COVID-19 as a game-changer? The impact of the pandemic on urban trajectories

Will the COVID-19 pandemic interrupt the recent European urbanization trends – and if so – what is the magnitude of this sudden shock, and how deaths, births, and net migration contribute to this disruption? Until now, most discussions on the topic have circled either around the anecdotal evidence of city center decline, or contrarian speculations about residential inertia and the forthcoming business-as-usual. Bringing clarity to the uncertainty and confusion surrounding COVID-19, this paper seeks to detect overarching patterns in and the magnitude of its sudden shock to long-term urban trajectories, understood as a reversal of the pre-pandemic population development trend, across European cities in the early 2020s. It reveals that during the first year of COVID-19, population growth in European cities significantly slowed down to −0.3 % per annum, with 28 % of all European cities having experienced a U-turn from population growth to loss. Out-migration was the main driver of such rapid urban shrinkage, while excess mortality associated with COVID-19 has also contributed to population loss in several European city-regions; some, especially, smaller cities suffered from a significant drop in birth rates. Based on the factorial, hierarchical, and temporal dimensions of the COVID-19 crisis, the paper provides a plausible forecast about the future of Europe's post-coronavirus city.


Introduction
The COVID-19 pandemic continues to challenge the world as we know it since its beginning in early 2020, focusing one's attention on the role of cities. Being instrumental in the spread of the novel coronavirus SARS-CoV-2 due to their manifold transport connections overall, cities have also experienced a severe impact of the pandemic itself (UN Habitat, 2021;European Union, 2020;Sharifi & Khavarian-Garmsir, 2020).By embedding cities within their long-term development trajectories, a number of urban scholars (Dodds et al., 2020;Florida et al., 2021;Siedentop & Zimmer-Hegmann, 2020) have recently questioned to what extent the pandemic could lead to a growth slow-down or even an interruption to hitherto known urbanization trends?This paper intends to contribute to this research agenda by focusing on European cities.
Cities have been afflicted by pandemics throughout time and space.Historically, the scientific literature would predominantly concentrate on the number of deaths caused by a pandemic as the relevant demographic factor, leading to tremendous deviations from the previous urban trajectories (e.g.Rieniets, 2009;Krüger, 2006).During the Black Death of 1347-1351, causing the loss of 39 % of the affected population across Europe, cities were the hardest hit because of high population densities and crowded living quarters, making disease transmission easier (UN Habitat, 2021).During Black Death, between 60 and 70 % of Florence, Hamburg, or Bremen's population perished; while many people fled from cities to the countryside in hope of saving themselves from the disease, seeking jobs, or due to persecution (Cohn, 2002).This out-migration, in turn, facilitated the plague's rapid spread between different communities, resulting in severe depopulation across the continent.
Migration plays a central role in the transmission of a virus.For example, the Black Death first entered Messina in Sicily via a fleet of Genovese trading ships in 1347, and spread farther via the city's shipping routes.Similarly, the first reported cases of the Third Plague in 1899 were found in port cities and shipping harbors such as Lisbon, Marseille, Paris, and Piraeus (Bramanti et al., 2019).Cholera, for people, preventing the establishment of new relationships, postponing marriages, and reducing the frequency of sexual intercourse (Lehmiller et al., 2021).Urban residents also have demonstrated a stricter preventive behavior during lockdowns than rural ones (Chen & Chen, 2020), leading to a sharp drop in urban fertility recorded in Brazil, Italy, Spain, France, and Germany (Lima et al., 2022;Rose et al., 2022;Arpino et al., 2021).
Third, several European studies have documented a high increase in migration outflows and decline in migration inflows during the pandemic, especially across larger cities in Germany (Stawarz et al., 2022), Spain (González-Leonardo et al., 2022), Sweden (Vogiazides & Kawalerowicz, 2022), the Netherlands (Klopper & Kooiman, 2021), and the UK (Rowe et al., 2022).Moreover, public policy responses to the COVID-19 pandemic, including travel restrictions, work-from-home (WFH) decisions, and lockdowns have had a significant impact on international and domestic migration behavior.International travel bans and border closures, introduced to reduce the spread of COVID-19, caused a significant drop in international migration, especially evident in larger cities relying on international in-migration (Wolff et al., 2020).In addition, many work migrants have had to return to their home countries, having lost their jobs in retail and hospitality sectors, which were decimated by lockdowns and sharp economic downturns (IOM, 2022).Government-mandated business shutdowns and social distancing measures have made cities less vibrant and attractive (Honey-Rosés et al., 2020).In addition, the fast spread of COVID-19 infections in dense urban areascombined with the newly-discovered WFH opportunitieshas motivated many households to leave larger cities (Wong & Li, 2020) and cut on commuting travel (Fielding & Ishikawa, 2021).For example, net migration losses in German cities were most pronounced amongst young adults aged 18-29 years (Stawarz et al., 2022).Overall, outmigration from cities has been driven by people relocating to more remote and affordable places; at the same time, high-income households were able to leave cities by relocating to their second homes in the countryside (Pitkänen et al., 2020).
What is the medium-term impact of these three demographic factors health, society, policyon cities, in general?Will COVID-19 interrupt the recent European urbanization trendsand if sowhat is the magnitude of this sudden shock event?Alternatively, how many cities will continue on their previous growth trajectory, irrespective of the COVID-19 pandemic and its repercussions?Until now, most discussions about 'cities after COVID' have either circled around the often sensationalist mass-media coverage of city center decline, or contrarian speculations by panic-driven real estate players about residential inertia and the forthcoming business-as-usual, punctuated by occasional futuristic architectural musings on 15-min neighborhoods.From an academic perspective, the expectation has been for the pandemic shock to hit cities particularly hard, with variable long-term effects, for "the lockdown has become a massive forced experiment in teleworking, remote shopping, dependency on home deliveries, and even in keeping and developing personal relationships" (Florida et al., 2021, p. 11, emphasis added).In the most popular scenario, the COVID-19 pandemic was forecast to reverse the recent Western trends towards (re-)urbanization in favor of population decline and desurbanization (Garrett, 2020;Hernández-Morales et al., 2020;Nathan & Overman, 2020).Some authors have gone as far as to suggest thatif not the coronavirus per se public policy responses to it, including the WHF and hybrid working patterns, and the associated economic impact, could well make urban living obsolete (Batty, 2020;Hernández-Morales et al., 2020).
Others, looking back in history, have argued that "cities have rebounded from the devastation of epidemics and many other types of crisis and catastrophes" and should bounce back yet again (Florida et al., 2021, p. 3).This resilience is typically attributed to the main reason for why cities existthe primary economic role that urban agglomerations continue to play (Scott & Storper, 2015).For instance, a recent survey of 250C-suite executives of the world's largest multinational enterprises carried out by the Global Centre on Healthcare & Urbanisation at M. Wolff and V. Mykhnenko Kellogg College, Oxford, has discovered that 94 % of corporate leaders believe cities in 2030 would either be more important (64 %) or as important (30 %) to their overall business strategy as today (Mykhnenko et al., 2021).At the same time, a study of US cities by Morgan Stanley (2021) has detected a widespread exodus from city centers as a consequence of the pandemic, though not all cities have lost residents.In Germany, according to Wolff et al. (2021), 75 % of all cities experienced population loss in 2020 compared to 25 % in 2019.Indeed, for the first time since the mid-1990s, the average growth rate of German cities has turned negative, with every second large city turning from long-term population growth to population loss in 2020.
A few previous studies have already highlighted that interruptions of urban trajectories are a common phenomenon.Turok and Mykhnenko (2007), in their paper on urban trajectories in Europe between 1960 and 2005, identified recent decline and growth setback amongst the four most common urban trajectories, characterized by a recent period of decline after a longer period of population growth and by a zigzag growth pattern, respectively, encompassing the experience of up to 22 % of European cities, in total.In a follow-up study, Wolff and Wiechmann (2018) showed that the majority (63 %) of European shrinking cities have experienced population loss in at least one fire-year period since 1990.Amongst these temporarily shrinking cities, the largest share experienced losses in the most recent period on record, between 2005 and 2010.In a recent position paper, the European Commission has forecast a rise in the number of shrinking cities, though not necessarily caused by COVID-19 (EC, 2022).By contrast, a paper on the potential impacts of COVID-19 on cities and regions in the EU presented by the European Committee of the Regions could not provide any definite conclusion on their potential longer term urban trajectories (EU, 2020).
To bring clarity to the uncertainty and confusion surrounding COVID-19, this paper seeks to detect overarching patterns in and the magnitude of its sudden shock to long-term urban trajectories, understood as a reversal of the pre-pandemic population development trend, across European cities in the early 2020s.In particular, the following three research questions will guide the paper consequently: 1) To what extent have European cities experienced the sudden shock of COVID-19 to their population growth trajectories?2) How do the three demographic factorshealthcare, societal, and public policyshape the temporal and spatial impact of the COVID-19 shock on European urban trajectories?3) What firm conclusions can one draw from juxtaposing the sudden shock of COVID-19 to Europe's urban trajectories with the long-term interplay of these three demographic factors?

