Economics, Business & Finance: Field disagreement mapping

What you'll be able to do: Find out, rather than guess, whether a claim is one economists broadly agree on or one they genuinely dispute, which is the single most important skill for using AI safely in this subject.

Why this method exists

The primary article for this subject opens with a warning: AI gives a confident, complete, well-organised answer whether the underlying question is settled or genuinely contested, and the tone of the answer gives you no way to tell which one you got. This method is the direct response to that warning. It doesn't ask the AI whether something is true. It asks the AI to map who in the field believes what, on what evidence, and then tells you to go check that map against a real source, because the AI's own claim about how contested something is needs verifying just as much as any other claim it makes.

A real instrument to anchor on

Economists have actually built a tool for exactly this: panels of prominent economists across the political spectrum, regularly surveyed on specific policy and empirical propositions, with their agree/disagree/uncertain responses published along with confidence weights. The best-known running example is the IGM Forum survey run out of a US business school, which has polled its panel on propositions from rent control to tariffs to fiscal multipliers for over a decade.

That's the model to imitate. Instead of asking "do economists agree that X", ask what such a poll would likely show, then go verify it, the AI's guess about the state of the field is a starting hypothesis, not a finding.

For this claim ([state it precisely]) tell me: (a) what standard economic theory implies under its usual assumptions, (b) whether that implication survives when the key assumption is relaxed, (c) what the empirical literature finds, separately from the theory, and (d) how confident you are that this is actually contested versus settled among specialists, and why,
Then tell me the name of a real source, a survey, a literature review, a specific paper. I could check this against myself.

Layering theory, robustness, empirics and confidence separately matters because they can point different directions. A theoretical prediction can survive at the textbook level and fail once you relax one assumption; an empirical literature can be mixed even where the theory is clean.

The pair worth memorising

Rent control. A price ceiling on rents. Close to the nearest thing empirical economics has to a professional consensus: it tends to reduce the quantity and quality of available rental housing over time, by discouraging new construction and maintenance of existing units in the controlled market, And it remains one of the most fiercely contested political issues in city and state politics anywhere it's proposed.

Minimum wage. A price floor in the labour market. The structural mirror image of rent control on a supply-and-demand diagram, and the empirical picture is the mirror image too: a genuine, ongoing dispute among labour economists about how large the employment effect is, whether it's distinguishable from zero at moderate levels, and how it changes as the floor rises relative to the local wage distribution. Meanwhile plenty of public debate on both sides treats it as obviously settled.

These sit one diagram apart and point opposite ways on the consensus/controversy axis. If you only remember one worked pair from this whole subject, make it this one, it's the cleanest demonstration available that "looks the same in the textbook" and "same state of professional agreement" are unrelated facts.

Tariffs are a good third anchor: fairly strong empirical consensus, particularly from recent direct measurement of tariff pass-through, that the cost of a tariff lands substantially on the importing country's consumers and firms rather than being absorbed by foreign exporters, a finding regularly contradicted in political rhetoric on all sides.

How confidence has moved over time: and why that matters

The method isn't just "poll the current state." Part of mapping disagreement is knowing that the field's confidence on some questions has shifted hard within living memory. The belief that policymakers could reliably trade off a bit more inflation for a bit less unemployment along a stable curve was widely held until the stagflation of the 1970s broke it; then the reverse lesson (that this trade-off might not exist at all in the long run) held for decades until the post-2021 inflation episode, where unemployment stayed surprisingly low even as inflation rose sharply, unsettled that too. A claim that was "settled" in one decade can be an open question in the next, and treating a model as timelessly agreed risks importing a confidence the field itself has since walked back.

Across the disciplines

Macroeconomics. The size of fiscal multipliers, the neutrality of money in the long run, the appropriate response to asset-price bubbles, all have real, ongoing professional disagreement, not just political disagreement.

Trade. Broad consensus that voluntary trade increases the total size of the economic pie; much less consensus on how large the transition costs are for displaced workers and how well those costs get addressed in practice.

Corporate finance. Genuine, live dispute over how much of the corporate income tax ultimately falls on shareholders versus workers via lower wages, an empirical magnitude, not a settled split, with estimates varying widely across studies and methods.

Management. Whether a given management fad, a particular organisational structure, a particular incentive scheme, actually improves performance, versus reflecting survivorship bias in which firms got studied and written up.

Pitfalls

1. Trusting the AI's own "this is contested" or "this is settled" verdict without checking it. That verdict is exactly the claim this method exists to verify, not to accept.
2. Confusing political controversy with professional disagreement. They correlate poorly, which is the entire point of the rent-control/minimum- wage pair.
3. Treating a poll from a decade ago as current. Confidence moves; check the date.
4. Asking a yes/no question instead of asking for the shape of the disagreement. "Is X true" invites a verdict; "who believes what, on what evidence" invites a map.
5. The tell: every claim you've run through this method has come back either "broadly agreed" or "hotly disputed": never anything more specific, like "the theory is clean but the empirics are thin" or "settled at moderate levels, contested at extremes." A tool giving you the same two verdicts regardless of topic is pattern-matching your question, not reporting the field.

Try this today

Take one claim from this subject you've been treating as obviously true or obviously false. Run it through the four-part prompt above. Then find one real source (a survey, a review article, a textbook's footnote) and check whether the AI's account of the field matches it.