D.5Who Holds the Power: Suppliers or Landlords?

Ingredients are the largest share of every plate, and their price rises are shared with diners. Landlords take less, but their rises land on the hawker. The model measures this with four indicators.

Take

share of each sales dollar received.

Squeeze

how much of the hawker's take-home income is lost if that party raises its prices by 10%, after the hawker has raised his own prices as far as he can and diners have responded.

Capture

how much of any help paid to the stall the landlord takes back at the next renewal (our assumption, by type of landlord).

Renewal risk

the typical size of a rent change at renewal (one standard deviation, a usual measure of spread), after the part the hawker passes on to diners, as a share of his income.

The table shows the 2026 baseline. The table-cleaning fee subsidy (D.14) does not change the landlord's take: the Government pays part of the stall's fee to the cleaning contractor, and the landlord is paid as before. At an operator-run centre the landlord's take includes the operator's share, about $182 a stall a month, derived from the official operating surplus (Method, Levers and Data, F.2).

D.5.1Why the two kinds of cost land differently

MTI's 2015 study of hawker prices found that a 1% rise in raw-material costs went with a 0.56% rise in hawker food prices, while a 1% rise in rent went with only 0.03%. Put in dollars, that means each extra dollar of ingredient cost raised the plate's price by about $1.25, and each extra dollar of rent raised it by about 38 cents (ingredients are about 45% of the plate's price; rent is about 8% in an NEA stall). The model uses $1.25, and 35 cents for rent at every kind of landlord, a little below the NEA figure. MTI (Ministry of Trade and Industry) studied NEA centres only (2012–14), so for coffeeshops and food courts the rent figure is our extension of it, and the 5,000 simulated stalls vary it from 12 to 45 cents. MTI followed the same stalls from July 2012 to April 2014 and found the same: about two in three stalls that raised prices had no change in rent; subsidised stalls, paying about $200 a month against $1,250, charged broadly the same prices as the rest, so hawkers appear to charge what the market will bear; stalls in the Central Business District passed on more of their costs; and each extra stall selling the same dish went with prices about 8.4% lower (official, MTI). In a tendered NEA stall, a 10% rise in ingredient prices raises the plate by $0.25. Diners pay all of the extra cost and a little more, yet the hawker still loses 2.6% of income, because some customers stop coming; a 30% rise in rent and fees at renewal raises the plate by $0.04, and the hawker loses 10.4%.

The landlord's squeeze is larger than the supplier's in five of the six kinds of stall: tendered NEA stalls, operator-run centres, HDB-rented coffeeshops, private coffeeshops and mall food courts. Only the subsidised NEA stall, whose rent is a few hundred dollars, is more exposed to its suppliers. One caution: an ingredient rise usually hits every stall at once, so all can raise prices together, while a rent rise hits one stall, which cannot. Comparing equal 10% rises understates the landlord's advantage, if anything.

D.5.2Suppliers: large, fragmented and unmeasured

More than 90% of Singapore's food is imported. Between the farm and the wok sit importers, wholesalers at Pasir Panjang and Jurong Fishery Port, distributors and, increasingly, central kitchens selling pre-made pastes, stocks and par-cooked items. The import price index for food stood at 102.4 in August 2026, down 2.6% on a year earlier, while meat rose 5.4%. No public source shows the margins at each step or how much of a stall's cooking is now bought in. The split in Figure D.1 is therefore an assumption, labelled as such. The policy point survives the uncertainty: hawkers buy alone, in small lots, from many sellers, and have no collective bargaining power. The Buying Circles are the plan's answer (D.16): through their own association, small groups of hawkers at each centre post their pooled weekly volume on an open board, wholesalers and distributors offer terms, and the circle chooses. The Seed Bank never buys, stocks, sets prices or picks suppliers. The model credits the circles with 3% off ingredients, an assumption no Singapore data yet tests; some hawker-centre operators already offer bulk buying, and few stallholders use it (MSE, 10 Jan 2024), so their offers simply join the board. The saving is not taken out of anyone's profit. It comes from a lower cost of serving many small buyers: one pooled weekly drop instead of a dozen, predictable volume and prompt payment. The supplier and the cook share it. At the model's take-up (the share of those eligible who join) of one stall in four, circle members would pay suppliers about $11.8M a year less (D.7.4), about 0.75% of the $1,577.3M the trade spends on ingredients, with more suppliers competing for the orders, not fewer.

