Walk through a hill village in Gorkha or Baitadi during planting season, and you will notice something: fewer hands in the fields than there used to be. A son is in Qatar, a daughter-in-law is in Kuwait, and the previously double-cropped terraces are now planted once, if at all. The house has a new tin roof and a motorcycle parked outside, paid for by money wired home every month. But the fields around it are quietly doing less than they used to.
This is the story most people tell about Nepal’s two biggest economic facts sitting side by side. Remittances make up more than a quarter of the country’s entire economy, among the highest shares anywhere in the world. Agriculture, meanwhile, still employs more than six in ten workers, and yet its productivity has barely moved in years. The obvious explanation is that migration hollowed out the farms, the workers left, thus the yields fell.
I wanted to know if that story actually holds up once you look at the whole country, district by district, rather than just at the villages where it’s most visible.
Building a fuller picture
Using government census data on migration and labor, matched with the agriculture ministry’s district-level crop data, I compared 75 districts within Nepal in the dataset on two things: how migration-dependent each district is, and how its cereal yields compare to the national picture. To capture migration dependency properly, I didn’t just look at how many people had left. I built a fuller score that also weighs how many households have a member abroad, and the education level of the departing workers, since a district that has lost its most skilled young people is a different story from one that has lost mostly unskilled seasonal labor. I built a second score for something related but distinct: how structurally vulnerable a district is, combining migration with poverty, literacy, and human capital indicators. (Full data and code.)
Cereal yields across the country ranged enormously: from just over 1 tonne per hectare in Humla, tucked in the far northwest mountains, to more than 5 tonnes per hectare in Bhaktapur, in the fertile Kathmandu Valley. Household migration rates ranged just as widely, from under 10% of households in some districts to over 40% in others.
The first-pass answer looked like it confirmed the common story: districts with more structural vulnerability tended to have lower yields. But something else stood out at the same time, something that didn’t fit the simple version of the story at all. Districts with higher migration dependency, on their own, were not showing lower yields. If anything, they leaned slightly the other way.
The real pattern: geography, not migration
Digging further, the clearest pattern in the data wasn’t about migration at all. It was about location. A district’s farm output tracked its neighbors’ farm output far more closely than it tracked anything about who had left to work abroad. Districts in the flat, irrigated Terai plains did well, whether or not migration was high, because flat land, good soil, and existing irrigation infrastructure make farming productive almost regardless of who is working the fields. Districts in the steep, remote hills and mountains did poorly, whether or not migration was high, because terracing, poor road access, and thin soils cap what farming can produce even with a full workforce.
In other words, Nepal’s yield map looks less like a map of who has migrated and more like a map of altitude and terrain.
That single insight reframes the whole “vulnerability” story. Once geography is accounted for, the poverty-and-vulnerability side of the picture mostly stops predicting yields on its own. However, this doesn’t mean vulnerability is irrelevant. Rather, it means that in Nepal, vulnerability and difficult terrain are closely intertwined almost completely: the same mountain districts that are poor, have lower literacy, and have thin human capital are also the districts where farming is structurally hard no matter who is present to do it. Migration dependency, on the other hand, kept an independent relationship with yield even after accounting for geography, though not in the direction the popular story assumes.
Four kinds of districts, not one story
Sorting every district by its migration level and its yield produces four distinct groups. (You can explore the classification district by district in the interactive atlas.)
There are 11 districts concentrated in the mid-hills of Gandaki and the far west that display the combination of high migration with weak yields. These are the places the “migration hollowed out our farms” story is really describing, and terrain is working against them at the same time.
In contrast, 16 districts combine high migration with strong yields. These districts, mostly in parts of the Terai, show that heavy outmigration and healthy farms can coexist just fine when the underlying land is good. This is the group the simple story completely misses, and it’s proof that migration alone doesn’t doom agriculture.
Meanwhile, 28 districts show weak yields despite low migration, mostly for the same terrain reasons as the 11 districts with high migration but low yields. Similarly, 20 districts show low migration and decent yields.
Six districts, including Baitadi, Gorkha, Salyan, Dailekh, Panchthar, and Bajura, sit right on the edge between categories, close enough to the migration threshold that a small policy nudge could shift them either way.
Why this matters for policy
If migration were really the cause of Nepal’s stagnant farms, the fix would be to bring migrants home, or to pour remittance money into agriculture through savings schemes and cooperatives. But if geography is the real constraint, that fix won’t reach the districts that need it most.
The 11 districts that fit the regular narrative, however, do not need programs aimed at reversing migration. Instead, they need mechanization support suited to steep terrain, irrigation investment where it’s geographically possible, and cooperative extension services built around hill farming, not flatland farming. The 16 districts where migration coexists with strong yields are proof that this is achievable if the geography cooperates. And the 28 districts, where migration is actually low, need basic infrastructure, roads, irrigation, and market access, far more than they need anything remittance-related at all.
Treating all of rural Nepal as one undifferentiated “migration problem” risks sending the same blunt policy tool to districts with completely different underlying causes. A hill district and a Terai district with the same migration rate can have entirely different farming futures, and the current policy conversation rarely makes that distinction.
What this doesn’t prove
This is a snapshot of one year, comparing districts to each other rather than tracking any single district over time. It shows an association, a pattern that holds up across the country, not a proven cause-and-effect chain. Confirming that would need to follow the same districts over several years as migration levels change. It’s also worth noting that districts near the Indian border likely undercount migration to India specifically, since that flow is harder to track than migration to the Gulf or Malaysia; excluding those districts from the comparison didn’t change the overall picture, which is reassuring but not conclusive.
What the data does support is a simple reframing: Nepal’s remittance economy and its agricultural stagnation are both real and both large, but treating the second as a straightforward consequence of the first oversimplifies a problem that is, underneath it all, mostly about land, terrain, and where investment has and hasn’t reached.
Prasanna Thapa is a recent graduate of Kathmandu Model Secondary School. He is interested in economics, mathematics, and technology, particularly in how economic and social issues shape people's lives in Nepal. His work explores topics such as development, agriculture, migration, and access to financial services, and he enjoys using numbers and data to better understand these challenges.
