Measuring Income Inequality (Deininger and Squire) Database 1890-1996
This file contains data on Gini coefficients, cumulative quintile shares, explanations regarding the basis on which the Gini coefficient was computed, and the source of the information. There are two data-sets, one containing the "high quality" sample and the other one including all the information (of lower quality) that had been collected.
The database was constructed for the production of the following paper:
Deininger, Klaus and Lyn Squire, "A New Data Set Measuring Income Inequality", The World Bank Economic Review, 10(3): 565-91, 1996.
This article presents a new data set on inequality in the distribution of income. The authors explain the criteria they applied in selecting data on Gini coefficients and on individual quintile groups’ income shares. Comparison of the new data set with existing compilations reveals that the data assembled here represent an improvement in quality and a significant expansion in coverage, although differences in the definition of the underlying data might still affect intertemporal and international comparability. Based on this new data set, the authors do not find a systematic link between growth and changes in aggregate inequality. They do find a strong positive relationship between growth and reduction of poverty.
Kind of data
Aggregate data [agg]
The data included in the data base are a slightly updated version of the full version of the data-base described in Deininger and Squire (1996). The “high quality” data-set described in this paper can be obtained by utilizing only the data marked with “accept” in the quality column.
Compared to the earlier version we have added a number of African countries for which additional data have recently become available. As a consequence, the number of countries and the descriptive statistics will be slightly different from those reported in the paper.
In what follows, we provide brief descriptions of main features for individual countries that are included in the data-base. Without being comprehensive, these notes are intended to indicate some of the considerations underlying our decision to include or exclude certain observations.
Various permanent household surveys, all covering urban centers only, have been regularly conducted since 1972 and are quoted in a wide variety of sources and years, e.g., for 1980 (World Bank 1992), 1985 (Altimir 1994), and 1989 (World Bank 1992). Estimates for 1963, 1965, 1969/70, 1970/71, 1974, 1975, 1980, and 1981 (Altimir 1987) are based only on Greater Buenos Aires. Estimates for 1961, 1963, 1970 (Jain 1975) and for 1970 (van Ginneken 1984) have only limited geographic coverage and do not satisfy our minimum criteria.
Despite the many urban surveys, there are no income distribution data that are representative of the population as a whole. References to national income distribution for the years 1953, 1959, and 1961(CEPAL 1968 in Altimir 1986 ) are based on extrapolation from national accounts and have therefore not been included. Data for 1953 and 1961 from Weisskoff (1970) , from Lecaillon (1984) , and from Cromwell (1977) are also excluded.
Household surveys, the result of which is reported in the statistical yearbook, have been conducted in 1968/9, 1975/6, 1978/9, 1981, 1985, 1986, 1989, and 1990.
Data for 1962 (Cromwell, 1977) and 1966/67 (Sawyer 1976) were excluded as they covered only tax payers. Jain's data for 1970 was excluded because it covered income recipients only. Data from Podder (1972) for 1967/68, from Jain (1975) for the same year, from UN (1985) for 78/79, from Sunders and Hobbes (1993) for 1986 and for 1989 were excluded given the availability of the primary sources. Data from Bishop (1991) for 1981/82, from Buhman (1988) for 1981/82, from Kakwani (1986) for 1975/76, and from Sunders and Hobbes (1993) for 1986 were utilized to test for the effect of different definitions. The values for 1967 used by Persson and Tabellini and Alesina and Rodrik (based on Paukert and Jain) are close to the ones reported in the Statistical Yearbook for 1969.
In addition to data referring to the employed population (Guger 1989), national household surveys for 1987 and 1991 are included in the LIS data base. As these data do not include income from self-employment, we do not report them in our high quality data-set.
Data for Ginis and shares are available for 1973, 1977, 1979, 1986, 1988, 1989, 1991, 1992, and 1993 in government reports on population censuses and household budget surveys, and for 1973 and 1975 from UN (1981). Estimates for 1970 (Jain 1975), 1973, 1975, 1977, and 1979 (Fields 1989) have been excluded given the availability of primary sources.
Data from household surveys for 1973/74, 1976/77, 1977/78, 1981/82, and 1985/86 are available from the Statistical Yearbook, complemented by household-survey based information from Chen (1995) and the World Development Report. Household surveys with rural coverage for 1959, 1960, 1963/64, 1965, 1966/67 and 1968/69, and with urban coverage for 1963/64, 1965, 1966/67, and 1968/69 are also available from the Statistical yearbook. Data for 1963/64 ,1964 and 1966/67, (Jain 1975) are not included due to limited geographic coverage, We also excluded secondary sources for 1973/74, 1976/77, 1981/82 (Fields 1989), 1977 (UN 1981), 1983 (Milanovic 1994), and 1985/86 due to availability of the primary source.
