世界经济千年史
2004 年第 54 期（总第 470 期）
2004 年 9 月 28 日荷兰格罗宁根大学教授、《世界经济千年史》作者 Maddison 先生在中国经
济研究中心万众楼做了题为“世界经济千年史”的讲座，集中讨论了宏观经济计量历史研究的
发展进程。下面是他报告的主要内容。
宏观经济计量历史的发展经历了三个历史阶段：20 世纪 50 年代以来的“收入和财富国际研究
协会时期”即 IARIW 时期（International Association for Research in Income and Wealth）；1820－
1950 年的库兹涅斯研究为代表的时期或“Kuznetsian 时期”；1500－1820 年商人资本主义年代。
50 年代以来，宏观计量的主要目的是提供可能的政策选择以改善国家的增长表现以及进行国
家间差异分析。我们现在有 50 年代以来世界大部分国家的增长和收入的官方估计。而在
Kuznetsian 时代，数量经济史学家在测量世界经济增长和量化增长因素方面取得了巨大进展。
虽然仍有一些领域需要改进，但对这个时期经济发展大致轮廓没有实质争议。相反，关于商
人资本主义年代的世界经济的描述则存在很大分歧，作为代表的是亚当、斯密的乐观看法和
马尔萨斯的悲观论。
20 世纪 50 年代以来的 IARIW 时代
虽然 50 年代以来标准国民经济核算体系逐渐被采用，但是现存估计中仍有很多问题。使用苏
联 MPS 体系（material production system）的国家增长被高估，需要做大量调整。非洲许多新
生国家缺少必要的技术和资金进行国民收入增长核算，虽然国际组织工作在一定程度上弥补
了这种不足，但是非洲数据也需要做大量调整。还有 1995 年以来一些高收入国家开始引入享
乐指数（hedonic index）来反映产品质量的改善，但是这么做并没有充分的根据并且会高估经
济增长，比如美国用新的计量技术重新进行 1929－1950 年国民经济核算，结果使国内生产总
值(GDP)的增长率由 %增加到 %。最后关于教育和知识的计量也还没有成熟理论。
另一问题是购买力平价转换与国家间 GDP 水平比较。政府统计官员会提供真实价格计算的总
产出和总支出数据，经济学家、新闻记者和官员们都将其作为经济增长和波动的主要指标。
购买力平价方法与用真实价格计量的目的是一样的，即修正价格差异使真实产出和支出水平
的有效比较成为可能。以汇率转换的 GDP 与以购买力平价转换的 GDP 区别非常大。以购买
力平价计算的发展中国家的 GDP 要大大高于以汇率转换的 GDP，差距可能达到 3-5 倍。同时
发达国家 GDP 以汇率计算则容易被高估。更为明显的一个例子是 1950 年中国和印度以汇率
计算人均 GDP 分别是$85 和$175（1990 年价格），这两个数据低得让人难以置信。那么为何
汇率转换还如此经常的被使用呢？一方面出于无知和支持自己观点的需要，另一方面许多发
展国家不愿意接受购买力平价方法，因为担心这会不利于他们申请世行的优惠贷款和资助。
这导致了经济间比较分析的明显错误，媒体经常称日本是世界第二大经济虽然其 GDP 还不到
中国的 60％，英国的政客们也始终相信英国的经济规模比中国大。
1820－1950 年的 Kuznetsian 时代
Simon Kuznets 教授对现代经济增长的研究成果，使可以数量化研究的时间界限从 20 世纪 50
年代推进到 19 世纪 20 年代。关于这段时期我们现有的几个结论是：加速的经济增长始于 1820
年而非 Kuznets 认为的 1760 年；经济学家对西欧各国研究表明西欧的崛起是同时而非交错发
生的；这段时期快速发展加大了西方与其他国家的差距；Kuznets 研究推翻了 Kondratieff 长期
循环的说法和 Schumpeter 循环发展论点。事实上 1820 年以来的技术变革不是浪潮式的而是连
续推进的，目前有足够证据把 1820－2001 这段时间划分为 5 个不同的发展阶段。其中 1950－
1973 年是一个无可比拟的繁荣的黄金时代，这个时期世界 GDP 年均增长 5％，世界贸易年均
增长 8％，人均收入有明显趋同趋势，大部分地区经济增长都快于美国。1973 年以后，世界
经济增长明显下降，各地区的差异加大，但是基于世界的角度，最近这个阶段仍然是增长表
现名列第二的阶段。区分领导国家和跟随国家对于研究技术的动态传播和赶超过程是非常重
要的，“领导”国家指那些处于技术前沿的国家，而“跟随”国家指劳动生产率较低的国家。1500
年以来共有 4 个领导国家，16 世纪的意大利，16 世纪至拿破仑战争时期的荷兰，此后的英国
和 1890 年后的美国。
对 1820 年以来英国、美国和日本的增长分析，最早考虑的是劳动力投入和生产率，战后资本
成了一个非常重要的增长因素，之后又有观点将“人力资本”看作生产要素。Denison 扩展增长
分析方法指出 1820 年以来英国、美国和日本经济增长表现出以下一些显著特点：物质资本大
量增长，非居住建筑和机器设备增加非常明显，伴随在机器设备中加速的技术进步。教育水
平大大提高，英国提高了 8 倍，美国和日本 11 倍。人均劳动投入在英国和日本下降了 40％，
在美国下降了 20％。对外贸易占 GDP 的比重在英国由 3％增至 25％，日本 ％至 13％，美
国 2％至 10％。自然资源的缺乏对增长并不构成限制，人均土地在美国下降了 14 倍，英国和
日本下降了 4 倍。能源的投入增长比较温和，美国人均增长了 3 倍，英国 6 倍，日本 8 倍。
1500－1820 年的商人资本主义年代
关于商人资本主义年代（1500－1820）经济表现，在 18 世纪末就已经存在两种不同的解释。
亚当、斯密乐观地认为，美洲大陆和新航线的发现给大型经济，国际贸易和专业化的发展创
造了新的机会，虽然因为对贸易的共同限制，这些机会不能被充分利用。马尔萨斯则悲观论
则认为，经济表现取决于人口增长和固定的土地供给间的平衡，技术进步、资本形成和国际
贸易专业化的因素被忽略，因此只有通过灾难（战争、饥荒、疾病）才能实现平衡。这两种
观点对立一直存在。此外悲观的看法有了一些其他的支持，LeRoy Ladurie 认为法国 1300－
1720 年的经济是停滞的，真实工资论者则更加悲观他们有人认为英国 1820 年的生活水平比
1500 年下降了 44％等等。
关于 1500－1820 不同国家和地区的经济表现，Kuznets 在 1965 年提出了关于西欧人口率和人
均 GDP 增长的一个非常有影响的假说，认为西欧发达国家 1500－1750 年可能的（并且或许
是最大的）长期人均产出年均增长率为 %。对上述假说验证表明，西欧 1500－1820 年人
均年增长率为 ％，显著小于 Kuznets 估计，美洲国家整体而言人均 GDP 增长快于西欧，
非洲南部和北部的发展差异比较明显，亚洲国家整体而言收入水平是停滞的，日本的人均表
现要好于中国和印度，但是非常明显中国这个时期有着广泛的增长，它在人口大量增长的情
况下生活水平并没有下降并且 GDP 的增长与西欧一样显著。
在分析商人资本主义时代的增长时，增长分析的方法就不再适用了。研究表明，人均劳动投
入、资本和人力资本、知识在这个时期都有增加，但是非常明显的是全球化在这个时期起到
了比以往更加重要的作用，相对来说全球化在这段时期的作用比在 20 世纪还要重要。西方造
船业和航海的巨大发展，使世界贸易在 1500－1820 年增长了 20 倍。欧洲国家还从殖民地，
非洲奴隶贸易中获取了很多利益。美洲国家则经历了生态、技术和人口的转型。引进新的农
作物提高了粮食的产量；马和其他牲畜的引入则改善了交通条件；欧洲各种技术的引入也有
助于经济的发展；欧洲疾病使 2/3 土著居民死亡，加上欧洲人和非洲奴隶大量涌入，改变了美
洲的人口结构。非洲和亚洲国家也引入了美洲农作物。此外还存在大陆间的技术传递，欧洲
向美洲输出武器、工具、车辆、船和造船技术、印刷、文字、教育和政治经济机构，欧洲采
矿技术在美洲应用生产出大量金银供欧洲与亚洲进行贸易，同时期欧洲从亚洲引入了纺织和
陶瓷技术。
与亚当斯密和马尔萨斯关于商人资本主义的对立解释相对应，关于现代化的根源有非常不同
的看法。有一个学派认为现代经济增长源于工业革命，而之前则是几个世纪的马尔萨斯停滞。
很多人支持这个观点，但是这些观点在根本上是错误的。事实上向现代资本主义过渡经历了
长时间的准备。首先是教育和知识的传播由于印刷术的发明和推广而变革；其次造船业和航
海业到 1820 年也发生了变革，船只的设计、装备和天文知识都大大改善，有了精确的航海指
导等等。这些进展都是科研努力的结果。无疑，欧洲的这些发展是 19 和 20 世纪经济更快发
展的前奏，欧洲的现代化不是一蹴而就的。
（任丽达、卢锋整理）
--------------------------------------------------------------------------------
发贴时间： 2004-10-17 2:28:56 .*.*
errantking
等级：贫农
财产：
经验：
魅力：
注册：2003-9-7
文章：55
鉴定：保密
--------------------------------------------------------------------------------
Measuring and Interpreting World Economic Performance 1500-2001*
Angus Maddison
This is a suitable occasion for surveying the progress achieved, in the past 60 years, in quantifying
world economic development, and analysing the causal influences which determine the pace and
pattern of growth. This was a major objective of the founding fathers of the International
Association for Research in Income and Wealth (IARIW). The initiative for creating an association
including both academics and official statisticians came from Simon Kuznets (1901-85), the pioneer
of quantitative economic history. Milton Gilbert (1909-79) and Richard Stone (1913-1991) were
strategic partners with enormous international leverage in creating and diffusing standard
procedures for construction of comparable national accounts by official statistical offices.
