Records · Verification table
Inequality Do page: source verification, 2026-11-06
The working record behind the pages, published as written. Source file source/verify/inequality-do-2026-11-06.md. All records.
Scope. Every claim in a line of content/inequality-do.html that was added or changed between ee6f7b8 and HEAD (4764ebf; the file is unchanged in 8235279) and has no row in source/verify/inequality-2026-11-05.md. That covers each number, date, quote, evidence label and what it summarises, named study (authors, title, year, venue, sample, effect), organisation, URL and source grade. The help-now box and the languages line were checked against the quotes in source/impact/needs-2026-11-06.md and their pages re-read. Claims that already have a 5 November row are listed in section 7 and not counted. Where a fix touched them anyway, it says so there.
Method. Read 2026-11-06, 09:36–10:05 EST (file timestamps). Pages and PDFs were fetched with curl and a browser user agent. Quotes are copied from the text: HTML through a tag-stripper, PDFs through pdftotext -layout. Sites that refuse curl were read in headless Chrome (probe.mjs): Springer's script challenge, the IRS site locator's language menu and LSC's finder. Two hosts block both curl and headless Chrome (Cloudflare), codeforamerica.org and governor.ny.gov, so they were read with WebFetch, twice each. Those quotes are the tool's copy. For the governor's page, today's copy matches the 5 November row D10 word for word. Publication status comes from the DOI registry (Crossref API) and the authors' or publishers' own pages. Two web searches were run, only to find URLs, never as evidence: one for later TIGTA volunteer-program audits, one for an open copy of Frandsen (2021). Working files are in the session scratchpad (…/scratchpad/ineqdo/raw/), not kept in the repo. Nothing below is from memory.
Verdicts. CONFIRMED = the source's words support the page as written. CORRECTED = the page needed a wording, figure, grade, title or URL change (made). NOT IN SOURCE = the cited source doesn't say it (what it does say is given; fix made). NEWER DATA = a later edition gives a different figure (applied, see D15). UNREACHABLE, BLOCKED, UNFETCHABLE: no row ended in these. Every blocked host was read another way, as listed under Method. Rows with two verdicts are counted under the one that needed an edit.
Grades (the About page's scale): primary · research (peer-reviewed) · preprint (not yet) · evaluator · X's own figures · news · our arithmetic · unread · unverified.
---
1. Help-now box and languages line (checked against needs-2026-11-06.md and re-read)
| # | Page text (before) | Source | Quote from source | Verdict | Grade |
|---|---|---|---|---|---|
| H1 | "in New York City, free lawyers are available in every ZIP code" | NYC HRA, legal services for tenants (needs A5.1; re-read, curl) | "Right-to-Counsel (or Universal Access) legal services are free and available in every ZIP code, regardless of immigration status. Tenants facing eviction in Housing Court or NYCHA administrative proceedings have access to free legal representation or advice provided by legal services organizations". OCJ FY2025 report (needs A5.2): "44% of all respondents in eviction cases who appeared in court received full legal representation … during the fourth quarter of FY25." | CORRECTED. The City promises "legal services", meaning "representation or advice", not a lawyer for everyone. Now: "free legal help is available in every ZIP code". | primary |
| H2 | "whatever your immigration status" | same | "regardless of immigration status" | CONFIRMED | primary |
| H3 | "say 'I would like an attorney' at your first court date" | same | "Go to your initial court appearance and say "I would like an attorney" when you appear in court." | CONFIRMED | primary |
| H4 | "or call 311 and ask for the Tenant Helpline" | same | "Call 311 and ask for the Tenant Helpline." | CONFIRMED | primary |
| H5 | "Elsewhere, find free legal aid by ZIP" | LSC finder (needs A5.5; re-read, curl + headless Chrome) | "LSC is an independent nonprofit established by Congress in 1974 to provide financial support for civil legal aid to low-income Americans." / "enter an address or city below to find an LSC-funded legal aid organization near you." The only "free" on the page is LawHelp's "find free legal forms". Our test (needs A5.5, repeated today on the new URL): ?address=10001 → "Legal Services NYC"; ?address=nonsense-zzqx → "Sorry, we couldn't find a match for this location." | CORRECTED. "free" is in neither the page nor the needs file's quotes; "by ZIP" is confirmed by the test. Now: "find legal aid by ZIP". URL at U1. | primary; our test |
| H6 | "IRS volunteers help people who generally make about $70,244 or less" | IRS, free tax return preparation for qualifying taxpayers ("Page Last Reviewed or Updated: 19-Oct-2026") | "While the IRS does not set a program-wide income limit, individual VITA sites may establish their own." / "People who generally make about $70,244 or less" | CONFIRMED (no season named, as s16 says) | primary |
| H7 | "find a site" → freetaxassistance.for.irs.gov/s/sitelocator | same; locator (HTTP 200) | "Attention: The locator is updated regularly from February through April." | CONFIRMED | primary |
| H8 | "NYC's 311 answers in more than 175" | NYC 311 languages (KA-03541; re-read, curl) | "If your language is not available online, you can get help in more than 175 languages by calling 311." | CONFIRMED | primary |
| H9 | "the IRS's free tax site locator lets you filter for 36 languages" | IRS site locator, headless Chrome (the "Language Available?" combobox read from the component) | 36 options: Arabic, Armenian, Bengali, Cantonese, Chinese, English, Farsi, Filipino, French, French Creole, German, Greek, Hebrew, Hindi, Hungarian, Inuit, Italian, Japanese, Korean, Kurdish, Lakota, Mandarin, Navajo, Other, Other - See Comments, Polish, Portuguese, Russian, Serbian, Sign Language, Spanish, Swedish, Tagalog, Turkish, Vietnamese, Virtual Interpreter Services (same list as needs §B). | CORRECTED. 36 options, not 36 languages: one is English and three aren't languages. Now: "32 languages besides English" (our count). | primary; our arithmetic |