D.5.3The Middlemen: Who Takes a Cut Between the Farm and the Plate

Before a plate of kway chap reaches the counter, the pork has passed through four or five pairs of hands: a farm across the Causeway or further, an importer, a processor, a distributor whose van pulls up at five in the morning and, often, a central kitchen that sells the braising sauce ready-made. Each takes a cut. Many readers have asked whether those cuts are why the plate costs more than it did. I went looking. The short answer is no. The long answer matters for what the plan should do.

Start with the rise itself. Hawker food prices rose about 23% from 2019 to the first nine months of 2026 (derived from SingStat's indices, across the 2024 change of base year). Most of it came in two years: 5.7% in 2022 and 6.1% in 2023, the steepest rise since 2008 (official, SingStat).

The test. If the middle of the chain were taking a bigger slice, the price of food in Singapore's shops would have pulled away from the price of food at the border. It did not. From 2019 to 2025, food import prices rose 20.9% and retail food prices 18.5% (derived from SingStat). The Singapore end of the chain charged roughly the same percentage on a bigger bill. A second test points the same way. If coffeeshop and food-court operators were marking up harder, their prices would have risen faster than those at hawker centres. From 2019 to 2023 they rose 15.1%, against 15.8% (derived from SingStat's yearly rates, which ran almost in step: 1.2, 1.5, 5.6 and 6.1% against 1.6, 1.4, 6.0 and 6.0%). A third test uses the operators' own books. Across food and beverage businesses, hawkers excluded, rent fell from 17.0% of revenue in 2019 to 10.8% in 2024, and the operating surplus from 3.2% to 2.3% (derived from DOS figures given to Parliament, 7 April 2026). The landlords behind the operators did not take more, and the operators earned less.

The level is another matter, and IPS (Institute of Policy Studies) measured it. In late 2022 the same item cost from 4% to 41% more at a food court than at a hawker centre, a median of about a quarter across IPS's 18 items (derived, Makan Index 2.0, Table 3.1). Among the dishes: chicken rice $4.64 against $3.72 (+25%), wanton noodles $4.90 against $3.72 (+32%), fishball noodles $4.51 against $3.46 (+30%), economic rice $4.17 against $3.36 (+24%), nasi lemak $4.49 against $3.29 (+36%), mee rebus $4.02 against $3.41 (+18%) and kopi-O $1.26 against $0.99 (+27%). Kopitiams sat close to hawker centres on dishes (chicken rice $3.68, economic rice $3.32, fishball noodles $3.71) but 11–14% dearer on drinks (kopi-O $1.13), and the drinks stall is often the operator's own. Granted, the premium is not all margin: a food court carries air-conditioning, mall rents and turnover rents, and IPS took menu prices at face value, not adjusted for portion or quality, as it says itself. Still, it is the cleanest measured price of a layer we have, and it sits in the level of prices, not in their rise.

A tendered NEA stall and its $4.50 plate, except the operator rows ($4.40 and $5.20 plates). Official: a published figure. Derived: our arithmetic on official figures. Assumption: not grounded in Singapore data. The Singapore-side share of the ingredient bill, 35%, is the model's assumption (Method, Levers and Data, F.2), the middle of a 25–45% range; only its change over time can be tested, and the test above finds none. The GST row assumes that 5–8% of the plate's price is input value that bore GST the hawker cannot reclaim. Delivery: GrabFood's published service fee is 30% of an order, plus GST; foodpanda does not publish its rate.

What the middle takes. It is not small. Ingredients take about 45¢ of every dollar a diner pays (D.3), and if Singapore's importers, wholesalers, distributors and central kitchens take a third of that bill, as the model assumes, they take about 16¢ of each dollar. No one publishes margins by stage, so that is a range and not a measurement. Those cents buy real things in a city that imports about 90% of its food: consolidation from hundreds of origins, cold chains, trade credit to stalls too small for a bank to bother with, food-safety compliance, and the national rice stockpile, which importers must hold in government-designated warehouses (Enterprise Singapore). When floods in Malaysia sent wholesale vegetable prices up by 30–80% in December 2019, wet-market sellers held their prices and took the loss (TODAY). That is a service too.

