Dimensions of shrinkage: Evaluating the socio-economic consequences of population decline in two medium-sized cities in Europe,using the SULD decision support tool

Abstract

All over Europe, it is a known fact that cities are shrinking. One of the main causes is population decline, but the consequent reduction of urban area is neither immediate nor easy to foresee spatially. Questions arise such as where do cities start to ‘shrink’ first? What are the most fragile areas that face the risk of becoming derelict? What are the most vulnerable social groups? And how does this affect real estate values across the city? Existing models for projecting the effects of shrinkage have been criticized for lacking spatial-explicitness, being excessively data-dependent, and failing to incorporate various socio-economic, urban and environmental aspects in the assessment of attractiveness of urban areas and of decisions by households. In this article, we attempt to overcome this criticism by applying the spatially-explicit Sustainable Urbanizing Landscape Development decision support tool (SULD), based on hedonic pricing theory, in two cities in southern Europe (Aveiro, Portugal and Imperia, Italy). SULD is used to project, assess and compare changes in land-use, household type distribution, real estate values and household densities, in three different scenarios of population decline (−5%, −10% and −15%). Results quantify the amount of contraction of urban area, housing quantity and living space; highlight the most problematic areas; and uncover low income households as the least affected, whereas the relocation of high income households may cause gentrification of medium income households in some areas of the historical city centre.

Keywords

Shrinking cities population decline decision support tool hedonic pricing simulation model

Introduction

Population grows and cities grow, it is a fact. But when population declines cities do not instantly shrink. In fact, whereas shrinking cities have most surely lost population more or less prevalently, severely, and persistently (Beauregard, 2009), they might have not yet started to spatially contract (Sousa and Pinho, 2013). Urban growth and urban shrinkage are ‘fraternal twins’, and durable housing is the chief reason their nature is so disparate, explaining slow physical reduction and notable persistence (Glaeser and Gyourko, 2005).

Shrinkage can happen everywhere in the world though within different contexts and circumstances. The phenomenon affects all urban scales, from metropolitan to rural areas, yet, at least at a first glance, does not seem to follow a standardized pattern. It manifests itself through doughnut patterns (in the centre), its reverse (in the suburbs) or mosaic and perforation patterns (mixed type) (Sousa and Pinho, 2013). The literature provides some typologies of shrinking cities, most of which of a national nature (Beauregard, 2009; Cunningham-Sabot and Fol, 2009; Sousa, 2010), albeit there have also been attempts to propose a global (mostly Western) typology (Wu et al., 2013). Haase et al. (2014a), for one, have conceptualized a heuristic model on urban shrinkage. Regardless, it is consensual that there are more fragile and volatile areas that face the risk of becoming derelict; archetypes of shrinkage such as traditional/old/non-gentrified city or metropolitan centres, first generation suburban neighbourhoods, newly built ‘ghost’ residential developments and remote rural villages. Population decrease can be caused by low fertility, related to the second demographic transition, and/or out-migration, which in turn can be triggered by countless reasons: post-industrial transformation (deindustrialization and shift from manufacturing activities to services); rapid economic breakdowns; political transformation and conflicts; natural hazards or other sources of economic decline, unemployment and destruction, or, simply, by changes in residential preferences.

Shrinkage affects social groups differently. The most vulnerable stay for lack of choice. Those who are willing and able to pay either stay or move to affluent urban areas; those who aren’t, move to the periphery or elsewhere. Glaeser and Gyourko (2005) show why declining cities tend to have lower levels of human capital (as cheap housing is comparatively more attractive to the poor) that create negative externalities or result in lower levels of innovation. This is a self-reinforcing process in which an initial decline causes concentrated poverty and segregation, which then pushes the city further down and discourages growth (Glaeser and Gyourko, 2005), causing in turn a spiral of decay of social and community life, public sphere, creativity and innovation (Keenan et al., 1999; Sousa, 2010). Shrinkage implies dramatic land-use impacts, including underutilization of infrastructures and facilities, considerable oversupply of dwellings and resulting residential vacancies or brownfield sites, which can lead to demolition and de-densification (Haase et al., 2010, 2014b). In addition to these impacts, Sousa (2010) identifies declining purchasing power, falling real-estate values (in most countries) and a slowing housing market.

Consequently, recent urban shrinkage is a paramount issue in worldwide planning agendas. Haase et al. (2012) urge for a proper knowledge of the nuances of shrinkage. Likewise, Hollander (2011) discusses the importance of understanding how smart decline and shrinkage strategies/policies might affect neighbourhood change and quality. If land consumption due to residential development, economic growth and transportation is one of the most serious environmental pressures on landscapes (Nuissl et al., 2009), land abandonment is one of the most challenging planning problems facing shrinking cities, both in the United States (Hackworth, 2014) and in Europe, where, according to Kroll and Haase (2010), it actually should be regarded as one of the main components for future land use development. Based on Swedish and OECD panel results, Lindh and Malmberg (2008) show how, although not strictly conclusive, theoretical considerations and empirical data support a clear effect of demographic change on residential construction even if the price effect for several reasons may be less clear-cut, especially in Mediterranean countries, such as Portugal and Italy (Hoekstra and Vakili-Zad, 2011). Likewise, in East Germany, Kroll and Haase (2010) argue that a decreasing population does not mean less land consumption for housing and transportation.

Shrinking, therefore, is not just a question of losing inhabitants. Shrinkage affects economic, social, land-use, morphological and even environmental aspects of urban life. Existing models for projecting the effects of shrinkage, however, as we will discuss in the next section, have been criticized for lacking spatial-explicitness, being excessively data-dependent, and failing to incorporate various socio-economic, urban and environmental aspects in the assessment of attractiveness of urban areas and of decisions by households. This article aims to address and overcome this criticism by applying the spatially-explicit Sustainable Urbanizing Landscape Development (SULD) decision support tool (Roebeling et al., 2007), a simulation model that builds on hedonic pricing theory and can predict changes in land-use, household type distribution, real estate values and household densities. SULD is applied to two cities in Southern Europe (Aveiro, Portugal and Imperia, Italy) in three possible (and probable) scenarios of population decline, namely −5%, −10% and −15%.

New tools for a new problem

Cities are very complex urban systems. Simulation models on urban land-use change processes are an attempt to better understand and explain these systems, namely development paths, causal driving forces and their expected impacts, as to support urban and regional planning. Up until very recently, urban growth was the main focus of planning and especially of land-use research. Nevertheless, new challenges have arisen since urban shrinkage and ensuing unanticipated vacancy entered the research agenda (Blanco et al., 2009; Schwarz et al., 2010; Sousa and Pinho, 2013).

With the purpose of establishing a conceptual framework for model enhancement, Schwarz et al. (2010) reviewed 21 recent urban land-use-change simulation models, stemming from four different modelling approaches/methodologies: (a) system dynamics, (b) integrated transport-urban models, (c) cellular automata (CA) and (d) agent-based modelling (ABM). Their results show that simulation models are not homogeneous in the causalities and feedbacks considered in explaining urban land-use dynamics. The criteria required to model urban shrinkage in a spatially explicit way was not satisfied by one single model. Nevertheless, according to Schwarz et al. (2010), system-dynamic models documented in the literature can serve as a good point of departure for spatially non-explicit simulation, and one case was found for linked transport-urban models, which comprised aspects of urban shrinkage. The potential of CA was imprecise as spatially explicit data on vacancy to feed it is typically inaccessible.

