Builder confidence is at a 2012 low. Construction output is not.

The survey says 2012. The output books do not.

Monday’s NAHB/Wells Fargo Housing Market Index dropped to 29, the weakest reading since June 2012. The 30-year mortgage averaged 7.40 percent last week, buyer traffic is the worst since April 2012 outside the pandemic, and 43 percent of builders are cutting prices. If you lived through the last housing crash, that combination of numbers is enough to make a person check the attic for unsold drywall.

The official output books tell a different story. Real U.S. construction GDP — houses, highways, data centers, the lot — ran at an $870 billion seasonally adjusted annual rate in the second quarter of 2026, in chained 2017 dollars. That is 30.5 percent above the second quarter of 2012, the last time builder morale looked like this. It is also 3.6 percent below the 2021 Q2 peak of $902.5 billion. Construction’s share of real GDP has slipped to 3.56 percent, from 3.82 percent in 2012 Q2 and more than 4 percent in the mid-2010s. The industry is smaller relative to the economy than it was when the last crash ended. It is not small in the 2012 sense.

These figures come from the Bureau of Economic Analysis’s quarterly state GDP accounts, published at a seasonally adjusted annual rate. They measure the whole construction sector, not just single-family housing. A data-center slab and a highway interchange count. That is the point: a homebuilder survey can look like 2012 while the industry’s real output does not.

U.S. construction output is 30% above 2012, not back in the last crash

Explore chartDownload exact data

View exact chart values

period	period_start	construction_saar_bil_ch2017	total_gdp_saar_bil_ch2017	construction_share_pct	index_2012q2
2012 Q1	2012-01-01	668.46	17367	3.85	100.3
2012 Q2	2012-04-01	666.75	17444.5	3.82	100
2012 Q3	2012-07-01	665.19	17469.7	3.81	99.8
2012 Q4	2012-10-01	680.31	17489.9	3.89	102
2013 Q1	2013-01-01	689.7	17662.4	3.9	103.4
2013 Q2	2013-04-01	691.94	17709.7	3.91	103.8
2013 Q3	2013-07-01	704.8	17860.5	3.95	105.7
2013 Q4	2013-10-01	703.85	18016.1	3.91	105.6
2014 Q1	2014-01-01	707.2	17954	3.94	106.1
2014 Q2	2014-04-01	719.88	18185.9	3.96	108
2014 Q3	2014-07-01	721.91	18406.9	3.92	108.3
2014 Q4	2014-10-01	735.9	18500	3.98	110.4
2015 Q1	2015-01-01	746.14	18666.6	4	111.9
2015 Q2	2015-04-01	763.34	18782.2	4.06	114.5
2015 Q3	2015-07-01	770.8	18857.4	4.09	115.6
2015 Q4	2015-10-01	777.35	18892.2	4.11	116.6
2016 Q1	2016-01-01	796.48	19001.7	4.19	119.5
2016 Q2	2016-04-01	795.88	19062.7	4.18	119.4
2016 Q3	2016-07-01	805.28	19197.9	4.19	120.8
2016 Q4	2016-10-01	819.8	19304.4	4.25	123
2017 Q1	2017-01-01	836.14	19398.3	4.31	125.4
2017 Q2	2017-04-01	833.16	19506.9	4.27	125
2017 Q3	2017-07-01	837.16	19660.8	4.26	125.6
2017 Q4	2017-10-01	854.42	19882.4	4.3	128.1
2018 Q1	2018-01-01	864.94	20044.1	4.32	129.7
2018 Q2	2018-04-01	859.86	20150.5	4.27	129
2018 Q3	2018-07-01	865.26	20276.2	4.27	129.8
2018 Q4	2018-10-01	864.95	20304.9	4.26	129.7
2019 Q1	2019-01-01	877.3	20431.6	4.29	131.6
2019 Q2	2019-04-01	881.54	20602.3	4.28	132.2
2019 Q3	2019-07-01	887.6	20843.3	4.26	133.1
2019 Q4	2019-10-01	882.72	20985.4	4.21	132.4
2020 Q1	2020-01-01	870.08	20709.2	4.2	130.5
2020 Q2	2020-04-01	826.2	19078	4.33	123.9
2020 Q3	2020-07-01	873.39	20558.9	4.25	131
2020 Q4	2020-10-01	884.09	20791.9	4.25	132.6
2021 Q1	2021-01-01	894.21	21106.6	4.24	134.1
2021 Q2	2021-04-01	902.5	21474.1	4.2	135.4
2021 Q3	2021-07-01	885.62	21641.7	4.09	132.8
2021 Q4	2021-10-01	862.54	22001.6	3.92	129.4
2022 Q1	2022-01-01	873.51	21945.7	3.98	131
2022 Q2	2022-04-01	838	21976.3	3.81	125.7
2022 Q3	2022-07-01	807.01	22124.4	3.65	121
2022 Q4	2022-10-01	803.19	22272.5	3.61	120.5
2023 Q1	2023-01-01	802.7	22456.9	3.57	120.4
2023 Q2	2023-04-01	808.64	22577.6	3.58	121.3
2023 Q3	2023-07-01	832.41	22828.8	3.65	124.8
2023 Q4	2023-10-01	844.49	23033.2	3.67	126.7
2024 Q1	2024-01-01	858.29	23116	3.71	128.7
2024 Q2	2024-04-01	865.03	23320.1	3.71	129.7
2024 Q3	2024-07-01	867.49	23501.8	3.69	130.1
2024 Q4	2024-10-01	869.29	23641.8	3.68	130.4
2025 Q1	2025-01-01	857.3	23650.2	3.62	128.6
2025 Q2	2025-04-01	867	23884.6	3.63	130
2025 Q3	2025-07-01	866.22	24113	3.59	129.9
2025 Q4	2025-10-01	864.25	24125.5	3.58	129.6
2026 Q1	2026-01-01	863.12	24274.4	3.56	129.5
2026 Q2	2026-04-01	870.04	24408	3.56	130.5

