F.2Limits, Assumptions and How This Enters the Plan

F.2.1What the model cannot tell us

It is static. It shows a year's effect on a stall that stays open. It does not model entry, exit, or how many young cooks a better-paid trade would attract.
How diners respond to hawker prices is barely measured in Singapore. The model assumes that a 10% price rise loses 5% of custom (an elasticity, the measure of that sensitivity, of −0.5), and tests anything from −0.3 to −1.0. Studies of restaurant meals abroad find −1.2 and below; hawker meals are assumed less sensitive, because they are the cheapest cooked option. The one Singapore estimate we found is early and only a guide: a survey of 90 diners, not chosen at random, asking what they say they would do (Poh, 2026). Sensitivity is highest among the young (2.81 at 18–24) and the old (3.92 at 60 and over, from only nine people), and lowest in the middle (1.01 at 45–60). Diners on small meal budgets are more sensitive too (2.09, against 1.1–1.2). If it holds, diners are more sensitive than the model assumes, which would make price freedom (Follow the Plate, D.15) costlier for the hawker. That is a reason to measure it properly, not yet to change the parameter.
Supplier margins are unpublished. The split of ingredient spending along the supply chain is an assumption, not data: 35% of the bill to Singapore-side intermediaries, the middle of a 25–45% range. Only its change can be tested, and from 2019 to 2025 it did not widen (Follow the Plate, D.5.3).
The coffeeshop and food-court archetypes (a typical stall of one type) rest on listings and company reports, not official data. The food-court turnover rent is the weakest parameter. Operator-run centres are the most stylised archetype.
Stall counts by type are estimates. No official count of coffeeshop cooked-food stalls exists. The model's 5,800 sits in the middle of a range of about 5,600–7,200, and SFA's 13,893 licensed food stalls outside the centres NEA (National Environment Agency, which runs the public hawker centres) runs cap the total (Follow the Plate, D.3).
How much of a cost rise stalls pass on to diners comes from one study of NEA centres in 2012–14. We apply it to every kind of landlord and vary it widely.
The share of stallholders under 55 is not published. The retirement match applies only to them; from 55 the existing MRSS (Matched Retirement Savings Scheme, which matches top-ups from 55) matches top-ups for those who pass its income and home tests, and the plan helps hawkers claim it. Half of all hawkers are 60 or older (MSE, Nov 2024), so fewer than half are under 55; the model counts 50%, an upper bound (the most it could be), so the match's costs are ceilings (Follow the Plate, D.7.5).
Hawker income is measured once, and not by type of stall. MOM's figure for 2021 is one median (the middle figure: half are above it, half below), for full-time hawkers who employ no one, read from a chart; the model's estimates are set beside it in Follow the Plate, D.4, and the reconciliation is ours.
A closed stall's cost assumes the regulars come back. The model keeps the assistant on and lets the regulars return after an illness; if they drift away, the loss is larger than it shows.
CPF (Central Provident Fund, Singapore's compulsory savings for retirement, health and housing) does not publish how Workfare (the Government's top-up for low-wage workers) builds up and tails off as income changes. The model assumes it rises and falls in a straight line. That matters because many hawkers earn close to the $3,000 income limit.

