Suburbanisation and socio-spatial segregation in the Tallinn metropolitan area

Transcription

1 Workshop 1 - East European Housing & Urban Policy Suburbanisation and socio-spatial segregation in the Tallinn metropolitan area Anneli Kährik Anneli.Kahrik@ut.ee Tiit Tammaru Tiit.Tammaru@ut.ee Paper presented at the ENHR conference "Housing in an expanding Europe: theory, policy, participation and implementation" Ljubljana, Slovenia 2-51July 2006

2 Paper presented at the ENHR conference, Ljubljana, Slovenia, 2-5 July 2006 Suburbanisation and socio-spatial segregation in the Tallinn metropolitan area Anneli Kährik and Tiit Tammaru Institute of Geography, University of Tartu Abstract. Transformations in social stratification order and the growth of new settlements around major cities in Central and Eastern European transition countries lead to significant changes in the social and spatial distribution of population. However, our knowledge about this emerging sociospatial segregation is still poor. The aim of the current paper is to examine the characteristics of residents in new settlements in the Tallinn metropolitan area, Estonia, by focusing on the selectivity of surburbanisation with regard to the origin population as well as on its impact on the destination population. Using data from the Household Panel Survey (2004) and New Residential Areas Survey (2006), we apply logistic regression to study the socio-economic and demographic differences between three research populations comprising residents of Tallinn core city, inhabitants of old suburban settlements and those living in new residential areas. The main results of the study indicate that people younger than 35 who are well educated and earn considerably higher than average incomes have the highest odds of moving to new settlements. The odds of living in new residential areas are also higher for Estonians as compared to ethnic minorities. INTRODUCTION The fall of the Berlin Wall and demise of the former Soviet Union brought along major transformations in the political, economic and social structures in formerly centrally planned societies. One of the results of the reforms was an increase of (until now partly latent) social and spatial inequalities across countries in transition. Relevant studies point to changes in stratification order and the emergence of winners and losers in the transition process both in terms of population subgroups and regions. As regards the population, it has been argued that an individual s life chances during the transition period are shaped by previous social capital (e.g., being a member of the Communist Party) and human capital (e.g., age and education level) (Szelényi and Kostello, 1996). Most importantly, there is increasing evidence of a return to valuing education (Brainerd, 1998; Chase, 1998; Li, 2003; Titma et al., 1998) in the overall neoliberal atmosphere (Bockman and Eyal, 2002). Much of the discussion on social polarization is centred on the theory of market transition (Cao and Nee, 2000; Nee, 1989; Nee and Matthews, 1996), being 2

3 mainly based on studies in Russia (e.g., Brainerd, 1998; Gerber, 2000, 2003) and China (e.g., Bisogno and Chong, 2003; Cao and Nee, 2000; Nee, 1989; Nee and Matthews, 1996). Despite their size and importance, neither Russia nor China can be considered as being representative of countries that have undertaken profound reforms with at least three dimensions, i.e. the communist party no longer having monopoly power, the dominant part of the means of production being privately owned, and the market being the dominant coordinator of economic activities (Kornai, 1992, 2000). This calls for studies in countries that have carried out much more radical reforms and serve as role models of transition. This group of countries includes Estonia (Bunce, 1999; Korhonen, 2001). Previous research on spatial inequalities and migration has presented mixed results. For instance, studies that employ similar methodology indicate that in Hungary, counterurbanisation dominated in the 1990s (Brown and Schafft, 2002), while the prevailing migration trend in Estonia was urbanisation (Tammaru et al., 2004). Both countries went through radical economic reforms and there were no major differences in the population age structure (US Census Bureau, 2006). However, the differences in migration patterns are explained by the fact that unlike Hungary, Estonia has a significant minority population a characteristic compositional difference that exists between the Baltic countries and other Central and Eastern European countries. Ethnic minorities including the Russian-speaking population represented one third of the total population of Estonia and 50 per cent of the urban population in 1989 (Tammaru, 2001a). Also, 90 per cent of non-estonians lived in urban areas (mainly in the largest cities, including the capital city Tallinn), as compared to only 60 per cent of Estonians. The emigration of Russianspeaking population after Estonia restored its independence in 1991 freed considerable housing space in urban areas that favoured concentration of population within cities. This was different from the situation in other Central and Eastern European countries, where low availability of housing constrained migration (especially of young people) into the major cities (Sýkora and Čermák, 1998). While migration characterising a given settlement system diverged in countries in transition, suburbanisation prevailed in most major metropolitan areas (Krisjane, 2005; Kupiszewski et al., 1998; Ouředníček, 2006; Ravbar, 1997; Tammaru et al., 2004; Timár and Váradi, 2001). Although it has been argued that a close relationship exists between social stratification order and migration (Enyedi, 1998; Kostinskiy, 2001; Timár and Váradi, 2001), the relationship has been poorly studied due to the unavailability of relevant data (Ladányi and Szelényi, 1998). The aim of the current paper is to examine the characteristics of residents in new settlements in the Tallinn metropolitan area, Estonia, exploring both the selectivity of suburbanisation with regard to the origin population (residents of Tallinn) and the effect of suburbanisers on the destination (suburban) population. We are thus particularly interested in how the evolution of these new settlements influences the socio-spatial segregation in a metropolitan area of a country that has gone through a profound political, economic and social transformation. The paper is structured as follows. The next section outlines the results of previous studies on post-soviet suburbanisation and its socio-spatial impacts. Subsequently, we 3