Conceptual framework
This paper's conceptual framework is broadly based on urban resilience theory.It is contended that resilience thinking is beneficial by focusing on "the ongoing capacity of cities to absorb, adapt, transform and prepare for shocks and stresses along the economic, social, institutional and environmental dimensions, with the aim of maintaining the functions of a city and improving response to future shocks" (Figueiredo et al., 2018: 31).Resilience theory distinguishes between exogenous variables and the system as such, e.g., the urban system, with its endogenous variables (Mykhnenko, 2016).Geographically, exogenous variables operate on a macro scale, while the urban system is localspecific, as depicted in Fig. 1 below (Wolff & Haase, 2019).On the macro scale, the drivers of urban change can be slow and gradually progressing, such as ageing, industrial restructuring, rising income disparities, and environmental degradation, or they can be rapid, such as sudden shocks of wars, earthquakes, and pandemics (Haase et al., 2014;JRC, 2019;Walker et al., 2012).Furthermore, the dynamics of the drivers of rapid change are shaped by the slow variables, e.g., when a pandemic hits an ageing and highly urbanized society.
Central to urban resilience theory is the term 'shock' which is used in order to describe rapid macro-processes defined as events that can cause sudden, drastic changes of "partly unexpected nature that represent challenges for existing […] systems and mechanisms" (Haase et al., 2012, p. 3).Examples include demographic shocks, with rapidly dropping fertility rates (Kohler et al., 2002), economic shocks, with galloping inflation (Everson, 2013), fiscal shocks, with a sudden loss of tax revenues (Calomiris et al., 2010), as well as environmental catastrophes, wars, refugee crises, and epidemics (JRC, 2019;Rieniets, 2009).While the COVID-19 pandemic undoubtedly represents a severe crisis (OECD, 2021), we prefer the term shock for evaluating the pandemic's impact on urban trajectories.Unlike urban crises caused by the so-called slowburn events like de-industrialization or urban sprawl, sudden shocks: • are surprising in scope and impact; • can have either positive or negative consequences for a particular urban trajectory; • can have either temporary or permanent effects on cities; and • are next to impossible to be integrated into 'normal' urban planning and strategic thinking due to their unpredictable nature (Figueiredo et al., 2018;Haase et al., 2012;UN-Habitat, 2017).
Most importantly, according to the ecological reading of resilience, there are interrelations between exogenous and endogenous variables, as the system responses to slow and/or fast drivers of stress (stressors) leading to a new steady state of the system post-shock (Walker et al., 2012).Hence, it is the relationship between the endogenous variablesin our conceptualization the three demographic factors of healthcare, societal, and public policywhich evolves as a response to exogenous variables (Fig. 1).This response shapes the direction and magnitude of changes that the system has to undergo, measured as a deviation from the pre-shock system's performance trajectory.As shown in Fig. 1, city A suffers a severe hit, resulting in a radical deviation from its previous steady state, while city B is able better to absorb the disturbance triggered by the exogenous variables and, consequently, experiences just a slight dip in its urban trajectory.Moreover, the way exogenous factors impact upon local social systems and the way cities respond to sudden shocks and slow-burn events largely depends on the urban governance in place, which acts as a membrane between the social system and its global, regional, and local scale stressors (Haase et al., 2014).For a city, the governance mechanisms can be particularly decisive in its ability to return to the original steady state, depending on the 'rotation' or 'spin' the local governance actors have to handle initially, and, subsequently, within the feedback loop of shocks and after-shocks of different intensity and duration.

