QMV, DOUBLE VOTING: explaination needed

This is taken directly from the referendum commissions site.

Referendum Commission wrote:

Proposed Changes – Council (of Ministers)


Meetings

At present, the Council takes decisions in private and holds some policy debates in public. If the Treaty of Lisbon is ratified, the Council will meet in public when a law is being debated and approved.

Voting

At present, some decisions by the Council must be made unanimously while others are made by Qualified Majority Voting (QMV). Under the QMV system, each Member State’s vote is given a weighting. This is not directly proportionate to the population. The weighting arrangements give smaller countries a greater share than their populations would warrant if a directly proportional arrangement existed.

If ratified, the Treaty would:

• increase the number of areas where QMV will apply;
• change the QMV voting system within the Council from 2014.

Areas to which Qualified Majority Voting applies

At present, QMV applies to decisions on a wide range of issues including agriculture, competition rules, consumer protection, environment and judicial co-operation in civil matters. It is proposed to apply QMV to a number of new areas – these include energy, asylum, immigration, judicial co-operation in criminal matters and sport.

Certain decisions will continue to be made unanimously – they include decisions on defence and taxation. This means that any Member State may veto a proposed change in these areas.

Changes within Qualified Majority Voting

At present, each member state has a weighted vote as follows:


France, Germany, Italy, United Kingdom
29

each
Spain, Poland
27

each
Romania
14

The Netherlands
13

Belgium, Czech Republic, Greece, Hungary, Portugal
12

each
Austria, Bulgaria, Sweden
10

each
Denmark, Finland, Ireland, Lithuania, Slovakia
7

each
Cyprus, Estonia, Latvia, Luxembourg, Slovenia
4

each
Malta
3

Total
345



A Qualified Majority Vote (QMV) at present requires 255 weighted votes – that is almost 74% of the total. This system would remain in place until 2014. If the Treaty is ratified, from 2014, a qualified majority (also to be known as a double majority) will require that decisions must meet two conditions:

a) 55% of the Member States must agree: (for example, while there are 27 Member States, 15 Member States must agree);
b) those Member States supporting the decision must represent 65% of the EU population.

In addition, at least four Member States must be opposed to a decision in order for it to be blocked. This ensures that decisions cannot be blocked by just 3 of the larger Member States acting together.

If there are fewer than 4 Member States opposed to a decision then the qualified majority will be deemed to have been reached even if the population criterion is not met.

Presidency of the Council

The present six month rotating Presidency will continue for all Councils except the Foreign Affairs Council which will be chaired by the High Representative for Foreign Affairs and Security Policy.

http://www.lisbontreaty2008.ie/lisbon_treaty_changes_gov.html

This is my opinion of the new QMV system.

sink wrote: »

New areas will become subject to QMV rather than the unanimous decision making process currently operated. Two areas which remain under unanimous decision are Taxation, Foreign policy & Defence. The Lisbon treaty reorganises the voting system of the Council. The council at the moment has a QMV system which give each state a weighted vote which does not match their population size (e.g. Germany has 16% if the population but only has 8% of the vote, Ireland has 0.8% of the population but has 2% of the vote) and a 75% majority is required. It replaces this with a double QMV system whereby there are two requirements for legislation to pass. First it has to have 55% of member states in agreement, this currently gives each member state an equal 3.75% say. Second those member states in favour must represent at least 65% of the population, so here we have 0.8% weight and Germany has 16% weight. This double QMV the voting system roughly balanced in favour of smaller countries as it did before, in that an individual Irish citizens vote is still slightly more powerful than a German citizens.

You can find my overall opinion on the Lisbon treaty here

That is some nice work. Did you calculate the difference for all permutations of countries voting for and against? If so that would be 27! (factorial) if my leaving cert maths is still with me. That's 1.0888869450418352160768e+28 permutations. How long did it take to calculate?

IRLConor wrote: »

It's 2^27 which is 134.2 million. It takes a little over a minute on my laptop. It could probably be made faster, but that's fast enough for my needs.

If you have a background in computer science, the reasoning goes like this: Represent each yes/no vote with a single bit. There are 27 member countries. That means that any vote can be represented with a 27 bit number. Iterate through all 27 bit numbers and you get all 134.2 million possibilities.

Clever. I did do computer science for 2 years at DIT but found it wasn't for me so I dropped out to spend a year in Canada.

Hope you don't mind me adding your post to my sig.

I have previously been referred to this thread during a discussion on QMV here. ConorIRL last year did some interesting simulations comparing various voting methods. His simulations showed little major difference between country voting powers for QMV under Nice and under Lisbon. I had some somewhat technical doubts (perhaps really too technical for a discussion board like this but anyway! ) about how relevant such simulations really were. I've done a little reading since in the the theory behind voting systems. I just want to attach a health warning here to claims made using such simulations.

ConorIRL from what I can see basically generates enormous numbers of voting configurations and tests and compares different voting systems using these. Essentially he generates a voting configuration by randomly tossing 27 coins corresponding to the different countries. My concern was that there would be quite a strong statistical bias in generating voting configurations in this manner. Statistically this would be a binomial distribution and the number of yes's versus no's would concentrate quite strongly around the middle (27/2 = 13.5) with voting configurations with roughly equal numbers of yes's and no's predominating. The extra rules with QMV would complicate this further but such random sampling of voting configurations would still introduce a certain type of strong uniform sampling bias. I wondered whether voting patterns in real world situations would really be similar to this.

