Cobb-Douglas in pre-modern Europe 1

Simulating early modern growth

Jan Luiten van Zanden

International Institute of Social History/University of Utrecht
[email protected]

Working paper IISH, May 2005

Please do not quote without the permission by the author

1
The idea for the paper was born after listening to a presentation by Bob Allen at the TARGET workshop in
Oxford (in October 2004) on Capital Accumulation, Technological Change, and the Distribution of Income
during the British Industrial Revolution: An Analysis with Solow.

1
1. Introduction.

The question how much growth there was in Europe in the centuries before the Industrial

Revolution has become the subject of systematic research since the early 1990s. This began

with a number of papers by Graeme Snooks (1990; 1994; 1995) about the long term

evolution of the English economy, in which he argued that its GDP per capita had increased

almost eightfold between the Domesday book of 1086 and 1800. In fact, in his view growth

before the Industrial Revolution had been almost as rapid as after 1780. Next followed a

series of papers brought together by Angus Maddison and Herman van der Wee at the XIth

World Economic History Congress in Milan in 1994 which addressed the same issue for a

number of other European countries. Their conclusions were in general much more

conservative; growth between 1500 and 1800 was substantially slower if there was growth

at all than after 1800 (Blomme and Van der Wee 1994; Malanima 1994; Yun 1994; Van

Zanden 1993). The outcome of this sur vey was that, apart from one or two growth spurts such

as during Dutch Golden Age, per capita growth had been slow (Belgium) or non-existent

(Italy, Spain). In an attempt to synthesize these results I concluded that on average GDP per

capita in Western Europe may at best have increased by about 20% to 30% between ca 1500

and ca 1820; because labour input per capita probably increased substantially, labour

productivity may have remained the same (Van Zanden 2001).

These pessimistic conclusions contrast with the picture that can be derived from the

recent synthesis of thousand years of economic growth in the world economy by Angus

Maddison (2001). In The world economy, a millennial perspective he presented his own set

2
of estimates of the economic growth in Europe between 1000 and 1820 (and beyond): GDP

per capita in Western Europe trebled between 1000 and 1820, the result of a near doubling of

income levels between 1000 and 1500 and another 56% increase during the three centuries

before the Industrial Revolution. Giovanni Federico (2002) in his detailed review of

Maddisons 2001 book already pointed at a number of weaknesses in his estimates. They

result, Federico argued, in medieval levels of GDP per capita that are unrealistically low

(only just above subsistence levels), and growth rates between 1500 and 1800 seem to be

inconsistent with independent estimates of the growth of labour productivity in agriculture.

This paper present, firstly, an update of the research carried out recently, focusing

mainly on the country for which estimates of GDP (before 1688) are still rather shaky,

England. Also new research concerning a number of other countries (Sweden, Spain, the

Netherlands and Italy) will be presented briefly. The main focus of the paper is on testing the

results of these country-studies by experimenting with a Cobb Douglas production function.

The question is whether the available estimates for the development of the population, of real

wages and real rents can be made consistent with the benchmark estimates of the

development of GDP per capita in the centuries before 1800. In other words, assuming a

Cobb Douglas world (with constant returns to scale and an elasticity of substitution of 1), are

the estimates of GDP (per capita) consistent with what we know about the development of

factor prices and of population levels (and of the labour input per capita). This means

experimenting with the different variables that enter into the production function, with the

share of GDP that is invested (which determines the long-term growth of the capital stock),

with the number of days worked per year, with the absence or presence of technological

progress etc. As I hope to show, this produces new insights into the long-term development

3
of the European economy in the period before the Industrial Revolution, and also makes it

possible to discriminate between the two datasets at hand.

The most important result of such an approach is twofold. First, it bring together two

strand of research which have developed more or less independently, i.e. the new research

focusing on the measurement of national income and product, and the more traditional

approach, the study of real wages, which has dominated the analysis of the early modern

economy since the days of Abel, Braudel and Slicher van Bath. In the latter field a number of

attempts to synthesize the findings of local and national studies and analyse the European

patterns of real wage changes between the 14th and 19th centuries have recently been

published (Van Zanden 1999; Allen 2001). In particular the systematic study of real wages of

labourers and skilled craftsmen by Allen (2001) has produced new insights into what he

termed the great divergence within Europe, eg. the growing apart of real wages between the

North Sea region (where they did not decline much between the 15th and the 18th century),

and the rest of Europe (where the decline was quite strong, and real wages more or less fell

by 50% between the middle of the 15th century and the end of the 18th century. The first aim

of paper is to find out if these two approaches to the development of the early modern

economy can be recon ciliated using a Cobb Douglas production function.

