Developers do not build farms for the vegetables

The Cannery in Davis, California, is 547 homes built on the site of a tomato cannery. The pitch was simple: a housing development with a working farm attached. The farmland was announced at 7.4 acres, and a three-cornered arrangement was set up — the developer deeded the land to the city, and the city leased it to a non-profit that trains new farmers. So far, a handsome story. Look inside the arrangement, though, and the farm's working capital was being handed straight from the developer to the non-profit. The party selling the houses was paying the party doing the farming. If the produce paid for the farm, that transfer would not be needed. That it was needed means the farm does not stand up as agriculture.

The pattern is not peculiar to Davis. A 2018 Urban Land Institute report gave the name agrihood to residential and mixed-use projects built around a working farm, counted more than two hundred of them in the United States, and set out the developer's advantages plainly: lower amenity costs, greater marketability, faster residential sales. Against a golf course or a pool, a farm is cheaper to install and lighter to maintain. That sentence very nearly settles the question of why private capital plants vegetables. It is not investing in agriculture. It is investing in an amenity — and among amenities, a farm is one of the cheap ones.

What follows dissects eight developments with the same three questions. First: whose accounts carry the cost of building the farm and running it each year — the developer's marketing budget, the residents' association dues, a tenant's rent, the income of an endowment? Second: where is that spending recovered? The route is never the vegetables. It is lot prices, sales pace, tenant attraction, rent, and above all the terms on which a given number of units or a given use was permitted at all. Third: if the farm disappeared, what would pull the trigger — a lease expiring, a subsidy ending, an operator walking away, a change of owner, a decision to build on the site. Putting that third question up front is deliberate: the fragility of city farms is something this database has had to record again and again.

The conclusion, stated up front: in none of the eight cases is the cost recovered by selling produce. What is recovered is permitted density, a fee levied on every resale, footfall for a restaurant tenant, differentiation for an exhibition venue — and, in the two Japanese cases, the cost of holding land that cannot yet be developed without letting it go derelict. That last one sounds duller than rental premiums or public-relations value, but for Japanese practitioners it is probably the most usable logic of the set. It also settles the ending: for land held that way, the day the site becomes developable is the farm's last day.

Prairie Crossing sits in Grayslake, Illinois, about an hour north of Chicago by rail: 677 acres carrying 359 single-family houses and 36 condominiums. What matters about those numbers is that they represent deliberate under-building. An earlier proposal for the same ground had 2,400 homes on it. After that plan collapsed, the group that took the land on set out to combine responsible development, the preservation of open land, and commuting by train — placing roughly 60 per cent of the site under a conservation easement and leaving about 70 per cent of it open. At the centre of that open land is Prairie Crossing Farm. Note the direction of causation: the number of homes was not cut because there was a farm. There was a farm because a plan that cut the number of homes left a great deal of land that needed a use.

Land dialled down to low density is, left alone, a liability that only consumes money. Turn it to lawn and it must be mown forever; leave it and it goes to scrub, and the residents complain. Put a farm on it and three things happen at once: somebody becomes responsible for the ground, visitors acquire a reason to come, and the site can be described as in use rather than vacant. This is the context for the ULI report's observation that a farm is cheaper to install and to keep than a golf course or a pool. Eighteen holes occupy tens of hectares and demand permanent turf management, with revenue that depends on running a club. A farm holds the same area in a state of visible use on far less capital.

The logic has a limit, though. What holds the ground in that state of visible use is the farm operator, not the developer. Prairie Crossing's farm has itself passed between operators over the three decades since it opened. From the development's side the farm is still there; from agriculture's side the business running it can be replaced. What the developer recovered — an approved low-density plan and the story that carried it — was recovered by the time the last lot was sold. Who carries the next thirty years is a separate question, and it stays open.

