CAIRN-INT.INFO : International Edition

1In Northern countries, the cities of the past were unhealthy places to live, and the mortality of city-dwellers was higher than in rural areas. This imbalance was reduced by the development of agglomerations and their infrastructures, accompanied by a growing concentration of qualified jobs and wealth, such that in the twentieth century, the differences between town and country disappeared. But with the growth of periurban areas and residential mobility from city centres to the agglomeration belts, intra-urban mortality differentials are now emerging. Using the example of Switzerland, Mathias Lerch, Michel Oris and Philippe Wanner describe changes in the geography of mortality in the last quarter of the twentieth century. They first show that the mortality gradient between rural, periurban and urban areas has changed considerably over time, then analyse the mechanisms whereby inhabitants of agglomeration belts now enjoy the lowest mortality levels.

2Geographic variations in mortality have been extensively analysed in developed countries as a means to assess population health and to better target preventive measures (Caselli and Vallin, 2006). Rural-urban differences have attracted less attention from researchers, despite their historical importance in the demographic process of urbanization (Vries, 1990). Before the onset of the demographic transition (i.e. the sequential fall in death and birth rates), urban areas were characterized by high mortality due to population pressure on the city environment (Ramiro-Fariñas and Oris, 2016). Urbanization was essentially the result of a massive rural exodus. After the urban death rates declined to levels below those of rural areas thanks to modernization and sanitary improvements, urbanization was sustained by natural increase. Yet in the context of low birth rates, urban demography is again increasingly determined by spatial differences in mortality (and migration). In this article, we investigate the urban geography of mortality in Switzerland.

3In 1920-1921, three decades after the onset of the demographic transition in the Swiss Confederation, 29.2% of the country’s population lived in cities. Life expectancy was up to 12 years higher in the more urbanized than in the less urbanized cantons because of higher infant mortality related to infectious diseases (Fei et al., 1998). In 2013-2014, 73% of the 8.3 million inhabitants of Switzerland lived in urban areas. Life expectancy in the country reached world record levels (80.7 years for men and 84.9 years for women), and the intercantonal gradient fell sharply, to only 2.2 years for men and 2.4 years for women. [1] Previous research stressed the importance of behavioural risk factors, and hence of mortality differences at older ages (Wanner et al., 1997). With the progression of this epidemiological transition towards old-age mortality related to chronic diseases (Omran, 1971), a new geographic gradient emerged in the 1980s. Life expectancy in more urbanized regions fell below the level observed in the Swiss periphery (Bopp and Gutzwiler, 1999; Wanner et al., 1997, 2012).

4This reversal is related here to the changes in the patterns of population redistribution that underlie the historical process of urbanization (Geyer and Kontuly, 1993). The initial phase of population concentration led to diseconomies resulting from industrial agglomeration and congestion (in terms of traffic and pollution) in city centres. In the early to mid-twentieth century in Europe, this development resulted in the spatial extension of cities and the delocalization of jobs; a process referred to as suburbanization. Later, with the shift from an industrial to a post-industrial economy, the development of transport and communication technologies reduced the significance of distance to the workplace as a residential determinant. These changes led to a second phase of urban sprawl into formerly rural areas located on the more distant urban periphery; a process referred to as peri- or counter-urbanization (Champion, 1989). This trend reflected a desire for environmental amenities in less congested and more natural settings (Geyer and Kontuly, 1993), and was sustained by intra-urban differentials in housing costs and the allocation of activities (Frumkin et al., 2004).

5The process of periurbanization started in the 1960s in Switzerland, and would have resulted in a sharp decline in city-centre and suburban populations had outmigration not been offset by the inflow of young adults – international migrants especially. This outflow was notably selective, with an overrepresentation of highly qualified individuals and wealthier families. Populations in city centres, by contrast, remained relatively diverse because of the emergence of a dual labour market for both highly specialized skills and low-paid jobs in the service sector (Cunha and Both, 2004; Rérat et al., 2008).

6Given that in Switzerland the reversal of the urban geography of mortality coincided with the intensification of periurbanization, the link between the two processes deserves to be analysed. Based on the international literature, we start by presenting the pathways through which periurbanization may be expected to influence urban mortality. After introducing our data and methods, we describe the differences in life expectancy, as well as their underlying age- and cause-specific patterns, across the urban-rural continuum at the national and city levels since 1969. We then apply a multilevel model to the entire adult resident population in 2000-2008 to investigate whether the spatial concentration of population characteristics and/or the spatial differences in living contexts explain the urban mortality gradient.

I – Potential causes of a new urban mortality gradient

7The analysis of urban mortality is challenging because cities are changing so rapidly. As cities grow in size and in population, the resulting changes in urban environments have multiple and often competing physical and socioeconomic consequences for the health of urban populations (Ramiro-Fariñas and Oris, 2016).

8First, urbanization refers to a process of population growth and agglomeration that transforms the built environment, with direct consequences for health. The health of urban populations may be adversely affected by air pollution, smog, traffic accidents, lack of time for physical activity because of long commutes, and the heat island effect [2] (Frumkin et al., 2004). Studies in western countries have shown that mortality is higher in more densely populated areas than in less densely populated ones. Urban populations have especially high rates of respiratory disease, lung cancer, obstructive pulmonary disease and ischaemic heart disease (Chaix et al., 2006; Fan and Song, 2009; Gartner et al., 2011; O’Reilly et al., 2007; Pearce and Boyle, 2005). In Switzerland, mortality from cancer of the trachea, bronchus or lung also increases with exposure to particulate matters and residential proximity to a major road (Huss et al., 2010). Excess mortality during the 2003 heat wave was also higher in city centres and suburban areas, possibly due to an urban heat island effect (Grize et al., 2005).

9According to this environmental hypothesis, mortality should decrease linearly from the city centre to suburban and periurban areas, with the lowest level being found in the countryside. Periurbanization exacerbates the environmental effects on health because it extends the built environment, as Fan and Son (2009) observed in metropolitan areas of the United States.

10Second, in a historical perspective, periurbanization has transformed local population structures by spatially (re-)distributing population characteristics associated with health behaviours and mortality. Effects of selective migration on geographic differences in health and mortality have indeed been found in several European countries (Eggerickx and Sanderson, 2010; Kimbele and Janssen, 2013; Maguire and O’Reilly, 2015; Verheij et al., 1998). In Switzerland, educational attainment and marital status are the two factors most clearly related to health status (Burton-Jeangros, 2009). Results on mortality are coherent, as 30 year-old men with tertiary education can expect to live up to 7.1 years longer, on average, than their lowest skilled counterparts; and married people have similar advantages relative to single people (Schumacher and Vilpert, 2011; Spoerri et al., 2006).

11This structural hypothesis expects that people who live in periurban areas will have a health advantage, as the migrants to these areas tend to be selected among the better educated and are – at least when they make their residential change – more often married than the individuals who live in either the city centre or in the countryside. In rural areas, mortality should be relatively high because the inhabitants tend to have low socioeconomic status. Expectations are more difficult to establish for city-centre populations, however, as the relatively high mortality of the less educated native residents may be offset by the influx of highly qualified and/or healthy immigrants, as evidence for Montreal appears to suggest (Choinière, 1991).

12While these two hypotheses are clearly distinct, disentangling the impact on mortality of compositional changes associated with periurbanization from the impact of the environment remains a delicate exercise. In describing the health determinants of city dwellers, a challenging distinction has been made between urbanization and urbanicity (Vlahov and Galea, 2002), both of which are affected by the contemporary process of periurbanization. Urbanization, here, basically refers to the above-presented environmental hypothesis. The term urbanicity, on the other hand, refers to interactions and redundancies of (dis)advantages affecting the spatial distribution of risks factors (Vlahov and Galea 2002). Urbanicity differs according to the local living context since it is shaped by the spatial aggregation of individuals and their interactions, as well as by exogenous influences (social policies, urban planning). These collective characteristics of the living environment may affect the lifestyles and health of all residents, regardless of their socioeconomic status (Macintyre et al., 2002).

13These place-specific group properties are often proxied by area-level material deprivation, which exerts physiological influences on health through poor infrastructure and housing (Cyril et al., 2013). Inhabitants of poor areas generally have higher mortality, including in Switzerland (Moser et al., 2014). Yet these contextual effects are usually modest compared to the much larger effects of individual characteristics, and they mainly concern men (Pickett and Pearl, 2001). Nevertheless, controlling for area-level material deprivation has explained the urban mortality gradient in the United States and in England and Wales (Gartner et al., 2011; Singh et al., 2011).

14In addition to socioeconomic disparities across different areas, the unequal distribution of resources and wealth within local populations may also be associated with higher mortality because of the psychosocial stress induced by social comparison (Wilkinson, 1996). This could explain why the situation is worst in areas where poor and wealthy people cohabit. Risky health behaviours related to nutrition, smoking and alcohol consumption are not only higher among the lower social classes, but may even increase when they live in affluent areas. Although the negative effect on mortality associated with unequal living contexts was robust to controls for geographic differences in population structures in the U.S., this hypothesis has not been systematically confirmed elsewhere (Jen et al., 2009; Subramanian and Kawachi, 2004).

