Population age structure: An underlying driver of national, regional and urban economic development

Malmberg, Anders; Malmberg, Bo; Maskell, Peter Article Population age structure: An underlying driver of national, regional and urban economic development ZFW - Advances in Economic Geography Provided in Cooperation with: De Gruyter Brill Suggested Citation: Malmberg, Anders; Malmberg, Bo; Maskell, Peter (2023) : Population age structure: An underlying driver of national, regional and urban economic development, ZFW - Advances in Economic Geography, ISSN 2748-1964, De Gruyter, Berlin, Vol. 67, Iss. 4, pp. 217-233, https://doi.org/10.1515/zfw-2023-0040 This Version is available at: https://hdl.handle.net/10419/333181 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. https://creativecommons.org/licenses/by/4.0/ ZFW– Advances in Economic Geography 2023; 67(4): 217–233 Anders Malmberg*, Bo Malmberg, Peter Maskell Population age structure– An underlying driver of national, regional and urban economic development https://doi.org/10.1515/zfw-2023-0040 Received: 2 May 2023; accepted: 7 November 2023 Abstract: This paper argues that population age structure plays a significant role alongside institutional, technological, political, and cultural factors when it comes to explaining shifts in urban, regional and national economic development. The paper demonstrates how demographic transitions lead to changes in population age structure which in turn correlate with global shifts in economic development from 1950 onwards. It then analyzes the role of population age structure at the sub-national level by reviewing some prominent cases of regional and urban shifts in Western Europe and North America. Population size, population density and migration have always been an integrated part of economic geography, and the consequences of ageing in national and regional economies are increasingly being studied. The specific role of population age structure as a driver of economic development has, however, so far largely been ignored in the field. Keywords: working age ratio, age structure, economic growth, regional shifts, demography, aging 1 Introduction In recent years, macro-economic analysis has increasingly started to investigate how demographic shifts impact economic trends (e. g. Eggertsson et al.2019). In a way this is old news for economic geographers. From Smith (1776) through Weber (1909) to Krugman (1991), population volume and density have been seen as a determinants of market size, in turn allowing increased division of labor, attracting economic activity and forming the basis of agglomeration economies. While many demographic aspects are generally well-studied and understood by economic geographers, including aging or migration and gender, one aspect has been strikingly absent: the economic impact of changing population age structures1. Here– and this is the core argument to be advanced in the paper– geographers have a lot to learn from recent research in economics and demography showing how shifts in the age structure during a demographic transition, from high to low death and birth rates, correlate with economic performance on different geographic levels. This growing body of literature2, has demonstrated that age structure was a major factor behind increasing GDP per capita in the historical making of today’s economically advanced countries, and it is a crucial component when contemporary emerging economies experience strong economic growth. Age structure refers to the relative distribution of a given population across age cohorts or age groups (such as youngsters, young or older adults and seniors). While most studies of economic effects of population change have focused on national economies, only few have considered 1 Clark’s 1987 Roepke lecture was an early harbinger of these insights (Clark 1987), which hark back to Thompson (1929) and Myrdal (1957). Yet, two major readers in economic geography (Clark et al.2018, Sheppard & Barnes, 2000), with a total of around 70 chapters covering a quite comprehensive range subthemes in the field, contain no considerations regarding population age structure as a factor that might influence economic development at national or regional levels. A bibliometric search of the three journals Economic Geography, Journal of Economic Geography and Regional Studies 2000–2022, with a total corpus of 6000 papers, returns no papers using the terms demography, demographic transition or dividend. The term population can be found in the title, abstract or keywords of 260 papers, while age or aging appears in 90 papers. Of these, only a handful deal with aspects of the topic in focus in this paper. For recent examples of such contributions, see Andersson (2001); O’Leary (2001); Gregory & Patuelly (2015); Polèse & Shearmur (2006); Bönte et al. (2009); Kurek (2011); Baerlocher et al. (2019); Prenzel & Iammarino (2021); Vendemmia et al. (2021). A couple of recent papers include age diversity as a control factor when modelling spillovers and productivity effects of immigration (Kemeny and Cooke 2028, Buchholz 2021). 2 Early contributions upon which much of this literature in economics and demography builds, include Brander & Dowrick (1994); Kelley & Schmidt (1995); Crenshaw et al. (1997); Bloom & Williamson (1998); Lindh & Malmberg (1999); Bloom et al. (2003). *Corresponding author: Anders Malmberg, Department of Human Geography, Uppsala University, Sweden, E-Mail: [email protected] Bo Malmberg, Department of Human Geography, Stockholm University, Sweden, E-Mail: [email protected] Peter Maskell, Department of Strategy and Innovation, Copenhagen Business School, Denmark, E-Mail: [email protected] 2023 Article Note: Order of authors is alphabetical, all three authors contributed equally to this work. Open Access. © 2023 the author(s), published by De Gruyter. This work is licensed under the Creative Commons Attribution 4.0 International License. 