How the numbers are made

Every number on this site comes from the Penyata Kewangan Kerajaan Persekutuan (Federal Government Financial Statements), published each year by Jabatan Akauntan Negara Malaysia (JANM) and audited by the National Audit Department. The documents are public and freely downloadable from the JANM website.

Operating spending comes from Lampiran 1, Penyata Perbelanjaan Mengurus (Statement of Operating Expenditure). For each vote it shows, by programme, activity and object group, the original estimate (Asal), supplementary allocation (Tambahan/(Kurangan)), virement (Viremen), revised total (Jumlah) and actual spending (Sebenar). Variance (Varian) and utilisation (% Prestasi) are derived from those columns and not stored. The printed % Prestasi sometimes contradicts its own row; the site computes it instead.

Development spending comes from Lampiran 2, Penyata Perbelanjaan Pembangunan (Statement of Development Expenditure). Each row is printed across a two-page spread. The left page gives the project code, name, total cost and spending under the current Malaysia Plan up to the end of the previous year. The right page gives the year's estimate and actual spending, each split into direct (Cara Langsung) and loan-funded (Pinjaman), and plan spending to date. Within each row, previous plan spending plus the year's actual must equal plan spending to date, which also proves the two halves of the spread were paired correctly. 2021 is the first year of the Twelfth Malaysia Plan: its left column counts the Eleventh Plan to 2020 and its right column the Twelfth Plan, so there the year's actual must equal plan spending to date.

From 2020 to 2023 the statements also include Lampiran 3, Penyata Perbelanjaan Kumpulan Wang COVID-19: the COVID-19 Fund's programmes, each with its ceiling, the year's allocation and spending, and spending since 2020. The fund was financed outside the operating account, so its spending adds to operating and development spending.

Total federal spending is operating plus development, less the transfer from the operating account to the Development Fund (Kumpulan Wang Pembangunan). That fund pays for development spending, so the transfer would otherwise be counted twice. Operating, development and COVID-19 Fund spending for every loaded year match data.gov.my's published figures to the million, with one exception: for 2023 development, data.gov.my reports RM96,091 million and the statement prints RM96,089.7 million.

2020 is not loaded yet. Its operating statement is printed as drawn shapes with no text layer, so it can only be read by optical character recognition, which this site does not yet rely on.

Numbers are stored in integer sen (1 RM = 100 sen). Every total the documents print is stored alongside the rows and recomputed from them before a load is accepted. A load with even one mismatch is rejected. Where the PDF's text cannot be read literally (an empty cell, a misprinted code, text the page does not draw), a named rule handles that one pattern, every use is logged with its page, and the totals check still covers the result. Each number links to the PDF page it was read from.

The budget pages come from the Ministry of Finance's Anggaran Perbelanjaan Persekutuan 2027 (SHA-256 5caf1cc1a4416e21912e4a450f66add2e7d34ee684d8e10de4e5717be42457a9). Its summary pages give each vote's actual spending for two years, the current budget and the proposed one. Its detail tables give every operating vote's programmes and activities by object group, with established posts, and every development vote's projects with total cost and the proposed direct and loan-funded amounts.

• Operating detail: 2,354 of 2,354 printed totals equal the sum of their activity lines.
• Development detail: 85 of 85 group and vote totals equal the sum of their projects, and every vote agrees with the book's development appendix (Lampiran D).
• 136 of 136 vote totals in the detail tables equal the same votes in the summary pages.
• MOF's printed actuals agree with the audited statements for 272 of 272 vote figures, all within RM 1.45: MOF prints whole ringgit, and rounds either each figure or the parts it adds.

Activity codes are reused for different activities from one year to the next (B.64 020400 is Universiti Putra Malaysia in the 2025 statement and Universiti Sains Malaysia in Budget 2027), so the site compares budgets with spending by vote, where the budget restates past actuals itself, and never matches activities across documents by code.

