Rob Richie continues to flail

Rob Richie, after Doug Greene pointed out to him on the election law blog our criticism of his contrived pro-IRV/anti-competitors argument, responded as follows on that blog (verbatim, in full):

Very lame, Doug.

What you and the range voting crew are missing is that [my] three criteria are spot-on for gauging the political viability of your proposed systems, whether the oddly quoted person thinks they are legitimate "social choice theory" standards or not.
[end message]

To provide a brief response: There are reasons that the social choice theory community focuses more on certain criteria than others for the purpose of comparing voting systems. (We find it rather odd how Richie in his essay cites, e.g. Tideman's book in an effort to "ground" himself in "scholarship," yet fails to appreciate the most important theme in that book, does not mention the criteria in that book, and invents his own criteria.) Preferred are criteria which:

• Tend to discriminate between commonly proposed voting systems. As opposed to criteria which are always true, or always false (or effectively always) for all voting systems, or for all voting systems ever seriously proposed.
• Tend to sound fundamental to democracy, as opposed to sounding contrived and cooked up to make one system appear better than another.
• Are precisely mathematically definable (so that we can decide objectively whether a system satisfies them).

As examples of widely-accepted such criteria, we mention:

1. Monotonicity: increasing your vote for candidate X, should increase or at least never decrease, X's winning chances. This indeed seems very fundamental in order for your vote to have meaning, and it indeed discriminates between systems such as Coombs, IRV, and Nanson which fail it, versus systems such as Approval, Range, Tideman-ranked-pairs, and Schulze-beatpaths, and Plurality voting, which satisfy it.
2. Participation: providing an honest vote will yield an election result better (or at least, never worse) than not voting at all. ("Better" in the view of that voter.) Again, this indeed seems very fundamental in order for your vote to have meaning, and it indeed discriminates between systems such as Condorcet systems and IRV, which fail it, versus systems such as Approval, Range, Borda, and Woodall-DAC, which satisfy it.
3. Favorite-betrayal criterion: Voting your favorite topmost, is never strategically sub-optimal for a voter. This seems very important for making democracy work. If voters will not vote for their true favorites, what good is democracy? Satisfying FBC automatically gets rid of the "spoiler" effect (if that is regarded as: voters who honestly vote "S is my favorite" cause the election winner to change from A to B.") Richie himself seemed to appreciate this idea when he said

   Does the system meet the common sense principle of rewards for sincere voting? A voter should not likely be punished for voting sincerely under the system's rules.

   And FBC indeed discriminates quite nicely among seriously proposed systems, since, e.g. IRV, Borda, Plurality, and all Condorcet systems fail it, whereas Range Voting and AntiPlurality obey it.
4. Clone-immunity: If an additional "clone" Y of a candidate X enters the race, that should not affect who wins (except up to cloning). This agains seems very fundamental to democracy. If a party can simply sponsor clones of their opponent ("vote-splitting") or of themselves ("teaming") to assure themselves victory, then it isn't "democracy," it is a "joke." Cloneproofness (invented by Richie's hero Tideman) also is very useful for discriminating amongst voting systems. Obeying it include IRV, Range, Schulze-beatpaths, and Tideman-ranked-pairs; disobeying it include Approval, Borda, and Plurality voting.

Of the voting systems mentioned so far, range voting is the only one which meets all four (or even merely the last two) of these criteria. That is not a coincidence. We have proven the theorem that no voting system based on rank-order ballots – and this includes the infinite number of systems nobody has ever invented yet – can satisfy the above conditions. Range voting is thus provably better than every rank-order ballot voting system, in some senses that social choice theorists have long-agreed to be important.

In contrast, of Richie's three criteria, the first two were non-discriminatory since, effectively, every voting system ever seriously proposed obeys them both. (Richie carefully worded them to try to argue that Approval, Range, and Condorcet failed them while IRV obeyed them, but this was not a realistic measure of anything, but rather an exercise in carefully contrived wording; and even as such Richie failed since we showed IRV in fact fails his criterion 1 by a factor of 100 more than Approval voting fails it, and IRV fails his criterion 2 if its threshold is not "one" vote but rather "two" votes.) And his third was not mathematically defined. That is why social choice theorists would not be interested in Richie's three criteria.

But as an exercise in deceptive and misleading – yet clear sounding – writing, we can only express admiration for Richie's essay.

Here is another criticism of Richie's essay, by A.Lomax.

Return to main page