This paper's analysis covers 915 European cities in 24 European
Economic Area (EEA) member states and the United Kingdom.The cities under investigation represent local administrative units with a population of at least 50,000 inhabitants (21 cities were smaller in 2021) as defined by the Urban Audit delineation (Eurostat, 2020).We have created an urban database with open access data derived from the Urban Audit dataset, combined with the national databases based on Eurostat figures, for all the countries with the most recent data.The occasional data gaps were filled by using the population statistics from the relevant national statistical agencies, after carefully checking for comparability.The urban demographic data on which the paper is based were collected for each individual year between 20121 and 2021, both for the cities and their countries; subsequently, the data were transformed into indicators for further processing (see Fig. 2).
The first data-processing step has involved an assessment of how well cities were able to absorb the sudden COVID-19 shock to their growth trajectory in 2020.This was achieved by analyzing the variations of individual urban trajectories both in relative and absolute terms.For the relative assessment, we have calculated the difference between the mean annual population change rate between 2016 and 2020, and the  population change rate between 2020 and 2021.We have chosen 2016 as the starting point, for European cities were on a continuous population growth trajectory of 0.25 % per annum (p.a.) since then.The emerging delta value 2016to2020-2020to2021 in percentage points (p.p.) was plotted on a map together with its significance.The delta value is deemed significant, if it is outside the range defined as Mean 2016to2020 ± Standard_Deviation 2016to2020 .For the absolute assessment, we have developed a taxonomy consisting of six categories; each of the growth change rates for the period between 2016 and 2020 was classified as follows: cities characterized by population growth (≥3 % change rates >0 %), population decline (≥3 % change rates ≤0 %), or oscillating between positive and negative growth (2 change rates >0 % and 2 change rates ≤0 %).These three pre-COVID19 urban growth trajectories have been contrasted with the trajectory 2020-2021, divided into growth (>0 %) and decline (≤0 %).For the emerging (3 × 2) six urban trajectory types, we have calculated the average 2020 population growth rate as well as the delta value 2016to2020-2020to2021 and checked for statistically significant differences using Games-Howell post-hoc test, with a significance level of 0.05 (i.e., the confidence interval of 95 %).
During the second data-processing step, we have assessed the stress impact of the three demographic factors by calculating the variation between births, deaths, and net migration statistics in time and across the relevant spatial scales (see Fig. 2).The temporal variation was calculated as the percentage point (p.p.) difference between 2019 and 2020 for each of the three indicators, referred thereafter as p-score.The spatial variation was calculated as the p.p. difference between the pscore of a city and its national p-score.For the temporal variation, the following three groupings were assembled for each of the three demographic indicators (births, deaths, and net-migration): decrease (≤0 p.p. -½ Standard Deviation), stable (0 p.p. ± ½ Standard Deviation), and increase (≥ 0 p.p. + ½ Standard Deviation).In a similar vein, three groupings were created for the spatial variation, too: slower than the national average (≤0 p.p. -½ Standard Deviation), similar to the national average (0 p.p. ± ½ Standard Deviation), and faster than the national average (≥ 0 p.p. + ½ Standard Deviation).Furthermore, summary graphs for all analyzed indicators, with the mean values for cities between 2012 and 2021, as well as the corresponding national values were generated to contextualize the variations detected.Finally, we have contrasted the three demographic indicators with the population change rate in 2020, using bivariate scatterplots and the coefficient of determination (R 2 ) as a measure for the relation between the two sets of values.This was done to test each of the indicator's explanatory power vis-à-vis urban population growth or decline.In addition, using the scatterplots analysis, we have also tested the interrelationship between births, deaths, and net-migration versus city population size to detect differences in the magnitude of the COVID-19 shock on different city sizes by 2021.
During the final third step of the data analysis undertaken here, we have contextualized the interplay of demographic factors between 2012 and 2020.Furthermore, we have accounted for differences in terms of regional belonging, city size, and the major demographic driver behind each city's population change.For the regional classification, we have used the World Unified Territorial System (WUTS) Level 4 regions (ESPON, 2014).City size was classified following the OECD (2022) into small cities (<200,000 inhabitants), medium-sized cities (200,000-500,000), and metropolitan areas (>500,000).For the emerging regional clusters as well as the three city size classes, we have calculated the average rates of natural balance (birth-death ratio) and migration balance (net migration) per 1000 persons for each year.As a result, we have developed a Europe-wide urban typology, distinguishing between growing cities, which are either dominated by in-migration or by natural gains, and shrinking cities, which are either dominated by out-migration or natural losses.Using the natural and migration balance figures, we have grouped all the European cities into one of these four types for the period 2012-2020, before producing maps and frequency diagrams, as visual aids and analytical tools.
As Fig. 3 shows, about one in five European cities experienced the sudden shock-induced decline in population growth of more than − 1.0 p.p..This has been particularly evident in Spain (64 cities), Germany (34), Italy ( 14), Bulgaria (11), and, to some extent, also in Hungary (6 cities), Poland (4), and Sweden (3).Amongst the 66 European metropolitan areas (i.e., cities with ≥500,000 inhabitants in 2021), 61 metro areas, or 93 % in total, faced a drop in population growth rates.Barcelona and Madrid (<− 2.0 p.p.), Zaragoza, Valencia, and Milan (<1.5 p. p. each), Budapest, Stuttgart, Berlin, Stockholm, and Amsterdam (<1.0 p.p. each) were the top 10 European metro areas with the most decelerating population growth rates.Other capitals such as Rome, Oslo, Copenhagen, Sofia, Brussels, and Warsaw experienced a <− 0.5 p.p. decline.Only the Lithuanian capital city, Vilnius, managed to buck the trend, accelerating its population growth rate during the pandemic.An increase in urbanization rates in other cities across the Baltic region was detected in 2020, averaging at +0.11 p.p. in Latvia, and +0.93 p.p. in Lithuania.These cities received a sudden inflow of returning migrant workers, escaping protracted lockdowns, which decimated retail, tourism, and hospitality businesses in western and northern Europe.
Other individual examples of accelerating urban growth under COVID-19 were detected in Portugal (7 cities), Southern Italy (12 cities), and the Netherlands (11 cities).
Whereas only 23 % European cities (211 out of 915) have been affected by a sharp deceleration in urban growth rates, as many as 63 % suffered absolute population loss under COVID-19.Amongst these shrinking European cities, 28 % had been shrinking pre-COVID-19 and, thus, continued alongside their trend of population loss, while ×27 % of the cities suffered a severe trend disruption, shifting from population growth to shrinkage in 2020.This sudden shock-induced trend disruption has affected more than half of Spanish, German, Belgian, and Italian cities (see Fig. 3).Berlin, Roma, Madrid, Sofia, and Zagreb have all now become newly-shrinking capital cities.In other countries faced with a rapid slowdown in urbanization, e.g., Bulgaria and Hungary, urban trajectories did not really alter in absolute terms, for almost 80 % of these cities have simply continued loosing population.By striking contrast, in Sweden, where the average population growth rate has slowed down, all cities have continued growing.Furthermore, in other Scandinavian countries, as well as in Austria, the Netherlands, Switzerland, and the UK >2/3rd of cities have continued along the previous upward growth trajectory.In Latvia, where all cities had been shrinking pre-pandemic, negative population growth rates persisted in 2020, though at a slightly slower pace than before COVID-19.Nine European cities, including Kaunas and Siauliai in Lithuania, Coimbra and Viseu in Portugal, Trencín in Slovakia, and Aberdeen and Luton in the UK, have experienced a rare trend reversal during this period from population loss to growth.Noticeably, most of the newly-growing cities have seen some green shoots of recovery in 2019 already.Amongst the newly-shrinking cities, however, the vast majority began losing population only under COVID-19 in 2020.

