Portal de Periódicos da CAPES

Urban–rural population changes and spatial inequalities in Sweden

2022; Elsevier BV; Volume: 15; Issue: 4 Linguagem: Inglês

10.1111/rsp3.12602

1757-7802

Martin Henning, Hans Westlund, Kerstin Enflo,

Urban, Neighborhood, and Segregation Studies

This paper documents regional population changes in Sweden since 1860 and investigates how these changes link to regional economic development (regional GDP). We combine long-term decade population data for the historical counties (1860–2020) with detailed annual population observations for municipalities (1968–2021). As industrialization picked up speed, this benefited regions all around the country in terms of production, at the same time as regional population patterns started to diverge. After a slowdown in the regional GDP convergence processes during the low-growth period of the 1980s, 'double divergence,' in both population and regional GDP per capita, has characterized Swedish growth patterns since the 1990s. Este artículo documenta los cambios regionales de población en Suecia desde 1860 e investiga cómo se relacionan estos cambios con el desarrollo económico regional (PIB regional). Se combinan datos de población de muchas décadas para los condados históricos (1860–2020) con observaciones detalladas de población anual para los municipios (1968–2021). A medida que se aceleraba la industrialización, esta beneficiaba a las regiones de todo el país en términos de producción, al mismo tiempo que los patrones regionales de población empezaban a divergir. Tras una ralentización de los procesos de convergencia del PIB regional durante el periodo de bajo crecimiento de los años ochenta, la 'doble divergencia', tanto en población como en PIB regional per cápita, ha caracterizado los patrones de crecimiento suecos desde la década de 1990. 本稿では1860年以降のスウェーデンにおける地域人口の変化を記録し、この変化が地域の経済発展(地域のGDP)とどのように関連しているかを検討する。我々は、歴史的な郡の長期的な10年単位の人口データ(1860~2020年)と市町村の毎年の詳細な人口観測(1968~2021年)を組み合わせた。工業化は生産という点で全国の地域に利益をもたらしたが、それと同時に工業化が急速に進むにつれ、地域の人口パターンの多様化が始まった。1980年代の低成長期に地域のGDPの収斂の過程が減速した後、人口と地域の一人当たりGDPの両方における「二重の多様化」が、1990年代以降のスウェーデンの成長パターンの特徴となっている。 This paper documents regional population changes in Sweden since 1860 and investigates how these changes link to regional economic development. In the literature on the topic, it has already been thoroughly documented that Sweden experienced a long period of regional economic convergence (in regional GDP per capita) during the post-war period, followed by regional economic divergence after about 1980. Sweden shares this trajectory with many industrialized countries (Enflo et al., 2014; Rosés & Wolf, 2019). Many explanations for this recent economic divergence have been proposed, including structural change and an increasingly sophisticated spatial division of labor, the effects of technological change, the introduction of two-earner households, and changes in social norms (Enflo & Henning, 2016). Analysis of economic convergence and diverge in terms of regional GDP per capita describe important but partial aspects of long-term regional economic change. How regional economic convergence and divergence link to changes in population across Sweden is not as well described, nor analyzed from a long-term point of view. While we know that urbanization was a dominant feature of Sweden's late industrialization and that there was a strong inclination for the economically active population to migrate into the expanding metropolitan cities (Stockholm, Gothenburg, and Malmö) during the post-war decades (Schön, 2010), the details of regional population change and especially how it varied across time for different categories of regions and how it relates to economic divergence, has not been systematically covered. Nevertheless, the outcomes of the population change process are of great economic, political, and socio-economic concern: while the Swedish population increased by 33% in the 1950–2008 period, municipality growth rates varied between a 1,300% population increase to a 60% decrease (Holm et al., 2013). Declining municipalities risk a depletion of resources and negative cumulative economic spirals, while expanding municipalities risk overuse of local infrastructure and rapidly increasing factor costs. To create a longitudinal analysis of the regional population changes and spatial inequalities in Sweden from the 1860s to 2020 and analyze their determinants and links to regional economic convergence and divergence, we combine long-term decade population data for the historical counties (1860–2020) with detailed annual population observations for municipalities (1968–2021). Based on this data, we ask how regional population changes developed in Sweden during our investigated period, how have they varied across time, and how regional population changes and spatial economic inequalities have been empirically related across time. In the analysis, we find that, since the early 1900s, regional divergence has overall characterized regional population growth in Sweden – as the national population grew larger, the differences between regions have increased. Before 1910, the regional differences grew smaller, both in terms of population and in the production of economic value. As industrialization picked up speed, this benefited regions all around the country in terms of GDP per capita, but under the condition of a population that was becoming increasingly concentrated. After a slowdown in this process during the low-growth period of the 1980s, 