f(Y002) / MARKET INSIGHTS / PYTHON PIPELINE + PUBLISHED RESULTS
Electric cars across 50 states, 2025 to mid-2026.
How did electric car buying change in every state when the federal tax credit ended? Which states buy more electric cars, and why? Do people who search for electric cars buy them? And do electric car owners complain about different things than owners of similar gas cars?
How it started
Three short, anonymous conversations in Metro Detroit gave me the ideas to test: repair worries, charging costs, no place to charge at an apartment, cold winters, state policy, too few public chargers, and truck-heavy markets. Public data then tested which of those show up at the state level.
What this is not
It is not a sales forecast, a survey of buyer motives, or a causal estimate. State averages say nothing about any single buyer.
- Markets
- All 50 states and Washington, DC
- Period
- Q1 2025 to Q2 2026: six quarters spanning the end of the federal credit
- Main measure
- Battery-electric + plug-in hybrid + hydrogen share of new light-duty registrations
- Other data
- Census income and housing, truck share, chargers, gas prices, state fees, Google searches, federal safety complaints
- Unit of analysis
- State × quarter: 306 rows
01 / SNAPSHOT
A rush, then a cliff.
Buyers rushed to beat the September 30, 2025 deadline, and the electric share fell almost everywhere afterward. By spring 2026 it had recovered only part of the way.
02 / ELECTRIC SHARE OF NEW REGISTRATIONS
The credit ended. The share halved.
National electric share of new registrations by quarter, against any state you pick. The dashed line marks the end of the federal credit.
Source: Alliance for Automotive Innovation, Get Connected EV Quarterly Reports, Q1 2025 to Q2 2026. The Q1 2025 state tables were images; they were typed in by hand and checked two ways (components add to totals; every state matches the year-on-year change in the next year's report).
03 / EVERY STATE
Almost every state fell by a similar proportion.
Percent change in electric share from Q3 to Q4 2025. Richer or poorer, truck-heavy or not, the typical state lost about half its share. Texas and Hawaii are the outliers on the small side, Michigan on the large side.
All 51, sortable
Click a column header to sort. Year-on-year change is in percentage points (5% to 7% is +2 points); the credit-end drop is in percent.
04 / MICHIGAN
The home market depended on the credit.
Based on its income, housing, truck share and fees, Michigan "should" have averaged far less electric buying than it did while the credit lasted. After the credit ended, most of that extra disappeared.
05 / WHAT GOES WITH ELECTRIC BUYING
Income and trucks, mostly.
Each bar is the difference in average electric share (percentage points) that goes with a typical, one-standard-deviation difference between states in that trait, holding the others steady. Lines show 95% confidence intervals; a line crossing zero means no clear link.
- The base model explains about — of the differences between states, almost all of it from income and trucks.
- Apartments, renters and special registration fees show no clear link once income is included.
- Public chargers line up strongly with electric buying on their own (correlation 0.69), but not once income is included. Chargers also get built where electric cars already are, so the arrow can point both ways.
- Within each state over time, quarters with bigger gas price jumps did not bring a higher electric share (—). Six quarters dominated by one national shock is too short to see much more.
n = 51; robust (HC3) standard errors; all traits standardized. Re-run without California and DC to check that two places weren't driving the result.
06 / INTEREST VERSUS BUYING
States that search, buy.
Each dot is a state, ranked on Google search interest and on electric share in Q2 2026 (1 = highest). Dots near the diagonal search and buy in the same order. State rankings agreed closely in every quarter.
07 / SEGMENTS
Three kinds of state market.
Two patterns run underneath ten state traits: electric readiness (searches, electric share, gas prices, chargers, the zero-emission rule, income) and urban density (renters and apartments). States group into three segments. A second method agreed only moderately, so states near the edges could fall into a neighboring group.
08 / OWNER COMPLAINTS
It depends on the car, not on being electric.
Share of federal safety complaints that involve each part of the car, for each electric model against the gas model from the same company that is closest in size, same model years on both sides. A complaint can involve more than one part.
