Theory & Prediction Development (jar-theory-development)

When to trigger

• Predictions are descriptive ("X is associated with Y") with no economic channel
• A referee asks "what is the mechanism?" or "why would a rational agent do this?"
• The sign of the effect is ambiguous and needs an explicit argument
• You need to motivate an empirical prediction from an analytical/agency model

How JAR theorizes: economics first

JAR's dominant tradition is positive, economics-based accounting research. Predictions are derived from the behavior of informed economic agents — managers, investors, analysts, auditors, regulators — under asymmetric information, contracting frictions, and proprietary costs. The relevant toolkit is information economics (signaling, disclosure, adverse selection), agency/contracting theory (incentives, monitoring, debt covenants), market price-formation logic (how information enters prices and expectations, in the Ball-Brown lineage), and standard-setting/political-economy arguments. This is not the latent-construct, mediator-moderator style of psychology-based fields: a JAR mechanism is an economic argument about incentives and information, often disciplined by a simple model.

Build the mechanism

• Name the friction. Information asymmetry between whom and whom? A contracting incompleteness? A proprietary cost? The friction is what makes accounting information matter.
• Specify the agent's problem. What does the manager/investor/auditor maximize, and how does the accounting variable change their optimal action or belief?
• Sign the prediction. State the predicted direction and why. If two channels push opposite ways (transparency vs. proprietary cost; recognition vs. disclosure), say so and let the data adjudicate — an ex-ante ambiguous sign is acceptable if argued, but it must be argued.
• Derive observable implications. Beyond the main effect, predict cross-sectional variation: where should the effect be stronger (high information asymmetry, weak governance, binding covenants)? These conditional predictions make the channel testable and separate it from alternatives.
• Consider a model. Analytical work is welcome at JAR; a stylized model that yields the comparative static you test sharpens the prediction and pre-empts "story-time" critiques.

Registered Reports note

On the Registered Reports track, the theory and signed predictions are locked in the Stage 1 protocol before data are seen — strong a priori derivation is essential because in-principle acceptance is granted on the protocol's quality, not the eventual results.

Checklist

• [ ] The friction (information asymmetry / agency / proprietary cost) is named explicitly
• [ ] The agent's optimization problem and the accounting variable's role are specified
• [ ] Each prediction has a stated sign and an economic reason (or a justified ambiguous sign)
• [ ] Cross-sectional / conditional predictions sharpen the channel
• [ ] Alternative (non-causal or mechanical) explanations are anticipated
• [ ] If analytical: comparative statics map to the empirical predictions

Anti-patterns

• Construct-chain theorizing: psychology-style mediators with no economic agent or friction.
• Sign by hand-waving: predicting a direction with no incentive argument.
• Kitchen-sink hypotheses: many associations, no unifying mechanism.
• Mechanism only in the title: a channel asserted in the intro but never tested cross-sectionally.
• Rational/behavioral muddle: invoking efficiency and mispricing interchangeably without committing.

Output format

【Friction】information asymmetry / agency / proprietary cost (between whom)
【Agent's problem】who maximizes what; role of the accounting variable
【Main prediction(s)】sign + economic reason
【Conditional predictions】where stronger/weaker (and why)
【Model?】comparative static derived / not needed
【Alternative explanations to rule out】[...]
【Next step】jar-literature-positioning or jar-methods

Resources

• ../../resources/official-source-map.md — official JAR/Chicago Booth/Wiley URLs (accessed 2026-06-01)
• ../../resources/external_tools.md — empirical-accounting data sources and analytical/econometric software