2. The whole thing (three steps)
| # | Page text (before) | Source | Quote from source | Verdict | Grade |
|---|---|---|---|---|---|
| S1 | Step 1 label strong: for recipients | Egger et al., NBER w26600 page; impact review A3 | "We provided one-time cash transfers of about USD 1000 to over 10,500 poor households across 653 randomized villages in rural Kenya." / "We find large impacts on consumption and assets for recipients." Published Econometrica 90(6), 2603–2643, Nov 2022. | CONFIRMED. A large randomised trial covers the chain from gift to recipient. | research |
| S2 | "Recipients' assets were still 40% higher three years on." (placed right after the 653-village trial) | Haushofer & Shapiro, J-PAL-hosted PDF, "This version : January 2018" | "Comparing recipient households to non-recipients in distant villages, we find that transfer recipients have 40% more assets (USD 422 PPP) than control households three years after the transfer"; within villages "a 40 percent increase in asset holdings (USD 416 PPP)". The trial: "Between 2011 and 2013, GD sent unconditional cash transfers averaging USD 709 PPP … in western Kenya". | CORRECTED (attribution). The figure holds, but it comes from an earlier, different trial, and the sentence read as the 653-village one. Now: "In an earlier Kenyan trial, recipients' assets were still 40% higher three years on." | preprint (see R1) |
| S3 | "Four of six other studies found no spillover to neighbours, or a slight loss." | GiveWell, GiveDirectly cash for poverty relief ("Published: November 2024, Updated: October 2025") | "Of the 7 randomized evaluations we found of GiveDirectly's lump sum program, 6 try to measure consumption spillovers to non-recipient households. 4 of these papers find null or slightly negative effects, while two (both the Egger et al. papers) find positive spillovers." | CORRECTED. The six include the 653-village trial's two papers, so "other" is wrong, and they are "papers", not studies. Now: "Of six papers that measured spillovers to neighbours, four found none, or a slight loss; the two that found gains both come from the 653-village trial." s2 now carries the count. | evaluator |
| S4 | "This is poverty, not inequality: no trial finds it moves a country's wealth gap." | Banerjee, Faye, Krueger, Niehaus & Suri, UBI paper (15 Sep 2023), the one cash trial read that measures inequality | "There are no significant changes in either the 75-25 ratio or the Gini coefficient for either wealth or income from any of the interventions. On the other hand, the 75-25 ratio for non-land wealth goes down substantially and significantly in all three…" These are village-level measures. No trial read measures a country's distribution. | CONFIRMED (absence; our reading). Village-level evidence only, now cited in s2 (R18). | preprint; our reading |
| S5 | Step 2 label moderate | Bergan & Cole 2015; Butler & Nickerson 2011; impact review A4 | Two field experiments on state legislators. The authors' own caveat: "Anti-bullying is a relatively low salience issue" (Bergan, LSE blog, 29 Apr 2015). | CONFIRMED | research |
| S6 | "the only thing on this page with experimental evidence of changing a rule" | same | Springer abstract: "estimates of the effects on legislative voting"; LSE: "The outcome variable in the study is the final vote on the anti-bullying bill." Butler & Nickerson: "much more likely to vote in line with constituent opinion". | CORRECTED. Both experiments measured votes, not changed rules. Now: "experimental evidence of moving lawmakers' votes". | research |
| S7 | "a few constituent messages made state legislators about 12 points more likely to back a low-profile bill" | Bergan & Cole, Political Behavior (Springer page via headless Chrome; curl got a script challenge) | Abstract: "Michigan state legislators are randomly assigned to be contacted by their constituents about a specific bill or to a control group … being targeted by constituent contacts increases the probability of supporting the relevant legislation by about 12 percentage points." Notes: "Calls were evenly spaced over the 2 weeks for legislators in each of the treatment groups"; "treatment group (0, 22, 33 or 65 intended calls)". | CORRECTED. The contacts were phone calls, 22 to 65 per legislator, not "a few messages". "About 12 points" and "low-profile" hold. Now: "calls from constituents made state legislators about 12 points more likely to back a low-profile bill"; s3 adds "the contacts were phone calls". | research |
| S8 | Step 3 heading "Do free tax returns this winter." and the description's "free tax returns this winter" | NYC DCWP volunteer page ("Page updated 10/2026"); IRS Pub 5347 | DCWP: "now accepting volunteers for the 2027 tax season!" (no months). Pub 5347: "QSS reviews were conducted throughout the 2026 filing season, beginning the week of February 8th and concluding on the week of April 5th." No source gives the 2027 season's dates. | NOT IN SOURCE. Now "in tax season", in both places (as the overview already reads). | primary |
| S9 | "in New York the fee they keep averages $346" (and s4 "($346 average fee)") | NYC DCWP Free Tax Prep FAQ ("This page was updated 03/2026") | "How much can I save with NYC Free Tax Prep? It depends. On average, tax preparation fees are about $346 in New York." | CONFIRMED | primary |
| S10 | Step 3 label moderate: certain per family | impact review A1 | Graded there "moderate for the families served; weak for 'raises EITC take-up'". The page carries the weak part in words (S12). | CONFIRMED | — |
| S11 | "About forty households a season per volunteer." (cited to s4 only) | NYC DCWP, Become an NYC Free Tax Prep volunteer (in s6) | "In just one tax season, you'll complete about 40 tax returns, helping your neighbors claim more than $90,000 in refunds and tax preparation fee savings." | CORRECTED. The unit is returns, not households; the figure is New York City's; and its source is s6. Now: "About forty returns a season per volunteer in New York City", cited 4 and 6. | primary |
