by Tom Burrows
April 2, 2012
Version 2.2
Introduction
Problems with Colorado's Current Primary Election System
Proposed Primary Election System for Colorado
Making Caucuses More Useful
Conclusion
Appendix I: Presidential Delegate Allocation Alternatives
Appendix II: Crowd Psychology and the Caucuses
Do you feel that straw polls and delegates you do not know give you more influence over and confidence in election results than having a primary where you get to vote and your vote actually counts?
Colorado's current primary election system—with its associated pre-primaries (assemblies) and pre-pre-primaries (caucuses)—is broken.
Although a few people have learned to exploit this nearly incomprehensible system and will not hear of its abolition, the public may be sufficiently fed up to demand its replacement. In the Mesa County 2012 Republican caucuses, only 2,800 out of 30,000 (counting "active" only) to 44,000 (counting all) registered Republicans attended; and attendance elsewhere is reported to have been even worse. Recent news articles about primary turnouts define a "low primary turnout" as about 20% and an "extremely low turnout" as about 15%. The level of the Colorado caucus turnout was off the bottom of this scale.
Any new system needs to avoid certain technical mistakes in order to be a true improvement. This paper, along with a companion paper on approval voting, proposes a new primary election system for Colorado which meets the following requirements:
All voters registered with a party can participate and have equal influence. There is no caucus packing; and there are no assembly delegates to "game the system" and wield 50-100 times the influence of a regular voter.
No "vote splitting," "spoilers" or related problems. (See the aforementioned paper on approval voting.) Briefly, with approval voting you vote for all candidates you approve of; and the winner is whomever is approved by the most voters. Without the vote splitting problem, we can tolerate a larger field of candidates on the primary ballot; so 1) we don't need assemblies to pre-prune the field and 2) the bar can be set low for prospective candidates who want onto the ballot.
Ballots are trivial and fault-tolerant. Even if you ignore the instructions and simply do something reasonably sensible, your vote will be valid and will be interpreted in a way that does, in fact, make sense.
There is minimal impact on clerks & recorders: Same ballots; same machines; and similar, if not identical, software. (This isn't the complicated "IRV" stuff.)
Caucuses can be retained, if desired, but are re-purposed to serve as "political block parties," "meet-n-greets" or signature-gathering fairs for prospective candidates.
The presidential part of the primary can be either winner-take-all or proportional. This proposal allows for either; and you vote your ballot the same way in either case.
Colorado has voluntarily made itself relatively irrelevant in the national presidential primary process. Colorado voters have so little input into the selection of the state's national delegates that the result is, at best, random and, at worst, possibly corrupted.
With the current system, about all you as a Colorado voter can do to impact the presidential nominee selection is to contribute to candidates—so they can advertise in states that do consider the selection of the most powerful officeholder in the world to be worth having a real primary.
As explained in the companion paper on approval voting, a technical mistake known as plurality voting tends to spoil elections because of "vote splitting" and related problems if there are more than two candidates on the ballot. This problem is exacerbated in primaries, where there can be many candidates splitting an even more homogeneous voter base than in a general election. Instead of nipping this problem in the bud by simply switching to approval voting, the current system attempts to fix this by a complicated system of pre-primaries (assemblies) and pre-pre-primaries (caucuses) to pre-prune the field to a size that doesn't break the plurality voting system.
This complexity has spilled over into the caucus system. Precinct committee-people need to attend hours of training, read multi-page manuals and ask follow-up questions about things the training and manuals don't cover. It is not reasonable to expect that this can be explained to a caucus room full of people who are eager to get out and go home.
One would assume that if complete, authoritative, readable documentation of this system existed, it would have been distributed to precinct committee-people before the last caucus. The apparent inability of anyone to prepare such documentation can be taken as another symptom of our problems.
The layers of delegates between the actual voting public and the decisions coming out of the conventions is analogous to the telephone game, where a message becomes distorted beyond recognition as it is relayed from person to person.
The first problem is that the people who attend caucuses are not representative of the public. Specifically,
Getting people to attend caucuses is like pulling teeth.
People who would like to attend are not adequately notified unless the parties spend either money (they don't have) on postage or huge amounts of time (their members don't have) pestering people.
Since only a small percentage of registered voters attend caucuses (compared to the turnout at full-fledged elections), organized factions only need to pack the caucuses with a few people to overrule the general public.
A "perfect storm" of the above problems occurs when a faction packs a caucus, gets their people elected as precinct committee-persons and then takes advantage of the lack of official caucus notifications to only prod people of their faction to attend the next caucus, which then becomes super-packed.
