Rebuttal Strategy (expecon-rebuttal)
When to trigger
- An R&R or rejection arrived and you must decide what to concede, contest, or re-run
- A referee asks for a new treatment or more sessions — the costliest and most common ExpEcon request
- A reviewer raised a deception or power/inference-unit concern that could be decisive
- Two referees disagree, or the two editors' priorities differ, and you need a coherent plan
The ExpEcon revision reality: new data is on the table
Unlike journals where revisions are textual, an ExpEcon R&R often requires collecting new data — an added control treatment, more matching groups for power, or a robustness condition. Plan the response around that fact.
- Triage every comment into: (a) new data needed (new treatment / more sessions), (b) re-analysis (correct inference unit, MHT correction, new test), (c) clarification (already in the paper, surface it), (d) respectful disagreement (with evidence). Most letters are mostly (b)–(c); isolate the (a) items because they set the timeline and budget.
- Address the gate concerns first and decisively. A deception worry is existential: either prove the procedure is not deception under the ESA definition, or re-run without it — there is no middle path. A power/unit concern: re-analyze at the session/matching-group level and report MDE; if truly underpowered, run additional sessions rather than argue.
- For each new-data request, decide and justify. Run it if it sharpens identification or power; if not, explain precisely why the existing design already answers the question (and offer a cheaper analysis that addresses the underlying worry). Never silently ignore a request for a treatment.
- Re-deposit the package. New sessions mean updated data, instructions, and z-Tree/oTree code in the repository; keep the replication deposit in sync with the revised paper.
Triaging the common ExpEcon requests
| Referee request |
Default response |
Cost |
| "Add a control treatment that isolates X" |
usually run it — it is the cleanest answer |
new sessions |
| "Underpowered / wrong inference unit" |
re-analyze at group level; add sessions if MDE not met |
re-analysis ± data |
| "Could be deception / demand effects" |
prove non-deception per ESA definition, or re-run |
possibly fatal |
| "Comprehension/confusion confound" |
report pass rates; re-analyze excluding failers |
re-analysis |
| "Why not JEBO/GEB?" |
sharpen the methods contribution in framing |
text only |
| "More structural estimation" |
add if it identifies a parameter; else decline with reason |
analysis |
Writing the response letter
- Point-by-point, quote-then-respond. Reproduce each comment, then state the change, the location (section/table/figure), and the new result. Editors and referees verify against the manuscript.
- Lead with what you did, not why the referee was wrong. Even when contesting, open with the action taken or the evidence, in a collegial register.
- Report new numbers honestly. If a new treatment weakened the effect, say so and interpret it; an experimentalist audience values the clean result over the convenient one.
- Separate primary from exploratory in everything you add, mirroring the pre-registration logic.
- Mind the two-editor structure. Where referees conflict, address both and let the response help the editors converge; do not pick one referee and ignore the other.
Worked vignette (illustrative)
A referee says a coordination-game result "might be a demand effect — subjects guessed the hypothesis." Three moves: (1) note the abstract/neutral framing already used (no leading language — quote the instructions); (2) report the post-experiment belief question showing subjects did not infer the hypothesis; (3) if still unconvinced, add a treatment with an obfuscated cover task and show the effect persists (illustrative: 0.7 SD, unchanged). The letter leads with these three actions, not with "the referee is mistaken," and updates the instructions and data in the repository to match.
Budgeting the new-data round
An added treatment is not just bench cost — it is a timeline the editors will weigh. Before committing, estimate: sessions needed for the new arm at the group-level power you must hit, lab availability, IRB amendment time, and re-deposit effort. If a requested treatment is genuinely infeasible, say so explicitly and offer the strongest available substitute (a re-analysis, a bound, or a focused robustness condition) rather than a vague promise. Editors respond far better to "we ran 10 of the 12 sessions you asked for; here is why 12 was infeasible and what the 10 show" than to silence or hand-waving.
Checklist
Keeping pre-registration credible through revision
A subtle ExpEcon trap: a referee asks for a new analysis or treatment, you run it, and your "pre-registered" framing quietly erodes because the new work was not pre-specified. Protect the distinction in the revision. Anything you add at the referee's request is, by definition, not part of the original confirmatory plan — label it as referee-requested and exploratory, and keep the original pre-registered primary test reported exactly as filed. For a substantial new treatment, consider pre-registering its analysis before collecting it, and say so in the letter. Editors trust authors who keep the confirmatory/exploratory line clean under revision pressure; blurring it to make new results look pre-planned is the fastest way to lose that trust.
