Can we automate
policy evaluation?

AI may soon be capable of producing rigorous economic research. If that happens, policy evaluation could scale dramatically: highlighting what works, what fails, and what harms, far faster than human researchers alone.

We want to find out whether an autonomous system can generate, replicate, and revise empirical policy research, with everything made public.

This is an experiment in building reliable AI research systems. For a global overview, click here.

2,878Ideas+414 this week

1000Papers+10 this week

18k+Matches

Last updated: April 16, 2026

Top APE papers

Loading top APE papers...

Most policies — probably millions of them globally — are never rigorously evaluated. Data is plenty but there aren't enough researchers. Could AI help? We genuinely don't know. So we're running an experiment. An AI system attempts to produce economics research at scale, , using publicly available data. Will any be good? How would we even know? Ideally, we'd want PhDs or editors of top journals to evaluate all of them. But they are busy. We run an automated tournament evaluating the papers against human benchmarks from top journals. This could help triage. Get to a "you know it when you see it" moment, faster. Most importantly, everything is : papers, code, data, failures. The more people look, the faster mistakes get caught. And we want feedback! In fact, the core thesis is that recursive self-improvement is possible and can be enhanced by human feedback. The next milestone: generate a 1000 papers, evaluate, and share lessons in a report. Can policy evaluation be automated? Or is hallucinated slop unavoidable? Let's find out!

Ranking Metrics

▼

Rank

Position based on conservative rating (lower bound of estimated skill)

48h

Change in rank over the past 48 hours

Estimated skill level — higher means the paper wins more often

Uncertainty — decreases as the paper plays more matches

Cons.

Conservative rating (μ − 3σ) — papers need consistent wins to rank highly (TrueSkill)

Elo

Familiar chess-style rating for comparison (1500 = average)

Number of head-to-head matches played

Reviewed

Human expert review — Yes for journal papers (peer-reviewed), No for APE papers (pending)

Average Elo by Source

Elo Rating Distribution

Conservative Rating Distribution

TrueSkill Conservative: Top-3 Papers

Papers & Matches

Tap chart to expand

Data as of 2026-04-17 01:09:13 CET

Review Status

▼

Benchmark:

Automated Review of APE Papers

Code review:

Replication:

Referee Reviews:

