Spatial Distribution of Population by Age in France, 1851–2014
The issue of regional planning was central in French political debates during the second half of the 20th century. This can be explained by two phenomena at this time: the share of the population living in the Paris region was increasing sharply for 100 years – especially because of the rural exodus – and the share of the population living in the “empty diagonal” – a geographical area connecting Ardennes in the North-East to Ariège in the South-West (see Oliveau and Doignon, 2016) – was declining. These two phenomena contributed to the popularity of Gravier (1947)’s book entitled « Paris and the French Desert »: the author showed how the rural areas gradually emptied in favor of a territorial organization centered around the capital. This question led in 1963 to the creation of the “Delegation for Regional Planning and Regional Attractiveness” (DATAR in French), in charge of implementing the interministerial policy of spatial planning. This policy involved, for example, incentives for companies which settle in depopulated or impoverished territories, or the promotion of “balancing metropolises” in order to reverse the hyper-centralization.
In this paper, I clarify the question of spatial planning by providing an in-depth analysis of the spatial distribution of the French metropolitan population. This study is mainly based on Bonnet (2018b)’s database in which the annual populations by age and département since 1901 are available. I add to this database population-by-age data retrieved in censuses between 1851 and 1896 to cover a period of more than 150 years. With these data I analyze the spatial distribution of the population according to age and sex. This question is not clearly understood nowadays. Combes et al. (2011) for example analyzed the spatial distribution of the total population between the French metropolitan départements for only five years (1860, 1896, 1930, 1982, 2000). Ayuda et al. (2010) analyzed the evolution of the spatial distribution of total population for 9 European countries including France, but in 1850 and 2000 only. Talandier et al. (2016) propose a cartographic analysis of this issue since 1806 within the framework of French cities, but for the total population and without using an analysis by indicators (see Le Mée, 1989 for a presentation of the raw data used). Finally, other papers have studied this issue in the recent period, without a historical perspective (see in particular Breton et al., 2017).
From a methodological point of view, indicators used to analyze the evolution of the spatial distribution of the population are crucial. So far, the literature has mainly used indicators aggregating the departmental distribution of densities per km2 into a single indicator. Combes et al. (2011) for example used the Theil indicator, while Ayuda et al. (2010) based their analysis on indicators such as the standard deviation, the coeﬃcient of variation or the Gini index. These indicators may hide evolutions in specific parts of the departmental distribution. For example, the Gini index may decrease while the share of the less populated départements decreases. This occurs if this phenomenon is more than oﬀset by a population transfer from the most populated départements to the “a-little-less” populated départements. In this paper, I try to provide an answer to this issue. In order to analyze the spatial distribution of the population and its evolutions, I use a complete set of indicators containing both the Gini index and indicators specific to each part of the departmental distribution of densities per km2.
In addition, Bonnet (2018b)’s database provides annual flows of births and deaths by sex and département. With these data one can dissociate population variations due to natural or migratory movement. In this paper, I highlight the départements combining demographic imbalances: an imbalanced demographic structure that leads to a birth deficit, and a very low attractiveness for migrations. Consequently, I participate in the literature on “Shrinking Regions”, which emerged with Oswalt and Reniets (2006)’s work: they listed all the cities in the world whose population decreased over time. This literature analyses the territories too, particularly within the works of Bontje et al. (2012), Fol (2012) or Galjaard et al. (2012). My contribution shed new light on this issue since I do not only look at the absolute variations of population but I also compare it to the national evolution. I therefore consider that a territory is on the decline if its share in the national population decreases, since this evolution leads to a loss of political and economic power.
Finally, with the departmental population structures by age and sex, one can analyze how these structures are increasingly diﬀerentiated between the territories and where older or younger people are overrepresented. As far as I know, this issue of diﬀerences in age structures has never been treated in the literature but deserves further consideration. Indeed, if age structures are more and more diﬀerentiated, the territories are becoming more and more interdependent. This would result in significant transfers of income from the most active to the oldest territories. Here we meet the distinction between productive and residential economies in line with Blanc (2007), Davezies (2008) or Beyers and Nelson (2008). These potential transfers of income implies that fiscal decentralization should be conducted with caution: local budgets must not be aﬀected by demographic imbalances. It also implies that public policies have to be driven by the specificities of each territory: towards education where young people are overrepresented, towards health and dependency where the older are. In this paper, I therefore propose both an overall analysis of diﬀerences in age structures by using a single indicator, but also a cartographic analysis to better know territories in wich each age group is overrepresented or underrepresented.
