Ungameable Outcome Incentives
Tie an institution's money to the one outcome it cannot fake, and it redesigns itself
- Difficulty
- Moderate
- Time to result
- ~months to results
- Steps
- 5
- Confidence
- 74%
Lonsdale's institutional-reform model is to stop funding inputs and instead tie an organization's money to the single outcome it exists to produce, chosen so the operator cannot game it. When you rate technical schools by the average salaries of their graduates over three years, the schools spontaneously ask what skills to teach and which employers to partner with, because their funding now depends on it. He applies the same pattern to student loans, only lending for majors that on average let students repay, and to prisons, measuring probation and parole staff on rehabilitation and reduced recidivism. The power of the method is that it does not prescribe the method: it sets the right outcome and lets the system's own incentives drive the redesign, which he claims more than doubled Texas vocational-graduate salaries over a decade.
Origin
Extracted from Lonsdale's policy work through his Cicero organization, which operates in roughly twenty states to make government 'smarter, dynamic, and less stupid.' He offers vocational schools, student loans, and prison reform as live applications of the same incentive-design pattern.
Core principles
- 01Systems reorganize around whatever you actually measure and fund
- 02Pick an outcome the operator genuinely cannot game
- 03Funding-by-input rewards the same broken behavior; funding-by-outcome forces change
- 04The right metric makes both the left and right want the same result
- 05Let the incentive do the redesign instead of dictating the method
How to run it
- 1
Define the real outcome
Identify the true end result the institution exists to produce, such as graduate earnings or reduced recidivism.
Watch out Confusing an input like enrollment for the outcome keeps the system broken.
- 2
Pick an ungameable metric
Choose a measure of that outcome the operator genuinely cannot manipulate, like average salaries of graduates.
Pro tip Outcomes tied to third parties, such as what employers pay, are far harder to fake.
- 3
Fund in proportion to the metric
Allocate money based on the outcome metric rather than headcount, requests, or historical budgets.
Watch out Giving a failing operator 'more money' with no outcome tie changes nothing.
- 4
Average over time
Compute the metric over a multi-year window so operators cannot game a single cohort or quarter.
Pro tip A three-year salary average smooths noise and blocks short-term manipulation.
- 5
Let operators find the how
Do not dictate curricula or methods; the funded outcome pushes operators to discover their own path.
Pro tip Schools start asking which skills and partners produce results once pay depends on it.
In the wild
Instead of handing failing vocational schools more money, Lonsdale describes funding roughly twenty-seven Texas technical schools in proportion to the three-year average salaries of their graduates. Because that number cannot be gamed, schools began asking which skills to teach and which businesses to partner with. He says salaries coming out of these schools more than doubled over the following decade.
→ Graduate salaries more than doubled without prescribing any curriculum.
Lonsdale argues probation, parole, and prison staff should be measured partly on rehabilitation: teaching skills, preparing inmates in the year before release, and reducing the odds they return. He frames this as a bipartisan fix, since both left and right should want released prisoners to succeed, and credits reformed incentive structures with improving outcomes in the communities affected.
→ Aligned incentives around lower recidivism instead of mutual misery.
Common mistakes
Throwing more money at inputs
Increasing a failing institution's budget without tying it to an outcome rewards the same broken behavior.
Choosing a gameable metric
If the operator can manipulate the measure directly, the incentive drives cosmetic gaming rather than real improvement.
Is it for you?
Best for
Leaders and reformers redesigning schools, agencies, or programs whose incentives are misaligned.
Not ideal for
Outcomes that are genuinely unmeasurable or where the best proxy is easily gamed.
From the transcript
“We're gonna only give them money in proportion to the salaries of the students coming out. That's not something they can game.”
“When you start giving them all their money tied to that, the school starts saying, wait a second, what skills do we need to teach.”
From the episode
How To Win The War Of The Future - Joe Lonsdale - #934
Joe Lonsdale