Post-colonial inequality trends:From the “Four Old Colonies” to Overseas Departments

of France

Yajna Govind ∗†

(PSE, INED)

Abstract

The existing socioeconomic disparities in the French overseas departments arewidely acknowledged. However, despite the recurrent occurrence of protests and riotsin these territories, there has sparsely been an in-depth long-run economic analysis onthis assertion. This study estimates the evolution of income inequality in the four old-est French colonies, now overseas departments; La Reunion, Guadeloupe, Martiniqueand Guyane since their decolonisation in 1946 until recent years. It is the first long-runcomparative analysis of inequality at the regional level in France and is among the fewthat analyses all four overseas departments together. Drawing on a new income taxdataset put together in this paper, the results reveal an inverted-U shaped curve of in-equality over the period. Interestingly, the top 10% income share is consistently higherin the overseas departments compared to the metropolis while top 1% stabilises at thesame level as the metropolis in the recent period. Finally, it is worth noting that thereare differences between the four territories. I argue that these differences could be dueto differences in the colonial heritage, as well as economic and demographic differences.

Keywords: Inequality, Top incomes, La Reunion, France, Overseas Departments, Taxation

JEL classification: D63, H20, N30

∗Contact: yajna.govind@gmail.com†Version: 06/08/2018

1. Introduction

There is a general awareness of the socio-economic disparities that exist in the overseas

departments. The widespread discontentment takes the form of recurrent protests, strikes

and riots1 against major socio-economic difficulties. The assertion of higher level of inequality

in the overseas departments has been assimilated in the common speeches and in the political

sphere. Are the claims of high inequality in the overseas departments legitimate grievances?

Contemporary economic situation of former colonies is affected by two main historical

factors. The first is the colonial setting. Historically, colonies have been built on inequality

in the social and political spheres. Together, these two factors have forged societies with

marked economic divide. In this aspect, three centuries of colonial domination has undoubt-

edly left its imprints on the “four old” colonies of France. The second aspect is the type

of decolonisation process. The decolonisation wave in the mid-20th century took two forms:

the first consisted of granting independence2 to ex-colonies thus ending western domination

and the second, as in the case for the “old four” colonies, was a process of decolonisation by

integration to the metropolis. While the former could go down any path in terms of post-

colonial inequality, the assimilation process generally aimed at progressively eradicating the

existing colonial settings by aligning to the metropolis.

In view of this, I analyse the case of the overseas departments of France. Known as the

four “old colonies”, these territories present an interesting setting given their long common

history with France, the political status of the colonised and later their choice to integrate

the French Republic. Despite being part of France, these territories are marked by various

cultural, ethnic, demographic and linguistic divergences given their historical paths (sec-

tion II). While the juridical assimilation of these overseas territories occurred almost as a

quasi-natural transition from colonisation to departmentalisation, it is agreed upon that this

process has not led to a complete alignment with the metropolis. France has witnessed a

two-speed development process since the decolonisation period, with its overseas counterpart

left behind on various fronts.

In this paper, I construct a novel income tax dataset at the overseas departmental-level

1List of major riots/protests in the overseas departments: Violent riot in 1959 Martinique leading toanti-colonial protests against oppression, 13 years after departmentalisation; Riot May 1967 in Guadeloupeerupted due to racism, leading to workers protests demanding improved economic conditions; Riot in 1996 inGuyane which started with demands for the local education system; Riots in La Reunion in 2005 and 2012and general strike in the Antilles in 2009 against the high cost of living and the unacceptable low standardof living

2Either through an agreement or with conflicts

1

covering the period 1950 - 20143. The paper has three main goals: to estimate the evo-

lution of income inequality in the overseas departments since their departmentalisation; to

perform comparative analysis between the departments and the french national level; and to

identify potential mechanisms that could be at play. I analyse the extent to which the gap

in the level of inequality between the ex-colonies and the metropolis has narrowed down in

the post-colonial period. By doing so, I evaluate the success, or lack thereof, of France to

remedy the unequal setting that the overseas departments inherited from their colonial past.

Most of the existing literature on the evolution of inequality in France have been con-

ducted at the national level (Piketty, 2001; Garbinti et al, 2018). This paper is the first to

take an interest in a departmental-level analysis. Moreover, given the lack of data on the

overseas departments, there are unsurprisingly few long-term series constructed for these ter-

ritories. Most of the surveys that exist in metropolitan France were only recently extended

to the overseas departments. Similarly, the establishment of post-colonial administrative

bodies in these territories took time given the facilities at hand at the time. These factors

considerably limited the work that could be done over the whole post-colonial period. In

this sense, the construction of this fiscal database represents a major contribution to the

literature on the overseas departments.

It needs to be kept in mind, however, that while these territories are grouped under

one umbrella term, it hides very different realities one from the other. Moreover, due to

the lack of comparable data or collaboration between the different local research institutes,

until recently these departments were mostly analysed separately or quite artificially juxta-

posed. This paper thus also constitutes a first attempt at performing a coherent comparative

analysis- between departments and in comparison with the French national level.

It is generally known that there are significant socio-economic disparities between main-

land France and the overseas departments today. This can be easily deduced from the

recurrent protests and strikes over crime, economic difficulty and poor quality of social ser-

vices that have erupted in all four territories over the last decade. It has led to mounting

pressures on the French government to recognise the plights of the far-off territories. In this

light, various action plans such as the enactment of the bill on “Real equality for overseas

department4 have been undertaken in the recent years. These are signs of acknowledgement

of the unacceptable situation but a lot still remains to be done.

3With some breaks in the series4Loi no 2017-256 passed on the 28th February 2017 on Egalite reelle des outre-mers

2

The distress of the overseas departments is partly reflected in the stark levels of inequality

that persists.5 A study by Bernier and Maurin (2013) the Gini Index6 at the departmental-

level in 2011. They find that La Reunion notoriously ranks first as the most unequal French

department (0,53), overtaking Paris (0,50), followed by Martinique (0,47), compared to an

average Gini index of 0,31 in metropolitan France. While it gives an idea of the current

situation in the overseas departments, there is a call for more in-depth and a longer-run

analysis. In fact, Angeles (2007) argues that colonial history is a major explanatory factor

behind todays large differences in inequality. He finds that the percentage of European set-

tlers is associated with significantly higher values of todays Gini coefficient.

The interest in examining the trends in inequality in the overseas departments of France

draws on a body of literature that looks at the colonial period. Atkinson (2014) studies the

distribution of top incomes in former British colonies in Africa. He identifies three factors

that led to the current evolution of inequality: the presence of a colonial elite and the power

of the ruling class; the evolution of the colonial income concentration and; the legacy at the

time of independence. Similarly, there are studies on the evolution of income inequality in

French colonial societies (Alvaredo et al, forthcoming)7 and British India (Alvaredo et al,

2017). The aim of this paper is to focus solely on the post-colonial period, thus complement-

ing the literature on inequality in former colonies.