How deaths, births, and migration shaped cities in 2020
For all European cities the number of deaths per capita (death rate) increased by 13.4 % between 2019 and 2020 (from 9.2 to 10.4).The explanation power of death rate for population change weakened from 2019 to 2020, for in 2020 fast declining cities experienced low death rates, while growing cities showed high deaths rates (R 2 dropping from 0.34 to 0.25, Fig. A1).The increase in the number of deaths highlights high regional variations (Fig. 3): death rates increased fastest in cities across Spain, Slovenia, Belgium, Romania, Italy, Bulgaria, and Poland (>15 %).By contrast, the death rate increases were lowest in Norway, Finland, Estonia, and Germany (<5 %), while in Danish cities the death rates even decreased (− 1.8 %).In most European nations, death rates increased faster in cities than in the rest of the country, especially across Spain, Romania, Northern Italy, Latvia, Croatia, and Western Hungary.By contrast, in the Czech Republic, Switzerland, and the Netherlands the death rate of cities scored below their national averagetogether with several cities in Southern Italy, Western Poland, and Spanish costal regions (Fig. A1).
The birth rate across all European cities dropped by − 3.6 %, from 10.4 in 2019 to 10.0 in 2020 (Fig. 3).Compared with 2019, a decreasing birth rate became quite influential in driving population loss in 2020 (R 2 dropped from 0.10 to 0.25; see Fig. A1).The drop in both rates between 2019 and 2020 was particularly rapid in cities across Romania, the Baltic states, Spain, Poland, and Belgium (<− 5 %).Moreover, in Romania, Belgium, Latvia, Austria, and Croatia, urban birth rates dropped faster than their corresponding national average.By contrast, and, especially, in Lithuania, United Kingdom, and Slovenia the decline in urban birth rates was slower than the national average.During the pandemic, the average birth rate increased only in Finland (by 3.3 %), while some sporadic examples could also be found in Germany, Italy, Hungary, and the Czech Republic.
In 2020, net migration continued to be the main driver of urban growth, though not as much as in 2019 (R 2 dropped slightly from 0.81 to 0.79, see Fig. A1).The average net migration in European cities dropped significantly by − 137 % from a net gain of 3.1 migrants per capita in 2019 to a net loss of − 1.1 migrants per capita in 2020.This decrease was particularly fastest in cities across Bulgaria, Romania, Hungary, and Poland (<400 %), but also in Germany, Spain, and Italy (<− 130 %).In several countries, we have detected a bipolar migration tendency, with a negative net migration in cities compared to a positive migration balance for the corresponding nation, as a whole, including Bulgaria and Croatia, and, to some extent, Slovakia, Switzerland, Denmark, Slovenia, and Finland.By contrast, in Latvian, Lithuanian, and British cities the average net migration increased between 2019 and 2020: here the cities have been the main destination of international in-migration.There are also cities in Germany, Italy, Poland, Portugal, and Bulgaria, with improved net migration figures in 2020, compared to 2019 (Fig. 4).
The findings about population change rates between cities of different sizes suggest that the inter-city differences were high in the early 2010s, but were continuously decreasing past 2016, before dissipating almost entirely by the end of 2020 (R 2 = 0.06, Fig. A1).This was a result of the fast decline in net migration across almost all city size classes.In terms of fertility, however, the difference between city size classes has remained almost constant since the early 2010s: very low birth rates are becoming dominant the smaller a city is (R 2 = 0.12, Fig. A1).In addition, average death rates have significantly diverged between cities of different size since 2012, with mortality levels increasing faster in smaller cities than in larger ones.In 2020, this difference in death rates peaked, but the relation between mortality and city size under COVID-19 became very weak, on average, as almost all cities experienced an increase in the death rate (R 2 = 0.04, Fig. A1).

