Living in a Ghost Town: The Geography of Depopulation and Aging

Yves here. Some of the mill towns, or towns in which I lived that had mills as significant employers, have shrunk even though they are not yet ghost towns. Escanaba, MI had a population of over 15,000 in the 1970s and is now below 13,000. Chillicothe, Ohio had a population of 29,000 in the 1970s and is now under 22,000. And as for aging, it is hard to beat Bailey Island, Maine, where my father’s family lived from when it was first settled. The median age is 59. Cranky Yankees!

By Elisa Giannone, Junior Researcher Center for Research in International Economics (CREi): Yuhei Miyauchi, Assistant Professor of Economics Boston University; Xinle Pang, Assistant Professor of Economics University At Buffalo; Yuta Suzuki, Antai College of Economics and Management Shanghai Jiao Tong University; and Nuno Miguel Marques da Paixao, Senior Economist Bank of Canada. Originally published at VoxEU

Policy debates about aging and shrinking populations often treat the problem as a national one. This column presents evidence from Japan showing that it is also profoundly spatial: aging and depopulation hit rural regions first. Outmigration from shrinking regions today reduces births tomorrow, which makes the origin region older still. However, policies that reduce spatial inequality also reduce aggregate efficiency, presenting a trade-off that leaves policymakers with the challenge of choosing among different goals.

Governments are again asking how much to invest in places that are aging, shrinking, and losing young people. Japan has restarted its regional revitalisation agenda, explicitly linking regional policy to low fertility and population aging (Cabinet Secretariat of Japan 2024). Spain has mobilised more than €10 billion across 130 measures to address the ‘demographic challenge’ (Government of Spain 2021). Italy’s National Strategy for Inner Areas (OpenCoesione 2026) and the European Commission’s long-term vision for rural areas (European Commission 2021) similarly put essential services, connectivity, and local development at the centre of policy.

The central question is not simply whether shrinking places deserve support, but what such support can achieve and at what cost. Economists warned that shrinking municipalities can face declining amenities because public services exhibit scale economies (Heinemann et al. 2007), and that population aging can weigh on growth (Kotschy and Bloom 2023). The missing link is geography: who leaves, who stays, and how these choices reshape the fiscal and economic base of local communities.

In Giannone et al. (2026), we study this question using Japan as a laboratory. Japan is useful not only because it is one of the world’s oldest and fastest-shrinking countries, but also because its demographic future foreshadows problems already visible in parts of Europe and East Asia. By 2015, 26% of Japan’s population was aged 65 or older, and the share is projected to reach 37% by 2050. Japan’s age wave can be considered a warning case for the social and fiscal pressures associated with population aging (Stawasz et al. 2018). But these national figures hide a striking spatial pattern. Some rural municipalities already have elderly shares close to 50%, while large metropolitan areas remain younger and denser.

The first fact is simple but important: depopulation and aging are not evenly spread across space. Between 1980 and 2010, Japanese municipalities that were already relatively old lost population and aged faster. Municipalities that were younger in 1980 tended to grow and remain younger. Figure 1 makes this historical divergence visible.

Figure 1 Historical demographic divergence across Japanese municipalities between 1980 and 2010

Notes: Older municipalities in 1980 lost population and aged further, while younger municipalities tended to grow and remain younger.Source: Giannone et al. (2026), Figure 4.

It also highlights an important timing point: Japan’s aggregate population decline begins only around 2010, but the regional divergence in aging and depopulation was already well underway long before then. This divergence is not just a story of different fertility rates across places. It is also a migration story. Young adults continue to move toward large cities, especially Tokyo, and their children are then born there rather than in the places their parents left. The result is a cumulative demographic mechanism: outmigration today reduces births tomorrow, which makes the origin region older still.

This mechanism matters because local economies are built around local scale. Shops, clinics, schools, childcare providers, public transport, and municipal offices all depend on enough users and taxpayers to remain viable. Using municipality-level data and an instrumental-variable strategy based on historical migration links, we find that areas losing working-age population experience declines in several measures of local amenities, including retail and medical services. At the same time, the per-person cost of local public spending rises. A road network, a school building, or a municipal office does not become proportionally cheaper when the population falls. This is the fiscal arithmetic of the ghost town: fewer residents, fewer workers, fewer children, and higher costs per person.

