What Drives Residential Land Expansion and Densification? An Analysis of Growing and Shrinking Regions

Abstract

While the driving factors of urban growth and urban sprawl have repeatedly been studied, the implications for residential densities presumably differ in growing and shrinking regions. Thus far, those differences have received little attention. This paper examined the dynamics of urban growth and shrinkage across EU regions, using residential densities as an explanatory factor to analyse the underlying dynamics. To do so, detailed spatial data on various potentially relevant factors were used in regression methods to establish the relevance of those factors for residential expansion and densification in growing and shrinking EU regions between the years 2000 and 2010. We found that expansion and densification processes are affected by population size, prior residential density, land supply, accessibility, agricultural land rent, physical factors, public regulation, and regional characteristics. The results of this study can confirm that residential expansion is driven differently in declining regions than in regions with population growth. Models explaining residential density changes also yield different results in declining regions.

1. Introduction

As ever more people live in cities around the globe, the sustainability of urban expansion becomes an ever more pressing issue [1]. The concern here is that cities that grow horizontally (i.e., expand) are inefficient. Excessive land consumption has repeatedly been associated with the irretrievable loss of greenfields and ecosystem services [2,3], longer travel, higher car dependence, higher transport energy consumption [4,5], and loss of social sustainability [6,7]. Vertical growth of cities (i.e., densification) is often repeated as the preferred pathway to provide sufficient housing for the world’s population [8], although some argue that evidence in favour of compact urban development is still wanted [9,10].

The benefits from density have been at the centre of urban studies and political goals or policy given that it boosts prosperity and innovation capacity of economic systems through agglomeration economies [11], increases accessibility to goods and services [12], reduces travel needs [13], encourages more energy efficient buildings and transport modes [14], and allows for sharing of scarce urban amenities [15]. A further issue is that there are negative effects associated with high-density development such as low air quality and thermal island effect, which contribute proportionally to the progressing climate change [16,17]. For instance, the vulnerability of high-density urban areas to summer heat waves and its associated health effects are higher than that observed in suburban or rural areas [18]. There is empirical evidence pointing to transport system congestion associated with high densities and also high unit rental prices being attributed to the properties located in high-density population centres [19,20]. There is no consensus in the literature on the optimum density that creates positive returns for all the cities in general [21,22], but in particular, policymakers have increasingly based their policies on promoting high-density development. Here, the assumption is that compact urban forms are more sustainable than dispersed ones [23,24].

Clearly, urban expansion and urban densification are interlinked processes that are often considered substitutes of each other. Any outcome urban density is driven by factor substitution between land and real property [25], typically causing, due to the greater demand for land in central locations, urban land density decreases from the urban centre to the outskirts (see [26,27,28] among others). The resulting negative density gradients are observed in most cities throughout the world, or ‘at least in those in which private land markets determine development’ [29,30]. Urban densities may increase through changes in internal density, typically by redevelopment of brownfield sites or existing properties in the inner city [31]. Such redevelopment is often more costly than new development, so that it often requires political involvement and public funding [32].

Europe is one of the world’s most developed and urbanised regions. In the EU, around 75% of the population lives in cities, suburbs, and towns, and further urbanisation is expected [33]. Despite the decreasing rate of population growth, a significant increase in the number of households have led to increase in land expansion to accommodate the new housing demand in the cities [34]. Progressive urbanisation in Europe has caused significant population changes with simultaneous shrinkage and growth observed in different cities and regions (Figure 1). Europe, in particular, is the continent where shrinkage processes are the most prevalent that affect the deindustrialised regions as well as the peripheral regions located in the Central and Eastern part of Europe [35]. The demographic and environmental factors, deindustrialisation, suburbanisation, and post-soviet transformation are the leading factors resulting in urban land abandonment and land use perforation in the rapidly declining cities [36,37]. Besides shrinking cities, there are others that have experienced re-densification processes in the inner-city areas [38]. Therefore, different density pathways are observed simultaneously in Europe. This was also confirmed by Wolff et al. [39], who asserted that de-densification has occurred in growing as well as shrinking urban areas [40], whereas densification was observed only in growing urban areas in Europe after the year 2000.

The predicted 83.7% rate of urbanisation by 2050 [41] in Europe underlines the significance of sustainability of the urban environment to the forthcoming changes. This means that the city governments will come up with significant challenges related to economic, social, and environmental pillars of sustainability due to the issues caused by urban growth and shrinkage processes accompanied by de-densification. As in many regions of Europe, cities in developed countries are likely to experience more rapid decline in residential densities through more sprawling patterns, thus putting more pressure on city governments to pursue compact development strategies [42]. Understanding the factors effecting population densities and its variation across growing and shrinking regions is therefore essential for urban and regional planning. Although the determinants of urban expansion are well known, there is no consensus on the key drivers of residential densification in the literature, as these remain largely unknown for many cities and regions, particularly the shrinking regions in Europe. Identifying the driving forces of urban densification processes will contribute to our understanding of their impacts on the environment and social well-being, as well as for supporting related decision making.

While expansion and densification processes are modelled extensively, in general, they are modelled separately [43,44,45], which is a major shortcoming of the current studies. Among the few studies that examine expansion and densification processes simultaneously, we can refer to Luck and Wu [46], Haase and Nuissl [47], Broitman and Koomen [48], and Mustafa et al. [49]. These studies used statistical methods and landscape metrics to quantify the spatial pattern of urbanisation at the regional or country levels; however, none of them paid attention to the urban growth and shrinkage processes observed in their case study areas. Other studies examine only densification [39,50] or specifically the redevelopment of brownfield sites to promote urban intensification [51,52]. Many previous studies in Europe have focused on the examination of factors influencing urban land expansion or densification at the country [48,53], inter-regional [49,54], or city [55,56] levels. Investigation of driving forces of both urban expansion and densification via sampling at the regional scale of the growth and shrinkage processes, which are necessary for EU policymaking at the regional level, is still lacking. The studies at the pan-European level are mainly centred on urban land use patterns and densities [39,57,58], land use change dynamics [59,60,61], and conceptualisation of land consumption in the EU context [33].