National household surveys have been conducted in 1951/52 and 1978/79 (Downs, 1988). Estimates based on personal tax returns, reported consistently for 1951-1981 (Holder and Prescott, 1989), had to be excluded as they exclude the non-wage earning population. Jain's figure (used by Alesina and Rodrik) is based on the same source.
Household surveys with national coverage are available for 1978/79 (UN 1985), and for 1985, 1988, and 1992 (LIS 1995). Earlier data for 1969, 1973, 1975, 1976 and 1977 (UN 1981) refer to taxable households only and are not included.
The only survey with national coverage is the 1990 LSMS (World Development Report). Surveys for 1986 and 1989 cover the main cities only (Psacharopoulos et al. 1992) and are therefore not included. Data for 1968 (Cromwell 1977) do not refer to a clear definition and is therefore excluded.
The only survey with national coverage was conducted in 1985-1986 (Chen et al 1993); surveys in 74/75 and 85/86 included rural areas only (UN 1981). We excluded Gini estimates for 1971/72 that refer to the economically active population only (Jain 1975), as well as 1974/75 and 1985/86 (Valentine 1993) due to lack of national coverage or consistency in definition.
Data from 1960, 1970, 1974/75, 1976, 1977, 1978, 1980, 1982, 1983, 1985, 1987 and 1989 are available from the statistical yearbook, in addition to data for 1978 (Fields 1987) and for 1979 (Psacharopoulos et al. 1992). Other sources have been excluded as they were either not of national coverage, based on wage earners only, or because a more consistent source was available.
Data from household surveys are available for 1963-69 (in two year intervals), for 1970-90 (on an annual basis) from the Statistical yearbook and for 1991 - 93 from household surveys by the World Bank (Milanovic and Ying).
A priority survey has been undertaken in 1995.
Central African Republic:
Except for a household survey conducted in 1992, no information was available.
The only data are from a 1983/4 household budget survey (World Bank Poverty Assessment).
Gini- and share data for the 1950-61 (in irregular intervals), 1961-81 (biennially), and 1981-91 (annually) are available from official sources (Statistical Yearbook for years before 1971 and Income Distributions by Size in Canada for years since 1973, various issues). All other references seem to be based on these primary sources.
An estimate for 1958 is available in the literature, and used by Alesina and Rodrik and Persson and Tabellini but was not included due to lack of primary sources.
The first nation-wide survey that included not only employment income was carried out in 1968 (UN 1981). This is complemented by household survey-based data for 1971 (Fields 1989), 1989, and 1994. Other data that refer either only to part of the population or -as in the case of a long series available from World Bank country operations- are not clearly based on primary sources, are excluded.
Annual household surveys from 1980 to 1992, conducted separately in rural and urban areas, were consolidated by Ying (1995), based on the statistical yearbook. Data from other secondary sources are excluded due to limited geographic and population coverage and data from Chen et al (1993) for 1985 and 1990 have not been included, to maintain consistency of sources..
The first household survey with national coverage was conducted in 1970 (DANE 1970). In addition, there are data for 1971, 1972, 1974 CEPAL (1986), and for 1978, 1988/89, and 1991 (World Bank Poverty Assessment 1992 and Chen et al. 1995). Data referring to years before 1970 -including the 1964 estimate used in Persson and Tabellini were excluded, as were estimates for the wage earning population only.
Data on Gini coefficients and quintile shares are available for 1961, 1971 (Cespedes 1973),1977 (OPNPE 1982), 1979 (Fields 1989), 1981 (Chen et al 1993), 1983 (Bourguignon and Morrison 1989), 1986 (Sauma-Fiatt 1990), and 1989 (Chen et al 1993). Gini coefficients for 1971 (Gonzalez-Vega and Cespedes in Rottenberg 1993), 1973 and 1985 (Bourguignon and Morrison 1989) cover urban areas only and were excluded.
Data based on national-level household surveys (LSMS) are available for 1985, 1986, 1987, 1988, and 1995. Information for the 1970s (Schneider 1991) is based on national accounting information and therefore excluded
Official information on income distribution is limited. Data from secondary sources are available for 1953, 1962, 1973, and 1978, relying on personal wage income, i.e. excluding the population that is not economically active (Brundenius 1984).
