The Texas Housing Machine

Texas issued building permits for 108,433 single-family homes in the first nine months of 2026, 3.1% fewer than a year earlier. That puts the state on pace for roughly 136,000 for the year, about a quarter below the 2021 peak of 179,620 (Census Building Permits Survey, state year-to-date file, September 2026¹).

Building the house still earns little. PulteGroup reports its Texas business as a separate segment. In the first nine months of 2026, that segment earned pre-tax income equal to an annualised 5.6% of its segment assets (the land, lots, houses and other assets the company assigns to that segment), down from 8.7% in 2025. Century Communities' Texas segment earned 2.9% (PulteGroup 10-Q, Q3 2026²; Century Communities 10-Q, Q3 2026³). What has changed is where Texas ranks. At most builders, Texas no longer comes last among their regions, because their other regions fell faster. PulteGroup's West and Century's West now earn less than their Texas segments. Lennar's South Central region, which is mostly Texas, has the second-best margin of Lennar's four regions, and D.R. Horton's equivalent region ranks first of its six.

Mortgage rates rose, which is the wrong direction for a recovery. Freddie Mac's average rate on a 30-year fixed-rate mortgage rose from 6.58% on 23 July to a peak of 7.40% in October and stood at 7.24% on 29 October, after the Federal Reserve raised its policy rate (the short-term interest rate it sets) on 16 September (Freddie Mac PMMS weekly history⁴; FOMC statement, 16 September 2026⁵). Meanwhile the serviced lot in a master-planned community kept its price. The developer Howard Hughes sold residential land in Bridgeland, one of its master-planned communities, at $708,000 an acre in the first half of 2026. That was 12.9% more than a year earlier (Howard Hughes 10-Q, Q2 2026, MPC land sales table⁶). Texas builders are being bought for their size, and the buyers are not paying up for their margins. Berkshire Hathaway bought Taylor Morrison, Dream Finders bought Beazer, and Century bought Dallas's HistoryMaker Homes.

The original July study began with a physical sequence: permission, pipes, pad and product. A field must first become a place where homes can legally be built. Roads, drainage, water and sewer then make it usable. The developer prepares finished lots, and the homebuilder constructs and sells the houses. Each stage ties up a different amount of capital and carries a different risk. The business with the largest gross margin need not earn the highest return on the capital it holds.

Permission. Entitlement means the public approvals that allow development. Texas permits substantial new supply, and its recent housing laws make some local restrictions harder to impose. Those laws have specific population, parcel and infrastructure conditions; they do not make every parcel immediately buildable. The economic implication is competition: when developers can respond to demand with more lots and homes, volume can rise without giving every existing owner lasting pricing power. The current statutory detail appears in Section 17.

Pipes. Developers can advance money for eligible water, sewer and drainage works and later seek reimbursement through a municipal utility district, or MUD. A district is a public body that can issue bonds and levy taxes. Reimbursement converts some eligible development spending into a district receivable; it does not eliminate the cost of the infrastructure. Residents ultimately support the district through their bills and property taxes. Eligibility, district finances and bond approval determine what gets reimbursed and when.

The July study's essential qualification remains: the infrastructure advantage is facility-specific and conditional. The ordinary developer-contribution rules contain exemptions, including qualifying large-area facilities. The interest-reimbursement rules also distinguish large service areas. This can favour a large master-planned development over a smaller subdivision, but it is not a blanket subsidy on every dollar of construction. Section 12 updates the rules, debt figures and tax-rate evidence.

Pad. A finished lot includes access and services needed to build a house. Commodity lot production can be expanded by competitors. A serviced lot inside an established master-planned community carries more than those physical inputs: a coordinated development, amenities, schools, buyer recognition and a controlled release of land. These features can support a price premium, but they require time, capital and successful execution. The comparison is between specific positions, rather than a claim that every land developer has an advantage.

Product. The homebuilder takes the lot through construction and closing. It coordinates subcontractors, carries inventory, pays financing costs and bears selling and warranty risk. If mortgage rates rise before the buyer closes, the builder may cut the price, pay to reduce the buyer's mortgage rate, or change the mix of homes it sells. Those incentives lower the builder's realised economics. They may reach the land seller later, when builders renew contracts or reduce purchases. Section 9 walks through the updated lot economics and labels its assumed per-lot calculations.

A master-planned community owner can receive an upfront land price and, under some contracts, builder price participation: a share of a home's selling price above an agreed threshold. Those two receipts must be considered together. Less participation revenue can accompany a higher upfront price; it does not by itself prove the land owner lost bargaining power. The October evidence tests that distinction in Section 10.

The July study also supplied two limits that matter throughout this edition. First, a land bank is finite: selling appreciated land releases value accumulated over years, and buying replacement land at today's price may not reproduce that return. Second, an attractive asset is not the same as an attractive listed company. Corporate overhead, debt, acquisitions and other divisions affect what shareholders receive. Land-sale gross margin, segment earnings and consolidated return on assets answer different questions.

The practical test is whether a well-funded competitor can recreate the position. Houses, ordinary lots and distribution capacity can often be added. A permitted quarry near demand, a secured water connection or a mature community in a particular location can be harder to replicate. Even those assets face cyclical demand and cannot charge any price they choose. The updated sections retain that test while revising the July study's figures, company rankings and open questions.

The three months since the July study brought four kinds of news: borrowing costs rose, the builders reported weaker quarters almost everywhere, two large companies left the stock market, and the owner of the master-planned communities kept raising its prices. The table lists the dated events in order and shows whether each one supports or weakens the July thesis. "Supports" means the event is what the July study would have expected. "Weakens" means it cuts against a specific July claim, and the table names that claim.

