1. Series needed in preprocessor. Gross domestic product (GDP); gross domestic product per capita at PPP (GDPPCP); exports and imports as a percent of GDP (EXPORTS, IMPORTS); personal consumption, government consumption and investment as a percent of GDP (PCON, GCON, INVEST). Have numerical functions that fill holes in GDPPCP, EXPORTS, IMPORTS, PCON, GCON, INVEST that should be removed from code and switched between 1960 and 1995 values. Preprocessor balances global trade and normalizes expenditure components to GDP. Preprocessor reads merchandise exports and imports (MERCHEX and MERCHIM) in billion dollars and computes service exports and imports. Preprocessor reads crop and meat imports and exports in physical units (AGXCROP, AGMCROP, AGXMEAT, AGMMEAT) and uses world prices (WAP) to compute values (need to fix to read historic WAP values). Agricultural trade is constrained by merchandise trade and normalized. Preprocessor reads energy exports and imports in physical units (ENX, ENM) and uses world prices (ENPRI) to compute values (need to fix to read historic ENPRI values). Energy trade is constrained by remaining merchandise trade and normalized. (GDPPC, SAVINGS exist in the data file but appear not to be used; dropped out in historic load.)
Preprocessor reads percentages of total exports in metals, minerals, and other primary categories and percentages of total imports in mineral/material and other primary categories, removes energy trade already computed, and constrains remainder by remaining merchandise trade and normalizes. This may need to be changed fundamentally for historic load.
Manufacturers trade is computed as a residual of merchandise trade after above computations of agricultural, energy, and raw materials trade.
The preprocessor next reads the percentages of value added in agriculture, manufactures, industry, and services. It fills holes with cross-sectionally estimated functions. It splits industry minus manufacturers into energy and raw materials, based on export volumes (very crude) and normalizes value added to 100 percent.
The preprocessor next splits personal consumption across all sectors except raw materials. It similarly splits government consumption into manufactures and services and identifies the origin of investment as manufactures. The personal and governmental consumption splits are crude functions that should be improved with sector-specific data.
The preprocessor next reads external debt (XDEBT) in billion dollars and allocates that debt across developed countries.
The preprocessor next fills holes in the income share of the lowest 20% of the population (INSHAREL) using a numerical function that should be put into the analytic function file and similarly fills holes in the initial economic growth rate (IGDPR).
Finally, the preprocessor stores newly computed/adjusted/normalized values for sectoral value added (VADD), exports (XS), imports (MS), consumption (CS), government consumption (GS), and investment (INVS).
2. Series needed in data load. IO matrices (IOMAT) by GDP per capita level, also providing sectoral labor requirements. The load also draws on the variables prepared by the preprocessor, including especially gross domestic product (GDP) at both exchange rates and purchasing power parity; GDP growth rate (IGDPR).
3. GDP
We began with World Bank series for GDP in constant 1995 dollars (from the WDI 1999 CD). That provided basically complete series for about 97 countries and partial data for many others. The major holes for larger countries were:
Afghanistan, no years
Bolivia, 1980+
Bulgaria, 1980+
Ethiopia, 1981+
Germany 1991+
India, 1965+
Iran, 1974+
Iraq, all years
Kuwait, 1962+
Libya, all years
Mongolia, 1981+
Morroco, 1966+
Peru, 1965+
Phillipines, 1965+
Poland, 1980+
Romania, 1975+
Russian Fed, 1989+
Taiwan, all years
Turkey, 1968+
UAE, all years
Vietnam, 1990+
Hole-filling strategies:
a. We used GDP in current dollars (also World Bank) with a conversion to constant dollars from US dollar CPI to fill what we could (e.g. Afghanistan). We used other World Bank GDP series (from the WDI CD) to fill some. We went back to various issues of the World Development Report or World Development Indicators (especially 1982) to fill additional holes, focusing on getting as much data for 1960 as possible. Using the WDR there were a few values put in for 1960 that left holes over time that we filled with interpolation. GDP for 1995 in Lebanon and for 1980 in Tanzania came from the World BankWorld Development Indicators 1997. When we finished use of this strategy, we a considerably more compete set of data (notably missing the FSU republics, other communist countries, and a number of other countries).
b. The second strategy involved use of various volumes of the U.S. Arms Control and Disarmament AgencyWorld Military Expenditures and Arms Transfers to fill as many holes as possible. The most used reports were the 1965, 1986, and 1995. The time-series in those reports span from 1965-1994:
1965…1974 1965 WMEAT report
1974…1984 1986 WMEAT report
1984…1994 1995 WMEAT report
Unfortunately, we were unable to gather data from WMEAT for 1960-64.
