Navigation – Plan du site
2019
763

Is the French diagonal emptying? An exploratory spatial analysis of demographic decrease in France for the past 50 years

Sébastien Oliveau et Yoann Doignon
Cet article est une traduction de :
La diagonale se vide ? Analyse spatiale exploratoire des décroissances démographiques en France métropolitaine depuis 50 ans

Résumé

This article’s purpose is to draw attention to the results of the latest censuses in France, and to the dichotomy that remains between space and population in terms of settlement dynamics. A detailed examination on different levels – from counties (“départements”) to villages/towns (“communes”)- shows territories where the demographic dynamics are opposed. These dynamics don’t fit with commonplace conceptions, even in the scientific world. Beyond a mark that’s still visible and quantifiable, the “diagonal of emptiness” (a line drawn between the Pyrenees and the Ardennes) has actually not totally emptied. However, its geography is varied and can only be evaluated with thorough consideration removed from preconceptions. This is the point of our article, based on sharp cartography and measurements of spatial patterns. These indicate that some spaces in France seem to be literally abandoned by human beings.
population geography, demographic change, spatial autocorrelation

Haut de page

Texte intégral

We would like in particular to thank Jean-Philippe Antoni and the organizers of the 2013 Théoquant conference, where we were able to have a first discussion with our geographer colleagues. This work was conducted as part of the Demographic Observatory of the Mediterranean - DemoMed (http://demomed.org), led by Isabelle Blöss.

Introduction

1In the 1970s and 80s, French geography saw an important debate about the existence and qualification of a "diagonal of emptiness". The term, whose exact origin is not known (Fumey, 2009), describes a diagonal band which crosses from south-west to north-east France and is characterised by lower population densities than elsewhere. The main focus of the debate was the interpretation of the social and economic dynamics of low density regions. This preoccupation is an old one: as far back as 1884, the theme of the Congress of the Society of French Geographers was the depopulation and sterilisation of vast regions (Mathieu, Duboscq, 1985). JF Gravier (1947) establishes a link between population and economy by representing the unequal distribution of men as an obstacle to obtaining the economic optimum. From the 1960s, in scientific texts and official speeches, the issue of desertion is increasingly associated with rural outmigrations, depopulation and desertification (Mathieu, Duboscq, 1985). Desertification of a territory was defined as the destruction of its social life and permanent degradation of its vegetation and soils. With the development of theories of a "sociability threshold" and "irreversible threshold of de-settlement", desertification takes on a pessimistic and alarmist connotation. In the 1970s, the seriousness of the problem grew with the closure of public services in "deserted" regions and a deterioration in the living conditions of these territories. Regional planning policies then became more concerned with regions with low population densities. Yet some researchers take a stand against the discourse of emptiness and desertification (Faudry, Tauveron, 1975, Larrère, 1976, Bontron, Mathieu, 1976). They oppose the notions of desertification and the threshold of social viability, as well as the pejorative and alarmist vocabulary (in particular the symptomatic expression La France du vide by R. Béteille (1981)) by advocating the use of a more neutral vocabulary.

2The purpose of this article is not to assume a position in this debate, but to introduce, 30 years later, an element of demographic and spatial clarification.

3In recent years, a number of authors have once again become interested in the subject. They rely on the interpretation of recent censuses to show that these regions are once again experiencing positive population growth. Didier Boutet (2006) thus speaks of "demographic reconquest of isolated rural areas", Jean Laganier and Dalila Vienne (2009) of "rediscovered growth". The analyses proposed by Hervé Lebras (2012) are more measured, but no one seriously believes that the population decline is continuing. However, although some areas have experienced a revival, in particular the cities (Dubuc, 2004), the latest census results persist in showing a population decline for a large part of this "diagonal". The scale of the phenomenon is, moreover, large enough to be observed at the scale of several counties ("départements"). This is why we have chosen to explore the phenomenon at different scales, by adopting an exploratory approach to the data (Tukey, 1977; Banos, 2001).

4We believe that the latest census data may not have received the attention it deserves. Even if the French population continues to grow as a whole, its growth is uneven. When we take a closer look, we see that some regions are not experiencing population growth at all, but are continuing to lose population. Therefore, we considered it useful to delve further by proposing a systematic description of the French territory which distinguishes the components of population dynamic at work, particularly the regions experiencing depopulation (population decrease linked to a negative natural increase) and / or de-settlement (population decrease linked to negative net migration). We relied on figures from censuses since 1968 compiled by INSEE. Our study thus resituates regions and their population in time, so as not to focus on the most recent trends.

  • 1 This was done using GeoDa software (Anselin, 2005).

5These data were georeferenced in order to firstly propose a cartography, and to secondly explore their spatial configuration using spatial autocorrelation measurements1.

  • 2 The data used were harmonised by INSEE over the entire period. We therefore assume the territory is (...)