First, a correction. Berkshire Hathaway completed its purchase of Taylor Morrison on 24 July 2026, five days before the July study was issued, at $72.50 a share, about $6.8 billion of equity value and $8.5 billion of enterprise value (the equity plus debt, less cash) (Berkshire Hathaway and Taylor Morrison closing release, 24 July 2026¹⁵). The July study graded Taylor Morrison as a listed builder and did not mention the deal. Taylor Morrison is now a wholly owned Berkshire subsidiary and will be combined with Clayton Properties Group, Berkshire's 15 regional builders, into what the companies call the fourth-largest US homebuilding operation, with nearly 23,000 site-built closingsA closing is the completed sale of a home: the buyer pays and takes ownership, and the builder books the revenue. Builders also report orders, which are signed sales contracts that have not yet closed. in 2025 (Taylor Morrison 8-K, 24 July 2026²²). It no longer publishes its own results, so it drops out of the builder comparison.

The forecasters the July study cited were wrong again, in the same direction as before: they expected rates to be lower than they turned out. In January, the Texas Real Estate Research Center's forecast assumed a 30-year mortgage rate of 5.0–5.6% by December 2026 (TRERC, 2026 Texas Real Estate Forecast, 15 January 2026³⁸). Its September forecast moved that range to 6–7% for the year to summer 2027, and moved its federal funds target to 3.4–3.8% (TRERC, Texas Real Estate Forecast Through Summer 2027, 10 September 2026²⁸). Six weeks later the actual mortgage rate was above the top of the new range, and the actual policy rate was above the top of the new target.

The builders' own releases say the same thing. D.R. Horton wrote that conditions "became more challenging during the fourth quarter as rising mortgage rates, affordability constraints and cautious consumer sentiment continued to pressure new home demand" (D.R. Horton FY2026 release, 29 October 2026⁷). PulteGroup's chief executive said "the rapid increase in interest rates has stretched affordability even further, made selling an existing home even harder, and has pushed even more consumers to the sidelines" (PulteGroup Q3 2026 release, 22 October 2026³⁹). Lennar put the 30-year rate at "approximately 6.8% at quarter end and even higher since", and linked it to inflation "driven by geopolitical tension and higher oil prices" (Lennar Q3 2026 release, 16 September 2026⁴⁰).

The July study's worked example was a home priced at $306,600, D.R. Horton's FY2025 South Central average, bought with a 10% down payment. At a 6.58% mortgage rate, the principal and interest on that loan is $1,759 a month. At 7.24% it is $1,881, which is $122 more. To keep the payment at $1,759 with the rate at 7.24%, the price would have to fall to about $286,700, a cut of 6.5%. A builder's incentive budget is now trying to close that gap, on top of the gap it was already closing in July. D.R. Horton's own South Central average fell from $308,500 in fiscal 2025 to $297,200 in fiscal 2026, computed from its table of closings by region. That suggests part of the cut has already been made, through lower prices and a shift in the mix of homes sold toward cheaper ones.

Texas Realtors publishes a quarterly report based on multiple-listing service data (the shared databases that agents use to list homes for sale). It shows the same pattern in months of inventory, which is how many months the current listings would last at the current pace of sales. Statewide it was 5.5 months in the third quarter, compared with 5.4 a year earlier. By metro, Dallas–Fort Worth had 4.5 months, Austin 5.4, Houston 5.6 and San Antonio 6.1. Median prices were down about 1% in each of the four: $330,990 in Houston, $385,535 in Dallas–Fort Worth, $420,000 in Austin and $305,000 in San Antonio. Closed sales fell 5.6% in Houston and 4.5% in Dallas–Fort Worth, and rose 0.9% in Austin and 2.9% in San Antonio (Texas Realtors, 2026 Q3 Texas Quarterly Housing Report, 20 October 2026³²).

In Texas, prices measured on repeat sales (comparing what the same house sells for each time it changes hands) have stopped falling. The Federal Housing Finance Agency's purchase-only indexThe version of the agency's house price index built only from sale prices, leaving out the appraisals made when owners refinance. for Texas rose 0.45% in the year to the second quarter of 2026, which ranked Texas 45th of the states. In the first quarter Texas had been 50th, at −1.63%, as the July study reported. Houston was up 0.20%, Fort Worth up 0.33%, Dallas down 0.06%, Austin down 0.27% and San Antonio down 3.04% (FHFA House Price Index report, 2026Q2²⁶). Texas prices started falling earlier than in most states, and Texas is now one of the first states where they have flattened. So far the index shows only that the fall has stopped, not that prices are recovering.

Nationally, the supply of new homes for sale is a little lower than in July. The Census Bureau estimated 476,000 new homes for sale at the end of September, which is 8.4 months of supply at the current sales pace. It put the median new-home price at $400,400 (Census, New Residential Sales, September 2026, released 27 October³⁵). The July study cited 9.3 months of supply for June, and the same release now revises June to 8.5 months. In the South region, which includes Texas, new-home sales ran at an annual rate of 413,000 in September, 0.7% below a year earlier.

That reading has a consequence for the July thesis, and it cuts both ways. The July study argued that Texas earns thin returns on the house because its supply is elastic: permissive entitlementThe set of government approvals, such as zoning, a recorded plat and building permits, that gives a landowner the legal right to build a given number of homes. Permissive entitlement means those approvals are quick and cheap to get., district-financed pipes and many competing builders keep the house replicable. Supply is elastic when builders can add homes quickly whenever prices rise, so extra profit draws in more building until the profit is gone. On that argument the thinness is structural, a lasting feature of how Texas builds rather than a phase of the housing cycle. If the West can fall to 1–4% on the same builders' books when rates rise, then some of what looked like Texas's structural thinness in 2025 was Texas being earlier in the cycle. The evidence does not settle how much. What it does settle is the narrower claim: on the latest numbers, the house earns thin returns in Texas, and the gap has narrowed because the rest of the country moved closer to Texas. The test that would separate structure from timing is the recovery. If Texas segment returns lag the other regions on the way back up, as they did not lag them on the way down, the structural reading is right.

