Introduction

In the past two decades, the history of the European population witnessed several watershed changes. First, the Eastern Enlargement of the European Union (EU) in 2004 and 2007 expanded the EU population from 381 to 494 million (Eurostat, 2020a). This expansion entailed that an additional 100 million people could more easily move across countries as EU citizens, providing a strong source of internal migration in Europe. Another momentous change for the European population came from external migration, coinciding with the European Migrant Crisis that started in 2015. Only in 2015 more than a million refugees reached Europe constituting the largest ever inflow of asylum seekers in Europe (PEW, 2016). In addition to the structural change introduced by the EU expansion and the mass migration flows that necessitated great adjustments for many European countries, the Great Recession of 2008 had a significant impact on the lives of many people in Europe, which in turn had consequences for the general population.

On no account European population dynamics are limited to the unique events of the beginning of the twenty-first century, but are also defined by the long term processes of population ageing, mortality, fertility, and migration. Europe has one of the highest proportions of older persons in its populations in the world (Vobecká et al., 2013; European Commission, 2020). The increase in the proportion of the elderly in European populations has drawn scholarly efforts to identify challenges in socio-economic development of the continent (Bloom et al., 2015). With increasingly ageing populations, the national pension, social, and healthcare systems in the EU are expected to face sustainability challenges (Christensen et al., 2009; Ediev, 2014; Rechel et al., 2013). Hand in hand with ageing comes fertility of the EU member states. In general, the fertility rate of the EU has been below the replacement rate since the beginning of the twenty-first century (Frejka & Sobotka, 2008). Yet regional fertility rate differences exist. While Western Europe and the Nordics have relatively higher fertility rates, some of the Southern and Central and Eastern European countries barely pass the lowest-low fertility threshold of 1.3 (Billari & Kohler, 2004; Goldstein et al., 2009; Human Fertility Database, 2023). Migratory tendencies are heterogeneous across Europe as well. First, demographers have shown that free movement in the EU has magnified intra-European migratory flows (Castro-Martín & Cortina, 2015) and changed cohort migration patterns (Bernard, 2017). Second, international migration in Europe has been growing steadily (Fassmann & Munz, 1992; van Mol & de Valk, 2016). Extra-EU migration is expected to either contribute to population growth or slow down depopulation in Europe (van Nimwegen & van der Erf, 2010). However, internal migration is a crucial component defining subnational populations (Rees et al., 2012, 2017).

Population policies have also played a role in defining the current demographic state of Europe. Facing declining fertility rates and ageing societies, many countries in Europe have introduced various family policies in order to smoothen the socio-economic challenges of the future and in some cases to bolster their demographic resilience of their populations (Ainsaar & Rootalu, 2016; De Souza, 2015). Countries with the prospect of population decline may experience changes ranging from national identity (Teitelbaum & Winter, 1998) to long term economic growth prospects (Bloom et al., 2015). Population ageing is closely connected to social security and pension concerns that are frequent in industrialised countries (Kinsella & Phillips, 2005). Policies that address ageing have mostly focused on fiscal aspects of the matter whilst more controversial ideas such as replacement migration policies have faced public scrutiny (May, 2012). On the other hand, fertility policies that focus on the pro-natalist aspect receive support from governments, but usually fail to deliver expected results as not only the fertility rates do not increase, but the decline in fertility barely slows down when policies are enacted (Grant et al., 2004). Immigration policies seem to have contributed the most with respect to mediating the population ageing or decline in Europe. Yet the European governments aim to balance the need of immigration and concerns it raises among native populations by introducing quotas or qualification requirements (Martin & Zürcher, 2008; van Houtum & Pijpers, 2007).