Contextualization of demographic factors
This sub-section seeks to provide a synthesis of the interplay of the three demographic factors by focusing on natural and migration balance.First, we will trace the interplay between the two, accounting for regional and city-size differences.Second, we will display the dominant demographic factors driving population change in urban Europe, and uncover major changes in this process over the past decade.
In 2020, cities in Europe experienced the fastest change in both natural and migration balance since 2012, but with very significant regional variations (Fig. 5).Cities in West-Central and Northern Europe were experiencing positive natural change and migration balances before the sudden shock of COVID-19.In 2020, out-migration from cities in West-Central Europe increased, while, in Northern Europe, cities faced the drop in natural population balance faster than in terms of migration balance.In addition, for the first time since 2012, in 2020, West-Central European cities experienced a fast drop of natural population change, while in Northern Europe, the natural balance remained positive, though continuously decreasing.Cities in East-Central Europe and the Balkans have revealed a similar average profile, with positive migration balances between 2017 and 2019, in parallel with an evergrowing negative natural balance since 2012 (Fig. 5).While netmigration in the Balkan cities have been declining since 2018, cities in East-Central experienced increasingly positive net migration figures between 2014 and 2019.In 2020, in both regions, these two population change drivers, in particular, net-migration, decreased and turned negative.In fact, across all European regions, natural population balance decreased, accompanied by net migration losses.In addition to the Balkans, this was particularly visible in Southern Europe.Between 2015 and 2019, South European cities had experienced population growth only because of positive net migration.From this perspective, it is clear that in most regions some fundamental demographic trends were accelerating even before the COVID-19 pandemic.
In terms of city size, major differences can be observed both before and during the pandemic.In West-Central Europe, throughout the observed period 2012-2020, especially metropolitan areas enjoyed positive natural balance; and these metropolitan areas were the places where the shift towards a decrease in migration balance was at its fastest in 2020.By contrast, a shift towards decreasing natural balance were similar for all North European cities, while only metropolitan areas there experienced a significant drop in migration balance in 2020.In a similar vein, metropolitan areas in Southern Europe experienced the fastest drop in net migration in 2020 amongst all city size classes and regions, while for small cities this drop was less sharp.In East-Central Europe, Fig. 5 reports a big contrast between city size classes in respect to migration: while small cities have experienced negative migration balance since 2012, metropolitan areas have seen positive balance, with migration balance in medium-sized cities growing and turning positive just before COVID-19.In 2020, however, the decrease in population migration balances was similar throughout the urban hierarchy across East-Central Europe, whereas natural population balance decreased fastest in small cities and slowest in metropolitan areas (Fig. 5).Cities in the Balkans have demonstrated an even clearer contrast with respect to city size: natural balance remained negative across all three size classes between 2012 and 2020, but was decreasing faster the smaller the city was.In terms of migration balance, metropolitan areas have demonstrated a high and positive rate since 2012, while medium-sized cities had experienced an even stronger growth than metropolitan areas before 2015.In 2020, migration balance dropped faster the smaller the city size, while remaining positive in Balkan metropolitan areasa significant difference in contrast with other European regions.This paper's findings reported above have already pointed to a change in the constellation of powerful demographic factors driving   M. Wolff and V. Mykhnenko urban population change in Europe during the COVID-19 pandemic, to be further analyzed and broken down here.First, European cities, where population change was dominated by in-migration, experienced a fast Uturn in 2020.Before the pandemic, growing cities were becoming increasingly reliant on in-migration (53 %, in total); in particular, this refers to the European economic powerhouses, including Germany, Northern Italy, the UK, the Benelux, as well as Spain, Northern Europe, and several rapidly growing regions in East-Central Europe (Fig. 6).In 2020, the share of migration-driven urban growth halved to 25 %, remaining persistent only in the Netherlands and, to some extent, across Northern Europe.The 230 European cities belonging to this type accounted for a net migration gain of 253,000 persons in 2020.By contrast, during the COVID-19 pandemic, the share of shrinking cities, driven by out-migrationwhich was diminishing until 2019 -exceeded the share of growing cities dominated by in-migration.In particular, this type was present in Western Germany or Southern Spain, where cities had grown through in-migration before 2020, and in East-Central Europe and the Balkans, where cities were shrinking as the result of out-migration, even before COVID-19 (Fig. 6).In 2020, Europe's 280 shrinking cities driven by out-migration accounted for a migration balance of − 315,000 (Table A.1).
Furthermore, amongst both growing and shrinking cities, the trend towards net natural loss has remained unbroken since 2012, peaking in 2020.The share of growing cities dominated by natural gains has been continuously decreasing, reaching just 8 % in 2020.In 2020, 73 cities across Europe, particularly, the larger ones still grew, although net migration balance turned negative in Amsterdam, Oslo, Munich, Brussels, Stockholm, Frankfurt-on-Main, and Copenhagen (Fig. 6).In 136 cities, e.g., in Zagreb, Dortmund, Gdańsk, and Bologna, net migration gains experienced during the pandemic were eaten up by natural population losses.At the same time, the share of shrinking cities driven by natural population loss has been continuously increasing since 2012, peaking at 36 %, in total, in 2020.These 332 shrinking cities have accumulated a death surplus of 242,000 personsa number almost equal to the number of migration gains of growing migration-driven cities (Table A.1).In particular, urban population growth dominated by migration gains in 2019, turned into population loss dominated by natural losses in 2020, across Germany, Northern Italy, and Spain (Fig. 6).