'double divergence,' in both population and regional GDP per capita, has characterized growth since the 1990s. Smaller regions, and especially peripheral ones, lag substantially behind both in terms of regional GDP per capita and population. Two regions, in particular, forge ahead: Stockholm, with its top position in the urban hierarchy, and Norrbotten, with its vast natural resources. To the regional system at large, the double divergence effects are potentially problematic. In essence, smaller and more peripheral regions become less populous and relatively poorer production-wise. The last time we saw this was in the 1930s. We believe that such a mechanism is an important background to what some have called the 'geography of discontent' (McCann, 2020). In our time, this creates regional welfare challenges for policymakers that are very different from those posed during the post-war persistent growth period, where the regional system also diverged in terms of population, but where production was still regionally distributed in a more equal way to the benefit of the remaining population. How and why spatial inequalities vary across time (frequently analyzed as the processes of regional convergence and divergence), and how they are linked to economic change, are long-debated issues, from both a theoretical and empirical point of view. While mainstream neoclassic theory explains that economies have a general tendency towards equilibrium (such as in Solow, 1956) and that this also includes regional economies, a second class of theories, working in the traditions of Myrdal, Kaldor, and Pred but also Lucas and Romer's endogenous growth models, claim that spatial development is largely a process of cumulative causation and increasing returns. Such mechanisms make spatial gaps increase across time and prevent geographical economic equalization. Among the most powerful accounts in the last decades of how regional inequalities emerge and persist are those connected to new economic geography, pioneered by Paul Krugman (1991). It essentially states that the more regional integration processes continue (by means of, for example, improved means of transportation), the more pronounced the benefits provided to firms in core regions will be. The counterintuitive results from these models show how regional integration will not lead to convergence, as the main sentiments about the impact of communication technologies normally suggest, but rather to regional divergence and an economic strengthening of the core regions. Empirically, while localized assets and individual preferences may put restrictions on the divergence process, new economic geography models still seem to provide an explanatory framework for tendencies in our time of development (Krugman, 2000). Indeed, new economic geography has provided an important framework to explain why economic development becomes increasingly concentrated in a few places within countries at the very same time as it becomes far cheaper and easier to transport goods and information alike. In the wake of new economic geography, and partially as a response to it, evolutionary economic geography has elaborated further explanations for regional divides on the basis of primarily path dependency mechanisms but also learning and institutional change (Boschma & Frenken, 2018). In contrast to the focus on regional economic divergence in our time, there are also accounts that aim to explain the long-term empirical observation that saturation mechanisms over time tend to moderate (regional) cumulative causation processes. They eventually lead to a turning point, after which an equalizing process – including, among other things, migration – dominates (Williamson, 1965). In essence, these long-term accounts teach us that trends of regional development may shift across time, depending on economic, technological, and institutional factors. Also, new economic geography views have been fruitfully combined with a more long-term economic history perspective, where, across longer periods of time, the locational advantages provided by localized and fixed assets are complemented by advantages offered to firms in urban agglomerations with their rich markets (Crafts & Mulatu, 2005). The notion that relations between convergence and divergence are not static but change across time also links to empirical work on what type of agglomeration advantages are important to economic performance, and at which point in time. From a vantage point in a rather conventional view on agglomeration advantages, of which there are different variations – Marshall–Arrow–Romer (MAR)/Localization (specialization), Jacobs (diversity), and Urbanization (size) – Duranton and Puga (2001) as well as Neffke et al. (2011) show that the impacts of different types of agglomeration advantages vary across the evolution of a firm's activities, and across the industry life cycle. This recalls the spatial product cycle model (Norton & Rees, 1987; Vernon, 1966), which assumes that the location of activities in the development of a product varies over time, depending on the supply of inputs and access to markets. While the time perspective of such studies is relatively limited, they still show that the ways in which different regions offer advantages to firms tend to shift with economic, technological, and institutional development. Indeed, 'steady states' of regional development are highly theoretical and, at best, temporary constructs. Suddenly, it becomes more important