The central rule used was that we accepted existing (World Bank) data in the matrix, when available, as “best estimate” and used ACDA values to extend the values forward or backward in constant 1995 dollars. The process for extending the data series was to use overlapping years between the existing series and those in the ACDA volumes (or between different series in different ACDA volumes) in order to compute ratios that (i) convert the series added to 1995 dollars and that simultaneously (ii) update earlier series to the values at the beginning of later series (in some cases we anticipated that more recent series would, in essence, be corrected relative to earlier ones).
For instance, to convert values from the 1986 ACDA report to 1995 dollars, we computed a conversion coefficient as follows (also converting from million to billion dollars):
Conv Coef = GNP value for 1984 from the 1995 report/GNP value for 1984 from the 1986 report/1000
To convert values from the 1965 report we had to extend the logic:
Conv Coef = GNP value for 1984 from the 1995 report/GNP value for 1984 from the 1986 report* GNP value for 1974 from the 1986 report/GNP value for 1974 from the 1965 report /1000
In some cases, such as Mozambique, we extended the World Bank series back from a year in the middle of a series in the ACDA volume (1980 for Mozambique). We used the same conversion coefficient calculation, centered on 1980.
The ACDA volumes allowed us to fill in Afghanistan (1982-88), Albania (1965-76), Angola (1975-84), Bulgaria (1965-69), Cuba (1965-93), Cyprus (1965-93), Ethiopia (1965-69), Guinea (1965-85), Iran (1966-73), Jordan (1965-69), North Korea (1965-94), Lebanon (1965-87; 1989-94), Libya (1990-94), Mozambique (1975-79), Myanmar (1965-92), Poland (1965-79), Romania (1965-74), Somalia (1965-89), Swaziland (1969), Taiwan (1965-94), Tanzania (1965-87), Uganda (1965-81), and Yugoslavia (1965-91).
c. Third, we turned to the CIAdata set on GDP at purchasing power parity (obtained courtesy of the Strategic Assessment Group of DI). That data set represents GDP at purchasing power parity. Instead of putting in purchasing power parity values, however, we used the values to extend again the series available to use from the combined World Bank and ACDA sources. Thus the combined file is intended to represent GDP at exchange rates.
The CIA data set allowed us to fill in Albania (1960-79; we replaced earlier 1965-76 values from ACDA because they appeared rougher), Angola (1960-74), Bosnia and Herzegovina (1960-97), Bulgaria (1960-64), Croatia (1997), Cuba (1960-64, 1995-97), Cyprus (1960-64, 1994-97), Czech Republic (1960-83), Ethiopia (1960-65), The Gambia (1960-64), Greece (1997), Guinea (1960-64), Guinea-Bissau (1960-69), Guyana (1996-97), Iceland (1997), Iran (1960-64, 1996-97), Jordan (1960-64), North Korea (1960-64, 1995-97), Kuwait (1996-97), Laos (1960-87), Libya (1995-97), Luxembourg (1997), Mongolia (1960-80), Mozambique (1960-74), Myanmar (1960-64, 1993-97), Namibia (1960-79), Oman (1996-97), Panama (1997), Poland (1960-64), Puerto Rico (1996-97), Qatar (1960-69, 1996-97), Romania (1960-64), Somalia (1960-64), Suriname (1960-69, 1996-97), Swaziland (1960-68), Taiwan (1960-64, 1995-97), Tanzania (1960-64), Uganda (1960-64), United Arab Emirates (1960-71, 1996-97), Vietnam (1960-83).
d. The fourth strategy was used to fill the holes for the former Soviet Republics, for the former Yugoslavian Republics, and for the constituent parts of Czechoslovakia (the Czech Republic and Slovakia). Because the currently independent republics were parts of larger entities prior to the early 1990s, we do not have data for them from any of the above sources in earlier years (it might be that we will ultimately find relative data from earlier years). Therefore we used their size in the first year for which we do have data (for the former Soviet Republics that year was 1990) to compute their portions of the former aggregate entity in that year. We applied those portions to the former aggregate entity in earlier years (we had data for the Soviet Union, Yugoslavia, and Czechoslovakia from the above sources) to crudely estimate the individual republic sizes prior to the breakup of each country. With respect to Germany we reversed this procedure, adding up East and West Germany before 1989 to calculate the size of an aggregate Germany back to 1960.
e. At this point we had at least complete data except: Bhutan (1960-1979, 1996-97), Djibouti (1960-1990), French Guiana (all), Greenland (all), and Lebanon (1960-64). [also note bad data for Lebanon, prior to 1987]. The Information Please Almanac and the New York Times Almanac provide the following data: Bhutan GDP was $1.3 billion in 1995 and $1.9 billion in 1998; Djibouti GDP was $500 billion in 1995 and $530 billion in 1998.
