What Matters is a guest opinion column written by a different MIT alumnus or alumna each month. The views expressed in What Matters are entirely those of the author and do not necessarily represent the views of the Association or of MIT. For previous columns, please see the archives.

Range Voting: the best way to select a winner?

by Warren D. Smith '84

The world is approaching some major crises such as the end of cheap oil, the exhaustion of important "fossil water" reserves, climate change, overpopulation, and nuclear and bioweapon proliferation. So it is more important than ever that the world make the right decisions.

So: what is the "decision-making algorithm" for the world? With the USA now the sole superpower, the closest simple approximation to the answer unfortunately is "the USA's appalling voting system." It features

• 98% year-ahead predictability of all important races (by far the largest among the nations usually called "democratic"), plus about one-third of statehouse races are uncontested and congressmen are more likely to die in office than be defeated,
• Rampant gerrymandering which looks to the eye worse than in 90% of democracies,
• An elected partisan official in charge of supervising elections in 33 out of 50 states (plus in many other states he is appointed by the governor, which may be even worse) – such as J.Kenneth Blackwell who supervised the Bush-Kerry 2004 election in Ohio while serving simultaneously as the Bush-Cheney campaign co-chair there, and who now is supervising his own election for the 2006 Ohio governorship,
• Massive two-party domination (over 99.5%) assuring that voters have at most two choices even in the few percent of times when the race is unpredictable, and causing other views to be nearly completely blocked out of both power and the media. E.g. all the USA's minor parties in 2004 stood against NAFTA, WTO, high budget imbalance, Iraq war, "PATRIOT act," Bush's medicare prescription drugs-payment plan, and the "war on drugs" – an amazing unity considering their very different philosophies. However, since both major-party candidates (Bush/Cheney and Kerry/Edwards) supported, cosponsored, and/or had voted for all these things, USA voters had no effective way to express their feelings on any. (And polls showed the public was roughly 50-50 split on each one of them.)
• The "electoral college" inspires presidents to cater only to a very small number of "swing states" while ignoring the rest.
• Many states have highly restrictive ballot-access laws designed to ensure 2-party (or 1-party) domination. E.g. Georgia, in the heyday of single-party domination in 1943, enacted a law preventing any independent or new-party candidate from getting on ballot for any office unless they submit a petition signed by a huge 5% of the registered voters. (And all petition signers are subject to subpoena to determine if they actually signed!) Result: since 1943, zero independent or third-party candidates have ever managed even to get on the ballot for a Georgia U.S. congress seat.
• Most states (and the nation) lack citizen "initiative and referendum" making it essentially impossible to pass laws beneficial to citizens but not presently-in-power legislators, such as anti-gerrymandering laws, freer ballot-access laws, or freedom-of-information laws.

So the questions then become: how can we change this system, and what should we change it to? There are many obvious improvements one can immediately suggest, but unfortunately their obviousness in no way implies that it will actually be possible to make them happen.

In 2005 I founded the Center for Range Voting () which is currently suggesting two less-obvious improvement ideas. One is a simple district-drawing algorithm called the shortest splitline algorithm. (Basically: find the shortest line that splits the state into two parts with the right population ratios, then continue recursively.) This produces far nicer district maps for far less cost than present methods, and more importantly is completely unbiased. This stops gerrymandering, period. It should be feasible to enact this via citizen initiative and referendum in the states that allow that (but that costs about $1 million per state).

The second, more important, idea is range voting (RV). Essentially, Joe Voter rates every candidate on an 0-9 scale (e.g. Gore=9, Nader=9, Browne=5, Bush=0 in the USA 2000 election) as his vote. (You can also intentionally leave blank a candidate's score, such as Phillips.) The candidate with the greatest average score wins. Essentially the same system is used by Olympics judges to select gymnastics gold medalists and by teachers in academia to select valedictorians.