Population studies have largely overlooked Nomenclature of Territorial Units for Statistics level 3 (NUTS-3) as a unit of analysis at the European level. NUTS-3 constitute relatively smaller regions that are deemed useful for socio-economic analyses that focus on specific diagnoses (Eurostat, 2020b). This is because NUTS-3 provide granular regional data that can be used for more precise analyses than those relying on the larger NUTS-1 and NUTS-2 regions. NUTS-3 regions vary in both population and area providing a considerable degree of variation to explore. There exists a considerable body of research that has looked at subnational population dynamics in Europe, the work either focused on one of the components of population change (e.g. Kashnitsky et al., 2017, 2021) or has utilised higher level statistical units (e.g. Rees et al., 2012). To the best of our knowledge, we are the first to explore the patterns of natural population change, measured as the difference between the number of live births and deaths in a year, and net migration at NUTS-3 level using the longest period time frame from 2000 to 2017. Analysing the population dynamics at NUTS-3 level provides an overview of the recent population status of the continent, highlights the potential of areas for policies ranging from subnational to supranational levels in Europe, and expands on the literature by using all NUTS-3 units rather than selected NUTS-3 or more local level samples in European countries (Gregory & Patuelli, 2015; Gutiérrez Posada et al. 2018; Sabater et al., 2017). We also propose a four-point typology that allows us to identify NUTS-3 regions as highly depopulating, depopulating, exhibiting population growth, and highly growing. Using this typology we produce a map that captures the snapshot of average population dynamics in Europe for the period of 2000—2017. Our four-point typology helps to address policy matters in Europe. This is especially relevant in the light of demographic resilience debate that juxtaposes population size (and related policies) against a government’s ability to capitalise on the population structure (e.g. De Souza, 2015). As population policies are largely crafted at the national level, we assign countries rather than NUTS-3 regions to the four types of population dynamics to check if there exists a connection between population policies adopted in the 1990s and the current average demographic state of European countries. In this way we contribute to the efforts evaluating population policies in Europe (e.g. Ainsaar & Rootalu, 2016). We also expand the broader literature on population policies in fertility (e.g. De Silva & Tenreyro, 2017; Tsui, 2001) and migration (e.g. May, 2012).

This paper outlines the differences in population growth or decline across 2239 NUTS-3 regions across Europe for which Eurostat data on population size, natural population change, and net migration exists over the period of 2000–2017. It also looks at the associations between the population policies adopted in the 1990s and the current state of population dynamics in Europe by employing the data from the United Nations (UN) World Population Policy database, the UN Population Division and the World Bank’s World Development Indicators. This study encompasses EU-28 countries (including the UK) and countries in the European Economic Area (Norway and Iceland), Switzerland as well as potential EU candidate countries (Albania, North Macedonia, Serbia, and Turkey).

Background of Population Dynamics in Europe

Mortality

Mortality patterns in Europe are often discussed through the lens of divergence between Eastern and Western European countries (Hertzman & Wiens, 1996). It is argued that Europe can roughly be divided into two groups of higher life expectancy in the West and lower in the East (Meslé & Vallin, 2017). In more detail, the demographic literature focuses on mortality tendencies within a selected sample of countries. For instance, Vallin and Meslé (2004) have indicated that not only there exists a difference in life expectancy between Western and Eastern Europe, but have shown that a similar tendency is also present within the region of Central and Eastern Europe where Central European countries (e.g. Poland, Czechia) have higher life expectancy and faster life expectancy improvement in years than Eastern Europe (e.g. Russia, Ukraine). Looking at post-Soviet countries Grigoriev et al. (2010) have found life expectancy differences between countries where more radical socio-economic changes have led to higher life expectancies. Jasilionis et al. (2011) investigated the mortality rates in the Baltic sub-set of post-Soviet countries to find that countries with relatively similar mortality rates have diverged with an improvement in Estonia, stagnation in Latvia, and worsening in Lithuania. A recent study that incorporated countries from Central Europe (Poland), Eastern Europe (Russia), and the Baltic states (Lithuania) has shown that mortality rates not only vary between countries, but within countries as well (Grigoriev et al., 2020).

Population Age Structures

At the European level, population age structures have been found to become more similar over time in Western, Eastern, and Southern Europe. Yet the population age structure differences between subnational regions across Europe have gotten starker (Kashnitsky et al., 2021). These differences can either hold in terms of the centre-periphery, country border territories or migration origin and destination regions. A large proportion of the working age population is found to live in urban areasFootnote 1 while peripheries suffer from out-migration (Kashnitsky & Schöley, 2018). Similar findings have been confirmed by studies that looked at the subnational regions in selected countries. Sabater et al. (2017) have found an increasing segregation over time between older and younger groups across neighbourhoods in England and Wales. The divergence suggested a more pronounced gap between north and south of the countries. It is worth noting that all European populations are experiencing ageing, however differences in the speed of ageing exist across countries and regions (Kashnitsky et al., 2017).

Literature has shown that the main determinant of population ageing in Europe was fertility decline in the first half of the twentieth century, which shifted to mortality improvement and increasing life expectancy starting from the second half of the century (Murphy, 2017). Countries in later stages of demographic transition, which is the case for the developed countries as well as most developing countries, exhibit the same pattern and the importance of mortality improvement overshadows the importance of fertility decline in terms of population ageing (Preston & Stokes, 2012).