Household surveys for 1993 and 1994 were obtained from Milanovic and Ying. While it is in principle possible to go back further, splitting national level surveys for the former Czechoslovakia into their independent parts, we decided not to do so as the same argument could be used to justify introduction of distributional data from states within countries such as the US, Brazil, or provinces in China, an issue that would require a separate effort. Information on 1989, 1990, 1991 and 1992 (Cornia 1994) was therefore excluded.
Household data are available for 1958, 1965, 1970, 1973, 1976, 1980, 1985, and 1988 (Atkinson and Micklewright 1993), 1991 and 1992 (Milanovic and Ying) .
Data for and 1981 and 1987 are available from the statistical yearbook, complemented by information for 1976 (ILO 1984). Data on income share are also available for these years. Household expenditure surveys in 1955, 1963, 1966, and 1971 were limited to wage earners' taxable income (UN 1981) and have therefore not been included.
Data are available from a priority survey undertaken in 1996.
National household surveys are available for 1976 (UN 1981), 1984 and 1989, and 1992 (IDB 1994, Chen et al. 1995, and World Bank Poverty Assessment). Earlier surveys covered urban areas only and are thus excluded.
The only survey with national coverage is the 1993 LSMS (World Bank Poverty Assessment). Data with urban coverage only are available for the 1967/68 - 1975/76 period and for 1986-92. A rural income surveys conducted in 1965 is reported by Jain.
Consumer (or family) budget surveys have been conducted in 1958/59, 1964/65, and 1974/75 (Levy 1986; Hansen and Radwan 1982) for urban and rural areas. Data from a survey in 1991 is available from the World Development Report.
El Salvador :
The only household survey with national coverage was undertaken in 1976/77 (UN 1985), followed by a household survey in 1990, the coverage of which was limited to San Salvador (Psacharopoulos et al. 1992). Other estimates available in the literature are thus excluded.
A nationally representative survey was carried out in 1996. Other surveys do not have national coverage although rural surveys such as the one undertaken in 1981/82 covered most (80%) of the population (Chen et al 1993).
Household surveys were undertaken in 1968 and 1972 (UN 1981) and 1977 (Fields 1989).
Income distribution statistics are available in the Statistical yearbooks for the years of 1971 (LIS 1995), 1978, 1979 1980, 1982, 1983, 1984, complemented by information for 1987 and 1991 (LIS data base). The information used by Alesina and Rodrik and Persson and Tabellini, with Ginis more than 15 points higher than the ones with national coverage, seems to be based on the distribution of earnings only.
Household surveys for 1956 and 1962, 1965, 1970, and 1975 are available from UN (1981). Data for more recent years (1975 and 1984) are covered by the LIS data base.
Estimations based on national household surveys in 1975 and 1977 are reported by Kervyn (1980). Information for earlier years appears to be based on the economically active population and therefore not included.
A priority survey from 1992 is available.
Reliable data are available from household income and expenditure surveys for 1962/63 and 1969 (UN 1985) and for 1973, 1978, 1981, 1983, and 1984 (LIS data base). Data for the same years from other sources is excluded to maintain consistency.
Gini coefficients are available from World Bank sources from 1988-92.
There were three national level household surveys (in 1974, 1981/82 and 1987/88), results for which are published in the statistical yearbook as shares. Surveys in 1957/8 and 1962/63 covered only the urban areas (UN 1981) and had thus to be discarded. Similarly, another widely quoted study (Lianos and Kyprianos 1974) relies only on the distribution of taxable family income, leaving out a large number of families who fall below the threshold level.
Information from national household surveys in 1979 (UN 1981) as well as 1987 and 1989 (Chen et al 1993) is based on primary sources. An estimate for 1947/48 (Adler et a. 1952) is based on the combination of, and extrapolating from, two surveys, one (n = 222) covering indigenous households and another one that was conducted in Guatemala city (n=179) to construct a cost of living index. Although ingenious, it does not satisfy our criteria and has therefore been excluded.
There has been a nationally representative survey in 1995.
Data from national household surveys are available for 1967/68 (Jain 1975), 1989 (Chen et al 1993), 1990 (CEPAL 1993), 1992 and 1993 (World Bank 1994) in Gini and in forms of percentage shares. A household survey in 1986 covered urban areas only and was therefore excluded (World Bank 1992).