To study where these forces may lead, we build a dynamic spatial model in which people of different ages choose where to live, taking into account wages, housing costs, amenities, migration costs, pensions, and taxes. Wages and amenities can rise with local population because dense places support productivity and services; housing costs rise when many people want to live in the same place; and local public-service costs are higher per person in smaller places. The model is calibrated to Japanese data and then used to project future regional outcomes.

The baseline projection is sobering. Tokyo’s share of Japan’s population rises from about 10% in 2015 to roughly 26% over the next two centuries. The combined population share of the five oldest prefectures – Kochi, Shimane, Tokushima, Tottori, and Yamagata – falls from about 3% to below 1%. Their elderly share approaches 60%, while Tokyo’s remains close to one-third. When we shut down internal migration or hold fixed the spatial distribution of newborns at its 2015 level, much of this regional divergence disappears. In other words, the future geography of aging is not determined mechanically by national fertility and mortality alone. It depends critically on the spatial choices of younger cohorts and on where the next generation is born.

Figure 2 Projected population shares and elderly shares for Tokyo and the five oldest prefectures

Source: Giannone et al. (2026), Figure 9.

This finding connects aging to a broader urban-economics literature on agglomeration and regional divergence (Moretti 2012, Diamond 2016). Large cities can offer higher productivity and more amenities, but also higher housing costs. In our simulations, the productivity and amenity advantages of Tokyo dominate the congestion effect from housing. As more people concentrate there, the city becomes relatively more attractive, while declining regions lose scale. This reinforces regional inequality in flow utility, especially for working-age residents. The elderly are partly insulated by pensions, which are national rather than local, but they still depend on local services and amenities.

There is, however, a difficult twist. The same reallocation that worsens spatial inequality can raise aggregate efficiency. Concentrating more people in high-productivity, lower-cost locations raises average labour income and reduces the national fiscal burden of providing local public services. Preventing migration, or implicitly freezing the geography of births, keeps more people in smaller and less productive regions where public services are more costly per person. The policy problem is therefore not a simple choice between ‘good’ support for declining regions and ‘bad’ laissez-faire abandonment. It is an equity-efficiency trade-off.

Place-based policies sit exactly on this margin. Recent work on place-based industrial policy shows that subsidies may have important spillovers and only modest effects on regional inequality (Atalay et al. 2023). Another recent study shows that retiree mobility in France can bring economic gains to poorer and more rural areas (Badilla-Maroto et al. 2026), reminding us that not all migration flows reinforce metropolitan concentration. In aging Japan, however, the central force is the movement of young workers and future parents away from already old regions.

We simulate one transparent policy: transfers to residents of the five oldest prefectures, financed by taxes on Tokyo residents. This is not meant to replicate a particular programme; it is a benchmark for understanding the economic forces behind regional revitalisation policies. Policy duration matters in this setting because migration, births, and local-scale effects unfold slowly. A temporary subsidy may cushion decline for a while, but a long-lived policy changes expectations about where it is worth living, working, and raising children.

A transfer equal to 5% of income, maintained for 100 years, therefore has large effects. By 2065, it nearly doubles the population of the five oldest prefectures relative to the baseline and reduces their elderly share from about 45% to roughly one-third. The gain in local real income is larger than the transfer itself because more residents raise productivity and amenities through local scale.

But the policy is not free. The same 5% transfer lowers aggregate labour income per capita by more than 1% and raises aggregate fiscal spending per capita by about 0.5%. These costs arise because the transfer keeps more people in places where productivity is lower and public services are more expensive to provide. The policy reduces spatial inequality, but it also reduces aggregate efficiency.

Figure 3 Effects of transfers to the five oldest prefectures on population shares and elderly shares in 2065

Source: Giannone et al. (2026), Figure 13.

The main implication is that national debates about aging and depopulation need a spatial lens. A country can shrink while its largest city gains population share. An aging society can become regionally more unequal even as people rationally move toward opportunity. Subsidies can slow the hollowing-out of rural regions, but they also reshape where people live, work, and have children. That means their benefits and costs accumulate over decades.

For policymakers, the relevant question is not whether every village can or should be restored to its past population. It is how to choose among different goals: protecting access to essential services, preserving communities, supporting mobility, encouraging birth and settlement in declining regions, and maintaining aggregate productivity. Our results suggest that these goals cannot all be maximised at once. Living in a ghost town is not only a demographic concern. It is a spatial economic problem – and it requires policy to confront the geography of aging directly.

See original post for references