Fertility

Overall, fertility patterns and trends in Europe are extremely diverse (Frejka & Sobotka, 2008). Comparative analyses looked at total fertility rates and cohort fertility trends across European countries to establish patterns in childbearing and fertility timing across Europe (Billari & Kohler, 2004; Kohler et al., 2002, 2006). The recent pan-European fertility trends have been generalised as a ‘dual reproductive system’ where Southern European and German speaking countries exhibit lower fertility than that of Western, Northern and some of the Central and Eastern European countries (Lesthaeghe, 2020).

Other work sought to show that countries and regions play a specific role in determining fertility rates and types across Europe (Coale & Watkins, 1986; Klüsener et al., 2013; Lesthaeghe, 1980; Thomson et al., 2014). In Central and Eastern as well as Southern Europe fertility rates have been low since the 1990s (Billari & Kohler, 2004). Western Europe and Nordic countries, however, had been exhibiting close to replacement fertility rates until recently. Nonetheless, many countries with relatively high fertility have been witnessing a decline in period fertility rates since 2010 (Human Fertility Database, 2023). This holds with respect to Nordic countries where declining period fertility is expected to turn into a smaller cohort fertility (Hellstrand et al., 2020). The return of the lowest-low fertility (Billari, 2008) to Southern Europe has also been documented with evidence from subnational regions (Caltabiano et al., 2019; Comolli, 2017).

The fertility differences across European societies are often attributed to a plethora of aspects (see Balbo et al., 2013). Welfare regimes for example have a complex link to fertility through a variety of labour markets and educational systems that translate into fertility differences (Neyer, 2013). For example, in low fertility countries educational, labour market and housing policies have had a positive effect in advancing childbearing (Rindfuss & Brauner-Otto, 2008). Economic uncertainty has been found to not only affect people’s fertility through policies and welfare regimes, but also through the way people perceive the future (Vignoli et al., 2020). Increasing gender equality has been seen as a factor that could converge fertility rates in the future (Esping-Andersen & Billari, 2015). The UN’s projections indicate that the sub-replacement fertility will become a norm everywhere but in Africa, in turn quickening population ageing (Lutz et al., 2008).

International and Internal Migration

The scholarship on migration in Europe is divided into two general strands. First, demographers have shown that free movement in the EU has magnified intra-European migratory flows (Castro-Martín & Cortina, 2015) and changed cohort migration patterns (Bernard, 2017). For instance, nearly two million EU citizens changed their countries of residence in 2008. The largest migration flow has been found to be from the new member states to the EU-15, in real terms migrant flows from Poland to Germany made for the biggest share of intra-EU migration (van Mol & de Valk, 2016). During the Great Recession outmigration from Bulgaria and Romania to EU-15 remained the highest even in comparison to Central and Eastern European countries that joined the EU in 2004 (Kahanec et al., 2016). This migratory pattern was reciprocated by outmigration from Southern European countries to the rest of the EU. For instance, outmigration of Italian and Spanish residents grew during the Great Recession (Anelli & Peri, 2017; Izquierdo et al., 2016).

Second, another strand of demographic literature looks at extra-EU migration. To start with, extra-EU migration is expected to either contribute to population growth or slow down depopulation in Europe (van Nimwegen & van der Erf, 2010). Predominantly immigrants come to the EU from China, India, and Morocco while the nationality of non-EU migrants residing in the EU most often is Turkish, Moroccan or Chinese (van Mol & de Valk, 2016). The non-EU migrants not only add up to the total population, but also have been found to have more children than locals at least in the first generation (Milewski, 2010; Garssen & Nicolaas, 2008). There is also evidence that shows the importance of immigration to sub-regions in Europe. Sub-regions receiving immigrants are predicted to have larger populations in the future (Rees et al., 2012).

Fertility and Migration Policies in Europe

Fertility

The determinants of fertility rate differences translate into difficulties when constructing fertility policies. Challenges to counter the reasons why people choose not to have children are complex. Among other things they consider that marital status, the cost of raising a child, the household income, childcare, the opportunity costs, and compatibility with workforce participation for women are important factors defining the (lack of) success of fertility policies (Chawla et al., 2007; Gauthier & Philipov, 2008; Vobecká et al., 2013). Public policies have not been seen as completely effective because they slow down the demographic trends as opposed to stopping or reversing them (Grant et al., 2004). May (2012) argues that the policy specialist consensus orbits around the negligible effect of transfer-based pro-natalist policies. Money transfers and tax breaks are only perceived as monetary benefits that are not sufficient to raise a child from the viewpoint of families. These measures are also very costly for governments and temporary benefits of such policies do not deliver expected long term results. In addition, current fertility policies may have a potentially detrimental impact on sexual and reproductive health and gender equality (Gietel-Basten et al., 2022). For fertility policies to be effective they need to incorporate long-term measures together with short term tools. More precisely these policies should combine financial incentives with work and family arrangements as it has been done in Scandinavian countries and support sexual and reproductive health as well as human rights as in France, Germany, and Estonia (Chawla et al., 2007; Gietel-Basten et al., 2022; May, 2012; McDonald, 2002).

