B.9Follow the Gift: Who Pays for the Plan

The plan's own money gets the same test as the diner's dollar: where it comes from, where it is kept and what it buys. On our central assumptions the campaign reaches $40M on time if major donors give about $5.7M. The fund (the corpus) then grows, because diners and food-and-beverage (F&B) partners keep giving. While the endowment is being raised, Government puts in about $1.08 for every private dollar. Afterwards it reverses: about $3.6 of private money for every public dollar in a typical later year.

The model (funding_model.py) runs the B.6 plan year by year for twenty years. Four terms first: the draw is the share of the fund spent each year; a close is a fundraising milestone; the bond ladder (bonds timed to mature in the years the money is needed) is government bonds bought to mature in the years the money is needed; and the operating match (the Government's yearly grant towards running costs) is Government's yearly grant towards running costs. Each year: (1) gifts arrive, and during the campaign every cash gift joins the fund and is matched dollar for dollar until the $10M pool is used, with major gifts making up whatever each close still needs; (2) the draw (the share of the endowment spent each year) is taken; (3) spending is met from the draw, the operating match, gifts for spending and the bond ladder; (4) from Year 3 continuing gifts meet any gap first, and the rest joins the fund. It matches B.6 exactly: draws of $0.42M, $1.05M and $1.41M in Years 1–3, and the printed gaps, which we cover in advance with $12.93M of bonds bought in Year 0 – $577,000 less than the gaps add up to, because the bonds earn interest. Spending comes in the ledgers (the plan's year-by-year cost tables)' two parts: the operating budget, which is the institution's own running cost, and the programme budget, the grants and expenses that reach hawkers at each centre. Over Years 0–3 the programme budget is 72% of the net requirement (what the plan must raise that year, after its own income), and from Year 4 the model indexes both parts to prices.

B.9.1Where the money comes from

In the diagram, the "Foundations, bridge grants" node ($13.6M) is the table's "Foundations and bridge grants" ($13.3M) plus the NHB (National Heritage Board) research grant ($357K), which the table lists as public: all of the Year 0 research grant and the bridge grants (a temporary grant that covers a gap until the fund is full). The bond ladder takes in $12.93M and pays out $8.47M in Years 1–2; the remaining $4.47M is held for Year 3. Rent rebates go from partner operators straight to the heritage stalls they host. At 20 hosted stalls they come to about $427 a month each, which lifts an HDB-rented coffeeshop stall from $7.56 to $8.86 an hour; the rebate is now one of the model's policies, and the simulator can switch it on for any coffeeshop or food-court stall. They are not matched and earn no deduction, so a partner's cash goes further at the Seed Bank; but a rebate from the landlord itself is the one stall-attached support (paid to the stall, so a landlord can take it back at renewal) that cannot leak back to the landlord.

B.9.2Twenty years

After the campaign, the money in each year is small beside the corpus (the endowment's invested capital), but it is what keeps the corpus whole. At the central assumptions the corpus ends Year 20 at 121% of its $40M target in today's dollars, and the draw reaches $1.77M in Year 10. Of all the giving, the dollar at the till matters most. Across the 5,000 simulated futures below, the biggest drivers of the fund in Year 20 are, in order: how many diners add the dollar at the till (a typical swing, one standard deviation, adds 25 percentage points to the fund as a share of its target); how many outlets take part (24); whether Government keeps its operating match (9); how many partners pledge (6); and the Heritage Plate (5). The Heritage Plate is licence income, not giving: from Year 3 the Seed Bank's 20% share, about $400K a year from Year 7, joins the continuing gifts, and lifts the Year 20 corpus from 106% of its target without it to 121%. Without the till dollar the corpus ends at 83% of its target and 17 of the later years cannot be paid in full. If Tekka's own hawkers vote against joining (B.9.3), the fund ends 14 points higher for a typical swing, because Tekka's later costs never arise. For the campaign itself, the order reverses: what major donors give decides whether the final close (a fundraising milestone) lands on time ($0.75M per standard deviation, against $0.26M for the till and $0.17M for partner take-up).