218 Anders Malmberg, Bo Malmberg, Peter Maskell: Population age structure subnational regional and urban economies, even though intra-national spatial differences in age structure can be substantial (see Cruz & Ahmed, 2018 for examples). In this paper we will argue that economic geography would benefit from explicitly considering the role of changes in population age structure in the analysis of shifts in local and regional economic development. Thus, the overall aim of this paper is to explore the impact of population age structure on regional and urban economic development. We do this by introducing and discussing the literature and by presenting population data that displays the close relationship between demographic transition and economic development. The remainder of the paper is divided into four main sections. Section 2 presents the theoretical argument of why we should assume that population age structure impacts economic development, and describes the data and methodology of the paper. In section 3 we present a generalized model of demographic transition in four stages and argue that this transition must now be understood as a driver of economic development, departing from entrenched views of demographic changes as a pure reflection of economic change. We consider how the broad global shifts in economic-geographic development since 1950, as described e. g. by Dicken (1998, 2015), correlate with the stages of demographic transition that have taken place, at uneven pace, in different countries around the world. The theoretical discussion of the demographic transition is supplemented with descriptive data on how population age structure has changed over time in different countries at different stages of economic development. We present a regression model that shows a statistical connection between age structure and growth. This model is not aiming to formally test causality– this will remain the task of future research– but merely to state that there is a correlation. Having asserted that there is a connection, at the national level, between the historic undergoing of the demographic transition and its resulting effects on the population age structure on the one hand, and global economic growth patterns during the last 50–60 years on the other, we raise the question of whether a similar connection could be assumed to play a role when it comes to uneven growth patterns within countries. Section 4 thus focuses on sub-national regions in Western Europe and North America. In this section, we revisit some of the most prominent cases of regional and urban shifts that have been studied by leading economic geographers in the last 30–40 years, and investigate if and how incorporating data on population age structure might add to their explanation. Even though all the cases in point concern regions where the grand demographic transition was completed many decades ago, we show that differences in age structure can still be large enough to play an important role. In this context we also discuss the effect of migration on population age structure. In section 5, we consider the main findings in the setting of economic geography and raise some general points for future research. 2 Theoretical considerations and methods The impact of a population’s age structure on macro-eco- nomic outcomes can be linked to the economic life-cycle of individuals at the micro level. Young children rely on the care and assistance of guardians or other caregivers. As they grow, their need for immediate care decreases, but they still require support to engage in education before joining the workforce. Upon transitioning into adulthood and the labor market, typically between 20–25 years of age, individuals often establish partnerships and have children. Under favorable circumstances this results in a relatively extended period where they are responsible for supporting themselves and their children, usually with limited opportunities to save. Around the age of 45–59, parental obligations typically decline, allowing for more savings and increased working hours (among all genders). Individuals approaching 65 often start retiring from the workforce and begin to rely more on their savings, including pension-system transfers. In the event of a long life, the final years may again necessitate dependency on care and support from others. An age structure model of economic development can, in its most simple form, be seen and the mere aggregation of such individual life cycles. In any society, at any point in time, the population is distributed over ages in a specific way. This aggregate age composition has both direct and indirect effects on economic outcomes at the macro level (for an overview see Lee & Mason, 2010). First, there are consequences for aggregate savings and aggregate investment (upper middle-aged people on aggregate save more than younger or older people). As an extensive literature in economics has clarified, changes in the absolute and relative size of age groups will affect capital flows, currency exchange rates, real interest rates, and inflation. Other important channels consist of demographic effects on capital accumulation, labor supply, skills and capabilities, innovation, and entrepreneurship. Recently, it has been suggested that shifts in age structure will have consequences for the balance between the Anders Malmberg, Bo Malmberg, Peter Maskell: Population age structure 219 production of tradable and non-tradable goods (Papetti, 2021). Dependent age groups, i. e. children or elderly, induce a higher demand for non-tradables like personal services. In contrast, an