F.2 Limits, Assumptions and How This Enters the Plan: table 1
Parameter	Basis
Plate price, plates a day	Calibrated so the tendered NEA stall reproduces NEA's 2022 cost shares; coffeeshop and food-court prices sit above hawker centres, as in IPS's 2023 price survey.
NEA rent	The typical tendered stall pays about $1,250 a month, the median non-subsidised cooked-food rent across NEA's centres, stable for about a decade (MSE, 18 Sep 2023 and 5 Feb 2025; NEA, 4 Nov 2024), and the median assessed market rent given to Parliament. $1,450 is the 2025 median at the NEA centres comparable with the operator-run ones (MSE, 8 Sep 2026): the model uses it only where the text sets NEA centres beside operator-run ones, and as a quoted rent in the rent test (D.4.1).
Centre fees	NEA and operator-run centres: 2025 median conservancy and table-cleaning fees at comparable centres (MSE, 8 Sep 2026). At NEA centres 30% of the fees is treated as landlord income and the rest as services: an assumption, probably high, since MSE says the fees fund 'a significant part' of a centre's running costs (5 Feb 2024); tested at 0-30% with no change to the findings. At operator-run centres the table-cleaning fee is a pass-through of the contractor's price, which operators may not profit from (NEA, 21 Jul 2026), so only the conservancy charge carries the 30%. Coffeeshop O&M fees of $1,000-1,200: commercial sources only; 60% treated as landlord income, the rest as services. Assumption.
Food-court turnover rent	25% of sales; the only data point is about 20% (2014). The weakest parameter. Assumption.
Manpower	$2,920 a month: one assistant at the $2,220 PWM benchmark plus 17% employer CPF ($2,597), and about $320 of part-time help. Matches NEA's 20% manpower share. The PWM does not bind hawker stalls, so actual pay may be lower.
Delivery	Commission of 30% of a delivered order: GrabFood's published service fee, before GST (Grab, Oct 2026); foodpanda does not publish its rate. Varied 25-35% in the Monte Carlo. Deliveroo went offline in Singapore after 4 March 2026. Grab said in 2020 that it kept about 5% of an order as its own fee, the rest paying riders (Vulcan Post, 17 Apr 2020). Delivery is 8% of sales: an assumption, since no share is published for hawker stalls.
Pass-through	MTI 2015: a 1% rise in raw-material cost goes with a 0.56% price rise, and 1% rent with 0.03%. To turn these into dollars: ingredients are 44.5% of the plate's price, so 0.56 / 0.445 = 1.25, and each extra dollar of ingredient cost raises the plate by $1.25. Rent is 8% at an NEA stall, so 0.03 / 0.08 = 0.38, about 38 cents for each extra dollar of rent; the model rounds this down to 0.35 and applies it to every kind of landlord. MTI sampled NEA centres only (2012-14), so the coffeeshop and food-court rates are our extension of it.
Landlord capture	Share of any help paid to a stall that its landlord takes back at the next renewal: 0% subsidised, 5% tendered, 15% operator-run, 35% HDB-rented coffeeshop, 85% private coffeeshop, 95% food court. assumption by type of landlord.
Unwanted closures	10 days a year lost to illness or exhaustion; varied 4-16. Assumption.
Demand elasticity	One early Singapore estimate (Poh, 2026: a survey of 90 diners, asking what they say they would do, not a random sample) finds diners more sensitive, 1.0-3.9 by age; not yet used. Restaurant studies abroad find -1.2 and below; hawker meals assumed less sensitive at -0.5, varied -0.3 to -1.0. Assumption.
Workfare	Work Year 2025: the self-employed maximum is $3,267 at age 60 and over, for incomes of $500-3,000 a month, and MediSave must be paid; 10% is paid in cash and 90% into MediSave. CPF does not publish how Workfare builds up and tails off with income, so the model assumes straight lines. One hawker in twenty is assumed to be behind on MediSave, with or without the Protection Floor. A hawker cannot renew the SFA stall licence while behind on MediSave unless he is on a GIRO plan, an automatic monthly bank deduction (SFA; CPF, 4 Nov 2025), so the one in four of all self-employed persons who are behind (MOM, Feb 2025) does not describe hawkers. The Protection Floor therefore adds no Workfare in the model.
Supply-chain split	Assumption, not data. No published margins for importers, wholesalers, distributors or central kitchens (SingStat publishes no margins by food sub-industry). The model puts 65% of the ingredient bill at the border (farm or import value and freight) and 35% with the Singapore-side intermediaries, the central case of a 25-45% range. Only its change can be tested: from 2019 to 2025 retail food prices rose 18.5% and food import prices 20.9% (SingStat; our arithmetic), so the middle's share did not widen (D.5.3). A Plate Ledger indicator (D.8).