6The first part of our study focuses on a demographic description of counties, with a resulting typology of the different trajectories they may have followed. This typology is then applied to two other administrative levels, districts (Cantons) and municipalities (communes)2, emphasising the complementary relationship between population and region. Next, the description is supplemented with an analysis of the spatial structure of the phenomenon at different levels. In order to understand the dynamics underlying the spatial structures observed, we focus on the population dynamic components of districts, which appear to be the most appropriate level for our study.

7Finally, we show that a large area of French territory is still experiencing a loss of population, even if the population concerned is minimal. This article focuses on population decline and does not intend to address the issue of economic dynamics, which are not systematically linked.

What does the "diagonal of emptiness" look like?

  • 3 The striking term "diagonal of emptiness" marked the collective imagination of French geography. It (...)

8Before we enter into a geographic description of the relevant regions3, it is essential to describe their demographic evolution in detail. We have data for three administrative levels: counties, districts and municipalities. We will start with a study at the most aggregated level (i.e. counties) and then refine the analysis with the district level, which will remain our baseline, although municipalities will also be referenced regularly. In the geographic literature of the last 35 years it is well acknowledged that the change of administrative level causes variations in the results obtained, related to the shape and size of the grid (Mathian, Piron, 1999). This phenomenon is usually described by the term MAUP (Modifiable Areal Unit Problem, Openshaw, 1981). We therefore expect to encounter the issue during our analysis when changing level.

9For 50 years or more, counties have been losing population and this phenomenon continues. The first observation that we were able to make is the steady decline that some counties are experiencing. Six of them (in dark red on map 1) have experienced a steady decline in population since 1968 (Creuse, Haute-Marne, Cantal, Allier, Nièvre, Ardennes). Their location is not contiguous, but in a way they form the skeleton of the diagonal. 6 other counties (in light red) have had less regular trajectories (decline and growth according to the censuses), but are characterized throughout the period by a final decline in their population. Moreover, none of them has experienced population growth above the national average (Paris, Meuse, Lozere, Aveyron, Vosges, Indre). Paris is a specific case. It is more of a construct, due to its administrative delimitation rather than the real dynamics observed in the metropolis: intra-mural population decline and population growth of the city far beyond the walls.

10These 12 counties (in red) accounted for 10.6% of the population in 1968 and only 7.5% in 2009. If we discount Paris, these 11 counties only actually accounted for 5.4% of the population in 1968 and only 4% of the population today.

A typology of observed trajectories

  • 4 Despite varied intercensal growth (decrease and increase), the units have the same population in 20 (...)

11It appears that a number of territories have followed similar population growth trajectories for half a century. We therefore sought to define the different routes taken. In order to do this, we systematically examined intercensal growth for each of them. The resulting typology describes the trajectory of the territories between each census and gives an overview of the changes between 1968 and 2009 (figure 1). This method yields a robust typology, since the initial choice of 1968, 1975 or another year does not fundamentally change the results. Indeed, a unit that is in "absolute and steady decline" is characterised by a population decline between censuses. The starting year has no impact. On the other hand, the split between positive and negative growth seemed too reductive. This is why we have chosen to use intercensal mean growth as a way of distinguishing, within the group of growing territories, territories continuing to increase in population concentration from territories which, despite an increase in population, are experiencing a slowdown compared to the whole group. The result is a typology that distinguishes three main types of trajectories: absolute loss of population, increase in population but at a slower rate, increase in population and reinforcement of population concentration. Each of these types is then divided into units with uniform paths (same behaviour between each intercensal period) and those with heterogeneous paths (possibility of change between intercensal periods). A seventh, marginal category consists of units whose population growth between 1968 and 2009 is zero4.

Figure 1: Segment tree of trajectory classification

Figure 1: Segment tree of trajectory classification

Map 1: Population growth: variations in dynamics at county level

Map 1: Population growth: variations in dynamics at county level

12Map 1 highlights the concentration of the most dynamic counties around 3 main hubs: the Paris region, the Atlantic coast and a large part of the south-east stretching from the Toulouse region to the Alps, excluding the Massif Central. These counties show a hollow diagonal that has been empty for 40 years or at least barely remains populated. One notable group is the most northerly counties of France which, while not actually losing any population, show a decrease in their relative weight over the entire period.

1322 counties experienced population growth over the entire period from 1968 to 2009 (in dark green), but only during phases where population growth is below the national average, or phases of absolute decline (negative growth). The result is a steady loss of their relative weight over time. They represented 26.7% of the population in 1968 and represent only 22.8% of it in 2009.

1421 counties (in light green) also experienced population growth below the national average despite periods of increase (their weight went down from 19.1% of the population in 1968 to 17.8% in 2009). As in the previous case, they experienced a slowdown, but this is sometimes kept in check by larger increases than average, especially during the last decade.

15There are therefore 41 counties whose population has grown faster than average over the period 1968-2009. Of these, 26 (in light blue) experienced periods of population growth, but also one or more periods of population decline (absolute or relative) during the period. Their demographic weight increased from 29.3% of the national population in 1968 to 32.99% in 2009. Finally, only 15 counties (in dark blue) experienced a steady increase above the national average over the entire period. These counties still represent almost 1/5th of the population today, their relative weight having increased from 14.24% of the national population to 18.94%.