KB Home is the exception. Its Central segment, Colorado and Texas together, earned $21.5 million of pre-tax income in the nine months to 31 August, an annualised 3.1% on segment assets and the lowest of its four segments; its housing gross margin of 14.7% was also the lowest, and Central deliveries fell 25% (KB Home 10-Q, fiscal Q3 2026¹⁶). Toll Brothers' South segment, which mixes Texas with Florida, Arkansas and South Carolina, earned an annualised 11.4%, fourth of five (Toll Brothers 10-Q, fiscal Q3 2026⁵⁹). Meritage reports a Central region of Tennessee and Texas but publishes regional profit only in its 10-Q, which had not been filed by 29 October; its release shows Central closings of 3,586 in nine months at an average of about $337,000 (Meritage Q3 2026 release¹²).

The figure shows why that fall is weaker evidence of a Texas recovery than it looks. Lennar's incentives fell almost everywhere: in the East from 17.6% to 11.5% of revenue, and in the Central region from 11.7% to 10.0%. Company-wide they fell from 14.3% to 12.0%. Only the West's rose. In the West, gross margin fell from 16.3% to 12.1% over the same year. That put the Texas region's 16.9% above both the West and the Central region's 15.3%, and second only to the East's 20.9%. It is the same convergence from above that Section 6 found at PulteGroup and Century.

Lennar is also cutting back its homesites in the Texas region. Homesites are the lots it owns or controls for future homes. South Central homesites fell to 146,523 at 31 August from 166,768 a year earlier. That 12% fall was larger than in any other region; company-wide, homesites fell 6.7%, to 487,961. Lennar owned only 1,478 of the South Central homesites itself and controlled the rest through options, contracts that give it the right to buy the land later. In the quarter, South Central took 6,100 new orders, down 13.5%, from 479 communities, up from 411. Lennar was selling in more communities and taking fewer orders in each. That is the pattern of a builder holding volume in a soft market by spreading it across more communities, and the fall in incentives is consistent with it.

The quarter cannot show the effect of the rise in mortgage rates. It ended on 31 August, when the 30-year mortgage rate was 6.66%; by mid-October the rate was 7.40%. When Lennar cut its delivery target, it cited "continued pressure on interest rates and the deterioration in market conditions through the quarter". It expects a fourth-quarter gross margin of 15.5–16.0%, about where the third quarter came in, at 15.8%. D.R. Horton, whose fiscal year ended 30 September, said it expects "incentives to remain elevated in fiscal 2027" (D.R. Horton FY2026 release⁷), and PulteGroup said the same (PulteGroup 10-Q, Q3 2026²). On the evidence, Lennar's Texas-region incentives peaked a year ago at 20.5%, and every builder that spoke in October expects incentives to stay high.

D.R. Horton's fiscal 2026 results were weakest in the fourth quarter. Net orders fell 7% from a year earlier to 18,619 homes and the cancellation rate rose to 23% from 20%; in the South Central region (Arkansas, Oklahoma and Texas), fourth-quarter orders fell 15%, to 4,548 (D.R. Horton FY2026 release, 29 October 2026⁷). The homebuilding gross margin on home sales, computed from the segment tables, was 19.5% in the quarter, against 21.5% for fiscal 2025. The company forecast 79,000–84,000 closings for fiscal 2027, against 84,202 in fiscal 2026. It also took a $91.4 million impairment, writing down the value of nine multifamily rental projects. Consolidated pre-tax income for the year was $3.9 billion, or 11.1% of average total assets, against 13.2% in fiscal 2025.

Even so, South Central is the company's best region. In the nine months to 30 June 2026, the latest period with regional detail, it earned $609.3 million pre-tax on $4,868.7 million of revenue. That is a 12.5% margin, against 12.3% in the North and 10.5–10.9% in the other four regions (D.R. Horton 10-Q, Q3 FY2026, homebuilding results by reporting segment⁸). D.R. Horton counts only inventory as segment assets, and South Central's annualised return on those assets was 20.3%, also first of six. In fiscal 2025 the region was second of six on margin, at 14.0% on $6,899.3 million of revenue, behind the Northwest's 14.7% (D.R. Horton 10-K FY2025, homebuilding results by region⁶³). The Northwest now ranks last. The full-year regional table will come with the 10-K in November.

A second explanation also fits: D.R. Horton sells the cheapest homes in the market. Its South Central average closing price was $297,200 in fiscal 2026, against $366,000 for PulteGroup's Texas homes and about $337,000 for Meritage's homes in Tennessee and Texas. A buyer whose limit is the monthly payment looks for the cheapest new house. But Lennar's South Central average was lower still, $230,000, and its margin was lower too, so price alone does not explain the ranking.

The waterfall chart shows how little room D.R. Horton has, even in its best region. Of about $296,900 of revenue per home, $232,800 went to the lot, the construction and capitalised interestInterest on money borrowed to buy land and build. The builder adds it to the cost of each house while the house is being built and expenses it when the house is sold, so it shows up in the cost of sales., $26,000 to selling and overhead and $1,400 to land and option charges, leaving about $37,200 before tax. On the July study's example house, a 6.5% price cut would offset the rise in the mortgage rate from July to October. On this house, that cut is about $19,300, half the profit. Builders do not cut list prices that far. They spread the cost through rate buydowns, where the builder pays to lower the buyer's mortgage rate, and through closing-cost incentives, and those show up in the same cost line. The South Central margin fell from 14.5% in the first nine months of fiscal 2025 to 12.5% in the same months of 2026, before most of the rate rise arrived.

D.R. Horton ended the year with 570,600 owned and controlled lots. It owned 130,500 and controlled the other 440,100, or 77%, through purchase contracts. South Central held 146,300 of those lots, of which 115,400, or 78.9%, were controlled. At 30 June, D.R. Horton had $2.5 billion of deposits on its purchase contracts, against $27.5 billion of purchase price still to pay, and $210.1 million of the deposits were on contracts with Forestar (D.R. Horton 10-Q, Q3 FY2026, land purchase contracts note⁸). It wrote off $58.5 million of earnest money and pre-acquisition costsMoney spent on land before buying it, such as engineering, surveys and permit work. A builder that walks away from a land contract writes these off along with its deposit, the earnest money. in the nine months, against $76.3 million a year earlier. It returned $4.1 billion to shareholders in fiscal 2026 and plans $3.0 billion of buybacks in fiscal 2027. In a year when its closings fell 1%, D.R. Horton, the largest builder in Texas, cut its share count by 8% and its owned and controlled lots by 3.6%, from 591,900 to 570,600.