15In addition, urbanicity also encompasses the opportunities and constraints which are specific to cities. These include the quantity and quality of health and educational facilities which, historically, were the main cause of the transition from urban excess mortality to an urban advantage in the first half of the twentieth century (Ramiro-Fariñas and Oris, 2016). In a dynamic perspective, selective outmigration to the urban periphery from the 1960s left poorer population groups behind. This impoverishment had a direct impact on tax revenues and threatened the urban health advantage. [3]

16This income reduction occurs when urban city centres have to face most of the so-called “new social risks” which include not only precariousness, structural unemployment, the rise of elderly isolation, etc., but also new family forms, especially single-parent families who tend to be associated with a high risk of poverty (Ranci, 2010). Moreover, in Finland, it has been shown that mortality is higher in urban areas characterized by greater heterogeneity in living arrangements, even after controlling for the individuals’ family situations (Martikainen et al., 2003). Negative contextual impacts on health linked to more limited social interaction are thus likely to be observed in environments with high concentrations of non-conventional family forms.

17Finally, if we consider that the inflow of both low-skilled and highly qualified immigrants increases local inequality in urban centres, it seems clear that those areas will accumulate disadvantages for survival. The lowest mortality levels should, by contrast, be found in periurban regions because these areas tend to have relatively high living standards and amenities, including more family-friendly social environments (Eggerickx et al., 2002). These contextual effects may be competing in rural areas; while material deprivation may be higher, local inequality tends to be lower than in city centres.

18In summary, the environmental effects of urbanization should lead to a general urban disadvantage in mortality. The structural and contextual hypotheses, by contrast, predict a more differentiated mortality gradient. Mortality levels should shift from being highest in city centres, to being lowest in the periurban belt, and then to being relatively high in rural areas.

II – Data, definitions, and methods

1 – Analytical approach and data

19We describe urban trends and gradients in mortality between 1969 and 2002 at the national and the agglomeration level, and explore the associations of these gradients with the demographic and socioeconomic processes of periurbanization. We use data from the Swiss Federal Statistical Office on the average annual number of deaths registered by vital statistics over the periods 1969-1972, 1979-1982, 1989-1992, and 1999-2002; and the populations at risk enumerated in the censuses taken in December 1970, 1980, 1990, and 2000, respectively.

20We then examine whether the urban gradient in 2000-2008 can be explained by the socioeconomic consequences of periurbanization, applying multilevel analysis on individual follow-up mortality data. We use the Swiss National Cohort (SNC) database, in which 94% of registered deaths occurring at ages 25-89 were linked to the individuals enumerated in the 2000 Census. A mix of deterministic and probabilistic methods of record linkage was used (Bopp et al., 2008). This analysis also includes unlinked deaths, which were imputed by means of a stratified random technique. [4]

2 – The urban classification

21We adopted Schuler et al.’s (2005) functional approach to space based on the concepts of agglomeration and metropolitan areas (i.e. city-centred labour markets). Their urban classification is based on geolocalized data from the 2000 Swiss Population Census and regroups the official municipalities based on the following criteria: the continuity and form of the built-up area, population density, demographic growth, and the scale of commuting to the agglomeration’s central municipality, which is represented by an official city (i.e., with at least 10,000 inhabitants). As can be seen in Figure 1, the Swiss urban system is composed of 50 agglomerations and five isolated cities situated on a plateau along a south-west to north-east axis (in parallel to the Alps; in white), which is where the majority of the population live. The largest agglomerations are Zürich (with more than one million inhabitants) in the north-east, Geneva and Basel on the south- and north-western borders (about 480,000 inhabitants each), followed by the political capital of Bern in the middle of the country, as well as Lausanne in the west (more than 300,000 inhabitants each).

Figure 1

Urban typology of Swiss municipalities in 2000

Figure 1

Urban typology of Swiss municipalities in 2000

Source: Schuler et al. (2005).

22Within these agglomerations, three broad classes of municipalities are distinguished in order to represent the urban-rural continuum of space (Figure 1): central municipalities (hereafter also referred to as city centres); suburban municipalities in the first agglomeration belt which was initially urbanized in the early twentieth century; and periurban and high-income municipalities (hereafter referred to as periurban), which have been urbanized since the 1960s. City centres play a central political or functional role. Suburban municipalities are defined according to minimum thresholds of population, building densities, and local employment. [5] Periurban municipalities do not meet these criteria. High-income municipalities are a special category, characterized by high average per capita tax levels. They are generally localized in the urban periphery (see Schuler et al. 2005). Areas located outside of these agglomerations are referred to as rural.

23For the analysis of mortality trends at the aggregate level, we applied the urban typology defined in 2000 to population and mortality data for periods back to 1969. Municipality codes were harmonized to take into account the administrative regrouping since 1969, as documented by the Federal Statistical Office (BFS, 2007). The demographic trends across this urban-rural continuum over the three decades of observation are those expected, with a decline in city-centre population numbers and strong gains in periurban areas (a 1.5-fold increase; Table 1). This delineation of the boundaries of agglomerations and intra-urban zones at the end of the observation period enables us to focus on the demographic dynamics and the spatial differentiation of populations over time within constant spatial units. However, there is a risk of misclassifying peripheral areas on the outskirts of cities in earlier periods of urban sprawl. The descriptive statistics in Table 1 do not confirm such a bias in Switzerland, as the age and the socioeconomic composition of the populations in areas that had a periurban status in 2000 were already more similar to urban than to rural areas in 1970.

24For the multilevel analysis of individual data in 2000-2008, the official municipalities were regrouped into 207 higher-level spatial clusters. In large agglomerations (with at least 40,000 inhabitants), clusters of municipalities are defined by the agglomeration and the intra-urban zone they belong to. Due to the low number of deaths in less populated areas, clusters in smaller agglomerations and in rural areas are defined by the “spatial mobility regions” of residence, which group municipalities according to their structural characteristics and commuting patterns (Schuler et al., 2005). The clusters representing small agglomerations are considered suburban, whereas all of the regions located outside of urban agglomerations are considered rural.

3 – Methods and indicators

25For each urban zone, we calculated male and female life expectancies at birth based on the death rates of the abridged life tables at successive census dates. The differences between each urban zone and the national average were decomposed for 1970 and 2000 according to the contribution of broad age groups. We used the method developed by Arriaga (1984, 1989), which estimates the total effect of mortality differences in each age group on the differential in life expectancy (by summing the years gained [lost] in each age group and the years gained [lost] due to the higher [lower] number of survivors exposed to different mortality conditions at older ages). The relative contributions of the main causes of death to the differentials can also be estimated for 2000; we used the classification applied by the Federal Statistical Office (Kohli, 2007), distinguishing between cancers, cardiovascular diseases, respiratory diseases, external causes, and residual causes (which are dominated by diseases of the nervous and digestive systems, and mental and behavioural disorders).

Table 1

Population characteristics according to Schuler et al.’s urban typology, Switzerland 1970 and 2000

Table 1
Centre Suburban Periurban Rural Total 1970 Population 2,303,961 1,626,036 695,079 1,633,063 6,264,086 Age structure Pop. aged 65-89 / Pop. aged 25-64 0.19 0.13 0.18 0.23 0.18 Marital status (%) (a) Single 14.2 9.8 11.1 14.4 12.8 Married 70.2 79.0 77.4 73.4 74.0 Divorced, widowed 15.6 11.2 11.5 12.2 13.2 Educational attainment (%) (b) None, lower or upper secondary 38.7 44.3 44.5 59.1 45.8 Tertiary 12.9 13.0 14.8 8.9 12.1 Nationality (%) (b) Foreign 18.7 20.4 14.2 10.1 16.5 2000 Population 2,145,800 2,117,604 1,066,367 1,953,695 7,283,466 Age structure Pop. aged 65-89 / Pop. aged 25-64 0.30 0.24 0.24 0.27 0.27 Marital status (%) (a) Single 15.8 10.3 9.8 10.6 12.0 Married 62.1 71.6 74.2 73.0 69.6 Divorced, widowed 22.1 18.1 16.0 16.4 18.5 Educational attainment (%) (b) None, lower or upper secondary 29.8 27.5 21.7 33.3 28.8 Tertiary 23.9 20.4 26.0 15.4 21.0 Nationality (%) (b) Foreign 26.0 21.6 14.0 12.0 19.3

Population characteristics according to Schuler et al.’s urban typology, Switzerland 1970 and 2000

(a) Population aged 30-89 years.
(b) Population aged 25-89 years.
Sources: 1970 and 2000 censuses.

26Multilevel survival models were estimated to investigate the structural and contextual effects of periurbanization on individual all-cause mortality. Individual exposure to the annual conditional risk of mortality starts at the time of the census in 2000 and ends with death or with truncation in November 2008 (or emigration for foreigners, who are linked to the register of foreign residents).

27The models were estimated separately for broad age groups (25-64 and 65-89) for two reasons. First, the cause-of-death structure of mortality differs by age. Second, the past migration flows that shaped the intra-urban differences in population composition in 2000 also selected different subpopulations at different ages (Wanner, 2005), with potentially contrasting effects on geographic differences in mortality. Movements of young families and workers can be expected to select more healthy populations, and thus to reduce (increase) period measures of mortality at destination (origin). Second, at older ages, mobility due to widowhood and increased frailty has an inverse effect. The models were also stratified by sex to account for women’s progress in the epidemiological transition, and for lower levels of responsiveness to contextual factors among women than among men (Pickett and Pearl, 2001).