ABM appeared to be the most capable approach for spatially explicit simulations of urban shrinkage, enabling researchers to explicitly incorporate household-location choice, housing-market development and urban planning. As an alternative to calibrating the model for (lacking) historical data on residential vacancy, explicit decision rules which determine where vacancies take place and where building stock should be demolished could be included. Moreover, it had the potential to simulate innovative land-use projects as an emergent feature of urban communication and governance (Schwarz et al., 2010).

In this respect, the literature points out several research gaps to tackle in the future. For one, there is lack of time series and cross-section data (Lee et al., 2001) for calibrating and validating simulation models, especially regarding decisions made by households (Schwarz et al., 2010). Lee et al. (2001) believed that a better understanding of the user cost and capital stocks to obtain and recover more precise and meaningful estimates of structural parameters was needed. Planning and redevelopment models that offer a holistic approach to the challenges arising from continual population loss and the increasing presence of aspects of urban shrinkage (e.g. vacant abandoned properties or demolition measures) are rare (Schilling and Logan, 2008; Schwarz et al., 2010).

Nuissl et al. (2009) and Schwarz et al. (2010) argue for the inclusion of socio-economic aspects (e.g. infrastructure costs, socio-spatial segregation) in the evaluation of land use transition processes in conditions of decline. Furthermore, Schwarz et al. (2010) concluded that only a few models integrate environmental factors to determine the attractiveness of cells or regions. For the authors, the conceptual framework should be expanded to more ‘impact dimensions’, whereas Haase et al. (2010) believe that it should be expanded to incorporate (a) scenarios of decreased or modified infrastructure supply; (b) supplementary economic variables considering the economic constraints of individual households and their choices concerning transport modes and travel distances and (c) a more detailed local housing market as well as contextual constraints for land use policy and planning in the form of scenario alternatives.

Consequently, Hollander (2011) highlights the importance of researching the relationships between population and housing change to better understand and evaluate how smart decline and shrinkage strategies/policies might affect neighbourhood change and quality. He asserts the need to test this relationship through more sophisticated multivariate statistical analysis (Hollander, 2011). Therefore, more primary research is necessary, through qualitative fieldwork, to explore how residents of shrinking cities perceive their changing milieus and how they construe shrinkage. Similarly, Haase et al. (2012) stress the true knowledge of the advantages and disadvantages of shrinkage in terms of land use, sustainable resource use and quality of life.

The impact of housing prices, however, is one of the most important concerns regarding shrinkage, especially in hedonic pricing modelling. Keenan et al. (1999) argue that, in shrinkage situations, housing prices fall as a result of elements such as landlord competition for tenants, clearance threats, shop closure, service reduction or disinvestment and blight. Regardless of several ‘shock absorbers’, abandonment is an inevitable consequence of over-supply in some parts of the housing market, which spreads as demand drops and as a neighbourhood loses its status, and not even good, newly built and refurbished housing stock is insusceptible to this problem (Keenan et al., 1999). Furthermore, Glaeser and Gyourko (2005) find that durable housing predicts that exogenous shocks lead to different asymmetric responses of housing prices and population sizes – negative shocks affect housing prices more than population, while positive shocks affect population more than housing prices. An implication is that the distribution of housing prices is a predictor of future population growth (Glaeser and Gyourko, 2005). In fact, the authors’ data show that growth is quite uncommon in cities with large shares of their housing stock valued below the cost of new construction.

In addition, Levin et al. (2009) use a difference-in-differences methodology to explore the effect of population decline and ageing on housing prices in the UK. Their comparative analysis suggests that differences in housing prices are consistently related to differences in demography, i.e. population decline, size and ageing substantively and significantly determine the value of and put downward pressure on housing prices, supporting previous findings of Mankiw and Weil (1989). Consistent with this view is the assumption that the long-run trend of rising real housing prices will not carry on into the future. These results contrast those from Hoekstra and Vakili-Zad (2011) regarding Mediterranean countries. They highlight the so-called ‘Spanish paradox’, which stands for both rising real estate prices and a high or even rising vacancy rate, explained by the fact that not all vacant dwellings are actually available on the housing market.

Last, regarding the transferability of simulation approaches, authors such as Schwartz et al. (2010) urge for simpler, easily understandable applications, with more recognizable benefits for those who are not experts in modelling.

The SULD decision support tool

The SULD decision support tool (Roebeling et al., 2007) is a GIS-based optimization model, that builds on hedonic pricing theory and that is based on a classic urban-economic model with environmental amenities (Mills, 1981; O'Sullivan, 2000; Wu and Plantinga, 2003). In essence, SULD determines the value of housing given its location relative to urban activity generators and environmental amenities – i.e. the equilibrium price for which demand for and supply of housing are equal (Roebeling et al., 2007, 2016). In this sense, SULD is able to address most of the questions the literature review has raised. SULD is a spatially explicit model, which requires only a limited amount of input data (as compared with above cited models), uses housing prices as a predictor of demographic change, includes socio-economic aspects (such as household types, transportation costs, environmental amenities and urban activity generators) to determine the attractiveness of locations, it is a simulation, rather than a regression model (thus it does not require, for example, past property sales and market data), allows for the assessment and comparison of scenario alternatives and, last, enables results to be presented not only in tables and graphs but also in colour-maps that are easily interpreted by non-modelling experts.

In SULD, the demand side (equation 1) is represented by households, characterized by their preferences for a certain set of goods and services: residential space S, other goods and services Z, and environmental amenities e. The utility obtained by households in each location is a function of their preferences, distance to environmental amenities and income. Households maximize their utility U at location i subject to the budget constraint y, which is spent on housing S, other goods and services Z, and transportation between the residential area and the urban activity generator (p_xx):

{Max}_{S_{i}, Z_{i}} U_{i} (S_{i}, Z_{i}) = S_{i}^{μ} Z_{i}^{(1 - μ)} e_{i}^{ɛ}

(1)

subjectto y_{i} = p_{i}^{h} S_{i} + Z_{i} + p_{x} x_{i}

where U_i is household utility, S_i is residential space, Z_i are other goods and services, e_i the environmental amenity value, y_i is household income, p_i^h is the rental price of housing, p_x are annual transportation costs and x_i is the distance to the urban activity generator.

The environmental amenity value e_i that the household experiences at location i is decreasing with distance from the amenity source, and is determined by:

e_{i} = 1 + a \cdot \exp - β z_{i}

(2)

where a is the level of amenity provided by the environmental amenity, β is the amenity distribution factor and z_i is the distance from location i to the environmental amenity. The household’s bid-rent price at a given location can now be derived (Roebeling et al., 2007, 2016), and gives the household’s maximum willingness to pay for housing (p_i^h*) at location i. Considering that households face a given rental price (they are ‘price-takers’), they can select the faced rental price and obtained utility/welfare from the environmental amenities and other goods by choosing the residential location i.

The supply side (equation 3) is represented by developers, who maximize their profit by trading off returns from housing development density net of associated development costs, subject to households’ willingness to pay for housing. Developers aim to maximize their profit π at location i, which is given by the revenue of construction (p^hD) net of incurred development costs (l + D^η):

{Max}_{D_{i}} π_{i} (D_{i}) = p_{i}^{h} D_{i} - (l_{i} + D_{i}^{η}) with D_{i} = n_{i} S_{i}

(3)

where π_i is the developer’s profit, D_i is the development density, p_i^h is the rental price of housing, l_i is the opportunity cost of land, D_i^η is the construction cost function, η is the ratio of housing value to non-land construction costs (Wu, 2006), n_i is household density and S_i is residential space. The developers bid-price for land can then be derived (Roebeling et al., 2007, 2016) and they will develop when residential land rents (p_i^hD_i) are larger than the opportunity cost of development (l + D_i^η; which is equal to the forgone land rents l_i and the costs of converting land D_i^η).

