Regional income inequality in France 1860–1954: Methods and findings

Abstract

This paper explores regional (département or NUTS3) income inequality in France between 1860 and 1954. To this end we first document the existing evidence, evaluate the estimation methods and findings, assess the suitability of each approach and address potential concerns. We then present our own subnational estimates of per-capita income, derived following Geary and Stark. Overall, we find that indirect estimation (or top-down) methods provide somewhat differing results, thereby calling into question the validity of some of these approaches. Our estimates, meanwhile, appear to fit reasonably well with direct estimations (or bottom-up) of value-added and income and provide evidence in support of a decline in regional inequality over the period of study.

Keywords:

• Economic history
• methods
• regional inequality
• France

JEL codes:

• N93
• N94
• 018
• R11

Acknowledgment

We are grateful to the editor and referees for their comments and advice. Funding was provided by Ministerio de Economía y Competitividad (ECO2015-65049-C12-1-P; EC2015-71534-REDT)

Notes

1 Recent empirical studies strongly support the hypothesis of an inverted U-shaped relationship between spatial inequality and economic development (Barrios and Strobl 2009; Lessmann 2014).

2 Rosés and Wolf (2018) have compiled a set of studies for 16 European nations based on estimates of regional GDPs which comprises 173 regions at NUTS2 level corresponding to 11 benchmark years between 1900 to 2010. In the United States regional specialization rose substantially between 1860 and the turn of the 20^th century, flattened out during the interwar years, and then fell continuously since the 1930s (Kim, 1995 and 1998).

3 The physiocrats employed a scientific method based on data and analysis in their work. More recently, Toutain (1987) constructed séries historiques of the domestic product and several of its components for 1815 to 1938, while Lévy-Leboyer and Bourguignon (1990) estimated annual time series for the period 1820-1913. See Chadeau (1989) for an overview of long-term time series of macroeconomic aggregates.

4 Toutain (1981) reconstituted the physical product per inhabitant (and per active person) for 21 regions in 1784-85, 1840, 1865, 1896, 1931 and 1970. Also, Toutain (1961, 1992) derived agricultural production for each department in 1840, 1892 and 1929. On the other hand, using agricultural product per inhabitant and per active person between 1851 and 1970, Lévy-Leboyer (1982) describes regional dynamics.

5 In doing so, a myriad of sources was used. For 1864, agricultural income was derived from labour incomes and capital revenues published in the agricultural survey of 1862. In manufacturing, labour income was estimated using average wages from the industrial survey 1861-65 and the labour force using the population census of 1866. Average profits for industrialists and merchants and incomes for liberal professions were reconstituted using fiscal information and the population census of 1866. Administrative sources made it possible to estimate the labour incomes of public servants, army and navy personnel, domestic workers and transport workers, while the cadastre was used to reconstruct real estate revenues. For 1954, DM1959 initially uses the same sources that INSEE. However, some adjustments are made to enable the transition from gross product to net income. The Annuaire Statistique de la France 1954 and the Statistique Agricole Annuelle 1954 jointly with fiscal sources, specifically the taxe proportionelle, are used to reconstitute capital incomes in agriculture and manufacturing. For labour incomes several sources are employed.

6 Although Brigitte Desaigues’s work was cited in Combes et al. (2011) as forthcoming, it has not yet been published. The Industrial Surveys of 1860 and 1861/65 tended to exclude small-scale industries and craftsmen. In order to overcome this, Desaigues imputed the value-added of a representative sample of small industries (fewer than five workers).

7 Fiscal sources, state and local community account records, cadastre information and data provided by the Statistique Générale de la France (1930) are some of the sources used by Toutain to reconstruct the revenues of five service sectors (housing, public services, retailing and wholesaling, professions and household services). In Combes et al. (2011), see the appendix for a complete description of the sources.

8 In France, agricultural surveys were conducted in 1840, 1862, 1892, 1911 and 1929. Their rich information let to reconstitute intermediate consumption and enabled the estimation of agricultural revenues without falling into double accounting. Toutain (1992) uses these surveys to impute agricultural output by department for 1840, 1862, 1892, 1911 and 1929. Only the population censuses of 1861 and 1931 include an industrial survey with detailed information by department: Statistique de la France. Industrie: Résultats généraux de l’enquête effectuée dans les années 1861-1865 and Enquêtes annexes du recensement de 1931: Enquête industrielle. The population census of 1896 contains a census of professions with figures only for employment and wages by department.

9 In order to avoid time inconsistency, Caruana-Galizia (2013), following Marchand and Thélot (1991), made some adjustments to the employment data (Caruana-Galizia 2013, 75-76).

10 Although Caruana-Galizia (2013, 72-73) acknowledges the potential role of labour productivity, it is argued that “a more efficient allocation of labour among economic activities” is what “really counts”.

11 According to Toutain’s (1987) estimates of historical GDP.

12 Value-added from Toutain (1987) and total active population in manufacturing from Marchand and Thélot (1991) for 1860 and the population census for 1930.