Area and population

16A population study cannot be made independently of its region. The region can be viewed in several ways. At different geographic levels, the perception and measurement of the phenomenon are not necessarily the same.

17The INSEE data make this multiscalar investigation possible. The population dynamic of regions can thus be observed at a more detailed level (see appendix 1). We can observe a higher incidence of population decline than when measured at county level.

18While 12.5% of counties have experienced an absolute population decline since 1968, the same applies to 30% of districts. They represent only 26% of the total population in 2009 (compared to 37% in 1968) and above all 36% of the metropolitan area. A third of the country has seen its population decrease in the last forty years.

19As expected, the MAUP is in evidence here: the change in level causes a change in how the phenomenon is perceived. The perception of the latter becomes even more nuanced if we look at the individual trajectories.

Map 2: Population dynamic of French districts between 1968 and 2009

Map 2: Population dynamic of French districts between 1968 and 2009
  • 5 INSEE defines community as "the smallest territory in which the inhabitants have access to the most (...)

20At district level, the map of France becomes more legible (see map 2). We can see the fragmentation of France between a coastal fringe that has experienced continuous population growth for 40 years, and inland regions, which decrease as soon as one leaves the big metropolises and their periphery. Furthermore, this geography is reminiscent of Pierre Pistre's study (2011), conducted at community level5, in which he evokes the idea of "fragile countryside".

21At municipal level, the proportion of territories in absolute population decline is increasing: 36% of municipalities have a 2009 population lower than that of 1968. While they accommodated 41% of the population in 1968, they include only 28% today, losing more than 3 million inhabitants (15% of their population over the period). It can always be argued that most municipalities have been picking up in the last decade, however 4.8% of municipalities did not experience any period of population growth between 1968 and 2009. This still represents 1.7 million people. In addition, these municipalities are often contiguous and thus form pockets of population decline.

Map 3: Population dynamics of French municipalities between 1968 and 2009

Map 3: Population dynamics of French municipalities between 1968 and 2009

22The previously observed spatial structure continues to be clearly visible at a global level (see map 3). The distribution of municipalities according to their demographic dynamics does not appear to be random. The municipalities which have lost population during the last forty years continue to form a long ribbon which starts at the Pyrenees and goes up to the Ardennes. In addition, at municipality level, a larger number of areas in population decline appear around the edges of small urban hubs in growth compared to the district map.

23Even though a large number of rural towns and villages have shown a recovery, as have their direct neighbourhoods (see the work of Sylvie Dubuc on Aveyron and Lozère, 2004), traces of what some call the "diagonal of emptiness" can still be clearly seen in their dynamics. Further afield, inland Brittany, with the exception of Rennes and its hinterland, also experienced constant losses. Corsica shows a similar dynamic, with a contrast between coastal and inland regions, with the exception of Corte. Municipalities experiencing a slowdown are at the interface of regions of declining population and those who have experienced the largest increases.

  • 6 We refer this point to the interesting 2013 article by Wolff et al. focusing on French urban dynami (...)

24Locally, gradients are often observed, rather than discontinuity. Some municipalities in large urban centres also experience population decline at their centre (which seems to correspond to the "density craters" described by the Newling model, 1969)6. Paris and Marseille appear clearly on the map. Nevertheless, they represent rather small areas, and if the population amounts affected by population decline are to be taken seriously, these urban centres only moderately vary the conclusions in terms of surface area.

25Appendix 1 gives a detailed summary of the results mentioned above. Methodologically, we observe in this appendix a nice aggregation effect from the MAUP. The transition from one level to another causes great variation in the trajectories observed. Thus, at the highest level of aggregation, we can no longer see any units whose population is stagnating. Similarly, the "permanent slowdown" category represents almost a quarter of counties, but only 2.66% of districts and 0.42% of municipalities...

Measurement of the spatial autocorrelation of population growth: between continuity and spatial fragmentation

26From our maps, can we be justified in believing that the declining population areas form a continual "diagonal"? Indeed, reading maps is often insufficient or misleading as regards the strength of correlations (Müller, 1977). Their interpretation may differ among readers, and we now have robust methods for measuring local clusters, such as local indicators of spatial association (also known as LISA, Anselin 1995). There are restrictions nonetheless, since with these numerical methods do not allow working on a qualitative typology or integrating different dates. They are restricted to continuous quantitative values. So we chose to work only on population growth rates between 1968 and 2009. The three levels considered are examined one after the other to see if any spatial structures stand out. In other words, we have tried to see if there are spatial groupings of units at particular levels whose population growth is statistically significant. We also know that statistical measures of spatial phenomena are sensitive to the level of analysis (Mathian, Piron, 1999) and that LISAs are no exception to the rule (Oliveau, 2010).

27LISAs are disaggregated forms of spatial autocorrelation indicators. We can therefore first observe the global behaviour of spatial structures through the study of global indices, then analyse the effects of changes in the type of neighbourhood. In order to be able to compare levels, we used neighbourhoods defined by distance. Comparing different levels of aggregation prevents the definition of neighbourhoods using contiguity, since there would be no correspondence between the types. Conversely, distance makes it possible to present different levels in the same graph.

