beamer-presentation

name	beamer-presentation
description	Builds academic economics presentations in LaTeX Beamer that share the same `notation.tex` macros and `tabs/`/`figs/` outputs as the underlying paper, so a single rebuild keeps slides and paper in sync. Defaults to clean themes (`metropolis` / `default` with custom colors), 16:9 aspect ratio, time-budget-aware structure (15/20/45/60/90 min), incremental reveals, and DIME-aligned reproducibility (figures/tables produced by code, never screenshotted). Cross-links to `latex-econ-model`, `latex-tables`, `econ-visualization`, and `academic-paper-writer`. Use when the user asks for conference talk slides, seminar presentations, job-market talks, thesis defense slides, or wants to convert a paper into Beamer.
workflow_stage	communication
compatibility	["claude-code","cursor","codex","gemini-cli"]
author	JonasWeinert
version	2.0.0
tags	["LaTeX","Beamer","presentations","slides","reproducibility","dime"]

Beamer Presentation

Build clean, content-first academic slides in LaTeX Beamer that are reproducibly linked to the paper. The default style is metropolis-themed, 16:9, with the same notation.tex and tabs//figs/ artifacts as the paper.

Operating Principles

1. Slides share the paper's source of truth. Same notation.tex, same tabs/*.tex, same figs/*.pdf. Rerun the analysis once; both PDF and slides update together. Never re-typeset a number on a slide.
2. One claim per slide. If you need more, split.
3. Headline, then evidence. The frame title states the claim; the body shows the evidence. Reviewers in the audience are skimming.
4. Structure to time. A 20-minute talk is fundamentally different from a 90-minute seminar. Pin the time budget early; it determines depth, not just length.
5. Backup, don't overload. Move detail into the appendix; reach for backup slides during Q&A. Better to be asked than to drown the talk in detail.

Decision Policy

This skill follows the repo-wide Agent Policy.

ASK before proceeding (blocking):

1. Time budget (15-20 min conference / 45-60 min seminar / 75-90 min job-market talk / 10-15 min policy briefing).
2. Audience expertise (specialists / broader economists / non-economists).
3. The headline (one sentence the audience must remember).
4. What goes in the appendix vs. the main deck.

DEFAULT + flag (use this default; tell the user how to override):

• Theme metropolis, 16:9 aspect ratio, metroset{numbering=fraction, progressbar=frametitle}.
• Frame title states the claim, not the topic ("Treatment increases outcome by 8%" not "Results").
• Imports notation.tex, tabs/, figs/ from the paper directory — single source of truth.
• Backup slides start at \appendix; one tightly-focused slide per anticipated question.
• End on takeaways; do not end on "Questions?".

DOCUMENT and proceed (write into the deck preamble comment):

• Time-budget allocation (slides per section).
• Which paper artifacts the deck imports.
• Any slide-only number that does not come from tabs/ or figs/ (rare; flag for review).

PROCEED items: slides imported from the paper's notation file; latexmk build; \Cref{} for cross-refs; \alert{} for the headline color (matched to figure palette).

Pre-flight Checklist

• Format. Conference (15-20 min), seminar (45-60 min), job-market talk (75-90 min), policy briefing (10-15 min), thesis defense?
• Audience. Specialists in the same subfield, broader economists, or non-economists?
• What's the headline? State it now in one sentence.
• What evidence is irreducibly part of the talk? (Identification, main result, robustness summary at minimum.)
• What goes in the backup? (Detailed robustness, mechanism evidence, alternative specifications.)
• Visual aids. Are there figures more compelling than tables? Almost always: yes.

Time Budget by Format

15-20 min conference talk
└── ~12-15 content slides
    Motivation (2)  Question (1)  Strategy (2)  Results (4-5)
    Robustness (1)  Conclusion (1)  +backup
    Rule of thumb: 1-2 minutes per slide; never more.