| S12 | "six trials found that telling people about free sites barely moves how many claim the credit" | Linos, Prohofsky, Ramesh, Rothstein & Unrath, AEJ: Economic Policy 14(4): 432–452, 2022 (NSF repository PDF; AEA page "November 2022") | "We ran six preregistered, large-scale field experiments with 1 million observations to test whether "nudges" could increase EITC take-up. … we find no evidence that they affected households' likelihood of filing a tax return or claiming the credit." Arms include "Local in-person free tax preparation information" and "Address of local in-person free tax preparation assistance". | CONFIRMED on Linos et al. The page credited it to Code for America (R7). | research |
| S13 | "because most who miss it don't file at all" | Goldin et al., NBER w28398; Linos et al. | Goldin: "approximately two-thirds of individuals who qualify for but fail to claim the EITC do not file a tax return (Census, 2013)". Linos: "even the most behaviorally informed low-touch outreach efforts cannot overcome the barriers faced by low-income households who do not file returns." | CONFIRMED (two-thirds = "most"; the "because" is Linos et al.'s own reading) | research |
3. Details: cash, bill, tax, tenants, home
| # | Page text (before) | Source | Quote from source | Verdict | Grade |
|---|---|---|---|---|---|
| D1 | "GiveDirectly's own research page overstates its trials in places" | givedirectly.org/gdresearch/ (the cited URL 301s here); Aggarwal et al., AEJ: Applied 18(3), 2026 (GiveDirectly-hosted PDF) | GiveDirectly: "with the value of harvest output increasing by 50% in Liberia and 20% in Malawi". The article: "The value of harvest output increased by about 35 percent in Liberia and 16 percent in Malawi". | CONFIRMED (also D2) | GiveDirectly's own figures; research |
| D2 | "(it reports a fall in inequality where the paper finds no change)" | GiveDirectly research page; Banerjee et al. 2023, text and Table 8 | GiveDirectly: "-29% Inequality within the village" (source: "Banerjee et al., 2023"). Paper: no significant change in the 75-25 ratio or the Gini for wealth or income, and no significant change in the Gini for non-land wealth or consumption (Table 8), but "the 75-25 ratio for non-land wealth goes down substantially and significantly in all three". Lump sum: −1.63 on a control mean of 5.55 = −29.4% (our arithmetic), which matches GiveDirectly's −29%. | CORRECTED. The paper does find a fall, in one narrower measure; it finds no change in the Gini. Now: "It headlines a 29% fall in inequality within villages; the paper finds no change in the Gini, and a fall only in narrower measures." Banerjee et al. is now cited in s2. | GiveDirectly's own figures; preprint |
| D3 | "It disclosed about $1.2 million diverted by staff in the DRC in 2022, 0.8% of that year's transfers." | GiveDirectly, DRC case ("Published June 5, 2023"; updates 16 Nov 2023 and 24 Sep 2024) | "Update as of November 16, 2023: … Our investigation concluded that just under $1.2M was illegally diverted from our D.R.C. operations in 2022. This updated number is based on an exhaustive investigation of SIM fraud and other staff malfeasance, and amounts to 0.8% of the $144M we delivered globally last year." / "The fraud occurred when our D.R.C. staff conspired with outsiders to divert mobile money payments". | CONFIRMED. Markup fixed too: this item's <sup> was never closed. | GiveDirectly's own figures |
| D4 | "Find your state legislators at usa.gov" | usa.gov/elected-officials (curl) | "Find the names and current activities of state and territorial legislators." | CONFIRMED | primary |
| D5 | "a call counts too" | CMF, Citizen-Centric Advocacy (PDF via its 302) | Chart: "Phone Calls 84%". Bergan & Cole's contacts were calls (S7). | CONFIRMED | primary; research |
| D6 | "the IRS's own … reviews put volunteer sites at 98.5%" | IRS Publication 5347 (Rev. 9-2026) | "the estimated overall accuracy for federal tax returns prepared at eligible VITA/TCE sites nationwide was 98.47%." | CONFIRMED | primary |
| D7 | "announced" (the IRS's "own announced reviews") | Pub 5347; TIGTA 2016-40-045 (17 June 2016; not cited on the page) | Pub 5347 says nothing on notice. TIGTA, of the 2015 season: "For most of the site visits and return reviews, volunteers are aware that a QSS review is being conducted because the QSS analysts introduce themselves at the beginning of the site visit." That is introduced on arrival, not announced, and it describes 2015. | NOT IN SOURCE. Word removed. Proposal 5 restores the point with its source. | primary |
| D8 | "the last independent audits, in 2009 to 2013, found far lower rates on anonymous visits" (cited to Accounting Today, 21 Sep 2011) | Accounting Today, 21 Oct 2009, 21 Sep 2011, 5 Nov 2013; TIGTA 2013-40-110 (16 Sep 2013) | 2009: "41 percent of the tax returns were prepared inaccurately by volunteers when TIGTA auditors anonymously visited volunteer sites". 2011: "Of the 36 tax returns prepared for TIGTA auditors, only 14 of them, or 39 percent, were prepared correctly" (the prior year was "a 90 percent accuracy rate", when "the IRS was aware last year of the site visits"). 2013: "Anonymous visits by government auditors…"; "Of the 39 tax returns prepared for auditors during the 2013 filing season, 20 of them (or 51 percent) were prepared correctly … over the 49 percent accuracy rate … in the 2012 filing season." TIGTA 2013 confirms 20 (51%). | CORRECTED. The rates hold, but TIGTA audited the program again later (D9), so these were the last anonymous tests, not the last audits. The cited 2011 article covers only 2010–11. Now: "the last independent tests on anonymous visits, from 2009 to 2013, found far lower rates". s4 links the 2013 article, the last test. | news; primary |
| D9 | "Nobody independent has audited since." | TIGTA 2016-40-045; TIGTA 2017-40-088 (20 Sep 2017); TIGTA 2020-40-004 (13 Nov 2019); search of tigta.gov, gao.gov and oversight.gov (URLs only) | 2016: "TIGTA's evaluation of a statistically valid sample of 44 of these returns identified that 42 (95 percent) were accurately prepared by the volunteers." 2017: the VITA grant program. 2019: "TIGTA's unannounced visits to 20 volunteer sites identified multiple security weaknesses at each site". TIGTA 2025-108-017 is about Taxpayer Assistance Centers, a different program. No anonymous accuracy test after 2013 was found. | CORRECTED. Independent auditors did audit after 2013, just not with anonymous accuracy tests. Now: "We found none since." | primary |