A few housing states have already slipped

The national +0.4 percent year-over-year change from 2025 Q2 to 2026 Q2 hides a split. Florida’s real construction output fell 2.0 percent over that year. New York was down 1.7 percent. California and Georgia were each down 1.0 percent. Arizona slipped 0.6 percent. Texas rose 2.2 percent. Nevada, a much smaller market, was up 5.9 percent.

That is not a 2012-style collapse in the Sun Belt. It is a stall with a couple of large housing states already in reverse, while Texas — still pouring slabs for warehouses and power as well as houses — kept adding. Because the BEA series does not split residential from nonresidential construction in this extract, a Texas gain is not proof that Texas homebuilding is fine. It is proof that the state’s construction industry, broadly defined, had not followed Florida down by mid-2026.

Florida and California construction output slipped. Texas did not.

Explore chartDownload exact data

View exact chart values

state	q2_2025_mil_ch2017	q2_2026_mil_ch2017	pct_change
Florida	63735.8	62469.2	-2
New York	44127.1	43363.2	-1.7
California	99611.5	98592.1	-1
Georgia	30701.7	30381.9	-1
Arizona	24443.5	24295.9	-0.6
United States	867000	870037	0.4
Texas	92120.5	94120.4	2.2
Nevada	13900.3	14723.1	5.9

Jobs follow the same split

Payrolls, which run more current than quarterly GDP, rhyme with that map. From January 2025 through August 2026, seasonally adjusted construction employment in California fell 16,100, or 1.8 percent, to 886,200. Texas added 39,800 jobs, or 4.5 percent, to 924,900. Florida was essentially unchanged (up 200). Arizona was up 900, or 0.4 percent. The BLS state employment series stop in August 2026 in this snapshot, so they do not include whatever October’s 7.40 percent mortgage did to crews after Labor Day.

Builders told NAHB that more than a third of them saw immigration enforcement in their markets last month, delaying jobs and keeping legal workers home. The payroll books cannot test that claim. They can only show that California has already been shedding construction jobs while Texas has not, and that Florida’s construction workforce stopped growing. A labor-shortage story and a demand-slump story can both be true in different states. The national survey is averaging them.