I found that ConorIRL's simulations are quite similar to the measuring of voting systems according to the Banzhaf power index (one of the standard indices for measuring voting power). This measure seems slightly more sophisticated than (if not that different to) Conor's approach. Rather than just counting winning votes it measures critical winning votes for a country (where if that country changed its vote to a no the overall vote would fail and therefore this counts where its vote really matters). The Banzhaf method has previously been used to measure voting power under the QMV systems with results not that dissimilar to Conor's.

However, my doubts regarding the potential bias from generating voting configurations by random sampling (tossing coins) do turn out to be valid. In papers by Gelman, Katz and Bafumi and Mulligan and Hunter it was been shown that when real voting patterns were used the actual voting weights/powers observed were substantially different from that predicted by Banzhaf and similar voting power indices and simulations. For example such voting index theory would suggest that in general the actual voting power of a country or it citizens would be closely related to a formula based on the square root of its voting weight in the voting system. This is the basis of the Penrose method which the Poles unsuccessfully tried and failed to make the voting system under Lisbon. However in these papers it was found in real world voting systems with real voting pattern data that a country's actual voting power did not correspond to what would be predicted or expected from voting indices or such simulations. They found that most often a country's effective voting power was simply proportional to its actual weighting in the voting system.

Therefore one has to be wary when interpreting the results of such simulations. The results they give can incorporate rather subtle biases. I would argue that the seemingly less sophisticated method of comparing voting weights may be just as valid!

turgon wrote: »

@finbar10, those doubts would be legitimate if an element of chance existed in the script but it doesnt as far as I know. ConorIRL evaluates every possible voting pattern (134.2 million possibilities) of the 27 countries so there is no element of chance.

....or maybe not - 27! = 10,888,869,450,418,352,160,768,000,000. Perhaps the man himself could clear it up.

It wouldn't matter. Whether he randomly samples and tests a huge number of votes or just goes through them all from start to finish wouldn't essentially make any difference. I would guess he goes through them all one by one. 2^27 steps wouldn't probably be too big a problem to computers these days.

PopeBuckfastXVI wrote: »

If he goes through them all, his program will indeed accurately answer the question, what is the percentage difference between the number of 'wins' for Ireland (or any other country) under Nice, and under Lisbon.

Yes, I'm sure that's pretty much what his program does. And that it gives that answer. I'm just questioning the usefulness of that answer. His answer assumes that all possible voting configurations are equally likely. In the real world that's just not the case.

To give an example. In IRLConor's simulations above one could actually safely remove the rule that at least four countries must vote against a proposal to block it. This obviously helps curb the power of the bigger countries. Removing this rule in the simulation would have almost no effect on the results generated. Why?

There are 2^27 possible voting configurations. The four country rule only makes any difference to voting configurations with 3 or less countries.
There would be only
Binomial(27,0) + Binomial(27,1) + Binomial(27,2) + Binomial(27,3)
= 27/(!0!27!) + 27!/(1!26!) + 27!/(2!25!) + 27!/(3!24!)
=1+27+351+2925 = 3304 of these.
But they only form 3304/2^27 = 0.000025
of all possible votes.

If one assumes all voting configurations are equally likely then one would have to assume that a situation where only 3 or less countries would vote against a proposal would almost never occur (or only 0.000025 of the time). These votes would only form a negligible part of the simulation results above. So IRLConor's simulations wouldn't capture this important characteristic of the Lisbon QMV system.

But the 4 country minimum rule is there. Obviously someone must think it is important and that otherwise small groups of large countries blocking votes would happen more often than just 0.000025 of the time. That's just a simple example of the bias inherent in treating all voting configurations equally. When real voting pattern data is used the results are often significantly different from such simulations.

the four country rule actually already is
built into Nice QMV. It needs at least four
countries to block anything.

My understanding was the blocking in Nice was only 3 countries.

Nice article

Article 205 (*)
1. Save as otherwise provided in this Treaty, the Council shall act by a majority of its Members.
2. Where the Council is required to act by a qualified majority, the votes of its Members shall be
weighted as follows:
Belgium 5
Denmark 3
Germany 10
Greece 5
Spain 8
France 10
Ireland 3
Italy 10
Luxembourg 2
Netherlands 5
Austria 4
Portugal 5
Finland 3
Sweden 4
United Kingdom 10.
For their adoption, acts of the Council shall require at least:
— 62 votes in favour where this Treaty requires them to be adopted on a proposal from the
Commission,
— 62 votes in favour, cast by at least 10 members, in other cases.
3. Abstentions by Members present in person or represented shall not prevent the adoption by the
Council of acts which require unanimity.

searched the term block and nothing came back so I assume there is no blocking minority. But I'll keep looking.


Seems it was my mistake.

I don't think that is any different to Lisbon.
In Lisbon one needs at least four countries to block.
But four countries alone may not be enough. Their
combined population needs to exceed 35%.

No that only applies if not all member states are involved in the vote. Mostly because depeneding on the number of members taking part 4 might be extremely difficult to achieve (say 5 states are voting on something) So instead its 35% of the population of the member states taking part.

3. As from 1 November 2014 and subject to the provisions laid down in the Protocol on
transitional provisions, in cases where, under the Treaties, not all the members of the Council
participate in voting, a qualified majority shall be defined as follows:
(a) A qualified majority shall be defined as at least 55 % of the members of the Council representing
the participating Member States, comprising at least 65 % of the population of these States.
A blocking minority must include at least the minimum number of Council members representing
more than 35 % of the population of the participating Member States, plus one member, failing
which the qualified majority shall be deemed attained;

and my misunderstanding of the blocking minority was not that it was a system only applying to 3 states, but that the 3 biggest states of the EU alone have the voting weight to block any legaslative in the EU. Lisbon raises it to 4 as a requirement regardless of population.