Previous studies have focused on the relationships between urban land densities and their determinant factors through the application of global regression models. For instance, Seto et al. [44] highlighted the existence of infrastructure, economic, political, and demographic factors to explain the conversion of agricultural land to urban uses using a global analysis approach. A similar approach was followed by Plieninger et al. [60], which examined political/institutional, cultural, natural, and spatial drivers of landscape change in Europe. Ustaoglu and Williams [62] is another Europe-wide study that conducted regression analysis to explain urban expansion on agricultural land on the basis of the socio-economic, locational, geological, climatic, and policy-related factors. There are also local studies researching the land-use change process and drivers of land-use change at the regional or more local levels. Among these, Oueslati et al. [63] investigated the causing factors of urban sprawl in selected European regions and cities, finding that transport infrastructure and socio-economic and environmental factors are central in explaining sprawled urban developments. Li et al. [64], Achmad et al. [65], Munthali et al. [66], and Alijani et al. [67] assessed the impact of socio-economic, biophysical, transportation, and neighbourhood factors on urban land expansion in relation to different geographical locations using a similar modelling approach.

In contrast to urban expansion studies, the literature on the drivers of densification is scarce. There are studies at the country or local/regional levels. A study by Broitman and Koomen [48] is an example of the former group exploring the determinants of residential land expansion and densification in the Netherlands. Similarly, Wang et al. [68] and Koning et al. [69] used accessibility, location, and geo-physical characteristics to model urban densification dynamics in Wisconsin (USA) and Bergen (Norway), respectively. Yunda and Sletto [70] and Scheba et al. [71] assessed only the role of socio-economic factors and the informal, public, and private sectors in densification processes in the cities of Cape Town and Bogota (Colombia), respectively. Similar to the urban expansion literature, the latter literature on densification applied global regression approaches to model urban densification dynamics. A final example is that of Mustafa et al. [49], who conducted densification modelling at the regional scale through estimating the impacts of physical, accessibility, socio-economic, political, and neighbourhood factors on densification in Wallonia (Belgium). While the driving factors of urban expansion and densification vary across space and time, we can question whether there are variations of the driving factors across growing and shrinking regions in Europe. Much of the urban growth literature focuses on the description and prediction of urban patterns in the case of demographic and economic growth. To our knowledge, there is no study that has researched the factors influencing residential expansion and densification across the growing and shrinking regions at the pan-European level.

This paper aims at researching the determinants of urban densification and expansion with respect to population growth and decline, through focusing on residential land use in Europe. In this study, we describe and analyse various spatiotemporal characteristics of residential development in EU regions between the years 2000 and 2010. The areal units used in the analysis consist of 1300 regions at the so-called NUTS3 level in the EU. It is well known that spatial aggregation of data will affect the results, and data capturing the individual choices driving urban development processes ought to be preferred [72,73]. Unfortunately, relevant socio-economic and policy-related data were not available at a finer spatial resolution than the selected regional level, and thus we had to revert to this aggregate level. It must be noted that spatial aggregation is typically less influential on model outcomes than, for instance, model specification [74]. To illustrate the patterns of change of residential development in the studied time period, the driving factors behind processes of urban densification and expansion will be researched in growing and declining regions separately. This will allow us to quantify and evaluate the relative contribution of factors contributing to density changes and land expansion processes separately.

2. Data and Methods

2.1. Dependent Variable

In order to quantify the residential expansion that took place between the years 2000 and 2010, residential surfaces were computed using 300 m GHSL project grids [75], which indicates percentage built up in $\%B_{g,t}.$ Those data are the results of analysis of global, fine-scale satellite image data; census data; and crowd sources or volunteering geographic information sources. One very attractive aspect of these data is that they were gathered with a consistent methodology for a longer range of years, which makes these data a reliable source for analysing spatial and temporal dynamics. Additional information was obtained from the LUISA (Land-Use Integrated Sustainability Assessment) base map product [76], a considerably refined version of the CORINE land cover grids [77] that is available for 2012. For our purposes, the LUISA base map was simplified into a Boolean grid, indicating that a grid cell is used for residential purposes, $R_{g,2012}=\left\{\begin{array}{c}1 if land use is residential\\ 0 if land use is other\end{array}\right.$. Total residential land cover is subsequently summed for every NUTS3 region i through $R{B}_{i,t }=\sum _{g \in i}^n\%B_{g,t}{R}_{g,2012}$, so that hectares of built-up land coverage are summed if they are located in grid cells that are identified as residential land cover. Expansion of land is subsequently computed as $\Delta R{B}_i=R{B}_{i,2010}-R{B}_{i,2000}$. Thus, land expansion is only considered relevant if the use was likely to be residential two years after the observed changes. Underneath that decision is the assumption that land use changes slowly, so that grid cells identified as residential in 2012 most likely did not have another use in prior years that would involve constructed buildings. We therefore assume that any change in land covered by buildings in a residential grid cell must have been due to expansion of land for residential purposes. Alternatives, for example using land-use changes observed in the CLC project, most likely are less accurate, in particular because necessary detail lacks in other land-use sources. For example, the 25 ha minimum mapping unit requirement in CLC data would likely neglect a substantial amount of small-scale residential expansion.

To measure residential densities, regional population counts in $P_{i,t}$ were obtained from Eurostat. Residential population densities were observed as $R{D}_{i,t}=P_{i,t}/R{B}_{i,t}$, and the observed changes in residential density $\Delta R{D}_{i,t}=R{D}_{i,2010}/R{D}_{i,2000}.$ It must be clear that changes in density can then occur because of changes in population or because of changes in residential land. In declining regions, land use densities may dwindle because of urban expansion, or simply because of population decline.

2.2. Empirical Strategy

We focused on explaining variance in three sets, namely, the entire group of observations $I$, the set of growing regions G for which $P_{i,2010}\ge P_{i,2000}$, and the set of declining regions D for which $P_{i,2010}<P_{i,2000}$. The approach taken consisted of two stages. In a first stage, extensive principal component analyses were conducted to establish variables relevant for the explained variables $\Delta R{B}_i$ and $\Delta R{D}_i$. All the variables in Table 1 were used here. We found that factors related to economic output, accessibility, land availability, climate, governance, and geographical position all were relevant determinants. In the second stage, we applied a number of multivariate OLS regressions to quantify the determinants of regional residential densification and expansion in Europe. Equations were fitted to explain the variance in $\Delta R{B}_i$ and in $\Delta R{D}_i$ on subsets $i\in I, i \in G, i \in D$ using straightforward ordinary least square (OLS) techniques, with the form

$$Y_i=\alpha 0+{\beta }_k{X}_{k,i}+{\epsilon }_i$$

(1)

Those regressions were performed using a number of selected independent variables in X. Here, α is the constant term and β represents the estimated coefficients of the independent variables. The selection of those variables was based on the correlations identified by the principal component analyses. The correlations and principal component results are shown in Figure A3 and

Table A1. The eigenvalues of the components summarised for all the data covering the EU countries.

