Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Apply public choice theory to analyze political decision-making as rational self-interested behavior. Use this skill when the user needs to evaluate government policy failures, rent-seeking costs, voting outcomes, or bureaucratic incentives, especially when the assumption of benevolent government is questionable.
.claude/skills/asgard-ai-platform-grad-public-choice/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-08 | ✗→✓ | ▲ Improved | 18% | 0% |
| case-09 | ✗→✓ | ▲ Improved | 122% | 0% |
| case-11 | ✗→✓ | ▲ Improved | 54% | 0% |
| case-20 | ✗→✓ | ▲ Improved | 137% | 0% |
| case-04 | ✓→✗ | ▼ Worse | 120% | 0% |
Public choice applies economic reasoning — rational self-interest, strategic behavior, and equilibrium analysis — to political decision-making. Politicians, bureaucrats, voters, and lobbyists are modeled as utility maximizers, not benevolent social planners. The theory explains phenomena such as rent-seeking, logrolling, pork-barrel spending, regulatory capture, and the systematic divergence between public interest and political outcomes. Buchanan and Tullock's foundational work treats constitutional rules as the ultimate mechanism design problem.
IRON LAW: Public officials are NOT benevolent social planners — they
respond to incentives just like market participants. Policy outcomes
reflect the preferences of those with political power, not the
preferences of society at large.Step 1 — Identify the Political Market Map the actors: voters, politicians, bureaucrats, interest groups. Specify what each actor maximizes and the constraints they face (electoral cycles, budget rules, information costs).
Step 2 — Apply the Relevant Model Choose from: (a) Median Voter Theorem — in single-dimensional, single-peaked preference space, the median voter's preferred policy wins under majority rule; (b) Rent-seeking model — agents spend real resources to capture a transfer, dissipating up to the full value of the rent; (c) Logrolling / vote trading — minorities trade votes across issues to pass legislation that fails majority support on each issue individually; (d) Bureaucracy model — budget-maximizing bureaus produce beyond efficient output.
Step 3 — Estimate Government Failure Costs Quantify: (a) Tullock rectangle — resources spent on rent-seeking; (b) Allocative distortion from policies that reflect political rather than economic efficiency; (c) X-inefficiency within government agencies lacking competitive pressure. Compare against the market failure the policy aims to correct.
Step 4 — Propose Institutional Remedies Recommend constitutional or institutional design changes: supermajority requirements, sunset clauses, independent agencies, fiscal rules, transparency mandates, or decentralization (Tiebout competition). Evaluate trade-offs between flexibility and constraint.
markdown## Public Choice Analysis: [Policy / Institution] ### Political Actors | Actor | Objective | Key Constraint | |----------------|-----------------------|------------------------| | Voters | | | | Politicians | | | | Bureaucrats | | | | Interest groups | | | ### Model Applied - **Framework**: Median voter / Rent-seeking / Logrolling / Bureaucracy - **Prediction**: [what the model predicts will happen] - **Observed outcome**: [what actually happens — consistent?] ### Government Failure Costs | Cost Category | Estimate / Description | |-----------------------|----------------------| | Rent-seeking expenditure | | | Allocative distortion | | | X-inefficiency | | ### Market Failure vs. Government Failure - **Market failure being addressed**: [externality / public good / monopoly] - **Government failure introduced**: [rent-seeking / capture / inefficiency] - **Net assessment**: [intervention improves welfare? or worsens it?] ### Institutional Recommendations [Specific reforms with rationale]
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | pass→pass | 36,071 | 44,032 | +22% | 1 | 1 | 0% | 5,869 | 5,262 | -10% | 0 | 0 | — |
case-02 | pass→pass | 30,638 | 21,952 | -28% | 1 | 1 | 0% | 4,434 | 4,557 | +3% | 0 | 0 | — |
case-03 | pass→pass | 37,380 | 29,763 | -20% | 1 | 1 | 0% | 5,890 | 5,129 | -13% | 0 | 0 | — |
case-04 | pass→fail | 9,101 | 14,486 | +59% | 1 | 1 | 0% | 1,533 | 3,370 | +120% | 0 | 0 | — |
case-05 | pass→pass | 20,298 | 37,210 | +83% | 1 | 1 | 0% | 3,360 | 6,837 | +103% | 0 | 0 | — |
case-06 | pass→pass | 13,102 | 16,726 | +28% | 1 | 1 | 0% | 2,267 | 4,025 | +78% | 0 | 0 | — |
case-07 | pass→pass | 19,490 | 19,644 | +1% | 1 | 1 | 0% | 3,004 | 3,934 | +31% | 0 | 0 | — |
case-08 | fail→pass | 22,968 | 19,045 | -17% | 1 | 1 | 0% | 3,489 | 4,117 | +18% | 0 | 0 | — |
case-09 | fail→pass | 16,982 | 32,669 | +92% | 1 | 1 | 0% | 2,641 | 5,854 | +122% | 0 | 0 | — |
case-10 | fail→fail | 18,479 | 19,426 | +5% | 1 | 1 | 0% | 2,723 | 4,180 | +54% | 0 | 0 | — |
case-11 | fail→pass | 19,691 | 21,125 | +7% | 1 | 1 | 0% | 3,158 | 4,874 | +54% | 0 | 0 | — |
case-12 | pass→pass | 16,533 | 17,747 | +7% | 1 | 1 | 0% | 2,467 | 3,971 | +61% | 0 | 0 | — |
case-13 | pass→pass | 20,801 | 21,581 | +4% | 1 | 1 | 0% | 3,196 | 4,296 | +34% | 0 | 0 | — |
case-14 | pass→pass | 17,909 | 25,033 | +40% | 1 | 1 | 0% | 2,510 | 4,645 | +85% | 0 | 0 | — |
case-15 | pass→pass | 16,730 | 21,035 | +26% | 1 | 1 | 0% | 2,347 | 4,149 | +77% | 0 | 0 | — |
case-16 | pass→pass | 22,534 | 19,986 | -11% | 1 | 1 | 0% | 3,284 | 4,054 | +23% | 0 | 0 | — |
case-17 | pass→pass | 18,899 | 21,825 | +15% | 1 | 1 | 0% | 2,788 | 4,346 | +56% | 0 | 0 | — |
case-18 | pass→pass | 17,581 | 18,445 | +5% | 1 | 1 | 0% | 2,630 | 3,772 | +43% | 0 | 0 | — |
case-19 | pass→pass | 18,232 | 24,289 | +33% | 1 | 1 | 0% | 2,265 | 4,787 | +111% | 0 | 0 | — |
case-20 | fail→pass | 22,249 | 40,817 | +83% | 1 | 1 | 0% | 2,889 | 6,848 | +137% | 0 | 0 | — |
case-21 | pass→pass | 18,233 | 20,531 | +13% | 1 | 1 | 0% | 2,912 | 4,406 | +51% | 0 | 0 | — |
case-22 | pass→pass | 19,627 | 24,231 | +23% | 1 | 1 | 0% | 2,835 | 4,304 | +52% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 22 cases were attempted. The headline lift of +14 percentage points is the difference between those two pass rates over the 22 comparable cases. 1 case got worse with the skill loaded, and it is included in that figure.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
Other measured skills in the registry, with their headline benchmark lift.