Anti-patterns
- Arguing a deception concern away instead of removing the procedure
- Promising a treatment "in future work" when the referee asked for it now
- Burying a weakened effect from a new treatment instead of interpreting it
- A defensive letter that disputes referees before reporting any changes
- Updating the manuscript but leaving the repository deposit stale
- Satisfying one referee while ignoring the other under a two-editor decision
When the decision is reject, not R&R
A flagship reject is not the end of the experiment — the data are clean and the design is real. Triage the report for what is fixable (positioning, power, a missing control) versus fatal at this venue (a deception concern that cannot be removed, or a contribution the editors judge topical rather than methodological). If the core is sound but the contribution was read as too narrow, JESA is the natural next ESA home (replications, null results, shorter design notes). If a referee's deception read is correct, the only path is a re-designed follow-up. Do not resubmit the same paper to the flagship after a reject without a substantive change to the dimension that sank it.
Output format
【Journal】Experimental Economics (ESA method flagship)
【Skill】expecon-rebuttal
【Decision type】R&R / reject-resubmit / conditional accept
【Triage】new-data / re-analysis / clarification / disagreement (counts)
【Gate items】deception / incentive concerns — resolution
【New-data items】treatment/sessions to run or justified omission
【Letter form】point-by-point, quote-then-respond, locations cited? [Y/N]
【Package sync】repository updated to revision? [Y/N]
【Next step】re-run expecon-identification/robustness for new data → expecon-submission for resubmission
Source: brycewang-stanford/Awesome-Journal-Skills → Experimental-Economics-Skills/skills/expecon-rebuttal/SKILL.md
1---2name: expecon-rebuttal3description: Use when responding to a revise-and-resubmit or referee report for an Experimental Economics (ExpEcon) manuscript — including when a referee demands a new treatment, more sessions, or raises a deception/power concern. Plans the response; it does not run the new analysis.4---5
6
7# Rebuttal Strategy (expecon-rebuttal)
8
9## When to trigger
10
11- An R&R or rejection arrived and you must decide what to concede, contest, or re-run
12- A referee asks for a **new treatment or more sessions** — the costliest and most common ExpEcon request
13- A reviewer raised a **deception** or **power/inference-unit** concern that could be decisive
14- Two referees disagree, or the two editors' priorities differ, and you need a coherent plan
15
16## The ExpEcon revision reality: new data is on the table
17
18Unlike journals where revisions are textual, an ExpEcon R&R often requires **collecting new data** — an added control treatment, more matching groups for power, or a robustness condition. Plan the response around that fact.
19
201. **Triage every comment** into: (a) **new data needed** (new treatment / more sessions), (b) **re-analysis** (correct inference unit, MHT correction, new test), (c) **clarification** (already in the paper, surface it), (d) **respectful disagreement** (with evidence). Most letters are mostly (b)–(c); isolate the (a) items because they set the timeline and budget.
212. **Address the gate concerns first and decisively.** A deception worry is existential: either prove the procedure is not deception under the ESA definition, or re-run without it — there is no middle path. A power/unit concern: re-analyze at the session/matching-group level and report MDE; if truly underpowered, run additional sessions rather than argue.
223. **For each new-data request, decide and justify.** Run it if it sharpens identification or power; if not, explain precisely why the existing design already answers the question (and offer a cheaper analysis that addresses the underlying worry). Never silently ignore a request for a treatment.
234. **Re-deposit the package.** New sessions mean updated data, instructions, and z-Tree/oTree code in the repository; keep the replication deposit in sync with the revised paper.
24
25## Triaging the common ExpEcon requests
26
27| Referee request | Default response | Cost |
28|-----------------|------------------|------|
29| "Add a control treatment that isolates X" | usually run it — it is the cleanest answer | new sessions |
30| "Underpowered / wrong inference unit" | re-analyze at group level; add sessions if MDE not met | re-analysis ± data |
31| "Could be deception / demand effects" | prove non-deception per ESA definition, or re-run | possibly fatal |
32| "Comprehension/confusion confound" | report pass rates; re-analyze excluding failers | re-analysis |
33| "Why not JEBO/GEB?" | sharpen the methods contribution in framing | text only |
34| "More structural estimation" | add if it identifies a parameter; else decline with reason | analysis |
35
36## Writing the response letter
37
38- **Point-by-point, quote-then-respond.** Reproduce each comment, then state the change, the location (section/table/figure), and the new result. Editors and referees verify against the manuscript.
39- **Lead with what you did, not why the referee was wrong.** Even when contesting, open with the action taken or the evidence, in a collegial register.