What explains the 23-point rise. About 9 percentage points come from dearer imported ingredients (derived). This agrees with MTI's 2015 finding that a 1% rise in raw-material costs goes with a 0.56% rise in hawker prices; MTI measured those costs with import price indices, so the 0.56 already includes whatever Singapore's wholesalers add. Assistants' wages explain 2–5 points, on an assumed wage path before 2026. Stalls run by their owners, hawkers' included, are exempt from the Food Services Progressive Wage Model, which is enforced through the rules for hiring foreign workers on work passes (MOM, March 2026), and a hawker may not hire work-pass holders at all; since January 2025 he may hire long-term visit pass holders (MSE). So the Progressive Wage sets the pay he has to match to keep an assistant; it is not a minimum he is bound to pay. Table cleaning, utilities and the rise in GST explain about a point each (utilities held at about 7% of a stall's costs in 2022, 2023 and 2024, by NEA's surveys). Wider margins in the middle of the chain explain, as far as the data can tell, between minus one and plus one point: about none of it. The residual, about 5 points, is the hawker's own margin and whatever the accounting misses.

Where the tolls are real. Three deserve the name. The first is delivery. On a delivered $4.50 plate GrabFood's 30% fee is $1.35. Granted, most of it pays the rider. Still, a hawker cannot bargain alone, and since Deliveroo went offline after 4 March 2026 there are two large platforms, not three; analysts expect fees to rise, though CCCS (Competition and Consumer Commission of Singapore) had seen no systematic shift in commissions by April (MTI, 7 Apr 2026). The second is the tied supplier. Some operator-run centres require stallholders to buy gas or crockery from a supplier the operator chose (D.16), and MSE told Parliament in September 2025 that one operator would look for cheaper gas when its contract ends. A captive buyer is the one place where a middleman can charge what he likes. The third is concentration. In 2018 CCCS fined 13 fresh-chicken distributors, who together supplied more than 90% of the market, about $26.9M for coordinating the amount and timing of their price rises between 2007 and 2014; hawker stalls were among their customers. The law worked, but only after seven years.

The operators. Between the State and the hawker sits another layer. At an operator-run centre the average operating surplus over 2019–2023 was about $100,000 a centre a year, and at least half must be spent on the centre or its stallholders (MSE, 15 Oct 2024). At about 48 stalls a centre that is about $182 a stall a month (derived; the stall count is assumed), a few per cent of what stallholders pay, not a fortune; the model now books it so (Method, Levers and Data, F.2). At HDB-rented coffeeshops the margin is larger and unmeasured. Kimly's outlet segment earned about $1,820 a stall a month before tax from all sources, drinks and cleaning included (derived from its FY2025 report): an upper bound. HDB found rents unchanged at 90% of its rental coffeeshops over three years, said it was concerned that some operators impose "high mark-ups" on stallholders, and said in January 2026 that it would begin collecting stall rents (CNA). HDB rents out 374 coffeeshops, 45% of them to chains (MND, May 2024); the other 402 were sold before HDB stopped selling them in 1998 (HDB, July 2024). At those, the landlord is a private owner and nothing is published. From January 2026 budget meals are optional at HDB-rented coffeeshops, which keep a 5% rent discount for three years if they offer them (HDB). KF Seetoh, Makansutra's founder, pointed in March 2023 to two coffeeshops sold for more than $40M in June 2022, one in Tampines for $41.68M, and said stall rents there had reached about $12,000 a month; at the Yishun shop at least 10 of 14 stalls left (The Online Citizen, citing 8 World News). That is his view and a press report, not a measured series. He also called for an end to highest-bid tendering for public stalls, which the bands and the ballot answer at the pilot centres (D.14). As D.5.1 shows, that layer squeezes the hawker's income far more than the diner's price.