Some people cannot attend a caucus because of logistical problems (out-of-town, family obligations, etc.) or health.
Advocates of the current system boast that the most "active" people in the party are the ones who tend to go to the caucuses, as if these people vote more intelligently than the general public. The first problem with this is that a system that only attracts the most active people is a failure. The second problem with this, as explained below in Appendix II: Crowd Psychology and the Caucuses, is that caucuses attract a number of well-meaning but frightened and angry people who may not have the best judgement and knowledge. Combine this with "group think" and you get situations like the internet hoax which was exalted as one of the top three resolutions at (at least) one county assembly.
These problems are compounded at the county assembly level:
Delegates are chosen by the same limited sample of the public attending the caucuses.
Delegates are selected from unvetted persons you usually do not know and might not trust.
Secret ballots prevent you from knowing how the delegates representing you voted. (And delegates in small precincts can request what amount to "double-secret" ballots.)
Many people can't or won't attend county conventions for the same reasons they can't or won't attend caucuses.
Even in a caucus that is not "packed," activists who do not represent the majority can easily become delegates because so few people can or want to volunteer.
Since there is roughly one delegate for each 50-100 registered voters in the precinct, this is why I pointed out earlier that each of these unknown people casting an unknown vote wields 50-100 times the influence of a regular voter.
These problems are repeated and compounded yet again at the congressional district, legislative district and state-level conventions. In addition to the problems just listed:
Some of these higher-level conventions are held at remote locations. The necessity of spending days on the road to attend some of these further limits the pool of willing and able delegates.
The remoteness of these conventions relative to where many delegates live results in reduced representation for their regions due to no-shows.
Most of what follows is rationale and discussion of some additional options. The actual proposal amounts to about five sentences, which would replace a system that is almost too difficult to describe at all.
To save money and people's time, and to maximize turnout, the non-presidential and presidential primaries should ideally be held as one event. Adjusting the timing as necessary will obviously require some legislation. In consideration of the presidential primary timing war being waged between the states, the non-presidential primaries are probably more movable.
For all offices other than the President, the proposed primary election system can be described in its entirety as follows:
Those who wish to run as candidates petition onto the primary ballot. (See the remarks below in Making Caucuses More Useful on streamlining the signature gathering process.)
Closed primary elections are held using approval voting.
With approval voting, we can have a large number of candidates on the ballot without the election being spoiled by vote splitting. Therefore,
There is no need to pre-prune the field of candidates with pre-primary events like conventions, which are now eliminated.
Caucuses can be like meet-n-greets instead of delegate-selecting bureaucracy fests.
The threshold for petitioning onto the ballot can be set low; so it will not be necessary for candidates to be backed by political "machines" to enter the race.
The other part of the proposed system is a real presidential primary, also closed, in which voters vote for the desired candidate(s) by-name. There are at least two ways to do this that, unlike the current system, achieve the goals of simplicity and guaranteed reflection of voter intent:
Winner-take-all, using approval voting.
Proportional allocation, using quasi-approval voting. (If you vote for more than one presidential candidate, this means you want the state's delegates to the national convention to be divided equally among those for whom you voted.)
For a more detailed discussion of the rationale and weighing of tradeoffs which went into this proposal, see Appendix I: Presidential Delegate Allocation Alternatives.
Whether winner-take-all or proportional allocation is used, the actual delegates to the national convention who will be supposedly voting for the selected candidate(s) should indeed do so. Instead of voters voting directly for delegates who they do not know and may not trust, it is recommended that the presidential candidates themselves (or their state campaigns) select the actual persons. They presumably know who can be trusted to represent them.
For many rank-and-file party members who do not have politics as a principal hobby, the caucuses are their only opportunity to meet with a concentrated group of politically like-minded people. Unfortunately, this opportunity goes to waste when almost all of the allotted time for the event is spent on the bureaucratic procedure for appointing delegates. As we have seen, a clean-up of the primary process would make the delegate appointment process unnecessary.
This creates the opportunity to re-purpose the caucuses in ways that will make them more attractive and informative for the rank-and-file, such as team-building and candidate meeting. They would be more like "political block parties" or "meet-n-greets". In particular,
Caucuses could be staggered over different days or times. This would give Executive Board members, elected officials and candidates time to make the rounds and speak with all precincts. In addition, people who are unable to attend their own precinct's caucus due to schedule constraints could attend another caucus (at least as an observer).
