Methodology

Go to search page

The data series available on MOxLAD draw mostly from the studies and statistical abstracts of the Economic Commission for Latin America, but also rely on Mitchell's International Historical Statistics, International Monetary Fund's International Financial Statistics, and the World Bank's World Development Indicators. The sample of countries covered has been widened to twenty countries in the region and the time series presented have been extended to cover the whole of the twentieth century. The time series published in MOxLAD include new comparable GDP and value-added series in constant purchasing power parity prices, century-wide series for nominal exchange rates, prices indices of manufactures and a wide range of commodities, and new and more reliable estimates of life expectancy. The following discussion highlights some of the methodological problems encountered in the compilation of the data published in MOxLAD.

Reliability, consistency, and comparability

All comparative analyses, particularly when extended over many countries and many years, encounter methodological problems such as the reliability, consistency, and comparability of the data series.

The reliability of statistical data depends to a large extent on the quality of the source material, but also on the quality of data adjustments and methods, which cannot always be vouched for. This is particularly true for source material dating from the earlier part of the century. Even if the reliability of data can be established, changes in the statistical concepts and methods employed in the data collection and dissemination process methods within countries are not always fully accounted for in the source material. This may be due to: 1) intentional data omissions and honest mistakes on the part of the reporting statistical agency or office; 2) differences in the area of responsibility of diverse statistical offices (e.g., national statistics offices, central banks, census bureaus, etc.) and in the period to which the data refer (e.g., fiscal or calendar years); 3) to discrepancies in the way statistical concepts have been defined arising from frequent changes and variations in definition, accounting practices, data adjustment methods, procedures, and classifications, and lastly; 4) differences in data coverage due to the inclusion or exclusion of certain components (e.g., enrolment in public and/or private schools). Consequently, despite best efforts to ensure full disclosure, series may not always be strictly consistent or comparable for the region—let alone for one country—and may result in misleading international comparisons of long-run statistics. With the publication (and subsequent revision) of the United Nations System of National Accounts and the International Monetary Fund Balance of Payments Manual, national and international data collection and dissemination practices have become increasingly standardised. For data preceding the mid-century, regional comparisons may require the supposition that the margin of discrepancy in the data reported by original sources is more or less comparable for each country, making these statistics reasonable, albeit imperfect, indicators.

The researchers involved in the recompilation of the data presented in MOxLAD have made every effort to ensure meaningful intercountry comparison. This has involved the inclusion of data series and the selection of data sources in conformance with a conceptual framework with emphasis on consistency in statistical concepts over time, data coverage for the broadest range of countries and time series, and intercountry comparability. These issues have required that adjustments be made to the original data, particularly the series for GDP and life expectancy.

Gross domestic product

The key source for the data on gross domestic product (GDP) is ECLAC CE (1978), which gives estimates at factor cost using 1970 prices. This source includes all Latin American countries (except Cuba) beginning at different dates according to the country in question, and ending in 1976. Broadly, the next step has been to work backwards from this post WWII foundation to 1900 by using national sources or authoritative compendia of comparative historical statistics for GDP volume. In most cases, estimates for GDP after 1976 are obtained by applying an index (usually with a subsequent base year) for GDP at constant market prices. The “splicing” procedure implicit in joining series in this way means that although the rate of change in the series reported here and in the corresponding sources are almost always the same, the resulting levels of the variables may be different in some cases. This procedure gives greater weight to comparability between countries and change over time than to consistency in levels over time for any one country. This approach appears to be justified for the study of regional trends, but the series should not be seen as a substitute for the original sources where the study of an individual country is concerned.

Purchasing power parities

International income comparisons require the conversion of output levels in different countries into a common unit, usually by employing exchange rate or purchasing power parity (PPP) converters. PPPs are generally preferred over the exchange rate for several reasons: first, because exchange rates tend to underestimate the value of non-traded services in low income countries, where wages are commonly cheaper than in high income countries; second, because exchange rates are often misaligned due to intervention in foreign exchange markets; and third, because capital movement volatility, particularly in recent decades, can affect the reliability of exchange rate converters, see Maddison (2001), p. 162.

