This is something that comes up from time to time. I think we're pretty far off on doing this _in general_ ... the important piece that's missing is the sensor group for it. It would be easy enough to track who gets what bills from an ATM, but much beyond that it gets very dicey; think of who, even if you use a lot of credit/debit cards, you do cash transactions with: small transactions, small merchants, friends, informal things.
Don't forget: technically, all credit card and debit card transactions are "tracked" -- and yet some would say up to 2% of all transactions are fraudulent. So how good of tracking is that? I've seen numbers that indicate that far more "cash" is stolen through the electronic networks (i.e., cards) than in the paper money world.
I did notice that my bank has changed their ATM hardware/software to allow you to deposit checks and cash without counting it; the cash gets counted by the machine, and the checks get imaged and/or MICR scanned. It also takes much longer to do the transaction, which is a general trend in outsourcing customer service: get the customer to do the work. So it's probably cheaper to process deposits because they don't have to pay someone to do it. And in this case, probably more accurate because of the techno-assist.
But anyway: I don't think you have to worry about cash transactions being tracked that accurately for a while. It's still pretty easy to make untracked phone calls, for instance; and if they haven't closed that hole in the name of terra, cash is pretty far down on the list.
For an informal project, see http://wheresgeorge.com/ ... I hear it's been used to make a model of movement that's being applied to disease migration.
/jordan