All of these analyzes bring a number of new results. Firstly, I show that the population is more and more unevenly spread. In broad outline, this process can be described with three phases. For example, the increase of inequalities from 1851 to 1901 is the result of the concentration of population in the most densely populated territories to the detriment of all others. Indicators also show that the increase in inequality from 1968 onwards hides a drop in the share of the most densely populated départements. Second, I reveal that départements which cumulated imbalances according to both natural and migratory movement changed between the first and the second half of the 20th century: they were mainly in the central and western parts of the country between 1901 and 1968, while they were in the North-East and the south of Massif Central between 1968 and 2014. As such, I name « wide belt of attractiveness » the départements surrounding Seine and Seine-et-Oise insofar as their population increase is due to both a strong natural movement and migrations. This contrasts with Seine and Seine-et-Oise whose population growth rate due to migrations is lower than the national rate. Third, the analysis of the spatial distribution of the population according to age shows an inverted U-shaped profile over the recent period: the “20–39” age group is the most unevenly distributed. This profile has changed over time: in the second half of the 19th century, the elderly were the most unevenly distributed. Forth, the analysis of departmental age structures reveals that they are more and more diﬀerentiated since the end of the Second World War. This process is mainly explained by an overrepresentation of young adults in urban départements, while retirees are overrepresented in rural départements. In particular, the rural South West is gradually becoming a land of exclusion for young workers.
The rest of the paper is organized as follows. In Section 2 I present the data as well as the methods used in this study. In Section 3 I present the results. The fourth section concludes.
Data and Methods
In this paper I analyze the evolutions of the spatial distribution of the population between the French départements since 1851. The choice of this geographical unit is explained by the stability of their administrative boundaries since their creation in 1789. For the purpose of this study, I will confine the results to French metropolitan départements. Overseas départements are not included because available data are too recent to carry out a long-term analysis. In this study the term “national” refers to these French metropolitan départements.
In order to analyze the evolution of the spatial distribution of the population, I use diﬀerences in population densities, in line with Ayuda et al. (2010) for example. The population density per km2 is defined as the ratio of the population to the total of km 2. I use population densities since départements are not of equal size. For example, Gironde in the South-West has an area of 10,375 km2 while the one of its neighbor Tarn-et-Garonne is only 3,718 km2.
In stage 1, I present the density diﬀerences for the total population. My analysis, however, goes further than most of the current works on the spatial distribution of population over long periods: I analyze in stage 2 inequalities for each major age group (0–19, 20–29, 30–39, 40–49, 50–65, 65–79, 80 and over). In the remaining of the paper, I analyze the evolution of the spatial distribution of the female population, even if I have populations for both sexes. This choice is explained by two main reasons. The first concerns the data available to recontextualize the evolutions. Lifetables are crucial to do so : thanks to Bonneuil (1997)’s work, I have female lifetables for the 1851-1900 period, but not male lifetables. The second reason concerns the readibility of the historical trends. France has experienced three major wars during this period, and men have been more widely aﬀected than women by their consequences: forced migration and excess mortality make long-term developments less readable. Nevertheless, when there are noticeable diﬀerences, I present the results for the men in the appendix.
There are a large number of inequality indicators to capture inequalities. These indicators can also be used to analyze the spatial distribution of the population. Mackenbach and Kunst (1997) in the field of health studies, or Cowell (2011) more generally, make a non-exhaustive list of these indicators; they show how each of them provides diﬀerent informations on the issue. One can use the indices based on extreme ranks – the diﬀerence or the ratio between the highest value and the lowest value – or on the interquantile interval (the diﬀerence or the ratio between the x % of the higher values and the (1 − x) % of the lowest values). There are also Gini or Theil indices that reduce the distribution in a single indicator, or the indices of dissimilarity which express the part that should be distributed among the observations so that the values are similar for all. One may analyze the evolution of inequalities through σ−convergence or β−convergence (Barro & Sala-i-Martin (1992) for the most known) . The σ−convergence studies the evolution of an inequality indicator between two periods. There is β−convergence if the relationship between the variation of a variable between two dates t0 and t1 and the values in t0 is positive. Finally, inequalities can be analyzed in absolute or relative terms: a density diﬀerence of 10 inhabitants per km2 between two territories represents 20% of the average density when the mean is equal to 50, but only 10% for a value of 100.
In what follows, I use the Gini index as it is easily readable. In order to deepen the analysis, I use however other indicators which target specific parts of the departmental population distribution. I split this distribution in six parts and calculate the share of each of them. I get the shares of the 10% most densely populated km2 in the national population (namely P90–100), but also the shares of the second (P80–90), third and forth (P60–80), fifth and sixth (P40–60), seventh and eighth (P20–40), ninth and tenth (P0–20).