| D10 | Tenant label moderate | Seron et al. 2001; Cassidy & Currie; impact review A/C1 ("moderate–strong") | an RCT and a quasi-experiment, both on outcomes for households | CONFIRMED (the page takes the review's lower bound) | research |
| D11 | "tenants with a lawyer got a judgment against them 32% of the time, against 52% without" | Seron, Van Ryzin, Frankel & Kovath, Law and Society Review 2001 (PDF) | Body: "although judgments were issued against 52% of control cases, only approximately 32% of the treatment cases had judgments against them", a "conservative estimate … because … a substantial minority of tenants in the treatment group was never provided with an attorney". Table 1: represented "75 (56%)" of treatment, "5 (4%)" of control. Abstract, comparing by representation: "only 22% of represented tenants had final judgments against them, compared with 51% of tenants without legal representation." | CORRECTED. 32/52 compares tenants assigned a lawyer with controls, not tenants who had one. Now: "tenants picked at random to get a lawyer had a judgment against them 32% of the time, against 52% of the rest." | research |
| D12 | "evictions in covered ZIP codes fell by somewhere between 45% and 78%" | Cassidy & Currie, NBER w29836 ("March 2022, Revised July 2022") | "we find large reductions in: … the probability of ultimately being evicted (between 44.8 and 77.8 percent). The high end estimates are upper bounds…"; "legal representation greatly improved outcomes among tenants who received services due to the program." | CORRECTED. The range is the fall in eviction risk for tenants who got a lawyer through the program, not a fall in evictions across covered ZIP codes. Now: "tenants who got a lawyer through it were somewhere between 45% and 78% less likely to be evicted". | research |
| D13 | "at about $3,200 a household" | same | "the Department of Social Services FY2021 budget for tenant legal programs was $136 million and 42,000 households were served by UA, implying a cost of about $3,200 per household (Office of Civil Justice, 2021)." | CONFIRMED | research |
| D14 | "in 2022 only 63% of tenants facing eviction had a lawyer" (s7: OCJ "Universal Access annual report 2022 (63% represented)") | OCJ FY2022 report (PDF created 9 Mar 2023); OCJ FY2025 report ("10th Annual Report", PDF created 24 Dec 2025) | FY2022: "more than 63% of tenants who appeared in Housing Court for eviction cases in the fourth quarter of FY2022 were represented by attorneys in court." FY2025: "44% of all respondents in eviction cases who appeared in court received full legal representation … during the fourth quarter of FY25", and "the statewide eviction moratorium … sharply reduced the number of new and active eviction cases, temporarily inflating the representation rate. Once the eviction moratorium was lifted, filing surged again and the need for representation quickly outpaced the supply of attorneys, driving the rate back down." The City's fiscal year runs July to June (FY2022 report, footnote 5), so Q4 FY25 = April–June 2025. | NEWER DATA. 63% holds for April–June 2022, but the city's newer report gives 44% for April–June 2025 and says the earlier rate was inflated by the moratorium. Now: "Even so, from April to June 2025 only 44% of tenants in court had full representation", with s7 and the inline link pointing at the FY2025 report. Fallback if you prefer the cited edition: "in spring 2022 just over 63% of tenants in court had a lawyer" (FY2022 link). | primary |
| D15 | "because the money ran short" | OCJ FY2022 and FY2025 reports | FY2022 cause: "UA citywide expansion, vestiges of COVID-19, and the "Great Resignation", meant that, for the first time in the program's history, providers did not always appear at housing court." FY2025: "the need for representation quickly outpaced the supply of attorneys". Neither mentions money. | NOT IN SOURCE. Clause removed. | primary |
| D16 | "New York City's program is run by the Office of Civil Justice" | OCJ FY2022 and FY2025 reports | FY2022: the law "tasks the Office of Civil Justice (OCJ) of the New York City Human Resources Administration (HRA) with planning and implementing programs"; FY2025: OCJ "oversees and monitors City-supported civil legal services for low-income tenants". | CONFIRMED (link now the FY2025 report) | primary |
| D17 | "Community land trusts keep homes affordable when resold, and the evidence for the people housed is real" | Cited s5 (City Limits, NYCCLI, ProPublica): silent. Wang, Cahen, Acolin & Walter, Lincoln Institute WP19RW1, April 2019 (impact review A5) | "Ninety-five percent of shared equity homes are priced affordably for families earning 80 percent of area median income (AMI) or below across all housing market periods."; "Affordability is achieved for both first purchases and resales of shared equity homes."; shared-equity models include "community land trusts (CLTs), deed-restricted homes, and limited-equity cooperatives". Working paper, "not been" reviewed. | NOT IN SOURCE (cited); true on Wang et al., which covers shared-equity homes, land trusts among them. Fix: Wang et al. added to s5 as an unlinked citation, graded preprint. | preprint |
| D18 | "Western Queens CLT, is new" | wqclt.org home (curl); ProPublica | WQCLT: "We've spent the last 6 years collaborating with the Queens community to create permanently affordable buildings". ProPublica: "Tax exemption issued: Nov. 2024". | CORRECTED. Six years of work; only the tax exemption is new. Now "is small", which the receipts support (D20). | WQCLT's own figures; primary |
| D19 | "with no completed project" | wqclt.org home | Three buildings, all in progress: "one of the buildings we are pursuing, dubbed the Queensboro People's Space"; "working to acquire a mixed-use building in Astoria"; "working with other local groups to acquire an unfinished apartment building in Elmhurst". None is described as acquired. | CONFIRMED (absence on its own site) | WQCLT's own figures |