Data on income distribution are available from household expenditure surveys in 1957, 1963/64, 1973, and 1979 as well as census data from 1971, 1976, 1981, 1986 and 1991. However, coverage of the surveys before 1973 was limited to about 6-% of the population (Lin 1985), leading us to exclude these observations due to lack of national coverage.
Income surveys have been carried out every 5 years from 1962. Gini estimates and data on income shares are available for 1962, 1967, 1972, 1977, 1982, and 1987 (Atkinson and Micklewright 1992). Data for 1991 and 1993 (Chen et al. 1995 and LIS) indicate a considerable increase in inequality although the large shift in 1991 may have been more of temporary character .
We use national Gini coefficients for the years from 1951 to 1991, calculated by Datt (1995) that are more consistent than the other sources available.
Our data include estimates of Gini coefficients and share information for 1976, 1978, 1980, 1984, 1987, 1990 and 1993 (from the Statistical Yearbook and the WDR) as well as estimates for 1964/65, and 1966/7, and 1969/70 from surveys covering all Indonesia except Maluku and West Irian reported in Fields. The 1968/69 HES, conducted by the Central Bureau of Statistics covered only eight cities, while the 1969,1970 integrated Agricultural and Socio-economic Survey and the 1971 Census did not provide data on income distribution.
Gini estimates and data on income shares based on survey information have been found for 1969/70, 1970/71, 1971/72, 1972/73 (Pesaran 1976), and 1984 (Behdad 1989). Data for 1959, 1968 (Jain 1975) cover only a limited population and are thus excluded.
National household surveys were conducted in 1973, 1980 (Murphy 1985), and 1987 (Report of Household Budget Survey 1987).
All household surveys (available for 1948, 1963/64, 1970, 1977, 1979, 1987, and 1992) cover the urban population only and are thus not excluded. Given the low population threshold (about 2000) required for a settlement to be considered “urban” they are often considered to be nationally representative.
Gini coefficients for 1974 through 1977 are presented in UN (1981), based on surveys conducted by the Banca d'Italia. Shares and Ginis from the same source are available annually for 1978-84 , 1986, 1987, 1989 and 1991 (Brandonlini 1994).
The earliest data with national coverage is the 1958 household budget survey analyzed by Ahiram (1964). Additional sources include Boyd (1988) for 1971 and 1975. LSMS were conducted regularly since the late 80s (Chen et al. 1995).
We used information for 1962 - 1982 that is directly based on the survey of people's living conditions (Mizoguchi 1985 and Mizoguchi and Takayama). From 1985 to 1990, information is based on Oshima (1994). Information for years prior to 1962 was excluded due to the lack of nationally representative household surveys in this period.
Household survey based data of income distribution are available for 1980 (Haddad 1990), 1987 (Sha'ban 1990) and 1991 (Chen et al 1993). Information from the 1966 HES was excluded as the survey covered only urban areas.
Results from the 1981/83 and 1992 LSMS (Chen et al 1993) are the only information based on national coverage. There were no nationally representative survey before this, forcing us to discard all observations pre-dating this date, such as 14 data-points for 1914-1976 that were based on extrapolation from tax accounts (Bigsten 1986), as well as observations for 1961 (Cromwell 1977), 1968/69 (Jain 1975), and 1969 (Lecaillion 1984; Jain 1975). A Gini for 1977 derived from a Social Accounting Matrix (van Ginneken 1984) and estimates for 1976 (ILO 1986) have been eliminated due to poor data quality..
Data for 1980, 1982, 1985, 1986, and 1988 are based on a nationally representative household survey (Social Indicators in Korea). All other primary sources refer to the urban and rural populations separately (UN 1985) and have therefore to be consolidated to yield a single national estimate. Such consolidation is available in a number of secondary sources from which we derive Ginis for: 1953, 1961, and 1964 (Lau 1986); 1965, 1970, 1976 (Choo 1985); 1966, 1968, 1969, and 1971 (Jain 1975).
The only observation is for 1986/87 Chen (1993), based on the Statistical Yearbook.
A household-based estimate is available only for 1993. Earlier estimates (Pryor 1982) are of a synthetic nature and were therefore not included.