40- **Report new numbers honestly.** If a new treatment weakened the effect, say so and interpret it; an experimentalist audience values the clean result over the convenient one.
41- **Separate primary from exploratory** in everything you add, mirroring the pre-registration logic.
42- **Mind the two-editor structure.** Where referees conflict, address both and let the response help the editors converge; do not pick one referee and ignore the other.
43
44## Worked vignette (illustrative)
45
46A referee says a coordination-game result "might be a demand effect — subjects guessed the hypothesis." Three moves: (1) note the **abstract/neutral framing** already used (no leading language — quote the instructions); (2) report the **post-experiment belief question** showing subjects did not infer the hypothesis; (3) if still unconvinced, add a treatment with an obfuscated cover task and show the effect persists (illustrative: 0.7 SD, unchanged). The letter leads with these three actions, not with "the referee is mistaken," and updates the instructions and data in the repository to match.
47
48## Budgeting the new-data round
49
50An added treatment is not just bench cost — it is a *timeline* the editors will weigh. Before committing, estimate: sessions needed for the new arm at the group-level power you must hit, lab availability, IRB amendment time, and re-deposit effort. If a requested treatment is genuinely infeasible, say so explicitly and offer the strongest available substitute (a re-analysis, a bound, or a focused robustness condition) rather than a vague promise. Editors respond far better to "we ran 10 of the 12 sessions you asked for; here is why 12 was infeasible and what the 10 show" than to silence or hand-waving.
51
52## Checklist
53
54- [ ] Every comment triaged: new-data / re-analysis / clarification / disagreement
55- [ ] Gate concerns (deception, incentive compatibility) resolved decisively, not argued around
56- [ ] Power/inference-unit concerns met with group-level re-analysis or additional sessions
57- [ ] Each new-treatment request is run or its omission is precisely justified
58- [ ] Response is point-by-point, quote-then-respond, with exact locations of changes
59- [ ] New results reported honestly, including any that weaken the original claim
60- [ ] Replication deposit (data, instructions, z-Tree/oTree) updated to match the revision
61
62## Keeping pre-registration credible through revision
63
64A subtle ExpEcon trap: a referee asks for a new analysis or treatment, you run it, and your "pre-registered" framing quietly erodes because the new work was not pre-specified. Protect the distinction in the revision. Anything you add at the referee's request is, by definition, **not** part of the original confirmatory plan — label it as referee-requested and exploratory, and keep the original pre-registered primary test reported exactly as filed. For a substantial new treatment, consider pre-registering *its* analysis before collecting it, and say so in the letter. Editors trust authors who keep the confirmatory/exploratory line clean under revision pressure; blurring it to make new results look pre-planned is the fastest way to lose that trust.
65
66## Anti-patterns
67
68- Arguing a deception concern away instead of removing the procedure
69- Promising a treatment "in future work" when the referee asked for it now
70- Burying a weakened effect from a new treatment instead of interpreting it
71- A defensive letter that disputes referees before reporting any changes
72- Updating the manuscript but leaving the repository deposit stale
73- Satisfying one referee while ignoring the other under a two-editor decision
74
75## When the decision is reject, not R&R
76
77A flagship reject is not the end of the experiment — the data are clean and the design is real. Triage the report for what is *fixable* (positioning, power, a missing control) versus *fatal at this venue* (a deception concern that cannot be removed, or a contribution the editors judge topical rather than methodological). If the core is sound but the contribution was read as too narrow, **JESA** is the natural next ESA home (replications, null results, shorter design notes). If a referee's deception read is correct, the only path is a re-designed follow-up. Do not resubmit the same paper to the flagship after a reject without a substantive change to the dimension that sank it.
78
79## Output format
80
81```text
82【Journal】Experimental Economics (ESA method flagship)
83【Skill】expecon-rebuttal
84【Decision type】R&R / reject-resubmit / conditional accept
85【Triage】new-data / re-analysis / clarification / disagreement (counts)
86【Gate items】deception / incentive concerns — resolution
87【New-data items】treatment/sessions to run or justified omission
88【Letter form】point-by-point, quote-then-respond, locations cited? [Y/N]
89【Package sync】repository updated to revision? [Y/N]
90【Next step】re-run expecon-identification/robustness for new data → expecon-submission for resubmission
91```
92
93---
94
95**Source:** [`brycewang-stanford/Awesome-Journal-Skills`](https://github.com/brycewang-stanford/Awesome-Journal-Skills) → `Experimental-Economics-Skills/skills/expecon-rebuttal/SKILL.md`
thedixitjain