Avoiding the Two Party Duopoly
By Removing a Well-Hidden Bias in the Voting System
Voters have all sorts of different concerns. For example, one voter's major concern might be that the military simply takes up too large a share of the federal budget. But that voter knows that there are only two candidates with any chance of winning election and both come from parties with no interest in restraining military spending. Would that voter appreciate more options? Is that voter even so unusual? Other voters surely have other concerns neither of the two viable candidates share.
Surely there is some debate on the question, but at least there seems widespread agreement that it would count as an improvement if somehow we could have more than just two viable parties. With only two candidates to choose from, voters simply do not have enough choices and even three parties might not be enough. Ideally, there should never be a barrier for a new party to form and compete fairly with some reasonable chance to win so that all views at least have a chance to be represented.
So why not put an end to our two-party duopoly? The most obvious impediment might appear to be that the two dominant political parties are no more interested in ending the two-party duopoly than they are in restraining military spending. As evidence it might be noted that Democrats and Republicans have conspired to erect and maintain barriers designed keep smaller parties from even appearing on ballots. That is a problem to be sure, but that is not the fundamental problem.
Limits on ballot access may have seemed perfectly justified so as to avoid a seemingly inherent problem with elections. When there are more than just two candidates anomalous election outcomes can and do occur at times, for example, when there are three candidates but voters find two of them to be quite similar. Perhaps 40% of the voters prefer the one distinctive candidate while 60% do not much care which of the other two wins so long as that one distinctive candidate loses. But in the election, what will probably happen is for the two similar candidates to both lose with roughly 30% of the vote each, disappointing a clear majority of voters. This is the spoiler effect, so called because each of the two similar candidates spoil the election for the other; but really it is the election that is spoiled and the voters who suffer. To remedy this problem, politicians decided to limit ballot, ideally to restrict elections to just two candidates. That works to some extent of course, but this solution now may seem worse than the problem. Instead, might we adopt a voting system that will work well despite there being three or more candidates?
One alternative voting system, instant runoff voting (IRV), does in fact avoid the spoiler effect. This alternative has been actively promoted precisely on that basis and has gained considerable interest in recent years. Maine adopted IRV through a citizen ballot initiative after experiencing what appeared to be spoiled elections. But there is no reason to think IRV to be the best available alternative and in fact, there are worrisome issues that should temper enthusiasm for IRV.
A less widely familiar system that also avoids the spoiler problem is approval voting (AV). AV is an evaluative voting system, meaning that voters are asked only to evaluate each candidate individually, avoiding a need for voters to favor one candidate over another. With evaluative voting systems, it is not necessary for voters to find some extraneous reason for preferring just one of them (such as some poll or some pundit's opinion about which is more electable). A voter can simply give identical ratings to equally suitable candidates.
Evaluative voting systems make voting easier, in part because they avoid introducing such extraneous horse-race issues; voters are not even forced to make hard choices between candidates. And it is those issues tend to prompt voters to opt for big-party candidates over candidates they actually believe would be better choices. With AV, voters can concentrate purely on the suitability of each candidate. That should allow an AV election to take a more accurate reading on which candidate the voters judge as best suited for office.
But just as with IRV (or for that matter any other voting system), it is entirely possible that there are voting systems that would be preferable to AV. Such systems might even bring an end to the two-party duopoly. For several years, I have been considering this very issue, producing a series of articles that might be considered a journal of insights as they occurred to me. None of these articles are particularly lengthy but there are now sixty-four of them and some of the articles do drift away into related topics. What prompted me to write this article is to provide a single article focusing only on putting an end to the two-party duopoly while summarizing the relevant ideas from the entire series of articles. I plan to add other articles to take the same approach for other topics addressed in that series.
Of the two alternatives to plurality voting described above, AV clearly seems the more promising. An AV ballot displays a list of candidates and the voter is asked to check a box to indicate support for candidates when appropriate. When the ballots are counted, a tally is made of the number of ballots supporting each candidate; the winner of the election is the candidate with the largest tally of supporting ballots.
