Urban growth effects on rural population, export and service employment: evidence from eastern France

Abstract

This paper examines how the spatial pattern of urban growth in functional economic regions influences the interplay of rural export employment, rural services employment, and population change in rural areas. Using an extension of the Boarnet’s model (Papers in Regional Science 73:135–153, 1994), we find that urban spread effects to rural areas in France are more likely than urban backwash effects, and that spatial urban (both dynamic and static) externalities affect rural population and employment growth. In the functional economic regions where the urban core is declining and the urban fringe is expanding, urban population growth involves an increase in rural export employment, and larger change in service employment favors rural population growth. However, urban export job growth reduces the growth in rural service jobs and expanding urban service jobs reduce rural export jobs, suggesting that expanding urban employment opportunities draws employees away from proximate rural communities. Conversely, where both urban core and fringe are growing, we observe an urban spread effect from the urban export sector to rural services—an export base multiplier effect with a spatial dimension—and from urban population growth to rural service employment.

Appendix

1.1 FER delineation and canton selection

For defining the FERs, we used a hierarchical classification based on commuting flows between communes (1990 Census data). The link between each couple of communes (A, B) is calculated as:

$$Link{\left( {A,B} \right)} = {Commuting\;flow{\left( {A \to B} \right)}} \mathord{\left/ {\vphantom {{Commuting\;flow{\left( {A \to B} \right)}} {Pop{\left( A \right)} + {Commuting\;flow{\left( {B \to A} \right)}} \mathord{\left/ {\vphantom {{Commuting\;flow{\left( {B \to A} \right)}} {Pop{\left( B \right)}}}} \right. \kern-\nulldelimiterspace} {Pop{\left( B \right)}}}}} \right. \kern-\nulldelimiterspace} {Pop{\left( A \right)} + {Commuting\;flow{\left( {B \to A} \right)}} \mathord{\left/ {\vphantom {{Commuting\;flow{\left( {B \to A} \right)}} {Pop{\left( B \right)}}}} \right. \kern-\nulldelimiterspace} {Pop{\left( B \right)}}}$$

It allows taking into account for double relationships between communes. After classification, the 4,084 communes making up the six selected regions are grouped into 414 FERs. Each FER has an urban core (the greatest urban unit of the FER), an urban fringe (all the communes that send more than 30% out of their active residents working in the urban core and fringe), a rural hinterland (all the other communes).

In contrast to previous studies (Schmitt and Henry, 2000), we cannot use the commune level in this paper because the employment division between export and services employment uses data from a one-fourth sample of the French census. The reliability of data is not sure under a threshold of 2,000 people and 4,007 communes among the 4,084 introduced in the analysis having less than 2,000 inhabitants. Thus, we used the canton level that is an intermediate administrative level. In our 414 FERs, there are 482 cantons where more than 2,000 inhabitants live and where at least one commune belongs to the rural hinterland. Among these, there are 225 cantons of which more than 50% of their population live in the rural hinterland. After eliminating cantons, which belong to FERs having a small urban core (less than 5,000 people), only 191 cantons can be introduced in the analysis.

These 191 cantons belong to 64 FERs, of which some main characteristics are given in Table 1 below. Our analysis takes into account a large part of the land area of the six selected regions (about 50%) and a large part of their rural hinterland population. Furthermore, we focus on the rural hinterland related to large and medium-sized cities: 37 FERs have a core size larger than 20,000 inhabitants, and 27 FERs have between 5,000 and 20,000 inhabitants. This corresponds to the largest and medium-sized FERs: their mean size is 140,000 inhabitants but with a large standard deviation (about 260,000). Our analysis tends to exclude many small FERs or/and FERs with small rural hinterland.

Table 1 Comparison between FERs included in the analysis and FERs excluded to the analysis.

		FERs with rural hinterland <5,000 inhabitants		FERs with rural hinterland ≥5,000 inhabitants		All the FERs
		Mean	SD	Mean	SD	Mean	SD
FERs with urban core <5,000 inhabitants
With cantons included in the analysis	Number of FERs	0		0		0
Without cantons included in the analysis	Number of FERs	206		18		224
	Population in FER	4,092	2,446	10,770	2,373	4,628	3,040
	FER land area (ha)	14,667	11,341	37,891	19,945	16,533	13,733
FERs with urban core between 5,000 and 20,000 inhabitants
With cantons included in the analysis	Number of FERs	5		22		27
	Population in FER	19,353	4,844	24,942	8,597	23,907	8,258
	FER land area (ha)	50,025	28,454	71,344	31,440	67,396	31,531
Without cantons included in the analysis	Number of FERs	63		27		90
	Population in FER	13,802	5,780	24,804	12,746	17,103	9,820
	FER land area (ha)	23,150	14,313	49,259	29,612	30,983	23,316
FERs with urban core ≥20,000 inhabitants
With cantons included in the analysis	Number of FERs	0		37		37
	Population in FER	–	–	221,037	314,693	221,037	314,693
	FER Land Area (ha)	–	–	190,313	129,522	190,313	129,522
Without cantons included in the analysis	Number of FERs	13		23		36
	Population in FER	46,161	17,049	118,503	72,689	92,380	68,284
	FER land area (ha)	36,662	20,352	86,461	46,640	68,478	45,802
Total FERs
With cantons included in the analysis	Number of FERs	5		59		64
	Population in FER	19,353	4,844	147,917	265,785	137,873	257,383
	FER land area (ha)	50,025	28,454	145,952	118,900	138,458	117,217
Without cantons included in the analysis	Number of FERs	282		68		350
	Population in FER	8,201	10,499	52,781	63,811	16,862	34,387
	FER land area (ha)	17,576	13,665	58,833	39,652	25,592	26,822