With these indicators I cannot study the inter-departmental diﬀerences in population distribution according to age. To do so, I use the Kullback-Leibler Divergence (Kullback and Leibler, 1951) as d’Albis et al. (2014) did to analyze the international dissimilarities of age-specific mortality rates. This Kullback-Leibler Divergence (KLD) is based on Shannon’s entropy (1948). Formally, the KLD between two population distributions by age P and Q is calculated as follows: KLD =log � Q(a) � P (a), (1.2.1) Ω P (a) with a the age and Ω the maximum age. To get an index summarizing departmental dissimilarities, I calculate the national KLD: N Ω Pi(a) K LDN at = log � � Pi(a), (1.2.2) i=1 a=0 PN at(a) with i the département and N the total of départements.
The national KLD is an aggregate indicator for all distributions. I also calculate distortion indices (ID) which highlight the departmental distortions according to age structure. Thus:
I Di(a) = PI (1.2.3)
PN AT (a) .
Aggregation of Data Sources
This study combines two specific sources of data: the first is raw data collected in 19th century censuses. The second is Bonnet (2018b)’s departmental database.
For the period 1851–1900, populations by age group were recorded every 5 years.1 These data were collected by a team of Franco-American researchers from the University of Ann Arbor and formatted by INSEE.2 They are available by quinquennial age group, sex and département. These data are not as reliable as the 1901–2014 ones, because of the quality of censuses in the 19th century. Bonneuil (1997) explained this point in his study on the demographic transition in French départements. Most of the biases come from respondents’ poor specification of age (attractiveness for round ages), lack of internal consistency in tabulations (the sum of départements is sometimes not equal to the national figure) and bad transcription of the data in tabulations. The impact of these biases is limited because the study focuses on populations by broad age group, and not by single age. Nevertheless, it is important to bear in mind that only the major trends are totally reliable for this period. In order to get populations on January 1st of each year, I assume that the population at the date of the census is similar to the population on January 1st of the census year, and that populations by age group during intercensal periods can be interpolated linearly.
For the period 1901–2014, data come from Bonnet’s database (2018b). This companion paper explains in detail the methodology used to estimate age populations. It relies mainly on the protocol of the Human Mortality Database developed by Wilmoth et al. (2007). The raw data that feed the estimation process come for the most part from the archives of the French statistical institutes. They consist of censuses, vital statistics (births and civilian deaths by age) as well as statistics on military deaths and deportations during the two World Wars. In this paper, I use deaths and populations by age as well as births, for each département and year.
Unification of the Geographical Framework
French metropolitan borders have little changed over the period 1851–2014. The variations are due to changes in the eastern borders, but also to the integration of new territories. I apply the departmental classification in force from 1918 to 1967, which includes 90 départements (see map in Appendix 1.5.1), in order to get a unified geographical framework and compute the inequality indicators. I rebuild the missing départements data in this classification. The methods used are diﬀerent, depending on the database.
Three main territorial modifications took place during the period 1851–1900. Savoie and Nice’s comté integrated France in 18603, which resulted in the creation of Savoie, Haute-Savoie and Alpes-Maritimes. Alsace-Moselle integrated Germany in 1870; this led to the creation of Territoire de Belfort and Meurthe-et-Moselle, that remained under French administration, while Moselle, Bas-Rhin and Haut-Rhin passed under German administration.4 Table 1.1 presents these departmental issues and the periods concerned. To rebuild these data, I assume that population changes had been synchronized between the missing département and a geographically close département. I therefore associate with each missing département a reference département. The latter is used as a support to estimate missing data. Table 1.1 reveal that these periods are short and the impacts on the overall results are therefore limited. Alsace-Moselle is somewhat diﬀerent as the missing period is longer.
Since reliable data are available before and after, I keep these départements in the study.
The period 1901–2014 presents two kinds of missing data. The first concerns Bas-Rhin, Haut-Rhin and Moselle. These départements were reintegrated in 1921 in Bonnet (2018b)’s database. I therefore estimate these data during the period 1901–1920. To do so, I proceed as before: I associate a reference département (namely Meurthe-et-Moselle) to each of these missing départements , then I estimate yearly deaths and population by age by assuming that their evolution were synchronized. The second concerns the reorganization of Ile-de-France in 1968. Until that date, Ile-de-France contained three départements: Seine, Seine-et-Oise and Seine-et-Marne. Seine-et-Marne remained the same, but the other two turned into seven départements (Essonne, Hauts-de-Seine, Seine-Saint-Denis, Val-de-Marne, Val d’Oise, Paris and Yvelines). I rebuild Seine and Seine-et-Oise from these new geographical units, by dividing yearly deaths and population by age as well as births pro-rata their distribution in 1968.5 The sum in the old classification thus remains equal to the sum in the new classification. The relative positions are fixed at their 1968 level.