This paper is organised as follows. In Section 2, I lay down the peculiar historical

trajectories, demographic evolutions and economic situations in the overseas departments.

To estimate the long-term series of income inequality in the overseas departments of France,

I then describe the data sources and the methodology that is used in this paper in Section 3.

I present the evolution of income inequality of the overseas departments and a comparative

analysis with the French national level in Section 4. It is followed by a discussion of the

potential underlying mechanisms in Section 5.

5See Jonzo (2009) for the case of La Reunion between 2001 and 20066Varies between 0 (highly equal) and 1 (highly unequal)7On French Algeria, Cameroon, Indochina and Tunisia

3

2. Background

2.1. Historical Context

Figure 1 depicts the historical timeline since the outset of the colonisation of the four

“old colonies”. These territories have been part of the French empire as early as the 17th

century8. They have been managed by different entities- the French East India Company or

the French West India Company and later, directly by the French Crown. They have since

remained part of French for most of the period. After a brief British domination during the

French Revolution and the Napoleonic war, these four territories were ceded back to France

in 1814 with the Treaty of Paris. Thereafter, these territories were marked with major inter-

nal conflicts with respect to the emancipation of the oppressed population. As seen in Figure

6 to 9, slaves constituted around 60-80% of the population, the rest being white population

in the mid-18th century. By 1842, the white population constituted only around 6-10% of

the population in the Antilles and Guyane compared to around 20% of the population in La

Reunion.

Fig. 1. Historical timeline

The growing share of freed coloured population together with the emancipation slavery

in neighbouring British colonies has led to mounting pressures on the local colonial forces

to give in to the abolition of slavery in the four “old colonies”. In 1848, the ex-slaves in

these colonies were all emancipated and acceded a pseudo-citizenship status. The constant

need for cheap labour led to the immigration of Africans and Indians on these territories.

It is only a century later, in 1946 that these territories were fully transformed into French

departments. This rather rare form of decolonisation process was thought in a logic of in-

stitutional, judicial and cultural assimilation. Three centuries of colonial domination was

deemed ample to instill French values in the population.

8Even before some metropolitan territories such as Nice, Savoy and Corsica

4

Fig. 2. La Reunion

Fig. 3. Guadeloupe

5

Fig. 4. Martinique

Fig. 5. Guyane

In the mid-20th century, these four territories were marked with deep social divides in

their own respects. There existed, for instance, a segregation between the white economic

elite and the African and Asian descendants in la Reunion. The latter advocated for a

decolonisation by integration to France as a way to end the local elite domination who

supported independence. This divide also existed in the Antilles between the economically

6

powerful white descendants, Bekes and the rest of the population. Guyane faced a different

challenge in terms of its border disputes, mass migration inflows and dismal inequality. In

this sense, the assimilation process was thus an indirect cry for increased legal, social and

economic equality within these territories as well as with their metropolitan counterparts.

The immediate post-colonial period witnessed a generally alarming socio-economic situ-

ation in the overseas departments. Marked by high illiteracy rates, low sanitation level, low

life expectancy and almost no productive sector after the decline of the sugar industry, these

newly-turned departments were in no way comparable to the metropolis. In fact, well into

the mid-20th century, the economy of these territories (except for Guyane which was mostly

dependent on gold extraction) was still controlled by the small plantocracy. In the face of

this reality, the French government undertook a step-by-step action plan to gradually tackle

the problems at hand.

The socio-economic development of these territories has occurred in three phases (Drozin,

2001). Starting modestly in the 1950s, the first phase consisted in addressing the urgent san-

itary and social situation. For a decade, the new departments witnessed the setting up of

basic medical, familial and child supports, the creation of the Caisse Generale de Securite

Sociale, basic labour rights. These were set at a comparably lower level than metropolitan

France. Given the prevailing level of poverty, a large section of the population was depen-

dent on the social system. This period has also seen the consolidation of formal migration

policies9 providing incentives to immigrate to the metropolis.

The second phase, from the 1960s to the 1980s, has known an increased focus on DOM-

specific policies. This period has also witnessed the setting up of the post-colonial French

administration and the mass creation of public jobs. In La Reunion for instance, these ef-

forts led to a 26-fold increase in the average real income between 1950 and 1970 (Coder et

al, 2016). However, the social benefits put in place were very restrictive and set at a lower

level than in the metropolis due to the burden of high birth rates and rapidly growing pop-

ulation. Given the predominantly low levels of literacy rates and the lack of qualified local

population, incentives were provided to metropolitan French to take up civil service jobs in

the overseas territories.

The third phase starting in the early 1980s, marked an intensified effort to tackle the

9Creation of the BUMIDOM- Bureau pour le developpement des migrations dans les departementsd’outre-mer in 1963

7

persistent levels of inequality. It took the form of decentralization of power in the mid-

1980s. The devolution of power from the central government to the regional level led to

the revamping of the local social system. There were the implementation of departmental-

level programs. Family allocations and minimum old-age pensions were extended to a larger

share of the population. This period was also marked by the gradual alignment of benefits

to the metropolitan level- the Revenu Minimum d’Insertion (RMI) in 1989, followed by the

unemployment insurance in 1991, family allocations in 1993, the alignment of the minimum

wage in 1996 and the facility of Youth Employment in 1997 among others. This social benefit

alignment process was completed by the beginning of the 20th century.

2.2. Economic Situation

The overseas departments are among the richest territories compared to their neigh-

bours. It is unsurprisingly so since they are attached to and their economy sustained by the

metropolis.

Decolonisation in the overseas departments of France took the form of institutional as-

similation rather than independence as was the case around the same period in other colonial

settings, namely the British framework and the setting up of the Commonwealth. The four

oldest colonies of France, La Reunion, Martinique, Guadeloupe and Guyana acceded to the

status of department following the Assimilation Law in 1946 and became European entities

since 1957. The unified State of France through the integration of the colonies in France

was inherently an extension of a colonial setting whereby assimilation was primed over the

recognition of the local cultural and ethnic differences and the diverse institutional needs

present in those territories. The framework introduced had little regard for these differences

and hence could not achieve its purpose of complete assimilation. There was a need for a

reconciled path between equality and diversity.

2.3. Demographic Conditions

From a demographic point of view, the composition of the population have also changed

a bit over the years and are different from one dept to the other. Taking the case of Guyane,

it can be seen that there is a very significant share of foreigners (mostly migration from

neighbouring countries- Suriname and Brasil) which explains, along with the lower adminis-

trative infrastructure, the lack in terms of fiscal data. As for the other overseas dept, there is

a growing share of metropolitan French since the 1950s. The increase in the earlier periods

might have been driven by the creation of public admin jobs which had various benefits- a

premium on the national public servant wage and a reduction in the taxation paid. The

8

metropolitan French increased from being a minority of 1% of the popn in 1954 to 10% in

the most recent years.

