The receipts
DraftBefore ParliamentWe read every page of both official documents: 186 pages of the English memorandum and 337 of the Dutch submission to Parliament. A second, independent reading re-checked 1,630 key figures against the page and caught 179 transcription slips before anything was published.
A machine then ran 3,180 arithmetic checks over the numbers book, and comparing every English figure with the Dutch original found no differences. Where the documents themselves disagree, we say so, and show which figure we use.
Watch it balance
Draft ordinance · 2027
The legal totals in the draft budget law, to the guilder. Take spending from revenue and the surplus comes out exact, no rounding anywhere.
114 of 114checks pass · re-run in your browser
Each check adds up a set of figures and compares the result with the official total. They run every time the site is built; if one fails, nothing is published.
Pick a figure
Cg 670.3M
Exactly Cg 670,306,505
| Taxes | Cg 497M |
|---|---|
| Licences & permits | Cg 22M |
| Fees & concessions | Cg 64M |
| Other revenue | Cg 87M |
| Total shown | Cg 670.3M |
The parts are rounded one by one, so they can differ from the total by Cg 0.3M , rounding, not a missing amount.
Cg 662.2M
Exactly Cg 662,158,099
| Education | Cg 133.1M |
|---|---|
| Justice | Cg 113.8M |
| Health & Social | Cg 103.7M |
| General Affairs | Cg 101.1M |
| Finance | Cg 97.9M |
| Tourism & Economy | Cg 43.9M |
| Housing & Infra | Cg 39.8M |
| Parliament | Cg 21.6M |
| Special Entities | Cg 7.4M |
| Total shown | Cg 662.2M |
The parts are rounded one by one, so they can differ from the total by Cg 0.1M , rounding, not a missing amount.
Cg 8.1M
Exactly Cg 8,148,406
| Revenue | Cg 670.3M |
|---|---|
| − Spending | Cg 662.2M |
| Total shown | Cg 8.1M |
Cg 497M
| Wage tax | Cg 194M |
|---|---|
| Turnover tax | Cg 184M |
| Profit tax | Cg 47M |
| Transfer tax | Cg 20M |
| Room tax | Cg 15M |
| Excise duties | Cg 13M |
| Motor vehicle tax | Cg 12M |
| Income tax | Cg 6M |
| Time-share fee | Cg 3M |
| Rental-car tax | Cg 2M |
| Other taxes | Cg 1M |
| Total shown | Cg 497M |
Cg 133.1M
Exactly Cg 133,079,014
| Subsidies & transfers | Cg 74.3M |
|---|---|
| Personnel costs | Cg 34.3M |
| Other operating costs | Cg 17.3M |
| Student costs | Cg 3.7M |
| Accommodation & transport | Cg 2.8M |
| Not itemised in the numbers book | Cg 0.6M |
| Total shown | Cg 133.1M |
The parts are rounded one by one, so they can differ from the total by Cg 0.1M , rounding, not a missing amount.
Cg 289.8M
Exactly Cg 289,845,608
| Justice | Cg 81.8M |
|---|---|
| Health & Social | Cg 55.7M |
| General Affairs | Cg 39.1M |
| Education | Cg 34.3M |
| Finance | Cg 26.4M |
| Tourism & Economy | Cg 20.2M |
| Parliament | Cg 14.3M |
| Housing & Infra | Cg 12.7M |
| Special Entities | Cg 5.4M |
| Total shown | Cg 289.8M |
The parts are rounded one by one, so they can differ from the total by Cg 0.1M , rounding, not a missing amount.
Cg 53M
The law allows up to Cg 52,986,502
Cg 222M
| Other equipment | Cg 79M |
|---|---|
| Non-residential buildings | Cg 61M |
| ICT equipment | Cg 36M |
| Land improvements | Cg 22M |
| Land | Cg 6M |
| Research & development | Cg 6M |
| Transport equipment | Cg 5M |
| Computer software | Cg 3M |
| Residential buildings | Cg 2M |
| Other structures | Cg 2M |
| Total shown | Cg 222M |
Cg 63M
Where the documents disagree
The official documents run to 523 pages, and in a few places they don't line up: the text gives one figure and the table beside it another, or a printed total doesn't equal its lines. Each one is listed here with both sides, the figure this site uses and why, and the pages.
This list was re-checked against the pages on 2 October 2026. The legal totals in the draft budget law still add up to the guilder. When an amendment or the adopted budget deals with an item, it will be marked “explained” or “corrected”, with a note saying what changed and when.
Showing 20 of 20
One page gives one figure; another page, sometimes the table right beside it, gives a different one.
The text says “a balance of Cg negative 8 million is expected at the end of the year”. The liquidity table beside it ends 2027 at +Cg 2M. The Dutch text says the same.
What we show
We show the table's +2, because every row of the table adds up; the −8 matches nothing in it.
§3.2.1 puts total debt incl. short-term debt at Cg 1,127M in 2027, 32.2% of GDP. The loan overview gives Cg 1,078M of loans at end-2027. §3.1 says the ratio stays “around 40–41 percent”, which matches the Cft's own estimate of Cg 1,424M incl. all short-term debts, 41% of GDP. The memorandum never reconciles the three.
