The role of cultural heritage in attracting skilled individuals

2.1 Agglomeration economies and urban amenities

Agglomeration economies and the related concept of urban amenities are commonly pointed out as a key drivers of regional growth (Glaeser et al. 2001; Duranton and Puga 2004; Rosenthal and Strange 2008). In theory, there are various channels through which agglomeration may influence the location decisions of individuals and firms. Studies that focus on the later tend to emphasize that agglomeration economies benefiting colocated firms (Porter 2000; Parr 2002). In this view, agglomeration is closely associated with localization or ‘Marshallian’ economies that arise from colocation between firms within the same industry. Firms that colocate can benefit from sharing (e.g., of fixed costs, inputs and customers), matching (e.g., of workers with relevant skills), and learning due to knowledge accumulation and spatial spillovers (Duranton and Puga 2004). This view dates back to Marshall (1920), who argued that colocation is able to spur economic activity, which increases growth both in the sector and the region as a whole. The type of agglomeration economies that arise as a result of knowledge spillovers is, however, somewhat different compared to the ‘Marshallian’ in that it refers to a pure externality that is bounded in space (Jaffe et al. 1993; Breschi and Lissoni 2009). In the view that agglomeration is a result of knowledge interconnections, others have also pointed at the importance of industrial diversity as a source of knowledge spillovers and clustering, commonly referred to as Jacobs’s externalities (Glaeser et al. 1992).

Another strand of the literature focuses on the location decisions of individuals, rather than firms and use the concept of urban amenities to signify the type of agglomeration externalities that make cities desirable places to work and live in (cf. Andersson 1985; Quigley 1998; Glaeser et al. 2001; Florida 2002; Clark et al. 2002). This literature is highly influenced by the work of Glaeser et al. (2001) and Florida (2002) and their idea of urban amenities as significant attractors of skilled and creative workers. Studies that follow along the line of Florida (2002) tend to argue for an understanding of agglomeration based on individuals and their knowledge characteristics, rather than the characteristics of firms. Accordingly, the so-called creative workers have locational preferences for different types of amenities or quality-of-life attributes, e.g., esthetics, culture, lifestyle, consumption possibilities and natural amenities. Although this theory has gained widespread popularity among policy makers, it has also been questioned on various grounds including its originality (Rausch and Negrey 2006).

2.2 Natural and cultural amenities

Amenities have long been seen as a central part in theories on urban growth and in explaining individual location patterns. Since the work of Yang and Fujita (1983), there has emerged a growing literature that focuses on location patterns and the formation of the urban spatial structure resulting from interactions between natural and cultural amenities and agglomeration (Wu and Plantinga 2003; Irwin and Bockstael 2001; Chen and Rosenthal 2008; Wang and Wu 2010). Brueckner et al. (1999) present a theoretical framework on the role played by different types of amenities in explaining location patters of high income earners. They identify three categories: (1) natural amenities (climate, topographical features, rivers, coastline, etc.), (2) historical amenities (cultural infrastructure from past eras in terms of monuments, historical buildings and architecture) and (3) urban amenities (restaurants, theaters, shops). While the first two categories are seen as largely exogenous, urban amenities are seen as endogenous and dependent on current economic conditions and income levels (Brueckner et al. 1999, p. 4). In their model, location by income depends on the spatial distribution of exogenous amenities such as water access, topography or historical amenities. Their analysis shows that when the urban center has a strong advantage over the suburbs in exogenous amenities, and when the valuation of such amenities rises with income, high-income households are likely to cluster in central locations. They also conclude that this effect occurs regardless of the presence of endogenous amenities, e.g., restaurants, theatres and other urban amenities that depend on current economic conditions. What follows from this outline is that there exist a number of potentially important locational characteristics that are able to attract high-skilled labor, not only in terms of urban and natural qualities, but also in terms of tangible and intangible cultural qualities. Andersson (2006) refers to physical cultural capital as a form of infrastructure that is able to ensure the durability and sustainability of cities. Throsby (2001) interprets cultural heritages as local public goods that generate positive spillover effects through their esthetic, authentic and historical values, which are intrinsic to their location.