Start in the field. The Texas Real Estate Research Center put the statewide price of rural land at $5,218 an acre in the second quarter of 2026, 3.27% higher than a year earlier, while the acreage that changed hands fell 12.57% (TRERC, Texas Rural Land Markets, Q2 2026, 20 August 2026²⁵). In the Gulf Coast–Brazos Bottom region around Houston the price was $11,369, up 3.05%; around Austin, Waco and the Hill Country it was a record $8,040; in Northeast Texas, which includes the land east of Dallas, it fell 5.96% to $8,604. The center describes "price stickiness and a persistent psychological standoff": sellers holding out for peak prices, buyers held back by rates. A landowner in that position owns an option on growth and earns nothing until somebody exercises it.

The gap between the field and the finished lot is where the money is, and it is mostly not land. On illustrative assumptions, a 0.15-acre lot, the size under which half of new spec homesA spec home is a house a builder starts before it has a buyer. in the West South CentralA Census Bureau division of four states: Arkansas, Louisiana, Oklahoma and Texas. states sat in 2025 (NAHB analysis of the Census Survey of Construction, 21 July 2026⁶⁴), plus streets, detention ponds and open space at a gross-to-net ratioTotal acres in a development divided by the acres that end up inside sellable lots. At 1.6, each acre of lots needs 1.6 acres of land once streets, ponds and open space are added. of 1.6, uses about 0.24 acres of farmland. At the Gulf Coast price that is roughly $2,700 of raw land. Forestar sold its average finished lot for $111,200 in fiscal 2026 (Forestar FY2026 release⁹). Forestar's lots are national, not Texas only, and the gross-to-net ratio is an assumption, so the comparison is an order of magnitude. It says the raw acre is a few per cent of what a builder pays for the lot. The rest is engineering, pipes, roads, permits, carrying costThe cost of holding land or unfinished work until it is sold: interest on the money tied up, property taxes and upkeep., the developer's margin and, inside a trusted community, the premium for the place.

The pipes are financed by the public. At Bridgeland, Howard Hughes spent $138.8 million on development in the first half of 2026, including capitalised interest and property taxes, and transferred $116.6 million of it, 84%, to MUD receivables, the claims it holds on future district bond proceeds (Howard Hughes 10-Q, Q2 2026, MPC asset roll-forward⁶). Section 12 follows that money to the homeowner's tax bill.

The builder is the next hand, and its margin is the thinnest in proportion to the capital and risk. D.R. Horton's South Central homes produced about $37,200 of pre-tax profit on $296,900 of revenue each (Section 8). The last hand is the lender. D.R. Horton's financial services segment earned $258.1 million pre-tax on $817.3 million of revenue in fiscal 2026, a 31.6% margin, down from $278.7 million and 33.1% (D.R. Horton FY2026 release⁷). The July study could not measure that segment's capital. D.R. Horton's fiscal 2026 release now shows its balance sheet: $3.25 billion of assets at 30 September, mostly $2.57 billion of mortgage loans held for sale, funded partly by $1.37 billion of notes payable. On average segment assets the pre-tax return was about 8.2%; on the segment's net assets, assets less liabilities, about 19%. A high margin on a warehouse-fundedFunded with a warehouse line, a short-term bank credit line that a mortgage lender draws on to make loans and repays when it sells those loans to investors. book is a moderate return on its assets.

Timing matters for the film. A Texas house takes about four months to build; Lennar reported a record cycle time of 116 days in the quarter to August. The land before it takes far longer. Bridgeland, with 1,142 residential acres left at 30 June, is not expected to sell out until 2032 (Howard Hughes Q2 2026 supplemental, land bank overview⁶⁵). The time from raw acre to the first closing on a new Texas subdivision was not found in any primary source opened for this update, and it is listed as not verified.

Howard Hughes sold 92.1 acres of Bridgeland residential land to builders in the first half of 2026 at an average of $708,000 an acre, against 77.3 acres at $627,000 in the first half of 2025. In the second quarter alone it sold 30.2 acres at $751,000 an acre, against 40.3 acres at $648,000 (Howard Hughes 10-Q, Q2 2026, MPC land sales closed⁶). Bridgeland's earnings before tax were $58.0 million for the half, up 52%, and $23.2 million for the quarter, up 9%. The July study's figure for 2025 as a whole was $669,000 an acre. The first-half 2026 average was 5.8% above that and 12.9% above the same half a year earlier. Over the same months, PulteGroup's Texas pre-tax income fell 46% and Century's fell 14%.

Howard Hughes's other Texas community moved the same way. The Woodlands Hills sold 26.5 acres at $486,000 an acre in the half, against 7.5 acres at $491,000 a year earlier, and its earnings before tax nearly doubled to $8.0 million. Across all of Howard Hughes's master-planned communities, segment earnings before tax were $134.7 million in the second quarter, up 32%, and $529.5 million over the trailing twelve months, up 44% (Howard Hughes Q2 2026 supplemental⁶⁵). Builders' net new home sales in the quarter rose 17% in Bridgeland and 34% in The Woodlands Hills.

One figure went the other way. Builder price participation fell to $6.9 million in the quarter from $14.1 million. Participation is a contractual share of whatever a builder's home sells for above an agreed price, called the breakpoint, and the builder pays it to Howard Hughes. The July study called it the developer's call optionA call option gives its holder the gains above a set price without the losses below it. Participation works the same way: Howard Hughes collects when a home sells above the breakpoint and pays nothing back when it sells below. on the builder. The company gave two reasons for the fall: the mix of homes closed, and fewer homes clearing the breakpoint.