What we show
We show the loans (Cg 1,078M, §6.6) and the memorandum's total incl. short-term debt (Cg 1,127M, §3.2.1), and note the Cft's broader Cg 1,424M.
The text says “In 2027, Cg 17 million in repayments will be paid”. The loan overview says Cg 22.0M and the numbers book Cg 22,002,692. 17 was the 2026 figure.
What we show
We use 22.
The interest section projects 3.75% “on a loan volume of Cg 56.5 million”. Everywhere else the 2027 loan is Cg 53M, and the draft ordinance authorises Cg 52,986,502. (The Cft reviewed an earlier draft with a Cg 57M loan.)
What we show
We use the ordinance's Cg 52,986,502.
The text says spending rises Cg 27M (personnel +9, other operating costs +8, subsidies −4). The table shows 637 → 662, which is +25 (personnel +7, other operating +4, subsidies 0), although its own difference column prints 27.
What we show
We use the exact amounts from the numbers book: +Cg 26.2M.
The results table shows revenue 648 / 666 / 688 and spending 640 / 642 / 668 for 2028–2030. The detailed tables show revenue 647 / 665 / 682 and spending 640 / 642 / 660.
What we show
We show the results table as printed and flag the gap.
The text says the new financial-assistance law is “an important explanation for the increase in estimated expenditure in 2027 compared to 2026”. The table shows social provisions going from 37 to 36.
What we show
We show the table's figures.
The text says “For the year 2026, a dividend payment of Cg 9 million from the CBCS has been budgeted”. The table shows 6 for 2026.
What we show
We show the table's figures.
VROMI's capital table prints “Generators GEBE 75.000” (thousands), but the table's own total of 95.212 only adds up if that line is 75.600.
What we show
We use Cg 75.6M, which the total requires and the 2026 budget also carried.
The Council of Advice found it “not plausible” that Justice needs no or only limited changes and strongly advised revising the chapter. The government replied that the necessary changes had been made. The submitted chapter still says “There are no significant differences between the approved 2026 budget and the (draft) 2027 budget.”
What we show
We report both statements; Justice's 2027 total is essentially unchanged from 2026.
The memorandum prints the approved 2026 budget as revenue 647 and spending 637, a result of 10. The numbers book's 2026 column adds up, line by line, to spending of Cg 635,929,870; with 2026 revenue of Cg 647,140,798 that is a result of Cg 11,210,928. Three chapters (General Affairs, Education, Health) also print 2026 totals that differ from their own lines.
What we show
We compare 2027 with the exact 2026 amounts from the numbers book (Cg 635.9M spending, +Cg 11.2M result), so every change is worked out to the guilder.
The memorandum's table (EN 165) prints 2027 personnel costs as 288 and other operating costs as 112. Adding up the numbers book's own lines by account gives Cg 289.8M and Cg 109.1M; the consolidation statement (Staat H) also prints personnel costs of Cg 289,860,608. The memorandum appears to group a few lines differently.
What we show
We show the numbers book's sums, which match each ministry's line items and the Staat H personnel total.
The financial chapter prints 21 million; Chapter 7 prints 22. The exact amount is Cg 21,632,335, which rounds to 22. Chapter 7 also prints “222” for Parliament in 2028.
What we show
We use the exact amount, Cg 21,632,335 (Cg 21.6M), which rounds to 22.
The four revenue lines on the headline page (497 + 25 + 65 + 84) add up to 671, not the 670 shown, and licences are 25 there but 22 in the detailed revenue table.
What we show
We use the detailed chapters, which add up to the legal total of Cg 670,306,505.
Chapter 7 prints Finance 99 and VROMI 41 for 2027; the financial chapter and the exact amounts give 98 and 40.
What we show
We use the financial-chapter figures, which match the exact amounts to the guilder.
A printed total that is higher or lower than the printed lines it is meant to add up.
In 34 places a printed total is higher or lower than the lines printed above it, most likely hidden spreadsheet rows. Three of them leave a department's total higher than its own lines; the others are subtotals, chapter totals or capital-budget totals that don't match their parts. Example: the Youth department's 2027 total is Cg 642,124 more than its printed lines; General Affairs' 2026 chapter total is printed as Cg 89,988,040 while its departments add up to Cg 89,974,905.
What we show
Those three appear in our searchable ledger as visible “not itemised” lines, so the ledger still reconciles to the ordinance total to the guilder.
Two subtotals don't equal their lines (printed 627.428 where the lines add to 672.429; printed 18.692 where they add to 20.111), the 2026 interest total is Cg 0.8M less than its parts, and two loan rows carry no label.
What we show
None of these change the headline loan balances, which we take from the loan overview.
Nothing contradicts here, but anyone reading the totals should know it.
The memorandum says every 2026 capital item “has been included again in 2027” because “it is currently not known whether the relevant loan for 2026 will be obtained”. That includes the Cg 75.6M GEBE generators, also in the 2026 budget.
