1 Introduction

1Worldwide rural-urban migration continues to show a trend in favor of cities, particularly large metropolitan areas [Moreno-Monroy, Schiavina and Veneri, 2021]. However, in Western countries, the residential attractiveness of large cities is stagnating, to the detriment of smaller cities [Milet, Maisetti and Simon, 2022 ; Peng, 2024]. The COVID-19 crisis has brought to the front of the debate the image of a better quality of life in medium-sized cities that were previously struggling in a globalized, competitive world [Bolay and Rabinovich, 2004] and therefore lacking in attractiveness [Kaufmann and Arnold, 2018]. As public policies continue to favor metropolitan areas for job creation [Bouba-Olga and Grossetti, 2015], they intensify a dissonance between types of territory and therefore a growing distance between workplace and residential location. The widespread use of teleworking enables certain categories of workers to project themselves far from their workplace [Peng, 2024]. This dynamic of distance is often studied in the light of urban sprawl issues [Nechyba and Walsh, 2004] and on the scale of large cities [Brueckner and Fansler, 1983]. This raises the question of how people who have left these large metropolitan areas to move to these smaller cities can revisit the major urban issues of links between cities and inequalities in access to employment on a national scale. These residential dynamics also call into question regional planning and the increasing distances between home and workplace.

2In this article, we aim to understand how residential migration from metropolitan areas to medium-sized cities affects spatial disparities and creates links between these forms of city. We will try to identify whether there is a different relationship with the workplace between those who move far from their previous place of residence and those who move fairly close to it. The underlying hypothesis is that there is a social and commuting divide between those who have made a “spatial split” with their previous place of residence and those who remain within the relative attraction of their original metropolitan area. The first would choose a place to live close to their aspirations in the case of chosen immigration, while the second would be constrained to live even further from the metropolitan area.

3To understand the migration dynamics of the working population and commuting practices on a national scale, we are using the French population census of 2023, i.e. surveys carried out between 2018 and 2022. We will focus our analyses on the most recent migrations from major metropolitan areas and the Paris region to medium-sized cities. The distinction between migrations of less than 50km and more than 50km enables us to distinguish between two patterns of migration more or less at break with one's previous living space. Analysis cross-referenced with socio-demographic characteristics provides a social picture of migration and associated mobility practices. The confirmation of these two main forms of migration is highlighted in statistical models [Kim, 2014].

4This article is structured as follows. Part 2 presents the state of the art, beginning with the relationships between workplaces and residential choices, then on job locations at a regional scale, and finally on the social impacts of job-residence disparities through spatial mismatch. Part 3 details the method, the scope of the study and the census data used. In the following section, descriptive statistics are presented to better understand the links between new and old places of residence for new residents from major French metropolitan areas. The final section outlines the social disparities in location choices and their link with disparities in commuting practices.

2 Literature review

5Finally, to the best of our knowledge, very few studies highlight the dynamics of spatial mismatch in residential migration on a regional or national scale. With the exception of one study, based on an urban economic model, which shows that residential migration can also be associated with longer commuting times without job loss [Bar-El, 2006]. This can be explained by the search for more affordable housing, even on an inter-regional scale. The literature related to our study can therefore be divided into two main areas of research: the link between migration and commuting, and the spatial mismatch.

2.1 Migration and commuting

6The link between relocation and workplace relationships has long been explored in the literature.

7On a regional scale, which shows that working people who make long-distance commuting prefer to remain in non-optimal situations, whereas they would have an economic interest in moving close to their workplace [Schmidt, 2014]. But other research shows that commuting costs in metropolitan areas can become an argument for leaving. For example, research shows that when workers are far from their workplace, they tend to relocate closer to it [Clark, Huang and Withers, 2003].