Component	Eigenvalue	Difference	Proportion	Cumulative
Comp1	4.10848	1.17225	0.1956	0.1956
Comp2	2.93623	0.364062	0.1398	0.3355
Comp3	2.57217	0.746115	0.1225	0.4579
Comp4	1.82605	0.231704	0.087	0.5449
Comp5	1.59435	0.503992	0.0759	0.6208
Comp6	1.09036	0.0918696	0.0519	0.6727
Comp7	0.998486	0.0363305	0.0475	0.7203
Comp8	0.962155	0.0698244	0.0458	0.7661
Comp9	0.892331	0.117723	0.0425	0.8086
Comp10	0.774608	0.15234	0.0369	0.8455
Comp11	0.622267	0.0966038	0.0296	0.8751
Comp12	0.525664	0.026642	0.025	0.9001
Comp13	0.499022	0.129004	0.0238	0.9239
Comp14	0.370017	0.038467	0.0176	0.9415
Comp15	0.33155	0.0927882	0.0158	0.9573
Comp16	0.238762	0.0157336	0.0114	0.9687
Comp17	0.223029	0.0727235	0.0106	0.9793
Comp18	0.150305	0.0364094	0.0072	0.9865
Comp19	0.113896	0.0137852	0.0054	0.9919
Comp20	0.100111	0.0299409	0.0048	0.9967
Comp21	0.0701697		0.0033	1

Note: Eigenvalues that are greater than 1 are considered significant. The eigenvalues for the data of growing and declining regions were computed and can be obtained from the authors on request.

and

Table A2. Principal components (eigenvectors) for all the data covering the EU countries.

Variable	Comp 1	Comp 2	Comp 3	Comp 4	Comp 5	Comp 6
IND_COM_2000	0.1983	0.3416	0.2542	−0.0764	0.0665	−0.0551
URBAN_RECREAT_2000	0.292	0.2152	0.0586	0.2393	0.0589	−0.2565
DENSITY_2000	−0.0591	0.0717	−0.0003	0.1914	−0.0462	−0.4357
ACCESS_2000	0.3154	−0.2631	−0.0683	−0.1014	0.1159	0.2132
GDP_CAP2000	0.3366	0.0939	0.3064	0.0571	0.0525	−0.058
POP_2000	0.2761	0.2629	0.352	−0.0781	0.0755	−0.0721
AGRI_LAND_RENT_2000	0.0554	−0.0632	−0.0061	−0.0002	−0.1351	0.0144
NATURA2000	−0.1138	0.3223	0.0636	0.2903	−0.035	0.5459
AVAIL_LAND_2000	−0.0639	0.4006	−0.0303	0.3896	−0.0595	0.3388
PUBLIC_REG	0.1912	−0.3681	0.0004	0.3464	−0.2516	0.0338
METROPOL	0.3196	−0.0167	0.0367	−0.1439	0.0945	0.1364
URBAN	0.3372	−0.0852	0.097	−0.058	0.0841	0.0614
RURAL	−0.3013	0.0723	−0.0483	0.0983	−0.0528	−0.1483
EASTERN_EU	−0.1224	0.3578	−0.1114	−0.3632	0.2654	−0.1104
NORTHERN_EU	0.1852	0.0341	−0.1273	0.4898	0.012	−0.253
SOUTHERN_EU	−0.1544	−0.0389	0.4577	0.0153	−0.3669	0.0218
ASPECT	0.0107	−0.1766	0.0881	0.0052	0.2782	0.3777
PLAIN_RATIO	0.2564	0.1066	−0.3856	−0.0883	−0.297	0.0811
PRECIPITATION	−0.0481	−0.2695	0.1506	0.2474	0.4036	−0.0346
TEMPERATURE	−0.0177	−0.0721	0.3538	−0.2058	−0.502	0.0201
SLOPE	−0.2792	−0.13	0.386	0.0862	0.2778	−0.0372

Note: The principal components for the growing and declining regions were also computed and can be obtained from the authors on request.

in the Appendix A and Appendix B. The descriptive statistics of all included variables are presented in Table 1.

Table 1. Variables explaining density increase in residential expansion and densification areas (NUTS3 data).

Variable (Abbreviation)	Type	Range	Mean
Dependent variables
Residential expansion (RES_EXP)	Continuous	0–3471	212.72
Residential densification (RES_DEN)	Continuous	0.56–1.33	0.973
Socio-economic variables
Gross domestic product (thousand Euro) per capita in 2000 (GDP_CAP2000)	Continuous	0.045–7.993	1.00
Gross value added in agriculture (in Euro)/total agricultural area (Ha) (AGRI_LAND_RENT)	Continuous	0–121,346.5	783.1
Population in 2000 (POP_2000)	Continuous	19,214–5,345,542	370,757
Spatial planning policies
Total area of available land for residential development (Ha) in 2000 (AVAIL_LAND_2000)	Continuous	957–6,966,745	240,928.4
Total area of NATURA2000 sites (NATURA2000)	Continuous	0–3,524,306	59,593.7
Public regulation and policy strictness of a country (PUBLIC_REG)	Continuous	1–6	4.36
Accessibility
Potential accessibility indicator, 2000 (ACCESS_2000)	Continuous	144.5–349,309	48,942.7
Geological and climatic characteristics
Ratio of land with a slope less than 8 degrees in a region (PLAIN_RATIO)	Continuous	14.21–100	109.1
Average slope of a region (%) (SLOPE)	Continuous	1–4.16	1.77
Average annual temperature of a region (°C) (TEMPERATURE)	Continuous	−0.80–17.9	9.04
Average annual precipitation of a region (mm/year) (PRECIPITATION)	Continuous	97–1687.9	745.3
Location characteristics
Residential density in 2000 (DENSITY_2000)	Continuous	41–81,092	741.13
Urban recreation areas (green urban areas, sports and leisure facilities, Ha) (URBAN_RECREAT_2000)	Continuous	0–13,786	824.48
Business, retail, and commercial areas (Ha) (IND_COM_2000)	Continuous	0–14,736	1502.09
Located in Eastern Europe (EASTERN_EU)	Dummy	0/1	0.21
Located in Southern Europe (SOUTHERN_EU)	Dummy	0/1	0.22
Located in Northern Europe (NORTHERN_EU)	Dummy	0/1	0.22
Base category: WESTERN_EU
Located in a predominantly urban region (URBAN)	Dummy	0/1	0.24
Located in a predominantly rural region (RURAL)	Dummy	0/1	0.38
Base category: predominantly intermediate region
Located in a NUTS3 coinciding with a metropolitan region (METROPOL)	Dummy	0/1	0.39

Table 1. Variables explaining density increase in residential expansion and densification areas (NUTS3 data).