An important perspective set forth in prior research is that amenities operate at different spatial ranges or limits (Cho et al. 2008). Therefore, while the benefits of some amenities attenuate slowly in distance and might be poor predictors of location choices within a region, e.g., weather conditions, other amenities are very much placed based (e.g., built amenities and landscape features) and might be critical predictors of locations choices within a region (Nilsson 2014). It has, for example, been shown that open spaces have an influence on house and land values over small geographic areas and that the effects diminish within relatively short distances (Geoghegan et al. 1997). This suggest that the contribution of amenities in attracting skilled labor may not be reflected accurately when modeled at the more aggregated regional level, which has been the dominating approach thus far. In order to account for heterogeneity among different types of place-based amenities and the local nature of their spillover effects, this study employs a spatially disaggregated approach and distinguish between three types of amenities: (1) amenities that are primarily natural, such as landscape and seascape; (2) amenities that are primarily urban and related to regional size and agglomeration; and (3) amenities which are primarily cultural, such as the local supply of built heritages (monuments, buildings and sites) and cultural environments.

Another empirical regularity that has been identified is that amenities are valued differently depending on local and regional contexts, e.g., they tend to be driven by regional size and density (Chi and Marcouiller 2011; McGranahan and Wojan 2007). It has been found that marginal valuations of certain types of natural amenities (e.g., open spaces) are high in areas characterized by high population density and low or insignificant in areas where undeveloped lands are abundant (Bates and Santerre 2001). The implication is that amenity-led growth might not be a one-size-fit-all solution and that areas may have varying potential to exploit amenities depending on their degree of urbanization. Much of the research that focus on natural amenities and regional growth base their evidence on the type of land use patterns found in highly urbanized cities in the USA or Europe (Mieszkowski and Mills 1993; Andersson et al. 2004). Under these conditions, individuals regard high density as a dis-amenity and are willing to trade accessibility to the center against access to open space and natural amenities. As noted by many, the type of land use patterns found in most European cities are reversed and high-income households predominantly occupy centrally located expensive land (Brueckner et al. 1999). Thus, it is not evident that the same type of spatial patterns in amenity assessment exists in less urbanized regions that have relatively low density levels.

Although the local supply of natural amenities and regional attractiveness has received a lot of attention in the literature and there are many recent studies motivated by explaining their relation, there are still relatively few studies that focus on the role played by built heritages in attracting high-skilled labor. Falck et al. (2011) focus on Baroque opera houses in Germany and find that well-educated workers prefer to live geographically close to such cultural amenities. They apply an instrumental variable approach and find that it is the local level of high-human-capital employees who value proximity to a Baroque opera house and that this in turn has a positive effect on regional growth. Moro et al. (2013) analyze a related question, e.g., if housing markets reflect cultural heritage values. They estimate hedonic price equations to establish whether distance to, and density of, cultural heritage is capitalized into housing prices in Greater Dublin, Ireland. They find that some types of cultural heritage sites, such as historic buildings and memorials, provide positive spillovers to property prices and interpret these price premiums as capturing esthetic beauty. These findings support the hypotheses that high-human-capital individuals value the availability of cultural amenities (Moretti 2004); they also support the idea of cultural heritage as an important locational attribute that can increase the attractivity of an area. However, as discussed previously, most of the studies with this focus tend to be geographical delimited to the study of one specific region or to specific types of built heritages or environments. One limitation is that their results are not generalizable.³ This paper contributes to the literature by addressing the role played by listed buildings and cultural environments in attracting high-skilled individuals across the whole Swedish geography and over time. This is made possible by having access to unique spatially disaggregated data that enable us to connect the local growth of high-skilled labor with the local supply of built heritages (e.g., historical buildings, monuments and sites), a perspective of amenity-led regional growth that has received little attention in the literature.⁴ Although we anticipate a positive relationship between the local supply of cultural heritages and the growth of high-skilled individuals, it is also important to note that there could also be negative consequences of proximity to these heritages. For example, if they attract many tourists, congestion could prevent skilled people from locating in these areas. The expected outcome can thus be ambiguous. Besides location-specific amenities discussed above, there exist several additional factors that may influence location decisions of high-skilled individuals including the presence of institutions for higher education (Faggian et al. 2007; Faggian and McCann 2009), labor market characteristics and the demand for high-skilled labor (Andersson and Mantsinen 1980; Acs and Audrestch 1990).⁵

