Background Note on Historical Statistics in www.ggdc.

net/Maddison (March 2010)

Standardized estimates of GDP growth provide a coherent macroeconomic framework
covering the whole economy, which can be crosschecked in three ways. GDP is by definition
equal in each of the three approaches. From the production side, it is the sum of value added
in different sectors (agriculture, industry and services) net of duplication. On the demand side,
it is the sum of final expenditures by consumers, investors and government. From the income
side, it is the total of wages, rents and profits. In all three dimensions these measures need to
be adjusted to eliminate changes in the price level in the period they cover, so that they show
changes in volume. The volume movement is easiest to measure for production and
expenditure.
The first SNA manual for standardizing the coverage and modes of estimation of GDP
was prepared by Richard Stone and published by OEEC in 1952. Soon after, he produced a
similar manual for the United Nations Statistical Office. The latest version is a joint product
of EU, IMF, OECD, UN and the World Bank (1993), System of National Accounts , Brussels,
New York, Paris and Washington DC.
Most official accounts back to 1950 have been standardized according to these
guidelines; this is also true for most estimates for earlier years by quantitative economic
historians. Until the 1990s, official statistics in communist countries used the Soviet material
product system which took a narrower view of the scope of economic activity than the SNA,
because it excluded many service activities regarded as non-productive. It also involved
double counting (measuring gross output without deducting inter-sector transfers of inputs),
which exaggerated economic growth. Since then all ex-communist countries and China have
switched to the SNA system in principle, but implementation was complicated by massive
changes in ownership, in the level and structure of prices, allocation of resources between
consumption and investment, and statistical reporting procedures. It will take some years
before these problems can be fully resolved.
Intercountry comparison of economic performance relies on purchasing power parity
converters (PPPs) to measure GDP levels, rather than exchange rates. Measures of economic
growth over time must be corrected to exclude the impact of inter-temporal price change. The
purpose of PPP conversion is precisely analogous: the elimination of inter-country differences
in price level, so that differences in the volume of economic activity can be compared across
countries. By merging time series for economic growth with the cross-country estimates of
GDP levels now available we can make a coherent set of space-time comparisons.
There are several alternative measures of PPPs. For our purpose, the method
developed by Kravis, Heston and Summers in their International Comparisons Program (ICP)
is by far the best for multilateral comparisons. They adopted the GearyKhamis (GK) method
invented by Roy Geary (1896-1983) and Salem Khamis (1919-2005), which provides
transitivity and other desirable properties (see their 1982 volume, World Product and
Income). In this study, I have taken 1990 as the benchmark year, so my estimates are in 1990
GK $.
Table 1 shows the sources of the PPPs for the benchmark year 1990. There were ICP
indicators for 69 countries and an ICP type estimate for China made by myself; these
estimates covered 93.7 per cent of world GDP; for 84 countries, representing 5.6 per cent of
world GDP, I used the Penn World Tables (PWT), which is a short-cut measure of Alan
Heston and Robert Summers. For the other 48 countries, representing 0.6 per cent of world
GDP, I made proxy estimates (see Table 1).
Table 1 Nature of PPP Converters Used to Estimate GDP Levels in 1990

(billion 1990 Geary-Khamis dollars and number of countries)
Europe & W.
Offshoots Latin America Asia Africa World
ICP 15,273 (28) 2,131 (18) 8,017 (24) 0 (0) 25,421 (70)
PWT 59 (3) 71 (14) 524 (16) 891 (51) 1,516 (84)
Proxies 16 (10) 38 (15) 87 (17) 14 (6) 155 (48)
Total 15,349 (41) 2,240 (47) 8,628 (57) 905 (57) 27,122 (202)

