USA Baby Names Analysis

• Top-10 share fell sharply: Female 24.6% → 7.1%; male 44.2% → 8.0%.
• Top 100 tells the same story: 76.5% → 31.0% for females; 80.1% → 38.0% for males.
• Male naming was more concentrated but converged: The 1880s male top-10 share (44%) was nearly double the female (25%). By 2024 they've nearly met (8.0% vs. 7.1%).
• Steepest decline post-1960: Gradual 1880–1960, then a sharper break coinciding with the cultural shifts toward individualistic naming.

For a given year, let $p_i$ be the proportion of births with name $i$. Shannon entropy is:

$$H = -\sum_{i=1}^{k} p_i \log_2 p_i$$

measured in bits. Intuitively, $H$ is the average number of yes/no questions you'd need to guess a random baby's name. The effective number of names — the number of equally-popular names that would produce the same $H$ — is $2^H$. An entropy of 10.5 bits ≈ 1,448 effective names.

• Diversity grew sharply: Female 7.6 → 11.1 bits; male 6.9 → 10.5. 3–4 extra bits of surprise ≈ 8–16× effective variety.
• Female names have always been more diverse, and the gap widened over time.
• Unique-name count diverges from entropy after 1980 — the 20k+ unique female names today include many used only a handful of times.
• A brief 1950s plateau: The baby-boom era briefly concentrated choice around Linda, Mary, Barbara / James, Robert, Michael.

• Male top names have always lasted longer: In the 1880s, male top-10 half-life was ~45 years vs. ~32 for females.
• Both declining: By the 1990s, male ~19 years, female ~15.
• Steepest female drop: 1920s–30s, from ~33 to ~20 years in two decades — a post-WWI cultural shift toward novelty.
• The gap is narrowing: 13-year male-female gap in the 1880s → 4 years by the 1990s.
• Chart cuts off at the 1990s because newer names haven't finished decaying yet.

• Female names got longer, then shorter: 5.4 chars (1880) → 6.4 peak (~1990) → 5.9 by 2024.
• Male names are stable: 5.5–6.0 chars across the whole period, mild 1990–2000 peak from Christopher/Alexander/Nicholas.
• Strong correlation: r = 0.954 between male and female length year-by-year. Year-over-year changes correlate at r = 0.706.
• Crossover around 1910: Female names have been longer than male ever since.
• What drove the 1990 peak: Brittany, Ashley, Jessica, Samantha surged as Amy, Lisa, Tracy collapsed. What reversed it: Christopher fell from 2.6% to 0.3% of male births, replaced by Liam, Noah, Leo.

• Character trigrams (3-letter subsequences) capture the sound of each era more precisely than starting letters alone.
• Trigram dominance collapsed alongside name-level concentration — no more "mar".
• Male trigrams are stickier — slower, broader waves vs. the sharper crashes of the female chart.

• Strict definition matters. Loose definitions catch noise — Liam, given to ~22K boys and ~500 girls in 2024 (~2% female share), is not meaningfully gender-neutral. The 20% minority-share threshold demands substantive cross-gender use; the 100-count floor rules out one-off statistical noise.
• An early-1900s male spike (~3.3% in 1910) driven almost entirely by Willie — hugely popular and substantively cross-gendered.
• Flat 1920–1980 plateau at ~2% for both genders, then a strong post-1980 acceleration to ~5–5.5% by the early 2020s.
• Diversity expansion drove the post-1980 explosion. As parents picked from a wider pool, more names simultaneously cleared the substantive cross-gender bar.

Top by single-year minimum prevalence (highest and most evenly balanced):

• Willie (1910): 1,796 F + 2,897 M — 0.45% F / 1.49% M
• Willie (1900): 1,351 F + 2,113 M — 0.45% F / 1.40% M
• Willie (1909): 1,549 F + 2,175 M — 0.45% F / 1.33% M
• Taylor (1992): 14,952 F + 8,239 M — 0.81% F / 0.41% M
• Jordan (1997): 7,166 F + 14,761 M — 0.41% F / 0.78% M

Willie's dominance reflects how concentrated naming was in the early 20th century: a single hit name could simultaneously be top-tier and substantively cross-gendered. The modern era is more diffuse — Taylor and Jordan are the cleanest contemporary analogues, but neither reaches Willie's combined prevalence.

