Where Remote Workers Go—and Why
A city can lose net migration without losing every newcomer
This chapter follows the spatial consequence of the WFH–migration relationship: where workers move when the job no longer requires them to live near the employer. The key task is to separate out-migration, in-migration, and net migration, then interpret destination characteristics without assuming that remote workers all seek the same kind of place. That distinction lets us see what a lower-cost move may trade away and why the U.S. and Melbourne evidence should be compared carefully rather than combined into one headline.
The distinction between the three migration measures is simple but consequential. Out-migration is the share of people who leave a city or MSA. In-migration is the share of people who arrive. Net migration is the difference between arrivals and departures. A city can have a high out-migration rate and still gain population if even more people arrive. Conversely, a large increase in departures can produce a net loss if arrivals do not rise enough to offset it.
The Dallas Fed study finds that U.S. MSAs with larger increases in WFH also experienced larger increases in out-migration. The relationship is not just a visual impression. Among MSAs with at least 200,000 residents in 2019, a ten-percentage-point increase in an MSA’s WFH share was associated with an increase of about 0.52 percentage points in out-migration. The relationship runs in the opposite direction for net migration: the same increase in WFH was associated with a decrease of about 0.39 percentage points in net migration.
The two estimates tell a more precise story than “remote workers left the city.” High-WFH cities did receive newcomers, but those arrivals did not fully replace the people leaving. The strongest WFH expansion occurred disproportionately in large, expensive cities, so a change in their outflow could have a visible effect on national and regional patterns. Yet the estimates are relationships across cities, not a complete causal account of why each MSA changed. A city with a large increase in WFH may also have experienced changes in housing demand, office use, local amenities, public safety, immigration, or preferences about density.
The destination comparison adds another layer. Among people who moved across MSAs, WFH workers went to places with lower average rents, lower average wages, and lower state income-tax rates than their origins, relative to comparable commuters. The estimated destination-origin difference was roughly 3.5% lower for rents, 1.8% lower for wages, and 0.35 percentage points lower for state tax rates. These figures do not say that every WFH worker moved to a cheaper place, or that a destination was cheaper in every respect. They describe average differences after controlling for worker characteristics and year.
The lower-wage result may seem counterintuitive until the job–residence link is kept in view. A commuter who changes cities generally needs a job in the new local labor market, so the local wage is part of the reason for moving and a constraint on where moving is possible. A fully remote worker can sometimes carry the current employment relationship into a place where the local wage is lower. That does not mean the worker is indifferent to income. It means the relevant income may come from a remote employer or a broader labor market rather than from the destination’s typical local job.
The destination evidence is therefore best read as a trade-off. Lower rent and tax can make a move attractive, while a worker may accept a different local wage, distance from services, or set of amenities. A city with a lower average wage can still be a desirable destination if the worker’s job remains portable and the cost of housing is meaningfully lower. But a lower-cost location can also become less attractive if it lacks accessibility, health care, schools, social connections, or the services a household values. The data reveal the direction of several trade-offs; they do not rank every place for every worker.
It helps to read the city-level coefficients as changes in rates, not as a count of people who must move. Suppose, only as an illustration, that an MSA has 10,000 workers and its WFH share rises by ten percentage points. The reported relationship would associate that change with about 52 additional departures per year relative to the study’s comparison, and about 39 fewer net migrants after accounting for arrivals. Those are not predictions for that particular MSA: the estimates are averages from a selected sample of larger cities, and the relationship is not a guaranteed response. The illustration does show why net migration cannot stand in for out-migration. The city could lose 52 more people through departures while gaining some of them back, yet still finish with a smaller net inflow.
The destination-origin estimates need the same care. A result that WFH movers reach places with mean wages 1.8% lower does not mean that each mover takes a pay cut of exactly 1.8%, or that remote work makes earnings unimportant. It compares average characteristics of the origin and destination cities. The worker’s own wage may come from the current employer, from a remote job found after moving, or from a new local job. The city’s mean wage is a proxy for the local labor-market environment, not a complete measure of a household’s income. The rent and tax differences are also averages: a household’s actual housing payment and tax liability depend on the home they choose and the rules that apply to them.
That is why the word “choice” should not be used too casually. A worker may prefer a lower-rent city but remain because moving costs are high, a partner’s job is local, a child’s schooling is settled, or a remote employer later requires office attendance. Another worker may move farther than the average because family support or a particular amenity outweighs a higher rent. The observed destination pattern records the choices that survive these constraints. It is not a map of the locations workers would choose in a frictionless world.
The city also responds to the arrival of remote workers. If newcomers bring a job and income from elsewhere, they may be able to bid for more housing than a typical local worker. Their demand can raise rents in receiving neighborhoods, reducing part of the original cost advantage. If departures reduce demand in an expensive origin, rents may ease there, but the city may also lose customers, tax revenue, or the density that supports services. A destination that looks cheaper before the move may become less cheap after many households make the same calculation. The spatial effect is therefore a feedback process, not a one-time sorting of people onto a fixed map.
