Permits Are Not Homes

# Permits Are Not Homes ### How long until new housing reaches renters? ![[permits-pipeline-aug-2026.png]] > [!abstract] TL;DR > - In August 2026 the U.S. authorized homes at a **1.39M annual rate**, but a permit is ==permission to build, not a key to an apartment==. > - The homes you can actually move into sit at the **end** of the pipeline: **1.27M units under construction**, and **280K** permitted but not started. > - Timing depends heavily on the project type. Single-family averages **8.8 months** from permit to completion. Apartment buildings average **18.9 months**, and much longer in big cities. > - **For a one-year decision** (renew your lease or wait?), permits are the weakest signal. Watch local buildings that are nearly finished, vacancies, and rents for your unit type. | At a glance | | |---|---| | **Question** | If a region issues more permits, should households expect relief within a year? | | **Data** | Census/HUD New Residential Construction (Aug 2026), Survey of Construction durations, Fed and academic studies, rent and price indices | | **Methods** | Stock vs. flow decomposition · ratio analysis · benchmark triangulation · reconciling indices built with different methods | | **Answer** | Probably not from permits alone. The near-term signal is further down the pipeline. | --- ## 1. The question > [!question] Does a permit boom mean cheaper rent next year? > Headlines often treat permits as a measure of "new housing." But a renter deciding whether to renew, or a buyer deciding whether to wait, needs to know **when** homes arrive, **what kind** they are, and **where**. This post follows the housing pipeline from left to right to see how much of today's permit number could plausibly become a home a household can use within the next year. ## 2. How I approached it ```mermaid flowchart LR A[Frame the decision question] --> B[Separate flows from stocks] B --> C[Descriptive ratios with cohort caveats] C --> D[Segment by structure type] D --> E[Triangulate timing from published benchmarks] E --> F[Reconcile rent & price signals across methods] F --> G[Translate into a monitoring framework] ``` > [!info]- Method toolkit (click to expand) > | Step | Technique | Why it matters here | > |---|---|---| > | Flows vs. stocks | Separate *rates* (permits, starts, completions per year) from *levels* (units sitting in a stage) | Adding or dividing across these two kinds of measure gives misleading answers | > | Ratios | Same-period ratios between stages | Useful for comparing stages, **not** a conversion probability | > | Segmentation | Split by structure type (5+ units vs. single-family) | Renters and buyers are served by different pipelines | > | Triangulation | Compare durations from Census, the Atlanta Fed, and UCLA | Each measures a different "clock" | > | Reconciliation | Compare indices built with different methods (ZORI vs. Apartment List) | Explains why the signals appear to disagree | > > Related notes: [[Exploratory Data Analysis]] · [[Statistical Knowledge]] · [[Survival Analysis]] --- ## 3. Reading the pipeline correctly The August 2026 national numbers[^nrc]: | Pipeline stage | Type | August 2026 | |---|---|---:| | Permits | Flow (SAAR) | **1,394,000** | | Starts | Flow (SAAR) | **1,275,000** | | Completions | Flow (SAAR) | **1,128,000** | | Authorized, not started | Stock | **280,000** | | Under construction | Stock | **1,271,000** | > [!note] What is a SAAR? > A **seasonally adjusted annual rate** turns one month of activity into a yearly pace: > $\text{SAAR}_t = 12 \times \text{SA}(x_t)$ > where $\text{SA}(x_t)$ is the seasonally adjusted monthly count. So "1,394,000 permits" does **not** mean 1.39M homes were permitted in August. It means August's pace, if held for a year, would reach that total. ### The ratio trap It is tempting to divide one stage by another and call the result a "conversion rate": $ \frac{\text{Starts}}{\text{Permits}} = \frac{1{,}275}{1{,}394} \approx 0.915 \qquad \frac{\text{Completions}}{\text{Starts}} = \frac{1{,}128}{1{,}275} \approx 0.885 \qquad \frac{\text{Completions}}{\text{Permits}} \approx 0.809 $ > [!warning] These are not probabilities > A project completed in August may have been permitted a year or more earlier. A project permitted in August may not start for months. The numerator and denominator describe **different cohorts of projects**, so the ratios compare activity levels. They do *not* say what share of August permits will become homes. > > Estimating a real conversion rate needs **project-level data** that links each permit to its start and completion. See [[#8. What I would build next]]. ### From permit to occupancy ```mermaid flowchart LR P["🗂️ Permit authorization"] --> S["🏗️ Start construction begins"] S --> U["🔨 Under construction"] U --> C["🏠 Completion structure finished"] C --> O["🔑 Available & occupied inspected · marketed · leased"] style P fill:#cde2fb,stroke:#2a78d6 style O fill:#2a78d6,color:#fff,stroke:#2a78d6 ``` Completion is the **closest measured national stage** to delivery, but it still isn't occupancy. Completed units may still need inspections, certificates of occupancy, marketing, and lease-up, and some are held off the market. --- ## 4. The near-term pipeline is smaller than the headline At the end of August: - **1,271,000** units were under construction - **280,000** authorized units had **not started**, about **0.22×** the under-construction stock ($280 / 1{,}271$) Before a permit turns into a construction project, several things can still go wrong: > [!failure]- Why permitted projects stall > - Financing doesn't close > - Construction costs change > - Market demand weakens > - Local approvals or inspections take longer > - Labor or materials are unavailable > - The project is redesigned, delayed, or abandoned > > Census removes canceled, abandoned, expired, and revoked permits from the not-started stock[^nrc], but a permit that remains still has no delivery date. > [!tip] Monitoring principle > For a **one-year decision**, a project that is already *under construction* is more useful to track than one that is only *authorized*. That is a practical rule of thumb, not a statistical finding about individual projects. --- ## 5. It isn't one pipeline Of the 1,271,000 units under construction: | Structure type | Units | Share | |---|---:|---:| | 5+ unit buildings (proxy for rental supply) | 666,000 | **52.4%** | | Single-family | 589,000 | **46.3%** | | 2–4 unit buildings (remainder) | 16,000 | 1.3% | > [!quote] The national housing pipeline is not one pipeline. A renter needs a **nearby apartment**, available **on the right date**, at a **manageable rent**. A buyer cares about single-family inventory, mortgage rates, and prices of existing homes. So a large national permit number can exist alongside local shortages of: - rental units in a particular city - smaller apartments - moderately priced homes - housing near jobs or transit - units available *within the next year* --- ## 6. Timing: there is no single national countdown ![[permits-time-to-completion.png]] ### Average time from permit to completion (Census, 2025)[^time][^nahb-sf][^nahb-mf] | Structure | Permit → start | Start → completion | **Permit → completion** | 2024 | |---|---:|---:|---:|---:| | Single-family | 1.4 mo | 7.4 mo | **8.8 mo** | 9.1 mo | | 2+ units | — | 16.5 mo | **18.9 mo** | 19.6 mo | ### The spread matters as much as the average[^pct] | Start → completion | Single-family | 2+ units | |---|---:|---:| | Completed within 6 months | **59%** | **10%** | | Took 13+ months | — | **60%** | > [!example]- Regional and project-size differences > **Multifamily permit → completion, projects completed in 2025** > > | Region | Months | > |---|---:| > | Northeast | 21.9 | > | West | 20.6 | > | South | 17.5 | > | Midwest | 16.6 | > > **By project value** (Census construction spending survey[^c30]): average start → completion ranges from **12.1 months** for multifamily projects under $3M to **36.3 months** for projects of $100M or more. ### When the clock starts earlier Census measures from the **permit**, but apartment projects start well before that, with design and entitlement work. - **Atlanta Fed / Cal Poly**[^atl]: from announcement or planning to completion averaged **27.6 months** (2008–2022). Projects from 2022 took about **30 months**. - **Los Angeles** (UCLA Anderson)[^ucla]: project-weighted mean development time of **1,413 days (≈46 months)**. The unit-weighted mean is **1,784 days (≈59 months)**, and the interquartile range is **946–1,739 days (≈31–57 months)**. > [!info] Reading the LA numbers > The **unit-weighted** mean is longer than the project-weighted mean, which suggests that **larger projects take longer**. That matches the project-value pattern above. The LA study also includes unfinished projects, so it is not a national forecast. It does show how poorly a permit count predicts availability in the near term. > [!important] Takeaway > For multifamily, **12 months is often too short** a window for a new permit to turn into a leasable apartment. --- ## 7. Conditions that can speed up or slow down the pipeline These indicators describe the **broad construction environment**. They don't directly measure residential delays. | Indicator | Latest reading | Read it as | |---|---|---| | Construction input prices (PPI)[^ppi] | **+8.8%** YoY, **+1.3%** MoM (Aug 2026) | Cost pressure on project feasibility (excludes labor, imports, capital investment) | | Construction payroll employment[^ces] | **8,239K → 8,359K** (Aug 2025 → Aug 2026) | Industry capacity | | Construction job openings[^jolts] | **240K** avg 2025 (vs. 301K in 2024, 400K in 2022) | A cooling labor market | | 30-yr fixed mortgage