Jmis Methods

Research Design & Methods (jmis-methods)

When to trigger

• You have a mechanism or propositions but no defensible way to test/evaluate them
• The method may not match the claim (a causal IT-value claim resting on a cross-sectional correlation)
• A reviewer asks "what identifies this effect?" or "how do you know the artifact is useful?"
• You need to decide what evidence fits inside the 50-page complete-manuscript ceiling

Match the design to the JMIS research style and the strength of the claim

JMIS is methodologically broad but the design must earn the causal/economic verb in the claim.

IT-value and platform empirics: identify, do not just control

IT investment and platform participation are chosen, not random. Anchor identification in a real source of exogenous variation — a policy change, a staggered system rollout, a platform redesign, a security breach, a pricing shock — and pre-commit the comparison and the assumptions you will defend. With staggered adoption, plan a modern estimator (Callaway–Sant'Anna, Sun–Abraham, de Chaisemartin–D'Haultfœuille) rather than naive TWFE, and design the event-study leads up front. Endogeneity that is only "controlled for" with covariates will draw reviewer fire.

Behavioral IS: design out the threats before you collect data

Build procedural separations against common-method bias — temporal/source/psychological separation, validated and pretested scales, attention and manipulation checks — because statistical fixes (e.g., a marker variable) alone will not convince reviewers later. For experiments, make the IT manipulation realistic and the estimand explicit; report power.

Design-science / data-science: plan the utility evaluation up front

A JMIS artifact paper lives on managerial utility, not algorithmic novelty alone. Decide before building how you will demonstrate utility: held-out benchmarks against credible (not strawman) baselines, a controlled experiment or A/B field deployment, simulation, or expert evaluation — each tied to the artifact's design rationale and to a real managerial decision. State the problem's relevance and the evaluation criteria so reviewers judge rigor and relevance.

Scope the evidence to the 50-page budget

The complete manuscript is capped at ≤50 pages (12pt, double-spaced). Online appendixes are permitted, but the main paper must be self-contained and the core claims established in the body — do not design a study whose key evidence only fits by exporting it. Survey instruments go as separate anonymized attachments. (检索于 2026-06；以官网为准.)

Worked vignette: identifying IT business value (illustrative)

A team wants to claim that an ERP go-live raised plant productivity. A cross-section of ERP-adopters vs. non-adopters cannot carry that claim — adopters differ systematically (larger, better-managed firms self-select). The JMIS design uses the staggered go-live timing across plants as the variation, estimates with Callaway–Sant'Anna (not naive TWFE), pre-specifies the event-study window, and checks that pre-trends are flat before go-live. The identifying assumption — that go-live timing is not driven by anticipated productivity shocks — is argued from the institutional rollout schedule and falsified with a placebo on plants whose go-live slipped. That is the difference between "ERP correlates with productivity" and "ERP go-live raised productivity by X%."

Referee pushback mapped to a design fix

• "Selection — adopters are not comparable to non-adopters." → Switch from cross-section to within-firm timing variation or a defended IV; show balance/pre-trends.
• "How do you know the artifact is actually useful?" (design-science) → Add an evaluation against credible baselines tied to a real managerial decision, not a benchmark of convenience.
• "Common-method bias undermines your survey." → Show the procedural separations you built in ex ante, then the statistical test; do not rely on the test alone.

Execution bridge (StatsPAI / Stata MCP)

For the empirical / causal lane, estimate and audit rather than only specify. Full map: execution-with-mcp. JMIS is empirical IS — survey-based SEM and econometric panels; the chain below serves causal / quasi-experimental designs and many-outcome corrections.

• detect_design → recommend → fit with as_handle=true → audit_result to enumerate the checks the design owes.
• Panel / staggered DiD: callaway_santanna / sun_abraham + bacon_decomposition
  • honest_did_from_result. IV: effective_f_test + anderson_rubin_ci. RDD: rdrobust + mccrary_test.
• Experiments: randomization-based inference and romano_wolf for the many-outcome family-wise correction reviewers expect.

Match the toolchain to the reviewer pool, and report the effect size the venue wants. A run end-to-end (synthetic data, real returns) is in the JF execution walkthrough.