Assume the object of an election is to select one of n options (e.g., to elect one of n candidates for some position). Under the conventional voting system (Plurality Voting), each voter chooses one of n options and the option receiving the most votes is the winner. Under Approval Voting (AV), each voter can select any subset of the n options; the winner, again, is the one who received the most votes.
AV is very simple and not harder to implement than conventional--vote for one candidate--voting. But it is superior in that it allows each voter to express support for any number of candidates, rather than having to choose just one. For example, suppose a voter considers A to be best candidate, but that A is unlikely to win. Suppose another candidate, D, clearly inferior to A in that voter's opinion, but significantly better than all the other candidates, has a good chance of winning. Then, under our present system (vote for one candidate) many, probably most, voters in that situation would vote for D.
Under AV such voters would find it more satisfying to vote for both A and D. The key point is that, while voting for A in addition to D slightly increases the remote possibility of A winning, it does not increase the likelihood of a win by any other candidate. The only candidate that might win as a result of a thousand voters adding votes for A to their ballots is A. It may seem strange to say this, but such a statement would not be valid for some other voting systems, such as Instant-Runoff Voting (IRV).
In an Instant Runoff Voting (IRV) election, the voter ranks all the candidates. A candidate receiving a majority of the first place votes, is the winner. If there is no such candidate then the candidate with the fewest first place votes is eliminated and that candidate's votes are added directly to the totals for the remaining candidates. If this produces a candidate with a majority of the votes we have a winner; otherwise the process is repeated until such a winner is found. Sounds good? Think about it more. Consider the following simple election with candidates A, B, and C, and 9 voters (the example would still be valid if we multiplied the number of voters by 10,000--or any other number). 4 CAB 4 CB 3 BAC -------> 3 BC ----> C wins 1 ACB 1 CB 1 ABC 1 BC The first column above shows the results of the first round of voting. The entry "4 CAB" indicates that 4 voters ranked the candidates as C > A > B. Rows 2, 3, and 4 indicate the rankings assigned by the other 5 voters. (The other 3 possible rankings, such as A > B > C, were not chosen by any of the voters.) Since none of the 3 candidates was the first choice of a majority of the 9 voters, the candidate ranked first by the fewest number of voters (in this case A, ranked first by only 2 voters) is dropped by the IRV algorithm--a bad idea. The resulting modified array is shown in the second column, headed by 4 CB. Since 5 of the 9 voters, a majority, chose C as their first choice, C is declared the winner.
That is too bad, since A dominated both B (first and third rows) and C (second and third rows), and so should have been declared the winner.
Score (also called Range) voting: each voter gives each candidate an integer score from 0 to k (a finite integer). A candidate's score is the sum of the scores.
Approval voting (score voting where k = 1): each voter gives each candidate a 0 or 1. If a thousand voters add votes for A to their ballots, the only candidate that may win as a result, is A. A consequence of this is that, over a period of several elections, A might gain enough support to win.
Consider an election with 3 candidates, A, B, and C. Suppose A and B agree on all the important issues, disagreeing on only a few less important matters, while C's views on all the important issues differ sharply from those of A and B. Then, in a conventional election, we would expect A and B to receive similar numbers of votes. Consider a case where 40% of the voters strongly support C, while 25% support A and 35% support B. Then C would win. Under AV, if the voters vote the way they feel, we would expect that the great majority would vote for both A and B, so that both A and B might receive more votes than C. If all those voting for A or B vote for both of them, then there would be a tie and the choice between A and B would be made by the flip of a coin. More likely a small number of people would vote for just one of the two, so that the A and B tallies would differ slightly--all that is needed to avoid the tie.
In general, AV would encourage people to cast votes for candidates that they think are clearly the best, even tho they appear to have little chance of winning, while also voting for less desirable, but acceptable, and more electable, candidates. If the favored candidates do well, they are more likely to continue to run for office and eventually might succeed in some future election.
Both Range Voting and its simplest version, Approval Voting, give voters a lot of power, are easy to understand, easy to implement, and are not vulnerable to tricky manipulation.
Range (Score) voting allows the voter to support various candidates to different degrees. E.g., under a Range voting system with 4 candidates and 5 possible ratings (0, 1, 2, 3, 4) a voter might give 4 candidates scores of 0, 3, 0, 4.
A fundamental difference between Range Voting and IRV is that IRV requires voters to rank candidates, but not to assign numerical values to them. So if, in an IRV system, a voter ranks 4 candidates A, B, C, D as B > C > D > A, the same voter, if Range Voting is used, might assign to the same candidates one of many scores (e.g., integers ranging from 0 to 4) corresponding to the same ranking. A voter can assign the same score to several candidates.
Two of the many possible votes corresponding to the above example, are: B = 4, C = 3, D = 3, A = 1, and B = 4, C = 1, D = 1, A = 0.
This illustrates how, using Range Voting, the voters can make much finer distinctions among the candidates. The same is true under Approval Voting, to a somewhat lessor extent. In the above example, the same 2 voters, using Approval Voting, might have rated the 4 candidates as B = 1, C = 1, D = 1, A = 0, and B = 1, C = 0, D = 0, A = 0.
Suppose there are 4 candidates and a voter considers only A to merit support. In an IRV election, that voter can't do better than ranking the candidates as A>B>C>D. If another voter thinks that A, B, and C are all very good and that D is unacceptable, that voter can't do better than rank them as A>B>C>D, the same as in the previous case! In AV elections voters can express approval only for A with the vote 1000, while other voters can express approval for all except D, with the vote 1110.
On the other hand, a voter in an IRV election wishing to express maximal support for A, a little less for B, still less for C and none for D can vote A>B>C>D. The same situation can be handled in a Range Voting election (with range 0-4) with a vote: 4321 (or 4310, etc.).
AV can't do as well. The best that can be done is to partition the candidates into 2 sets; acceptable and not acceptable, assigning 1 to each acceptable candidate and 0 to the others. So if only A and C are acceptable the vote would be 1010. []
[1] , "Instant Runoff Voting: Looks Good--But Look Again", , March 26, 2007 "We Can Do Better than Instant Runoff Voting", Stephen Unger, OpEdNews_Op_Eds 08/20/2008
[] , "", ,
[] , "", ,
[] , "", ,
[] , "", ,
[] , "", ,
[] , "", ,
[] , "", ,
[] , "", ,
[] , "", ,
[] , "", ,
[] , "", ,
[] , "", ,
Comments are welcomed and can be sent to me at unger(at)cs(dot)columbia(dot)edu
Don't forget to replace (at) with @ and (dot) with the symbol .
Return to Ends and Means to see other articles that you might find interesting. ***insert statcounter code***