to understand economic change and why it happens than to understand the short-term developments. This has led us to argue elsewhere that there even are cohesive eras of regional development and that those in essence can be connected to a geography of long swings (Enflo & Henning, 2016; Henning et al., 2011). Technological, institutional, and, arguably, territorial factors (Storper, 1997) jointly determine the characteristics of these 'structural periods,' 'paradigms' (Freeman & Louçã, 2001), or 'technology shifts' (Schön, 2010). The impacts of localized resources, agglomeration externalities, regional path dependencies, and patterns of convergence and divergence seem to be consistent within the economic, technological, and institutional frameworks of each era. Yet, the exact spatial characteristics of each era are not repetitive but empirically distinct. In sum, theories about regional convergence, divergence, and cumulative causation may all be relevant, but at different points in time and at different places. This is true because economies fundamentally change across time, and they are also less predictable than we often would like. So far, this type of long-term work has developed largely based on regional production measures (regional GDP per capita) or income measures (regional incomes per capita), that is, measures where population acts as a denominator. The general tendency for this research has also been to focus on the numerator, production or income, while regional population and internal migration have been considered merely as an adaption to the conditions that changes in regional production, such as localized natural resources or technology, have brought. However, migration and other demographic changes of course play an important, although not always recognized, role in the processes of regional convergence and divergence. While the neoclassical Solow model sees migration as an equalizing force, the new economic geography models regard migration as driving divergence. On the basis of the economic research about regional development eras discussed above (Schön, 2010), it is reasonable to assume that the forces inducing migration also vary over the long run and are particularly linked to structural change and the creation of job opportunities in growing industries that vary in their dependence on the localized factors of production that are most important in each era. In essence, the majority of within-country migrants are drawn to regional job opportunities in each era. Once people find opportunities in new regions, increasing returns set in and make the regions grow even more by the logics of agglomeration. There is one complication to such a diverging process, of course – the increasing regional factor costs that arise in agglomerations. However, in our case, because of the availability of land in Sweden and self-reinforcing regional path dependency mechanisms, we find it rather unlikely that increased regional factor costs, on average, induce a dramatic turning point in regional growth patterns. Rather, such turning points set off processes of structural change in the economy and the introduction of a new era of economic development, where the regional availability of production factors set off new structures of regional growth. In Table 1, with inspiration from research on technological paradigms (Freeman & Louçã, 2001) and technology shifts (Schön, 2010), we try to collect the intuitions from some previous research to identify some core production factors and determine which types of overall location patterns they produce in each era, from a Swedish perspective (Berger et al., 2012; Enflo & Henning, 2016). The specific time periods should indeed be regarded as overall approximations, but they give a general sense of the timeline. The agriculturally dominated economy was generally characterized by a spatially distributed growth regime in Sweden, partly because of the need to access arable land and partly because people simply did not have much choice to stay by their plot of land, if they were fortunate enough to have one. This period gradually transcended into the early manufacturing economy by the late 1800s, with increasing manufacturing employment (Figure 1). However, because of the nature of early Swedish industrialization, which was predominantly based on localized raw material or access to power (Berger et al., 2012), the new era of economic change was still met by a predominantly spatially distributed growth regime. Industrial activities grew across the country, as well as in relatively peripheral locations. Also, it can be noted that this would last well into the 1900s until Sweden, or indeed Scandinavia, became endowed with cities of any internationally relevant size. In effect, well into the industrialization process, Scandinavia consisted of rural societies with a very limited bourgeoise class. Only with a more mature manufacturing economy in the very late 1800s and first half of the 1900s came a real expansion of traditional 'industrial cities,' such as Stockholm, Gothenburg, Norrköping, and Malmö. However, big cities were actually not the overall dominating issue here – the industrial economy continued to grow also in smaller cities and seemingly peripheral places. Source: Elaboration on data from Enflo et al., 2014 After the industrialization peak in the 1960s (Figure 1), the industrial cities suffered considerably because of deindustrialization and even recorded negative population growth for a limited