Methods & Identification (mksc-methods)
When to trigger
- You must choose between a structural, analytical, or reduced-form/causal-ML approach
- The model is written but not yet estimable (parameters, moments, normalization)
- Identification is hand-waved ("we use instruments") without specifics
- A reviewer says "the design cannot identify the structural parameters"
Choose the genre that fits the claim
Marketing Science is methodologically plural around a modeling core: structural econometrics, analytical models, econometric/statistical analysis, ML tools, surveys, and experiments — all judged by whether they develop, test, or rigorously apply a formal model.
| Claim / goal |
Approach that earns it |
| Quantify demand and simulate a policy |
Structural demand (BLP/mixed logit), supply FOCs, counterfactual |
| Forward-looking behavior, adoption, churn |
Dynamic discrete choice / dynamic games (Rust, BBL, CCP) |
| Strategic-interaction insight, comparative statics |
Analytical (game-theoretic) model |
| Bidding, sponsored search, marketplaces |
Auction/structural-IO model with equilibrium bidding |
| Heterogeneous treatment effects tied to a model |
Causal ML (double/debiased ML, causal forests) disciplined by theory |
| Causal effect from field variation |
Field experiment / quasi-experiment as identifying variation |
A field experiment or quasi-experiment is welcome when it identifies a model primitive or validates a mechanism, not as a stand-alone reduced-form result.
Make the model estimable and identified
- Estimator: match it to the model — GMM (BLP moment conditions), MLE/SMLE, simulated method of moments, or hierarchical Bayes (MCMC) for rich heterogeneity.
- Instruments / identifying variation: name them concretely (cost shifters, BLP/Hausman/differentiation instruments, exclusion restrictions, randomized or discontinuity variation) and defend exogeneity.
- Normalizations and functional form: state outside-good normalization, scale/location normalizations, and which assumptions are substantive vs. convenience.
- Computation: specify the solver, equilibrium/inner-loop fixed point, starting values, and how you handle multiple equilibria or local optima.
For analytical papers
Specify the equilibrium concept, solve it, and prove the claims; relegate long proofs to an appendix but state the key steps. Plan to validate counterintuitive predictions and discuss robustness to the modeling assumptions that drive them.
Execution bridge (StatsPAI / Stata MCP)
For the empirical / causal lane, estimate and audit rather than only specify. Full
map: execution-with-mcp. Marketing Science is heavily structural/analytical; the chain below serves its reduced-form / field-experiment lane — structural demand and analytical modeling are outside this causal-inference toolchain.
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.
Checklist
Anti-patterns
- A structural model "estimated" with no stated moments or instruments.
- Reduced-form regressions presented as the whole contribution at MKSC.
- Ignoring multiplicity of equilibria or optimizer convergence.
- Treating convenience assumptions as if they were innocuous.
Methods pass for Marketing Science
Use this as a second-pass capability check. First lock the demand/supply mechanism, fit evidence, and counterfactual decision margin; then test whether the manuscript addresses quantitative marketing reviewers who read the model through the managerial counterfactual it makes possible.
- Primary move: Name assumptions, diagnostics, robustness, falsification, and failure modes; do not accept a method section that hides the decisive validity threat.
- Decision ledger: return
claim / evidence / blocker / next edit rows so the next pass can patch the manuscript directly.
- Neighbor test: compare against Journal of Marketing Research for empirical marketing breadth, Management Science for wider OR/MS reach, Quantitative Marketing and Economics for specialist modeling; if the neighboring outlet has the stronger audience claim, recommend re-routing before polishing.
- Verification floor: before submission-ready advice, re-open
resources/official-source-map.md for volatile rules and name the one unresolved fact that could change the recommendation.
Output format
【Genre】structural / analytical / causal-ML / experiment
【Model→estimator】GMM / MLE-SMLE / SMM / hierarchical Bayes
【Identification】instruments/variation → parameters; exogeneity defense
【Normalizations/assumptions】substantive vs. convenience
【Computation】solver, fixed point, starting values, multiplicity
【Next step】mksc-data-analysis
Source: brycewang-stanford/Awesome-Journal-Skills → Marketing-Science-Skills/skills/mksc-methods/SKILL.md
1---2name: mksc-methods3description: Use when the empirical/analytical approach is the bottleneck for a Marketing Science manuscript — choosing among structural econometrics, analytical modeling, and model-disciplined causal/ML methods, and making the model estimable and identified. Designs the approach; it does not execute the estimation and counterfactuals (mksc-data-analysis).4---5
6
7# Methods & Identification (mksc-methods)
8
9## When to trigger
10
11- You must choose between a structural, analytical, or reduced-form/causal-ML approach
12- The model is written but not yet estimable (parameters, moments, normalization)
13- Identification is hand-waved ("we use instruments") without specifics
14- A reviewer says "the design cannot identify the structural parameters"
15
16## Choose the genre that fits the claim
17
18Marketing Science is methodologically plural around a modeling core: structural econometrics, analytical models, econometric/statistical analysis, ML tools, surveys, and experiments — all judged by whether they develop, test, or rigorously apply a formal model.
