Metropolitan migrational trends in post-pandemic

context: analysis of the Madrid Region based on

mobile phone network data
Running title: Post-pandemic migration with mobile phone network data

Keywords: Mobile phone network data, migration, metropolitan areas, Madrid,

COVID-19

Abstract:

This paper aims to enrich the current literature on the study of migratory
movements in the context of covid and the post-pandemic period. While most
of the studies on this topic measure migration using official registers, we
propose a new methodology based on the leverage of mobile phone network
data, taken from the Madrid region, as a case study. While the use of such
data is common in other fields, such as transport and mobility planning, we
demonstrate their usefulness in the study of migration. Analysing the case of
Madrid, we find new evidence of the changes in migratory trends during the
COVID-19 pandemic: increased immigration into rural and outer suburban
areas and emigration from core urban areas. A geographical description of the
new migration trends is provided on different scales, from the national level to
the metropolitan and municipal scales, including small urban transportation
zones.
1. Introduction
Migration from the city centre to enlarged peripheries or exurban spaces, has
been an ongoing process over recent decades, from the industrial era to the
new city regions. This process has been described by various terms depending
on the context or motivation (such as suburbanisation, counterurbanisation,
deurbanization and neoruralisation) the actual fact is that outbound migration
has been a consistent driver for the construction of metropolitan areas
(Champion, 1997). Although this trend is not new in the configuration of urban
systems, the study of this process is receiving significant attention from academia
in the context of the COVID-19 crisis.
The media widely perpetuated the idea of an 'urban exodus' and its associated
changes during the crisis (Weisbuch, 2021). This migration from the central areas
of cities has become a consistent trend in the post-pandemic era (Whitaker,
2021), described and analysed by Ramani and Bloom (2021) using the concept
of the so-called "donut effect". However, the persistence of such changes in the
post-pandemic era requires a longer-term perspective. It requires a
comprehensive analysis of the trends observed across various cities and domains
to better understand their implications (Rowe et al., 2023).
Although COVID-19 could just have triggered a one-off crisis, other elements that
had already been in place may have also affected and fuelled the urban
consequences of the new scenario:
(1) From an urban point of view, a desire for a less dense, greener and healthier
built environment (Glaeser, 2022) tied to the ecological crisis many cities are
experiencing due, in part, to climate change and human activities (for example,
due to air pollution or climate comfort). This desire may have been intensified
during lockdowns as less dense areas were less prone to contagion.
(2) The housing affordability crisis, that many cities are also experiencing, with a
clear tendency for a center-periphery logic. This case is even more important
given the gentrification in tourist cities. In urban centres like Madrid, Barcelona,
Paris and London, housing affordability has become a widespread problem,
especially for younger people. This has resulted in a shift towards suburban areas
as individuals seek more affordable residential options.
(3) The ongoing digitalisation of work environments and specifically the possibility
of telecommuting has enabled some type of workers to stay at home and
choose their residential location without the limitation of having a specific job
location. The option to work from home became widely used during COVID-19
and has remained important during the post-pandemic era.
None of these elements are unique and universal across the entire population,
as biases related to socio-economic status, age or gender may have impacted
the geographical distribution of those migratory movements. Some personal
factors, such as owning a second residence, the possibility for working remotely
or the socio-economic constraints of families may be determining factors.
From the methodological point of view, analysis of migration has typically been
conducted based on official census data, which is usually collected on a regular
basis and at a municipality level. However, over the last decade, new data
sources and methodologies for studying migration have emerged, particularly
those based on big data (Vich et al., 2017). While many methodologies use data
from social media (Rowe et al., 2023), postal services (PUCA, n.d.) or property
browsers (Breuillé et al., 2022a), there are currently no migration studies based on
geolocated data from mobile phone networks. With these data, the presence
of individuals in different places at different times can be estimated, regardless
of whether they have a social media account or are registered in a municipal
census. Geolocated data from mobile phone networks are becoming common
in other fields, such as mobility and transportation (Chrétien et al., 2019), but are
rarely used for migration purposes (Rowe et al., 2023).
This research has two main objectives:
(1) First, it seeks to describe and analyse the changes in migration structure in
the metropolitan area of Madrid, with special regard to its periphery. This
analysis considers three different time periods: before, during and after
the COVID-19 pandemic, allowing to analyse the extent to which people
are moving out of the core city. The region of Madrid is an excellent case
study for the scenario of new internal urban migration among its
metropolitan area, as it has experienced a significant suburbanisation
process over the last few decades (Bayona-I-carrasco & Rúbies, 2019).
The growing importance of the periphery over the central city (Gil-Alonso
et al., 2021) and the clear continuity of its strong centralistic geographical
structure creates a solid framework in which to analyse the donut effect.
(2) Second, it aims to conduct the migration analysis using data on the
presence of people on different dates, inferred from mobile phone
networks user location data. We aim to test this methodology and
evaluate its advantages and limitations.
This paper is structured into five main parts:
(1) First, we present a theoretical framework positioning the donut effect
within the literature on suburbanisation. We provide an overview of key
case studies in recent research on the evaluation of the donut effect,
selected to cover different data sources and methodologies. This justifies
the appropriateness of the dataset we propose to use as a complement
to previous methodologies.
(2) Second, a description of the data processing, its features and its limitations
compared to other data sources previously used in the study of migration
is provided. Our dataset is structured geographically and enriched with
existing public datasets.
(3) Third, we provide a general description of the geography of migration in
the Madrid Region during the COVID-19 pandemic. An analysis on the
national level is taken to evaluate where migrants went when they left
Madrid during the pandemic. The presence of data from the post-
pandemic period also allows us to explore whether these changes are
temporary or persistent.
(4) Fourth, an exploration of results regarding how migration is taking place in
terms of the relationship between the city core and the different rings in
the Madrid metropolitan area is conducted, to assess the potential
existence of the donut effect, and to understand how the geography of
migration changed during lockdowns and in the post-pandemic period.
(5) Finally, we provide a summary of the findings and describe the gaps and
limitations of the dataset used and highlight some research opportunities
based on the results.
2. Literature review
The term “donut effect” was coined by Ramani and Bloom (2021) after the
COVID-19 pandemic. They observed a shift in residential and business property
demand from central areas to low density suburbs and exurbs in large cities in
the USA. The study provided significant evidence that households leaving city
centres were mainly moving to the suburbs or, to a lesser extent, to smaller cities
or rural areas. Beyond the USA context, this trend has also been identified in
European metropolitan areas, such as in the United Kingdom (Rowe, Calafiore,
et al., 2023), France (Breuillé et al., 2022) and Spain (González-Leonardo et al.,
2022).
A review of the literature covering the donut effect, or outbound migration from
urban centres, over recent years, shows increasing interest in the topic among
many authors in different contexts. This phenomenon has been addressed
through a wide range of frameworks The Table A - Annex 1 presents a summary
of the common outcomes of these studies.
A significant number of studies on how the pandemic is impacting migration in
urban contexts have been published since the end of the crisis. Our review
includes six cases: four in Europe (González-Leonardo et al., 2022; Rowe et al.,
2023; Vogiazides & Kawalerowicz, 2022), one in Japan (Fielding & Ishikawa, 2021)
and one in the USA (Ramani & Bloom, n.d.). With the aim of analysing the gaps
to fill, these articles have been reviewed from three different perspectives: (i)
methodological, to understand the methods and source dataset; (ii)
geographical, to examine the specific features of each case and understand
the global trends; and (iii) theoretical, to understand the conceptual framework
through which the issue has been approached.
First, from a methodological point of view, the main data source used by most of
the migration studies is a form of official national statistical record, where a
registration of the population is offered, aggregated to the municipal scale. In
our review, we find this for Spain, Sweden and Japan. Studies with national data
benefit from national coverage and provide an easy comparison with previous
years, but sometimes lack a full record of migrants since some groups are not
officially registered (students, retirees, temporary migrants etc.). Studies in the UK
and US, using data from Meta and from postal services, can give a more
accurate picture of the situation, even if the representativeness of the sample is
never complete.
Second, concerning the geographical context, most the cases revealed an
overall migration flow from city centres and urban environments to less dense
areas, with this trend being more pronounced in large and capital cities (as seen
in Spain, Stockholm and the US). It is also true that sometimes is not the growth of
outbound migration, which is revealed, but rather the reduction in inbound
migration. Two studies examine the sociological segmentation of the issue,
looking at the scenario of patterns for retirees, in the case of Japan, and students,
in the case of Sweden.
Finally, regarding the conceptual framework, most studies frame the research in
a story of the relationships between urban areas and less urban areas, based on
well-known concepts such as counterurbanisation and residential choices. All
the studies study geographical hierarchies between large, medium and small
metropolitan areas or by density classification. After reviewing all these studies,
some interesting trends have been identified. First and foremost, the fact that
during the COVID-19 pandemic, migration from the rural areas to the cities has
decreased. As a result, net-migration was negative in the cities during the early
stages of the pandemic. In contrast, the areas that became attractive are linked
to second homes on the coast or in the mountains, in mid-sized cities or in some
rural areas suffering depopulation. Within metropolitan areas, the most dispersed
suburban areas were the preferred destinations.
Leaving aside these common outcomes, some interesting nuances were
identified in the early stages of the pandemic. For example, there was an insight
into the relationship between migration and demographic or sociological
determinants. For age, some unlikely trends were identified, specifically related
to old people. Belonging to an older age group, as opposed to the 18–24 age
group, is associated with a lower probability of moving out of the Stockholm inner
city (Vogiazides and Kawalerowicz, 2022), while in Spain, retired people were
characterised by a flow of outmigration from urban areas to rural or coastal
areas (González-Leonardo, Rowe, et al, 2022). In general, an individual’s stage
of life is a determining factor in this change in residential choice (Breuillé et al.,
2022) and for income, there is a trend for high income migrants to move to
second homes. The possibility of working from home is also an important
determinant in location choice, as pointed out by Vogiazides and Kawalerowicz
(2022) in the case of Sweden, where the self-employed moved more towards
small cities and rural areas. This was also observed in Britain by Rowe et al. (2023).
However, the urban exodus from large cities seen in the wake of the pandemic
seems to have lost some of its intensity in the current post-pandemic period.
Instead, the COVID‐19 pandemic can have accelerated existing mobility trends,
with cities expected to bounce back and remain major centres of population
attraction after the pandemic. A longer timeframe is needed to study any
migration changes triggered by the pandemic. The short observation period and
lack of available data make it difficult to evaluate the consistency of these
changes (Rowe, González-Leonardo, et al., 2023; Wolff and Mykhnenko, (2023).
Some trends are still unclear. For example, there are questions about changes in
the attractiveness of city centres and how central areas are responding to those
changes (Florida et al., 2023b). There is also uncertainty about the reality and
scope of the urban exodus movement and to what extent it is fuelling a new
period of counterurbanisation, with a positive impact on rural communities
(PUCA, n.d.), as well as the limits of this process with suburbanisation and
colonisation of the extended rural hinterland of cities (Champion, 1997). At the
same time, multifactorial processes such as the housing affordability crisis, labour
market changes relating to working from home, and the influence of COVID, are,
among all the factors, also aspects that require further study (Wolff & Mykhnenko,
2023).
A final key element in the study of migration and the donut effect in the context
of COVID-19 is the availability of data. Most of the previous studies use data from
official census and population registers. These official sources typically do not
have enough time granularity to analyse population movements over short time
intervals. Therefore, alternative data sources can facilitate the generation of
more accurate insights of the evolution of trends in internal urban migration.
Digital trace data derived from mobile phone applications offers a unique
opportunity to capture these movements with unprecedented spatial and
temporal resolution (Green et al., 2021).
However, little research has used these new sources to study the effect of COVID-
19. Rowe et al. (2023) rely on an unconventional dataset: anonymised data
aggregated from Meta‐Facebook users comprising 21 million observations for
Great Britain. Elejalde et al. (2023) leverage mobile phone network data to unveil
the complex dynamics of internal migration and daily mobility in Santiago de
Chile during the global COVID-19 pandemic.
The advantage of these new sources is that they enable more up-to-date
analysis, often at finer spatial scales than official data. However, these sources
do not usually enable a full characterisation of the migrants from a sociological
perspective, such as age, gender and other demographic attributes.
The pandemic has altered migration patterns in large cities. Understanding the
extent and durability of population movements to and from cities in this new
scenario and the factors underpinning these is key to informing spatial planning
and developing appropriate policy responses (Rowe et al. 2022).

