tm-vocabulary · papers · online appendix · the fifty themes · IPO viewer · 500-theme explorer · code
Online appendix to two papers built on the same corpus and measure: An Event-Dated Corpus of US Trademark Prosecution and a Two-Sided Measure of Vocabulary Position (SSRN 7520758), which builds and validates the measure, and Arrows in Their Backs: Vocabulary Lead and Product Survival in the US Trademark Record, which asks whether being early pays. Both develop the earlier combined working paper, SSRN 6954598.
The measure in brief. Every goods/services description is reduced to a mix of fifty themes (lists of words that tend to appear together; what they are and how they were found). Each filing's theme mix is compared with its own Nice class's filings in the five years before and the five years after its filing date. Atypicality is the average of the two comparisons: how unusual the language is for its industry. Lead is their difference: positive when the industry's language later moved toward the filing's, negative when the filing was written in language the industry was leaving.
Everything here is computed on the same corpus and the same per-filing reference windows as the papers. The papers report pooled estimates because they are what the design identifies cleanly; this appendix reports the class level because the pooled numbers average over real and large industry variation, and a reader who cares about one industry should be able to see it.
The contrast is the failure rate at the five-year proof of continued use of the most leading fifth of registrations minus the most lagging fifth (2002–2018 registrations), in percentage points: a positive number means leading marks were cancelled more often. Panel A of each class figure is a linear probability model with registration-cohort fixed effects and log description length held fixed, estimated within that class alone — simpler than the papers' pooled specification, which adds drafting and filer controls and clusters errors on the owner (pooled: +1.4 points, t = 8.3; with fixed effects alone, +2.3). Panel B is raw registration completion rates by quintile; panel C places the class in the cross-industry distribution. Shaded bands and whiskers are 95% intervals; the ± in the table is a 95% interval.
Per-class estimates are noisier than the pooled figure by construction, and small classes are noisy enough that individual signs should not be read as findings. The cross-industry exhibits below are the honest summary: 33 of 44 classes show a positive raw contrast, and 32 of 44 do so under the cohort-fixed-effects specification, which is the claim the papers make. The spread around it is what this appendix shows, and it is wide: the fixed-effects contrast runs from about −6pp to +19pp across classes.
Nice 023 (yarns) falls below the volume floor the pooled analysis uses and has no raw entry; its figure is still generated.
The scatter matters because the two obvious mechanical explanations for cross-class variation are class size (more registrations, tighter estimate, possibly different sign) and base failure rate (a class where most marks die has less room to move). Neither organises the variation.
Open the IPO viewer — every class’s registrations (1996–2018, a longer span than the 2002–2018 cohorts the estimates use) as dots on the lead/atypicality plane, colored by the furthest outcome the record reaches: cancelled at the five-year proof of continued use, passed it, owner in SEC financial reporting, owner with an IPO marker. Later outcomes draw on top. Filter by Nice class and by registration year range.
per_class_estimates.csv — one row per class:
scored filings, registrations, base failure rate at the five-year proof, raw
and fixed-effects leading-minus-lagging contrasts with standard errors, and
registration completion at both tails of each axis.
Click a column heading to sort; click again to reverse. The Nice number links to that class’s figure below.