8Residential relocation or job change appears to be central to commuting trip patterns [Clark, Chatterjee and Melia, 2016]. It would thus facilitate modal changes, especially if habits are recent [Zarabi, Manaugh and Lord, 2019]. The dissonance between chosen transport mode and residential environment would be temporary, and over the long term individuals would adapt their trips to their urban context of residence [De Vos, Ettema and Witlox, 2018]. By changing their workplace and place of residence at the same time, workers will travel much longer for commuting [Prillwitz, Harms and Lanzendorf, 2007]. Using a biographical approach, these researchers demonstrate the influence of socio-demographic characteristics and family events on the relocation decision. Changing jobs increases commuting distance. Leaving from the center to the periphery increases commuting distances [Prillwitz, Harms and Lanzendorf, 2007].

9Few studies have focused on this dynamic between different types of territory, particularly those leaving large metropolitan areas. In the UK's rural areas, new residents from the major cities would be prepared to commute longer [Champion, Coombes and Brown, 2009]. On an inter-regional scale, in the Sao Paulo region, those who relocate farther away would make fewer commutes, and there would be a trade-off between relocation distance and commuting distance [De Castro Lameira and Golgher, 2022]. Another study in Spain confirms that by relocating beyond the city limits, working people make shorter commutes [Romaní, Suriñach and Artiís, 2003]. These studies, often restricted to metropolitan or regional scales, demonstrate the complexity of the interplay between residential relocation, commuting distance, transport mode and change of workplace.

10Several studies have examined the French situation and show a form of social segregation at play through residential migration to rural areas [Lépicier and Sencébé, 2007]. The workplace is not examined in these studies. Study shows that residential migration to rural areas in France continues to reinforce the marginalization of larger cities [Pistre, 2011]. However, dynamic rural areas attracting more affluent populations appear to be emerging. Another highlights areas far from the city of Lyon with a population that is becoming increasingly vulnerable, and the role of metropolitanisation in the territorial rejection of a certain population [Charmes, 2021]. Residential migration reinforces spatial segregation in large cities. It is primarily the exodus of the upper classes from urban centers that creates this segregation [Charlot, Hilal and Schmitt, 2009]. The study of the Atlantic coastline highlights different forms of migration and segregation between the peripheralization of the poorest and the coastal residential choices of the wealthiest [Vye, 2011].

2.2 Spatial mismatch

11The labor market would be significant in explaining residential migration, but largely shared with other factors [Kim, 2014]. Studies suggest that people also relocate far away for reasons other than joining a job [Coulter and Scott, 2015]. They would also be attracted by a better quality of life and a different environment. Others demonstrate that wage differentials and social ties are trade-offs when deciding to leave a residential location [Haas and Osland, 2014].

12When the commuting distance is excessive, households tend to be more inclined to relocate [Van Ham and Hooimeijer, 2009]. This phenomenon is inversely proportional to age [Brueckner and Št’astná, 2020]. If, on the contrary, these working people adapt to the long-distance constraint, they may choose to have a second home close to the workplace [Petzold, 2020]. But those who live on the periphery and have the longest commutes would be less sensitive to the hardship of transport than those who live in the center. They would therefore decide to live in the center to avoid long trips [De Vos and Witlox, 2016]. Men are more likely to commute than women, and women are closer to work than men when the couple relocates [Plaut, 2006]. Research shows that residential relocation is associated with the birth of a child or a change of job, and is accompanied by greater car use [Tao et al., 2023].

13Other analysis keys are also central to understanding the links between workplace and residential choices or constraints, notably social position. For example, Musterd et al. show that people leave their residential location more often because they are too wealthy compared to the average resident [Musterd et al., 2016].

