A catalog of alternative proposals

EngineerDemocracy seeks to promote good and responsive government by reducing partisanship and extremism and increasing the diversity of represented viewpoints. In addition to these goals, our selection of proposals was guided by the following criteria: that proposals be robust to strategic voting, encourage honest voting by making it the most strategically effective approach, be robust to election tampering, disadvantage the Democratic and Republican parties roughly equally, and require only typical levels of knowledge and civic devotion among the general populace. The proposals we have advanced are those that we believe meet these criteria and most effectively contribute to our primary goals.

Alternative ideas abound, however. For interested readers, we describe here some of the more popular ones and explain why we have not chosen to promote them (though we applaud the spirit of their proposal).

Multi-member districts

One simple way to increase the diversity of viewpoints represented in legislative bodies such as congress is to establish multi-member districts with plurality elections in which each voter can cast a single vote. Such a system encourages the formation of at least as many political parties as there are seats to be filled in the election because the best strategy for each party is to tailor their platform to appeal strongly to just a large enough percentage of the population to win. In a district that elects six members to congress, for example, any candidate receiving just over one seventh of the vote is elected so candidates will typically try to appeal to one seventh of voters and there will be six or seven political viewpoints represented. Such elections still suffer from the spoiler effect common to all simple plurality elections with more than two candidates, but it tends to have less of an impact and could, at any rate, be eliminated by replacing simple plurality with transfer elections.

Our objection to multi-member districts is that while they certainly increase diversity they also encourage extremism and dysfunction. Candidates have no incentive to appeal to a larger percentage of the population than necessary to win election and the most effective way to appeal to a small segment of society is often to campaign on providing them with special treatment under the law or to focus on divisive issues that only they support. Candidates in districts that elect many members can focus on niche concerns to the complete disregard of a comprehensive plan or philosophy of government and as elected representatives can disregard important legislation or even disrupt it to advance their limited agendas. The classic cautionary tale about multi-member districts is the Weimar Republic, whose dysfunction due to an utter lack of unity ultimately led to the rise of Hitler, but there are more recent and less extreme cases such as the current Italian government.

All of that being said, we must admit that apparently well-governed countries such as the United Kingdom and modern Germany utilize multi-member districts. While not as representative as sortition and not as unifying as transfer elections, multi-member elections in districts that elect three or four members to the legislature might be a good balance between diversity and unity.

Multi-member districts are promoted by FairVote.

Approval Voting

In an election using approval voting, voters are allowed to vote for more than one candidate and the candidate with the most votes wins. Since voting for your favorite and voting for your favorite frontrunner are not exclusive, this somewhat weakens the spoiler effect, where voting for your preferred candidate can cause your least preferred candidate to win. Approval voting is not strategy free, but the simple strategy of voting for every candidate that you like better than your least preferred of the two frontrunners is fairly effective and appealing.

Our main objection to approval voting is that we do not think that switching to it would have much effect on the outcome of elections. The reason is that with approval voting it is still important to vote for whichever of the two frontrunners you like best. Otherwise you have no say in the contest that is likely to decide the election. But if everyone who prefers a candidate that is not a frontrunner also votes for one of the two frontrunners then only the frontrunners can possibly win. In elections where three or more candidates that are competitive approval voting can have more of an impact, but then voters are faced with the strategic choice of more strongly opposing the candidates that they like least, by voting for more candidates, or more strongly supporting the candidates that they like best, by voting fewer candidates.

Approval voting is clearly preferable to standard plurality voting, but it is incompatible with transfer elections. The latter is strategy-free (for voters) and seems more capable of translating voter's preferences into wins for the preferred candidates. Consequently, we do not promote approval voting, though we believe it to be the next best option to transfer elections.

Score Voting

In score voting, voters score candidates on a fixed scale, e.g., 0 to 9, based on their perceived utility to the voter. The winning candidate is determined simply by summing all the scores for each candidate and selecting the candidate with the highest total score. If all voters vote according to their true beliefs the winning candidate is the one that has the highest utility to the electorate as a whole. Thus, if 70% of voters in a two-candidate race rank the liberal candidate as a 6 and the moderate candidate as a 5, but 30% of voters rank the liberal candidate as a 0 and the moderate candidate as a 6, then the moderate will win since his average score is 5.3 while the liberal candidate's average score is 4.2.