Here are some properties of range voting:

1. Voter provides information – and quantitative information – about all candidates he wants to. (In the present plurality system you provide qualitative info only about one candidate – least possible.)
2. That information tends to be comparatively honest. With plurality, honestly voting for a third-party candidate is strategically stupid, so (polls showed) about 90% of Nader-favorers dishonestly voted for somebody else in 2000. But with range voting, there never is any incentive to "betray your favorite" by scoring him below top. In many systems based on rank-order ballots like "instant runoff" and "Borda" voting, not only does Joe Voter provide less information (cannot express different intensities of preference) but also simple example-elections show that Joe by honestly ranking his favorite Nader top, can (and is quite likely to) cause both Nader and Joe's second-favorite to lose. That kind of disincentive to honesty is exactly why every instant-runoff country is 2-party dominated – same thing that happens ("Duverger's law") under plurality voting. (Some US third parties have endorsed instant runoff, suicidally not recognizing this fact.)
3. Many voting systems are manipulable by candidate "cloning" – sponsoring a "clone" of some candidate can cause both clones to lose thanks to "vote splitting." Range voting is immune to cloning. No more bitter emnity between alike candidates.
4. Range voting can be used on every voting machine in the USA, right now. No modification needed, no reprogramming needed. (But that's not true for some other systems like Borda, Condorcet, and instant runoff. The trick is we can transform an RV election into "several plurality elections.") This makes RV comparatively painless to adopt.

If RV is enacted, then a lot of problems such as 2-party domination, gerrymandering, 98% predictability, rubberstamping approval of unqualified party hacks as judges and agency heads, and media blackout of non-major-party views, all will automatically cure themselves at varying rates. Meanwhile a direct attack on these problems simply hits a stone wall.

There is strong evidence from Monte Carlo computer simulation studies that RV is actually the best among all commonly-proposed single-winner voting systems. Specifically, there is a statistical yardstick called "Bayesian Regret." The BR value of any voting system can be measured by simulating a zillion elections with artificial voters and candidates. The BR of a voting system depends not only on what voting system it is, but also on how many voters there are, how many candidates there are, how the simulation makes each voter "feel" about each candidate, how "ignorant" the voters are, how "strategic" they are versus how "honest" they are, etc. So there are many "knobs" on the side of the simulator you can "turn." I performed a comprehensive study of all the most common voting system proposals, for each one measuring the BR value in each of 720 different knob-settings. Range voting had as good or better BR than every competitor at every one of the 720 knob settings. Six years later I also later did a new simulation of "three candidate left-middle-right scenarios" and again RV clearly came out superior to all commonly proposed competitors, and electrical engineer Peter L. Bielawski tells me that he too has written his own simulator which reached the same conclusion.

Not only that, BR, since it is a quantitative measurement of voting system quality, can be roughly translated into both "human happiness" as well as tangible units such as "dollars" or "lives lost." From that we estimate that adopting RV would cause an quantitative improvement in the lot of humanity comparable to the improvement previously caused by replacing "elect random winner" with "plurality voting democracy." It is also possible to come to similar conclusions in other ways (e.g. via economic statistics).

Sounds great – but why would the USA's power-holding two parties want to end their fun by enacting RV? Several reasons:

• They'll still have a huge advantage in resources, organization, reputation, and experience over all the minor parties in all near-term races, but RV will eliminate the "spoiler" problem which is one of the biggest threats to current office-holders. Long term, RV will allow the gradual rise of third parties which then will offer an "insurance policy" to office-holders who lose a future primary race in their own party (also one of the biggest threats they face) because they could then run third-party and have a reasonable chance to win without risking being a "spoiler" and hence a "traitor."
• If the Democrats or Republicans enact RV in their state primaries, especially Iowa 2008, then RV's quality advantage over Plurality will benefit that party's presidential election chances in the real election, without hurting them in any way. Plus it will make them look like "reformers," generate free publicity for both them, voting reform, and the Iowa politicians that push for it and otherwise would remain unknown on the national stage, and be appreciated by Iowan voters who now benefit from increased expressivity. I.e. every major power group wants it, with no group motivated against it. And all it takes is a party internal rule change, not even a law-change. That means this can really happen.

But it cannot happen if the CRV organization stays small, underfunded, and unknown. The CRV needs a heck of lot of help. Please check the CRV web site and click "join," "endorse," and/or "ways to help." Contact the CRV by posting to the RV bulletin board http://groups.yahoo.com/group/RangeVoting or emailing CRV members (see the CRV's online directory).

About the Author

Warren D. Smith '84 got a double SB from MIT in physics & math, then got a PhD in applied math from Princeton. He is the author of an 85% written but unpublished book on "mathematics and democracy."

Return to main page