Internet Appendix (finman-internet-appendix)

When to trigger

• The main paper is bloated with secondary tables that crowd the headline result
• A referee asked for a check whose result is reassuring but not central to the story
• Variable-construction detail, sample screens, or derivations are too long for the main text
• You are about to submit and need the supplementary materials organized and labeled

The FM internet-appendix bar

FM's "people actually read" brand and its less weight on trivial robustness stance make the internet appendix a strategic tool, not a dumping ground. The main paper should carry only what a reader needs to be convinced; everything that is reassuring but not load-bearing belongs in the internet appendix. The discipline is traceability: every internet-appendix exhibit must map to a specific main-text claim or referee concern, and the main text must point to it where relevant. An appendix that is a disorganized pile of leftover tables undercuts the paper; one that is a clean, indexed companion makes the main paper lean and the author look in command of the evidence. (FM's exact archiving/hosting and data-and-code policy — 待核实 against the Wiley author guidelines at submission time.)

What goes where

Organizing sequence

1. Index it. Open the appendix with a short table of contents mapping each IA exhibit to the main-text section or referee point it supports.
2. Cross-reference both ways. The main text says "see Internet Appendix Table IA.3"; the IA exhibit's note says which main-text claim it backs.
3. Keep numbering distinct. Use an IA prefix (Table IA.1, Figure IA.1) so the main and supplementary sets never collide.
4. Make each IA exhibit self-contained. Same standard as the main exhibits — sample, units, clustering, period in the note — because referees read the IA without re-reading the paper.
5. Resist scope creep. If an IA exhibit is doing real persuasive work for a binding concern, promote it to the main paper; if it persuades no one of anything, cut it.

Checklist

• The internet appendix opens with an index mapping each exhibit to a main-text claim or referee point
• Every IA exhibit is cross-referenced from the main text where relevant
• IA numbering is distinct (IA.1, IA.2 …) and never collides with the main set
• Each IA exhibit is self-contained (sample, units, clustering, period in the note)
• Load-bearing material was promoted to the main paper; dead weight was cut, not parked
• Full variable definitions and sample screens live here, summarized in the main text

Anti-patterns

• An internet appendix that is an unindexed pile of leftover tables
• IA exhibits referenced nowhere in the main text (orphans)
• A genuinely decisive check hidden in the appendix where the editor may miss it
• IA numbering that overlaps the main paper and confuses cross-references
• Treating the appendix as a place to satisfy the urge for trivial robustness FM under-weights
• IA exhibits with no self-contained notes, forcing the referee back into the paper

Worked vignette (illustrative)

A draft's main paper has eighteen tables, several of them winsorization and control-set sweeps; the headline gets lost. The FM fix: keep five exhibits in the main paper (summary stats, the headline table, the event-study figure, the one Oster sensitivity check, and the economic-magnitude exhibit), move the thirteen sweeps and secondary subsamples to an indexed internet appendix (Tables IA.1–IA.13), and add cross-references both ways. One placebo table that actually rules out a confound gets promoted into the main paper. The main paper is now readable in one sitting and the appendix is a clean, navigable companion.

The promote / park / cut decision

For every candidate supplementary exhibit, make a deliberate call rather than defaulting to "park it":

• Promote to the main paper if it defuses a binding concern (the editor or a likely referee would reject without it). Decisive evidence buried in an appendix is a wasted asset.
• Park in the internet appendix if it is reassuring and a referee may want it, but no reader needs it to be convinced (alternative measures, secondary subsamples, full variable tables).
• Cut entirely if it persuades no one of anything — FM's under-weighting of trivial robustness means a sweep that varies nothing material is dead weight even in the appendix.

This three-way call is the actual craft; an appendix that is purely "everything that didn't fit" signals the author never made it.

Referee pushback mapped to the appendix fix

• "The paper is too long / I lost the thread." → Move secondary and sweep tables to an indexed internet appendix; keep the main paper lean.
• "Where is the check you mention?" → Add two-way cross-references between the text and each IA exhibit.
• "This appendix is a mess." → Open with an exhibit→claim index and give every IA exhibit a self-contained note.

Code and data alongside the appendix

The internet appendix usually travels with the replication materials. Keep them in sync from the first submission: the IA exhibits should be reproducible from the deposited code, variable definitions in the IA must match the code's construction, and any data-access restrictions (vendor licenses on Compustat, CRSP, governance feeds) should be stated so a referee or the editorial office knows what can and cannot be shared. (FM/Wiley's exact data-and-code policy — 待核实; confirm before depositing.)

Output format

【Main vs. IA split】load-bearing in main; reassuring in IA? [Y/N]
【Index】IA opens with an exhibit→claim map? [Y/N]
【Cross-references】both directions present? [Y/N]
【Numbering】distinct IA prefix, no collisions? [Y/N]
【Promotions/cuts】decisive checks promoted; dead weight cut? [Y/N]
【Policy check】Wiley data/code & hosting policy verified or 待核实
【Next skill】finman-writing-style