3. Case study
3.1 Madrid Metropolitan Area
The demographic structure of Spain has experienced a significant trend towards
urban areas and the coastline during the first years of this century. This was
particularly evident for Large Urban Areas, defined by the Spanish Ministry for
Urban Affairs (Ministerio de Transportes y Agenda Urbana, 2022), and it continued
up to the COVID-19 pandemic. According to this body, in 2020-2021 the
migratory consequences of COVID-19 were already notable: negative dynamics
in large urban areas, which lost 130,000 inhabitants (0.5%), and positive ones in
medium sized cities and non-urban municipalities, which gained 12,000 and
52,000 more individuals, respectively. In general terms, a reduction in migration
movements was registered by the Spanish Statistical Office (INE): the total
number of internal migrations during 2020 (when lockdowns were implemented)
was 1,519,000, 130,000 fewer than in the previous year. The latest figures from
2021 show the post-pandemic era, with an interesting spike up to 1,678,000
internal migration movements, regaining pre-pandemic levels.

The Large Urban Area of Madrid is home to 6,265,362 inhabitants across over 52
municipalities, representing one in five urban dwellers in Spain. The Large Urban
Area of Madrid is part of Madrid Region (officially called the Community of
Madrid), which is the study area for this publication, and is home to 6,750,336
inhabitants (INE, 2022).