increase in the proportion of working age adults will enable a shift towards the production of tradables which might produce export incomes. This, in turn, could generate national and regional growth processes along the lines suggested by Kaldor (1966, 1970), where increasing output in the export sector generates higher productivity through a mechanism of increasing returns3. Throughout the paper we will use the working age ratio, i. e. the proportion of working age adults in the population, as our main empirical indicator of population age structure, where working age adults are defined to be in the 20–64 age span. We have chosen 20 years instead of 15 years as the lower cut-off value based on the increased proportion of young adults that are in upper secondary and tertiary education. The working age ratio will be used to explore both global and regional differences in age structure, as well as to explore global shifts in age structure over time, and regional shifts in age structure for a selection of countries. What we attempt to show is that this simple measure is closely related to cross-sectional differences in economic development and also to historical shifts in development patterns. This will be done both formally, using regression estimates, and informally, by relating differences in age structure and changes in age structure to well-documented patterns in economic development that have previously been analyzed in the economic geography literature. Since the main aim of our paper is to argue that age structure effects should be given more consideration in economic geography research, the empirical results presented in this paper should be regarded as illustrations rather than original contributions to the empirical literature on age structure effects. The regression estimates used should be seen as descriptive, and as providing an indication of the type of statistical correlations that exists between demographic variables and income growth trajectories. A formal examination of possible causal relationships requires more advanced econometric modelling than the estimations we present in this paper. Our data sources are presented in the Data appendix. 3 See Antenucci et al. (2020) and Deleidi et al. (2021) on how shifting demographic structures through the Kaldor mechanism can have more pronounced effect on per capita income growth than suggested by mainstream growth models. 3 Demographic transition– a driver of global change The idea that most countries at some point experience a demographic transition from high death rates and high birth rates to low death rates and low birth rates was developed by demographers at Princeton University in the 1940s (Dyson, 2010). Back then it was assumed that the transition was driven by economic development. Later research has demonstrated that the transition process cannot be explained in this way. In most countries, the decline in (child) mortality predates any increase in per capita income, and the same is the case also for the subsequent reduction in fertility. This is a significant finding for economic geography, since it implies that demographic change cannot be seen as mere reflection of economic change. It opens up for considering a different direction of causality, namely viewing demographic change as a driver of economic change. What, then, are the mechanism through which the demographic transition brings about economic development? This is explained by the now renowned stage model, showing how the proportion of dependents relate to people in the working ages throughout the transition process (Bloom and Freeman, 1988). The presence of clearly distinguishable stages of age structure change is due to the strong uniformity of the demographic transitions process, non-withstanding differences in the details: “when dealing with the demographic transition we are focusing on a phenomenon that, in very long term perspective, is fundamentally uniform” (Dyson, 2010: 79). The model has by now been extensively tested to explain divergent patterns of global economic development, and as such it deserves to be included alongside more familiar accounts of international trends in economic geography and related fields.4 In Figure 1, we use UN data to present how the working age ratio has changed from 1960 to 2020 and how it is expected to change until 2050. The heatmap, used several times in the paper, shows gradually warmer colors for higher working age ratios: shades of blue up to 40 per cent, green/yellow up to 50 per cent, yellow/orange/light red up to 60 per cent and dark red above that. In this figure, the countries have been grouped according to their current stage in the transition model. It shows how the four stages in the demographic transition can be 4 For example, capital mobility (Sassen, 1991), information technology (Castells, 1996), a search for new spatial fixes (Jessop, 2000), rescaling (Brenner, 2004), re-allocation of surplus capital (Harvey, 2010), or generalized, regional agglomerative processes (Scott & Storper, 2003). A summary of this literature is provided in Perrons (2004). 