4 focus on factors related to housing availability and affordability in Estonia by examining the underlying socio-economic circumstances and mortgage supply, housing and urban policy situation, as well as the activities of private developers in the evolution of new residential areas in the suburbs of Tallinn. We proceed by presenting research data and methods used. The study is based on two sample surveys, pooled for data analysis: the Household Panel Survey 2004 carried out by the Statistical Office of Estonia, providing data on the population of Tallinn and its suburbs, and the New Residential Areas Survey 2006 conducted among the residents of new settlements established in the Tallinn metropolitan area since The subsequent data analysis focuses on the socioeconomic and demographic differences between the three populations of residents studied those of Tallinn, old suburban settlements and new suburban areas and on changes in the composition of inhabitants in new residential areas that have occurred between the 1990s and the 2000s. The term suburbanisation is used in the current study to denote moving into these new residential areas. POST-SOVIET SUBURBANISATION AND ITS SOCIO-SPATIAL IMPACTS There seems to be a widespread agreement that socialist cities experienced less sociospatial segregation than their Western counterparts (Smith, 1994; Szelényi, 1996). Most scholars agree that the extent of suburbanisation was considerably smaller in socialist countries, and that resident groups who lived in the suburbs tended to have lower socioeconomic status as compared to the core city population (Ladanýi and Szelényi, 1998; Tammaru, 2001b). Ladanýi and Szelényi (1998) consider the high share of migrant workers in the villages surrounding Budapest as an important factor in explaining this tendency (the workers were unable to enter the housing market in the city due to administrative or other reasons), whereas Marksoo (2005) refers to the high level of agricultural engagement of the suburban population around Tallinn. Higher social status groups, however, tended to occupy the most desirable housing within the city or build quality condominiums for themselves in desirable inner urban areas (Gentile, 2003; Gentile and Tammaru, 2006; Kulu, 2003; Ladanýi and Szelényi, 1998). There are compelling reasons for anticipating changes in suburbanisation processes during the transition period. The post-socialist changes related to the re-emergence of market structures as a result of housing and land privatisation radically reduced the role of the public sector in influencing the socio-spatial outcomes of the suburbanisation process, while the concurrent differentiation of the financial opportunities of households created preconditions for selective migration (Sýkora, 1999a; Węcławowicz, 1998). New key actors in the system replaced the ruling stakeholders under socialism, e.g. private developers and commercial banks substituted central planners in the former system. Węcławowicz (1998) argues that the social positions acquired in the pre-transition period could be used as residual assets in the transition phase, and claims that the new market processes accelerate trends of differentiation already visible in the former socialist society. Other studies demonstrate the role of human capital characteristics (Kulu and 4

5 Billari, 2004; 2006). The suburbanisation process is initiated by a demand for singlefamily dwellings in environmentally attractive and accessible suburban areas along with the spread of car transport (Pichler-Milanovich, 2001). According to Szelényi (1996, p. 314), the expanding new bourgeoisie and the professional classes whose living standards are rapidly increasing are beginning to look for new residential locations outside cities. The increasing insecurity of down-town areas, environmental concerns and the desire of the better-off to escape from poverty areas are mentioned as factors influencing the decision of the wealthiest people to move to the urban fringe or areas beyond the city boundaries (Sailer-Fliege, 1999; Szelényi, 1996). In some of the countries with former central planning, migration from cities to rural areas in the suburban belt did not become a dominating trend until the 1980s (Ladanýi and Szelényi, 1998; Tammaru, 2001b). New residential development continued throughout the 1990s, but slowed down during the first half of the decade due to housing privatisation taking place at the time and the low purchasing power of households (Brown et al., 2005; Sailer-Fliege, 1999; Sýkora, 1999a). In fact, studies reveal that there was instead a trend of people with lower social status migrating from cities to cheaper existing suburban housing (Kulu and Billari, 2004, 2006; Ladanýi and Szelényi, 1998; Tammaru, 2005a). Migration to the suburbs accelerated over time, representing one of the main migration processes in post-socialist metropolitan areas (Ladanýi and Szelényi, 1998; Kok and Kovács, 1999; Pichler-Milanovich, 2001; Sýkora, 1999a; Szelényi, 1996; Tammaru et al., 2004). The growing wealth of households and improved access to mortgages were the most important factors that speeded up development of new residential areas in the late 1990s and 2000s (Ouředníček, 2006). It has been argued that only the wealthiest can afford a new single-family house in the suburbs (Kok and Kovács, 1999; Sýkora, 1999a, 1999b; Sýkora and Ouředníček, 2006; Szelényi, 1996; Tasan, 1999), and thus suburbanisation correlates with moving to new housing. This means that suburbanisation has an increasingly important impact on socio-spatial residential segregation in post-communist metropolitan regions, as housing is almost exclusively being built by and for the winners of the transformation process (Sailer-Fliege, 1999). While suburban areas were often inhabited by groups with lower social status during the communist period, new residential developments increasingly attract people with higher social status (Sýkora, 1999a, 1999b; Sýkora and Ouředníček, 2006). Consequently, the cities loose their best taxpayers, which has a negative impact on city management (Pichler-Milanovich, 2001). HOUSING AVAILABILITY AND AFFORDABILITY IN ESTONIA The financing, construction and allocation of housing in the socialist Estonia was subordinated to central planning. The majority of the housing stock was state-owned while private housing construction was modest (Kõre et al., 1996; Pihlak, 1994). Highrise housing estates became the most common building type during that period 5