Evolutions of Departmental Densities of Population since 1851
I start with the evolution of the French population between 1851 and 2014. Table 1.2 shows the population by sex for several dates. One can see that this evolution has not been linear: two phases can be identified to describe it. I choose the year 1946 as it seems to be the turning point in natural movement trend. During the period 1851–1946, the total population increased by only 10% (Line 3), and only by 5% for men (Line 2). Two causes can explain this evolution. Due to a very early demographic transition, the birth rate in France in the beginning of the 20th century already reached low levels compared to its European neighbors. Coale (2017, p.38) showed that the fertility index dropped sharply since 1820 while this phenomenon appeared rather around 1900 in other developed countries (1890 in Germany, 1913 in Italy). In addition, the mortality increased dramatically during the three major conflicts: the two world wars, and to a lesser extent the war against Prussia in the early 1870s. In addition, these wars impacted strongly the sex ratio: in 1946, the male population is 10% lower than the female population; this diﬀerence was only 1 % in 1851 (Line 4). The period 1946–2014 is radically diﬀerent. In 70 years, the population increased by 60%: the annual growth rate was 65 times greater than the period 1851–1946 one. The baby boom of the post-war years (according to Bonnet (2018b)’s database, the crude birth rate increased from 66 to 90 births per thousand of women between 1936 and 1946) created much larger cohorts than cohorts born before 1946. Moreover, the sharp rise in life expectancy during the second half of the 20th century (female life expectancy at birth increased from 65 years in 1946 to 85 years in 2014 according to the same source) allows older people to live longer, which increases the population. Finally, international migrations (coming mainly from the North-Africa in the 1960s) contribute to increase the population too.
In a second stage I analyze the population density at national level but also for each département. The results for several years are presented in Table 1.2. The national density of population in 1851 and 2013 was 67 and 117 inhabitants by km2, respectively. This increase hides local specificities. The first statement is that the density was multiplied by more than a factor 5 in the most densely populated département (namely, Seine), while it decreased by 30% in the less densely populated département (Basses-Alpes in 1851, Lozère in 2013). These variations are mainly due to the rural exodus, already highlighted by Ariès (1948). The second is that these evolutions have not been similar for all départements during these 150 years. The maximum density increased continuously between 1851 and 2013, but this is not the case for all the others, whose density followed a U-shaped curve: it decreased from 1851 to 1946, then increased from 1946 onwards.
In order to better understand the changes of these departmental population densities, Figure 1.3.1 maps the absolute departmental values for 1851 and 2014.6 Northern France was in 1851 globally more densely populated than the rest of the country. The Channel coasts, the German borders as well as Paris and Lyon regions had a population density of more than 80 inhabitants per km2, whereas these densities were less than 20 inhabitants per km2 in the Alpine départements, Lozère and Landes. Overall, this finding continues today. One can add the Atlantic and Mediterranean coasts as well as the Swiss border in the densely populated regions, namely where the densities are greater than 95 inhabitants per km2. The diﬀerence between the two maps is mainly in relation to the relative positions: the second revealed a broad band sparsely populated from Meuse in the North-East to Aveyron in the South-West. This band is called the “empty diagonal” in Oliveau and Doigneau (2016), among others. In 1851, it was not as marked as it is today.
The Three Phases in the Evolution of Spatial Distribution of Population
I begin by analyzing the spatial inequality indicators, as presented in Section 1.2.1. Figure 1.3.3 presents the evolution of these indicators since 1851 for the population of women. There was a sharp increase in inequality of population densities since the beginning of the period: the Gini index has more than doubled over the period (from 0.232 in 1851 to 0.320 in 1900 and 0.478 in 2014). While Combes et al. (2011) found a stagnation of the Theil index between 1982 and 2000, the Gini index was still increasing along this period. This can be explained in two ways. Combes et al. (2011) did not weight the départements by their area, and the Gini and Theil indices do not give the same importance to each part of the departmental distribution. This illustates why the analysis conducted with my indicators is particularly interesting. Moreover, these indicators highlight three phases in the increase of inequalities for 150 years.
