Wealth Inequality in Paris 1807-1945

Project Summary :

A research team composed of Thomas Piketty (EHESS, DELTA), Gilles Postel-Vinay (EHESS, INRA-LEA), and Jean Laurent Rosenthal has been working on long term changes in wealth inequality in France

To that effect we have collected estate tax records from archival documents. These data allow us to reconstruct the distribution of wealth at death of Parisians over the 19^th century.

Currently we have colleted data for all individuals who died in 1807, 1812, 1817, 1822, 1827, 1832, 1847, 1857, 1867, 1877, 1887, 1902. We are in the process of collecting data for 1912 and 1922 and will expand the range to WWII in the coming year thanks to a grant from the National Science Foundation (See proposal).

The data sets for 1807 to 1867, 1877, 1887, 1902 are documented below and available for download. Their analysis is documented in (Piketty et al 2006).

I. Archival Sources

All the archival sources are currently stored at the Archives de Paris (Bd Serrurier, Paris XVIII) and come from the French fiscal administration. The Estate administration has changed its recording procedures over time.

For 1807 we used the tables de successions payées (call series DQ8) we collected ages from the tables des deces (call series DQ8) and cross checked the missing wealth information from the déclarations de successions (call series DQ8).

For the years 1812 to 1867 we relied upon the tables des successions et absences (henceforth TSA). The TSA holdings of the Archives de Paris now run through the 1970s. These records include name, occupation, residence, marital status, age, and for individuals with wealth, information about heirs and the date at the declaration was filed. Up to 1870 the TSA also include a summary of the individual’s estate broken down into moveable wealth and real estate. After 1870, the administration stopped recording any wealth information.

For the years 1877 and 1887, we began with the TSA and collected gender, first three letters of last name, day and month of death, age and date at which the estate declaration had been filed. We then turned to the registres de mutations par décès (henceforth RMD, call series DQ7), and collected every declaration that was filed within a two and half years span (e.g. between January 1877 and June 1879).

For 1902, we were able to concentrate on the déclarations de mutation par décès only because they give both wealth information and all the required individual information about the deceased (in particular age). We collected all declarations that related to individuals who died in 1902 and filed before December 1903. When we collected the data, the registers were at the Ministry of the treasury’s depot in Aubervillier, they have since been moved to the Archives de Paris and are in the call series DQ7)

The exact list of registers opened is in (link to Sources.xls). Each sheet in this file contains the call numbers for a particular regression.

II. Data Collection.

For each cross section from 1807 to 1867, we opened all the TSA registers where there might be some estate declarations for a given year. For 1877, 1887 and 1902

We collected declaration on a window of 2 or 2 and 1/2 years starting in January of the sample year.

After 1902, the estate tax was progressive and if individuals filed multiple returns the treasury had to link them. So it was relatively easy to reconstitute an individual’s wealth at death.

Prior to that date we faced two problems, for how many individuals would we collect detailed information about gender, age, profession, marital status, and residence? How would we reassemble the different declarations that related to a given person.

To economize on resources, we collected detailed information only above a threshold (though we have since been collecting age information for all decedents with positive wealth). We chose to collect enough information to be able to know to whom belonged 90% of the wealth left behind that year. Because we did not know what the threshold would be we missed the target in 1807 and 1817 and reopened a large number of registers. Table 1, displays the summary data.

Table 1: Data collection threshold

						Individuals with Names
	Threshold	Adult death	TSA	RMD	Individuals Wealth>0	Number	Share of Total	Share of real estate
1807	12500	11675	3835		3647	859	90%	96.4
1817	17000	11653	12725		3287	637	90.2	96.1
1827	25000	13834	10980		3877	805	90.5	96.9
1837	30000	16582	18176		4922	1046	91	97.1
1847	40000	17831	19737		4814	976	90	96.9
1857	40000	18906	25742		6048	1302	92	98.6
1867	40000	27988	36402		7370	1757	91	95.8
1877	40000	28771		9503	8250	2426	93.3	97.0
1887	40000	34226		10337	9817	2772	94.9	97.3
1902	100000	36265		11750	9830	1676	90.6	93.7

Although the method is economical, we only have to collect nominal information on about 6% of the population, it is not exact. Since we do not know the names of individuals whose heirs filed declarations below the threshold we have to treat all of these as separate individuals. This builds in a downward bias in the estimation of inequality. Nevertheless the bias is small because the main source of multiple declarations were individuals who owned real estate in multiple bureaus but in Paris at least real estate was extremely concentrated and expensive, so the likelihood of under matching is limited.

II.1 Assembling the returns

We aggregated the returns principally based on last name and day and month of death. We accepted some variation in given names because many individuals had multiple given names, we also accepted some variation in address but nearly always a match on day month and last name leads to matches in the other categories.

