Quadratic Voting

Thinking Tools5.2 Selecting the Right Decision Tools

Originated by Aanund Hylland and Richard Zeckhauser (A Mechanism for Selecting Public Goods When Preferences Must Be Elicited, 1980) and developed by E. Glen Weyl (Quadratic Vote Buying, 2012), with Steven P. Lalley (2018) and Eric A. Posner (Radical Markets, 2018).

Gives each voter a budget of credits to spread across issues, where casting several votes on one issue costs the square of their number, so strength of preference counts.

Each extra vote on one item costs more than the last, so strong views count without taking over

Twelve managers set next year's priorities, each with 100 credits. You are one of them. Votes on an item cost the square of their number: one vote costs 1 credit, two cost 4, six cost 36. Spend your credits for or against the five items: the bars show each item's total with your votes added to the other eleven managers' (the black line), and the next vote on any item always costs more than the one before. Illustrative.

3
−3
4
6
0
Credits you have spent, of 100
70
Credits left (below zero means over budget)
30
Cost of one more vote on the customer portal
13 credits
Customer portal's total, all twelve managers
11

Total votes, yours includedThe other eleven managers

−2002017−9131111New dataplatformOffice moveTwo moreengineersCustomer portalTraining budget−2002017−9131111New dataplatformOfficemoveTwo moreengineersCustomerportalTrainingbudget

An example

The twelve managers of a logistics software company had five candidates for next year's discretionary budget: a new data platform, an office move, two more engineers, a customer portal, and a training program. In past years a show of hands had produced a list nobody felt strongly about.

The COO gave each manager 100 credits and one rule: casting several votes on one item costs the square of their number. The head of customer success, who heard every week from clients asking for a portal, spent 36 credits on six votes for it and spread the rest thinly. Others did the same with their own priorities. The office move, which a few people opposed firmly and nobody wanted much, came out negative. The data platform ranked first and the engineers second; the portal, carried by a few intense votes, drew level with training. The managers accepted the list because each had been able to show how much they cared, and nobody could buy the result alone.

Why it works

One person, one vote counts only which side people are on. It cannot tell a shrug from a conviction, so a mild majority beats an intense minority every time, and the result says little about where the value lies. Letting people pile up votes fixes that, but with a flat price the most determined voter simply spends everything on one item.

Quadratic pricing sits between the two. Each extra vote on an item costs more than the last (the first costs 1, the sixth 11), so a voter spends on an item until how much they care no longer justifies the rising price. That is the point where a vote reflects the strength of a preference. Strong views count, a voter who cares about one thing can show it, and the square makes it expensive for any one voter to dominate.

The thinking behind it

Aanund Hylland Aanund Hylland, a Norwegian economist and decision theorist, wrote A Mechanism for Selecting Public Goods When Preferences Must Be Elicited (1980) with Richard Zeckhauser while he was taking his doctorate at Harvard's Kennedy School. The paper gave each voter a budget of influence points whose votes grow with the square root of the points spent, an early form of the pricing Glen Weyl later developed and named quadratic voting.

Richard Zeckhauser Richard Zeckhauser, a Harvard Kennedy School economist known for work on decisions under uncertainty and on status quo bias, gave with Hylland an incentive argument for making the cost of votes rise with the square of the number cast, and proposed a procedure for choosing public goods on that basis. Lalley and Weyl credit Groves and Ledyard (1977) with first suggesting quadratic pricing for collective decisions; Hylland and Zeckhauser were the first to justify it uniquely.

E. Glen Weyl E. Glen Weyl, an economist at Microsoft Research, coined quadratic voting: voters buy votes from a budget at a cost equal to the square of the number cast on an issue, so they can show how strongly they care as well as which side they take. He set it out with Steven P. Lalley in AEA Papers and Proceedings (2018) and with Eric A. Posner in Radical Markets (2018). Critics note that with real money it favors the wealthy and invites collusion, which is why most uses give every voter the same budget of credits.

How to run it

  1. Give every voter the same budget. Credits, never money; 100 is easy to reason about and leaves room for a strong vote.
  2. List the items to decide. Five to fifteen items, each described in a sentence so everyone votes on the same thing.
  3. Allow votes for and against. Opposition counts too, and it costs the same as support.
  4. Charge the square of the votes on each item. A ballot or a spreadsheet that shows the running cost keeps voters inside their budget.
  5. Vote privately. Open ballots invite trading and pressure, which the mechanism cannot price.
  6. Total the votes per item and rank. Report each item's net votes, and how many voters cast at least one vote on it.
  7. Treat the ranking as advice. Use it to set the agenda and the order, then decide on the merits with the result in view.

Where it goes wrong

The common failure is collusion. Two voters who agree to split their strong votes across each other's items get more influence than the price intends. Keep ballots private, and watch for blocs whose votes mirror each other.

A second failure is poorly defined items. Voters spend credits on what they imagine an item to be, and an item that means different things to different people collects votes nobody can act on. Write each item precisely before the vote. The third is using it where everyone cares about the same few items equally. With no difference in intensity, quadratic voting gives the same answer as a show of hands at more effort; save it for lists where people care unevenly.

When to use something else

  • Nominal Group Technique. When the group first needs to generate and discuss the items before anyone ranks them.
  • Weighted Decision Matrix. When the choice should rest on agreed criteria and scores more than on how strongly each person feels.
  • Mechanism Design. When the task is designing the rules of a whole process so people's incentives line up with the outcome wanted.

Where it sits in SOLVE eX

Go to the source

  • A Mechanism for Selecting Public Goods When Preferences Must Be Elicited, Kennedy School of Government discussion paper, 1980
  • Quadratic Vote Buying, working paper, 2012
  • Quadratic Voting: How Mechanism Design Can Radicalize Democracy, AEA Papers and Proceedings, 2018
  • Radical Markets, 2018

Back to the catalog