The Building Starts Survey (建築着工統計調査) is a monthly statistic published by Japan’s Ministry of Land, Infrastructure, Transport and Tourism (MLIT), compiled from the construction notifications (建築工事届) required under Article 15 of the Building Standards Act. It is the only public dataset that tracks, on a notification basis, when, where and how much hotel and ryokan construction has begun. Extracting the mid-level use category “52 Lodging” shows that 2024 saw 2,836 buildings, 1,756,286 ㎡ of floor area and ¥1.0384 trillion in estimated construction cost. Floor area rose +94.5% YoY, recovering to 57.8% of the 2018 peak. This article is the third in our series of reading guides, following those on the Japan Tourism Agency’s Overnight Travel Statistics Survey and the Inbound Consumption Trend Survey. It covers how the statistic is built, how to extract the hotel portion from e-Stat, what the three reported metrics mean, how to read the monthly and prefecture-level figures, and how it complements our own hotel opening pipeline data.

HotelBank has published two articles that use this statistic to analyze hotel supply: “Hotel Construction Starts Doubled in 2024 to 57% of the Peak”, which projects 2025–2028 supply from prefecture-level floor area trends and construction periods, and “16.7% of Hotel Supply Is Non-New Construction”, which breaks down the share of new builds, extensions and reconstructions using the tables by type of construction. This article does not repeat those analyses; it focuses on the background knowledge needed to read them.

Overview of the Building Starts Survey — Built on Construction Notifications, Published Monthly in Three Parts

The first thing to understand is that this statistic is not a questionnaire-based survey. It is closer to an administrative statistic that directly tallies legally required notifications. Article 15(1) of the Building Standards Act requires any building owner who intends to construct a building of more than 10 ㎡ in floor area to submit a construction notification to the prefectural governor. MLIT’s Construction Economics Statistics Office (Policy Bureau) collects these notifications via the prefectures, aggregates them by month of start, and publishes the results at the end of the following month. Figures for August 2026 were released on September 30, and September figures are scheduled for release on October 30.

The statistic consists of three parts. The first is the “Building Starts Statistics” (建築物着工統計), which tabulates all buildings started — residential or not — by owner, use, structure, prefecture and so on. The second is the “Housing Starts Statistics” (住宅着工統計), which tabulates the number of new dwelling units by tenure (owner-occupied, rental, for sale) and type of building. The third is the “Adjustment Survey” (補正調査), which estimates the gap between the cost estimated at notification and the cost actually incurred after completion through field surveys of a sample, and derives coefficients to adjust estimated costs to an actual basis. To track hotel construction starts you use the first part, the Building Starts Statistics; lodging facilities are not included in the Housing Starts Statistics.

On e-Stat the government statistics code is 00600120, and tables are provided at three frequencies: monthly, calendar year and fiscal year. The calendar-year figures sum January–December, the fiscal-year figures sum April through the following March, and the “annual total” in press releases refers to the calendar year. Mixing these up can shift floor area for the “same” 2024 by more than 10%.

How to Extract Hotels — Use Category “52 Lodging” and Two Pitfalls

So where in the vast set of tables are the hotels? The use classification of the Building Starts Statistics follows the Japan Standard Industrial Classification: under the major category “M Buildings for accommodation and food services” sit the mid-level categories “52 Lodging” and “53 Food services.” Lodging construction starts can be extracted by selecting category 52. In the e-Stat table “By use and owner / number of buildings, floor area, estimated construction cost” (annual table 0003114490), the use classification code (cat02) is 57, owner “Total” is 11, and the nationwide area code is 00000.

Source: Prepared by HotelBank Editorial Team based on the use classification of MLIT “Building Starts Survey”

There are two pitfalls to watch for. The first is how mixed-use buildings are treated. Buildings combined with housing are counted in a separate bucket — mid-level category “20 Residential combined with accommodation and food services” under major category “C Residential buildings combined with industrial use” — and are not included in “52 Lodging.” This category recorded 627 buildings and 550,183 ㎡ in 2024, equivalent to just over 30% of lodging-use floor area, but the tables do not break it down into lodging, food service and residential portions. The second is the counting rule for multi-use buildings: a complex housing offices, retail and a hotel is counted entirely under the use with the largest floor area (the principal use). A hotel inside a redevelopment tower whose principal use is office does not appear in lodging-use floor area at all. The more large-scale urban mixed-use developments there are in a given year, the more the lodging-use figures are biased toward understating reality.

