← No Poll Asks to Abolish Social Security

Social Security online versus the polls: the poll items, the codebook, the counts and the limits

Companion to the editorial, part two of the desk’s Boomers poll audit, and the full record; the editorial’s figure carries the poll cells and this page carries the stance counts by platform, query type and age signal. Poll numbers are from the surveys named; the online counts, the codebook and the hand checks are the desk’s own and are labelled as such. Everything here was collected on 11 October 2026 (Pacific; source timestamps are UTC).

Limits. The desk ran no poll and read no respondent-level data. Poll numbers are as printed in each survey’s report or topline, or read by the desk from the crosstab workbook or a chart, as stated in each row. The online sample is not random: it is what public web searches returned, coded by a model and checked by the desk’s model, and it supports no statement about any population or any age group. Only aggregates appear here; no private person’s name, handle, link or text is on this page.

The poll items, with wording, field dates and ages

Pew Research Center, American Trends Panel Wave 146

Fielded April 8 to 14, 2024; online panel; full sample 8,709 adults, margin of sampling error plus or minus 1.5 points; this item was asked of Form 2 only (N=4,360). Item SOCSEC_FUT: “Thinking about the long term future of Social Security, do you think …” Answers: some reductions in benefits for future retirees need to be considered (19% of adults: phased out as a government program 6%, maintained at a reduced level 13%) or Social Security benefits should not be reduced in any way (79%: cover more people with greater benefits 40%, be kept about as it is 38%). Age cells were read by the desk from the report chart; the two phased-out cells marked ~ are not labeled on the chart and are derived by subtraction. Pew prints no Ns for the age groups.

Answer18-2930-4950-6465+
Phased out as a government program8%9%~3%~2%
Maintained at a reduced level16%13%11%15%
Net: some reductions need to be considered24%22%14%17%
Be kept about as it is21%30%46%55%
Cover more people, with greater benefits51%45%39%26%
Net: should not be reduced in any way73%76%85%82%

Cato Institute / YouGov, December 2025 Social Security National Survey

Fielded October 9 to 14, 2025; YouGov interviewed 2,402 respondents matched down to 2,000; margin of error plus or minus 2.59 points for the full sample only. Unweighted Ns by age as printed in the crosstab workbook: 18-29 N=399, 30-44 N=521, 45-54 N=266, 55-64 N=361, 65+ N=452. Items marked half-sample (Q38a, Q38b, Q38g, Q46, Q47) were asked of half the respondents; the half-sample Ns are not printed. All cells below were extracted programmatically from the converted crosstab workbook and Q35 was checked against the age figures in Cato’s report text. Q35, the forced choice behind the viral boomers-and-taxes post, is covered in the desk’s earlier editorial.

ItemWhat it measures (desk label)All18-2930-4445-5455-6465+
Q1Favorable view of Social Security83%83%81%80%78%91%
Q1Unfavorable view of Social Security17%17%19%20%22%9%
Q5Creation of Social Security was a bad idea5%4%8%6%6%3%
Q22Social Security will exist when I retire: yes34%34%25%28%42%74%
Q22Social Security will exist when I retire: no30%40%37%27%14%5%
Q22Social Security will exist when I retire: not sure36%26%38%45%44%21%
Q27US spends too much on Social Security11%13%12%12%11%9%
Q27US spends too little on Social Security66%57%59%67%74%74%
Q28Social Security benefits are too high5%10%6%4%2%2%
Q35Protect retiree benefits even if higher taxes on younger workers69%47%57%74%84%89%
Q35Protect younger workers from higher taxes even if reducing retiree benefits31%53%43%26%16%11%
Q36Willing to reduce own benefits if guaranteed to get back at least what paid in plus growth35%53%42%39%24%17%
Q37Prefer Congress cut benefits (vs raise taxes or borrow)14%25%17%16%9%5%
Q38aFavor reducing benefits for current and future retirees23%47%28%24%10%6%
Q38aOppose reducing benefits for current and future retirees77%53%72%76%90%94%
Q38bFavor reducing benefits for only future retirees33%53%27%28%24%36%
Q38gFavor paying benefits only to seniors in financial need33%50%36%27%33%17%
Q45Support letting workers put part of their Social Security taxes into personal investment accounts60%63%58%58%60%58%
Q46Favor letting young workers opt out of Social Security taxes if benefits for current seniors would be reduced31%55%35%33%20%14%
Q47Favor letting young workers opt out of Social Security taxes if benefits went only to seniors in financial need35%57%41%35%26%17%
Q52Calling Social Security a Ponzi scheme is a fair characterization33%38%37%37%30%25%

The 47% under 30 for Q38a (favor reducing benefits for current and future retirees) and the 47% under 30 for Q35 (protect retiree benefits, even if higher taxes on younger workers) are different items.

Data for Progress, Polling on Privatizing Social Security

Fielded August 1 to 3, 2025; 596 U.S. likely voters; web panel; margin of error plus or minus 4 points for the full sample, larger for subgroups. Q1: “Do you think the U.S. should privatize Social Security or leave it as it is?” Topline: leave 72%, privatize 20%, don’t know 8%. Under 45 (weighted N 205): leave 72%, privatize 23%, don’t know 5%. 45 and over (weighted N 391): leave 72%, privatize 18%, don’t know 10%. No under-30 cut is printed.