Variable (Abbreviation)	Type	Range	Mean
Dependent variables
Residential expansion (RES_EXP)	Continuous	0–3471	212.72
Residential densification (RES_DEN)	Continuous	0.56–1.33	0.973
Socio-economic variables
Gross domestic product (thousand Euro) per capita in 2000 (GDP_CAP2000)	Continuous	0.045–7.993	1.00
Gross value added in agriculture (in Euro)/total agricultural area (Ha) (AGRI_LAND_RENT)	Continuous	0–121,346.5	783.1
Population in 2000 (POP_2000)	Continuous	19,214–5,345,542	370,757
Spatial planning policies
Total area of available land for residential development (Ha) in 2000 (AVAIL_LAND_2000)	Continuous	957–6,966,745	240,928.4
Total area of NATURA2000 sites (NATURA2000)	Continuous	0–3,524,306	59,593.7
Public regulation and policy strictness of a country (PUBLIC_REG)	Continuous	1–6	4.36
Accessibility
Potential accessibility indicator, 2000 (ACCESS_2000)	Continuous	144.5–349,309	48,942.7
Geological and climatic characteristics
Ratio of land with a slope less than 8 degrees in a region (PLAIN_RATIO)	Continuous	14.21–100	109.1
Average slope of a region (%) (SLOPE)	Continuous	1–4.16	1.77
Average annual temperature of a region (°C) (TEMPERATURE)	Continuous	−0.80–17.9	9.04
Average annual precipitation of a region (mm/year) (PRECIPITATION)	Continuous	97–1687.9	745.3
Location characteristics
Residential density in 2000 (DENSITY_2000)	Continuous	41–81,092	741.13
Urban recreation areas (green urban areas, sports and leisure facilities, Ha) (URBAN_RECREAT_2000)	Continuous	0–13,786	824.48
Business, retail, and commercial areas (Ha) (IND_COM_2000)	Continuous	0–14,736	1502.09
Located in Eastern Europe (EASTERN_EU)	Dummy	0/1	0.21
Located in Southern Europe (SOUTHERN_EU)	Dummy	0/1	0.22
Located in Northern Europe (NORTHERN_EU)	Dummy	0/1	0.22
Base category: WESTERN_EU
Located in a predominantly urban region (URBAN)	Dummy	0/1	0.24
Located in a predominantly rural region (RURAL)	Dummy	0/1	0.38
Base category: predominantly intermediate region
Located in a NUTS3 coinciding with a metropolitan region (METROPOL)	Dummy	0/1	0.39

2.3. Explanatory Variables

A number of secondary variables were selected in order to determine their influence on residential expansion and densification processes. Potential accessibility has repeatedly been noted as a key factor in explaining urban patterns [73,78,79,80]. According to the classical representation of urban pattern, the central locations are the most accessible and therefore are more densely developed compared to peripheral locations of the least accessibility [81,82]. In today’s modern cities, areas of highest accessibility may or may not be centrally located, given that urban activities have been decentralised in metropolitan areas [83,84]. A high accessibility level in the periphery refers to reduction in transportation costs, which makes far-reaching land more accessible, resulting in low-density settlements [85]. To take these into account, EU-wide regional average potential accessibility values were obtained from Jacobs-Crisioni and Koomen [80]. These values were computed for all municipalities in Europe on the basis of

$$A_{i,t}={\displaystyle {\sum }_{i\ne j}^m{P}_{j,t}{M}_{i,jt}^{\gamma }}$$

(2)

where A is the accessibility at origin i; P is the population in destination j in decade t; M represents the distance-decayed travel times in minutes, obtained from historical road networks [86]; and γ is assumed to be −1 as the real value of distance–decay is unknown for the study area. Both the origins in i and the destinations in j consist of municipalities. Accessibility levels were averaged to the NUTS3 level to be included in the current study. Accessibility values for the years 2000 and 2010 were assumed to be equal to the values originally obtained from Jacobs-Crisioni and Koomen [80] for the years 2001 and 2011.

Population, economic income, and agricultural land rent are expected to affect residential expansion and densification processes as well. Economic activities may lead to increased density of populations in urban areas, which increases the pressure on demand for housing and housing prices at the central locations. Therefore, it would be cheaper to develop land in the outskirts of the city where population density is lower [49]. Similarly, highly populated regions are expected to be more densely developed compared to less populated counterparts; the more prosperous regions are expected to have (re)-densification of the inner-city area through re-development of existing urban uses and brownfield sites; and the lower the agricultural land rent, the lower the development densities are expected in the urban–rural fringe. According to the literature, in the regions with lower agricultural land rent, the cheaper it will be to convert agricultural land to residential uses [87,88]. However, through application of land use plans, land use zoning, and farmland protection programmes, many European countries aim to support farming and limiting agricultural land consumption in the urban fringe. These polices are successful in some regions, whereas farmlands that face higher pressure may be less protected, for instance in the Mediterranean Region including the south of France, central Italy, and Portugal [89,90,91]. To take into account socio-economic characteristics of the regions in the current study, we considered three variables, i.e., population, gross domestic product (GDP) per capita, and agricultural land rent (Table 1). Following Jiang et al. [92], the value of gross agricultural output divided by total agricultural land area in a NUTS3 region was used as a proxy for agricultural land rent (see [62]). The GDP, agricultural output, and population data of the regions were collected from Eurostat’s online database (http://ec.europa.eu/eurostat/data/database, accessed on 23 August 2022).

Regional area characteristics that we consider relevant for residential development are the existence of a metropolitan area in a region, and whether a region is predominantly considered urban, rural, or intermediate. The existence of a metropolitan area in a region influences the demand for residential land use, as metropolitan regions typically have much higher population densities than non-metropolitan areas, and thus residential areas may be expected to be denser. On the contrary, residential expansion may be more frequent in regions specified as rural or intermediate, where there is presumably more land ready for development at a cheaper price. A dummy variable representing the metropolitan regions is therefore included in the demand equation. To indicate whether a region is urban, rural, or intermediate, predominant degrees of urbanisation have been obtained from Dijkstra and Poelman [93] and used as dummies here. The existing metropolitan regions linked to a NUTS3 region or combination of NUTS3 regions representing ‘agglomerations of at least 250,000 inhabitants’ were obtained from the online statistics for Eurostat Metropolitan Regions [94].