Finally, equilibrium occurs where supply for housing equals demand for housing. The equilibrium land rent price r_i at a given location i can then be derived (Roebeling et al., 2007, 2016), and development patterns for a certain population size are determined given the location of urban activity generators and environmental amenities.

The SULD decision support tool builds on a numerical application of the above-described model (Roebeling et al., 2007, 2016), using the General Algebraic Modelling System (GAMS 21.3; Brooke et al., 1998). The objective function maximizes, for a given household population Q_t, benefits B from residential land uses L_i^res and non-residential land uses L_i^nres net of development costs (l + D^η) over all locations i, so that:

{Max}_{L_{i}} B (L_{i}) = \sum_{i} (l_{i} L_{i}^{nres} + (r_{i} - l_{i} - D_{i}^{η}) L_{i}^{res})

(4)

subject to

Q_{t} = \sum_{i} n_{i}

and

L_{i}^{res} + L_{i}^{nres} = a_{i}

, and where l_i is the opportunity cost of land, r_i is the land rent price and a_i is the grid-cell area at location i. Note that land use conversion can take place between residential and user-defined non-residential land uses – the remaining land uses are fixed.

Introduction to the case studies – Modelling and the reality of population decline

Aveiro (Portugal) and Imperia (Italy) are fairly similar in size and dynamics. Aveiro (∼26,000 inhabitants and ∼1,219 inhabitants/km²) is located about 250 km North of the capital Lisbon, and faces the Ria de Aveiro lagoon on the Atlantic Ocean. Imperia (∼37,000 inhabitants ∼920 inhabitants/km²) is located in the North of Italy, less than 100 km from Nice, and faces the Mediterranean Sea. Both receive a considerable amount of seasonal tourists, which accounts for an important slice of their economies, both depend on seaside activities and both are famous for their food industry (traditional sweets in Aveiro; olive oil and pasta in Imperia). Aveiro also hosts one of the most important Universities in Portugal, whilst Imperia hosts a regional Campus of the University of Genoa.

The Aveiro case study focuses on the city’s larger urban area, comprising a total area of 21.4 km² (Figure 1(a); source: EEA, 2009). The City of Aveiro itself (∼3 km²) – the smaller area that we shall focus on this analysis – is surrounded by the Ria de Aveiro lagoon (to the Northwest), three satellite villages to the East (Aradas, São Bernardo and Esgueira) and several agricultural areas (also to the East). The city is serviced by one highway (A25), one provincial road (N109) and an intercity railway station. The city of Aveiro has four major environmental amenities, including three urban parks (Rossio Garden (#1), the University of Aveiro campus gardens (#2) and the Santo Antonio Park (#3)) as well as water elements (#4). Six main urban activity generators have been considered, including the historical city centre, three shopping centres (Forum, Glicínias and Taboeira), the railway station and the University of Aveiro (see white dots in Figure 1(a)).

The Imperia case study focuses on the area East and West of the Impero river, and comprises an area of 2.6 km² (Figure 1(b); source: Citta D'Imperia, 2012). The city is surrounded by olive-grown hillside areas to the North and the Mediterranean Sea to the South. It is serviced by one major highway (A10), two provincial roads (Via Nazionale) and one railway station. There are five major environmental amenities, including one urban park (Villa Grock (#1)), one neighbourhood park (San Leonardo Park (#2)), one local park (Arturo Toscanini garden (#3)) and two water elements (Impero river (#4) and the Mediterranean Sea (#5)). Seven main urban activity generators have been highlighted, including a shopping district, a stadium, a library, the railway station, a museum and several schools, as well as numerous bus stops (see white dots in Figure 1(b)).

Both cities, however, risk population decrease. In Aveiro, this is expected to occur in the near future. While the Aveiro region underwent severe urbanization and industrialization over the last decades (population in the city increased from around 24,766 in 1991, to 26,197 in 2001, to 27,913 in 2011), population growth is currently stagnating. In combination with falling fertility rates (1.3 in 2009; OECD, 2011) and increased emigration (Hugh, 2013), it is expected that the population will decrease over the next decade. Furthermore, the financial crisis has resulted in the closure of numerous small and medium enterprises as well as a contraction of the University of Aveiro students and workforce.

In Imperia, population decrease is already a reality. According to ISTAT (2015), Italy has the second oldest population in Europe after Germany, and in 2013, the highest (negative) difference ever between birth and deaths was recorded. This decline is worse in the Italian Islands and in the North. Accordingly, the same source (ISTAT, 2015) shows a population decline in the city of Imperia, from 40,689 inhabitants in 1992 to 38,395 inhabitants in 2002, to 37,633 inhabitants in 2011. Furthermore, a large food factory is in the eminence of closing, causing a sudden unemployment problem that may also contribute to population decline. During the summer season, however, Imperia’s population can increase considerably (up to +75%), due to the amount of second homes, holiday homes and tourist establishments.

For this study, the definition of household socio-economic characteristics was based on available statistics on population, household size, expendable income and expenditure distribution (see Table 1). For Aveiro, the total population living in the study area is 26,078 and the total number of households equals 10,911 (INE, 2012), resulting in an average household size of 2.39 persons/household. For Imperia, the total population living in the study area is considered to be 61,405 (taking into account full occupancy of second homes, holiday homes and tourist establishments) and the total number of households equals 29,240 (SISTAN, 2012), resulting in an average household size of 2.10 persons/household. Full occupancy needs to be considered when modelling Imperia to achieve realistic results because a city’s constructed area (offer) is determined by the total amount of inhabitants (both residents and seasonal visitors) that seek real estate in that city (demand). By procuring real-estate, seasonal home owners also exert pressure on the real estate market, increasing demand. For example, they may purchase or let homes that are left vacant in the off-season.

Table 1.

Household characteristics for the Aveiro case study area (based on INE, 2012) and the Imperia case study (based on Comuni Italiani, 2010; SISTAN, 2012).

	Unit	City	Household type 1	Household type 2	Household type 3	Total
Demographics
Population	#	Aveiro	4,929	18,046	3,103	26,078
Population	#	Imperia	17,973	39,446	3,987	61,406
Household size	#/hh	Aveiro	2.39	2.39	2.39	2.39
Household size	#/hh	Imperia	2.10	2.10	2.10	2.10
Households (Q)	#	Aveiro	2,062	7,551	1,298	10,911
Households (Q)	#	Imperia	8,559	18,784	1,899	29,242
Household budget
Expendable income (y)	€/year	Aveiro	9,473	19,849	49,155	233.2 × 10⁶
Expendable income (y)	€/year	Imperia	9,599	23,998	85,431	695.2 × 10⁶
Housing expenditures (µ)	%	Aveiro	23.5%	23.0%	21.4%	22.6%
Housing expenditures (µ)	%	Imperia	30.0%	29.0%	28.0%	29.0%

Based on income data for the Centro region in Portugal (INE, 2012), and the Imperia region in Italy (Comuni Italiani, 2010), we distinguish three social groups: low, middle and high income households. For Aveiro, the low income household type (HHtype1) corresponds to 19% of the population that earns 8% of total income, and comprises student and retiree households. The middle income household type (HHtype2) corresponds to 69% of the population that earns 64% of total income, and comprises working young families with children. Finally, the high income household type (HHtype3) corresponds to 12% of the population that earns 27% of total income, and comprises highly educated professionals and business men.

