Table of contents

I. Population and GDP per capita

II. Country grouping

III. Income distribution data

IV. Life expectancy data
 
 
 

I. Population and GDP per capita

All available data on population and GDP in Maddison (1995) have been used directly or indirectly. For 1950-92, his data was collected from publications by various international organizations (OECD, World Bank, etc..). For previous years, Maddison carried out a huge work of critical assessment of existing data and consistency checks. He also adjusted all data in order to take into account changes of national frontiers. All GDP data are expressed in Geary-Khamis 1990 US Dollars.

For several countries Maddison?s series begin after 1820 (for example, 1870 for Argentina, 1900 for Chile, Colombia, Peru and Venezuela, etc.). For each of these countries and each missing period, it was assumed that both GDP and the population grew at the same rate as in the nearest country in the same region for which data was available. For example, we estimated GDP per capita in Portugal in 1820 using the rate of growth of Spain's GDP per capita between 1820 and 1850 and Portugal?s GDP figure for 1850 (a figure given by Maddison).

II. Country grouping

The annual series of Maddison relate to 56 countries, including some very small ones like Norway. Maddison regrouped all other countries into 4 groups of countries - 15 in Europe, 37 in Latin America, 45 in Asia and 46 in Africa - representing altogether less than 10% of world GDP. Estimates for these groups are available for 1820, 1870, 1900, 1913, 1929 and every year since 1950.

We started with Maddison's definitions of countries and country groups and made some additional aggregation so as to obtain country groups or single countries- representing at least 1 per cent of the world population. Indeed, it was assumed that single countries smaller than this threshold could not have any significant influence on the world distribution of income. In some other cases, the aggregation was done for more obvious reasons. For instance we put all Scandinavian countries together and we merged East Germany, which Madison included in the 15 European countries mentioned above, with West Germany. Likewise, we put together Austria, Hungary and Czechoslovakia because they all belonged to the Habsburg empire from 1820 to 1918  and had a rather similar economic evolution until 1950. We also added Ethiopia, Tanzania, Zaire to the 46 African countries, but we took 2 countries - Algeria and Tunisia - out of this group to set up a North Africa group together with Morocco.

In some cases, this clustering process was convenient for other reasons. When missing data for some countries were systematically replaced by data referring to some specific country, it made sense to consider all these countries as a single group. This is the reason why Bangladesh, Myanmar and Pakistan are considered as a single country group.

At the end of this grouping process we eventually obtained a set of 15 single countries and 18 country groups.

For analytical purposes -i.e. decomposing the evolution of world income inequality - it was necessary to consider broader country groups. We thus defined 6 "classes" or "regions" :  Asia (excluding Japan, Korea and Taiwan) ; Japan Korea and Taiwan; Latin America (excluding Argentina and Chile); Eastern Europe (Russia, Poland , Bulgaria, Greece, Rumania, Yugoslavia and Turkey); Western Europe and "Western offshoots". This last "region" comprises all Western European countries and all countries in the rest of the world where a very large majority or the whole population were of European origin still at the beginning of the 20^th century. These countries are the U.S.A , Canada, Australia, New-Zealand, Argentina and Chile. Considering them together with Western Europe makes very much sense in an economic history context. European labor and European capital were the main factors of growth for these countries between 1820 and 1914. They also all had a very close relationship with Western Europe in trade, finance, technology, etc. In other regions, the only departure from a geographical definition of regions is our setting aside of Japan , Korea and Taiwan. The reason for this choice is that we wanted to evaluate the effect of the huge gap between economic growth in these countries and in the rest of Asia after 1950. No such heterogeneity was present among Africa or Latin America countries.
 
 

III. Income distribution data

It is not too difficult to find estimates of income distribution for the 1950-92 period. The literature on international comparisons of income distribution is abundant, even though definitions are not always consistent. In general all distribution figures in this paper refer to household disposable income per capita, or possibly per adult equivalent, households being weighed by their size. In some cases, however, other definitions are used. This is true mostly for developing countries. The difference with the previous definition does not lead to big discrepancies, however - on this point see Lecaillon et al. (1984). Things are much more difficult for the 1820-1950 period. We indicate in what follows the data source or the approximation made for each group of countries in the data base.

