[EM] stratified renormalisation for elections

[Howard Swerdfeger]

raphfrk at netscape.net wrote:
>  It is often possible effect who actually votes elections by selecting when the voting occurs. For example, the general election in Ireland is being held on Thursday.
>  
>  However, university exams are being held at the moment. This means that students are much less likely to vote. Also, even the fact that the elections are on Thursday is likely to suppress that demographic as students tend no to re-register when they move to go to university. They stay registered in their home constituency. 
>  
>  The elections are sometimes held on Saturday so that students have the option of going home for the weekend to vote (some would go home for the weekend anyway).
>  
>  Similar tricks can be used for lots of demographics.
>  
>  One solution would be to do stratified renormalisation. This is where you split the population up into sub-groups. If a sub-group is over-represented by the number of voters, the vote of each member of that group would be reduced in weight. Similarly, if the demographic is under-represented, it would have its votes increased in weight.
>  
>  This would mean that differential turn-out would be corrected. If a demographic is 20% of the population, it will count for 20% of the votes.
>  
>  This is already done for regions. A region/district gets seats on the basis of population not on the basis of number of votes.
>  
>  The initial split could be based on population. The voters could be split into 4 equal groups starting at 18 and going upwards. Votes from each group would be coloured slightly differently, if one group is over/under-represented, then renormalisation could be applied.
>  
>  Obviously, the characteristics for the groups would have to be clearly defined.

This is an interesting idea, one I never thought of before.
but there are many many issues you need to address before you could even 
think of implementing something like this

The obvious problem is how you define the groups
Some of these groups may be obvious

Age, sex, race, income, current net worth, marital status, education 
level, Many sub groups for each common disability

But some groups that I would think would be important to have would be 
hard to prove that you are a member of that demographic. Two that spring 
to mind are religion and sexual orientation.
Both of these demographics are self identified (ie. can be faked), but 
both could affect voting patterns.

Worse some demographic category that you missed may effect voter turn out.
Example :
Bed ridden 80+ year olds with dementia, are much less likely to vote 
then an 80 year old in good health. and it is likely that there voting 
patters would be different.
as the bed ridden one might favour more money to permanent long term 
care, and the other more money to medicare, and homecare.