When Nepal promulgated its federal constitution in 2015 and held its first subnational elections in 2017, the ambition was transformative. Power would flow from Kathmandu to seven provinces and 753 local governments: municipalities, rural municipalities and metropolitan cities. These were closer to citizens and better placed to understand their needs. The logic of fiscal federalism is elegant: local governments know local problems; give them resources and let them solve those problems.
The reality has proven considerably messier. Budget under-execution, declining grant allocations and a cascade of audit irregularities have caused observers to question whether federalism is working. But many of the failures diagnosed as political or institutional have a simpler, less noticed cause: Nepal's subnational governments are making decisions with barely any reliable data to inform them.
The gap between responsibility and revenue
Under Nepal's constitutional framework, spending responsibilities are broadly shared across three tiers. Local governments handle health posts, primary education, local roads and agricultural extension. Provinces manage secondary schools, provincial roads and some social programmes. Yet the power to raise revenue has barely moved from the centre.
In FY2024–25, the federal government accounted for 92.1% of all government revenue collected, while provinces and local governments together managed just 7.9%. On the spending side, the picture is more balanced: local governments executed 24.6% of consolidated public expenditure, while provinces spent 9.8% and the federal government 65.6%. The mismatch is structural and large.
This fiscal gap is filled by intergovernmental transfers: four categories of grant that flow from Kathmandu to the provinces and on to local governments. The transfers are constitutionally mandated and represent the lifeblood of subnational finance. But they depend on data that, in large part, does not adequately exist.
The fiscal equalisation grant is the most important unconditional transfer. It is meant to be allocated according to each government's expenditure needs and revenue capacity. In practice, computing these figures requires reliable subnational revenue data, current population estimates, and some measure of the cost of delivering services in different localities. Nepal has none of these in a form adequate for the purpose.
Grants that miss their mark
The data problem manifests most visibly in how conditional grants are spent. A study of four federal ministries found that conditional grants are supposed to fund specific local development projects. They are overwhelmingly diverted to recurrent costs. In FY2024–25, 78.21% of conditional education grants covered teacher salaries, leaving infrastructure, digital learning and training badly underfunded. Without data on local service-delivery gaps, it is difficult to set grant conditions that direct spending toward where it is actually needed.
The fiscal equalisation grant's share of the federal budget has been shrinking. In FY2018–19 it represented 10.5% of the federal budget; by FY2023–24 this had fallen to 8.3%, and by FY2024–25 to 7.9%. Meanwhile the Financial Comptroller General's Office (FCGO) cut the third-quarter instalment by more than two-thirds, from 25% to just 13.24%. The cut was blamed on fiscal pressure. The National Association of Rural Municipalities has condemned the decision.
Underspending as a data symptom
The most tangible consequence of weak data infrastructure is budget underspending. The World Bank's Nepal Fiscal Federalism Update 2024 found that provincial governments underspent their budgets by 34% in FY2022–23, with local governments underspending by 24.4%. In FY2024–25, utilisation improved marginally. Provinces reached 71.3% and local governments 76.4%, but remained well below complete execution.
This is usually attributed to weak procurement capacity or administrative bottlenecks. Both are real. But a neglected explanation is that budgets are poorly calibrated to actual local needs in the first place. Without data on the spatial distribution of poverty, infrastructure deficits or demographic pressure, local governments must guess at what to budget for. Guessing poorly leads to budgets that either over-allocate to programmes that cannot be absorbed, or under-allocate to genuine needs.
What needs to be done
The IMF's 2023 Article IV consultation recommended establishing a fiscal risk monitoring system for subnational governments by April 2024. This structural benchmark would require, at minimum, a consolidated and timely view of subnational finances. The World Bank's 2024 Fiscal Federalism Update called explicitly for a 'consolidated public financial management performance database that includes data from the subnational levels.' Neither has been fully delivered.
The reforms most needed are unglamorous: harmonising chart-of-accounts coding across the three tiers so that expenditure data are comparable; building a consistent local own-source revenue register; and training ward-level officials in basic data recording. The census, now that it has been conducted along federal lines, should be made a live resource for grant formulae rather than a decade-old baseline. The National Natural Resources and Fiscal Commission, which recommends the formula for fiscal equalisation grants, has consistently called for better data to underpin its calculations.