While the French Antilles have completed their demographic transition in a much shorter

span of time than mainland France, La Reunion is still in its phase of transition with a

low death rate but high birth rates owing to the high fertility rate. The distinct demo-

graphic structure of these far-flung territories owe to the migration flows they have witnessed

throughout their history. Recent works have identified a major pattern of rapid ageing of

the population in the DROM (Marie and Rallu, 2012). However, this common ageing trend

with mainland France does not share the same underlying factors. In fact, in the overseas

departments, this pattern emerges from interesting DROM-specific aspects. Firstly, in the

face of the high unemployment rate3, especially for young adults, many of them are driven

to move to the mainland for their education and jobs and return later in their lives. In

addition to this, some of the overseas territories are also viewed as retirement destinations,

be it by metropolitan French or by native-DROM who chose to return to their native land

much later in their lives (Marie and Rallu, 2004).

Fig. 6. La Reunion- Composition of the population

9

Fig. 7. Guadeloupe- Composition of the population

10

Fig. 8. Martinique- Composition of the population

11

Fig. 9. Guyane- Composition of the population

12

3. Data & Methodology

3.1. Data

In order to establish the long-term evolution of income inequality in the overseas depart-

ments, income tax data published annually by the tax administration coupled with national

accounts and population census are used. The income tax data consists of tabulated tax data

at the departmental level. This contains information about the number of tax filers and the

total income per brackets of income, spanning over the period of 1950 to 2014 intermittently

for the different departments. These income tax data are obtained from different sources

and can be accordingly categorised in three different periods:

i) 1950 - 1985: Annuaire Statistique of the Overseas Departments and INSEE La Reunion

ii) 1986 - 1998: Etats 1921- Centre des Archives Economiques et Financieres (CAEF)

iii) 2000 - 2014: Direction Generale des Finances Publiques (DGFiP)10

Starting from the most recent period (2001 - 2015), income tax data is obtained from

the Direction Generale des Finances Publiques (DGFiP) for all four departments. Similarly,

data for the period 1986 - 1998 for all four territories are obtained in the forms of tables

annually published in a fifty-pages long pamphlet. This is known as the ”Etats 1921” origi-

nally published for internal use by the Ministry of Finance. These data do not violate any

statistical confidentiality rule as it includes a large number of taxpayers.11 As for the pre-

ceding period, the data is obtained from the various Annuaire Statistiques of La Reunion,

Guadeloupe, Martinique and Guyane, published by INSEE over the period 1950 - 1974.12

The availability of data for the different department are more or less sparse and do not

cover the entirety of the period for all departments. For the period 1972 - 1985 and 1988,

partial tax data for La Reunion is retrieved from an INSEE summarised publications. Un-

fortunately, the latter are not as detailed as the previously-mentioned sources as it was used

for expository purposes only. These summary reports only report the number of taxable

taxpayers per income brackets with no information about the corresponding incomes in the

brackets.A summary of the availability of the data over the whole period is presented in

Figure 20 in Appendix A.

Prior to 1999, the income tax data for the year n normally corresponded to the income

10As of this date, data for 2004-2014 can be retrieved online from www.impots.gouv.fr11Data concerns a group of more than 11 persons12The latest year corresponding to income perceived in 1972

13

of the same year. This changed as from 2000, whereby the income tax data for year n is

based on the income of year n-113 The tax administration normally publishes income tax

data based on income perceived in year n in both the following year 31/12/(n+1) and the

year after 31/12/(n+2). The latter is in principle the most up-to-date data as it takes into

account tax audits, tax reliefs and changes in family status which occurs in the year after the

imposition. The latest tabulations available in all years are thus used. As of now, the latest

publication available is that of 2015 tax tabulations, which as explained above, corresponds

to 2014 incomes. The comparability of the publications across time is generally consistent,

except for changes in income definition used over the years, which is described in (#REF).

It is worth noting that there has been noticeable changes in terms of the number of thresh-

olds reported in different years. The aim for this frequent update of the number of threshold

is normally to provide more detail at the upper end of the distribution as taxpayers reports

increasingly higher taxable income over the years.

As explained in more details in Appendix B- Control Total for Population, the unit of

analysis in the tabulation tax data is the tax unit. It is conceptually close to a household

unit which the preferred unit of analysis in economic surveys referring to people living in

the same household. In contrast, tax units refers to a person or group of persons that fills

a unique tax form. In France, adults above 18 years old can choose to declare their income

separately from their parents’ declarations. This becomes mandatory as from 21 years of

age.14 Among them, married or PACSed15 couples are required to fill a single tax form.

Hence, there may be more than one tax unit in a household. For instance, a cohabiting

unmarried couple would constitute a single household but two tax units. As is done in this

literature and for the sake of consistency over time, a tax unit is estimated as an adult above

20 years of age or a married couple (see Appendix B).

Due to changes in tax laws, an attempt to create a homogeneous series requires various

corrections as explained in Appendix C. Apart from income tax data, this analysis also relies

on population and income data. Demographic data are primarily obtained from population

censuses over the whole period. These data exist in the form of result reports for the following

years: 1954, 1961, 1967, 1974 and 1982. The later censuses, namely for the years 1990, 1999,

2009 and 2014 are digitalised databases. These contain information on different demographic

13This mechanically led to an artificial gap in the publication in income tax data in the year 1999, sincethe income tax data of the year 1998 corresponds to the income of the same year while the income tax dataof 2000 refers to the income of the preceding year, 1999.

1425 years of age for students15A civil solidarity pact- a contractual form of civil union

14

factors of the population such as age, gender, birth place, marital and socio-economic status.

Departmental-level income estimates are primarily obtained from national accounts compiled

by INSEE. This covers the entire period for La Reunion and unfortunately exists only as from

the 1970s for Guadeloupe, Martinique and Guyane. For the previous period, the national

income series are estimated based on some assumptions (for more details see Appendix C).

Advantages and Limitations of fiscal data

A large part of the literature on top income shares has made use of tax data. The

analysis of the top part of the distribution provides a lot of insight for political economy

debates as in some cases, they concern a huge proportion of total income. However, given

the various limitations of fiscal data, an alternative to this source of data would be to use

household budget surveys. However, the latter has various limitations since it is subject

to misreporting especially at the upper end of the distribution. This may take the form of

under-reporting or top coding for confidentiality reasons and precisely to avoid the issue of

misreporting. Hence fiscal data represents a more reliable source of data when looking at

top incomes. Moreover, fiscal data has been collected and annual records are thus available

since the onset of the imposition of income tax, which dates back to the begining of the

20th century in metropolitan France and as from the 1940s in the overseas departments. A

similar analysis using household budget survey would not have allowed for such a historical

series. For instance, the French household budget survey (Enquete Budget de Famille) was

first conducted in the DOM in 1995 and are available every five years, compared to annual

tax data. Perhaps an in-depth, yet more tedious work would require an analysis of income

inequality which combines both tax data and household budget surveys.