period. With the revival of knowledge-intensive manufacturing and growth of knowledge-intensive producer services from the 1990s onwards, the biggest cities, especially Stockholm, experienced a massive economic revival. The most important turning point came, however, with the economic crisis in the beginning of the 1990s. More traditional manufacturing industries were particularly hard hit, and the crisis weakened some of the most important remains of the distributed town and city location regime of the mature manufacturing economy. While the three metropolitan regions and some university cities that supply local economies with well-educated labor for the service economy have recorded great economic success since then, many older industrial regions have not yet reached the economic levels they had in 1989, and it is maybe unlikely that they ever will. While this transition story is nowadays uncontroversial, there are good analytical reasons to consider the dispersion and concentration of population at various spatial levels and its regional implications as an integral part of this story (and not only the 'denominator') and link them to the longitudinal patterns of institutional, technological, and regional production changes documented so far in the literature. During the different eras and spatial growth regimes, what did population development actually look like? To analyze the regional population dynamics in Sweden across time, we rely mainly on two sets of data from Statistics Sweden. First, for the period 1860–2021, we rely on county population data from the Swedish Historical Regional Accounts. We use the 24 historical counties, 'län' (Enflo et al., 2014). We track the net population development over time and explore the changes in the distribution of population. We also compare the changes in population of the counties with their economic performance across time and explore the temporal co-variation of the measures. For data on the economic performance of the regions (regional GDP per capita), we rely on calculations from Enflo et al. (2014) and Enflo & Henning (2022; see also Rosés & Wolf, 2019). Because of changes in administrative units over time on lower levels than the county, it is difficult to create complete longitudinal population time series back to the pre-industrial era on spatial aggregation at levels lower than counties. Second, to create a more detailed account of more recent developments in the spatial population distribution, we therefore explore Statistic Sweden's publicly available longitudinal data, and we follow population development in about 290 municipalities between 1968 and 2021. We describe the population changes, using appropriate classification groups. In particular, we make use of the categorization of municipalities provided by the Swedish Agency for Growth Policy Analysis, where municipalities are categorized into very remote countryside municipalities, remote countryside municipalities, countryside municipalities close to cities, remote but dense municipalities, dense municipalities and close to bigger cities, and big city municipalities. Also, we classify municipalities into labor market regions, defined as functional analysis (FA) regions by the Swedish Agency for Economic and Regional Growth. We use a structured and simple path of analysis, relying on identification of breakpoints in regional population development and convergence–divergence in the time series and analysis of the link between long-term regional population growth and regional economic growth; analysis of cumulative alternative catch-up effects by means of beta-convergence analysis; and, lastly, exploration of the relation between regional structural change, regional size (agglomeration), and population growth. During the period that we study, the Swedish population grew from 3.8 million in 1860 to 8 million in 1970 to almost 10.5 million in 2020. Patterns of international migration have changed drastically during this period – from the American exodus of 1850–1920, where more than 1 million Swedes left, to 2021, when Sweden had about 90,000 inward migrants. Figures 2 and 3 give a first record of the relative regional population changes during the period 1860–2020. Figure 1 (above) describes the relative population (shares of national total) by geographical partitions. 1 All parts of the country record population growth since the 1860s in absolute terms. However, the South and the West show relative decreases in their shares of national population. The North experienced a boom both in economic and population growth during the late 1800s because of the international demand for wood products and iron ore, but a substantial relative population decline since then. Even in absolute numbers, population in the Northern counties declined from the 1960s onwards. The East, which includes Stockholm, increased its share of the Swedish population from 25% to almost 40% at the end of our investigated period. Figure 2 (below) complements this picture, documenting the relative shares for the 'metro' counties (which include Stockholm, Gothenburg, and Malmö) across the period. They increased their relative population shares substantially from the late 1800 onwards, with an intermission period during the international (and Swedish) structural crises in the 1970s and 1980s. Source: Own elaboration on data from Statistics Sweden, Enflo et al. (2014), and Enflo & Henning (2022). Source: Own elaboration on data from Statistics Sweden, Enflo et