Abstract
Empirical evidence on the effect of urbanisation on individual self-reported well-being generally points to a negative effect of urbanisation and city size, at least in most developed economies. This article aims to re-assess this conclusion by claiming that this approach overlooks the fact that a city’s positive externalities may expand well beyond the urban boundaries, as postulated in urban economics theory. Based on survey data on self-reported life satisfaction derived from different waves of Eurobarometer surveys in the period 2005–2010 covering 21 European Union member countries, the article empirically verifies the existence of a positive association between urbanisation and individual well-being, depending on the proximity to settings of higher rank in the urban hierarchy. In particular, it shows that the higher the distance from a city larger than the one of residence of the respondent, the lower the probability of being satisfied with life.
Introduction
A long stream of research has investigated the determinants of individual life satisfaction (LS) (Dolan et al., 2008). 1 In fact, by representing a proxy for experienced utility, the study of LS allows testing of some of the assumptions on which economic theory rests (Easterlin, 1973).
A branch of this literature has focused on the relationship between urbanisation and LS. The empirical findings are, however, largely mixed if not in contrast with the theoretical expectations: while cities are the place where the most intense processes of economic growth take place (Glaeser, 2011), urbanisation is frequently associated with lower levels of well-being when compared with less dense settings (Graham, 2012). This spatial mismatch between objective and subjective dimensions of well-being especially applies to most developed and affluent countries in the Western part of the globe (Burger et al., 2020). Many scholars have interpreted these results as the empirical demonstration that the negative externalities of large cities on well-being (cost of living, pollution, etc.) overcome the positive ones, like job opportunities, amenities, etc. (Fischer, 1973; Okulicz-Kozaryn, 2015).
This interpretation, however, fails to take into account a relevant aspect. The externalities of a city, in fact, are not constrained within the urban boundaries, as claimed in the ‘borrowed size’ concept proposed by Alonso (1973). Rather, such externalities are likely to spread out to the surrounding areas, that is, cities are sources of externalities affecting urban residents (within effects) as well as of externalities affecting residents of surrounding areas (between effects). The intensity and spatial range of such between effects are proportional to the characteristics (i.e. size and, thus, functions and rank) of the city generating them. At the same time, the recipient areas will be affected by such urban externalities with different intensities, that is, depending on their proximity to the originating city.
This idea is certainly not new, and a broad literature has focused on the spatial structure of cities, the diffusion of urban externalities and their effect on different dimensions of economic prosperity (Parr, 2014). Yet, to the best of our knowledge, there is no evidence available on the effects of urbanisation on LS across cities. The present article aims to fill this gap. Specifically, the goal of this work is to demonstrate that less-urbanised settings are not a source of well-being per se but, rather, that the proximity to higher-rank cities (and therefore to the urban functions they host) matters.
The discussion and empirical analysis presented here originally merge intuitions from traditional fields of investigation in urban economics. In particular, agglomerations economies theory (Duranton and Puga, 2004; Glaeser, 2011), the concept of borrowed size (Alonso, 1973; Camagni et al., 2016; Meijers and Burger, 2017; Meijers et al., 2016) and the Central Place Theory (CPT) framework (Christaller, 1933) offer important insights on households’ behaviour and conclusions on the hierarchical distribution of cities that fit well with the testing of our hypotheses, as commented in the next section.
The discussion is organised as follows. The next section reviews the literature on urbanisation and LS and elaborates the original contribution of this article. The presentation of data and methods follows in the third and fourth sections, respectively. The fifth section comments on the results of the empirical analysis, while conclusions and indications for further research can be found in the final section.
Urbanisation and LS: Evidence from the literature and research hypotheses
A broad and multidisciplinary literature, ranging from psychology to public health studies, from planning to economics, has studied the determinants of LS. Relatively recently, some scholars have started investigating its association with territorial factors and urbanisation in particular (Morrison, 2007). In the last decade, empirical evidence has been collected on this issue, adopting alternative statistical techniques (Ballas and Tranmer, 2012), typologies of data (Diener et al., 2013) and spatial levels of analysis ranging from regions to cities (Ballas, 2013), and from urban form (Mouratidis, 2019) to neighbourhood as well as buildings and houses (Ala-Mantila et al., 2018).
