Conclusion — the acquisition-cost advantage survives where land prices did not move, and is disappearing in Karuizawa

Start with the conclusion. Three numbers now decide whether a corporate retreat conversion works. First, acquisition cost. Two “idle onsen-town retreats” can look identical, yet in an area like Karuizawa, where published land prices are running at +10.7% a year, the entry price for an asset generating the same income is close to double what it is in Shiga Kogen, where land is sinking at -1.5%. Second, the level of renovation capex. A guestroom-only light refit (¥450k per tsubo) and a full renovation (¥1.20M per tsubo) differ by a factor of two in total investment — but by a factor of 3.2 in the ADR they can support. Third, zoning. Of 23 zoning parcels sampled within 1km of Atami Station, 9 (39%) lie in districts where a new hotel or ryokan cannot be built; for a retreat standing on such a site, converting the existing building is the only route available.

In our modelling, Scenario A — a light refit into simple lodging — returns the highest figures at a GOP yield of 8.3% and a levered IRR of 16.9% after construction inflation, followed by Scenario C, a full renovation into small-scale luxury, at 7.1% and 12.6%. Building a new hotel on the same site yields just 4.8%. Pushed to its essence, the investment appeal of conversion lies not in avoiding the purchase of a building, but in avoiding the need to rebuild one.

The idle stock — 54 sites reached the market in two years, and supply is thinning society by society

First, the population. According to the annual tally of directly operated health facilities published by the National Federation of Health Insurance Societies (Kenporen), society-run retreats numbered 295 at the end of March 2021 (held by 178 societies), 269 at the end of March 2022 (166 societies) and 241 at the end of March 2023 (152 societies). Over two years the facility count fell by 54 and the number of holding societies by 26. Deteriorating society finances driven by rising medical cost burdens sit behind this, and the trend is structural. For scale: directly operated retreats are reported to have numbered 1,581 at the end of fiscal 2000 (per Reloclub), meaning the base has shrunk to less than one-sixth in a quarter of a century.

One caveat matters here: these figures cover health-insurance-society-operated facilities only. No public statistic comprehensively captures the total private idle stock once you include retreats and training centres owned by operating companies, facilities held by mutual aid associations and labour unions, and seminar houses run by universities and school corporations. The 241 society facilities are only the part visible above the waterline. Read the other way, that means for investors the main battlefield is off-market negotiated transactions, before anything reaches the open market.

Seller circumstances also shape conversion economics. Because retreats were treated as an employee-benefit expense even when occupancy was low, many sellers have no experience pricing them as income-producing assets. On top of that, depreciation has run its course on the building, so land value effectively sets the floor on the transaction price. This is the source of the “acquisition-cost advantage” discussed below — and equally the reason that advantage evaporates quickly in areas where land prices have risen.

Location analysis — land price movement splits the five areas in two

For the five retreat-dense areas (Atami City / Ito City–Izu-Kogen / Karuizawa Town / Hakone Town–Gora / Yamanouchi Town–Shiga Kogen), we extracted standard sites within roughly a 4km radius of each area centre from MLIT’s Real Estate Information Library and tracked year-on-year change rates from 2021 to 2025. The result splits cleanly in two.

Karuizawa Town accelerated from +5.16% in 2021 to +10.67% in 2025, lifting the median standard site price 44.7% over those five years, from ¥64,900 to ¥93,900 per m². Permanent-residence demand for second-home land plus new luxury hotel development is driving land higher. Yamanouchi Town (Shiga Kogen), by contrast, has not turned positive once in five years and was still at -1.50% in 2025. Atami City recovered gently from +0.12% in 2021 to +3.34% in 2025, and Hakone Town swung from -2.33% to +3.89%. Ito City (Izu-Kogen) only just escaped decline in 2025, at +0.01% — its first non-negative print in five years.