al. (2014), and Enflo & Henning (2022) In Figures 4 and 5, the coefficient of variation (CV) of population in Sweden is recorded for two periods and regional divisions (black lines) – counties (long-term) and municipalities (medium term). The coefficient of variation is a global variation measure (s/m) – the higher the value, the greater the variation in the distribution (among the regions). Thus, a lower CV represents a more equal value dispersion among the regions – or, in a time perspective, rising CVs signal regional divergence and decreasing CVs signal convergence. Source: Own elaboration on data from Statistics Sweden, Enflo et al. (2014), and Enflo & Henning (2022) Source: Own elaboration on data from Statistics Sweden, Enflo et al. (2014), and Enflo & Henning (2022) From an initially low regional variation in the population between the regions in the late 1800s (and even slight convergence), a long period of increasing differences in regional (county) population distribution lasted until the peak of the manufacturing era (1960/1970), where it came to a temporary halt. After that, from 2000 onwards, the population divergence process in Sweden continued. In this perspective, the regional population divergence that Sweden experienced in the last decade is just a continuation of a process that has lasted for more than 130 years and is thus closely linked with the entire Swedish progression towards a rich welfare state. Figure 5 (below) gives a more detailed picture of the latter part of the process. Among the municipalities, the stagnant period in the 1970s and 1980s described in the county data even signals relative population convergence, followed by the stagnant period. The force of the divergence process from 1990 onwards is, again, demonstrated by increasing CVs toward the very end of the period. This divergence trend seems robust regardless of which spatial unit is used for measurement. This could be compared with the CV of regional GDP per capita (the grey line in Figure 4) that record a strongly decreasing trend before 1910 and during most of the post-war era. Overall, until the 1980s, production of economic value had become more equally distributed across the country since industrialization, but with greater accompanying imbalances of population. This indicates that, even though some of the population left for growing population centers, those who actually stayed in other regions still gained from a wide distribution of productive activities across the country. During such circumstances, population divergence would not pose much of an economic problem – there are no obvious 'left behind' places on average. In two periods, 1910–1940 and in our time, economic divergence in regional GDP per capita and regional population divergence have, however, temporarily coincided. During these periods, economic production of wealth became (has become) more concentrated, as did (has) population. This means that the regional divergence process we experience today is much more qualitatively profound than the population divergence trajectory in the first decades of the post-war era until 1980. Source: Own elaboration on data from Statistics Sweden, Enflo et al. (2014), and Enflo & Henning (2022). Moving from a global measure to a more detailed view, it is, of course, interesting to see if there are structures in terms of which regions have gained overall, and which have not, from this long-term population divergence process. Is it the counties that house industrial cities that have, by virtue of agglomeration, especially benefited from population growth, or did it enable weaker regions to catch up over time ('beta-convergence')? In the long term and for the regions, it seems difficult to identify a real structure in the divergence process (Figure 7). In the real long term since industrialization, no convincing long-term tendencies of cumulative growth in terms of population, nor beta-convergence, can be found (beta-coefficient 36.5, not significant). Instead, in the long run, small as well as large counties have recorded substantial growth, and overall regional growth for this period seems to have little to do with established historical conditions in the late 1800s – except for Stockholm (top growth observation). One obvious reason for this is the vast transformation of Swedish society that industrialization brought, where initial conditions for growth were rather like a tabula rasa, without any pre-existing agglomerations or agglomeration advantages to speak of. Industrialization made all kinds of places in Sweden grow, and not only those that could nurture embryonic industrial cities. Cities come with blessings, but not all of the time. In a shorter time perspective and when we apply a more detailed regional picture (using the municipalities) and cut the time series by the shift in regional development patterns of the 1980s as identified above, the picture becomes a bit different (Figures 8, 9). Source: Own elaboration on data from Statistics Sweden Source: Own elaboration on data from Statistics Sweden For the period 1968–1990, where there was a short period of population convergence and economic stalemate, there was, again, no real size structure when it comes to which of the municipalities that grew or declined and which did not (Figure 8). While a simple regression has a slightly negative beta coefficient (−2.8), it is insignificant on all standard levels. Indeed, ample examples of both positive and negative population growth could