Evidence, however, is far from conclusive. Several studies boldly claim that, in developed economies, agglomeration leads to lower levels of LS (Berry and Okulicz-Kozaryn, 2011; Glaeser et al., 2016; Morrison and Weckroth, 2018). Still others are unable to establish statistically significant relationships between territorial and urban factors and people’s well-being. For example, Ballas and Tranmer (2012) do not find statistically significant variation in well-being across areas; Mouratidis (2019) does not report a significant association between city compactness (i.e. density) and LS. Besides the complexity in the empirics, there are also difficulties in disentangling the actual conceptual mechanisms at the basis of the empirical findings obtained. In particular, sorting and self-selection can make it difficult to separate out, both conceptually and empirically, whether urbanisation generates lower LS or rather if less-satisfied individuals tend to migrate to cities. In this respect, Morrison and Weckroth (2018) suggest that lower levels of LS in large cities depend on the disproportionate concentration of people with extrinsic and personally focused values (i.e. values that correlate negatively with well-being) in large cities, thus supporting the sorting argument. 2
Confronted with this multitude of approaches and findings, this article takes a specific perspective on the relationship between urbanisation and LS. In particular, the main claim of the article is that the debate on the possible existence of a negative net balance of positive and negative agglomeration externalities in the largest urban settings has been so far simplistic and insufficient, for at least two reasons.
First, cities are not all the same; each provides very different kinds of amenities and externalities to the resident population. Urban areas can be classified in different hierarchical ranks, proportional to their size. Larger cities at the top of the hierarchy occupy higher ranks, because they can offer a larger variety of agglomeration benefits and amenities (but also disadvantages, in terms of pollution, congestion and the like) to their own inhabitants compared with smaller cities at the bottom of the scale. The theory of agglomeration economies has largely emphasised the advantages stemming from agglomeration in terms of increased opportunities for matching, sharing and learning (Duranton and Puga, 2004). Moreover, cities are also increasingly seen as a source of consumption amenities, in terms of increased variety of goods and services, aesthetics and physical settings, high quality public services and speed of transport services (Glaeser et al., 2001). Following this line of reasoning, some authors have studied LS in urban settings of different sizes (Lenzi and Perucca, 2019). Interestingly, the negative effects of urbanisation on LS arise significantly only when contrasting the largest cities 3 with the least urbanised areas, supporting sociological theories of urban malaise (Fischer, 1973): above a certain population threshold, urbanisation diseconomies prevail on the positive externalities of cities, at least as far as their impact on residents’ LS is concerned.
Second, urbanisation externalities are not restricted within the urban boundaries, but propagate to the areas outside the city. Yet, surrounding areas are not alike either, and the extent to which they are affected by these externalities crucially depends on their distance from the closest cities (Van Oort, 2007). The relevance of spatial interdependencies and spillovers in the geography of well-being has been recently documented in empirical works (Okulicz-Kozaryn, 2011). This empirical result can be conceptually interpreted through the borrowed size concept developed by Alonso (1973) and rejuvenated and enriched in very recent contributions (Burger et al., 2015; Meijers and Burger, 2017). The borrowed size concept suggests that smaller places (i.e. not placed at the top of the urban hierarchy) can take advantage of some of the urbanisation benefits (e.g. accessibility to markets, larger and diversified labour markets, amenities) of proximate larger cities without incurring the related disadvantages (e.g. because of lower rents, less congestion). In short, urbanisation effects on LS are not bounded within a city’s administrative/functional borders (within, i.e. agglomeration, effects). Rather, urbanisation effects can spread across the regional urban system and filter down the urban hierarchy (between, i.e. borrowed size, effects), signalling the existence of a spatial range within which residents external to the city perceive the effects of urbanisation economies. The shorter the spatial range (i.e. the shorter the distance), 4 the greater the possibility to reap the agglomeration advantages stemming from nearby cities. 5 In particular, smaller cities embedded in larger (multi-centric as well as polycentric) 6 metropolitan areas and well connected within broader urban systems are the best candidates to borrow size from larger cities. Moreover, these cities are more likely to perform better (whatever the performance variable considered) than isolated ones because of a more favourable balance between agglomeration benefits and costs (Meijers and Burger, 2017; Meijers et al., 2016).