The most recent sources of PPP conversion factors for Latin America include Heston and Summers (1991) and subsequent revisions, and the International Comparison Programme (ICP) of the UN and the OECD (these sources are used by Maddison (2001)), but their coverage for Latin America is limited to 7 countries. MOxLAD relies on PPP conversion factors for 1970 reported by ECLAC CE (1978). These PPP rates are based on a common basket of goods reflecting consumption patterns in the region during the 1960s, see United Nations (1968), and were later updated by ECLAC to 1970. The ECLAC PPP rates were chosen to preserve consistency, as MOxLAD uses the ECLAC series for GDP at constant prices, but also to present greater coverage of comparable income data, since the ECLAC PPP estimates include conversion factors for all of Latin America with the exception of Cuba. By merging the 1970 benchmark levels with the GDP time series in 1970 local currency units, MOxLAD also presents level comparisons for every year in terms of this 1970 benchmark. Table 1 below compares the ratio of alternative estimates of real GDP per capita to real US GDP per capita given by different PPP converters; it suggests that ICP3, ICP4, and Summers and Heston estimates of real GDP in PPP prices are generally higher than ECLAC estimates for all countries but Argentina.

Table 1
Ratio of alternative PPP estimates of real GDP per capita in Latin America to real US GDP per capita in 1990

Sources: calculated from ECLAC CE (1978) for col. 1; Maddison (2001), p. 174 for cols. 2 –5, and for real US GDP per capita.

Life expectancy

The estimates of life expectancy were produced by Barbara and Shane Hunt for the Thorp volume with data from the Centro Latinoamericano de Demografía (CELADE) (1996), Arriaga (1968), and Palloni (1990). CELADE's estimates are the official estimates of the respective member countries because they are produced in cooperation with member governments. They are also reproduced by the United Nations and other international agencies. CELADE's estimates refer to a 5-year period, beginning with 1950-55. To get year-specific estimates, the figures of adjacent life tables were averaged; as a result, the earliest year-specific CELADE estimate refers to 1955.

The main source of data for figures for 1950 and earlier years is Arriaga (1968). Arriaga's method inferred mortality rates from the survival of cohorts from one census to the next. Consequently the method was not applicable in cases where immigration was substantial. Thus Arriaga did not attempt estimates for Argentina, Uruguay, and Cuba. For lack of census data, neither did he attempt estimates before 1950 for Haiti and Ecuador. Arriaga's estimates refer to specific years, are generally estimated for each year that begins a new decade, and extend through either 1950 or 1960, which overlaps with the earliest CELADE estimates. Table 2 below shows the correspondence between the two sources for 1960.

Table 2
Life Expectancy, 1960: Comparisons between Arriaga (1968) and CELADE (1996)

(1) Figure is from UN DY (1951).
(2) Figure refers to 1950. CELADE estimate is derived from linear extrapolation of figures for 1955 and 1960.

Table 2 shows that the match is quite good: what discrepancy exists appears to be largely systematic (i.e., the Arriaga estimates are slightly higher). A 2% discrepancy was considered insignificant. In such cases, which cover 10 of the 20 Latin American republics, the Arriaga figure was tacked on to the CELADE series without adjustment. In two other cases (Cuba, Uruguay), Arriaga made no estimates, and in one (Haiti) Arriaga had an estimate for 1950 only. In that case, the estimate diverged significantly from that of a linear projection of CELADE figures for 1955 and 1960, and the latter estimate was used.

There were, however, seven cases of discrepancy. In six of the cases where the Arriaga figures for 1960 exceeded those of CELADE, the discrepancy ranged from 12% for Honduras to 4% for Nicaragua and Venezuela. In these cases, the CELADE figure was considered to be the more reliable for 1960, and Arriaga's earlier estimates were adjusted accordingly. However, it was considered inappropriate to apply a downward adjustment to the earliest Arriaga estimate for each country, since the low values of those early estimates were seen as the most noteworthy result of the statistical exercise. Because this result should not be exaggerated by a downward adjustment coming from discrepancies arising many years later in 1960, the downward adjustment applied to Arriaga's estimates was based on a sliding, linear scale, ranging from the actual discrepancy in 1960 to zero in the earliest year for which Arriaga had an estimate in each given country. This discrepancy in the case of Paraguay is, however, sui generis, both with respect to magnitude and direction, as will be discussed further below.

Alberto Palloni's (see Palloni 1990) more recent estimates for some countries and some time periods rely on vital statistics data, adjusted to allow for incomplete reporting. Palloni's estimates fill in several of the gaps in Arriaga's work. These estimates provide comparisons with those of CELADE (1996) for four different five-year periods. If it is assumed that a discrepancy of 3% or less indicates general agreement between the two sources, then 71% (54 of 76) of the comparisons are in agreement. In general, discrepancies greater than 3% do not persist from one time period to the next in the given country. Thus, they appear accidental and not reflective of an underlying bias in one or the other series. Some of the largest discrepancies are, however, quite persistent. They refer to Bolivia, Brazil, Colombia, and Paraguay, and are set forth in further detail in Table 3 below.