The Household income and expenditure surveys for urban areas and agricultural estates in 1968 covered only 7% of the population and is thus excluded. Results reported for 1968/69 by Pryor (1980) is from the National Sample Survey of Agriculture, and the Household income and expenditure surveys for urban areas and agricultural estates in 1968 and can not be regarded as nationally representative either. His 1984/85 results are an estimate based on combining the 1979/81 urban and the 1984/85 rural survey. There was, however, a large and nationally representative survey in 1993.
Following the Post-Enumeration Survey (PES 1970) with national coverage, household income surveys covering peninsular Malaysia were conducted in 1976, 1979, 1984, 1987, 1989. A nation-wide Household Budget Survey from 1957/58 is, according to Meesook (1975) fraught with serious problems regarding data quality and coverage of population as well as income and has thus been excluded. A Federal Savings Survey in 1960 included employees only, and a socio-economic sample survey of households in 1967/68 was limited to information regarding cash income (Meesook, 1975).
A nationally representative survey was carried out in 1994.
Mauritania: Data reported are from a nationally representative survey undertaken in 1995.
Data were obtained from UN (1985), World Bank country operations (Khan 1995), and Chen et al. (1995).
Existing sources vary dramatically in the way in which they adjust for underreporting and we therefore chose consistent sources as far as possible.
Household survey-data are available for 1984/85 and 1991 (Chen et al 1995). Earlier data are of doubtful quality (Adelman 1972) or cover wage income only (Bourguignon and Morrisson 1989) and are thus excluded.
Data from national household surveys are available for 1976/77 (Fields 1989) and 1984/84 (Chen et al 1993) and 1989 (World Bank 1990) that have been included although it is widely held that the income data reported are problematic (UN 1981). Another survey in 1973/75 was limited to eighteen urban centers and is therefore not included.
The Statistical Yearbook provides data on income shares by household with nation-wide coverage for 1979, 1981, 1982, 1983, 1985, and 1986, complemented by LIS information for 1987 , 1988, 1989, and 1991. Before 1977, data on size income distribution were based on taxation statistics (Statistical Yearbook: 1957, 1958, 1960, 1962, 1964, 1972). As they cover only wage earners, we have excluded them from our data-base.
The statistical yearbook only reports income distribution results from population surveys in 1973, 1974, 1976, 1977,1979, 1980, 1982, 1983, 1986, 1987, 1989, and 1990. Annual estimates for the 1952-1972 period (Easton 1983) are based on taxable income and have thus beendiscarded. The same deficiency led to the elimination of Jain's (1975) data for 1966, 1967/68, 1968/69. 1969/70, 1970/71, and 1971/72.
Data from national household surveys are available for 1975 and 1980 (Aigbokhan 1988) but were excluded due to doubtful data quality, leaving us with three observations for 1986, 1992 and 1993. Data for years before 1975 have been excluded due to limited population coverage.
A consistent series of data for 1962, 1967, 1973, 1976, 1979, 1984, 1985, 1986, and 1991 is available from the Statistical Yearbook. We complement them with a number of other sources.
Data on size distribution of income by households are available 1969/70, 1970/71, and 1985/86 (Economic Survey Report 1990), 1971/72, 1979, 1986/87, 1987/88 from official reports, and from other sources for 1963/64, 1966/67 (Mahmoud 1984 in Fields 1989), 1971/72 (UN 1981), and 1991/92 (Chen et al. 1995). This includes all the years in which household surveys were conducted. Data for 1963/64, 1963/65, 1966/67, 1968/69, 1969/70, 1970/71 (Jain 1975) were excluded due to limited income and geographic coverage and for 1964, 1969/70, 1971, 1972, 1984 (Mahmoud 1984 in Fields 1989) due to availability of primary sources.
National household surveys were conducted in 1970 (van Ginneken 1984), 1979, and 1989 (Psacharopoulos et al. 1992). We add information on income from the 1980 Census Report. Observations for 1960, 1962, 1968, 1969, 1970, 1972 (Jain 1975), for 1969 (Adelman, 1970), and for 1970 (Lecaillion 1984) are excluded due to limited income and geographic coverage.
All available data are for urban areas only.
We have information from surveys in 1971/72, 1981 (Statistical yearbook), and from LSMSs in 1985/6 and 1994. Estimates for 1961 (Webb, 1977) cover only wage earners while data from 1961/63, 1970/71 (Jain 1975), and 1962 (Adelman 1972 in Cromwell 1977) refer only to part of the population. Data for 1981 (Berry 1989) and 1985 (Webb 1988) have been excluded due to availability of better primary sources.