At present, however, the relevance of such between (i.e. borrowed size) effects on LS has been largely disregarded. Preliminary evidence was provided by Lenzi and Perucca (2018), who pointed out that within the most urbanised regions, residents in less dense settings are more likely to be satisfied, keeping other things constant, than those living in the main cities. However, the opposite holds for those living in regions with a low degree of urbanisation: in this case, residing in a less dense setting is associated with lower well-being than living in the main city. These findings suggest that urbanisation per se does not hamper LS. Rather, it is expected to be positively associated with LS only when residents external to the main city have access to the positive urban externalities, avoiding at the same time most of the typical urbanisation diseconomies, like pollution, criminality and high cost of living.
Stemming from these considerations, the goal of this article is to provide evidence on both the within and the between effect of urbanisation on LS. While a long stream of research has documented the former, even if with contrasting results (Burger et al., 2020), very few studies have focused on the between (i.e. borrowed size) effect.
Accordingly, the first hypothesis to be tested consists in the empirical verification of the findings from previous literature, as follows:
H1: LS is negatively associated with the rank of the city of residence (within effects).
As discussed above, the borrowed size concept stresses also the relevance of the impact of urbanisation on the LS of individuals living in neighbouring cities. As far as cities of higher rank provide greater agglomeration benefits, a broader variety of goods and larger consumption opportunities, physical proximity to them can have a positive effect on individuals’ LS. In fact, individuals value greater agglomeration benefits, consumption opportunities and variety in terms of utility – and, thus, LS. Therefore, individuals are likely to benefit from the proximity to those cities (and markets) that guarantee these conditions. Accordingly, it is reasonable to expect a negative association between LS and the distance from cities of higher rank, since remote locations are those characterised by higher transport costs and fewer opportunities to enjoy their greater agglomeration advantages and to access their superior goods’ markets. However, as postulated by the borrowed size concept, the positive externalities of large cities are not assumed to be in strong antagonism with their negative urbanisation economies since, by commuting, consumers can enjoy only the benefits of larger urban areas without sensibly undergoing their costs.
The second hypothesis to be tested, then, is as follows:
H2: LS is negatively associated with the distance from cities of higher rank (between effects).
At the same time, this positive effect may fade out and be superseded if a city of higher rank than the one taken into consideration is located closer, that is, borrowed size effects depend on the relative ranking of the originating and the destination cities. Keeping other things constant, proximity to top-rank cities is more valuable, in terms of individual utility and LS, than proximity to lower-rank urban settings, since the former generate greater agglomeration advantages and host a broader variety of markets. As a matter of example, consider the residents of a city of fourth rank. The distance to the closest third-rank city is expected to be negatively associated with their LS. Yet, if the distance to the closest top-rank city (second or first rank) is shorter than the distance to the closest third-rank one, it is reasonable to expect that the effect of the distance from the third-rank city becomes negligible. In this case, agglomeration shadow effects from the first- or second-rank city on the third-rank city are at work, meaning that the advantages of the smaller cities are outpaced by those of the first- or second-rank city (Burger et al., 2015). 7 In fact, there is a closer centre providing all agglomeration advantages, functions and goods typical of the third rank and, in addition, some of higher order. On the other hand, let us consider the resident population of a city of second rank. As noted above, the relationship between distance to the closest first-rank city and LS is expected to be negative. This effect is expected to hold even if a third-rank city is interposed between the residence of the individuals and the closest first-rank city, because the agglomeration advantages and goods provided at the third rank do not cover the whole agglomeration advantages and variety available at the top of the hierarchy. In this case, as well, agglomeration shadow effects from the first-rank city on the third-rank city are at work, meaning the advantages of the smaller cities are outpaced by those of the first-rank city (Meijers et al., 2016).
