Methodology

How the Winter Home Emergency Index and the Summer AC Emergency Index are built, and what they can and can't tell you.

Cost estimates

Cost pages are Home Stats Data estimates, built the same way for every job, from the same data, on the same page template, and labeled as estimates. Where the American Housing Survey reports what homeowners paid for that kind of project, the page shows it next to the estimate as a check.

  1. The trade's billing rate per hour of work, from the Census Bureau's 2022 Economic Census: the trade's net value of construction work divided by its construction-worker hours. It includes materials, overhead and profit at the trade's average mix.
  2. Moved to current dollars with the BLS producer price index for that trade (2022 average to the latest month). Where BLS has no residential index, we use the closest one and say which.
  3. Times worker-hours for the job: our own estimate, given as a low–high range with the reasoning on each page. No free official source publishes labor hours per job, and paid estimating guides are not used.
  4. Adjusted by state: the trade's payroll-and-benefits share of work value is scaled by the state's BLS wage for the trade, relative to the US average. Materials and the rest stay at the US average.
  5. Checked against reality: the American Housing Survey 2023 median for the matching project type (homeowner-reported, 2021–2023, including do-it-yourself jobs), moved to current dollars with the same price index.

Each cost page lists every input with its source and shows the formula, so anyone can check the estimate or rerun it with their own labor hours.

Tornadoes

The Tornado Alley map uses every tornado in NOAA's Storm Events Database for 2021–2025. NOAA logs a tornado once for each county it crosses, so figures are county tornado reports. Each county's yearly average is divided by its area, computed from Census map boundaries. States with fewer than 10 reports in five years are not ranked per square mile, because a few reports in a small place swing the figure.

Hail

Hail figures come from NOAA's Storm Events Database (National Centers for Environmental Information), 2021–2025: every hail report to the National Weather Service with its county, date and stone size. We average each year's counts. "1-inch+" is the National Weather Service's severe-hail threshold. A hail day is a date with at least one report. Reports depend on people seeing and reporting hail, so populous places are overcounted relative to rural ones. Property damage is not used because most hail reports list none.

Electric bills

State figures come from the US Energy Information Administration's annual electric utility data for 2025 (the Average Monthly Bill table, from forms EIA-861, 861S and 861U, which every electric utility files). EIA publishes each year's data the following October. For the current year, we compare EIA's year-to-date average residential price (Electric Power Monthly, Table 5.6.B) with the same months of the year before, and apply that change to the full-year bill. That figure is labeled as our estimate, and it assumes usage stays the same. The average monthly bill is residential revenue ÷ residential customers ÷ 12, so it includes delivery charges and fees. To show whether price or usage drives a state's bill, each state page also gives its bill at US-average usage (its price × US usage) and at the US price (US price × its usage).

Property taxes

State and county figures come from the Census American Community Survey 2020–2024 five-year estimates: table B25103 (median real estate taxes paid by owner-occupied households) and B25077 (median value of owner-occupied homes). The effective rate is median taxes ÷ median value. It shows the typical burden, not a statutory rate. Census reports taxes of $10,000 or more as "$10,000+" and values of $2 million or more as "$2,000,000+"; those figures are lower bounds and are marked. Rates differ from some other rankings because of method: the Tax Foundation, for example, divides total taxes paid by the total value of owner-occupied homes from the one-year 2024 survey (B25090 and B25082), which gives New Jersey 1.88% against our 2.11%. Totals give more weight to expensive homes and medians to the typical one, and a single year reflects higher recent home values than a five-year average. County estimates whose margin of error is over 30% of the estimate are flagged as small samples.

Home Emergency Indexes: what they measure

Both indexes measure search demand: how often people in a city search Google for help with heating, frozen or burst pipes, emergency plumbing and air conditioning. It is not a count of breakdowns, repairs or insurance claims.

Cities

The 100 largest US cities by Census 2024 population estimates. "City" means the location as Google Ads defines it. Households are for city limits, from the Census American Community Survey (ACS) 2023 5-year estimates.

Search data

Monthly Google Ads search volume for each city, 2022-10 to 2026-09: four full winters and four summers. The two studies use separate keyword baskets: 30 winter terms and 21 summer terms.

Winter Home Emergency Index: furnace repair, furnace repair near me, emergency furnace repair, furnace not working, furnace not turning on, furnace service, heater repair, heater repair near me, heater not working, heating repair, heating repair near me, emergency heating repair, heat pump repair, heat pump repair near me, heat not working, no heat, no heat in house, emergency plumber, emergency plumber near me, 24 hour plumber, 24 hour emergency plumber, frozen pipes, frozen pipe, frozen pipe repair, how to thaw frozen pipes, burst pipe, burst pipes, pipe burst, burst pipe repair, water pipe burst.

Summer AC Emergency Index: ac repair, ac repair near me, ac repair service, air conditioner repair, air conditioner repair near me, air conditioning repair, air conditioning repair near me, central air repair, ac service, ac service near me, emergency ac repair, emergency ac repair near me, 24 hour ac repair, ac not blowing cold air, ac not cooling, ac not working, ac unit not cooling, air conditioner not working, air conditioner not cooling, ac blowing warm air, ac stopped working.

Two fixes to the raw data

  1. Close variants counted once. Google Ads often reports one shared number for close variants, such as "ac repair" and "air conditioner repair". Within each city, identical monthly series count once.
  2. A typical month, not one month. Some city-months come in at several times the same month in other years, with no matching weather event. Each headline figure uses the median of that calendar month across the four years, so one bad year drops out.

The measures

  • Searches per 1,000 households: the typical December–February (winter) or June–August (summer) total, divided by households. This sets the rank.
  • vs weather: searches compared with what a city's normal cold (or heat) predicts. We fit one line through all cities: the log of searches per 1,000 households against the log of 1991–2020 NOAA heating degree days (cooling degree days for summer). A city's figure is its actual searches divided by the line's prediction for its climate, so 2.00x means twice the predicted level. Climate explains about 36% of the variation between cities in winter and 46% in summer (R²), so the measure is a rough guide. Honolulu has almost no heating degree days, so it has no winter comparison.
  • Spike ratio: the typical peak month divided by the yearly average. For a few cities, mostly warm and low-volume, the typical peak lands in an unusual month; we don't report a spike ratio or peak month for them.
  • Urgent share: the share of winter searches for urgent terms (burst or frozen pipes, no heat, emergency furnace or heating repair).

Ranking check

Dividing by city-limits households can overstate cities whose limits hold a small share of their metro, because Google's city location covers a wider area. Dividing by metro population has the opposite bias: it favours cities that make up most of their metro. So every city is ranked both ways, and a city is confirmed only if it ranks in the top 25 on both. Headlines, charts and per-household claims use confirmed cities only; the full ranking is published as measured, with each city's check result.

We also compared the winter results with Google Trends metro data for the 2024-25 and 2025-26 winters. Confirmed leaders such as Pittsburgh, Cleveland and Denver rank high there too; cities that fail the check, such as Atlanta, rank low.

Weather

NOAA data via the Regional Climate Centers' ACIS service, from each city's nearest long-running airport station. Three cities use a station 25 to 35 km away: Chesapeake, Winston-Salem and Corpus Christi.

Limits

  • Google Ads volumes are estimates rounded into buckets. Summing a basket of terms and using four-year medians reduces the noise but doesn't remove it.
  • Google's city locations don't match city limits exactly.
  • Searches reflect need and also habits: some people call a plumber they know instead of searching.