6 multifamily panel houses made up 74 per cent of total housing construction in the 1980s (Kõre et al., 1996). The advantaged groups in the central housing allocation system were those belonging to the nomenclature, i.e. those holding higher level positions or having favoured occupations (Kährik, 2000; Org, 1989). In addition to social capital, human capital also played an important role, as people with a university degree enjoyed the best housing conditions (Gentile and Tammaru, 2006; Kulu, 2003). The pre-soviet housing stock was badly maintained and inhabited by people with lower social status. The existence of housing market segregation was evident by all major population characteristics, including age, education and ethnic origin (Gentile and Tammaru, 2006; Kulu, 2003; Org, 1989; Raitviir, 1990). During the first half of the 1990s, with the prevailing of overall liberal values, the housing and land reforms undertaken led to privatisation and restitution of most of the housing stock and land, creating premises for the development of private housing market (Kährik, 2000; Kährik et al., 2003). Sitting tenants in public housing mostly became homeowners, unless they occupied restituted buildings (Kährik, 2000). The voucher-type privatisation resulted in a high level of home-ownership; the share of private ownership reached 96 per cent of the total housing stock. The tenants who occupied the best housing segments gained the most they were the principal winners of the housing reform. Social segregation (e.g., by education) that was previously latent became visible. On the other hand, young generations had nothing to privatise and were forced to pay the market price in order to enter the housing market (Kährik, 2000). The first half of the 1990s the period of housing privatisation can be described as an initial stage of housing market development. The housing and land reforms were still in progress, and the availability of financial resources was restricted due to high mortgage rates (around 20 per cent until 1995). By the end of the 1990s, construction of new housing had dropped to its lowest level. New housing construction intensified in the 2000s, boosted by a combination of low mortgage rates (dropping from per cent in 2000 to 3 4 per cent in 2005), a favourable overall economic environment, increased wealth of people and a demand for modern houses (Raudsaar et al., 2006). Due to unmet demand for new dwellings, real estate prices also rapidly increased their growth rate in Estonia in 2005 was higher than in any other European Union member state. The majority of real estate transactions and construction activities are taking place in the capital city region (in 2000, 74 per cent of the total value of all transactions was ascribed to the Tallinn metropolitan area) (Kährik et al., 2003). This makes an analysis of sociospatial segregation in the Tallinn metropolitan area especially relevant. The spatial evolution of the metropolitan area is shaped by private developers. The public sector is not involved in construction activities, and even the detailed plans needed for residential development are not initiated and implemented by local authorities. Private developers buy up land for residential developments and initiate detailed plans, which typically involve changes in the purpose of land use (conversion of land from other uses to residential use). Housing construction is started after the plans are formally accepted by local authorities (Metspalu, 2005). The minimal role of local authorities in regulating housing construction stems from competition between municipalities for affluent 6

7 residents, and as a result, requirements for private developers are made as simple as possible. However, not all population groups have equal access to the new housing stock. Income disparities in Estonia are among the largest in the European Union, and only those belonging to the upper third of income deciles have earnings exceeding the national average income level (Table 1). Only one fifth of all households those belonging to the top two income deciles are considered eligible for mortgages needed to acquire a living space in new suburban residential areas, and this only if they have sufficient mortgage guarantees and resources to make a down-payment. Table 1. Average monthly disposable income by income deciles and overall average income per household member, in EEK (1 EUR = 15.6 EEK) (Income deciles are derived by dividing all households into ten equal groups according to the average income per household member) Income decile Average income Std. deviation 1 st nd rd th th th th th th th Average Source: Household Panel Survey 2004 The national government also plays a marginal role in the construction and housing allocation system in contemporary Estonia, in stark contrast to the previous centrally planned system (Kährik et al., 2003; Ruoppila, 2005). As stated in the National Housing Development Plan for , the role of the state is to create legal, institutional and subsidy-related conditions that would enable owners and tenants to act individually on the housing market (Ministry of Economic Affairs and Communications, 2002). State interventions are limited to creating a minimum safety net for households receiving social benefits. There are also some measures to help young families with children enter the housing market: state guarantees enable to reduce the down-payment for housing purchase and to extend the mortgage period. These measures, however, have some drawbacks as they further elevate real estate prices. This could be one of the reasons why the national government s intervention to the housing market is minimal. 7