Between 1851 and 1900, the more uneven distribution of the population is explained only by the share increase of the 10% most densely populated km2: the share of this decile in the total population increased from 21.5% to 30.2%, an increase of about 40%. Inside this decile, the Seine’s share7 – including Paris and its surrondings – doubled (9.1% in 1900) while the one of the remaining decile increased by only 20%. In contrast, the share of all other territories declined, without exception. These evolutions are in line with the rural exodus, whose beginning dated back to the 18th century (Ariès, 1948). Goreux (1956) for example showed how Paris and the other big urban centers attracted agricultural workers during the 19th century; the choice of emigration place was largely explained by the distance. The spatial distribution of the French population can no longer be explained by « first-nature » advantages as in pre-industrial societies8: agglomeration eﬀects became the main force explaining the changes at work.9 Thereby, the development of railway during the second half of the 19th century could have favored this process. Mojica and Marti-Hennberg (2011) revealed that in 1880, 90% of agglomerations were connected to the railways. For the authors, the train facilitated migrations of rural people to cities and therefore spatial concentration. Moreover, Fletcher (1961) and Schwartz et al. (2011) showed how lower transport costs at the national level led to the importation of cheaper US wheats. Consequently, the French and the English agricultural sector were plunged into crises between 1870 and 1900. Figure 1.3.3 shows that the share of the four least densely populated deciles strongly decreased from 1870 too: for example, the share of the 20% less densely populated km2 decreased by only 2% between 1850 and 1870, compared to 11% between 1870 and 1900 (from 11% to 9.7%). I name this phase “hyper-centralization” since the 10% most densely populated km2 expanded to the detriment of all the others.
Between 1900 and 1968, the 10% most densely populated km2 were still concentrating the popu-lation, even if the World War Two was a temporary break in this process. This break can be fully explained: Bonnet (2018a) showed that internal migrations during this conflict were strong, espe-cially from the North of the country (occupied by the Germans) to the South (in the free zone until
1942). According to this study, the scars left by the conflict in the country’s demography were deep: the refugees who fled densely populated regions such as Northern and Eastern borders, Bretagne, Normandie and Seine-et-Oise, did not fully come back. Nevertheless, over the whole period 1900– 1968, the share of the total population who lived in this first decile went from 30.2% to 40%; the cumulative increase reached 80% since 1851 . Unlike the previous period, this hyper-centralization was no longer at the expense of all other territories: the share of the second decile stagnated or increased slightly between these two dates, at around 12%. The population of the départements comprising second-tier cities increased as quickly as the national population. On the other hand, the share of the less populated départements were still declining. Since 1851, this decrease fell between -20 and -50% according to the deciles. Thus, while the rural exodus aﬀected all territories except the most densely populated between 1851 and 1900, the decline was over in fairly densely populated départements. I call this second phase “hyper-centralization thwarted”.
Finally, between 1968 and 2014, the share of the first decile in the total population decreased from 40% to 38.8%, around 3%. The decline is a little more sharp for Seine, around 4%. While the share of this département in the total population had increased three-fold in a little more than 100 years, the cumulative process stopped for the capital and its suburbs. The second-nature advantage (Krugman, 1993) was no longer enough to attract the national population. This can be explained by congestion costs in the Paris region, in line with the results of Puga (1999), Graham (2007) or Combes et al. (2012). Conversely, the share of the second, third and fourth deciles increased quite strongly. The share of the population living in the second decile went up from 12.3% in 1968 to 14.8% in 2014, an increase of 20%. The share of lower deciles continued to decrease, even if the pace was less sustained. Overall, the rise in global inequalities hides two phenomena: a decrease in the share of the most densely populated départements (pushing down inequalities), and a decrease in the share of the least densely populated départements (pushing up inequalities). I name this phase “multipolarization” of the French population. The multipolarization is geographical and not statistical: départements belonging to the second, third and fourth deciles increased over time. One can find in this cate-gory départements such as Haute-Garonne, Loire-Inférieure, Haute-Garonne, Haute-Savoie, Isère or Gironde, scattered throughout the landscape. It is within these départements that the second-tier cities such as Toulouse, Bordeaux, Nantes and Grenoble are located. The expression “Paris and the French desert” developed by Gravier (1947) gradually loses its importance. The policy pursued by DATAR since the 1960s succeeded. It allowed the displacement of the most mobile jobs towards regions where mass unemployment threatened at the end of the “30 glorious”.
Notes: Computations based on the population of women. P90-100 refers to the share of national population who lived in the 10% of km2 with the highest density values. All inequality indicators are weighted by km2 and normalized by 1851 values. Sample includes 90 départements.