28We used discrete-time random-intercept logistic regression to adjust standard errors of the covariates’ effects for the spatial clustering of individuals at the level of agglomerations and intra-urban zones. The variance of mortality is partitioned into variation between individuals within the spatial clusters (e0ij) on the one hand, and variation in average mortality between clusters (u0j) on the other. We obtain

30with individuals i nested in j clusters of municipalities, where x1ij and W1j are the effects of individual and contextual variables, respectively, and b0(t) is the baseline hazard of mortality.

31By step-wise adjustment of the model, this method also allows us to assess whether the geographic variation in mortality arose from socioeconomic differentiations in urban populations and living contexts. A first set of models only includes individual age and the cluster’s urban status (i.e. central, suburban, periurban, or rural), providing a non-standardized measure of the urban mortality gradient. In a second set of models, we control for differences in the population structure of the clusters by adjusting for the effects of individual socioeconomic characteristics. We then assess whether this standardization lowers the urban mortality gradient relative to that of the first models. Population structures are standardized for marital status (single, married, divorced, or widowed), nationality (Swiss, citizen of the European Union, other countries), and the level of educational attainment using broad categories of the International Standard Classification of Education (ISCED): none to lower secondary; upper secondary, which can be either on-the-job training or general school education; and tertiary. As shown in Table 1, it is city centres that have the highest rates of demographic ageing and of international immigration since 1970, as the significant increase in the proportion of older persons and foreigners indicates. The proportions of single, divorced, and low-skilled individuals are also higher in city centres than in other urban zones. These trends reflect the selective out-migration from city centres to the urban periphery since the 1960s. Accordingly, periurban areas have seen the strongest increase in the share of the population with tertiary education since 1970 (the same phenomenon is observed in city centres, but is partly explained by international immigration of increasingly qualified people). In 2000, a relatively high proportion of the people living in both periurban and rural areas were married. The populations in rural areas have been ageing, and generally have the lowest proportions of foreigners and of individuals with tertiary education.

32In a third set of models, contextual effects are also tested, providing an estimate of the urban mortality gradient that is standardized for both demographic structure and living context. The contextual variables are derived from the census of 2000 for the 207 clusters of municipalities. Inter-cluster material deprivation is estimated using the Townsend index (Townsend et al., 1988), i.e. an unweighted sum of standardized percentages (i.e. z-scores) of overcrowded private households (i.e. more than one person per room), non-owned private dwellings, unemployment rates, and the share of the population aged 25 and higher with no more than a lower secondary qualification. A high index value for a given cluster reflects a higher than average level of deprivation. The Townsend indicator is widely used and has been validated for Great Britain and Northern Ireland, at least among the working-age population (Gordon, 1995; O’Reilly, 2002). [6]

33In addition to differences in the level of deprivation between clusters, intra-cluster material inequality was estimated by an unconventional Gini index based on the cumulative distribution of local populations in terms of wealth, approximated here for each individual by an unweighted sum of the inverse attributes used for the Townsend index. [7] The Gini index gives a score ranging from zero to 100, with higher scores indicating a more unequal distribution of wealth in a given spatial cluster.

34To consider the diversity of family forms, we estimated inter-cluster differences in living arrangements using a weighted summary index of the proportion of non-conventional household configurations (i.e. unmarried and without children) among the adult population aged 30-49 (weights were estimated by factor analysis; see Hermann et al., 2005). The higher the spatial cluster score on this index, the higher the proportion of the population in this cluster that lives in a non-conventional household structure. It is worth noting that in European comparative perspective, Switzerland has a high rate of premarital cohabitation but a low rate of out-of-wedlock births, a contrast interpreted by researchers as reflecting a still highly dominant normative vision of what a family should be (Le Goff et al., 2005).

35Models were estimated in MLwiN version 3.32 using the Markov chain Monte Carlo method (MCMC; Browne, 2003; Rasbash et al., 2005). The relevance of structural and contextual factors for the geography of mortality in Switzerland was evaluated against the step-wise decrease in inter-cluster variance as the successive variables were introduced. The statistical significance of each factor in the explanation of individual mortality differentials was evaluated against the step-wise improvement of the model according to the Bayesian Information Criterion (BIC).

4 – Sensitivity tests

36We tested different functions of the effects of continuous contextual variables (linear, quadratic, and using five dummies for each quintile), and retained those functions that provided the best quality models, according to the BIC. Although the contextual indicators were found to be significantly correlated with each other (especially intra-cluster inequalities and inter-cluster differences in deprivation; r = 0.80), they measure different dimensions of the context that are important to consider when seeking to determine the cumulative effects of disadvantages. We adjusted the model through the step-wise inclusion of these variables and checked whether confounding effects were at work. When this was shown to be the case, the interaction effects between the two variables in question were tested. The outcome of this test indicated that these effects were either not statistically significant or did not improve the quality of the models (not shown). The robustness of our results to the definition of spatial clusters was also assessed by running the models on the urban population only, and by using a different spatial classification (based on the municipality rather than the urban zone). The results remained qualitatively the same (not shown).

37As inter-cluster migration during the follow-up period (2000-2008) was documented for deceased individuals only, we had to assume that the place of residence observed at the date of the census remained constant over time. This did not necessarily correspond to the place of death among the elderly who still lived independently in 2000, as the majority of deaths in Switzerland occur in nursing homes or hospitals (Hedinger et al., 2015). This is not problematic for our analysis because a certain time period has to elapse between exposure to the living context and its effect on mortality. However, the place of residence at the time of the census may not be the place where an individual spent most of his or her life. Migration between urban zones is the root cause of the spatial differentiation of population and living contexts that potentially influences the urban mortality gradient. Yet migration also biases our results, particularly at older ages, as mortality is endogenous to the in-migration of frail people into care institutions (Jonker et al., 2013; Kimbele and Janssen, 2013). Unfortunately, the census provides little information that can be used to address these issues (i.e. only the respondents’ place of residence five years prior to the enumeration is known). As a sensitivity test, we ran models that controlled for the most recent inter-cluster mobility by considering, for internal migrants (international immigrants were excluded), the cluster of residence in 1995 (rather than that in 2000), which might be the socioeconomic environment to which they had been exposed over a longer period. At older ages, we also excluded from the models the population living in nursing homes and hospitals.

III – The urban mortality gradient in Switzerland, 1969-2002

38Table 2 shows the trends in life expectancy at birth for each urban zone between 1970 and 2000 (see also Wanner et al. 2012). Over the three decades of observation, life expectancy in Switzerland increased by 7.0 years for men and by 6.6 years for women, reaching 77.3 years and 82.9 years, respectively.

39While the increase occurred in all the urban zones, the rate of increase varied, leading to a transformation of the urban mortality gradient. The geographic analysis by period, as well as the place-specific trends over the three decades, provide a similar picture for both men and women, although the patterns are more marked among men. In 1970, life expectancy in rural areas was around 1.5 years lower than in cities. Male life expectancy was similar in the central, suburban, and periurban municipalities (between 70.5 to 70.8 years). For women, a linear gradient congruent to that in the early twentieth century was observed, with the lowest mortality found in city centres (76.9 years).

40In 1970, the male suburban and periurban mortality advantage and the female advantage in city centres are explained by lower mortality at ages 40 and above. Decomposition reveals a life expectancy gain for suburban men (of 0.25 years at ages 40-64), periurban men (0.14 years) and women in city centres (0.53 years at ages 65 and over). Male excess mortality in rural areas stemmed from higher death rates among children and young adults (aged 20-39), which were mainly caused by childhood diseases and high rates of death from external causes, respectively. Young people’s rural-to-urban migration before 1970 may also have been selective, leaving behind members of the more disadvantaged social strata who tend to engage in more risky health behaviours. Among women, however, mortality at retirement ages explains the bulk of the rural disadvantage.