Migration

In terms of policies, immigration is arguably the most controversial topic in social policy public debates. Unsurprisingly migration remains a constrained policy field despite the need for global mobility because of both population and labour decline (Pritchett, 2006). Immigrants are often seen as a threat to receiving countries’ low-skilled workers and even cultural and religious values, an argument often used by the extreme political right (Le Bras, 1998; Martin & Zürcher, 2008). Many immigration policies in European countries have taken these concerns into account by embracing restrictions and greater selectivity of immigrants in both number and skill (May, 2012). Nonetheless, immigration is a measure that could address and mitigate not only the labour needs but also low fertility and population ageing in industrialised settings such as Europe, as suggested by the replacement migration framework (Craveiro et al., 2019; Marois et al., 2020). The United Nations (UN) has prepared estimations of a yearly number of immigrants necessary to maintain the population size throughout the period of 2000–2050 (UN, 2000). The estimates for the EU have shown that 949,000 immigrants would be necessary every year to keep the EU population constant whilst 1.5 m would be needed annually to maintain constant the 15–64 age group, and 13.5 m would be needed every year to maintain a constant ratio of the 15–64 to the 65 + age groups. These estimates were later revisited, evaluated and virtually confirmed to conclude that the replacement migration remains a valid option to address population decline in Europe (Bijak et al., 2008). For a set of Western European countries recent immigration flows were found to be higher than predicted by the UN and yet not sufficient to slow down the population change (Craveiro et al., 2019). These results were unexpected in Europe and have indicated an urgent need to address population issues (May, 2012). In fact, in recent years, governments have adjusted policies in order to promote the immigration of the particular types of individuals they need and limit other forms of immigration. The UN’s inquiry into governments’ immigration policies has indicated that governments aim at managing migration rather than curb it (United Nations, ). The data has also confirmed that over the time migration policies became more liberal and yet more sophisticated in terms of selectivity (Czaika & Parsons, 2017; de Haas et al., 2019). For example, the EU has opened its borders to selected migrants due to an increasing need for certain types of skills (van Houtum & Pijpers, 2007).

Data, Typology, and Methods

Data

The descriptive part of this paper uses Eurostat data on natural population change and net migration plus statistical adjustment (Eurostat, 2020c) across Europe. This study encompasses EU-28 countries (including the UK) and countries in the European Economic Area (Norway and Iceland), Switzerland as well as EU candidate and potential candidate countries (Montenegro, North Macedonia, Serbia, Turkey and Albania). The data is aggregated at the NUTS-3 level, version of 2016 (Eurostat, 2020d); NUTS refers to Nomenclature of Territorial Units for Statistics. Our data covers the NUTS-3 regions for the period of 2000–2017.

We do not exclude the non-European territories of France, Portugal, and Spain,Footnote 2 but we acknowledge they are geographical outliers (Kashnitsky et al., 2017). In addition, the data does not equally cover all the NUTS-3 regions throughout the period of the study. Some NUTS-3 of EU countries and candidate countries lack data for selected years.Footnote 3 Our final working sample consists of 2239 NUTS-3 regions across Europe and its territories for the 2000–2017 period.

The analytical part of this paper employs the UN World Population Policy database, the UN Population Division and the World Bank’s World Development Indicators data for 1996 together with the Eurostat data for 37 countries in Europe for which we can connect the aforementioned datasets.

The UN World Population Policy database collects data on governmental policies regarding key population issues such as population size growth, fertility, reproductive health and family planning, health and mortality, internal and international migration. The data is collected for the period 1976–2015, and it is revised every two years starting from 2001. The policy variables under the UN World Population Policy database are constructed based on the analysis of four types of sources. First and foremost of these sources is the United Nations Inquiry among Governments on Population and Development, which includes detailed information provided by each country. The UN makes further use of official position documents of governments, information from international organisations and UN agencies as well as news articles and academic publications to determine the position of each country for each indicator (UN, 2020). From this large pool, we focus on fertility and migration policies and make use of two indicators, namely “policy on fertility level” and “policy on immigration”. In particular, while policy on fertility level indicates the availability of governmental policies influencing fertility level, policy on immigration states the existence of governmental policies targeting the level of documented immigration in a given country. Both indicators contain five response categories; “raise”, “maintain”, “lower”, “no intervention” and “no official policy”.