The share of the less densely populated départements declined since 1851, with a temporary im-provement due to World War Two. This evolution, which is explained by a population growth rate below the national average, can come from two factors. The first is related to natural movement: the growth rate of the population from a surplus of births to deaths is lower than the national rate. The second is related to migratory flows: the growth rate of the population coming from net migrations is lower than the national rate. Départements can be classified in four categories as presented in Table 1.4. Class 1 départements are those with both rates below the national average. Class 3 départe-ments are those with both rates above the national average. Classes 2 and 4 are mixed situations.
Data from the period 1851–1900 do not allow for this classification since they do not include the total births and deaths of each département. Conversely, Bonnet (2018b)’s database is suﬃcient. By diﬀerence between the population change between two dates and the natural movement, it is possible to compute the apparent net migratory flow. Figure 1.3.4 presents the classification for periods 1901–1968 and 1968–2014, for the population of women. These two phases correspond to what I have called “hyper-centralization thwarted” and “multipolarization”.
According to the period 1901–1968, Figure 1.3.4 reveals a deep diﬀerence between the North and the South according to the natural movement (“Skyrocketing” and “Fertile-but-Repulsive” départements in green and yellow against “Shrinking” and “Infertile-but-Magnetic” départements in red and blue). In the North (generally above an arc connecting Doubs, Seine and Vendée), the growth rate of population due to the natural movement is higher than the metropolitan rate, diﬀerent from the South. This region was named « croissant fertile » (see for example Francart, 1983) and was visible since the Second World War. This phenomenon compensated a net migration rate lower than the national average: these regions were not attractive for migrations (whether internal or coming from abroad). In the South, départements along the Mediterranean coast, the Rhone Valley and those which host major cities compensated this impairment by a strong attractiveness: net migration rates are higher than the national average. For all the others, in red on the map, the disadvantages cumulated: their shares fell because of a weak natural movement and a lack of attractiveness on the migratory side.
The situation changed during the period 1968–2014. For this period, researchers have studied migration flows, and in particular interregional flows. This work has been carried out on each intercensal period (Baccaïni (2001), Baccaïni and Lévy (2009) for the most recent studies), but also dynamically between 1954 and 2008 (Baccaïni and Dutreuilh, 2007). The first striking result concerns Seine and Seine-et-Oise and feeds the conclusions already stated above. Migration rates fell below the national ones, which explains the decline in the share of the first decile (Figure 1.3.3). This statement is supported by the fact that the neighboring départements became “Skyrocketing” ones: they draw a green belt around Seine and Seine-et-Oise. These départements, more and more eﬃciently linked by transport to the capital, became attractive for migrations. I name it the “wide belt of attractiveness”. While Baccaïni and Dutreuilh (2007) put forward the reversal concerning Ile-de-France, the authors could not mention this departmental belt since they worked at the regional level. The second observation is about the North-South gradient concerning natural movement, which is no longer as readible as before. It became stronger than the national average in the South-East. The reverse occurred in Bretagne and in départements like Meuse and Ardennes, while they were at the heart of the baby boom following the Second World War. The North-West and the North-East diﬀered on the migratory side: the former was attractive, which was not the case of the latter. These results support the idea that Aisne, Ardennes, Meuse or Somme constitute a « shrinking region » in the sense of Oswalt and Rieniets (1984). This region, which had enjoyed a first-nature advantage (Krugman, 1993) for much of the 20th century (mines, heavy iron and steel plants), is still facing diﬃculties in converting its productive capital. Laménie (2016) showed, for example, that the population of Ardennes decreased by 10% between 1968 and 1999, due to the emigration of young people. This process led to a drop in the birth rate and an aging population. Finally, the South-West became attractive for migratory flows, which was not the case previously: Dordogne, Lot-et-Garonne, Tarn-et-Garonne or Corrèze are examples of départements that switched from “Shrinking Regions” to “Infertile-but-Magnetic” class. Baccaïni and Dutreuilh (2007) noted this attraction for the South and West from the 1960s, which contributed to the spatial redistribution of the population. Nevertheless, one can see on Figure 1.3.4 that some départements remained on the margins of this process. This is the case in the south of Massif Central (Aveyron, Cantal, Lozère), but also in Deux-Sèvres and Charente which did not benefit from the amenities of Atlantic coast.
Notes: Computations based on the population of women. Classification built according to the theoretical classification presented on Figure 1.3.4. Sample includes 90 départements.
