[Search for users] [Overall Top Noters] [List of all Conferences] [Download this site]

Conference rusure::math

Title:Mathematics at DEC
Moderator:RUSURE::EDP
Created:Mon Feb 03 1986
Last Modified:Fri Jun 06 1997
Last Successful Update:Fri Jun 06 1997
Number of topics:2083
Total number of notes:14613

1679.0. "Another paradoxical voting problem" by --UnknownUser-- () Thu Oct 22 1992 15:09

T.RTitleUserPersonal
Name
DateLines
1679.1HERON::BUCHANANThe was not found.Thu Oct 22 1992 15:2130
	In France, regional elections are run in the following way.   There
are a number of parties, P; voters, V; and seats, S.   Each party consists
of an ordered (arbitrarily large) list of candidates.   Each voter votes for
exactly 1 party.

	k = ceiling(V/S) denotes the number of votes required to win a seat.  
Let V_i be the number of votes for party i.   s_i = floor(V_i/k) seats are
immediately allocated to party i.   The top i candidates from the party i list
are hence elected.   party i has r_i = V_i - k*(V_i/k) votes remaining, being
considered to have "spent" the rest.   There are r = S - sum(i)(s_i) seats
remaining to be allocated.   They are allocated to the r parties who have the
greatest r_i. (ie, the parties effectively bid with their remaining votes for
the remaining seats.)   Each party can thus win at most 1 bonus seat this way,
on top of the s_i seats they won earlier.   Ties in competition for the bonus
seats are resolved by lot.

	Let's make it concrete with an example.   Suppose 4 parties are
fighting for 6 seats, and have received 7,8,9,10 votes respectively.   It will
cost ceiling(34/6) = 6 votes to buy each seat.   This means that each party
gets a seat, and they have 1,2,3,4 votes over.   Two of the parties can buy
the remaining two seats with 3&4 votes respectively.

	Suppose that the votes have been cast.   Can you construct a situation
where increasing by 1 the total number of seats would have the effect of 
*reducing* the number of seats that a certain party, the Unlucky Party, would
win?   Can you construct a situation where increasing by 1 the number of
votes won by a certain party, the Very Unlucky Party, would *reduce* the number
of seats the VUP would win?

Andrew.
1679.2CSC32::D_DERAMODan D'Eramo, Customer Support CenterWed Oct 28 1992 23:2236
>	Suppose that the votes have been cast.   Can you construct a situation
>where increasing by 1 the total number of seats would have the effect of 
>*reducing* the number of seats that a certain party, the Unlucky Party, would
>win?   Can you construct a situation where increasing by 1 the number of
>votes won by a certain party, the Very Unlucky Party, would *reduce* the number
>of seats the VUP would win?
    
    The first situation can happen.
    
    Let there be three parties with V=111 voters and S=11 seats.
    Call the parties A (the unlucky party), B, and C.  Let the
    vote count be A - 23, B - 44, C - 44. (That does total V=111.)

    There will be ceiling(V/S) = ceiling(111/11) = ceiling(10+) = 11
    votes needed per seat, so the seats parcel out as
    
    	A - 2 remainder 1
    	B - 4 remainder 0
    	C - 4 remainder 0
    
    That's only 10 of the S=11 seats, so A gets a third seat, and the
    seat distribution ends up A - 3, B - 4, C - 4.
    
    Now add a seat to get S' = 12 seats.  The new "price" per seat
    is now ceiling(V/S') = ceiling(111/12) = ceiling(9+) = 10 votes
    per seat.  So the same vote count A - 23, B - 44, C - 44 yields
    
    	A - 2 remainder 3
    	B - 4 remainder 4
    	C - 4 remainder 4
    
    The full seats only total 10, but the 2 left over now go to B and C,
    with a final seat distribution of A - 2, B - 5, C - 5.
    
    Dan
    
1679.3CSC32::D_DERAMODan D'Eramo, Customer Support CenterWed Oct 28 1992 23:3430
>	Suppose that the votes have been cast.   Can you construct a situation
>where increasing by 1 the total number of seats would have the effect of 
>*reducing* the number of seats that a certain party, the Unlucky Party, would
>win?   Can you construct a situation where increasing by 1 the number of
>votes won by a certain party, the Very Unlucky Party, would *reduce* the number
>of seats the VUP would win?
    
    I can interpret the second situation two different ways.  One
    interpretation has two completely different scenarios with the
    same V (number of voters) and S (number of seats).  The first
    scenario, party A gets m votes yielding n seats, and in the second
    scenario, party A gets m+1 votes yielding n-1 seats.  Other aspects
    of the scenarios, such as the number of parties and how many votes
    they get, can differ.
    
    The second interpretation is to fix a scenario with V voters,
    S seats, and P parties with vote counts v1 + v2 + ... + VP = V.
    Now change this by adding one new voter who votes for party 1,
    resulting in V+1 voters, S seats, and P parties with vote counts
    (v1+1) + v2 + ... + vP = V+1.  Can party 1 end up with fewer
    seats?
    
    Which of the two interpretation did you have in mind?  An example
    for the first would be rather easy: 41 voters, 2 seats, 2 parties
    with vote totals A-11, B-30 (resulting in one seat each).  If party
    B splits and loses a vote to A, then A-12, B1-15, B2-14 results in
    A getting one more vote but no seats (one fewer seat).
    
    Dan
    
1679.4clarificationHERON::BUCHANANThe was not found.Thu Oct 29 1992 07:4411
>    The second interpretation is to fix a scenario with V voters,
>    S seats, and P parties with vote counts v1 + v2 + ... + VP = V.
>    Now change this by adding one new voter who votes for party 1,
>    resulting in V+1 voters, S seats, and P parties with vote counts
>    (v1+1) + v2 + ... + vP = V+1.  Can party 1 end up with fewer
>    seats?

	This is the interpretation that I had in mind.   I agree, the other
one is trivial.

Andrew.
1679.5YesSUBURB::STRANGEWAYSVision without Plan = DreamTue Jul 20 1993 13:4145
    
    Yes, it is possible for a Very Unlucky Party to exist.
    
    Let the VUP have V votes out of a total of nS for some integral n.
    Then seats = V/k.
    
    If we increase VUP's votes by 1, we have (V+1) votes, but the divisor
    is now k+1.
    
    If V/k - (V+1)/(k+1) > 1 then we have lost seats.
    
    This happens if V > k^2 + 2*k.
    
    For example, suppose 25 seats and 6 parties.
    
    Case (A): 275 votes cast, as follows:
    
    v_i = (250,5,5,5,5,5)
    
    k = ceiling(275/25) = 11
    
    s_i = (22,0,0,0,0,0)
    
    r_i = (8,5,5,5,5,5)
    
    seats = (23,1,1,0,0,0)
    
    Now suppose the first party gets an extra vote:
    
    v_i = (251,5,5,5,5,5)
    
    k = ceiling(276/25) = 12
    
    s_i = (20,0,0,0,0,0)
    
    r_i = (11,5,5,5,5,5)
    
    seats = (21,1,1,1,1,0)
    
    And the party is two seats worse off.
    
    Interestingly, if you scale this example up to (say) v_i =
    (1000,20,20,20,20,20) and seats = 100 then you end up with vacant seats
    that cannot be allocated within the terms of the problem.