Every year, in the last weeks before the middle of July, the diggers come out. Trenches open along the bazaar, gravel is tipped on the road, and blacktop goes down in the rain. It happens so reliably that it has a name, असारे विकास, Asar’s development, after Asar, the last month of Nepal’s fiscal year. Anshu, whose tea shop runs through these chapters, has watched it from her doorway: customers stepping over a ditch with their glasses, a road laid in a downpour and cracked by Dashain. (Anshu and her shop are made up; the trenches are not.)
Nobody chooses the monsoon because it is good weather for roadworks. The crews are racing a date. A budget is permission to spend for one year, and what isn’t spent by the end of Asar can’t be spent under it. The Auditor General counted when the federal government’s capital spending, its money for building, was spent in the year 2024/25 (2081/82): 13% of it in the first three months, 12% in the next three, 21% in the third quarter, and 54% in the last. To see why a government runs its year like that, and why it borrows to do it, start with what a budget is.
Three ways to spend
A budget is a government’s plan of what it will spend in the coming year and where the money will come from. Anshu makes one too, if only in her head. She pays rent, wages, milk and gas every month. Once in a while she buys something that lasts, a new stove or a fridge. And she pays back the loan she took to open the shop. Nepal’s budget sorts its spending the same three ways.
The first is current spending, चालु खर्च: what has to be paid again every year to keep things running. Salaries of teachers, soldiers and clerks, pensions, electricity, fuel, medicines, the allowances paid to the old and to widows, and the interest on the government’s loans all go here. So does the money the federal government passes down to the seven provinces and 753 local levels, which they then spend for themselves.
The second is capital spending, पुँजीगत खर्च: money spent once on something that lasts and goes on being useful, such as a road, a bridge, a school building, a hydropower plant, or the land under them. This is what Nepalis mean by development spending, the part of the budget that builds.
The third is the one that puzzles people, वित्तीय व्यवस्था, financing. It is neither running costs nor building. It is the money that settles past borrowing, the principal of old loans, the sum borrowed itself, falling due, plus loans the government makes to public companies and shares it buys in them. The interest on the government’s loans is not here; it is a cost of the year, counted in current spending. In 2024/25 the federal government paid Rs. 291 billion of principal and put about Rs. 41 billion into loans to and shares in public companies, which together make the Rs. 333 billion of financing in its accounts almost exactly.
That year the federal government spent Rs. 1,513 billion in all. Of every Rs. 100, Rs. 63 was current spending, Rs. 15 capital and Rs. 22 financing. Some Rs. 37 of every Rs. 100 of the current spending was money handed down to the provinces and local levels, so a share of what looks like running costs in Kathmandu becomes a school roof or a ward’s road elsewhere.
The law guards some of this spending more than the rest. Under the Constitution’s Article 118, the money to pay the interest and principal of the government’s loans is charged on the state’s main account, the Consolidated Fund, and parliament doesn’t vote on it at all. The lenders are paid without waiting for a vote.
Surplus, deficit, balance
When Anshu’s takings in a year are more than she spends, she has a surplus; when they are less, a deficit. A government is the same. A budget that plans to take in more than it spends is a surplus budget, बचत बजेट. One that plans the two to be equal is a balanced budget, सन्तुलित बजेट. One that plans to spend more than it takes in is a deficit budget, घाटा बजेट, and the gap has to be borrowed.
Nepal’s federal budgets are deficit budgets, year after year, and so are most countries’. There are two ways to measure the gap, and the budget gives both. The deficit itself is current and capital spending minus what comes in from revenue and from grants other countries give. In 2024/25 it was about Rs. 100 billion. But financing has to be paid too, Rs. 333 billion, nearly all of it old loans falling due. So the government needed about Rs. 433 billion beyond its revenue and grants, and it borrowed Rs. 362 billion of that, at home and abroad. Much of a year’s new borrowing goes on paying off earlier borrowing.
Is a deficit bad? Not in itself, any more than Anshu’s loan for the shop was. Chapter 17 showed one reason a government lets it grow: in a slump, its takings from tax fall and what it pays out in help rises without anyone deciding it, and that steadies the economy. Keynes went further and argued for spending more on purpose in a slump. The other side of his idea is a surplus in the good years, to pay the slump’s borrowing back, and at such times the surplus budget is the one suited to economic stability, आर्थिक स्थिरता.
What the borrowing pays for matters more. Nepal’s Financial Procedures and Fiscal Responsibility Act, 2076, allows a deficit budget, but its section 15 adds a rule a household could live by: no deficit budget may be presented to meet administrative costs. Borrow to build, that is, never to pay the clerks.
How a budget is made
Nepal has presented a budget since 2008 BS, when Subarna Shamsher read the first one on 21 Magh 2008, early in 1952. Today the steps are set down in the Constitution and in that Act of 2076.