| D20 | "receipts under $50,000" | ProPublica Nonprofit Explorer, EIN 85-3774663 | "Fiscal Year Ending Dec. 2025 / Gross receipts / Less than $50,000 / 990-N"; the same for Dec. 2024. | CONFIRMED | primary (IRS data via ProPublica) |
| D21 | Label weak: for this organisation | impact review A5 | "weak as a lever for one reader"; "a small bet on an early-stage group" | CONFIRMED | — |
| D22 | Label strong: by construction, for that worker; "changes one person's pay and nothing about inequality" | impact review A2 and section 0 | "strong by construction for that worker; none for inequality" | CONFIRMED (the review's grade; our reading) | — |
| D23 | Union label moderate: contested | s11 studies | Close-election studies disagree (D25). | CONFIRMED | research |
| D24 | "Unions compressed pay gaps for most of the last century" | Cited s11 (DiNardo & Lee, Frandsen, Knepper): silent. Farber, Herbst, Kuziemko & Naidu, NBER w24587 page (rev. April 2021; published QJE 136(3), 1325–1385, 2021) | "U.S. income inequality has varied inversely with union density over the past hundred years." / "we find consistent evidence that unions reduce inequality, explaining a significant share of the dramatic fall in inequality between the mid-1930s and late 1940s." | CORRECTED. Not in the cited sources, and broader than the best one. Now: "Unions explain a significant share of the fall in US inequality from the mid-1930s to the late 1940s." Farber et al. added to s11 (unlinked, "abstract read"). | research |
| D25 | "whether winning a vote raises wages at that workplace is disputed, with one study finding no effect and others finding narrower pay and better benefits" (s11: Frandsen "(pay compression)") | DiNardo & Lee, QJE Nov 2004 (author PDF); Frandsen, JOLE 39(4), 861–894, 2021 (EconPapers abstract; full text paywalled, none open found); Knepper, REStat 102(1), 98–112, 2020 (EconPapers) | DiNardo & Lee: "estimates for wages are close to zero." Frandsen 2021: "unionization substantially decreases payroll, employment, average worker earnings, and establishment survival. Payroll and earnings decreases are driven by composition changes, with older and higher-paid workers leaving unionizing establishments and younger workers joining or staying. Worker-level effects on earnings are small". Knepper: "Following unionization, average employee compensation and employer pension contributions increase". Frandsen's 2010 working paper (not cited) did find compression: "unions raise the lower end of the distribution by around 25 percent". | CORRECTED. The cited 2021 paper does not report narrower pay; that was the 2010 draft. Now: "one study finds no wage effect, one finds average pay falls as better-paid workers leave, and one finds pay and pensions rise." The s11 notes match. | research |
4. Questions people ask
| # | Page text | Source | Quote | Verdict | Grade |
|---|---|---|---|---|---|
| Q1 | "For inequality, no study measures what one person's deposit or purchase changes" ("deposit" is row D37 of 5 Nov; "or purchase" is new) | impact review A7 | "No evidence a buyer's choice changes a worker's income." | CONFIRMED as the review's absence finding. Not searched again today. | our research |
| Q2 | Label none but plausible on credit unions and co-op buying | impact review A6 (credit unions: none-but-plausible), A7 (co-ops: weak) | The label sits on the inequality claim (one person's deposit or purchase), where no outcome evidence exists. That fits the site's definition. The review's "weak" for co-ops grades firm-level correlations, which the answer covers in words. | CONFIRMED (note the difference) | — |
5. Sources list and footer
| # | Page text (before) | Check | Quote | Verdict |
|---|---|---|---|---|
| R1 | s2 Haushofer & Shapiro, "the three-year follow-up, 2018 (assets +40%)" research | PDF; the author's research page (johanneshaushofer.com/research); Crossref | PDF "This version : January 2018". Author: "Haushofer, Johannes, and Jeremy Shapiro. "The Long-Term Impact of Unconditional Cash Transfers to the Poor: Experimental Evidence from Kenya." Working Paper (2018)." Crossref lists the 2016 QJE short-term paper but no journal version of this one. | CORRECTED (grade): preprint, "a January 2018 working paper". |
| R2 | s3 Bergan & Cole, "Call your legislator", Political Behavior, 2015, Michigan | Springer (headless Chrome); Crossref | "Call Your Legislator: A Field Experimental Study of the Impact of a Constituency Mobilization Campaign on Legislative Voting"; "Published: 01 June 2014 / Volume 37, pages 27–42 (2015)"; Daniel E. Bergan & Richard T. Cole. | CONFIRMED (short title; "phone calls" added, see S7) |
| R3 | s3 Butler & Nickerson, "Can learning constituency opinion affect how legislators vote?", 2011, New Mexico | ISPS page (the URL 301s, U4) | "Daniel M. Butler and David W. Nickerson (2011) … Quarterly Journal of Political Science: Vol. 6:No 1, pp 55-83"; "Peer Reviewed Article"; "We surveyed 10,690 New Mexicans". | CONFIRMED |
| R4 | s4 Pub 5347, "September 2026 (2.9 million returns, 84,500 volunteers, 98.5% …)" | "Publication 5347 (Rev. 9-2026)"; "During the 2026 filing season, dedicated SPEC partners supported over 9,000 VITA/TCE sites, with more than 84,500 volunteers. VITA/TCE sites prepared and filed over 2.9 million federal tax returns". | CORRECTED: "(in the 2026 filing season, over 2.9 million returns and more than 84,500 volunteers; …)" | |
| R5 | s4 participation rate by state "(80.8%)" | IRS page ("Page Last Reviewed or Updated: 19-Aug-2026") | Table: "National … 80.8%" (tax year 2022). Footnote 2: "Comparably, the tax year 2022 national estimate, based on the Current Population Survey (CPS) is 78%. The 78% participation rate based on the CPS is the reported national estimate of the EITC participation rate." | CORRECTED: both, in the Practice page's wording: "(80.8% in its table for tax year 2022; 78% is the page's "reported national estimate")". |