The third hypothesis to be tested, then, is as follows:
H3: LS is not associated with the distance from cities of higher rank when there is another city, of higher rank than the previous one, which is closer (spatial hierarchical ordering of between effects).
Finally, the borrowed size concept as well as CPT highlight that cities of the same size are expected to provide similar varieties of goods and amenities, and then utility and LS levels. Accordingly, cities of the same rank can be considered as comparable in the functions they host and in the general level of agglomeration advantages they generate, at least to a certain extent. Even if cities located in polycentric urban systems are more favourably placed to borrow size than isolated ones (Burger et al., 2015; Meijers et al., 2016), 8 the pair-wise distance from cities of the same size is expected to have a negligible if not nil effect on LS. In this case, in fact, distance would capture the proximity to cities that are very similar (in terms of markets, functions, amenities) to the one of residence, that is, agglomeration shadow effects are likely to prevail on borrowed size effects.
The last hypothesis to be tested, then, is as follows:
H4: LS is not associated with the distance from cities of the same rank as the one of the city of residence (spatial hierarchical ordering of between effects).
The next sections present data and methods applied to test these hypotheses.
Data
The database employed in the empirical analysis is made up of several waves of Eurobarometer (EB) survey studies. A recurrent question concerns the degree of LS of European Union (EU) citizens: respondents are asked to choose whether they are ‘not at all satisfied’, ‘not very satisfied’, ‘fairly satisfied’ or ‘very satisfied’ with their life. The degree of self-reported LS is frequently used in the literature as a measure of LS (see, among others, Deaton, 2008; Hoogerbrugge and Burger, 2018). The dataset used in the present empirical analysis collects the evidence from 15 EB waves (more than 250,000 observations) on individual LS conducted between 2005 and 2010. 9
On conceptual grounds, the data collection followed the general approach adopted by most of the studies on LS: LS is assumed to depend on a set of individual characteristics and other variables characterising the context in which the respondent is living. A complete description of these data is available in Supplemental Material B. 10 Among the former, respondents provide information on their demographic and socioeconomic status (age, sex, occupational and marital status), identified by previous literature as among the most significant determinants of LS (Dolan et al., 2008). Among the latter, respondents are asked about their NUTS3 11 region of residence, allowing to test whether the characteristics of the local setting are associated with LS. 12 In particular, as discussed above, the focus of this article is on the relationship between LS, the degree of urbanisation and the proximity to urban settings of higher rank. Notice that most existing studies on the link between urbanisation and LS in the EU classify the urbanisation level of individuals’ region of residence at the NUTS2 level at best. The availability of data at the NUTS3 level implies some advantages. First, in the European context, NUTS3 is the classification level that best approximates urban areas, although defined according to administrative criteria. NUTS3 can therefore be considered as a fair proxy for cities. Second, the NUTS3 level captures in a much finer way the degree of urbanisation of the setting in which respondents live. In fact, while within NUTS2 regions urban settings of different kinds may coexist, at the lower level of the nomenclature it is fair to assume that each area is characterised by its own specific positioning in the urban hierarchy. Therefore, even if within each NUTS3 the largest city might not completely cover the surface area, it is, however, fairly reasonable to assume that the rank of the main city in each NUTS3 strongly characterises the functions and markets supplied in the NUTS3 itself.