Zoning directly governs whether conversion is possible at all. Under Appended Table 2 of the Building Standards Act, hotels and ryokan cannot be built in Category I or II low-rise exclusive residential districts, Category I or II mid/high-rise exclusive residential districts, industrial districts or exclusive industrial districts (in Category I residential districts they are capped at 3,000 m²). Of the 23 parcels extracted within a 1km radius of Atami Station, 7 were Category II mid/high-rise exclusive residential, 1 was Category I mid/high-rise exclusive residential and 1 was Category I low-rise exclusive residential — 9 parcels in total (39.1%) where a new hotel or ryokan cannot be built. In Karuizawa, 3 of the 5 parcels extracted are Category I low-rise exclusive residential.

Here lies a point specific to conversion schemes. Zoning restrictions apply in principle to new construction and to changes of use, so converting an existing retreat into a lodging-business facility faces the same constraint. That said, the 2019 amendment to the Building Standards Act (in force June 2019) removed the requirement for a building confirmation application where the floor area subject to the change of use is 200 m² or less. A retreat sitting in a low-rise exclusive residential district will, as a rule, not be able to obtain a lodging business licence; in areas with no designated zoning district, such as Izu-Kogen and Shiga Kogen, no use restriction under the Building Standards Act applies at all. Cheap land plus loose use restrictions — that double condition is precisely what the residual acquisition-cost advantage in these two areas consists of.

The absence of zoning does not, however, mean the absence of constraint. Shiga Kogen falls partly within Joshinetsu-Kogen National Park and Izu-Kogen within Fuji-Hakone-Izu National Park, so construction, extension or alteration of structures requires permission or notification under the Natural Parks Act. Restrictions on exterior appearance, colour and height reach existing-building renovations too, which makes “how much of the existing condition, including legally non-conforming elements, can be retained” the single largest practical variable.

Area distribution map — land price movement overlaid on lodging stock

The trade area data sharpens the contrast between the five locations further. Atami holds 1,500 establishments and 11,614 employees within 1.5km, with an office-sector share of 11.1% — the most urban of the resource areas. Hakone Gora, at 3.0% office sector, is overwhelmingly tourism employment. Izu-Kogen has only one small district covered in the Economic Census (89 establishments, 370 employees), which makes it a location where demand simply cannot be argued from trade area data.

The population outlook is harsher still. Within 2km of Atami, the 15,937 residents recorded in 2020 are projected to fall 27.1% by 2040, and the aged share has already reached 50.0%. Izu-Kogen shows an aged share of 58.0% and a 31.2% decline by 2040. In these areas, demand for a converted asset cannot be sought from the local population. Conversely, the post-conversion business plan must be built on the assumption of 100% out-of-area and inbound demand. The single exception is Karuizawa, where the 2040 decline is just 3.9% and the aged share 37.4% — second-home land turning into permanent residence is holding up the population structure. For how trade-area ageing constrains the exit options available to an owner, see Aging Trade Areas Constrain Hotel Exits: Atami 49.1% vs Takayama 38.8%, which examines Atami and Takayama in depth.

What the mesh map shows is a structure in which Atami’s resident population clings to the seaward slopes below the station and thins rapidly on the inland hillsides. Most corporate retreats sit precisely on those inland, elevated slopes. They hold an irreplaceable asset in the view, but they are distant from everyday convenience, and staffing plus the logistics of food and linen determine the operating cost of a converted property. This feeds directly into the differences in GOP margin by scenario discussed below.

Post-conversion revenue — estimated settled ADR is up 10.4% in Atami and 12.7% in Karuizawa

After acquisition cost comes the revenue side. Tracking estimated settled ADR across the five areas in MetroEngines Research data, the first-half 2026 (January–June) average was ¥26,200 in Hakone Town (+4.8% year on year, N=173–187 properties), ¥24,000 in Atami City (+10.4%, N=100–105), ¥21,300 in Karuizawa Town (+12.7%, N=49–52), ¥20,200 in Ito City (-1.3%, N=123–126) and ¥12,800 in Yamanouchi Town (+7.2%, N=89–118). Ito City is the only one below the prior-year period, yet in level terms it still sits above the ryokan category for Shizuoka Prefecture as a whole (¥18,700, N=453 properties).

