Applied Futarchy

[psztorc]

IMHO, the "weak point" of Futarchy is the 'goal metric'.

In corporate governance ("Fire the CEO" as advocated by Hanson), the stock price is the goal metric. This is convenient because the spot price of a corporation's stock is itself resistant to manipulation (the dividend payments, and market for loanable funds, should keep it in lockstep with the market value of the assets of the corporation). A parallel metric for a non-profit might be some metric set by its board of trustees, or by a nation's voters, but this metric would probably not be resistant to manipulation. In the business world, you can't fake what you've produced: it's either there (to take home to your family, or sell) or it's not. Advocates of Futarchy mention GDP, leisure, education, inequality, and health.

GDP (PPP) per capita is a highly reliable indicator of national welfare....today. However, if we adopted GDP/Pop as our 'goal metric', the reliability might plummet. Futarchy would demand that we adopt the policies which increase pc GDP as measured by X, so it is likely easiest for a tyrant to go straight for X, and through that, seize control of the nation's policies and the nation itself.

Most other ideas seem open to Sybil attacks. The best idea I have so far is:
1] Have every Bitcoin (or CashCoin) account sign a message answering: "From 0 to 20, how satisfied are you with your government?"
2] Average the satisfaction numbers, with the BTC balance as weights.

We can have several different metrics ("...how satisfied are you with respect to 'educational availability', or "your health").

The most obvious problem is that: This is a plutocracy (the government is doing whatever the wealthy want [some consider this a "pro" instead of a "con"]). A second obvious problem is that this is self-report (you can have a terrible educational outcome, but lie or exaggerate your experience), which is why I bite the bullet and ask something directly subjective ("how satisfied are you").

To solve problem 1, we might try something like this: http://www.ted.com/talks/david_bismark_e_voting_without_fraud?language=en
[1] n = nations_population (quantity of ballots)
[2] Cryptographic set up of 'the vote' (on this day, this many keypairs, this voter-eligibility).
[3] Go to election HQ on election day, randomly select your sheet with a keypair.
[4] Go home to Namecoin, use your keypair to extract the "ballot form", use it to cast your votes.
[5] Check the outcomes after the fact.

I have no idea how to solve the second (self-report) problem.

Does anyone have any ideas in general, about how to reliably measure a nation's success?

[zack]

Everyone measures success differently, we all care about different things.
Each person should either use a popular metric, or combine several into a custom metric of what they care about most.

Politicians look at popular metrics, and attempt to make decisions that offend the least people.
It is an optimization thing. Everything you do makes some people less happy, and other people more happy.

[psztorc]

Ok, but consider each metric in your multitude of metric. How do you reliably measure any of them? How do you keep that measurement reliable after the PM about it is made?

[zack]

Well, we can only ask questions that are easily verifiable, otherwise all the voters will say it is a bad question, and it gets thrown out.

So we are somewhat limited in what metrics we can use. You cannot ask how long the average person will sleep next year since statisticians are so incentivized to cheat, but you could ask for the market price of black beans.

[dhcmrlchtdi]

Well, we can only ask questions that are easily verifiable, otherwise all the voters will say it is a bad question, and it gets thrown out.

So we are somewhat limited in what metrics we can use. You cannot ask how long the average person will sleep next year since statisticians are so incentivized to cheat, but you could ask for the market price of black beans.

If limiting Futarchy to "black bean optimization" is the best you can do, I think you just shouldn't say anything at all. Writing is for people who actually have thoughts.

[psztorc]

Ideally, there would be some kind of "econ experiment" that regular people could verify. For example, the unemployment rate is calculated by the CPS of BLS, which doesn't really seem that difficult for a few independent people to do ( http://www.bls.gov/opub/hom/homch1_itc.htm ).

Life expectancy and infant mortality (whatever those are worth) are pretty auditable.

GDP leaves me scratching my head.

[zack]

Researchers in political economy have viewed the trade-off between military and consumer spending as a useful predictor of election success
http://en.wikipedia.org/wiki/Guns_versus_butter_model

The price of things like black beans and butter and homes is very important to predict, and very hard to fake.

There is already popular futures markets for food stuffs, because people care a lot about how much it will cost in the future:
http://www.cmegroup.com/trading/agricultural/grain-and-oilseed/rough-rice.html
http://www.barchart.com/commodityfutures/Orange_Juice_Futures/OJ?search=OJ*
http://www.cmegroup.com/trading/agricultural/grain-and-oilseed/soybean.html

Correlating price predictions with political decision predictions is sufficient for some aspects of futarchy to appear.

[psztorc]

For GDP, the BEA drills down to MSA and by industry and year: http://www.bea.gov/iTable/iTable.cfm?reqid=70&step=1&isuri=1&acrdn=2#reqid=70&step=1&isuri=1

...which might be audit-able enough. Perhaps we can get more replicable results once the ball gets rolling.

[psztorc]

Robin Hanson comments on a brand new paper arguing for more effective 'quadratic voting'.

Very interesting. Of course, while giving some direction, it does not help with the practical challenge (which it demands): "how to get everyone registered in the system exactly once".

That would probably need to be centrally managed.