In the present study, urbanisation is therefore captured by a set of dummy variables controlling for the rank of the largest city in the NUTS3 of residence of the respondent. Urban ranking comes from the official classification of cities provided by Eurostat (n.d.). According to this definition, EU cities are classified into six groups, from first rank to sixth rank. 13 As far as the proximity among settings with different degrees of urbanisation is concerned, the use of NUTS3 allows us to measure with reasonable precision the distance between cities of different ranks. The greater the distance, the lower the possibility to borrow size, in terms of agglomeration advantages and accessibility to superior goods’ markets. In particular, the distance to the closest city of higher rank is measured, in terms of travel time by car (as in Polèse and Sheamur, 2006).14,15
Finally, EB surveys do not provide consistent information across waves about the income and/or wealth of the respondents. In order to mitigate the potential risk of omitting this individual characteristic, per capita income in the city (i.e. NUTS3) of residence represents the control variable for the overall level of wealth in the area of residence, while at the individual level we included information on whether the respondent owns the apartment where she is living.
The methods applied for the investigation of the association between LS and urbanisation are discussed in the next section.
Methods
The relationship between LS and the other characteristics, whose empirical measurement was presented in the previous section, can be formalised as follows:
where i stands for the individual, r and c respectively for the NUTS3 and country of residence and t for the wave of the survey study. As mentioned in the previous section, EB surveys do not have a panel structure, that is, it is not possible to observe the same individual in different periods. Therefore, we estimate equation (1) on a pooled dataset joining together the EB data in the period 2005–2010. 16 The variables of interest for addressing the research hypotheses are the rank of the NUTS3 of residence (urbanisation) and the distance, in terms of travel time by car, from the closest cities of higher rank (distance from cities of higher rank). We are interested in the estimation of the coefficients θ and γ once controlling for individual characteristics (X), NUTS3 characteristics (Q) and survey-specific effects (τ).
The empirical estimation of equation (1) poses a number of methodological issues, related to: (i) the coding of the dependent variable and the choice of the most appropriate statistical techniques to be adopted; (ii) the hierarchical structure of the data; and (iii) the potential sorting and self-selection of individuals.
For what concerns the first issue, EB data provide the level of individual LS defined on a four-point scale. In the literature on the topic, some studies keep this coding of the data, while some others prefer to dichotomise the level of LS, distinguishing between satisfied and unsatisfied respondents (Boyd-Swan and Herbst, 2012). The transformation of the dependent variable into a binary variable facilitates the interpretation of the results, but at the expense of the loss of some information. Therefore, in the context of this article we chose to maintain the definition of the dependent variable in four categories: very satisfied with life (4), fairly satisfied (3), not very satisfied (2), not at all satisfied (1). The definition of the dependent variable highlights another issue, related to the most appropriate statistical techniques to be used for the estimation of equation (1). Also in this case, previous literature includes works employing either models for categorical dependent variables or linear models (Okulicz-Kozaryn, 2011). In this study, linear regression analysis is adopted. Neither the alternative coding of the dependent variable (dichotomous vs categorical) nor the statistical techniques adopted (for continuous vs categorical variables) lead to findings significantly different from those reported in the next section. 17
A second methodological issue concerns the hierarchical structure of the data and the most adequate technique to capture the within and between effects of urbanisation discussed in the second section. Individuals in the sample are nested within countries, at a first level, and within cities (i.e. NUTS3) at a second level. This hierarchical structure of the data may imply that two randomly selected individuals from the same area are more similar, in terms of LS, than two people randomly chosen from other groups. If these group effects are not taken into account, the independence assumption of the residuals from model (1) will not hold. In order to address this issue, we estimate a random effect model to explicitly treat the hierarchical structure of the data and to correct standard errors. Specifically, we estimate a linear random intercept model, where the intercept of the group regression lines is allowed to randomly vary across areas. Therefore, in equation (1),
Finally, the third potential methodological issue concerns sorting and self-selection, which may arise if satisfied people, for instance, tend to relocate in cities that are more satisfied in general, and vice versa. To a certain extent, this issue can be mitigated by controlling for the origin of the respondents. A recurrent question in Eurobarometer surveys differentiates bet-ween the respondents born in the country of current residence and those born elsewhere. The estimates of equation (1) for different samples of individuals (natives vs non-natives) suggest rejecting the hypothesis about the occurrence of a worrying issue of sorting. Results of these robustness checks are reported in Supplemental Material C.