The Three Reported Metrics — Reading Buildings, Floor Area and Estimated Construction Cost

The main tables of the Building Starts Statistics all report three items: “number of buildings,” “total floor area (㎡)” and “estimated construction cost (¥10,000).” These measure the same starts from different angles, and the picture changes depending on which you treat as primary. The building count indicates volume of cases. In 2024 lodging use, Yamanashi ranked second nationally with 253 buildings, behind Kagoshima (291), yet its floor area was only 39,402 ㎡ — 156 ㎡ per building — showing that small vacation rentals and cottage-type facilities dominate. Tokyo, conversely, had just 67 buildings but 530,755 ㎡, or 7,922 ㎡ per building, centered on large hotels with many rooms. If you want to gauge supply in terms of room count, look at floor area rather than building count.

Estimated construction cost is the anticipated amount stated by the building owner at notification. It is not the actual figure after completion and does not reflect changes in material prices or design. That is why the Adjustment Survey mentioned above exists; the estimated amounts in the statistic are most accurately read as “the size of the investment intent at the time of start.” That said, dividing by floor area to get “estimated construction cost per ㎡” provides a convenient indicator of construction cost trends. Calculated nationwide for lodging use, it rose 1.8× from ¥332,000/㎡ in 2018 to ¥591,000/㎡ in 2024. Converted to tsubo, that is a rise from roughly ¥1.1 million to ¥1.95 million, likely the combined result of soaring material and labor costs over the period and higher specifications in large urban projects.

One more point that is easily overlooked is the difference between the “number of projects” and the “number of buildings.” The tables by type of construction (annual table 0003114493) include a column for number of projects; in 2024, lodging use recorded 1,288 projects against 2,836 buildings. The more dispersed-type lodging facilities there are where one notification covers multiple buildings (cottage clusters, glamping, etc.), the more the building count exceeds the project count. A year in which the gap between projects and buildings widens can be read as a sign of increased small, dispersed-type starts.

Lodging-Use Actuals 2011–2024 — Floor Area at 57.8% of Peak, Unit Cost Up 1.8×

Now let us look at the actual figures. The table below shows lodging use (owner = total, nationwide) for 2011–2024, taken from e-Stat annual table 0003114490. Floor area doubled from 929,353 ㎡ in 2015 to 1,956,867 ㎡ in 2016, peaked at 3,039,844 ㎡ in 2018, and then declined for five consecutive years from 2019 through 2023. The 902,844 ㎡ of 2023 was below the 2015 level, but 2024 rebounded to 1,756,286 ㎡, +94.5% YoY, recovering to 57.8% of the peak.

Source: Prepared by HotelBank Editorial Team from MLIT “Building Starts Survey,” Building Starts Statistics annual table (by use and owner, nationwide, owner = total, use = 52 Lodging). Estimated construction cost converted from the original ¥10,000 units to ¥100 million; per-㎡ figures calculated by the editorial team as estimated construction cost ÷ floor area

Source: Prepared by HotelBank Editorial Team from MLIT “Building Starts Survey” (e-Stat table 0003114490)

Focusing on years when building counts and floor area move in different directions deepens your reading of the statistic. In 2022, floor area fell 9.0% YoY while the building count rose 66.9%, shrinking floor area per building from 927 ㎡ to 506 ㎡. Downsizing continued to 406 ㎡ in 2023 before recovering to 619 ㎡ in 2024. Compared with 1,435 ㎡ per building in 2018, recent starts remain small on average. Rather than a full-fledged return of large hotel starts, the picture is one of rising small, dispersed-type starts plus a handful of mega-projects pushing up floor area.

Source: Prepared by HotelBank Editorial Team from MLIT “Building Starts Survey” (e-Stat table 0003114490). Estimated construction cost per ㎡ calculated by the editorial team

Estimated construction cost reached ¥1.0384 trillion in 2024, exceeding the ¥1.0086 trillion of 2018 and marking the highest level since 2011, the earliest year available in the annual tables. The amount hit a record even though floor area remained just under 60% of the peak because the unit cost rose to ¥591,000 per ㎡. That is 2.8× the ¥212,000 of 2011, with a sharp jump in 2024 from ¥410,000–420,000 in 2022–2023. The per-㎡ cost was ¥609,000 in Tokyo, ¥969,000 in Osaka and ¥980,000 in Kanagawa, with large projects in major cities pulling up the average. When estimating construction costs for investment decisions, the per-㎡ cost from this statistic can serve as a starting point for a sense of regional and annual market levels, bearing in mind that it reflects “estimated amounts at notification.”