A. Africa

1) Cote d'Ivoire, Ghana and Kenya

Data on income distribution in each country have been combined to estimate the distribution for the whole group. GDP per capita is comparable across countries. Kenya's share amounts to 50 per cent of GDP for the group. Income distribution data for Kenya are drawn from Bingsten (1986) who reconstructed the evolution of the distribution from 1914 to 1976. The distribution in Cote d'Ivoire in 1960 is from Morrisson (1968) and from Deininger and Squire for 1990 - this estimate is included in their so-called 'high quality data set'. Data for Ghana are also from Deininger and Squire. It was assumed that the variation of the distribution during the period 1914-1970 in Ghana and Cote d'Ivoire was the same as that estimated by Bingsten for Kenya during the same peiod. Then the distribution was assumed to have remained the same ever since 1820.

2) Egypt

Estimates are those from Bourguignon and Morrisson (1990) for 1970. The distribution was assumed to have remained the same ever since 1820.

3) Nigeria

The distribution is assumed to be the same as in the Cote d'Ivoire-Ghana-Kenya group prior to 1970. For the 1970-1992 period, the distribution is assumed to be more unequal than in these countries because of the incidence of the oil sector. The distribution data gathered by Deininger and Squire show less inequality, but they are debatable. In particular, they point to unexplainable differences between 1986 and 1993. Moreover, they are inconsistent with estimates by Canagarajan et al. (1997), which are consistent with our own estimates.

4) North Africa

Two data points are available for the period 1970-1992 in Bourguignon and Morrisson (1998). Estimates for 1820-1960 are based on data about average income ratios between Europeans and Moslems (1950-1960) and the percentage of Europeans in the population before 1950. These figures are avilable in Algeria since 1856, in Tunisia since 1881 and in Morocco since 1921.

5) South Africa

As South-Africa in 1950 resembled very much Kenya, as studied by Bingsten (1986), the distribution was assumed to be the same. However, it is assumed that had remained constant since 1820 instead of becoming more unequal as with Bingsten's data for Kenya. According to Lachman and Bercuson (1992), inequality has slightly decreased after 1960. Such a trend was introduced in the data. The distribution obtained for the terminal year, 1992, fits well the figures given by Deininger and Squire for 1993.

6) 46 African countries

Available within country distribution for the 15 most populated countries were combined with information on population and GDP per capita in 1970 to obtain a single estimate for the whole 1910-1992 period. Prior to 1910, the distribution was assumed to be less unequal due to less inequality both between and within countries before the colonization and the development of a modern sector.

B. Asia

1) China

The distribution was estimated for 1890 and was assumed to have stayed constant from 1820 to 1910. For 1890, the distribution of total income and active individuals across sectors of activity is available. In addition, Chang (1962) gives detailed information on the income of the gentry -2 per cent of the population - which receives income from various sectors.

In 1929, detailed data - originating from the Cambridge Economic History of China, Feuerwerker (1968), Maddison (1998) - on the distribution of income and the labor force by sectors and on the size distribution of land permit computing the share of the first quintile - landless agricultural workers - and the top decile - people employed in the modern sector and large landowners. The evolution between 1950 and 1992 is based mainly on Rawski (1982) and on Khan et al. (1992). The increase of inequality between 1980 and 1992 is taken from Ying (1995) and Chai and Chai (1994).

2) India

We assumed no change between 1820 and 1929. The Cambridge Economic History of India (CEHI) gives the distribution of national income and labor force in agriculture and traditional and modern non-agricultural sectors in 1875 and 1895. Maddison (1989) gives the population and income shares on English expatriates. In order to obtain an estimate of the share of the top decile out of these informations , it was further assumed that land was unequally distributed - with 40 per cent being owned by the top decile - an assumption that is consistent with various evidence found in the CEHI about provincial land distribution. For 1950-70, the estimates of Bardhan (1973) are used. They are based on a comparison of all available estimates for this period. Bardhan's analysis was supplemented by Bhalla and Vashistha (1988). The slight decrease of inequality observed by Datt (1995) - quoted by Deininger and Squire - after 1970 has been taken into account.