When house prices are flat and builder margins are falling, participation above a fixed breakpoint pays less. So Howard Hughes has moved more of its share into the upfront land price and taken less through participation at the back end, and builders paid the higher land price. That is the strongest single piece of evidence in this update that the owner of the master-planned communities kept its pricing power while the builders lost theirs. It is also first-half evidence, from a period when mortgage rates were around 6.1–6.4%. Howard Hughes had not reported its third quarter by 29 October, so the first sales made with mortgage rates at 7% are not yet visible.

The July study's point that the land will run out still holds, and the figures are now more precise. At 30 June Bridgeland had 1,142 residential acres left to sell, down from 1,234 at the end of 2025, and was expected to sell out in 2032. The Woodlands Hills had 597 acres left and was expected to sell out in 2035. The Woodlands itself had 63, expected to sell out in 2031. Howard Hughes's own five-year model prices Bridgeland's remaining residential land at an average of $661,000 an acre, with no inflation assumed. That is below the $708,000 it just achieved, and the model projects an 88.7% cash gross margin on the land (Howard Hughes Q2 2026 supplemental, land bank overview⁶⁵). The company itself is not assuming that land prices keep rising. Its estimate of the residual valueThe company's estimate of what its unsold land will still bring in, after the cost of developing it. of all its wholly owned communities was $5.73 billion at 30 June, against $6.11 billion a year earlier. Every sale of land reduces that figure automatically, so the fall does not by itself mean the company has marked its land down.

RCLCO, a real-estate consultancy, found in its mid-year 2026 ranking that sales at the 50 best-selling master-planned communities were up about 2.9% on a year earlier. Texas had 20 of the 50 and 31% of their sales. Houston had nine of the 50, with 3,049 sales and about 16% of the total (RCLCO, Top-Selling Master-Planned Communities of Mid-Year 2026, 30 July 2026²⁴). A year earlier Houston had eleven, with 3,666 sales and about 22%. Houston's best communities are selling, but fewer of them rank among the national leaders. That weakens the July prediction that top Houston communities would sell homes faster than the national top-50 average.

Two things in the quarter bear on the second half of the July prediction, that at least one more large builder would hand a material land portfolio to outside capital. Dream Finders completed its $2.2 billion purchase of Beazer Homes on 20 October (Dream Finders closing release³¹). Millrose supplied about $1.1 billion of it: roughly $580 million of land-banking capital across about 7,000 homesites, more than half finished, with a weighted average lifeThe average time until the money is repaid as builders buy the lots, with each lot counted in proportion to its dollar amount. of about 2.5 years, and about $488 million of preferred equity in Dream Finders carrying a 12% dividend (Millrose Q3 2026 release, 27 October 2026¹¹). That is a builder financing an acquisition partly by selling the target's land to a land bank on the day it buys it. It counts toward the prediction, though it is an acquisition structure rather than an existing builder moving its own book. Lennar also sold Millrose 1,233 completed homes for a single-family rental pilot at a cost basis of about $333 million.

Millrose itself is growing. Its total assets were $10.6 billion at 30 September, its homesites 156,729 across 834 communities, and 22 builder and developer counterparties, 21 of them other than Lennar. Its Lennar capital was about $5.9 billion at a weighted average yield of 8.5%; it funded $714 million of new non-Lennar business in the quarter at 10.8%, bringing that book to $3.4 billion, and its whole portfolio yielded 9.3%. Pro forma for Dream Finders, about 42% of its assets sit outside the Lennar agreement. On the funding side, it priced $500 million of 6.500% notes due 2029 and $500 million of 6.750% notes due 2031 in September, against 6.250% and 6.375% on its earlier issues, and ended with $3.1 billion of corporate debt. The July study predicted that Millrose's non-Lennar yields would fall below 10.7% if credit spreadsA credit spread is the extra interest a lender charges over a safe benchmark rate, such as a government bond yield, as pay for the risk of not being repaid. When spreads narrow, lenders accept less extra interest. narrowed. Spreads did not narrow and the yield on new business did not fall. The gap between the 9.3% portfolio yield and the 6.5–6.75% cost of the new notes is about 2.6–2.8 points before the external managerAn outside firm that runs the company under contract for a fee, here a percentage of assets, in place of management the company employs itself.'s fee, which the July study put at 1.25% of tangible assets a year. Lennar borrows for less than it pays Millrose. On 15 October it sold $300 million of 5.700% notes due 2029 and $300 million of 5.950% notes due 2032 (Lennar 8-K, 15 October 2026³⁰). The 8.5% yield on Millrose's Lennar capital is about 2.6–2.8 points above that. Lennar pays the difference to keep the land off its balance sheet.

Forestar, the lot developer D.R. Horton owns 62% of, is the cleanest test of whether producing finished lots is a scarce position. It earned $192.1 million pre-tax in fiscal 2026 on average total assets of about $3.22 billion, 6.0%, against D.R. Horton's 11.1% on the same basis (Forestar FY2026 release⁹). A year earlier the figures were 7.3% and 13.2%. The July study's cleanest falsifierA result, named in advance, that would show the July thesis wrong if it happened. A falsifier that has not fired is one whose result has not happened., Forestar out-earning its parent for two years, has not fired. Forestar sold 13,545 lots, 85.9% of them to D.R. Horton, and expects 12,000–13,000 in fiscal 2027. The Delaware derivative suitA lawsuit shareholders bring on a company's behalf, usually against its directors or controlling shareholder. This one concerns the prices at which Forestar sells lots to D.R. Horton, its majority owner; a sale to a company's own owner is called a related-party sale. over its related-party lot pricing, Mississippi Public Employees' Retirement System v. D.R. Horton, was still pending at 30 June, with Forestar still unable to estimate a loss (Forestar 10-Q, quarter ended 30 June 2026⁶⁶).