Secondly, as I hope to demonstrate, such an experiment creates new insights into the

determinants of growth in the centuries before the Industrial Revolution. The question, for

example, when technological progress (eg. a sustained increase in total factor productivity)

began in England and in Holland, for example, and how important it was as a source of

productivity growth before 1800, can tentatively be answered with this approach. Similarly,

the issue of the importance of Malthusian forces cans be analysed as well: did the growth of

population lead to lower wages and GDP per capita, as many have argued in the 1950s and

4
1960s? Was there a technological ceiling beyond which the economy could not develop, as

stated by the famous Annales-school and by Emmanuel Leroy Ladurie in particular?

Before addressing these and similar questions, we will first turn to the recent research

on long-term economic growth in England.

2. Recent research: England between 1086 and 1688

Modern work on growth before and during the Industrial revolution began with the seminal

study by Deane and Cole (1962), who produced estimates of the economic growth of Great

Britain in the 18th century. Their work has been extended and refined by Crafts (1985) and

Harley (1993), which resulted in a set of estimates of the development of GDP starting in

1688 that have been widely accepted (Crafts and Harley 1992).

As already mentioned in the introduction, Snooks (1990; 1994; 1995) pioneered

research estimating national income before 1688 with his work on the Domesday Book of

1086. He ingeniously fills out the data from this venerable source, which gives detailed

information on `demesne income', with estimates of income from outside the `demesne'

sector, such as that generated by towns. The result is some fairly conserva tive figures; it is

assumed, for example, that the consumption of farmers was on a subsistence level. He then

combines his estimates of total income and income per head in 1086 with Craft's revision of

Deane and Cole's estimates for 1688 and onwards, and finally makes several (unspecified)

estimates for the intervening period. The result is a set of estimates of economic growth

between 1086 and 1688 which are astonishingly high: an average growth rate of 0.29% per

capita per year over a period of 600 years, leading, according to Snooks, to a quadrupling of

5
real income per head over the whole period. In some of the component periods - especially in

the first half of the 16th century - he even finds annual growth rates higher than 1.5% per

head. These results mean that growth between 1086 and 1688 would not have been much less

than during the 18th century (1688-1760: 0.31% per capita per year); the pace of growth in

the first half of the 16th century would only just be matched by that between 1830 and 1870.

Snooks' optimistic results have met with a great deal of criticism. There is a fairly

general objection to his 1086 level, which is probably underestimated, and to the fact that he

does not clearly explain how he has apportioned the overall growth between 1086 and 1688

between the various centuries. Neither is it clear how he solves the problem of price changes

between 1086 and 1688; here too there is insufficient explana tion of the method he has used

in his calculations. His estimates have been revised and improved by a number of authors,

most notably Nicholas Mayhew (1995) and Bruce Campbell (2000: 406-10). The latter

carefully compares Snooks figures with his much more detailed estimates of the output of

arable farming (about which more below), and showed that in view of what we know about

the subsistence needs of the population Snooks estimates are untenable (he systematically

underestimates the cost of peasant subsistence and fails to allow for the production of a

disposable surplus for the unfree). For 1086 Campbell agrees with the much higher estimates

produced by Mayhew, who corrected for some of the problems with Snooks work.

In another recent study Campbell (2005) produces detailed estimates of national

income in the 1290s, which can be compared with Gregory Kings very similar estimates (as

corrected by Lindert and Williamson (1982)) (Table 1). These estimates show a per

household increase of real income of 110%; because the average size of households also

declined, per capita income grew by about 150%. This can be compared with the relatively

small increase in income per head between 1086 and 1300 (or the 1290s), which Campbell

6
(in the footsteps of Mayhew) estimates at 10%. Because population growth was strong before

the 1290s, total GDP increased by 130-150% between 1086 and 1300 (Campbell 2000:409).