Figure 2: Spatial autocorrelation of population growth between 1968 and 2009.

Figure 2: Spatial autocorrelation of population growth between 1968 and 2009.
  • 7 At 50 km, municipalities can already have more than a thousand neighbours - which may no longer rea (...)

28Figure 2 presents measurements of the local spatial autocorrelation of population growth for different distance steps (measured from the centroids of spatial units) at the three levels of our study. As for counties, Moran's I could not be calculated for distances less than 100 km, because the numbers are insufficient for a spatial autocorrelation measurement. Equally, measurements at municipality level start at 20 km, but could not be made beyond 60 km, since the neighbourhood matrices became too large for the software7. District measurements were carried out to allow comparisons between the different levels.

  • 8 We note in passing that two "died for France" municipalities of the Meuse (Fleury-devant-Douaumont (...)

29It is essential to note the importance of the level of spatial autocorrelation for short distances: 0.47 for municipalities8 and 0.48 for districts in a neighbourhood of 20 km. In other words, almost half of the observed variability in population growth is related to the values observed in the neighbourhoods of the units. Spatial autocorrelation nevertheless decreases rapidly enough to drop below 0.2 after 80 km. The spatial structure of population growth is very strong at the scale of the immediate neighbourhood, but its regional structure is not very marked. This corresponds with the previous reading of local gradients around central places that polarise population growth. The measurements performed at county level (0.34 in a neighbourhood of 100 km) confirm a tendency towards regrouping, but this is not the level at which the standard trends are most strongly manifested

Map 4: Spatial structure of growth of French districts between 1968 and 2009.

Map 4: Spatial structure of growth of French districts between 1968 and 2009.

30The map of LISA values (map 4) clearly shows the strong spatial structure of the trend in population evolution on a national scale. We can see the "diagonal of emptiness" running north along the Belgian border. We can also see inland Brittany. The dynamics of large cities also appear clearly in the form of large areas of strong growth, with small central municipalities that are becoming depopulated in favour of their surrounding regions. From this perspective, Toulouse creates a real rupture within the diagonal, whereas Clermont-Ferrand appears more troubled.

31Since some authors emphasise the resumption of population growth during the last intercensal period, it seemed appropriate to explore the spatial dimension. The spatial structure of population decline has changed over the entire period, from an overall Moran's I of 0.22 for the period 1968-1975 to 0.36 for the period 1999-2009 (see LISA map in appendix 2). During the period from 1968-2009, however, the spatial structure measured by Moran's I is 0.38. This means that spatially related units have tended to increase in similarity over time. The regions that have gained the most population are located in the same areas (particularly Toulouse, the Mediterranean and Atlantic coasts, Rhône-Alpes). On the other hand, declining or slowing regions show remarkable spatial homogeneity over time: from the Massif Central to the north and east of France, inland Brittany, etc.

The components of population decline

32Without taking the analysis of the processes of population decline too far, it is important to dwell on its components. In the same situation of population growth/decline very different demographic dynamics can be observed. Two components can be used to explain the growth of a population: the natural component (subtraction of births from deaths) and the migratory component (subtraction of immigrants from emigrants). There are therefore four population dynamic scenarios: both components are positive, both components negative, one component positive and the other negative. A population with a migratory deficit has a positive natural increase (more births than deaths), but a negative migratory balance (more emigrants than immigrants). A population with a natural deficit is in the opposite situation. However, it is important to remember that the resumption of population growth through migratory surplus or natural surplus does not have the same implications for future demographic growth. Indeed, a natural surplus carries the potential for renewal of future population growth (Chesnais, 2010). Children born in these regions are potential future parents once they reach the age of maternity. The impact of the migration phenomenon depends on the age of the migrants. Young migrants permanently rejuvenate the population and can induce a revival of natural growth. Conversely, older migrants transform the structure of the population by predisposing it to a future sustainable population decline. In addition, immigration is an uncertain and fluctuating phenomenon, in contrast to a natural increase which is very closely dependent on the structure of the population. One can thus consider, without it being a value judgment, that population growth by natural surplus is more durable and less prone to fluctuation than that related to migration.

33We decided to study the population dynamic components at district level (figure 3). This choice is justified by greater precision than at county level, while eliminating some of the statistical noise at municipal level. However, caution must be exercised when interpreting the results presented. It should be recalled that the calculations involve relative data, which may, in absolute terms, represent small numbers.

34Over the period as a whole, the profile of the district's population dynamic changes profoundly, with a strong increase in the districts in natural deficit, and a decrease in the districts in migratory deficit. A similarity can be seen between districts having two positive components and those in migratory and natural deficit.

35There is a significant change in the dynamics of districts over time (see figure 3). Between 1968 and 1975, the two components are positive for 37% of the districts, compared to 22% of the districts having both negative components. Between 1999 and 2009, these proportions are respectively 45% and 4%. These trends can be explained by the decrease in the phenomenon of de-settlement. There are fewer districts in both natural and migratory deficit. However, another development must be emphasised: the ever greater proportion of districts with only a natural deficit. From 10% during the period 1968-1975, this increases to 32% of all districts for the last intercensal period. At the same time, the percentage of districts with only a migratory deficit decreased by 10 points between 1968-1975 and 1999-2009.

