The short version. We measured how long 33,955 City of Austin building permits took from application to issuance, found what actually predicts the wait, and tested the forecast on recent permits it had never seen. The biggest factor is not the lot. It is who applies.
The data
- Permits: the City of Austin's Issued Construction Permits records (data.austintexas.gov), building permits issued in the last 34 months. Wait = calendar days from the application date to the issue date.
- Year-by-year trends: every building permit issued from 2021 through 2025 (65,994 permits), used only for the year-by-year figures on the District benchmarks page. They are not used to fit the forecast, because review times have changed too much since 2021 for older permits to predict today's.
- Excluded: 320 permits with waits over five years, which are 1980s and 1990s permits re-issued in recent years (record artifacts, not real waits).
- Site constraints: the City's own GIS layers (zoning, FEMA and City fully developed floodplains, waterway setbacks, watershed regulation areas, the Edwards Aquifer recharge zone, local historic districts, historic landmarks, 2022 tree canopy, council districts), joined to every residential permit with map coordinates (9,705 permits).
- Builder experience: the same general contractor's issued residential permits in the 365 days before each application, counted from permit history going back about four years, with no look-ahead.
What predicts the wait
For new homes, district, year and site constraints together explain about 19% of the variation in wait time. Adding the applicant's experience raises that to 37%. First-time applicants wait a median 103 days; builders filing 25 or more permits a year wait 28.
Site constraints, measured
Multiplier on the typical wait with the constraint present, holding district, year, builder experience and the other constraints fixed (regression on log wait, robust standard errors). A constraint is called a major trigger in a check only when it adds at least 30% with p < 0.05 for that job type.
This table is the average-wait model on every permit in the window. The forecast itself is a median model fit on permits before the latest 12 months, so the effect a check quotes ("What drives this forecast") can differ from this table; each check quotes its own model's number, with the same 30% and p < 0.05 test.
Raw medians overstate these effects because constrained lots cluster in slower districts and are more often built by small builders. The aquifer recharge zone, for example, looks slower in raw numbers and is not, once location is accounted for.
The forecast
New homes and residential additions and remodels use quantile regression: the 10th, 50th and 90th percentiles of log(days + 1) as functions of council district, the four site constraints that matter (100-year floodplain, waterway setback, historic district or landmark, drinking water protection zone) and builder experience. The band is then widened or narrowed so that about 80% of permits issued in the most recent 12 months fall inside it (conformal calibration). Other job types use the district's own 10th, 50th and 90th percentiles, or citywide figures where a district has fewer than 20 permits.
How it tested
Rolling test: the model was fit on older permits and asked to forecast the newest 12 months, which it never saw. Typical error is the median absolute error on the log scale (relative error), the right yardstick for waits that run from days to years.
For new homes the forecast's typical error is 6% smaller than a district-only estimate. For remodels it is no better than district alone. Both are reported because a forecast you cannot check is not worth much.
Limits, plainly
- Only issued permits are published. Applications still in review, withdrawn or denied are invisible, so real waits run longer than these figures, most of all for the newest applications.
- About a quarter of residential permits have no map coordinates and cannot be tied to site constraints; they skew fast (often new subdivision addresses).
- Roughly two-thirds of the variation is still unexplained. The likeliest cause is submittal quality and review cycles, which public records do not show. Saved lots let you record your own review cycles and dates.
- Year-to-year drift is real (volume builders slowed in 2026), which is why the model is rebuilt and recalibrated monthly.
- The City's map has no tree-by-tree inventory, so tree findings say "likely" and need an arborist.
Advisory only. Dirt to Permit is not affiliated with the City of Austin, which makes every permitting decision.