45-60 min seminar
└── ~25-35 content slides
    Add: literature, institutional setting, mechanism, validation,
    extended robustness. Spend longer on the identification frame.

75-90 min job-market talk
└── ~40-55 content slides + 15+ backup
    Audience expects the full chain: motivation, contribution, model,
    data, identification, results, mechanism, robustness, policy,
    conclusion. Plan for 30+ min of Q&A.

10-15 min policy briefing
└── ~6-10 content slides; figures only.
    Lead with the policy-relevant headline; minimize regression tables.

Project Layout (paper + slides share assets)

project/
├── paper/
│   ├── paper.tex
│   ├── tex/
│   │   ├── notation.tex      # SHARED with slides
│   │   ├── preamble.tex
│   │   └── (sections)
│   ├── tabs/                 # SHARED
│   ├── figs/                 # SHARED
├── slides/
│   ├── slides.tex            # \input{../paper/tex/notation.tex}
│   ├── preamble-beamer.tex
│   └── (frame files, optional)
└── code/

The slide deck imports the same notation and references the same tabs/ and figs/ paths. This is DIME's single source of truth principle applied across artifacts.

Frame Recipes

Frame title = the claim, not the topic

BAD:   "Results"
BETTER: "Treatment increases outcome by 8% (p < 0.01)"

Build complex claims incrementally

\begin{frame}{Treatment effect grows over time}
  \only<1>{\includegraphics[width=\textwidth]{figs/fig_event_static.pdf}}
  \only<2>{\includegraphics[width=\textwidth]{figs/fig_event_with_ci.pdf}}
  \only<3>{\includegraphics[width=\textwidth]{figs/fig_event_full.pdf}}

  \medskip
  \only<3>{\textbf{Takeaway:} effect rises monotonically through year~5.}
\end{frame}

Tables on slides should fit in 6-8 rows

For talks, regenerate a slide-friendly table from the same regression with esttab keep()/drop() options or a separate slides_tabs/ folder.

Common Pitfalls

• Reading the slide. The slide is for the audience; the script is for you.
• Tables with 12 rows of fixed effects. Cut to the headline coefficient + clearly labeled FE/N.
• Equations no one will read. Show the estimating equation if it carries the identification; otherwise drop it.
• Figures from screenshots of regression output. Always export from code.
• A title slide with five logos and seven affiliations. One affiliation, one date, one venue.
• "Questions?" as the final slide. End on the takeaway; questions are implicit.
• 9-point font in a footnote no one will read. Cut the footnote.

Additional Resources

• reference.md — extended patterns: theme tweaks, metropolis customization, equation reveal patterns, Q&A appendix structure, multi-author title page, animated coefficient plots.
• examples/ — copyable Beamer files:
  • examples/preamble-beamer.tex — packages, theme, custom colors, frame title style
  • examples/slides.tex — master deck \input{}-ing notation + frame files
  • examples/frame_motivation.tex — motivation/question slide
  • examples/frame_strategy.tex — identification slide with equation
  • examples/frame_results.tex — main result slide with figure + takeaway
  • examples/frame_robustness.tex — robustness summary slide
  • examples/frame_conclusion.tex — takeaways slide

Cross-Skill Routing

• For shared notation and the paper-side LaTeX scaffolding → latex-econ-model skill.
• For slide-friendly versions of regression tables → latex-tables skill (keep(), drop(), alternative mtitles()).
• For figures used on slides → econ-visualization skill (width = 10 in, height = 5.5 in for 16:9).
• For the underlying paper draft → academic-paper-writer skill.

References

• Shapiro, How to Give an Applied Micro Talk — still the canonical reference.
• McKenzie, Tips on Giving an Effective Conference Presentation.
• Beamer User Guide — https://ctan.org/pkg/beamer.
• Metropolis theme — https://github.com/matze/mtheme.
• DIME Analytics, Reproducible Research — single-source-of-truth idea applied to slides.