14The link between social constraint and living and working locations is most often studied in terms of spatial mismatch, which is the spatial mismatch between place of work and place of residence for the poorest households, originally studied in the USA [Kain, 1968]. In this case, the poor are predominantly located in the city center, while the rich and employed are concentrated on the peripheries [Gobillon, Selod and Zenou, 2007]. Research has studied residential dynamics according to social category. On the contrary, these dynamics point to gentrification in the urban centers of large and medium-sized cities (Bereitschaft, 2020). In Los Angeles, spatial mismatch is seen to be intensified for poor households in the inner suburbs due to a lack of housing [Hu and Giuliano, 2011]. Blumenberg and King show that low-income workers would have longer commutes which could be explained by the fact that they relocate to low-density areas [Blumenberg and King, 2019]. "High-skilled" populations would be the only ones to succeed in reducing their commuting time due to their high wages and education [Xiao, Wei and Chen, 2023]. "Low-skilled" populations would be constrained by their low skills. Still in the United States, commuting distances increase for rich and poor populations when they relocate to peripheral urban areas [Blumenberg and King, 2019]. Wealthier people would be able to change their workplace more easily and thus adapt to a change of workplace and residential change, even if household structure limits this possibility [Kronenberg and Carree, 2012].

3 Method, scope and data

3.1 Scope

15Our study area is the set of medium-sized cities in France, as defined by Santamaria [Santamaria, 2000]. INSEE (national statistics institute) has defined functional perimeters for each city, which it calls “attraction areas”. These are connected territories where, for each municipality, at least 15% of residents work in the city center. The cities selected are those whose attraction area population is less than 200,000 and greater than 20,000 [Milet, Maisetti and Simon, 2022 ; Santamaria, 2000]. For each city, the studied territory will be the central municipality and its attraction area. For the remainder of this document, we refer to the entire area as a “mid-sized city”. Figure 1 shows these cities, distinguishing between the central municipality and its attraction area. There are 416 of these cities, representing 30% of the French population. We study residential migration in relation to the workplace. Thus, the population studied will be the territory's working population (40% of the territory's population). The region contains 7.6M jobs for 7.7M working people. Of these, only 83% live and work in the area.

16We focus our analysis on residential migration from France's major metropolitan areas. Our definition of a metropolitan area refers to the top of the pyramid in the hierarchy of French cities, but also to the economic symbol they represent [Gaschet and Lacour, 2007]. Public authorities and some researchers still believe that the trickle-down effect of wealth and jobs from metropolitan area should be supported [Guieysse and Rebour, 2022]. These include the Paris metropolitan area and cities whose metropolitan area has a population of over 700,000 [Cusset and George, 2024]. There are 10 of these (Figure 1). This corresponds to 11M inhabitants in the Paris region (23% of metropolitan jobs) and 12M in the major metropolitan areas, representing a total of 35% of the French working population.

3.2 Data, the 2023 Census in France

17By construction, the 2023 population census corresponds to respondents from 2018 to 2022. Each year, one tenth of the French population is surveyed at home. By aggregating the 5 years, the 2020 census therefore corresponds to a quarter of the French population. In this study, new residents are respondents who have been in their long-term residence for less than 2 years prior to their survey [Poulhès and Brachet, 2024]. In the Census, they number 1,500 from the Paris region and 5,570 from major metropolitan areas to medium-sized cities in France.

3.3 Method

18Our research focuses on new working residents of all French medium-sized cities whose previous residence was in a large French metropolitan area. The population census provides a sufficiently representative and recent sample, but does not allow us to precisely analyze each city independently. For this reason, all our analyses will be aggregated for all medium-sized cities. We compare the behavior of these new residents with other new residents of medium-sized cities. Several territorial divisions are studied, including the commune of residence and the attraction area associated with each city, in order to understand potential specificities linked to urban forms. The distinction between the Paris region on the one hand, and other major metropolitan areas on the other, is motivated by the specificity of the Paris metropolitan area in terms of size, the characteristics of its working population and the mobility patterns of its residents. To be able to characterize the differences in population and practices between one form of sprawling relocation and another involving a spatial break, we use an identical threshold of 50 km between previous and new residential areas. This threshold was constructed by evenly distributing relocations in relation to this value. The distance of 50 km is arbitrary but corresponds to the radius of the Paris metropolitan area's catchment area. It is also the threshold chosen by an INSEE study to show a break in the number of commutes beyond this threshold in Paris, Lyon, and Marseille [Gascard and Van Lu, 2019]. Distances between municipalities are calculated between municipal centroids as the shortest route on the road network.