The taxes that ride along. A hawker selling $200,000–320,000 a year (D.3) is far below the $1 million GST threshold (IRAS). He charges no GST and claims none back, so the 9% inside his inputs stays with him: about 17 cents on a $4.50 plate (derived, on an assumption about how much of the ingredient bill bore GST), and perhaps 4–5 cents more, because NEA quotes rent and conservancy (the monthly charge for cleaning and upkeep of a centre) before GST (NEA tender notice, August 2026). I will not call this unfair. Were the same hawker registered, he would owe 9% on the whole plate, and the diner would pay about 18 cents more, net of his input credits (derived). The threshold is a benefit, not a cascade. The carbon tax has risen from $5 to $45 a tonne since 2019 (NCCS), about a cent a plate. The cleaners' Progressive Wage has risen from $1,312 in July 2022 to $2,080 in July 2026 (MOM), and the table-cleaning fee pays for it. Together with GST these explain about an eighth of the rise. All three are deliberate public policy, and the plan opposes none of them.

What I will not claim. I will not say that middlemen are profiteering, or that they drove hawker inflation; the evidence says otherwise. I agree with those who say the hawker is squeezed. I disagree that the squeeze comes mainly from the chain behind him. It comes from the rent in front of him and the platform beside him. Rents, wages and the price of pork at the border will keep moving. The point of the Plate Ledger is that next time someone asks who took the cut, we can answer with a number rather than a suspicion.

D.5.4The Plate Over Time: How a Stall's Costs Changed, 2013–2026

Ask a hawker why the plate costs more than it did, and he will tell you that his costs went up first. A sharper claim travels with it: that government, or bodies close to it, have slipped new layers between the State and the wok, each taking a cut. D.5.3 asked who takes a cut along the supply chain. This section asks how a stall's own costs moved over thirteen years, and whether those layers drove them. Mostly, they did not. But part of the claim is true, and the plan should say which part.

Two periods, two stories. Below, "points" means percentage points. From 2013 to 2019 hawker food prices rose about 12.5%. The official rates were 1.9% in 2013 and 3.0% in 2014 (DOS, in MTI's reply to Parliament, 13 February 2015); 2015–19 averaged about 1.8% a year (derived from SingStat's 2012–22 average of 2.2%). A second measure agrees: the prices implied by SingStat's sales index for cafés, food courts and other eating places sit within a point of the hawker-food index in every year from 2012 to 2025 (derived).

What costs did. Over the same years food import prices fell about 3% from 2014 (derived, linked across SingStat's changes of base year), electricity tariffs fell from 26.3 to 23.6 cents a kWh (official, EMA), and the rent of a non-subsidised NEA stall held at about $1,250: its average in 2012–13 (official, MTI) and its median from 2015 to 2024 (official, MSE and NEA).

What that explains. On MTI's 2012–13 cost shares, measured costs explain only about 3 of the 12.5 points (derived; anywhere from none to 8, on an assumed path for wages). That is a weak result, and I present it as one.

Where the rest may lie. The rest sits where the data cannot see: assistants' wages, which no series we could read covers before 2023; coffeeshop stall rents, which HDB said only in January 2026 that it would begin collecting; and margins. Coffee shops and food courts carry 80% of the weight in the hawker-food price index (official, SingStat), so their rents and margins land in that unexplained part.

My reading. The timing is telling. Rents in the food and beverage trade outside hawker centres rose 7.4% a year in 2010–14 and fell 2.6% a year in 2015–19 (DOS, in MTI's reply to Parliament, 1 August 2022), and the hawker-food index rose fastest in 2014, at 3.0%. After that, food import prices, electricity and those rents all fell, yet hawker prices kept rising by about 1.8% a year. Hawker prices, it seems, do not come down when costs do; the stall keeps the difference until the next shock. That is my reading, not a measurement.

From 2019 the story is simpler. To the first nine months of 2026 prices rose about 23% (derived), and this time the costs moved with them: from 2019 to 2025 food import prices rose 20.9%, shop prices for food 18.5% and hawker food 20.7% (derived from SingStat). SingStat put 2023's rise of 6.1%, the steepest since 2008, down to dearer raw ingredients after COVID-19 broke supply chains and the war in Ukraine (official). D.5.3 takes the 23 points apart; in short, food and wages explain most of them, and rent none. Figure D.6 shows both periods.