8 DATA AND METHODS The Tallinn metropolitan area, the capital city region of Estonia, is located by the Baltic Sea. It has a total population of about half a million people (approximately four fifth of whom live in Tallinn core city), comprising 37 per cent of the population of Estonia. The population of Tallinn decreased by 16 per cent during the 1990s, mainly due to the emigration of Russian-speaking residents after Estonia regained its independence in 1991, as well as natural decrease and suburbanisation. The population in the nearby municipalities, however, increased by 10 per cent as a result of in-migration, mainly from Tallinn (Leetmaa, 2003, 2005). The New Residential Areas Survey 2006 indicates that an estimated individuals have migrated to new suburban dwellings built since New suburban settlements are located very close to Tallinn core city, being stretched out along the coastal areas to the east and west of the city (Figure 1). This shows that the new suburbanisers opt for the most accessible and environmentally attractive locations. The new settlements have mostly emerged as extensions to the already existing suburban satellite towns, rural settlements or summer cottage areas. A few have been built further away from existing settlements, on former agricultural land or replacing forests. New settlements Old settlements Figure 1. Distribution of old (pre-1991) and new (post-1991) settlements in the Tallinn metropolitan area. Source: New Residential Areas Survey 2006 The data used in the current study are derived from the Household Panel Survey, conducted by the Statistical Office of Estonia in 2004, and the New Residential Areas Survey (2006). The aim of the Household Panel Survey was to provide comprehensive data on the financial behaviour of the population of Estonia, from which we extracted data relating to people living in Tallinn or its suburbs. Analysis of the distribution of population by the age of dwelling indicates that most of the housing units, especially in Tallinn, were built during the period of Soviet mass housing construction spanning the 8

9 1960s through the 1980s (Figure 2). The drop in housing construction during the 1990s and early 2000s has been significant. As a result, the number of people living in dwellings built since 1991 is very small in the nationally representative sample of the Household Panel Survey, and this group has therefore been excluded from the database. In order to explore the population living in those new settlements, we conducted a New Residential Areas Survey in the Tallinn metropolitan area in early 2006, using a classification similar to that employed in the Household Panel Survey The current study is based on a pooled database derived from these two surveys. Figure 2. Distribution of people by construction year of dwelling. Source: Household Panel Survey 2004 Our dataset includes 1684 people, of whom 876 live in Tallinn, 244 live in old (pre-1991) suburban residential areas and 564 live in new (post-1991) residential areas. As expected, there are significant differences in the housing conditions between these three populations. Virtually all the people residing in new residential areas own their living space, while 20 per cent of the inhabitants of Tallinn and 9 per cent in old suburban settlements are tenants (Table 2). The differences in dwelling types and living conditions are more pronounced. The share of people living in single-family houses is highest among residents of new settlements less than a third of them live in multifamily houses (Figures 3, 4). In contrast, the latter figure stands at 89 per cent for Tallinn and 51 per cent for old residential areas in the suburbs. However, the share of people living in multifamily houses in new settlements has increased over time. Houses in new residential areas are also the most spacious and with the largest number of rooms. As regards infrastructure in new suburban residential areas, sewage systems are better but central heating is worse than in the old suburban areas. Finally, car ownership is most widespread in new settlements: 95 per cent of the households own at least one car and half of the households have two cars. This partly reflects the differences in housing composition between new residential areas (with the lowest share of multifamily houses and a less concentrated settlement pattern) on the one hand and Tallinn and old suburban residential areas on the other. 9

10 Table 2. Living conditions of the research populations (values given as percentages). Tallinn Old suburban settlements New suburban settlements Dwelling ownership Owner status Tenant Dwelling Single-family type Semi-detached Multifamily Number of rooms or more Living space (m 2 ) Less than or more Sewage system Local Central Heating system Local Central Car in household No car or more cars Sources: Household Panel Survey 2004, New Residential Areas Survey

11 Figure 3. New single-family housing estate in Järveküla. Source: authors photo Figure 4. New apartment housing estate in Järveküla. Source: authors photo In order to analyse the differences between the three studied research populations people living in Tallinn, in old (pre-1991) or in new (post-1991) suburban settlements we have used binary logistic regression. The regression model can be formalised as follows: p(y i = 1) K log = α + β k X i,k, p(y i = 0) k=1 where 11