The other missing countries are related to country-specification issues in IFs. Greenland (like the Faroe Islands and about the same size demographically/economically) is part of the Danish realm. The CIA Factbook (1992) said it had a population of 57,407 and a GNP of $500 million ($9,000 per capita). French Guiana is an overseas department of France (in South America). The CIA Factbook (1992) said it had a population of 127,500 and a GNP of $186 million ($2,240 per capita). Two larger countries are missing from the model: (1) Equatorial Guinea (in 1999 population of 465,746; in 1997 GDP of $660 million) and Eritrea (in 1999 population of 3.98 million; GDP of $453 million in 1998).
We should perhaps replace Greenland with Eritrea and French Guiana with Equatorial Guinea in the model. But probably we should just add Eritrea and Equatorial Guinea. We should also consider adding Palestine.
4. GDP at purchasing power parity
Used CIA data from Strategic Assessment Group, which is pretty complete from 1960-1998. Have missing data for all years for Afghanistan, Belize, Bhutan, Cambodia, Djibouti, French Guiana, Greenland, Lebanon. Have missing data before 1990 for all former Soviet Republics, including Russia.
5. GDP growth rate
Used the GDP data set described above to compute GDP growth rate for the longest possible period for all countries. Using those values and the values for the 1995 data load (from the most recent data), and using judgment only for countries without GDP data (including all of the former Soviet republics), estimated values for IGDPR (initial GDP growth rate).
6. Economic data added to historical data tables
GDP in 1995 dollars, GDP95, 1960-97 (In process of being updated, filled out)
GDP per capita at PPP in 1998 dollars, 1960-1998 (from DI)
Value added in agriculture, percent of GDP, 1970-95 (from WRI)
Value added in manufactures, percent of GDP, 1970-95 (from WRI) – subset of industry
Value added in industry, percent of GDP, 1970-95 (from WRI) - broadest category
Value added in services, percent of GDP, 1970-95 (from WRI)
Private consumption, 1970-90, 5-year intervals (appear nominal values)
Private consumption as a portion of GDP, 1960-98 (WDI)
Government consumption, 1970-90, 5-year intervals (appear nominal values)
Government consumption as a portion of GDP, 1960-98 (WDI)
Investment, 1970-90, 5-year intervals (appear nominal values)
Investment as percent of GDP, 1970-90, 5-year intervals - can extend with WDI back to 1960
Savings, 1970-90, 5-year intervals
Savings percentage of GDP, 1970-98 (WDI)
Savings (genuine, special measure) percentage of GDP, 1970-98. Appears to discount raw material use, including forest depletion and add educational expenditures
Exports of goods and services as percent of GDP, 1960-98 (WDI)
Imports of goods and services as percent of GDP, 1960-98 (WDI)
Exports, Services, 1970-98, nominal (WDI)
Imports, Services, 1970-98, nominal (WDI)
Exports, Merchandise, 1970-98, nominal (WDI)
Imports, Merchandise, 1970-98, nominal (WDI)
Agricultural goods, raw exports as % of merchandise exports, 1980-98 (WDI)
Agricultural goods, raw imports as % of merchandise imports, 1980-98 (WDI)
Agricultural goods, food exports as % of merchandise exports, 1980-98 (WDI)
Agricultural goods, food imports as % of merchandise imports, 1980-98 (WDI)
Arms exports as % of merchandise exports, 1985-97 (WDI)
Arms imports as % of merchandise imports, 1985-97 (WDI)
Fuel exports as % of merchandise exports, 1980-98 (WDI)
Fuel imports as % of merchandise imports, 1980-98 (WDI)
Manufactures exports as % of merchandise exports, 1980-98 (WDI)
Manufactures imports as % of merchandise imports, 1980-98 (WDI)
Ores and metals exports as % of merchandise exports, 1980-98 (WDI)
Ores and metals imports as % of merchandise imports, 1980-98 (WDI)
Communications, computer, etc. imports as a percent of service imports, 1970-98 (WDI)
Communications, computer, etc. exports as a percent of service exports, 1970-98 (WDI)
Insurance and financial exports as a percent of service exports, 1970-98 (WDI)
Insurance and financial imports as a percent of service imports, 1970-98 (WDI)
Transportation exports as a percent of service exports, 1970-98 (WDI)
Transportation imports as a percent of service imports, 1970-98 (WDI)
Travel exports as a percent of service exports, 1970-98 (WDI)
Travel imports as a percent of service imports, 1970-98 (WDI)
Travel exports as a percent of service exports, 1970-98 (WDI)
Important note: For the United States, the sum of communications, computer etc, insurance and financial, transportation, and travel exports percentages equals 100; that is, those four categories are an exhaustive breakdown of service exports.