Reading the Monthly Data — Flash Releases Stop at the Major Category; Mid-Level Data Is in the e-Stat Database

When tracking monthly trends, the level of detail you can reach depends on which table you use. MLIT’s end-of-month press release, the “Building Starts Survey Report,” publishes floor area by building use at the major-category level, “Accommodation and food services.” For August 2026, this category recorded 250,000 ㎡, +67.4% YoY, or +69.5% for private-sector owners. Monthly tables broken down to mid-level category “52 Lodging,” meanwhile, can be found in three tables in e-Stat’s database function: “By use and owner” (monthly table 0004000400), “By use and structure” (0003114407) and “By use and type of construction” (0003114342). However, as of this writing (October 2026), the latest month in these database tables is December 2024; mid-level monthly figures for 2025 onward have not yet been loaded. To follow the latest mid-level trends, you need to refer directly to the Excel time-series tables attached to the press releases or to the statistical table files (PDF/Excel) on e-Stat.

The basic procedure for reading monthly data has three steps. First, compare the latest month with the same month of the previous year. Second, compare the cumulative total from January through the latest month with the cumulative total for the same period of the previous year. A single month can swing 2–5× on one mega-project, so you will misread the trend unless you look at cumulative totals. Third, do not rely on seasonality. Even the monthly medians of the major category “Accommodation and food services” for 2017–2025 show only that January, April and October tend to be high and March, September and December low; year-to-year swings are larger. Month-on-month changes carry almost no meaning.

Source: Prepared by HotelBank Editorial Team from the time-series tables (floor area by owner, use and structure) attached to MLIT “Building Starts Survey Report (August 2026).” Major category “Accommodation and food services” (includes restaurants); unit: thousand ㎡

Applying this procedure to 2026, the January–August cumulative total for the major category “Accommodation and food services” was 1,505,000 ㎡, down 19.9% from 1,879,000 ㎡ in the same period of the previous year. However, April 2025 had an outlier single-month value of 821,000 ㎡; excluding it, the remaining seven months show 1,287,000 ㎡ in 2026 versus 1,058,000 ㎡ in 2025, turning the comparison to +21.7%. This is a textbook case of a single mega-project reversing an annual comparison, so even when comparing cumulative totals it is worth checking for single-month outliers. The calendar-year totals for the major category were 2,692,000 ㎡ in 2024 and 2,537,000 ㎡ in 2025 (-5.8%), and 2026 had reached 59% of the 2025 total as of August. Note that because this major category includes food services, the share of lodging use (mid-level) fluctuated between just over 40% and just over 80% during 2011–2024 (65.2% in 2024, 43.1% in 2023).

By Prefecture — Tokyo Holds 30.2% of 2024 Lodging-Use Floor Area; Top Three Prefectures 49.3%

Annual table 0003114490 also provides values for all 47 prefectures by area code, so ranking 2024 lodging-use floor area by prefecture under the same conditions yields the table below. Tokyo accounted for 30.2% of the national total with 530,755 ㎡, followed by Hokkaido at 191,797 ㎡ (10.9%) and Osaka at 143,401 ㎡ (8.2%). The top three prefectures reach 49.3% and the top ten 68.6%, indicating strong regional concentration of starts. Year-on-year, Tokyo was up 687%, Hokkaido 171% and Osaka 165%, showing that the nationwide doubling in 2024 was driven by large projects in major cities and Hokkaido.

Source: Prepared by HotelBank Editorial Team from MLIT “Building Starts Survey,” Building Starts Statistics annual table 0003114490 (2024 and 2023, owner = total, use = 52 Lodging, 47 prefectures). Share, per-building and per-㎡ figures calculated by the editorial team

Source: Prepared by HotelBank Editorial Team from MLIT “Building Starts Survey” (e-Stat table 0003114490, 2024)

When reading the prefecture table, we recommend looking at the building-count ranking and the floor-area ranking together. Yamanashi ranks second nationally by building count with 253 (Kagoshima leads with 291) but ninth by floor area, and Chiba, third by building count with 209, ranks sixth by floor area — both centered on small facilities of around 200 ㎡ per building. Ishikawa jumped 23-fold from 2,147 ㎡ the previous year to 50,167 ㎡, and its per-㎡ cost of ¥1,059,000 is the highest among the top ten prefectures. This may reflect reconstruction after the 2024 Noto Peninsula Earthquake coinciding with the start of large projects in Kanazawa, but individual projects cannot be identified from the tables, so interpretation requires cross-checking against local development information.

Building Starts vs. Opening Pipeline Data — Starts for Two to Three Years Out, the Pipeline List for One to Two Years Out

Building starts record the point at which “construction began,” and opening follows one to four years later. In our supply pipeline research, we measured the median hotel construction period at 24 months for standalone buildings and 49 months for mixed-use complexes, so the 2024 doubling in building starts can reasonably be taken to correspond to openings in 2026–2028 (for details on construction periods, see “Hotel Construction Periods: Median 24 Months for Standalone Buildings, 49 Months for Mixed-Use Complexes” and the article on the doubling of construction starts mentioned above).