The existence of NATURA2000 sites [95] in which natural functions are conserved is considered a development constraint. Total areas of NATURA2000 sites were therefore included as an explanatory variable. The NATURA2000 is a network for natural protected zones in the EU countries that has been established under 1979 EU Birds Directive and 1992 EU Habitats Directive. The NATURA2000 consists of Special Protection Areas and Sites of Community Importance whose objectives are to assure long-term protection of Europe’s most valuable species and habitats [95]. Total area of suitable land available for residential development is a restriction on the supply of land. All of a NUTS3 region’s area is considered suitable for residential development except water bodies, mining sites, and the land covered by transportation network. The suitable land is calculated by subtracting unsuitable areas from total land area of a region.

The public regulation and planning institutions of the analysed countries have significant influence on land-use change decisions at local, regional, and country levels. Sustainability of land use not only depends on socio-economic processes that shape the spatial structure but also on effectiveness of the spatial governance instruments that regulate the processes [96]. To promote sustainable land use development, planners and other stakeholders seek to explore possible alternatives to land take such as brownfield development, mixed use development, urban reconversion, and densification practices [97]. In the EU, types of spatial planning policy (i.e., regional or national level) range from non-interventionist to controlled systems. Tosics et al. [98] quantified the potential influence of public regulation over land-use change in different EU countries, which was used in the current study as a proxy of regulation (see also [96]). These values indicate the relative strength of public regulation in each EU country. In the explanation of Tosic et al. [98], the Northern EU countries (e.g., Denmark, the UK, the Netherlands, Sweden) are governed with higher levels of potential control. By contrast, the Southern EU countries (e.g., Italy, Spain, Portugal, Greece) were shown to have higher potential values of having more fragmented local government systems with a stronger control mechanism, both at regional and national levels. The potential strength of public regulation over land use change in different countries was quantified as follows [98] (the bigger the number the higher the levels of control): 6: Denmark, the Netherlands, Portugal, the UK; 5: Belgium, Cyprus, France, Germany, Greece, Ireland, Lithuania; 4: Italy, Spain, Sweden; 3: Austria, Bulgaria, Finland; 2: Estonia, Latvia, Luxembourg, Malta, Poland, Slovenia; 1: the Czech Republic, Hungary, Romania, Slovakia. This index of potential strengths of control of the government in different EU countries was included as an independent variable in the regressions.

Biophysical and climatic conditions are considered relevant for urban development [62,99]. Given that land use activity contributes to climate change, climate change has in turn direct and indirect impacts on land use [100]. Changing temperatures or rainfall patterns can change crop suitability, forest harvesting patterns, and land use/cover (e.g., changes from agriculture to abandoned land or to urban land). Therefore, climatic variables should be included in the studies of land use change modelling. Rehdanz [101] focused on the amenity value of climate on households in Great Britain, arguing that people’s preferred climate would influence their residential location choice. For instance, the locations with mild climate (e.g., fewer rainy days and more hours of sunshine) are generally more preferable, and therefore households wishing to live in these locations may pay a premium for it [101,102]. According to Maddison [103], British people value higher temperatures as an amenity while higher precipitation levels are considered as a disamenity. There are other studies with different findings performed for Italy and Germany where high summer temperatures are a significant disamenity for these countries [101,104]. To capture climatic factors, two variables are accounted for, namely, mean annual precipitation in (mm/year) and mean annual temperature in °C. Both variables were obtained from the European Food Safety Authority [105]. To represent biophysical characteristics, two variables were included: slope and plain ratio of the regional landscape, which were obtained from NASA [106]. Average slope in a region is used here, alongside the plain ratio, which is defined as the percentage area of a region that has slopes at less than 8 degrees. The latter cutoff value was chosen because construction costs on steep lands are costlier, and thus generally only slopes below 8 degrees are considered to be suitable for residential property developments [107].

Cheshire [108] showed the different urbanisation phases observed in the regions in Europe where Northern and Western Europe and Southern Europe showed different trajectories and a clear shift in time according to urbanisation phases. The Eastern European countries experienced a shift during the 1990s and 2000s through depopulation and economic centralisation after the collapse of the socialist era. Compared to Western and Northern Europe, cities in Eastern and Southern Europe have experienced more recent suburbanisation, possibly due to rapid demographic changes observed recently [109]. On the basis of the literature identifying the characteristics of European regions at the macro level regarding socio-economic trends, housing characteristics, and planning and institutional structures [108,110], we grouped Europe into four macro regions: (1) Eastern Europe, including Bulgaria, Croatia, the Czech Republic, Hungary, Poland, Slovenia, Slovakia, Romania, Estonia, Lithuania, and Latvia; (2) Western Europe, including Austria, Belgium, Germany, France, Luxembourg, and the Netherlands; (3) Southern Europe, including Spain, Portugal, Italy, Greece, Malta, and Cyprus; and (4) Northern Europe, including Denmark, Sweden, Finland, Ireland, and the UK. Therefore, three dummy variables were employed for the regions located in Eastern Europe, in Southern Europe, and in Northern Europe, with Western Europe being the base category. We included all the variables explained above in the regression models describing residential expansion and densification. However, in the models explaining densification, two additional variables were used, namely, the areas of urban recreation sites (green urban areas, sports and leisure facilities) and the areas of business, retail, and commercial sites in a NUTS3 region. While these variables presumably have a negligible influence on urban expansion, they can be considered potential drivers of residential densification, as they possibly represent the attractivity of existing urban locations and may thus encourage brownfield development. The inclusion of such variables for densification models has been suggested repeatedly [48,111].

3. Results

3.1. Urban Expansion

Table 2. Distribution of residential areas developed between 2000 and 2010 across countries in the EU.

Country	Newly Developed Residential Area		Total Residential Area 2000	Share of New Developed Area in Total Area
	Ha	%	Ha	%
AT	6179	2.23	131,753	4.69
BE	13,731	4.97	227,552	6.03
BG	2283	0.83	104,590	2.18
CY	874	0.32	16,239	5.38
CZ	5623	2.03	144,404	3.89
DE	41,288	14.93	1,270,623	3.25
DK	3797	1.37	103,870	3.66
EE	236	0.09	7650	3.08
EL	1984	0.72	90,146	2.20
ES	22,018	7.96	299,445	7.35
FI	3716	1.34	33,245	11.18
FR	48,307	17.47	973,073	4.96
HR	1460	0.53	49,982	2.92
HU	7051	2.55	180,994	3.90
IE	4016	1.45	46,508	8.64
IT	23,921	8.65	604,175	3.96
LT	440	0.16	18,531	2.37
LU	416	0.15	8749	4.75
LV	261	0.09	8963	2.91
MT	297	0.11	4667	6.35
NL	19,214	6.95	222,785	8.62
PL	21,056	7.61	381,828	5.51
PT	7893	2.85	120,505	6.55
RO	15,924	5.76	278,373	5.72
SE	3664	1.33	97,724	3.75
SI	646	0.23	19,915	3.24
SK	3614	1.31	95,926	3.77
UK	16,614	6.01	850,501	1.95
Total	276,522	100	6,392,716	4.33

presents country totals of the residential land areas existing in 2000, as well as newly developed residential land in the observed period. Percentage changes in residential land between 2000 and 2010 in the EU territory and in declining EU regions specifically are shown in Figure 2 and Figure 3, respectively. Finally, Figure A1 in the Appendix A shows the spatial distribution of change in regional residential land.

