Jfm Tables Figures

Tables and Figures (jfm-tables-figures)

When to trigger

• The main table reports many liquidity coefficients but the reader cannot see the headline result
• A market-structure event has no event-study figure showing the spread/depth path around the date
• Units are ambiguous (basis points vs. cents vs. percent; per-share vs. per-dollar) across exhibits
• Intraday or cross-sectional patterns are described in prose where a figure would carry them
• Summary statistics for the liquidity measures and the sample are missing or buried

What a JFM exhibit set must carry

Microstructure papers live or die on whether the trading-process pattern is visible. The exhibit set should let a referee verify the mechanism without re-reading the text.

Craft rules JFM referees reward

1. State units and horizon on every liquidity exhibit. "Effective spread (bps)", "5-min permanent impact", "depth at touch (shares)". Inconsistent units between tables is an immediate credibility hit.
2. Make the headline coefficient unmissable. One column or one panel is the contribution; the rest is support. Bold the takeaway in the caption, not with significance stars on every cell.
3. Report standard errors and the clustering in the table notes — say it is two-way stock×time / Newey-West, not just "robust SE."
4. Event studies need leads. Show the periods before the market-structure change so pre-trends are visible; without them the design is unfalsifiable on the page.
5. Self-contained notes. Each table/figure note states sample, period, measure construction, and inference — a referee should not have to hunt.
6. Plot what is intrinsically dynamic (intraday U-shape, impulse responses, order-book replenishment) rather than tabulating it.

Worked exhibit set (illustrative)

For a tick-size-change paper, a clean four-exhibit core: Table 1 summary stats (quoted/effective spread in bps, depth in shares, daily volume, N stocks × days, with percentiles not just means); Figure 1 event study — effective spread plotted from t−10 to t+10 days with confidence bands, leads visibly flat, the level shift at t=0 obvious; Table 2 the DiD with the single treatment coefficient in column 1 isolated, columns 2-4 adding controls, notes stating two-way stock×day clustering; Figure 2 heterogeneity — the spread effect by adverse-selection quintile, rising monotonically, showing the mechanism. A reader can verify the mechanism from the exhibits alone. Anything beyond this goes to the Internet Appendix.

Summary statistics that pre-empt questions

The summary-statistics table is the first thing a referee reads, and a good one answers questions before they arise. For a microstructure paper, report not just means but the distribution (percentiles, SD) of each liquidity measure, the cross-sectional spread across stocks, and the sample dimensions (N stocks, N days, N trades). Show the liquidity measures' pairwise correlations so the reader sees whether your alternative constructs actually agree. Flag the skew that liquidity variables always exhibit — an unreported fat tail invites the "is this driven by a few illiquid names?" objection that a percentile table would have closed.

Reading the exhibit like a referee

A microstructure referee scans, in order: Are the units sensible (a 200-bps "effective spread" on a liquid large cap is a red flag)? Do the event-study leads look flat, or is there a pre-trend the authors did not mention? Does the magnitude make economic sense relative to the average spread? Is the clustering in the notes, and does it match the panel structure? Build the exhibits to pass this scan — surprising-but-unexplained numbers and missing leads are what trigger the first skeptical comment.

Captions and notes do real work

In microstructure papers the table/figure notes carry as much information as the cells. A complete note states: the sample (universe, period, N), the exact construction of each measure (effective spread formula, impact horizon, sign rule), the estimator and the standard-error / clustering scheme, and the units. A reader should be able to understand an exhibit without leaving it for the text. Vague notes ("robust standard errors in parentheses") invite the clustering objection; precise notes ("standard errors two-way clustered by stock and trading day") close it on the page. Treat the note as part of the result, not boilerplate.