220 Anders Malmberg, Bo Malmberg, Peter Maskell: Population age structure translated into changes in the working age ratio, which will typically rise gradually all the way up to the final stage where an ageing population will eventually lead to a decreasing ratio. In stage one of the demographic transition model, diminished child mortality results in swelling numbers of surviving newborns, leading to a rapidly increasing population size while lowering the working age ratio. Families are typically struggling to provide for their many children and per capita income remain modest. Countries in stage 1 today are mainly found in Sub-Saharan Africa. When this is followed– often after several decades– by a decline in fertility5 in stage two, a “generational bulge” is created that will affect society over the coming decades. The generational bulge consists of all the large cohorts of surviving children and youngsters that were born after the decline in child mortality but before the decline in fertility. In stage two, smaller cohorts of new born children ease the dependency burden on older generations and provide more room for capital accumulation. Many women, previously homeworking, might enter the labor market for the first time (Becker & Lewis, 1973). At the same time the early cohorts of the “bulge” are entering adulthood. In contrast to dependent children these young, mobile adults are positioned to support themselves with their labor, although still with limited capacity to generate savings. It is even likely that a large increase in the share of young, less experienced labor will temporarily push down the relative wages of this age group, while the growing need for investments, not least in housing and infrastructure, will drive up the price of capital. Income inequality and inflationary pressures may follow. Faced with limited options, restless unemployed youngsters may also take to the streets causing civil unrest (Urdal, 2006, Canning et al., 2015) or attempt to migrate to places where the perceived prospects for building a future are greater (Malmberg & Sommestad, 2000). Stage 2 countries are today predominantly found in northern Africa, the Middle-East, southern Asia and Latin America. In stage three, as time passes, the “bulge” continues upward into the middle-ages and things tend to change. With growing age follows increased experience and self-re- liance. Competitiveness intensifies, incomes rise, savings grow, and personal networks expand as individuals in the bulge reach middle age (Mason & Lee, 2004). Idea generation, creativity, new firm formation, and productivity are stimulated. This is where a society benefits from the 5 Evidence now suggests that lower infant mortality triggers lower fertility and that both are results of simple knowledge dissemination (World Bank, 1984; Kalemli-Ozcan, 2002; Galor, 2010; Cervellati & Sunde, 2011). so-called demographic dividend, and often experiences an extended period of exceptionally strong economic growth. Ireland and the Asian Tiger Economies of South Korea, Taiwan and Singapore reached this stage in the early 1980s while China followed somewhat later. This growth-stimulating demographic environment is typically maintained for several decades, but inexorably “gold turns to silver” as the now aging “bulge” reach retirement age in stage four. Firms are deprived of access to the former employees’ personal networks and experience. The retirees will start spending their savings and demand more services that can drive up wages when competing with other sectors of the urban, regional, or national economy. The cohorts making their way into adulthood are much smaller, born as they are after the decline in fertility. With fewer to carry the growing burden of aging dependents, further economic progress can become increasingly difficult. Countries that have since long reached stage 4 include most European countries as well as the US, Canada, Japan, Argentine and Uruguay. From this point on, changes in the working age ratio will mainly be determined by baby booms and baby busts, i. e. cyclical variations in fertility levels, alongside migration (Crenshaw et al., 1997; Feyrer, 2011). The almost 40 “mature economies” classified as having reached stage 4 have passed their peak working age ratio, but it is notable that they have all maintained ratios well over 50 per cent for the last 70 years (see Figure 1), as the growing number of aging and substantially healthier citizens is balanced out, so to speak, by declining numbers of newborn (Kurek, 2011; Van der Gaag & Beer, 2014; Pool, 2016). Still, rapidly ageing countries, such as Italy, Spain and Greece display modest growth during the last few decades, and the United Nation’s (2019) forecasts indicate that several of these will see a further decrease in working age ratios, to levels substantially under 50 per cent, in the decades to come. In Asia, Japan has become the current showcase of demographic contraction, displaying the world’s largest proportion of centenarians– almost double in size of its closest followers (France, Italy, US). China has perhaps benefitted more than any other country from a demographic dividend, with a working age population almost twice as large as the rest of the population altogether stemming from its one-child-only domestic policy 1980–2015, supported by rapid skilling and attraction of foreign investments (Wei & Hao, 2010). But now the consequences of this harsh child policy are being felt and by 2030 China’s population will on average be older than Europe’s and much older than that of the US. This foreseeable country-wide aging will in all likelihood take place before China has closed the economic gap to the richest economies. Future attempts to close this gap depends on Anders Malmberg, Bo Malmberg, Peter Maskell: Population age structure 221 Figure 1: Working age ratio and GDP growth 1960–2015 in 111 countries. 