rate[^fm] | **6.95%** (week ending Sep 17, 2026) | Buyer affordability and project financing | | Banks' multifamily lending standards (Q2 2026)[^sloos] | **9.4%** tightened · **15.1%** eased · **75.5%** unchanged | Net slight easing for developers | ### What prices and vacancies say ![[permits-rent-price-signals.png]] > [!question]- Why does one rent measure fall while the other rises? > They measure different things: > - **Zillow ZORI**[^zillow][^zori] is a *smoothed asking-rent* index: **$1,948/mo**, **+2.5%** YoY. Multifamily asking rents rose **+1.9%**. > - **Apartment List**[^al][^al-m] estimates *new-lease* rents from **repeat-leased units** with a 3-month moving average: **−0.8%** YoY. > > Different methods, segments, and sampling can point in different directions. Neither is "wrong," and neither forecasts rents in *your* neighborhood. - **Rental vacancy**[^hvs] rose from **7.0%** (Q2 2025) to **7.3%** (Q2 2026). That's a bit more slack, but the data don't say whether the vacant units are in the right places or price ranges. - **House prices**[^fhfa] (FHFA purchase-only): **+2.13%** YoY and **+0.35%** QoQ in Q2 2026. That doesn't show that permits cause, or will cause, lower prices. Demand can push up permits and prices at the same time. - **Lead–lag evidence**[^frbsf]: the San Francisco Fed found that pipeline measures led new-lease rent indicators by about **one year** in its study period. That is a reason to watch the pipeline, but not a universal timeline or a causal effect. --- ## 8. What I would build next The aggregate data can't answer the question households actually care about: ==what is the probability that a given project delivers units by a given date?== Answering it needs **linked project-level data** (permit → start → completion → occupancy, plus cancellations and rents). With that data, this becomes a **time-to-event** problem. See [[Survival Analysis]]. > [!example]- Sketch: modeling delivery as a survival problem > Let $T$ be the time from permit to completion. Projects still unfinished at the data cutoff are **right-censored**, and treating them as failures or dropping them would bias the estimates (a real issue in the LA data). > > **Kaplan–Meier** estimate of the probability that a project is *not yet complete* after $t$ months: > $\hat S(t) = \prod_{t_i \le t}\left(1 - \frac{d_i}{n_i}\right)$ > where $d_i$ is the number of completions at $t_i$ and $n_i$ is the number of projects still at risk. > > **Cox proportional hazards** model, to see how structure type, size, region, and financing shift the completion rate: > $h(t \mid x) = h_0(t)\,\exp(\beta_1\,\text{units} + \beta_2\,\text{region} + \beta_3\,\text{5+ units} + \dots)$ > > A **competing-risks** extension would treat *cancellation* as a separate outcome from *completion*. > > **Lead–lag check** between pipeline stocks and rents, using cross-correlation at lag $k$ months: > $\rho(k) = \operatorname{corr}\big(\text{UnderConstruction}_{t},\ \Delta\text{Rent}_{t+k}\big)$ --- ## 9. What this means for you > [!tip]- 🏢 Deciding whether to renew a lease > Don't assume a recent permit level means comparable apartments will be available within a year. Look closer to delivery: > - Projects already under construction, and buildings nearly finished > - Units being pre-leased > - Local vacancy rates > - Asking rents **for the unit type you need** > - New supply in the neighborhood where you'd actually live > [!tip]- 🧳 Considering a move > Compare the local pipeline with local demand: > - Is population or employment growing? > - Are vacancies rising? Is growth in asking rents slowing? > - Are new projects concentrated in luxury segments? > - Are units being built near the jobs and neighborhoods that matter to you? > - Are projects **under construction** or only **permitted**? > [!tip]- 🏡 Considering buying > Separate the new-home pipeline from the wider ownership market. Single-family construction matters, and so do mortgage rates, existing-home inventory, local prices, income, employment, and the mix of entry-level and higher-priced homes. ### A three-tier monitoring dashboard (6–18 month horizon) | Tier | Stage | Signal strength for a 1-year decision | Watch out for | |---|---|---|---| | 🟢 1 | **Completed units** | Strongest | Still may not be occupied, affordable, or well located | | 🟡 2 | **Under construction** | Medium | Delivery date depends on type, size, financing, and local processes | | 🔴 3 | **Authorized, not started** | Weakest | The most exposed to delay, redesign, or cancellation | This is a **monitoring horizon, not a delivery forecast**. --- ## 10. What this analysis does and doesn't claim > [!warning]- Limitations & methodology notes > - **No trend claim.** The August levels are verified, but the **direction of year-over-year change** in national permits, starts, and completions is not established here. That would require checking the data vintage, comparison months, seasonal treatment, and revisions. > - **Ratios ≠ conversion rates.