Quite often, voters are not so much concerned with supporting a candidate as they are with avoiding the election of a different candidate. When there are only two candidates it is clear, even with plurality voting how to vote for this; but as noted, our objective is to allow for more candidates. If defeating one specific candidate were a voter's sole objective then with AV a voter could simply indicate support for each of the other candidates. But in an election with many candidates voters are apt to support some candidates while opposing others and neither supporting or opposing still other candidates. Given this more complex agenda, even an AV voter can have no easy plan.
Does not opposition to a candidate seem every bit as important as support? So could a voting system not allow a voter to explicitly show either opposition or support for candidates? Importantly though, there probably will be candidates that a voter neither supports nor opposes (for example, the voter may first be learning of the candidate when reading the ballot). So voters actually need three choices for each candidate, support, oppose and neither.
AV can easily be modified to make this possible. Instead of a single check box next to each candidate's name, two are needed, one for indicating support and the other for indicating opposition. When a voter checks neither box (or both boxes) then the voter is effectively saying neither. When ballots are tallied, both supporting ballots and opposing ballots are counted to produce, for each candidate, two separate tallies. For each candidate, the net-tally (or net-vote) is computed as the support-tally minus the opposition-tally and the candidate with the largest net-tally is declared the winner. We call this very system, balanced approval voting (BAV).
Polling companies often take exactly this style of measure of likely voter opinion and often report both support level and opposition level along with their estimate of net support. BAV elections effectively try to take that same measurement.
BAV, just like AV, is an evaluative voting system so a voter simply considers each candidate in turn and chooses whether to support or oppose that candidate (or, perhaps even in many cases, neither). As with other evaluative voting systems, BAV voters are free to ignore how other people might vote and free not to take account projections of how the election might turn out. In some measure, this should improve the election's ability to measure which of the candidates is viewed as the best choice.
It is worth considering what might happen if some state or locality, long accustomed to plurality voting, switches to BAV. For the sake of this example we can assume that at least one other party, say the Libertarian party, manages to get onto the ballot. Many voters will probably not even recognize the name of the Libertarian party candidate so how will they evaluate that candidate? Typically, these voters will mark neither support nor opposition. Meanwhile, Democrats will generally mark support for the Democratic candidate and opposition for the Republican candidate. Likewise, Republican voters will mostly mark support for the Republican candidate and opposition for the Democratic candidate. This brings down the net-tally for both the Republican and the Democratic candidate to well below their respective net-support tallies.
So, given a polarized electorate, BAV significantly lowers the bar for minor party candidates, especially when the two major parties are about equal in size. This happens not because any special advantage it grants to smaller parties but because BAV removes the advantage that other familiar voting systems provide for the very large parties by not allowing voters to express their heartfelt opposition to any candidate.
Voters would learn from just one win by a minor party candidate not to assume elections are limited to just the two major parties; there are now other possibilities and simply knowing knowing that will affect how they vote in the future. They will pay attention to candidates from smaller parties and the media may, in time, also feel obligated to treat these other candidates more seriously. The major parties can and likely will still win elections when that is what the voters show they want, but they will have to face more competition, In time, small parties will grow and the distinction between large and small parties will not continue to be such a controlling issue. Candidates will continue to feel pressured to attend to their supporters but to avoid antagonizing other voters will come to seem equally important. Campaigns will likely become more civil once candidates realize that negative ads and bombastic speech could prove counterproductive.
Confined thinking inside the box (where voting is considered synonymous with plurality voting) has not only created the two-party duopoly, it has also influenced our language. We talk of the other party, for example whereas with more viable parties there will be no single other party. With a view beyond the confines of the plurality voting box comes the realization that specifying only your first choice among potentially many candidates cannot possibly provide adequate information for a proper election. With more candidates, voters' opinions become more complicated and cannot be adequately described as an answer to which candidate is your favorite? With IRV, AV and BAV we have seen examples of some ways additional opinion detail might be collected. While collecting more detail about voter preferences is necessary, there is a countervailing need to keep voting reasonably simple. Overly complex voting systems surely have the potential of driving voters away.