3) Indonesia

The estimates for 1820-1929 are partly based on Maddison (1989) who supplies the population and income shares of Europeans and foreign Asians from 1820 to 1929. This provides a reasonable estimate of the income share of the top vintile. The rest of the distribution is based on the assumption that the distribution in Indonesia excluding the Dutch expatriates was the same as that in India excluding the British expatriates. The data on the distribution of consumption expenditures quoted by World Development Indicators 1998 and Deininger and Squire, are used for the 1950-92 period, after a rough adjustment in order to take into account the difference between expenditure and income distribution

4) Bangladesh, Myanmar, Pakistan

It is assumed that these three countries had the same distribution in 1820-1992 as India between 1820 and 1929. The inequality within the group is larger than inequality in each country as a consequence of an increasing gap between Pakistan and Bangladesh in terms of average income - the proportion was 1.2 in 1950-60 and it rose to 2 in 1970-92. The weight of Burma in the group is small - 42 millions inhabitants in 1992 vs. 230 millions in Pakistan and Bangladesh. Also, the distribution in each country is more unequal than the estimate given by Deininger and Squire who consider expenditure rather than income distribution.

5) Philippines, Thailand

The estimates for 1960-92 are from Bourguignon and Morrisson (1998), which are themselves consistent with the estimates of Chen et al (1995) for 1985. The distribution within country is assumed to have been constant prior to 1960. The inequality within the group is slightly larger than in each country because of differences in average income. From 1900 to 1970 , the Philippines/Thailand average income ratio fluctuates between 1.1 and 1.95 from 1900 to 1970 and then falls to 0.48 in 1992.

6) 45 Asian countries

The 17 countries with more than 2 millions inhabitants in 1970 were selected. Among them, the share of the 4 richest amounts to 9 per cent of the population and 28.5 per cent of income, whereas the corresponding figures for the 4 poorest were respectively 32 per cent and 10 per cent. If inequality within countries, which is important in the Middle East, is taken into account, the overall inequality within the group reaches a high level. The distribution in Lebanon and Iran in Lecaillon (1984) was used as a reference. As the distributions within countries remained very unequal since the beginning of the XIX century, it was assumed that the distribution remained constant throughout the whole 1820-1992 period.

C) Japan , Korea, Taiwan

1) Japan

The estimates for 1820-1992 are based Takahashi (1959), Okhawa (1968) , Wada (1975). The distribution was assumed to be unchanged between 1970 and 1992.

2) Korea and Taiwan.

Two elements must be taken into account : population and GDP in Korea exceed the double of Taiwan's ; the average income gap between the two countries is low : it fluctuated between 1.16 and 1.37 in favor of Taiwan between 1970 and 1992. In both countries, the distribution was assumed to be the same as in China for the 1820-1910 period. Chang ( 1983) shows that inequality has slightly decreased in 1929, as a result of the Japanese occupation and colonization which had a positive impact on farmers. The estimates for 1950-92 are based on the data of Deininger and Squire for Taiwan, and of Choo (1978) for Korea, after some adjustments. The changing average income gap between the two countries was taken into account. Also, a correction was made for differences in income unit. Deininger and Squire report data on persons rather than households. This justified a small increase in the share of the top decile in comparison with their estimates.

D) Latin America

1) Brazil

Estimates for 1950-92 are based on estimates in Deininger and Squire, Maddison (1992) and Altimir (1996). The distribution is assumed to have remained the same between 1820 and 1950. This is justified by the absence of strong evidence suggesting the distribution was much less unequal during the 19th century.

2)  Mexico

Estimates for 1950-92 are based on the data of Deininger and Squire , Maddison (1992) and Altimir (1996), which are mutually consistent. Regarding 1820-1950 , we made the same assumptions than for Brazil for the same reasons. The resulting distribution data are consistent with Malthus' suggestion in 1820 that inequality was much larger in Mexico than in England.