| R6 | s4 Goldin, Homonoff, Javaid & Schafer, "Tax filing and take-up: experimental evidence on tax preparation outreach", 2021 "(56,015 non-filers; small effects)" | NBER page and PDF; Crossref | Title: "Tax Filing and Take-Up: Experimental Evidence on Tax Preparation Outreach and EITC Participation", NBER WP 28398, "January 2021". "The experimental population consisted of taxpayers who did not file a tax return for the prior tax year (2017)"; "collectively, 56,015 letters"; "statistically significant but fairly modest effects … on filing rates (an increase of approximately 1 percentage point)". Peer-reviewed version: Journal of Public Economics 206, 104550 (Feb 2022), titled "…and benefit claiming". | CORRECTED (title truncated mid-subtitle). Now "…outreach and EITC participation", and "(letters to 56,015 non-filers; small effects)". research stands. |
| R7 | s4 "Code for America's summary of six trials" | codeforamerica.org page: curl 403, headless Chrome stopped by Cloudflare; WebFetch, two passes | "Our Research into Closing the Tax Benefits Coverage Gap"; "March 8, 2023"; "The Tax Benefits team ran nearly a dozen randomized controlled trials in 2022." "Six" doesn't appear. The trials test its own GetCTC outreach. | NOT IN SOURCE. The six trials are Linos et al. (S12). Replaced with "Linos, Prohofsky, Ramesh, Rothstein and Unrath, Can nudges increase take-up of the EITC?, American Economic Journal: Economic Policy, 2022 (six trials, 1 million observations; no effect found)", linked at aeaweb.org, a host already on this page. |
| R8 | s5 "Western Queens CLT's filings via ProPublica's Nonprofit Explorer" → projects.propublica.org/nonprofits/ (the search home) | ProPublica | The organisation's record is /nonprofits/organizations/853774663 (D20). | CORRECTED (URL to the record) |
| R9 | s7 Seron et al., title, Law & Society Review, 2001, "(a randomised trial)" | "Law and Society Review / 2001"; "THE IMPACT OF LEGAL COUNSEL ON OUTCOMES FOR POOR TENANTS IN NEW YORK CITY'S HOUSING COURT: RESULTS OF A RANDOMIZED EXPERIMENT"; Carroll Seron, Gregg Van Ryzin, Martin Frankel, Jean Kovath. Manhattan Housing Court, 1993–94, 134 treatment and 134 control. | CONFIRMED (subtitle omitted) | |
| R10 | s7 Cassidy & Currie, title, 2022, research | PDF; Crossref | "March 2022, Revised July 2022"; peer-reviewed version Journal of Public Economics 222, 104844 (June 2023). | CONFIRMED (subtitle "Evidence from New York City's Universal Access Program" omitted) |
| R11 | s11 DiNardo & Lee, "Economic impacts of new unionization on private sector employers", QJE, 2004 "(no wage effect)" | author-hosted PDF | "ECONOMIC IMPACTS OF NEW UNIONIZATION ON PRIVATE SECTOR EMPLOYERS: 1984 –2001"; "The Quarterly Journal of Economics, November 2004"; "estimates for wages are close to zero". | CONFIRMED (subtitle omitted) |
| R12 | s11 Knepper, "From the fringe to the fore", REStat, 2020 "(benefits)" | EconPapers | "From the Fringe to the Fore: Labor Unions and Employee Compensation", "The Review of Economics and Statistics, 2020, vol. 102, issue 1, 98-112". | CONFIRMED (note now "(compensation and pension contributions rise)") |
| R13 | s14 mapping.ncua.gov "(choose 'Select Low Income Designation' for one that lends in its neighbourhood)" | NCUA, Low-Income Credit Union Designation (curl) | "a credit union may be designated low-income if more than 50 percent of its members have a family income of 80 percent or less than the median family income for the metropolitan area where they live or national metropolitan area, whichever is greater"; "About half of all federally insured credit unions have the low-income designation". | CORRECTED. The designation is about who the members are, not where it lends. Now "for one where more than half the members have low incomes". (5 Nov row D24 fixed only the click path.) |
| R14 | s15 Alaska's News Source, "dividends start this week", 30 September 2026 | URL (200) | "By James Brooks / Published: Sep. 30, 2026"; "This year's dividend was set at $1,000 by the Alaska Legislature in the state budget signed by Gov. Mike Dunleavy"; "This story has been republished with permission from the Alaska Beacon." | CONFIRMED (the origin is the Alaska Beacon, which blocked fetchers on 5 Nov) |
| R15 | s15 Anchorage Daily News, 19 September 2026, "(the amount is set by the Legislature in the budget)" | URL (200) | "Published: September 19, 2026"; "Alaska lawmakers approved a $1,000 dividend per eligible Alaskan, along with a one-time $200 energy relief payment, when passing the state budget in May."; "The Legislature has not followed the statutory formula to calculate the dividend since 2016". | CONFIRMED |
| R16 | s16 dates and grades: IRS "reviewed 19 October 2026; no season named"; NYC 311 "read 6 Nov 2026"; HRA and LSC primary | pages above | as H1–H8 | CONFIRMED |
| R17 | s2 GiveDirectly "the DRC case, 2023" and "its research page" charity's own figures | pages above | "Published June 5, 2023"; the $1.2M figure is from the 16 Nov 2023 update. | CONFIRMED (URLs at U2, U3) |
| R18 | s2 "the paper" behind D2, not cited before | Banerjee et al. PDF; Crossref | "15 September 2023"; Crossref lists no journal version, only an Unjournal evaluation summary (2024). | Fix only: added "Banerjee, Faye, Krueger, Niehaus and Suri, Universal basic income: short-term results from a long-term experiment in Kenya, working paper, September 2023, read 6 Nov 2026 (…) preprint". Not counted separately (part of D2). |
| F1 | Footer: "New York's wage floor is next decided by 1 March 2027." | governor.ny.gov (curl 403, headless Chrome stopped by Cloudflare; WebFetch); dol.ny.gov/minimum-wage (curl) | Governor: "Governor Hochul has proposed a change to the law that could allow minimum wage rates to rise with inflation starting March 1, 2027, if the Legislature approves it."; "If approved by the State Legislature, new rates would take effect on March 1, 2027." DOL: "On January 1, 2027, New York's minimum wage will stay the same." | CORRECTED. 1 March is when a rise would start if approved; the source sets no decision date. Now: "A rise in New York's wage floor could start on 1 March 2027, if the Legislature approves." (data-until and data-expired kept.) |
6. Links that redirect permanently (301), pointed where they land
Every one of the page's 62 external links was requested without following redirects.