Table 2 shows that more than 275,000 hectares of new residential land were developed in the EU. The highest share of new development in the total residential area in a country was observed in Finland, followed by Ireland and the Netherlands, then Spain, Portugal, and Belgium (Figure 2). The UK and Bulgaria showed the lowest percentages of new residential development (Figure 2). Residential expansion occurred even in regions that declined considerably in population, for example, in Austria, Germany, Portugal, Romania, and Finland, among others. (See Figure 3). This may point to sprawled development, even under conditions of shrinkage [34,112]. Wolff et al. [39] claim that this type of urban development is dominant mainly in post-socialist countries and in Germany. From these results, it follows that processes of sprawl are much more widespread in Europe, confirming other studies that propose that in Europe, processes of urban sprawl are dominant and persistent, leading to low-density developments in many European countries if left unchecked [39,57,88,113].

Urban expansion and sprawl are considered to be two different terms explaining the physical development process in developing and developed countries/regions, respectively. Urban sprawl is generally characterised as a low-density, scattered, and leapfrog development pattern. By contrast, urban land expansion mainly observed in developing regions does not possess the characteristics of leapfrog or low density [114]. Therefore, the process of urban land development in some cities and regions cannot be specified as sprawl, which was also the case in our study. There was no clear demarcation between sprawled and compact residential development in our analysis. It would therefore add value to the current research if the regions that experienced a sprawl pattern of development or high-density expansion between the years 2000 and 2010 are identified. This would help in developing policy measures in limiting the effects of sprawl in the subject regions.

Table 3. Results of the regression analysis explaining residential expansion density in EU countries.

Dependent Variable: Residential Expansion (ha)
	All Regions		Declining Regions		Growing Regions
Variables	Standardised Coefficient	SE	Standardised Coefficient	SE	Standardised Coefficient	SE
Constant	1.779 **	0.6	−0.292 *	0.29	3.639 **	0.91
GDP_CAP2000	-	-	−0.148 *	0.08	-	-
AGRI_LAND_RENT_2000	−0.009 **	0.02	-	-	-	-
POP_2000	0.741 **	0.03	0.594 **	0.05	0.681 **	0.04
NATURA2000	-	-	-	-	-	-
AVAIL_LAND_2000	0.083 **	0.02	-	-	0.202 **	0.04
ACCESS_2000	0.341 **	0.05	0.125 **	0.05	0.387 **	0.06
DENSITY_2000	−0.031 **	0.02	0.047 **	0.02	-	-
METROPOL	-	-	-	-	-	-
URBAN	-	-	0.344 **	0.12	-	-
RURAL	-	-	-	-	−0.168 *	0.11
EASTERN_EU	−0.684 **	0.15	−0.409 **	0.10	-	-
SOUTHERN_EU	−0.965 **	0.14	-	-	−0.981 **	0.18
NORTHERN_EU	−0.601 **	0.09	-	-	−0.762 **	0.11
PUBLIC_REG	−0.111 **	0.04	0.154 **	0.03
PLAIN_RATIO	1.224 **	0.42	0.711 **	0.17	−2.961 **	0.66
SLOPE	−1.088 **	0.19	-	-	−1.999 **	0.31
PRECIPITATION	0.870 **	0.14	0.417 **	0.12	0.989 **	0.19
TEMPERATURE	1.078 **	0.16	0.165 *	0.09	1.318 **	0.23
Number of observations	1300		494		806
R-squared	0.45		0.42		0.46
F (probability)	79.19 (0.00)		35.19 (0.00)		65.76 (0.00)

Note: * Significant at 10%; ** significant at 5%. Standardised coefficients represent the regression outcomes of the variables that were standardised by dividing the values of each variable to the mean of the corresponding variable.

presents the results of regression analyses aimed at explaining the cross-sectional distribution of residential land expansion for all EU regions, as well as for declining and growing regions separately. In all regressions, land expansion is positively and significantly correlated with population size and accessibility. This highlights the importance of these factors in steering residential land development both in growing and declining regions. Population has been repeatedly identified as one of the most significant drivers of land expansion in Europe [57,113,115]. However, this relationship varies greatly across and within countries. From our analysis, in many of the European countries, residential land grows faster than population (Romania, Slovakia, the Czech Republic, the Baltic countries, Hungary, Croatia, Bulgaria, Finland); in others, the relationship varies across regions within a country. Siedentop and Fina [116] pointed out that population growth is a poor indicator of land take at the country level. Because spatial unit of our analysis is NUTS3 regions, we believe that the population variable clearly explains its geographically differentiated effects in Europe.

Accessibility is the other indicator that has been widely utilised in the urban land development studies, pointing out that transport improvements and increased accessibility has led to outward expansion of urban areas mainly in the form of low density and sprawled developments. Numerous studies found that the probability of new urban development is higher in the locations close to highways and road network [117], rail-based networks [5], and transit corridors [118]. There is consensus in the literature that private-car-oriented transport policies and investments attract urban growth to outward locations and reduce overall density while integrated land use and public transport policies can increase urban density as the latter supports more compact urban developments. Through taxes, subsides, and regulations, public policy has played a major role in promoting public transport and fostering more compact patterns of development. This should be considered in the policymaking context by governments and policymakers.

Plain ratio has an unexpected negative coefficient in growing regions, whereas it has a positive coefficient for all EU and declining regions. The positive relationship between plain land and the probability of urban expansion was confirmed by various studies [119,120]. The negative relationship may occur occasionally; for instance, in regions where there is high risk of flood, or flat areas less suitable to land development. Population dispersion relates to slope as well: built-up areas in high slopes are more dispersed than in flatter areas, given that suitable land for urban development is dispersed in locations of high slopes [121]. These can possibly explain the negative coefficient of plain ratio estimated for growing regions. The presence of NATURA2000 areas is not found to be significant in any of the regressions. One explanation is that such areas are more likely appointed in regions that have ample space. Available land does not yield significant estimators in declining regions, presumably because land availability is not a bounding factor for urban expansion in such regions. Since we applied regression analysis at the NUTS3 level, it is evident that the land suitable for development is abundant almost in every one of the NUTS3 regions. However, at the more local scale analysis, the literature has shown that the presence and size of NATURA2000 sites prevent land take in different locational contexts [122,123]. At the regional or local level, even some studies indicated that there is landscape fragmentation and urban sprawl observed in NATURA2000 sites [124]. These impacts can be analysed at more local levels using our data on densification and expansion, but this can be considered as a future work.