The Uneven Spatial Distribution of Population According to Age
The long-term studies conducted so far could not analyze diﬀerentiated trends in the spatial distri-bution of the population by age group. The data available in this study allow this analysis. Overall, I find that the trends of population aged 0 to 19, 40 to 49 and 50 to 64 are the same as the trends observed for the whole population. On the other hand, they are significantly diﬀerent for women aged 20 to 29, 30 to 39, 65 to 79, and 80 and over. Figure 1.3.5 reveals the inequality indicators for women aged 30 to 39 and 65 to 79.
According to women aged 30 to 39, the main diﬀerence with national trends comes from the period 1990–2014, during which the share of the first decile increased while the share of the third and fourth deciles stagnated. The evolution of the first decile was almost completely due to the increase of Seine’s share (Paris and surrondings). Consequently, France is facing a new phase of “hyper-centralization thwarted” concerning this age group. It gathers individuals at the heart of their working lives, usually with high salaries and stable work situations. This result supports Combes et al. (2011)’s paper, which show that the Gini index of the spatial distribution of tertiary value added followed an inverted U-shape from 1860 to 1982 and increased from 1982 to 2000. The parenthesis in the aggressive spatial planning policy counducted during the 1960s could explain this process. Following works on endogenous growth, economists and politicians were aware of how economically strong regions have to be supported in order to redistribute income to poorer geographic areas (see, for example, Jayet et al (2006) and Davezies (2008)).
For the older ones, changes are diﬀerent for two reasons. Overall, the Gini index increased only by 60% over the period 1851–2014, compared to 110% for the whole population. From 1851 to 1910, this index remained stable, hiding contrary evolutions: both the share of the first decile and the share of the less densely populated départements expand to the detriment of the second, third and fourth deciles. Thus, there was a deconcentration of the elderly population in France from 1851 to 1900, unique in my statistics of population by age. Between 1900 and 1968, the hyper-centralization thwarted was at work: the share of the 80% least densely populated km2 fell, while the share of the first decile increased and that of the second stagnated. Finally, between 1968 and 2014, the multipolarization appeared: the share of the most densely populated territories in old people decreased, while the one of the départements of the second, third and fourth deciles increased.
These results are quite the same for men (See Appendix 1.5.2 for the corresponding graphs). Beyond the evolution of the spatial distribution of population by age, it is interesting to know which
are the most unevenly distributed populations, and whether these relative positions have evolved over time. Figure 1.3.6 presents the age profile of the Gini index for women and several dates along the 150 years of this study.
First of all, inequalities have increased for all age groups, which is consistent with what was presented earlier. These increases are significant: if one consider the age group 0–19, the Gini index went from a value of 0.25 in 1851 to a value of 0.48 at the end of the period. With regard to the age profile of the Gini index, Figure 1.3.6 reveals that the population density inequalities in 1851 were similar for all age groups between 0 and 64-year-old, with a Gini value of about 0.25. Beyond these ages, the values were growing. In other words, women aged 65 and over were much more unequally spread than the others. This specificity of the oldest ages gradually disappeared during the end of the 19th century: the inequalities observed for this age group became the weakest from 1901 onwards. Conversely, at this date, the flat profile between age 0 to 65 disappeared too. The profile reveals an inverted U-shape, more and more pronounced over time, where the most uneven age group is 20–29.
Table of contents :
Part 1: Essays in French Demographic and Economic History
1 Spatial Distribution of Population by Age in France, 1851–2014 21
1.2 Data and Methods
1.2.1 Inequality Indicators
1.2.2 Aggregation of Data Sources
1.2.3 Unification of the Geographical Framework
1.3.1 Evolutions of Departmental Densities of Population since 1851
1.3.2 The Three Phases in the Evolution of Spatial Distribution of Population
1.3.3 The Uneven Spatial Distribution of Population According to Age
1.3.4 Differences in Departmental Structures of Population by Age
1.5.1 Map of the 90 French Départements Used to Calculate Inequalities