Discussion and conclusions
Learning from the ecological and adaptive evolutionary perspectives on urban resilience, as conceptualized in this paper (see Fig. 1), one has to focus on the interrelation between the exogenous and endogenous variables at different scales, as the first step towards building a predictive capacity vis-à-vis the social phenomena under investigation.In the context of urban Europe today, to uncover the systemic response of individual cities to the sudden shock of COVID-19, this paper has concentrated on the pandemic as an exogenous (macro-) scale variable, documenting its impact upon the three local endogenous (micro-) variables of health, society, and policy.By applying the demographic indicators of population change, death rates, birth rates, and net migration flows (see Fig. 2), this paper i) has uncovered, in depth, the direction and magnitude of change that European cities experienced during COVID-19, and ii) has measured effectively the pandemic-related disruption to the continent's pre-shock urban trajectories.In this final section, we summarize the key findings of this study, draw up three distinctive analytical dimensions required to make a plausible forecast about the future of the post-coronavirus European city, before sketching out avenues for further research.

Summary of key findings
During the first year of the pandemic, population change in European cities significantly slowed down, turning into decline at − 0.3 % p.a.
Only in the Baltic States have the urbanization rates increaseda process which had started long before the pandemic.Overall, in 2020, 28 % of all European cities experienced a U-turn in population trends from growth to loss.Adding the previously shrinking cities to the total, the share of shrinking cities in Europe during COVID-19 has reached quite an extraordinary figure of 63 %.This latest indicator greatly exceeds any previously recorded measurement of urban shrinkage in Europe (cf., Turok & Mykhnenko, 2007;Wolff & Wiechmann, 2018).
Yet, to what extent does the pandemic-instigated urban shrinkage indicate a fundamental turnaround in European urbanization, in the long run?Could this not just represent a temporary phenomenon, similar to recent decline (Turok & Mykhnenko, 2007) or temporary shrinkage (Wolff & Wiechmann, 2018) of the 2000s?One could approach this question from three different angles: hierarchical, in terms of differences between cities of different size; temporal, in terms of differences between the timing of urbanization processes and effects, and factorial, in terms of differences between the constituting demographic elements.