What we show
We show the 2027 capital budget as printed and say plainly that it largely repeats 2026.
Loans approved for 2023, 2024, 2025 and 2026 (Cg 134.2M together) appear again as financing for the 2027 investment budget, alongside the new Cg 53M loan.
What we show
We show the financing as printed.
A template carry-over in both the English and Dutch memoranda.
What we show
Cosmetic; no figures affected.
The numbers book comes with statements, subsidies, loans, rent, staff, vehicles, buildings and agreements. Reading them line by line turned up 50 more items: places where they don't line up with each other or with the ledger, and gaps or labelling problems worth knowing. Most are small; 6 are material. Open any item to read both sides.
The statement of loans taken up (Staat E, NL 78) has a line 'Government TrustFund Loan to PJIA' at 4.50% from 30-06-2020, with a balance printed as a negative, (167,440,000), in every year-end column from 2025 to 2030, and interest 'P.M.' (no amount). The overview of loans the government has issued (NL 108) lists two 'Trustfund' loans to the airport holding company: 91,000,000 and 39,130,000, both at 4.50%, together 130,130,000, both still fully outstanding at 31-12-2025. Both pages report the position at 31 December 2025, so timing does not explain the difference of 37,310,000 (computed). Re-reading both pages: there is no other Trust Fund line on either page. Adding the airport's two interest-free loans from NL 108 (9,100,000 and the 27,000,000 bridge loan) gives 166,230,000, still 1,210,000 short of 167,440,000, so that does not explain it either. Statement E's own 'Bridge loan PJIA' rows are empty. Because the Trust Fund interest is 'P.M.', accrued interest cannot be checked: 4.5% of 130,130,000 is 5,855,850 a year, and 37,310,000 is not a whole number of years of that (computed). The English memorandum does not itemise the line (EN 21 groups it under 'Interest on old loans'). Statement E also treats the two airport on-loans differently: the EIB loan counts as debt (+91,000,000), while the Trust Fund loan is subtracted (-167,440,000), which lowers the printed debt total (Totaal generaal 1,077,513,322 at end-2027, NL 79).
What we show
Unresolved: the documents do not reconcile the two figures. Show each with its own citation and do not combine them. If Statement E netted the 130,130,000 from the loans-issued overview instead, its end-2027 total would be 37,310,000 higher, 1,114,823,322 (computed). Without any netting it would be 1,244,953,322 (computed). The documents do not say which figure is correct.
Staat G lists the Integriteits Kamer (Integrity Chamber) as a recurring government transfer of 2,160,639 (2025), 2,303,672 (2026) and 2,298,280 (2027) (NL 82). In the Special Entities ledger the Chamber's 2025 and 2026 money is a single transfer (2,459,728 and 2,390,569). For 2027 the transfer is removed and the same 2,298,280 is itemised as the Chamber's own staff costs (1,489,780) and running costs (808,500) (NL 13-14; also Staat H, NL 63). Consistent with this, the memorandum's Special Entities category 44 (subsidies, transfers and commission fees) falls from 3 to 1 (Cg million) in 2027 (EN 165). So the 2027 Staat G amount equals the Chamber's total ledger budget, which in 2027 is not booked as a transfer; it is not an additional payment on top of that budget. The earlier years also disagree: Staat G 2,160,639 / 2,303,672 against ledger transfers 2,459,728 / 2,390,569. Separately, the personnel statement shows the Integrity Chamber with no staff and no personnel cost at all (Staat B, NL 67), although the ledger budgets 1,489,780 for its staff in 2027.
What we show
The row is kept because Staat G prints it and includes it in its 107,438,419 total. Treat it as 'not a transfer in the 2027 ledger'. Never combine Staat G with the ledger's staff and running costs (for example in a spending-by-type chart): the Chamber's 2,298,280 would be counted twice. Without it, the 2027 Staat G rows total 105,140,139 (computed).
'Refinancing liquidity loans (2023)', 313,909,260, is printed at 3.40%, but its yearly interest of 9,103,369 is exactly 2.90% of the balance. At 3.40% the interest would be 10,672,915, which is 1,569,546 more a year (computed). A second row with the same name is labelled 2.90% and contains only dashes (NL 78). The memorandum's overview repeats the 3.40% / 9.103 pairing (EN 21).
What we show
Show the printed rate and interest and say that they do not match; the documents do not explain which is correct. If 3.40% applied for the whole year, the 2027 interest would be about Cg 1.57 million higher (computed).
Prison (Justice): cash wages are printed as 8,693,615.09, but total wages and personnel cost are built on 6,693,615.09, a gap of exactly 2,000,000 (NL 70). The memorandum separately says the remand centre's personnel budget was raised by Cg 2.0 million (EN 83); the source does not say it is the same money. General Audit Chamber: wage components add up to 605,935 but total wages are printed as 988,743, and personnel cost (758,231) does not follow from either (NL 66). Meteorological Service +85,000.10 (NL 73) and Beheersdienst (management/maintenance service, VROMI) +201,000.00 (NL 74) between components and total wages. One TEATT total is printed as '##########' (NL 73). The holiday-allowance column is headed 6%, but the amounts equal 8% of total wages (for example Griffie 118,994 on 1,487,419, NL 66).