The UN Population Division data is used for the median age in a country in 1996 while the World Bank’s World Development Indicators database is the source for the urban population as a percent of total population, the GDP per capita, and the unemployment rate.

Population Dynamic Score Typology and Methods

We establish a novel typology which permits placing subnational European regions (NUTS-3) and countries within a specific group according to population dynamics (population growth or decline) and drivers (natural population change and migration). A NUTS-3 region or a country is considered to be extremely depopulating if both negative natural population change and negative net migration occur in the unit of analysis. A region or a country is treated as depopulating if it has a negative natural population change even if the net migration in these units is positive, but not big enough to counter the negative natural population change. If a region or a country experiences population growth via positive natural population change and its tendencies in outmigration are smaller than the natural population growth, a unit is considered populating. A NUTS-3 or a country is marked as extremely populating if it exhibits both positive natural population change and positive net migration.

Using the aforementioned definitions, each NUTS-3 region-year and country-year is assigned a number: (1) high depopulation, (2) depopulation, (3) population growth, and (4) high population growth. An arithmetic average is derived for each NUTS-3 region-year and country-year assigned to a NUTS-3 region and a country to indicate average population dynamics of each unit for the timeframe of 2000–2017. The average population dynamic scores of the NUTS-3 regions over the years 2000–2017 are plotted in Fig. 1, using the tmap package on R software. The population dynamics scores of NUTS-3 regions for each year in the above mentioned time frame are also plotted separately and merged as a short animation on R software. A video is included in the supplementary materials for the online edition. The categorisation of countries based on the dominant score of NUTS-3 regions can be observed in Table 1.

Fig. 1
figure 1

Average population dynamics in NUTS-3, 2000–2017. 1 (purple)—high depopulation, 2 (turquoise)—depopulation, 3 (green)—population growth and 4 (yellow)—high population growth. Source: Eurostat (2020a, 2020b, 2020c, 2020d)

Full size image
Table 1 Average population dynamics in European countries, 2000–2017
Full size table

In the analytical part, we use the UN World Population Policy database (policy on fertility level coded as missing if there was no official policy at the time, − 1 = lower; 0 = no intervention; 1 = maintain; 2 = raise and policy on immigration coded as missing if there was no official policy at the time, − 1 = lower; 0 = no intervention; 1 = maintain; 2 = raise), median age in population, available for 1995 by the UN Population Division data and log GDP per capita, percentage of urban population, and unemployment expressed as a share of total labour force for year 1996 by World Development Indicators of World Bank as main determinants. We also include a categorical variable capturing whether a country is part of the EU-15 (more developed economies in EU in Western, Northern and Southern Europe) and European Economic Area (EEA), EU-28 (all EU countries before 2020) and EEA, or not as a control. We estimate the association between fertility and migration policies enacted in 1996 and average population dynamics in the period of 2000–2017 via OLS regression with robust standard errors. The time frame between 1996 and 2017 is chosen purposely in order to be able to capture a connection (if any) between the population policies and their possible influence on population dynamics.

Results

Average Population Dynamics in 2000–2017

Here we refer to groups of NUTS-3 regions that coincide with regional definitions provided by the EU's multilingual and multidisciplinary thesaurus, the EuroVoc (2020).Footnote 4 Four regions are defined: Western Europe, Central and Eastern Europe, Southern Europe, and Northern Europe. These regions encompass both countries inside and outside of the EU.

In Fig. 1 we present descriptive results of population dynamics on NUTS-3 in Europe from 2000 to 2017. Over the period of 17 years population patterns emerge. To start with, we find descriptive evidence that the centre-periphery concept holds whether at national level where urban areas exhibit population growth whilst rural areas depopulate, or at European level where Western European and Scandinavian NUTS-3 regions have been growing whereas Central and Eastern European and Southern European NUTS-3 have been mostly depopulating (Kashnitsky & Schöley, 2018).