Willowsford lies on the outer edge of the Washington region, in Loudoun County, Virginia. Most of its land is not held by the homeowners' association but by a separate non-profit, the Willowsford Conservancy, which owns and manages some 2,300 acres of open space, more than 200 acres of farmland and over 40 miles of trail. The interesting part is how money reaches it. A buyer of a home at Willowsford pays the Conservancy 0.25 per cent of the purchase price — not only on the first sale from the builder but each time that house changes hands thereafter. Alongside that sit draws from an endowment, and in some years the budget carries no separate capital assessment at all.

That quarter of a per cent rarely appears in discussions of urban agriculture, but it is one of the most instructive details in the field. First, the cost has been shifted from developer to residents in the form of a charge at each transaction rather than an annual subscription. Second, the obligation passes automatically to future owners: the developer can leave and the body's income continues. Third, because the income is indexed to house prices, the upkeep grows as the area's value grows — a loop premised on green space lifting value, with the lift then funding the green space, written into the paperwork. How well that premise holds empirically is Thursday's subject; what matters here is that even if it fails, the income simply tracks house prices instead of collapsing. As our earlier piece on who pays for the roof argued, green is held up not by principle but by somebody's ledger; here that ledger is embedded in the deed.

It is worth seeing the ground the arrangement stands on. Loudoun County's zoning ordinance demands a high proportion of open space from cluster development — the form that pushes dwellings onto part of a site and leaves the rest unbuilt. In the transitional residential districts the requirement is 50 per cent, rising to 70 per cent in the Lower Bull Run sub-district and in TR10. Leaving a great deal of land open, in other words, was not a choice but a condition. The developer's decision was never whether to keep open space but what to do with open space it had to keep. Reverse that order and the farm starts to look like an act of goodwill; in fact it is the most legible use anyone could find for an area produced by a rule. Note too that marketing material speaks of more than 300 acres of farmland while the Conservancy describes more than 200. Whether the denominators differ or the cultivated area has shrunk cannot be settled from the public documents.

Back to The Cannery. Some 100 acres of former factory land, 547 homes, and farmland announced at 7.4 acres, deeded to the city and leased to a non-profit that trains beginning farmers. Accounts from the site, however, describe about three acres of cropland plus a one-acre orchard, a greenhouse and a barn. Part of the reason was the soil: the ground of a former cannery was not in the state assumed, and had to be built up with organic matter before it would carry crops, which delayed the farm's establishment by that much again. The lesson is the same one the rooftop pieces kept running into, and it holds at ground level too — the area announced is almost always the site, not the growing beds.

The second friction was expectation. People involved on the ground return again and again to the same point: many residents held unrealistic ideas about how the farm would be run and how they would get to take part. This is not a joke at the residents' expense. The more a sales brochure paints the farm as part of daily life, the more it will be received as a shared garden. A commercial farm, though, has harvest logistics, chemicals and materials to handle, and areas people may not walk into. In 2017 residents in Davis raised concerns about the possible use of herbicide near the farm and about the site's lack of organic certification. The expectations manufactured during the sales phase and the operations required during the farming phase do not mesh on their own.

Then in 2024 the food-hub operator running a farm stand at the entrance to the development announced its immediate closure. The reason given was plain: over two years it had not been able to grow its customer base or make the stand profitable, and it would refocus on farm and food events instead. Translate that into the developer's vocabulary and it reads like this. Once 547 homes are sold, the developer's rational interest in the farm is largely spent. Whoever is left operating it must balance the books against a single small market — the residents. Five hundred and forty-seven households is not a large market for a farm stand. The trigger here was neither a planning decision nor land value. It was a profit-and-loss statement after the support had run out.

Agritopia, in Gilbert, Arizona, is a 166-acre planned mixed-use community. It began not as a development plan but as a family farm — land bought by Jim and Virginia Johnston in 1960 and worked for decades in cotton, wheat and hay. In the 1990s, as the Phoenix metropolitan area swelled and the Loop 202 freeway advanced on the property, most neighbours sold to conventional subdivision builders. This family chose to develop the land themselves. After his father's retirement, the son, Joe Johnston, assembled the plan; house construction broke ground in 2003. Today there are roughly 460 homes, a 120-unit senior living building, a commercial quarter of makers and craftspeople, and at the centre an 11–12-acre USDA-certified organic farm.