• Stable neutrals (top row). Jessie has been gender-neutral for 122 of the last 145 years, bobbing in the 30–80% female range. Jamie (since 1910) and Casey (since 1968) show the same pattern — modest drift, never an exit.
• M → F flippers (second row). Madison passed through the cross-gender band in just 3 years (the Splash effect). Lauren and Lindsay show the same shape on different timescales.
• F → M flippers (third row). Rarer and quieter. Lavon and Robbie drifted gradually from ~90% female to male-dominant over decades. Samar is the modern example.
• Brief crossers (bottom row). Aidan, Flynn, Juno briefly bumped above the 20% threshold for a year or two, then receded — they never flipped, just brushed against the band.
• Pre-GN history is universal: only 9 of 1,653 names were gender-neutral from their first year of meaningful use. Parker had 130 years of male-dominant use before crossing into GN territory.

• MI declined 27%: 0.64 bits (1880) → 0.47 bits (2024).
• But strong polarities persist: -la, -ia, -na, -da, -sa remain near-100% female; -rt, -hn, -rd, -es near-100% male.
• Big shifts in the middle: -yn went from 10% female to 80%. -ee and -ah similarly shifted toward female.
• 1950s–60s bump corresponds to the baby boom's traditional, strongly-gendered naming (Linda/Barbara vs. James/Robert).

• Female migrations: -yn, -ee, -ah, -gh all S-curved toward female dominance.
• Transfer is bidirectional: Endings migrating toward male roughly match those migrating toward female.
• S-curve suggests tipping-point dynamics: Once an ending "sounds" like a gender to enough people, the remaining cross-gender names feel incongruent and get replaced.

• CV dominates: ~42–46% for females, ~36–39% for males — the Ma-ry, Jo-hn rhythm.
• Female names are more liquid: Higher CV, lower CC share — a persistent pattern.
• Consonant clusters declining: CC fell 16% → 12% (female), 22% → 14% (male).
• Vowel sequences rising: VV share has nearly doubled since 1980 (Mia, Ava, Aria, Noah, Liam, Isaiah).
• Male convergence: Male phonotactics are drifting toward the female pattern — smoother, more vowel-rich.

• J dominance in male names: Peaked mid-century, declined since the 1990s.
• The rise of A: Surging since 1960s–70s for both sexes; dominant for girls today.
• M's endurance for girls: Consistently large share across the whole period.
• K's boom and bust: Karen, Kimberly, Kelly, Kristen surged 1960s–80s, then faded.
• Male naming is more concentrated — fewer letters command large shares at any time.

• Strong negative correlations reveal generational shifts — as one letter-block faded, another rose to replace it.
• Male names show a clear "old vs. new" block structure.
• Persistently niche letters: Q, U, X, Z show weak correlations with everything — they've never been popular enough to participate in large-scale shifts.

• Mary held #1 longest: ~80 years for girls.
• Male #1s rotate more slowly than females overall but at lower absolute counts today.
• Modern churn is rapid: Emma, Olivia, Liam cycle in and out faster than classic names.
• Raw counts dropped even as total births held: Michael's ~92K peak vs. Liam's ~22K.

• Edges connect names whose Metaphone encodings are within edit distance 2. Node size = total all-time births.
• Dense cores = sound families recycled across generations.
• Male graphs are tighter: Popular male names draw from a smaller phonetic repertoire.
• Isolated nodes are phonetically distinctive: Kimberly, Rebecca, Virginia, Patricia (F); Christian, Richard, Scott, Albert (M).

• Female contagion rose 7×: ~10 new names per new top-100 trigram (1900s) → ~71 (2000s).
• Male contagion rose modestly: ~3 → ~20.
• Much of the diversity is remix: Unique-name growth tracks systematic phonetic recombination, not random invention.