While score voting offers voters much more freedom, however, voters have little incentive to use it. A voter can maximize his impact by giving each candidate either the highest or the lowest score. Given that this is significantly easier than deciding precisely where on a scale to rank a candidate and, while somewhat exaggerated, does not feel particularly dishonest, we expect that it is what most voters will do. In this case, score voting reduces to approval voting. As stated in the previous section, we think that approval voting is actually a rather good system, so this is not disastrous, but it does defeat the purpose of admitting a variety of scores. Given this, it's hard to justify the extra complexity of score voting compared to approval voting.

Proponents of score voting argue that some voters will provide intermediate scores even if it is not in their best interest and in doing so will cause score voting to outperform approval voting. This is an interesting argument, but since using intermediate scores effectively reduces a voters impact on the election, it is worth considering who is likely to use them. In our judgement it is voters who are less impacted by the outcome of the election, are especially honest, do not know the best voting strategy, or are more moderate in temperament. It is not obvious to us that the impact of these voters ought to be reduced.

Though we do not promote score voting, we do believe it to be one of the best voting systems and would be interested to see how it performed in real world elections.

Score voting is promoted by RangeVoting.org.

Instant Runoff Voting (a.k.a. Choice Voting and Single Transferable Vote)

Instant runoff voting is a kind of transfer election in which voters submit ballots that rank candidates in order of preference and the ballot rankings, not the candidates, determine how votes are transferred. For a single member district, the results of the election are resolved as follows: If any candidate has more than 50% of the votes that candidate is elected. If not, the candidate with the least votes is eliminated and each of his votes is transferred to the next most highly ranked candidate marked on the voter's ballot. The process repeats until some candidate has a majority of the remaining votes (some votes can be wasted if voters fail to rank all of the candidates). For multimember districts, the election threshold is 1/(c+1), where c is the number of candidates and any excess votes for a winning candidate are transferred to the remaining candidates proportionately to the frequency with which they are the next most highly ranked candidate on ballots for the winning candidate.

Our primary objection to instant runoff voting is that it does not eliminate the spoiler effect. Consider again the example election described in our page on transfer elections. In that example, 40% of the population favored the liberal, 39% favored the conservative, and 21% favored the moderate. Assume further that two thirds of moderate voters prefer the conservative to the liberal. In an election using instant runoff voting, no candidate will win outright since none received a majority of the votes. The moderate will then be eliminated because he received the fewest first preference votes. If all moderate voters indicated their second highest preference, then two-thirds of the moderate's votes will be distributed to the conservative and one third to the liberal, resulting in a win by the conservative with 53% of the vote to the liberal's 47%. Thus, the conservative wins even though the moderate would have beaten him in a two candidate race by 61% to 39% (assuming that liberal voters prefer the moderate to the conservative). By giving their top ranking to their preferred candidate, liberal voters caused their least preferred candidate to win. This is the spoiler effect.

Instant runoff voting can malfunction in other ways. Consider an election in which 40% of the population favors the liberal candidate, 28% favors the conservative candidate, 32% favors the moderate candidate, and moderate voters are indifferent to the other two candidates. If all voters vote their true preferences in an election using instant runoff voting, the conservative will have the fewest first preference votes and will therefore be eliminated causing the moderate to win 60% to 40% (assuming that conservative voters prefer the moderate candidate to the liberal candidate). However, if instead liberal voters making up 5% of all voters were to vote for the conservative candidate, the moderate candidate would have the least first preference votes and would therefore be eliminated causing the liberal candidate to win 51% to 49% (or 35% to 33% if the indifferent moderate voters marked no second preferences on their ballots). Thus, liberal voters could cause their preferred candidate to win by voting for their least preferred candidate. Voting systems with this property are said to fail monotonicity. The odds of a real election exhibiting non monotonicity are hotly debated, but we suspect that it is reasonably common. How much this matters in practice is harder to judge.

Instant runoff voting has other issues as well. The process of determining the winner of an instant runoff election can get quite complicated and non-intuitive and so many different combinations of voter preferences are possible that it can be difficult even to accurately summarize the votes received in a given election. Worse, as we saw in the proceeding paragraph, small changes in the vote tally can dramatically change the outcome of the election, which poses a serious risk of election fraud. Combined, these facts seem likely to engender distrust of the system.

