Thailand Housing Supply and Household Formation
Let’s explore this page together
Published · Updated
Why the household count is not a new-home demand figure
A city can have plenty of people looking for somewhere to live and still have apartments that developers struggle to sell. That is why a census figure cannot be read as a ready-made sales forecast. Thailand’s 2025 Population and Housing Census reports 25,873,003 households nationwide. The Real Estate Information Center (REIC), meanwhile, counted 214,083 homes within its marketed-housing supply for Bangkok and five adjoining provinces in the second quarter of 2026. These are official figures, but one is a resident-household stock for the whole country and the other is a market measure for six provinces at a later date.
Consider what a household represents. Someone living alone counts as a household, just as a larger group sharing a home does. A household might rent, own a house purchased years ago, share accommodation with others or live in a dwelling that never enters a developer’s sales inventory. Population counts people: NSO records 70,275,751 residents nationwide and 8,890,112 in Bangkok alone. Occupied living arrangements are different from housing units, and the entire dwelling stock includes properties that may not currently be offered for sale. A count of developer-market listings or units offered cannot stand in for any of those totals. It also leaves out parts of the existing-home market.
The term household formation adds another distinction. Formation is a process taking place over time: people establish separate households, combine them, migrate or cease to maintain them. A census gives a stock on a particular date. Even the difference between two census stocks is a net outcome, not the gross number of households formed. The housing-market equivalents also need separating: new project launches are not completed homes, new sales are not fresh supply, and the unsold balance at the end of a quarter sits inside the broader offer reported by REIC rather than alongside it as a second pile to add.
There are two NSO versions to watch, too. A January 2026 early-results announcement gives 26.30 million households, while the July 2026 Preliminary Report’s Table 3 records 25,873,003 for the 1 April 2025 census date. The material reviewed does not explain the difference, which cannot be dismissed as a rounding of the same precise figure. NSO also issued a final report in September 2026, giving the national total at the rounded level of 25.9 million. The exact province figures displayed here are explicitly sourced to July’s Table 3 and should not be passed off as a verified extraction from the later edition. None of these releases tells us how many households entered the purchase market in 2026.
Households and housing supply: area, date and measure
NSO figures count households on 1 April 2025 (Preliminary Report, Table 3). REIC figures count homes in its surveyed offer in Q2 2026. The six-province household sum is calculated from the NSO rows. Remaining unsold supply is part of the relevant total supply, not an additional stock; none of these figures is an oversupply ratio.
| Measure | Area · period | Figure |
|---|---|---|
| NSO households | Bangkok · 1 Apr 2025 | 3,364,773 |
| NSO households | Samut Prakan · 1 Apr 2025 | 842,701 |
| NSO households | Nonthaburi · 1 Apr 2025 | 558,936 |
| NSO households | Pathum Thani · 1 Apr 2025 | 602,914 |
| NSO households | Nakhon Pathom · 1 Apr 2025 | 375,924 |
| NSO households | Samut Sakhon · 1 Apr 2025 | 379,390 |
| Sum of 6 NSO rows | Same 6 provinces · 1 Apr 2025 | 6,124,638 |
| REIC total supply | 6 provinces · Q2 2026 | 214,083 |
| REIC unsold supply | 6 provinces · Q2 2026 | 202,414 |
| NSO households | Thailand · 1 Apr 2025 | 25,873,003 |
| REIC total supply | 27 surveyed provinces · Q2 2026 | 359,433 |
Which geographies and periods can be placed side by side
There is a sensible geographic bridge between the two datasets, although it is narrower than “Thailand”. REIC’s Greater Bangkok market comprises Bangkok, Nonthaburi, Pathum Thani, Samut Prakan, Samut Sakhon and Nakhon Pathom. NSO’s census lists each of these provinces separately. Adding their six household counts in Table 3 gives 6,124,638 households as of 1 April 2025. That sum is reproducible from the census, but it is not a count of potential buyers, a new-household flow or a demand estimate issued by REIC.
REIC’s second-quarter 2026 release describes 214,083 units in total marketed supply across the same six provinces, of which 85,249 are condominium units and 128,834 are homes in housing-estate projects. The breakdown adds up exactly to the regional total. The release separately reports 202,414 units of remaining unsold supply. The distinction matters: “total supply” is the volume offered within the monitored market, whereas “remaining supply” tells us what was still available at the quarter’s end. They are not independent housing stocks to be summed. Nor do the 202,414 units consist exclusively of unsold condos.