This paper analyses the development of macro-measurement in three epochs:
a) for the IARIW epoch, back to 1950, the main purpose has been to illuminate policy options to
improve growth performance at the national level and to analyse inter-country divergence in real
income levels to help devise policies for catch-up. We now have official estimates of growth and
levels for the vast bulk of the world economy from 1950 onwards.
b) For the Kuznetsian epoch of “modern economic growth” back to 1820, quantitative economic
historians have made great progress in measuring world economic growth and in quantifying the
forces determining performance. There is scope for further research to fill gaps and crosscheck
existing estimates, but the broad contours of development in this period are not under serious
challenge.
c) for the “merchant capitalist” epoch, 1500-1820, there are sharply divergent interpretations about
world, and particularly European performance. The dichotomy between positive and negative
assessments started with Adam Smith and Malthus at the end of the 18th century. In my view, the
evidence for the Malthusian view is shoddy.
Historians usually start at the beginning of their chronology. Quantitative economic historians have
to work backwards from the present, proceeding from what is known and accepted, to earlier epochs
where evidence is weaker and there is greater reliance on clues and conjecture.
*This is the first Ruggles Lecture, delivered at the 28th IARIW General Conference, Cork, Ireland
August 2004
I
(i) Standardised Estimates of GDP Growth for 1950 onwards
The standardised accounts provide a coherent macroeconomic framework covering the whole
economy, crosschecked in three ways. From the income side, they are the total of wages, rents and
profits. On the demand side, they are the sum of final expenditures by consumers, investors and
government. From the production side, the sum of value added in different sectors-agriculture,
industry and services, net of duplication.
Milton Gilbert had been responsible for the official US accounts during the war and from 1950 to
1961 was head of statistics and national accounts in OEEC. The Marshall Plan required criteria for
aid allocation, and NATO needed them for its burden-sharing exercises. Gilbert met these
requirements by pushing official statistical offices of the 16 OEEC member countries to adopt the
standardised system of accounts designed by Richard Stone.
Stone set up a programme in Cambridge to train official European statisticians to implement the
standardised system. A set of national handbooks was prepared to explain the problems of
adjustment to the standardised system, and a first comparative set of accounts for the 16 countries
for 1938 and 1947-52 was published by OEEC in 1954, with extensive notes explaining the
adjustments which had been made to achieve comparability.
In 1953, Stone became chairman of a United Nations commission which established a system of
accounts for worldwide application. The UN could not exert as much leverage on its member
countries to conform as was possible in OEEC. The communist countries used the Soviet SMS
(system of material accounts) which had a narrower definition of productive activity (excluding
many service activities), involved serious double counting (measuring gross output without
deducting inter-sector transfers of inputs) and exaggerated economic growth. The price system and
tax-structures were different from those in capitalist countries, and measurement conventions gave
incentives to exaggerate quality change when new products were introduced. Abram Bergson
(1914-2003) pioneered procedures for re-estimation of Soviet GDP on a basis corresponding
approximately to Western conceptions in coverage, with elimination of double-counting, and
repricing on an “adjusted factor cost” basis with imputation for capital costs which were not
considered in Soviet accounting. These corrective procedures were applied to Soviet statistics by a
team of CIA Sovietologists in Washington. In New York, Thad Alton and his colleagues did the
same for Bulgaria, Czechoslovakia, East Germany, Poland, Romania and Yugoslavia. This work
was financed for intelligence purposes, but was publicly available in annual reports to the US
Congress (see Maddison 1998b).
In the 1990s most of these countries adopted the standardised SNA system in principle, but
implementation was complicated by the massive change 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. The IMF continues
to use exaggerated measures of GDP growth for these countries. As a result, it shows a growth in
world GDP averaging percent a year for 1970-200, compared with my estimate of percent.
For China it shows growth averaging per cent a year, whereas my adjusted measure shows a
growth rate of per cent (see Maddison, 2003, p. 231).
Another area of weakness is Africa, where there was and still is a great shortage of skills and money
for such work in a large number of newly created countries. The gap in estimates of GDP growth
was filled in substantial degree by the OECD Development Centre which compiled annual estimates
of real GDP growth 1950-90 for 51 African countries. The Centre benefited from the expertise of
Derek Blades, who had been chief statistician in Malawi for eight years, and by David Roberts who
had similar experience in Gambia.
A third problem in the assessment of GDP growth performance in the higher income countries
derives from recent changes in measurement conventions from 1995 onwards, involving adoption of
hedonic indexes to adjust for assumed changes in quality of product, use of chain indices, and
treatment of consumer software as investment.
Hedonic indices are perfectly respectable in small doses, but one can be skeptical about the
widespread assumption that quality changes have been so large and monotonically positive. In the
USA, where the switch to hedonics was most significant, their net impact was to raise the measured
rate of growth to a somewhat greater degree than in Western Europe and Japan. US official
estimates go back to 1929, and the changes in measurement technique had their biggest impact for
1929-50, raising the GDP growth rate for that period from percent a year to . There was no
counterpart to this long retrospective readjustment in other countries, and I have continued to use
the earlier US official measure for 1929-50 (for reasons explained in Maddison, 2001, p. 138, and
Maddison, 2003, -80). More than 40 years ago, Milton Gilbert warned that such adjustments
could open Pandora’s box: “In the end, they would make it impossible to construct measures of
output and price changes that are useful to the study of economic growth” (Gilbert, 1961, p. 287).
The danger which arises from an overdose of hedonics is discussed in Appendix 3.
Ed Denison (1915-1992) expressed opposition to changes in national accounting which treat
accretions of knowledge as investment. He considered this a “misclassification” which made
“growth analysis chaotic” (see Denison, 1989, p. 10). A major justification for his complaint was
that his growth accounts included “human capital”, . increments in the quality of the labour force
due to increases in the level of education.
In fact, the only form of knowledge which is now treated as investment is computer software. It is
odd to treat this rapidly depreciating knowledge as investment, whilst ignoring the more durable
influence of books and education.
(ii) Purchasing Power Converters for Cross-country Comparison of GDP Levels
Once standardised accounts of real GDP growth in national currencies had been established for all
OEEC countries, the next step to facilitate inter-country comparison and multi-country aggregation
was the development of purchasing power parity converters (PPPs) to measure real GDP levels,
rather than relying on exchange rate comparison. The first OEEC study, co-authored by Milton
Gilbert and Irving Kravis (1916-92). appeared in 1954, a second in 1958. They compared real
expenditure levels in 8 OEEC countries. A third volume by Paige and Bombach (1959) compared
real output levels in the UK and USA. Kravis and his colleagues, Alan Heston and Robert Summers
improved the methodology of PPP estimation in their ICP project at the University of Pennsylvania
from 1968 onwards.
The OEEC studies were binary comparisons between pairs of countries. The three options were i) a
Paasche PPP, with “own-country” quantity weights; ii) a Laspeyres PPP with the quantity weights
of the numeraire country-the United States; iii) a compromise geometric (Fisher) average of the first
two measures. The corresponding measures of real expenditure were: i) Laspeyres comparisons of
GDP levels based on the prices (unit values) of the numeraire country; ii) Paasche level
comparisons based on “own-country” prices (unit values); iii) a Fisher geometric average of the two
measures. Binary comparisons, . Germany/USA and UK/USA, could then be linked with the
USA as the star country. Such star comparisons could provide a proxy Germany/UK comparison,
but it was not “transitive” (. the result would not be identical to that derived from a direct
Germany/UK comparison). This was not a great drawback for OEEC countries where the
inter-country deviation in performance levels was not too wide. But Kravis, Heston and Summers
were engaged in comparisons over a much wider range of per capita income. They therefore
adopted the Geary-Khamis method, invented by Roy Geary (1896-1983) and Salem Khamis, which
multilateralised the results, provided transitivity and other desirable properties. They used it in
conjunction with the commodity product dummy method (CPD), invented by Robert Summers, for
filling holes in the basic dataset. Their masterpiece was their third study, the 1982 volume World
Product and Income, which contained estimates for 34 countries (in Africa, the Americas, Asia and
Europe) in 1975 prices and international Geary-Khamis dollars. These countries accounted for 64
per cent of world GDP in 2001.