Forestar's lot price is the one figure in this section that runs against the July thesis, and the explanation matters. Its average lot sold for $111,200 in fiscal 2026, 2.6% more than in fiscal 2025, while Lennar cut its construction cost per square foot 6%. The July study set a falsifier on exactly this: Texas finished-lot prices rising faster than builders' cost per square foot through 2029 would mean land rentIn economics, rent is profit above the cost of producing something, earned because the thing is scarce. Land rent is that extra profit earned by holding a scarce location or permission. The July thesis puts it with the trusted master-planned lot, not with the ordinary finished lot. returning to the commodity lot. One year points that way. Yet Forestar's gross margin, computed from its income statement, fell from 22.3% to 20.9%, and its pre-tax margin from 13.2% to 11.7%. Lot prices rose less than the cost of making lots. That pattern fits development cost being passed on to builders, and PulteGroup, Meritage and Green Brick all named rising lot and land-development costs in October. A Texas-weighted measure points the other way. In the West South Central states, which include Texas, the median lot under a spec home started in 2025 was worth $56,000, after a second year of easing (NAHB analysis of the Census Survey of Construction, 28 July 2026⁶⁸). The falsifier is weakened for now. It should be read again with Forestar's fiscal 2027 results, and with its fourth-quarter average of $108,300, 6.4% below a year earlier, in mind.

Water districts borrowed more each year while permits fell. Their new-money borrowing, which pays for new projects and does not refinance old bonds, was $3.60 billion in fiscal 2021, $5.18 billion in 2022, $5.80 billion in 2023, $6.28 billion in 2024 and $6.37 billion in 2025. Over the same period, Texas single-family permits fell from 179,620 to 140,579. Dividing fiscal 2025 new-money borrowing by calendar 2025 single-family permits gives about $45,300 a home. As the July study said, that is an order of magnitude and cannot be read as each home's share: district bonds also fund commercial and multifamily works, and the years do not line up. A district sells bonds to repay the developer for pipes laid earlier, against homes that have since been built and sold, so the public debt keeps growing after the building slows.

In the November 2024 bond elections, voters approved $23.81 billion of water-district bond propositions; in May 2025 they approved another $20.15 billion (Bond Review Board annual report, Appendix A, Tables A.3 and A.5⁶⁹). Together that is about $44 billion of borrowing authorised by voters, against $30.37 billion of tax-supported debt outstanding. Not all of it is new borrowing. Many district ballots carry separate refunding propositions, which let the district refinance existing bonds. Travis County's Triada MUD, for example, had $210 million approved for water, sewer and drainage, $184 million for roads and $315 million for refunding. Williamson County MUD 56 had $265 million, $150 million and $397.5 million approved in the same three categories. Even leaving out the refunding authority, the borrowing that voters have approved but districts have not yet issued is a multiple of what they issue in a single year.

One developer's accounts show how this works. At Bridgeland, Howard Hughes spent $138.8 million on development in the first half of 2026 and booked $116.6 million of it, 84%, as MUD receivables, the amounts it expects the districts to repay once they sell bonds (Howard Hughes 10-Q, Q2 2026⁶). It borrows against those receivables. Its Bridgeland notes, with $600 million of capacity at 5.93%, are secured by MUD receivables and Bridgeland land, and $85 million was drawn at 30 June. Waiting for the districts to pay can be expensive. In the first half of 2025 Howard Hughes sold MUD receivables for $180.0 million and recorded a $48.2 million loss on the sale; in the first half of 2026 it sold $17.5 million at a $0.6 million loss. Those are one developer's choices, so they do not give a statewide discount rate. They do show that the developer bears a real cost for the time between laying the pipe and being repaid.

The rules did not change. The Texas Commission on Environmental Quality's rule still requires a developer to contribute "not less than 30%" of district water, wastewater, drainage and recreational construction costs, and still exempts lines serving 1,000 acres or more, drainage serving 2,000 acres or more, treatment plants and districts whose debt is 10% or less of assessed value (30 TAC §293.47⁷¹). The Legislature did not meet in regular session in 2026, and this update found no 2026 statute or rule that changes district reimbursement. Both chambers did, however, put districts on the agenda for the 2027 session. The House Land and Resource Management committee met on 21 July 2026 to "consider the proliferation of municipal utility districts (MUDs) and evaluate their impact on housing" and to "identify best practices to ensure sufficient oversight exists" (House committee hearing notice, 21 July 2026⁷²). The Senate Local Government committee met on 2 September to take up the implementation of SB 15, SB 840, SB 2477 and SB 2038, and "utility connection charges, permit and inspection fees, and drainage or stormwater improvement fees" (Senate committee hearing notice, 2 September 2026²⁷).

The chance that these rules change has risen. The July study argued that the district statute is a scale subsidy to the large master-planned developer, meaning a subsidy that is worth more the bigger the project. Its reason was that the statute's acreage thresholds let the largest projects recover more of their infrastructure cost and carry (the cost of financing the work until the district repays it). In 2027 a Legislature looking at "proliferation" and "oversight" could narrow those thresholds, make districts harder to create or cap tax rates. Any of those would move cost off the districts' public books and back to the developer, and the master-planned owner, which benefits most from the thresholds, would bear most of it. No such bill had been filed by 29 October.

The three positions at the top are the same as in July. The price of master-planned lots rose while the builders that buy them lost margin (Section 10). Aggregates producers kept raising prices while volumes fell. Vulcan's freight-adjusted priceThe price per ton with delivery charges taken out, so it shows what the quarry itself earns on the stone. rose 4% to $22.84 a ton in the third quarter while shipments fell 3%, its aggregates cash gross profit per ton rose to $11.95, and it told investors to expect "mid- to high- single digit pricing gains" in 2027 (Vulcan Q3 2026 release, 29 October 2026¹⁸).