While the use of tax data has been adopted by various works contributing to the WID

database, some caveats should be acknowledged. In particular, fiscal income is not devoid

of measurement and technical problems. First, fiscal income might diverge from national

income and be biased due to tax exemptions, tax avoidance and evasion. Hence, the esti-

mations obtained might understate the income accruing to the top to the extent that part

of their income are not declared. This is especially the case since the rich have more incen-

tive to under-declare their incomes or redirect their income to tax havens to avoid paying

taxes. Moreover, there are various factors which makes analysis of income inequality esti-

mates based on tax data in different countries and across time not perfectly comparable. In

fact, since these data are primarily for administrative purposes, the income definitions and

unit of analysis might differ, in which case the comparability of two series relies solely on

15

the ability of corrections to reach a common definition. In addition, the focus on pre-tax

and pre-transfer income inequality do not take into account re-distributive effects of public

policies. This limitation is especially true in the case of the overseas departments where

attempts in correcting income inequality is largely made through social transfers (#REF).

3.2. Methodology

This paper follows the work of Piketty (2001) and Garbinti et al (2018) which establish

a thorough study of top incomes at the national level in France using fiscal data. These

papers have made use of exhaustive tax tabulations published by the tax administration to

construct the income distribution in France and estimate top income shares. This paper

aims to construct an analogous series at the departmental level for the overseas departments

of France. Prior to 1985, only tax units subject to taxation were subject to income tax

declarations. While it becomes mandatory as from the mid-1980s to fill in a tax form, it is

only gradually applied in the overseas departments. Hence there is a need to estimate the

total number of tax units and their corresponding income over the whole period to better

estimate the whole distribution. These concepts, known as the control total for population

and income respectively, are explained in this section. I also explain go over the corrections

made to harmonise the series over time. I then describe the interpolation technique employed

this study.

Population Estimation

As explained above, in order to estimate top incomes shares, we need to estimate the

total number of tax filers. It is the total number of tax units which would have been observed

had everyone been required to fill in a tax form. Over time, a greater proportion of tax units

is captured in the tax data. A summary of the evolution of the adult population (above 20

years old), the total estimated number of tax units and the proportion of declaration are

presented in Figure 10 and 11.

16

Fig. 10. Adult population

The control total for population estimate is based on information about the section

population that is liable to pay taxes, defined as the population above 20 years old and data

on married couples obtained from the french population census. The hypothetical control

population is simply the difference between the total population above 20 years of age and

the number of married couples (equation 1). As population data are available every ten years

on average, simple interpolation techniques are applied to estimate an interrupted population

series. Note that young adults are legally allowed to detach from their parents’ fiscal unit

as from the age of 18 and are not obliged to do so until the age of 21 (or 24 in the case

of students). Moreover, marriages and divorces during the year of imposition might affect

the number of tax declarations16. Hence, these two factors imply that this method will not

provide a perfect estimate of the control population but is nevertheless precise enough. A

discussion on the choice of the age of the adult population and a detailed explanation of the

steps in the estimation of is made in (Appendix B). The estimate of tax units are shown in

Figure 11.

16For instance, a marriage during the year would entail three declarations in total- two separate declara-tions for the income received before the marriage and one declaration for the couple thereafter.

17

Fig. 11. Total number of tax units

As seen in Figure 13 the fraction of tax unit subject to positive income tax has hov-

ered around 20-30% in La Reunion, a much lower level than in metropolitan France which

stabilised around 50-60% (Garbinti et al., 2017). As expected, the proportion of tax unit

subject to income declaration has increased over the years and was made mandatory for all

tax units as from 1985 in France. However, as observed in Figure #REF, it is only after

the 2000s that there is 100% declarations in the overseas departments, hence the importance

of constructing a series of control total for population for the preceding period. It is assumed

that all tax units declare their income as from the year 2003.

18

Fig. 12. Proportion of tax declarations

Income Estimates

Similarly, there is a need to estimate the total income which would have been reported

if all the tax units were required to fill a tax form. As assumed in the previous section, all

tax units are required to declare their income as from 2003 and hence, the control income

for those years corresponds to the total taxable income from the tax data. However, for

the years prior to 2003, a control total for income needs to be estimated. Two approaches

with two different starting points have been employed in the literature to make such an

estimation. The first one consists of starting from the income observed in the income tax

data and adding an imputed income for the non-filers. The second approach uses an external

control, typically national income accounts correcting for non-household income and other

non-relevant incomes to obtain the total taxable income. The relationship between the

national income and the taxable income is shown in table 1.

19

Table 1: Relationship between National income and Taxable income

Balance of Primary Income

(-) Non-household incomes

Household sector total income

(-) Items not included in the tax base

Household Gross income

(-) Non-declared income

(-) Non-filers

Declared taxable income of filers

Fiscal income may hence diverge from national income due to production taxes and the

part of income not subject to taxation and thus not declared in the tax data. The latter

may include imputed rent (rental income from owner-occupied housing), employers and em-

ployees social security contribution, tax-exempt life insurance income and other tax-exempt

income, for instance interest paid to deposits and savings accounts and non-taxable transfer

payments. On the capital front, fiscal income also excludes corporate retained earnings and

corporate taxes. In this paper, we adopt the second approach of estimating a control income.

In addition, due to the lack of detailed national accounts data at the departmental level in

France, especially so for the overseas departments, I build the estimation of a control total

for income based on the series established for France. The steps used in the estimation of

this taxable income series for the overseas departments are detailed in Appendix C - Control

Total for Income.

TO REVIEW: PARAGRAPH ABOUT GDP- maybe put the diagram in

the economic situation section The ratio of the GDP per capita between La Reunion and

France is shown in #REF. We then estimate the ratio between the average taxable income

per adult population of France to its GDP per capita. Based on the assumption that all the

tax units declare their income as from 2001 in La Reunion, we estimate the ratio between

the average taxable income per adult population of La Reunion to its GDP per capita for the

years 2001 to 2014. We then assume that the relationship between these two ratios remains

constant at 1,26 (See Appendix C for more details) and are thus able to estimate the ratio

for La Reunion and consequently, the taxable income for La Reunion for the years before

2001.}

20

Definition of Income and Corrections

The definition of income used by the administrative tax is the net taxable income which

is the fiscal income less some deductions. However, since the income definition of interest is

the income reported in tax declarations before any adjustments, namely the fiscal income,

there is a need to correct the taxable income for these deductions to obtain the fiscal income.

This is especially so since these deductions change over time and may lead to biased estima-

tions of trends. In France, the tax law allows for various main deductions which are as follows:

i) A 10% lump-sum deduction for professional expenses of wage earners, currently capped

at 12 183 e per member of the tax unit.

ii) An additional 20% deduction for wage income, up to a ceiling which has been repealed

in 2006.

Apart from corrections for these deductions, the series are also corrected for capital gains

based on the corrections made in the French national series. Figure #REF shows the trend

in taxable income based on the estimation described in the previous section and fiscal income

based on the above-mentioned corrections. There is a clear jump in taxable income in 2006

due to the repeal of the 20% deductions for wage income.