What deserves attention here is less the level than the seasonal amplitude. Across full-year 2025, Karuizawa Town’s ratio between its cheapest month (June, ¥16,900) and its most expensive (August, ¥37,000) reaches 2.18x. Hakone Town, by contrast, sits at 1.27x, Atami City at 1.49x and Ito City at 1.36x. Where amplitude is large, the majority of annual GOP concentrates into just two or three months, so the same average annual ADR implies a completely different working capital design and a different tolerance for off-season fixed costs. A Karuizawa retreat conversion has to be designed to “take it all in summer”, or the levelled funding plan will not hold.

A second question is the balance between revenue and acquisition cost. Dividing first-half 2026 estimated settled ADR by the 2025 median published land price (¥/m²) gives an index of 0.49 for Ito City, 0.49 for Yamanouchi Town, 0.41 for Hakone Town and 0.39 for Atami City — against a conspicuously low 0.23 for Karuizawa Town. Even though Karuizawa’s ADR is growing at +12.7%, land is rising faster still, so its entry price relative to income is the heaviest of the five areas. Ito City, conversely, has ADR below the prior year yet ranks at the top of the index because land has barely moved in five years.

Development scenario modelling — renovation capex moves the yield by two points

Building on the location and revenue conditions above, we model investment economics on a standardised conversion. The assumed asset is a retreat with a 1,500 m² site, 1,000 m² of gross floor area (302.5 tsubo), RC construction, completed in the 1990s, with around 20 original guestrooms. Using Atami City’s median published land price (¥61,600/m²), land is valued at ¥92M and the building at ¥28M, for an acquisition price of ¥120M. Renovation capex is set at three levels with reference to prevailing per-tsubo ranges (guestroom-only refit ¥500k–1.0M/tsubo; full renovation including common areas and building services ¥1.2M–2.0M/tsubo).

Reading the results: Scenario A posts the highest GOP yield at 8.27%, but it does so by trading low ADR and minimal service for a 45% GOP margin. An operating model built on room-only stays, self check-in and no food service fits well with the “hard-to-staff hillside location” identified in the previous section. Scenario C, on the other hand, generates the largest GOP in absolute terms (¥37.15M), and that is what drives the exit sale price. On a 10-year hold at a 6.0% exit cap rate, the levered IRR ranking runs A at 16.9%, C at 12.6% and B at 9.9%.

Scenario B looks weakest because the GOP margin of a two-meals-included ryokan is structurally low (an industry rule of thumb of 5–15%; we adopt a conservative 20% here). That does not, however, mean B has little room to improve. Lift the GOP margin by a few points through better food and beverage gross margin, repurposing the banquet hall for events, or selling room-only plans alongside the full-board product, and the yield improves immediately. Scenarios A and C, by contrast, face structural constraints — a low ADR ceiling and heavy capex respectively — that operational effort cannot easily overturn.

The most important comparison is against a new build. Delivering the same location and the same product (10 rooms, ADR ¥45,000) through new construction costs ¥613M on the ¥2.026M/tsubo RC basis from MLIT’s 2025 “Statistics on Building Construction Starts”, or ¥678M after inflation, dropping the GOP yield to 4.82%. That is 2.3 points below conversion Scenario C, and on total investment per room the gap is ¥52.20M versus ¥77.00M — 32%. Rising construction costs suppress new supply and raise the scarcity of existing assets, which makes them a double tailwind for conversion schemes. Where exactly the boundary falls between renovating and rebuilding depends on how much ADR lift each yen of capex can buy, and has to be tested project by project against recent large-scale renewal precedents.