Madrid’s metropolitan area has been developed based on two parallel

migratory trends. First, it has attracted migrants from all over the country, seeking
the advantages of agglomeration economies and the concentration of
opportunities offered by a capital city. This process is linked to population loss in
inland regions. Second, there has been an internal redistribution towards an
expanding metropolitan area driven by suburbanisation processes (Bayona-I-
Carrasco & Pujadas-I-Rúbies, 2014), which in the case of Madrid sees the edges
of the administrative Community of Madrid spilling out towards neighbouring
provinces, such as Toledo to the south and Guadalajara to the east.
The latest research on internal migration in Spain provides evidence of increased
migration outflows and a decrease in migration inflows for Madrid and Barcelona
(González-Leonardo et al., 2022), demonstrating the donut effect in those cities.
This trend appears to be temporary, but evidence is still limited.

3.2 Mobile phone network dataset

We use a residential migration dataset derived from anonymized data provided
by one of the main mobile phone network operators in Spain, which holds
around a 30% market share.

The process followed to infer migration data is divided into the following
processes:
1. Identification of valid users who are present in the study area in different
periods: T0, T1 and T2 (before, during and after COVID-19), measured in
the month of February across three different years:
o T0 measures the number of movements between 2019 and 2020
(before COVID-19),
o T2 between 2020 and 2021 (during COVID-19), and
o T3 between 2021 and 2022 (after COVID-19).
2. Identification of the location of residence: the place where the individual
spends the night. The destination is identified as the area where the
individual spends several hours during the day.
3. Socio-demographic characterisation (age, gender and income) of the
valid users, using predictive and probabilistic methods which take into
account other official data sources, such as the census.
4. Zoning of users based on antenna locations and corresponding Voronoi
areas. These locations are geographically distributed using the 210
transport zones that dived the Madrid region, for the locations inside
Madrid’s region, while areas outside Madrid are distributed by Spanish
regions. This process allowed the creation of an origin-destination matrix.
5. Expansion of the sample using public data and national population
registers from the Spanish Statistical Office (INE, 2020), associated with
each transport zone.

For this data processing, a minimum of 10 people in each OD pair was set to
ensure anonymity. Additionally, a migration is considered only if there is a
minimum distance between Voronoi polygons of 1 kilometre. Census districts with
a sample of less than 100 times their population are discarded.

The result is a dataset representing each origin and destination and the number
of “migrants” and “stayers” for each period (T0, T1, T2), in the form of an OD
matrix. In addition, the migration OD matrix is segmented according to age,
gender and income.

To properly distribute the data in Madrid’s metropolitan area, we have split the
data into two parts.
1. The first dataset explains the internal changes inside the region, taking all
the OD pairs with origin and destination inside the boundaries of the
Community of Madrid autonomous region, which contains nearly the
whole Madrid Region (internal-internal trips). It is considered that the
 region contains most of the functional urban area.
2. The second dataset contains all the pairs with an origin or destination
outside the Madrid region, so it is possible to assess the number of migrants
arriving from other regions in Spain or leaving the urban area (internal-
external trips). These regions have also been classified into a coastal ring
and the remaining intermediate regions inside Spain.

In this second set, two sub-regional zonings have been introduced for classifying
origins and destinations in a geographical sense. First, a subdivision based on
different rings around the city was adopted: (i) Madrid’s central districts, (ii) the
rest of the municipality of Madrid, (iii) the first metropolitan ring, and (iv) the rest
of the community. A fifth zone, on the southeast of the region, has been added
to account for the adjacent provinces around the Madrid region, including Avila,
Guadalajara, Toledo, Cuenca and Segovia. Second, a subdivision is established
for the region, based on sub-regional NUTS4, divided into 11 zones, to help the
understanding of residential movements inside the region. These two divisions are
presented on Figure 1.

Figure 1 - Madrid region and its rings (left) and Madrid NUTS 4 zones (right)
Source: ESRI and INE

To understand flows, each OD pair has been classified as “outbound flows” and
“inbound flows”, according to its direction with respect to Madrid city centre. We
also identify the number of rings each flow involves. For example, an OD pair
from Madrid city centre to the regional ring is classed as a level 3 outbound flow.
An OD pair from the regional ring to the metropolitan ring is classed as a level 1
inbound flow. This allow us to understand which flows changed before and
during the crisis, giving an idea of distance, and also to separate migrants staying
in the Madrid Region from others leaving the region.

In addition, the mobile phone network dataset has been enriched and
supplemented with other publicly available data. In our case, two sets of such
data have been used:
• The Residential Variation Statistics (“Estadística de Variaciones
Residenciales”), made available by the INE. This takes into account of
changes in all municipal population registers. It is a traditional source of
information, but only captures official and declared residential changes,
therefore excluding information on an important proportion of migrations
(students, retirees, temporary workers etc.). This dataset will be used to
validate the size and order of magnitude of the mobile phone network
data and also to understand the previous trends before the mobile phone
 data periods (prior to 2019). Only the 2013-2019 period has been
consulted.
• The official register of residents (“Padrón Municipal”) also made available
by INE. This is a standard population register broken down into census
sections. These sections have been geographically mapped to the zones
used by the mobile phone network datasets.

With the aim of identifying who is permanently leaving the metropolitan area
and moving to areas where daily commuting to Madrid is impossible, we
separated origins and destinations into the expanded Metropolitan Area and
the rest of Spain.

4. Results
This section presents the results of the study at four different scales. First, the results
are analysed by transport zone. Next, the results are aggregated by ring to
evaluate general trends. Then, the results are examined for the 11 NUTS4 zones
of Madrid. Finally, the destinations of emigrants within the national territory are
presented.

4.1 Migration by transport zone

As the first purpose of this study is to offer a full characterisation of migration flows
inside the Madrid region during the COVID-19 pandemic, a map showing
emigration and immigration at the transport zone level is firstly shown (Figure 2).
The map shows both, the total number of immigrants and emigrants during
COVID-19 (Feb. 2020 to Feb. 2021), and the change compared to the previous
year’s immigration and emigration figures (Feb. 2019 to Feb. 2020).

Figure 2 shows an increase in immigration to the peripheral areas compared to

the previous year. Some of the largest increases are observed in the outer rings.
Since most of these areas are more rural and populations and flows are small,
small changes in absolute values means the proportions reported can be high.
In parallel, a surge in emigration rates is shown for emigrations from central and
intermediate metropolitan areas, although emigration rates are less significant.
This supports the idea that a change in migration destinations but not in the
amount of migration or the distribution of emigration: migrants originated in the
same geographical distribution, but it is the attractiveness of the destination that
is evolving. The change in central spaces reflect less immigration rather than
more emigration. These results are similar to the findings of Bayona (Bayona-I-
carrasco & Rúbies, 2020).
 Figure 2 - Total and evolution rates of immigrants (left) and emigrants (right) between the pre-COVID and COVID
periods

Figure 3 - Total and evolution rates of immigrants (left) and emigrants (right) between the period pre-covid and
post-Covid

Table 1 shows that the number of emigrants originating in the municipality of

Madrid did not change significantly from the year prior to the pandemic and the
year of the pandemic. During this period, the number of immigrants arriving in
Madrid significantly decreased, resulting in a negative net migration of almost
130,000 people. This is similar to evidence provided by some studies on American
cities during COVID-19, where negative net-migration was mainly due to a
decrease in immigration rather than an urban exodus (Whitaker, 2021).