19
20| Claim / goal | Approach that earns it |
21|------------------------------------------------|-------------------------------------------------------------------|
22| Quantify demand and simulate a policy | Structural demand (BLP/mixed logit), supply FOCs, counterfactual |
23| Forward-looking behavior, adoption, churn | Dynamic discrete choice / dynamic games (Rust, BBL, CCP) |
24| Strategic-interaction insight, comparative statics | Analytical (game-theoretic) model |
25| Bidding, sponsored search, marketplaces | Auction/structural-IO model with equilibrium bidding |
26| Heterogeneous treatment effects tied to a model| Causal ML (double/debiased ML, causal forests) disciplined by theory |
27| Causal effect from field variation | Field experiment / quasi-experiment as identifying variation |
28
29A field experiment or quasi-experiment is welcome **when it identifies a model primitive or validates a mechanism**, not as a stand-alone reduced-form result.
30
31## Make the model estimable and identified
32
33- **Estimator**: match it to the model — GMM (BLP moment conditions), MLE/SMLE, simulated method of moments, or hierarchical Bayes (MCMC) for rich heterogeneity.
34- **Instruments / identifying variation**: name them concretely (cost shifters, BLP/Hausman/differentiation instruments, exclusion restrictions, randomized or discontinuity variation) and defend exogeneity.
35- **Normalizations and functional form**: state outside-good normalization, scale/location normalizations, and which assumptions are substantive vs. convenience.
36- **Computation**: specify the solver, equilibrium/inner-loop fixed point, starting values, and how you handle multiple equilibria or local optima.
37
38## For analytical papers
39
40Specify the equilibrium concept, solve it, and prove the claims; relegate long proofs to an appendix but state the key steps. Plan to validate counterintuitive predictions and discuss robustness to the modeling assumptions that drive them.
41
42## Execution bridge (StatsPAI / Stata MCP)
43
44For the **empirical / causal lane**, estimate and audit rather than only specify. Full
45map: [`execution-with-mcp`](../../../shared-resources/empirical-methods/execution-with-mcp.md). Marketing Science is heavily structural/analytical; the chain below serves its reduced-form / field-experiment lane — structural demand and analytical modeling are outside this causal-inference toolchain.
46
47- `detect_design` → `recommend` → fit with `as_handle=true` → `audit_result` to
48 enumerate the checks the design owes.
49- **Panel / staggered DiD:** `callaway_santanna` / `sun_abraham` + `bacon_decomposition`
50 + `honest_did_from_result`. **IV:** `effective_f_test` + `anderson_rubin_ci`. **RDD:**
51 `rdrobust` + `mccrary_test`.
52- **Experiments:** randomization-based inference and `romano_wolf` for the many-outcome
53 family-wise correction reviewers expect.
54
55Match the toolchain to the **reviewer pool**, and report the effect size the venue
56wants. A run end-to-end (synthetic data, real returns) is in the
57[JF execution walkthrough](../../../Journal-of-Finance-Skills/resources/worked-examples/02-execution-walkthrough.md).
58## Checklist
59
60- [ ] Genre (structural / analytical / causal-ML / experiment) matches the claim
61- [ ] Estimator matches the model (GMM/MLE/SMM/Bayes) and is stated
62- [ ] Identifying variation/instruments named and exogeneity defended
63- [ ] Normalizations and substantive vs. convenience assumptions separated
64- [ ] Computation (solver, fixed point, starting values, multiple equilibria) planned
65- [ ] Any experiment/quasi-experiment tied to a model primitive, not free-standing
66
67## Anti-patterns
68
69- A structural model "estimated" with no stated moments or instruments.
70- Reduced-form regressions presented as the whole contribution at MKSC.
71- Ignoring multiplicity of equilibria or optimizer convergence.
72- Treating convenience assumptions as if they were innocuous.
73
74
75## Methods pass for Marketing Science
76
77Use this as a second-pass capability check. First lock the demand/supply mechanism, fit evidence, and counterfactual decision margin; then test whether the manuscript addresses quantitative marketing reviewers who read the model through the managerial counterfactual it makes possible.
78
79- **Primary move:** Name assumptions, diagnostics, robustness, falsification, and failure modes; do not accept a method section that hides the decisive validity threat.
80- **Decision ledger:** return `claim / evidence / blocker / next edit` rows so the next pass can patch the manuscript directly.
81- **Neighbor test:** compare against Journal of Marketing Research for empirical marketing breadth, Management Science for wider OR/MS reach, Quantitative Marketing and Economics for specialist modeling; if the neighboring outlet has the stronger audience claim, recommend re-routing before polishing.
82- **Verification floor:** before submission-ready advice, re-open `resources/official-source-map.md` for volatile rules and name the one unresolved fact that could change the recommendation.
83
84## Output format
85
86```
87【Genre】structural / analytical / causal-ML / experiment
88【Model→estimator】GMM / MLE-SMLE / SMM / hierarchical Bayes
89【Identification】instruments/variation → parameters; exogeneity defense
90【Normalizations/assumptions】substantive vs. convenience
91【Computation】solver, fixed point, starting values, multiplicity
92【Next step】mksc-data-analysis
93```
94
95---
96
97**Source:** [`brycewang-stanford/Awesome-Journal-Skills`](https://github.com/brycewang-stanford/Awesome-Journal-Skills) → `Marketing-Science-Skills/skills/mksc-methods/SKILL.md`
thedixitjain