traded corporations. Many corporations allow shareholders to vote by
proxy, but in many cases, shareholders do not know much about the proxies
whom they designate. If corporations were required to allow for delegable
proxy voting, then I as a small stockholder could potentially delegate my
votes to nonprofit organizations that shared my values with regard to
corporate policy. (I should retain proxy vote rights even if my ownership
of stock is brokered by an intermediary agent.) This could help corporate
policy to conform more closely to the values of ordinary citizens, just as
the general proposal could help government policy to do the same. Hence,
the use of the delegable proxy system could bring us closer to both
political and economic democracy.
More information about the Election-Methods