Table 3
Life Expectancy, 1950-1985: Comparisons between Palloni and CELADE for Selected Countries

For Bolivia and Colombia, discrepancies arise only in the early years, but they are substantial. Palloni's estimates show life expectancies that are between two and a half and five years less than those of CELADE. For Brazil, Palloni's figures are consistently lower than those of CELADE, right through to the latest period. Without further knowledge of the different sources and methods used in the estimates, the CELADE estimates were preferred because its figures are official and in better agreement with those of Arriaga, and because it has greater coverage.

These arguments could not be applied to the discrepancy in the case of Paraguay because the CELADE figures, whilst official and complete, do not seem credible. CELADE's life expectancy figure for Paraguay in 1950-55 was 62.6 years, a figure equal to that of Argentina and exceeded in Latin America only by Uruguay. This would indicate a significantly greater length of life and superior health conditions than could be found in countries such as Chile, Costa Rica, and Venezuela, all of which had significantly longer life expectancy than Paraguay in later years. In consequence, Palloni's figures for Paraguay were used from 1950-55 through 1980-85; CELADE estimates were used only for interpolation and for extrapolation since 1980-85.

For the period before 1950, two alternative sources are available: Arriaga (1968) and Palloni (1990). The advantage of Palloni is that it is the more recent and, to some extent, builds on Arriaga's earlier work. Arriaga's advantage is that his estimates refer to the decennial years used in this study. Also, his work is more complete; Palloni offers estimates for only three points in time: 1900-05, 1930-35, and 1940-45. For these reasons, the Arriaga estimates were preferred, and Palloni's estimates used to fill in existing gaps and as a check on Arriaga. The devices for checking were the ratios shown in Table 4, which compares figures for those countries and points of time where both authors offer estimates.

Table 4
Life Expectancy, 1900-1950: Comparisons between Palloni and CELADE
(Ratio of Palloni estimate divided by Arriaga estimate)

Table 4 shows that the correspondence is fair in 1940-45, better for 1930-35, and quite good for 1900-05. More specifically, 7 of the 16 comparisons for 1940-45 diverged by more than 5%, 4 of 14 for 1930-35, and only 1 of 7 for 1900-05. The one country with more than a 5% discrepancy in 1900-05 was Costa Rica, with a 10% gap between estimates of Arriaga and Palloni. In addition, Palloni's 1900-05 figure for Colombia is 8% higher than Arriaga's earliest figure, which refers to 1910. It should also be noted that Palloni's 1930-35 figure for the Dominican Republic is 19% above Arriaga's earliest figure, which refers to 1930.

This suggests that while the century-long trends in increase of life expectancy were very little affected by choice of source, the timing and magnitude of the acceleration of improvement around mid-century were affected by choice of source. Thus in the case of Brazil, using Palloni as a source would show an increase in life expectancy of only some 6 years from the early 1940s to the early 1960s, whereas using Arriaga would show an increase of 19 years over approximately the same period. In the case of Bolivia, however, the story was reversed. Palloni estimates an extremely low life expectancy--only 28 years--as late as 1940-45, and then a very rapid rise in the 20 years following, presumably the result of improved health conditions following the Bolivian Revolution. In both the Bolivian and Brazilian cases one source is not intrinsically more plausible than the other, but Arriaga's estimates were preferred in part for source uniformity.

There were, however, instances where Palloni includes data not available in Arriaga (or in the U.N. Demographic Yearbook, in the case of Argentina). In the case of Argentina 1900-05, Palloni's figure of 40.3 was used, along with the UN figure for 1915, for constructing estimates for 1910 and 1900. In the case of Cuba 1900-1950, Palloni's figures from 1900-05 to 1950-55 were used to interpolate estimates for 1900 through 1950. This series fit well enough with the CELADE series that begins in 1955 that no linking adjustments were attempted. In the cases of Honduras 1900-05 and Panama 1900-05, it was decided not to rely on the Palloni figures to fill the gaps in Arriaga's estimates for these countries and periods without more information on Palloni's estimation methodology. Both estimates seem dubious, Honduras 1900-05 because it is identical to Honduras 1930-35, and Panama 1900-05 because it is only slightly lower than Panama 1930-35—this despite the fact that mortality rates in Panamanian cities dropped sharply at the time of the building of the Panama Canal.

Data definitions

The data series included in MOxLAD are defined as follows:

Population and Demographics

Go to search page

Population: Midyear estimates of total residents from census data, with interpolations between census years; expressed in thousands.

