Solar permits that never got installed: 595,060 bare roofs

A residential solar permit is a promise that something will be bolted to a roof. It is possible to check whether it was, because the roof is photographed from above and the permit is a public record. Matching the two across 2,836,257 single-family homes with a solar permit and an evaluated roof, 595,060 — 21.0% — have no array visible today. All figures derived August 29, 2026; the methodology matters on this page and is worth reading before quoting the number.

The rate follows permit status, which is how you know it is real

The first thing to ask of a measurement like this is whether the detector is simply wrong. If it were missing panels at random, the no-array rate would be roughly the same regardless of what the permit record says. It is not. It moves by a factor of four in exactly the direction it should.

Homes whose solar permits all expired show no array 65.2% of the time. Homes with a permit the jurisdiction signed off on show no array 15.7% of the time. The gradient is the validation: the imagery and the permit clerk independently agree about which jobs happened.

It also identifies the cleanest cohort in the data. 101,033 homes hold a solar permit that expired and nothing else, and 65,857 of them have a bare roof. Those are contracts that were signed, permitted, and abandoned.

By permit vintage: a curve with two different causes

Plotted against the year of the home’s most recent solar permit, the rate is a U. Both ends are high and they are high for opposite reasons.

The right-hand end is imagery lag. A permit pulled in 2026 may describe a system installed last month over a photograph taken last year, and the 49.2% figure for part-year 2026 is mostly that. Do not quote it as a failure rate. By the 2023 vintage the lag has largely washed out; by 2018–2022 it is gone, and those years sit tightly between 16.6% and 17.8%. That band is the honest structural number.

The left-hand end is different, and more interesting. Permits from 2010 and 2011 show 28.1% and 24.3% no-array rates against imagery that is entirely current. Nothing is lagging there. Those are systems that were never installed, or were installed and have since come off — through a re-roof that never got the panels back, a repossession, or a failure nobody paid to fix. Fifteen years is long enough for a residential PV system to have a history.

By state, and why the map is partly about weather

The state table below is the one place on this page where the number is measuring several things at once, and it should be read with that in mind rather than as a league table of installer conduct.

Sort that table and a climate line appears: Nevada 12.1%, Arizona 13.9%, California 17.6% at the bottom; Minnesota 40.8%, Ohio 38.9%, Maryland 32.6%, Connecticut 31.6% at the top. Snow-belt states cluster high, and a roof photographed under snow reads as a roof without panels. Part of the northern excess is real and part of it is a season, and this data cannot cleanly separate the two.

Hawaii at 35.3% is a different story with a specific cause: a large share of Hawaii’s residential “solar” permits are for solar water heating, which the state has effectively required on new homes since 2010. A solar water heater is a small collector, not a photovoltaic array, and it is not what the imagery is looking for. The permit is genuine, the absence of PV is genuine, and reading it as an abandoned installation would be wrong.

California carries the volume. 1,332,042 permitted homes, 233,954 with no visible array — 39% of the national total, at a below-average rate, because California is 47% of the permitted stock.

Why this cohort is worth working

For a solar or roofing company, a permitted home with no array is one of the more informative addresses available. Somebody in that house went through a full sales cycle for a solar system — sat through the pitch, signed, got as far as a permit — and does not have one on the roof. Whatever stopped it, the qualification step is already done and the objection is known to be surmountable.

The expired-permit cohort is the sharpest version: 65,857 homes with a dead permit and a bare roof, most of which correspond to a contract somebody walked away from. The 2010–2013 vintages are a second, quieter cohort — homes where a system may have come off during a re-roof and never went back on, which is a reinstall conversation rather than a new sale.

This is the mirror image of orphaned solar systems, where the panels are on the roof and the company that installed them is gone. Here the company may still exist and the panels never arrived. The overlap between the two — a permit from an installer that later failed, and no array — is the hardest-hit group of all, and it is why the bankrupt solar installer list is worth reading alongside this page.

Methodology

• Cohort. Single-family residential building permits that pass our photovoltaic-solar classification rule and whose own description text contains “solar”, “photovoltaic” or “PV”, on homes that carry a rooftop imagery verdict. That is 2,836,257 homes. Percentages are of that cohort, never of all homes.
• Why the description gate exists. Permit classifiers tag ordinary trade work as solar. Without the gate, Wisconsin reads a 86.8% no-array rate, and its highest-volume “solar” permit descriptions are “New single family home” (2,909), “Sf home” and “Basement drainage system”. The gate removes Wisconsin from the table entirely rather than publishing a number about basements. Any national figure computed without it is inflated.
• The denominator excludes unevaluated roofs. A home with no imagery verdict is not counted as “no panels” — that mistake would add millions of homes and roughly double the headline. 99.3% of solar-permitted homes do carry a verdict, because permitted roofs are prioritised for evaluation.
• Known detection error, stated. A panel-segmentation audit of this pool identified 7,971 homes that did have panels the classifier had missed — adjudicated by human review at 95.8% precision and since corrected in the underlying data. A further 47,464 homes are flagged as possible misses and not yet adjudicated. If every one of them turned out to have panels, the headline rate would fall by at most a point and a half. We would rather publish that bound than imply the detector is perfect.
• Permit coverage is a floor. Permit records vary in completeness by jurisdiction, so this measures homes we have a permit for, not every permitted home in the country. A home with an unrecorded permit is simply absent from the cohort — it is not counted as an install or a failure.
• Status buckets. A home counts as having a completed permit if any of its solar permits reached a completed status; “expired only” means every solar permit on the home expired and none is open. Status vocabularies are normalised across jurisdictions and a small share resolve to unknown.
• What this is not. It is not a fraud measure, an installer scorecard, or a claim about any individual company or household. Six ordinary explanations sit behind the same observation and this data cannot always tell them apart.

Derived August 29, 2026. Spot something that looks wrong? Tell us — we’d rather correct a number than defend it.

Companion pages: solar permit statistics covers the permit volumes this cohort is drawn from, what percentage of homes have solar counts the arrays from the other direction, and how to get solar leads covers turning cohorts like this one into a route.

Frequently asked questions