Table 2

Life expectancy at birth (e0) for men and women according to a functional urban classification, and age and cause-specific contributions (in years) to differences in e0 with reference to the national average, 1969-2002

Table 2
Men Women Total (Ref.) Centre Suburban Periurban Rural Range Total (Ref.) Centre Suburban Periurban Rural Range 1970 e0 70.3 70.5 70.8* 70.7 69.3* 1.4 76.3 76.9* 76.3 76.1 75.2* 1.7 Difference area category – total + 0.2 + 0.5 + 0.4 – 1.0 + 0.6 0 – 0.2 – 1.1 Age-specific decomposition 0-19 0.11 0.13 0.12 – 0.29 0.07 0.1 – 0.04 – 0.16 20-39 0.13 0.15 – 0.02 – 0.41 – 0.02 0.04 0.03 – 0.05 40-64 – 0.08 0.25 0.14 – 0.19 0.07 0.05 – 0.01 – 0.18 64+ 0.04 – 0.05 0.16 – 0.09 0.53 – 0.13 – 0.17 – 0.70 1980 e0 72.5 72.4 73.1* 73.0* 71.7* 1.4 79.1 79.2 79.4* 79.1 78.5* 0.9 1990 e0 74.3 73.6* 74.8* 75.3* 73.8* 1.5 81.1 80.8* 81.3 81.4 81.0 0.4 2000 e0 77.3 76.7* 77.8* 78.5* 76.8* 1.6 82.9 82.4* 83.3* 83.3* 82.9 0.4 Difference area category – total – 0.6 + 0.5 + 1.2 – 0.5 – 0.5 + 0.4 + 0.4 0.0 Age-specific decomposition 0-19 – 0.01 0.03 0.12 – 0.07 0 0.01 0.03 – 0.02 20-39 – 0.03 0.04 0.12 – 0.09 – 0.10 0.02 0.09 0.06 40-64 – 0.43 0.17 0.54 – 0.06 – 0.25 0.02 0.24 0.12 64+ – 0.16 0.28 0.37 – 0.28 – 0.12 0.33 0.02 – 0.18 Cause-specific decomposition Cancer – 0.13 0.07 0.29 – 0.10 – 0.10 0.04 0.02 0.08 Cardio-vascular 0 0.09 0.19 – 0.21 0.07 0.09 0.06 – 0.25 Respiratory system 0 0.07 0.10 – 0.12 – 0.02 0.03 0.03 – 0.02 External causes 0.02 0.14 0.23 – 0.29 – 0.09 0.03 0.10 0.03 Other – 0.52 0.15 0.35 0.22 – 0.33 0.19 0.17 0.13

Life expectancy at birth (e0) for men and women according to a functional urban classification, and age and cause-specific contributions (in years) to differences in e0 with reference to the national average, 1969-2002

Note: * Indicates a statistically significant difference at 95%.
Interpretation: Men in periurban areas live 1.2 years longer than the national average in 2000 (78.5 years versus 77.3) because they gain 0.12 years of life due to lower mortality at ages 0-19, 0.12 years at ages 20-39, 0.54 years at ages 40-64 and 0.37 years at ages 64+. In terms of causes of deaths, the differential in life expectancy results from 0.35 years gained because of lower mortality from residual causes, 0.23 years due to lower external cause mortality, 0.10 years due to lower mortality from diseases of the respiratory system, 0.19 years due to cardiovascular diseases and 0.29 years due to cancers.
Source: Author’s estimates based on vital statistics and population censuses.

41Between 1970 and 2000, rural municipalities caught up in terms of life expectancy, with increases of 7.5 years for men and 7.7 years for women. In periurban areas it increased by 7.8 years for men and by 7.2 years for women, compared with only 6.2 years for men and 5.5 years for women in city centres. By 2000, the mortality levels in city centres had become similar or slightly inferior to those in the countryside. Consequently, the intra-urban mortality gradient widened, especially between 1980 and 1990, when gains in life expectancy were twice as large in periurban areas as in city centres. This period corresponds to the peak phase of periurbanization, during which more affluent families from central areas were selected.

42Male life expectancy in 2000 was lowest in the city centres and rural areas, and highest in the periurban zone (76.7 versus 78.5 years). Suburban municipalities occupied an intermediate position. Among women, by contrast, inhabitants in both agglomeration belt areas had the highest life expectancy (83.3 years compared with 82.4 years in city centres). Although the nationwide urban gradients of 1.6 years for men and 0.4 years for women were statistically significant, they were small compared to the differences based on individual characteristics, such as education and marital status. Yet the non-linearity of the gradient mirrors the spatial differences in the socioeconomic compositions of populations (see Table 1).

43The age and cause-of-death patterns of the rural excess mortality in 2000 indicate a role for infrastructural factors (i.e. proximity to emergency services) and also suggest differences in risk-taking behaviours. Compared to mortality in urban areas, male and female mortality in rural areas was higher in the older population (aged 65 and over). Cardiovascular diseases, as well as more deaths from external causes and fewer deaths from residual causes (among men), were the main reasons for this differential.

44The higher male and female mortality levels in city centres in 2000 mainly concerned adults aged 40-64, and were attributable to diseases of the nervous and digestive systems, and mental and behavioural disorders (i.e. residual causes). This contrasts with the health advantage in the agglomeration belts, which also concerned adults aged 40 and over and stemmed from the same causes of death in addition to cancers (among men). This age- and cause-specific gradient suggests that differing lifestyles in city centres and periurban areas (i.e. urbanicity) play an important role.

45To explore whether the new intra-urban differentiation in mortality is related to the demographic process of periurbanization, we plotted for each agglomeration the intra-urban difference in male life expectancy at birth in 2000 against a simple measure of urban sprawl between 1970 and 2000; i.e. the average difference in the intercensal demographic growth rates of the agglomeration belt areas and of their central municipalities (Figure 2). To enhance the robustness of the results, we subtracted growth rate and life expectancy in the city centre from the estimates for the suburban and the periurban areas combined.

Figure 2

Correlation between the average differential in demographic growth of agglomeration belts versus city centres (1970-2000) and the intra-urban gradient in male life expectancy at birth (e0) in 2000, for Swiss cities with more than 50,000 inhabitants

Figure 2

Correlation between the average differential in demographic growth of agglomeration belts versus city centres (1970-2000) and the intra-urban gradient in male life expectancy at birth (e0) in 2000, for Swiss cities with more than 50,000 inhabitants

Note: Agglomeration belts group suburban and periurban municipalities, whereas centres are the central municipality of each agglomeration (see section on data and methods).
Statistical significance: * at 0.05% level.
Source: Author’s estimates based on vital statistics and population censuses.

46The correlation coefficient is positive and statistically significant at the 95% level (r = 0.69*). The cities that experienced the most intense periurbanization over the previous decades had a wider intra-city life expectancy differential in 2000. Essentially the same picture emerged when we tested the correlation between the agglomerations’ periurbanization and changes in intra-urban differentials in life expectancy over the period 1970-2000 (r = 0.58*); similar but less extreme patterns were found for women (not shown). Periurbanization and the intra-urban mortality gradient were more significant in the cities that are the main hubs in the Swiss urban system (Zürich, Basel, Bern, Lausanne and, to a lesser extent, Geneva), and in the cities that had recently undergone strong expansion (such as Fribourg, Bienne, Will, Winterthur and, to a lesser extent, Luzern, and Lugano; see Rérat et al. 2008) [8]. Differentials in male life expectancy in these urban agglomerations reached more than three years, which exceeds the recent inter-cantonal gradient.

47Further explorations at the agglomeration level also point to an association between the socioeconomic consequences of periurbanization and the intraurban mortality gradient. The differences in male life expectancy at birth in 2000 indeed correlate positively with the increasing concentration in agglomeration belts of married individuals (r = 0.59*) and of people with at least an upper secondary education (r = 0.41*); as estimated by the changes in the respective locational indexes between the 1970 and the 2000 censuses. [9] To avoid potential ecological biases (i.e. false inference of individual behaviours based on demographic changes observed at the population level) we assess in the next section whether this socioeconomic interpretation of the urban mortality gradient is confirmed in a multilevel analysis of individual-level data.

IV – Structural and contextual effects of periurbanization on the urban mortality gradient

48Tables 3 and 4 show the independent effects of differential socioeconomic structures and living contexts on the urban mortality gradient among the enumerated individuals in 2000, for whom we know the survival status in 2000-2008 (results appear in the form of odds ratios). The age-standardized urban mortality gradients among men and women of working age are in line with the aggregate estimates at birth for 1999-2002 (Table 3, Model 1). Introducing socioeconomic variables substantially improves the quality of the model (Model 2). The effects of these individual characteristics are found to be consistent with what we know about the socioeconomic mortality differentials in Switzerland. Single and divorced/widowed individuals have higher mortality levels, as do individuals with low educational levels. People who are married and better educated – especially those with a tertiary education – have lower mortality levels. Swiss nationals have higher mortality than foreigners from the European Union and from non-European countries especially. This gap can be explained by the selection of emigrants with more favourable health profiles and less risk aversion among population subgroups in the countries of origin (Zufferey, 2014).

Table 3

Contextual and individual factors of mortality by sex, odds ratios, adults aged 25-64, Switzerland 2000-2008

Table 3
Men Women M1 M2 M3 M1 M2 M3 Age 25-44 0.21* 0.19* 0.19* 0.20* 0.20* 0.20* 45-64 (Ref.) 1 1 1 1 1 1 Urban status Centre 1.11* 1.14* 1.04 1.21* 1.20* 1.08 Suburban 0.98 1.02 1.00 1.07* 1.08* 1.03 Periurban 0.88* 0.95 1.00 0.93* 0.98 1.01 Rural (Ref.) 1 1 1 1 1 1 Civil status Single 1.63* 1.63* 1.58* 1.58* Married (Ref.) 1 1 1 1 Divorced, widowed 1.80* 1.80* 1.68* 1.68* Nationality Swiss (Ref.) 1 1 1 1 EU 0.78* 0.78* 0.75* 0.74* Non-EU 0.55* 0.55* 0.56* 0.56* Educational attainment Lower secondary 1.38* 1.37* 1.44* 1.44* Upper secondary (Ref.) 1 1 1 1 Tertiary 0.66* 0.66* 0.82* 0.81* Regional context Inter-cluster deprivation 1.03* 1.00 Intra-cluster inequality 0.99 1.02 Family diversity (inter-cluster) 1.00 1.005* σ0.017* 0.010* 0.008* 0.004* 0.004* 0.003* j BIC 58,265 51,379 51,372 43,483 40,703 40,699 N person-years 14,725,121 14,714,796 N events 50,177 28,040

Contextual and individual factors of mortality by sex, odds ratios, adults aged 25-64, Switzerland 2000-2008

Note : σj = geographic variance of inter-cluster mortality.
Statistical significance: * at 0.05 level.
Source: Swiss National Cohort database (Census 2000 and vital statistics 2000-2008).