The shape of the stall barely moved. In 2012–13 ingredients took 59% of a stall's operating costs, manpower 17%, rent 12% and utilities 9.3% (official, MTI). In 2022 the shares were 56%, 20%, under 10% and about 7% (official, MSE, reporting NEA's survey), and NEA's surveys put utilities at about 7% again in 2023 and 2024 (official, MSE, 4 Aug 2026). Rent's share for a non-subsidised stall slid from 14% in 2012–13 to 12.5% in 2019, 11.6% in 2020 and under 10% in 2022 (official, MTI and MSE). The model's tendered stall in 2026 is 54%, 20%, 9% and 7%, with 3% to delivery and payments (modelled, D.3). Delivery is the only line that did not exist a decade ago.

Testing the claim, layer by layer. Granted, new parties did arrive, four of them by the State's doing.

• Operator-run centres. The first, Ci Yuan, opened on 20 September 2015 with a social enterprise as "managing agent and master tenant" (official, NEA). Sixteen of the 123 centres are now run this way (official, NEA, 17 Aug 2026). At some, stallholders fund a 10% discount for diners who pay through the operator's app, by their own account about $1,000 a month (The Straits Times, 22 Aug 2025; not an official figure), and value meals make up 5–20% of meals sold (NEA, as reported).
• Centralised dishwashing, compulsory at every new centre since January 2019 and running at 16 centres by October 2022, with NEA co-funding the fee in the early years (official, NEA and MSE). No fee series is published; one hawker's account puts it near $650 a month (unverified).
• Outsourced cleaning under the Progressive Wage Model. A cleaner at a hawker centre earned a basic $1,312 a month in July 2022 and earns $2,080 from July 2026 (official, MOM).
• Tray return, enforced from 1 September 2021 at hawker centres and from late 2021 at coffee shops and food courts (official, NEA), which moved work to diners rather than adding a fee.

A fifth, the delivery app, is private, and D.5.3 deals with it. Method, Levers and Data (F.4) tabulates every layer since 1966.

Yet the layers that matter most moved little. NEA, the tenancy landlord at 107 centres, kept the rent of a non-subsidised stall at about $1,250 from 2012–13 to 2024, rising 0.3% a year over 2015–20 (official, MTI and NEA).

Still, the average stall's rent did rise, for a quieter reason: fewer stalls pay the subsidised rent each year. In March 2013, 53% of cooked-food hawkers paid subsidised rents, then $160–320 a month (official, Dr Vivian Balakrishnan, Committee of Supply, 12 Mar 2013); about 40% did in 2020, and about 30% in 2023 (official, MSE).

And about 2,000 hawkers at 15 centres, who had bought their stalls on 20-year leases under the Stall Ownership Scheme of 1994–97, saw those leases run out in 2014–17. Grace Fu set out their options in the same debate: those formerly on subsidised rent could return to it, any owner could hand the stall back for an ex-gratia payment (one made as a gesture, not because the law requires it), and the rest carried on at market rent (official). About 600 of the stalls sold were cooked-food stalls (official, MTI, 2015). No landlord raised a rent to do it. The pioneers simply retired.

At the 27 centres it manages itself, NEA kept the conservancy charge for a cooked-food stall at $140 from 2005 until 2024, then raised it by a quarter, to $155 in January 2024 and $175 in July, saying that cleaning and maintenance costs had "increased significantly" (official, NEA). At the 80 Town Council centres the Town Council sets the charge, and the amounts vary (NEA tender notice, August 2026). The food stall licence was $32 a year when last given to Parliament (official, MSE, 2021).

At four NEA-managed centres the table-cleaning fee was the same, before GST, in December 2022 and August 2024 (derived from NEA's tender notices), although the cleaner's Progressive Wage rose by a third over those two years, from $1,312 to $1,740 (official, MOM). And at the HDB hawker centres that Town Councils run, NEA's notices say the table-cleaning service "is managed by the respective Hawkers Association" on the stallholders' behalf (official, NEA, Dec 2022 and Aug 2024). That is the hawkers' own body, not a new government agent.

One caution on rent. The three NEA measures are separate series and must never be drawn as one line: the system-wide median, about $1,250 for 2015–24 (and the 2012–13 average, official, MTI); the median at centres comparable with the operator-run ones, which fell from $1,700 in 2019 to $1,450 in 2025; and the median winning bid, about $2,000 in 2022 and $1,800 in 2023 (all official, MSE). NEA put the median at about $1,250 again in November 2024, "roughly constant for 10 years" (official, NEA, 4 Nov 2024); none has been published since, so Figure D.6 stops there.