7
Table 1 Estimates of the structure and level of income in England, ca 1290 and 1688,
according to Campbell and Lindert & Williamson (in percent) 2

Share of Share of Share of Share of Real

households households income income increase
1290s 1688 1290s 1688 income
per
household
1290-
1688
Landowners, aristocracy, 2,3 2,9 15,8 22,0 132
gentry, high clergy
Tenants, smallholders 41,8 16,4 43,6 22,4 175
Cottagers, (rural) labourers, 36,2 37,9 19,4 9,1 - 6
vagrants
Non-agrarian sector 19,8 42,8 21,2 46,4 112
(commerce, professionals,
craftsmen etc.)
Total 100 100 100 100 110
Total population (millions) 4,32 4,9*
Total GDP (current prices, 3,67 54,44**
million pounds)
* according to Lindert and Williamson 1982 ; more recent estimates are gene rally slightly
higher, up to 5,06 million
** Lindert and Williamson 1982; Maddison 2004 gives 54,04 million

The comparison shows a strong polarization of the rural income distribution: the middle

classes of tenants and smallholders which still dominated the agricultural sector in 1290s,

were greatly reduced in relative and absolute terms, whereas the share of landowners in

income increased substantially (from less than 16% to 22% of GDP, or from about 20% of

value added in agriculture to about 50%). Moreover, the real income of agricultural labourers

(and cottagers) declined, in spite of the more than doubling of GDP per capita. Finally, the

share of non-agricultural activities in GDP more than doubled in these 400 years;

employment in the non-agrarian sector increased from less than 20% in the 1290s to almost

2
Assumptions: price level increased by factor 5 between 1290s and 1688 (according to Campbell); one-third of
labourers in 1688 is non-agricultural, the rest is agricultural.

8
43% in 1688, and their share in income increased in a similar way. These figures may

however still underestimate the growth of the secondary and tertiary activities; because of the

spread of proto- industrial activities on the countryside the part of the income of tenants,

smallholders, cottagers and rural labourers with non-agricultural sources of income must have

increased as well. Crafts estimated that the share of agriculture in GDP in 1700 was as low as

37% (declining to 26% in 1800 (Crafts 1985: 16-7)), whereas in the 1290s this must have

been more than 70%.

Perhaps the comparison between the estimates for the 1290s and the detailed

reconstruction of the social tables of 1688 is somewhat biased, however. The estimates

produced by Mayhew and Campbell are still quite rough, and have not been compared with

other data on the socio-economic composition of the population. The main contribution of

Lindert and Williamson (1982) was to correct the initially perhaps also rather rough

estimates by Gregory King, which resulted in a strong increase in estimated total income (by

25.1%). This is an example of what can maybe called the first law of historical national

accounting: the more we know about pre modern societies, the more detailed our estimates

of historical national accounts are, the higher income per capita appears to be. In other words:

major revisions of historical national accounts almost always lead to a strong upward

correction of the level of income per capita (as the revisions of Snooks by Mayhew and

Campbell, or the work by Lindert and Williamson revising Gregory King demonstrate). 3 The

comparison of a rough first generation estimate for the 1290s with a very detailed and

corrected second generation estimate for 1688 may therefore lead to an overestimate of the

3
This is in fact still true: even most recent revisions of national accounts result in upward adjustments of levels
of GDP, which may be as high as 10% or more; in these cases also the reason for this is that more detailed
information and improved estimation methods almost always lead to upward adjustments.

9
growth rate between these two dates. 4 Also, income per household increased much less than

income per capita because household size was falling from 4.4 in the 1290s to 3.5 in 1688;

but Lindert and Williamson (1982) may have underestimated population in 1688 somewhat

(it was probably 5,06 million instead of the 4,9 million they used), which may also lead to

slightly lower per capita GDP estimates for 1688 (on the other hand, the recent revision of

GDP in 1688 by Maddison (2004) did not result in higher estimates).