Landlord split	Head landlord vs operator shares of coffeeshop rents: assumption; HDB said in January 2026 that it would begin collecting stall-rent data. Operator-run centres: the operator keeps about 10% of rent and landlord fees, derived from MSE's average operating surplus of about $100,000 a centre a year (2019-23, 15 Oct 2024) at an assumed 48 stalls; at least half of the surplus must be reinvested in the centre or its stallholders, so 10% is an upper bound.
Coffeeshop rents	HDB-rented $3,400 and private $6,000 a month: assumptions. No official stall-rent data exist; a commercial guide puts HDB coffeeshop stall rents at $4,000-10,000 (Home & Decor, 1 Apr 2026), so the base is conservative. HDB rents 374 coffeeshops and 402 are privately owned (Enterprise Singapore, MND and HDB, 1 Jul 2024): at about 7.5 cooked-food stalls each, about 2,800 and 3,000 stalls.
Relief cooks	A Corps cook is an employee, so the Employment Act limits his working day to 12 hours. In a week of more than five days, as a stall's is, hours beyond the normal 8 a day are overtime, paid at 1.5 times the hourly rate (MOM). He is paid the Progressive Wage rate for a Cook (QS), $12.95 an hour from July 2026: 8 hours at that rate and 4 at 1.5 times, equal to 14 hours' pay, plus the employer's 17% CPF, about $212 a day in 2026, rising with the PWM path ($13.69 from July 2027, $14.42 from July 2028). Where the stall's day is longer than 12 hours (13 in the typical case), the stall's own assistant does the first hour's prep, as on any day. The hawker co-pays $80 a day.
Cost shocks (D.6)	Wages: every wage the stall pays follows the PWM path, +12.6% from July 2026 to July 2028, employer CPF included; 30% passed on (assumption). Energy: electricity rose from 27.27 to 31.91 cents a kWh in Q3 2026 (SP Group) and town gas from 21.92 to 23.48 cents (City Energy, as reported) after the Middle East conflict; in Q4 they fell to 28.59 and 21.45 cents; SP Group has not said where they go next, and one analyst expects a single-digit rise in Q1 2027 (Rystad Energy, as reported, 30 Sep 2026). Utilities are 60% electricity and 40% gas (assumption). A rise hits every stall at once, so we assume stalls pass half of it on to diners: MTI 2015's gas-price effect (0.0486) divided by a utilities share of about 9.3%. A fall is assumed not to lower prices (assumption).
Closed-stall cost	A day shut for illness loses a day's sales; rent, centre fees, the assistant, utilities and other running costs continue, as for every unwanted closure in the model. The assistant is kept on and the regulars are assumed to come back; if regulars drift away the true loss is larger, and the model does not count it. NEA may waive the rent of a stall closed for medical reasons (MSE, 10 Mar 2025); a Corps cook on medical leave costs the hawker $40 a day (the Plan, §2.7). Derived.
Newcomer's rent path	NEA's renewal rule from November 2024, for new tenancies: the winning bid for three years, halfway to the assessed market rent at the first renewal for the next three, the assessed market rent from the seventh year (The Online Citizen, 15 Sep 2026; NEA, Nov 2024). NEA does not publish a stall's valuation, so the $1,450 and $1,250 set against the $1,800 bid are illustrative.
Hawker income check	MOM's labour force survey put the median gross monthly income of full-time resident hawkers and stall holders who employ no one at about $1,470 in 2021 (in the Advisory Committee on Platform Workers' report, 2022, Fig. 1.1), read from a bar chart, so derived. For the self-employed it is profit before tax, as the model's net is. We converted it to 2026 money using SingStat's hawker-food price index (102.8 in 2021 and 122.8 in January-September 2026, where 2019 = 100) and compared it with the model's simulated stalls of every type, counting each type by its number of stalls (D.4).
Retirement match under 55	The Protection Floor's $1,200 match applies only to stallholders under 55; from 55 the existing Matched Retirement Savings Scheme already matches cash top-ups up to $2,000 a year for those who pass its income and home tests, and the plan helps hawkers claim it. The share of stallholders under 55 is not published; the median hawker is 60 (MSE, Nov 2024), so fewer than half are under 55. The model counts half, an upper bound, so the match's cost is a ceiling. Hospital-cash cover is not scaled.