When it comes to stayers, the figures indicate that some people moved within
the city. In relation to the metropolitan ring, there is no significant change during
the pre-pandemic, pandemic and post-pandemic periods, with very similar net-
migration rates across those three periods. The same is true for the “Rest of the
community” category, which represents the outer rings.

2019-2020
Emigrants Stayers Immigrants Net migration
Municipality of Madrid 567,953 2,740,908 551,494 -16,459
Metropolitan Ring 383,390 2,431,910 382,108 -1,282
Rest of Community of Madrid 64,266 417,718 72,226 7,960
Regional Ring 447,656 2,849,628 454,333 6,678

2020-2021
Emigrants Stayers Immigrants Net migration
Municipality of Madrid 584,565 2,600,827 455,100 -129,465
Metropolitan Ring 344,746 2,469,397 344,588 -158
Rest of Community of Madrid 54,516 428,773 79,237 24,721
Regional Ring 983,827 5,498,997 878,925 -104,902

2021-2022
Emigrants Stayer Immigrants Net migration
Municipality of Madrid 496,296 2,801,548 509,622 13,326
Metropolitan Ring 340,105 2,491,797 342,645 2,540
Rest of Community of Madrid 61,889 430,299 63,004 1,115
Regional Ring 401,994 2,922,096 405,649 3,655
Table 1 - Number of immigrants and emigrants by metropolitan ring

Evolution of the trends by transport zone

To explore the migration processes over a longer period, a comparative analysis

has been conducted between net-migration rates during COVID and gross net
migration in the period 2013-2019.

The figures have been classified according to six types of evolution trend,
allowing us to understand whether COVID-19 intensified a negative or positive
existing trend (1 and 2), COVID reversed a positive or negative former trend (3
and 4) or COVID did not change the trend but lowered its significance (5 and 6).
This analysis is inspired by the description conducted by Wolff and Mykhnenko
(2023) at a European level, and it is useful to understand the role of COVID-19 as
a “game changer”.

Figure 4 shows that the outer ring is becoming more attractive to migrants than
more central spaces. Attractive areas maintain their trends (mostly in the north-
western mountain area), and areas that underwent a population decline are
now gaining habitants. Some exceptions can be seen, such as in the south-east
where the large housing developments in the municipality of Madrid continued
gaining population during lockdowns, and some parts of the intermediate rings
with very high density were also losing population in the same way as the central
core.
We can see from this map that the impact of COVID has not been homogeneous
across the metropolitan area, but a clear trend can still be identified: COVID
made the centres less attractive (even if they were not already), and peripheral
spaces more attractive. The peripheral areas that were already attractive
became even more attractive and some previously unattractive zones changed
their trend, as in the case of the rural outer ring experiencing depopulation.

Figure 4 - Changing trends in net migration between 2019-2020 (left) and 2019-2022 (right)

With the perspective of the third period analysed (Feb. 2021 to Feb. 2022), we
can gain an insight into the post-pandemic period. The first and foremost
outcome is that most changes are temporary, but permanence varies
significantly across different geographic regions.

Regarding the inner rings, we observe a change from negative to positive net
migration, or at least a slowing of the negative trend. A strong acceleration in
the positive trend is observed in the southeast of Madrid, due to the important
housing developments in the area. In contrast, the outermost rings show a dual
trend: some areas are going back to their original negative trend (the mountain
area in the north and some areas in the rural south-east), while others are
maintaining the positive trend or seeing this slow, typically in the south-east, the
east and the south, places that are well-connected to the city. The position for
the intermediate ring remains unclear.

4.2 Migration by ring

While we see that the overall population of the Madrid metropolitan area is not
changing, we can observe a relocation of this population over a wider area due
to the attractiveness of these outer rings. The result is a re-densification of the
whole metropolitan area and the occupation of former rural hinterlands which
are now integrated into the urban continuum.

The classification of inbound and outbound flows results in the total figures over
the three time periods studied presented on Table 2. It should be noted that the
available flows are only those with an origin or destination in the Community of
Madrid, so the relationships between the adjacent provinces, inland Spain and
coastal Spain are not assessed.

A clear decrease in inbound flows is seen during the year of COVID-19 (-27%
when comparing 2019-2020 and 2020-2021 figures), when an important increase
in outbound flows occurred (+17%). Outbound flows were significantly larger
than inbound ones, which indicate a negative net migration figure: Madrid lost
population during COVID. Nevertheless, in the following year (post-COVID), the
proportion between inbound and outbound migration is similar to the pre-COVID
levels.

Change
Change Change
T0: T1: T2: during-
pre-during pre-post
Migrations Migrations Migrations post
COVID COVID
2019-2020 2020-2021 2021-2022 COVID
(T0-T1) (T0-T2)
(T1-T2)
Regional
inbound 169,064 122,987 149,672 -27.2% 21.7% -11.4%
migrations
Regional
outbound 191,656 225,122 165,789 17.4% -26.3% -13.5%
migrations

Outbound
140,231 180,910 132,115 29.0% -26.9% -5.7%
from the region

Inbound 134,687 112,634 152,635 -16.3% 35.5% 13.3%

Stayers 6,113,151 6,064,227 6,180,388 -0.8% 1.9% 1.1%

Table 2 - Inbound flows and outbound flows

Comparing pre-pandemic and post-pandemic flows, we observe stability in

inbound flows, but a significant decrease in outbound ones: approximately
100,000 more people than previously chose to reside in the metropolitan area.
Since the Madrid municipality has not changed, the main possible explanation
is a relocation of the population inside the metropolitan area in favour of the
peripheral areas, especially those in outer rings.

To focus on movements during the pandemic, the following diagram shows the
flows between rings during the period 2020-2021 and the change compared to
the previous period. It shows the positive and negative change in outbound and
inbound migration along with the size of each flow.
 Figure 5 - Flows between rings during the period 2020-2021 and the change compared to the previous period

Some conclusions can be drawn from this. First, there was a net increase in
outbound migrations, specially from the municipality of Madrid. These new
migrants mostly went to the Spanish coast, with fewer going to the Spanish
hinterlands or the outermost rings, although the change in the migrations to the
hinterlands or the outermost rings (such as 64% to Spanish Inlands) is larger than
the change to the coast (+37%). Migration to intermediate metropolitan rings
stagnated.