19The first results present descriptive analyses of the results according to the main variables studied. The first are spatial variables, with comparisons of residential migration aggregated by individual characteristics, but distinguished by migration distance threshold, commuting distance and mode used.

20Thus, we propose comparisons according to several major characteristics of the working population, based on their distribution by transport mode, origin (metropolitan area or Paris region) and the 50 km threshold for residential migration. We also attempt to socially characterize these populations by comparing migration and commuting by socio-professional category [De Castro Lameira and Golgher, 2022].

21Finally, we propose several statistical models to better understand the determinants of location and commuting decisions. Locating within or beyond 50km is explained by a binomial logit model that integrates the variables presented above, adding the modal choice present in the census. From these same variables, we attempt to explain the logarithm of the commuting distance with a multilinear model. Finally, another binomial logit model aims to characterize the profiles of driver commuters. Interaction effects are added between the variables: socio-professional categories and distance from previous place of residence.

4 Results

4.1 New residents from metropolitan areas who keep their jobs

22In France, between March 2020 and March 2021, the cities in our study area gained 41,000 inhabitants in net residential migration, while the largest cities lost 60,000 inhabitants [Milet, Maisetti and Simon, 2022]. Based on data on letter redirection requests, this study also points out that 8.4% of migration is from large cities to medium-sized cities (including intermediate-sized cities, which are not included in this study), and only 4.5% in the reverse direction.

23These new residents would be 4400 from the Paris region and 17,000 from the metropolitan areas. This would represent around a quarter of migration to medium-sized cities. According to the 2023 census, Figure 2 presents the link to residents' workplaces, with each scheme corresponding to the origin of residential migration. This is divided according to origin from the Paris region or other metropolitan areas and whether migration is over 50 km or not.

24Leaving from their previous residential location is more likely to result in having (or presumably keeping) a workplace in the metropolitan area, representing 60% of this category of new residents. This would mean a residential break while maintaining a strong link with the metropolitan area of origin [Bar-El, 2006].

25Differentiation by distance from their previous residence clearly shows differences in behavior. Relocating far from one's previous place of residence is more frequently associated with relocating to the city center. (67% from the Paris region and other metropolitan areas). This proportion can be compared with the 33% of "existing" residents who live in the city centers of medium-sized cities. The periphery are chosen by the majority of working people who remain close to the Paris region (50%), which is not the case for those from other major metropolitan areas (35%). And yet, 61% of long-term residents live in the inner suburbs. This highlights the attractiveness of the city center for new working residents from the major metropolitan areas.

4.2 A clear difference in modal choice depending on the distance to the metropolitan area of origin

26To clarify these results, Figure 3 disaggregates new residents in relation to their workplace and the associated modal share. A comparison between long-term and new residents from the same medium-sized cities shows a high degree of stability in behavior [Poulhès and Brachet, 2024]. Only the share of public transport increases for those working in a metropolitan area.

27When relocating below the 50km threshold, new residents are more likely to use the car than those who relocate farther away, especially those from the Paris region (53% only) and those from other metropolitan areas (69%). When moving closer to home, 85% use the car, whatever their origin. New residents from large metropolitan areas who relocate only slightly have, on average, the same modal relationships to their workplace as new and long-term residents of medium-sized cities.

28If we cross-tabulate urban types of employment and residence, it appears that the average car modal share for workers from metropolitan areas and the Paris region is always at least 10% lower (and even 20% lower for those from the Paris region) when they relocate beyond 50 km of their previous residence. And this is the case regardless of changes in urban contexts. The spatial break with their previous residential location is not associated with a break with their workplace for 30% of these residents.The average commuting distance for those moving within 50km is 23km. This value remains stable whatever the type of territory and the size of the metropolitan area of origin. This average distance is close to the average values for long-term residents (21 km). However, once this threshold is exceeded, the average distance to the workplace is much higher, 115km from the Paris region and 45km from other metropolitan areas. The modal share of the car follows the same logic of strong distinction with the 50km limit. Those who relocate closest have modal shares close to those of long-term residents. The modal share drops significantly with distance, with only a 50% modal share for those coming from the Paris region (close to the Paris region average). The further they move from the center, the greater the car modal share (+15% modal share points for all types of moves). The literature confirms that urban forms of residence influence our mobility, and in particular reinforce car use [Cervero and Kockelman, 1997]. Living in a suburban area increases car use. Those who move far away logically increase their average commuting distance, carried by those who keep a job in their previous metropolitan area, as we'll see later. And these average distances vary quite little with the urban context.