1.2 Estimation results

The first two columns in Tables 1, 2, and 3 reported below are parameter estimates for observations on rural cantons in FERs with a declining urban core and the next two are for rural cantons in FERs with a growing urban core. First, consider the proximity and amenity effects and impacts of earlier growth (from 1975 to 1982). Since the impact of urban spatial externalities on rural communities can be expected to decay with distance from the urban core, more remote rural villages (greater distance and/or no autoroute access), can be expected to grow slower than villages proximate to urban cores. Alternatively, villages with lower quality amenities are expected to grow more slowly than their more appealing counterparts.

We find few strong proximity or amenity effects on rural population. In each regression in Table 1 reported below, larger increases in the rural canton population over the 1975 to 1982 period (VDP7582) are associated with more rural people from 1982 to 1990. Since the natural balance (births less deaths, NATBAL_P) is a control variable, these population changes can be viewed as a result of in- or out-migrants.

In Table 2, the determinants of rural export employment growth indicate that again only a handful of amenities and initial conditions matter. Rural cantons that had more prior period growth (VDO7582) tended to have less export employment growth. However, larger beginning period export employment (DOTHE82) increased export employment (the parameter estimate in Table 1 reported below is multiplied by −1, as indicated in Eq. 4b). Higher initial period unemployment rates (RCHOM82) also tended to boost export employment growth. Increasing distance to the nearest urban area with more than 200,000 residents (DAGGLO) increases export employment in rural cantons.

In Table 3, we find that larger beginning period service sector employment (DSHOP82) decreases rural canton service employment (again the parameter is multiplied by −1 as shown in Eq. 4c). Moreover, in FERs with growing urban cores, prior period service employment growth (VDS7582) slows rural service employment growth. Higher density tourist accommodations (DCAPAC80) also slow subsequent serves sector employment.

Table 1 Estimation results for rural population change equation

	Cantons belonging to FERs with declining urban core		Cantons belonging to FERs with growing urban core
Variable	Parameter	t	Parameter	t
N	84		107
INTERCEP	−2.177	−0.88	−8.478	−1.75
DPOP82	−0.017	−0.55	0.076	1.43
DPBAS90C	−0.092	−0.76	0.158	1.00
D2PBA90C	0.001	0.55	−0.001	−0.47
NATBAL_P	−0.090	−1.33	0.068	0.54
DEQU80S	−2.738	−0.67	−0.525	−0.08
DIAUTORO	−0.001	−0.22	0.014	0.90
SCHO80D	0.055	0.83	−0.068	−0.59
HOSP80D	−0.035	−0.82	−0.077	−0.99
INCOM84	0.047	1.20	0.115	1.79
VDP7582	0.291	1.69	0.318	1.29
DWOTH82 (δ2)	0.027	0.15	−0.221	−0.61
DWOTHF1 (δ3)	0.002	0.48	−0.015	−1.38
DWOTHF2 (δ4)	0.001	0.24	0.004	1.15
WVDOTH (δ5)	−0.787	−1.02	3.584	2.13
WVDOF1 (δ6)	−0.035	−0.76	−0.003	−0.03
WVDOF2 (δ7)	0.006	0.22	−0.080	−1.31
DWSHO82 (δ8)	0.195	0.67	1.094	1.00
DWSHOG1 (δ9)	−0.007	−0.74	−0.045	−1.13
DWSHOG2 (δ10)	0.002	0.38	−0.020	−0.82
WVDSHO (δ11)	1.302	1.04	−2.082	−0.43
WVDSG1 (δ12)	0.046	0.43	0.101	0.47
WVDSG2 (δ13)	0.004	0.11	0.107	0.77
Adj. R 2	0.57		0.75
Sargan test (Qs)	χ2(14)=18.27		χ2(15)=4.88
Exogeneity test	F(10,50)=0.24		F(10,73)=0.64
I−Moran: W c	0.220	1.05	−0.274	1.17
W 30	−0.075	0.54	−0.096	0.52
W d	−0.028	0.75	−0.019	0.60