Urban Population: Midyear population of areas defined as urban in each country; expressed in thousands. Please note that the definitions tend to vary between countries and over time, as the process of urbanisation occurs. Figures are calculated from their share of total population reported in the original data sources.

Life Expectancy at birth: The number of years a newborn infant would live if prevailing patterns of mortality at the time of birth remained unchanged throughout his or her lifetime; expressed in years.

Illiteracy rate: The percentage of the population of or above a certain age (usually 15) who cannot read and write a simple statement about everyday life. In practice, however, the illiteracy rate can sometimes refer to the percentage of the population of or above a certain age with less than two years of primary school enrolment, or can be self-defined. The reported figures come from census data, with interpolations between census years.

Public Spending on Education: Current operating expenditures in education, including wages and salaries and excluding capital investments in buildings and equipment; expressed as a percent of gross domestic product.

Primary, Secondary, and Tertiary School Enrolment: Gross enrolment in primary, secondary, and tertiary levels of education regardless of age; expressed in thousands. Secondary school generally includes vocational and teacher training. Users should be aware that enrolment data may not be very precise due to lack of uniformity of definition of enrolment, (e.g., pupils in attendance on a particular day, in regular attendance, or on the register), reorganisation of school systems, differences in the period constituting the school year between countries and over time, and the inclusion or exclusion of private school enrolment. Users are advised to consult the country notes and cautioned that these may not always fully account for changes in data coverage.

Labour Force

Go to search page

Economically Active Population (EAP): The number of employed and unemployed persons. Reported figures come from census data with interpolations between census years; expressed in thousands. The number in brackets preceding the (+) sign (e.g., [C 14+]) refers to the lower age limit of those persons considered to be economically active; the higher end is usually assumed to be 65 years. Figures are calculated from their share in total population reported in the original data sources.

EAP in Manufacturing: The number of employed and unemployed persons in the manufacturing industry; expressed in thousands. Figures are calculated from their share in total EAP reported in the original data sources.

EAP in Agriculture: The number of employed and unemployed persons in agriculture; expressed in thousands. Figures are calculated from their share in total EAP reported in the original data sources.

Industry

Go to search page

Electricity Output: Gross output of electricity, including electricity consumed in power stations and transmissions losses; expressed in million gigaWatt hours. Figures generally include electricity generated by hydropower (hydroelectric power plants), coal, crude oil and petroleum products, natural gas, and nuclear power plants, as well as geothermal, solar, wind, and tide and wave energy, and combustible renewables and waste.

Cement Production: Manufacturing production of hydraulic cements used for construction-purposes, including aluminous, metallurgic, natural, and Portland cement; expressed in thousand metric tonnes.

Beer Production: Manufacturing production of beer; expressed in thousand hectolitres. Figures include all types of beer except home-brewed beer.

Transport and Communications Infrastructure

Go to search page

Roads: Generally the total network of permanent traffic; expressed in thousand kilometres, but for earlier periods may widen to include non-permanent roads, or narrow to roads suitable for motor vehicle transit only. Users are advised to consult the country notes for changes in data coverage.

Railways: Route length of line open at the end of each year; expressed in kilometres. Figures generally include narrow gauge line and exclude mountain railways. Purely industrial lines, (i.e., closed to public traffic), are not generally included, although the distinction cannot always be made for some countries, (e.g., Central American countries).

Telephones: Number of telephones in use; expressed in thousand lines. For later years, may include cellular subscribers. Users are advised to consult the country notes for changes in data coverage.

Passenger Cars and Commercial Vehicles: Official estimates at year-end; expressed in thousands. Passenger cars are defined as cars seating fewer than eight persons, and include jeeps and station wagons but exclude special-purpose vehicles. Figures for commercial vehicles include buses and taxis but exclude trailers and farm tractors.

External Trade

Go to search page

Exports: Total value of merchandise exports valued FOB (free on board); expressed in million US dollars. The FOB valuation comprises the price of goods at the border of exporting country and includes the value of the commodity, all outlays on transport to the shipping point, and all requisite fees for loading. Reported trade figures are generally of ‘special’ rather than ‘general’ trade, i.e. they refer only to commodities that have been produced within the country.

Index of Unit Value of Exports: The fixed-weight index of price of exports, generally constructed using the Laspeyres formula. Series with different base years are linked by ratio splicing at the first annual overlap; linked series are then shifted to a common base period (1970=100). These series differ from export price indices, which are compiled from survey data for prices at the wholesale level or from direct pricing (i.e., directly from the exporter); the latter are generally considered preferable to unit value indices because of biases introduced by the aggregation of commodity classification categories with no meaningful unit of quantity (e.g., categories related to machine tools, capital equipment, and certain consumer goods).