• A clear hierarchy: name entropy > phoneme-sequence entropy > trigram entropy.
• The crossover matters: Name entropy overtook phoneme entropy around 1960 — the start of the "creative spelling" era.
• Pre-1960, trigram entropy was actually higher than name entropy for females — a few names (Mary, Dorothy, Helen) dominated but contained diverse trigrams.

• Trigram space filled up fast: Novelty rates dropped to near-zero by the 1920s–30s.
• Beginnings slightly more innovative than endings, every era — endings carry gender/phonetic identity.
• Modern near-zero novelty + exploding name count is the signature of remix.

• Heavy right skew: Most Soundex codes map to 1–3 distinct names; a few map to hundreds.
• Top groups share simple phonetic frames: S050, J050, K050, M020 — common consonant-vowel templates.
• 104,819 names → ~13,000 codes: Most "new" names sit on top of an existing phonetic slot.

• Four layers of diversity: name > phoneme-sequence > trigram > phoneme.
• Phoneme-sequence entropy crossed below name entropy around 1960 — the quantitative signature of the creative-spelling era.
• Phoneme trends: "AH" (schwa) and "N" dominate female names; "R" declined (Mary/Margaret/Barbara fading); "L" rose (Olivia, Emily, Ella). For males, "N" rose steadily, tracking Aiden/Jayden/Ethan.
• Phoneme-identical groups are huge: John/Jon = 5.4M births. Same sound, different spelling.

Cumulative adoption:

$$F(t) = \frac{1 - e^{-(p+q)t}}{1 + (q/p) \cdot e^{-(p+q)t}}$$

$p$ controls how quickly first adopters appear (media-driven early awareness). $q$ controls how steeply the curve accelerates (word-of-mouth amplification). A high $q/p$ ratio means the name spread through social networks; a high $p$ means a single external spark drove it.

• Imitation dominates: Median $q/p \approx$ 10–20 across 175 fit names.
• Peak in 1940s–60s: The conformist era, when Michael/Jennifer/Linda spread through tight social networks.
• $p$ stayed flat: External media influence has been roughly constant.
• $q$ declined post-1960: Individualist culture = weaker social contagion.
• Male and female diffusion patterns are remarkably similar — same social dynamics, different specific names.

• Madison is the clearest cultural fingerprint — from near-zero to #2 female name after Splash (1984).
• Drops can be steeper than rises: Katrina (hurricane, 2005), Isis (terrorist group, 2014–15), Alexa (Amazon, 2015), Karen (meme, 2020).
• Algorithmic spike detection found historical events: Dewey +6.5× in 1898 (Admiral Dewey), Grover +5.8× in 1884 (Cleveland elected), Woodrow +8× in 1912 (Wilson), Marlene +8.9× in 1931 (Dietrich's Hollywood debut). Elections and military heroes were the movies of their era.
• Modern spikes are numerous but smaller — a more diverse naming pool absorbs them.

• West is most diverse, South is most concentrated — ranking stable across 1910–2024.
• Interior states peak before coastal states — average lead of 1.9 years across 80 recent hits. 44 peaked interior-first, 3 coast-first, 33 tied.
• Regional signatures (2015–2024 overrepresentation):
  • Northeast: Rachel, Esther, Maeve, Sara, Nicholas, Sienna — heritage and European-influenced.
  • South: Landyn, Hunter, Ryleigh, Kingston, Khloe, Messiah, Bryson — creative spellings, aspirational, religious.
  • Midwest: Beckett, Emmett, Graham, Ryker, Griffin, Bennett, Lincoln — surname-as-first-name.
  • West: Emiliano, Ximena, Damian, Emilio, Jesus, Santiago, Natalia, Camila — Hispanic/Latino influence.

• Spread speed compressed 3×: 10–12 years (1960s) → 3–4 years (2010s) for the 5→30-state transition.
• Small but real neighbor effect: Neighboring states adopt 0.4 years earlier than non-neighbors.
• Small states are trend incubators, not trend setters: A few babies with a novel name register as a high share of a small state, making early adoption visible. The actual trend mechanism is national.