What we show
These affect personnel costs, not the head-counts used in 'personnel'. Do not quote Staat B row costs for these units without the caveat.
The rental overview (NL 100) covers only six ministries. Eight rent lines appear instead among the service agreements: Parliament building rent, Caribbeam Palm NV, Cg 1,706,400 (NL 115); General Audit Chamber office lease, Vineyard Private Fund Foundation, Cg 108,000 (NL 115); Ombudsman rent, Commsystem, Cg 153,450 (NL 116); Social and Economic Council office lease, Harbor View building, Cg 170,878 (NL 116); Security Service rents of Cg 226,800, Cg 97,200 and Cg 205,200 (the last for a third office marked "Pending") (NL 116); Integrity Chamber office rent Cg 152,880 (NL 116). Together Cg 2,820,808 a year (computed), or Cg 2,615,608 without the pending office.
What we show
The rentals list mirrors the official overview on NL 100 so it matches the printed total. The eight rent lines stay under serviceAgreements. Any 'total government rent' figure should say it covers ministries only, or add these lines.
The English memorandum shows long-lease (ground rent) revenue realised in 2025 of Cg 17 million (EN 163); the leasehold statistics give the 2025 annual ground rent as NAf 6,853,647.00 (NL 97).
What we show
Not resolvable from the documents; neither explains the gap. Both figures are shown with their own labels.
Printed totals: 101,790,557 (2025) and 108,311,143 (2026). The 83 rows add up to 101,618,861 and 108,064,303, which is 171,696 and 246,840 less. For 2027 the rows equal the printed 107,438,419 exactly (NL 87). Both gaps match the General Affairs advisory-committees budget (cost centre 3350, 489,328 in 2025 and 683,963 in 2026, NL 17) minus the Staat G rows under 3350 (317,632 and 437,123). So the totals include an amount with no row of its own. In 2027 cost centre 3350 falls by exactly 246,840 (NL 17) and a VROMI Rent Committee row of 246,840 appears (NL 87); the memorandum says the Rent Committee was transferred from AZ to VROMI (EN 146). The documents do not say what the 2025 amount of 171,696 was for.
What we show
Use the printed totals for headline figures and the rows for breakdowns. The gap is an unitemised amount, not a recipient that can be named.
The 2025 column is headed 'Vermoedelijk uitgaven' 2025 (estimated expenditure 2025). The 'realised' field in this file holds those estimates.
What we show
Label 2025 figures as 'estimated 2025' rather than 'actual'.
Staat G gives 200,000 for the Raad van State (Council of State) in 2025 (NL 82); the Special Entities ledger shows transfers of 99,174 (NL 13). 2026 and 2027 agree (300,000).
What we show
Show the Staat G figure for the Staat G list and note the ledger figure; the documents do not explain the difference.
Blank in all three years: Youth Program, Dutch Quarter Community Center (NL 85) and Gaming Control Board (NL 87). Blank for 2025 only: Mortgage Guarantee Fund and Infobiz (NL 83), Uniformed Organization (NL 85), Turning Point under VSOA (NL 86) and Committee of Experts (NL 87).
What we show
Blank 2025 and 2026 cells are stored as null; blank 2027 cells as 0 (the field needs a number). None of this changes any total.
Each row's ministry is inferred from the first digit of its cost centre (2 = Special Entities, 3 = General Affairs, 5 = Justice, 6 = ECYS, 7 = VSOA, 8 = TEATT, 9 = VROMI). Three rows print no cost centre: Mortgage Guarantee Fund (code 44201.41) and Infobiz (code 44201 40 40018) are placed under Finance, whose budget uses 4xxx cost centres and 400xx project codes (NL 23); Gaming Control Board (code 44201.8270.80037) is placed under TEATT. All three carry no money in 2026 or 2027.
What we show
Treat the 'area' of these three rows as an inference.
'Turning Point' appears twice with the same decision reference: under Justice with 1,885,281 (NL 83) and under VSOA with 0 (NL 86). 'Foster Home' and 'Foster Home / New Start Foundation' share the same account codes (NL 83). The 'Electoral Reform Committee' name is printed in the decision column and its recipient column is empty (NL 82).
What we show
Both Turning Point rows are kept as printed; only the Justice row carries money, so nothing is double-counted.
Staat G totals 107,438,419 for 2027 (NL 87). The memorandum's liquidity table shows 'Subsidies 106' (Cg million) for 2027 (EN 22), and its economic category 44 'Subsidies, transfers and commission fees' is also 106 (EN 165). Even without the Integrity Chamber (105,140,139, computed) the figures do not match.
What we show
Present Staat G as 'the statutory list of subsidies and transfers' and do not call it the same figure as the memorandum's 106 million; the documents do not reconcile the two.