3.1 Cultural and natural amenities

Data from three geocoded databases are used to create measures that reflect educational infrastructure, industrial composition and the supply of different types of amenities at the neighborhood level. These contain (1) data on local labor market structure and educational infrastructure obtained from Statistics Sweden, (2) nature-based amenities (climate, protected nature, open spaces, lakes and coastline) provided by The County Administrative Boards and The Swedish Meteorological and Hydrological Institute and (3) built heritages and cultural environments provided by the Swedish National Heritage Board. The amenity variables in focus are the ones that reflect local supply of cultural heritages and include built heritages (e.g., buildings, monuments and other types of physical structure that are appointed as having significant historical values) and cultural environments (e.g., cultural heritage sites that are designated as protected areas).⁹ Built heritages are obtained from the Data Base of Built Heritage (Bebyggelseregistret, BeBR), which is a national registry that holds information about the built heritage in Sweden. The represented built heritages range from traditional log-cabins and old industrial quarters to modern city buildings. There are around 80,000 searchable building records, with around 13,000 listed as National Monuments, Historical Buildings and Church Monuments. The content of the database is publicly available.¹⁰ Data on cultural environments are obtained from the County Administrative Boards. There are 1700 areas classified as national interests from the point of view of heritage protection, varying in their size and cultural characteristics. They include built environments, agricultural villages, city centers and mines that reflect the Swedish history and historical values from certain eras.

The variable in focus is the total number of registered built heritages and cultural environments in each neighborhood (SAMS) constructed using spatial joins (in ArcView). Since built heritages are not randomly distributed across geography and reflect places that were attractive for settlement a long time ago, the variable is weighted with population to account for this possible endogeneity bias. The hypothesis is a positive association between the local supply of cultural heritages that have an inherent historic and cultural significance and the growth of high-skilled workers (Falck et al. 2011; Andersson 2006). Figure 3 in “Appendix” shows the spatial distribution of built heritages and cultural environments across Sweden, and Fig. 2 shows their number aggregated to the neighborhood level and weighted by population.

Open image in new window

Following the outline in Brueckner et al. (1999), a distinction is made between natural and cultural amenities; seven variables are constructed to control for the local supply of nature-based amenities using the same GIS-based methodology.¹¹ These variables measure the local supply of: (1) recreation areas and preserved open spaces¹² (Geoghegan et al. 1997), (2) lakes, rivers, mountain and coastal areas (3) and measures of local climate (average temperature and average precipitation).

3.2 Neighborhood characteristics related to size, education and labor market

The presence of universities is important, not only because they produce high-human-capital individuals and employ higher-skilled labor, but also because of the resulting positive externalities to the private sector in the form of knowledge spillovers (Andersson et al. 2009; Abel and Deitz 2011). A dummy variable is therefore included to indicate if the municipality, in which the neighborhood is located, hosts a university. The ability of regions to attract human capital also depends on labor market characteristics in terms of industrial composition (Bresnahan et al. 2002; Berry and Glaeser 2005). To capture these effects, we create three variables that measure (1) the share of firms with a workforce where half of the employees have at least three years of studies at the university level, (2) the share of knowledge-intensive business firms (KIBS) and (3) the share of high technology manufacturing firms (HTMF) at the neighborhood level. Furthermore, since large regions tend to attract more high-skilled workers, we control for the size of neighborhoods in terms of population density. The regional level of human capital is also a function of the demographic structure, as human capital in many ways is a cohort phenomenon, we therefore include the average age in the neighborhood to control for age structure.

3.3 Estimated model

The data used in this study are structured as a cross-sectional dataset, and we model the growth in the number of high-skilled workers as a function of a set of initial conditions. Although it would have been preferable to use a panel setting, the variables that reflect the local supply of natural and cultural qualities are time invariant, implying that we need to rely on data from one year. This gives rise to the problem of unobservable heterogeneity, which is more challenging to mitigate when dealing with cross-sectional data. A concern in this study is the presence of level-2 endogeneity with regard to built heritages and cultural environments, e.g., when the local supply is influenced by factors that are common in the region (e.g., regional size and density). Although the spatial distribution of cultural heritages in Sweden does in some instances correlate with locations that have grown large and prosperous, this is not valid for the majority of cases. An indication of this is the low bivariate correlation between the local supply of built heritages and cultural environments and local population density. To account for unobserved regional heterogeneity and endogenous covariates, the empirical approach is to estimate a multilevel model where neighborhoods are nested in higher-level regional units.

Hence, the fixed part of the model contains variables that can be either variable within j (ij) or invariant (j). In the case, where the cluster sample mean coefficient (γ) is significant, this indicates an endogeneity bias across levels. However, as this is absorbed by the instrumental variable, it does not affect the estimated coefficient of the built heritage variable \(\left( \theta \right)\) (Snijders and Berkhof 2008).

