Does planned Seattle supportive housing cost about $500,000 an apartment?

1. Pass Title matches the evidence

   The title asks about planned Seattle supportive housing at about $500,000 an apartment. The evidence gives a median of $503,128 across the four Seattle all-supportive buildings in state applications from 2023 to 2025, and the answer narrows the scope to those four. The sample is small, but that is a judgment call the limitations already disclose, not a mismatch in direction or number.

2. Pass No obvious contradictions

   Key points 1 to 3 and 6 are marked established and match the conclusion and evidence figures. Key point 6 states the $357,000 finished median as final-report figures, which the limitation qualifies without contradicting. Key point 1 and the conclusion both call the $503,000 an estimate. Per-building figures recompute to the stated per-apartment costs, the 38% rise and the median. The sample-objection median of about $553,000 also recomputes from the four listed buildings.

3. Pass Plausible justification

   The conclusion explains the True rating from the state application figures, gives the reasons it is only moderate confidence (developer estimates, four buildings, three by one developer, only applicants appear), and notes the rise and inflation limits. A reader would accept that the rating follows from the evidence.

4. Pass Sources are relevant

   Each cited source is on topic for its use: Commerce Housing Trust Fund application lists and cost reports for the planned and finished costs, the Seattle Office of Housing for awards and cost pressures, King County's Health Through Housing report for purchase costs, the BAE memo for a typical apartment cost, DESC pages only for apartment type and Clover Place status, and the PPI series and 2020 council release for inflation and design-review estimates. The San Francisco Tahanan memo fits the modular objection.

5. Pass Common questions answered

   The article has twelve common questions. Each response addresses its question directly, including the affordable comparison, typical-apartment comparison, services space, finished buildings, estimate changes, inflation, outside-sample costs, buying versus building, public share, modular savings, operating costs and the causes of cost.

6. Pass Headline and verdict agree

   Title is a question. Finding True gives short answer Yes. Step b sentence: 'Does planned Seattle supportive housing cost about $500,000 an apartment? Yes.' Step c: the reader takes away that planned Seattle supportive housing costs about $500,000 an apartment. Step d: this matches the claim 'New supportive housing planned in Seattle costs about $500,000 per apartment', the True finding and the answer. The claim is not negated, so there is no polarity problem.

Coverage: I audited the article JSON as given: title, claim, finding, answer, summary, conclusion, context, method, key points, evidence items with limitations and takeaway IDs, twelve objections with responses, and the source list with takeaways and reliability notes. I did not read any source snapshot or the web. Every takeaway in the sources block is marked 'not yet checked', and no takeaway is marked established. I therefore treated source-level statements as unchecked and judged only the article's own text. Key points 1, 2, 3 and 6 are labelled established by the article, and I relied on them only as the article's text, not as verified. I recomputed the article's own figures from the takeaway numbers: per-apartment costs, the planned median of $503,128, the 2021-2022 median of $390,562, the 38% Woodland rise, the finished median of $356,813, the Lake City and Belmont shares of cost and City awards (about 61% and 45%), the sample median of about $553,000 and the 1.5 ratio to the King County figure. All matched within rounding. The inflation objection says the 2023-2025 index averages were about 4% above 2021-2022; my recomputation from the annual averages gives about 5%. That is a rounding-level difference, so I did not flag it. The chart and image were not assessed beyond their captions, since they are not among the checks.

1. Concern (minor) Confidence matches the evidence

   Add 'though its apartments are usually smaller' to the summary's typical-apartment comparison, and 'so far' to the description's King County purchase comparison.

   Moderate confidence fits the evidence. The article itself lists the reasons in its conclusion and in limitation:ev-planned: developers' estimates, four buildings, three of them DESC's, only state-fund applicants, nominal dollars, and land inclusion not verifiable on the 2024-2025 lists. The gap is in the summary and description. The summary says the cost "is about a typical new apartment's cost." That rests on a single 2023 draft consultant estimate (ev-market) for a generic building that is larger, about 700 sq ft per unit, and includes an underground garage. The summary drops the article's own caveat that supportive apartments are usually smaller. A summary-only reader could take the comparison as cost parity, when per square foot the supportive buildings likely cost more. The description also says buying "has cost King County about a third less" without the "so far" and "repairs not final" qualifiers that appear in ev-bought and key-point:4. Neither gap changes the main finding. They are editorial fixes.

   About summary, description, evidence:ev-market, limitation:ev-market, evidence:ev-bought The finding

2. Pass Rates the real claim

   The claim is "New supportive housing planned in Seattle costs about $500,000 per apartment, including land and financing." A reader who repeats it means that a new supportive housing apartment now costs roughly half a million dollars all-in. The article rates exactly that. It narrows to "the four all-supportive Seattle buildings that sought state Housing Trust Fund money since 2023" and says so in the answer, summary and method:1. The narrowing does not make the claim easier to pass. ob-sample shows that Seattle buildings with only some supportive units cost more per apartment (median about $553,000). Key points 4 and 6 give the readings a person might also mean: finished buildings at $357,000 and county purchases at $331,000. The conclusion stays within scope: "the $500,000 figure describes what new buildings cost now, not what the city's existing supportive housing cost." The time window (2023 onward) is a choice, but it matches "planned" and is stated openly. The 2021-2022 median is given beside it.