12 p(y i = 1) is the probability that an individual i = 1, I lives in an old suburban settlement built before 1991 in Models 1 and 2, in a new suburban settlement built since 1991 in Models 3 through 6, or in a new settlement built since 2000 in Models 7 and 8; p(y i = 0) is the probability that an individual i = 1, I lives in Tallinn in Models 1 though 4, in an old suburban settlement built before 1991 in Models 5 and 6, or in a new settlement built between 1991 and 2000 in Models 7 and 8; α is a constant; X i,k is the value of variable k for individual i; and β k is a parameter describing the impact of variable k, with K variables. RESULTS First, we compared populations living in pre-1991 dwellings in Tallinn and in the suburbs (Table 3, Model 1). First of all, this analysis highlights the differences between origin and destination populations of suburbanisers. In addition, the comparison sheds some light on the suburbanisation process during the Soviet time. It becomes evident that there are no major differences in the basic demographic characteristics of the two populations, i.e. in gender, age, household and family composition. In contrast, ethnic and educational differences are notable. Tallinn was the most important destination of immigration and internal migration of immigrants in Estonia during the Soviet period. Therefore, ethnic minorities residing in the Tallinn metropolitan area are more likely to live in the capital city than in the suburbs. Likewise, the educational composition differs significantly: the odds of having acquired only primary education are much higher in the suburbs as compared to Tallinn. This is expected, as it reflects the legacies of the Soviet time agriculture- and industry-related suburbanisation, where people with lower social status played a significant role (Ladányi and Szelényi, 1998; Tammaru, 2001b). There are no differences in income; disaggregation of income into more detailed categories yields similar results. The majority of the housing stock in Tallinn was built during the period of mass construction of standardised multifamily housing, whereas in the suburbs, singlefamily housing was more common. Therefore, residents in the suburbs are more likely to live in detached houses. As expected, people living in the suburbs are more likely to own a car, but this conclusion cannot be generalised from the research sample to the total population, as the results are not statistically significant. In contrast to car ownership, the inhabitants of Tallinn have better access to the Internet. Table 3. Comparison of populations living in Tallinn (0) and in old residential areas in the suburbs (1) (odds ratios). Model 1 Model 2 Gender (reference = Male) Female Age (reference = <35)

13 * * > Household size (reference = 1 2 members) >2 members Children (reference = No) Yes Ethnic origin (reference = Minority) Estonian *** *** Level of education (reference = Secondary) Primary *** *** Tertiary *** Income per household member (reference = Low) High Dwelling type (reference = Multifamily) Single-family *** *** Tenure status (reference = Tenant) Owner *** *** Internet at home (reference = No) Yes ** ** Car in household (reference = No) Yes Interactions Single-family house and Secondary education Single-family house and Tertiary education *** Single-family house and Car in household * -2 LL * Significant at p < 0.1 ** Significant at p < 0.05 *** Significant at p <

14 We also looked for interactions between individual characteristics (Table 3, Model 2). Statistically significant results were obtained for the correlation of living in a singlefamily house with education level as well as with car ownership. People with primary education are not only more likely to live in the suburbs, but they also have higher odds of living in single-family houses. This again reflects the effects of the Soviet legacy on housing outcomes. First, living in multifamily houses with full modern facilities was desirable in countries under central planning (Rykiel, 1984), and people with a university degree were more likely to end up in those houses (Gentile and Tammaru, 2006; Kulu, 2003). Second, the priority status of agriculture in late Soviet Estonia and the tradition of constructing single-family houses by oneself had the effect of redistributing people with primary education to these types of dwellings in rural areas. As late as in the 1980s, almost half of the suburban workforce was still engaged in agriculture in the Tallinn metropolitan area (Kulu and Billari, 2004). People living in single-family houses in the suburbs also have the highest odds of owning a car. This is probably related to the relatively poor public transport system in the Tallinn metropolitan area, which makes it almost obligatory to have a car when living outside of satellite towns and other major settlements in the suburbs. We next performed a comparison of people living in Tallinn with the population of new residential areas built since The results of this analysis indicate much larger differences (Table 4, Model 3). First, we assessed the selectivity of out-migration from Tallinn to the new suburbs. The results are interesting in many regards. The very young age of people living in new settlements comes as a surprise. People who live in expensive newly built houses in new residential areas are most likely to be younger than 35. Furthermore, families with children do not have a significantly higher probability of moving from Tallinn to new suburban settlements, which is also unexpected. Thus, even though bivariate analysis indicates that the share of families with children is highest among residents of new suburban settlements when compared to Tallinn and old suburban settlements, having a child does not have an independent impact on suburbanisation. Table 4. Comparison of populations living in Tallinn (0) and in new residential areas in the suburbs (1) (odds ratios). Model 3 Model 4 Gender (reference = Male) Female *** *** Age (reference = <35) *** *** *** *** > *** *** Household size (reference = 1 2 members) >2 members ** Children (reference = No) Yes Ethnic origin (reference = minority) Estonian *** *** 14