49These structural factors confound the urban mortality gradient (compare Model 2 with Model 1): the advantage in periurban areas disappears for both men and women, especially after controlling for differentials in educational attainment. However, the higher odds ratio of death in city centres remains almost unchanged.

50In line with observations in other countries, the mortality effects of social and material living context among adults aged 25-64 are found to be modest compared with individual socioeconomic determinants (Model 3). These contextual effects are also shown to be larger for males than for females, as revealed by the model’s quality criterion and the reduced spatial variance in mortality (compared with Model 2). Living in a deprived area is significantly associated with increased mortality among men, and controlling for this confounding effect accounts completely for the excess mortality among men in city centres (compare Model 3 with Model 2). The similar excess mortality among women, by contrast, is explained by a higher heterogeneity in family forms – although the low odds ratio calls for caution in interpreting this result. Intra-cluster inequality does not matter significantly for either sex.

51The analysis controlling for recent migration (by considering for migrants the cluster of residence in 1995) generally confirms these results (not shown). The only difference is in the unadjusted mortality advantage in periurban zones, which is significant but weaker for men and not significant for women. Thus, migration to the urban periphery indeed selects more healthy individuals.

52The results for the older population (Table 4) are less consistent with our hypotheses. The age-standardized urban mortality gradients are less marked. The male inhabitants of all urban zones have significantly lower mortality levels than their counterparts in rural areas (Model 1). However, male periurban residents have the largest survival advantage, and this advantage is also significant among women. Socioeconomic characteristics have the same effect on old-age mortality as on mortality at working ages (Model 2).

Table 4

Contextual and individual factors of mortality by sex, odds ratios, adults aged 65-89, Switzerland 2000-2008

Table 4
Men Women M1 M2 M3 M1 M2 M3 Age 65-74 0.28* 0.30* 0.30* 0.21* 0.25* 0.25* 75-89 (Ref.) 1 1 1 1 1 1 Urban status Centre 0.96* 1.03* 1.06* 0.98 1.01 1.09* Suburban 0.95* 1.00 1.00 0.99 1.01 1.02 Periurban 0.88* 0.95* 0.95* 0.91* 0.96* 0.93* Rural (Ref.) 1 1 1 1 1 1 Civil status Single 1.51* 1.51* 1.77* 1.76* Married (Ref.) 1 1 1 1 Divorced, widowed 1.48* 1.48* 1.73* 1.73* Nationality Swiss (Ref.) 1 1 1 1 EU 0.84* 0.84* 0.86* 0.86* Non-EU 0.62* 0.62* 0.71* 0.70* Educational attainment Lower secondary 1.23* 1.23* 1.21* 1.21* Upper secondary (Ref.) 1 1 1 1 Tertiary 0.79* 0.79* 0.83* 0.83* Regional context Inter-cluster deprivation 0.99 0.98* Intra-cluster inequality 1.01* 1.02* Family diversity (inter-cluster) 1.00 1.00 σ0.004* 0.002* 0.002* 0.004* 0.004* 0.003* j BIC 56,001 47,271 47,274 53,837 41,838 41,830 N person-years 3,714,686 5,083,403 N events 156,175 156,271

Contextual and individual factors of mortality by sex, odds ratios, adults aged 65-89, Switzerland 2000-2008

Note : σj = geographic variance of inter-cluster mortality.
Statistical significance: * at 0.05 level.
Source: Swiss National Cohort database (Census 2000 and vital statistics 2000-2008).

53Controlling for these structural effects (individual-level sociodeomographic variables) substantially increased the model’s quality. These effects also explain the lower male mortality levels in suburban areas, but do not entirely account for the periurban health advantage for both sexes: the associated odds ratios approaches but remains significantly inferior to unity (compare Model 2 with Model 1). Moreover, the lower mortality in city centres turns into excess mortality after controlling for differences in educational attainment. Better educated individuals with lower mortality levels are overrepresented among the older population in city centres. This can be explained by the historical centre-periphery diffusion of higher education in Switzerland.

54The main differences in the determinants of the urban gradient in old-age mortality, compared to the younger age group, pertain to the effects of contextual factors (Table 4, Model 3). First, for both sexes, mortality significantly increases with the level of inequality within spatial clusters. This effect is not confounded by local differences in diversity of family forms, which do not affect old-age mortality at all; or by colinearity with inter-cluster deprivation levels. Second, inter-cluster deprivation only matters significantly for women, and is associated with a lower (rather than a higher) mortality level. We do not have an explanation for this paradox. [10] Third, controlling for differences in the socioeconomic context among men does not explain the geographic variance in mortality, and does not improve the quality of the model (compare Model 3 with Model 2). Thus, socioeconomic context is not an important factor in old-age mortality for men. After controlling for differences in living contexts among women, the model improves and the geographic variance in mortality decreases. Yet the urban gradient widens significantly: older women with identical structural characteristics and living in similar socioeconomic contexts are particularly vulnerable in city centres, and are most protected in periurban areas.

55The sensitivity analysis that controls for migration since 1995 essentially provides the same results (not shown). Only one qualitative difference is observed when we exclude the population in nursing homes and hospitals: intra-cluster inequality no longer significantly increases female mortality (not shown). Thus, old-age mobility into care institutions appears to be particularly detrimental to survival when these institutions are located in areas characterized by strong socioeconomic inequalities.

V – Discussion and conclusion

56Given the world-wide decline in birth rates, the geography of mortality is having ever greater effects on urban demography. Based on censuses and vital statistics, we described trends in the mortality gradient between 1969 and 2002 using consistent spatial delineations of agglomerations and urban zones in Switzerland. Relying on exhaustive individual-level follow-up mortality data, we then investigated whether the urban gradient in 2000-2008 can be explained by the socioeconomic consequences of periurbanization.

57Differentials in levels of life expectancy across the urban-rural continuum have not substantially widened since 1969, as mortality has also been declining in rural areas. However, an intra-urban differentiation has appeared since the 1980s, when periurbanization became the dominant form of population dispersal in Swiss cities. The geography of mortality went from showing a consistent urban advantage over rural areas in the early twentieth century to displaying a non-linear gradient across the urban-rural continuum in the 2000s. Our findings indicate that mortality is higher in city centres and in the countryside than in urban agglomeration belt areas, in line with results for Belgium and Canada (Eggerickx and Sanderson, 2010; Ostry, 2009). These findings contradict the environmental hypothesis which expects mortality to be lower in rural areas.

58The excess mortality in rural areas is concentrated among the elderly population, especially in 2000; and is driven mainly by deaths from external causes and cardiovascular diseases. These causes suggest a role for proximity to emergency services (see also Lopez-Rios et al., 1992). Cardiovascular diseases may also be related to more risky health behaviours. The mortality gradient between periurban areas and city centres, by contrast, is largest among men, and stems mainly from differential death rates at ages 40-64. It can be attributed to diseases of the nervous and digestive systems, mental and behavioural disorders, and cancers. These findings suggest the existence of specific lifestyles among the population of city centres compared to the inhabitants of periurban areas. Our argument that this differentiation in urbanicity is related to selective out-migration from the centres to the urban periphery is backed up by our finding that the mortality gradients are larger in the agglomerations with higher levels of urban sprawl and a spatial redistribution of the population by socioeconomic characteristics.

59The role of these socioeconomic consequences of periurbanization is confirmed in the multilevel analysis based on individual-level data, and entirely accounts for the urban gradient among adults of working ages in 2000-2008. The mortality advantage in periurban areas is related to the spatial concentration of highly educated people and of families. We also find evidence for a positive health selection of migrants to the urban periphery. Excess mortality in the city centres, by contrast, is attributable to local deprivation among men and (to a minor extent) to a higher level of diversity in family forms among women.

60The living context is found to be more important for men’s than for women’s mortality levels. Moreover, the gender-specific contextual effects we found are in line with previous findings for Switzerland, which showed that income affects men’s health to a greater extent than women’s health, while social dimensions are more important for the physical and mental well-being of women than of men (Burton-Jeangros, 2009). The stronger link between mortality and the social (rather than the economic) context among women is consistent with evidence that women have reached a more advanced stage in the epidemiological transition than men. Another difference between the female and male mortality gradients is that life expectancy levels among women are not lower in the suburbs than in the periurban areas. This may indicate a negative effect among working-class men, who may, in the past, have been more exposed than women to workplace health risks in suburban industries.

61These results suggest a role for the socioeconomic composition of populations and living contexts in determining urban differences in health behaviours which ultimately led to the observed mortality gradient. The absence of urbanrural differences in current health levels, as observed in Swiss surveys (Heeb et al., 2011), does not necessarily challenge this interpretation. Undocumented intra-urban differentials in behaviours (mirroring the non-linear mortality gradient shown here) may blur the dichotomous comparison between urban and rural areas. Moreover, current mortality is determined by behaviours of cohorts in previous decades, which have not been surveyed.