Source: Maddison, The World Economy: Historical Statistics, 2003, p. 230

The World Bank (2008) presented new PPP estimates for 2005 for 146 countries. The
estimates were made in five regional groups, and then aggregated. There are several good
reasons for being sceptical about the new World Bank results. 1) They use the EKS rather
than the Geary-Khamis procedure. EKS is appealing to bureaucrats because the ranking of
countries within each region cannot be modified in the linking process, because the regions
insist on fixity. It gives all countries the same weight, whatever their size, putting
Luxemburg on a par with the USA. 2) EKS produces a lower relative standing of low-income
countries than the Geary-Khamis measure I used. In the 1982 study of Kravis, Heston and
Summers, p. 96, the average Geary-Khamis GDP result for the lowest income group was 16
per cent higher than the EKS measure. However, the discrepancy between the WB and my
results is bigger than can be explained by the bias of the EKS procedure. 3) The World Bank
estimate for Chinese GDP is only 43 per cent of the USA, whereas my estimate is 80 per cent;
their per capita estimate is only 9.8 per cent for China whereas mine is 18.3 per cent. There is
also a strong downward bias (see Table 2) in the WB estimates for India, Indonesia, Korea,
Thailand and Vietnam. 4) The Bank does not test the plausibility of its new results. For
several years its own Development Indicators contained estimates for China similar to mine.
There were five previous ICP global studies. These are all dismissed (p. 13) as being based
on very old and very limited data, implying that any discrepancy with earlier findings cannot
cast doubt on its weird results for China, India and some other Asian countries. 5) Kravis,
Heston and Summers (1982) contained a detailed sophisticated analysis explaining the
sensitivity of PPP results to different measurement techniqes which is completely lacking in
the recent World Bank study; 6) World Bank results for China are highly implausible when
one considers their intertemporal implications. My growth estimate shows Chinese per capita
income increasing 12.5-fold between 1950 and 2005. If we merge the WB level estimate for
2005 with my growth estimate, one gets a per capita GDP $4,091 in 2005, and $326 (well
below subsistence) in 1950. If one believes the official estimate of per capita GDP growth
(21-fold over 55 years), the 1950 level would be $196.
Table 2 compares my estimates of per capita GDP levels for 2005, in 1990 Geary-
Ghamis dollars, with the new World Bank estimates in 2005 EKS dollars from World Bank
(2008), Global Purchasing Power Parities and Real Expenditure, 2005; International
Comparison Program, Washington DC. The 130 countries covered in the table represent about
95 percent of world GDP.
Table 2 Maddison & World Bank per Capita GDP Relatives in 2005
Maddison 1990 GK W. B. ICP, 2005 EKS $
% of USA % of USA
USA 30,474 100.0 41,674 100.0
Canada 24,485 80.3 35,078 84.2
Australia 24,064 79.0 32,798 78.7
New Zealand 18,078 59.3 24,554 58.9
4 W. Offshoots 29,413 96.5 40,360 96.8
Austria 22,049 72.4 34,108 81.8
Belgium 22,131 72.6 32,077 77.0
Denmark 24,130 79.2 33,626 80.7
Finland 22,169 72.7 30,469 73.1
France 21,513 70.6 29,644 71.1
Germany 19,434 63.8 30,496 73.2
Greece 14,841 48.7 25,520 61.2
Ireland 26,606 87.3 38,058 91.3
Italy 19,303 63.3 27,750 66.6
Luxembourg 37,177 122.0 70,014 168.0
Netherlands 22,819 74.9 34,724 83.3
Norway 27,384 89.9 47,551 114.1
Portugal 14,093 46.2 20,006 48.0
Spain 18,197 59.7 27,270 65.4
Sweden 23,292 76.4 31,995 76.8
Switzerland 23,215 76.2 35,520 85.3
UK 22,438 73.6 31,580 75.8
30 W. Europe 20,497 67.3 30,137 72.3
12 E. Europe 7,255 23.6 12,260 29.4
15 Former 6,311 20.7 9,646 22.4
Argentina 9,019 29.6 11,063 26.5
Brazil 5,750 18.9 8,596 20.6
Mexico 7,486 24.6 11,317 27.2
Iran 5,737 18.8 10,692 25.7
Turkey 7,699 25.3 7,786 18.7
15 West Asia 6,123 20.1 9,738 23.4
J apan 21,978 72.1 30,290 72.7
Hong Kong 27,771 91.1 35,680 85.6
Taiwan 19,018 62.4 26,069 62.6
Singapore 24,610 80.8 41,479 99.5
S. Korea 17,526 57.5 21,342 51.2
China 5,575 18.3 4,091 9.8
India 2,419 7.9 2,126 5.1
Pakistan 2,084 6.8 2,396 5.7
Indonesia 3,868 12.7 3,234 7.8
Thailand 7,878 25.9 6,869 16.5
Vietnam 2,456 8.1 2,148 5.1
11 Asia-Pacific 5,183 17.0 4,895 11.7
53 Africa 1,604 5.3 2,223 5.3