222 Anders Malmberg, Bo Malmberg, Peter Maskell: Population age structure its ability to reduce the current high value placed on male children and to reverse its by now institutionalized 4–2–1 model, with one child supporting two parents and four grandparents (Fang, 2016). Some countries in stage 1, mainly in Sub-Saharan Africa, experience a prolonged interval before completing stage two, resulting in a quite extended “bulge” (Canning et al., 2015). So even when the demographic transition may induce gains in terms of improved health and human development (Eloundou-Enyegue & Giroux, 2013), today’s around 30 pre-transitional countries mainly in Sub-Saharan Africa (Ahmet et al., 2016; Bloom et al., 2017) might face tougher prospect of converting reduced fertility to economic growth than countries already well underway such as the countries that are today in stage two or three (Bleakly, 2006; Eastwood & Lipton, 2011). In Figure 1, Nigeria stands out as an exception with strong economic growth in the 2000s, with a still low national working age ratio. Despite these qualifications, important as they are, broad and strong empirical evidence suggests that the demographic transition constitutes an important component in a country’s economic development (Galor, 2012; Ranganathan et al., 2015). Industrialization and increased economic growth are typically associated in time with the demographic stimulus accompanying the third stage of the transition, and a slow-down of economic growth is typically associated with the ageing process that takes place in stage four. The statistical correlation between the working age ratio and log GDP per capita is very strong. An OLS regression including 111 countries with GDP data from 1960 to 2015 using the working age ratio as the only explanatory variable beside fixed time effect gives an adjusted R-square of 0.679. The parameter for the working age ratio is 12.58 implying that that a one percentage point (0.01) increase in the working age ratio between two periods corresponds to an 12.58 % (0.01*12.58 = 0.1258) increase in GDP per capita between the two periods. This corresponds to a 2.5 % annual growth rate over a five-year period. See Table 1, below. To visualize this correlation, the two rightmost columns in Figure 1 show the annual growth rate in GDP per capita for countries in the different stages for two periods, 1960 to 1990 and 1990 to 2015. During the first period, income growth was very slow in stage 1 countries and in most stage 2 countries. Growth 1960–1990 was stronger in stage 3 countries, especially in those countries where the working age ratio increased early. Also most stage 4 countries had stable income growth during this period. Through the years 1990 to 2015 the growth rate in GDP per capita intensified, especially in stage 2 countries that went from low to high working age ratios, while almost all stage 3 countries now experienced fast growth. In contrast, the countries in stage 4 proceeded on a downward slope in income growth as their populations got older and working age ratios consequentially declined. The pattern for stage1 countries is more mixed with some displaying higher growth rates whereas others continued along a low growth track. The data suggest a link between demographic development and trends in global between-country income inequality. During the period up until 1990, developed countries increased their relative income as they experience an increasing working age ratio. During the second period, countries in stage 2 and stage 3 further reduced the global income gap as their working age ratio reached or even surpassed the now aging countries in stage 4. At the same time, the growth patterns show that the fit between age structure change and per capita income growth is far from perfect. Clearly there are other factors at play (including institutional, technological, political, and cultural), which can modify or restrain the ways by which changes in working age ratios influence economic development. What we would maintain, though, is that the general correspondence between shifts in the working age ratio across countries and shifts in global income structure is of such extent and magnitude that it seems important to consider the demographic transition as a driver of economic development in its own right. Table 1: Effects of age structure (population in working ages, 20–64 years, as proportion of total population) on GDP per capita, cross-country 5-year panel data. Models with only fixed time effects, and with both fixed time and fixed country effects.  Response ln GDP per capita (time effects) 1960–2015 Response ln GDP per capita (time effects and country effects 1960–2015 Estimate for Prop in work age 12.58*** 5.19*** t-value 49.55 18.7 Adj R sqr 0.679 0.924 Effect Prop work age (SSQ) 1143.3 38.5 Effect Year (SSQ) 27.6 53.5 Effect Country (SSQ) – 481.2 N of observations 1332 1332 N of countries 111 111 Note: Shifts in world technology levels over time have been controlled for by using period fixed effects in column 1. By additionally controlling for country level factors that are constant over time, the parameter estimates in column 2 measure the extent to which shifts in the proportion of working ages for a country that already is in a specific stage of the transition will affect GDP per capita (5.19 over five years gives a 1.04 % annual growth rate in GDP per capita). SSQ is Sum of squares. See the Data Appendix for sources. Anders Malmberg, Bo Malmberg, Peter Maskell: Population age structure 223 Existing frameworks in economic geography are quite successful in explaining how economic development tends to be spatially polarized and also to demonstrate the dynamic character of capitalist development. But these frameworks are not equally good at explaining why and when broad shifts appear in the relative economic fortune of nations, regions or cities, and it is here that the demographic dividend framework can be of help. That is, by exploring the role of changing age structure, economic geography could become better at explaining– and even predicting– shifts in regional and urban economic development. Moreover, changes occurring in working age ratios provide an account of global income trends that differs markedly from a more neoliberal interpretation suggesting that post-1990 shifts in income growth largely reflects the success of market-oriented reforms. 