62Results for the older population point to a more complex set of causes of the urban mortality gradient. The relevance of the socioeconomic context for mortality among older women, but not among men, may be explained by the higher prevalence of widowhood among the former. The loss of the husband’s emotional support may inflate the role of context. [11] As women live longer than men, they may also have been less selected at older ages and, thus, constitute a more heterogeneous group, more prone to contextual influences. Although sensitivity analyses confirmed our finding that female mortality in more deprived areas is surprisingly low, potential biases related to migration before 1995 cannot be excluded. An older person’s place of residence may not be representative of the living context in which he or she spent most of his or her life, as the person may have migrated in retirement or in response to increased frailty. Alternatively, the Townsend index may not accurately reflect the material living context of older adults, as has been found in Northern Ireland (O’Reilly, 2002). Finally, the most unexpected result among the elderly is that controlling for the spatial differentiation of population and living contexts does not substantially account for their urban mortality gradient. Thus, the gradient may be related to factors other than that of urban living.

63Differences in health status and behaviour that are not consistent with socioeconomic status could explain mortality differences at older ages, as has been shown for the United States (House et al., 2000). The residual excess mortality in city centres and the advantage in periurban areas may also indicate that environmental effects of urbanization contribute to these differences, since people who are older and chronically ill tend to be especially vulnerable to environmental effects, such as heat (Grize et al., 2005). In conclusion, our results for old-age mortality raise more questions than they provide answers, and suggest that there is a need for more research on the vulnerability of older populations in densely built environments.

64At the same time, the results for the working-age population clearly indicate that periurbanization has played a role in the historical re-emergence of urban excess mortality in advanced phases of urbanization. Although longevity was shown to vary within some agglomerations to a greater extent than it did between the Swiss cantons in 2000, the underlying factors are not the same as those in the early phases of the epidemiological transition. Environmental stress related to overpopulation was the primary cause of excess mortality in urban areas in the past, whereas the recent disadvantage mainly stems from the spatial differentiation of urban living.

65The constantly changing face of urbanization does indeed matter. Large metropolitan centres in the United States and western Europe have recently experienced a demographic revival that can be attributed to various factors, including the upgrading of poor neighbourhoods through the in-migration of upper-class populations (i.e. gentrification), delays in life course transitions, and the spatial concentration of a highly specialized service economy (Buzar et al., 2007; Guest and Brown, 2005; Kabisch and Haase, 2011). In line with these trends, urban gradients in mortality have even been reversed in the United States, i.e. mortality levels are now lower in metropolitan centres than in agglomeration belt areas or smaller cities (Cosby et al., 2008; Singh et al., 2012). Given the recent (albeit limited) demographic evidence of an urban revival in Switzerland (Rérat, 2012), the urban mortality gradient in the country may be expected to change in the future because of the structural and contextual effects of population mobility. The challenge for future research is to provide a better understanding of the independent effects of migration flows, looking at both internal migration and international mobility, which select population subgroups with different levels of health, depending on age. Gaining greater insight into the complexity of these mobility trends could make it possible to design spatially focused preventive measures.


We are grateful for the support from IP213 of the Swiss Centre of Competence in Research LIVES-Overcoming Vulnerability: Life Course Perspectives, financed by the Swiss National Science Foundation. We also would like to thank the Swiss Federal Statistical Office and the Swiss National Cohort for access to the data.


  • [*]
    Max Planck Institute for Demographic Research.
    Correspondence: Mathias Lerch, Max Planck Institute for Demographic Research, Konrad-Zuse-Str. 1, 18057 Rostock, Germany, e-mail:
  • [**]
    University of Geneva.
  • [1]
  • [2]
    Epidemiologists and environmentalists coined the term “urban heat island effect” about half a century ago. Dark-coloured built environments absorb more heat than green areas, cool down to a lesser extent at night, and high-rise buildings lower the speed of winds that refresh urban air. Moreover, the concentration of human activity exacerbates meteorological phenomena through the effects of air pollution and heat wastage associated with high energy use (Anderson and Bell, 2011).
  • [3]
    An extreme illustration is provided by Liège, the largest city in the French-speaking part of Belgium. According to fiscal statistics, in 1977, the ratio of tax on household incomes to population size ranked the city as the sixth wealthiest municipality of the province. But 20 years later, Liège had fallen back to 62nd position. In the meantime, the city became bankrupt and was forced to close many institutions (see Eggerickx et al., 2002).
  • [4]
    A test showed that this did not change our results.
  • [5]
    Suburban municipalities have a population of at least 500, a proportion of high-rise buildings of at least 40% and a ratio of individuals actively employed within the municipality relative to all actively employed residents of at least 75% (Schuler et al. 2005).
  • [6]
    In Switzerland, an alternative indicator of people’s socioeconomic environment has recently been developed based on neighbourhood areas which are centred on individuals, and considers rents rather than home ownership (Panczak et al., 2012). However, this spatial approach does not allow us to decompose the variation of mortality at the contextual and individual levels (because spatial areas are not shared by the individuals). To ensure the international comparability of our results, we therefore kept the home ownership variable.
  • [7]
    Each individual had a score of one if he or she was living in a home that was not overcrowded or that he/she owned, was employed or inactive, and had an upper secondary educational level at least. The Gini index measures the area between the Lorenz curve formed by plotting the cumulative percentages of wealth against the cumulative number of recipients (starting with the poorest individual), on the one hand, and the hypothetical straight line assuming perfect equality, on the other.
  • [8]
    The small city of Zug is an outlier because of its specific population composition. Zug’s fiscal advantages have led many international firms to establish their administrative headquarters in the city, so it is inhabited by large numbers of wealthy people. Life expectancy in the city centre was higher than in the agglomeration belt in 1970, but the differential has been narrowing since then.
  • [9]
    The locational index for the agglomeration belt areas is obtained by dividing the shares of the married/higher educated population by the same indicators computed for the whole urban agglomeration.
  • [10]
    This may be due to the time that has to elapse between the exposure to physiological effects and death at older ages. We therefore tested whether relative area-level deprivation in 1970, 1980, or 1990 had an independent negative impact on mortality, or whether the upward or downward trends in the rankings of the areas on this dimension over the last 30 years affected the phenomenon. Neither of these two hypotheses could be confirmed (not shown).
  • [11]
    The loss of the economic support could also be considered. But while widows’ poverty was a topic of public debate and empirically measured in the late twentieth century in Switzerland, this relation disappeared in the early twenty-first century following a reform of the pension system that benefited widows (Gabriel et al., 2015).

While regional differences in life expectancy have flattened out in Switzerland, we investigate the effect of periurbanization on the geography of mortality. Using data from vital statistics and censuses, we find an increasing intra-urban differentiation of mortality since 1980, especially in the largest and most recently sprawling cities. A non-linear gradient, in which life expectancy is lower in city centres and rural areas than in urban agglomeration belts, has emerged. Age- and cause-specific mortality profiles suggest that lifestyles specific to the population of the city centres and related to the spatial concentration of disadvantaged groups play a dominant role in shaping this pattern. Considering mortality at ages 20-64, a multilevel model applied to census-linked mortality data shows how the mortality advantage observed in periurban areas can be explained by a concentration of highly educated individuals and of families. Excess mortality at ages 20-64 in city centres, by contrast, arises from more deprived material and social living environments. However, these socioeconomic consequences of periurbanization fail to account for the urban mortality gradient observed among older people.


  • urban mortality
  • urbanization
  • periurbanization
  • urban living
  • multilevel analysis
  • Switzerland

Périurbanisation et transformation du gradient de la mortalité urbaine en Suisse

Alors que les différences régionales d’espérance de vie se sont estompées en Suisse, quels sont les effets de la périurbanisation sur la géographie de la mortalité ? À partir des données de l’état civil et des recensements, on a pu observer un accroissement des différentiels intra-urbains de mortalité depuis 1980, en particulier dans les villes les plus grandes ou qui se sont récemment étendues. Un gradient non linéaire émerge : l’espérance de vie est plus faible dans le centre des villes et les zones rurales que dans la ceinture des agglomérations urbaines. Les profils de mortalité par âge et cause suggèrent que cela tient à la fois aux styles de vie propres aux populations des centres-villes et à la concentration spatiale des groupes défavorisés. Pour la mortalité entre 20 et 64 ans, un modèle multiniveau appliqué à des données de mortalité couplées aux recensements montre que la moindre mortalité observée dans les zones périurbaines résulte de la concentration d’individus très instruits et de familles. À l’inverse, la surmortalité des 20-64 ans dans les centres-villes reflète des désavantages matériels et sociaux. Cependant, ces conséquences socioéconomiques de la périurbanisation ne suffisent pas à rendre compte du gradient de la mortalité urbaine observé chez les personnes âgées.