Regarding net migration, we observe that there are more people arriving in the
metropolitan ring, the rest of the community and the regional ring than leaving
it, in contrast to the situation with the municipality of Madrid. This is likely to be
related to the development of new housing units, such as the ones on the
southeast of Madrid.

The role of distance

Based on the previous results, we can identify a clear restructuring of migration

inside the metropolitan area. The recovery of migration trends in the post-
pandemic period suggests that this restructuring has been non-uniformly
distributed across the territory. The fact that flows to the first ring are stable, while
flows to the outer rings change substantially, leads us to the hypothesis that
migration distance is a determining factor in the recovery of flows.

To assess the evolution of different types of migration, we have classified

migration by distance, breaking it down into the number of rings each migration
crosses. As a result, we have four different types of migration: 1, 2, 3 and 4,
ranging from the shortest distance (migration to the neighbouring ring), to the
longest one (to/from the coast from/to the municipality of Madrid), presented
on Figure 6. This approach allows us to identify to what extent inbound and
outbound migrations have changed during the three time periods depending
on their distance (such as suggested by Florida et al., 2023c). Short migrations
clearly outnumber long migrations, but while long or intermediate migrations
completely recovered to pre-pandemic levels, short migrations were lower in
both directions in post-pandemic scenarios, as also observed by Bayona i
Carrasco & Rúbies, 2019.

Figure 6 - Number of inter-ring flows broken down by distance (number of urban rings travelled through)

4.3 Migration by NUTS4 zones

Beyond the distribution of migration according to rings, the dataset also allows
to study movements in a more specific context. To explore the distribution of
changes in the metropolitan area, we have analysed flows broken down into
smaller units (corresponding to NUTS4). This provides a suitable scale for studying
the relationship between the different areas inside the metropolitan area.

Specifically, we have examined the logic of corridors that define Madrid’s

metropolitan area following the main transportation routes, radial highways and
railways which divide the territory, with a decreasing population density as we
move from the city centre to the rural areas. This segmentation allows us to study
the migrant emission/reception level of each area, as well as the movements
between different points of each corridor and between rings.

Tables 3, 4 and 5 present the permanence, flows and net migration for the three
analysed periods for the scale of NUTS4 zones. Some zones, such as Sur
Metropolitano, have seen decreases on the three periods.
 Net
Group Location Stay Outbound Inbound
Migration
MR MUNICIPIO DE MADRID 2,740,908 567,953 551,494 -16,459
MR ESTE METROPOLITANO 549,487 88,333 89,341 4,917
MR OESTE METROPOLITANO 421,698 73,292 74,815 1,523
MR SUR METROPOLITANO 1,150,716 162,134 148,066 -12,068
RR NORTE METROPOLITANO 250,539 44,397 42,052 -2,345
RR NORDESTE COMUNIDAD 56,334 10,710 11,268 576
RR SIERRA CENTRAL 116,937 23,921 57,256 3,325
RR SIERRA NORTE 34,204 5,200 6,116 916
RR SIERRA SUR 75,870 12,925 4,270 -8,655
RR SUDESTE COMUNIDAD 91,182 14,713 11,502 -2,211
TOTAL COMMUNITY OF MADRID 2,849,624 447,656 454,328 -6,678
* MR: Metropolitan Ring; RR: Regional Ring
Table 3 – Permanence, flows and net migration in the 2019-2020 period for NUTS 4 zones

Net
Group Location Stay Outbound Inbound
Migration
MR MUNICIPIO DE MADRID 2,600,827 584,565 455,099 -129,466
MR ESTE METROPOLITANO 562,442 96,111 80,828 417
MR OESTE METROPOLITANO 424,600 78,198 78,416 2,262
MR SUR METROPOLITANO 1,141,806 148,454 148,340 -112
RR NORTE METROPOLITANO 255,099 38,304 32,781 -5,023
RR NORDESTE COMUNIDAD 58,662 8,319 11,637 8,318
RR SIERRA CENTRAL 117,504 24,158 57,940 33,782
RR SIERRA NORTE 34,326 5,231 5,543 312
RR SIERRA SUR 75,742 12,903 4,378 -8,525
RR SUDESTE COMUNIDAD 118,260 15,739 14,124 -1,615
TOTAL COMMUNITY OF MADRID 2,898,165 399,257 423,822 -24,565
* MR: Metropolitan Ring; RR: Regional Ring
Table 4 – Permanence, flows and net migration in the 2020-2021 period for NUTS 4 zones

Net
Group Location Stay Outbound Inbound
Migration
MR MUNICIPIO DE MADRID 2,801,548 496,296 509,622 -13,326
MR ESTE METROPOLITANO 579,353 78,459 79,836 4,030
MR OESTE METROPOLITANO 424,123 70,240 66,457 3,693
MR SUR METROPOLITANO 1,164,653 151,481 150,352 -1,129
RR NORTE METROPOLITANO 255,940 33,805 34,530 745
RR NORDESTE COMUNIDAD 58,614 8,107 12,048 -2,361
RR SIERRA CENTRAL 117,574 24,485 57,922 347
RR SIERRA NORTE 34,700 5,369 5,542 173
RR SIERRA SUR 76,540 13,646 4,193 -1,889
RR SUDESTE COMUNIDAD 125,904 13,190 19,944 554
TOTAL COMMUNITY OF MADRID 2,922,092 401,990 405,646 -3,556
* MR: Metropolitan Ring; RR: Regional Ring
Table 5 – Permanence, flows and net migration in the 2021-2022 period for NUTS 4 zones

On the Figure 7 we present an OD analysis for the 11 NUTS4 zones and important
cities in the Madrid regions. This analysis shows a correspondent strong pattern of
corridors in the structure of migration. A strong preference to stay in the same
corridor is shown through the presence of adjacent provinces. For example, a
significant proportion of arrivals in Guadalajara come from the eastern
metropolitan area of Madrid. The same was true for Toledo, coming from the
south, and for the Sierra de Madrid (northwest), coming from the west
metropolitan area. This also reflects the socio-economic pattern of the region,
with the North and West areas being comparably wealthy compared to the
popular and industrial south and east. As a result, an important link following a
logic of corridors is translated into outbound flows that remain in the same
corridor by proximity.

It is worth mentioning the historical migration ties within this migration structure,
as metropolitan areas in intermediate rings saw an important amount of
immigration from the countryside in the previous generation, during the 1970s
and 1980s. Families with origins in Toledo (in the south) and Guadalajara (in the
east), settled in southern and eastern metropolitan areas, continue to maintain
a strong link to their original regions, probably an important factor in the
migration of their second or third generations.

Figure 7 - Total number of migration flows during 2020-2021. The origin of flows is indicated by the
colour of the ring. Four rings are represented by a colour palette: blue for Municipality of Madrid,
red for the first metropolitan ring, green for the regional ring.