Table 1 Nature of PPP Converters to Estimate GDP Levels in the Benchmark Year 1990
(billion 1990 Geary-Khamis dollars and number of countries)
Europe & Latin Asia Africa World
W. Offshoots America
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 (2003), p. 230
The UN Statistical Office extended the ICP work and had covered 84 countries by 1985. UNSO
then dropped this endeavour, though some of the regional UN bodies continued with it. The OECD
recommenced its comparisons on a regular basis in 1982. Its latest work covered the 28 OECD
countries and 20 others (in Eastern Europe, the 15 successor states of the USSR, and Mongolia).
Alan Heston and Robert Summers produce short-hand estimates of PPPs and real income levels for
countries for which full-scale ICP type measures are not available. As a result, we now have
reasonably acceptable PPP adjusted measures available for over 99 percent of world GDP,
There were three Eurostat estimates (for 1980, 1985 and 1993) of PPPs for 22 African countries, but
the results were erratic, and I preferred to use the more comprehensive and plausible results of the
Penn World Tables (see Summers and Heston, October 2002). Table 1 summarises the nature of the
PPP estimates I used to create my 1990 benchmark estimates of world GDP.
Table 2 World's 10 Largest Countries: Comparative Ranking, 1950 & 2001, at constant
1990 prices, using 1990 Geary-Khamis PPP converters and 1990 exchange rates
1950 2001 1950 2001
GDP $billion, with 1990 PPP conversion $billion, with 1990 exchange rate
USA 1,456 7,966 1,456 7,966
China 240 4,570 47 886
Japan 161 2,625 206 3,358
India 222 2,003 62 558
Germany 265 1,537 337 1,951
France 221 1,258 261 1,491
UK 348 1,202 363 1,253
Italy 165 1,101 191 1,272
Brazil 89 990 58 638
Russia 315 791 154 388
GDP per head 1990 PPP $ 1990 ER $
USA 9,561 27,948 9,561 27,948
China 439 3,583 85 541
Japan 1,921 20,683 2,458 26,466
India 619 1,957 172 545
Germany 3,881 18,677 4,928 23,717
France 5,271 21,092 6,244 24,985
UK 6,939 20,127 7,266 20,985
Italy 3,502 19,040 4,046 21,996
Brazil 1,672 5,570 1,077 3,588
Russia 3,086 5,435 1,515 2,669
Source:Maddison (2003)
Table 2 shows the difference between PPP and exchange rate conversion for the world’s 10 largest
economies (which represented 65 percent of world GDP in 2001). The exchange rate conversions on
the right hand side show much lower levels for the poorer countries (China, India, Russia and
Brazil) and somewhat higher levels for the west European countries and Japan relative to the USA
than the PPP converters. In the case of China the deviation the exchange rate and the PPPs was very
large-purchasing power was more than 5 times higher than the exchange rate. In India the ratio was
more than 3 times higher, in Russia twice as high and in Brazil more than 50 per cent higher. In
Japan and the west European countries, the exchange rate overvalued the purchasing power relative
to the US dollar. In fact the big differential for poorer countries is a fairly systematic outcome in
such comparisons. For the west European countries and Japan the differential is smaller and has
varied above and below parity in the past two decades. The implausibility of exchange rate
conversion is clear when we at the results for 1950 where exchange rate conversion implies a per
capita GDP of $85 in China and $172 in India (both in 1990 prices). These levels are much too far
below subsistence to be credible.
There has been reluctance on the part of many poorer countries to accept PPP conversion, because
they felt it might weaken their case for foreign aid or favourable loan programmes of the IDA type
(the cheap loan window of the World Bank). In fact, the World Bank provided substantial financial
support for the ICP programme, and continues to support this type of research, but has generally
avoided explicit use of PPP adjusted measures in its analytical work or in its loan decisions.
In spite of the creeping acceptance of PPP adjusted estimates, there continues to be significant error
in comparative economic analysis because of ignorance of the pitfalls of exchange rate conversion.
This is true in journalism, in political discourse, and also amongst some economists. Newspapers
frequently refer to Japan as the world’s second largest economy, though its GDP is less than 60% of
the Chinese, and some British politicians continue to believe that their economy is bigger than
China’s (1).
In this situation, it is highly desirable that statistical offices be more vigorous in explaining the
merits of PPP adjustment and in pushing for reinvigoration of this work on a worldwide basis.
(iii) Reasons for the Worldwide Adoption of Macro-measurement since 1950
The main reason for the massive increase in coverage and quality of official national accounts from
1950 onwards was the realisation of their usefulness as a tool of macroeconomic policy. Denison,
Gilbert, Kaldor, Kuznets, Ruggles, Stone, and others in the UK and USA, knew from personal
experience that such accounts were also an extremely important tool for resource mobilisation in
wartime (3). In the 1950s, Keynesian analysis had a powerful influence on economic policy in many
Western countries and its fundamental concern was with macroeconomic magnitudes (Keynes was
the godfather of the first British accounts created by his pupils, Meade and Stone). Harold
Macmillan, soon to be prime minister, discovered national accounts in 1956, when he became
Chancellor of the Exchequer. He compared them to a railway timetable, without which you
wouldn’t know when the trains were running.
This new macroeconomic perspective was very different from pre-war conceptions, echoed by
Hayek and Schumpeter. The latter considered “total output a figment which, unlike the price level,
would not as such exist at all, were there no statisticians to create it. We seem indeed to be faced by
a meaningless heap--- for most purposes, a highly inconvenient composite” (Schumpeter, 1939, pp.
484, 561).
The operational significance of national accounts became obvious in OEEC, when Milton Gilbert
became responsible for economic policy analysis from 1955 to 1961, and greatly improved its
quality. His earlier work on national accounts was the bedrock on which analysis of comparative
growth performance was based. It provided a yardstick for assessing the success of policy which
had never existed before. We served as the secretariat for a new Group of Economic Experts which
included Otmar Emminger from the Bundesbank, Etienne Hirsch, head of the French Plan, Jan
Tinbergen from the Netherlands, Arthur Burns, chairman of the US Council of Economic Advisors,
and Robert Hall, chief advisor to the UK Treasury. In 1955, Hall described the significance of their
work as follows: “These meetings are really something quite exceptional for economists and I
should think are quite new in the history of the world, in the sense that economic experts, if they
existed at all as Government advisers, were not generally very important people until Keynes's ideas
had been commonly accepted in the West. So that there were not the people to meet as we do: now
we have 7 or 8 or 9 people who are by and large the chief professional advisers of the main Western
Governments --- all have more or less the same professional training in that they understand how to
maintain the level of activity and what forces operate on it.” (Cairncross, 1991, ).
II
Quantifying and Interpreting World Economic Growth in the Capitalist Epoch
Simon Kuznets (1901-85) did more than anyone else to push back the quantitative time horizon
beyond 1950 by promoting the development of historical evidence on “modern” economic growth
and interpreting its driving forces. He convinced many of his distinguished students, and an
international network of scholars, that such an exercise was feasible, exciting and important. In
1940, Colin Clark had estimates of GDP growth for 16 countries. Most were rough proxies, and the
average number of years covered was 19. By 1971, Kuznets had assembled estimates for 21
countries, for an average around 75 years. Maddison (2003) included 163 countries for 1950, 52 for
1913, and 25 for 1820. As a result of these efforts we now have quantitative evidence for 79% of
world GDP for 1820, and 99% in 1950. This has revolutionised the analytical scope of economic
history by giving it a quantitative underpinning.
Table 3 Coverage of Maddison (2003) GDP Sample, percent of Regional and World GDP
1500 1700 1820 1870 1913 1950
W. Europe
W. Offshoots 0 99 100
E. E. & f. USSR 0 0
Latin America 85
Asia
Africa 0 0 0 0
World
Source: Maddison (2003), p. 226
(i) Characteristics of the Kuznetsian Epoch back to1820: What have we learned?
a)Kuznets’ evidence was fairly Eurocentric. He was not able to measure world performance. We
now have a much broader range of evidence on growth and levels. Table 4 shows the long-term
divergence in income levels between the advanced capitalist group (the “West”) and the Rest.