Factorial
The pandemic has had unequal impact on different urban demographic factors.The major factor of the slow-down of urbanization during COVID-19 was the fast drop in net migration, by as much as 137 %, especially, in Bulgaria, Romania, Hungary, Poland, Germany, Spain, and Italy.Given the data available, it is difficult to infer to what extent non-urban regions have benefited from this out-migration.Yet, in Bulgaria, Croatia, and, to some extent, in Slovakia, Switzerland, Denmark, Slovenia, and Finland we have uncovered negative net migration in cities, compared to positive migration balance for the respective nation in 2020, implying population growth in the countryside.Thus, the COVID-19-induced decline in net migration was responsible for disrupting and reversing previous urban population growth trajectories towards decline.In half of European cities, the new shrinkage trajectory was caused by out-migration, whereas in the other half, it was driven by natural population loss (i.e., deaths exceeding births).Therefore, excess mortality associated with COVID-19 has been an important demographic factor in European cities, as the number of deaths increased by 13.5 % during the first year of the pandemic.Indeed, in the majority of countries, death rates have increased faster in cities than in the countryside: this has been most pronounced in Spain, Romania, Northern Italy, Latvia, Croatia, and Western Hungary.In addition, in many cities across Romania, Belgium, Latvia, Austria, and Croatia, it was a sharp drop in births that had fostered natural population loss.Overall, the average birth rate in European cities decreased by 3.6 % between 2019 and 2020.

Hierarchical
In terms of city size effects, the pandemic has been a great equalizer: while the difference between the average growth rate for different city size classes has been declining since 2016, it almost entirely disappeared in 2020.The primary reason was the sudden drop in net migration in 2020, which affected cities of all size classes to a similar extent: 61 out of 66 metropolitan areas experienced a slow-down in population growth rates, resulting in absolute population loss in large European capitals like Berlin, Rome, and Madrid.The COVID-19-induced decline in net migration has been especially pronounced in metropolitan areas across West-Central, Southern, and Northern Europe; smaller cities in these regions have experienced a less dramatic shock.By contrast, in East-Central Europe, and in the Balkans, migration balance dropped faster in 2020 the smaller the city was, while remaining positive in metropolitan areas.It appears the sudden change in pre-pandemic migration patterns has accentuated some long-term natural population growth trajectories (Wolff et al., 2022).In particular, the difference in the number of births between city size classes has remained almost constant since 2012, with lower the birth rate the smaller the city was.At the same time, the COVID-19-associated death surplus has hit harder the smaller cities, especially in Spain, Poland, and Romania.Consequently, in East-Central Europe and the Balkans, the natural population change in 2020 saw the sharpest fall in smallest cities, and the flattest fall in metropolitan areas.

Temporal
The sudden shock of COVID-19 has comprehensively altered Europe's urbanization in 2020, accelerating the long-term natural population change trends, and reversing the pre-pandemic net migration patterns.Between 2015 and 2019, the total number of deaths in European cities averaged around 1.6 million, jumping to 1.9 million in 2020.With the number of births in European cities significantly decreasing since 2016, the pandemic's excess mortality has effectively driven major parts of Europe into natural population decline.Overall, natural population growth has been continuously declining across all European cities since 2012.During the pandemic, the drop was especially rapid in smaller cities, generally, and in cities across Southern, East-Central Europe, and the Balkans.At the same time, during COVID-19, West-Central European cities have experienced natural population loss for the first time in almost a decade, and in North European cities natural balance decreased faster than the decrease of net migration.Temporally, it was the exceptional pandemic-induced drop in net migration that has generated the largest sudden shock to urban Europe.In cities across East-Central and Southern Europe, where urban growth until the pandemic had relied exclusively on significant positive net migration, COVID-19 has derailed the previous growth model.

Learning from theory, looking ahead
Looking ahead, one has to evaluate the factorial, hierarchical, and temporal dimensions of the sudden shock of COVID-19 to make a plausible forecast about the future of the post-coronavirus European city.Writing back in April 2021, Florida et al. (2021, p. 18) stressed the temporal dimension of the pandemic: If the duration is short, cities may return to a new normal that is not very different from the previous normal.Some operational and regulatory changes will persist but the overall structure of urban life will remain mostly unchanged […].A century from now, we may think of COVID-19 as yet another 'forgotten pandemic', whose impact was horrific in the moment but quickly washed away in the tide of history.
Indeed, the international business community has overwhelmingly backed the "business-as-usual" scenario in its assessment of the importance of (largest) cities for economic growth in North America, Europe, and Asia-Pacific: with the post-lockdown revival of many urban economies, people ought to flock back to the buzz of city life (see Mykhnenko et al., 2021).Net-migration inflows to metropolitan areas, fully supported by local governance actors, is the scenario that the global C-Suite leaders expect to see during the post-COVID-19 decade (see the City A post-shock big deviation trajectory in Fig. 1).Yet, for the post-shock system's State 2 to approximate the pre-shock system's State 1 (see Fig. 1), the urban system has to provide the level of response commensurate with both the sudden shock (fast drivers) and chronic stress (slow drivers) deviations from its trajectory.In this context, one finds the post-COVID-19 urban revival task ahead frankly daunting, for the following four reasons: 1) The magnitude of net migration, which is need to compensate for the pandemic-induced urban population loss, is huge: while in 2018-2019, European cities, on aggregate, experienced a net migration inflow of >620,000 people, in 2020, net migration balance flipped, registering an outflow of 153,000 inhabitants, in total (for details, see Table A.1). Thus, to pull population back into European cities, the local economies have to recover and/or create a vast number of new well-paid jobs.
The magnitude of net migration, which is need to compensate for the pandemic-induced urban population loss, is huge: while in 2018-2019, European cities, on aggregate, experienced a net migration inflow of >620,000 people, in 2020, net migration balance flipped, registering an outflow of 153,000 inhabitants, in total (for details, see Table A.1). Thus, to pull population back into European cities, the local economies have to recover and/or create a vast number of new well-paid jobs.
2) In addition to the scale of the necessary job creation, any net migration gains might appear at a very gradual pace, unlike the rapid population losses caused by COVID-19.This is because in many sectors, working-from-home (WFH) and hybrid work options will continue after the pandemic.This new social phenomenon may herald further decoupling of one's place of work from one's place of living, and, thereby, seriously altering economic productivity and demographic potential of different localities.At the same time, the WFH and hybrid work options will continue to be relevant mostly to employees of advanced sectors and largest firms, with limited effect on the return-to-city migration trends (Florida et al., 2021).Furthermore, it is not the pandemic itself, which has restricted migration, but public health policy measures: while in 2020, 14 out of 23 European countries issued restrictions on international travel (culminating in total border closures and travel bans), the figure reached 23 countries in 2021 (JRC, 2022).Given that many, especially the largest, cities essentially depend on international inmigration for continuous population growth (Wolff et al., 2020), it is uncertain how long the impact of international migration bans and restrictions would last.3) Apart from COVID-19, other exogenous sudden shocks may derail the wanted economic and demographic revival of urban Europe.While the EU average GDP per capita has seen a considerable recovery in 2021, to Euro 32,270 (in current prices), after the sharp drop to as low as Euro 29,810 in 2020 (Eurostat, 2022), the Russian invasion of Ukraine on 24 February 2022 is projected to result in a significant economic slow-down (Borrell, 2022).So far, the war has already had detrimental impact on inflation, pushing up the price of many commodities, including energy, fuel, and food. of SARS-CoV-2 and a rather haphazard nature of the global vaccination campaign have led to a large increase in excess mortality over time: the number of deaths in the 25 most affected countries increased by 5.4 times, from 55.6 million in 2020 to 298 million in 2021 (JHU, 2022).Although COVID-19 excess mortality should cease over time, European societies face the ever-growing impact of declining birth rate on the local communities and localities (Andersson, 2021).Across European cities, birth rates have been declining increasingly faster with every year since 2017, with the pandemic's economic uncertainties accelerating the low fertility trends even further (UN, 2019).This slow-burning demographic process is essentially responsible for the increasing number of shrinking cities in almost all European regions.The migration gain will be increasingly eaten up by natural losses in the future: the share of declining cities driven by natural losses increased from 18 % in 2019 to 36 % in 2020 (Table A.1).