4.3 “Low-skills” and “high-skills” who relocate more or less far away?

29Table 1 compares the distributions of relocations by executive called “high-skill” workers and clerical and Manual worker (“low-skill” workers). The types of relocation are compared according to the mode used. The average distance to the workplace is also specified.

30The first result of this table expresses the population differences between those leaving metropolitan areas and those already living there or coming from other territories. Indeed, 31% of immigrants from the Paris region are “high-skill” workers and 21% from other metropolitan areas, whereas only 12% of long-term residents are “high-skill” workers, and only 10% from other areas. By contrast, over 50% of long-term residents are “low-skill” workers and 40% come from all metropolitan areas. The most original point is the distribution of this new population in relation to distance from the previous residence. Those moving over 50 km away are equally divided between “high-skill” and “low-skill” workers. On the other hand, more than four times as many “low-skill” workers relocate less than 50km away, which is in line with the average distribution of medium-sized cities.

31In terms of relationships with the workplace, “high-skill” and “low-skill” workers from the Paris region use other transport modes almost as much as the car when they relocate far from their previous place of residence. This is slightly less the case when they come from other major metropolitan areas. Logically, this is supported by a greater distance to the workplace, driven by those who keep their jobs in the metropolitan areas, as we saw earlier. Whatever their origin and the distance of their relocation, “high-skill” workers commute greater distances than "low-skilled" workers.

32Both in terms of population structure and behavior, the results for long-distance migrants show an overall contrast with the population of residents of medium-sized cities. This is not the case for those who relocate close by.

Table 1: Distribution of new residents from metropolitan areas by origin and socio-professional category, INSEE 2023

Population distribution in % by place of origin
	Relocation distance	High-skill workers non-drivers	High-skill worker drivers	Low-skill workers non-drivers	Low-skill worker drivers
Paris region	>50 km	12%	15%	10%	10%
	<50 km	1%	3%	3%	15%
Metropolitan areas	>50 km	5%	10%	5%	11%
	<50 km	1%	5%	3%	20%
All new residents		2%	8%	9%	47%
Old residents		2%	10%	10%	42%
Average commuting distance (km)
	Relocation distance	High-skill workers non-drivers	High-skill worker drivers	Low-skill workers non-drivers	Low-skill worker drivers
Paris region	>50 km	150	88	106	67
	<50 km	54	29	33	28
Metropolitan areas	>50 km	55	38	43	49
	<50 km	34	34	30	25
All new residents		47	32	18	19
Old residents		46	33	15	20

Note: There are 9% of High-skill workers who do not commute by car and who relocate farther than 50 km from their previous residence. 100% corresponds to all new residents coming from the Paris region, some categories are not represented in the table.

4.4 Two major forms of migration and a confirmed connection to the metropolitan area

33Table 2 presents the results of logistic regressions to explain whether or not one uses the car to commute, and to explain whether one has made a residential migration more than 50 km from their previous home. Linear regression explains distance to the workplace. The results confirm the distinction between workplace connection and migration distance. The even larger coefficient in this model for the “Relocation over 50 km” variable confirms the lower use of the car for metropolitan residents who move far from their previous area of residence. The greater distance from the place of employment for those who have kept a job in their previous metropolitan area also increases the distance for those who move further away. Relocating further is also associated with a much greater use of alternatives to the car - active modes for short distances, and transit for long distances.