The main advantage with this model is that it is able to mitigate the problems associated with unobserved heterogeneity and spatial dependencies by controlling for unmeasured heterogeneity between and within the two geographical levels. We therefore assume a higher correlation between neighborhoods within a given municipality than between and the varying intercepts can be interpreted as unobserved heterogeneity at the different levels as they induce dependence among units at the lower level in the same higher-level unit (Skrondal and Rabe-Hesketh 2004). Moreover, by estimating a growth equation where a set of initial conditions (in 2001) influence the development over the following 5 or 10 years (2006 and 2010), we mitigate the problem of reversed causality.

4.1 Contextual variation and control variables

Turning to the control variables and the contextual variation, the results show that a higher share of knowledge-intensive firms in both the business service (KIBS) and manufacturing sector (HTMF), i.e., the number of individuals that perform advanced and knowledge-intensive tasks is positively related to the growth of individuals with a higher education. These results are robust over the two studied time periods and to the inclusion of population density; they are also in line with prior findings stating that labor market characteristics are important determinants of local knowledge accumulation (Bresnahan et al. 2002; Berry and Glaeser 2005). Similarly, the variable indicating the presence of institutions for higher education, universities and university colleges is significant and robust. These results are intuitive considering that the presence of educational infrastructure increases the number of individuals with a higher education (supply effect); at the same time, they also give rise to a demand effect since individuals working in the university are generally highly educated. These results may also reflect that there exists a multiple effect working through other organizations and firms that are complements to universities such as Science parks and research laboratories, which also demand highly educated individuals. The link here is likely primarily via knowledge spillovers between universities and firms in the local economy as argued, for example, in Abramovsky et al. (2007). Related to the location-specific characteristics at the neighborhood level, we observe that there is a positive relationship between larger neighborhoods, in terms of population density, and the inflow of human capital-intensive individuals. It is interesting to note that the average age of the inhabitants already living in the neighborhood tend to have a discouraging effect. Given the assumption that younger and older highly educated individuals have different location preferences overall, this result can be interpreted as it is mainly younger highly educated individuals that change location and that they are attracted to neighborhoods that are overrepresented by younger individuals. The focus on how the relocation pattern differs of human capital-intensive individuals across the life-cycle is highly important but unfortunately outside the scope of this paper and is recommended for future studies.

Turning to the contextual variation, we find that the ICC for the municipality level ranges around 6–10% across the specifications. The rather low ICC for municipalities indicates that approximately 6–10% of the total variance is explained by characteristics at the more aggregated municipality level, given the independent variable and that the remaining variation is explained by neighborhood effects. We also estimated an unconditional model, excluding the regressors, to indicate the direct effect from the two geographical levels.¹⁴ The unconditional model shows that the neighborhood level is able to explain approximately 85% of the total variance in human capital growth and that the more aggregated municipality level explains the remaining part of approximately 15%. This indicates that most of the variance in the growth of highly educated individuals is captured by the variance at the local level suggesting that determinants are indeed local in nature (Cho et al. 2008).

4.2 Regional heterogeneity in the outcomes

Starting with the dummies for each regional category (E–G), it can be observed that the metropolitan dummy is positively correlated with the growth of human capital in each SAMS (specification E). This result is confirmed in other studies showing that individuals with a higher level of human capital are attracted to large dense regions (Backman 2014). The other side of the coin is reflected in the negative influence from the dummy variable for the most rural SAMS (specification G). Although Sweden has a history of decentralization policy of institutions for higher education, no institutions were established in the most rural parts of Sweden but in the semi-rural parts. Those semi-rural locations, wherein there exist an institute of higher education, have experienced an outflow of educated individuals as graduates tend not stay in the region of study (Bjerke 2012).

The results of including interaction terms for amenities (open spaces, built heritages and cultural environments and recreation areas) show mixed results. Regarding the interaction term for the variable capturing the Built heritages and cultural environments and the regional category, a positive relationship is observed only for city municipalities. While the results show a significant and positive coefficient of the interaction term between the coefficient reflecting the local supply of recreation areas and the metropolitan dummy, we find no significant interaction effects between open space and neither of the regional categories. Nevertheless, these results are interesting as they reflect that the slope of the recreational areas coefficient, with regard to growth in human capital, changes as the level of agglomeration changes, supporting our hypotheses concerning moderation effects among the determinants. They also reflect that valuations of natural amenities (in this case recreation areas) are valued higher when located in agglomerated regions. Similar results have been found in other studies, i.e., that the supply of recreational areas is valued, marginally higher, in locations that are characterized by high population and housing densities and low, insignificant or even negative in areas where these natural types are abundant (Bates and Santerre 2001).