The next section discusses the results of the empirical analysis on the relationship between urbanisation and LS.
The within and between effects of urbanisation on LS
This section presents the empirical findings on the testing of the four research hypotheses elaborated in the second section. Table 1 reports the unstandardised coefficients from the models investigating both the within (H1) and between (H2) effect of urbanisation on LS.
The within and between effect of urbanisation: LS as a function of the rank of the city of residence and of the distance to cities of higher rank.
Notes: Reference categories: sixth-rank regions, student (occupation), married (marital status). Robust standard errors in parentheses. ***p < 0.01, **p < 0.05, *p < 0.1.
All models include, as controls, individual characteristics, GDP per capita and EB-wave dummies. 19 Four equations are estimated, whose findings are reported in columns [a]–[d]. Taking cities with fewer than 50,000 inhabitants as the reference group (sixth-rank), the first equation (column [a] in Table 1) aims to test the association between LS and living in a first-rank city. Therefore, these results test for the occurrence of what we defined as the within effect of urbanisation in the second section. Columns [b], [c] and [d] in Table 1 provide evidence on the between effect of urbanisation, that is, whether the distance respectively from a city of first (column [b]), second (column [c]) and third rank (column [d]) has a significant association with LS.
Table 1 shows that the LS associated with top-rank cities tends to be, other things constant, lower than that in areas at the bottom of the urban hierarchy. This result is to a certain extent consistent with the findings from previous literature as well as with H1, that is, urban areas at the top of the hierarchy are generally associated with lower LS compared with cities of lowest rank. Therefore, urbanisation negatively correlates with LS only in the largest cities, while the same does not apply to urban areas of lower size.
In order to test H2, about the between effects of urbanisation, the travel time distance by car to the closest city of higher rank was added to each model specification. As discussed in the second section, individuals are expected to benefit from living close to cities larger than their own, since proximity provides them with potential access to urbanisation benefits not supplied in the place of residence.
Results reported in column [b] of Table 1 show the relationship between LS and the proximity to first-rank cities. Therefore, the regression was run on the individuals living in cities of second and lower rank. 20 The same procedure was applied in order to investigate the impact on LS of the distance from, respectively, second-rank (column [c], thus excluding individuals living in first- or second-rank cities) and third-rank (column [d], thus excluding individuals living in first-, second- or third-rank cities) cities. 21 These results show that the coefficient associated with the distance from cities of higher rank is always negative and statistically significant. This implies that the higher the distance from a city larger than the one of residence of the respondent, the lower the probability of being satisfied with life. This result confirms our hypothesis on the between effects of urbanisation on LS: while living in first-rank cities is likely to reduce LS compared with settings characterised by a lower level of urbanisation (within effect), the proximity to the top-ranked urban centres is associated with higher levels of LS. The same applies for the proximity to the cities of second and third rank. For individuals living in less-urbanised NUTS3 (i.e. fourth-rank or lower), LS decreases with increasing distance from a relatively larger city.
Yet, the intensity and ranking of the between effects are likely to vary along the urban hierarchy according to the rank and relative proximity of the origin and destination cities (H3), because of the operation of agglomeration shadow effects (Burger et al., 2015; Meijers et al., 2016).
Empirically, this expectation was tested by estimating the effect of the distance to the closest third-rank city on the LS of individuals living in cities of fourth rank or lower. This effect, which was found to be negative and statistically significant (see column [d] in Table 1), is expected to vanish when a city of higher rank (first or second) is closer than the closest third-rank city. This hypothesis is tested in Figure 1 (Panel 1a). The figure plots the coefficient, and the corresponding 95 per cent confidence interval, of the effect on LS of the distance from third-rank cities in two alternative scenarios. Scenario [d] corresponds to the between effect of third-rank cities estimated on the whole sample (i.e. the coefficient reported in column [d], Table 1). Scenario [e] corresponds to the situation in which the distance from the closest third-rank city is higher than the distance from the closest first- or second-rank city. 22 Results show that the coefficient associated with scenario [e] is not statistically significant (i.e. the extreme values of the intervals of confidence show positive and negative signs). Put differently, in this case the distance from the closest third-rank city does not affect individuals’ LS. This finding supports the expectation that proximity to cities of higher rank influences LS only when the borrowed size effects, that is, the possibility to take advantage of the urbanisation economies of larger cities, prevail on the agglomeration shadow effects.

