Banzhaf power index

The Banzhaf power index, named after John F. Banzhaf III (originally invented by Lionel Penrose in 1946 and sometimes called Penrose–Banzhaf index), is a power index defined by the probability of changing an outcome of a vote where voting rights are not necessarily equally divided among the voters or shareholders.

Computer Model of the Banzhaf power index (The Wolfram Demonstrations Project)

To calculate the power of a voter using the Banzhaf index, list all the winning coalitions, then count the critical voters. A critical voter is a voter who, if he changed his vote from yes to no, would cause the measure to fail. A voter's power is measured as the fraction of all swing votes that he could cast.

The index is also known as the Banzhaf–Coleman index after James Samuel Coleman or the Penrose–Banzhaf index.

Examples

Voting Game

Simple voting game

A simple voting game, taken from Game Theory and Strategy by Phillip D. Straffin:

[6; 4, 3, 2, 1]

The numbers in the brackets mean a measure requires 6 votes to pass, and voter A can cast four votes, B three votes, C two, and D one. The winning groups, with underlined swing voters, are as follows:

AB, AC, ABC, ABD, ACD, BCD, ABCD

There are 12 total swing votes, so by the Banzhaf index, power is divided thus:

A = 5/12, B = 3/12, C = 3/12, D = 1/12

U.S. Electoral College

Consider the U.S. Electoral College. Each state has more or less power than the next state. There are a total of 538 electoral votes. A majority vote is considered 270 votes. The Banzhaf power index would be a mathematical representation of how likely a single state would be able to swing the vote. For a state such as California, which is allocated 55 electoral votes, they would be more likely to swing the vote than a state such as Montana, which only has 3 electoral votes.

The United States is having a presidential election between a Republican and a Democrat. For simplicity, suppose that only three states are participating: California (55 electoral votes), Texas (34 electoral votes), and New York (31 electoral votes).

The possible outcomes of the election are:

California (55) Texas (34) New York (31) R votes D votes States that could swing the vote
R R R 120 0 none
R R D 89 31 California (D would win 86-34), Texas (D would win 65-55)
R D R 86 34 California (D would win 89-31), New York (D would win 65-55)
R D D 55 65 Texas (R would win 89-31), New York (R would win 86-34)
D R R 65 55 Texas (D would win 89-31), New York (D would win 86-34)
D R D 34 86 California (R would win 89-31), New York (R would win 65-55)
D D R 31 89 California (R would win 86-34), Texas (R would win 65-55)
D D D 0 120 none

The Banzhaf power index of a state is the proportion of the possible outcomes in which that state could swing the election. In this example, all three states have the same index: 4/12 or 1/3.

However, if New York is replaced by Ohio, with only 20 electoral votes, the situation changes dramatically.

California (55) Texas (34) Ohio (20) R votes D votes States that could swing the vote
R R R 109 0 California (D would win 55-54)
R R D 89 20 California (D would win 75-34)
R D R 75 34 California (D would win 89-20)
R D D 55 54 California (D would win 109-0)
D R R 54 55 California (R would win 109-0)
D R D 34 75 California (R would win 89-20)
D D R 20 89 California (R would win 75-34)
D D D 0 109 California (R would win 55-54)

In this example, the Banzhaf index gives California 1 and the other states 0, since California alone has more than half the votes.

Cartel game

Five companies (A, B, C, D, E) sign an agreement for the creation of a monopoly. The size of the market is X = 54 millions units per year (i.e. petroleum barrels) for a monopoly. The maximum production capacity of these companies is A = 44, B = 32, C = 20, D = 8 and E = 4 millions of units per year. Therefore, there is a set of coalitions able to provide the 54 millions of units necessary for the monopoly, and a set of coalitions unable to provide that number. In each of the sufficient coalitions we may have necessary members (for the coalition to provide the required production) and unnecessary members (underlined in the table below). Even when one of these unnecessary members goes out of the sufficient coalition that coalition is able to provide the required production. However, when one necessary member leaves, the sufficient coalition becomes insufficient. The monopoly's profit to be distributed among the coalition's members is 100 millions of dollars per year.

Sufficient coalitions ABCDE, ABCD, ABCE, ABDE, ACDE, ABC, ABD, ABE, ACD, ACE, BCDE, BCD, BCE, ADE, AB and AC
Insufficient coalitions CDE, BDE, AD, AE, BC, BD, BE, CD, CE, DE, A, B, C, D and E

The Penrose–Banzhaf index may be applied to the calculation of the Shapley value, which provides a basis for a distribution of the profit for each player in the game in proportion to the number of sufficient coalitions in which that player is necessary. The player A is necessary for 10 of the 16 sufficient coalitions, B is necessary for 6, C also for 6, D for 2 and E for 2. Therefore, A is necessary in 38.5% of the total cases (26 = 10 + 6 + 6 + 2 + 2, so 10/26 = 0.385), B in 23.1%, C in 23.1%, D in 7.7% and E in 7.7% (these are the Banshaf indexes for each company). The distribution of the 100 millions of monopoly profits under the Shapley value's criterion has to follow those proportions.

History

What is known today as the Banzhaf power index was originally introduced by Penrose (1946) and went largely forgotten. It was reinvented by Banzhaf (1965), but it had to be reinvented once more by Coleman (1971) before it became part of the mainstream literature.

Banzhaf wanted to prove objectively that the Nassau County Board's voting system was unfair. As given in Game Theory and Strategy, votes were allocated as follows:

This is 30 total votes, and a simple majority of 16 votes was required for a measure to pass.

In Banzhaf's notation, [Hempstead #1, Hempstead #2, North Hempstead, Oyster Bay, Glen Cove, Long Beach] are A-F in [16; 9, 9, 7, 3, 1, 1]

There are 32 winning coalitions, and 48 swing votes:

AB AC BC ABC ABD ABE ABF ACD ACE ACF BCD BCE BCF ABCD ABCE ABCF ABDE ABDF ABEF ACDE ACDF ACEF BCDE BCDF BCEF ABCDE ABCDF ABCEF ABDEF ACDEF BCDEF ABCDEF

The Banzhaf index gives these values:

Banzhaf argued that a voting arrangement that gives 0% of the power to 16% of the population is unfair, and sued the board.

Today, the Banzhaf power index is an accepted way to measure voting power, along with the alternative Shapley–Shubik power index.

However, Banzhaf's analysis has been critiqued as treating votes like coin-flips, and an empirical model of voting rather than a random voting model as used by Banzhaf brings different results (Gelman & Katz 2002).

See also

References

External links