B.9.3Five thousand futures

The 5,000 simulated futures change, at random, everything the plan cannot control:

• how many partners pledge (20–80% of the profit pool) and renew (40–80%);
• how many bills add the dollar (3–15%), at how many outlets (2–10% of F&B sales);
• how many households give, and what other companies give;
• how much major donors have to give (median $6M);
• whether Government keeps its operating match (70%);
• whether Tekka, the pilot centre that opens in Year 2, passes its own hawkers' vote at the end of Year 1 (80%; voted down, it keeps what it was given in Year 1 and costs nothing after; the two Lighthouses vote in Year 0, and one that declines is replaced by a centre with a similar cuisine strength);
• how far the Heritage Plate gets (a quarter of the central case to double it);
• and the markets: yearly returns with occasional very bad years, and inflation around 1.5%.

Fundraising decides the timing. The second close lands by the end of Year 1 in 91% of futures, the final close by the end of Year 2 in 59%; in the worst tenth the campaign is $3.2M short at the end of Year 2 and needs more time. Investment policy decides how the corpus fares afterwards:

Returns after fees, assumed: fixed-income core (60% bonds, 40% equities) 4.1% a year, s.d. 7.3%; bonds and bills only 2.7%, s.d. 4%; growth (70% equities) 5.2%, s.d. 11%. Inflation 1.5%, s.d. 1 point.

B.9.4What it means for the plan

• From Year 4 the costs follow the heritage logic. By the end of Year 3 every pilot centre's planning count has been offered (B.9.6), so new Master Fellowships continue at a maintenance rate fed by the at-risk count (the Plan, §3.6), which comes from what hawkers choose to tell and from public notices: about 10% of each centre's planning count a year, at least one, an assumption until the count gives real numbers (3, 1 and 1 a year at Chinatown Complex, Geylang Serai and Tekka). The programme at three centres then costs $4.13M a year in Year 3 prices ($4.19M in Year 4). The draw and the operating match meet $2.02M of it and continuing gifts the rest ($2.17M of $3.26M), so at the central assumptions no year runs short, the remaining gifts join the corpus and it ends Year 20 at 121% of its target in today's dollars. The two Lighthouses (the first two pilot centres, Chinatown Complex and Geylang Serai) alone ($3.38M a year) would leave it at 158%.
• The number of new Master Fellowships each year is the figure to watch. The plan assumes about 10% of a centre's planning count a year. At 20% (9 Fellowships a year) the programme costs $5.07M and runs $0.1M short over Years 4–10 and $5.8M over Years 4–20; at 30% (13 Fellowships a year) the programme costs $6.06M and runs $6.9M short over Years 4–10 and $24.8M over Years 4–20; at 40% (18 Fellowships a year) the programme costs $7.28M and runs $16.0M short over Years 4–10 and $48.5M over Years 4–20. Up to about 10% a year the endowment keeps its real value (value after inflation) at the central assumptions; at 20% it no longer carries three centres.
• The add-ons. On the same rule Old Airport Road would add $1.14M a year from Year 4 and Maxwell $0.71M: four centres would leave $9.4M short over Years 4–20 and the corpus at 86%, five $23.3M short over Years 4–20 and the corpus at 86%, and they need $5.30M and $9.49M more of bridge grants in Year 0.
• The draw rate is not the lever (a setting the plan can turn up or down) while giving continues. Continuing gifts meet the year's spending first and the rest joins the corpus, so a higher draw only changes which pocket pays. At 3.5% from a fixed-income core the corpus keeps its real value in 71% of futures.
• It becomes the lever if giving stops. If every diner, partner and company stopped after Year 5 (Heritage Plate licence income, which is not a gift, continues), the programme would be about $2.2M a year short, and at 3.5% the corpus would keep its real value in only 34% of futures, against 45% at 3%. So the draw rule needs a trigger: if continuing gifts fall below half their Year 3–5 average for two years running, the draw falls to 3% until they recover.
• Bonds only would cost the programme about a third of its draw (a 2.5% draw, $1.04M in Year 10 against $1.55M) and protect the corpus no better. The fixed-income core with the bond ladder is the safer design: money needed within three years never meets the market, and the rest can earn enough to last.
• Major donors set the timetable: about $5.7M at the central assumptions, $6.4M if only a fifth of partners pledge, $7.7M with no till dollar. Pledging 1% across all of F&B would cut it to $5.0M.
• Public money carries the campaign; private money carries the future. $31.8M of public money (anchor, match, tax forgone, operating match and NHB's half of the Year 0 research grants, $357K) against $29.4M of private money in Years 0–2; in Year 6, $0.94M public against $3.43M private. The tax forgone is the largest public cost after the anchor and the match: $5.7M over the campaign.
• The public cost does not end with the campaign. From Year 4, Government's operating match, if it continues, is $0.50M a year (fixed in nominal terms), and the tax forgone on continuing gifts through the 250% deduction (donors deduct two and a half times their gift from taxable income) is $0.44M–$0.54M a year over Years 4–10 ($0.44M in Year 6): about $0.98M a year in all. Two more public costs arise only if adopted: Government's 1:1 match on Heritage Plate licence income in its first three years ($1.36M in all, Years 3–5) and HDB's forgone tender rent (up to $1.7M a year once fully phased in). Only the under-55 part of that rent, about $0.8M a year, counts towards the match: operators pay $1,200 for every stallholder, and the part for those 55 and over goes into their own retirement savings beside MRSS (Matched Retirement Savings Scheme, which matches top-ups from 55). The largest would be the retirement match itself, if Government is asked to fund its shortfall (B.9.5).