Figure 3: Population dynamic of districts from 1968 to 2009

Figure 3: Population dynamic of districts from 1968 to 2009

36However, to stop at this general picture would be misleading in the sense that it is misleading to attribute these profiles to districts experiencing de-settlement. Among the districts with natural and/or migratory deficit, some registered a decrease in population and others did not. A natural deficit can be offset by a high migratory surplus, and vice versa. It is then necessary to focus only on the districts in absolute population decline over the period and to study the population dynamic (figure 4).

37As noted above, the population dynamic profile changes radically between 1968 and 2009. In the first intercensal period, very few districts are exclusively in natural deficit (4%), or with both components positive (5%). On the other hand, almost half of the districts (48%) have a natural and migratory deficit, while 43% have a migration-only deficit.

38In the last intercensal period, the share of districts in only natural deficit increases significantly to 58%. There is a doubling of the proportion of districts with both components positive (12%), the other two scenarios (both negative and migratory deficit) being sharply reduced.

39The same general observation can be made when we study only the districts of each intercensal period in demographic decline (see figure 5).

Figure 4: Population dynamic of districts with a population in 2009 lower than in 1968

Figure 4: Population dynamic of districts with a population in 2009 lower than in 1968

Figure 5: Population dynamic of districts in population decline between 1968 and 2009

Figure 5: Population dynamic of districts in population decline between 1968 and 2009

40The map illustrates this change in population dynamics between 1968 and 2009. It shows that population dynamics are not randomly distributed across the region. We have selected the first and the last intercensal period, which are typical of the observed changes (see maps 5 and 6).

Map 5: Population dynamic (1968-1975) of districts in population decline between 1968 and 2009.

Map 5: Population dynamic (1968-1975) of districts in population decline between 1968 and 2009.

41The period 1968-1975 (see map 5) is characterised by both a natural and migratory deficit for most of the districts observed. A diagonal starting from the Pyrenees and going up to the South of Champagne is marked by a joint deficit in net migration and natural increase, just like the heart of Brittany. Beyond that, it is mainly net migration that is negative. For the period 1999-2009 (see map 6), the situation has changed dramatically. The whole of France, with the exception of Nord-Pas de Calais and Champagne-Ardenne, is characterised above all by a natural increase deficit, mainly due to a positive recovery in migrations. Lorraine, Nord-Pas de Calais and Champagne-Ardenne are relatively specialised areas with an industrial economy and thus experiencing deindustrialisation. This phenomenon can be observed even for a large number of small and medium-sized industrial cities, such as the coalfield of Lorraine, the Meuse valley in the Ardennes, the industrial region of Lens, etc.

Map 6: Population dynamic (1999-2009) of districts in population decline between 1968 and 2009.

Map 6: Population dynamic (1999-2009) of districts in population decline between 1968 and 2009.

42Some of the selected districts have certainly experienced a population recovery since 1968 or more recently, but this is primarily due to migration (57%). The countryside attracts new inhabitants, mostly retirees (Pistre, 2011). The migration observed here, which relies heavily on retirees, therefore projects an image of demographic recovery whose sustainability is questionable. Retirees breathe new life into these regions, but we can doubt the sustainability of this population growth in the coming decades. It would seem more appropriate to speak of a rebound than a renewal. Moreover, from an economic perspective, Emmanuelle Cunningham-Sabot et al. (2010) include in the category of shrinking regions, regions whose economy depends on the immigration of the elderly. There is a growing residential economy (Davenzies, 2009), but it is directly related to the survival of the newcomers. Moreover, Claude Grasland identifies the arrival of elderly people, combined with the departure of young workers, as an accelerator in the downturn of areas in population decline (2010). Thus, among all the districts in population recovery at some point during the period, 24% of them are again in a situation of population decline during the next intercensal period. 52% of them were no longer in a state of de-settlement because of migrations. This finding shows how fragile the recovery of population growth can be. The return of population growth after a phase of decline is not a guarantee of sustainability. Population decline may resume, especially when the change is a result of elderly migration.

43In addition, the north-east part of the diagonal continues to experience a local dynamic of population decline linked to natural and migratory deficits, which raises questions about its future (Milian, Barthe, 2012).

44Beyond the purely demographic questions, these results prompt geographic, social and political reflections. The inability of the current populations to sustain the long-term management of these regions raises important points: the disappearance of landscapes (Sgard, 1991, Le Floch, 2005), uncontrolled return of wildlife, maintenance of roads, etc. Moreover, what will become of this countryside when this population has aged and requires social support? These populations have little in the way of services, whereas as they get older some of them will be increasingly dependent on them. Many services have been closed in recent decades in declining areas, such as health services (Savignat, 2013) or the schools needed to accommodate the workforce required for elder care. As Olivier Véran says: "Medical deserts strike in places the state has already deserted: transport, schools, public and private services" (2013: p.82).