Peri-urbanización y transformación del gradiente de la mortalidad urbana en Suiza

Mientras que las diferencias regionales de mortalidad en Suiza han prácticamente desaparecido ¿qué efectos produce la peri-urbanización sobre la geografía de la mortalidad? A partir de datos del estado civil y de los censos, se observa un aumento de los diferenciales intra-urbanos de mortalidad desde 1980, particularmente en las ciudades más grandes o en las que han crecido recientemente. El gradiente que aparece no es linear: la esperanza de vida es más baja en el centro de las ciudades y en las zonas rurales que en la cintura de las aglomeraciones urbanas. Los perfiles de mortalidad por edad y causa sugieren que este fenómeno se debe tanto a los estilos de vida propios a los residentes del centro de las ciudades como a la concentración espacial de los grupos desfavorecidos. Un modelo multinivel aplicado a los datos de la mortalidad a 20-64 años asociados a los censos, muestra que la menor mortalidad observada en las zonas periurbanas se debe a la concentración de individuos muy instruidos y de familias. Por el contrario, la mortalidad excesiva de los 20-64 anos en el centro de las ciudades refleja desventajas materiales y sociales. Sin embargo, las consecuencias socioeconómicas de la periurbanización no son suficientes para explicar el gradiente de la mortalidad urbana observado en las personas mayores.