The route of recovery here differs from the two previous cases. Where Willowsford funds its farm through a fee on every transaction, Agritopia funds it through commerce. Restaurants converted out of the farm's own buildings and a quarter of workshops and shops sit inside the site, and their rent and footfall are what make the farm economically defensible. The farm is at once a supply of ingredients and the warrant for a food-and-retail brand. The vegetable revenue on its own is small; the sentence “from the organic farm on the site” is not, and its effect shows up in what the food tenants can charge and how many people come, which returns to the landowner as rent. Look at the farm only as a production facility and this circuit is invisible.

On the question of what would end the farm, Agritopia differs decisively. There is no lease to expire, no subsidy to be cut, no developer to walk away, because the family still holds the land and the developer, the landowner and the beneficiary of the farm are very nearly the same party. As this database has recorded elsewhere, the single greatest source of fragility in city farms is that they are held on borrowed ground; Agritopia dissolves that fragility through ownership. Ownership, though, imports a trigger of its own: succession. The farm continues as long as the family's intent continues, and if that intent changes there is no clause anywhere protecting it. An institutional guarantee and a personal intention are not strong in the same way.

The same shape appears on the commercial side. When the Funan mall in Singapore reopened in 2019 after being rebuilt, it carried a rooftop farm of more than 5,000 square feet, run by an urban-farming operator, with produce supplied to a food-and-beverage tenant in the same building. The developer's reasoning is easy to read: putting a farm on the roof gives the restaurant a line no competitor has — grown upstairs — and gives visitors an outdoor space they can walk through for nothing. The physical constraints of a roof (load, waterproofing, irrigation) were set out in earlier pieces here; as a business decision they are preconditions, not motives. The motive is being able to say what distinguishes this mall from the others at the same address.

A larger case sits in Paris. As part of the modernisation of the Paris Expo exhibition centre at Porte de Versailles, a 14,000 m² farm was announced for the roof of Pavilion 6 and opened in the spring of 2020, run by private operators including Agripolis, with catering alongside it. The headline — Europe's largest rooftop farm — travelled the world. Read the record a few years on and the area actually in production is upwards of 4,500 m², worked as 696 growing columns and 1,428 cultivation gutters. The roof is 14,000 m²; what is being farmed is about a third of that. This is less an exaggeration than a change of subject: the announced number was a roof, the present number is a farm.

From the property side that contraction is not necessarily a failure. Measured against the goal — differentiating a venue and generating attention — a good share of the return was banked in the opening week's coverage alone. A third of the area can still make business sense. And there lies the trap for anyone tracking city farms by the numbers. When the developer's purpose is publicity, the farm delivers its peak value on opening day, and everything after that is handled as cost optimisation. So long as we keep quoting the opening announcement, we are counting the yield of a communications budget as the yield of agriculture. If a figure is to be quoted, it should be the growing area in operation, with the date it refers to attached.

Japan has almost no equivalent of the American agrihood — a for-sale housing estate built around a farm. One reason is simple: plots are small and few projects generate hundreds of acres of surplus land in a single scheme. What appears instead is a pattern of interim use by railway companies and landholding firms. In front of the west exit of Seijōgakuen-mae station on the Odakyu line, on the deck created when a grade-separation project put the tracks underground, a membership allotment garden of about 5,000 m² and more than 300 plots opened in May 2007. Tools are provided, staff are on site daily, and the clubhouse has a lounge and showers. This is not a farming business; it is a decision to run a service business on a horizontal surface that appeared on prime land beside a station.

In Kitakagaya, in Osaka's Suminoe ward, the landholding company Chishima Tochi applies the same logic at the scale of a district. Once a shipbuilding quarter, the area accumulated vacant lots and empty buildings as factories moved out and the population aged. The company owns roughly 230,000 m² — about a third of the district — and has been letting land and buildings returned to it through the dissolution of leasehold rights at low rents, building a base for younger makers. As a use for vacant lots it launched a creative-farm scheme in 2011, and the community farm Kitakagaya Minna-no-Uen opened the following year; it is now run by a non-profit association based in the district. What the landowner recovers is not rent from a particular plot but the value of a district that people have a reason to come to.