Data on percentage shares for 1957, 1961, 1965, 1971, 1975, 1985, and 1988 are available from the Statistical Yearbook, complemented by Fields (1989) and Oshima (1994). Given the availability of the primary source, we excluded data reported in the literature for these dates as well as data referring only to the rural or urban population.
In addition to data for 1976 (UN 1981) we have a consistent series for the years 1978-1993 (Milanovic and Ying) which we complement with information from the LIS data base (for 1992) and Atkinson and Micklewright (for 1989).. Data for the years 1956, 1960, 1962, 1964 (Jain 1975), and 1965 (Cromwell, 1977) have not been included due to limited coverage of the population (wage earners only) whereas information for 1985 and 1989 (Chen et al, 1993; Cornia 1994) was excluded to maintain consistency of definitions.
Data for a 1973/74 survey with national coverage are available from UN (1981). They have been complemented by information based on the 1980/81 and 1989/90 Surveys of Family Budgets by the National Statistical Institute (Gouveia and Tavares 1992) and 1991 data for the size distribution of income from the Statistical Yearbook .
The US Bureau of Census reports data for 1969, 1979, and 1989. These differ considerably from estimates based on household surveys conducted in 1953 and 1963 (Mann 1985; Fields 1989) which are excluded as we were unable to obtain information regarding their coverage and sample size.
While the statistical yearbook is of limited use, we obtained information for the early 1990s that indicates some increase in inequality from Chen et al. 1995 and Milanovic and Ying 1996.
There was a nationally representative survey in 1978 (Fields 1989), followed by one in 1984 (Khan 1995).
The only useable piece of information is by Kansal for 1968. Fields (1989) reports an estimate for 1967/9 that seems to be based on a survey with national coverage but with monetary income only. ILO (1982) reports information for 1975/76 without indicating the primary source and could thus not be included.
Information based on analysis of the household expenditure survey data for 1972/73, 1977/78, 1988 (Rao 1976, 1988, 1990), 1980 (UN 1985), 1983 (Fields 1989), and 1989 (Oshima 1994) is included in our data-set. Gini coefficients from other sources were excluded because they are based on the same primary data or because of limited coverage of the population.
A 1992 survey is included both in the LIS database and in the WDR. Pre-independence sources have been excluded for reasons mentioned earlier.
The only data source that met our criteria is the 1993/94 World Bank LSMS. Other data are based on extrapolations from subsets of the population and thus not included. While there have been decennial censuses, as well as a household expenditure survey in 1975, the information is not provided in a form that could be used to estimate income distribution parameters and, in addition, is subject to serious problems concerning nation-wide (racial) coverage. For similar reasons, information for 1959 (Adelman 1972 in Cromwell 1977) and 1965 (ILO 1984; Jain, 1975) has been excluded. Nel and Wyk (1984) and Lachman and Bercuson (1994) attempt to provide estimates of the income distribution for seven years during 1960 -87 that are not based on a household survey and that differ considerably from the information of the nationally representative household survey.
Data from Atkinson and Micklewright have been complemented by information from a World Bank LSMS (Chen et al. 1995).
Results of household surveys (encuestas de presupuestos familiares) is available for 1964/65, 1973/74, 1980, 1986, 1987 , and 1989 from the Statistical Yearbook, and for 1985 and 1988 from separate reports. Among the excluded data are estimates for 64/65 (Jain, 75) that cover only the urban population, and for 64/65 (Jain, 75), 73/74 (Ginneken, 84), and 88 (Milanovic, 94) due to availability of the primary sources.
We have information for 1953, 1963, 1973, 1978/79, 1981/82, and 1987 (Datt 1994) that is based on primary sources (Report on Consumer Finance and Socio-economic Surveys 1953, 1963, 1973, and 1978/79, and for 1980/81 from UN 1981). We also included information from 1970 (Central Bank) and for 1985/86 and 1990/91 from Chen (1995), based on expenditure rather than income. All other data are excluded due to availability of a consistent primary source.
Results of a 1967/68 survey, the only one with national coverage, are reported in UN (1981). We exclude estimates for 1963 (Jain 1975) since they cover only urban population, and for 1969 (Cromwell 1977) due to unclear sources definitions.
The only available observation (Lecaillion 1984) is based on wage earnings only and thus excluded.
We have shares for net household income from a total of 15 years between 1967 and 1992 from either the LIS data base or the Statistical Yearbook. The clearly higher Gini coefficients based on wage income (Spant 1979) were excluded.