1. Entries in boldface indicate significance at the .10 level

Table 2 Estimation results for rural export sector employment change equation

	Cantons belonging to FERs with declining urban core		Cantons belonging to FERs with growing urban core
Variable	Parameter	t	Parameter	t
N	84		107
INTERCEP	−1.055	−1.12	−4.486	−1.90
DOTHE82	−0.267	−4.87	−0.066	−1.08
DPBAS90C	0.048	0.70	0.056	0.71
D2PBA90C	−0.001	−0.70	−0.001	−1.04
DAGGLO	0.012	2.03	0.011	0.93
DIAUTORO	0.001	0.13	0.006	0.60
NOBLU82	0.008	0.69	0.002	0.09
SKIWO82	−0.004	−0.99	0.003	0.86
SELFJ82	−0.003	−0.33	0.000	0.01
RCHOM82	−0.070	−1.17	0.253	2.16
VDO7582	−0.425	−1.86	−0.170	−0.55
DWPOP82 (θ2)	0.058	1.69	0.057	1.21
DWPOPH1 (θ3)	0.000	0.08	0.000	0.07
DWPOPH2 (θ4)	−0.001	−0.84	−0.003	−1.51
WVDPOP (θ5)	0.298	0.69	0.236	0.36
WVDPH1 (θ6)	0.010	0.17	−0.054	−1.11
WVDPH2 (θ7)	0.008	0.31	0.010	0.34
DWSHO82 (θ8)	−0.193	−1.10	−0.456	−0.75
DWSHOG1 (θ9)	0.005	0.78	0.006	0.17
DWSHOG2 (θ10)	−0.001	−0.41	0.016	1.54
WVDSHO (θ11)	−0.438	−0.61	0.332	0.08
WVDSG1 (θ12)	−0.082	−1.26	0.131	0.58
WVDSG2 (θ13)	0.043	1.87	−0.098	−1.76
Adj. R 2	0.48		0.25
Sargan test (Qs)	χ2(14)=10.02		χ2(15)=21.59
Exogeneity test	F(10,50)=0.99		F(10,73)=2.36
I−Moran: W c	−0.091	0.28	−0.134	0.93
W 30	−0.164	0.88	−0.129	1.05
W d	−0.042	0.99	−0.036	1.28

1. Entries in boldface indicate significance at the .10 level

Table 3 Estimation results for rural service sector change equation

	Cantons belonging to FERs with declining urban core		Cantons belonging to FERs with growing urban core
Variable	Parameter	t	Parameter	t
N	84		107
INTERCEP	1.206	1.31	−1.981	−1.99
DSHOP82	0.415	3.46	0.164	2.40
DPBAS90C	−0.030	−0.47	0.046	1.16
D2PBA90C	0.000	0.19	−0.001	−0.85
DAGGLO	0.003	0.50	0.001	0.23
DIAUTORO	0.001	0.18	0.007	1.54
DCAPAC80	−0.028	−1.90	−0.008	−1.55
NOBLU82	0.002	0.14	0.001	0.06
SKIWO82	−0.003	−1.03	0.000	0.12
SELFJ82	−0.007	−0.53	0.007	1.24
RCHOM82	−0.032	−0.56	0.021	0.47
VDS7582	−0.005	−0.02	−0.359	−2.12
DWPOP82 (τ2)	0.006	0.17	−0.007	−0.42
DWPOPH1 (τ3)	0.001	0.68	0.002	1.20
DWPOPH2 (τ4)	0.001	0.96	−0.001	−1.53
WVDPOP (τ5)	0.470	1.74	0.474	1.76
WVDPH1 (τ6)	0.006	0.11	−0.042	−2.00
WVDPH2 (τ7)	−0.050	−1.94	−0.003	−0.27
DWOTH82 (τ8)	−0.176	−0.96	0.086	1.69
DWOTHF1 (τ9)	−0.004	−1.14	0.002	0.76
DWOTHF2 (τ10)	0.002	1.02	0.001	0.71
WVDOTH (τ11)	−0.976	−2.08	0.210	0.61
WVDOF1 (τ12)	−0.049	−1.50	0.001	0.03
WVDOF2 (τ13)	0.031	1.69	−0.007	−0.52
Adj. R 2	0.61		0.51
Sargan test (Qs)	χ2(13)=17.22		χ2(14)=17.28
Exogeneity test	F(10,49)=0.86		F (10,72)=2.21
I−Moran: W c	−0.012	0.06	−0.128	0.82
W 30	0.002	0.01	−0.037	0.33
W d	−0.010	0.25	−0.032	1.10

1. Entries in boldface indicate significance at the .10 level