4 Age structure and regional shifts in developed countries– Reexamining four “paradigmatic” cases We have seen that age structure changes can contribute to explain global shifts in economic development. Can it also explain regional trend shifts within industrialized countries? Having so far looked at population age structure and economic development globally and at the national level, we will now focus on spatially uneven economic development within countries, previously much studied by economic geographers. In order to investigate the impact of population age structure on regional economic development, we will review four classical historical cases where economic geographers have paid a lot of attention to document and explain major regional shifts in industrial and economic growth patterns within industrialized countries. The cases have thus been selected based on the established fact that during a given historical period one previously successful region has been “outcompeted” by another. The cases selected are not random. They do belong to the most well-studied, in their time “paradigmatic”, cases that economic geographers have investigated during the past 50 years (cf Scott 2000), and some of the classical analyses of these cases still belong to the most highly cited in the economic geography literature. The cases selected are: – The industrial growth of the so called Third Italy during the 1970s and 1980s as compared to the relative stagnation of the so called First Italy. – The economic downturn of the traditional industrial core regions of West Midlands and the Northwest in the United Kingdom during the 1970s as compared to the growth in Southeast England and The London-Bristol axis. – The industrial and economic crisis of the American Manufacturing Belt during the 1970s, as compared to the industrial and economic rise of the “Sunbelt” in the US. – The relatively weaker economic growth of Southern California (Los Angeles) as compared to the strong economic growth of The Bay Area (San Francisco) 1970– 2010. In each of these cases, earlier research has documented how one region (or group of regions) displayed superior economic performance compared to the other. This difference in economic performance was typically explained with reference to changes in the ways economic activity were organized (technological development, industrial restructuring, rise of new production regimes) and by how well such changes conformed with varying social, cultural and institutional traits of the regions in question. Even though the selected cases are historic, and the initial key studies were published quite some time ago, they still form an important background to contemporary economic geography, characterized as it is by more focus on contextuality, path dependency and contingency. Despite differences in time frame, scope, theoretical approach, and methods applied, the studies selected all share two common features; they have had great impact on the field of economic geography and they do not include or even refer to population age structure in the array of explanatory factors considered. In the following we will examine whether there can also be demographic explanations to those shifts, i. e. whether it can be that, in each of these cases, the regional shift coincided with a change to a more unfavorable population age structure in the “losing region”, and a more favorable age structure in the “winning region”. We do this by calculating the working age ratio for each pair of regions in point, and illustrating with heatmaps how it changes over time during the periods in focus. 224 Anders Malmberg, Bo Malmberg, Peter Maskell: Population age structure 4.1 The economic rise of the Third Italy The industrial districts of the Third Italy,6 rose to global fame in the late 1980s and early 1990s not least through Piore and Sabel’s (1984) extremely well-cited analysis.7 The relevance of the case is derived from the claim that the industrial growth in Third Italy signified the rise of a new form of industrial organization, often referred to as flexible specialization, that was proving to be more competitive than the rigid Fordist production regime then dominating the traditional industrial core of the so called First Italy. Even critics of this claim had to admit that the idea of the Third Italy had “achieved an iconic status in geography” and that “perhaps the major theoretical thrust of Anglo-Ameri- can economic geography in the 1980s and 1990s concerned the previously hidden potential of industrial districts for stimulating regional economic development” (Agnew et al., 2005:83; see also Boschma, 2008). 6 The term “the Third Italy” was first described in a book by Bagnasco (1977). It alludes to the regions left out from the traditional division between the First Italy (the industrialized core in the Northwest) and the Second Italy (the less developed South). 