For practitioners the important thing is that this pattern's trigger is explicit. An interim-use farm ends on the day the land becomes usable for the purpose it was held for. A deck over a railway can be swept into a station-area redevelopment; a vacant lot disappears when a building goes up. That is not a flaw but the design. So a proposal should separate two things from the start. First, how many years this farm is a story about. Second, what remains in the participants' hands when it ends — the soil cannot be carried out, but the operating know-how, the relationships among growers and the standing to negotiate for the next site all can be. Naming the term and deciding in advance what to carry away lasts longer, in practice, than calling a time-limited project a permanent community hub.

There are objections that land squarely on the account given so far. The first concerns what the farm is being called cheap relative to: a golf course and a pool. That comparison belongs to large American suburban development, a market in which hundreds of acres are built out at once and shared facilities do the differentiating. There, a farm looks inexpensive. In a Japanese condominium or a small subdivision, the comparison set is a shared lounge and a parcel locker, and against those a farm is not cheap at all. Holding land that earns nothing per square metre in a dense city can be the most expensive amenity on the list. Paste an American case book into a Japanese proposal and it fails at exactly this point.

The second objection is that the party paying and the party recovering keep drifting apart. Line the eight up and the period during which the developer bears the cost is remarkably short. At The Cannery it effectively ends when the last home is sold; at Porte de Versailles most of the return was banked in the opening coverage; at Prairie Crossing the purpose was served the moment a low-density plan was approved. The only parties carrying a long obligation are Willowsford's conservancy, funded by a quarter per cent on every resale, and the family that owns Agritopia. So the phenomenon is not quite “developers build farms”. It is “developers build farms and hand them to somebody”. A scheme with no named recipient runs out of road within a few years of opening.

Third, announced areas need to be doubted. Among the figures in this article, almost none match between announcement and operation: 7.4 acres of farmland with about three under cultivation; a 14,000 m² roof farming upwards of 4,500 m²; the same development described as having more than 300 acres of farmland and more than 200. This is less a matter of anyone lying than of two different measurements sharing a word — the real-estate area (site, roof, protected zone) and the agricultural area (growing beds). Anyone handling numbers in this field would do well to ask, at every citation, which of the two is on offer. As Thursday's instalment showed, the credibility of studies measuring effects on land value turns on the same choice of denominator.

To bring this down to Japanese practice. First, identify the counterparty's route of recovery before anything else: sales pace for a housing scheme, tenant attraction for a commercial one, a substitute for the cost of holding idle land, or the value of a whole district. Different routes imply different areas, different opening hours, different answers on whether outsiders may walk in. Second, name who bears the cost and for how long. A scheme that cannot write the sentence “the developer carries it for three years, and from the fourth year onward it is carried by —” is walking The Cannery's road. Third, distinguish time-limited support from permanent machinery: grants and sponsorships are the former; a transfer fee attached to every resale, as at Willowsford, is the latter.

Fourth, write the trigger for removal in advance. What ends the farm — a lease expiring, construction starting, the operator withdrawing, the land changing hands? Leaving it out looks more amicable and guarantees that the ending, when it comes, is abrupt. Alongside the conditions, settle the extent of restoration required, what happens to soil and equipment, and how much notice growers receive. Fifth, state areas in growing beds. Agree on site area and everyone discovers after opening that paths, storage and buffer strips have eaten more than expected. None of these five is about agriculture; all are about contracts. But what the eight cases here show is that a farm's lifespan is settled by the contract, not by the farming.

Finally, the territory this article did not enter. Every case here involves private capital placing a farm on land that was not farmland. Japan has the opposite problem as well: a great deal of land inside its urbanisation promotion areas that is farmland already, where tax and regulation put the same question to the owner every year — hold it as farmland, or convert it. What became of the productive green land whose thirty-year clock was due to run out in 2022 is where tomorrow's fifth instalment takes up Japan's present position.