Table 4 Levels of Per Capita GDP: World and Major Regions, 1500-2001
(1990 international dollars)
1500 1820 1870 1913 1950 1973 2001
Western Europe 771 1 204 1 960 3 458 4 579 11 416 19 256
Western Offshoots 400 1 202 2 419 5 233 9 268 16 179 26 943
Japan 500 669 737 1 387 1 921 11 434 20 683
West 702 1 109 1 882 3 672 5 649 13 082 22 509
Asia (excluding Japan) 572 577 550 658 634 1 226 3 256
Latin America 416 692 681 1 481 2 506 4 504 5 811
Eastern Europe & f. USSR 498 686 941 1 558 2 602 5 731 5 038
Africa 414 420 500 637 894 1 410 1 489
Rest 538 578 606 860 1 091 2 072 3 377
World 566 667 875 1 525 2 111 4 091 6 049
Interregional Spread :1 :1 :1 :1 :1 :1 :1
West/Rest Spread :1 :1 :1 :1 :1 :1 :1
Average per capita income of the West rose 20-fold between 1820 and 2001, and less than nine-fold
in the Rest. The spread between the two groups rose from 2:1 to nearly 7:1and the inter-regional gap
increased much more from 3:1 to 18:1. Nevertheless, it is clear from Table 5 that the Western share
of world GDP has peaked and will in all probability fall considerably more if the two Asian tigers,
India and China maintain a high growth momentum;
Table 5 Levels of GDP: World and Major Regions, 1500-2001
(billion 1990 international dollars)
1500 1820 1870 1913 1950 1973 2001
Western Europe 1 396 4 096 7 550
Western Offshoots 1 635 4 058 9 156
Japan 161 1,243 2,625
West 1 3 193 9 398 19 331
Asia (excluding Japan) 823 2 623 11 481
Latin America 416 1 398 3 087
E. Europe & f. USSR 695 2,064 2,072
Africa 203 550 1 222
Rest 1 2 137 6 626 17 862
World 1 2 5 330 16 024 37 194
% West/World
Source for tables 4 and 5: Maddison (2003), pp. 259 & 262. W. Offshoots are Canada, USA,
Australia & New Zealand.
b) the evidence now available suggests that the transition to accelerated growth started around 1820,
not 1760 as Kuznets thought. The work of Crafts and others on British performance in the
eighteenth century helped demolish the old notion of a sudden take-off in the second half of that
century. The important point about Britain’s exceptionalism is not an industrial revolution, but a
much longer process of ascension, with per capita growth faster from 1500 onwards than in the rest
of Europe;
c) the acceleration in Western Europe was synchronous, not staggered as Gerschenkron and Rostow
believed. Hansen's (1974-76) work on Denmark showed evidence of substantial advance in the early
nineteenth century; Tilly (1978) found the same for Prussia; Levy-Leboyer & Bourguignon (1985)
and Toutain (1987) for France; Hjerppe and Associates (1987) for Finland; Krantz (1988) for
Sweden; Hodne and Grytten (1994) for Norway; Smits, Horlings and van Zanden (2000) for the
Netherlands. Their research strongly suggests that the acceleration of economic growth was quite
general in Western Europe after the Napoleonic wars. It was slower in 1820-70 than it became in
1870-1913. Nevertheless the pace of advance in Western Europe in 1820-70 was clearly much faster
than in the eighteenth century and earlier.
d) Kuznets (1930) demolished the Kondratieff notion of long cycles and Kuznets (1940) found
Schumpeter’s cyclical schema unacceptable-“a mere recording of impressions of charts, impressions
with which it is often difficult to agree.” Technical progress did not come in big Schumpeterian
waves, but was a smoother more diffused process: “flowing in a continuous stream, a stream
magnified in a constant proportion by the efforts of imitators” (Kuznets 1940). This way of thinking
he transmitted to his students, Schmookler and Fogel, who gave it fuller articulation. Kuznets
concentrated on performance in the period as a whole, but we now have enough evidence is possible
to discern five phases from 1820 to 2001 in which the momentum of growth and fashions in
economic policy differed substantially (see table 6).
Table 6 Per Capita GDP Growth : World and Major Regions, 1500–2001
(annual average compound growth rates)
1500–1820 1820–70 1870-1913 1913–50 1950–73 1973–2001
Western Europe
W. Offshoots
Japan
West
Asia (ex Japan) – –
Latin America
E. Europe & USSR –
Africa
Rest
World
Source : Maddison (2003), p. 263.
The years 1950-1973 were a golden age of unparalleled prosperity. World GDP rose at an annual
rate of 5 per cent, per capita GDP near 3 per cent and world trade almost 8 per cent a year. There
was a significant degree of convergence in per capita income, with most regions growing faster than
the USA (the lead economy). After 1973, there was a marked slowdown in world growth, with
substantial divergence between different regions, and performance in many of them below potential.
Nevertheless, on a world basis this latest phase was the second-best since 1820. It is clear that
“modern economic growth”, in all its phases, has been much faster than in the preceding centuries.
From the year 1500 to 1820, world per capita income rose .05 per cent a year. From 1820 to 2001, it
averaged per cent, nearly 25 times as fast;
e) it is important to distinguish between lead and follower countries to understand the dynamics of
technological diffusion, and analyse processes of catch-up and falling behind. “Lead” countries are
those whose economies operate nearest to the technical frontier; “follower” countries have a lower
level of labour productivity (or GDP per capita). Since 1500 there have been four lead countries,
Northern Italy in the sixteenth century, the Netherlands from the sixteenth century until the
Napoleonic wars, when the UK took over. The British lead lasted until around 1890, and the USA
has been the lead country since then.
(ii) Quantifying the Causes of Growth
As quantitative evidence on comparative GDP growth has accumulated, it has became feasible to
sharpen analysis by quantifying the reasons for inter-temporal and interspatial variance in
performance.
The first step in growth accounting was to measure labour input and productivity. Labour input has
grown unevenly over time and between countries. It has been very different from the movement of
population. Since 1820, labour input has increased less than population; and labour productivity a
good deal faster than GDP per capita.
Early post-war analysts laid great stress on the role of capital in economic growth, though for lack
of accurate information, some assumed that the capital-output ratio was stable, some used
incremental investment-output ratios (ICORs), or wealth surveys, insurance valuations or company
book-keeping or stock exchange values as a proxy. A major breakthrough came when Goldsmith
(1951) pioneered the “perpetual inventory” method in which stock estimates were derived by
cumulating historical series on past investment at constant prices, and deducting assets scrapped,
written off or destroyed by war. In the course of the 1970s and 1980s, several OECD countries
developed official stock estimates on a perpetual inventory basis, when they had accumulated a long
enough run of investment data to permit their construction. Academic researchers such as Feinstein
(1988) and Gallmann (1986 and 1987) pushed these capital stock estimates much further back in
time.
These modern official estimates are usually similar conceptually but are need adjustment because of
different assumptions about asset lives. I made standardised estimates of fixed non-residential
capital for France, Germany, Japan, the Netherlands, UK and USA in Maddison (1995c) broken
down into structures, and machinery. This is a very pertinent distinction, as the rate of growth of the
latter component has been much faster than the former, and technical progress is probably more
rapidly embodied in machinery investment than in structures.
Schultz (1961) suggested that inputs of “human capital” should also be regarded as a factor of
production. The main component he had in mind was the increase in the educational levels, but
improvements in skill through working with sophisticated equipment, and improvements in health
were also relevant. The idea proved attractive and measures of joint factor productivity were soon
constructed in which education was treated as part of factor input. In growth accounts, the normal
procedure is to treat increases in education as an improvement in labour quality, rather than as an
independent factor of production analogous with physical capital (3).
Table 7 a Determinants of Growth: UK, USA and Japan, 1820-1998
UK USA Japan UK USA Japan
Gross Stock of Machinery and Equipment Per Capita(1990$) Gross Stock of Non-Residential
Structures Per Capita (1990 $)
1820 92 87 . 1,074 1,094 .
1870 334 489 94a 2,509 3,686 593
1913 878 2,749 329 3,215 14,696 852
1950 2,122 6,110 1,381 3,412 17,211 1,929
1973 6,203 10,762 6,431 9,585 24,366 12,778
1998 11,953 25,153 29,987 21,066 35,810 49,042
Primary Energy Consumption Per Capita (tons of oil equiv.) Average Years of Education Per
Person Employed*
1820 .61
1870
1913
1950
1973
1998
Land Area Per Capita (hectares) Exports Per Capita (1990 $)
1820 53 25 0
1870 390 62 2
1913 862 197 33
1950 781 283 42
1973 1,684 824 875
1998 4,680 2,755 2,736
Hours Worked Per Head of Population GDP Per Manhour (1990 $)
1820 1,153 968 1,598
1870 1,251 1,084 1,598
1913 1,181 1,036 1,290
1950 904 756 925
1973 750 704 988
1998 657 791 905
Source for tables 7a and 7b: Appendix K of MADDISON (1995,PP. 252-55), amended and updated.
a) 1890; b) 1850; * equivalent years of primary education.