On the other hand, for Imperia HHtype1 corresponds to 29% of the population that earns 10% of total income, and comprises, in particular, retirees. HHtype2 corresponds to 64% of the population that earns 70% of total income, and comprises working families with children. Finally, HHtype3 corresponds to 6% of the population that earns 20% of total income, and comprises business men and non-resident population.

Finally, housing expenditures are obtained for the identified household types based on household expenditure data for the Centro region in Portugal (INE, 2012) and the Imperia region in Italy (Comuni Italiani, 2010; SISTAN, 2012). Households spend on average 22.6% (in Aveiro) and 29.0% (in Imperia) of their income on housing, with low income households spending relatively more (23.5% in Aveiro and 30.0% in Imperia) and high income households relatively less (21.4% in Aveiro and 28.0% in Imperia) than average.

Modelling, calibrating and validating the base scenario

The numerical application of SULD provided spatially explicit information on the consequences of population decrease (−5%, −10% and −15%) in Aveiro and Imperia, in terms of changes in land-use, household types distribution, real estate values and household densities. As described in the previous section, in Aveiro, the entire study area encompasses 4.625 km by 4.625 km (=21.39 km²) and in Imperia 1.60 km by 1.60 km (=2.56 km²) – see Figure 1. In the case of Aveiro, in order for the model to consider just the area corresponding to the City of Aveiro (3 km², see previous section), and thus for the case studies to be comparable, the area corresponding to the satellite villages has been catalogued as ‘open space’ (see Figures 2 and 4). In both cases, using ArcGIS software, the study area was covered by a grid layer of 185 by 185 (= 34,225) cells, such that the cell size was 25 m by 25 m in Aveiro, and 8.65 m by 8.65 m in Imperia. Although differences in cell size influence the detail included in and obtained from the scenario simulations, it does not influence the overall outcomes and conclusions obtained from the case study analyses nor their comparison.

To create the base-scenarios for both cities (i.e. reproductions of existing conditions), an extensive process of model parametrization, calibration and validation has been performed. Parameter values for the numerical application of SULD to the case studies are thus based on a population comprising three household types (low, middle and high income households), differentiated by number of households (Q), levels of expendable income (y) and shares of housing expenditures (µ; see Table 1) as well as levels of utility u. The latter were, for Aveiro, u = 3,175 for HHtype1, u = 6,652 for HHtype2 and u = 16,475 for HHtype3; and for Imperia, u = 2,000 for HHtype1, u = 5,000 for HHtype2 and u = 17,800 for HHtype3. All household types share the same appreciation for environmental amenities; ɛ = 0.08; a = 10.0/7.5/5.0; β = 1.0 (based on Wu, 2001, 2006) – that reflect exponential distance decay (in line with Tobler’s (1970) first law of geography) and are differentiated between park categories (following Lutzenhiser and Netusil, 2001).¹ Annual transportation costs, p_x, are 250 €/km for both Aveiro and Imperia, that share similar public and private transport costs (0.25 €/km and 0.45 €/km, respectively) as well as commuting expenditures (∼11% of expendable income) (INE, 2012; SISTAN, 2012). Opportunity costs of land l are 530 €/ha/year in Aveiro (Roebeling et al., 2014) and 1000 €/ha/year in Imperia (SISTAN, 2012), and construction costs are η = 1.50 for Aveiro (given average real estate values of 876 €/m² and construction costs of between 450 and 650 €/m² for Central Portugal; INE, 2012) and η = 1.65 for Imperia (given average real estate values of 2,520 €/m² and construction costs of between 1,400 and 1,600 €/m² for the Imperia municipality; SISTAN, 2012). In Aveiro, all environmental amenities have an equal amenity value (a = 10.0), whilst in Imperia, the environmental amenities possess varying amenity values (a = 10.0/7.5/5.0). Model supply, demand and equilibrium conditions are properly scaled based on straight-line distances to environmental amenities and road-network distances to urban centres, so scale or cell size differences will not influence the overall direction of the results.

Model calibration took place with respect to levels of household utility u (the only non-observable parameter), thereby aiming to best approximate real world patters. The degree of this approximation has been validated by (a) comparing modelled and observed patterns and (b) by discussing all model results with researchers and stakeholders in regular meetings. In the first case, base scenario simulation results are assessed against the most recent land use data (Citta D'Imperia, 2012; EEA, 2009, see Figure 1) using quantity and location disagreement indicators (Pontius Jr et al., 2004). For Aveiro, results show that the number of urban land use cells is overestimated by 6% (quantity disagreement) and that the location of urban cells is correct in 79% of cases (location agreement). For Imperia, results show that the number of urban land use cells is underestimated by 4% (quantity disagreement) and that the location of urban cells is correct in 88% of cases (location agreement). Validated results for the base scenario are displayed spatially in Figures 2 and 3, for Aveiro and Imperia, respectively, and numerically in Table 2, which also displays the results for Scenarios 1, 2 and 3 (−5%, −10% and −15% population, respectively), discussed in the next section.

Table 2.

Base run and scenario simulation results for the Aveiro and Imperia case studies.

		Aveiro				Imperia
Variable	Unit	Base scenario	Scenario 1	Scenario 2	Scenario 3	Base scenario	Scenario 1	Scenario 2	Scenario 3
Land use
Forest	ha	230	0.0%	0.0%	0.0%	6	0.0%	0.0%	0.0%
Water	ha	156	0.0%	0.0%	0.0%	44	0.0%	0.0%	0.0%
Open space	ha	973	2.8%	5.5%	8.1%	24	23.7%	47.0%	69.9%
Industry/commerce	ha	250	0.0%	0.0%	0.0%	43	0.0%	0.0%	0.0%
Park_urban	ha	56	0.0%	0.0%	0.0%	7	0.0%	0.0%	0.0%
Roads	ha	128	0.0%	0.0%	0.0%	33	0.0%	0.0%	0.0%
Urban	ha	346	−8.0%	−15.6%	−22.6%	99	−5.6%	−11.2%	−16.6%
Total	ha	2,139	0.0%	0.0%	0.0%	256	0.0%	0.0%	0.0%
Population
HHType1	#	4,929	−5.0%	−10.0%	−15.0%	17,973	−5.0%	−10.0%	−15.0%
HHType2	#	18,046	−5.0%	−10.0%	−15.0%	39,446	−5.0%	−10.0%	−15.0%
HHType3	#	3,103	−5.0%	−10.0%	−15.0%	3,987	−5.0%	−10.0%	−15.0%
Total	#	26,078	−5.0%	−10.0%	−15.0%	61,405	−5.0%	−10.0%	−15.0%
Housing quantity
HHType1	1000 m²	66.4	−6.1%	−12.0%	−17.9%	143.0	−5.3%	−10.6%	−15.7%
HHType2	1000 m²	733.2	−9.7%	−18.5%	−26.2%	10,27.2	−5.9%	−11.6%	−17.3%
HHType3	1000 m²	165.6	−8.5%	−17.4%	−27.2%	670.9	−6.8%	−13.1%	−19.1%
Total	1000 m²	965.2	−9.2%	−17.9%	−25.8%	1,841.1	−6.2%	−12.1%	−17.8%
Living space
HHType1	m²/hh	89.3	−0.2%	−0.4%	−0.6%	56.8	−0.1%	−0.2%	−0.2%
HHType2	m²/hh	168.6	−1.1%	−2.1%	−2.9%	116.5	−0.2%	−0.5%	−0.7%
HHType3	m²/hh	277.9	−0.9%	−2.1%	−3.9%	337.9	−0.5%	−1.0%	−1.4%
Average	m²/hh	166.6	−0.9%	−1.9%	−2.9%	113.4	−0.3%	−0.5%	−0.8%
Real estate value
HHType1	€/m²/year	23.9	0.3%	0.6%	1.0%	50.5	0.1%	0.2%	0.3%
HHType2	€/m²/year	25.9	1.4%	2.7%	3.9%	59.5	0.3%	0.5%	0.8%
HHType3	€/m²/year	37.0	1.0%	2.2%	4.0%	70.6	0.6%	1.0%	1.4%
Average	€/m²/year	26.8	1.2%	2.3%	3.3%	59.4	0.3%	0.5%	0.7%
Total	m€/year	50.7	−4.8%	−9.6%	−14.5%	200.0	−5.0%	−10.0%	−15.0%