All authors agree that income per capita increased strongly between the 1290s and

1688, an increase that was perhaps as much as 100% to 150% (Campbell 2000: 411). As a

compromise I suggest 130%, but this is obviously a very rough approximation. To bridge the

gap between 1300 and 1688 I use the results of estimates of the performance of English

agriculture in the period 1086-1871 by Overton and Campbell (1996). Starting from data on

population, food consumption, yields, imports and exports of agricultural products and the

development of arable acreage from a large number of different sources, they could give a

rather detailed series of estimates of the total output of grain and potatoes in England in this

extended period. The resulting estimates of the development of output per head of the

agricultural population show a long-term stagnation in this measure of labour productivity

between 1086 and 1600 - with some significant swings - followed by a strong increase in

labour productivity between 1600 and 1800 (Table 2). Using a different approach mainly

based on the demand for agricultural products (as determined by real wages) and by the

structure of the labour force Allen (2000) produced another set of estimates of the

development of labour productivity in agriculture which are also reproduced in Table 2. Both

series show similar trends: an increase in labour productivity in this vital sector in the 14th

4
Another argument is that the 1688 revisions are so good that the resulting estimates may lead to the conclusion
that there was no growth during the 1700-1830 period; this is the result of a study by Lawrence Officer into the
same issue available via the eh.net; see http://www.eh.net/hmit/ukgdp/ukgdpstudy.pdf

10
and 15th centuries (following the population decline set in by the Great Famine and the Black

Death), a strong decline during the 16th century, which is followed by growth in the next two

hundred years. The main difference is that Overton and Campbell are more optimistic about

agricultural growth during the 18th century.

These estimates are used for the following conjectures about growth between 1300 (in

fact the 1290s) and 1700 (or 1688). It is clear that in the long run GDP per capita increases

much more rapidly than labour productivity in agriculture, which of course means that labour

productivity growth in the rest of the economy is much faster than in the primary sector. If it

is assumed that labour productivity growth outside agriculture is about twice as fast as in

agriculture (and that the share of agriculture in GDP declined from about 75% in 1300 to

67% in 1380) it is possible to estimate in a very rough way the development of GDP per

capita between 1300 and 1500. During the 16th century GDP per capita remained constant, in

spite of a strong decline of labour productivity in agriculture due to the renewed expansion of

population. This is consistent with recent estimates by Greg Clark (2004) who also concluded

that English GDP per capita did not increase during the 16th century. In the 17th and 18th

centuries the pattern changes again then, productivity growth in agriculture was almost as

rapid or, if we are to believe the Overton and Campbell estimates, even more rapid than in the

rest of the economy, which was a characteristic of the transformation of English agriculture

in these years (Crafts 1985). The overall picture that emerges from these estimates is one of

slow growth during the High Middle Ages, followed by relatively rapid growth in the 1300-

1500 period, stagnation during the 16th century and rapid growth in the next two hundred

years. Overall GDP per capita increased by about 250% in these 700-odd years.

11
Table 2 Estimates of the development of the output per head of the agricultural population
and of GDP per capita in England, 1086-1800 (indices 1800 = 100)

1086 1300 1380/1400 1500 1600 1700 1800

Crafts/Harley and 29 32 - - - 76 100
Campbell*
Labour productivity 43 41 45 - 39 64 100
in agriculture:
Overton & Campbell
Idem, Allen - 56 64 70 53 80 100
GDP p.c. Snooks 13 - - 20 54 76 100
GDP p.c. this study 29 32 42 48 48 76 100
*1300 (is the 1290s) is linked to 1700 by assuming 4% growth between 1688 and 1700,
and 130% between 1300 and 1688 (as explained in the text)

Sources: Campbell 2000; Crafts 1985; Crafts and Harley 1992; Overton and Campbell 1996;
Allen 2000; Snooks 1994; 1995.

3. European economic growth between 1000 and 1800

In 2001 I published a compilation of estimates of GDP per capita in six European countries

Great Britain, the Netherlands, Italy, Spain, Belgium and Poland taken from case-studies by

different authors, which resulted in the conclusion that growth was rather slow and in fact

normally absent in most of them during the three centuries before 1800 (Van Zanden 2001).

The new research carried out since tends to confirm this conclusion:

- Olle Krantz made a detailed reconstruction of Swedish national income in 1571,

concluding that despite the margins of uncertainty, it could be inferred that GDP per

capita was about the same in the 16th century as around 1800. Thus, Sweden, like

12
 other peripheral countries was characterised by stagnation during the period between

the 16th and the 19th centuries. (Krantz, 2004);

- Paolo Malanima (2003) in a recent paper produced more detailed estimates for Italian

GDP in the very long run (from 1300 to 1861), which confirm the trend of a gradually

declining level of income between the high Middle Ages and the early 19th century;

- Albert Carreras (2003) reviewed estimates of Spanish GDP in the early modern

period, adding to the estimates previously published by Yun (1995) his own estimates

for the beginning of the 16th century;

- Finally, I published detailed estimates of the structure and level of GDP in Holland in

the years 1510/14 (Van Zanden 2002b), which can also be compared with estimates

for the early 19th century; during these three hundred years, GDP per capita increased

by about 50%, almost all of which was concentrated in the period between the 1570s

and the 1650s (in fact, between 1650 and 1807/08 levels of real income slowly

declined again; see also De Vries (1984), Van Zanden (1987; 1992)).