be found in all municipality size classes (Figure 8). Things become very different from 1990 onwards (Figure 9). Here, there is certainly a premium for the more populous municipalities. The municipalities are more neatly clustered around a regression line with a beta coefficient of 16.1 and are highly significant. Compared with the earlier period, none of the municipalities grow except the ones that are already bigger. This picture again speaks to a convincing population divergence process taking place since 1990, and one in fact not only driven by the very top municipalities in the regional system but along the regional distribution as a whole. How did these different periods of regional growth change the size distribution of the regional system, and are the regional tendencies above robust to a more function-based (labor-market) definition of regions? Figure 10 records rank-size graphs for the current 60 Swedish labor market (LA) regions for 1968, 1990, and 2021 (for the sake of comparison, we use the 2015 definition of the functional regions). Sweden has a comparatively flat rank-size curve (Stockholm's functional region is, for example, roughly twice the size of the second largest, Gothenburg). Between 1968 and 1990, the characteristics of the regional hierarchy in terms of population also did not change that much (a simple linear regression obtains an explained variance of 34%). There are tendencies for a downward drop in the end of the distribution, but it does not affect the overall slope much. Then, between 1990 and 2021, the slope drops owing to the weak growth in the end of the distribution and high growth in the top, among the smaller regions. This starts to change the entire functional shape of the distribution (the explanatory value of a linear regression drops to 27% in 2021). Again, the regional system diverges. Source: Own elaboration on data from Statistics Sweden So far, our analysis has shown that regional divergence from a population point of view has taken place in Sweden ever since the late 1800s. Via a temporary convergence period during the post-war era, divergence picked up new speed, especially with the decline of smaller regions towards the right end of the rank-size distribution. Geographically, which regions are these? Finally, in Figure 11, we use the regional categories of the Swedish Agency for Growth Policy Analysis to investigate how relative population shares changed across times in different regional categories. Source: Own elaboration on data from Statistics Sweden The interpretation is rather straightforward: metropolitan municipalities (including suburban municipalities) increase their shares from the late 1990s, while dense municipalities (which include several university cities) remain more or less stable. Those municipalities with already low shares declined even more. As we have seen before, this is not a temporally concentrated disaster scenario but rather a slow and steady decline, possibly accelerating toward the end of our investigated period. However, the 1990s mark a new link between regional population change and spatial distribution of production. This new 'double divergence' will be our main point in the concluding discussion. The results are shown in Table 2. As expected for the entire timeline (1860–2020), there is a strong and significant GDP-per-capita convergence. This has, however, no specific link to the population density of the provinces (Models 1, 2, and 3). The post- versus pre-1970 results are even more informative. Before 1970, while Models 4, 5, and 6 demonstrate a clear regional convergence and catch-up effect, there is only weak evidence that county population size, per se, was closely linked to growth. This reinforces our previous indication – that growth was comparatively spatially dispersed in the aftermath of the second industrial revolution. After 1970, there are only signs of very weak convergence in the regional hierarchy below Stockholm (Models 8 and 9), but Stockholm records a very strong and significant growth effect, as does Norrbotten. Successful regions in this later period are, in other words, top hierarchy regions or endowed with unique and rich natural research resources. In this paper, we use a variety of methods and regional aggregations to show that regional divergence overall characterized regional population growth in Sweden since the early 1900s – as the national population grew larger, and the economy too, the population differences between regions soared. Concentration of population and regional population divergence have been some of the defining features of Swedish development, as the country within less than two generations turned from an enormously poor country, receiving considerable foreign aid during the 1877–79 famine, to one of the richest industrialized countries in the world. We believe that, from the early 1900s until the late 1980s, the growing spatial population inequalities were not an adverse outcome of growth – they were rather a sign that people became materially better off, on average, and were increasingly able to move within and outside the country as they desired. There is nothing romantic about the economic situation of the much more widely dispersed population in Sweden of the 1860s. We also show that, as industrialization picked up speed, this benefited regions all around the country in terms of GDP per capita. This means that, while population was concentrating in the regional system until 1980, the people that moved