B.9.5The retirement match: what giving could cover

The Plan lists the retirement match as an optional ask of Government (B.7). Before making that ask, it is worth knowing how much of the match giving and licensing could pay. The match is costed here, not asked for: it starts only if and when it is adopted, so nothing is spent on it in Years 1–3. If adopted, it goes national from Year 4, reaching a quarter, half, three-quarters and all of its full cost in Years 4–7, and rises with prices after Year 3. (An early match at the two Lighthouses, $1,200 a year for their stallholders under 55, at most $173K a year, is a switch in the model, off by default, because no funder is named for it.) Full cost is $4.14M a year for the base and $7.92M for the stretch (Follow the Plate, D.7.5). Both are ceilings: the match is for stallholders under 55, fewer than half of them, because from 55 the existing MRSS already matches top-ups, up to $2,000 a year, for those within its means tests (Follow the Plate, D.7.5), and the plan helps hawkers claim it rather than paying twice.

Four sources could pay part of it. The Heritage Plate, a certification mark licensed to restaurants, hotels, caterers, airlines and packaged-food makers, charges 50 cents a certified plate, priced in and shown on the menu, plus a 3% royalty on packaged goods. Licensed hawker brands have mostly done badly: of the licensed or acquired expansions in Nigel Chua's study, most shrank or closed, and the one licence that paid a share of profit paid nothing (IPS, 2024; Succession Economics, E.1.1). So the model assumes that in any year a third of the licensed outlets and packaged lines fail and pay nothing, and that the mark signs new ones to hold the outlet counts (assumption). At the central assumptions it earns $60K in Year 3, $340K in Year 4, $1.0M in Year 5, $1.8M in Year 6 and $2.0M a year from Year 7; 40% of it goes to the match. Government could match licence income 1:1 for its first three years, up to $2M a year: $1.4M in all. Continuing gifts above the programme's needs, which would otherwise join the corpus, could be earmarked (set aside for one use) in part. HDB (Housing & Development Board, Singapore's public-housing landlord), if it agrees, could count the tender rent it forgoes under the tender lever. This is the same money Follow the Plate, D.7.5 treats as a way to pay for the stretch's match at HDB-rented coffeeshops, not an extra benefit, so it offsets the stretch only. It can bite only as coffeeshops are re-tendered, because HDB holds rents for six years (Hawkers and Social Protection, C.12), so the model phases it in: 25%, 50%, 75% and 100% of $1.7M in Years 4–7. Operators pay $1,200 for every stallholder, but only the part for those under 55 can stand in for the match, so the model counts 50% of it, at most $0.8M a year; the part for those 55 and over goes into their own retirement savings beside MRSS and is not counted. It never counts more than that year's match for HDB-rented coffeeshop stallholders under 55 ($1.18M at full roll-out), a limit that does not bind at these assumptions.