45Finally, and from a national perspective, it seems fair enough to question the political weight of the populations that live in these regions, who are underrepresented in national bodies (National Assembly and Senate). On the other hand, the composition of these districts is very specific, marked by under-representation of young and working people (Morin, 2011).

Conclusion

46This article proposed to explore the spatiality of population growth over the past 40 years, viewed from the perspective of regions losing population. If the processes behind this population decline are simple (natural increase and net migration), their interaction makes their analysis more complex due to the variety of possible trajectories. Nevertheless, we were still able to observe the resulting spatial forms: coastal France (the "litturbanisation" of Dumont, 1996) versus inland France, urban spread and "para-urbanization" (Dumont, 1996) versus a downturn in isolated rural areas. We did not wish to enter into a study of differentiated dynamics by region type (rural/urban).

47The composition of this population growth dynamic therefore calls for a careful assessment of the population "recovery" of the rural world. It could be believed that the overriding tendency is towards urban spreading, accompanied by a downturn in the most remote rural regions which, due to a lack of inhabitants and a reserve of migrants, will certainly end up disappearing. This is not without consequences, both social and territorial. First social, because still more than a million people are affected by what we must call desertification, making life ever more difficult (loss of services, increased isolation) (Bontron, 2013). Then territorial, because these regions in de-settlement are also regions over which we have less and less control and for which a recovery will be increasingly difficult as they become depopulated, even if this situation is more nuanced (Maigrot, 2003).

Haut de page

Bibliographie

Anselin L., 1995, "Local indicators of spatial association - LISA", Geographical Analysis, Vol. 27, N°2, 93-115.

Anselin L., Syabri I., Kho Y., 2005, ""GeoDa: An Introduction to Spatial Data Analysis",Geographical Analysis, Vol.38, N°1,5-22.

BanosA., 2001, "A propos de l’analyse spatiale exploratoire des données", Cybergeo : European Journal of Geography, [En ligne], Systèmes, Modélisation, Géostatistiques, N°197, 18 octobre 2001. URL : http://cybergeo.revues.org/4056 ; DOI : 10.4000/cybergeo.4056

Béteille R., 1981, La France du vide, Paris, Editions Litec, Coll. "Géographie économique et sociale".

Boutet D., 2006, "L’importance d’une dynamique résidentielle dans le rural isolé", RERU, N°5, 781-798.

Bontron J.-C., Mathieu N., 1976, La France des faibles densités, II : Documentation – bibliographie – analyses d’études, Paris, SEGESA-ACEAR.

Bontron J.-C., 2013, "L’accès aux soins des personnes âgées en milieu rural : problématiques et expériences", Gérontologie et société, Vol.3, N°146, 153-171.

Chesnais J.-C., 2010, La démographie, Que sais-je, Paris, PUF.

Cunningham-Sabot E., Fol S., Grasland C., Roth H., Van Hamme G., 2010, "Shrinking cities etShrinking regions. Définitions et typologies", in Baron M., Cunningham-Sabot E., Grasland C., Rivière D., Van Hamme G., Villes et régions européenne en décroissance. Maintenir la cohésion territoriale, Paris, Hermès Lavoisier, 67-95.

Dubuc S. 2004, "Dynamisme rural : l’effet des petitesvilles", L’Espace Géographique, N 1, 69-85.

Dumont G.-F., 1996, Les spécificités démographiques des régions et l’aménagement du territoire, Rapport du Conseil économique et social, Paris, Editions des journaux officiels.

Faudry D., Tauveron A., 1975, Désertification ou réutilisation de l’espace rural ?, Grenoble, IREP-INERM.

Fumey G., 2009, "la France en diagonales", Café-géo.net, Vox Geographica, [En ligne] http://cafe-geo.net/wp-content/uploads/france-en-diagonales.pdf

Grasland C., 2010, "Les mécanismes démographiques de la décroissance : 3 boucles de rétroaction", in Baron M., Cunningham-Sabot E., Grasland C., Rivière D., Van Hamme G., Villes et régions européenne en décroissance. Maintenir la cohésion territoriale, Paris, Hermès Lavoisier,97-116.

Gravier J.F., 1947, Paris et le désert français, décentralisation, équipement, population, Paris, Le Portulan.

Laganier J., Vienne D., 2009, "Recensement de la population 2009 ; la croissance retrouvée des espaces ruraux et des villes", Insee Première, 1218.

Larrère G., 1976, Dépeuplement et annexion de l’espace rural : le rôle de la théorie des seuils de sociabilité, Toulouse le Mirail, Coll. Géodoc, N°7, Institut de Géographie.

Lebras H., 2012, "Mobilité et attractivité territoriales : éléments pour 2040", in DATAR, "Des facteurs de changement 1", Territoires 2040, N°6, 37-48.

Le Floch S., Devanne A.-S., Deffontaines J.-P., 2005, "La "fermeture du paysage" : au-delà du phénomène, petite chronique d’une construction sociale", L’Espace Géographique, N°1, 49-64.