With these alternative voting systems, the meaning of the noun, “vote” can become confusing. With AV, for example, does it refer to the entire list submitted by a voter or does refer to just one of the entries in the list? Within that list is there one vote or several? To make such distinctions clear we will try to avoid using vote as a noun and instead use either the word, votelet for an individual entry or the word voteplex for the full list of votelets compiled by a voter. This new terminology suggest an updated principle for elections of one person, one voteplex.
Just as there are many evaluative voting systems, not just AV and BAV, there are many different balanced voting systems. But we have not yet explained what that means (we first had to learn this new terminology). A voting system is balanced if whenever a voter can choose a votelet indicating support for a candidate that voter can just as easily choose a votelet of opposition that would effectively cancel an support votelet on a different ballot. Put another way, the votelets available to voters must come in pairs that balance out each-other.
It should seem clear that if we want to adopt a better voting system then we probably should look among systems like BAV that are both balanced and evaluative. There surely are many such systems (an infinite number of them, actually), but the one I would point to is Balanced-Randomized Voting (BRV). As the name suggests, this system uses a statistical sampling approach. Whether the general public would find a resort to statistical sampling acceptable is a matter of conjecture, but it does have some advantages.
BRV seems best suited for winnowing down the number of candidates when there are simply too many candidates to ask voters to make a final decision on at once. For elections with a manageable number of candidates, BAV would likely be a better choice. But surely there will always be the possibility that someone might discover a way to improve on whatever voting system we might use.
FAQ's
Q: Is there any way this could happen? How could we possibly get a system like BAV adopted for our elections.
A:
Adopting BAV will not happen suddenly, at least for this country, but it could soon be adopted on a smaller scale. It will not be easy to get elected officials, particularly in Washington, behind changing a system that has in the past worked to their benefit. But then again they are not necessarily in control. This is a state issue and sometimes even a local issue.
Voting methods can and have been changed in the past. Most notably, Maine recently adopted IRV. This this came about through a citizen initiative to put the question on the ballot and the voters then approved it. Such is possible in a number of states. Nationally, this could only be a gradual state-by-state process, but that slowness could be helpful. It would make it possible to work out unforeseen problems before they become really serious scale.
Where BAV could be adopted quickly, even at the national level, is in the nominating process. Multiple candidates, not just two, often stand for elections in primaries. Election outcomes would become more representative of actual voter preferences if they used a more suitable voting system.
Political parties are, at least in concept, private organizations that can choose their candidates however they choose. They do not have to participate in the federally sponsored state-by-state primaries and they do not have to feel confined to using plurality voting. Smaller parties in particular have nothing to lose and much to gain by adopting an improved approach for choosing their candidates, perhaps in nation-wide elections by mail or using the internet.
Q: With BAV, might not a candidate win election with a negative net-vote?
A:
Yes, that is possible. But should that happen, it would be apparent that the winner lacks a mandate. It is a virtue of a BAV election that it will make that clear.
There happen to be easy cosmetic solutions that can hide this from sight if indeed it is seen as a blemish. Instead of choosing the winner using the net-vote we could reach the same election outcome by instead using what we might call the preference index, defined as the net-vote plus the number of voters. Election outcomes will remain unchanged but the winner might be more inclined to claim a mandate.
Q: Which voting system, PV or BAV, makes it easier to vote?
A:
That is a good question to think over and discuss, along with the question of which system would make voters feel more satisfied with the way they voted.
Q: Is BAV not the same thing as score voting with three scores?
A:
No, not quite, but that is a great question. It has often been asked and the answer brings out some illuminating subtleties about both BAV and about score voting.
Score voting names a collection of abstract voting systems and BAV rightfully deserves a place in that collection. But BAV has some additional characteristics that go beyond merely being a score voting system, different even from score voting systems using the very same three scores.
Abstract models often make simplifying assumptions and that in fact is why the term, abstract, is used; the abstraction retains only what is important. Sometimes, these assumptions even conflict with reality, on the assumption that the conflict is harmless. But in large part, the failure to pay attention to the possibility of it not being harmless is precisely what is at issue with this question.
Score voting (as commonly described) requires a voter to specify, for each candidate, a votelet to select one of the available scores, each score being an integer. In real elections using score voting, however, score voters are even encouraged to skip over some candidates whenever they wish. This puts real-world score voting a bit at odds with the abstract model. Does this matter? What happens when the voter fails to specify a score?