3)  Colombia, Venezuela, Peru

Taking into account differences in average income among the three countries - Colombia/Venezuela and Peru/Venezuela income ratios fluctuate between 0.28 and 0.55 from 1929 to 1992 - and Bourguignon and Morrisson (1998) estimates for each country led to distribution data for 1950-92 similar to Brazil's. It was finally assumed that both distributions were the same throughout the 1820-1992 period.

4)  37 Latin American countries

The 15 countries with more than 1 million inhabitants in 1970 were selected. Among them, the share of the 4 richest amounts to 13.6 per cent of the population and 27 per cent of income, whereas the corresponding figures for the 4 poorest were respectively 28 and 13 per cent. Combining that inequality with the very high inequality observed in several of these countries - in several countries the share of the last quintile amounts to 60 or 70 per cent (see Bourguignon and Morrisson, 1998) - led to the estimates in the data base. Without further information, the distribution is assumed to have remained over the whole 1820-1992 period.

E) Eastern Europe

1) Bulgaria, Rumania, Greece, Yugoslavia.

The 1950-92 estimates are based on data in Morrisson (1984), excluding Greece, and on estimates for Russia, where inequality is slightly higher. Differences of average income between countries have been taken into account. The data for 1820-1929 replicate estimates for Russia in 1820-70.

2) Poland

Estimates for 1950-1992 are from Morrisson (1984). As this country shares many features with Russia, they replicate distribution data in that country. The data for 1820-1929 are based on estimates for Russia in 1820-70 - part of Poland actually belonged to Russia until 1914 and the economic structures were rather similar.

3) Russia

There are data on the total income received by the top centile of the population in 1905-10. They are drawn from studies undertaken by the Ministry of Finance but adjustments are necessary because incomes of the church and of the royal family are omitted and because land rents are underestimated. The Cambridge Economic History of Europe (1965 and 1978) gives the distribution of national income between agricultural and non-agricultural sectors. The share of the top decile in each sector may be estimated from various sources: rural census in 1905 and data on wages in the non-agricultural sector. Data for 1890 and 1910 are based on all that information. The large increase of property income after 1880, evidenced in CEHE, was taken into account to extrapolate the preceding estimates to 1820-1870. 1929-92 estimates are from Morrisson (1984)

4) Turkey.

Deininger and Squire estimates for 1968,1973 and 1987 are rather different. An average of these data was used for the whole 1950-1992 period. This average is consistent with Hansen's (1991) estimate for 1978. Hansen gives other estimates for this period which are close to 1978 figures. As no information on the 19th century is available - Hansen thinks it is impossible to make even a rough guess about distributional changes then owing to the fall of Ottoman empire - the distribution was assumed to have remained the same throughout the 1820- 1992 period.

F) European countries and their offshoots in America and the Pacific

1. Argentina, Chile.

1950-92 estimates are based on Altimir (1996) and Bourguignon and Morrisson (1998). Both sources give very close estimates, with the same inequality increase between 1970 and 1980-92. Estimates for 1820-1929 replicate the estimated 1950 distribution. Inequality is substantially less than in Brazil throughout the 1820-1950 period. This seems reasonable in view of the lesser ethnic heterogeneity in Argentina and Chile than in Brazil and in other Latin America countries where the relative income of Indians and slave descendants was very low.
 

2. Australia, Canada, New-Zealand.

Most recent estimates are based on data of Gottschalk and Smeeding (2000) and OECD (1995). These studies compare distributions in these three countries with the distribution in the UK and the US. It was assumed that the difference between the distribution in these countries and that in the US was constant ever since 1820, thus reflecting constant structural differences.
 

3. Austria, Czechoslovakia, Hungary

1820-1929 figures replicate estimates for Germany because economic structures are rather similar. For 1950-1992, the difference between Austria and the two other countries, where inequality is lower (Morrisson 1984) was taken into account. This resulted in a less unequal distribution than in Germany. Changes over time are linked to the increasing gap of average incomes between Austria and the two socialist countries.
 

4. France

Estimates are from Morrisson (2000).
 

5. Germany

Estimates are from Morrisson (2000) and Kraus (1981).
 