Detailed Source Notes by Country for updates to 2008

There are detailed source notes by country in three earlier OECD publications by
Maddison: Monitoring the World Economy, 1995, The World Economy: A Millennial
Perspective, 2001, and The World Economy: Historical Statistics, 2003. Estimates are shown
for the years 1, 1000, 1500, 1700, and annually from 1820 onwards when available. They
were updated to 2003 in Maddison, Contours of the World Economy, 1-2030 AD: Essays in
Macroeconomic History, Oxford University Press, 2007. For updates and revisions of GDP
volume movement for earlier years, the sources are shown below.
GDP volume movement for 29 OECD countries updated fromNational Accounts of
OECD Countries, Volume 1, Main Aggregates, 1995-2006, pp. 344-5. Updates up to 2008
stem from the IMF, World Economic Outlook, October 2009. They cover the USA, Canada,
Australia, New Zealand. J apan, the Czech and Slovak republics, Hungary, Korea, Mexico,
Poland, Turkey and 17 West European countries (Austria, Belgium, Denmark, Finland,
France, Germany, Greece (where statistics for 1914-20 were derived from S. Kostelenos,
2003, Historical Estimates of National Accounts magnitudes in Greece, 1830-1939 and
Kostelenos, 1995, Money and Output in Modern Greece, Centre of Planning and Economic
Research, Athens), Iceland, Ireland, Luxembourg, Netherlands, Norway, Portugal, Spain,
Switzerland and the UK). These 29 countries accounted for 52 per cent of world GDP in 2006
(24.6 trillion Geary Khamis 1990 dollars). Figures for Cyprus and Malta are available from
the Conference Board estimates on the Groningen website. GDP movement in 9 small
countries (Andorra, Channel Islands, Faeroe Islands, Gibraltar, Greenland, Isle of Man,
Liechtenstein, Monaco and San Marino), 0.015 percent of west Europes population, was
assumed to move parallel to the average for the 12 west European countries shown on my
website. Estimates of Norwegian GDP for 1820-1990 are from Ola Grytten, The Gross
Domestic Product of Norway 1830-2003 in Etreheim, Klovland and Qvigsted (eds.)
Historical Monetary Statistics for Norway, 1819-2003, Norges Bank.