This is especially important because these adjacent regions are increasingly

considered as part of the functional metropolitan area of Madrid. As urban daily
mobility is a determining factor in choice of residential area, it seems that the
option of commuting is key in migrating without changing job or switching to
teleworking. This can be difficult to measure as there are no data that can split
those adjacent regions between the rural areas and the outer rings of the Madrid
functional area.

4.4 Migration on the national level

The geographical coverage of the dataset has also allowed us to identify the
preferred destinations for those leaving not only the city centre but the whole
metropolitan area and reveal their preferences. From a general perspective, the
data reveal that 264,000 people left Madrid during 2020, while 154,000 people
arrived.

As shown in Figure 8, a rise in the number of immigrants from Madrid is seen in

almost all regions, the top recipient regions being Andalucia, Valencia and
Castilla y Leon. In contrast, migrants with the Community of Madrid as their
destination decreased by 10%. This trend needs to be understood in the context
of the even more remarkable changes in the closest regions adjacent to Madrid:
Cuenca, Segovia and Avila received 60% more migrants from Madrid than in the
previous year, and Toledo received around 40,000 people.

Figure 8 - Number of immigrants from the Madrid Region during the COVID period (2020-2021) and the change
compared to 2019-2020
5. Conclusions
With this publication, we aim to introduce the case study of Madrid into the
debate on whether the pandemic was a “game-changer” for migration trends
or not, especially for the flows related to the “donut effect”. New evidence to
support this trend has been found but with significant nuances: a group of
people permanently leaving the city and moving to the coast or the countryside
and a group of people leaving the city centre and the metropolitan rings and
moving to an outer ring. We provide evidence on the enlargement of this ring:
neighbouring peripheral regions, such as, Toledo in the south and Guadalajara
in the east, are important in the residential choice and this was also the case
during the pandemic, when this trend was intensified. These two processes are
occurring in parallel and are probably related to life cycles. COVID may have
prompted the exit of families and single workers to the peripheries or other rural
spaces and delayed the arrival of young students. As the post-pandemic figures
show, the previous trends have generally returned, albeit not to the same extent.

The figures for ‘stayers’ are not dramatically changing, in that the number of
migrants is not evolving and net migration for the Community of Madrid is not
changing significantly, but the structure of this migration has evolved. We found
(1) areas formerly showing population decline that since the pandemic have
been (and still are) receiving population; (2) areas that have only received
temporary population; and (3) areas that are now less attractive than before,
commonly city centres.

Explaining migration behaviour during a shock event is key in understanding the

resilience of metropolitan areas. This coexists with other long-term processes
mediating this resilience in an interdependent way, as previously seen with
counter-urbanisation and suburbanisation. As both processes (people quitting
the metropolitan areas and people simply relocating) are taking place in
parallel, it is difficult to differentiate the impact of each of them.

The big question remains unanswered: are the big cities still an attractive place
to live or is this changing? A preliminary answer could be the appeal to the public
of areas that can offer an intermediate way of living: while still having all the
opportunities and resources of the cities, also benefiting from less dense areas,
rural environments and natural resources that are supposed to offer a better
quality of life. That might have already been occurring, but COVID has clearly
highlighted that desire. In addition, as socio-economic restrictions also affect
residential choices, periurbanisation is closely linked to other processes already
occurring in Madrid: access to housing (which especially affects young people),
touristification of urban cores and job availability. The evolution of working from
home and other digital processes remain a strong socio-economic selector for
certain types of migration. Relationships between those general processes and
migration as explanatory variables is revealed as an interesting research path in
the case of Madrid or in other cases.

As a complementary conclusion, it is interesting to note the extent to which the

boundaries of a given metropolitan region can fluctuate depending on the
context, which in this case is COVID. Madrid’s hinterland evolved before, during
and after COVID. An enlargement of Madrid’s hinterland, if measured as the
suburban space with a commuting population working in the city, has been seen
and this now includes former rural areas in the exurbs and outer suburbs of the
city. In addition, there is a discontinuity in this metropolitan area for people living
between Madrid and another region. As a general conclusion, we can state that
the metropolitan region of Madrid is still attractive, but the region is both
expanding in terms of area and restructuring its residential distribution with a
reduction in the desire for the core city.

Finally, additional methodological evidence comes from the dataset being

useful for the analysis of urban migration. In contrast to the census data, our data
reveal that 264,000 people left Madrid during 2020, whilst 154,000 people arrived.
In the official census, the same figures are 108,000 and 70,000, respectively. This
difference shows the extent to which official registers are probably not reflecting
the actual migration flows, albeit still reflecting the real proportions.

Nevertheless, methodologies should still be developed to eliminate statistical

biases and the typical data collection issues (representativeness of certain
groups). A development of the methodological processes to expand the
statistical sample could in the future include machine learning processes using
land use data or national statistics, making the mobile phone networks sample
even more representative.

Concerning the theoretical framework for the donut effect in the context of post-
pandemic geographical outcomes, this study remains a descriptive analysis of
the pre-during-post scenarios for pandemic migrations. Future research covering
the period 2021-2023 should help to discover whether these trends are being
maintained or not, to identify whether changes are more stable or temporary.

Apart from methodological concerns about the analysis of Madrid’s

metropolitan region, the descriptive insight of this study may drive further
research on the development of policies on contemporary topics such as
regional planning in the context of the Spanish urban agenda, metropolitan
governance, mobility and transportation supply/demand, or housing.
Concerning the case of Madrid, one of the most interesting outcomes of this
study, already pointed out by other authors, is the fact that the boundaries of the
Madrid Region no longer correspond to the functional metropolitan space. This
is important because some aspects are often hard to analyse as many official
data sources are restricted or measured differently in different regions.

This methodology remains a quantitative approach. It can feed other studies on

migration, inequalities, housing and so on. Future research taking a qualitative
approach could help explain issues such as residential choice or mobility
choices, as a complement to this quantitative data. Research on the
explanatory variables would offer a more comprehensive understanding of
territorial trends such as: urban shape, accessibility, house prices, housing types,
urban density and migration behavioural conditions. Future research on the
causes or consequences of this migration phenomenon could complete the
picture. Among other parallel processes, there is an important one: the
development of teleworking and digital culture, which may lead to a significant
change in the factors of mobility and residential choices.
REFERENCES
Ahrend, R., Bétin, M., Caldas, M. P., Cournède, B., Diaz Ramirez, M., Pionnier, P.-A.,
Sanchez-Serra, D., Veneri, P., & Ziemann, V. (2022). Changes in the geography housing
demand after the onset of COVID-19: First results from large metropolitan areas in 13
OECD countries. https://doi.org/10.1787/9a99131f-en

Bayona-I-Carrasco, J., & Pujadas-I-Rúbies, I. (2014). Movilidad residencial y

redistribución de la población metropolitana: los casos de Madrid y Barcelona. EURE
(Santiago), 40(119), 261–287. https://doi.org/10.4067/S0250-71612014000100012

Bayona-I-carrasco, J., & Rúbies, I. P. (2019). Las grandes áreas metropolitanas en

España: del crecimiento y la expansión residencial al estancamiento poblacional.
Documents d’Anàlisi Geogràfica, 66(1), 27–55. https://doi.org/10.5565/rev/dag.576

Bayona-I-carrasco, J., & Rúbies, I. P. (2020). Large metropolitan areas in spain: From
growth and residential expansion to population stagnation. Documents d’Analisi
Geografica, 66(1), 27–55. https://doi.org/10.5565/rev/dag.576

Breuillé, M. L., Gallo, J. Le, & Verlhiac, A. (2022). Residential Migration and the COVID-19
Crisis: Towards an Urban Exodus in France? Economie et Statistique, 2022(536–537), 57–
73. https://doi.org/10.24187/ecostat.2022.536.2084

Champion, T. (1997). The facts about the urban exodus. Town and Country Planning,
66(3), 77–79.