1.5.2 Spatial Distribution of Men Aged 30 to 39 and 65 to 79: Inequality indices
1.5.3 Evolution of the Distribution of Population by Age Group : Decomposition Method
1.5.4 Departmental Contributions in KLDNat Evolution
1.5.5 Distorsion indices in 1856, 1896, 1946 and 2011
2 Spatial Inequalities in French Life Expectancy, 1806–2014
2.2 Data and Methods
2.2.1 Data Sources
2.2.2 Geographical Scope
2.2.3 Indicators of Inequality
2.2.4 Convergence Indicators
2.3.1 The Three Phases in Reduction of Spatial Inequalities of Life Expectancy
2.3.2 Role of Infant Mortality in Shrinking Spatial Inequalities
2.3.3 Major Changes in the Geography of French Longevity
3 Spatial Inequalities of Income and Welfare in France, 1922–2014
3.2 Data and Methods
3.2.1 Existing Databases Used
3.2.2 Income per Adult between 1922 and 2014: Estimation Method
3.2.3 Income and “Mortality Adjusted Income”
3.2.4 Analysis of Spatial Income Inequalities
3.3.1 The Three Phases in the Decrease of Spatial Inequalities of Income per Adult
3.3.2 The Virtuous Convergence of Income per Adult during the Second Half of the 20th Century
3.3.3 The Evolution of Spatial Inequalities of “Mortality Adjusted Income” per Adult
3.3.4 Changes in the Geography of Development in France
3.5.1 Publications Used to Compute Departmental Incomes
3.5.2 Map of the 90 French Départements in 1967
3.5.3 Methodology Used to Compute “Mortality Adjusted Incomes”
3.5.4 Spatial Inequalities of Income Densities
3.5.5 Shrinking Regions: A Classification
3.5.6 Departmental Classification According to Relative Income per Adult
4 Beyond the Exodus of May-June 1940: Internal Flows of Refugees in France
4.2 Data and Methods
4.2.1 Departmental data
4.2.2 Framework of the Study
4.2.3 Estimation of Annual Departmental Populations and Internal Migrations
4.3.1 The Global Consequences of the War
4.3.2 Annual Monitoring of Internal Population Migrations
4.5.1 Map of the 86 French Départements
4.5.2 Cause of Death Classification
4.5.3 Evolutions of Departmental Populations by Component, 1939–1946
4.5.4 Yearly Variations of Population due to Migratory Movement
Part II: Data and Methods 125
5 Computations of French Lifetables by Département, 1901–2014
5.3.1 HMD Protocol Methods
18.104.22.168 Raw Data Adjustments
22.214.171.124 Splitting Deaths into Lexis Triangles
126.96.36.199 Computations of Populations by Age at 1st January of each Year
188.8.131.52 Adjustment of Computed Mortality Rates
5.3.2 Specific Departmental Methods
184.108.40.206 Specific Methods Due to Data Quality
220.127.116.11 Specific Methods Due to the Two World Wars
18.104.22.168 Specific Methods Due to Territorial Changes
22.214.171.124 Specific Methods Due to Missing Data
5.3.3 Reliability of the Data and Comparison with Other Studies
5.4 Available Results and Discussion
5.4.1 Available Results
126.96.36.199 Census Reliability
188.8.131.52 Interdepartmental Migrations
184.108.40.206 Domiciliation of Deaths during the Two World Wars
220.127.116.11 Small Département Figures
5.6.1 Computations of Population on 1st January
18.104.22.168 Intercensal Survival
22.214.171.124 Precensal Survival Method
126.96.36.199 Extinct Cohorts Method
188.8.131.52 Survivor Ratio Method
5.6.2 Set of Different Lifetables
5.6.3 Census Adjustments
184.108.40.206 Distribution of Deaths of Unknown Age in 1901 Census
220.127.116.11 Addition of Age Group for Pre-1946 Censuses
18.104.22.168 Adjustment of Censuses by Cubic Splines
5.6.4 Estimates of Military Deaths during the Two World Wars
5.6.5 Estimates of Deportees
22.214.171.124 Born-abroad Deportees
126.96.36.199 French Deportees
5.6.6 Missing Data During the Two World Wars
188.8.131.52 Births and stillbirths
5.6.7 Reorganization of Ile-de-France in 1968
5.6.8 Computations of 1st January Populations by Class of Départements
5.6.9 Sources of Raw Data
6 Computations of French Income Distributions by Département, 1960–2014
6.2 Data and Spatial Framework
6.2.1 Fiscal Data at the Departmental Level
6.2.2 Departmental Demographic Data
6.2.3 National Data
6.2.4 Spatial Framework
6.3.1 Period 1986–2014
184.108.40.206 Raw Fiscal Statistics Available
220.127.116.11 Taxable Income and Fiscal Income Distributions
6.3.2 Period 1960–1969
18.104.22.168 Computations of Total Tax Units by Département
22.214.171.124 Computations of Fiscal Income by Département
126.96.36.199 Computations of Fiscal Income Distributions for all Tax Units
6.3.3 Template of Fiscal Income Distributions by Département
6.4.1 Fiscal Income per Adult
6.4.2 Intra-departemental Inequalities
6.4.3 Spatial Distribution of Tax Units Belonging to each Fractile
6.6.1 Computations of Departmental Fiscal Incomes Using Regional Accounting
6.6.2 Supplementary Materials