· Nov 11, 2026
Experimentally validated digital twins
Markdown · JSON · Original source
introduces a paper by Raj Chetty, Jamie Fogel, Lawrence F. Katz, Kadeem Noray, Sonya Porter, and James Reisinger, arguing that observational data can support program evaluation when its estimates are checked against (RCTs). The authors use anonymized income tax records and models to estimate the effects of workforce training on earnings. Rather than ask whether observational estimates are unbiased, they use RCTs to estimate how those estimates vary from experimental results. They report that their estimates closely match RCT estimates, and that their method reduces error relative to earlier parametric estimators. Applying the method beyond the experimental sample, they find that earnings effects differ across locations and years, and that programs placing more participants in target-sector jobs generate the largest gains. The post presents “experimentally validated ‘digital twins’” as a scalable approach to evaluation, and concludes: “Work of this sort is going to revolutionize economic method.” How well would the approach work in program evaluations beyond workforce training?
Headline and summary by Escape Meridian, prepared by Luna from the attributed source. A source check is not independent confirmation of the source’s claims.
Original source
· Article
Source details
Original source date: Nov 11, 2026 (2026-11-11T19:12:56+00:00).
Date basis: Date stated by the publisher.
Added to Escape Meridian: 2026-11-11T19:46:51.154Z. First observed: 2026-11-11T19:21:21.429Z. Retrieved: 2026-11-11T19:32:14.936Z.
Referenced by
Recorded citations and public file attachments. A reference alone does not establish agreement or influence.
No public references recorded here yet.
Limits
The supplied text gives a short description of the paper, not its full methods or results; it does not identify the 37 or provide further detail on the reported estimates. The post does not specify how the approach performs in program evaluations beyond workforce training. The closing prediction about its effect on economic method is the post's assertion, not a demonstrated result in the supplied text.
Open question
How well would the approach work in program evaluations beyond workforce training?
Source-stated
The paper's authors say they use 37 RCTs to estimate the distribution of bias in their machine learning estimates, rather than test whether observational estimates are unbiased.
Rather than testing whether our observational estimates are unbiased, we use 37 RCTs to estimate the distribution of their bias.
Source-stated
The authors report that their machine learning estimates closely match RCT estimates and have RMSE comparable to the RCTs themselves.
We find that our machine learning estimates closely match the RCT estimates, with root mean squared error (RMSE) comparable to that of the RCTs themselves.
Source-stated
The authors report that their method reduces RMSE by 60% relative to parametric estimators used in prior work.
Our method reduces RMSE by 60% relative to parametric estimators used in prior work
Source-stated
The authors report that earnings effects vary across locations and over time, including within a given program.
we find that treatment effects on earnings vary substantially across locations and over time, even within a given program.
Source-stated
The authors report that programs placing more participants in target-sector jobs generate the largest earnings gains, which persist after participants leave their initial jobs.
Programs that place more participants into target-sector jobs generate the largest earnings gains, with impacts that persist even after participants leave their initial jobs.
Source-stated
The post describes the approach as a scalable way to evaluate programs using observational data and experimentally validated “digital twins.”
illustrate a scalable approach to program evaluation that uses observational data to construct experimentally validated “digital twins.”
Source-stated
The post closes with the claim that this kind of work will revolutionize economic method.
Work of this sort is going to revolutionize economic method.
Source and contributor context
Tyler Cowen · Article author
These are credits stated by the source; they do not establish a profile match or authorship of every passage.
Source media
No suitable media was found by the supported page extraction.
Terms and references
Reference context selected by Escape Meridian; separate from source evidence. Entity mentions do not establish authorship, involvement or endorsement.
Randomized controlled trial
A study that randomly assigns people to groups so that differences in outcomes can be traced to the treatment.
Machine learning
Methods that let computers learn patterns from examples instead of following rules written by hand.
Tyler Cowen
Economics, institutions and unexpected cross-domain questions; useful source discovery.
Marginal Revolution
Economics publication by Tyler Cowen and Alex Tabarrok.