21

Fig. 13. Average Fiscal Income

Generalised Pareto Interpolation Technique

Since the income tax data is in the form of tabulations and the given thresholds do not

generally coincide with the percentile of the population that is of interest (for instance the top

1%, top 0.1%...), there is a need to apply an interpolation technique to obtain the shares of

top incomes. As lengthily explained in Atkinson et al. (2011) and confirmed in the literature

on top incomes, the top tail of the income distribution is usually well approximated by a

Pareto distribution (power law) and has been widely used in the literature on top income.

In order to estimate annual series of fiscal income for the whole distribution, the generalised

non-parametric Pareto interpolation technique developed by Blanchet et al. (2017) is applied

to the tabulation tax data.

22

4. Results

4.1. Top Income Thresholds

Figure 14 and 15 depicts the minimum income required to be part of the top 10%, top

1% and top 0.1% of the distribution respectively.

Fig. 14. Threshold income of top 10%

The top 10% income threshold in the overseas departments have consistently been lower

than the corresponding threshold in France. Notice that the gap between the two has dras-

tically reduced over the years, especially so for Guyane and the Antilles. In fact, the 7 000 e

- 16 000 e gap in the mid-1980s narrows down to 6 000 e - 10 000 e today. When looking

at the income threshold to be in the top 1% of the distribution, in figure 15, the threshold

for La Reunion is very close to that of France for most of the period. In the period from

the 1960s to the 1980s, the top 1% threshold of La Reunion is higher than the threshold of

France. At the very top of the distribution, the income threshold for La Reunion is close to

that of France until the mid-1980s and decreases thereafter. As for the Antilles and Guyane,

23

Fig. 15. Threshold income of top 1% and top 0.1%

the top 0.1% for the entire period is much lower than the one at the national level. Taken

together, these income threshold trends suggest that those who are part of the top incomes

are less rich in the overseas departments than at the national level17.

4.2. Top Income Shares

This section presents the results obtained from the estimation of the top income shares

using the generalised Pareto method. The results for the overseas territories are put in per-

spective by comparing with the trends in top income shares in France obtained by Garbinti

et al (2018). In order to understand the following series, one needs to grasp the concept of

top income shares. As an illustration, in a perfectly egalitarian economy, the top 10% of the

distribution would own 10% of total income. Similarly, the top 1% would own 1% of total

income. If the share of the top 10% is estimated to be 20%, then the top 10% own twice the

income they should have owned under a perfectly egalitarian economy. Evidences from the

WID series suggest that countries with the lowest levels of inequality have top 10% income

shares of around 20%.

Overall, the top income shares series spans over a 60-years period from the 1950s to 2014.

The beginning of the period until 1986 is intermittently covered in the different departments,

La Reunion having the most complete data. Thereafter, an almost uninterrupted series is

established up to 2014 for all four overseas departments.

17Except for the top 1% and the top 0.1% in La Reunion from the mid-1960s to the mid-1980s.

24

Top 10%

Figure 16 shows the estimation of the top 10% in the overseas departments compared to

France. The top 10% series has been computed since the 1950s in Guadeloupe, Martinique,

La Reunion and as from the late 1980s in Guyane (See Appendix A for details about data

coverage). Three main points stands out from this graph. First, the top 10% income shares

seems to follow an inverted U-shaped curved over the period. Based on the series of La

Reunion, which provides the most complete picture, there is an initial upwards trend in the

shares of top 10% income shares until the 1970s, peaking at a share of 80%. This is one of

the most extreme level of inequality witnessed in the world. It is higher than most unequal

regions in the world- the Middle-Eastern regions and South Africa which have top 10% shares

as high as 65- 70% (Alvaredo et al, 2018; Alvaredo and Atkinson, 2010). This is followed

by a general declining trend since the 1980s in the overseas departments. The period as

from the mid-1990s is marked with a relative stabilisation followed by a slight decline at the

beginning of the 21st century and again a period of stabilisation.

Second, the level of inequality is consistently higher in the overseas departments com-

pared to France. Despite the significant drop in the level of top 10% income shares in the

overseas departments, it is still higher than the national level at its lowest points. In the

late 2000s, the top 10% share is around 35% in France compared to around 40-44% in the

overseas departments. This is in line with the INSEE analysis which ranks the overseas

departments among the most unequal departments of France, based on survey data. The

third point to be noted is that despite the common factor that the overseas departments are

more unequal than France, the levels of inequality are different between the four territories.

Among the four, La Reunion seems to have the highest top 10% shares, followed by the

Antilles and Guyane.

25

Fig. 16. Top 10% income shares

26

Top 1%

Fig. 17. Top 1% Income shares

Figure 17 shows the top 1% income shares in the overseas departments in comparison

with France. Similar to the trends observed for the top 10% income shares, we can discern

a downward trend in the top 1% shares in the overseas territories since the beginning of the

period. The bulk of the reduction seems to occur as from the 1980s from shares as high as

27% in La Reunion and to the level of France at around 10% in the recent years. France’s

top 1% income has been at its highest point of 23% in the inter-war period, in the 1920s.

As noted before, there are significant differences between the four territories prior to the

1990s. La Reunion is the territory with the highest inequality, reaching levels higher than the

most unequal countries in the world- e.g South Africa and the US. There are also differences

between the Antilles territories at the beginning of the period, with a top 1% income share at

around 18% in Martinique compared to arounf 14% in Guadeloupe. This is not so surprising

as is discussed in the next section. Interestingly, contrary to the case of the top 10% income

27

shares, there seems to be a complete ”catch-up” of the overseas departments to the French

level of top 1% income shares. Despite this sizeable distance between the top 1% shares

between the overseas departments and France, they reach an almost comparable level in the

recent years. This being said, the absolute level of income required to be in the top 10%

(P90) and top 1% (P99) is on average lower in in the overseas departments than in France.

Top 0.1% and Top 0.01%

Fig. 18. Top 0.1% and top 0.01% income shares

Figure 18 shows the evolution in the income concentration at the very top of the distri-

bution, namely the top 0.1% and the top 0.01% in the overseas departments and France. For

the top 0.1%, we again find that the share is relatively much higher in La Reunion compared

to the other territories and France in the 1960s. It stands at around 8%, and reduces sig-

nificantly to approximately 3% in the mid-1980s with a declining trend thereafter until the

2000s. After 2000, the top 0.1% income shares of all four overseas departments hover around

2-3%. Again, France had comparable shares of 9-10% in the early 20th century, between

1915 to 1923. With the rapid catch-up phase in La Reunion and milder in the Antilles and

Guyane since the mid-1980s, they reach a lower level of top 0.1% shares than France as from

1996.

The top 0.01% also follows the same pattern with a very high top 0.01% share of 1.7%

in La Reunion in the 60s thereafter falling to around 0.8% in the mid 1980s. On the other

hand, the Antilles seems to be at a comparable level with France. As is the case above, as

from 1995, the top 0.01% shares in the overseas departments are on average lower than that

28

of France. Taking a longer historical perspective, France has witnessed a much higher top

0.01% income shares in the range of 2-5% during the period from 1910 to 1930.