In Table 3 sets out the results for the seven countries for which point estimates based on a

detailed reconstruction of GDP are now available. The relative levels of GDP per capita in

1820 taken from Maddison (2001) are used to render comparable these attempts to quantify

economic growth before 1800.

13
Table 3 Estimates of the development of GDP per capita in six European countries, 1500-
1820 (Great Britain 1820=100)

c. 1300 c.1400 c. 1500 c. 1570 c. 1650 c. 1700 c. 1750 1820

Great Britain 29 38 43 43-45 54 69 84 100
Netherlands - - 58 58 95 94 94 92
Belgium - - 46 55 53 55 61 62
Italy 71 71 67 65 60 57 61 53
Spain - - 43-48 43-48 39-48 39-44 40-41 48
Sweden - - - 51 - - - 56
Poland - - 45-53 42-48 42-49 35-40 30-33 41
Average* - - c. 51 c. 52 c. 58 c. 59 c 62 c 66
(unweighted)
Average** - - c. 54 c. 54 c. 55 c. 56 c. 56 c.58
(weighted)

* without Sweden

Sources: Van Zanden 2001; Maddison 2001 ; Carreras 2003 ; Malanima 2003; Krantz 2004.

14
The results of this comparison can be briefly summarized. Long-term stagnation is

revealed on the periphery; i.e. Italy, Spain and Poland. Between 1500 and 1750 GDP

per capita in these countries first fell, after which a certain recovery set in (at least in

Spain and Poland). Only in Spain, however, was GDP per capita possibly somewhat

higher in 1820 than in 1570. Compared with the stagnation of southern and eastern

Europe, the countries bordering the North Sea show relatively gradual (Belgium) or

rapid (England) growth in the early modern period. These estimates show a rough

doubling of GDP per capita in England between c 1600 and 1820; the estimates for

Holland imply a much more modest rise of only about 50% for the same period, and the

extent of the increase in Belgium was probably even smaller.

Going back in time the margins of error of these estimates increase. The ranking

for the 16th century Italy being the richest country, followed by the Low Countries is

plausible. Only English income levels appear to be lower than expected at least in

international perspective. The minimum level Poland in the 18th century, or England

in 1300 - was probably about 30% of the base year.

The general picture emerging from these data can be looked at in two ways. It is

clear that, over the long term, population growth was more than compensated by the

increase in production whereby in `Europe' - the average of these six countries -

production per capita between c.1500 and c.1820 increased by an average of some 25%

(of which almost 10% occurs after 1750). In total the population in these six countries

increased by 91% between 1500 and 1800 (see Table 1). The Malthusian pessimists,

who saw a growing tension arising between population and resources, were therefore

not right, if we go by these data: the growth of population was clearly matched by a

somewhat larger increase in output. But likewise the hopes of the optimists are only met

15
in patches: economic growth, in the sense of growth in per capita produc tion, is not

normal in western Europe, but rather an exception to the rule - certainly before 1700. In

Holland there occurred only one `growth spurt' in a period of 300 years, and this was

probably also the case in Belgium; moreover, growth in Holland was partly achieved at

the expense of Flanders/Brabant, whose economy declined at the same time (when the

center of gravity of the economy of the Low Countries moved from Antwerp to

Amsterdam). On balance, growth is very modest indeed in these six countries, taken as

a whole; the weighted average GDP per capita increases by less than 10% in a period of

300 years, and this growth is mainly due to the inclusion of the most dynamic parts of

Europe (England and Holland) in our sample.