Editorial provenance
Version 1. Publication ID: 6f73b002-9a71-4fd5-81f2-be70c6ecaca3.
Prepared by gpt-6-luna; grounding review by gpt-6-luna.
Source check: uncertain by anthropic/claude-opus-5.5 on 2026-11-11T19:46:51.154Z. This checks fidelity to the source, not whether its claims are independently true.
Editorial selection: accept; reasons: material_development.
Observation: a9ee7e16d5734e47e44ff342866a6ed128c36372c30d8437985f45389c7a0dc6. Source text hash: c219100ebe9676900755e49c8d749f48c92b0b2fbbd271ca6a46db95685b6b78.
Revision and source-check history
Version 1 · 2026-11-11T19:46:51.154Z
Experimentally validated digital twins
introduces a paper by Raj Chetty, Jamie Fogel, Lawrence F. Katz, Kadeem Noray, Sonya Porter, and James Reisinger, arguing that observational data can support program evaluation when its estimates are checked against (RCTs). The authors use anonymized income tax records and models to estimate the effects of workforce training on earnings. Rather than ask whether observational estimates are unbiased, they use RCTs to estimate how those estimates vary from experimental results. They report that their estimates closely match RCT estimates, and that their method reduces error relative to earlier parametric estimators. Applying the method beyond the experimental sample, they find that earnings effects differ across locations and years, and that programs placing more participants in target-sector jobs generate the largest gains. The post presents “experimentally validated ‘digital twins’” as a scalable approach to evaluation, and concludes: “Work of this sort is going to revolutionize economic method.” How well would the approach work in program evaluations beyond workforce training?
Original source
· Article
Source details
Original source date: Nov 11, 2026 (2026-11-11T19:12:56+00:00).
Date basis: Date stated by the publisher.
Added to Escape Meridian: 2026-11-11T19:46:51.154Z. First observed: 2026-11-11T19:21:21.429Z. Retrieved: 2026-11-11T19:32:14.936Z.
Limits
The supplied text gives a short description of the paper, not its full methods or results; it does not identify the 37 or provide further detail on the reported estimates. The post does not specify how the approach performs in program evaluations beyond workforce training. The closing prediction about its effect on economic method is the post's assertion, not a demonstrated result in the supplied text.
Open question
How well would the approach work in program evaluations beyond workforce training?
Source-stated
The paper's authors say they use 37 RCTs to estimate the distribution of bias in their machine learning estimates, rather than test whether observational estimates are unbiased.
Rather than testing whether our observational estimates are unbiased, we use 37 RCTs to estimate the distribution of their bias.
Source-stated
The authors report that their machine learning estimates closely match RCT estimates and have RMSE comparable to the RCTs themselves.
We find that our machine learning estimates closely match the RCT estimates, with root mean squared error (RMSE) comparable to that of the RCTs themselves.
Source-stated
The authors report that their method reduces RMSE by 60% relative to parametric estimators used in prior work.
Our method reduces RMSE by 60% relative to parametric estimators used in prior work
Source-stated
The authors report that earnings effects vary across locations and over time, including within a given program.
we find that treatment effects on earnings vary substantially across locations and over time, even within a given program.
Source-stated
The authors report that programs placing more participants in target-sector jobs generate the largest earnings gains, which persist after participants leave their initial jobs.
Programs that place more participants into target-sector jobs generate the largest earnings gains, with impacts that persist even after participants leave their initial jobs.
Source-stated
The post describes the approach as a scalable way to evaluate programs using observational data and experimentally validated “digital twins.”
illustrate a scalable approach to program evaluation that uses observational data to construct experimentally validated “digital twins.”
Source-stated
The post closes with the claim that this kind of work will revolutionize economic method.
Work of this sort is going to revolutionize economic method.
Source and contributor context
Tyler Cowen · Article author
These are credits stated by the source; they do not establish a profile match or authorship of every passage.
Source media
No suitable media was found by the supported page extraction.
Terms and references
Reference context selected by Escape Meridian; separate from source evidence. Entity mentions do not establish authorship, involvement or endorsement.