The bill that actually dominates the household ledger

If the midterm argument is about what households pay, construction GDP is the wrong scoreboard. Personal consumption by state is closer. In 2025, Americans spent $11,159 per person on housing and utilities — tenant rent, an imputed rent for owner-occupied homes, and household power and water. That was 18.3 percent of all personal consumption. Groceries bought for home, including alcohol, were $4,459, or 7.3 percent. Gasoline and other energy goods — motor fuel and heating oils — were $1,225, or 2.0 percent. Health care was $10,442. Housing-and-utilities spending was 9.1 times the gasoline line.

Those are current dollars per person, not household averages, and they are 2025 annual figures, not October pump prices. They still rearrange the campaign poster. Housing and utilities per person were $6,409 in 2012 and $7,758 in 2019. By 2025 they had risen 43.8 percent from 2019. Gasoline per person was actually lower in 2025 ($1,225) than in 2012 ($1,342). The Iran war can still make a week at the pump miserable. On the official consumption books, it is not the line that ate the decade.

Among the large housing states in this extract, Florida’s per-person housing-and-utilities bill hit $14,278 in 2025, up 59.7 percent from 2019. California was $13,333, up 43.3 percent. Arizona $12,041, up 59.3 percent. Texas $10,353, up 46.6 percent. Kansas, where the president is barnstorming this week, was $8,758 — well below the national $11,159, and up 35 percent from 2019. A cheaper housing state is not a cheap one if the rest of the budget is diesel and groceries. It is still a reminder that the 7.40 percent mortgage is landing on bills that were already fatter in Florida than in Kansas.

The housing bill grew. The gasoline line did not catch it.

Explore chartDownload exact data

View exact chart values

year	housing_utilities_pc	food_offpremises_pc	gasoline_energy_goods_pc	health_care_pc	total_pce_pc	housing_share_pct	food_share_pct	gas_share_pct	health_share_pct	housing_to_gas_ratio
2012	6409	2692	1342	5794	35145	18.2	7.7	3.8	16.5	4.8
2013	6585	2749	1331	5884	35956	18.3	7.6	3.7	16.4	4.9
2014	6722	2852	1287	6094	37194	18.1	7.7	3.5	16.4	5.2
2015	6824	2927	991	6400	38213	17.9	7.7	2.6	16.7	6.9
2016	6996	2989	885	6674	39238	17.8	7.6	2.3	17	7.9
2017	7196	3094	992	6875	40693	17.7	7.6	2.4	16.9	7.3
2018	7486	3179	1116	7137	42415	17.6	7.5	2.6	16.8	6.7
2019	7758	3280	1067	7487	43720	17.7	7.5	2.4	17.1	7.3
2020	8106	3624	779	7100	42920	18.9	8.4	1.8	16.5	10.4
2021	8560	3889	1162	7956	48524	17.6	8	2.4	16.4	7.4
2022	9325	4174	1541	8438	52907	17.6	7.9	2.9	15.9	6.1
2023	10017	4271	1381	9098	55746	18	7.7	2.5	16.3	7.3
2024	10585	4349	1299	9786	58301	18.2	7.5	2.2	16.8	8.1
2025	11159	4459	1225	10442	61062	18.3	7.3	2	17.1	9.1

What that means for a midterm about $4 gasoline

Three things can be true at once. Builder confidence can look like 2012. Construction output can still be a third larger than 2012 in real terms, with a few big states already slipping. And the consumption line that grew without pause is housing and utilities, not gasoline. None of that says the 7.40 percent mortgage will not cut housing starts from here — the GDP books end in 2026 Q2, the jobs books in August, and NAHB is a survey of people taking orders now. It does say that treating this as a rerun of the last crash, or as a story that begins and ends at the pump, is a misread of the official record.

The political risk is the same mismatch. A sanctions waiver that knocks a few dollars off Brent can dominate a news cycle. A $11,159-per-person housing-and-utilities bill does not fit on a rally sign. Voters who tell pollsters the economy is the issue are not required to mean the thing the White House is trying to fix this week.