New Estimates for Sweden
There have been 6 major studies of the growth and level of Swedish GDP, by Lindahl,
Dahlgren and Kock in 1937; Lindahl in 1956, J ohansson in 1967, Krantz and Nilsson in 1975.
The two latest are by Rodney Edvinsson, Growth, Accumulation and Crisis in 2005, and
Krantz and Schn, Swedish Historical National Accounts in 2007. Both are published by
Almquist and Wicksell These latest studies show faster growth than the earlier estimates I
made (based on background material to a 1988 article in the Review of Income and Wealth by
Olle Krantz). In my earlier estimate GDP growth was 2.17 per cent a year for 1900-1990, and
1.09 per cent in 1820-1900. This compares with 2.35% and 1.25% for Edvinsson and 2.53%
and 1.67% in Krantz and Schn. I have opted for the Edvinnson estimate, as the downward
momentum of the Krantz and Schn estimate seemed excessive. It showed a 34 fold drop in
per capita income from 2000 to 1800, compared to 26 fold for Edvinsson.
Estimates of Swedish growth before 1800 are not well documented. I had assumed
that growth was 0.17 a year between 1500 and 1820, but this becomes excessive once one
accepts the Edvinsson estimate of the GDP level in 1820. I have now assumed an annual GDP
growth rate in 1500-1820 of 0.07 percent.
Edvinsson made an estimate of GDP growth of 0.18 per cent a year for 1720-1820, but
was chary of estimating earlier performance (see his memo Annual Estimates of Swedish
GDP in 1720-1800.Krantz speculated on early Swedish performance in his 2004 article An
Estimate of Swedish GDP in 1571 in Heikkinen and van Zanden, eds. Exploratons in
Economic Growth. He concluded that GDP was about the same in the sixteenth century as in
1800. It seems likely that he exaggerates the degree of stagnation in these centuries. Lennart
Schn also speculated on this early period in his paper Swedish Historical National
Accounts for the 2009 History Congress in Utrecht. His is a subtle analysis of Swedish
history. It stresses the rapid growth of urbanization in the 17
th
century, and a later slowdown
due to warfare.
Estimates of the 1999-2008 volume movement in East European countries not listed
above, were taken from IMF, World Economic Outlook, October 2009, for Albania, Bulgaria,
Romania and the component republics of the former Yugoslavia. This source was also used
for 14 of the component countries of the former USSR.
Estimates for 2002-2008 year-to-year GDP volume movement in 22 Latin American
and Caribbean countries were taken from ECLAC, Anuario Estadistico de America y el
Caribe 2009. GDP movement, 2003-2008 in 24 small Caribbean countries which represented
less than 1 per cent of the GDP of the region was assumed to be parallel to that in the 23
bigger countries, GDP for Chile 1820-90 from Rolf Luders (1998), The Comparative
Economic Performance of Chile 1910-1995, Estudios de Economia, 25, no. 2, with revised
population estimates 1810-1949 from J . Diaz, R Luders, and G Wagner (2005), Chili, 1810-
2000, La Republica en Cifras, mimeo, Instituto de Economia, Universidad Catolica de Chile.
Peruvian GDP 1896-1949 from Bruno seminario and Arlette Beltran (1998) Crecimiento
Economico en el Peru, 1896-1995, Universidad del Pacifico. Estimates for east and south
Asian countries, 2003-2006, were generally derived from year-to-year volume movements in
Asian Development Bank, Key Indicators of Developing Asian Pacicific Countries, Table 13.
For J apan, Korea and Turkey they are from the above OECD source. For Indonesia, annual
GDP volume movements, 1990-2008 are from the official national accounts. For India year-
to-year volume movements in fiscal years 2002-3 to 2006-7 are from the Reserve Bank of
India, Macroeconomic and Monetary Developments in 2003-4 to 2006-7, Table 1. For China,
1952-2003, they are from A. Maddison and H. Wu, Measuring Chinas Economic
Performance, World Economics, vol 12, no 2. April J une 2008, pp. 13-44. A rough estimate
of Chinese 2004-2008 movement was derived by adjusting downwards the official year-to-
year volume movement in GDP for these years in the China Statistical Yearbook 2009, using
the ratio (81 percent) of the Maddison-Wu ratio to convert the official estimate for 1978-2003.
GDP of the Philippines, 1902-40 from R. Hooley American Economic Policy in the
Philippines, 1902-40: Exploring a Dark age in Colonial Statistics J ournal of Asian
Economics, April 2005, pp.464-488. The GDP estimates for North and South Korea, 1911-74
were amended to correct an error in Maddison 2003. New estimates for Taiwan 1901-89,
were derived from Historical Statistics: Taiwan, Institute of Economic Research,
(Hitotsubashi University, Toyo Keizai, 2008, pp. 231-232, column 6. The movement in their
GDP volume series in 1960 prices was linked to my estimates for 1990 onward in million
1990 Geary-Khamis dollars. (See Maddison , The World Economy: A Millennial Perspective,
OECD, 2001, p.298. Midyear population estimates were supplied by Osamu Saito.
For 12 west Asian countries, the estimates are from the IMF, World Economic
Outlook, October 2009. Turkey 1998-2008 from OECD as above. West Bank and Gaza GDP
2004-8 assumed stagnant at 2003 level. Estimates of the year-to year volume movement for
50 African countries, 1999-2008, were derived from IMF, World Economic Outlook, October
2009. The IMF includes Egypt and Libya in the Middle East on p. 221. Here they are
included in the African total. The IMF does not provide estimates for Mayotte, Reunion, St.
Helena, and Western Sahara. Here I have assumed that their GDP movement was the same as
the total for the other 52 African countries.