Tourism expenditures as a percent of total imports, 1980-98 (WDI)
Tourism receipts as a percent of total exports, 1980-98 (WDI)
7. Economic data added to historical data load for 1960
GDP in 1995 dollars (GDP) – incomplete; needs to be fleshed out
GDP per capita at PPP (GDPPCP), 1960 in 1998 dollars
GDP per capita at PPP below 2000 (GDPPCPLT2000), transferred from GDPPCP
GDP per capita at PPP between 2000 and 4000 (GDPPCP2to4000), transferred from GDPPCP
GDP per capita at PPP below 4000 (GDPPCPLT4000), transferred from GDPPCP
GDP per capita at PPP greater and or equal to 4000 (GDPPCPGE4000), transferred from GDPPCP
Value added in agriculture, percent of GDP, used 1970 or earliest other value (WDI 2000 has agriculture value added, % of GDP 1960-98, skimpy but useful before 1970)
Value added in manufacturing, percent of GDP, used 1970 or earliest other value (WDI 2000 has manufacturing value added, % of GDP 1960-98, skimpy but useful before 1970).
Value added in industry, percent of GDP, used 1970 or earliest other value (WDI 2000 has industry value added, % of GDP 1960-98, skimpy but useful before 1970).
Value added in services, percent of GDP, used 1970 or earliest other value (WDI 2000 has services value added, % of GDP 1960-98, skimpy but useful before 1970)
Private consumption as a portion of GDP (PCON), average of 5 earliest years, often 1960-64 (WDI)
Government consumption as a portion of GDP (GCON), average of 5 earliest years, often 1960-64 (WDI)
Investment as % of GDP (INVEST), average of 5 earliest years, usually 1960-64 (WDI)
Exports as a portion of GDP (EXPORT), average of 5 earliest years, often 1960-64 (WDI)
Imports as a portion of GDP (IMPORT), average of 5 earliest years, often 1960-64 (WDI)
Merchandise exports and imports. Used earliest year available from WDI series for historical base file; 1960 was often not available.
Service exports and imports. Put earliest year available into historical base file; 1960 was often not available. But numbers Nyema gave me for service imports are wrong; for instance, Uruguay trade in services bigger than U.S.; fix this when Nyema provides correct data. Also, that suggests that the historical series files for service imports must be wrong and needs to be corrected.
Exports of ores and metals as a percent of merchandise exports (ExOresMets), 1980 filled with most recent year around 1980 – also went back and updated 1995 file in same way with this new series
Imports of ores and metals as a percent of merchandise imports (ExOresMets), 1980 filled with most recent year around 1980 – also went back and updated 1995 file in same way with this new series.
Note: although we now have values for food exports/imports as a percentage of merchandise exports/imports and also have the same for fuels and manufactures, the preprocessor uses physical data to compute agricultural and energy trade and then computes merchandise as a residual. Probably best to keep it that way and just use the monetary data as checks.
Note: when I set the energy price at $3 for 1960, all but one country disappeared from the economic data quality error messages obtained at run-time; work on this after put in better energy data, which may fix it. The message for 1995 data (region 41) disappeared when I upgraded the materials sector approach, as described above.
Reserves (monetary) are not now used. Need to rework financial representation; historic values are now the same as for 1995.
External debt in billion dollars (XDEBT) was filled with values in 1971 (WDI), the earliest available.
8. Next steps for economic data
· Move functions imbedded in code into table function(s)
· WDI 2000 does not have income share data table(s); will need to use what already have for 1960 data load; draw out earliest value we have from historic data series. Have not yet changed income shares for historic load.
· Do not have any data now on personal consumption (or governmental) by sector. Search for it.
· Need to compute historical matrix of GDP growth rates and then use 10-year averages for IGDPR historically. Have not yet changed IGDPR for historic load.