EN 21: (1) prints the 73.5 million Refinancing Bond's figures on a row labelled 2,90% and leaves a 3,62% row empty (NL 78 and EN 42 give 3.62%); (2) prints Loan Capex 2023's end-2030 balance as 46.640 (thousand) against 43,639,861 on NL 78; (3) prints the 2028 subtotal as 627.428 although its rows add up to 672.429; (4) prints 2026 'Interest on old loans' as 9.905 against 9,095,295 derived from NL 78-79; (5) prints the 2029 interest subtotal as 18.692 although its rows add up to 20.111. Its 2026 and 2027 grand totals agree with Statement E within rounding.
What we show
Use the Statement E figures (NL 78-79); the EN 21 cells don't match Statement E.
Statement E's 2027 interest is 29,872,239 (NL 79), and the consolidation statement shows the same figure as Finance 'Rentekosten' (interest costs) (Staat H, NL 63). The Finance chapter budgets 'Interest costs' of 30,116 (Cg thousand), 243,761 more (computed), and says this also covers an estimate for future loans; bank charges (650) are a separate line (EN 50). The liquidity table shows 'Interest and bank charges 32' (Cg million) (EN 22), more than 30.116 + 0.650, and the category table shows 31 for 'Interest and bank charges' (EN 165).
What we show
Quote Statement E for interest on existing loans and the Finance chapter for the interest budget; the documents do not bridge the two.
(1) The Sint Maarten 2010 tranche maturing in 2030 (78,571,000) still shows its full balance at end-2030 (NL 79). (2) 'Aankoop Emilio Wilson' repeats its 2029 figures in 2030, with no repayment (NL 79). (3) Loan Capex 2023's 2030 interest (1,376,507) breaks its otherwise steady interest-plus-repayment pattern (NL 78). (4) 'Budget tekort lening 2017' is printed as maturing in 2032 but still has 34,000,000 outstanding at end-2030 while repaying 2,000,000 a year (NL 78). Statement E's existing-loan balance falls by 21,459,812 in 2030 (computed), while the memorandum plans repayments of Cg 101 million that year (EN 22).
What we show
These affect only the 2028-2030 projections, not the 2026 or 2027 figures used on the site.
For all seven loans the balance at 31-12-2025 equals the original amount, together 264,230,000 (computed; no total is printed), of which 257,230,000 to the airport holding company. The table does not show whether anything has been repaid, including on the 4.50% EIB and Trust Fund loans, and it shows no interest received and no repayment schedule. The amount column is headed 'Oorspronkelijk bedrag Naf gewarborgde lening' (original amount of the guaranteed loan), the same wording as the guarantees table (NL 110), and amounts are labelled 'Naf', the old name of the guilder (NL 108).
What we show
Present balances as 'as recorded at 31-12-2025'; the documents cannot confirm whether any repayments were made.
The capital budget includes study loans issued ('Verstrekte studieleningen u/g') of 5,280,000 a year (NL 56), and the liquidity table shows 'Student loans 5' (Cg million) (EN 22). Neither appears in the loans-issued overview (NL 108). The memorandum says a Cg 75.6 million loan was concluded in 2024 'on behalf of the government for the benefit of GEBE' (EN 178). It appears neither as a loan issued nor as a guarantee; the capital annex lists 'Generatoren GEBE' (GEBE generators), 75,600,000, as 'Other Equipment' in VROMI's capital budget (NL 56).
What we show
Do not describe the NL 108 overview as a complete list of the government's lending.
The 'Restant van de lening Naf 12/31/2025' (outstanding balance) column is empty for all three guarantees, so the government's current exposure is not disclosed. Only the original amounts can be added: 26,900,000 (computed; no total is printed). The table has no description column, so none of the guaranteed loans, including the airport's 20,000,000 at 0%, is described. The table title is cut off at the margin (NL 110). TelEm (the government telecom company) has asked for a Cg 5 million guarantee, which has not yet been provided and is still being examined (EN 178); it is not listed.
What we show
Present 26.9 million as the original amount guaranteed, not as the current risk.
Staat B for Parliament and the High Councils of State prints 89 filled, 14 vacancies, 103 required, 8 budgeted vacancies and 97 budgeted (NL 66). The same page splits these into 'Fictieve werkplaats' (fictitious workplaces): 31 filled, 32 required and 31 budgeted, costing 1,685,244, the same amount as the 'Ex-Gezagdragers pensioen' row (pensions of former political office-holders). The 'Actuele werkplaats' (actual workplaces) are 58 filled, 71 required and 66 budgeted. The staffing plan's budgeted total (NL 58) uses the actual figure (66), but its establishment (103) includes the fictitious places.
What we show
The PARL row uses the printed Staat B totals (including the fictitious places). For people actually working for Parliament and the High Councils, use 58 filled of 71 positions (13 vacancies, 8 funded).
Filled / funded vacancies / budgeted / establishment: AZ 236/58/294/448 (NL 68) vs 236/59/295/439 (EN 37); FIN establishment 213 (NL 69) vs 214 (EN 55); JUS 590/46/636/1,096 (NL 70) vs 590/51/644/1,095 (EN 84), where EN's own department rows add up to 635 budgeted; ECYS establishment 367 (NL 71) vs 366 (EN 104); VROMI 106 filled / 13 funded (NL 74) vs 107 / 12 in the EN table (EN 156), whose text says 106 of 146 positions are filled and 12 vacancies are budgeted, while EN 144 says 13 vacancies are reserved. VSOA (EN 122) and TEATT (EN 141) match.