Maigrot J.-L., 2003, "Dépeuplement rural, maîtrise agricole et évolution des écosystèmes", L’Espace Géographique, N°3, 253-263.

Mathian H., Piron M., 2001. "Échelles géographiques et méthodes statistiques multidimensionnelles», in : Sanders L. (dir.), Modèles en analyse spatiale, Paris,Hermes Lavoisier, coll. «Information géographique et aménagement du territoire", 61-104.

Mathieu N., Duboscq P. (dir.), 1985, Voyage en France par les pays de faible densité, Toulouse, Editions du CNRS.

Milian J, Barthe L., 2012, "Les espaces de la faible densité", in : DATAR, "Des systèmes spatiaux en perspective", Territoires 2040, N°3, 141-159.

Morin B., 2011, "Qui habite en milieu rural ? ", Informations sociales, N°164, 11-22.

Müller J.-C., 1977, "Comparaison visuelle des cartes et groupements spatiaux", L’Espace Géographique, N°1, 59-72.

Newling B.E., 1969, "The spatial variation of population densities", Geographical Review, Vol.59,N°2, 242-252.

Oliveau S., 2010, "Autocorrélation spatiale : leçons du changement d’échelle", L’Espace Géographique, N°1, 51-64.

Openshaw S., 1981, "Le problème de l'agrégation spatiale en géographie", L'Espace Géographique, N°1,15-24.

Pistre P., 2011, "Migrations résidentielles et renouveaux démographiques des campagnes françaises métropolitaines", Espaces, Populations, Sociétés, N°3,539-555.

Savignat P., 2013, "Désert médicaux, vieillissement et politiques publiques : des choix qui restent à faire", Gérontologie et société, Vol.3, N°146, 143-153.

Sgard P., 1991, "Quelques aspects de la gestion paysagère de l’espace rural", Etudes rurales, Vol.121, N°1, 207-212.

Tukey J.W., 1977, Exploratory Data Analysis, Reading (Massachusetts), Addison-Wesley.

Véran O., 2013, "Des bacs à sable aux déserts médicaux : construction sociale d’un problème public", Les tribunes de la santé, Vol.2, N°39, 77-85.

Wolff M., Fol, S., Roth, H., Cunningham-Sabot, E., 2013, "ShrinkingCities, villes en décroissance : une mesure du phénomène en France", Cybergeo : European Journal of Geography [En ligne], Aménagement, Urbanisme, document 661, 08 décembre 2013. URL : http://cybergeo.revues.org/26136 ; DOI : 10.4000/cybergeo.26136

Haut de page

Annexe

Appendix 1: Summary of population dynamics on the different scales studied9.

Appendix 1: Summary of population dynamics on the different scales studied9.

Appendix 2: LISA cartography of the spatial autocorrelation of population growth of French districts between 1968 and 1975 and between 1999 and 2009.

Appendix 2: LISA cartography of the spatial autocorrelation of population growth of French districts between 1968 and 1975 and between 1999 and 2009.

Haut de page

Notes

1 This was done using GeoDa software (Anselin, 2005).

2 The data used were harmonised by INSEE over the entire period. We therefore assume the territory is constant according to the most recent grid.

3 The striking term "diagonal of emptiness" marked the collective imagination of French geography. It obviously lacks nuance and carries with it an ideology of economic growth that considers low densities as regions of non-development. For our part, we dissociate the notion of economic growth from that of population growth, which is why we reserve our analysis for the latter without considering the former.

4 Despite varied intercensal growth (decrease and increase), the units have the same population in 2009 as in 1968. While no county has experienced stagnation, the district of Saint-Pé-de-Bigorre (65) sees its population of 2009 exactly equal to that of 1968. This anecdotal situation also concerns 157 municipalities.

5 INSEE defines community as "the smallest territory in which the inhabitants have access to the most common equipment and services. We delineate it in several stages. First of all, a service centre is defined as a municipality or urban unit with at least 16 of the 31 intermediate facilities. The zones of influence of each service centre are then delimited by grouping the closest municipalities, proximity being measured in travel time by road at off-peak time. Thus, for each municipality and for each piece of equipment not present in the municipality, the closest municipality proposing this equipment is determined. Intermediate equipment but also local equipment are taken into account "(http://www.insee.fr/en/methodes/default.asp?page=zonages/ basin-vie-2012.htm).

6 We refer this point to the interesting 2013 article by Wolff et al. focusing on French urban dynamics.

7 At 50 km, municipalities can already have more than a thousand neighbours - which may no longer really make sense - and the neighbourhood matrix exceeds twenty million pairs of neighbours.

8 We note in passing that two "died for France" municipalities of the Meuse (Fleury-devant-Douaumont and Cumières-Le-Mort-Man), have a non-null population in 1968, but null in 2009. Their inclusion in the measurements creates a significant measurement bias. Moran's I increases from 0.34 to 0.47 when they are removed. So we decided not to take them into account.