NUTS-3 in Central and Eastern Europe as well as the Baltic States exhibit a pronounced tendency of depopulation by both negative natural population change and outmigration (1—purple). Similar tendencies can be observed in Southern European countries. However, they are less strongly pronounced. Depopulation takes place through the channel of negative natural population change, but net migration in these NUTS-3 is positive (2—turquoise). These regions are mostly found in all European regions but Central and Eastern Europe. In Central and Eastern Europe turquoise depopulating regions arguably coincide with urban areas (e.g. Romanian NUTS-3). NUTS-3 in Western Europe and parts of Northern Europe (Scandinavia) indicate opposite dynamics. Regions in green (3) show population growth via positive natural population change and outmigration. The most intensive population growth is marked with yellow (4) representing NUTS-3 regions that grow in size due to both positive natural population change and immigration. Yellow regions are predominantly major urban centres or national capitals. The latter especially stands out in Central and Eastern Europe in which a large majority of NUTS-3 regions have been experiencing depopulation. There also exists an evident West–East, North–South divide where many more NUTS-3 regions exhibit population growth in the West and North in comparison to the East and South of Europe. Turkey, however, stands out as an outlier in the context of Southern Europe. Its western NUTS-3 regions indicate a high population growth which is only characteristic to some parts of Northern Italy. Also, only a limited number of subnational regions in Turkey mostly located in the north-eastern part of the country depopulate strongly.

In Table 1 we summarise the average population dynamics by country using the same four-point typology as for the NUTS-3 regions. Some geographic patterns emerge where most Central and Eastern European countries experienced depopulation while remaining parts of Europe grew in 2000–2017.

Policies and Average Population Dynamics

In Table 2 we show the results of the association estimation between fertility and immigration policies in 1996 and average population dynamics score, defined in the previous section, in the following decades. The regressions are reported with and without controls for covariates traditionally used in the literature (Alvarez-Diaz et al., 2018; Luci-Greulich & Thévenon, 2013), such as mean population age, urbanisation, and GDP per capita. Columns 1 and 2 report a statistical association between fertility policies in 1996 and subsequent average population dynamics in Europe. Columns 3 and 4 show the connection between immigration policies and the average population dynamics in Europe. Columns 1 and 3 indicate the estimation of the relationship without controls while columns 2 and 4 include the covariates, columns 5 and 6 take into account whether a country belongs to the EU-15 or EEA, EU-28 or EEA, or not. We find a negative statistical association between fertility policies in 1996 and average population dynamics in Europe between 2000 and 2017. In practical terms this indicates a presence of a statistical relationship suggesting a connection between fertility policies and higher population decline in Europe. This relationship holds once the controls are included; the association does not change after controlling for a country’s status in terms of its membership in the EU and EEA. The estimation between immigration policies in 1996 and average population dynamics in 2000–2017 does not indicate a statistical association in neither of the models.

Table 2 Association between average population dynamics and population policies
Full size table

Sensitivity Analysis

In order to evaluate the robustness of the results we took the following steps. In a set of estimations reported in Table 3 columns 1 and 2 Turkey was excluded from the sample. As a country with a large number of populating NUTS-3 units and an outlier in terms of its fertility and migration terms when compared to other European countries, it may have led to result overestimation. We find that excluding Turkey from the sample does not change the previously established negative statistical relationship between fertility policies and population decline in Europe.

Table 3 Association between average population dynamics and population policies, sensitivity tests
Full size table

Additionally, connections between policies and population change can be time sensitive and perform divergently in different portions of the study timeframe that was set for 2000–2017. We have re-estimated the model by dividing the analytical period into two parts (2000–2009 and 2010–2017). We find that there is no statistical association between fertility policies and average population dynamics controlling for a set of covariates in 2000–2009, but there is a statistical association between the two focal variables in the period of 2010–2017. The findings for average population dynamics and migration policies remain unchanged in comparison to original estimations (see Table 3 columns 3, 4, 5, and 6).

Furthermore, policies on fertility or immigration may be linked to natural population change and migration measures respectively. To test that, we have introduced interaction terms between policies on fertility and population change as well as policies on immigration and net migration to the model. The estimations reported prove the main analysis to be robust (see Table 3 columns 7 and 8).

Table 4 reports the results of a robustness check relying on spatial regression models, focused on identifying spillovers across countries. We do so only for countries, rather than for NUTS-3 levels, as our focal covariates (fertility and immigration policies) are defined at the national level and uniform across sub-national units, and therefore do not warrant spatial regressions at the NUTS-3 levels.

Table 4 Spatial autoregressive model sensitivity test: GS2SLS estimates
Full size table

We rely on a spatial autoregressive model, with spatial lags on both the dependent variable and the error term. The results in Table 4 corroborate the main analysis results, the coefficient for fertility policy in 1996 is stably negative and statistically significant (at the p < 0.05 threshold) within the spatial autoregressive model. The decomposition of coefficients highlights that there is a direct effect of each country’s policy on its own average population dynamics (direct), while the indirect effect (gauging potential spillovers across countries) is not statistically different from zero, although it is similarly negative (see Table 5). Furthermore, the spatial lag coefficient for average population dynamics is not statistically significant, suggesting that there is little correlation between different countries’ average population dynamics over the study period.