Analyzing the base scenario results, it can be seen that the city of Aveiro comprises mostly urban residential (364 ha) and industry/commercial (250 ha) areas, surrounded by open-space/agricultural, forest and water areas. The population in the city is mainly middle income (HHType2; 69%). Lower income households tend to live closer to urban activity generators (particularly along the axis from the historical city centre to the railway station), while higher income households live in attractive areas close to the waterfront and the urban parks.

The total built area (housing quantity) equals ∼0.9 × 10⁶m², and the total floor-space (development density) equals twice this area (∼1.8 × 10⁶m²), with middle income household comprising the larger share in each (76% and 70%, respectively). Household density is highest in low income areas (up to well over 4.5 households per grid cell), lower in attractive high income areas (up to 3.5 households per grid cell) and lowest in middle income areas on the outskirts of the city (up to 2.5 households per grid cell). Available living space equals, on average, about 167 m² per household, ranging from 89 m² for low income to 278 m² for high income households. Real estate (rental) values equal, on average about 27 €/m²/year, varying between 24 €/m²/year for low and 37 €/m²/year for high income households. Largest values can be observed in attractive high income areas (up to well over 41 €/m²/year) and lowest values can be observed in low income areas close to urban activity generators and, in particular, the railway station (up to 20 €/m²/year). The total real estate (rental) value for the city of Aveiro equals just over 50 million Euros per year.

As for Imperia, the study area comprises an urban residential (99 ha) and industry/commerce (43 ha) area of about 142 ha, and is surrounded by open-space/agricultural areas and the Mediterranean Sea as well as crossed by the Impero river (see Table 2 and Figure 3). The population is distributed as 29% low income, 64% middle income and only 6% high income households. High income households are mainly located in the Northeast of the study area, away from urban activity generators and main roads, and to a minor extent in small strips near the sea. Low income households are located near transport hubs and urban activity generators.

The total built area (housing quantity) equals ∼1.8 × 10⁶m², distributed over low (8%), middle (56%) and high (36%) income households. Household density is highest in low income areas (up to over 3.7 households per grid cell) and lower in attractive high income areas (as low as 0.9 households per grid cell). Available living space equals, on average, around 113 m² per household, ranging from 57 m² for low income to 338 m² for high income households. Real estate (rental) values equal, on average, about 59 €/m²/year, varying between 51 €/m²/year for low income to 71 €/m²/year for high income households. Larger values can be observed in attractive high income areas (up to 73 €/m²/year), particularly in the Northeast near Villa Grock, and lowest values can be observed in low income areas (up to 47 €/m²/year), particularly along major roads. The total real estate (rental) value for the study area in Imperia equals 200 million Euros per year.

Simulating population decline of 5%, 10% and 15%

This section presents and discusses, for each city, scenario simulations for the decline of population of −5%, −10% and −15% (Scenarios 1, 2 and 3, respectively; see Table 2).

For Aveiro (see Table 2 and Figure 4), population decline results in a significant decrease in urban area and a corresponding increase in open space. Open space increases with about half the percentage of population decline (+2.8%, +5.5% and +8.1% for Scenarios 1, 2 and 3, respectively), whilst the urban area decreases with about 1.5 times the percentage of population decline (−8.0%, −5.6% and −22.6% for Scenarios 1, 2 and 3, respectively). In both cases, the greater impact, comparatively, is felt when population reduces only −5%. The contraction in urban area occurs at the edges of the city where mainly middle income households live, especially in the Northeast, and is intensified in these locations as population continues to decline. That is, the city becomes much more compact, particularly losing its urban continuity to the northern localities.

Overall across the entire study area, the total built area (housing quantity) decreases by −9.2%, −17.9% and −25.8%, and the living space decreases by −0.9%, −1.9% and −2.9%, for Scenarios 1, 2 and 3, respectively. For housing quantity, the greatest impact is felt, comparatively, when population decreases −5%, but for living space, the reduction is proportional to population decline. In both cases, this decrease is largest for middle and high income households, implying that these households relocate to attractive areas closer to the historical city centre and are willing to accept a smaller living space in so doing. For Scenario 1 (−5% population), the greatest decrease in housing quantity and living space is observed for HHtype2 (middle income households), reaching −9.7% and −1.1%, respectively. Yet in Scenario 3 (−15% population), the greatest decrease is observed for HHtype3 (high income households), respectively, −27.2% and −3.9% for housing quantity and living space. In this extreme scenario of population decline, high income households move from attractive peripheral areas, closer to the water yet farther from the contracting urban core, to more attractive areas near the city centre (i.e. closer to the historical centre as well as water and urban parks), that are vacated by low and particularly middle income households.² Low income households, living closer to the urban city centre and major road connections, are much less affected by the contraction in urban area associated to population decline.

Because of this contraction, there is a proportional reduction in total real estate (rental) value for the study area, culminating in losses of over 1.6 million Euros per year when the population declines by −15%. Yet looking at the real estate values per household type, there are notable increases. This can be explained by the reduction in living space and housing quantity as well as by the relocation of higher income households to areas closer to the historical city centre. On average, real estate values increase with +1.2%, +2.3% and +3.3% for Scenarios 1, 2 and 3, respectively. For Scenarios 1 and 2 (−5% and −10% population), real estate values increase most for middle income households (+1.4% and +2.7%), whilst for Scenario 3 (−15% population) values increase most for high income households (+4.0%). The maximum increase for low income households is observed for Scenario 3 (about +1%).

Summarizing, population decline in the city of Aveiro will lead to a more condensed city, higher real estate (rental) values, but a significant net decrease in total real estate value. Middle and high income areas at the edges of the city and in suburbia, near the water, will be abandoned as the households therein seek more attractive locations closer to the historical city centre, accepting smaller living spaces and higher real estate (rental) values. Lower income areas, near the traditional city centre and major road connections, are the least affected by population decline.

Considering now the city of Imperia (see Table 2 and Figure 5), population decline results in a more accentuated increase in open space and a somewhat similar decrease in urban area than that witnessed in Aveiro. Open space increases with almost 5 times the percentage of population decline (+23.7%, +47.0% and +70.1% for Scenarios 1, 2 and 3, respectively), whilst urban area decreases with about 1.1 times the percentage of population decline (−5.6%, −11.2% and −16.6% for Scenarios 1, 2 and 3, respectively). Again, in both cases, the greater impact, comparatively, is felt when population declines only by −5%. The contraction in urban area occurs at the edges of the city, in particular, in middle and high income areas in the west as well as high income areas in the northeast. As population continues to decrease, the contraction in these areas is intensified, in the direction of the historical city centre. The city becomes more compact, in a half-circle around the main water elements.