There are still problems, because of individuals with titles or with ‘de’ in front of a last name. These individuals may appears under different guises and in the TSA multiple times for the same declarations. For instance in 1837 de Pelage comte de Robinault appears both as de Robineau and as Pelage Comte de Robineau for the same estate return, we obviously eliminated one..

There is also the possibility that the same asset can appear in two bureaus (this was common in 1807), and we can only deal with this problem if the return is above the threshold. Hence we have eliminate all double occurrences above the threshold but not below. Again this builds in a downward bias in inequality measures.

II.2 Real Estate valuations

In 1807, the TSA registers give the value of the real estate. Starting in 1812 the registers report the rental value of real assets, at least early on there appears to be some confusion between the two types of reports. We verified suspect (large) declarations by examining the RMD whenever possible

Furthermore when individuals only owned a fraction of a building the values reported sometimes refer to the deceased’s share and sometimes to the value of the building as a whole. This problem persists until the middle of the 19^th century. The problem is that in the registers contain one column where the revenue from the real estate asset is noted and another with the address of the asset, and it may say that only a fraction of the building belongs to the deceased. But the mention of fractions is unsystematic.

Because France is a community property country, the default rule is that building acquired by couple are community property and a deceased individual will only own half (the other half belong to the spouse or to the children if the spouse is already dead) further there were some instance of properties that were not divided among siblings.

To limit the problem we have tried to note both the revenue and the share separately so as to be able to investigate whether our inequality measures a sensitive to these problems. When we just the revenue mentions (ignore the share information) we tend to create more larger fortunes and more inequality, while if we systematically divide by the mentioned share we tend to drive inequality down a bit (because real estate wealth was unequally owned)

III. Data sets

For all the sample years we make two data set available for down load

XXXX.tsa.xls are data sets with the raw data as it was collected for year XXXX

XXXX.indiv,xls are data sets where we have eliminated double entries and assembled the individuals.

1807

1812

1817

1822

1827

1832

1837

1842

1847

1852

1857

1862

1867

Individuals

Ind.

TSA

1872

1877

1887

1902

1912

ALL Years

Individuals

Soon

Ind.

Soon

Ind.

TSA

Soon

TSA

Soon

Specific information for sample years

The column heading are described in the tables below

1807

The sources we use are the registres de successions payées so we have no information on individuals who left no wealth behind. The registers are chronological by date of declaration. They contained no information on age, so the declarations were matched to the registres de deces on the basis of first three letters of last name and date of death.

There were lots of duplicates both within and between bureaus.

Gender information is unsystematic below the threshold.

Real estate is the value of the asset that belonged to the deceased.

The clerk who filled register 334 was sloppy so rather more information is missing than we expected.

1812 to 1827.

Around 1812 the administration put in place registers that were alphabetical so that individuals were sorted on the basis of the first three letters of their last name, and all individuals in the same category who died between 1812 and 1829 are in the same registers. In principal each category is arrayed chronologically by (time of notification or roughly time of death) but in practice they are jumbled, nevertheless all of the references to an individual’s declaration should appear on the same entry.

When we collected 1812 and 1822, we went back and collected all of the age, and gender information and added the first three letters of last name to all positive wealth entries. This allowed us to check our aggregation procedure.

Real estate wealth is given as revenue.

1827 : We are missing a dozen real estate values

1832 to 1857.

In these years registers group people in people in larger categories, but wealth information is systematic. Within a category individual are arrayed by date of report.

1832 : This is a Cholera year and is thus much larger than most

1837 : includes children

1862-67.

Those years are similar to previous ones, except that Paris had doubled in size. To be able to compare 1867 to 1857 we completely enumerated all individuals with a last name starting with T and the addresses of all individuals with last names starting with B and L.

Moreover we had to open a number of declarations because the TSA registers begin to fail to report wealth at death. Something they completely stop doing after 1870.

1877 et 1887.

These are the most recent (and thus the most complete) of our data sets. They were done a full year after the other years ending in 7 and benefit from our experience. We began with the TSA and collected the first three letters of last name, age, date of death, and the dates at which returns were filed by heirs. We then opened the RMD and collected wealth information.

1902.

This is the first full year of the progressive estate tax. And that implied a different data collection process. We were also subject to a 100 year rule for privacy which means we could not collect data past December 1903.

The creation of the progressive estate tax, simplified matters, relative to the procedure for 1877 or 1887: we did not need to open the TSA. Moreover progressive taxation implied that the administration had to track returns that belonged to the same person (so as to be able to apply the relevant marginal rates. Hence we could use the Administration tracking procedure to reassemble declarations. Furthermore individuals only filed returns in a single bureau which simplified matters further.

However upon analysis we realized that we our recovery of Age at death was inadequate relative to what could be accomplished with opening the TSA registers. We have since done so.