Index of Export Quantum: The fixed-price index of the quantity of exports, generally constructed using the Laspeyres formula. Series with different base years are linked by ratio splicing at the first annual overlap; linked series are shifted to common base period (1970=100).

Imports: Total value of import goods valued CIF (cost, insurance, and freight); expressed in million US dollars. The CIF valuation comprises the price of goods at the border of the importing country, and includes the component elements of FOB valuation, as well as the cost of insurance and the cost of international transport. Reported trade figures are generally of ‘special’ rather than ‘general’ trade, i.e., they refer only to commodities that are intended for internal use.

Index of Unit Value of Imports: The fixed-weight index of price of imports, generally constructed using the Laspeyres formula. Series with different base years are linked by ratio splicing at the first annual overlap; linked series are shifted to a common base period (1970=100). These series differ from import price indices, which are compiled from survey data for prices at the wholesale level or from direct pricing (i.e., directly from the importer); the latter are generally considered preferable to unit value indices because of biases introduced by the aggregation of commodity classification categories with no meaningful unit of quantity (e.g., categories related to machine tools, capital equipment, and certain consumer goods).

Index of Import Quantum: The fixed-price index of the quantity of imports, generally constructed using the Laspeyres formula. Series with different base years are linked by ratio splicing at the first annual overlap; linked series are shifted to common base period (1970=100).

Consumer Imports: These refer to miscellaneous manufactured articles (section 8 of the Standard International Trade Classification (SITC) Rev. 3); expressed as a percent of Imports.

Intermediate Imports: These refer to sections 0 to 6 and section 9 of the Standard International Trade Classification (SITC) Rev. 3); expressed as a percent of Imports.

Fuel Imports: These refer to mineral fuels, lubricants and related materials; (section 3 of the Standard Trade Classification (SITC) Rev. 3); expressed as a percent of Intermediate Imports.

Capital Imports: These refer to machines, transport, and equipment imports (section 7 of the Standard International Trade Classification (SITC) Rev. 3); expressed as a percent of Imports.

Index of US Inverse Net Barter Terms of Trade: Included as a proxy for world terms of trade (index of world export prices relative to world import prices) faced by Latin America. This index is obtained by dividing the price index of non-manufactured imports (to 1970, and general imports thereafter) by the price index of exports at a constant base year (1970=100).

Finance

Go to search page

Central Government Revenue: Total ordinary revenue exclusive of loan receipts; expressed in million local currency units. Depending on data availability, for some countries and time periods, figures are for current (tax and nontax) revenue excluding capital revenue and grants. Users are advised to consult the country notes for changes in data coverage.

Customs Tax Revenue: Customs tax revenue; expressed in million local currency units. Depending on data availability, for some countries and time periods, figures may include all or certain components of taxes on international trade, (e.g., import duties, exchange taxes, and other taxes on international trade and transactions). Users are advised to consult the country notes for changes in data coverage.

Income Tax Revenue: Tax on income, profit and capital gains; expressed in million local currency units, may include property tax for some countries and time periods. Users are advised to consult the country notes for changes in data coverage.

Taxes on Domestic Goods and Services: Expressed in million local currency units. Depending on data availability, for some countries and time periods the figures represent Value-Added Tax or excise taxes only. Users are advised to consult the country notes for changes in data coverage.

Central Government Expenditure: Includes all types of central government budgetary expenditure excluding debt redemption; expressed in million local currency units. Depending on data availability, for some countries and time periods, figures may include extra-budgetary expenditures, or entail budget estimates. Users are advised to consult the country notes for changes in data coverage.

National Accounts

Go to search page

Gross Domestic Product: Generally at factor cost (i.e., consists of GDP at market prices including subsidies but excluding indirect taxes); expressed in million local currency units at current prices, million constant local currency units at 1970 prices, and million constant international dollars at 1970 prices. Official PPP estimates for Cuba are not available (see Pérez-López, 1991), and ECLAC only reports GDP figures from 1985 onwards, see ECLAC SYLA (1998, 2000). Prior to 1989, official sources report only Gross/Global Material Product (GMP) figures for Cuba, in accordance with the Soviet Material Product System (MPS). Estimating GDP from GMP figures is further complicated by the exclusion of non-material services and 'gross value' approach in the GMP, see Brundenius (1984), pp. 19-40, Mesa-Lago (1981), pp. 33, 199-201, and Mesa-Lago (2000), pp. 564-5. Estimates of Cuban GDP, agricultural value-added and manufacturing value-added in current and constant 1980 prices were provided by Claes Brundenius.