A 1950s name rule had 4.53 percentage points more disagreement than a 2000s rule on shared names in listed 2010–2017 SSA counts

We tested two fixed rules for recorded sex in US Social Security Administration name counts. One rule used birth years 1950–1959. The other used 2000–2009. Both selected the larger listed sex count for each name. The target years were 2010–2017. On 9,745 shared names and 21,523,995 listed people, disagreement rates were 6.991% and 2.463%. The older rule had 974,644 more disagreements, a difference of 4.528 percentage points. Ten names carried 46.45% of the net difference. We also allowed each absent target name-sex-year count to range from zero to four. With exact matching applicant totals and no limit on names outside the shared dictionary, the sharp paired difference was 3.0472–3.0975 percentage points of all 31,640,853 eligible applicants. It stayed positive in each target year. An added control kept only names whose training majorities could not change under the omission rule. It retained 814,521 additional older-rule disagreements. Full dictionary coverage also favored the recent rule. This study gives a held-out cohort benchmark and exact target omission limits. Historical name change is established prior work. The result does not measure gender identity, a cause of name choice, or service performance. The package contains source records, all rules and scores, proofs, code, figures, checks, and repeat instructions.

name change · historical names · recorded sex · temporal validation · suppressed counts · partial identification · Social Security Administration