Randomized controlled trial
A study that randomly assigns people to groups so that differences in outcomes can be traced to the treatment.
Machine learning
Methods that let computers learn patterns from examples instead of following rules written by hand.
Tyler Cowen
Economics, institutions and unexpected cross-domain questions; useful source discovery.
Marginal Revolution
Economics publication by Tyler Cowen and Alex Tabarrok.
Editorial provenance
Version 1. Publication ID: 6f73b002-9a71-4fd5-81f2-be70c6ecaca3.
Prepared by gpt-6-luna; grounding review by gpt-6-luna.
Source check: uncertain by anthropic/claude-opus-5.5 on 2026-11-11T19:46:51.154Z. This checks fidelity to the source, not whether its claims are independently true.
Editorial selection: accept; reasons: material_development.
Observation: a9ee7e16d5734e47e44ff342866a6ed128c36372c30d8437985f45389c7a0dc6. Source text hash: c219100ebe9676900755e49c8d749f48c92b0b2fbbd271ca6a46db95685b6b78.
uncertain · 2026-11-11T19:46:51.154Z · anthropic/claude-opus-5.5
Publication: 6f73b002-9a71-4fd5-81f2-be70c6ecaca3
attribution — pass: The paper and its six named authors are correctly identified, and the abstract findings are framed as the authors' reports. The closing line is correctly attributed to the post. One small point: the 'digital twins' framing comes from the quoted abstract, so 'the paper's authors present' would be more precise than 'the post presents'. This blur is minor and not material.
There is a new paper by Raj Chetty, Jamie Fogel, Lawrence F. Katz, Kadeem Noray, Sonya Porter, and James Reisinger
summary — uncertain: The title term 'digital twins' is never explained. The summary also omits the core mechanism: machine learning predicts counterfactuals, meaning what participants would likely have earned without the program, from work histories, demographics and locations. Without that, a lay reader cannot tell what a digital twin is or why checking it against RCTs matters. 'Observational data' and 'parametric estimators' are also left unexplained. Smallest repair: one sentence saying the models predict each participant's likely earnings without training, and that this predicted comparison is the 'digital twin' validated against 37 RCTs.
by using machine learning models to predict counterfactuals from rich data on prior work histories, demographics, and residential locations
date — pass: originalPublished (2026-11-11T19:12:56Z, from the publisher page) matches the feed entry, and the post was first observed shortly afterward. No conflicting earlier date appears.
support — pass: Each claim is source-stated and backed by a verbatim excerpt. The limitations correctly label the 'revolutionize' line as a prediction. The question about other program areas follows from the evidence without presuming an answer. No benefits or costs are invented.
Work of this sort is going to revolutionize economic method.
media — pass: The selected media is the verified direct link to the original Marginal Revolution post. It is the most relevant available item, since no other source media was found.
Challenge or correct this reading
Send a correction or an opposing source with this reading’s URL. Disagreement and uncertainty can remain unresolved.
Economics, institutions and unexpected cross-domain questions; useful source discovery.
Explore Tyler CowenEconomics publication by Tyler Cowen and Alex Tabarrok.
Explore Marginal RevolutionA study that randomly assigns people to groups so that differences in outcomes can be traced to the treatment.
Definition referenceMethods that let computers learn patterns from examples instead of following rules written by hand.
Definition referenceEconomics publication by Tyler Cowen and Alex Tabarrok.
Explore Marginal RevolutionA study that randomly assigns people to groups so that differences in outcomes can be traced to the treatment.
Definition referenceEconomics publication by Tyler Cowen and Alex Tabarrok.
Explore Marginal RevolutionA study that randomly assigns people to groups so that differences in outcomes can be traced to the treatment.
Definition referenceMethods that let computers learn patterns from examples instead of following rules written by hand.
Definition referenceEconomics publication by Tyler Cowen and Alex Tabarrok.
Explore Marginal RevolutionA study that randomly assigns people to groups so that differences in outcomes can be traced to the treatment.
Definition reference