Evidence on the spatial hierarchical ordering of the between effect. Panel 1a: Effect on LS of the distance from the closest third-rank in the general model of Table 2 [d] and when a city of higher order (first or second) is closer than the closest third-rank [e]. Panel 1b: Effect on LS of the distance from the closest first-rank in the general model of Table 2 [b] and when a city of third-rank is closer than the closest first-rank [f].
Further evidence in favour of this conclusion is provided by the opposite experiment. Consider the population in cities of second rank, or lower. Table 1 (column [b]) shows a negative and significant association of their LS with the distance to the closest first-rank city. According to what we discussed in the second section, this effect is assumed to hold even if a third-rank city is interposed between the residence of the individuals and the closest first-rank. In fact, as explained above, the agglomeration advantages provided at the first rank of the urban hierarchy, also in terms of accessibility to superior goods and services markets, are not overshadowed by those generated by cities of third rank. As above, this hypothesis is tested in Figure 1 (Panel 1b). Again, the figure reports the coefficients and confidence intervals of two alternative scenarios. Scenario [b] corresponds to the overall sample, and is therefore analogous to the output reported in Table 1, column [b]. Instead, scenario [f] considers the situation in which the distance from the closest first-rank city is higher than the distance from the closest third-rank city. 23
As expected, the between effect of first-rank cities does not significantly differ between the two scenarios. Put differently, the proximity to cities of lower rank does not interfere with the borrowed size effects stemming from being close to centres of higher rank, while the opposite holds true because of the operation of agglomeration shadow effects (Meijers et al., 2016).
Finally, as far as cities of the same rank are comparable in terms of agglomeration advantages and typologies of goods and services markets, we may expect the minimum distance from cities of the same rank to have a negligible effect on LS. In this case, distance would capture the proximity to markets and functions very similar to the ones already present in the city of residence (H4).
This hypothesis is empirically tested in Table 2. The table reports the estimates of a model equivalent to the one used for the regression of Table 1. The sample, however, is restricted to the respondents living respectively in first- (column [b]), second- (column [c]) and third-rank (column [d]) cities.
The impact of the distance to cities of the same rank on LS.
Notes: Reference categories: sixth-rank regions, student (occupation), married (marital status). Robust standard errors in parentheses. ***p < 0.01, **p < 0.05, *p < 0.1.
As expected, none of the coefficients associated with the minimum travel time is statistically significant. This evidence implies that the relationship between proximity and LS holds only for cities of lower rank and their proximity to urban settlements of higher rank. In that case, distance mirrors the potential access to agglomeration advantages, services and goods that are not available in the place of residence.
Conclusion
This article presented an original analysis and novel evidence on the links between urbanisation and LS based on a rich dataset on individual LS collected at the NUTS3 level. Without contradicting key conclusions of the existing literature, that is, a lower level of LS in the case of residents in the largest urban settings, it demonstrates that the effect of urbanisation appears to be much more complex than this. Proximity to large cities, and therefore the accessibility to their agglomeration advantages, matters in the understanding of the LS of residents in smaller cities. In fact, the higher the distance from a city larger than the one of residence, the lower the probability of being satisfied with life. This result is important because it recognises that large cities are not just a source of dissatisfaction. Rather, they produce positive spillovers spreading beyond their boundaries and reaching the population living within a broad range. Moreover, consistent with the literature on borrowed size, and important pieces of traditional urban economics theory, the results seem to confirm the hierarchy of these between effects on LS, so that the proximity to higher-rank centres is valued more than the proximity to lower-level ones.