Bottom 50%

Fig. 19. Bottom 50% Income shares

I also estimate the share of income going to the bottom 50% of the tax units in the

overseas departments. Figure 19 depicts this share since the 2000s18. It can be seen that in

general, the share accruing to the bottom 50% is very low over the whole period. This implies

that the average income of the bottom 50% of the population was approximately 5% of the

average income in Guyane, 8-9% in La Reunion and Guadeloupe and 11% in Martinique.

18Refer to A to understand the period covered

29

5. Discussion

Periodisation

As explained before, the most complete data series exist for La Reunion while it covers only

a few years in the mid-50s for the Antilles and a full series as from 1986 in the Antilles

and Guyane. This long-term availability of data for La Reunion provides a comprehensive

picture of the case of that department. This combined with the data points available at

the beginning of the period for the Antilles will also allow to make inferences with respect

to these two overseas departments. Finally, due to lack of good quality data in Guyane,

the analysis will only focus on the post-1990s trends. The period from 1951 to 2014 can

be broken down in several sub-periods. The periodisation up to the beginning of the 21st

century follows closely the one laid down for La Reunion by Drozin (2001).

General Structure of the following paragraphs (which is divided in four pe-

riods):

1. First, more unequal than France colonial setting-poverty, literacy rate

2. Second, the differences between Antilles and La Reunion- education?, differ-

ence between MTQ and GUA- beke?

3. Third, the trend (evolution over time in this period- increasing trend)

4. Policies put in place in that period- 4.1 mostly social, at lower levels than

france + setting up of the public employment. evidence:

The first period consist of the immediate post-departmentalisation period until the

mid-1960s19. While departmentalisation is voted in 1946, there is no sharp break between

the colonial and post-colonial period until much later. There are no major changes in the

socio-economic situation in the newly turned departments until much later. The relation-

ship between the overseas departments and the metropolis relationship remains largely in

an imperial setting until the turn of the 60s (Blerald , 1986). Hence, the first data point

corresponding to the early 50s in the Antilles and La Reunion provide a fairly good picture

of the situation at the end of the colonial period.

To begin with, the overseas departments are more unequal than France at the beginning

of the period. This is not surprising given the inherent unequal nature of colonialism. In

fact, the widespread support for the law of departmentalisation relied largely on the promise

19The analysis for this period relies only on the series estimated for the top 1% and top 0.1% incomeshares

30

of equality within the overseas territories and with the metropolis. Various reports to the

government underline the extreme poverty in which the majority of the population in the

overseas departments live. For instance, in the industry and commerce sector, the Antilles

witness an hourly minimum wage (SMIG/SMIC) that is constantly lower than in France

as seen in the table 5. The minimum wage in the overseas departments stabilises at 20%

lower than the metropolitan rate. Moreover, while the minimum wage is merely a minimum

threshold in metropolitan France, the mass of the wage employees of the private sector in the

overseas departments earn close to the minimum wage prevailing there. In addition, there

existed another minimum wage until the 70s called SMAG at a much lower rate than the

minimum wage (SMIC) which was earned by the majority of the employees in the rural area

line with the assimilationist policy of France, the level of wages

Year Antilles France

1952 0.83

1955 1.04 1.25

1958 1.18 1.46

1962 1.53 1.72

1965 1.88 1.97

1966 1.99 2.06

1967 2.03 2.13

1968 2.45 2.68

1969 2.78 3.16

1970 2.86 3.42

1973 4.43 4.95

1975 6.16 7.27

1976 7.30 8.34

1977 8.21 9.40

1978 9.24 10.61

1979 10.56 11.94

1980 12.09 13.80

1981 13.30 17.76

1982 16.91 19.64

1983 18.62 22.33

Table 2: Values are in francs Source: INSEE, Blerald (1986)

This period presents a general increasing trend in inequality, especially so in La Reunion.

31

There is also a blatant difference between La Reunion and the Antilles, with the former

experiencing a much higher level of inequality. This period marks the transition from the

colonial past in these territories. The French government’s efforts were concentrated on

addressing the most urgent issues, in particular the sanitary and social situation. For almost

two decades after decolonisation, the policies put in place consisted only of basic medical,

familial and child supports and basic labour rights at lower levels than those prevailing in

the metropolis.

The second phase, between 1965 to 1980, is characterised by a stabilisation when look-

ing at the top 10% income share and a mild decreasing trend in the level of inequality when

looking at the income shares of very top of the distribution.

parskip 1. First, still higher level of inequality, be it top 10% or top 1% income shares.

2. More rapid decline in the top 1% and top 0.1% than in the top 10%.

3. Policies

A third phase of rapid declining trend can be discerned from 1980 to the mid-90s.

1. No discernable difference between the different overseas departments at the very top of

the distribution

2. Top 10% higher than France and no difference in top 1%. Why could that be?

3. Policies

The fourth and final phase consist in the relative stabilisation at least in terms of the

top 10% and top 1% income shares in the overseas departments from the mid-90s to 2014.

1. Stable

2. No difference

3. Policies

Maybe speak of no diff btwn DOM and top 10 higher than France while top 1 not higher

here. In common for the last two periods.

32

6. Conclusion

33

References

Alvaredo, F., Bergeron, A., Cassan, G. (2017). Income concentration in British India,

18851946. Journal of Development Economics

Alvaredo, F., Assouad, L., Piketty, T. (2017). Measuring Inequality in the Middle East

1990-2016: The World’s Most Unequal Region?.

Alvaredo F., Cogneau D., and Piketty T., (forthcoming) Income Inequality under Colo-

nial Rule. Evidence from French Algeria, Cameroon, Indochina and Tunisia, 1920-1960.,

Working Paper.

Alvaredo, F., Atkinson, A. (2010). Colonial Rule, Apartheid and Natural Resources: Top

Incomes in South Africa, 1903-2007.

Angeles, L. (2007). Income inequality and colonialism. European Economic Review, 51(5),

1155-1176.

Atkinson, A. B.; Piketty, T. and Saez, E (2011) ”Top Incomes in the Long Run of History”,

Journal of Economic Literature, 49(3), 3-71.

Atkinson, A. B. (2014). The colonial legacy: Income inequality in former British African

colonies (No. 2014/045). WIDER Working Paper

Baranger (2017), Laccs aux droits des etrangers en Guyane, etat des lieux des parcours, des

reglementations et dispositifs, etudes Perspectives, Coll. Integration et accs aux droits,

Centre de ressources de la Politique de la Ville - Guyane, 138p.

Besson, D. (1997). 25 ans de comptes economiques des departements d’outre-mer 1970-1994

(Vol. 153). Insee.

Blanchet T., J. Fournier, T. Piketty, (2017) Generalized Pareto curves: Theory and applica-

tion using income and inheritance tabulations for France and the US, WID.world Working

Paper.