Overall across the entire study area, the total built area (housing quantity) decreases by −6.2%, −12.1% and −17.8%, whilst living space decreases by −0.3%, −0.5% and −0.8%, for Scenarios 1, 2 and 3, respectively. Again, for housing quantity, the greatest impact is felt, comparatively, when population decreases −5%, while for living space, the reduction is proportional to population decline. For both variables and for the three scenarios, the decrease is largest for high income households, ranging from −6.8% and −0.5% for housing quantity and living space, respectively, in Scenario 1, to −19.1% and −1.4%, respectively, in Scenario 3. Like in Aveiro, as the city contracts, high income households relocate from the city’s edges to other attractive areas closer to the historic city centre (i.e. near water elements, urban parks and urban amenities), where the trade-off is a reduction in living space. Low and middle income households are in this case less affected. The decline in total built area is somewhat similar between HHtype1 and HHtype2 (about −5% in Scenario 1 to −16% in Scenario 3), and the reduction in living space is less than half of that observed for HHtype3.

The reduction in urban area associated with the decline in population leads to a proportional reduction in terms of total real estate (rental) value for the entire study area, culminating in losses of over 8.4 million Euros per year when the population declines by −15%. Looking at the real estate values for each household type, there are increases, due to the reduction of living space and housing quantity, and particularly due to the relocation of higher income households to more attractive areas closer to the historical city centre. On average, real estate values increase with +0.3%, +0.5% and +0.7% for Scenarios 1, 2 and 3, respectively. Real estate values increase more for high income households (HHtype3), with +0.6%, +1.0% and +1.4% for Scenarios 1, 2 and 3, respectively, as more households compete for the best locations in the city centre. Perhaps due to gentrification processes in these locations, real estate values for middle income households increase, particularly in Scenario 3 (+0.8%). Low income households, living around major road connections and urban activity generators, are hardly affected in this perspective (real estate value increases range from +0.1% to +0.3%).

Summarizing, population decline in the city of Imperia will lead to a more condensed city, higher real estate (rental) values, but a significant net decrease in total real estate value. Middle and high income areas at the edges of the city will be abandoned as households therein seek more attractive locations closer to the historic city centre and water elements, accepting smaller living spaces and higher real estate (rental) values. This process is more prominent for high income households than middle income households, which can be affected by gentrification processes, whilst low income households are the least affected by population decline.

Conclusion

The SULD decision support tool was applied to two cities in Mediterranean countries; Aveiro (Portugal) and Imperia (Italy), which risk further population decrease in the near future. Testing three scenarios of population decline (−5%, −10% and −15%), similar tendencies were found in both cities. As population declines, cities become more condensed. Urban area decreases by 1 to 1.5 times the percentage of population decline, whilst housing quantity and living space also decrease. Housing quantity decreases up to −26% and −18%, and living space decreases up to −2.9% and −0.8% in Aveiro and Imperia, respectively. Comparatively, this decline is more accentuated in the first 5% decrease of the population, i.e. population decline is immediately felt in the structure of these medium-sized cities. The most potential problematic areas identified are the edges of the city, farther from urban activity generators, environmental amenities and major road connections. Here, neighbourhoods have the risk of becoming abandoned, despite having been occupied as well by high and middle income households. As the city contracts, higher income households move closer to the city centre (nearer historical areas and environmental amenities) that are vacated by lower income households. On the other hand, low income households, living near the traditional city centre and major road connections, appear to be less affected by population decline. Nevertheless, gentrification in some parts of the city is a foreseen consequence of population decline, particularly in the scenarios of −10% and −15% decrease.

A point could be made as to whether these relocations can be a realistic thing to expect in cities of such comparatively small size that are also shrinking. We must remember, though, that size is relative. Aveiro, for example, is a relatively small sized city from a global, or even an European perspective, but it is actually one of the ten largest cities in Portugal. The inhabitants therein do not regard the size of their city in a global or European perspective; they regard it as their own reality and experience it with the same cultural preconceptions as other Portuguese inhabitants of larger cities. Patterns of living in smaller southern European towns such as these are characterized by an unusual amount of trips a day, far above the average home-work-home pattern of larger cities, precisely because the distances are smaller. These trips are made primarily using individual motorized transport, a use that increases with the increase in monthly income and professional status – a classic behaviour in smaller Mediterranean towns (Bento et al., 2015). Thus, traffic jams do occur at rush hours, even though the average length of the trips is generally small, the perception of distance is exacerbated for most inhabitants and the travel-time cost is relatively high (considering household incomes), i.e. it is much greater than what could be expected in such relatively small cities. Along with other reasons, such as the intangible quality of old/historical city centres, the requalification of green spaces or waterfront areas, the specific nature of South European housing markets, characterized by growing urbanization, high rates of home-ownership and low incidence of social housing (Allen, 2006; Balchin, 2013; Elsinga and Hoekstra, 2005), or the ‘Spanish Paradox’ theory discussed in the ‘New tools for a new problem’ section, it is then possible, and even probable, that in these cities middle and high income groups will make the move to the centre in a situation of shrinkage, contributing to higher housing prices there.

Consequently, although there is a decrease in total real estate values in the city (due to the decline in population), there is an increase in real estate (rental) values per household type (as relatively more households live in more attractive areas; i.e. closer to urban activity generators and/or environmental amenities). It must be noted, however, that SULD is a comparative-static simulation model that considers equilibrium of supply and demand in the medium to long-term. Excess supply of housing and associated possible decreases in real estate values in the short term are, thus, not considered – i.e. houses in abandoned areas are taken off the market in the medium to long term. For that reason, SULD is likely to provide inflated real estate value estimates in the short term and reliable real estate value estimates in the medium to long term. Further research is thus recommended to test whether that is the case, although, in truth, it is representing perfectly Hoekstra’s and Vakili-Zad’s (2011) ‘Spanish paradox’, discussed in the ‘New tools for a new problem’ section.

Furthermore, some additional caveats should be added to the work done. Urban amenities such as museums, theatres, monuments and historical city centres have been shown to influence residential development patterns (e.g. Brueckner et al., 1999), yet have not been considered as such in this study (i.e. providing direct utility, similar to the environmental amenities). As well, the quality of environmental amenities is considered exogenous, even though previous studies have shown that populations may have an impact on the quality of environmental amenities and subsequent residential development patterns (e.g. Dendrinos, 2000; Wu and Irwin, 2003). Finally, it is to be expected that changes in population may affect other land use types, such as the commercial, whose subsequent returns may affect, in turn, residential development patterns or the attractiveness of places (e.g. Bone et al., 2011; Wilson-Chavez and Rice, 2012).

Nevertheless, despite these model assumptions and limitations that should be overcome in following iterations of the model, SULD has been able to address most of the critiques previous models on urban shrinkage have faced in the literature. As a hedonic pricing simulation model, which requires only a limited amount of input data and uses housing prices as a predictor of demographic change, it has been able to model urban shrinkage in a spatially explicit way, including socio-economic aspects (such as economic characteristics of individual households of various types concerning decisions on where to live and how to travel) and also environmental amenities to determine the attractiveness of urban areas. By providing easily readable colour maps on land use, household density, real estate value and household types, with the possibility of testing several scenario alternatives, SULD could be a useful decision support tool both in shrinking cities and in growing cities, regardless of size. On one hand, based on evidence from SULD, politicians and practitioners are able to make informed planning decisions and to prepare and adapt to population change, and most importantly to the underrated population decline. On the other hand, it provides an opportunity to come closer to several urban planning ‘ideals’ such as compactness or a balanced real estate market. It could also potentiate a fair and equitable access to housing, urban and environmental amenities, the city centre and major access roads, by giving a likely picture of impending gentrification dynamics to cater for. Actually, Wolch et al. (2014) have precisely discussed that approaches on gentrification are lacking from economic and land-use studies. The outcomes of SULD, including household density, real estate values and household types, are, however, in tune with the indicators of gentrification presented by Kennedy and Leonard (2001): population density; social structure; average household income and real estate value, per neighbourhood.