What we show
Use Staat B (the statutory annex) and cite it; mention that the memorandum's chapter tables differ slightly.
Staat B's budgeted personnel cost adds up to about Cg 211.4 million (summary blocks, computed), while the consolidation statement gives total personnel costs of 289,860,608 for 2027 (Staat H, NL 63) and the memorandum gives Cg 283 million for 2027 personnel costs including the civil servants' share of OZR (EN 22). A footnote on every Staat B page says that for some departments an estimated amount for overtime is budgeted in the main budget but is not part of these salary details (NL 66-74). Staat B shows only 2027 (no 2025 or 2026 columns) and has no page for the Governor's chapter.
What we show
Use Staat B for staff numbers and per-unit costs only; never present its total as the government's personnel cost.
34 of the 36 rents convert USD to Cg at 1.80 (to within one cent). Police station Cole Bay: USD 11,000 -> Cg 20,196.00 a month (1.836; 1.80 would give 19,800.00, i.e. Cg 396 a month or Cg 4,752 a year less). Emergency dispatch room (meldkamer): USD 10,000 -> Cg 18,200.00 (1.82; 1.80 would give 18,000.00, Cg 200 a month or Cg 2,400 a year less).
What we show
The printed Cg amounts are used, because they make up the printed total. The source does not explain the different rates.
"Brown building/ICT staf" appears twice with the same rent (USD 2,966.67 / Cg 5,340.00 a month) and the same contents value ($ 300,000.00); one row gives Cannegietestreet 51, the other Cannegietestreet 52.
What we show
Both rows are kept, because both are counted in the printed total. The source does not say whether these are two separate leases.
Justice service agreements include "Coral Sea development" Cg 18,200 (NL 113), the same figure as the monthly rent of the emergency dispatch room owned by coral sea development (NL 100); "Questel NV" Cg 19,800 (NL 113), which equals USD 11,000 x 1.80 for the Cole Bay police station that Questel B.V rents out (NL 100 converts it to Cg 20,196 a month); and "Electricity Maho Substation: Strathclyde Hotel Holding" Cg 45,360 (NL 113), which equals 12 x Cg 3,780, the monthly rent of the Maho police station (NL 100). The General Audit Chamber's office lease with Vineyard Private Fund Foundation, Cg 108,000 (NL 115), equals 12 x Cg 9,000, the rent of the economic-licences office leased from Vineyard Private Fund (NL 100).
What we show
Both lists are kept as printed and nothing is netted out. The source does not say whether any of these are the same arrangement counted twice, or whether some service-agreement amounts are monthly.
NL 100 gives annual rent of Cg 16,031,975.99. The General Affairs budget line "Rent of buildings and premises" is Cg 16.045 million for 2027 (EN 29), and the project to buy government buildings speaks of annual rental expenditure of "approximately Cg 14 million" (EN 171).
What we show
Not reconciled. The EN 29 line is "Rent of buildings and premises", and its explanation says the increase covers the SZV lease "including maintenance" and "outstanding obligations from previous years"; EN 171 is a rounded estimate ("approximately"). The site shows the NL 100 figure with its own label.
For the building rented from SZV, the building value is printed "n.v.t" (not applicable) but the contents value is $ 6,944,444.44, far above any other contents value in the annex.
What we show
Not used anywhere. The source does not explain the figure.
Value cells read "see mark", "See remark" or "see remark" three times on NL 99 and three times on NL 100, but no remarks are printed anywhere in the annex. Several other cells read "niet bekend" (not known), and some names are cut off at the cell edge.
What we show
Shown as printed in the notes.
On NL 99 each yellow ministry label is printed after its group of buildings (the Receivers Office sits above "MINISTERIE VAN FINANCIEN", the police stations and prison above "MINISTERIE VAN JUSTITIE"); on NL 100 the labels are printed above each group. Read this way, no owned buildings are listed for TEATT (its label directly follows the Education label).
What we show
Buildings are assigned to the ministry label printed below them. This is an interpretation; the page has no headings that settle it.
Building and contents values carry a "$" sign but the currency is never named, and the page prints no total. 29 of the 32 buildings have a building value; together $ 47,702,202.00 for buildings and $ 3,299,857.28 for contents (computed). The same land-survey number "C 20-68" is given for both Prins Willem Alexander School and the pump buildings.
What we show
Values are quoted as printed in each building's note and not converted to Cg. Computed totals are not used as official figures.
The footnote says the fleet is "308 vehicle of which 246 is comprehensive and 62 third party". Counting the rows gives 308 vehicles, of which 245 comprehensive and 63 third party.
What we show
The vehicle counts per area add up to the printed 308. Insurance counts in the notes are counted from the rows.