  • OnlineAnderson G. Brooke, Bell Michelle L., 2011, “Heat waves in the United States: Mortality risk during heat waves and effect modification by heat wave characteristics in 43 U.S. communities”, Environmental Health Perspectives, 119(2), pp. 210-218.
  • OnlineArriaga Eduardo, 1984, “Measuring and explaining the change in life expectancies”, Demography, 21(1), pp. 83-96.
  • Arriaga Eduardo, 1989, “Changing trends in mortality decline during the last decades”, in Ruzicka L., Wunsch G., Kane P., Differential Mortality: Methodological Issues and Biosocial Factors, Oxford, Clarendon Press, 259 p.
  • BFS, 2007, Historisches Gemeineverzeichnis der Schweiz, Neuchâtel, BFS/OFS.
  • Bopp Matthias, Gutzwiler Felix, 1999, “Entwicklung der Mortalität in der Schweiz seit 1950 - II. Regionale Unterschiede innerhalb der Schweiz”, Schweizerische Medizinische Wochenschrift, 129, pp. 799-809.
  • OnlineBopp Matthias, Spoerri Adrian, Zwahlen Marcel, Gutzwiller Felix, Paccaud Fred et al., 2008, “Cohort profile: The Swiss National Cohort-a longitudinal study of 6.8 million people”, International Journal of Epidemiology, 38(2), pp. 1-6.
  • Browne William J., 2003, MCMC Estimation in MLwiN, Bristol, Centre of Multilevel Modelling, Universty of Bristol.
  • Burton-Jeangros Claudine, 2009, “Les inégalités face à la santé : l’impact des trajectoires familiales et professionelles sur les hommes et les femmes”, in Oris M., Widmer E., de Ribaupiere A., Joye D., Spini D. et al., Transitions dans les parcours de vie et construction des inégalités, Lausanne, Presse polytechniques et universitaires romandes, 464 p.
  • OnlineBuzar Stefan, Ogden Philip, Hall Ray, Haase Annegret, Kabisch Sigrun, Steinführer Annet, 2007, “Splintering urban populations: Emergent landscapes of reurbanisation in four European cities”, Urban Studies, 44(4), pp. 651-677.
  • Caselli Graziella, Vallin Jacques, 2006, “Geographic variations of mortality”, in Caselli G., Vallin J., Wunsch G., Demography: Analysis and Synthesis, Burlington, Elsevier, 2976 p.
  • OnlineChaix Basile, Rosvall Maria, Lynch John, Merlo Juan, 2006, “Disentangling contextual effects on cause-specific mortality in a longitudinal 23-year follow-up study: Impact of population density or socioeconomic environment?”, International Journal of Epidemiology, 35(3), pp. 633-643.
  • Champion Anthony G., 1989, Counterurbanization: The Changing Pace and Nature of Population Deconcentration, London, Edward Arnold, 266 p.
  • Chiang Chin Long, 1984, The Life Table and Its Applications, Malabar, Robert E. Krieger Publishing Company, 336 p.
  • OnlineChoinière Robert, 1991, “Les disparités géographiques de la mortalité dans le Montréal métropolitain, 1984-1988 : étude écologique des liens avec les conditions sociales, économiques et culturelles”, Cahiers québécois de démographie, 20(1), pp. 115-144.
  • OnlineCosby Arthur G., Neaves Tonya T., Cossman Ronald E., Cossman Jeralynn S., James Wesley L. et al., 2008, “Preliminary evidence for an emerging nonmetropolitan mortality penalty in the United States”, American Journal of Public Health, 98(8), pp. 1470-1472.
  • Cunha Antonio Da, Both Jean-François, 2004, “Métropolisation, villes et agglomérations. Structure et dynamiques sociodémographiques des espaces urbains”, OFS, Recensement fédéral de la population 2000, Neuchâtel, Office fédéral de la statistique.
  • OnlineCyril Sheila, Oldroyd John C., Renzaho Andre, 2013, “Urbanisation, urbanicity, and health: A systematic review of the reliability and validity of urbanicity scales”, BMC Public Health, 13(513).
  • Eggerickx Thierry, Sanderson Jean-Paul, 2010, “Les inégalités spatiales de mortalité en Belgique : 1980-2005”, Actes du colloque de la Cudep, Démographie et santé, Bordeaux, CUDEP.
  • Eggerickx Thierry, Capron Catherine, Hernia Jean-Pierre, Oris Michel, 2002, Démographie et développement durable. Migrations et fractures socio-démographiques en Wallonie (1990-2000), Liège-Louvain la Neuve, Laboratoire de Démographie – GéDAP – SSTC, 240 p.
  • OnlineFan Yingling, Song Yan, 2009, “Is sprawl associated with a widening urban-suburban mortality gap?”, Journal of Urban Health: Bulletin of the New York Academy of Medicine, 86(5), pp. 708-728.
  • OnlineFei Peng, Wanner Philippe, Cotter Stéphane, 1998, “Spécificités de la mortalité et de son évolution dans les cantons entre 1920/21 et 1988/93”, Geographica Helvetica, 1, pp. 3-16.
  • Frumkin Howard, Frank Lawrence, Jackson Richard, 2004, Urban Sprawl and Public Health: Designing, Planning, and Building for Healthy Communities, Washington, Island Press, 368 p.
  • Gabriel Rainer, Oris Michel, Studer Mathias, Baeriswyl Marie, 2015, “The persistence of social stratification? A life course perspective on old-age poverty in Switzerland”, Swiss Journal of Sociology, 41(3), pp. 465-487.
  • OnlineGartner Andrea, Farewell Daniel, Roach Paul, Dunstan Krank, 2011, “Rural/ urban mortality differences in England and Wales and the effect of deprivation adjustment”, Social Science & Medicine, 72(10), pp. 1685-1694.
  • OnlineGeyer Hermanus S., Kontuly Thomas, 1993, “A theoretical foundation for the concept of differential urbanization”, International Regional Science Review, 15(2), pp. 157-177.
  • OnlineGordon David, 1995, “Census based deprivation indices: Their weighting and validation”, Journal of Epidemiology and Community Health, 49(Suppl 2), pp. S39-S44.
  • Grize Leticia, Huss Anke, Thommen Oliver, Schindler Christian, BraunFahrländer Charlotte, 2005, “Heat wave 2003 and mortality in Switzerland”, Swiss Medical Weekly, 135(13-14), pp. 200-205.
  • Guest Avery M., Brown Susan K., 2005, “Population distribution and suburbanization”, in Poston D. L., Micklin M., Handbook of Population, New York, Springer, 918 p.
  • OnlineHedinger Damian, Braun Julia, Zellweger Ueli, Kaplan Vladimir, Bopp Matthias, 2015, “Moving to and dying in a nursing home depends not only on health. An analysis of socio-demographic determinants of place of death in Switzerland”, Plos One, 9(11), pp. e113236.
  • Heeb Jean-Luc, Inglin Sophie, Moreau-Gruet Florence, Schuler Daniela, Vilpert Sarah et al., 2011, “La santé en Suisse latine. Analyse des données de l’Enquête suisse sur la santé 2007”, Neuchâtel, Observatoire suisse de la santé, Osban Rapport, 156 p.
  • Hermann Michael, Heye Corinna, Leuthold Heiri, 2005, Sozikulturelle Unterschiede in der Schweiz: Vier Indizes zu raumlichen Disparitäten, 1990-2000, Neuchâtel, Office fédéral de la Statistique.
  • OnlineHouse James S., Lepkowski James M., Williams David R., Mero Richard P., Lantz Paula M. et al., 2000, “Excess mortality among urban residents: How much, for whom, and why?”, American Journal of Public Health, 90(12), pp. 1898-1904.
  • OnlineHuss Anke, Spoerri Adrian, Egger Matthias, Röösli Martin, 2010, “Aircraft noise, air pollution and mortality from myocardial infarction”, Epidemiology, 21(6), pp. 829-836.
  • OnlineJen Min Hua, Jones Kelvyn, Johnston Ron, 2009, “Compositional and contextual approaches to the study of health behaviour and outcomes: Using multi-level modelling to evaluate Wilkinson’s income inequality hypothesis”, Health & Place, 15(1), pp. 198-203.
  • OnlineJonker Marcel. F., van Lenthe Frank J., Donkers Bas, Congdon Peter D., Burdorf Alex, Mackenbach Johan P., 2013, “The impact of nursing homes on small-area life expectancies”, Health & Place, 19(1), pp. 25-32.
  • OnlineKabisch Nadja, Haase Dagmar, 2011, “Diversifying European agglomerations: Evidence of urban population trends for the 21st century”, Population, Space and Place, 17(3), pp. 236-253.
  • OnlineKimbele Eva, Janssen Fanny, 2013, “Distortion of regional old-age mortality due to late-life migration in the Netherlands?”, Demographic Research, 29(5), pp. 105-132.
  • Kohli Raymond, 2007, “Sterblichkeit nach Todesursachen, 1998/2003”, Neuchatel, OFS, Informationen aus der Demografie, 3.
  • Le Goff Jean-Marie, Sauvain-Dugerdil Claudine, Rossier Clémentine, Coenen Huther Josette, 2005, Maternité et parcours de vie. L’enfant a-t-il toujours une place dans les projets des femmes en Suisse ?, Berne, Peter Lang, 348 p.
  • OnlineLopez-Rios Olga, Mompart Ana, Wunsch Guillaume, 1992, “Système de soins et mortalité régionale : une analyse causale”, European Journal of Population, 8(4), pp. 363-379.
  • OnlineMacintyre Sally, Ellaway Anne, Cummins Steven, 2002, “Place effects on health: How can we conceptualise, operationalise and measure them?”, Social Science & Medicine, 55(1), pp. 125-139.
  • OnlineMaguire Aideen, O’Reilly Dermot, 2015, “Does conurbation affect the risk of poor mental health? A population based record linkage study”, Health & Place, 34(1), pp. 126-134.
  • OnlineMartikainen Pekka, Kauppinen Timo M., Valkonen Tapani, 2003, “Effects of the characteristics of neighbourhoods and the characteristics of people on cause specific mortality: A register based follow up study of 252 000 men”, Journal of Epidemiology and Community Health, 57(3), pp. 210-217.
  • OnlineMoser André, Panczak Radoslaw, Zwahlen Marcel, Clough-Gorr Kerri M., Spoerri Adrian et al., 2014, “What does your neighbourhood say about you? A study of life expectancy in 1.3 million Swiss neighbourhoods”, Journal of Epidemiology and Community Health, 68(12), pp. 1125-1132.
  • OnlineO’Reilly Dermot, 2002, “Standard indicators of deprivation: Do they disadvantage older people?”, Age and Ageing, 31(3), pp. 197-202.
  • OnlineO’Reilly Gareth, O’Reilly Dermot, Rosato Michael, Connolly Sheelah, 2007, “Urban and rural variations in morbidity and mortality in Northern Ireland”, BMC Public Health, 7(123), pp. 1-6.
  • OnlineOmran Abdel R., 1971, “The epidemiologic transition: A theory of the epidemiology of population change”, The Milbank Memorial Fund Quarterly, 49(4), pp. 509-538.
  • OnlineOstry Aleck S., 2009, “The mortality gap between urban and rural Canadians: A gendered analysis”, Rural and Remote Health, 9(4), pp. 1286-1296.
  • OnlinePanczak Radoslaw, Galobardes Bruna, Voorpostel Marieke, Spoerri Adrian, Zwahlen Marcel, Egger Matthias, 2012, “A Swiss neighbourhood index of socioeconomic position: Development and association with mortality”, Journal of Epidemiology and Community Health, 66(12), pp. 1129-1136.
  • OnlinePearce Jamie, Boyle Paul, 2005, “Is the urban excess in lung cancer in Scotland explained by patterns of smoking?”, Social Science & Medicine, 60(12), pp. 2833-2843.
  • OnlinePickett Kate E., Pearl Michelle, 2001, “Multilevel analyses of neighbourhood socioeconomic context and health outcomes: A critical review”, Journal of Epidemiology and Community Health, 55(2), pp. 111-122.
  • Ramiro-Fariñas Diego, Oris Michel (eds.), 2016, New Approaches to Death in Cities during the Health Transition, New York, Springer, 241 p.
  • Ranci Costanzo, 2010, Social Vulnerability in Europe: The New Configuration of Social Risks, Basingstoke, Palgrave Macmillan, 299 p.
  • Rasbash Jon, Steele Fiona, Browne William, Proser Bob, 2005, A User’s Guide to MLwin, Bristol, Center for Multilevel Modelling, University of Bristol, 306 p.
  • OnlineRérat Patrick, 2012, “The new demographic growth of cities. The case of reurbanisation in Switzerland”, Urban Studies, 49(5), pp. 1107-1125.
  • Rérat Patrick, Piguet Etienne, Söderström Ola, Besson Roger, 2008, “‘Back to the city?’ Études de l’évolution démographique et l’attractivité résidentielle des villes suisses”, Neuchâtel, Institut de Géographie, Université de Neuchâtel, PNR 54 Développement durable de l’environnement construit.
  • Schuler Martin, Dessemontet Pierre, Joye Dominique, 2005, Les niveaux géographiques de la Suisse, Neuchâtel, Office fédéral de la statistique, 232 p.
  • OnlineSchumacher Reto, Vilpert Sarah, 2011, “Gender differences in social mortality differentials in Switzerland”, Demographic Research, 25(8), pp. 285-310.
  • OnlineSingh Gobal K., Siahpush Mohammad, Williams Shanita D., 2012, “Changing urbanization patterns in US lung cancer mortality, 1950-2007”, Journal of Community Health, 37(2), pp. 412-420.
  • OnlineSingh Gopal K., Williams Shanita D., Siahpush Mohammad, Mulhollen Aaron, 2011, “Socioeconomic, rural-urban, and racial inequalities in US cancer mortality: Part I: All cancers and lung cancer and Part II: Colorectal, prostate, breast, and cervical cancers”, Journal of Cancer Epidemiology, 107497, 27 p.
  • Spoerri Adrian, Zwahlen Marcel, Egger Matthias, Gutzwiler Felix, Minder Christoph, Bopp Matthias, 2006, “Educational inequalities in life expectancy in the German speaking part of Switzerland between 1990 and 1997: Swiss National Cohort”, Swiss Medical Weekly, 136(9-10), pp. 145-148.
  • OnlineSubramanian Subu V., Kawachi Ichiro, 2004, “Income inequality and health: What have we learned so far?”, Epidemiologic Reviews, 26(1), pp. 78-91.
  • Townsend Peter, Phillimore Peter, Beattie Alastair, 1988, Health and Deprivation: Inequality and the North, New York, Croom Helm, Social Science, 211 p.
  • Van de Kaa Dirk J., 1987, “Europe’s second demographic transition”, Population Bulletin, 42(1), pp. 1-59.
  • OnlineVerheij Robert A., van de Mheen H. Dike, de Bakker Dinny H., Goenewegen Peter P., Mackenbach Johan P., 1998, “Urban-rural variations in health in the Netherlands: Does selective migration play a part?”, Journal of Epidemiology and Community Health, 52(8), pp. 487-493.
  • Vlahov David, Galea Sandro, 2002, “Urbanization, urbanicity, and health”, Journal of Urban Health: Bulletin of the New York Academy of Medicine, 79(4), pp. S1-S12.
  • Vries Jan de, 1990, “Problems in the measurement, description, and analysis of historical urbanization”, in van der Woude A., Hayami A., De Vries J. (eds.), Urbanization in History: a Process of Dynamic Interaction, New York, Oxford University Press, 392 p.
  • Wanner Philippe, 2005, “Mobilité et migration”, in Wanner P., Sauvain-Dugerdil C., Guillley E., Hussy C., Âges et générations, Neuchatel, Office fédéral de la statistique.
  • OnlineWanner Philippe, Fei Peng, Cotter Stéphane, 1997, “Mortalité par âge et cause de décès en Suisse : une analyse des disparités cantonales durant la période 1978/83 à 1988/93”, European Journal of Population, 13(4), pp. 381-399.
  • Wanner Philippe, Lerch Mathias, Kohli Raymond, 2012, “La géographie de la mortalité en Suisse depuis 1970”, Neuchâtel, Office Fédéral de la Statistique, 44 p.
  • Wilkinson Richard G., 1996, Unhealthy Societies - The Afflictions of Inequalities, New York, Routledge, 268 p.
  • Zufferey Jonathan, 2014, “Pourquoi les migrants vivent-ils plus longtemps ? Analyse des différentiels de mortalité des populations suisse et étrangères, 1990-2008”, Université de Genève, PhD thesis.
Mathias Lerch [*]
  • [*]
    Max Planck Institute for Demographic Research.
    Correspondence: Mathias Lerch, Max Planck Institute for Demographic Research, Konrad-Zuse-Str. 1, 18057 Rostock, Germany, e-mail:
Michel Oris [**]
  • [**]
    University of Geneva.
Philippe Wanner [**]
For the Swiss National Cohort
  • [**]
    University of Geneva.
Latest publication on cairn or another partner portal
Uploaded on on 30/05/2017
Distribution électronique pour I.N.E.D © I.N.E.D. Tous droits réservés pour tous pays. Il est interdit, sauf accord préalable et écrit de l’éditeur, de reproduire (notamment par photocopie) partiellement ou totalement le présent article, de le stocker dans une banque de données ou de le communiquer au public sous quelque forme et de quelque manière que ce soit.
Loading... Please wait