| # | Was | Now | Verdict |
|---|---|---|---|
| U1 | lsc.gov/about-lsc/what-legal-aid/get-legal-help (help-now box and s16) | …/i-need-legal-help (our test today: ?address=10001 → Legal Services NYC; nonsense → "couldn't find a match") | CORRECTED |
| U2 | givedirectly.org/research-at-give-directly/ | givedirectly.org/gdresearch/ | CORRECTED |
| U3 | givedirectly.org/drc-case-2023/ | givedirectly.org/drc-case-2023 | CORRECTED |
| U4 | isps.yale.edu/research/publications/isps11-016 | isps.yale.edu/resource/can-learning-constituency-opinion-affect-how-legislators-vote-results-from-a-field | CORRECTED |
| U5 | irs.gov/credits-deductions/individuals/earned-income-tax-credit/statistics-for-tax-returns-with-the-earned-income-tax-credit-eitc | irs.gov/tax-professionals/eitc-central/statistics-for-tax-returns-with-the-earned-income-tax-credit-eitc ("The average amount of EITC received nationwide in TY 2024 was $2,894.") | CORRECTED |
| U6 | dol.ny.gov/minimum-wage-0 | dol.ny.gov/minimum-wage | CORRECTED |
| U7 | wir2026.wid.world/?p=183 | wir2026.wid.world/insight/executive-summary/ | CORRECTED |
Left as they are: dol.ny.gov/facts-employers-domestic-workers, which 301s to a dated file path (/system/files/documents/2021/12/…_1.pdf), so the alias is the steadier link. The Oxfam 2018 release in s1 301s to oxfam.org/en/press-releases/richest-1-percent-bagged-82-percent-wealth-created-last-year, but that line is unchanged and outside this scope. The CMF link and both EconPapers links redirect with 302 (temporary). The Code for America link was replaced (R7). governor.ny.gov and nysenate.gov return 403 to curl and were read another way (F1; 5 Nov row D11). The Springer link returns 200 with a script challenge (read in headless Chrome).
7. Covered by 5 November rows (not counted)
653 villages, $2.40 and little inflation (L23, D5; re-read today, unchanged); GiveDirectly → Donate and 71–85% (D20, D21); GiveWell's top charities (D22); 92% against 56% "form email" (D30); New York bills and COPA "as of 5 November 2026" (D27, D31–D33); NYC Free Tax Prep recruiting, "over 16", the 3 December session, the IRS form, roles, $97,000/$68,000 (O3, D6–D8; the roles page re-read today, "Tax law certification is not required for this position" and, for interpreters, "Basic tax knowledge is helpful, but it is not required"); $2,894 (P5; re-read at U5); the wage floor and "a rise to $17.60 from 1 March 2027 needs the Legislature" (D10, P4; re-read today); overtime, rest days, sick leave, weekly pay, $3,000, disability and family leave, Hand in Hand, the cost box (D11–D19); credit-union rates, worker-owned layoffs, "stronger to begin with" (D23, D28); "no study measures what one person's deposit changes" (D37); the dozen, the growth question (D35, D36); California's billionaire tax "lost on 3 November" (D38: No 54.6% at 100% of precincts on 5 Nov, unofficial until certification); Kenya and US cash (D39); Alaska (L26–L29, D40); s1, s6, s8–s10, s12, s13 content (D41–D53).
One 5 November fix completed on the way through (not counted): row D17 asked for workers' compensation at "40 or more hours a week or live-in". The page had only the hours. WCB, re-read today: "Workers' Compensation coverage is required if the domestic worker is employed forty or more hours per week by the same employer", and coverage "is not required if … domestic workers … individually work less than 40 hours per week for that household and they do not live on premises." Now: "40 or more hours a week, or live-in: workers' compensation too."
8. Listed, not counted
- Lede: "Every action here carries a label … and says whether it touches poverty or inequality". Internal. Every action has a label, but steps 2, 3 and 4 don't say which they touch (proposal 4).
- "A lawyer in housing court is the best-tested housing lever one person can push for." Our reading of the review (an RCT plus a quasi-experiment, against land trusts' descriptive evidence).
