Lakh and crore are usually introduced as the South Asian number system, as though the region shared one ladder that simply continues past crore into arab, kharab and beyond. Converters print the whole sequence. Explainers list it.
But take it to the finance ministries of four countries that inherited it, and no two of them use it the same way. India stops at crore and compounds it. Nepal keeps climbing into the larger terms. Pakistan switches to “billion” where India would start compounding, and Bangladesh uses crore without ever compounding it.
India stops at crore, and compounds
Official Indian usage tops out at crore. Above that it does not reach for a new word — it multiplies the one it has, producing "lakh crore" for 10¹².
The Union Budget documents for 2026-27 are unambiguous. Crore and lakh appear throughout; "lakh crore" appears repeatedly. Arab, kharab, neel, padma and shankh appear zero times. Across four documents, not once.
That reframes a complaint often made about conversion tools — that they omit arab and kharab. There is no official Indian authority to omit. Those words belong to Hindi and Urdu vernacular, not to the register the Budget is written in.
Nepal does use them, and the proof is arithmetic
Nepal, by contrast, does continue up the ladder into the higher terms. The proof comes from Nepali institutions that publish the same documents in two languages.
The national statistics office's national accounts for 2025/26 gives gross output in Nepali as ९९ खर्ब २२ अर्ब ९० करोड. Its own English edition gives the same figure as 9,922,906 million rupees. Read kharab as 10¹¹, arab as 10⁹ and crore as 10⁷ and the Nepali expression computes to 9,922,900 million — matching to the crore, with the remainder being sub-crore rounding. Four figures in that one passage reconcile the same way.
The evidence here is arithmetic rather than lexicographic. One institution states the same quantity twice, and the two statements agree only if arab and kharab carry those values.
The finance ministry's budget speech shows the register split just as plainly: अर्ब appears seventy-nine times in the Nepali text, and the official English translation contains the word "billion" seventy-eight times and no arab, crore or lakh at all. The central bank's Nepali release captions its tables "amount in arab" and its English edition uses billion throughout.
There is internal variation, too. The central bank and the statistics office compound upward into kharab; the budget speech does not, writing 2124 अर्ब rather than breaking it into kharab and arab. Two conventions, in the same country and the same year.
Pakistan leaves the ladder exactly where India compounds it
Pakistan inherited the same vocabulary and its official English does not use it.
The Economic Survey for 2025-26 runs to several hundred pages. Across the whole document, "billion" appears hundreds of times, "million" more, "trillion" dozens — and lakh appears twice, crore once, arab and kharab never. All three of those lakh and crore occurrences sit inside a glossary rather than in the body.
The glossary is telling. It defines one lakh as a hundred thousand and one crore as ten million, then jumps straight to "one billion — one thousand million". Pakistan's own official ladder contains no arab and no kharab. It stops where India's stops and then jumps to the English scale rather than compounding.
This difference in official usage changes how large numbers are written. Pakistan states its GDP as roughly 126.9 trillion rupees. An Indian budget document expressing the same quantity would write 126.9 lakh crore.
Bangladesh uses crore and never compounds
Bangladesh takes a third position. Crore is everywhere — the central bank's monthly indicators caption tables "BDT in crore" while giving dollar figures in millions — but the "lakh crore" compound India relies on does not appear.
Instead Bangladesh simply counts crore past a hundred thousand. A figure printed as 684,575.30 crore taka is about 6.85 trillion, and an Indian document would render it 6.85 lakh crore. The budget statements do the same in Bengali, captioned "amounts in crore taka".
The classical ladder does not settle it either
Appealing to Sanskrit for the "real" values does not settle the matter; the lexicography is inconsistent.
The standard Sanskrit-English dictionary gives arbuda as ten million, then adds "or a hundred millions"; a compound built on it implies the larger reading. Kharva is given as either 10¹⁰ or, on another authority, one followed by thirty-seven zeros. Śaṅkha is glossed as both a hundred billions and a hundred thousand crores, which are different numbers.
The classical sources disagree with the modern ladder and with each other. Lakh and crore are solid; the terms above them are not stable even in the tradition they come from, which is part of why only one of these four countries uses them officially.
One etymological aside, since it circulates widely: arab as a numeral has no connection to the Arab people. The numeral is Indo-Aryan, and the Arabic word is a separate lexeme with a different consonant. The resemblance is a coincidence of transliteration.
The grouping is a practice without a specification
The other half of the system is the comma placement — 10,51,953 rather than 1,051,953. Across a budget document we counted twenty-two numbers in that grouping and none in the Western pattern, so the practice is not in doubt.
We could not find a written standard mandating it — no bureau specification, no government style manual. The practice appears to be a convention enforced by consistency, not by a written rule, a fact worth remembering before citing "the Indian standard" for it.
If you need to move between these notations, our currency notation converter handles lakh and crore — though it is worth knowing that it renders 10¹² as 100000 crore rather than the "1 lakh crore" every Indian budget document actually uses, which is a real difference in idiom rather than in value.
A methodological warning that nearly produced the opposite finding
A methodological note for anyone checking this work: one detail nearly led us to the opposite conclusion.
Nepal Rastra Bank's Nepali documents are not encoded in Unicode. They use a legacy typeface encoding in which the bytes for अर्ब are not the Devanagari codepoints for those letters. Searching such a document for अर्ब returns zero results, which wrongly implies Nepal does not use the word.
The word is there. It is stored under a different encoding, and only appears once the bytes are mapped back. A search returning nothing is only as reliable as the assumption that the text is encoded the way you expect.
Where this comes from, and what will date it
The Indian counts are from four Union Budget documents for 2026-27 as published by the budget portal. The Nepali evidence is from the national statistics office's national accounts, the finance ministry's budget speech in both its Nepali original and its official English translation, and the central bank's macroeconomic situation reports in both languages. The Pakistani figures are from the Economic Survey and the bureau of statistics. The Bangladeshi figures are from the central bank's monthly indicators and the finance ministry's budget statements.
We should state three gaps in this research. The Reserve Bank of India could not be reached; an RBI publication using arab would weaken our finding for India. That is the strongest outstanding test of this guide, and we are naming it ourselves. The State Bank of Pakistan is behind a firewall that refused every attempt. Pakistan's Urdu-language official documents could not be read at all, because the PDFs are built from subsetted fonts with no recoverable character mapping. Whether Urdu officialese keeps the older words is unknown to us. The English register is established; the Urdu one is not.
A budget in any of the four countries changing register would date this. The Indian finding in particular is a claim about current documents, and it is falsified the moment a Budget prints arab — which is exactly how it should be checkable.