Supply is also used in more than one sense within market commentary. New supply refers to newly launched or newly offered units in the reporting interval; new sales track purchases during that interval. Completions, building permits and registered title transfers belong to other statistical series. A developer can launch a condominium long before construction is finished. Conversely, a completed home may have been marketed in several earlier quarters. Adding quarter-end unsold stocks across successive releases would therefore count some of the same apartments repeatedly. These distinctions become useful when analysing a sales pipeline, but they do not turn housing-market figures into a census of the nation’s dwellings.
Time is the remaining mismatch even after the six provinces have been matched. Households were counted on 1 April 2025, while REIC describes the second quarter of 2026. During that interval households could relocate, divide, merge or change tenure; the two publications do not measure those transitions jointly. Price bands matter as much as calendars. Unsold houses in one suburban province do not necessarily compete with one-bedroom condominium units close to employment centres in Bangkok. The aggregate says something about the size of the monitored market. It cannot tell a buyer what is available within walking distance of an intended workplace.
At the national level the comparison is weaker still. NSO’s 25,873,003 households cover the kingdom; REIC’s 359,433 total-supply units cover only 27 surveyed provinces in Q2 2026. The REIC release uses a nationwide survey frame in ordinary language, but that does not turn its figure into the country’s entire housing inventory or even all homes offered by every seller. Neither zero supply nor zero demand can be inferred for unobserved areas. A national “oversupply percentage” produced by dividing these figures would have no defensible interpretation.
Another boundary hides behind the label Greater Bangkok. NSO’s 2025 Household Socio-economic Survey defines its income region as Bangkok plus Nonthaburi, Pathum Thani and Samut Prakan — four provinces, not REIC’s six. The reported regional income average can illuminate purchasing-power constraints, but it cannot be attached unqualified to the six-province household total. This is why the province list belongs next to the data, not in a footnote that disappears when somebody copies the chart.
Signals of a defensible comparison
A useful market comparison shows what was counted before asking the reader to draw a conclusion. These signals indicate statistical care; they do not certify that a particular property is well priced.
The six provinces are listed
Why it matters“Bangkok” can mean the city, a four-province income region or the six-province housing market.
Each figure has its own date
Why it mattersAn April 2025 census and a Q2 2026 sales-market snapshot describe different moments.
Total and unsold supply are separate
Why it mattersThe remaining unsold units are included in the broader monitored offer, not added to it.
The census version is identified
Why it mattersJuly’s Table 3 is distinguished from NSO’s January early announcement and registered addresses.
The missing market evidence is explicit
Why it mattersTenure, second homes, achieved sales, local prices and borrower affordability remain separate questions.
Does a smaller household mean demand for a smaller condo
The decline in household size is real within the published census series, but its commercial meaning is easy to overstate. The NSO reports 3.2 people per household in 2010 and 2.7 in 2025. Its provincial table puts Bangkok at 2.6 and nearby Samut Prakan at 2.1. Those averages describe living arrangements on the census date. They do not tell us whether people would prefer a studio over a family home, whether they can afford a mortgage or whether they plan to buy at all.
Two one-person households can have opposite needs. One resident may rent near a transport interchange to reduce commuting time. Another may occupy a long-owned family house and have no interest in a new apartment. Two people sharing a home might want a second room for work, or they might be priced out of new developments and stay in a rental. Age, housing tenure, children, proximity to work, available credit and disposable income affect the choice. The headline average contains none of that distribution.
Thailand’s census also describes far more housing types than the apartments marketed in a developer survey. In its breakdown of living quarters used by private households, 83.8% were detached houses and 3.3% were in the condominium-or-mansion category. Those are descriptions of existing living arrangements nationwide, not shares of 2026 purchases and certainly not a forecast of condo preferences near Bangkok’s transit lines. Treating an occupied-dwelling category as the next product people will purchase would produce a misleading development strategy. It is the local intersection of household needs, price and available stock that matters.
The historical count raises a related issue. NSO records 20,329,171 households in 2010 and 25,873,003 in 2025. A larger stock after fifteen years does not reveal the number of households formed along the way. New households can be created while others dissolve; people may consolidate households, migrate or change where they normally reside. The 2025 operation also used a digital-first collection approach alongside field enumeration. Changes in collection practice deserve attention before interpreting small differences as exact rates. Taking the net change and dividing it by fifteen would not yield the number of new-home customers per year, let alone an estimate for 2026.
