---

title: "Lieberson Null Model Results" description: "Full-scale Phase 5 analysis: 1.95M threshold rows across 46,412 names, 145 years of SSA data. Neutral drift + phonetic fashion null models with Wright-Fisher calibration." publishedAt: "2026-04-10" status: published authors: ["Mike West"] featured: true

Phase 5 — Lieberson Null Model Results

Generated: 2026-04-12 00:48 UTC Pipeline step: Phase 5 (null model) — body regenerated from null_model_summary.json + null_model_thresholds.parquet.

Key Findings

• Total name-year rows (thresholds parquet): 1,950,660
• Unique names (summary JSON): 46,412
• Names classified (≥10 SSA years, bucketed): 46,412
• Innovation rate (μ): 0.050455
• N_e (female): 9,852
• N_e (male): 22,320
• Average observed turnover per decade: 23.4%
• Share of name-years beating neutral drift at p95: 4.57%
• Share of name-years beating neutral drift at p99: 2.83%
• Share of name-years beating phonetic null at p95: 5.27%
• Share of name-years beating phonetic null at p99: 3.74%

Classification Breakdown (table)

Buckets use the share of years (aggregated across sexes) where beats_neutral_p99 is true, for names with at least 10 distinct SSA years:

Bucket	Names
drift-consistent	39,055
partially-cultural	6,743
culturally-influenced	390
strongly-cultural	224
Total	46,412

Methodology Notes

• Parquet columns used: name, year, beats_neutral_p99 on null_model_thresholds.parquet.
• Year-level beat: for each (name, year), beats_neutral_p99 is the logical OR across sex rows for that name-year.
• Per-name fraction: mean of year-level beats over all years with data for that name (among names with ≥10 years).
• Bucket thresholds: <5% drift-consistent; 5–19% partially-cultural; 20–39% culturally-influenced; ≥40% strongly-cultural.
• Summary metrics (μ, N_e, turnover, % beating nulls) are read from null_model_summary.json.