7 Piore and Sabel’s (1984) book on the second industrial divide, where the emergence of new industrial districts in the Third Italy plays a prominent role, has almost 20 000 citations in Google Scholar (captured autumn 2023). In order to check for a possible “hidden” demographic impact behind the shift from the First to the Third Italy during the 1970s and 1980s, we have calculated the working age ratio for Italian regions from 1950 onwards. Figure 2 shows that the regions making up the First Italy had a clear demographic advantage over the rest of the country in the 1950s and 1960s, a period when they displayed superior economic performance, while that gap had largely closed by the 1970s and 1980s. In particular it is worth noting how Emilia Romagna, arguably the most archetypical Third Italy region, had a higher working age ratio than all the core regions of the First Italy: Liguria, Piemonte and Lombardia, in 1970, 1975 and 1980. The figure also reveals how the longer-term problem of lacking economic development in Southern Italy was underpinned by a population age structure persistently unfavorable right up to the end of the 20th century. Together the data clearly support the claim that explanatory gains could be harvested by including information about the regions’ working age proportion– in this case when explaining the relative economic success of the Third Italy in the 1970s and 1980s. Figure 2: Proportion of population in working ages (20–65 year) 1955–2000 (per cent) in Italian provinces divided into First, Second and Third Italy. Source: ISTAT (2022). Regional division based on Boschma (1998). The Lazio region, dominated by Rome, occupies an intermediate position until the 1990s when it starts to display the very high share of working age population typical of capital regions. Anders Malmberg, Bo Malmberg, Peter Maskell: Population age structure 231 Scott, A.J. (2000) Economic Geography: The Great Half-Century. Cambridge Journal of Economics, 24(4): 483–504. Scott, A., Storper, M. (2003). Regions, globalization, development. Regional Studies, 37(6–7): 579–593 Sheppard, E. & Barnes T.J. (eds) (2000)A Companion to Economic Geography. Oxford: Blackwell. Smith, A. (1776) An Inquiry into the Nature and Causes of the Wealth of Nations. London: W. Strahan and T. Cadell. Storper, M., Kemeny, T., Makarem, N. & Osman, T. (2015) The Rise and Fall of Urban Economies: Lessons from San Francisco and Los Angeles. Stanford: Stanford University Press. Storper, M. & Walker, R. (1983) The Theory of Labour and the Theory of Location. International Journal of Urban and Regional Research, 7(1): 1–43. Thompson, W.S. (1929) Population. American Journal of Sociology, 34(4): 959–975. Urdal, H. (2006) A Clash of Generations? Youth Bulges and Political Violence. International Studies Quarterly, 50(3): 607–629. Van der Gaag, N. & De Beer, J. (2015) From Demographic Dividend to Demographic Burden: The Impact of Population Ageing on Economic Growth in Europe. Tijdschrift voor Economische en Sociale Geografie, 106(1): 94–109. Vendemmia, B., Pucci, P. & Beria, P. (2021) An Institutional Periphery in Discussion. Rethinking the Inner Areas in Italy. Applied Geography, 41(135): 1–11. Weber, A. (1909) Über den Standort der Industrie: Reine Theorie des Standorts, mit einem mathematischen Anhang von Georg Pick. Tübingen: JCB Mohr (Paul Siebeck). (Translated by Carl Joachim Freidrich and published 1929 as: ‘Theory of the location of industries’, Chicago: The University of Chicago Press) Wei, Z. & Hao, R. (2010) Demographic Structure and Economic Growth: Evidence from China. Journal of Comparative Economics, 38(4): 472–491. World Bank (1984) World Development Report 1984. New York: Oxford University Press. Wren, C. & Taylor, J. (1999) Industrial restructuring and regional policy. Oxford Economic Papers, 51(3): 487–516 Data appendix Global data on age structure Country level data on age structure is available from (United Nations, 2019). This dataset contains population estimates for every fifth year from 1950 to 2020 for a total of 201 countries. We compute the proportion of working age adults, aged 20–64 years, by dividing the number of individuals aged 20–64 with the total number of individuals in the population. The standard definition of the working age population is 15–64 years of age. However, we restrict the working age population to the 20–64 age group to account for the fact that an increasing proportion of the 15–19 years age group participate in upper secondary education or tertiary education (Lutz et al., 2017). Global data on real GDP 1950–2015 For data on GDP we use expenditure-side real GDP at chained PPPs from Penn World Table, version 10.0 (Feenstra et al., 2015). The Penn data contains time series for real GDP for 172 of the countries for which UN provides demographic data. 55 countries of these countries have data on real GDP starting in 1950. In 1960, a total of 111 of the 172 countries has GDP data in Penn. In 1970, and additional 37 countries have data. And from 1990, all the 172 countries have GDP data in Penn World Tables have data for real GDP. Per capita income is obtained by dividing real GDP with the total population as provided in (United Nations, 2019). Data on regional age structure for England and Wales 1961–2001 In England and Wales, changes in the geographical subdivisions that are used to report census data make it difficult to obtain data for regions with the same definition over time. The approach used in this paper is to aggregate local data to NUTS2 areas (Nomenclature of Territorial Units for Statistics) using digital vector boundaries for NUTS2 areas in England and Wales and for local districts. Such data is available at the Office for National Statistics on their Open Geography Portal, (https://geoportal.statistics.gov.uk). We have used data from three censuses: 1961, 1981, and 2001. For 1961 we have used the SH13 table: Age and marital condition by five year age groups [1961 Census] that reports aggregates for Local authority districts as defined in 1961. Available at https://www.nomisweb.co.uk/sources/ census_1961_sh. For 1981 we have used “1981 census– small area statistics”, Table 2, Age and Marital Status, Great Britain, with geography for “local authority: district / unitary (prior to April 2015)”. Available at: https://www.nomisweb.co.uk/ sources/census_1981. For 2001 we have used the 2001 census, Standard Table2 (ST002), Age by sex and marital status, with geography for “local authority: district / unitary (prior to April 2015)”, available from https://www.nomisweb.co.uk/sources/ census_2001_st. Census data is available for 5-year age groups. These age classes have been aggregated to working age adults in the age span 20–64 year, and to non-working age adults, age span 0–19 and 65+. Local authority districts as defined in 1961 and Local authority districts unitary, prior to 2015 have been assigned to NUTS2 areas using the “Join attributes to location” feature 232 Anders Malmberg, Bo Malmberg, Peter Maskell: Population age structure in QGIS version 3.22.7-Białowieża. This feature allows districts to be assigned to NUTS areas on the basis of with which NUTS area the district has the largest overlap. The aggregation to NUTS1 areas has been made using the map “UK: NUTS1 Levels 1 and 2, 2018”, available at https://geoportal.statistics.gov.uk/documents/ons::nutslevels-1-and-2-january-2018-map-in-united-kingdom/about. Data on regional age structure for Italy 1952–2001 For data on the age structure of Italian provinces we have used four data sets for inter-censuses populations: Estimated resident population – Years 1952–1971, Estimated resident population 1972–1981, Estimated resident population 1982–1991, and Estimated resident population 1991– 2001. The data is provided for one-year age groups. We have aggregated the age data into working age population (20–64 years), and non-working age population (0–19 years and 65+). These data contain information on 20 provinces corresponding to the NUTS2 regions of Italy, as well as data on 5 broader regional groups corresponding to Italy’s NUTS1 regions. The Inter censuses population data is available from ISTAT, https://www.istat.it/en/population-and- households?data-and-indicators, submenu Population-De- mographic indicators. Data on state level age structure for the United States 1940–2019 Regional data on age structure has been obtained from IPUMS (https://www.ipums.org). IPUMS now holds microdata from a large set of national censuses and provides researcher access to this data. In the present study we have used 1 % samples from the 1940 census, the 1950 census, and the census 1970, and 5 % samples from the 1960 census, the 1980 census, the 1990 census, and the 2000 census. In addition we have used the 1 % samples of the American community survey (ACS) for 2010 and 2019 (see Ruggles, S., Flood, S., Foster, S., Goeken, R., Schouweiler, M., & Sobek, M.(2022) IPUMS USA: Version 12.0 [dataset]. Minneapolis, MN: IPUMS). All these datasets contain information about the state where the samples individual lives as well as information of exact age, and on sample weights. Age structure data for the US states have been obtained by aggregating this data, using a weight that reflect the above sampling scheme, by year, state of residence, and age. A further age aggregation has then been made to obtain the size of the working age population (aged 20–64 years) and the size of the non-working age population (0–19 years, 65+), for each state and each year. Aggregation to Census Divisions has been done on the basis of the document “Census Bureau Regions and Divisions with State FIPS Codes”, (https:// www2.census.gov/geo/pdfs/maps-data/maps/reference/ us_regdiv.pdf). Data on county level age structure for the United States 1970–2010 The microdata from IPUMS does not, in general, contain information on county of residence. Here we have instead used two different data sources. First, Haines, M.R. (2018) U.S. County Populations by Single Years of Age, Sex, and Race, 1970, 1980, 1990. Inter-university Consortium for Political and Social Research Available online at https://www.icpsr. umich.edu/web/ICPSR/studies/37115. Second, a table provided by the Bureau of the Census: Intercensal Estimates of the Resident Population for Counties and States: April 1, 2000 to July 1, 2010. Available online at https://www.census. gov/data/datasets/time-series/demo/popest/intercensal- 2000-2010-counties.html. The Haines data provide numbers for one-year age groups. The intercensal data gives data for five-year age groups. For the Greater Los Angeles area and the Bay Area we follow the defintions given in Storper et al. (2015): “Los Angeles, in this context, means the Greater Los Angeles metropolitan region (known officially as the Combined Statistical Area [CSA] encompassing five adjacent, continuously urbanized counties (Los Angeles, Orange and Ventura, and parts of San Bernardino and Riverside)” p.3, although we include the whole of San Bernardino and Riverside. “The San Francisco metropolitan area, which is also known as the Bay Area, is a Combined Statistical Area that until 2010 comprised ten varied counties, from the Sonoma and Napa wine country in the north to Silicon Valley and the Santa Cruz Mountains and coast in the south, and from the wild Pacific coastline to the west inland to the mountains separating it from the Central Valley of interior California”, p.5. According to the map presented on p.6 (Map 1.2), these counties are: Alameda, Contra Costa, Marin, Napa, San Francisco, San Mateo, Santa Clara, Santa Cruz, Solano, Sonoma. Anders Malmberg, Bo Malmberg, Peter Maskell: Population age structure 233 References for data appendix Feenstra, R.C., Inklaar, R., & Timmer, M.P. (2015). The next generation of the Penn World Table. American Economic Review, 105(10), 3150–3182. Lutz, W., Butz, W.P., & Samir, K. (2017). World Population and Human Capital in the Twenty-first Century: An Overview. Oxford: Oxford University Press United Nations. (2019). World population prospects 2019, online edition. rev.1. File POP/7-1: Total population (both sexes combined) by five-year age group, region, subregion and country, 1950-2100 (thousands): United Nations, Department of Economic Social Affairs, Population Division. Available online at: https://population. un.org/wpp/Download/Standard/Population/