When no score is specified, perhaps the tally just skips over such ballots; in effect this is the same as making 0 the default value. But if the scores 1, 2 and 3 are specified then this in effect becomes score voting with four scores, namely 0, 1, 2 and 3. Instead perhaps the counting would adopt the smallest score, 1, as a default. This would seem the natural choice because the familiarity with plurality voting has established the habit of disregarding voter opposition to candidates for elections. It's not impossible that some score voting election has used the middle value as a default, but I do not know of one.
BAV allows voters to choose one of not three but only two scores (support or oppose) for candidates of the voters' choice. BAV is conceptualized to fit within the score voting model (with three scores) by assigning 1 for support, -1 for oppose and 0 otherwise. Effectively, BAV defines a default value (of 0) in cases where the voter fails to specify either of the other two scores. So real-world BAV actually fits nicely within the score-voting abstract model. What about real-world score voting?
The choice of specific numbers to act as scores does not actually matter much for score voting so long as the spacing between scores remains in the same proportions. But the choice of -1, 0, 1 seems a natural assignment of scores for BAV that makes it easy understand and appreciate how it works. That is what makes it what mathematicians would call the canonical form for BAV. These scores are not the only possible choices but for most purposes they are the best choices.
Defining the default value to fall mid-way between the other two available scores is the specific feature that distinguishes BAV from other score voting systems with three scores. BAV requires the default to be the middle score whereas score voting simply tends to ignore that there might be any need to specify a default value.
Whereas BAV denotes the three scores as oppose, neutral and support, the traditional score-voting/plurality voting thinking is that the three values represent neutral, support and strong-support. This difference in interpretations affects elections only because it determines the selection of the default value, something that score-voting simply fails to recognize even as a relevant concern. But in real-world elections, the choice of the default value has the power to change election results and likely even to influence the political climate.
In a polarized two-party duopoly, many voters want to support their own party while also wishing to oppose the other dominant party; of course they may feel inclined to support or oppose minor parties as well but they tend to consider that of little importance. If given the opportunity to express opposition explicitly, voters might exercise that option even with respect to minor party candidates but nearly always against the other major political party. It follows that to give voters that option removes a significant gift that the two dominant parties have long enjoyed. Enabling voters express opposition explicitly, as BAV does, removes a bias that works to the benefit of very big political parties and thereby to perpetuate a polarized duopoly.
While few voters will abstain from specifying either support or opposition to the major parties in a duopoly, many will not bother with minor party candidates. So treating voter indifference as if it were opposition (say by assigning the smallest score to be the default) is another way to understand how easy it is for a voting systems to perpetuate the duopoly.
This topic is explored in the article, Notation for Voting Systems.
Q: AV asks voters to put candidates into two boxes, support and non-support. BAV provides more options by allowing voters to put the candidates into three boxes, support, oppose or neither. Would it not be an even further improvement to allow voters to put the candidates into four, five or twenty boxes?
A:
Ranked voting systems, like IRV and score voting systems with more than three score do allow voters to express their views with more refinement than BAV does. Notably, score voting systems with any odd number of scores are both evaluative and balanced. But whether the extra refinement of voters' expressed opinions is worthwhile is less clear. The meaning of the three choices is quite clear with BAV but as the number of boxes increases their meaning becomes less clear.
If asked, probably most voters would say they like the idea of being able to make these more fine-grained evaluations. I suspect this is mostly what many people find so appealing about IRV. But voters might reconsider their answer if also asked whether the added flexibility would make it easier or harder to vote.
With score voting, voter satisfaction is not the only concern. Adopting a larger number of scores calls for more careful judgments from the voter that could lead the voter to compare candidates with similar scores. This would be another reason for horse-race issues to color their voting choices.
The perspective of election administration also matters and clearly the more refined approaches do complicate the ballot and the vote-tally. Moreover, it is not at all clear that election results would be improved. This consideration was discussed in the article, Can Less be Better, and in a different way in the article, The Perfect Voting System.