Martin Marietta paid about $7 billion in cash and 10,953,543 new shares worth $6.5 billion for Lhoist North America, citing "more than 2 billion tons of high-quality reserves" of limestone (Martin Marietta 8-K²⁰ and release, 24 August 2026⁷⁵). Lhoist makes lime, which goes mainly into steel and environmental markets rather than into houses. So the deal shows what permitted reserves sell for and says nothing about Texas housing demand. Earlier in the year, on 23 February, Martin Marietta transferred its Midlothian cement plant, its Texas ready-mixConcrete mixed at a plant to order and delivered wet by truck. plants and related terminals to QUIKRETE in an asset exchange, which left it with no cement business (Martin Marietta 10-Q, Q2 2026, divestitures note¹⁹). The July study described its 2024 sale of South Texas cement but did not record this second exit.

Cement softened slightly. Eagle Materials' average net cement price fell 2% to $154.09 a ton in the quarter to 30 June, and its cement operating margin per ton fell 16% to $34.31, while volume rose 8% (Eagle Materials 10-Q, quarter ended 30 June 2026²¹). Eagle's pre-tax income for the quarter, $131.7 million, comes to 13.6% of assets when annualised, against 15.3% for its last full fiscal year. Cement stays graded advantaged. Eagle had not reported its September quarter by 29 October, and a second quarter of lower cement prices would lower confidence in the grade.

Two grades hold on weaker evidence than in July. Green Brick is still graded advantaged, but its return fell from 18.3% to an annualised 13.4% and its margin lost 410 basis points in a year. What is left of its advantage is its land basis, the cost at which it holds its land, and that land is being used up. Confidence drops from medium-high to medium. Invitation Homes' Texas homes are still graded exposed, though their figures improved: 6,474 Texas homes at 94.1% average occupancy in the third quarter, with Houston at 94.6% and Dallas at 93.8%, against 91.1% at the end of 2025 (Invitation Homes Q3 2026 supplemental⁷⁶). D.R. Horton's $91.4 million impairment, a write-down of book value, on nine multifamily rental projects is evidence in the other direction for build-to-rent investment.

Federal law changed the rules for the largest rental buyers. The 21st Century ROAD to Housing Act became law without the President's signature on 11 July 2026. Its section 1001 bars any for-profit investor that controls 350 or more single-family homes from buying more for 15 years from 7 January 2027 (Congressional Research Service, 20 October 2026⁷⁷). The law exempts build-to-rent purchases. Treasury's proposed rules of 22 October define those as homes never occupied by an owner and bought within 18 months of their certificate of occupancy (the local sign-off that a new home is fit to live in) (Treasury release³⁴; proposed rule 2026-21675⁷⁸). The rules are still a proposal. If they stand, the largest rental owners can no longer buy in the resale market but can still buy in the new-home market, so the large rental investors that keep buying will buy from builders. That helps builders' bulk sales and the lots those homes are built on. It does nothing for a rental owner's return on the homes it already holds. Neither the build-to-rent grade nor Invitation Homes' grade changes.

Millrose, Forestar and D.R. Horton's mortgage arm earn their returns by supplying capital rather than by holding a permission, and all three did what the July study expected. Millrose earns a portfolio yield of 9.3% and pays 6.5–6.75% on new borrowing. That gap is a credit spread, which any well-funded lender can compete for, and Millrose is competing for it by adding builders. Forestar's 6.0% is the kind of return a manufacturer earns. D.R. Horton's mortgage arm, now that its balance sheet is visible, earns about 8.2% pre-tax on its assets. None of these changed grade, and none moved toward the top.

In 2025 Texas issued building permits for 140,579 single-family homes, about 385 a day (Census, 2025 annual⁴³). Houston alone accounted for 46,219, about 127 a day, and Dallas–Fort Worth for 39,790 (Census CBSA, 2025 annual⁴⁸). D.R. Horton closed sales on 22,045 homes in its South Central region in fiscal 2026, about 60 every day of the year. Forestar delivered about 75,000 finished lots (lots with roads and utilities in, ready for a builder) over five years, and invested more than $7.0 billion in land and development to do it (Forestar FY2026 release⁹).

A master-planned community is a small city built over decades. Bridgeland covers 11,506 gross acres in Cypress, northwest of Houston, and had about 29,000 residents at the end of 2025. The Woodlands covers 28,545 acres and has about 124,800 (Howard Hughes 10-K FY2025, master planned communities table⁷⁹). Both are Howard Hughes communities, and the company's communities together span about 101,000 gross acres in five states. Bridgeland had 1,142 residential acres left at 30 June 2026. It is expected to sell its last residential land in 2032 and its last commercial land in 2046.

A single neighbourhood shows how densely the land is built. Del Webb's first active-adultBuilt and marketed for older buyers, typically with an age requirement for residents. neighbourhood in Bridgeland, announced on 5 February 2026, puts 672 homes on 190 acres, on lots 50 and 65 feet wide (Bridgeland, 5 February 2026⁸⁰). That is about 3.5 homes an acre.

Even when everything goes to plan, it takes about two years to get from pasture to the first lot. The developer Hines bought 3,000 acres west of Houston in January 2025 for a community of more than 7,000 homes. It planned to start construction in early 2025 and deliver its first lots to builders in 2027 (Realty News Report, January 2025⁸¹). That timeline is one developer's published plan, as reported by the trade press.

The houses are mid-sized and sit on small lots. The median new single-family home completed in the South Census region in the second quarter of 2026 had 2,122 square feet of floor area (Census quarterly completions via FRED⁸²). Half of the new spec homes (houses a builder starts before it has a buyer) sold in the West South Central states in 2025 sat on lots smaller than 0.15 acres, about 6,500 square feet (NAHB analysis of Census Survey of Construction data⁶⁴). A house takes about four months to build: Lennar's cycle time, the days from starting a house to finishing it, was 116 days in its third quarter. Each house also uses a great deal of rock. The stone industry's trade association says almost 500 tons of aggregates (crushed stone, sand and gravel) go into a new home "including driveways, sidewalks, basements, cul de sacs and other neighborhood foundations", and about 38,000 tons go into a mile of highway (National Stone, Sand & Gravel Association, February 2024⁸³).