The census population is not restricted simply to citizens named in a registered house book. Its published coverage includes foreign civilians whose usual residence is in Thailand and those who had been temporarily present for three months or more as of census day. Foreign civilians staying for less than three months are outside that stated coverage, subject to the other enumerated exclusions. That matters when interpreting resort markets: some established foreign residents are counted, while transient visitors and many second-home users are not a census measure of resident household demand. Fewer people per household may increase the number of separate living arrangements, but the eventual choice between renting, owning, sharing and buying a newly built condo cannot be inferred from the average.
Where demographic assumptions diverge from housing demand
Smaller households will all want smaller condos
Household size says nothing by itself about purchasing power, tenure or required space.
TipUnit mix and local buyers matter more than a single average.
Every additional household becomes a new-home buyer
New living arrangements can be accommodated through rentals, sharing or existing homes.
TipNet household change is not gross formation or new-build purchases.
Unsold developer stock means local housing needs are met
Homes can stay unsold because their price, location or type misses local needs.
TipA sales inventory is not a census of habitable dwellings.
Resident households capture everyone using the local housing market
Short-stay visitors and some second-home buyers fall outside the resident count.
TipEligible foreign usual residents are included; the distinction is length and type of stay.
How much should these figures influence a property decision
For an owner-occupier, the census is a useful description of the wider community but a poor substitute for inspecting the immediate market. A household count cannot tell you how far a particular condominium is from the office, whether it is noisy at night or whether a better-priced completed unit is available across the street. Compare the property’s actual size and condition, transport links, running costs and competing homes within the same catchment area. Regional supply is background to that decision, not a verdict on whether the place will be comfortable or worth its asking price to its owner.
A long-term rental purchase puts tenant purchasing power closer to the centre of the analysis. NSO’s 2025 Household Socio-economic Survey reports average monthly household income of THB 28,308 nationwide and THB 37,505 for Greater Bangkok as defined in that survey. It also reports average monthly expenditure of THB 30,674 for the latter area. This Greater Bangkok includes only the capital and three surrounding provinces; it excludes two provinces in REIC’s six-province market. The figures are survey averages, not the typical earnings of tenants shopping in one building, and expenditure cannot simply be subtracted from income to produce an approved mortgage or rental budget. The distribution of incomes and the costs faced by the actual target group are essential missing pieces.
Rental demand is a flow of decisions, not the stock of people already housed. Many resident households renew an existing tenancy or stay in a long-held home; others move for work, education or family reasons. To judge a rental flat, achieved rents for comparable apartments, vacancy durations, competing buildings, management costs and the employment base of the neighbourhood matter more than a national household trend. If the only evidence is advertised rents, the distinction between asking prices and signed leases should remain explicit. A projection of occupancy cannot responsibly be reconstructed from the number of households alone.
A holiday property operates to a different rhythm. Visitors staying for a short trip do not necessarily enter the resident-household count, and someone may buy a second home in Thailand while maintaining their main household overseas. Other foreign residents fall within the census coverage where the residence rules are met. Tourism seasonality, permissible rental arrangements, condominium rules, service charges and the cost of empty months all change the economics. A province with a modest census household count might still host a substantial visitor market, but this material provides no local visitor-night or achieved-occupancy series from which to calculate that demand.
Buyers planning to resell an off-plan unit face a competition problem more directly connected to REIC’s evidence. Its six-province total includes 85,249 condominium units and 128,834 homes in housing-estate projects in Q2 2026. Only a subset would compete with a particular condo, and timing matters: a launch today, an unsold completed apartment and a future handover are different alternatives for the next buyer. Relevant evidence would include nearby units in the same price bracket, developer incentives, actual transaction volumes, construction schedules and the contract’s exit restrictions. An aggregate unsold stock of 202,414 across all monitored housing types cannot predict a resale date or margin for an individual unit.
Oversupply itself needs specifying before it can be measured. Is there a physical excess of usable dwellings, slow turnover in new projects, a mismatch between prices and household incomes, or a shortage of homes in employment-accessible neighbourhoods despite unsold stock elsewhere? These conditions can coexist. A credible district-level assessment therefore needs the age and composition of unsold projects, sales rates, future completions, second-hand availability, rental performance and financial access for the intended customer. The census provides a demographic foundation; the market study provides a partial sales-pipeline picture. Neither can carry the final investment decision on its own.