Chrétien, J., Néchet, F. Le, Leurent, F., & Yin, B. (2019). Using Mobile Phone Data to
Observe and Understand Mobility Behavior, Territories, and Transport Usage. Urban
Mobility and the Smartphone: Transportation, Travel Behavior and Public Policy, 79–141.
https://doi.org/10.1016/B978-0-12-812647-9.00003-2

Elejalde, E., Ferres, L., Navarro, V., Bravo, L., & Zagheni, E. (2023). The social stratification
of internal migration and daily mobility during the COVID-19 pandemic. arXiv preprint
arXiv:2309.11062.

Feria-Toribio, J. M. (n.d.). Problemas de definición de las áreas metropolitanas en

España.

Fielding, T., & Ishikawa, Y. (2021). COVID-19 and migration: A research note on the
effects of COVID-19 on internal migration rates and patterns in Japan. Population,
Space and Place, 27(6), e2499. https://doi.org/10.1002/PSP.2499

Florida, R., Rodríguez-Pose, A., & Storper, M. (2023a). Critical Commentary: Cities in a
post-COVID world. Urban Studies, 60(8), 1509–1531.
https://doi.org/10.1177/00420980211018072

Florida, R., Rodríguez-Pose, A., & Storper, M. (2023b). Critical Commentary: Cities in a
post-COVID world. In Urban Studies (Vol. 60, Issue 8, pp. 1509–1531). SAGE Publications
Ltd. https://doi.org/10.1177/00420980211018072

Florida, R., Rodríguez-Pose, A., & Storper, M. (2023c). Critical Commentary: Cities in a
post-COVID world. Urban Studies, 60(8), 1509–1531.
https://doi.org/10.1177/00420980211018072/ASSET/IMAGES/LARGE/10.1177_0042098021
1018072-FIG2.JPEG

Gil-Alonso, F., López-Villanueva, C., Bayona-i-Carrasco, J., & Pujadas, I. (2021). Towards
an Even More Spatially Diversified City? New Metropolitan Population Trends in the Post-
Economic Crisis Period. Urban Science 2021, Vol. 5, Page 41, 5(2), 41.
https://doi.org/10.3390/URBANSCI5020041

Green, M., Pollock, F. D., & Rowe, F. (2021). New forms of data and new forms of
opportunities to monitor and tackle a pandemic. In G. J. Andrews, V. A. Crooks, J. R.
Pearce, & J. P. Messina (Eds.), Covid-19 and similar futures (pp. 423–429). Springer.

Glaeser, E. L. (2022). Reflections on the post-Covid city. In Cambridge Journal of

Regions, Economy and Society (Vol. 15, Issue 3, pp. 747–755). Oxford University Press.
https://doi.org/10.1093/cjres/rsac039

González-Leonardo, M., López-Gay, A., Newsham, N., Recaño, J., & Rowe, F. (2022).
Understanding patterns of internal migration during the COVID-19 pandemic in Spain.
Population, Space and Place, 28(6), e2578. https://doi.org/10.1002/PSP.2578

González-Leonardo, M., Rowe, F., & Vegas-Sánchez, A. (2023). A ‘donut effect’?

Assessing housing transactions during COVID-19 across the Spanish rural–urban
hierarchy. Regional Studies, Regional Science, 10(1), 471-473.

González-Leonardo, M., Rowe, F., & Fresolone-Caparrós, A. (2022). Rural revival? The rise
in internal migration to rural areas during the COVID-19 pandemic. Who moved and
Where?. Journal of Rural Studies, 96, 332-342.

Le Néchet, F., Nessi, H., & Aguiléra, A. (2016). La mobilité des ménages périurbains au
risque des crises économiques et environnementales. Geographie Economie Societe,
18(1), 113–139. https://doi.org/10.3166/GES.18.113-139

Ministerio de Transportes y Agenda Urbana (2022). (n.d.). Atlas de áreas urbanas 2022

Plan Urbanisme Construction Architecture (2018). Exode urbain, mythe ou réalité ?

Ramani, A., & Bloom, N. (2021). The Donut effect of COVID-19 on cities (No. w28876).
National Bureau of Economic Research.

Rowe, F., Calafiore, A., Arribas-Bel, D., Samardzhiev, K., & Fleischmann, M. (2023). Urban
exodus? Understanding human mobility in Britain during the COVID-19 pandemic using
Meta-Facebook data. Population, Space and Place, 29(1).
https://doi.org/10.1002/psp.2637

Rowe, F., González-Leonardo, M., & Champion, T. (2023). Virtual special issue: Internal
migration in times of COVID-19. Population, Space and Place, 29(7), e2652.
https://doi.org/10.1002/PSP.2652

Singh, R., Manchanda, N., & Mishra, R. (2022). Internal student migration in India: Impact
of the COVID-19 crisis. Asian and Pacific Migration Journal, 31(4), 454-477.

Vich, G., Marquet, O., & Miralles-Guasch, C. (2017). Suburban commuting and activity
spaces: using smartphone tracking data to understand the spatial extent of travel
behaviour. Geographical Journal, 183(4), 426–439. https://doi.org/10.1111/GEOJ.12220

Vogiazides, L., & Kawalerowicz, J. (2022). Urban exodus in Covid times: Who moves out
of the inner city of Stockholm and where do they go? Stockholm Research Reports in
Demography, 2022, 4. https://doi.org/10.17045/STHLMUNI.19248190.V1

Weisbuch, G. (2021). Urban exodus and the dynamics of COVID-19 pandemics. Physica
A: Statistical Mechanics and Its Applications, 569, 125780.
https://doi.org/10.1016/J.PHYSA.2021.125780

Whitaker, S. D. (2021). Did the COVID-19 Pandemic Cause an Urban Exodus? Cleveland
Fed District Data Brief, 20210205. https://doi.org/10.26509/FRBC-DDB-20210205

Wolff, M., & Mykhnenko, V. (2023). COVID-19 as a game-changer? The impact of the
pandemic on urban trajectories. Cities, 134. https://doi.org/10.1016/J.CITIES.2022.104162
Annex I – References reviewed

Scope – Dataset - Source Evidence for the donut effect

Reference

Spain National residential variations All cities display overall negative net migration rates
(González- records / Estadistica de during the pandemic in 2020. These negative net
Leonardo et al., Variaciones Residenciales balances are exacerbated in Madrid, Barcelona and
2022) (Instituto Nacional de Valencia.
Estadistica).
Data from 2016-2021.
Sweden- Statistics Sweden (2022a). Individuals moved out of the Stockholm inner city
Stockholm Migration by region, age and area to a greater degree in 2020 than during the
(Vogiazides & sex. Years 1997-2021 years preceding the COVID-19 pandemic. The
Kawalerowicz, majority relocated to Stockholm’s suburbs, which
2022) experienced net migration gains, both in the period
2015-2019 and in 2020, which indicates an increasing
trend.

