[EM] On IRV compressability

Jonathan Lundell jlundell at pobox.com
Wed Mar 15 16:09:55 PST 2006

At 10:26 PM +0000 3/15/06, Gervase Lam wrote:
>  > Date: Tue, 14 Mar 2006 20:52:56 -0800
>>  From: Brian Olson
>>  Subject: [EM] On IRV compressability
>>  How summable or not is an IRV election?
>Here is the one attempt I remember reading at making IRV summable:

At 8:52 PM -0800 3/14/06, Brian Olson wrote:
>Take that data point for what you will. If we assume computer 
>calculation it doesn't matter. One representation of the total vote 
>set is 446082 bytes uncompressed and 10353 bytes after bzip2.

There were roughly 12,000,0000 ballots case in California in the 2004 
presidential election. If it required 100 bytes to represent each 
presidential vote (and that's got to be way high), we'd have 1.2GB. 
If Brian's roughly 50:1 compression ratio held, we'd have only 24MB 
of data to communicate to the vote counters in DC (or wherever).

So even in a very large election, Brian's right: it doesn't matter.

Recounts, now....
/Jonathan Lundell.

More information about the Election-Methods mailing list