Staggered caucuses would also be an appropriate venue for prospective candidates to gather signatures for petitioning onto the primary ballot. In the spirit of approval voting, each attendee should be able to approve more than one candidate to appear on the ballot if they find more than one candidate who is worth further study. This also eliminates bias because of the order in which candidates are seen at each location. You don't need to worry about a better candidate than one you already approved appearing later in the meeting. You can approve both.
There could still be a presidential straw poll—with the understanding it is for entertainment purposes only and will be followed by a real primary later.
Most of the activities that would now be the focus are still important and interesting even in "off years" when there is no presidential election; so off-year attendance may be improved.
For reasons discussed in Appendix II: Crowd Psychology and the Caucuses, it is desirable to limit group decisions at the caucus level to the minimum necessary. This might include only the selection of precinct committeepeople to represent the precinct on the county Central Committee. As many decisions as possible should be made by all registered voters on their own instead of in the caucus. This would also (and especially) include things like resolutions (as explained in the Appendix).
Under the proposed system, the county, state and other higher-level conventions are eliminated. Any remaining functions for something like county conventions can be dealt with by meetings of the Central Committee (i.e., precinct committee-people). Such meetings are already being called as needed and typically deal with organizational and administrative matters rather than policy decisions the voting public will want in on.
Having real primaries that work (i.e., with approval voting) will probably require some nontrivial legislation; but all parties need this. If the state leaders of the parties unite and speak with one voice, this legislation can be passed—with opposition only from isolated factions that may have gotten used to gaming the current system.
Suppose we have a delegate-allocation presidential primary in which there are two factions: The Liberty faction with, say, 60% support and the Tyranny faction with, say 40% support. The Liberty faction runs three equally attractive candidates: George Washington, Thomas Jefferson and James Madison; and the Tyranny faction runs one candidate: Joseph Stalin. Depending on the method of allocating delegates, we can get very different results:
Winner-take-all with plurality voting. The Liberty candidates split the Liberty vote and get about 20% each. The Tyranny candidate gets all of the Tyranny vote, which is about 40%. Result: Stalin is awarded all the delegates.
Winner-take-all with approval voting. The Liberty candidates are all pretty-much approved by the Liberty voters and get about 60% each. As before, the Tyranny candidate gets all of the Tyranny vote, which is about 40%. Result: Whichever of the Liberty candidates edged out the others gets all the delegates.
Proportional with quasi-plurality voting. (I.e., each voter votes for one candidate.) The Liberty candidates split the Liberty vote and get about 20% each. The Tyranny candidate gets all of the Tyranny vote, which is about 40%. Result: Each of the Liberty candidates gets 20% of the delegates and the Tyranny candidate gets 40%.
Proportional with quasi-approval voting. (I.e., each voter votes for all candidates they approve of. A vote for two candidates is considered 1/2 vote for each, etc.) The Liberty candidates split the Liberty vote and get about 20% each. The Tyranny candidate gets all of the Tyranny vote, which is about 40%. Result: Each of the Liberty candidates gets 20% of the delegates and the Tyranny candidate gets 40%.
The first method, which may be typical in most states using winner-take-all, is obviously the worst, since plurality voting is simply unsuitable for single-winner elections. The other methods are far superior, but in different ways. Choosing between these involves some tradeoffs:
In the winner-take-all method with approval voting, the most-approved candidate (by definition) gets all the delegates. Any credible but non-winning minority factions will have no representation. This method could be considered "fair" in the sense that it produces the same (most desirable) result which would happen if this were not a delegate-allocating primary, but the general election. An additional feature here is that winner-take-all states produce "more definitive" results, which may cause the whole nationwide process to arrive at a conclusion well before the national convention.
In the proportional methods, vote splitting can cause the most-disapproved candidate to get more delegates than any of the other more-approved candidates. This is true even with quasi-approval voting. However, the most-approved faction candidates taken together get more delegates than the most-disapproved faction; so delegates from the most-approved faction retain the opportunity to unite at the convention and defeat the most-disapproved faction. This method could be considered "fair" in the sense that the results simply mirror the makeup of the voting population. An additional feature here is that proportional allocation states produce "less definitive" results, which may cause the whole nationwide process to not arrive at a conclusion until the national convention.
Thus, the two methods which are recommended are either winner-take-all with approval voting or proportional allocation with quasi-approval voting.
As was hinted at above, a big difference between winner-take-all and proportional allocation is whether the nomination is resolved long before the national convention. There is disagreement among pundits and strategists as to which is preferable. Resolving the nomination early allows your party more time to campaign, but also allows the other party more time to destroy your candidate.
