What Economics Claims to Know
Explanation, Measurement, and the Hidden Architecture of Authority
Overview
Economics is at its best when it makes hard choices visible. It uses models that simplify, measurements that stabilise moving targets, and evaluative language that ranks outcomes. None of that is a defect. The defect begins when these different elements are quietly swapped for one another in the public-facing life of an argument—when an idealised representation is treated as the mechanism itself, when an administratively governed construct is treated as a mind-independent object, or when a normative ranking is treated as if it were compelled by technique. The result is a form of authority that looks technical while resting on contestable premises that are not where the reader can see them.
This summary integrates two complementary parts of a single methodological programme. The first part supplies a disciplined way to type what economics is doing when it explains, measures, and evaluates. The second part explains why the most disciplined typing is routinely crowded out even among competent practitioners, and how the profession can change the payoff structure so that candour becomes the easy, default strategy rather than a competitive sacrifice.
The Core Problem: When “Neutral” Discourse Stops Being Neutral
Economic discourse often advertises neutrality by speaking in a technical register. That register is not itself illegitimate; technical language is necessary for clarity and for the division of intellectual labour. The problem is a recurrent pattern of category slippage that turns contested premises into apparent inevitabilities.
The first move is the conversion of idealisation into literal story. A model introduces simplifications so that a dependence structure can be seen. That is the point of modelling. But the rhetorical temptation is to slide from “this assumption isolates a dependence under stated scope conditions” to “this is what is happening in the world”. When that happens, what began as a controlled representational device is treated as the operative causal mechanism. The audience is invited to treat a model’s architecture as the world’s architecture.
The second move is the conversion of a measurement construct into a mind-independent object. Economics routinely relies on indices, coding rules, survey instruments, administrative categories, and statistical conventions. These are not arbitrary; they are often the best available stabilisations of complex social facts. Yet the rhetoric of neutrality encourages a further slide: the construct is treated as if its meaning and invariance were independent of governance, institutional definition, and revision regimes. The construct becomes “the thing itself” rather than a governed artefact designed for particular purposes under particular constraints.
The third move is the conversion of welfare language into technical entailment. Economics contains a rich and legitimate tradition of welfare analysis. But welfare, efficiency, optimality, distortion, and improvement are not purely descriptive terms. They are predicates that rank states under a criterion. The neutrality problem arises when the criterion is not stated, when contestability is not identified, and when the conclusion is framed as what “the evidence requires”. Technique cannot remove the need to choose a criterion. It can only conceal that choice if the argument is allowed to glide from description to evaluation without marking the boundary.
These three slides can occur separately, but they often occur together in a single policy template. A stylised dependence story becomes “the mechanism”, a governed index becomes “the object”, and a welfare ranking becomes “the necessary conclusion”. The audience experiences the argument as a chain of technical entailment. In reality, it is a chain of type transitions—each legitimate when declared, each illegitimate when smuggled.
A Definition Framework: Typing What Economics Is Doing
The programme begins by fixing three categories that economics must keep distinct if it is to claim authority without laundering contestability.
An idealisation is an assumption introduced for tractability or explanatory leverage. Its legitimacy depends on what counterfactual target it supports, on whether the dependence it isolates is stable across plausible variations, and on whether its scope conditions are stated rather than implied. An idealisation is not a lie; it is a tool. But it is a tool that becomes dangerous when it is treated as literal description rather than as a disciplined representation.
A measurement construct is an administratively and statistically instituted object. It is formed through governance choices, data-collection practices, coding rules, index formulas, and revision regimes. Its legitimacy depends on definition stability, invariance claims that are explicitly defended, and portability limits that are declared rather than assumed. A construct becomes epistemically hazardous when it is treated as though it were a natural kind that remains the same across time, jurisdictions, and measurement regimes.
A normative predicate is a value-laden standard used to rank states—welfare, efficiency, optimality, improvement, loss. It is legitimate when the criterion is stated, the informational basis is declared, and the main margins of contestability are made visible. It becomes rhetorically dangerous when it is presented as if it were merely a technical descriptor.
Once these categories are fixed, the programme supplies operational misuse tests. A claim that presents itself as explanatory but cannot state its counterfactual target is not explanation. A claim that relies on a construct whose definition shifts across jurisdictions or datasets without addressing invariance and revision regimes is not a claim about a stable object. A claim that uses welfare language without stating a welfare criterion is not neutral. A claim that treats an idealisation as literal description without robustness or scope discipline lets rhetoric outrun warrant.