Finally, we note that instant runoff voting asks significantly more of voters than most other voting systems. To participate fully, a voter must determine their preference ranking of all candidates. This is quite a lot of effort given even a handful of candidates, but instant runoff elections in multi-member districts can easily have dozens of candidates, with ballots of a length and complexity to match.

In spite of our many objections, we must admit that instant runoff voting is used by many countries, including Australia and Ireland, with apparent success. There are plenty of real-world examples of instant runoff voting performing as it was designed to do and even of it selecting consensus candidates. Nevertheless, we feel that its flaws are too numerous and significant to overlook.

Instant runoff voting is promoted by FairVote.

Condorcet Method

The Condorcet method seeks to identify a candidate that would win in a head-to-head election against all other candidates for an office. Such a candidate is called a Condorcet winner. Choosing the Condorcet winner is often regarded as the best possible outcome of a multi-candidate election.

The Condorcet method is very old, and there are many variations and applications of the method. In modern elections, the method typically uses ranked preference ballots like those used in instant runoff voting, but the preference ballots are used to determine which candidate would win in a two-candidate election for every possible pairing of candidates. If one candidate is predicted to win against all other candidates then that candidate is declared the Condorcet winner. If not, then there exists a cycle of candidates such that each candidate loses against at least one candidate in the cycle. Many different rules for cycle breaking have been formulated, some of them quite complex.

When all voters fill out their ballots according to their true preferences the Condorcet method picks consensus candidates. Returning again to the example election described in our page on transfer elections, imagine a three-way election between a liberal, a conservative, and a moderate candidate which are the first choice of 40%, 39%, and 21% of voters, respectively. Assuming that both liberal and conservative voters prefer the moderate to the candidate at the other end of the political spectrum, the moderate would easily win if either of the other two candidates were excluded. Thus, the moderate is declared the Condorcet winner and wins the election. It is hard to argue that either of the other two candidates would have been a better choice since voters would have soundly rejected them in an election pitting them against (only) the moderate.

Unfortunately, voters may not vote their true preferences for strategic reasons. If the liberal candidate in the example above is the second highest preference for all moderate voters then liberal voters can cause a cycle of winners by marking the the conservative candidate (presumably inaccurately) as their second highest preference. The resulting cycle is conservative beats moderate 79% to 21%, liberal beats conservative 61% to 39%, and moderate beats liberal 60% to 40%. The most popular methods for breaking the cycle will then elect the liberal candidate since he has the weakest loss. Of course, if moderate voters hear of the liberal voter's plan they can foil it by ranking the conservative candidate higher than the liberal candidate, thereby causing the conservative to win. At any rate, the Condorcet winner according the voter's real preferences is not chosen.

The strategy of ranking a candidate lower than a less liked candidate is called burying. As the example illustrates, it is tricky to use and prone to backfiring badly. It also feels dishonest, so it is not clear how many voters can be convinced to do it. Consequently, it is unclear whether burying would be common in real Condorcet elections.

The results of a Condorcet election are unwieldy but far easier to report than for an election using instant runoff voting since it is enough to report the margins for each pairing of candidates. The amount of information required thus goes like the square of the number of candidates instead of increasing exponentially as for instant runoff voting. Nonetheless, the process of determining the winner can get quite complicated. This might easily erode faith in the system.

Finally, like all voting systems based on ranked preference ballots, the Condorcet method asks a lot from voters. In high-profile races such as the presidency the typical voter might know enough about the candidates to rank several of them, but this seems unlikely for races further down the ballot.

While we strongly approve of the goal of selecting a Condorcet winner, we feel that transfer elections offer a better mechanism for doing so. Nevertheless, among voting systems that utilize a ranked preference ballot, we consider the Condorcet method the best.

Abolishing the electoral college

Discussions of government reform often involve proposals that would clearly benefit one or the other of the two major political parties in the United States. In our view this is true of proposals to, for example, abolish the electoral college or make representation in the Senate proportional.

While there are often reasonable arguments to be made for these proposals, in our judgement, they are less about fundamentally changing the way our democracy functions than about shifting the balance of power toward either the Democrats or the Republicans. We make an effort to be non-partisan, and consequently do not promote any such proposals.