34Among the most interesting results, relocating closer to a previous home is correlated with greater use of the car and shorter distances to the workplace. Other research has reached similar conclusions for distances to the workplace [De Castro Lameira and Golgher, 2022]. Previous research, on the other hand, found a break in this pattern when workers relocated from one region to another, closer to their workplace [Clark, Huang and Withers, 2003]. In our study, the workplace in the metropolitan area of origin for many working people does not allow this observation to be confirmed, even if a large proportion of these immigrants also work in their new urban area (Figure 2). This difference can also be explained by changes in working practices over the last 20 years, and the increased use of teleworking. If we distinguish the Paris region from other French metropolitan areas, we can see that the behavior of new residents differs. Those from Paris use the car less and live further away from their workplace. Yet they relocate closer to their previous homes. This greater distance may be partly due to the larger size of the Paris region, which increases distances for those who continue to have their workplace there.

35Part-time work would enable workers to relocate further away from their previous home. But it is above all short-term contracts and being self-employed that correlate with relocating further away than those on long-term contracts [Bergantino and Madio, 2019]. Job instability can make residential choice more difficult. This is associated with lower car use. While men are further from the workplace [Bergantino and Madio, 2019 ; Clark, Huang and Withers, 2003 ; Surprenant-Legault, Patterson and El-Geneidy, 2013], gender is not a significant variable in residential and transport mode choices. Two-worker households are the ones that commute the longest distances, which can be explained by the difficulty of finding a job close to home for both workers. The literature shows that, on the contrary, two-earner households travel the least distance to work [Surprenant-Legault, Patterson and El-Geneidy, 2013]. In this case, at least one of the two active members of the household would commute to the workplace at a later stage. They are also the ones most associated with driving to commute. Households with children move closer to their previous home than other household types [Clark, Huang and Withers, 2003]. Members of the same dual-income household are likely to commute longer distances [Plaut, 2006]. Two-worker households, which are the most constrained, are ultimately the ones that follow the classic pattern of peri-urbanization by relocating close to their previous place of residence. This is coupled with a high dependence on the car (Figure 2) and longer distances travelled than other household types [Mitra and Saphores, 2019 ; Plaut, 2006].

36In our results, “Highskills” workers (as “Executive”) relocate further from their previous place of residence than other socio-professional categories, in line with other research on regional migration [Bergantino and Madio, 2019]. We can assume that this long-distance relocation is associated with a personal choice rather than a financial constraint. This suggests that, on the contrary, low-skilled workers are more often forced to leave metropolises where property prices are too high, and relocate to neighboring medium-sized cities. This spatial mismatch dynamic is not reflected in distances to the workplace. The socio-professional categories are not significant for distance [Mitra and Saphores, 2019]. Romani et al. find that "highskills" workers make longer commutes [Romaní, Suriñach and Artiís, 2003] and Plaut et al. associate higher wages with longer commutes for US couples [Plaut, 2006]. Similarly, car use seems to be more highly correlated with workers or intermediary businesses than employees. However, high-skills workers are more associated with car-using only when they move further away from their previous place of residence.

Table 2: Logit regressions and multi-linear regression on new residents to medium-sized cities from metropolitan areas and the Paris region