The whole round has four stages: drafting the budget, बजेट तर्जुमा; approving it, स्वीकृति; carrying it out, कार्यान्वयन; and judging how it went, मूल्याङ्कन. The drafting starts months before anyone sees a budget. The government first fixes a three-year plan of spending, the medium-term expenditure framework, so that each year’s budget fits into the next two (section 6 of the Act). The National Planning Commission and the Ministry of Finance, through a committee led by the Commission’s vice-chair, estimate how much money the coming year can bring in from taxes, grants and loans, and set a ceiling on what can be spent (section 7). The Ministry of Finance sends each ministry its share of that ceiling, with guidance on what to spend it on (section 8). The Act sets the dates: the estimate by the end of Magh, the ceilings by 15 Phagun.
Each ministry then writes its proposal within its ceiling, project by project, and argues it out with the Planning Commission and the Ministry of Finance (section 9). The Ministry of Finance puts the whole budget together. It must pay what the government already owes before anything new, and the Act says it must take into account how much each body is able to spend, a line that will matter in a moment (section 10). Fifteen days before the budget, the government lays its principles and priorities before parliament, whose members have seven days to suggest changes (section 11). And before the budget it lays the Economic Survey on the table, the Ministry’s account of how the economy and last year’s budget actually went (section 14).
Then comes the one fixed date. The Constitution’s Article 119 says the finance minister shall present the year’s estimates of revenue and spending to a joint sitting of both houses of parliament every year on 15 Jestha, in late May. The estimates must show the previous year’s money for each ministry and whether it met its targets. The budget speech is read that afternoon. Parliament debates the spending, ministry by ministry, and passes it as an appropriation bill (Article 120), which becomes law when the President authenticates it. The fiscal year begins on 1 Shrawan, in the middle of July. While the bill is still being debated, parliament can pass an advance, at most a third of the year’s estimated spending (Article 122).
Under the Intergovernmental Fiscal Arrangement Act, 2074, the provinces must present their own budgets by 1 Asar and the local levels by 10 Asar. That is the making. Then come the spending, which the ministries carry out and the Financial Comptroller General’s office records, a review of progress halfway through the year, and finally the audit. The Auditor General, whose duty under the Constitution is to audit the accounts of every government office (Article 241), reports to the President each year, and the President has the report laid before parliament (Article 294). Much of what this chapter knows about where the money went comes from those reports.
Why development money stays unspent
Here is the line from section 10 again: the budget should take into account how much can be spent. Year after year, Nepal’s federal budget sets aside more for building than its ministries manage to spend.
In four of these five years, about three-fifths of the capital budget was spent. In 2025/26, a year of street protests in September and an election in March, the government cut the capital budget halfway through the year from Rs. 408 billion to Rs. 243 billion, and still spent only Rs. 191 billion, less than half of the first plan. Over the longer run the slide is plain. In 2018/19 (2075/76) the federal government’s capital spending was about 6.3% of everything Nepal produced; by 2024/25, the Auditor General reported, it had fallen to 3.7%, while financing, mostly old loans paid off, had risen from 4.0% to 4.8%.
Why can’t a government spend money it has? A World Bank study of 2025, which followed Nepal’s projects through every step, found something wrong at nearly all of them.
Too many projects are begun. Nepal approves far more than it can pay for, many of them poorly prepared, so the money is spread thin across all of them. At the pace of funding of recent years, the study reckoned, Nepal’s 17 national pride projects, the biggest of all, would take 41 years to finish.
The ground isn’t ready. Before a road can be built, the land under it has to be bought, often from hundreds of owners, with old records and disputes over the price. The study found that buying land takes two to three years on average, and permission to cut the trees in the way about two years. On one road it studied, buying the land alone took 35 months, almost two and a half times as long as planned.
Buying is slow. Choosing a contractor through the rules of a public tender took longer in Nepal than anywhere else in South Asia for the projects the World Bank pays for: 231 days on average. The rules make the lowest bid win, and the World Bank found contracts stalling under that rule; it asked for bids too low to be real to be screened out.
And the money comes late. The Ministry of Finance releases cash a quarter at a time, and offices that can spend only what has been released spend it late. Rigid rules for moving money from a stalled project to a working one, and officials who are often moved to other posts, slow things further. The study found that more than 40% of capital spending happens in the last month of the year. The Auditor General’s report on 2024/25 told the government to make its budgets realistic, with an eye on the money there will be and the capacity to spend it.
Each of these is somebody’s job, and most have been known for years. The budget for 2083/84 promised, once more, to change “the pace and rhythm of capital expenditure”. What a student can see from the road outside Anshu’s shop is the result: the blacktop laid in Asar is the money that came last, spent fastest, on work that had the least time to be done well.
Borrowing at home and abroad
What the government can’t cover from revenue and grants, it borrows. It borrows in two places, and they are different kinds of debt.