As official income distribution statistics are based on samples of wage and salary earners, they are not included in the final data-base.
Data on size distribution of income are available for 1953 (Chang 1953), 1959/60, 1961 (Chang in Kuo 1975), and for 1964, 1966, 1968, 1970, 1972, and 1974-93 based on official sources (DGBAS). Given the limited sample of only about 300 households and the non-official character of the household surveys before 1964 (Kuo 1974), we excluded estimates for the first three years (1953, 1959/60, and 1961) although they may accurately reflect pre-reform income distribution.
There were national level household surveys in 1969, 1976/77, and a LSMS survey in 1993. Other sources cover only part of the population and are thus excluded.
Data based on national household income surveys are available for the years 1962/63, 1968/69 (Ikemoto and Limskul 1986), and 1975/76, 1981, 1986, 1988, 1989, 1990 and 1992 (Ikemoto, 1993).
Trinidad and Tobago
National household surveys were conducted in 1957/8 (Ahiram, 1966), 1971/2 (Household Budget Survey, Trinidad & Tobago), 1975/6 (Dookeran, 1981), and 1981/82 (Household Budget Survey, Trinidad & Tobago).
According to Government Sources (TMPDR), nation-wide household surveys on income and expenditure have been conducted in 1966/68, 1974/75, 1979/80, 1984/85 and 1990/91. Documentation in primary sources -though somewhat less disaggregated than one would wish- is available for all years. To avoid to the high level of aggregation we use a secondary source for 1966/68 (van Ginneken, 1984).
The Statistical Yearbook mentions household surveys with national coverage in 1968, 1973, and 1987 but provides data only for 1987, forcing us to rely on UN (1981) for the other two years. A Gini coefficient based on a consolidation of separated estimates from rural and urban surveys for 1978 is provided by Hansen (1991).
Two household surveys are available: the 1989/90 HBS and the 1992/93 Integrated Household Survey (Chen et al. 1995 and WDR).
Annual surveys based on analysis of primary data are available since 1962/3 - 1991 (Goodman and Webb, 1994). Definitions and measurement are consistent throughout these years. Data from other sources have been excluded to maintain consistency of definition.
While there are long and relatively reliable series on both urban and rural income distribution, lack of data prevented us from combining the two, implying that no data for the country are reported in our final data-base.
We include data published by the US Census Bureau, Current Population Reports, for the 1947 - 1992 period. Other years are excluded.
We include data for 1976, 1977, 1978, 1979 UN 1981), 1981 (World Bank 1992), 1987, 1989 (Chen, et al 1993), and 1990 (Hauseman and Rigobon 1993), all based on primary sources. While a 1962 survey with national coverage is mentioned (UN 1981), no data are provided and the 15-point difference between the two available estimates (Cromwell 1977 and Jain 1975) does not increase confidence in their reliability.
There was a LSMS survey, results of which are published in the WDR.
We had access to a consistent series from 1978 to 1990 (Milanovic and Ying) which we complemented by data on national household surveys for 1963 (UN 1967), 1973 (van Ginneken and Park 1984), 1985, and 1989 (Chen et al 1995). The resulting figures differ considerably from the data on workers' households' incomes reported in the Statistical Yearbook since 1963 (Flakierski 1989, Wiedeman 1984) thus justifying our decision to include only national level surveys.
The first survey with national coverage was conducted in 1974/76 (ILO 1982) and is referred to by van Ginneken and Park (1984). The 1991 and 1993 SDA surveys (Chen et al. 1995), as well as a Living Conditions Measurement Survey in 1996 are other sources with national coverage.
While a 1969 survey was limited to wage earners only, a 1990 survey is available from Chen et al. (1995).
Producers and sponsors
Klaus W. Deininger and Lyn Squire
Dates of collection
Mode of data collection
Use of the dataset must be acknowledged using a citation which would include:
- the Identification of the Primary Investigator
- the title of the survey (including acronym and year of implementation)
- the survey reference number
- the source and date of download
Klaus W. Deininger and Lyn Squire. Measuring Income Inequality (Deininger and Squire) Database (DS) 1890-1996. Ref. WLD_1996_DS_v02_M. Dataset downloaded from http://microdata.worldbank.org on [date].
Disclaimer and copyrights
The user of the data acknowledges that the original collector of the data, the authorized distributor of the data, and the relevant funding agency bear no responsibility for use of the data or for interpretations or inferences based upon such uses.