Tableau 2 : information sur les bases 1807-1867 :

1807	1817	1827	1837	1847	1857	1867
Total	Total	Total	Total	Total	Total	Total	Total calculé de la succession.
Cote	cote	cote	cote	Cote	Cote	cote	Cote du registre
Bur	bur	bur	bur	Bur	Bur	bur	Bureau fiscal ou la déclaration a été
Annee	Annee	Annee	Annee	Annee	Annee	Annee	Année ou la personne est morte
ID#	ID#	ID#	ID#	ID#	ID#	ID#	Numéro d’ordre
Matched	matched	matched	matched	Matched	matched	matched	Indique que cette observation correspond a une personne observée plusieurs fois (dans plusieurs bureaux ou sous plusieurs orthographes)
Page	Page	Page	page	Page	page	Page	Numero de la page dans les relevés
n° NN	n°	n°	n°	n°	n°	n°	Numero d’ordre l’interieur de la page
1enfant	1enfant	1enfant	1enfant	1enfant	1enfant	1enfant	1= mort enfant
	1indigent	1indigent	1indigent	1indigent	1indigent	1indigent	1= mort indigent
			1gratis	1gratis	1 gratis	1gratis	1= inhumé gratis
Sexe	sexe	sexe	sexe		sexe		Sexe= homme femme g=non noté
Valmob	>Valmob	>Valmob	>Valmob	>Valmob	>Valmob	>Valmob	Valeur mobilière de la succession
Val immob	>revimmob	>revimmob	>revimmob	>revimmob	>revimmob	>revimmob	Valeur (val) ou revenu Immobilier de la succession
					1ere let Nom	1ere let Nom	Première lettre du nom de famille
Nom	Nom	Nom	Nom	Nom	Nom	Nom	Nom de Famille
prénom	prénom	prénom	prénom	prénom	prénom	prénom
Prof	prof	prof	prof	prof	prof	prof	Profession
Adres	adres	adres	adres	adres	adres	adres	Addresse
J	j+	j+	j+	j+	j+	j+	Jour de la mort
M	m+	m+	m+	m+	m+	m+	Mois de la mort
	âge	âge	âge	âge	Age	âge	Age du mourant
Ldécès	Ldécès	Ldécès	Ldécès	Ldécès	Ldécès	Ldécès	Lieu de décès si pas Paris
situamaritale	situamaritale	situamaritale	situamaritale	situamaritale	situamaritale	situamaritale	Situation Maritale (veuf, veuve marié, célibataire)
Héritiers	Héritiers	Héritiers	Héritiers	Héritiers	Héritiers	Héritiers	Héritiers (directe collatéraux…)
DCL	DCL 1	DCL 1	DCL 1	DCL 1	DCL 1	DCL 1	Date de la première déclaration de succession
				Ndcl 1	Ndcl 1	NDCL 1	Numero de la première déclaration de succession
Part	Part	part	part	part	part	part	Part de l’immobillier déclaré qui appartenait au mourant
Lieu imob	Adresimm	adresimm	adresimm	adresimm	adresimm	adresimm	Addresse du bien immobillier
DCL2	DCL2	DCL2	DCL2	DCL2	DCL2	DCL2
				Ndcl 2	Ndcl 2	Ndcl 2
DCL3	DCL 3	DCL 3	DCL 3	DCL 3	DCL 3	DCL 3
End	NDCL 3	NDCL 3	NDCL 3	NDCL 3	NDCL 3	NDCL 3
	DCL 4	DCL 4	DCL 4	END	DCL 4	DCL 4
	NDCL 4	NDCL 4	NDCL 4		NDCL 4	NDCL 4
	End	End	End		END	Autres DCL
						Renvois
						invalides
						END

1902
Bureau	Bureau ou la déclaration a été passée
Volume	Volume ou la déclaration a été relevée
Adcl	Anne ou la déclaration a été passée
NDCL	Numéro de la déclaration
J Décès	Jour du décès
M Décès	Mois du décès
A Décès	Année du décès
H/F	Sexe
Age	Age
Mobilier	Valeur mobilière de la succession
Succession Brute	Valeur brute de la succession
Succession Nette	Valeur Nette de la succession (valeur brute moins le passif)
Imob Paris	Valeur Immobilier Parisien La division des immeuble en co-propriété est faite
Adresse Imob Paris	Adresse des bien ci-dessus
Imob Non paris	Valeur de l’Immobilier non Parisien
Adresse Imob Non Paris	pour le plus important
Passif	Passif de succession comprend souvent des reprises du conjoint survivant
Nom
Prénom
Profession
Adresse
Situation Marit	Situation Maritale (veuf, veuve Marié célibataire)
Héritiers	Héritiers (directe, collatéraux…)
Dcl supplémentaire	Déclaration supplémentaire (permet de relier les déclaration entre elles).
Ndcl1
Dcl2
Ndcl2
Nota
dcl3
Ndcl3
Dcl4
Ndcl4
Dcl5
Ndcl5
END