At home, आन्तरिक ऋण, it sells short-term treasury bills and longer development bonds, which banks and other investors buy, and smaller sums to people: savings certificates for citizens, and bonds for workers abroad. This borrowing is in rupees, so the rupee’s value doesn’t change what is owed. But every rupee the banks lend the government is a rupee they don’t lend to a business, the crowding out of Chapter 17. On the government’s rupee bonds the interest averaged 3.8% in 2024/25.
Abroad, बाह्य ऋण, it borrows mostly from the international banks set up to lend to poorer countries. In 2024/25, of the Rs. 130 billion it received in foreign loans, 46% came from the World Bank’s International Development Association and 40% from the Asian Development Bank; loans from single countries were under 5%. These banks lend to poorer countries largely on easy terms, and the interest on all of Nepal’s debt that year was Rs. 70 billion, about Rs. 2.60 for every Rs. 100 owed. But the loans are in foreign money, most of it in SDRs, the International Monetary Fund’s basket of the main world currencies, and in dollars. When the rupee falls, the same loan costs more rupees. In 2024/25 the rupee’s fall alone added about Rs. 66 billion to what Nepal owed abroad, without a single new loan.
Add up what is owed. By the middle of July 2025 the government’s debt was Rs. 2,674 billion, a little more than half of it owed abroad, by the reckoning of the Public Debt Management Office, which keeps the books. A year later it was Rs. 2,975 billion, 45% of the Rs. 6,600 billion that Chapter 13 found Nepal produced in 2025/26. In 2024/25 the government paid Rs. 362 billion to its lenders, Rs. 291 billion of principal and Rs. 70 billion of interest. Set against the federal government’s own revenue that year, that was about a third of everything it collected.
When debt builds, and when it sinks
So is 45% too much? A rule of thumb became famous in 2010, when two economists, Carmen Reinhart and Kenneth Rogoff, reported that countries whose public debt passed 90% of their output grew far more slowly than others; on their figures, rich countries above that line had shrunk on average. Leading policymakers in America and Europe cited the 90% line as they argued for cutting spending.
In 2013 Thomas Herndon, a graduate student at the University of Massachusetts Amherst, tried to reproduce their results and couldn’t. When the two sent him their working spreadsheet, he found that a formula had left out five countries, that some years of data for others had been left out, and that every country’s average counted the same however many years it spent in debt: New Zealand’s one year above 90%, when its output fell, counted as much as Britain’s nineteen. With his professors Michael Ash and Robert Pollin he worked it through again, in a paper the Cambridge Journal of Economics published. Rich countries with debt above 90% of their output from 1946 to 2009 had grown at 2.2% a year on average, not shrunk by 0.1%: slower than at lower debt, but not dramatically so. Growth didn’t fall off a cliff at 90%, and the link between debt and growth differed from country to country and decade to decade.
That leaves the plain questions that decide whether a loan builds or sinks, the same ones Anshu would ask. What does it buy? A loan for a stove that sells more tea pays for itself; a loan to pay this month’s rent only moves the problem to next month. Debt that builds roads, power and skills raises what the country can produce, and with it the taxes that repay the loan. Debt that pays running costs adds nothing to repay it with, the danger behind the Act’s rule against borrowing for administration. What does it cost, and in what money? Cheap, long loans are easier to carry than dear, short ones, and a loan in rupees can’t grow when the rupee falls. And does the economy grow faster than the debt? If output grows faster than the interest piles up, the debt shrinks against it even without being repaid.
On those tests Nepal sits in between. Nine-tenths of its foreign debt is owed to the international development banks, and the International Monetary Fund’s review of Nepal’s debt in August 2025 judged both its public and its foreign debt at low risk of distress, which means low risk of the country being unable to pay; in 2026 it still called the risk low. For the foreign debt it needed a judgment call: Nepal sells so little abroad that its usual test flashed a warning, and the Fund set the warning aside because money sent home by workers abroad brings in more than three times as much foreign currency as exports do. But the debt has grown faster than the economy, from 22% of its output in 2014/15 to 45% in 2025/26, while the share spent on building has shrunk. In 2024/25 the government borrowed Rs. 235 billion at home and paid Rs. 244 billion of earlier domestic loans back. Borrowing only sinks a country when what it bought can’t pay its way, and an unspent capital budget is borrowing for roads that weren’t built.
A budget is a year’s promise in numbers: what will run, what will be built, and what is owed. Nepal’s is made in steps the law sets out, from the estimate of what the year can bring to the speech on 15 Jestha, and it sorts every rupee into current spending, capital spending and the loans of the past. Its deficits are borrowed at home and abroad, and its debt is no longer small, though most of it is cheap. What decides whether that debt builds the country is less its size than what it buys, and there the budget’s weakest part is not the borrowing at all. It is the roads in the plan that aren’t built in the year, and the ones that are, in a hurry, in the rain.
Found a mistake? Please tell us.