15 Level of education (reference = Secondary) Primary *** *** Tertiary Income per household member (reference = Low) High *** *** Dwelling type (reference = Multifamily) Single-family *** *** Tenure status (reference = Tenant) Owner *** *** Internet at home (reference = No) Yes *** *** Car in household (reference = No) Yes *** *** Interactions Single-family house and Estonian origin *** Single-family house and >2 members per household ** Single-family house and Car in household *** -2 LL * Significant at p < 0.1 ** Significant at p < 0.05 *** Significant at p < 0.01 The results of our study are especially striking when one considers that sociologists refer to those born in the 1960s as the generation of winners in the post-soviet transition period, as they entered the labour market at the beginning of that period and are therefore regarded as being the most successful in securing a good employment (Titma et al., 2002). This generation also had an advantage in the housing career over younger people, since they were able to either privatise their housing free of charge or buy it at times when dwellings were cheap. We show that people born in the 1960s, falling into the age group of in our study, have three times lower odds of living in new residential areas than young people. As the 1960s generation has also reached the age of forming a family, we had strong reasons to believe that they would have the biggest financial resources and motives to lead the suburbanisation process by taking a next step in their housing career, i.e. moving out from Soviet-style multifamily houses of the capital city to new suburban detached houses. The surprising results of our analysis prompted us to run an additional model with a dummy variable, using the generation of winners as a reference group, but this did not change the outcome. Analysis of the interaction between the generation of winners and high income also produced similar results. However, this generation did suburbanise in the 1990s when they were younger than 35. This means that moving to new residential areas in the suburbs is related to being young rather than being born in the 1960s. Unlike the age structure, the social composition of suburbanisers is in accordance with expectations people living in new residential areas are well educated and earn a higher than average income per household member. Conversely, people with primary education have significantly lower odds of living in new houses in the suburbs. Differences in income are most pronounced: suburbanisers have nine times higher odds of belonging to 15

16 the high income group than residents of Tallinn. This completes the picture of the selectivity of the suburbanisation process: suburbanisers are well-educated young people with high income. The study also reveals that inhabitants of new suburban settlements are more likely to live in owner-occupied and single-family houses, and that they have a higher probability of owning a car and having an Internet connection at home. Finally, we analysed interactions between variables. Three interactions proved to be statistically significant (Table 4, Model 4). First, there is a correlation between ethnic origin and dwelling type. Although Estonians have three times higher odds of living in new settlements, the minorities, when they decide to move to the suburbs, have elevated odds of ending up in a single-family house. Second, interactions reveal that bigger households with at least three members have a considerably higher probability of moving into single-family houses than smaller households. People living in single-family houses in new residential areas are most likely to be car owners. Next, we assessed the impact of new residential development on the evolution of population structure in the suburbs, by comparing suburbanisers with people living in pre-transition settlements (Table 5, Model 5). The differences are marked and quite similar to the previous analysis. New residents are younger and better educated, their households are smaller and income per household member is much higher. People living in new residential areas are more likely to own a car and have an Internet connection at home. There is a weak indication of changes in the ethnic composition of the suburban population, since Estonians have a lower probability of suburbanising in our sample, but the results are not statistically significant and cannot be generalised from the studied sample to the total population. In this model, we also have one extra variable settlement type to measure the sprawl dimension in the suburbanisation process. The Soviet policy favoured very compact settlements (Marksoo, 2005; Tammaru, 2001b). This changed during the transition period, as the range of factors determining location choices for new residential areas diversified, including factors such as distance from Tallinn, availability and price of land, availability and development costs of infrastructure, and people s preferences. Due to a combination of all these factors, new residential developments were less likely to occur in areas of highest population concentration, i.e. in urban municipalities, shifting to previously rural areas instead. Some interesting interactions also emerged from this comparison (Table 5, Model 6). The income of people living in single-family houses in new settlements is the highest, while people with primary education have the lowest odds of living in detached houses. These results characterise a change in the social composition of suburbanisers. While primary education increased the probability of moving to a suburban single-family house in Soviet Estonia, the opposite is true today. 16

17 Table 5. Comparison of populations living in old (0) and new (1) residential areas in the suburbs (odds ratios). Model 5 Model 6 Gender (reference = Male) Female *** *** Age (reference = <35) *** *** *** *** > *** *** Household size (reference = 1 2 members) >2 members * Children (reference = No) Yes Ethnic origin (reference = minority) Estonian Level of education (reference = Secondary) Primary *** *** Tertiary ** Income per household member (reference = Low) High *** *** Dwelling type (reference = Multifamily) Single-family *** Tenure status (reference = Tenant) Owner *** *** Internet at home (reference = No) Yes *** *** Car in household (reference = No) Yes *** *** Settlement type (reference = Urban) Rural *** *** Interactions Single-family house and High income ** Single-family house and Primary education ** Single-family house and Tertiary education LL * Significant at p < 0.1 ** Significant at p < 0.05 *** Significant at p < 0.01 As a final step, we explored whether there are any changes taking place over time in the suburbanisation process during the transition period. To our knowledge, no such studies have yet been published; previous studies on Estonia (Kulu and Billari, 2004, 2006; Leetmaa, 2003a, 2005; Tammaru, 2005a) lack a temporal dimension. In order to start filling this gap, we have analysed changes in the development of new residential areas by distinguishing two periods, 1991 to 2000 (the 1990s) and 2001 to early 2006 (the 2000s). There are several factors expected to modify the structure of movers during these two decades. 17