Which option fits your situation
A provincial household count cannot tell whether the address works for your daily life.
Without local lease evidence, household growth is not a rental-yield forecast.
Resident households do not count visitor-nights or second-home use.
Six-province unsold totals cannot establish one project’s resale price or liquidity.
Follow-up questions
Why does NSO show both 26.30 million and 25.873 million households?
NSO’s 26 January 2026 early-results announcement gives 26.30 million households. Table 3 of its July 2026 Preliminary Report gives 25,873,003 for the same census date of 1 April 2025. The reviewed releases do not establish why the counts differ, so it would be misleading to attribute the gap to a specific correction or methodological change. A September 2026 final report also gives the national total as approximately 25.9 million. The exact provincial values shown here belong to July’s named table and should not be treated as verified values from the subsequent edition.
Can these figures tell us how many households formed in 2026?
No. The 2025 census gives a household stock on a single date, while the Q2 2026 REIC figures are about property offers. A difference between 2010 and 2025 census totals would measure net stock change over that span, not the gross flow of new households. An estimate for 2026 would need comparable household time-series data and a clear definition of formation. Annualising the earlier net change would not identify 2026 purchasers.
Why aren't Phuket and Pattaya included in the Greater Bangkok comparison?
The table uses a defined six-province Greater Bangkok footprint shared by the census comparison and REIC's regional release. Phuket is a separate province, while Pattaya is in Chon Buri; neither belongs to that footprint. A useful comparison for either destination would need its own matching population geography, housing category and reporting period. Bangkok's offer cannot be allocated to those markets by assumption.
Are registered houses the same as occupied housing units?
No. A registered house or address is an administrative record, not evidence that it is occupied by one household or represents one saleable condo unit. NSO’s census aims to count people and households under its usual-residence rules, which differ from house-registration records. One building can contain several households, while another dwelling may be vacant or used intermittently. Occupied homes, registered addresses and developer sales inventory need separate measures.
Expert view

The 202,414 unsold units in REIC’s Q2 2026 snapshot describe stock across several housing types and six provinces, not how many realistic alternatives one buyer has. A commuter choosing a condo near a station and someone shopping for a suburban house face different competition. I judge pricing pressure against nearby units of the same type and price, then look at achieved sales rather than just launches. For a rental purchase, tenant income and genuine leasing evidence would change the conclusion far more than the national household total.
Sources and check dates
Sources5
- National Statistical Office (NSO) — Preliminary Report, The 2025 Population and Housing Census (published July 2026; Table 3, pp. 66–67; Figure 2.9; coverage §1.4)
Primary census source for 1 Apr 2025: 25,873,003 national households, six provincial counts, household sizes, living-quarters categories and foreign-resident coverage rules. Preliminary report; the six-province total is an editorial sum.
- National Statistical Office (NSO) — Early results announcement for the 2025 Population and Housing Census (26 January 2026)
January official communication reports 26.30 million households from the 2025 census; the reviewed documents do not explain its difference from July Table 3.
- National Statistical Office (NSO) — The 2025 Population and Housing Census, Final Report (September 2026, preface and household summary)
Later official final census report, dated September 2026; its available household summary gives approximately 25.9 million households, including private and collective types. Exact provincial rows from this edition were not independently checked and are not substituted for July Table 3.
- REIC — official market statistics
Primary REIC release: 214,083 total-supply units in six provinces (85,249 condominiums and 128,834 housing-estate units), 202,414 remaining unsold, and 359,433 total-supply units across 27 surveyed provinces. Developer-market supply, not the total dwelling stock.
Original title: Real Estate Information Center (REIC) — Bangkok and Five Vicinity Provinces Housing Market, Q2 2026 (22 September 2026) - National Statistical Office (NSO) — 2025 Household Socio-economic Survey, Executive Summary (Figure 3.12; regional coverage footnote)
2025 sample survey: THB 28,308 mean monthly household income nationwide, THB 37,505 in Bangkok plus three provinces and THB 30,674 mean expenditure in that four-province region. Its geographic scope differs from REIC; averages do not measure affordability for an individual buyer.
Useful next reading
A practical second opinion
Need help assessing a specific property?
Send the property link, price and your goal. We will separate the documented facts from the points that still need a property-specific check.
What to send: link, budget and target date