to or stayed on in the regions further down the hierarchy performed economically well. This was, in a sense, an economically reasonable type of population divergence combined with economic convergence. In the 1980s, this picture changed. After a slowdown in the regional production convergence processes during the low-growth period of the 1980s, 'double divergence' (in both population and regional GDP per capita) characterized growth patterns. Smaller regions, especially peripheral ones, lagged substantially behind both in terms of population and regional GDP per capita. This is the second time that we observe such divergence since industrialization – the last time was in the turbulent time between the two world wars. In the period 1910–1940, regional population differences soared, and economic imbalances did, too. This was a time of economic uncertainty and crises. In particular, countryside industries were affected (Schön, 2010), and the link to population changes is easily identifiable. We believe that this story is one of structural change, varying forces of agglomeration across time, as well as of natural resource endowments. Agglomeration economies may at times have been important to the mature manufacturing society in Scandinavia, but probably in the form of local specialization (so-called MAR effects). These may arise in specialized regions of relatively humble size in, for example, textiles and furniture. This also allowed smaller regions to grow in the early post-World War II era. During early industrialization, peripheral growth was more likely connected to various localized resources than to agglomeration effects (Crafts & Mulatu, 2005). In our time, in line with the predictions of new economic geography, agglomeration effects seem to benefit growing service industries in the metropolitan areas and in the urban cores. Dense municipalities (medium-sized cities, which mainly consist of the regional centers) have defended their relative positions in terms of population relatively well since the 1990s. However, in terms of economic production, divergence during the 1990s and early 2000 can mainly be attributed to the growth of the most sizeable city regions – in particular, Stockholm – with a parallel decline in the smallest and most peripheral regions. Indeed, we conclude that cities come with blessings, but not all of the time. However, sparsely populated regions can also benefit in our time if they are endowed with natural resources that are in high demand in the global markets. This is the case for Norrbotten, with its uniquely rich, high-quality iron ore deposits. At the time of writing, it appears to be especially challenging to predict future regional population trends, as the COVID-19 pandemic has created unprecedented learning curves when it comes to the technologies of remote work and new meeting cultures have opened up to hybrid solutions. At least in theory, this questions previous assumptions about the necessity of geographical proximity and attractivity of big cities. At the same time, refugee crises not seen since after the Second World War shook the European continent. We do not know how this will affect settlement patterns in Europe and their connections to regional productivity and welfare. It is unlikely, but possible, that population trends will turn. As we have seen in this paper, they have changed before. In sum, it seems difficult to provide a consistent explanation for regional population growth – it is likely that this can be explained by referring to structural change in the economy (from a focus on agriculture to manufacturing and then to service) as it moves from one economic era to the next – and how this is met by the regional availability of production factors but also by changing institutions and technologies that have relaxed migration obstacles across time and migration self-selection. In fact, adjusting for international differences, it has been shown that, compared with many Organization for Economic Co-operation and Development (OECD) countries, Sweden is among the top most unequal in terms of geographical population concentration (OECD, 2003). However, this is not the whole story – to the regional system at large, the double divergence effects that we identify are potentially problematic. In essence, smaller and more peripheral regions become less populous and relatively poorer. The last time we saw this was in the 1930s. While we think that population divergence between regions is not a problematic issue per se as long as it is matched with per capita welfare and productivity development on par with the rest of the country, the double divergence of our time makes us slightly uneasy in the light of an emerging new structure of the 'geography of discontent.' In particular, we believe that it is important to stimulate productivity development and progressive structural change in more peripheral regions. While this is, of course, easier said than done, access to educational institutions throughout the country seems extremely important, as do persistent technology investments. The COVID-19 pandemic has also brought new remote work experiments, which actually allow people with high-productivity jobs to relocate to more peripheral locations that offer a high quality of life. In fostering a rise in productivity and creating a local sentiment, where peripheral regions actually actively participate in the national growth trajectory, we find such experiments embryonic but nevertheless promising.