6. Italy

1950-92 estimates replicate the distribution in France because Gottschalk and Smeeding (2000) report nearly the same figures in 1990 for the two countries. The distribution for 1820-1929 is constant and is close to the distribution in France, too.
 

7. Scandinavian countries.

Estimates are from Morrisson (2000)
 

8. Spain, Portugal

1950-92 estimates replicate the distribution in France because Gottschalk and Smeeding (2000) report nearly the same figures in 1990 for the two countries. The distribution for 1820-1929 is constant and corresponds to the average distribution in France during that period.
 

9. Switzerland, Benelux


The estimates for 1950-92 are based on Morrisson (2000) for the Netherlands, and on OECD (1995) for Switzerland, Netherlands and Belgium, and on Gottschalk and Smeeding (2000) who find nearly the same distribution for Belgium and the Netherlands. The 1820-1929 distribution is based on a historical study of the Netherlands (Morrisson 2000)
 

10. UK, Ireland

Estimates are from Lindert (2000). Kraus (1981) is also used for 1950-1970.
 

11. US

Estimates are from Lindert (2000) and OECD (1995).

F) References and sources for income distribution data
 

A. 1950-1992

All data for this period are from United Nations Demographic Yearbooks and the World Population Data Sheet (Population Reference Bureau).

B. 1820-1929

Original estimates were found for several countries at different points of time. Sources are listed below. For other countries, values had to be imputed on the basis of international analogy considerations. The corresponding assumptions are also listed below.

1) Direct sources by country or group of countries

- European countries and offshoots

* J-P Bardet and J. Dupaquier (1998), Histoire des populations de l'Europe, II, 1998

* Historical statistics of the US, part 1. Bureau of the Census (gives estimates for Massachusetts since 1850)

- Eastern Europe

* J-P Bardet and J. Dupaquier (1998), Histoire des populations de l'Europe, II, 1998

* Statistical Yearbook of the League of Nations 1938-39, League of Nations, 1939

* A. Blum and I. Troitskaja (1996), La Mortalité en Russie aux XVIIIe et XIXe siècles, Population - North Africa and Egypt

* Statistical Yearbook of the League of Nations 1938-39, League of Nations, 1939

* P. Fargues (1986), Un siècle de transition démographique en Afrique Méditerranéenne, 1885-1985, Population

* K. Kateb (1998), L'espérance de vie à la naissance et la surmortalité féminine en Algérie en 1954, Population (Data on Algeria are only available since 1880. A limited increase in life expectancy was assumed between 1820 and 1880).

- Japan

* Health and welfare Statistics in Japan (1997)

( Data are available since 1890. Same figure 1820 as in countries with similar life expectancy in 1890).

- South Africa

* Official Yearbook of the Republic of South Africa, 1975

* Statistical Yearbook of the League of Nations 1938-39, League of Nations, 1939

(Data are available only since 1910. It was assumed that the evolution between 1820 and 1910 was the same as in countries with similar life expectancy in 1910. )

- Thailand and Philippines

* Thailand Public Health Statistics, Ministry of Public Health, 1972

(Data are available only since 1929. It was assumed that the evolution between 1820 and 1929 was the same as in countries with similar life expectancy in 1929.)

- India

* Statistical Yearbook of the League of Nations 1938-39, League of Nations, 1939

(Life expectancy was assumed to be constant between 1820-1910 because the available estimate for 1910 - 23 years - seemed already very low.)

- Brazil

* Statistical Abstract of Brazil, 1946.

(This source gives an estimate for 1940. It was assumed that no change had taken place since 1929. The evolution since 1820 was then supposed to be the same as countries with comparable life expectancy in 1929.)

2) Assumptions for countries with no data

All the following assumptions are based on analogies observed in 1950.

- China

Constant difference with India between 1820 and 1950.

- Korea Taiwan

Same life expectancy as in Japan for 1820 and then constant difference with Japan between 1890 and 1950

- Colombia, Peru, Venezuela, Mexico, 37 Latin American countries

Same as in Brazil

- 45 Asian countries, 46 African countries, Côte d'Ivoire, Ghana, Kenya, Nigeria

Same as China

- Indonesia, Pakistan, Myanmar, Bangladesh

Same as India

Data