Denison (1962) created expanded “growth accounts” to explain twentieth-century American
economic performance. In 1967 he applied the technique to explain differences in growth rates and
levels of achievement in eight West European countries and the USA for 1950-64. Denison and
Chung (1976) incorporated Japan into the sample.
Tables 7a and 7b present Denison-type growth accounts for the UK and USA back to 1820, and for
the whole period of Japan’s “modern economic growth”. They show:
Table 7b Capital/Output Ratios, Growth of Labour &Total Factor Productivity
UK USA Japan UK USA Japan
Capital-Output Ratio
Machinery & Equipment/GDP Capital-Output Ratio
Non-Residential Structures/GDP
1820 .05 .07 . .63 .87 .
1870 .11 .20 .10a .79 .59a
1913 .18 .52 .24 .65 .61
1950 .31 .64 .72 .49
1973 .52 .64 .93 .80
1998 .64 .92
Labour Productivity Total Factor Productivity
(annual average compound growth rates)
1820-70 .
1870-1913
1913-50
1950-73
1973-98
a)1890; b) 1890-1913
a) A huge increase in the stock of physical capital, significant for non-residential structures, but
sensational for machinery and equipment. The ratio of the latter to GDP rose 13-fold in the UK and
USA between 1820 and 1998, and nearly 14-fold in Japan from 1890. This increase was linked to
the acceleration of technical progress, much of which had to be embodied in machinery;
b) The education level rose nearly eightfold in the UK, eleven-fold in the USA and Japan. This
increase in human capital, measured by years of formal educational experience of those in
employment (weighted by the earnings differential associated with years of primary, secondary and
tertiary) was also linked to technical progress. The increasing complexity of production processes
required better educated people to make it operational, and the involvement of educated people in
R&D helped institutionalise the process of innovation.;
c) Labour input per head of population dropped by 40 percent in the UK and Japan, 20 percent in
the USA;
d) International specialization: the ratio of foreign trade to GDP rose from 3 to 25 percent in the
UK, from to 13 percent in Japan and from 2 to 10 per cent in the USA;
e) Natural resource scarcities not a constraint; land area per capita fell fourteen-fold in the USA,
about four-fold in Japan and the UK;
f) The increase in energy inputs was relatively modest-a threefold increase per capita in the USA,
six-fold in the UK and eightfold in Japan. However, the composition of energy inputs changed
drastically. In 1820, 94 per cent came from organic matter. By 2001, this had dropped to 11 percent.
III
Economic Performance in the Merchant Capitalist Epoch: 1500-1820
Quantitative investigation for this earlier period has been relatively neglected for three reasons a) it
is clear that growth was much slower than it has been in the last two centuries; b) quantitative
evidence is harder to find, and c) many thought the results would be uninteresting-a long litany of
stagnation interrupted by catastrophe.
(i) Divergent Interpretations of Merchant Capitalist Epoch 1500-1820
There were already two very different views at the end of the eighteenth century. Adam Smith
(1776) took a mildly euphoric position and Malthus (1798) was deeply pessimistic.
Positive: Adam Smith (1776), argued that the discovery of the Americas and southern route to Asia
opened up new and significant opportunities for economies of scale and specialisation through
international trade. Though these possibilities were not fully exploited because of mutually hostile
trade restrictions, Smith was mildly euphoric about progress achieved. He did not quantify growth
performance explicitly, but arrayed countries in descending level of achievement: Netherlands,
England, France, North American colonies, Spanish America, China, Bengal and Africa. For him,
policy and institutions were a major reason for this inter-country variance.
Negative: For Robert Malthus (1798) performance depended on the balance between of population
growth and a fixed supply of land. Gains from technical progress, capital formation and
international specialisation were ignored. Balance was achieved by catastrophes-war, famine and
disease.
The dichotomy persists. Cipolla, Kuznets, Jones, Landes, Jan de Vries and Maddison take a view
similar in tone to Smith, but there has been a raft of latter-day pessimists.
LeRoy Ladurie, a French Malthusian (1966 & 1978), thought the French economy was stagnant
from 1300 to 1720. “Real” wage pundits are more pessimistic. Phelps Brown and Hopkins (1956)
suggested that English living standards in 1820 were 44 per cent lower than in 1500. Wilhelm Abel
(1978) suggested that such a drop was characteristic for the whole of Western Europe. These
judgements were endorsed by Bairoch, Braudel, Wrigley and Schofield, but later they switched
sides.
A new wave of real wage pessimism has been launched by Robert Allen (2001) and Jan Luyten van
Zanden (1999 & 2002). Allen finds negative per capita growth for 1500-1820, van Zanden less than
half the growth I find (see Appendix 1).
Most of the pessimist literature is Eurocentric, but Susan Hanley and Kenneth Pomeranz claim,
respectively, that Japan and China had living standards equal to those in the UK early in the 19th
century (4). They imply or suggest that there was no significant European ascension in 1500-1820.
(ii) Testing The Kuznets Conjecture
In 1965, Kuznets advanced an influential conjecture about the rate of population and per capita GDP
growth in Western Europe “from the end of the fifteenth to the second half of the eighteenth
century”. Judging from the demographic evidence then available (Carr-
Table 8a Per Capita GDP Levels 1500 and 1820
(1990 G-K $) (1990 G-K $)
Dynamic Countries & Regions Less Dynamic Countries & Regions
1500 1820 growth r. 1500 1820 growth r.
Belgium 875 1,319 China 600 600
France 727 1,135 India 550 533
Germany 688 1,077 Japan 500 669
Italy 1,100 1,117 Other Asia 565 584
Netherlands 761 1,838 All Asia 568 581
Portugal 606 923
Spain 661 1,008 Russian Empire 499 688
Ireland 526 880 Eastern Europe 496 683
Britain (ex Ir) 762 2,121
Other W. E. 650 1,051 Egypt 475 475
All W. Europe 771 1,204 Other N. Africa 430 430
Brazil 400 646 Black Africa 405 415
Mexico 425 759 All Africa 414 420
Caribbean 400 636
Other L. A. 412 683 Australia & NZ 400 490
USA & Canada 400 1,231
All Americas 415 1,148 World 566 667
Table 8b GDP Levels, 1500 & 1820
(billion 1990 G-K $) (billion 1990 G-K $)
Dynamic Countries & Regions Less Dynamic Countries & Regions
1500 1820 growth r. 1500 1820 growth r.
Belgium China
France India
Germany Japan
Italy Other Asia
Netherlands All Asia
Portugal
Spain Russian Empire
Ireland Eastern Europe
Britain (ex Ir)
Other W. E. Egypt
All W. Europe Other N. Africa
Brazil Black Africa
Mexico All Africa
Caribbean
Other L. A. Australia & NZ
USA & Canada
All Americas World
Source: Maddison (2001), Appendix B, and Maddison (2003), p. 242-263.
Saunders, 1936 and Urlanis, 1941), the work of Deane and Cole (1962) on British per capita income
growth in the eighteenth century, and adjusting for the likelihood of better-than average
performance in the UK, which had also had faster demographic advance, he “set the possible (and
perhaps maximum) long-term growth in per capita product for 1500-1750 in developed countries of
Western Europe at about per cent per year”. He felt that a higher rate was unlikely as the 1750
level was low, and that a lower growth rate was plausible. Kuznets did not advance a conjecture
about growth rates in the rest of the world, but he clearly thought that they were lower than in
Western Europe, and that their 1750 level was lower than that in Europe.
(iii)The Nature of the Quantitative Evidence on Economic Performance 1500-1820
Maddison (2001) involved a major effort to test the Kuznets’ conjecture and muster quantitative
evidence on world economic performance before 1820, as summarised below:
a) for Western Europe: I encouraged other researchers to extend their time horizon backward by
interactive networking of the type which proved so fruitful in building up evidence for 1820
onwards. Between 1985 and 1994, I organized six workshops on quantitative economic history (two
in Groningen, two IARIW, and two IEHA sessions). Most of the papers involved exploration of
pre-modern growth in Western Europe and evidence was mainly from production and expenditure
side. As can be seen in Table 8, the new estimates show per capita growth averaging % a year
between 1500-1820, significantly slower than Kuznets’ (1965) hypothesised. Details of the
estimation procedure and the conjectures used to fill gaps in the GDP database can be found in
Maddison (2001), Appendix B. Thanks to the work of modern demographic historians, the quality
and coverage of estimates of population levels and movement have been greatly improved, and also
yield useful corroborative evidence on changes in urbanisation and life expectation-where European
exceptionalism is clear.
