Route · /states

The states, one page each

16 states publish a budget this register can read, and every line of every one of them is on this site. 6 more publish one it cannot. Both facts are below, in the same table shape, because a state that cannot be read is a finding and not an omission.

States read

of 22 surveyed

Budget lines held

40,241

as each state published them

Funded and unlisted

1,848

across all 16

Together

3,48,377

crore of provision

The states this register can read

Ordered by how many schemes each one funds that no national portal lists · open a state for its whole book

State	Lines in the book	Clear the bar	Funded and unlisted	₹ crore	Precision	Recall	What its book is
Tamil Nadu	6,220	411	295	29,540	92.9%	48%	Fifty-five bilingual demand books with English in its own column and object heads against every line.
Telangana	2,039	275	267	85,087	98.9%	62%	The Pragathi Paddu is a plain English scheme table; the current cycle is on the IFMIS portal and not on the Finance Department's own budget page.
Uttar Pradesh	5,831	184	184	39,188	94.0%	44%	The only state here built WITHOUT clearing the English test. Its budget has no English. myScheme lists its schemes in romanised Hindi, so the join is a change of script. 111 of its 4,198 printed totals do not reconcile.
Punjab	2,961	144	140	22,059	97.2%	59%	Gurmukhi cannot share a codepoint with Latin, and the books check their own arithmetic at every level of the head of account.
West Bengal	9,024	217	138	12,526	91.2%	29%	English throughout the detailed demands, keyed on the head of account.
Jharkhand	852	142	126	29,404	95.8%	73%	It prints STATE SCHEME <name>(<code>) in English capitals, and the Hindi and English are separated by a slash.
Haryana	970	117	107	15,156	98.3%	52%	The Plan Memorandum names 970 schemes where the national portal lists 249.
Karnataka	969	104	99	19,728	94.2%	38%	The Gender, Child and SCSP/TSP books print no totals at any level, so the reading has nothing to check itself against.
Kerala	2,629	98	96	1,089	91.8%	26%	The Annual Plan statements carry a scheme code, an English column and Unicode Malayalam.
Delhi	1,578	97	85	14,517	94.8%	51%	The whole scheme budget is one plain English file. Its Grand Total does not close: it prints 62,55,000 lakh, and the rows sum to 64,55,956, 3.2 per cent more.
Maharashtra	1,956	95	80	15,700	97.9%	28%	It publishes an English EDITION of a Marathi book rather than a bilingual one, which is the single most useful thing a state does for anyone reading it.
Uttarakhand	2,324	96	74	1,230	99.0%	34%	Volume 5 names 2,302 schemes against the portal's 446, and 806 of its 4,492 printed totals do not reconcile against the rows beneath them.
Andhra Pradesh	552	90	69	47,600	96.7%	36%	Plain English and a single script, and the Gender Budget prints no totals to check a reading against.
Tripura	134	34	32	2,477	100.0%	62%	Its 134 state schemes and the 74 centrally sponsored ones beside them come to 208. DBT Bharat counts 209. No other state document here comes this close to a national count.
Chhattisgarh	574	29	29	1,344	96.6%	24%	Built from the WRONG BOOKS, and its own page says so. Its 44 department scheme books hold 2,562 rows and set every name in the Chanakya legacy font. That font maps its codes to Latin glyph names and draws Devanagari outlines in the slots. So neither the font nor anything else in the PDF can decode the names. They fail the text test, and this register does not publish them. Instead it reads the Outcome, Gender, Youth and Child books that 33 departments publish. These are in Kruti Dev, an ASCII font. Its table is checked against the state's own Unicode department index. Like Uttar Pradesh, the books are Hindi only. So the state does not clear the English test, and the join is a change of script, not a translation.
Odisha	1,628	29	27	11,734	96.6%	25%	Forty-four demand books with English above Unicode Odia and a printed total at every level.