Source: National Highway Traffic Safety Administration complaints filed January 2025 to June 2026. Names, addresses, vehicle identification numbers and dealer details are never read. Word-search flags such as "dealer" and "mechanic" are held back until a hand check of 120 complaints is done.
09 / WHAT THIS MEANS FOR THE CAR INDUSTRY
Plan state by state.
- Plan state by state. The same national message lands very differently where electric cars are 1 in 5 new sales and where they are 1 in 50.
- Treat incentives as a cliff, not a slope. When a credit ends on a known date, buyers move purchases forward. Expect a spike before the deadline and a hole after it, and plan inventory and pricing around both.
- Watch the hometown market. Michigan's electric buying was unusually dependent on the credit. Without it, Detroit brands need another reason for local buyers to choose electric.
- Use complaint data model by model. One recall on one charging part can make one electric model look much worse than gas while another looks better. Publishing that difference, or fixing its cause, answers buyers' repair worries directly.
- Search data is a cheap early warning. Interest rebounded well before registrations did, and it ranks states almost the same way. It can be watched every week for free.
- For researchers: measure national shocks with a comparison group. Comparing quarters within states cancels anything that hit everyone at once, so a credit that ended everywhere needs an unaffected group or a longer series.
10 / METHODOLOGY
Simple enough to explain.
1 · Describe
National and state trends quarter by quarter, with the dates of key events: the tax credit law, the deadline, and new cheaper models.
2 · Compare states
EVShares = β·z(traitss) + ε
Four or five traits at a time across 51 states, all on the same scale, robust standard errors.
3 · Follow states over time
EVShares,q = αs + γq + β·Xs,q + ε
State and quarter fixed effects, errors clustered by state. Cancels what is fixed about a state and what hit everyone at once, which is why it can't measure the credit itself.
4 · Segment
Factor analysis of ten traits, then k-means clustering (k = 3, best silhouette), checked against a Gaussian mixture.
5 · Interest versus buying
Interest = EV searches ÷ (EV + car dealership searches)
A ratio, because the "Hybrid vehicle" term collapsed in every state at once in late 2025 — a change on Google's side, not in people.
6 · Complaints
Matched gas twins (Equinox EV / Equinox, Ioniq 5 / Tucson, Mach-E / Escape), two-proportion tests with a Holm adjustment for many tests.
Known limitations
- Survivorship: complaints come only from owners who filed one; registrations count new cars only.
- Gas prices move for their own reasons, and registrations also reflect state rebates, dealer stock, model availability, fleets and financing. State rebates are not included.
- Cold weather was not tested: the cold quarters are the same quarters the credit ended. Electricity prices were not tested yet.
- Google groups the Equinox EV and Mach-E with their gas namesakes, so those are "model-family interest". Indiana's Q2 2026 jump was a fleet registration spike (128 to 3,080 cars).
- The three interviews gave ideas to test, nothing more.
11 / SOURCES AND REPRODUCIBILITY
Reproduce the work.
Every figure on this page comes from the research repository's published output (outputs/site_data/ev_insights_data.js); the page draws charts and tables only and calculates nothing. The Alliance reports and the daily gas file are not redistributed: download your own copies and follow their terms.
pip install -r requirements.txt(Python 3.10+, plus the freepdftotexttool)- Put the source files in
data/raw/as listed in the repository README python run_all.py— runsstage_1_data/thenstage_2_analysis/in order and stops at the first problempython -m pytest tests/
Sources
- Alliance for Automotive Innovation — Get Connected EV Quarterly Reports, Q1 2025 to Q2 2026 (e.g. 2026 Q2 ↗︎)
- Daily state gas prices collected from AAA (jgreathouse9/AAAGas) ↗︎ — independent; not affiliated with AAA
- U.S. Census Bureau population estimates (2025) and American Community Survey (2024): income, renters, housing
- Federal Highway Administration, Highway Statistics 2023, table MV-1: cars versus trucks
- ARTBA (Sept 2025) and NCSL (July 2026): special electric car registration fees
- NHTSA complaint database ↗︎ and Google Trends
Research repository ↗︎Full methodologyInterview notes
Data bundle generated —. Code and documentation MIT-licensed; source data keeps its own terms. Not for individual-level inference, sales estimates or causal claims.