Difference between revisions of "Arrow's Theorem"
(typo) |
|||
| Line 4: | Line 4: | ||
# ''Unrestricted domain.'' This condition means that the entire set of leaders (or whatever is being voted upon) must be decided by the voting system. It would clearly be both fair and satisfactory if the election, instead of determining the next [[President]], were to determine a Cabinet of 350 million people, a system known technically as [[direct democracy]] or [[referendum]]. This is not the situation being considered when we talk about Arrow's Theorem. | # ''Unrestricted domain.'' This condition means that the entire set of leaders (or whatever is being voted upon) must be decided by the voting system. It would clearly be both fair and satisfactory if the election, instead of determining the next [[President]], were to determine a Cabinet of 350 million people, a system known technically as [[direct democracy]] or [[referendum]]. This is not the situation being considered when we talk about Arrow's Theorem. | ||
| − | # ''Non-imposition.'' Non-imposition means | + | # ''Non-imposition.'' Non-imposition means that every outcome must be achievable by some combination of votes. |
| − | # ''Non-dictatorship.'' Clearly this condition is satisfied as well. | + | # ''Non-dictatorship.'' This condition means that the result must not just follow the preference of one individual. Clearly this condition is satisfied as well. |
| − | # ''Positive association,'' also known as the ''Peter principle.'' Now we come to one of the trickier principles. The criterion of positive association means that if one voter | + | # ''Positive association,'' also known as the ''Peter principle'' or ''monotonicity.'' Now we come to one of the trickier principles. The criterion of positive association means that if one voter changes his vote to rank a candidate higher, the new result cannot rank that candidate lower than in the original result. |
# ''Independence of irrelevant alternatives'' (known jocularly as the ''Nader corollary''). This condition means that a third-party candidate (such as [[Ralph Nader]] in the 2000 primaries) should not be able to "swing" the election to one side or another. In particular, a third party should be unable to draw votes away from the primary candidates, as Nader did to [[Al Gore]] in 2000, or as [[Pat Buchanan]] did to [[George W. Bush|Bush]]. This condition is perhaps the most relevant to the state of elections in the [[United States]] today. | # ''Independence of irrelevant alternatives'' (known jocularly as the ''Nader corollary''). This condition means that a third-party candidate (such as [[Ralph Nader]] in the 2000 primaries) should not be able to "swing" the election to one side or another. In particular, a third party should be unable to draw votes away from the primary candidates, as Nader did to [[Al Gore]] in 2000, or as [[Pat Buchanan]] did to [[George W. Bush|Bush]]. This condition is perhaps the most relevant to the state of elections in the [[United States]] today. | ||
| − | + | Another version of the theorem replaces monotonicity with a condition known as ''unanimity'' or ''Pareto efficiency''. This condition requires that, if every voter prefers candidate A to candidate B, candidate A must also be preferred to candidate B in the result. | |
| − | + | In 1951, Kenneth Arrow proved that no system of voting, however so fair, can possibly satisfy all five of these constraints. For example, the [[Electoral College]] system used in the United States satisfies all of them except the last two (and in fact even the first, unrestricted domain, is questionable, since [[federal judge]]s are not elected by the people, but rather ''appointed'' by the bench). | |
| + | The 1952 Presidential election demonstrated beyond doubt the correctness of Arrow's vision; incumbent [[Dwight D. Eisenhower]] beat his "egghead" Democratic challenger [[Adlai Stevenson]] in the largest landslide in human history. Nevertheless, the [[West Virginia]] Democratic primary was won by [[Averell Harriman]], who had never before held any elected office. This anomaly led to closer investigation of Arrow's results. For his work on modern voting methodology, Arrow was awarded the 1972 [[Nobel Prize]] in [[Economics]]. | ||
| + | |||
| + | ==External Links== | ||
| + | *http://www.nationmaster.com/encyclopedia/Arrow%27s-paradox | ||
[[Category:Mathematics]] | [[Category:Mathematics]] | ||
Revision as of 02:27, November 6, 2008
Arrow's Theorem, also known as Arrow's impossibility theorem or Arrow's paradox, is a seemingly paradoxical result in the mathematical science of voting theory. The theorem was invented by the economist Kenneth Arrow in 1951, and applied the following year in the 1952 Presidential Election.
The theorem states, quite simply, that no representative voting system can simultaneously be "fair" and satisfy the following five conditions:
- Unrestricted domain. This condition means that the entire set of leaders (or whatever is being voted upon) must be decided by the voting system. It would clearly be both fair and satisfactory if the election, instead of determining the next President, were to determine a Cabinet of 350 million people, a system known technically as direct democracy or referendum. This is not the situation being considered when we talk about Arrow's Theorem.
- Non-imposition. Non-imposition means that every outcome must be achievable by some combination of votes.
- Non-dictatorship. This condition means that the result must not just follow the preference of one individual. Clearly this condition is satisfied as well.
- Positive association, also known as the Peter principle or monotonicity. Now we come to one of the trickier principles. The criterion of positive association means that if one voter changes his vote to rank a candidate higher, the new result cannot rank that candidate lower than in the original result.
- Independence of irrelevant alternatives (known jocularly as the Nader corollary). This condition means that a third-party candidate (such as Ralph Nader in the 2000 primaries) should not be able to "swing" the election to one side or another. In particular, a third party should be unable to draw votes away from the primary candidates, as Nader did to Al Gore in 2000, or as Pat Buchanan did to Bush. This condition is perhaps the most relevant to the state of elections in the United States today.
Another version of the theorem replaces monotonicity with a condition known as unanimity or Pareto efficiency. This condition requires that, if every voter prefers candidate A to candidate B, candidate A must also be preferred to candidate B in the result.
In 1951, Kenneth Arrow proved that no system of voting, however so fair, can possibly satisfy all five of these constraints. For example, the Electoral College system used in the United States satisfies all of them except the last two (and in fact even the first, unrestricted domain, is questionable, since federal judges are not elected by the people, but rather appointed by the bench).
The 1952 Presidential election demonstrated beyond doubt the correctness of Arrow's vision; incumbent Dwight D. Eisenhower beat his "egghead" Democratic challenger Adlai Stevenson in the largest landslide in human history. Nevertheless, the West Virginia Democratic primary was won by Averell Harriman, who had never before held any elected office. This anomaly led to closer investigation of Arrow's results. For his work on modern voting methodology, Arrow was awarded the 1972 Nobel Prize in Economics.