** Same-period flow comparisons mix project cohorts. > - **Stocks ≠ guarantees.** "Authorized" and "under construction" are *potential* supply. > - **Proxy mismatch.** 5+ unit buildings stand in for rental supply. Duration benchmarks use 2+ units, which is a slightly different category. > - **No new duration model.** The timing figures are **published benchmarks**. My automated duration summary failed because the duration data couldn't be processed, so no new completion distribution or delivery countdown is claimed. > - **Context ≠ cause.** PPI, employment, job openings, mortgage rates, and lending surveys describe conditions. They don't directly measure residential delays. > - **National ≠ local.** None of these figures forecast rents, prices, vacancies, or delivery dates in a specific metro. > - **Completion ≠ occupancy.** > - **Preliminary data.** August 2026 figures may be revised. > - An earlier draft included a construction-cost chart whose underlying data couldn't be verified. I **removed** it rather than show unsupported numbers. --- ## Bottom line The August 2026 permit number signals **potential future supply**, not **near-term affordability relief**. The pipeline is large, but it moves at different speeds, and the homes people can use sit at the very end of it. > [!quote] The real question > Not *"how many homes were approved?"* > but ==**"how many suitable homes will be delivered, where and when households need them?"**== --- > [!note]- Further reading > - [Federal Reserve: H.8 Bank Loans](https://www.federalreserve.gov/releases/h8/current/) > - [Seattle SDCI: Construction Permit Performance](https://www.seattle.gov/construction-and-inspections/about-us/construction-permit-performance) ### More case studies ```dataview TABLE WITHOUT ID file.link AS "Case study", dateformat(file.mtime, "yyyy-MM-dd") AS "Updated" FROM "Analytics Cases" AND #projects WHERE file.name != this.file.name SORT file.mtime DESC ``` [^nrc]: U.S. Census Bureau & HUD, *New Residential Construction* ([current release](https://www.census.gov/construction/nrc/current/index.html) · [PDF](https://www.census.gov/construction/nrc/pdf/newresconst.pdf)). [^time]: U.S. Census Bureau/HUD, [Construction Time](https://www.census.gov/construction/nrc/data/time.html). [^nahb-sf]: NAHB, [Single-Family Home Construction Time Declines in 2025](https://eyeonhousing.org/2026/09/single-family-home-construction-time-declines-in-2025/). [^nahb-mf]: NAHB, [Shorter Apartment Construction Time in 2025](https://eyeonhousing.org/2026/09/shorter-apartment-construction-time-in-2025/). [^pct]: U.S. Census Bureau, [Percentage of Units Completed by Time from Start to Completion, 2025](https://www.census.gov/construction/nrc/pdf/pct_start_to_comp_2025.pdf). [^c30]: U.S. Census Bureau, [Construction Spending Survey](https://www.census.gov/construction/c30/pdf/t325.pdf). [^atl]: Federal Reserve Bank of Atlanta & Cal Poly, [How Long Does It Take to Build Multifamily Housing?](https://www.atlantafed.org/-/media/Project/Atlanta/FRBA/Documents/research/publication/policy-hub/2024/06/18/04-how-long-does-it-take-to-build-multifamily-housing.pdf) [^ucla]: UCLA Anderson, [Los Angeles Multifamily Development Study](https://www.anderson.ucla.edu/sites/default/files/document/2026-05/2026-03wp-rev.pdf). [^hvs]: U.S. Census Bureau, [Housing Vacancy Survey](https://www.census.gov/housing/hvs/current/). [^zillow]: Zillow, [August 2026 Rent Report](https://www.zillow.com/research/august-2026-rent-report/). [^zori]: Zillow, [ZORI Methodology](https://www.zillow.com/research/methodology-zori-repeat-rent-27092/). [^al]: Apartment List, [National Rent Data](https://www.apartmentlist.com/research/national-rent-data). [^al-m]: Apartment List, [Rent Estimate Methodology](https://www.apartmentlist.com/research/rent-estimate-methodology). [^fhfa]: FHFA, [House Price Index Report, 2026 Q2](https://www.fhfa.gov/document/d/hpi/fhfa-house-price-index-report-2026q2). [^ppi]: BLS, [PPI Detailed Report, August 2026](https://www.bls.gov/ppi/detailed-report/ppi-detailed-report-august-2026.pdf). [^ces]: BLS, [Construction Payroll Employment](https://www.bls.gov/news.release/charts/empsit-ces_rc_empconit.stm). [^jolts]: BLS, [JOLTS Job Openings](https://www.bls.gov/news.release/jolts.t15.htm). [^fm]: Freddie Mac, [Mortgage Rates](https://myhome.freddiemac.com/buying/mortgage-rates). [^sloos]: Federal Reserve, [Senior Loan Officer Opinion Survey](https://www.federalreserve.gov/data/sloos/sloos-202607-table-1.htm). [^frbsf]: Federal Reserve Bank of San Francisco, [When Is Shelter Services Inflation Coming Down?](https://www.frbsf.org/research-and-insights/publications/economic-letter/2024/09/when-is-shelter-services-inflation-coming-down/)