The point of this framework is not scholastic classification. It is auditability. When claims are typed, the reader can see what kind of warrant is required and where disagreement properly lives. Disagreement shifts from atmospherics (“you’re anti-growth”, “you’re politicising science”, “you don’t understand the model”) to premises (“your counterfactual target is underspecified”, “your construct is not invariant”, “your welfare criterion embeds a contestable ranking”). That is the difference between authority as aura and authority as procedure.
What Counts as Explanation: Prediction, Identification, Mechanism
Explanation in economics is often conflated with prediction or with credible identification. Prediction matters. Identification matters. But neither automatically licenses an explanatory upgrade.
Prediction alone does not settle ontology. A model can forecast well for reasons that do not match the world’s mechanisms. A predictive success can be a fragile alignment between a fitted structure and a particular regime. Explanation, by contrast, is a claim about dependence under change: if X were changed in a specified way, Y would change, under stated conditions, for stated reasons. It is not enough to show that X and Y move together, or even that an intervention on X shifts Y. The explanatory claim needs an explicit counterfactual target and a warranted account of what the result should be taken to mean.
Identification raises the credibility of dependence claims. It can warrant causal interpretations under assumptions. But it does not automatically give mechanism, portability, or policy necessity. The common misuse is to treat “identified effect” as “deep explanation”, and then to treat “deep explanation” as “policy entailment”. The disciplined move is to state the estimand, state the counterfactual, state the scope conditions, and then state what is and is not being claimed about mechanism.
Idealisation has a legitimate explanatory role when it clarifies dependence, when robustness supports portability, and when scope is explicit. It becomes harmful when it quietly fixes institutional or normative stakes. An assumption that “agents optimise” can be a modelling device; it becomes a psychological story when written as if it describes minds. An assumption that “markets clear” can be a simplifying structure; it becomes an institutional claim when written as if it describes constraints and power. The framework demands that these transitions be marked and warranted, not performed silently.
Measurement and Social Ontology: Why Constructs Are Not Passive
Many economic objects are not discovered in the way chemical elements are discovered. They are instituted. Inflation indices, GDP, unemployment, poverty lines, productivity series, regulatory capital ratios—these exist through administrative practice, statistical convention, and governance choices. That does not make them “made up”. It makes them governed.
This governance matters because measurement is not inert. Once a construct becomes a target—once it is used in budgets, contracts, performance frameworks, or political narratives—it feeds back into behaviour. Definitions reshape incentives; incentives reshape the measured world; the measured world reshapes the definition. Claims that treat governed constructs as mind-independent objects therefore invite a distinctive error: the error of treating policy disputes as disputes about facts when they are disputes about definitions, revision regimes, and administrative dependence.
The disciplined approach requires construct declarations that state definitional sources, revision regimes, portability limits, and target-reactivity risks. It also requires a more careful relationship between measurement and generalisation. Cross-time and cross-jurisdiction comparisons often fail not because anyone is incompetent, but because construct drift is real. If the definition changes, the object changes. Therefore, law-like claims require qualification: the claim must either be invariant under definitional shifts, or it must be explicitly local to a measurement regime.
Rationality, Agency, and the Normative Loading of “Preference”
Preference language plays multiple roles. It can be a modelling device, a parameterisation of choice under constraints. It can also be a causal story about psychology and motivation. The problem arises when economists move between these roles without signalling the shift. If preferences are treated as mere modelling primitives, they are not automatically claims about mental states. If preferences are treated as mental states, they require warrants from psychology and from institutional context.
Bounded rationality does not provide a licence for loosened standards. It increases the need for typological discipline: what is idealised, what is measured, what is moralised. Likewise, the term “irrational” often functions as covert moral judgement. It can mean “inconsistent with a specified model”, which is technical. It can also mean “bad”, which is evaluative. Disciplined discourse states which is intended. Welfare improvement is similarly loaded. It is always an improvement relative to a declared criterion, not a conclusion compelled by technique.
Why the Problem Persists: Incentives, Genre, and Selection
If the distinctions above are obvious, why does slippage persist? The integrated programme rejects the comforting answer—“because economists are careless”—and replaces it with an institutional explanation: slippage is often equilibrium behaviour under professional incentives.
Economic communication is selected and rewarded. Journals, referees, editors, policy institutions, and attention markets reward perceived generality, decisiveness, and apparent neutrality. A typed, conditional claim is epistemically stronger but rhetorically weaker. It takes more space. It creates more points of vulnerability. It makes contestability visible. Under those conditions, candour imposes private costs on authors while delivering benefits that are diffuse and delayed. Those benefits accrue to the profession, to downstream users, and to the public, not primarily to the author.