The six "died for France" municipalities (Beaumont-en-Verdunois, Bezonvaux, Cumières-le-Mort-Man, Fleury-devant-Douaumont, Haumont-near-Samogneux and Louvemont-Côte-du-Poivre) were totally devastated after the Battle of Verdun in 1916 and were never rebuilt, due to the excessive presence of unexploded ordnance and disturbed and polluted soils" (Wikipedia).

9 The differences observed between municipalities and districts relative to the populations of 1968 and 2009 and also the surface area correspond to the municipalities of Saint Symphorien (35317), Carolles (50102), Ruffiac (47227), Tremilly (52495), Plevenon (22201) which do not appear in the calculations because they do not have a population in 1999, due to administrative factors.

The differences observed (less than 0.1%) at different scales in the surface area relate to changes in geographic background in the GIS. We used the IGNF GEOFLA database for the municipalities. For the districts and counties, the data offered by the GADM of the University of California at Davis were used. There was no concordance between the GEOFLA data at district level and the INSEE data for this scale.

Haut de page

Table des illustrations

Titre Figure 1: Segment tree of trajectory classification
URL http://journals.openedition.org/cybergeo/docannexe/image/31641/img-1.png
Fichier image/png, 58k
Titre Map 1: Population growth: variations in dynamics at county level
URL http://journals.openedition.org/cybergeo/docannexe/image/31641/img-2.png
Fichier image/png, 235k
Titre Map 2: Population dynamic of French districts between 1968 and 2009
URL http://journals.openedition.org/cybergeo/docannexe/image/31641/img-3.png
Fichier image/png, 505k
Titre Map 3: Population dynamics of French municipalities between 1968 and 2009
URL http://journals.openedition.org/cybergeo/docannexe/image/31641/img-4.png
Fichier image/png, 509k
Titre Figure 2: Spatial autocorrelation of population growth between 1968 and 2009.
URL http://journals.openedition.org/cybergeo/docannexe/image/31641/img-5.png
Fichier image/png, 29k
Titre Map 4: Spatial structure of growth of French districts between 1968 and 2009.
URL http://journals.openedition.org/cybergeo/docannexe/image/31641/img-6.png
Fichier image/png, 430k
Titre Figure 3: Population dynamic of districts from 1968 to 2009
URL http://journals.openedition.org/cybergeo/docannexe/image/31641/img-7.png
Fichier image/png, 20k
Titre Figure 4: Population dynamic of districts with a population in 2009 lower than in 1968
URL http://journals.openedition.org/cybergeo/docannexe/image/31641/img-8.png
Fichier image/png, 22k
Titre Figure 5: Population dynamic of districts in population decline between 1968 and 2009
URL http://journals.openedition.org/cybergeo/docannexe/image/31641/img-9.png
Fichier image/png, 26k
Titre Map 5: Population dynamic (1968-1975) of districts in population decline between 1968 and 2009.
URL http://journals.openedition.org/cybergeo/docannexe/image/31641/img-10.png
Fichier image/png, 252k
Titre Map 6: Population dynamic (1999-2009) of districts in population decline between 1968 and 2009.
URL http://journals.openedition.org/cybergeo/docannexe/image/31641/img-11.png
Fichier image/png, 263k
Titre Appendix 1: Summary of population dynamics on the different scales studied9.
URL http://journals.openedition.org/cybergeo/docannexe/image/31641/img-12.png
Fichier image/png, 98k
Titre Appendix 2: LISA cartography of the spatial autocorrelation of population growth of French districts between 1968 and 1975 and between 1999 and 2009.
URL http://journals.openedition.org/cybergeo/docannexe/image/31641/img-13.png
Fichier image/png, 401k
URL http://journals.openedition.org/cybergeo/docannexe/image/31641/img-14.png
Fichier image/png, 402k
Haut de page

Pour citer cet article

Référence électronique

Sébastien Oliveau et Yoann Doignon, « Is the French diagonal emptying? An exploratory spatial analysis of demographic decrease in France for the past 50 years », Cybergeo : European Journal of Geography [En ligne], Espace, Société, Territoire, document 763, mis en ligne le 20 février 2019, consulté le 18 octobre 2019. URL : http://journals.openedition.org/cybergeo/31641 ; DOI : 10.4000/cybergeo.31641

Haut de page

Auteurs

Sébastien Oliveau

Univ. Nice Sophia Antipolis, Avignon Université, Aix Marseille Université, CNRS, ESPACE UMR 7300, 06204, Nice, France
Maître de conférences
Sebastien.oliveau@univ-amu.fr

Articles du même auteur

Yoann Doignon

Univ. Nice Sophia Antipolis, Avignon Université, Aix Marseille Université, CNRS, ESPACE UMR 7300, 06204, Nice, France
Aix Marseille Université, CNRS, LAMES UMR 7305, 13094, Aix-en-Provence
Doctorant
Yoann.doignon@univ-amu.fr

Haut de page

Droits d’auteur

Licence Creative Commons
La revue Cybergeo est mise à disposition selon les termes de la Licence Creative Commons Attribution - Pas d'Utilisation Commerciale - Pas de Modification 3.0 non transposé.

Haut de page