Where a layer is a margin, it is small and capped: operators may not profit from pass-through charges, and their surplus comes to about $182 a stall a month (D.5.3). The costs the State did push up, the cleaners' Progressive Wage, GST and the carbon tax, are deliberate policy, and the plan opposes none of them.

So the verdict is narrow. The broad claim, that government layers drove up hawker costs across the trade, does not survive the data. The narrow one does. In 2025 a stall at an operator-run centre paid a median of about $2,750 a month for rent, conservancy and table cleaning, against $2,170 at comparable centres (derived from MSE's medians, which MSE warns are "not directly comparable", since stalls there are generally larger). That is about $580 a month more, and the model's operator-run stall keeps 9¢ of each dollar against 18¢ at a tendered NEA stall (modelled, D.3). And a hawker who sells through an app hands over nearly a third of every delivered plate. Neither shows up much in the price index. Both show up in the hawker's purse, because a stall that raises its price alone loses its custom to the stall next door (D.6).

The other half of the purse. Costs are only half the story; a stall lives on price times plates. The price is well measured. The plates are not, and no one counts them at hawker centres. The nearest thing is SingStat's sales index for cafés, food courts and other eating places, which takes in coffee shops but not hawker stalls, and counts the operators rather than the stalls. After removing price rises, their sales in 2012 were 2.8% above 2019's, and in 2025 1.7% below. Over those thirteen years prices rose about 36% and the number of plates sold did not grow at all (derived from SingStat). This year they are falling: sales at food courts and other eating places were 1.5% to 5.3% lower in money terms from February to May 2026 than a year earlier, while restaurants grew and cafés held roughly level (official, SingStat). Restaurants took the share: 47% of the index's weight on its 2025 base against 41% on its 2017 base (official), and 42% of what households spend on eating out in 2023 against 35% in 2012/13 (derived, HES).

Put the two halves together. If a stall's plates moved like the average coffee shop's, its takings rose about 31% from 2015 to 2025 (derived). The median full-time worker's pay rose 46% (official, MOM). On the costs I can measure, the margin on each plate did not shrink: unit costs rose about 23–34% from 2013 to 2025 against prices up 36% (derived, on assumed wage paths). So the squeeze is not in the price of the plate. It is in the number of plates, and in what the same day's work earns against a wage job that keeps climbing. Figure D.6b sets the two against each other.

What diners spent. Households spent $435 a month at hawker centres, food courts and coffee shops in 2012/13, $437 in 2017/18 and $491 in 2023 (official, SingStat). In 2019 dollars that is a fall of about 13–14% (derived, on the official rates to 2014). The IPS Makan Index tells the same story from the diner's side. A kopi-O cost about 97 cents in 2017 (derived from IPS's area averages) and $1.09 in late 2022 (official, IPS): about 12% more, in line with the price index. Treat that as indicative only: the 2.0 report never refers to the 2017 edition, and no like-for-like comparison exists. SingStat's own average prices agree: chicken rice went from $3.13 in 2014 to $4.24 in December 2024, fishball noodles from $3.11 to $4.36 and kopi-O from 94 cents to $1.25 (official, SingStat, as reported; 2014 at each item's lowest month), about a third more against a price index up about 30% (derived). Its own short test was a revisit of 263 stalls in early 2023, and most had not raised prices; for most food items the average rise was no more than 10 cents, the largest iced Milo's 7% and the breakfast set's 6.8% (official, IPS). Three meals a day at hawker centres, kopitiams and food courts cost $16.89 (official, IPS Makan Index 2.0, prices of late 2022, published 2023).

To be sure, thirteen years of data with holes in it is not a verdict for all time. The holes are named in Method, Levers and Data, F.2, and SingStat's own tables would close most of them. Until then the fair reading is this: the plate got dearer mainly because food and wages did; the State as landlord held its rent still, and raised its conservancy charge by a quarter, citing cleaning costs. The new tolls are real, but they sit at the edges, and the plan can reach both. What the plan cannot reach by trimming tolls is the plates: a stall that sells no more than it did in 2012 falls behind the wage it could earn elsewhere, however fair its costs.