Blerald, A. P. (1986). Histoire economique de la Guadeloupe et de la Martinique: du XVIIe

siecle a nos jours. Karthala Editions.

Coder, Y., Duplouy, B., Fabre, E. (2016) La departementalisation de La Reunion. 1946-2016:

70 ans de mutations economiques.

34

Drozin G., (2001). Les politiques sociales en matiere d’action sociale a la Reunion depuis

1946 : un survol historique , Communication aux Journees partenariales de l’action sociale,

Institut Regional de Travail Social de la Reunion, Saint-Benoıt, 34 p.

Bertrand Garbinti, Jonathan Goupille-Lebret, Thomas Piketty, (2018) Income inequality in

France, 1900-2014: Evidence from Distributional National Accounts (DINA), Journal of

Public Economics

Jonzo A.-M., (2009) Niveaux de vie 2001-2006: les inegalites saccentuent. Economie de La

Reunion N134: pg610

Maurin and Bernier (2013), Des departements d’outre-mer marques par les difficultes sociales

et les inegalites , Compas Etudes n 9, Octobre

Piketty, T. (2001). Les hauts revenus en France au XXeme siecle. Grasset.

Frankema, E. (2010). The colonial roots of land inequality: geography, factor endowments,

or institutions?. The Economic History Review, 63(2), 418-451.

35

Appendices

Appendix A. Income Tax Data

Fig. 20. Availability of Tax Data

Also include a description of the data precision-when can we observe top 10%

and when top1%

36

Appendix B. Control Total for Population

In order to estimate the distribution of income, there is a need to estimate the total

tax units that should have been observed in the income tax data, had every tax unit been

required to fill a tax form. It should be noted that a person living in France can detach from

his/her parents’ tax unit and thus declare his/her income separately as from the age of 18.

However, it is not mandatory to do so until the age of 21, with the exception of 24 years

old for unmarried students or in a liberal profession. In addition, married people (including

PACSed couples) are required to fill a unique tax declaration. Given this setting, the control

total for population (TUit) is estimated as the number of adult population (Ait) deducting

the number of married couple (Mit) in order to avoid double counting married couple.

TUit = Ait −Mit (1)

These data are obtained from the Population Census in the overseas departments for

the following years of census: 1954, 1961, 1967, 1974, 1982, 1990, 1999, 2009 and 2014. It

is linearly interpolated for the years for which we don’t have these information. The age

threshold at which we define the adult population can be set in different ways, namely at 18,

19 or 20 years old. In this paper, the definition of adult population is taken as the population

above 20 years, as is widely done in this literature, for two main reasons:

i) given that the estimate of control population based on the definition of adult popula-

tion above 20 years seem to provide a good enough approximation of the total number of

tax declarations (See figure 26);

ii) given that the population census reports, which dates back to the 1950s, report pop-

ulation by pre-defined age groups. The age group are typically as follows: 15 - 19 years old,

20 - 24 years old and so on. Hence, a threshold of 18 or 19 years would require further hy-

potheses on the distribution of the population within the age group 15 - 19 years to estimate

the adult population of interest. Hence, to have the most consistent method in estimating

the control population, 20 years old is the threshold taken for defining adult population.

Similarly, the number of married couples is estimated from the population census data

by taking the average number of married individuals divided by 2. While this is not a per-

fect count of the number of joint fillings, it should nevertheless give a more or less precise

estimate of the latter.

37

Fig. 21. Number of taxable tax units in the overseas departments

Figure 21 and 22 show the number of taxable tax units since the 1950s and the number

non-taxable tax units since the mid-1980s respectively. There is a clear upward trend in

both the number of taxable and non-taxable units since the beginning of the period with a

slight downward turn at the end of the period for the number of taxables. Figure 23 shows

the total number of declarations to the tax office and the total estimated tax units (using

equation 1 over the years in La Reunion. We observe a very small number of declarations in

the years prior to 1986. Thereafter, with the requirement for non-taxables to declare their

income, there is a steady rise in the total number of declarations reaching the number of

estimated tax units in the early 2000s. Given this trend, we make the hypothesis that we

start to observe all the tax units in the income tax data as from the year 2003. The total

number of tax units estimated from the definition above is presented in figure 24.

38

Fig. 22. Number of non-taxable tax units in the overseas departments

Fig. 23. Total number of declarations and total estimated tax units in La Reunion

39

Fig. 24. Total taxable tax units in the overseas departments

40

In order to get a better understanding of the evolution of declarations over time, the

proportion of declarations (P) is estimated. P is simply the number of declarations (D)

divided by the total number of tax units (Tot):

Pit =Dit

Totit(2)

The numerator in equation 2 refers to the total number of tax units reported by the tax

authorities for an overseas department i at time t, while the denominator is our estimate

of tax units obtained from equation 1. We observe a general increase in the proportion of

declarations from the mid-20th century until recent years, partly due to the non-declaration

of non-taxable tax units at the beginning of the period. As from the mid-80s, both taxable

and non-taxable tax units are required to declare their income and we observe a steady

increase in the proportion of declaration from the mid-1980s until the early 2000s and a

stabilisation thereafter. This suggest that as from the beginning of the 21st century, we

observe more or less everyone in the tax data. In effect, we should be observing a proportion

of declaration of 100% in the recent years.

Fig. 25. Proportion of tax declaration for the overseas departments

41

Tax Units subject to Declaration(Different definitions of adult population)

Fig. 26. Tax units subject to tax declarations the overseas departments from 1986 to 2014

Figure 26 shows a the trend in the proportion of declaration from the year 1986 in all

four overseas departments. The following are estimated based on the three alternative adult

population: above 18 years old, above 19 years old and above 20 years. Note that the

estimates before 1990 with the alternative definition of above 18 and 19 years old are not

presented here as censuses before 1990 do not provide the population by age but rather age

groups (e.g 15 to 19 years old) as explained above and would require further hypotheses to

estimate the population of interest. We notice that irrespective of the definition used, there

seems to be a stabilisation in the proportion of declaration as from the early 2000s. During

that period, approximately 100% proportion of declaration, depending on the definition

chosen, is reached in La Reunion, Martinique and Guadeloupe, while Guyane reaches a

maximum of 90% during that period.

La Reunion, Guadeloupe and Martinique depict more or less similar situations as far

42

as the proportion of income tax declarations are concerned. However, the case of Guyane

seems to be a very peculiar one, as seen in figure 26. There is a steady increase in the

proportion of declaration reaching around 90% at its peak at the beginning of the 2000s and

there seems to be a slight decline thereafter. We argue that we never reach 100% declaration

in Guyane due to the nature of the data used in the construction of the control population

(the denominator of equation 2). Population census normally documents the population

living in the territory at the time of the census, without differentiating between legal and

illegal residents. Guyane has had a long history of illegal migration, mostly from poorer

neighbouring countries. However, since there are no estimate of the share of the population

within the French Guyanese territory that is illegal, there is a need to make some hypotheses.