Japan - Tokio Japan Household registration Japan's “one-point concentration” has been
(Fielding & system weakened by the pandemic; the largest
Ishikawa, 2021) Annual Report on Internal metropolitan city has lost out, and the emptier parts
Migration in Japan Derived of Japan, with previous depopulation trends, have
from the Basic resident gained.
Registers published by the
Statistics Bureau at the Ministry
of Internal Affairs and
Communications
Britain Anonymized aggregated Evidence of higher‐than‐average patterns of mobility
(Rowe et al., mobile phone location data from highly dense population areas to low densely
2023) from Meta‐Facebook users populated areas during lockdowns and overall levels
comprising 21 million of mobility declined across Britain.
observations for Great Britain,
covering an 18‐month period
from 23 March 2020 to 15
August 2021. Two data sets,
Facebook Movements and
Facebook Population, created
by Meta and accessed
through their Data for Good
Initiative
France 100,193 flows of residential The probability of an urban resident choosing to live in
(Breuillé et al., migration intention obtained an urban space rather than a rural space is 0.64 lower
2022) from the real estate browser than before COVID.
Meilleurs Agents during three
years before covid and during
COVID.

USA Zillow Group: residential Evidence supporting a donut effect in migration

(Ramani & Bloom, property rental rates and patterns. Population and business move from CBDs,
n.d.) property values at zip code mostly to suburbs but also to other smaller metro
and msa levels. areas.
WFH exposure: amount of
potentially teleworking work of
the residents of an area based
on the percentage of each
industry employing the
residents in an area.
CBD and msa density
characteristics to de zip code
NCOA dataset on migration
flows each month for
measuring pop inflows and
outflows.
Data Axle dataset on
household characteristics.

Table A - References reviewed for the datasets and methodologies and evidence of the donut
effect
Scope – Conceptual framework Key Results
Reference
Spain Early evidence of an urban During the pandemic in 2020, a pattern of population
exodus emerging. deconcentration emerged reflecting net migration losses in
González‐ Trends in Spain: Low internal high‐density areas and net migration gains in more sparsely
Leonardo, M., migration compared to other populated locations.
López‐Gay,A., countries By contrast, rural areas recorded a decrease in out‐migration of
Newsham, N., Three preexisting processes: 12.6% (296k–258k), which was met by a rise in in‐migration of
Recaño, J., & urbanization, suburbanization, 20.5% (288k–347k).
Rowe, F. (2022) counterurbanization. All cities displayed overall negative net migration rates during
the pandemic in 2020. These negative net balances were
greater in Madrid, Barcelona and Valencia.
Pattern of population deconcentration is particularly related to
movements from the large urban agglomerations to specific
locations involving mountain rural areas and certain coastal
towns.
Sweden- Counterurbanization before Individuals moved out of the inner city of Stockholm to a higher
Stockholm the pandemic: Social degree in 2020 than during the years preceding the COVID-19
Vogiazides, L., factors+Economic factors. pandemic. The majority of movers relocated to Stockholm’s
Kawalerowicz, Counterurbanization in times suburbs, which experienced a substantial increase in the
J. (2022). of covid. proportion of inflows from the Stockholm inner city
Other empirical evidence Suburban metropolitan municipalities and medium-sized
during covid pandemic: USA, city municipalities experienced net migration gains, both in
Japan, Netherlands the period 2015-2019 and in 2020: an increasing trend.
Interesting pattern: the fact that those born elsewhere in
Sweden have existing social networks in places other than the
Stockholm region, and work-related characteristics (self-
employed moved more to small cities and rural areas).

Japan - Tokio Context of covid in Japan: The shift is negative everywhere; not a single prefecture of origin
Fielding T, restrictions, migrations and sent more migrants to Tokyo in 2020 than in 2019.
Ishikawa Y. cultural features. Migration velocities: (1) the “distance-decay effect” whereby
(2022) Existing trends before covid in the rates of flow fall away with increasing distance from Tokyo
Japan’s migration system. and (2) the slight but distinctive bias towards flows to other
Contexts of the three major metropolitan and major city destinations.
metropolitan areas and their Japan's “one-point concentration” has been weakened by the
relations with rural spaces pandemic; the largest metropolitan city has lost out, and the
emptier parts of Japan, with previous depopulation trends,
have gained.
Britain Evidence for mobility patterns Evidence of an overall and sustained decline in human mobility
Rowe, F., during COVID. Key issues: between areas during lockdowns between March 2020 and
Calafiore, A., teleworking, housing February 2021.
Arribas‐Bel, affordability, job distribution, Declines in mobility levels between areas in Britain varied
D.,Samardzhiev, digital technologies, markedly across the urban hierarchy, with the most densely
K., & suburbanization and populated areas experiencing the largest reductions; that is, a
Fleischmann, M. counterurbanization. 60% decline from pre-pandemic levels.
(2023). City benefits for social Evidence of higher‐than‐average patterns of mobility from high
interaction, limits of rural population density areas to low population density areas during
areas. lockdowns and overall levels of mobility declined across Britain

France Residential choices during The probability of an urban resident choosing to live in the same
Breuillé, M., Le COVID from a real estate urban area is 0.87 lower.
Gallo, J., market perspective. The global probability of looking for a home in an urban area
Verlhiac, A. Importance of the sociology rather than a rural one is 0.92 lower than at the beginning of
(2022). and time of life of people for COVID.
this residential choice. The probability of an urban resident choosing to live in an urban
Some parameters changed area rather than a rural area is 0.64 lower than before COVID.
during COVID: WFH, distance
to work, etc.

USA The effect of Covid-19 and Evidence supporting a donut effect in migration patterns.
Arjun Ramani the rise of WFH on migration Population and business moves from CBDs, mostly to suburbs
and Nicholas patterns and real estate but also to other smaller metro areas.
Bloom (2023) markets in American cities. This effect is a large city phenomenon, clear evidence in the 12
Migration trajectories largest cities, no evidence in the rest.
Impacts of Covid on Approximative distribution of where people go: 3 in 5 stay in his
economic and geographical metro area but in the suburbs, the rest to other metro areas,
structure. only 4% to rural spaces.

Table B - References reviewed for the conceptual framework and key results.