Sensitivity analysis and market benchmarks — what still works at ADR -10% and OCC -5pt

To test the robustness of the modelling, we ran a sensitivity analysis on each post-inflation scenario with ADR and OCC shaded downwards.

Read across both axes (all on a full-year modelling basis), even the bottom-left cell of the grid — ADR ¥11,200 × OCC 48% — returns a GOP yield of 5.47%, which is 0.65 points above the 4.82% a new build of the same location and product would deliver. Let ADR fall 20% below base and hold OCC at 58% and the yield is 6.61%; let OCC fall to 48% while ADR holds at base and it is 6.84%. No single-factor downside breaks the advantage over a new build. The robustness of the conversion scheme derives not from the low ADR assumption itself but from the lightness of that ¥16.94M total investment per room. Note that the middle row of this grid matches the Scenario A downside sensitivity table (ADR -10% = 7.44%; ADR -10% with OCC -5pt = 6.80%).

Even under the harsher combined downside of ADR -10% and OCC -5pt, all three options hold a GOP yield above 5.3%. The Japan Real Estate Institute’s “53rd Real Estate Investor Survey” (as of October 2025) reports expected yields for limited-service hotels of 4.2% in Tokyo and 4.4% in Osaka; even allowing a risk premium for a resort location and small scale, the range in this modelling sits within investment-grade territory. That said, these figures assume the acquisition price can be held down to a land-valuation basis. Push the acquisition price from ¥120M to ¥200M and Scenario A’s yield falls from 8.27% to 6.39%, a drop of 1.9 points. Backing a maximum acquisition price out of the revenue side — dividing achievable GOP by a target yield — is the discipline that keeps this exposure bounded.

It is worth checking actual transaction levels as well. Extracting onsen and highland resort acquisitions from our ownership master (7 listed REITs, 301 properties, as of 20 July 2026) gives the following.

Per-room acquisition prices for the three onsen resorts bought in the last three years (2023–2025) run from ¥52.28M to ¥101.15M, with a median of ¥58.52M. The conversion scenarios modelled here come to ¥16.94M per room for A, ¥27.71M for B and ¥52.20M for C — so even the heaviest, C, lands below what listed REITs actually pay per room. That is the quantitative meaning of “IRR upside from an acquisition-cost advantage”. It should be said, though, that what a REIT buys is a stabilised asset with an operating track record and a management platform in place, whereas a conversion carries the risk of ramping up from zero. Part of that unit-price gap is precisely that risk premium.

As an operating benchmark, Hoshino Resorts REIT (星野リゾート・リート投資法人), which holds predominantly resort properties, disclosed portfolio-wide results for May 2026 of 80.1% occupancy (+1.1 points year on year), ADR of ¥22,950 (+2.4%) and RevPAR of ¥18,373 (+3.8%). The OCC assumptions of 58–63% used here sit almost 20 points below that — a conservative setting we regard as a fair reflection of realistic ramp-up performance.

Risk assessment and investment judgement — three conditions for viability

The investment case that follows comes down to three points. First, the initial screening gate is zoning. Before any discussion of profitability, confirm that the retreat sits in a district where a lodging business facility is permitted. That test alone eliminates 30–60% of candidates. Second, the acquisition-cost advantage can be measured by land price movement. In areas where land is moving at +10% a year the advantage disappears within one to two years; where land is flat or falling, it is preserved. Third, the level of renovation capex determines the shape of the return. A light refit maximises yield and IRR; a full renovation maximises exit value and absolute GOP. Which is correct depends on the investor’s holding period and exit strategy.

Corporate retreats, as an idle asset class, come to market with book value written off while still holding onsen rights, large communal baths and views — assets that simply cannot be newly acquired today. With construction costs staying high and new-build yields slipping below 5%, converting this idle stock can be a realistic route in for regional banks, regional developers and operating companies. Whether it works, though, has to be judged deal by deal on three numbers — zoning, land price movement and renovation unit cost — not on generalisations about location.