Table 5 Spatial autoregressive model sensitivity test: decomposition of coefficients
Full size table

We further test the robustness of our results by changing the dependent variable to population change rate. The population change rate data for the 37 countries included in the analysis come from the UN Population Division Data Portal. The results for the replication of our model using the average population change rate as the dependent variable are reported in Table 6 in the Appendix. The results corroborate with the results of our initial results (Table 2) and show a negative and statistically significant association between fertility policy status in 1996 and the average population rate between years 2000–2020. Immigration policy status is similarly observed to be insignificant.

In the final step of sensitivity checks, we expand the dataset over the years 2000–2020 instead of relying on the average values, in order to observe the association between the population change rate and policy status in a dynamic approach. The policy data are obtained from the UN World Population Policy database and includes fertility and immigration policy status by country for nine years between 1996 and 2015.Footnote 5 The dataset, thus, covers 37 countries and 21 years. We match the policy year and the observation year for the dependent variable and controls with two-, three- and four-year lags. In other words, we test the association of the fertility and immigration policy status in year P with the population change rate (and other controls) in year P + 2, P + 3 and P + 4 respectively. The results of the OLS regression with country and year fixed-effects are reported in Table 7 in the appendix. The negative and statistically significant relationship between fertility policy status and population change rate is also confirmed by this approach, in line with our initial findings. Additionally, the immigration policy status originally found as positive but insignificant becomes significant when observed over the years. We find no changes in the direction or significance of fertility and immigration policies with respect to the time lags between the year of policy status and observation of population change. However, the relationship is observed to be at its strongest for a three-year gap between fertility policies and population change, whereas it is four-years for the immigration policies and population change. This time difference may be attributed to longer decision-making processes needed for migration (especially international), however, future research may shed more light on this interpretation.

Discussion and Conclusions

In the beginning of the twenty-first century Europe has experienced several waves of population change. The Eastern Enlargement of the EU has established a channel of intra-EU migration that many have used. The Great Recession and the Sovereign Debt Crisis has encouraged European populations to move and work in other EU member states as well (e.g. Anelli & Peri, 2017). Also, the European Migrant Crisis has provided an unprecedented population influx into the continent. However, Europe has also been undergoing a period of ageing and low fertility. Even if age structures across European countries have been converging, age differentials are found at the subnational level (e.g. Kashnitsky et al., 2021). In addition, life expectancy in Europe is distributed unequally. Higher life expectancy is recorded in Western Europe in comparison to Eastern Europe (e.g. Meslé & Vallin, 2017). Fertility differential across Europe stands as an important factor of European population dynamics (e.g. Billari & Kohler, 2004).

This paper is an original effort to marry the measures of fertility and mortality through the concept of the natural population change as well as migration in order to provide a snapshot of how European sub-regional populations have changed since the beginning of the twenty-first century. For that matter we have established a typology that allows us to better understand the average population dynamics in Europe from 2000 to 2017. We have used Eurostat data on natural population change and net migration to demonstrate how the subnational NUTS-3 regions fare in terms of population growth or decline.

Our findings contribute to the literature on differences in fertility, mortality, ageing, and migration. We find evidence that the centre-periphery concept holds (e.g. Kashnitsky & Schöley, 2018). Nationally, urban centres exhibit population growth whilst rural areas have been found to be mostly depopulating. At the level of supranational European regions, more NUTS-3 in Western Europe and parts of Northern Europe (Scandinavia) have experienced population growth than decline. The opposite holds for Central and Eastern European as well as Southern European NUTS-3. Turkey stands out as an exception as a large number of its NUTS-3 regions has been growing on average. This can be partially attributed to the success of family and fertility policies through local welfare provisions (Aksoy & Billari, 2018). We have also established an existing negative statistical association between fertility policies in 1996 and average population dynamics in 2000–2017. There might be several underlying mechanisms that could explain this negative statistical association. First of all, fertility policies and high investment in pro-natalist means do not always ensure population resilience in the face of declining births (Botev, 2015). For instance, while policies regarding public childcare provision and generous parental leave are efficient in increasing birth rates (Bergsvik et al., 2021), other interventions such as cash transfers or tax incentives may fall short in reversing declining fertility trends (Andersen et al., 2018). We also know the influence of fertility policies may vary across different socio-economic groups within a population (Spéder et al., 2020) and relying on macro-level indicators may mask certain local and positive effects. Moreover, the countries which had implemented fertility policies in 1996, might be those that were already experiencing pressing concerns regarding fertility level, hence reversing the declining fertility trends could have been even more challenging. Lastly, the fertility effect of the Great Recession, both in terms of quantity and timing, was sizable among Southern, Central and Eastern European countries (Ayllón, 2019; Matysiak et al., 2021). As a result, any potential positive effects of a fertility policy implemented in 1996 could have been reversed by the substantial drop in fertility in the post-2008 period.