	Is driver?		Distance		Is > 50km
Size of city Medium-sized city (reference) Small city	0.08*	0.02 – 0.15	0.18***	0.12 – 0.24	-0.56***	-0.62 – -0.50
Residential origin
Metropole area (reference) Paris region	-0.79***	-0.86 – -0.72	0.32***	0.24 – 0.39	0.21***	0.13 – 0.29
Work time Full-time work (reference) Partial-time work	-0.22*	-0.31 – -0.14	-	.	0.16**	0.07 – 0.25
Socio-professional category
Executive	-0.26*	-0.35 – -0.17	0.17*	0.02 – 0.33	1.17***	1.07 – 1.26
Intermediary businesses	0.24***	0.15 – 0.32	-		0.52***	0.44 – 0.60
Employee (reference)
Manual Worker	0.63***	0.51 – 0.74	-		-0.31***	-0.41 – -0.21
Household structure One worker, w/o kids	-0.37***	-0.44 – -0.30	-0.47***	-0.54 – -0.40	0.59***	0.52 – 0.66
One-worker family with kids	-0.27***	-0.37 – -0.18	-0.21**	-0.35 – -0.08	-0.28***	-0.41 – -0.15
One-worker and other	-0.45***	-0.64 – -0.25	-0.35**	-0.55 – -0.15	0.62***	0.53 – 0.72
Two-worker household (reference)
Employment contract Permanent contract (reference) Learning(11)	-0.52***	-0.66 – -0.38	-0.17*	-0.32 – -0.02	0.53***	0.37 – 0.69
Fixed-term contract(15)	-0.29***	-0.38 – -0.20	-0.29***	-0.38 – -0.20	0.49***	0.40 – 0.58
Self-employed(21)	-0.38**	-0.52 – -0.24	-0.48***	-0.62 – -0.33	0.84***	0.68 – 1.00
Sex Female (reference) Male	-	-	0.16***	0.10 – 0.23	-
Relocation Under 50km (reference) Over 50km	-1.15***	-1.29 – -1.02	0.17***	0.05 – 0.30
Mode choice Car (Reference) No trip			-1.13***	-1.29 – -0.96	0.93***	0.76 – 1.11
Walk			-1.24***	-1.35 – -1.13	1.15***	1.03 – 1.27
Bike			-1.07***	-1.26 – -0.88	1.15***	0.94 – 1.37
Two wheelers			-0.46**	-0.75 – -0.18	1.17***	0.86 – 1.49
Transit			1.00***	0.90 – 1.09	0.96***	0.86 – 1.06
Employee* Under 50km (reference) Executive* Over 50km	0.41***	0.21 – 0.61	-0.22*	-0.41 – -0.03
N Pseudo R², R²	7,076 0.09		7,076 0.23		7,076 0.16

Notes: Each value in the table represents an estimate of the parameter along with its 95% confidence interval, * p<0.05, ** p<0.01, *** p<0.001 indicate the level of significance.

Discussion and conclusion

37The objective of this research was to answer the following question: how does residential migration from French cities raise questions about a new form of spatial mismatch? We used statistical methods based on census data to explore this research question. The hypothesis of differences in populations and commuting behaviors between those relocating closer to their metropolitan area of origin and those relocating further away seems to be confirmed. Urban relocation dynamics show a predominance of "low-skilled" workers who locate in medium-sized cities close to metropolitan areas, with a strong propensity to always have their workplace in the metropolitan area. This is associated with greater use of the commuter car, and ultimately with adaptation to the average practices of medium-sized cities. On the other hand, “high-skilled” workers are more likely to relocate far from their home metropolitan area and keep their jobs, but with less car use. In the end, this population was the one that was highlighted during the COVID-19 crisis [Knuepling, Sternberg and Otto, 2024], invisibilizing the other categories with a trajectory that follows the classic urban sprawl relocation of metropolitan areas. It also makes invisible new residents from the metropolitan areas, who in the end use their cars sparingly and work close to their new place of residence. These differences in population and practices confirm the gentrification underway in certain medium-sized cities [Bereitschaft, 2020]. But above all, considering the workplace confirms a certain type of spatial mismatch on a national scale, more than only a dynamic of spatial segregation [Lépicier and Sencébé, 2007 ; Pistre, 2011]. Even if the latter needs to be put into perspective, given the low number of migrations (~10,000/year).

38Migration to medium-sized cities, now often less dynamic than larger cities [Kaufmann and Arnold, 2018], is opening up new opportunities for non-metropolitan areas [Eurofound and European Commission Joint Research Centre, 2024]. However, they also open up a number of discussions that need to be addressed.

