The sample and approved exclusions

The All readers analysis includes 1,284 reader records from 2022 and 771 from 2026. It uses the full supplied 2022 dataset, including responses submitted after the original write-ups. Three approved duplicate records and four timestamp-only blank records were removed. The author’s final response is excluded from all reader statistics.

The less-certain duplicate pairs and other flagged responses remain included. Counts refer to response records, not independently verified unique people.

The age-label mistake

Respondents’ actual interpretation cannot be verified from the export. No attempt is made to assign corrected generation names or infer exact birth years.

Percentages, ratings, and missing answers

Results are unweighted. Percentages and means use the nonmissing answers to each question, so denominators vary. Missing answers are not treated as zero or as the midpoint.

Political self-placement runs from 1 (very liberal) to 5 (very conservative). The 246 ratings run from 1 (maximum disapproval or disagreement) to 5 (maximum approval or agreement). A change in percentage shares is reported in percentage points; a change in a mean rating is reported in scale points. Standard deviations describe response variation, not uncertainty around a mean.

Category interpretation

Race and career answers can contain multiple selections, so shares can sum to more than 100%. Combined tech and academia shares count a respondent only once within each combined group. The previously documented write-in mappings are retained; broad categories are not newly inferred.

Income brackets are nominal and not adjusted for inflation. The children chart uses biological children, not stepchildren or adopted children. The sexual-partner category 10–19 is treated as a range, not a calendar date.

Country (“Which country are you from?”) is self-reported and not verified as citizenship or residence. Conventional spellings are consolidated, including Netherlands, Mexico, and Turkey/Türkiye. Historical or regional names such as Siam, Swaziland, and EU are retained as given. Ambiguous multi-country and residence-only answers are pooled, with no inference about origin.

Discovery (“How did you discover Richard Hanania’s newsletter?”) is recalled, not tracked referral or growth data. Variants are consolidated: the same explicit source, Twitter/X, Substack (feed, Notes, recommendations), and family or friends. Named sources remain separate; multiple or unclear answers are grouped as Other. “Don’t remember” is a valid answer and stays in the denominator; blank answers are excluded. The 2022 and 2026 results are separate self-selected pools.

Not a matched-person study

These are self-selected respondent pools, not probability samples of all newsletter readers. There is no stable respondent identifier for reliable person-level matching across years. The question asking whether someone took the earlier survey does not itself establish a matched panel.

Differences may reflect audience composition, participation, or changes in individual views. These tables do not separate those explanations. Results are descriptive and exploratory; no statistical-significance claims are made.

Sources and privacy

The results are presented as supplied in the aggregate survey tables.

This website and its downloads contain aggregate reader results and Richard’s scores for the twenty featured comparisons. No individual reader response records or respondent-level audit files are included. The website does not collect submissions or use analytics, external fonts, or third-party scripts.

Joint age and sex cohorts; small samples

Male and Female use the exact selected sex answers only; sex is not inferred from free text. Other / self-described and missing-sex answers remain in all readers only.

The independent selectors apply both restrictions jointly using verified aggregates from actual selected answers, never estimates from marginal totals. Nonstandard and missing-sex answers remain eligible for All readers within a selected age bracket. Missing ages remain in All ages only. The numeric age brackets are used as listed; no generation identity is inferred and four years are not added. Reset filters restores All readers and All ages.

Means and sample standard deviations use verified aggregate counts, sums, and squared sums, with N minus one for sample variance. Missing answers are not zero. Statistics based on fewer than five valid answers are unavailable; answer counts remain visible and overall results are never substituted. Samples below 30 are flagged as small. All 246 items remain available in Ratings; ranks use eligible items within the selected cohort and changes require available means in both years. Disagreement rankings use the supplied independent cohort top twenties across all items Richard rated; his scores remain fixed.

These are descriptive comparisons between respondent pools, not matched-person changes, causal effects, or automatically statistically significant findings.