We found no statistical association between average population dynamics and immigration policies contrary to estimations that show slowing down of population decline when immigration is accounted for (e.g. Bermingham, 2001). The latter finding needs to be interpreted cautiously, considering the structural change these countries underwent, directly or indirectly, during the 1990s and 2000s, with the dissolution of the Soviet Union, ethnic conflicts in Balkan countries and the enlargement of the European Union. Observing the change in population dynamics in relation to fertility policies would take time, but immigration policies may have more immediate consequences. In this study, to enable consistency and observe long-term indications, we used 1996 as the year for both fertility level and immigration policies. But considering the insignificant results the latter has shown, future studies may analyse the potential effects of immigration policy at different time intervals, bearing in mind the turbulent nature of these two decades and changing perspectives towards migration across countries. Furthermore, future studies may explore the topic in more recent times, in light of the European Migrant Crisis. Furthermore, our statistical regression analysis focused on the effectiveness of policies at the national level in the 1990s for population outcomes in the 2010s, controlling for socio-demographic and economic covariates at the same geographic level. Relying on our four-fold typology for NUTS-3 units, future research may focus more specifically on how local-level dynamics in the domains of labour market, environment and climate, and welfare affect population change dynamics at the NUTS-3 level, grounding our national results with a more granular scope.

The relationship between migration policies and population dynamics appeared positive but statistically not significant in our analysis. This result may stem from our analytical strategy to use the policy status in 1996 as the benchmark for the next two decades. While this approach works and is corroborated by sensitivity checks for fertility policies, it is not the case for immigration policy. The reason for this could be that the 1996 policies are not sufficient to account for the migration flows within and from outside of Europe that occurred due to the EU expansion in 2004 (Fihel et al., 2006) as well as 2007 and 2013, the Great Recession in 2008 (Lafleur & Stanek, 2017) and the European Migrant Crisis 2014 onwards, in our period of focus. We tackled this possibility using a dynamic approach that includes the change of policies into our statistical analysis, and observed that the relationship between immigration policy status and population dynamics remained positive and became significant. These findings indicate that countries maintaining and increasing their policies to attract migrants, witnessed an increase in population change rates and population dynamics in general, while countries with more restrictive policies, less attractive for migrants experienced the opposite over time.

There are limitations that hamper the outreach of the paper. A number of NUTS-3 regions did not have data for the full period from 2000 to 2017. Therefore, labelling of such NUTS-3 regions should be done cautiously. This too counters the efforts to demonstrate meaningful year specific population dynamics on the whole continent as some countries or subnational regions are not represented due to data limitations. Moreover, the conceptualisation of depopulation or population growth is not unique and it may depend on the indicators used or the geographical context (Newsham & Rowe, 2023). In terms of assessing the population policies, this paper has presented associative results. The connection between population policy variables (fertility and migration) as reported in 1996 and average population dynamics in 2000–2017 needs to be handled with caution. That holds in particular taking into account huge economic and societal changes that were undergoing in countries of Central and Eastern Europe or Western Balkans at that time. These dynamics could not be captured by macro level variables used in this study. Methodological hazards such as reverse causality and endogeneity only allow for an interpretation that incorporates the direction and statistical association between the variables rather than causal links.

Another limitation of our study comes from the nature of official data on population change with respect to the coverage of migrants. We consider our results to be relevant for regular migration patterns, as our reference point is official data that only capture migrants with legal status. Asylum seekers may appear on official data with significant delays due to long bureaucratic processes and irregular migrants are not covered as part of the population data, if at all. Therefore, we are cautious extending this interpretation of our findings to the consequences of the European Migrant Crisis and the irregular migration flows that the region experienced in the second half of 2010s.

In conclusion, Europe has diverse population dynamics that are defined by differences in natural population change and migration. The distinction is particularly visible between urban and rural regions nationally. Supranationally, Western and Northern Europe differ from Central and Eastern Europe and Southern Europe in terms of average population dynamics. Knowing these population dynamics at the granular level of NUTS-3 can contribute to tailor made policies for specific populations of certain demographic characteristics as opposed to policies that can be associated with an even larger population decline.