The governance parameter (PUBLIC_REG), particularly strong in northern EU countries, has an interesting effect on land expansion. As may be expected, it is negatively associated with residential land expansion when analysing all regions, indicating policies to contain urban sprawl. However, in declining regions, PUBLIC_REG yields a significantly positive estimator, possibly indicating that policies to stimulate regional growth in declining regions are more effective. In fact, fragmented and decentralised settings where there is a higher degree of autonomy at the regional/local level are characterised by lack of coordination and higher competition between local authorities [125]. Each local authority shapes its policy according to its own interests and resource endowments and seeks to achieve more urban development than the optimum at global scale [121]. Sometimes these policies aim to protect the value of housing estates in a site by restricting growth, but also originate a growth of urban land resulting in sprawled development [121]. The policies that aim at promoting growth through lower land use regulations explains the positive coefficient of the PUBLIC_REG variable estimated for declining regions. In other circumstances, local authorities can limit urban development within their boundaries; however, it can be shifted to neighbouring authorities, thus having no or minor impact on global land take [121]. Wassmer [126] argues that urban containment policies can be effective, given that it does reduce sprawl in the long term. McDonald et al. [127] found that cities with greater conservation funding have higher densities in comparison to other cities and that effectiveness of urban containment depends on the scale and the density inside and outside of the city/region boundaries. Alternatively, Cunningham [128] suggests that urban containment policies are effective if managed in aggregate metropolitan areas. To put it differently, such policies applied at the local level may generate a positive net effect on urban land development at the aggregate level [129]. In many EU countries, there is a lack of coordination among municipalities to comply with regional plans. In the ESPON Report (2007), it is acknowledged that only nine EU countries (Austria, Denmark, Finland, Germany, Latvia, Luxembourg, the Netherlands, Norway, Sweden) envisage a certain degree of coordination among municipalities when drawing up regional plans [130]. Therefore, an effective coordination is required among the regional and local administration units by the vast majority of European countries. Solly et al. [97] analysed the effectiveness of spatial governance tools in achieving sustainable land use development across the EU countries at the country, regional, and more local levels and concluded that at the regional (NUTS3) level around 70% of interventions proved effectiveness, whereas it is around 30% at more local levels. Aside from this, a study by Cortinovis et al. [58] showed that although there is progress in compactness and functional mix in the examined 175 European cities, urban expansion is still continuing inefficiently, particularly in the growing cities, resulting abandoned urban areas and fragmented agricultural land use.

The higher the prior residential density in EU regions, the lower the land expansion; thus, confirming that denser regions maintain relatively high densities [57]. The sign of DENSITY_2000 coefficient is positive in the case of declining regions. Between declining regions, higher density regions have slightly more expansion, indicating that, presumably due to less pressure on space, in these regions’ expansion entails a clear loss of density. This is again confirmed by the fact that amongst declining regions, residential density is lower in urban regions compared to more rural counterparts.

3.2. Existing City Densification

The distribution of residential density changes is given in Figure 4. The highest relative density change observed in EU countries is the considerable density growth of 33% in the islands of Ibiza and Formentera (Spain), which have seen considerable population growth, while residential land consumption has been limited. The region with the largest density decrease was the city of Suhl in the east of Germany, where due to substantial population decline and some urban expansion, residential densities dropped by 23%. Residential densities changed in both directions between the years 2000 and 2010, with the majority of changes occurring between −25 and +25.

The percentage changes in densities from the years 2000 to 2010 are presented for all EU member states in Figure 5. Among the countries losing residential densities are countries that have seen substantial population loss (the Baltic states, Bulgaria, and Romania) and some countries that presumably have seen considerable urban sprawl (Germany, Finland). The majority of countries have experienced density increases. It may be expected, particularly in growing regions, that increasing densities are a result of refurbishment of vacant housing stock or residential redevelopment in inner cities [39,113,131]. For example, Bilbao and Munich have thus maintained compact development patterns, while their populations grew considerably [39].

Figure 6 shows density changes between the years 2000 and 2010 in declining regions. Figure A2 in the Appendix A shows the regional spatial distribution of residential density changes. All declining regions showed a decrease in density, which is consistent with the expectation that population sizes change faster than the built environment decreases its footprint.

Results of regressions explaining change in residential densities are provided in

Table 4. Results of regression analysis explaining residential densification in EU countries.

Dependent Variable: Residential Densification (ha)
	All Regions		Declining Regions		Growing Regions
Variables	Standardised Coefficient	SE	Standardised Coefficient	SE	Standardised Coefficient	SE
Constant	0.836 **	0.01	0.882 **	0.02	0.901 **	0.01
GDP_CAP2000	0.036 **	0.00	-	-	0.019 **	0.00
AGRI_LAND_RENT_2000	-	-	0.015 **	0.01	-	-
POP_2000	0.009 **	0.00	0.022 **	0.01	-	-
IND_COM_2000	-	-	−0.009 **	0.01	-	-
URBAN_RECREAT_2000	-	-	0.008 *	0.01	-	-
NATURA2000	0.002 **	0.00	-	-	0.004 **	0.00
AVAIL_LAND_2000	-	-	-	-	-	-
ACCESS_2000	0.001 *	0.00	0.031 **	0.01	0.006 **	0.00
DENSITY_2000	−0.068 **	0.00	-	-	−0.008 **	0.00
METROPOL	0.011 **	0.01	0.011 *	0.01	0.008 **	0.00
URBAN	-	-	−0.02 *	0.01	-	-
RURAL	−0.175 **	0.00	-	-	−0.005 *	0.00
EASTERN_EU	-	-	-	-	-	-
SOUTHERN_EU	−0.012 *	0.00	0.027 **	0.01	−0.021 **	0.01
NORTHERN_EU	0.059 **	0.01	0.063 **	0.02	0.049 **	0.00
PUBLIC_REG	-	-	−0.005 *	0.01	-	-
PLAIN_RATIO	-	-	-	-	-	-
SLOPE	0.021 **	0.01	0.019 **	0.01	0.042 **	0.01
PRECIPITATION	0.028 **	0.01	-	-	−0.013 *	0.01
TEMPERATURE	0.836 **	0.01	-	-	0.073 **	0.01
Number of observations	1300		494		806
R-squared	0.37		0.16		0.38
F (probability)	62.38 (0.00)		8.2 (0.00)		44.03 (0.00)

Note: * Significant at 10%; ** significant at 5%. Standardised coefficients represent the regression outcomes of the variables that were standardised by dividing the values of each variable to the mean of the corresponding variable.