The opening pipeline data maintained by MetroEngines Research & Consulting, by contrast, consolidates “where, how many rooms and when” from building-plan notices and press releases, and differs from building starts in both population and granularity. As of October 2026, we track 70 projects with 50 or more rooms totaling 15,504 rooms, with scheduled completion of 22 in 2025, 20 in 2026, 18 in 2027, 8 in 2028 and 2 in 2029 or later. Building starts, on the other hand, cover every building subject to the notification requirement but reveal neither room counts nor facility names. The table below summarizes the characteristics of the two.

Source: Prepared by HotelBank Editorial Team from MLIT “Building Starts Survey” / MetroEngines Research & Consulting (opening pipeline data as of October 2026, N=70)

The division of labor in practice is clear. To gauge total supply pressure two to three years out at the area level, use building starts; to count individual competitors or trade-area rooms one to two years out, use the opening pipeline list. In our count of new openings based on confirmed OTA listings, 1,749 facilities have been identified for 2026 and 27 for 2027 as of October, but the 2027 figure does not mean that supply is disappearing. Listings appear only a few months before opening, so counts for the following year and beyond will inevitably grow over time. Given the doubling of building starts in 2024, it is natural to expect 2027 openings to far exceed the number currently identified.

Limitations and Caveats — Estimated Costs, Mixed Use, Construction Type, Flash vs. Annual Figures

To summarize, while the Building Starts Survey excels in coverage and timeliness, four limitations must be kept in mind when using it for hotel analysis. First, estimated construction cost is the amount anticipated at notification, not the actual cost. Material price surges and design changes after the start are not reflected, and the gap with actual costs cannot be known without looking at the Adjustment Survey results. The per-㎡ time series should be read as the trend in “market levels at notification.”

Second, the hotel portion of mixed-use buildings is invisible. Hotels in complexes whose principal use is office or retail are counted in neither lodging-use floor area nor building count, and buildings combined with housing fall into the separate category “Residential combined with accommodation and food services.” The more hotels incorporated into large urban redevelopments in a given year, the more lodging-use figures are understated. A separate table exists for multi-use buildings, “Multi-use buildings by use and structure” (annual table 0003114494), but its breakdown is also based on principal use.

Third, extensions, reconstructions and changes of use (conversions) are only visible in the “By use and type of construction” table. In 2024, lodging use comprised new construction of 1,967 buildings and 1,597,784 ㎡, extensions of 853 buildings and 156,019 ㎡, and reconstructions of 16 buildings and 2,483 ㎡, with 9.0% of floor area being non-new construction. Conversions of existing buildings into hotels are subject to notification if the change of use requires building confirmation, but they are classified as “reconstruction” or “extension” by construction type, and the prior use is unknown. This point is covered in detail in our article on the non-new construction ratio.

Fourth, there is a gap between flash figures and annual totals. Monthly press releases are fixed as of the time notifications are tallied, but late submissions and corrections are reflected in the annual totals, so the sum of 12 monthly figures may not match the annual table. The e-Stat annual tables and mid-level monthly tables are loaded with a large lag: as of October 2026, the latest is 2024 data (updated January 2026), and 2025 data has not yet been posted. When citing annual figures in articles or internal materials, it is safest to treat annual table values as authoritative and to note monthly sums as “flash-based.”

Frequently Asked Questions

Summary

Because the Building Starts Survey rests on a legally defined population — construction notifications — it is the most comprehensive public statistic serving as a leading indicator of hotel supply. The key points for using it boil down to three. First, the category to extract is mid-level “52 Lodging,” which is biased toward understatement because of the principal-use rule for mixed-use buildings and the separate category for buildings combined with housing. Second, number of buildings, floor area and estimated construction cost each carry distinct information: read supply volume from floor area, the progress of downsizing from its ratio to building count, and construction cost levels from the per-㎡ cost. Third, understand the difference in granularity — monthly flash figures at the major-category level, mid-level data in the annual tables — and compare cumulative totals rather than single months. The 2024 figures of 2,836 buildings, 1,756,286 ㎡ and ¥1.0384 trillion indicate the total volume of supply that will open in 2026–2028. The practical approach is to move back and forth: use our opening pipeline data when counting individual competitors, and this statistic when projecting total volume.

Related Articles

• Japan Hotel Construction Starts 2024: 97,000 Rooms, 57% of Peak
• Hotel Supply: 16.7% of 5,707 Starts Are Not New Builds
• Overnight Travel Statistics Survey Guide: 61.6% OCC, Prelim vs Final
• Inbound Consumption Survey Explained: Q2 2026 ¥2.51tn, Lodging 36.6%
• Hotel Construction: 24 Months Standalone, 49 Mixed-Use; 2029 Deadlines
• Japan Hotel Pipeline 2028: 106 Projects, 48% of 2027 So Far