These results also convey some policy implications for urban and regional development. First, they point to the relevance of accessibility and connection between cities for individual well-being. What matters, therefore, is not simply the advantages and amenities in a single city but also the opportunities to enjoy those available in neighbouring ones. Boosting connectivity between cities can therefore highly promote individual well-being. Second, the results suggest that urban policies are likely to generate an impact also on the LS of individuals living in other administrative areas, typically not involved in the decision-making process. Hence, more integrated urban planning processes may be socially beneficial, maximising these externalities on LS. Third, our article contributes to the debate on the spatial variation of LS by indicating that it cannot be reduced just to an urban–rural divide. Rurality, when associated with remoteness, leads to even lower levels of LS than in first-rank cities. This result is fully consistent with the recent stream of research on the so-called ‘geography of discontent’, which identifies in peripheral, less-urbanised areas the core of increasing anti-system sentiments (Dijkstra et al., 2020). From a policy perspective, this implies that connectivity, but also the provision of services typical of higher-rank cities (Perucca et al., 2019), are particularly important in those regions that are currently excluded from the spillover effects of urbanisation described in this article.
As a first piece of evidence on this topic, the present article is certainly not exhaustive and future research should be aimed at investigating some of the issues raised by the present study. The first issue concerns the spillover effect arising from the geographical proximity of individuals within the same city. Unfortunately, the data used for this study did not allow for such measurement, which could shed light on the manifestation of the within effect at different distances from the closest city of one’s own region of residence. A second relevant topic involves the heterogeneity of the effects of urbanisation on LS across different groups of individuals, based on their socioeconomic status. Finally, an important issue is represented by the COVID-19 pandemic, and the way in which its consequences will reshape individuals’ behaviours, mobility and, more in general, the urban interactions and interdependencies. We hope to extend our future research in these directions.
Supplemental Material
USJ962397_Supplemental_Material_A – Supplemental material for Not too close, not too far: Urbanisation and life satisfaction along the urban hierarchy
Supplemental material, USJ962397_Supplemental_Material_A for Not too close, not too far: Urbanisation and life satisfaction along the urban hierarchy by Camilla Lenzi and Giovanni Perucca in Urban Studies
Supplemental Material
USJ962397_Supplemental_Material_B – Supplemental material for Not too close, not too far: Urbanisation and life satisfaction along the urban hierarchy
Supplemental material, USJ962397_Supplemental_Material_B for Not too close, not too far: Urbanisation and life satisfaction along the urban hierarchy by Camilla Lenzi and Giovanni Perucca in Urban Studies
Supplemental Material
USJ962397_Supplemental_Material_C – Supplemental material for Not too close, not too far: Urbanisation and life satisfaction along the urban hierarchy
Supplemental material, USJ962397_Supplemental_Material_C for Not too close, not too far: Urbanisation and life satisfaction along the urban hierarchy by Camilla Lenzi and Giovanni Perucca in Urban Studies
Supplemental Material
USJ962397_Supplemental_Material_D – Supplemental material for Not too close, not too far: Urbanisation and life satisfaction along the urban hierarchy
Supplemental material, USJ962397_Supplemental_Material_D for Not too close, not too far: Urbanisation and life satisfaction along the urban hierarchy by Camilla Lenzi and Giovanni Perucca in Urban Studies
Footnotes
Acknowledgements
This work would not have been possible without the support of GESIS - Leibniz Institute for the Social Sciences, which provided us with Eurobarometer survey data at NUTS3 level. In particular, we thank Meinhard Moschner for his time, availability and precious help. We also thank three anonymous referees for their useful comments on a previous draft of this article.
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) received no financial support for the research, authorship, and/or publication of this article.
Supplemental material
Supplemental material for this article is available online.
Notes
References
Supplementary Material
Please find the following supplemental material available below.
For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.
For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.