39A first limitation is to focus medium-sized cities in an aggregated way. However, our study already highlights the need to avoid a homogenous policy for medium-sized cities. Higher social categories who relocate far from their workplace, as opposed to lower social categories who move away from the center of the metropolitan area for economic reasons, do not define identical territories. Some medium-sized cities could become gentrified [Song and Chapple, 2024], while others close to large metropolitan areas could become impoverished as well as shrinking [Martinez-Fernandez et al., 2012]. Our aggregated approach doesn't enable us to identify the specific features of each medium-sized city, and only provides average results. A city-by-city analysis would not reveal any patterns, but other approaches based on specific city characteristics can provide a different perspective on the decisions made by new residents.

40In the article we discuss commuting distances, without considering the fact that working people sometimes do not have a fixed workplace. The census also limits us to the mobility associated with these trips. We don't know whether these journeys are really made. The literature suggests that long-distance trips are not travelled every working day, and that telecommuting replaces commuting [Helminen and Ristimäki, 2007]. Distance enhances the desire to change jobs [Clark, Huang and Withers, 2003]. The development of teleworking increases inequalities between different types of workers. High-skilled workers, for whom teleworking is much more accessible, can more easily move away from their workplace, reinforcing spatial mismatch [Zhu, 2013]. For some workers, who are totally free from their workplace - the digital nomads - this independence enables them to have a desired workplace that is constantly changing [Bozzi, 2024]. Others choose bi-residentiality to limit the number of long-distance trips [Petzold, 2020].

41To highlight the social disparities associated with moving, we used the socio-professional categories variable, which is specific to the working population in our study. We've associated executives with “high-skill” and employees and Manual worker with “low-skill”, which does indeed reflect differences in terms of level of education and social category, but not necessarily as much in terms of salary. The other social categories are much more heterogeneous, so are not analyzed through the prism of our categorization. Some research, notably on the spatial mismatch between place of work and place of residence, also studies low-skills and high-skills to highlight inequality phenomena [Gobillon, Selod and Zenou, 2007]. But much other research focuses on wage differences between individuals as a proxy for social inequalities [Bergantino and Madio, 2019 ; Blumenberg and King, 2019 ; Song and Chapple, 2024].

42The migration of residents to medium-sized cities while their jobs remain in their metropolitan area of origin raises issues of urban planning to meet household aspirations. Medium-sized cities close to metropolitan areas would become new peripheral areas, dependent on large metropolises and are related with relocation of “low-skill” population. And other medium-sized cities are then more related with attracting “high-skill” population that are also in a large part dependent on the metropolitan areas. Is this dependence and double imbalance sustainable [Bolay and Rabinovich, 2004]? The workplace is as important as the residential location in the well-being associated with commuting [Maheshwari et al., 2023]. Maly suggests developing metropolitan tools to plan urban extension and avoid competition between territories for new residents [Malý, 2024]. Our scale of analysis suggests planning on an even larger, national scale.

43The arrival of new residents whose commuting mobility is less car-centric than previous residents should support the implementation of local alternative mobility policies to the car, such as public transport and active modes, which are very scarce in these types of areas. Medium-sized cities in the U.S. that have an alternative transportation network to the car fare better economically and have less inequality [Frederick and Gilderbloom, 2018]. Their residents also have better quality of life where transit commuting is possible [Talmage and Frederick, 2019].

44A study of medium-sized cities according to their distance from metropolitan areas but also more qualitative on the choices of "high-skills" workers will help confirm the results obtained on the analysis of residential migration. Not only the resident population and its dynamics, but also the transport offer of the medium-sized cities studied [Frederick and Gilderbloom, 2018] would highlight what could be decisive in the attraction of populations and thus reduce the effects of spatial mismatch. Other geographical specificities may play a role in the attractiveness of territories, such as proximity to the sea coast, which would be more welcoming to metropolitan residents [Milet, Maisetti and Simon, 2022]. Migration analyses between all city sizes, and not just on the specific migrations studied, would help us to better understand the interactions in the city hierarchy and the new position of medium-sized cities in relation to all other types of territory.