As seen in figure 9, 30% on the population in 2014 are foreigners, mostly from Suriname,

Brazil and Haiti. Only a minority of asylum seekers are granted this status, accounting for

2.3% of applicants in 2009 (Baranger, 2017). Hence, the non-negligible share of the illegal

population in Guyane is part of the reason for the overestimation of the number of tax units

as that population is counted in the population census but do not declare their income to the

tax authorities. Moreover, the increasing and stabilising trend observed in figure 26, similar

to the other departments suggests that there is an increase in the number of tax units de-

tected by the tax office reaching almost full declaration as from the early 2000s. This could

mean that the remaining 10% that we do not observe in the recent period in Guyane are

either illegal immigrants20 (captured in the census) or population living in remote areas of

Guyane (not captured by the tax office). While we cannot entirely discard the latter, the

former seems to be a more important share in the Guyane context.

Hence, the evolution in the proportion of declarations over time tends to confirm our

hypotheses that:

i) The definition of adult population with a threshold of 20 years of age does a fair job in

estimating the total number of tax units

ii) We observe all the tax units as from 2003

Given these estimates of the control population, we then need to estimate the associated

control income. The step-by-step methodology employed to estimate this control income is

laid down in the next section.

20The Interior Ministry estimates the number of people in irregular situation to be between 30000 to60000 persons. ”Les etrangers en France”, Rapport du Comite interministeriel de controle de l’immigration,April 2014.

43

Appendix C. Control Total for Income

To estimate the share of income that accrues to the top groups, there is a need to es-

timate the total income that would have been declared had all the tax units been required

to declare their income. In other words, there is a need to estimate the income accruing to

the tax units who did not declare their income and hence who are not counted in the tax

data. As explained in Section 3.2, there exists different methods used in the literature to

construct a control total for income. In this paper, a national income approach is adopted.

This implies that the total taxable income is estimated by deducting all non-taxable income

and irrelevant factors (such as depreciation) from the national income or GDP of the territory.

The estimates of GDP for the overseas departments are obtained from INSEE publica-

tions. More specifically, GDP of La Reunion is obtained from INSEE-La Reunion for the

period 1950 - 2014, while these estimates are obtained from publications (See Besson (1997)

and INSEE website) for the other overseas departments, covering the period 1970 - 2014. In

order to have an uninterrupted series from the 1950s to 2014 for Guadeloupe, Martinique and

Guyane, a relationship between the different GDP per adult population is observed during

the period 1970 (1975 in the case of Guyane) to 1990 as shown in figure 27.

Fig. 27. GDP per adult

44

As far as the former two departments are concerned, it seems reasonable to assume that

the GDP per adult population has been the same as La Reunion for the whole period. This

assumption does not hold for Guyane, we assume a constant ratio21 between the GDP per

adult for La Reunion and that of Guyane throughout the period of 1954-1975. However,

in order to estimate the non-taxable income as explained above, we also need a detailed

breakdown of the national accounts. Since these are not available at the level of the overseas

departments, we will rely on the taxable income series observed at the French national level

to estimate its equivalent in the overseas departments.

Estimation of Taxable Income

First we establish the ratio R between average taxable income per tax unit in France,

T (tu) and GDP per adult population at the national level, GDP (a) for the period 1950 -

2014 as follows (where i = France):

Ri =T (tu)iGDP (a)i

(3)

From the previous section, based on the control total for population and the trends in

the proportion of tax units subject to declaration, we make assumption that everyone fills

a tax form as from 2003. As a result, we can also assume that we observe the totality of

the taxable income in the tax data as from the year 2003. Based on this logic, we can

thus observe the ratio between taxable income per tax unit and GDP per adult popula-

tion for the overseas departments for the period 2003- 2014, using equation 4, where i = La

Reunion, Guadeloupe, Martinique and Guyane. These estimations are presented in figure 28.

21An average over the period.

45

Ratio of Taxable Income per Tax Unit to GDP per adult population

Fig. 28. Ratio of Taxable Income per Tax Unit to GDP per adult population in France and

the overseas departments

We can observe in figure 28 that on average there seems to be a approximately parallel

trend between Rfr and the ratio for the overseas department. Note that there is a break in

the series for Guadeloupe in 2007 due to the detachment of the two islands: Saint-Martin and

Saint-Berthelemy. If we look at the pre-2007 and post-2007 trends separately, it is reasonable

to say that they closely relate to the trend in the ratio for France. There are exceptions for

some years in Guyane and Martinique but on average, it seems to fit relatively well. Based

on this scenario, we assume a constant relationship between the two ratios for each overseas

department, estimated as the average of the coefficient αi over the period 2003 - 2014.22

αi =

∑2014t=2003

Rfr,t

Ri,t

n(4)

22Note that for the case of Guadeloupe, we only take into account the period (t) 2003 - 2006 and for thecase of Guyane, the year 2011 is excluded.

46

i = La Reunion, Guadeloupe, Martinique and Guyane; t= 2003 - 2014 for La Reunion,

Martinique and Guyane (excluding 2011) and t=2003-2006 for Guadeloupe and n = number

of years.

Given αi, we can estimate the ratio between average taxable income and GDP per adult

for the period 1950 - 2002 for the overseas department based on the series of France, as

follows:

Ri,t = αi ×Rfr,t (5)

The estimation of this ratio for the entire period is presented in figure 29.

Fig. 29. Taxable Income per tax units to GDP per adult

Having estimated this ratio, an uninterrupted series for total taxable income, and hence

control total for income can be computed for the whole period.The total taxable income and

the average taxable income per tax units for the overseas departments are presented in figure

30.

47

Fig. 30. Total taxable income and Average Taxable income

Estimation of Fiscal Income

As explained in section 3.2, the income reported by the tax authorities are taxable income-

which is fiscal income deducting allowances. As the rules for allowances changes over time,

we would like to look at fiscal income instead. In order to go from taxable income to fiscal

income, various corrections have to be made to the series. The corrections made here follow

the ones in Garbinti et al (2018). For a more detailed explanation of these corrections please

refer to DINA Appendix D.2 of that paper and Piketty (2001). We apply the same correction

factors as used in the series for France. These include an upgrade rate due to previous-year-

tax deductions and other types of deductions such as the lump sum deductions for wage

earners. In 2006, the 20% deductions for additional professional expenses was repealed and

is accounted for in the corrections factors. Similarly, we assume the same aggregate taxable

income to fiscal income ratio as in Garbinti et al (2018).

48

Appendix D. Top Income Estimations

The next step consist in estimating the distribution of taxable and fiscal income in the

overseas departments and the top income shares. This is achieved by using the Generalised

Pareto Interpolation Technique (see Blanchet et al. (2017) for a description of this method)

and the gpinter web interface23. The income distribution are presented in g-percentiles,

divided as follows:

23It can be found on www.wid.world/gpinter

49