Nationwide Retirement Institute / The Harris Poll, 2025 Social Security Survey

Fielded June 2 to July 10, 2025; online; 1,812 adults who currently receive or expect to receive Social Security; Gen Z (18-28) n=301, Millennials (29-44) n=504, Gen X (45-60) n=505, Boomers+ (61+) n=502. Agreement that “I will not get a dime of the Social Security benefits I have earned”: Gen Z 38%, Millennials 34%, Gen X 26%, Boomers+ 9%, adults 60-65 14%, all 22%. Support for “Privatize the program” among measures to ensure long-term stability: 37% in all (26% somewhat, 11% strongly), 30% among current recipients, 41% among those who expect to receive it; the table splits by receipt, not by age. A 2026 wave (May 11 to June 4, 2026, n=1,823) exists; the desk could read its total (26% agree they will not get a dime) but not its generational cells from the converted deck.

AARP, June 18 to 23, 2025

3,599 adults 18 and older, online and by telephone. 96% consider Social Security important; confidence in the program’s future 36% overall, 25% among ages 18-49 and 48% among ages 50 and older.

General Social Survey (NORC)

The GSS item on spending on Social Security (too little, about the right amount, too much) is not about ending the program. The desk read a peer-reviewed analysis of the 1984 to 2022 waves (abstract and results paragraph only): older adults 65 and over are less likely than adults 18-64 to support more spending, and a weighted, adjusted estimate of more-spending support is 49.5% in 1984 and 57.9% in 2022 for the total sample. The desk did not pull GSS microdata by age. For comparison, Cato’s own spending item (Q27) puts “too much” at 13%, 12%, 12%, 11%, 9% from the youngest age group to the oldest.

Searched, nothing found on ending

Gallup, AP-NORC, the Reagan Institute, the Employee Benefit Research Institute, Harvard CAPS/Harris, KFF, Transamerica, the Peterson Foundation, Morning Consult and Napolitan News Service (June 1-2, 2026 toplines, read in full; items on receiving promised benefits, importance of protecting benefits, and delaying claims). The desk found no item in any of them that asks respondents to abolish Social Security, and read several only as search results. Gallup’s Social Security trend page lists tables on investing Social Security taxes and on the Bush approach; the embedded tables did not open. An absence found is not an absence proven.

The online sample

Public pages only. No logins, cookies, tokens or accounts. Requests at least 2.5 seconds apart. No attempt to get around a block. X and TikTok text is the excerpt a public web search returned (Firecrawl search restricted to x.com, twitter.com and tiktok.com, past-year filter); TikTok pages were not opened and no video downloader was run. Reddit’s JSON listing endpoint answered HTTP 403 and was stopped; the public search RSS feed answered three requests with no entries and then HTTP 429 and was stopped; a second pass at a ten-second gap drew HTTP 429 on its second request and was stopped. Reddit’s own listings contributed no items; Reddit items come only from public web search results restricted to reddit.com.

Search terms

Codebook

Counts

2,453 distinct public items surfaced (X 843, TikTok 766, Reddit 844). Model-judged voice: individual 748, unclear 867, public figure or organization 405, news or official 433. Items from individual or unclear voices: 1,615. Posts that state a view, counted once each after the evidence check and the desk’s adjudication: 114 (96 individual voice, 18 unclear). Of the 1,615, 21 (1.3%) state an age or generation. Where a date could be recovered, the 44 dated posts run from 4 December 2025 to 11 October 2026; the 70 undated posts are Reddit excerpts returned under a past-year filter.

RowPostsEnd itOpt out / private accountsCut / reformProtect / expand / fundWill not exist, no stance
All posts that state a view114171475422
X29841124
TikTok15100122
Reddit7081063016
Found by abolition-term searches64111242611
Found by all other searches506232811
Any age signal610014
No age signal108161475318
Under-30 or Gen Z signal210001
Millennial or Gen X signal400013
Age 55 and over signal000000

The six posts with any age signal: two from a stated age or generation, four from an age-defined community. The desk inferred no age from style, topic or tone.

Coding and hand checks

Coder: DeepSeek (deepseek-chat) through the desk’s LiteLLM gateway, three prompt versions; the third required an evidence quote copied from the post and downgraded any stance whose quote was not found. Hand checks were done by the desk’s model (Claude), not by a second person. Three blind samples of 60 items. Version 1: stance agreement 19 of 48 items the model called on-topic (40%); on-topic agreement 46 of 60. Version 2: on-topic agreement 54 of 60; six-way agreement 41 of 60 (68%). Version 3 on a fresh sample: six-way agreement 44 of 60 (73%, Cohen’s kappa 0.65), on-topic agreement 52 of 60; of the 40 items the model gave a stance, the desk agreed on 25; of the 28 the desk gave a stance, the model matched 25. Version 3 on the two earlier samples: 53 of 60 and 48 of 60. Then the desk read each of the 162 individual-or-unclear posts to which the version-3 model gave a stance: 112 kept (69%), 5 relabeled, 45 set to OTHER; three of the remaining 117 repeated text already counted, leaving 114. Token counts across the three versions: about 5.0 million input and 0.38 million output; about $1.5 at list price; the gateway counter for the seat reads $0.62 for the week; hard cap $2.

Files on the DGX under ~/jobs/ss_sentiment: items_coded.csv (sha256 a6b587378ba22374c55e22169b04cb6185e2863b3eef2af863b78cfc755e2505), codebook.md (sha256 c9459a48b25238fe40f5baf1f457bd92f6d08613488f4f67108fa7766bb78cab), SHA256SUMS for the raw, coded and adjudicated files.

Figure data

The two figures in the editorial draw only on the tables above: the poll small multiples use the Pew, Cato, Data for Progress and Nationwide cells; the stance figure uses the counts table.

What the desk could not verify or reach