. As has been found by Broitman and Koomen [48], R-squared values yielded that the variance in residential density change was more difficult to explain than variance in residential expansion; for unclear reasons, residential density changes are the result of a much more evasive process. Some results are reasonably monotone despite the set of observations on which the explanatory models are fitted. Higher slopes and higher accessibility are always associated with higher densities, no doubt reflecting market conditions. Ceteris paribus, North European countries and metropolis regions always yield larger density changes. Economic productivity and protected areas only appear to have structural effects on density change in growing regions, indicating that in richer growing regions, population growth outpaces residential expansion, while the restriction of land supply by conservation areas clearly does push densities up there. However, land availability does not have a structural effect in any of the regressions. In growing regions, and when taking the EU as a whole, prior population size and prior population densities have dampening effects on density change so that residential densities more likely go down in very populated or relatively high-density regions.

A number of found results pertain only to declining regions. Agricultural land rent structurally increases density changes; this may be either because population decline is less in declining regions with higher agricultural yields, or because less land is lost to urban expansion when the conversion costs of agricultural land are higher. These findings ground on the monocentric city model [82,132] where agricultural production represents the opportunity cost of urban land expansion. On the other hand, land rents are not a significant factor in explaining urban expansion in declining regions. We found that the areas of industrial and commercial land in declining regions result in lower residential densification values, probably due to the fact that population decline occurs predominantly in regions with ageing industrial systems. Residential densities decline less in declining regions that have more land dedicated to recreational purposes. We believe this is in line with our expectations that recreation areas are positively correlated to residential densification or, at least, regions with more recreational opportunities decline less. Counterintuitively, we also find that strong public regulation and control results in lower densities in declining regions, thus contrasting our assumption that residential densification is stronger with the existence of strong policy and regulatory systems; possibly, regardless of the existence of strong regulatory systems, urban sprawl cannot be contained sufficiently in declining regions.

4. Discussion

Previous studies have mostly focused on explaining urban expansion in growing regions. However, the EU territory provides an interesting case, as it has seen fairly polarised regional population growth patterns, with migration in particular causing substantial population growth in some regions at the expense of others. The simultaneous presence of regional growth and decline in the same cross-section has been used here to investigate to what degree urban expansion and density change had different driving motivations in growing and declining regions. Differences in model results are highlighted in

Table 5. Comparison of coefficient signs for the significant variables that explain expansion density.

Dependent Variable	Residential Expansion			Residential Densification
	Signs of Stand. Coefficients			Signs of Stand. Coefficients
Variables	EU	Declining Regions	Growing Regions	EU	Declining Regions	Growing Regions
Constant	+	−	+	+	+	+
Socio-economic factors
GDP_CAP2000		−		+		+
POP_2000	+	+	+	+	+	−
AGRI_LAND_RENT_2000	−				+
Area-based characteristics
ACCESS_2000	+	+	+	+	+	+
DENSITY_2000	−	−		−		−
IND_COM_2000	NA	NA	NA		−
URBAN_RECREAT_2000	NA	NA	NA		+
Supply constraints and planning policy
NATURA2000				+		+
AVAIL_LAND_2000	+		+
PUBLIC_REG	−	−			−
Regional characteristics
METROPOL				+	+	+
URBAN		+			−
RURAL			−	−		−
EASTERN_EU	−	−
SOUTHERN_EU	−		−	−	+	−
NORTHERN_EU	−		−	+	+	+
Geological and climatic factors
PLAIN_RATIO	+	+	−
SLOPE	−		−	+	+	+
PRECIPITATION	+	+	+	−		−
TEMPERATURE	+	+	+	+		+

Note: Grey-coloured areas represent insignificant coefficients.

. Clearly, residential expansion exists even in regions with dwindling population numbers, no doubt because of a mismatch between supply and demand. The results of this study can confirm that residential expansion is driven differently in declining regions than in regions with population growth. Although the results clearly confirm the work of Haase et al. [34] and Rienow et al. [40] that substantial residential expansion occurred in declining regions in the Eastern EU, similar processes can be confirmed in declining regions in most EU countries. Residential density and GDP are important factors driving expansion; our interpretation is that residential expansion in declining regions is a process of affluent citizens leaving inner cities for suburban developments, and similar conclusions have been drawn by Buzar et al. [133], Haase et al. [34,134,135], and Salvati [136]. Models explaining residential density changes also yield different results in declining regions. The factors that only affected residential densities in declining regions are agricultural land rent, industrial/commercial and recreational sites, public land control, and urban regions. Agricultural land rent and strong land control policies are influential in supporting residential densification in declining regions, whereas GDP, existing density, NATURA2000, rural regions, precipitation, and temperature are the factors responsible for residential densification in growing regions. Clearly, when formulating expectations on land consumption, declining regions are not to be neglected, and the differences in driving factors need to be taken into account.

All in all, the declining population seems related to loss of residential densities in cities and simultaneously of the extension of built-up areas outside the city centres [137,138]. The growth of built-up land area again leads to an increase in impervious surfaces, which has further implications on the sustainability of use of land, climate change, environment, and ecosystems. In addition, the degradation of urban centres contributes to residents searching for more desirable locations [139], potentially causing vicious circles of urban degradation.

5. Conclusions

In this study, the dynamics of regional residential land between the years 2000 and 2010 were examined in EU member states. To this aim, residential land-use changes were separated into processes of change of usage density and residential expansion. We uncovered significant determinants of residential land expansion and densification processes through fitting linear regression models on the basis of geo-physical, locational, socio-economic, and policy-related characteristics of the analysed regions. In general, expansion and densification processes were affected by population size, prior residential density, land supply, accessibility, agricultural land rent, physical factors, public regulation, and regional characteristics (Table 3 and Table 4). These results confirm the previous findings on urban expansion in Europe summarised in the paper.

There is quite some consensus in the literature that sprawled development is to be avoided. Results from the current study confirm the need for efforts to contain sprawl, not only in regions harbouring population growth, but also in regions facing population decline. No doubt because of substantial economic, cultural, and demographic differences, our results show that the characteristics of land use changes are different in the groups of countries across the EU that were accounted for. Given the heterogeneity of results across EU regions, it seems clear that urban containment policies require country- and region-specific strategies that are adapted to both local and regional contexts.