| Nice | Industry | Registrations | Base fail | Leading − lagging, raw (pp) | Leading − lagging, cohort FE (pp) |
|---|---|---|---|---|---|
| 1 | Chemicals | 55,166 | 39.2% | +1.9 ± 1.3 | +2.93 (0.66) |
| 2 | Paints | 16,070 | 39.9% | −0.3 ± 2.4 | −1.14 (1.24) |
| 3 | Cosmetics & Cleaning | 118,261 | 54.4% | +7.8 ± 0.9 | +7.32 (0.46) |
| 4 | Lubricants & Fuels | 16,708 | 48.0% | +7.1 ± 2.4 | +4.98 (1.22) |
| 5 | Pharmaceuticals | 124,888 | 53.4% | +2.8 ± 0.9 | +2.06 (0.45) |
| 6 | Metal Goods | 44,132 | 39.3% | +3.2 ± 1.4 | +2.80 (0.74) |
| 7 | Machinery | 74,433 | 39.1% | +3.8 ± 1.1 | +3.76 (0.57) |
| 8 | Hand Tools | 25,976 | 43.6% | +9.6 ± 1.9 | +7.89 (0.96) |
| 9 | Software & Electronics | 452,911 | 52.4% | +1.6 ± 0.5 | +3.28 (0.23) |
| 10 | Medical Apparatus | 70,401 | 45.5% | +9.5 ± 1.2 | +9.72 (0.59) |
| 11 | Lighting & Heating | 75,376 | 46.1% | +9.4 ± 1.1 | +7.52 (0.57) |
| 12 | Vehicles | 55,734 | 45.3% | +1.0 ± 1.3 | +0.39 (0.67) |
| 13 | Firearms | 9,516 | 37.3% | +2.7 ± 3.0 | −1.01 (1.56) |
| 14 | Jewelry | 55,321 | 55.6% | +4.1 ± 1.3 | +3.67 (0.67) |
| 15 | Musical Instruments | 7,622 | 40.5% | +0.1 ± 3.5 | −1.49 (1.78) |
| 16 | Paper & Printed Goods | 167,752 | 50.2% | +2.5 ± 0.8 | +1.40 (0.39) |
| 17 | Rubber & Plastics | 23,857 | 38.6% | +3.4 ± 2.0 | +2.81 (0.99) |
| 18 | Leather Goods | 58,194 | 55.0% | +5.0 ± 1.3 | +3.27 (0.66) |
| 19 | Building Materials | 34,332 | 44.0% | +1.1 ± 1.7 | +0.46 (0.85) |
| 20 | Furniture | 63,640 | 48.6% | +6.2 ± 1.2 | +3.35 (0.63) |
| 21 | Household Utensils | 66,993 | 50.2% | +7.7 ± 1.2 | +5.66 (0.61) |
| 22 | Cordage & Fibers | 9,556 | 43.5% | +3.4 ± 3.1 | +3.58 (1.60) |
| 23 | Yarns | 2,933 | — | — | — |
| 24 | Textiles | 32,519 | 51.1% | +7.4 ± 1.7 | +7.92 (0.88) |
| 25 | Clothing & Footwear | 252,116 | 58.3% | +5.0 ± 0.6 | +4.72 (0.31) |
| 26 | Lace & Embroidery | 13,857 | 53.8% | +4.2 ± 2.6 | +3.69 (1.34) |
| 27 | Carpets | 10,298 | 46.7% | +6.3 ± 3.0 | +6.18 (1.55) |
| 28 | Games & Sporting Goods | 111,839 | 55.6% | −0.5 ± 0.9 | −1.27 (0.47) |
| 29 | Meats & Processed Foods | 60,893 | 49.7% | +9.5 ± 1.3 | +9.15 (0.68) |
| 30 | Staple Foods | 102,594 | 51.9% | +2.9 ± 1.0 | +2.99 (0.51) |
| 31 | Agricultural Products | 32,292 | 43.9% | +2.7 ± 1.7 | +2.89 (0.87) |
| 32 | Beer & Soft Drinks | 48,341 | 53.3% | −9.1 ± 1.4 | −5.84 (0.72) |
| 33 | Alcoholic Beverages | 56,974 | 42.5% | +5.4 ± 1.3 | +1.32 (0.69) |
| 34 | Tobacco | 16,112 | 54.2% | +25.6 ± 2.3 | +19.03 (1.20) |
| 35 | Advertising & Retail | 428,569 | 54.4% | −0.6 ± 0.5 | −0.67 (0.24) |
| 36 | Insurance & Finance | 183,011 | 49.5% | −2.2 ± 0.7 | −0.58 (0.37) |
| 37 | Construction & Repair | 86,335 | 47.3% | −3.0 ± 1.1 | −3.15 (0.54) |
| 38 | Telecommunications | 64,322 | 60.6% | −1.3 ± 1.2 | −1.04 (0.61) |
| 39 | Transport & Storage | 54,593 | 49.2% | −3.9 ± 1.3 | −3.03 (0.68) |
| 40 | Material Treatment | 43,966 | 47.5% | −2.0 ± 1.5 | −2.32 (0.76) |
| 41 | Education & Entertainment | 381,509 | 53.1% | −0.6 ± 0.5 | +0.18 (0.25) |
| 42 | Scientific & Tech Services | 268,366 | 54.2% | +8.8 ± 0.6 | +7.76 (0.30) |
| 43 | Hotels & Restaurants | 86,774 | 48.6% | +4.3 ± 1.1 | +3.27 (0.54) |
| 44 | Medical & Beauty Services | 92,827 | 51.8% | +0.9 ± 1.0 | +0.72 (0.52) |
| 45 | Legal & Personal Services | 64,206 | 53.1% | −1.1 ± 1.2 | −2.65 (0.63) |