18 The first set of factors is related to the housing market. As mentioned earlier, the 1990s can be characterised as a period when the housing market started functioning (the housing was privatised by the mid-1990s), but at the same time, interest rates of mortgages remained high and housing development was modest. This limited people s opportunities to migrate to new settlements. We therefore assumed that only the wealthiest people (the so-called nouveaux riches ) not in need of mortgages had enough financial resources to construct housing at that time. The 2000s expanded the structure of opportunities, as banks liberalised their mortgage policy and private developers started providing new and diverse (detached, semi-detached, multifamily) housing in the suburbs. We thus expected an improvement in access to new suburban housing for less advantaged population and income groups. Secondly, differences in population structure may be related to somewhat different mechanisms leading to social stratification during these two decades. The wealth of the nouveaux riches who emerged in the very beginning of the 1990s was often based on illegal or semilegal business activities. Suburbanisation to new settlements became a matter of prestige for those people and for the business elite in general. In the 2000s, less wealthy but still affluent people were able to join the elite in new residential areas by taking advantage of favourable long-term mortgages. Following this logic, we assumed that people belonging to the highest income groups would have higher odds of suburbanising in the 1990s as compared to the 2000s. However, the analysis shows that this is not true: people who suburbanised in the 2000s earn higher incomes (Table 6, Model 7). One explanation for this finding is that the nouveaux riches of the 1990s may have lost their high socio-economic status by the 2000s. Unfortunately, there are no data on the income of migrants at the time of residential change. Table 6. Comparison of populations living in new residential areas built in the 1990s (0) and in the 2000s (1) (odds ratios). Model 7 Model 8 Gender (reference = Male) Female Age (reference = <35) ** > Household size (reference = 1 2 members) >2 members Children (reference = No) Yes *** Ethnic origin (reference = minority) Estonian Level of education (reference = Secondary) Primary * Tertiary Income per household member (reference = Low) High ***

19 Dwelling type (reference = Multifamily) Single-family *** Tenure status (reference = Tenant) Owner Internet at home (reference = No) Yes Car in household (reference = No) Yes Settlement type (reference = Urban) Rural Interactions Single-family house and Estonian origin ** Single-family house and High income ** Children and Age ** Children and Age > LL * Significant at p < 0.1 ** Significant at p < 0.05 *** Significant at p < 0.01 As regards the level of education, we do not find any statistically significant differences between the two population groups, although people with primary education appear less likely to suburbanise in the 2000s than in the 1990s in our sample. Likewise, there is no statistically significant difference in age composition at the time of moving. Even though bivariate analysis reveals that new migrants are of younger age (Figure 5), this change in the age structure is likely the result of the impact of other population characteristics on suburbanisation. The probability of having children when suburbanising is significantly higher in the 2000s than it was in the 1990s. This means that having a child rather than being young explains the increased share of younger suburbanisers. Thus, family motives seem to have gained importance in the suburbanisation process over time. Another significant difference between the two decades is a change in the structure of housing types while new suburban housing of the 1990s comprised mostly detached dwellings, the new millennium introduced a mixture of housing types. 19

20 25 20 Share, % < s 2000s Figure 5. Age at the time of moving to a new residential area. Source: New Residential Areas Survey 2006 The analysis was completed by assessing interactions between individual variables (Table 6, Model 8). It becomes evident that ethnic minorities have an elevated probability of moving into single-family houses in the 2000s as compared to the 1990s. As detached houses have traditionally been favoured by Estonians rather than by minorities, these results indicate that ethnic differences in housing outcomes are beginning to level off. People who move into single-family houses in the 2000s also enjoy higher incomes. Real estate prices have increased dramatically in the Tallinn metropolitan area since As a result, single-family houses are becoming increasingly exclusive, and suburbanisers and developers both opt for alternative dwelling types. The third significant interaction detected in the study indicates that middle-aged people with children have much higher odds of moving to new residential areas in the 2000s than they had in the 1990s. Thus, while the overall picture of suburbanisers highlights their young age, a shift towards a more expected pattern of suburbanisation can be seen among middle-aged people having families with children. SUMMARY AND DISCUSSION The results of previous studies on post-socialist cities indicate that suburbanisation processes accelerate over time (Ouředníček, 2006). This is also true for the Tallinn metropolitan area. As expected, people living in new suburban settlements built since 1991 enjoy better living conditions than the inhabitants of Tallinn and old suburban settlements. However, the most profound conclusions of our study are related to the 20