3. Pass Rating follows from the evidence

   I take ev-planned as stated. The four estimates are DESC Woodland $524,000, Plymouth Whittier $512,000, DESC Belmont $495,000 and DESC Lake City $402,000. Their median is $503,000 and their mean is about $483,000. The state fund's definition of the cost includes land and reserves, and ev-finished says the figure includes financing. Before reading the finding, I would have chosen True or Mostly true. Three of the four sit tightly around $500,000. One sits at $402,000, and the figures are estimates, but the claim says "planned," so estimates are the right measure. Even if Whittier's unchecked project type removed it, the remaining three have a median of $495,000. The article's True, with an answer that just states the median, is consistent with this evidence. The choice is defensible, and the answer does not contradict the finding.

4. Pass Strongest objections answered

   I would expect these objections to matter most: - The figures are only estimates (ob-estimates; also the 38% rise at Woodland). - The sample is small, DESC-dominated, and limited to state applicants (limitation:ev-planned, ob-sample). - The rise is just inflation (ob-inflation). - Supportive housing is pricey because of service space or small units (ob-services, ob-market). - Taxpayers don't pay it all (ob-public-share). - Buying is cheaper (ob-buy). - Operating cost and per-person framing (ob-operating). Each is stated and answered with evidence rather than assertion. ob-inflation compares the medians with a federal index over matching spans and admits the index leaves out labor and interest. ob-sample brings in the mixed buildings' costs. ob-services admits the cost can't be split out. A lesser gap: no objection asks whether the claim holds if all planned buildings from 2021 to 2025 are pooled. That pooled median is about $449,000. However, the 2021-2022 median is shown in the conclusion and key-point:3, so readers can see it.

5. Pass Evidence is independent

   Only ev-planned directly supports the claim, and it rests on one source family: Commerce's Housing Trust Fund application lists. The article does not present this as independent corroboration. method:4 says outright: "No second record gives these buildings' total costs... So the cost figures rest on the state's application lists." The City report (tk-oh-newpsh) only confirms building type and unit counts. ev-affordable comes from the same 2023 list, but it is used as a same-basis comparison, which is appropriate. The remaining items are separate datasets used for comparison, not presented as independent confirmation of the $500,000 figure: - ev-finished: Commerce and Housing Finance Commission final cost reports. - ev-market: BAE consultant memo. - ev-bought: King County report. The dependence is disclosed and reflected in the Moderate confidence.

6. Pass Lead questions hold up

   leadObjectionIds names only ob-affordable: "In Seattle, supportive housing costs more per apartment than other new affordable housing." Short answer: "No, other affordable housing costs more." 1. Surprising: passes. Most readers would guess supportive housing costs more. 2. Not the headline: passes. The title and answer say nothing about other affordable housing. 3. Settled as worded: passes. "Per apartment" is specific, and the evidence measures it the same way in the same 2023 list: a $576,000 median for 15 buildings against $402,000 and $524,000. ob-sample also shows mixed buildings at $553,000. The supportive side is only two buildings, and the response's caveat about family-sized units keeps the reading honest. 4. Neutral: passes. The article's main True rating confirms the critics' cost figure, and other objections correct claims on both sides (ob-public-share, ob-buy, ob-modular). This one does not read as one side corrected while the other is left alone. 5. Stands alone: passes. The first sentence, "Seattle's other new affordable buildings cost more per apartment: a median $576,000 in 2023 applications for state housing money," uses everyday words and makes sense on its own.

7. Pass Rating states its terms

   The finding is True, so part (a) does not apply. For part (b), the result depends somewhat on which buildings count as "new supportive housing" and what "cost" includes. key-point:1 states the reading used: buildings "where every apartment is supportive housing," counting "land, fees and financing." The likely alternative readings are given where the rating is seen: - Existing or finished buildings: key-point:6, $357,000. - Purchases of existing buildings: key-point:4, $331,000. - Earlier plans: key-point:3, $391,000. The mixed-building reading appears only in ob-sample. It is a less common reading and points the same way (higher cost), so it does not undercut the rating.

Coverage: I reviewed the full article payload: the claim, finding (True) and Moderate confidence, the answer, summary, description, key points, context, conclusion, all six evidence items with their methods and limitations, all twelve objections, the method steps, the editorial note and the source list. All takeaways in the sources block are marked "not yet checked"; none is established. Following the division of labor, I took every evidence item and figure exactly as the article states it and did not verify any against the sources. Whether the cost figures, the medians and the definitions match the source documents is unchecked here and left to the Source Checker. That includes the inclusion of land and financing on the 2024-2025 lists and Whittier's project type. I did not read source snapshots or use the web. My judgment on the rating assumes the four building estimates are as stated. If those turn out wrong, the rating could change, because the rating rests on a single source family (the state application lists).