Although the SULD analysis presented in this article was simplified for demonstration and comparison (for instance, population decline was considered homogenous across all household types), it would be possible to stratify population decline based on the uncovered cities’ trends. That is to say, the application of SULD could be refined according to changing aspects and planners’ needs. The SULD decision support tool could be especially helpful in the definition and justification of urban perimeters and priority areas, optimizing resources that range from land to infrastructures, and ultimately saving public and private investment as supply and demand even-up.

Recently, the literature on shrinkage (e.g. Haase et al., 2012; Hollander, 2011) has looked at it as a challenge rather than a woe, and has looked for ways to make it successful and, if possible, even beneficial to city areas. By modelling shrinkage with support tools such as SULD, we can understand the relationships between population decline and housing, social and economic change, and therefore be more equipped to propose smart decline policies that lead to the optimization of land use, sustainable resources and improved quality of life.

It must be made clear that the authors do not ambition to absolutely foresee the inescapable future. With that in mind, the authors believe that the possibility to analyze different scenarios based on available evidence is crucial to inform planning practice and match qualitative diagnoses. It allows to ‘play’ with different plausible realities before making irreversible (or not as easily reversible) strategic decisions.

Footnotes

Acknowledgments

The contribution and support of the Centre for Environmental and Marine Studies (CESAM) and the Department of Environment and Planning (DAO) at the University of Aveiro for facilitating this research is gratefully acknowledged.

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This article has been developed within the context of the Aqua-Add project (Deploying the added value of water in local and regional development; ), co-financed by the European Regional Development Fund (ERDF) and the 11 project partners through the INTERREG IVC program. This research was partially supported by the Portuguese Foundation for Science and Technology and the Human Potential Operational Programme/European Social Fund under Grant SFRH/BPD/86503/2012. In addition, this work was supported by European Funds through COMPETE and by National Funds through the Portuguese National Funding Agency for Science, Research and Technology (FCT) within the context of the projects PEst-C/MAR/LA0017/2013 and UID/AMB/50017/2013.

Notes

Miguel Saraiva is a Civil Engineer, with a doctoral degree on Urban Planning and Environment, who works as an Invited Assistant Professor and researcher at the Faculty of Arts of the University of Porto, - Department of Geography. As an integrated member of CEGOT (Centre of Studies on Geography and Spatial Planning), and of CITTA (Research Centre for Territory, Transports and Environment) of the Faculty of Engineering Porto, and former collaborator of the Universities of Lisbon and Aveiro, he has worked, published and participated in national and international projects in various areas which he has constantly entwined: crime-prevention, commerce, urban-morphology, accessibility, environmental amenity's optimization and most recently housing in a time of crisis. He is a member of COST ACTION TU1203 ‘Crime Prevention through Urban Design and Planning’.

Peter Roebeling is researcher with the Centre for Environmental and Marine Studies (CESAM) and the Department of Environment and Planning (DAO) at University of Aveiro, Portugal. He is an environmental and development economist who has led interdisciplinary research projects on resource use planning and management, and pathways for adoption of sustainable practices. Peter addresses the relationship between resource use, externalities, ecosystem state and function, and associated services and values in socio-ecological systems at the land-freshwater-sea interface. He develops approaches that support resource planning and management agents in the development, implementation and achievement of sustainable resource management plans.

Silvia Sousa is a Post-Doctoral Researcher from the Portuguese Foundation for Science and Technology and an integrated member of CITTA-Research Centre for Territory, Transports and Environment, Faculty of Engineering, University of Oporto. She has professional experience as team member, project manager, and co-author of several urban development plans, detailed local plans, sector studies, urban regeneration/design/architecture projects. She participated in national and international research projects in fields such as historic centers management, shrinking cities, housing policy, urban resilience, and evidence-based planning. She is a member of the Portuguese Association of Spatial Planners, Shrinking Cities International Research Network, and COST Action TU0803-Cities Regrowing Smaller.

Carla Teotónio is a PhD student at the University of Aveiro, Portugal, and a member of the Centre for Environmental and Marine Studies – CESAM. She is an economist who has participated in research projects on regional development and resource use planning and management. Her research interests focus on environmental economic topics, and on modelling approaches that support public policies pursuing sustainable development.

Anna Palla is Assistant Professor at the University of Genova in the field of Hydraulic Structures and Hydrology. Her research interests are mainly focused in the framework of sustainable urban drainage systems, in particular the research activity can be summarized in the following issues: storm water runoff quality, green roofs, permeable pavements, efficiency of treatment systems for storm water runoff and rainwater harvesting. The research activity is documented through more than 60 papers published on international/national journals and included in international/national conferences proceedings. She participates to several research projects supported by national and European funding.

Ilaria Gnecco is Assistant Professor of Hydraulic Structures and Hydrology at the University of Genova with research interests in various fields of urban hydrology and water resources management. Nowadays the research activity is mainly focused on sustainable water management in the urban environment and each topic is investigated based on both numerical approach and experimental activities carried out both in the laboratory and field. The research activity is documented through more than 80 papers published on international/national journals and included in international/national conferences proceedings. In the framework of the research activities she participated to several national and European research projects.

References

Allen

(2006) Welfare regimes, welfare systems and housing in Southern Europe. International Journal of Housing Policy 6: 251–277.

Balchin P (2013) Housing Policy in Europe. New York: Routledge.

Bateman

Brouwer

Ferrini

et al. (2011) Making benefit transfers work: Deriving and testing principles for value transfers for similar and dissimilar sites using a case study of the non-market benefits of water quality improvements across Europe. Environmental and Resource Economics 50: 365–387.

Beauregard

(2009) Urban population loss in historical perspective: United States, 1820-2000. Environment and Planning A 41: 514–528.

Bento R, Alves R, Vale D, et al. (2015) InLUT: Usos do solo e mobilidade – resultados. Presentation at the International Seminar InLUT – Integration of Land Use and Transport in Medium-Sized Cities. Lisbon. June 2015. Avaliable at: http://inlut.fa.utl.pt/media/wp-content/uploads/2015/07/Ap_Modulo3_v2.pdf.

Blanco

Alberti

Olshansky

et al. (2009) Shaken, shrinking, hot, impoverished and informal: Emerging research agendas in planning. Progress in Planning 72: 195–250.

Bone

Dragicevic

White

(2011) Modeling-in-the-middle: bridging the gap between agent-based modeling and multi-objective decision-making for land use change. International Journal of Geographical Information Science 25: 717–737.

Brooke A, Kendrick D, Meerhaus A, et al. (1998) GAMS release 2.5: a user’s guide. In: GAMS Development Corporation. Washington DC, USA: GAMS Development Corporation, p.262.

Brueckner

Thisse

J-F

Zenou

(1999) Why is central Paris rich and downtown Detroit poor? An amenity-based theory. European Economic Review 43: 91–107.

10.

Citta D'Imperia (2012) Piano Regolatore Generale della Citta di Imperia. Municipality of Imperia, Imperia, Italy. Available at: www.comune.imperia.it (accessed 28 June 2016).

11.

Comuni Italiani (2010) Information and Statistics on Municipalities, Provinces and Regions in Italy – 2010, Rome, Italy: Ministero Dell'Economia e Delle Finanze.

12.