Do not rank the states by the fourth column It is a floor set by how much evidence a state prints, not a measure of how much it hides. Kerala publishes 2,629 plan lines and this register will stand behind 96 of them; that is a fact about how much the Kerala plan document proves, not about Kerala. Precision is a count in every row, recall is measured on each state’s stratified sample alone, and both are explained on the state’s own page.

The states it cannot

Surveyed against the same five tests and stopped at one of them · DBT Bharat counts 1,223 state schemes between them and myScheme lists 1,368 · this register can read none of either

A machine cannot read these budgets, and they are the most important thing on this page. A citizen of each state cannot check the budget either. The cause is how the state saved the file.

Rajasthan

stops at bounded · 3 of 4 tests asked

191DBT Bharat counts

159myScheme lists

···this register can read

The Output-Outcome Budget is in English. It breaks names across lines mid-word with no hyphen, and geometry alone cannot undo that. It prints no totals to check a guess against.

What would change it. Print the totals. A machine could guess how to rejoin a broken name, but a guess that nothing can check is not a reading.

Gujarat

stops at bounded · 2 of 3 tests asked

388DBT Bharat counts

643myScheme lists

···this register can read

The Outcome Budget puts the name, the head of account and sometimes the target in one column, with no separator. The column boundaries move between pages. No rule can find the end of a name.

What would change it. Put the head of account in a column of its own, or any separator at all between it and the name. Nothing else about the document has to change.

Bihar

stops at text · 1 of 2 tests asked

189DBT Bharat counts

114myScheme lists

···this register can read

It converts its budget to vector curves before it publishes it. Its own file name is Demands For Grants Curve.pdf. Its 108 pages yield 108 extractable characters.

What would change it. Publish the same book with its text layer intact. The file is called Demands For Grants Curve.pdf, and the curves are the whole problem. The state typeset the document from text, then flattened it into drawings.

Assam

stops at list · 0 of 1 tests asked

127DBT Bharat counts

87myScheme lists

···this register can read

Its seven documents for this cycle include no scheme list. The FY 2024-25 Gender Budget is a good table and is two cycles behind what this register publishes.

What would change it. Publish a scheme list for the current cycle. The FY 2024-25 Gender Budget is a good table, so this is a question of when and not of how.

Himachal Pradesh

stops at list · 0 of 1 tests asked

175DBT Bharat counts

75myScheme lists

···this register can read

It publishes thirty documents and serves none of them: every link is a __doPostBack that 302s back to the portal home.

What would change it. Serve the thirty documents at addresses. They exist and the portal lists them; it just will not hand any of them over.

Madhya Pradesh

stops at list · 0 of 1 tests asked

153DBT Bharat counts

290myScheme lists

···this register can read

NOT the font any more. It is set in KrutiDev010, and parse/krutidev.py decodes it. A page of Volume 9 came out as fluent Hindi with no change to the decoder. Its demand books print a scheme code in brackets, a name and four money columns, so they would parse. What stops it is the CYCLE. Its latest published budget is 2025-26, and this register publishes 2026-27. The finance site's year filter offers 2026-27 but returns no documents for it. Every naming variant of the volume and demand book URLs gives a 404. So it stops at the FIRST test, which asks for a scheme list FOR THIS CYCLE, and the later tests are not asked. It would pass three of them. This note records that, but the score does not: a test the state was never asked cannot count as cleared.

What would change it. Publish the 2026-27 books at the same URLs the 2025-26 ones are already at. The format is readable; only the year is missing.

The two numbers above a refusal are not this register’s, and they disagree When a machine can read a state’s budget, this register is a third source. It can then say which of the other two is closer. For these states that is not possible. Only DBT Bharat’s count and myScheme’s remain, and they disagree about every state. myScheme lists 643 schemes for Gujarat and DBT counts 388. For Himachal Pradesh the numbers are 75 and 175. Each budget holds the number that would settle the question, and a machine cannot read it.

A refusal is about the DOCUMENT. This register does not accuse any of these states of hiding anything, and it makes no claim about what the documents hold. If a machine cannot read a budget one scheme at a time, it cannot read the budget in any other way. This register does not guess at the contents.