Skip to main content
← Back to Research

Phase 7c — Bass Diffusion Fits

Separating broadcast-driven from peer-driven adoption across 60,470 names via Bass diffusion (p, q) parameter fits.

By Namesake ResearchApril 13, 2026

Phase 7c -- Bass Diffusion Fits

Research question. For each name's adoption curve, what are the Bass (p, q) parameters, and does the p vs q mix differ by cultural-event type?

Summary

  • **104,819** unique names in annual_panel
  • **60,470** names fitted successfully (57.7%)
  • **44,349** names skipped (no valid segment >= 5 years or fit diverged)
  • Minimum segment length: **5 years**

Parameter distributions

Innovation coefficient (p)

StatValue
mean0.050979
std0.094888
min0.000001
25%0.006853
50%0.018949
75%0.047309
max1.000000

Imitation coefficient (q)

StatValue
mean0.118036
std0.138361
min0.000001
25%0.019606
50%0.088303
75%0.163316
max2.690257

Fit quality (R-squared)

StatValue
mean0.9666
std0.0689
min0.0154
25%0.9651
50%0.9916
75%0.9976
max1.0000

Classification breakdown

ClassificationCountPercentage
broadcast15,78726.1%
peer25,81842.7%
mixed16,22226.8%
unfit2,6434.4%

Classification rules:

  • **broadcast**: p > 0.01 and p/(p+q) > 0.4 (innovation dominant)
  • **peer**: q > 0.1 and q/(p+q) > 0.7 (imitation dominant)
  • **mixed**: both p and q non-trivial
  • **unfit**: no suitable segment or fit diverged

Notable examples

Broadcast

  • Denzil (p=0.0188, q=0.0189, R2=1.000, segment 1914-1966)
  • Burrell (p=0.0228, q=0.0292, R2=1.000, segment 1912-1956)
  • Harmon (p=0.0236, q=0.0278, R2=1.000, segment 1912-1962)
  • Franklyn (p=0.0159, q=0.0231, R2=1.000, segment 1913-1974)
  • Suzana (p=0.0253, q=0.0059, R2=1.000, segment 1970-2020)

Peer

  • Barbara (p=0.0111, q=0.1045, R2=1.000, segment 1926-1967)
  • Glenda (p=0.0138, q=0.1261, R2=1.000, segment 1934-1970)
  • Jackson (p=0.0136, q=0.1318, R2=1.000, segment 1996-2024)
  • Angela (p=0.0132, q=0.1430, R2=1.000, segment 1957-1991)
  • Susan (p=0.0180, q=0.1413, R2=1.000, segment 1942-1972)

Mixed

  • Waymon (p=0.0085, q=0.0537, R2=1.000, segment 1911-1984)
  • Nelda (p=0.0075, q=0.0917, R2=1.000, segment 1914-1965)
  • Melinda (p=0.0116, q=0.0942, R2=1.000, segment 1949-1993)
  • Brian (p=0.0131, q=0.0891, R2=1.000, segment 1954-2004)
  • Ian (p=0.0074, q=0.0666, R2=1.000, segment 1974-2024)

Stratification by event type

Event typeNp (median)q (median)R2 (median)Broadcast %Peer %Mixed %
book_character40.009170.039790.96825.0%25.0%50.0%
celebrity_birth220.004260.092670.9999.1%36.4%54.5%
celebrity_naming320.016900.036970.99840.6%28.1%25.0%
film_character3250.012210.067080.99823.4%33.8%37.5%
music_chart930.015090.088620.99817.2%34.4%48.4%
news_event1680.012880.083750.99813.7%35.7%45.2%
royal_event60.003160.039600.9960.0%0.0%100.0%
sports_moment830.027760.116750.99827.7%47.0%24.1%
tv_character3220.012250.087970.99811.8%37.6%43.5%
unknown60.026940.176130.9950.0%33.3%66.7%
video_game80.009220.063220.9940.0%25.0%75.0%

Diagnostic expectation. Broadcast-dominant names should be disproportionately associated with film/TV events (innovation via media exposure). Peer-dominant names should appear more in categories where social transmission matters (celebrity_naming, celebrity_birth).

---

Generated by `scripts/python/research/phase7_timeseries/bass_fit.py`