`Jan de Vries made 3 major contributions a) to the history of expenditure patterns; b) the
comparative study of European urbanisation; c) stressing the increase in per capita labour
inputs-productivity grew more slowly than per capita income.
b) For the USA & Canada; Australia & New Zealand; Brazil & Mexico, I adopted a multicultural
estimate of the type pioneered by Noel Butlin, making separate estimates for the indigenous
population, slaves, and white settlers. For the first two groups I used a stylised per capita income of
$400 intended to represent an income at near-subsistence level. For the Caribbean, with its highly
specialised export economies, I based the estimates on commodity production and exports. For the
Americas as a whole, per capita GDP growth was faster than in Western Europe.
c) For Africa, there was a sharp division north and south of the Sahara. Egypt’s production potential
was favoured because of the Nile, Egypt and the Maghreb had a higher degree of urbanisation and
literacy, more sophisticated economic and political institutions, and a greater participation in
international trade than black Africa. In spite of losses due to the slave trade, demographic
expansion was faster in black Africa, because agriculturalists were replacing hunter-gatherers, and
had new crops-maize and manioc from the Americas.
d) For China, India and Japan, the estimates are based on production, expenditure patterns, and
demographic evidence. Japanese per capita performance was better than Chinese or Indian, but for
Asia as a whole, income levels were stagnant. It is clear however, that China had extensive growth
in this period. It sustained a large increase in population, without a fall in living standards, and its
GDP growth was as big as in W. Europe.
China: China had a strong physiocratic bureaucracy which kept printed records on population and
agricultural performance back to the ninth century. There is also a great deal of scholarly work,
which I was able to use in writing Maddison (1998). This included Needham’s (1954-97 and 1970)
massive work on the development of Chinese technology, Ho (1959) on Chinese demography, the
interpretative analysis of Balazs (1931-33), Elvin (1973) on the economic history of the Tang and
Sung dynasties and Perkins (1969) on agricultural development from 1368. Grain output rose about
fivefold from 1400 to 1820-in line with population. The cultivated area rose threefold and yields
about 80 per cent. Ester Boserup (1965) demonstrated that this was achieved by increased labour
inputs per capita, and more intensive use of land by double-cropping, improved seeds, and
fastidious collection and application of manure and the introduction of new crops from the
Americas.. Rozman’s analysis of the demographic records shows no significant change in the
relative size of the urban population over this period. Earlier, it is clear that that China did
experience a growth in per capita agricultural output and GDP in the Sung dynasty (960-1280)..
India: Maddison (1971) contained an analysis of the social structure and institutions of the Moghul
empire and the British raj. For the Moghul period, I relied heavily on the economic survey made by
Abul Fazl for the emperor Akbar at the end of the sixteenth century (see Jarrett and Sarkar, 1949).
Between 1600 and the 1860s, the quantitative evidence is not so good, but the two leading historians
of Moghul India, Habib and Moosvi (at Aligarh Muslim University), adduce evidence which led
them to conclude, I think rightly, that there was some decline in per capita income after the collapse
of the Moghul Empire and the takeover by the East India Company.
Japan modelled its economy, society, literature and institutions on China from the seventh century.
The official commitment to catch up with the West started in 1867, but the Chinese model was
abandoned in the eighteenth century, at about the time Japan had caught up with China. In 1720 the
shogun lifted the ban on European books, and translations of Dutch learning (rangaku) had a
significant impact in transmitting knowledge of European science and technology to Japan. (see
Maddison (2001, pp. 204-206 and 252-260).
(iv) The Proximate Causes of Growth in the Merchant Capitalist Epoch
In analysing the proximate causes of growth in the merchant capitalist period, it is not possible to
present growth accounts, as I did in tables 7a and 7b for the Kuznetsian period. De Vries and
Boserup have shown that labour input per person employed increased in this period, instead of
declining, as it did later. There was a big increase in capital formation in shipping, an improvement
in human capital and knowledge, and it is very clear that the process of globalisation was much
more important in this period than it has been since. The evidence for this is summarised in Box 1.
Box 1 Trade, Ecological and Technical Interactions between Continents
Dramatic progress in Western shipping and navigation permitted a 20-fold increase in world trade
between 1500 and 1820. It brought gains from specialization of the type stressed by Adam Smith. It
provided European consumers with new products-tea, coffee, cacao. sugar, potatoes, tobacco,
porcelain, silk and cotton textiles. In relative terms this globalisation process was a more important
component of growth in these centuries than in the twentieth. European countries were also able to
extract a colonial surplus: the Spanish and Portuguese from the Americas in the sixteenth century,
the Dutch in Asia from 1600, the British and French in the eighteenth century. Spanish plunder was
mainly in the form of precious metals. These were very important in financing European trade with
Asians, who were not interested in buying European products. Most of the European trading nations
profited from the enslavement of Africans.
Globalisation: Comparative Growth in Volume of World Trade and GDP
(annual average compound growth rates)
World Trade World GDP
1500-1820
1820-70
1870-1913
1913-50
1950-73
1973-2001
1820-2001
Ecological, Technical and Demographic Transformation of the Americas
a) agricultural potential increased by introduction of wheat, rice, sugar, coffee, vines, olives, onions,
cabbages, lettuce, oranges, bananas, yams, cattle, pigs, chickens, sheep, goats;
b) traction and transport improved by introduction of horses, oxen, asses, mules;
c) European mining technology leads to production and export of 1,700 tons of gold and 73,000 tons
of silver 1500-1820 which financed European trade with Asia;
d) introduction of iron weapons, tools and ploughs, wheeled vehicles, ships and shipbuilding,
printing, literacy, education, political and economic institutions;
e) European diseases killed off two-thirds of the indigenous inhabitants. The continent was
repopulated by African slaves and European migrants attracted to a continent with much greater
land resources per capita. In 1820, 37 % of the population of the Americas were indigenous or
mestizo; 38% were white, and 22% black or mulatto.
Ecological Gains for Africa and Asia
Transfer of American crops: maize, manioc, potatoes, sweet potatoes, haricot, lima and string beans,
peanuts, tomatoes, chilis, pineapples and cacao enhanced production potential.
(v)Development of Secular Knowledge and Science
Looking beyond these proximate elements of causality, Kuznets stressed the importance of secular
knowledge and science as the basic cause of Western economic ascension. From about 1500 there is
evidence of a new awareness of human capacity to transform the forces of nature through rational
investigation and experiment which had no counterpart in other parts of the world in this period.
The first European university was created in Bologna in 1080. By 1500 there were 70 such centres
of secular learning in Western Europe (see Goodman and Russell, 1991, p. 25). Until the
mid-fifteenth century, most of the instruction was oral, and the learning process was similar to that
in ancient Greece. Things changed after Gutenberg printed his first book in Mainz in 1455. By 1500
220 printing presses were in operation throughout Western Europe and had produced eight million
books (see Eisenstein, 1993, -17). The productivity of universities and their openness to new
ideas was greatly enlarged.
Venetian publishers regularly had a print-run of 1,000 copies or more. By the middle of the
sixteenth century, they had produced some 20,000 titles, including music scores, maps, books on
medical matters, and a flood of new secular learning. Before printing, books were cherished for their
artistic or iconic value, and their content mainly reflected the wisdom and dogma of the past.
Printing made books much cheaper. Publishers were much more willing to risk dissemination of
new ideas and to provide an outlet for new authors. The proportion of the population with access to
books was greatly increased, and there was a much greater incentive to acquire literacy. With the
exception of China, the European printing revolution had no counterpart in most other parts of the
world until the beginning of the nineteenth century. The major difference between Europe and
China was the competitive character of European publishing, and the international trade in books.
This frustrated the attempts of the Papacy to achieve thought control through the Inquisition and
censorship. China was a centralized state, with vestigial foreign contacts. The education of its
bureaucracy was devoted to ancient classics, and they were able to exercise thought-control by more
subtle and effective methods than the Papacy in Europe.
Fundamental changes in intellectual horizons occurred between the sixteenth and seventeenth
centuries, when medieval notions of an earth-centred universe were abandoned. Thanks to the
Renaissance, the seventeenth century scientific revolution and the eighteenth century enlightenment,
Western elites gradually abandoned superstition, magic, and submission to religious authority. The
scientific approach gradually impregnated the educational system. Circumscribed horizons were
abandoned. A Promethean quest for progress was unleashed. The impact of science was reinforced
by the creation of scientific academies and observatories which inaugurated empirical research and
experiment. Systematic recording of experimental results and their diffusion in written form were a
key element in their success.
The Transition from Merchant Capitalism to Modern Economic Growth
Having considered the quantitative evidence on macroeconomic performance in the epochs of
modern economic growth and merchant capitalism, and the differences in the driving forces which
underlay their growth momentum (table 8 and Box 1), it is useful to consider the nature of the
transition between the two epochs. There is in fact a sharp divergence of views on the “roots of
modernity”, which echoes the divergence already noted between the Smithian and Malthusian
interpretations of what happened in the merchant capitalist epoch. Box 2 summarises this
dichotomy.