F.2.2Ranges in the Monte Carlo

Plates a day: lognormal centred on the base, σ 15%. Plate price: lognormal, σ 6%. Rent: lognormal centred on the sourced median (σ 12% tendered NEA, 10% operator-run, 20% HDB coffeeshop, 30% private coffeeshop), correlated 0.5 with plates a day; coffeeshop fees σ 25%; food-court turnover rent 20–30%. Ingredient share of sales 40–49%. Manpower and utilities 85–115% of base. Hours 11–15 a day; days 24–28 a month. Delivery 0–16% of sales at 25–35% commission. Unwanted closures 4–16 days a year. Elasticity (how many diners are lost when prices rise) −1.0 to −0.3. Pass-through: ingredients 1.05–1.45, rent 0.12–0.45. Landlord capture ±10 points. Triangular unless stated; 5,000 draws per stall type; fixed seed. An independent check of the model's code and parameters prompted the pass-through (a cost passed on to someone else, in whole or in part) calibration, the split between cash and savings, the voucher accounting and the centred distributions used here.

F.2.3How this enters the plan

F.2 Limits, Assumptions and How This Enters the Plan: table 2
Plan section	Change made or proposed
the Plan, §1.1	Puts a modelled estimate behind the "2nd decile" income cell: most hawkers earn less per hour than the benchmark for their assistant (Follow the Plate, D.4). Sets the one official income figure, about $1,470 a month in 2021 for full-time own-account hawkers, beside the model's $3,169 for the typical tendered stall in 2026, and reconciles them (Follow the Plate, D.4).
the Plan, §2.1, Follow the Plate, D.14	Treats the three NEA rents ($1,250, the median across NEA's centres and the typical tendered stall; $1,450 at the centres comparable with the operator-run ones, 2025; $1,800 bid, 2023) as a range: without the package the typical hawker falls below the assistant's benchmark above about $610 of rent, with it above about $2,450, or $2,150 without the named deputy, and the stall loses money only above about $4,420 (Follow the Plate, D.4.1).
Follow the Plate, D.14	Shows the newcomer's rent path under NEA's November 2024 renewal rule: a $1,800 bid against a $1,450 valuation pays about $18,900 above the valuation over six years before it reaches it (illustrative, Follow the Plate, D.4.1).
Hawkers and Social Protection, C.10, the Plan, §2.7	Prices a closed stall: each day the typical tendered stall is shut for illness costs its hawker about $372, because rent, fees and the assistant continue; six weeks cost $13,400, 35% of the year's net. A rent waiver saves about $48 a day of it; a Corps cook on medical leave turns it into the $40 co-pay (modelled, F.2).
Follow the Plate, D.15	Adds the voucher-cover test: at one stall in five, a $120 voucher covers an eligible household's extra cost two to five times over (Follow the Plate, D.7.3).
Follow the Plate, D.14	The table-cleaning fee subsidy: $200 a stall a month paid by the Government at the pilot centres, $0.98M a year at three, an upper bound tested by re-tendering the cleaning contracts with whoever holds them, the hawkers' association at a Town Council centre (Follow the Plate, D.2.2, D.5.4, D.7.4).
F.6	Adds the Plate Ledger and its twenty-one indicators, among them the margin spread and the floor cost by type of centre (Follow the Plate, D.5.4, D.8).
Follow the Plate, D.16	The Buying Circles' 3% is a lower cost of serving many small buyers, shared between supplier and cook, not a transfer from margins; delivery is counted at a 25% group rate against GrabFood's 30%, with 15% the target. Asks NEA to bar compulsory designated suppliers at operator-run centres (Follow the Plate, D.5.3).
Succession Economics, E.4	Adopted: each month the successor pays the lower of 8% of takings and 30% of verified profit, but never less than $600, and never so much that the successor is left with less than $1,500 a month. We have reworked the worked example and its expected cost (Succession Economics, E.1.4).
Succession Economics, E.5	Adopted: the royalty starts when the Succession Mortgage is paid off, respects the same protected income, and runs for ten years at most, with a buy-out at any time at a published multiple; within the test's ten-year horizon the cap never binds (Succession Economics, E.1.4).
Succession Economics, E.4, the Plan's Part VII	Proposed: the first-loss reserve is one lump sum, topped up as it pays out, never a yearly spend; it is kept at or above the amount that covers the open mortgages' losses in 95 futures out of 100 (P95) and topped up from operating surpluses and the escrow fee. The Years 1–3 money safely covers about 74 mortgages at P95 and 25 through a 2020-style crisis, enough for a pilot of 5–20 (Succession Economics, E.1.5).
the Budget Book, B.6	Tests the funding stack and adds a trigger to the draw rule: if continuing gifts fall below half their Year 3–5 average for two years running, the draw falls to 3% (the Budget Book, B.9.4). Adds the Heritage Plate's Seed Bank share to continuing gifts from Year 3 (the Budget Book, B.9.2).
the Budget Book, B.7	The HDB tender lever is not zero to HDB: operators bid less rent to fund the commitment, so HDB may give up up to about $1.7M a year at half take-up, offered as its possible annual contribution to hawker heritage. Only its under-55 part, about $0.8M, counts towards the Protection Floor's coffeeshop match, since the part for stallholders of 55 and over goes into their own savings beside MRSS; it is not an extra benefit, and is phased in as coffeeshops are re-tendered (Follow the Plate, D.7.5, the Budget Book, B.9.5).
the Budget Book, B.7	Lists the retirement match as an optional ask of Government, for stallholders under 55 only: from 55 the existing MRSS already matches top-ups (up to $2,000 a year) for those within its means tests, so the plan helps hawkers claim it rather than paying twice, and the costs are ceilings ($4.14M a year for the base, $7.92M for the stretch). Giving and the Heritage Plate cover about 40% of the base ($14.7M short over Years 4–10) and 21% of the stretch; the decision is left for later (Follow the Plate, D.7.5, the Budget Book, B.9.5).
Hawkers and Social Protection, C.7	Puts numbers on the rule that help should go to the person, not the property: about 37% of any help paid to the stall would go back to landlords nationally, 30% to operators and private owners (Follow the Plate, D.7.2).