Crompton

(2005) The impact of parks on property values: empirical evidence from the past two decades in the United States. Managing Leisure 10: 203–2018.

13.

Cunningham-Sabot E and Fol S (2009) Shrinking cities in France and Great Britain: A silent process. In: Pallagst K (ed.) Future of Shrinking Cities-Problems, Patterns and Strategies of Urban Transformation in a Global Context. California: University of Berkeley.

14.

Dendrinos

(2000) Land development and amenities: A predator-prey interaction. The Annals of Regional Science 34: 279–292.

15.

EEA (2009) Corine Land Cover (CLC2006) 100 m – version 12/2009. European Environment Agency, Copenhagen, Denmark. Available at: http://www.eea.europa.eu/data-and-maps/data/corine-land-cover-2006-clc2006-100-m-version-12-2009 (accessed 28 June 2016).

16.

Elsinga

Hoekstra

(2005) Homeownership and housing satisfaction. Journal of Housing and the Built Environment 20: 401–424.

17.

Glaeser

Gyourko

(2005) Urban decline and durable housing. Journal of Political Economy 113: 345–375.

18.

Haase

Kabisch

et al. (2012) Actors and factors in land-use simulation: The challenge of urban shrinkage. Environmental Modelling & Software 35: 92–103.

19.

Haase

Rink

Grossmann

et al. (2014a) Conceptualizing urban shrinkage. Environment and Planning A 46: 1519–1534.

20.

Haase

Rink

(2014b) Conceptualizing the nexus between urban shrinkage and ecosystem services. Landscape and Urban Planning 132: 159–169.

21.

Haase

Lautenbach

Seppelt

(2010) Modeling and simulating residential mobility in a shrinking city using an agent-based approach. Environmental Modelling & Software 25: 1225–1240.

22.

Hackworth

(2014) The limits to market-based strategies for addressing land abandonment in shrinking American cities. Progress in Planning 90: 1–37.

23.

Hoekstra

Vakili-Zad

(2011) High vacancy rates and rising house prices: The Spanish paradox. Tijdschrift Voor Economische En Sociale Geografie 102: 55–71.

24.

Hollander

(2011) Can a city successfully shrink? Evidence from survey data on neighborhood quality. Urban Affairs Review 47: 129–141.

25.

Hugh E (2013) The Great Portuguese hollowing out. Edward.Hugh.Blog.

26.

INE (2012) Main Indicators. Instituto Nacional de Estatistica (INE), Lisbon, Portugal. Available at: http://www.ine.pt/xportal/xmain?xpid=INE&xpgid=ine_main (accessed 28 June 2016).

27.

ISTAT (2015) Demografia in Cifre. Available at: http://demo.istat.it/ (accessed 28 June 2016).

28.

Keenan

Lowe

Spencer

(1999) Housing abandonment in inner cities - The politics of low demand for housing. Housing Studies 14: 703–716.

29.

Kennedy

Leonard

(2001) Dealing with Neighborhood Change: A Primer on Gentrification and Policy Choices, Washington, DC: Brookings Institution.

30.

Kroll

Haase

(2010) Does demographic change affect land use patterns? A case study from Germany. Land Use Policy 27: 726–737.

31.

Lee

Schmidt-Dengler

Felderer

et al. (2001) Austrian demography and housing demand: Is there a connection. Empirica 28: 259–276.

32.

Levin

Montagnoli

Wright

(2009) Demographic change and the housing market: Evidence from a comparison of Scotland and England. Urban Studies 46: 27–43.

33.

Lindh

Malmberg

(2008) Demography and housing demand - what can we learn from residential construction data? Journal of Population Economics 21: 521–539.

34.

Lutzenhiser

Netusil

(2001) The effect of open spaces on a home's sale price. Contemporary Economic Policy 19: 291–298.

35.

Mankiw

Weil

(1989) The baby boom, the baby bust, and the housing market. Regional Science and Urban Economics 19: 235–258.

36.

Mills

(1981) Growth, speculation and sprawl in a monocentric city. Journal of Urban Economics 10: 201–226.

37.

Nuissl

Haase

Lanzendorf

et al. (2009) Environmental impact assessment of urban land use transitions-A context-sensitive approach. Land Use Policy 26: 414–424.

38.

O'Sullivan A (2000) Urban Economics. New York: Irwin McGraw-Hill.

39.

OECD (2011) Doing better for families: Portugal. Available at: www.oecd.org/social/family/doingbetter (accessed 28 June 2016).

40.

Pontius

Jr Huffaker

Denman

(2004) Useful techniques of validation for spatially explicit land-change models. Ecological Modelling 179: 445–461.

41.

Roebeling

Fletcher

Hilbert

et al. (2007) Welfare gains from urbanizing landscapes in Great Barrier Reef catchments? A spatial environmental-economic modelling approach. Sustainable Development and Planning III, Volumes 1 and 2 102: 737–749.

42.

Roebeling

Rocha

Nunes

et al. (2014) Using the soil and water assessment tool to estimate dissolved inorganic nitrogen water pollution abatement cost functions in central Portugal. Journal of Environmental Quality 43: 168–176.

43.

Roebeling PC, Saraiva M, Palla A, et al. (2016) Assessing the socio-economic impacts of green/blue space, urban residential and road infrastructure projects in the Confluence (Lyon): a hedonic pricing simulation approach. Journal of Environmental Planning and Management, in press. DOI: 10.1080/09640568.2016.1162138.

44.

Schilling

Logan

(2008) Greening the rust belt a green infrastructure model for right sizing America's shrinking cities. Journal of the American Planning Association 74: 451–466.

45.

Schwarz

Haase

Seppelt

(2010) Omnipresent sprawl? A review of urban simulation models with respect to urban shrinkage. Environment and Planning B-Planning & Design 37: 265–283.

46.

SISTAN (2012) Annuario Statistico Regionale Liguria 2012. Sistema Statistico Nazianalo (SISTAN), Genova, Italy. Available at: www.sistan.it (accessed 28 June 2016).

47.

Sousa

(2010) Planning for Shrinking Cities in Portugal, Porto: University of Porto.

48.

Sousa

Pinho

(2013) Planning for shrinkage: Paradox or paradigm. European Planning Studies 23: 12–32.

49.

Tobler

(1970) A computer movie simulating urban growth in the Detroit region. Economic Geography 46: 234–240.

50.

Wilson-Chavez

Rice

(2012) Substance versus style: What is the role of new urban development in the restructuring of metropolitan retail? Applied Geography Conference 35: 384–392.

51.

Wolch

Byrne

Newell

(2014) Urban green space, public health, and environmental justice: The challenge of making cities ‘just green enough’. Landscape and Urban Planning 125: 234–244.

52.

C-T

Zhang

X-L

Cui

G-H

et al. (2013) Shrinkage and expansion in peri-urban China. In: Pallagst

TWK

Martinez-Fernandez

(eds) Shrinking Cities: International Perspectives and Policy Implications, New York: Routledge, pp. 164–185.

53.

Wu J and Irwin E (2003) Human-nature interactions and the spatial pattern of land use. In: AERE Workshop “Spatial Theory, Modeling, and Econometrics in Environmental and Resource Economics”. Madison, Wisconsin.

54.

(2001) Environmental amenities and the spatial pattern of urban sprawl. American Journal of Agricultural Economics 83: 691–697.

55.

(2006) Environmental amenities, urban sprawl, and community characteristics. Journal of Environmental Economics and Management 52: 527–547.

56.

Plantinga

(2003) The influence of public open space on urban spatial structure. Journal of Environmental Economics and Management 46: 288–309.