21 socio-economic status and demographic characteristics of suburbanisers. Our analysis confirms that people who belong to the highest income deciles and are well educated take advantages of new housing development, in contrast to the pre-transition (Ladányi and Szelényi, 1998) and early transition (Kulu and Billari, 2004, 2006) periods, when an opposite trend could be observed. The most affluent people have considerably higher odds of moving to new suburban settlements than other income groups, while people with primary education are least likely to suburbanise. As a result of improved mortgage conditions and more extensive new housing construction, housing in new suburban areas would be expected to be available to more diverse income groups in the 2000s as compared to the 1990s, but our findings did not substantiate this in-migrants in the 2000s tend to earn even higher incomes than earlier movers. With regard to demographic characteristics, the most significant and unexpected result of the study relates to the very young age of people living in new suburban settlements as the most expensive segment of the housing stock the odds of living in those dwellings are highest for people younger than 35 who are about to start their housing career. The results come as a surprise because young people were not able to take advantage of the housing privatisation process that favoured sitting tenants. Our findings indicate that this disadvantage may rather have served as an advantage in contemporary Estonia. Supported by the economic boom that has followed radical reforms in the 1990s and becoming a member state of the European Union in 2004, as well as increased personal wealth and favourable mortgages, young people take risks and move directly to the most desirable housing stock despite the fact that very quickly escalating real estate prices make entering the housing market increasingly difficult. One drawback of this is a very high income-to-mortgage ratio (cf. Neuteboom, 2004). Having a child does not increase the odds of leaving Tallinn for new suburban settlements, but it is nevertheless a significant factor in the suburbanisation process in the 2000s as compared to the 1990s. There are also significant differences in suburbanisation based on ethnic origin. In comparison to Estonians, ethnic minorities have a significantly lower probability of moving from Tallinn to new suburban residential areas. The findings of our research rise interesting issues for discussion and future research. The first important issue is related to the impact of suburbanisation on fertility. Studies indicate that people who move from urban to rural destinations have the lowest union dissolution risk as compared to all other groups of movers and non-movers (Boyle et al., 2006), and that migrants adapt to destination fertility levels which are lowest in major cities and highest in rural areas (Kulu, 2005, 2006). This would suggest a positive impact of suburbanisation on fertility. However, this issue deserves closer attention for two reasons: suburbanisers in the Tallinn metropolitan area are very young, and new dwellings in the suburbs are very expensive. On the one hand, young people take advantage of the economic boom coupled with low mortgage rates. On the other, it may have an important effect on family size due to the significant long-term financial responsibilities these people take, especially in the light of the overall birth postponement in Estonia (Katus, 2000), similarly to other countries in transition (Vikat, 2005). Although the government pays generous parental leave compensations during a child s first year of life, it is also evident that suburbanisation ties these young people very 21

22 strongly to the labour market. This calls for research in interdependencies of fertility and suburbanisation in countries in transition. Suburbanisation in Central and Eastern Europe is often compared to that in Western Europe. Although the process in those two regions was similar during the interwar period (Tammaru, 2001b), the trends diverged after World War II, as suburbanisation continued in Western Europe but was halted in the East. Therefore, one of the central research questions is whether the contemporary suburbanisation process is similar to what happened in the West since World War II, or whether there are also some diverging trends (Sailer-Fliege, 1999). Our current study presents mixed results. First of all, it seems that the social structure of suburbanisers has become closer to that of Western countries during the transition period (cf. Clark and Dieleman, 1996; Johnston, 1974; Mieszkowski and Mills, 1993). While primary education elevated the probability of suburbanisation during the Soviet and early transition periods, it is well-educated people who move to new suburban settlements today. However, our findings also support the thesis that suburbanisation trends in countries in transition bear similar traits with those in Western neo-liberal countries where the time-space trajectories and lifeworlds of wealthy suburban residents are quite distinctive, and interaction with other social status groups is modest; this has a negative impact on the overall social cohesion, and the life chances of the poor in particular (Atkinson, 2006). The economic boom in Estonia helps overcome such problems at the moment, but in the long-term perspective, they might emerge due to the high social selectivity of the suburbanisation process. As residents of new suburban settlements are also the top tax payers, suburbanisation influences the tax base, provision of public services and governance both in the origin and destination municipalities. People leaving Tallinn continue to use the city s infrastructure (e.g. roads) and public services (Tammaru, 2005b). Suburban municipalities support the in-migration of new affluent residents, but often lack resources or are reluctant to provide costly public services to the quickly growing population. All these emerging problem areas call for much more sophisticated planning activities and cooperation between the origin and destination municipalities. They also necessitate more detailed studies on socio-spatial segregation in the suburbs and the increasing interdependencies between central cities and suburban areas. Finally, there is a significant ethnic dimension to the suburbanisation process in the Tallinn metropolitan area. This dimension distinguishes the Baltic countries, including Estonia, from other Central and Eastern European countries, as the former have a sizeable Russian-speaking minority population that resides mainly in the major cities (Tammaru, 2001a). Since minorities have a lower probability of moving from Tallinn to the suburbs than Estonians, they continue to live predominantly in Tallinn, the main destination of internal migration of immigrants in Soviet Estonia (Marksoo, 2005). However, the latter trend has begun to change during the transition period, and both the suburbanisation and counterurbanisation tendencies of minorities, although modest in scale, are now evident (cf. Leetmaa, 2003; Tammur, 2003). Furthermore, our study also reveals that when minorities do move to new residential areas, they have higher odds of settling in a singlefamily house. This speaks of a development towards a convergence in housing outcomes 22