You may have noticed that both of the proportional methods (#3 and #4 above) produce roughly the same result and may be wondering why #3 is not recommended. To avoid voter confusion which would ruin the election, it is vital that the manner in which the ballot is filled out in the presidential part of the primary election be similar to that used for the other parts of the primary election (e.g., commissioners, coroner). I.e., overvotes in the presidential primary need to be allowed, counted and interpreted in an obvious manner. This is what I meant by "fault tolerance" in the Introduction.
You may also be wondering, in the proportional method with quasi-approval voting, what does the "quasi" mean and why does voting for n candidates give 1/n vote for each. The first way of explaining this is that this is not the same as standard approval voting because standard approval voting applies only to single-winner elections—not multiple-winner, proportional allocation elections. Approval voting must be tweaked to what I have called "quasi-approval" voting to apply to the latter. The second way of explaining this is that whereas plurality voting in single-winner elections can cause the faction running the most candidates to automatically lose the election; standard approval voting (without dividing by n) in multiple-winner, proportional allocation elections can cause the faction flooding the ballot with the most candidates to automatically win.
Once upon a time (2012) in a land not far away (Mesa County), some caucuses were deluded into approving a resolution for the Republican party's platform which happened to be an internet hoax. (Do a web search on the phrase "Congressional Reform Act of 2011" for detailed explanations from snopes.com, truthorfiction.com, factcheck.org and urbanlegends.about.com. This hoax is also known by buzz-phrases like "Make congress live under the same laws as the rest of us.") This hoax was then approved by the county resolutions committee and deemed one of the top three at the county assembly.
Before we vest too much decision-making power in the caucuses, it may be instructive to look at how this happened. Specifically, are there inherent features of the current system which create fertile ground for problems like this?
It is my observation that a certain portion of caucus attendees attend because they are fearful and another portion attend because they are enraged. There is also a third portion which could be called normal. Fearful people are uncertain about the state of things and want someone to tell them what is going on and how to vote. Enraged people have an inflated sense that they are right about whatever they believe in. They are eager to tell others what they think is going on and how to vote. Caucuses can now degenerate into forums where the fearful go to be misinformed by the enraged. And if the normal people are not on their guard, they will get sucked in and deluded as well.
Whatever psychological issues these people have as individuals are amplified when they are in crowds, where they can sympathetically resonate with one another. People with poor judgment and poor understanding of reality confirm one another's beliefs; and their judgement and understanding becomes even poorer. This causes the enraged to become angrier still, which encourages even more deluded beliefs. This vicious cycle feeds on itself like the howling feedback in a misadjusted P.A. system.
The people deluding the caucuses were well-meaning, but had already deluded themselves by participating in a "cyber-crowd" consisting of people all over the internet forwarding spam email chain letters to one another—typically prefixed with uncritical commentary like "LOOK AT THIS!!!!" and the traditional and obligatory "PLEASE FORWARD THIS TO EVERYONE YOU KNOW!!!!". People who are gullible enough to read and forward this stuff are in no position to be dishing out advice to others.
Online chat rooms and forums can also function as cyber-crowds providing additional opportunities for them to confirm and amplify the nonsense in which they believe. Examples of this can be found in the aforementioned internet search on "Congressional Reform Act of 2011." After the entries for the sites debunking this urban legend, one can find forums where people rubber-stamp one another's delusions—producing examples of the feedback loop mentioned earlier.
When the deluded enraged log off of their computers and attend caucuses, this howling feedback loop is resumed in a new crowd which now includes the fearful and normal members of the public. They bring their misinformation to the group and say "LOOK AT THIS!!!!" and insist "PLEASE PUT THIS IN THE PARTY'S PLATFORM!!!!"
In the caucuses, the actual population of the "deluded enraged" is small; but they are extremely noisy and it only takes one to lead a caucus astray. Because these same people tend to muscle their way into delegate positions (or more usually, waltz into delegate positions because of poor turnout), their numbers become more concentrated when we get to higher-level assemblies. I've witnessed disturbing examples of this at the last state Republican convention; and I'm told the Democrats have similar problems.
Attacking this problem directly would be very difficult, since we are up against human nature. One way to defuse such situations is to design the caucuses to minimize the decisions made during the meetings. And in the proposed system, the higher-level assemblies are abolished; so the tendency of the associated decisions to be made by concentrations of deluded enraged people there is eliminated.
The most important fact about crowd madness is that people in crowds tend to act in ways that would not occur to them as individuals in the privacy of their homes or voting booths.