The July study committed to eleven falsifiers, twelve dated predictions and thirty-two graded verdicts, some of them grouped. The table scores each one against what has happened since 29 July. A falsifier is a condition that would show the July thesis wrong if it occurred. For a falsifier, "holds" means it has not fired, so the July thesis survives it. For a prediction or a verdict, "holds" means the evidence since July is consistent with it. "Weakened" means the evidence has moved against the claim without settling it. "Broken" means the claim is wrong on current evidence. "Not yet testable" means either the data that would test the claim have not been published or its horizon has not arrived. It also covers four verdicts on companies this update did not re-check, and the table lists those as such.

Of eleven falsifiers, six hold, one is weakened and four cannot yet be tested; none has fired. Of twelve predictions, four hold, three are weakened and five cannot yet be tested; none is broken. Of the twenty-eight verdict rows, nineteen hold, four are weakened, one is broken and four cannot be tested. The broken verdict is Builders FirstSource's neutral grade, and Section 13 explains the downgrade. The three weakened predictions all have the same cause. The July study expected Texas to stay worst, and since then the rest of the country has fallen toward it.

The recovery case, argued fully. The builders enter 2027 leaner than they entered 2026. Lennar's construction cost per square foot is 14% below its late-2023 baseline, its cycle time (how long it takes to build a home) is a record 116 days, and it has cut completed, unsold homes to 1.8 per community (Lennar Q3 2026 release⁴⁰). D.R. Horton cut its homes in inventory from 29,600 to 28,200 in a year. Texas house prices, measured by repeat sales of the same homes, stopped falling in the second quarter, earlier than in most states. If the 30-year mortgage rate fell from 7.24% to 6.24%, the monthly principal and interest on the July example house would fall from $1,881 to $1,697, and builders could spend less on buydowns (payments that lower a buyer's mortgage rate). Operating leverage (profits rising faster than sales once fixed costs are covered) would be greatest at the low-price, high-volume Texas builders, so Texas segment returns would recover fastest. The July claim that Texas returns are thin for structural reasons would then turn out to have been a reading of one bad cycle.

That case fails today on its first premise, falling rates. The Fed raised rates in September, and two of its members wanted to raise them again in October. The research center now expects mortgage rates of 6–7%, up from the 5.0–5.6% it expected in January. The case might still come true later, and if it does, two of its claims can be tested. Texas builder returns could recover faster than those in other regions; Section 6 named that as the test that would show thin Texas returns to be a matter of timing rather than structure. And the master-planned land owner would gain at least as much as the builders, because a recovery raises the prices builders can pay for lots in the communities buyers want first. A recovery would lift Texas builder margins from a low base. The returns would stay with the owners of those scarce lots, because their price would rise too.

The downturn case, argued fully. The September rate increase landed on a market that was already weakening. D.R. Horton's South Central orders fell 15% in its fourth quarter, and its company cancellation rate rose to 23%. PulteGroup took $15.9 million of Texas land charges in one quarter. Forestar's fourth-quarter lot price was 6.4% below a year earlier, and it recorded $7.6 million of impairments (write-downs in the value of its land). Rural land changed hands 12.6% less often and fell 6% in Northeast Texas. Howard Hughes's builder price participation (its share of builders' home prices above an agreed level) halved in the second quarter. Houston, home of the leading master-planned communities, drew most of its 2025 growth from international migration, and that migration halved. Houston's share of the national top-50 communities fell. The Bridgeland prices in this report were set when mortgage rates were 6.1–6.4%. If builders stop buying at those prices, land priced for the next six years of growth is the asset most exposed to a slowdown.

That case has more evidence behind it than the first, and this update does not dismiss it. Three facts limit it. First, Bridgeland's sales and prices were rising in the half for which data exist, with net new home sales up 17% in the second quarter. Second, Howard Hughes's own model already values the remaining Bridgeland land below the prices it has achieved, at $661,000 an acre with no price inflation assumed. So the owner's valuation does not depend on continued growth. Third, the margin is wide: on a projected cash gross margin of 88.7%, a 20% fall in the price per acre, to about $566,000, would still leave the land highly profitable. A deeper downturn would cut the land owner's revenue, but only a far deeper fall would make the land a bad investment. Howard Hughes's third-quarter and full-year results will give the case its first real test. The July falsifier, the result that would show this report wrong, still stands: price declines of more than 10% for two consecutive years while the metro area grows.

Taken together, the downturn case is the one to watch, and the recovery case is the one the evidence currently rejects. Neither would move the returns from the master-planned land owners to the builders. Each would change how large those returns are.

The July study described an arrangement with three parts that matter for who gets paid: a permissive state that lets anyone build, public money that pays for the pipes up front, and a few places buyers trust enough to pay a premium for. If any one of them changes, the arrangement stops working.

The first part, the permissive state, is staying in place. SB 15 still bars cities of more than 150,000 people, in counties of more than 300,000, from requiring lots larger than 3,000 square feet on unplatted tracts (land not yet divided into recorded lots) of five acres or more (SB 15, enrolled⁸⁴). SB 840 still requires those cities to allow multifamily and mixed-use housing in commercial zones (SB 840, enrolled⁸⁵). HB 24, in force since 1 September 2025, made it harder to block a zoning change that allows more housing (HB 24, enrolled⁸⁶). Neighbours protesting such a change must now own 60% of the adjoining land within 200 feet, up from 20%. If a protest does qualify, a council can pass the change with a majority of all members instead of three-quarters. City-wide rezonings are exempt from the protest process altogether. The Legislature's committees are monitoring these laws and leaving them in place (House committee hearing notice, 20 July 2026⁸⁷).

The second part, public money for the pipes, is under review for the 2027 session. The third, the premium buyers pay in a few trusted places, is the part under the most pressure from 7% interest rates. The observations below are the ones that would change this report's answer.