Genre constraints intensify this. The journal article demands a clean narrative arc and a “main contribution”. Policy discourse demands legitimacy under time constraints. Both reward arguments that resemble entailments: the evidence implies we must. Typological discipline disrupts that entailment by making the inferential joints visible. Therefore, even sincere and competent authors face incentives to compress. They move caveats to footnotes, scope conditions to appendices, and welfare criteria into silence. The most visible parts of discourse become the most upgraded, because that is where rhetorical force matters most.
Refereeing and editorial selection enforce the payoff structure. This can happen in good faith. Referees audit what is cheapest to audit: identification, robustness, and technical execution. Those are valuable. But a substitution problem emerges: strong identification credibility becomes a master credential that is allowed to stand in for explanatory clarity, construct discipline, and normative disclosure. Once the design looks credible, the temptation is to treat other issues as settled infrastructure. Measurement is assumed stable. Welfare language is treated as harmless. Mechanism is narrated rather than warranted.
Transparency therefore loses without institutional support. It is a collective-action problem. Everyone benefits if disclosure is standard; individuals lose if they disclose while others do not. Under that equilibrium, exhortation is ineffective. Telling economists to “be transparent” asks them to bear private costs for public goods without changing the selection environment that punishes them for doing so.
What a Reform Looks Like: Minimal, Enforceable, Non-Sectarian
The integrated programme therefore ends with design, not sermon. The aim is to change payoffs at the points where claims are selected and certified, while remaining compatible with pluralism. A reform that requires agreement on substantive economics will fail. A reform that requires virtue will be ignored. The right target is procedural: make typological discipline cheap, routine, and enforceable.
A first reform is typed-claim declarations. A short, standardised box forces authors to state what is idealised and why, what constructs are being used and how stable they are, and what welfare criterion is invoked if welfare language appears. The point is not to add bureaucracy. The point is to prevent silent transitions. When typing becomes compliance rather than self-sacrifice, the private cost of candour falls.
A second reform is estimand-first and counterfactual-first reporting. If a paper claims causal or explanatory relevance, it should state what is being estimated and what counterfactual change is contemplated. This makes explanation claims earn their type. It does not force one method. It forces clarity about what a method warrants.
A third reform is construct stability disclosure. A construct declaration should state definitional source, revision regimes, portability limits, and target-reactivity where relevant. This shifts measurement from assumed infrastructure to audited premise. Again, the aim is not philosophical lectures; it is minimal information that allows a reader to evaluate whether an object is stable enough for the claims being made.
A fourth reform is welfare-criterion triggers. If welfare, efficiency, or optimality language appears in the abstract, introduction, conclusion, or policy section, the criterion must be stated, along with the main margin of contestability. This does not settle moral disagreement. It makes moral premises legible.
A fifth reform is lightweight referee and editorial templates. A one-page checklist makes enforcement feasible. It changes the substitution equilibrium by preventing identification credibility from laundering claim types. It also makes candour non-stigmatising: authors disclose because they must, and referees expect it.
These reforms are deliberately modelled on successful standardisation in adjacent domains of economics, where journal-level procedures have shifted norms without imposing theoretical consensus. The lesson is simple. If a profession wants transparency, it must build transparency into its certification pipeline.
How to Judge Success: Observable Changes in Discourse
Success should not be defined as agreement. Economics will not become value-free. Nor should it. The success condition is that disagreement becomes legible and auditability becomes routine.
The integrated programme yields observable predictions. If typing and disclosure templates are enforced, abstracts and conclusions should contain fewer rhetorical upgrades. Welfare language should more often carry an explicit criterion. Constructs should be defined and their portability limits stated. Dependence claims should be less often narrated as mechanism. Caveats should migrate from low-visibility appendices into short, standardised statements near the claims that rely on them.
These are not vague aspirations. They are changes that can be detected in texts and in review practice. They also allow journals and the profession to evaluate reforms without needing to adjudicate substantive policy debates. The aim is to improve the epistemic hygiene of economic authority, not to police what economists may conclude.
Boundary conditions matter. Some domains require abstraction. Some constructs are inherently governed. In such cases, the goal is not to eliminate idealisation or measurement dependence, but to keep them visible and properly typed. Likewise, normative disagreement cannot be removed by better econometrics. It can only be declared rather than laundered.
Closing Claim: Legitimacy by Procedure
Economics earns authority when it makes its warrants auditable. Models are legitimate when their idealisations are stated and their scope is disciplined. Measures are legitimate when their governance is disclosed and their invariance claims are defended. Welfare claims are legitimate when their criteria are stated and their contestability is acknowledged. None of this requires a new school of economics. It requires a procedural shift in how economics certifies claims.
The integrated message is therefore austere. The most damaging errors in economics are often not mistakes of calculation. They are mistakes of type. And the most effective remedy is not moral exhortation. It is institutional design that makes typological discipline the default condition of credible authority.