Swipe to see more columns

Rank ↑	48hRank change over the last 48 hours.	Paper ↕	μEstimated skill rating (μ). Higher values indicate better research quality based on pairwise comparisons. ↕	σUncertainty (σ). Lower values mean higher confidence in the rating. ↕	Cons.Conservative Rating (μ - 3σ), adjusted for integrity penalties. Used for ranking. ↕	EloElo rating. Standard chess-like rating where 400 points difference = 90% win probability. ↕	MPMatches Played. Valid head-to-head comparisons, excluding annulled matches against papers flagged with severe issues during automated code review. ↕	Status✅ Peer reviewed · 🔎 Awaiting review · 🧐 Issues detected · 🚫 Critical errors
1	—	Dynamics of the Long Term Housing Yield: Evidence from Natural Experiments AER	40.1	1.6	35.2	2104	190	✅
2	—	The Value of Clean Water: Experimental Evidence from Rural India AER	37.2	1.5	32.8	1989	171	✅
3	—	Vertical Integration and Cream Skimming of Profitable Referrals: The Case of Hospital-Owned Skilled Nursing Facilities AEJ: Policy	35.5	1.2	31.9	1920	185	✅
4	—	Are Complementary Policies Substitutes? Evidence from R&D Subsidies in the UK AEJ: Policy	35.8	1.4	31.6	1932	158	✅
5	—	Plata y Plomo: How Higher Wages Expose Politicians to Criminal Violence AEJ: Policy	35.3	1.3	31.6	1914	155	✅
6	▲3	Punishing Financial Crimes: The Impact of Prison Sentences on Defendants and Their Colleagues AEJ: Policy	35.0	1.2	31.4	1900	165	✅
7	—	Estimating the Economic Value of Zoning Reform AEJ: Policy	34.9	1.2	31.4	1896	195	✅
8	▲3	Why is Workplace Sexual Harassment Underreported? The Value of Outside Options Amid the Threat of Retaliation AER	34.9	1.2	31.3	1897	188	✅
9	▼3	Immigration, Innovation, and Growth AER	34.8	1.2	31.3	1892	191	✅
10	—	Abundance from Abroad: Migrant Income and Long-Run Economic Development AER	34.8	1.2	31.2	1893	185	✅
11	▼3	Invisible Wounds: How Mental Disability Benefits Shape Veteran Well-Being AEJ: Policy	34.6	1.2	31.0	1883	151	✅
12	—	Flood Risk Mapping and the Distributional Impacts of Climate Information AEJ: Policy	33.9	1.2	30.3	1856	168	✅
13	—	The Welfare Effects of Eligibility Expansions: Theory and Evidence from SNAP AEJ: Policy	33.8	1.2	30.1	1851	169	✅
14	▲1	Hacked to Pieces? The Effects of Ransomware Attacks on Hospitals and Patients AEJ: Policy	33.3	1.1	29.9	1830	166	✅
16	—	Short- and Long-Term Effects of Universal Preschool: Evidence from the Arab Population in Israel AEJ: Policy	33.0	1.1	29.6	1820	164	✅
17	—	Adjustable Product Attributes, Indirect Network Effects, And Subsidy Design: The Case of Electric Vehicles AEJ: Policy	32.6	1.1	29.4	1803	180	✅
18	—	Collective Bargaining Rights, Policing, and Civilian Deaths AEJ: Policy	32.6	1.1	29.3	1802	180	✅
19	▲1	Can You Erase the Mark of a Criminal Record? Labor Market Impacts of Criminal Record Remediation AEJ: Policy	32.6	1.1	29.2	1803	171	✅
20	▼1	Trade Protection Along Supply Chains AEJ: Policy	32.4	1.1	29.2	1795	173	✅
21	▲2	Cooking, Health, and Daily Exposure to Pollution Spikes AEJ: Policy	32.3	1.1	29.1	1792	179	✅
22	▲2	The Pass-Through of Retail Crime AEJ: Policy	32.2	1.1	28.8	1788	194	✅
25	▲2	Market Power and Capital Constraints AER	31.5	1.1	28.4	1761	199	✅
26	▲2	A Matter of Time? Measuring Effects of Public Schooling Expansions on Families AEJ: Policy	31.3	1.1	28.1	1750	177	✅
27	▲2	Harnessing Deductions to Increase Tax Compliance and Formalization AEJ: Policy	31.1	1.1	27.9	1742	186	✅
28	▲3	The Unintended Consequences of Academic Leniency AEJ: Policy	31.0	1.1	27.8	1741	203	✅
29	▲4	Childhood Health Shocks and the Intergenerational Transmission of Inequality AEJ: Policy	30.7	1.0	27.6	1727	175	✅
32	▲4	The Option Value of Municipal Liquidity AEJ: Policy	30.3	1.0	27.1	1710	168	✅
34	▲6	Zero-Sum Thinking and the Roots of US Political Differences AER	29.5	1.0	26.6	1681	188	✅
35	▲4	Perils of the Paperwork: The Impact of Information and Application Assistance on Social Benefit Take-Up in India AEJ: Policy	29.3	0.9	26.5	1672	213	✅
36	▲6	The Effect of Deactivating Facebook and Instagram on Users' Emotional State AEJ: Policy	29.4	1.0	26.4	1677	180	✅
37	▲13	Temporary Layoffs, Loss-of-Recall and Cyclical Unemployment Dynamics AER	29.2	1.0	26.2	1669	192	✅
38	▲8	Female Leaders and Intrahousehold Dynamics: Evidence from State Elections in India AEJ: Policy	29.1	1.0	26.1	1663	198	✅
40	▲5	Tax and Occupancy of Business Properties: Evidence from UK Business Rate Reliefs AEJ: Policy	28.9	1.0	26.0	1657	186	✅
44	▲12	Cross-State Strategic Voting AEJ: Policy	28.1	1.0	25.2	1625	184	✅
46	▲11	Black Lives: The High Cost of Segregation AEJ: Policy	27.6	1.0	24.7	1605	201	✅
47	▲14	Polling Place Location and the Costs of Voting AEJ: Policy	27.3	0.9	24.5	1593	223	✅
61	▲17	China's Nationwide CO2 Emissions Trading System: A General Equilibrium Assessment AEJ: Policy	26.1	0.9	23.4	1545	235	✅
67	▲33	Hurricanes, Climate Change Policies and Electoral Accountability AEJ: Policy	25.8	0.9	23.1	1533	217	✅
74	▲21	Polluted IPOs AEJ: Policy	25.4	0.9	22.7	1515	238	✅
104	▲53	Work From Home and the Office Real Estate Apocalypse AER	24.3	0.8	21.8	1471	236	✅
117	▲33	The Value of School Social Climate Information: Evidence from Chicago Housing Transactions AEJ: Policy	24.0	0.9	21.4	1461	233	✅
153	▲55	State Recreational Cannabis Laws and Racial Disparities in the Criminal Legal System AEJ: Policy	22.6	0.8	20.0	1403	256	✅
421	▲47	Fiscal Procyclicality in Commodity Exporting Countries: How Much Does It Pour and Why? AEJ: Policy	16.7	0.9	14.1	1169	288	✅

Total tokens used for tournament (excludes paper generation tokens): 1,476,570,316