The Melbourne evidence is valuable precisely because it makes these feedbacks visible at a smaller geographic scale. It asks whether workers whose jobs were more suitable for WFH changed the distance and direction of their residential moves, not whether they escaped an entire labor market. The answer can be meaningful even when the overall share of movers barely changes: the same number of moves can be redistributed among locations with different rent, access, and service profiles. This is a different margin from the U.S. result, but it speaks to the same question of which location utilities became more important once daily commuting was less binding.
A further caution follows from the unit of analysis. The U.S. regression compares the characteristics of a mover’s destination and origin MSA. It does not observe the household’s full decision process or prove that lower rent caused the move. A remote worker may have moved for family, health, or a new employer and happened to land in a lower-rent city. The consistent destination pattern is evidence that work mode and city costs are related; it is not a survey of motives.
A cheaper destination still has to work
The Melbourne study helps test what happens when the geography is narrowed from movement across labor markets to movement within one metropolitan area. Its researchers use census flows among SA2 areas, geographic units roughly comparable to suburbs. They separate all residents from people in occupations classified as suitable for WFH. The classification covers about 34% of Melbourne’s workforce, or 153 of 476 detailed occupations, but suitability is not the same as actual remote work. It identifies a group whose work could be more exposed to a shift in workplace practice.
Melbourne’s results do not show a simple story of everyone moving outward. The share of residents moving beyond an SA2 did not jump dramatically in 2021 compared with the earlier pattern. However, the distance traveled by movers increased substantially in 2021, especially among workers in WFH-suitable occupations. This is an important distinction. The number of people who move can remain fairly stable while the people who do move travel farther. A city may therefore experience a meaningful change in spatial structure without a dramatic change in its overall mobility rate.
The study’s difference-in-differences analysis reinforces that distinction. The estimated effect of the work-mode change on the residential mobility rate was weak, while the change in the outbound-to-inbound move ratio was more substantial over the five-year comparison. In practical terms, the evidence is stronger for a change in the direction of movement than for a large increase in the fraction of residents who moved at all. That is compatible with a gradual adjustment: households may wait for a lease to end, a job to change, a child to reach a new school stage, or a sufficient improvement in the value of another location.
Time horizon matters here. The one-year Melbourne data from 2020 to 2021 cover the middle-to-later period of lockdown and workplace rearrangement, but they do not capture the first wave of movement after the initial outbreak. The five-year data can capture a slower relocation pattern, yet they also mix the work-mode change with other changes over a longer period. The study reports that its spatial models fit the five-year migration patterns more robustly than the one-year patterns. That supports the idea that location utility adjustments may take time, but it does not isolate a single immediate pandemic effect.
The location-utility result is more nuanced than a simple “suburban flight” account. Melbourne’s WFH-suitable movers did not primarily target the fastest-growing, lowest-density outer zones. Their moves were more consistent with seeking a balance between housing affordability and accessibility or liveability. In the study’s framework, liveability includes access to amenities and services, while accessibility captures the ease of reaching opportunities and urban functions. If WFH reduces the need for a daily commute, accessibility becomes less dominant, but it does not become irrelevant. A household may still value public transport, shops, schools, health services, or occasional access to employment centers.
The Melbourne setting also illustrates why lower cost is not a complete explanation. During the period studied, changes in CBD vacancies and rents, border restrictions, and the timing of lockdowns affected where people could or wanted to live. An increase in movement toward a middle suburb might reflect a new balance of price and access, a temporary housing-market opportunity, or both. The aggregated census flows cannot tell us which motive applied to each household. They do show that the observed pattern does not require a universal preference for the farthest or least dense location.
The U.S. and Melbourne results are related at the level of mechanism but not interchangeable as measurements. The U.S. study asks whether workers cross MSA boundaries, usually leaving one labor market for another, and its WFH group is based on actual full-time work from home. Melbourne examines movement between SA2 areas within one city, and its focal group is based on occupational suitability. A U.S. inter-MSA move can involve carrying a job into a new city; a Melbourne move may preserve the same local labor market and change housing or commuting conditions. The shared conclusion is therefore conditional: loosening work and transport constraints can alter residential choices, but the destination depends on the scale of the move and the utilities available there.
This comparison also changes how a city should read its own migration statistics. Rising departures from a high-WFH area are not enough to infer decline; the city must examine arrivals, the worker groups involved, and changes in rents and services. A receiving city should not assume that every newcomer brings the same fiscal capacity or housing demand. A planning agency should distinguish a short-term movement spike from a longer-term shift in the balance among affordability, access, and liveability.
The most defensible conclusion is not that remote workers go to “the suburbs” or that cities have become obsolete. It is that full-time WFH expands the set of destinations some workers can consider, and those workers often use the new option to exchange one bundle of location costs and benefits for another. The U.S. evidence makes the cross-city cost pattern visible. Melbourne shows that even within a single urban area, the adjustment can be selective, gradual, and centered on balance rather than maximum distance. The next question is how these individual choices become a new equilibrium in jobs, housing, and city prices.