The budget pages come from the Ministry of Finance's Anggaran Perbelanjaan Persekutuan 2026 (SHA-256 e2eaeedf711be21dff09ef4409882de25b3e48f936bc3c9c07f95dc8fac7cdf0). Its summary pages give each vote's actual spending for two years, the current budget and the proposed one. Its detail tables give every operating vote's programmes and activities by object group, with established posts, and every development vote's projects with total cost and the proposed direct and loan-funded amounts.

• Operating detail: 2,482 of 2,482 printed totals equal the sum of their activity lines.
• Development detail: 86 of 86 group and vote totals equal the sum of their projects, and every vote agrees with the book's development appendix (Lampiran D).
• 136 of 136 vote totals in the detail tables equal the same votes in the summary pages.
• MOF's printed actuals agree with the audited statements for 272 of 272 vote figures, all within RM 0.59: MOF prints whole ringgit, and rounds either each figure or the parts it adds.
• The book's original estimate for 2025 equals the audited statement's original estimate (Asal) for 53 of 53 operating votes.Below vote level the book restates 2025 in its own structure (Budget 2026, for example, spreads the 2025 public service pay adjustment, SSPA, which the statement keeps as an activity of its own, across other activities), so its activity figures for that year are not the ones Parliament approved.

Activity codes are reused for different activities from one year to the next (B.64 020400 is Universiti Putra Malaysia in the 2025 statement and Universiti Sains Malaysia in Budget 2027), so the site compares budgets with spending by vote, where the budget restates past actuals itself, and never matches activities across documents by code.

The budget pages come from the Ministry of Finance's Anggaran Perbelanjaan Persekutuan 2025 (SHA-256 0a96d0a4c468b701482c2b1d0b00a1770c6c0d24391c52bc5d5bf28fda6710e6). Its summary pages give each vote's actual spending for two years, the current budget and the proposed one. Its detail tables give every operating vote's programmes and activities by object group, with established posts, and every development vote's projects with total cost and the proposed direct and loan-funded amounts.

• Operating detail: 2,488 of 2,488 printed totals equal the sum of their activity lines.
• Development detail: 86 of 86 group and vote totals equal the sum of their projects, and every vote agrees with the book's development appendix (Lampiran D).
• 136 of 136 vote totals in the detail tables equal the same votes in the summary pages.
• MOF's printed actuals agree with the audited statements for 278 of 278 vote figures, all within RM 0.98: MOF prints whole ringgit, and rounds either each figure or the parts it adds.
• The book's original estimate for 2024 equals the audited statement's original estimate (Asal) for 42 of 53 operating votes. The other differences add up to exactly zero: the book restates the year in its own ministry structure, after money moved between votes when ministries were reorganised.Below vote level the book restates 2024 in its own structure (Budget 2026, for example, spreads the 2025 public service pay adjustment, SSPA, which the statement keeps as an activity of its own, across other activities), so its activity figures for that year are not the ones Parliament approved.

Activity codes are reused for different activities from one year to the next (B.64 020400 is Universiti Putra Malaysia in the 2025 statement and Universiti Sains Malaysia in Budget 2027), so the site compares budgets with spending by vote, where the budget restates past actuals itself, and never matches activities across documents by code.

The published 2020 Statement of Operating Expenditure (Lampiran 1) has no text layer, so its figures were read from the page. The PDF still draws every character as a vector outline, so the position of each character, and the pattern of digits, commas, periods and dashes in each cell, come from the PDF itself rather than from OCR.

Each amount cell is read by OCR at 15 settings. A reading counts only if it fits the cell's drawn pattern character for character, and the cell takes a reading that at least 8 settings agree on, with no rival reading held by more than half as many. Where no reading clears that bar, the cell is read by eye from the page image and registered with an image of the cell. Every printed total is then recomputed from its rows exactly as in other years: it either matches or is a misprint confirmed on the page and listed below. Programme and activity names are OCR readings; the totals check does not cover them.

2020: tesseract 5.5.3; glyph boxes from pdftocairo version 26.10.0 -svg.