13 While the original or 1860 parameters would be $\left({\widehat{\mathrm{\beta }}}_1,{\widehat{\mathrm{\beta }}}_2,{\widehat{\mathrm{\beta }}}_3\right)=(1.037;\mathrm{ }0.326;0.149)$, the parameters for 1930 would be $\left({\widehat{\mathrm{\beta }}}_1,{\widehat{\mathrm{\beta }}}_2,{\widehat{\mathrm{\beta }}}_3\right)=(1.050;0.247;0.325)$. Caruana-Galizia (2013, Table 1) reported slightly different values for 1860: $\left({\widehat{\mathrm{\beta }}}_1,{\widehat{\mathrm{\beta }}}_2,{\widehat{\mathrm{\beta }}}_3\right)=(1.047;\mathrm{ }0.325;0.151)$.

14 The patente was a “general tax levied upon all individuals, whether French or not, carrying out non-agricultural activities on their own account in France” (Bazot 2014, 309). Delafortrie and Morice (1959) also used the patente for 1864. Lavergne and Henry (1908) criticized the use of the patente to estimate profits because of its poor coverage. The complementary use of other administrative surveys for 1844 and 1880 enabled them to conclude that the “main” part of the patente represents barely 3% of overall net profits. Using the number of small-business owners in the population census of 1866, Delefortrie and Morice (1959) estimate total net profits at national level. The contribution of each department to the patente was then used to calculate a figure for profits for each department. See Bazot (2014, 311-313) for more information.

15 Although Toutain (1992) only publishes agricultural value-added by department for 1840, 1860, 1892 and 1911, Bazot (2014) approximates intermediate years (1850, 1869, 1880 and 1900) keeping the share of agricultural value-added constant in 1840, 1860, 1892 and 1911 respectively.

16 See Geary and Stark (2002), Appendix: Method for a detailed explanation.

17 One of the criticisms of the Geary-Stark method is that it relies exclusively on labour incomes and does not include capital revenues (Crafts 2005). However, it uses wages only as an indicator of regional sector productivity (Geary and Stark 2015, 126) and not to estimate a multiple of total wage income as suggested by Crafts (2005). Hence the distribution of GDP across regions is based on sector labour productivity which depends on both the stock of capital and the state of technology in each department. By comparing their figures against official estimates for 1971 and 2001, Geary and Stark (2015) test that their method provides accurate estimates of regional GDP.

18 This method loses accuracy when moving back in time because wage coverage is reduced. Fortunately, French official statistics provide a wide set of wages for 1860, 1896, 1911 and 1929.

19 Although Combes et al. (2011) provide labour force figures for 1896, we use this information for 1892. Population data come from the harmonized national censuses published by the INSEE.

20 See sources for a complete description.

21 La France métropolitaine contains 96 départements or NUTS3 regions. After the Congress of Vienna in 1814/15, France had 86 departments. By 1861 there were 89 with the inclusion of Alpes-Maritimes, Savoie and Haute-Savoie. As a result of the Franco-Prussian War (1870-71), Bas-Rhin and parts of Meurthe, Moselle, Vosges and Haut-Rhin were ceded to the German Empire. These territories were recovered after the Great War (1914-18) and the Territoire de Belfort became the 90^th department in 1922. In 1968 Corse was divided into Haute-Corse and Corse-du-Sud, while Seine and Seine-et-Oise were reorganized into 7 departments. For a complete list of the departments included see Table A.1 and Map A. 1 in the Appendix.

22 For the sake of comparability, we also exclude Corse, as Combes et al. (2011), Caruana-Galizia (2013) and Bazot (2014). Notwithstanding this, it is worth mentioning that Corse was the bottom part of the regional income distribution. In 1911, for instance, we have estimated that Corse was the second-poorest department, just above Lòzere.

23 Gluschenko (2018) has recently argued that using indices weighted by population is inadequate since they are just a rough measure of interpersonal inequality.

24 Table A.2 in the appendix illustrates regional (NUTS2) income inequality, measured with a coefficient of variation, for a selection of European countries.

25 Duro and Esteban (1998) for several European countries, Caselli and Coleman (2001) for US states and Enflo and Rosés (2015) for Swedish regions in 1860-2010.

26 In terms of per-capita income, Meuse, Meurthe-et-Moselle, Pas-de-Calais, Vosges and La Manche would be the richest departments. Seine (including Seine-et-Oise) would not be in the top 5 and would fall below the national average (FRA =1), which seems somewhat puzzling.

27 Likewise, Bazot (2014, 323) points out that the “distribution of economic development among départements tends to homogenise –especially during the period of high economic growth (1900-1910)- even though the most developed départements deviate from the base –especially during the period of low economic growth (1860-1900)”

28 The SCV for Great Britain in 1861 was 0.192 and for the UK 0.277 (Geary and Stark 2015, Table 10).

29 The articulation of a transport network (railways…) radiating from Paris could have reinforced the “headquarter effect” during our period of study. In fact, this rationale was an integral part of the “Paris et le désert français” debate (Gravier, 1958). With a growing and well-integrated market, Seine benefited from agglomeration economies (Combes et al., 2011).

30 As Toutain (1992) points out in annexe B, agricultural value-added for 1860/64 is taken from Delefortrie and Morice (1959), who reconstituted these estimates using the Agricultural Survey of 1862.

31 Salaires et durée du travail dans l'industrie française. République Française. Ministère du Commerce (1897). Table XXXVII, p. 254 gives the average cost by department of running a family of 6. This information let us to replicate the average expenses for food, heating and housing for a family of 6.