Plate M-11255 (Toyota RAV-4, 2021, same chassis number) is listed under General Affairs (NL 90) and again under VSA (NL 95). Two pairs of police vehicles share chassis numbers: M-2693 and M-11332, and M-10023 and M-11336 (NL 92). If these are double entries, the fleet is 305 vehicles, not 308.
What we show
Counts are kept as printed (308). The source does not say whether these are double entries.
The legend defines only yellow ("Valid Inspetion"). 92 vehicles have their plate and inspection date printed in red, all with dates already past (the oldest 18 July 2019); none of VROMI's 58 vehicles is marked yellow. The red marking is not explained, and it is applied inconsistently: four vehicles with 2025 expiry dates are not red, and two of them are marked yellow.
What we show
The site reports the red count as 'marked in red (inspection date passed)' and does not state that the vehicles are uninspected.
Every sheet is titled "Wagenpark Gov. St. Maarten 2025", while the footnote says "as of 2026". There is no sheet for Parliament or the special entities, although Parliament insures a "Parliament fleet" (Cg 9,532, NL 115) and has a "Vehicle fleet" capital project printed as "Cg 100" (read as thousands, like the other project boxes; EN 169); the Ombudsman's two vehicles are listed on the General Affairs sheet (NL 90). An internal working note and staff contact details are also printed on the sheets.
What we show
The fleet counts cover only the vehicles printed on NL 89-95.
In the leasehold (erfpacht) statistics, every 2026 figure equals 2025 (3,192 leases; 5,274,575.00 m2; annual ground rent NAf 6,853,647.00) except the longest remaining term (76.0 vs 75.0 years). The 2027 change column compares against this repeated 2026 column. The printed average area for 2021 and 2022 (1,874 and 1,847 m2) does not match total area divided by number of leases (about 1,634 and 1,632 m2).
What we show
Not used as a trend. The source does not say whether 2026 was updated.
Annual ground rent (canon) rises from NAf 6,853,647.00 (2026) to NAf 9,015,646.33 (2027), +2,161,999.33, while the number of leases falls from 3,192 to 3,151. The highest single ground rent rises from 429,492 to 1,376,148 (+946,656, equal to 43.8% of the total increase, computed). The table does not say which lease this is, whether it is the same lease in both years, or how the 2027 figures were derived.
What we show
Reported as printed. The English memorandum's long-lease revenue (Cg 7 million in 2026, 9 million in 2027) agrees after rounding (EN 163).
The table titled "Government-owned Companies" gives only share percentages, with no capital, book value, dividend or results. It also includes three minority stakes (Saba Bank Resources 28.20%, DC-ANSP 18.75%, Ontwikkelingsbank 7.08%). Some names contain typos ("throug", "Televies").
What we show
Share percentages are shown as printed; financial context comes only from the English memorandum (EN 178-179), cited separately.
The 2027 annex prints "Collective Sector entities 2025" / "Overview of Collective Sector Sint Maarten 2025", in a different typeface from the surrounding pages. Marven N.V. is listed as "(inactive)" while also shown as a 100% participation (NL 102).
What we show
The 14 entities are listed as printed. The source does not say whether the list changed for 2027.
The airport concession yields Cg 0 in 2025, 2026 and 2027. Its description says it was extended in 2012 "for another 15 years" without saying from when, so it could end in 2027 (2012 + 15) or 2040 (2 January 2025 + 15).
What we show
Shown as printed; the end date is left open.
The 2027 budget for the GEBE fee (Cg 8,042,014) equals 2026, although the English memorandum says the fee "is indexed annually" and that this leads to "an average annual increase of approximately Cg 270" (EN 16; no unit is given for the 270).
What we show
The NL 106 figure is used. The difference is not explained.
NL 106 shows harbour receipts of 0 in 2025; EN 16 shows Harbour 4 (Cg million) for 2025. Telecom: NL 106 shows 0 / 2,500,000 / 2,500,000 (2025-2027); EN 162 "BTP result" shows 0 / 2 / 2 and EN 16 "BTP" shows 2 / 3 / 3 (Cg million), so the two English tables also differ from each other. EN 162 shows a separate "Concession fees" line of 8 / 12 / 12 (Cg million, 2025-2027), which matches GEBE plus harbour on NL 106 (7.66 / 12.04 / 12.04 million, computed), so EN 162 agrees with NL 106 that no harbour fee was received in 2025. NL 106 prints no total; the four concessions add up to Cg 7,659,061 (2025) and Cg 14,542,014 (2026 and 2027) (computed).
What we show
The NL 106 figures are used for each concession. The differences are not explained in the source.
None of the 14 blocks on NL 112-116 prints a subtotal, and no year or currency is given for the "Jaar bedrag" (annual amount) column. The 368 rows add up to Cg 59,540,131.90 (computed). An empty, unlabelled column runs through every page, and NL 115 uses a different (blue, colour-scanned) template.
What we show
Amounts are treated as annual amounts in Cg, as the column name suggests. This is an assumption; computed totals are not official figures.