Central assumptions: Heritage Plate central case; half of the continuing gifts above the programme's needs earmarked for the match from Year 4. Nominal dollars. Money not needed in a year is carried forward: Year 3's licence income and Government's match on it ($84K) help pay Year 4. Years 1–3 show no cost because the match starts only if adopted, nationally from Year 4. HDB's column, phased in as coffeeshops are re-tendered, is the under-55 part of its forgone rent ($210K in Year 4, $0.8M a year from Year 7, against forgone rent of $1.7M), and applies only to the last column. The outlet numbers, plates an outlet, packaged-goods sales, the failure rate and Government's licence match are assumptions.

Corpus, Year 20: in today's dollars, as a share of the $40M target. Shortfalls are summed over Years 4–10 in each year's money. With HDB, $4.62M of the $9.24M of tender rent HDB forgoes over Years 4–10 counts towards the match.

The corpus trade-off. Earmarking gifts for the match takes them out of the endowment. With no earmark the corpus ends Year 20 at 121% of its target in today's dollars. Earmarking half of the surplus gifts pays $3.9M of the match over Years 4–10 and leaves the corpus at 101%. Earmarking all of it pays $7.4M and leaves 86%: the endowment would no longer keep its real value. Half is about as far as the earmark can go without spending the endowment's future on today's match.

B.9.6Sizing each centre: heritage-worthy stalls, a proxy

The ledgers cost each pilot centre from its own shape (B.1). The heritage lines (Master Fellowships, apprentices, documented dishes and landings) follow how many of its stalls are worth saving; the critical-illness cover, the Relief-Hawker Corps, the table-cleaning fee subsidy and the Buying Circles' leads follow how many stalls it has. No one publishes the age of a centre's stallholders, so the number worth saving is estimated from a proxy: the stalls that guides and heritage records already name. The counts were made on 28 October 2026.

• Tier 1: a stall in the MICHELIN Guide Singapore (Bib Gourmand, Selected or Plate, in the current edition or an earlier one, or a former star), or a winner of a Hawker Heritage Award.
• Tier 2: a stall documented as running for 30 years or more, or in its second generation or later.

An assumption, labelled as a proxy. Each stall is counted once, in the higher tier. Tekka's stall count is data.gov.sg's. Padding corrects for what guides miss: they name 27–32% of the stalls at compact tourist centres but only 15–21% at big or heartland ones, so about half of that gap is treated as undercounting, ×1.5 at Chinatown Complex, Tekka and Old Airport Road, ×1.2 at Geylang Serai and ×1.0 at Maxwell, where the guides' coverage is densest. The planning count is the padded count, rounded. The typical further centre is a median (the middle figure: half are above it, half below) 100-stall centre with an 8% heritage share, padded ×1.3. Fellowships run from the year the centre opens: Years 1–3 at the two Lighthouses, Years 2–3 at the others. Sources: the MICHELIN Guide Singapore 2026 list as compiled by The Ranting Panda; the MICHELIN Guide's hawker guides to Chinatown Complex and Tekka Centre; NHB's Roots page on Geylang Serai; Mothership on Gerai Nenek Obek's Hawker Heritage Award; data.gov.sg's list of government markets and hawker centres; Eatbook on Maxwell Food Centre; and Seth Lui on Old Airport Road (Sources).

Popular is not the same as needing help. A stall with a MICHELIN plate and a queue may need nothing from anyone. The proxy is a starting point for a budget, padded for what the guides miss, not a list of who will be offered a Fellowship. A Fellowship is offered when a hawker wants to teach and the Registry's open test says the dish is fragile; the hawker decides, and declining changes nothing (the Plan, §3.6, B.13). From the planning count the ledgers take one Master Fellowship for each stall over the centre's three years, phased 30%, 35% and 35% at a centre that opens in Year 1, and 40% and 60% over the two years a centre that opens in Year 2 has in the window, so the first cohort (a group that starts together) is the smallest and each later one waits on the last; two apprentices a Fellowship; one documented dish a Fellowship; and landings for about 30% of the apprentices. The Fellowship line is a ceiling, not a target. Willing masters, not money, are the real limit: NEA's Hawkers Succession Scheme completed six handovers between January 2022 and September 2025 (MSE, 22 Sep 2025). Money a centre does not spend on Fellowships returns to the endowment. Once a centre's planning count has been offered, from Year 4 at the latest, new Fellowships continue at a maintenance rate, about 10% of the planning count a year and at least one, an assumption until the at-risk count replaces it (B.9.4).