Box 2 The Roots of Modernity, Technological “Takeoff” or Long Apprenticeship?
There is a school of thought which attributes modern economic growth to an “industrial revolution”
in Manchester, preceded by centuries of Malthusian stagnation. The metaphor was first popularised
by Arnold Toynbee in 1884, and has continuing resonance, . in Rostow’s (1960) “take-off”, and
Mokyr’s (2002) history of technology-“most techniques before 1800 emerged as a result of chance
discoveries…Before the industrial revolution the economy was subject to negative feed back…..the
best known of these negative feedback mechanisms are Malthusian traps.” Nordhaus (1997) and
DeLong (1998) overdosing on hedonics, have constructed fairytale scenarios which greatly
exaggerate progress since 1800, before which people lived like cavemen. These views are
fundamentally wrong.
In 1500, Western Europe already had 70 universities. Education and diffusion of knowledge were
revolutionised by printing. Venetian publishers regularly had a print-run of 1,000 copies or more.
By mid 16th century, they had produced 20,000 titles, with a flood of new secular learning. Forty
years earlier, a scribe would have taken a year to produce one volume.
By the end of the eighteenth century, great progress had been made in the design of ships and
rigging, in gunnery, in meteorological and astronomical knowledge and in the precision of
navigational instruments. Mariners acquired logarithmic tables, sextants, naval almanacs and
accurate watches. Maps were enormously improved and supplemented by detailed coastal surveys,
knowledge of winds and currents. Sailing had become safer, the duration of voyages more
predictable, the incidence of shipwreck had fallen, disease mortality was greatly reduced on long
voyages.
These changes were the result of scientific endeavour. In 1543 Copernicus rejected the notion that
the earth was the centre of the universe. Kepler and Galileo made detailed observation of celestial
bodies, the nature and mutability of their orbits. Newton in 1687 showed that the whole universe
was subject to the laws of motion and gravitation. Progress in astronomy and physics was
accompanied by major advances in mathematics and design of telescopes, microscopes,
micrometers, thermometers, barometers, air pumps, clocks and watches and the steam engine.
These developments in Europe were an essential prelude to the much faster economic development
that occurred in the 19th and 20th centuries. They had no counterpart elsewhere.
Footnotes
(1) Christopher Patten, the last British governor of Hong Kong, stated in an article in the Economist
newspaper of 4th January 1997. that “Britain’s GDP today is almost twice the size of China’s”. If he
had been briefed on PPP converters, he might have said that Britain’s GDP was one third the size of
the Chinese.
(2) Kaldor (1945) concluded that “Germany made no serious attempt to exploit her own war
potential fully, except for a brief period in August and September 1944, when it was too late to be of
any consequence”. Galbraith (1971) made the same point. Kaldor’s analysis was drawn from
material gathered as a staff member of the US Strategic Bombing Survey (1945), and interrogation
of Karl Otto Saur, Albert Speer’s deputy in the Armaments Ministry. The survey team was directed
by Galbraith, and included Paul Baran, Ed Denison, Burton Klein, and Tibor Scitovsky (under the
nom-de-guerre “Thomas Dennis”). Kaldor and Scitovsky interrogated Saur in Austria, the day the
war ended (see Scitovsky, 1999). He indicated where they could find the wartime production
records, and they whisked them away just before the Russians arrived. Denison and Haraldson made
a detailed estimate of German GNP, 1936-44, to put military mobilization in perspective. Richard
Ruggles served with the US Office of Strategic Services in London, inferred German production of
tanks, trucks, and planes by decoding information on serial numbers of captured equipment (see
Tobin, 2001). Stone worked with British intelligence and predicted the date of Italian entry in the
war, by tracking movement of ships in the Mediterranean. Kuznets used national accounting to help
organise the massive expansion of US military output in the Planning Committee of the War
Production Board (see Kapuria-Foreman and Perlman (1995), with help from Moe Abramovitz. The
moral of this digression is that the initial stage in construction of national and historical accounts is
not a boring bureaucratic business. It requires detective work and imagination, and can be as
exciting as the adventures of Sherlock Holmes.
(3) Estimation of the stock of human capital is analogous to the procedure for physical capital. A
useful starting point is scrutiny of successive population censuses where respondents report the age
at which their formal education ended (see estimates for 19 countries by sex and age-cohort in
OECD, 1975, vol. 1, pp. 31-108). These can be updated by annual cumulation of increments to the
stock, (see annual school enrolment in OECD, Education at a Glance and its predecessor volumes),
and deduction for people who retire from the labour force. The value of the stock can be derived
from estimates of earnings of people with primary, secondary and higher education (see
Psacharopoulos, 1975). This is the procedure I used in growth accounts for advanced OECD
countries (see Maddison, 1987a) and for 22 developing counties (Maddison 1970, -50).
(4) Kenneth Pomeranz (2000) asserted that China was ahead of Europe until 1800. He suggests that
Western Europe was “a non-too-unusual economy: it became a fortunate freak only when
unexpected and significant discontinuities in the late eighteenth and especially nineteenth centuries
enabled it to break through the fundamental constraints of energy and resource availability that had
previously limited everyone’s horizons”. I have explained my disagreement at length in Maddison
(2003, pp. 248-251). Hanley had a similar view about Japan, see my comment in Maddison (1999).
Appendix 1 Real Wage Revivalists-Robert Allen and Jan Luiten van Zanden
In an extensive review of Maddison (2001), Giovanni Federico (2002) suggested that I may have
exaggerated West European performance in the merchant capitalist epoch, citing alternative
estimates of Robert Allen which imply that aggregate West European income per capita actually fell
in this period, and those of Jan Luiten van Zanden which imply a growth rate of only per cent a
year. I do not regard their gloomier conclusions as a serious challenge to my estimates for the
reasons explained below.
Allen (2001) presents real wage estimates for Europe, 1500-1913. He shows nominal wages (in
grams of silver per day) for building craftsmen and labourers in 18 European towns, and consumer
price indices based on 12 items, two-thirds of which were bread, beer and meat. The results are
presented for fifty-year time segments. His basic data are for daily wages, which he converts to an
annual basis by assuming a working year of 250 days (this multiplier was apparently applied
uniformly to his inter-temporal and cross-country data). For the 14 towns where he has results for
craftsmen for both 1500-49 and 1750-99, the average real wage in the latter period was 66 per cent
of that in 1500-49, London was the only case where the real wage was higher, with a rise of less
than 1 percent. In the 12 towns for which he had results for building labourers for the two periods,
there was a rise of about 3 per cent in Amsterdam and a fall everywhere else. The average real wage
for labourers was 76 percent of the 1500-49 level in the end period (p. 428). This is less gloomy
than Phelps Brown, but the clear implication is that living standards in Western Europe declined
substantially from 1500 to 1800. Allen (2000) was “an exercise in historical reconstruction based on
simple economic theory” which presented estimates of the movement in agricultural output per
capita for nine countries for 1500-1800. He shows a fall in all nine countries between these two
points of time, and in most cases they are very substantial. For England, he shows a 32 per cent drop
(see p. 19). This is very different from Wrigley’s estimate that English agricultural output per capita
doubled in the shorter period 1600 to 1800. It also differs substantially from the estimates presented
by van Zanden and Horlings (1999b, p. 28). Allen does not measure agricultural output directly. He
derives it econometrically from his estimates of occupational structure and the assumption that his
real wage measures are a valid proxy for total output per capita.
Van Zanden (2002) presents estimates of real GDP for five countries which imply that, from 1500
to 1820, average West European per capita income grew less than less than half as fast as I
estimated in Maddison (2003). Table A compares his estimates with mine.
For the UK, he shows slightly slower growth, because he uses a different source for agriculture. We
both agree that growth was most rapid in the UK, and I see no reason to modify my estimate. For
the Netherlands, our estimates are quite similar for 1570 onwards. the main difference is that he
assumes Dutch per capita income have been stagnant from 1500 to 1570, whereas I assume a
substantial increase. Between 1470 and 1570 the Dutch merchant fleet increased nearly fourfold- a
growth rate of 14 per cent a year (see of Maddison 2001), and urbanisation was increasing
substantially.
Table A Maddison Van Zanden
1500 1820 1500 1820
Belgium 875 1,319 989 1,319
Italy 1,100 1,117 1,353 1,117
Netherlands 761 1,838 1,252 1,838
UK 714 1,706 792 1,706
Spain 661 1,008 946* 1,008
Average 882 1,345 1,116 1,345
Source: cols. 1, 2 and 4 from Maddison (2003), , col. 3 derived
from growth rates implicit in the index numbers in Van Zanden (2002), p. 76.