Examples: "CFATF (2024 Contribution)" and "Egmont Groep (20/21 Contribution)" (NL 113); "ESAS Technical Assistance Missions 2021", "Expenditure & Satisfaction Tourism Surveys 2024", "Maintenance and support fee for the year 2024 for MetLab2", "National Account 2023 actuals 2024 projections", "CPI tool for 2024-2025" (NL 114); "Rent main location (10.500$ in 2015)" and "approx. $ 4,500.00 pm 2022" (NL 116).
What we show
Shown as printed. Parts of the list appear to be carried forward from earlier years; the source does not say.
Gibson & Associates (TEATT) is listed with "p.m." (no amount) and Clean Communications copier maintenance (General Audit Chamber) with 0. Two TEATT rows show euro equivalents at different rates: 31,304 for EUR 16,000 (1.9565) and 79,043.90 for EUR 39,900 (1.9810). VROMI's numbering skips row 23. General Affairs row 11 shows a second figure, 1,592,024, inside the description.
What we show
p.m. is stored as amount null and the 0 row as 0; the printed Cg amounts are used for the euro rows.
Examples: an air-conditioning firm under "Cleaning Services" (General Audit Chamber), the exam bodies CXC and ETE and the academic-recognition centre NARIC under "Economic Support Services" (Education), newspaper subscriptions under "Educational Services" (Security Service), "Telem N.V." under "Other" (Integrity Chamber) but "ICT" elsewhere. Study allowances ("Studiebeurzen Toelage", Cg 3,500,000) are also in the list, although the line describes allowances rather than a contracted service. Labels are not standardised ("Other"/"other", "Financial Support"/"Financial Support Services").
What we show
Categories are quoted as printed in each subject.
In the 2025 column, the amount for the Foundation for Secondary Education reads '18,030.908' (NL 83), and the amount for Charlotte Brookson Academy (NL 84) also appears to use a point instead of a comma; at the scan's resolution the separators are hard to tell apart.
What we show
Read as whole guilders, 18,030,908 and 1,983,892; the column context confirms this.
The printed 'Totaal' covers existing loans only. The printed 'Totaal generaal' = 'Totaal' + the Stichting Overheidsgebouwen block + 'Totaal nieuwe leningen' (new loans). This was established by arithmetic; the table does not say so. Row sums match the printed totals to within 1-2 guilders (spreadsheet rounding). End-2027: Totaal generaal 1,077,513,322, interest 29,872,239 (NL 79).
What we show
Use the printed 'Totaal generaal' for headline debt figures. The 1-2 guilder differences are rounding.
Fourteen loan lines carry no amounts at all: eight expired 2020-2022 liquidity-loan lines (including the 2021 4th-tranche advance and the 2020-2022 refinancing), 'Bridge loan PJIA', 'PJIA Repayment of Bridge Loan', 'Sint Maarten 2018', two 'Herfinanciering' lines and the 2020 Sint Maarten 2010 tranche. Three more expired liquidity loans (2019; 2020 2nd tranche 1st part-payment; 2020 3rd tranche) show only nil balances (NL 78-79). Several names and dates are misprinted ('Refinancinng Bond', 'Jne-2021'), and rates are printed inconsistently ('3.75%', '4.50', '2.375').
What we show
These lines are kept in 'loans' with null amounts so the list matches the printed statement; they do not affect any total.
Establishment and budgeted vacancies match for every area (totals 2,854 and 232). The budgeted head-count does not: PARL 66 (NL 58) vs 97 (NL 66, fictitious places), AZ 293 vs 294 (NL 68), JUS 635 vs 636 (NL 70). In the NL 58 Justice row the education columns add up to 636, not the printed 635. Staat B's own education-level totals are 293 for AZ (NL 68) and 635 for Justice (NL 70), the same as NL 58.
What we show
Rows use the Staat B counts; 'filled' is Staat B's 'Current Personnel count'.
NL 100 prints a monthly grand total of Cg 1,335,998.00 (the 36 monthly rents add up to exactly this) and an annual total of Cg 16,031,975.99. 12 x 1,335,998.00 = 16,031,976.00.
What we show
Each rental's annual figure is its printed monthly Cg rent x 12; these add up to 16,031,976.00. rentalsTotal uses the printed 16,031,975.99.
How we checked
523
pages read
Both official documents, cover to cover: 186 pages of the English memorandum and 337 of the Dutch submission, including all 1,290 lines of the numbers book. Every figure is typed once, with the page it came from.
1,630
figures re-checked
A second reader, working separately, checked key figures against the page and caught 179 transcription slips, all fixed before publication. Comparing every English figure with the Dutch original found no differences.
3,180
sums checked
A validator re-adds every row, subtotal and total of the numbers book. In 34 places the printed book itself doesn't add up, and we report them instead of quietly fixing them. Where a department's printed total is higher than its own lines, the difference appears in the ledger as a visible “not itemised” line.
114
checks on every build
Each time the site is built, these checks confirm that the parts add up to the official totals, to the guilder for the draft budget law. If a single one fails, the build stops and nothing is published.
Last year's budget keeps its own receipts, checked against its own pages: the 2026 receipts in the archive.