Limitations
This paper has a number of data and methods-related limitations.Firstly, to proceed we have to use urban population data covering the respective municipal boundaries of cities, potentially raising the problem of over-and under-bounded cities (see Davoudi, 2009).To mitigate this methodological issue, this study uses the urban area definitions provided by the Urban Audit, which is the largest openly available and well-respected dataset (e.g.JRC, 2019), specifically designed for conducting the comparative analysis of European cities.It allows this paper's results to be cross-referenced with previous and follow-up studies, and, most importantly, is based on the Urban Audit's degree of urbanization classification aimed at avoiding data distortions caused by using administrative units of varied sizes and shapes (Dijkstra et al., 2021).This paper uses official population estimates which typically refer to resident population on a particular date (ranging between 1 January and 30 June).The study has been conducted when most countries under investigation were carrying out their general population census rounds, which means that previouspre-pandemicpopulation estimates have not yet been corrected and revised downwards or upwards, accordingly, based on the freshest census data (Pelletier, 2020).Consequently, this study represents a compromise between roughly estimating the magnitude of the pandemic's impact on urban trajectories and comparing it with the previous trends, using annualized comparable demographic data.Finally, from an analytical perspective, the paper is solely focused on selected key demographic processes in European cities; it does not engage with economic or employment shifts, happening across within and between these cities, or between the urban cores and hinterlands.

Conclusions
Writing in late 2022, it feels too early to conclude on the long-term impact of COVID-19 on Europe's urban trajectories.The growth of European cities has become almost exclusively dependent on in-migration; yet migration patterns are unpredictable, as the ongoing pandemic and Russo-Ukrainian war amply demonstrate.In 2020, out-migration from European cities was as sudden as it was unexpectedly substantial, causing even the largest of cities to shrink.By contrast, many of the demographic process, including decreasing birth rates, had already been observed before the pandemic -COVID-19 has just accelerated the existing trends.Four potential developments follow from this paper's findings.Firstly, due to a slow demographic recovery, Europe will suffer from deepening population decline in its continuously shrinking cities, combined with more stagnation and episodic shrinkage bouts in other cities (Wolff & Wiechmann, 2018).Secondly, the shrinking pool of migration-and working-age population caused by natural population decline will, in the long run, result in ever-growing competition between cities for talent: students, graduates, job-starters, and skilled workers.Third, Europe's ageing societies and the continent's natural population decline will not only have an irreversible impact on local labor markets, but will also increase the exposure to future healthcare crises.Finally, the post-COVID-19 economic revival and job creation will benefit the upper urban hierarchy in the first place, helping re-grow and expand the core metropolitan areas, while smaller cities will continue to suffer from death surplus and out-migration.To sum up, while COVID-19 has been a great leveler of urban fortunes, it will be the pandemic recovery that leads to an increasingly uneven demographic development of European cities, driven by size and regional location, and re-igniting urban shrinkage across the continent.

Declaration of competing interest
The authors report there are no competing interests or conflicts to declare.

Fig. 2 .
Fig. 2. Workflow highlighting input data and three analysis steps.Source: authors' own work.

Fig. 5 .
Fig. 5. Evolution of natural and migration balance of cities differentiated between classes of different size and regional belonging 2012-2020.