Van Zanden expresses his estimates as ratios to the UK per capita GDP level in 1820,
as estimated by Maddison (1984 or 1994). Here, I have benchmarked
his growth rates on my 2003 estimates of 1820 levels. Asterisked figure for Spain assumes that van
Zanden’s 1580-1820 growth rate also applies for 1500-1820. The figures in the last row are
weighted averages; for 1500-1820, the van Zanden average is per cent per annum, mine is
per cent for the same 5 countries.
Van Zanden agrees with me that Dutch per capita income in 1500 was lower than in what is now
Belgium. Nevertheless his estimates show the opposite situation in 1500. To mitigate this he
adjusted the Blomme-van der Wee growth rate for Belgium downward, whereas I accepted it. Van
Zanden and I agree that Italy was the highest income country in 1500, but he assumes an 18 fall in
per capita GDP from 1500 to 1820, whereas I assumed stagnation. He quotes Malanima’s (1994)
estimate for northern Italy as his source. In fact, Malanima (1994 and 1995) suggested a fall of 7 per
cent, but van Zanden’s estimate is near that of Malanima (2003), and shows a fall of about a fifth.
The evidence for Italy as a whole is not very good and there are two schools of thought on
performance in this period. Malanima’s judgement resembles that of Cipolla (1976), whereas Rapp
(1976) and Sella (1979) argued that per capita income was stagnant from 1500 to 1820. I lean
towards their judgement, but as the urbanisation rate was slightly higher in 1800 than in 1500, I
assumed a very slight rise. For Spain, Van Zanden shows a growth rate of per cent a year for
1570-1820, which he derives by modifying the estimate of Yun (1994) for Castile, 1580-1800. In
fact, Yun’s estimate (p. 105) shows growth twice as fast as this. In Maddison (2001, p. 249), I
explained my reasons for modifying Yun’s estimate, which omits the years 1500-80 when Spain’s
economy received a major boost from the conquest of the Americas.
Van Zanden (1999) presented real wage estimates for unskilled building labourers in 14 European
cities/regions, using cereal prices (rye or wheat) as a deflator. He shows a fall in all the 10 cases
where he had estimates for 1500-20 and 1780-1800, the average for the latter period was 60 per cent
of that in 1500-20, an annual average change of per cent a year. This is more pessimistic than
the Allen (2001) results for labourers, but is similar to the findings of Phelps Brown and Hopkins
(1956, -30) for English building craftsmen for the same period. Van Zanden feels that real wage
estimates are “an important source of information on living standards”, even though they are in
sharp conflict with a large body of other evidence. He suggests that a reconciliation may be
possible. His estimates are for daily wages, and he suggests that there may have been a substantial
increase in average annual working time of labourers over the period covered and that their family
income may have been supplemented by increased labour force activity of women and children. It
seems likely that there were changes in this direction as indicated in de Vries (1993), but van
Zanden does not attempt to quantify them, and it is highly doubtful that their effect would be big
enough to achieve a reconciliation. Van Zanden’s desire for reconciliation may have introduced a
downward bias in his (2002) estimates of real per capita GDP for 1500-1820.
The founder of real wage analysis, Thorold Rogers (1823-90), was professor of economics in
Oxford and a liberal member of parliament who argued that the condition of English wage earners
could be improved by extending the franchise and encouraging trade union activity. For him, low
wages were the result of exploitation of the labourer by the ruling elite. He made a sharp distinction
between wage income and national income, as is clear from his citation of Gregory King’s estimates
of inequality in 1688 (Rogers, 1884, pp. 463-465). He summarised his position, saying ():
“society may make noticeable progress in wealth, and wages remain low …relatively speaking, the
working man of today is not so well off as he was in the fifteenth century”.
Some of the real wage revivalists have forgotten this and use real wages for a small group of
workers as a proxy for GDP per head, without considering their representativity in macroeconomic
analysis. Lindert and Williamson (1982, p. 393) show that only per cent of families derived their
livelihood from the building trades in 1688. In the Phelps Brown-Hopkins study, whose sources
were meticulously documented, there were about 3 wage quotations a year for building labourers,
and for 1500 to 1800 there were 82 years without an estimate. Over such a long period there may
have been big changes in the nature of building work, . a shift from decorative ecclesiastical
stonework to bricklaying.
Appendix 2 Joel Mokyr and the “Industrial Revolution”
The most recent and sophisticated devotee of the “industrial revolution” is Joel Mokyr (2002) who
considers that modern economic growth derived from a sudden leap in industrial technology. He
provides a detailed, erudite, illuminating but complex history of the interaction of “propositional”
and “pres criptive” (useful) knowledge since the mid eighteenth century, with a more cursory
acknowledgement of what happened earlier. He suggests (-2) that “most techniques before
1800 emerged as a result of chance discoveries, trial and error”. He makes a grudging
acknowledgement of the importance of printing (), and only a fleeting reference to advances in
shipping and navigation technology but is dismissive about their impact: “these earlier
mini-industrial revolutions had always petered out before their effects could launch the economies
into sustainable growth. Before the Industrial Revolution, the economy was subject to negative
feedback; each episode of growth ran into some obstruction or resistance that put an end to it…The
best known of these negative feedback mechanisms are Malthusian traps, in which rising income
creates population growth and pressure on fixed natural resources”. He is very insistent on the
narrowness of the “epistemic base” before 1800, and argues that positive feedbacks between the two
types of knowledge have increased hugely in the course of three industrial revolutions since the
eighteenth century. There has been a cascading interaction (p. 100) and we have now arrived at a
point where modern information technology has produced “an immensely powerful positive
feedback effect from pres criptive to propositional knowledge” (p. 115). His analysis of the
economic impact of knowledge is based on assertions rather than quantitative evidence. These are
presented with characteristic fervour, . his assessment of the impact of his second industrial
revolution: “The pivotal breakthrough in the propositional knowledge set was the identification of
the structure of the benzene molecule by the German chemist August von Kekulé in 1865…the
discovery of the chemical structure is a paradigmatic example of a broadening of the epistemic base
of an existing technique” (p. 85). My problem with Mokyr’s analysis is with his judgement on the
impact of science and not with his model which can be useful in explaining why the scientific
revolution of the 17th century had a delayed payoff, and why the innovative impact of science and
technology accelerated in the past two centuries. The problem is that he assumes no net
improvement in living standards before 1800, and a constantly accelerating cornucopia since then.
This contradicts the quantitative findings of historical national accounts in the Kuznetsian tradition
for the period before and after 1800. Mokyr is of course aware of this. In his defence (pp. 116-117)
he suggests that “aggregate output figures and their analysis in terms of productivity growth may be
of limited use in understanding economic growth over long periods. The full economic impact of
some of the most significant inventions over the past two centuries would be entirely missed in that
way.” Instead he opts for the Silicon-valley serendipity of DeLong (see Appendix 3).
Appendix 3 Hallucigenic History (Nordhaus and DeLong)
Nordhaus (1997) was an ambitious attempt to measure long-term changes in the price of light using
the hedonic approach. He estimated that the “true” . hedonic price of artificial light fell by per
cent a year (about 3,450-fold) between 1800 and 1992 in the USA, whereas the annual rise, using
the conventional consumer price approach, was per cent a year. He neglects the fact that the
supply of daylight did not change between 1800 and 1992. If one were to assess the degree to which
natural light was augmented, the impact of articicial light would seem a good deal more modest.
He illustrates the implications of his approach in measuring real wages. The conventional measure
showed a 13-fold increase between 1800 and 1992. The “true” rise, he suggests, was between 40
and 190 fold. He derived this result by converting conventional price indices into hedonics for three
economic sectors. For “run-of-the-mill” activities, where the characteristics of goods and services
have changed relatively little, he adjusted conventional price indices downward to eliminate “bias”-
assumed to be per cent a year. For “seismic” sectors where the goods and services of 1800 have
changed, but are still recognisable, conventional price indices received a downward adjustment
equal to half of his measure of bias for light. For “tectonic” sectors, where the nature of the good or
service has changed drastically, or did not exist in 1800, he applied his bias adjustment for light. He
assumed that 75 per cent of goods and services were in the first category in 1800, and that this
proportion fell to 28 per cent in 1992, when 36 per cent were seismic and 37 per cent tectonic. I
estimate that US per capita GDP rose 21-fold from $1,087 in 1800 to $23,169 in increase
of 190 fold would mean an 1800 level of $122 which would be well below subsistence.
Taking his cue from Nordhaus, Bradford DeLong (1998), argued that the Maddison estimates of the
rise in world GDP per capita involved massive understatement because of uncaptured quality
improvements. To correct this alleged mismeasurement he assigned “somewhat arbitrarily…an
additional fourfold multiplication to output per capita since 1800”. He shows a 35-fold increase in
average world per capita from 1800 to 2000, against my 9-fold increase, For 1500, his preferred
estimate is $115 per capita (in 1990 PPP dollars), which he assumed to prevail back to 1 million BC.