- "Near you … The code stays in your browser, like the ticks." Site behaviour;
vercel.jsonsetsconnect-src 'none'. - Advice and costs: "Write five sentences…", "pick one with audited accounts…", "Volunteer if you like", "5 minutes · any amount", "15 minutes · $0", the prompts.
- s2: "see the site's impact review". The review is in the private repo and not published (proposal 3).
9. Summary by verdict
Counted rows: help-now and languages 9 (H1–H9), steps 13 (S1–S13), details 25 (D1–D25), questions 2 (Q1–Q2), sources and footer 18 (R1–R17, F1; R18 is part of D2), redirects 7 (U1–U7) = 74.
| Verdict | Count | Rows |
|---|---|---|
| CONFIRMED | 38 | H2, H3, H4, H6, H7, H8; S1, S4, S5, S9, S10, S12, S13; D1, D3, D4, D5, D6, D10, D13, D16, D19, D20, D21, D22, D23; Q1, Q2; R2, R3, R9, R10, R11, R12, R14, R15, R16, R17 (R12's note was reworded to match the body; R17's URLs are U2 and U3) |
| CORRECTED | 30 | H1, H5, H9; S2, S3, S6, S7, S11; D2, D8, D9, D11, D12, D18, D24, D25; R1, R4, R5, R6, R8, R13; F1; U1–U7 (seven URL-only fixes) |
| NOT IN SOURCE | 5 | S8 ("this winter"), D7 ("announced"), D15 ("because the money ran short"), D17 (land-trust evidence: the cited sources are silent; true on Wang et al., now cited), R7 (Code for America's "six trials") |
| NEWER DATA | 1 | D14 (63% in spring 2022 → 44% in spring 2025; applied) |
| UNREACHABLE / BLOCKED / UNFETCHABLE | 0 | Hosts that blocked curl or headless Chrome were read another way (Method). Frandsen 2021's full text is paywalled; its abstract was read. |
(38 + 30 + 5 + 1 = 74. H5's "free" sits inside a CORRECTED row and isn't counted again.)
Weakest links now on the page, by grade: two working papers carry figures (Haushofer & Shapiro's 40%, preprint; Banerjee et al.'s Gini, preprint); the land-trust clause rests on a working paper about shared-equity homes generally (preprint); the Frandsen summary rests on the abstract only; and the 98.5% is the IRS's own review, with no independent anonymous test since 2013.
10. Proposals, not applied (outside this file or this brief)
source/local/copy.jsoninequality[0]: "New York City has free lawyers in every ZIP code" → "New York City has free legal help in every ZIP code" (H1).source/local/copy.jsoninequality[1]: "{legalaid:Find free legal aid near you}" → "{legalaid:Find legal aid near you}" (H5). Infinders/legalaid:https://www.lsc.gov/about-lsc/what-legal-aid/i-need-legal-help?address={zip}(U1; the new path takes the parameter: 10001 → Legal Services NYC, nonsense → no match).inequality[6]says "a right to a free lawyer in {state}". NYC's right is full representation for income-eligible tenants or those 60 and over, and advice otherwise (needs A5.2). Check each listed city's rule before keeping "free lawyer".- "see the site's impact review" (this page's s2; child-hunger s13; trafficking s6) points at
source/impact/, which is in the private repo, so readers can't open it. Either publish the reviews or cite the papers directly. - The lede promises that each action "says whether it touches poverty or inequality". Steps 2–4 don't say. Using the review's "Moves" column: the bill touches the rules (inequality), tax returns touch poverty, and tenant counsel touches poverty and housing. Add one clause each, or soften the lede.
- Accuracy bullet: cite TIGTA 2016-40-045 to restore, with a source, that volunteers know they are being reviewed, and to keep the other side of the contest: TIGTA re-checked 44 of the IRS's sampled returns and found 42 (95%) accurate. tigta.gov is a new host, so the gate would lock.
- Link the three citations added without links: Banerjee et al. (econweb.ucsd.edu/~pniehaus/papers/UBI_main_paper.pdf), Wang et al. (www.lincolninst.edu/app/uploads/legacy-files/pubfiles/wang_wp19rw1_rev_0.pdf), Farber et al. (www.nber.org/papers/w24587).
content/inequality-practice.htmlfooter: "The wage floor is next decided by 1 March 2027." Use F1's wording.- s1's Oxfam 2018 link: point it at oxfam.org/en/press-releases/richest-1-percent-bagged-82-percent-wealth-created-last-year (301).
source/stale.tsv: OCJ's FY2026 annual report (about December 2026; replaces the 44%); NYC Free Tax Prep's 2027 limits (late November 2026).- Ticks are keyed by item text, so readers who ticked a corrected item will see it unticked once.
11. When this goes stale
- OCJ's next annual report (FY2026) is likely around December 2026, judging by the FY2025 PDF's creation date of 24 Dec 2025. It replaces the 44%.
- NYC Free Tax Prep: "Check this page in November 2026 for information for the 2027 tax season!" The $97,000/$68,000 limits are the 2026 season's.
- IRS VITA locator: "updated regularly from February through April"; the $70,244 names no season.
- Governor's proposal: a rise "would take effect on March 1, 2027" if the Legislature approves. Re-read before then (
data-until2027-03-01). - COPA's signature status is "as of 5 November 2026" (stale.tsv has 2026-11-21).
- GiveWell's GiveDirectly page was last updated October 2025.
Contest a fact, suggest something, or tell us what’s broken
If someone is in danger right now, don’t use this form. Call your local emergency number (911 in the US). For trafficking: 1-888-373-7888, or text 233733.
No name, no account. Your note goes into the site’s private records, nothing about your browser or connection is stored with it, and the email box is the only way we’d know who you are. The weekly review reads every note, re-reads the original source, and fixes the page if it should. A contested fact and its outcome are listed on Contested, in the review’s words, never yours.
No tracking on any link.Ticks stay on this device.What changedContested