Another possible conclusion is that differences in the level of economic activity

within Europe were small. The gap between the richest regions (Flanders and northern

Italy) and the poorest (England or Poland) in about 1570 was at most 30% (of the level

of the richest) and probably even smaller. Differences increased sharply during the 17th

century as a result of the rise of Holland and the decline of Poland and Spain, but this

was compensated partially by the rise of England and the decline of northern Italy and

Flanders relative to the `European average'. In the second half of the 18th century

international disparities seem to reduce slightly, due to the increase in GDP in Poland

and Spain (and to the stagna tion in Holland). In 1820 the spread around the mean was

even smaller than in 1700 or 1750, although at that time England was certainly running

increasingly ahead of the continental countries.

Growth between 1500 and 1800 was relatively slow, except for the countries bordering

the North Sea. The story for the five centuries before 1500 may be very different

though. It is plausible that English GDP per capita increased by as much as 50% during

16
the crisis of the late Middle Ages (see Table 2), after a more modest increase of only

10% per capita between 1086 and 1300. One wonders how exceptional this was.

There are different ways to address this issue. One way to approach it is to find

out how typical Medieval England was; are there, for example, reasons to assume that

English GDP per capita in 1086 was systematically higher or lower than on the

Continent? There is no doubt that compared to the Low Countries and southern Europe

(Italy, Spain, the Mediterranean in general perhaps) England in 1086 was a relatively

backward country, although the difference was not that big when one compares England

with Western Europe as a whole. Moreover, large parts of the Continent, such as

Flanders and northern Italy, developed more rapidly during the High Middle Ages than

England, which specialized on being a supplier of agricultural commodities (wool in

particular) during the great boom of 1000-1300, whereas Flanders and northern Italy

concentrated on industrial and tertiary activities with a higher value added. The result

was probably that growth was more rapid in the latter regions.

Another way to approach the same problem is to estimate some kind of

(subsistence) minimum below which income cannot fall. A reference point is

Maddisons estimate of British GDP per capita in 1820: 2121 (1990 interna tional)

dollars, the benchmark of Table 3. British GDP per capita at about 1500 was 910

dollars, and following the estimates presented in Table 2 this would be around 550

dollars in 1086. In a recent study of income and inequality in Byzantium around the

year 1000 Branko Milanovic (2004) estimated that the absolute minimum level of GDP

was probably about 400 (international 1990) dollars. This result seems plausible;

English GDP per capita in 1086 would be about 40% above the subsistence minimum,

which is consistent with the Campbell (2000) estimates that the value added of arable

output would be about 55-60% of GDP. The level of GDP per capita of Western Europe

17
as a whole would probably be somewhat higher than that in England, say between 550

and 600 dollar. In the same study Milanovic also estimates the GDP per capita of

Byzantium. He arrives at a figure of between 640 and 720 dollars, which sets the upper

limit of European GDP because at the turn of the millenium Byzantium was probably

the richest part of Europe. At about 1500 the average for Western Europe was about

1100 dollars (Table 3), which implies that it might have doubled in the preceding five

centuries (the increase in England was more modest, from 550 to 910 dollars).

Summing up, and taking into account that the two estimates used here to (of Campbell

for England in 1086 and of Milanovic for Byzantium in 1000) are highly tentative, we

can perhaps conclude that the five centuries before 1500 were more dynamic in terms of

per capita growth than the three centuries after 1500.

3. Holland 1500-1800

How plausible are the results presented here? Are they consistent with other evidence

on the long-term development of the European economy in these years? And what was

driving growth in the two cases that were more dynamic, the Netherlands and England?

To try to answer these questions, a number of experiments with a simple Cobb-Douglas

production function were carried out to simulate the relationships between real wages,

real rents (when data are available), population, labour input and land input.

The production function that is used has three inputs, land, labour, and capital,

which are all assumed to be homogenous and unchanging in terms of quality (an

assumption to which we will return). We start with Holland, an economy for which

detailed estimates of the structure of GDP in 1510/14 and 1807/08 are available. It is

18
possible to estimate the share of each factor in GDP in 1510/14 (land .07, labour .60 and

capital .33); 5 all simulations will be expressed in prices of that base period. Other

ingredients of the model are:

- factor prices: estimates of real wages of an unskilled labourer per day (from

Allen 2001) and of nominal and real rents (per hectare) (the latter deflated with a

CPI) (Van Zanden, work in progress)

- factor inputs: an annual series of the cultivated land (Van Zanden, work in

progress), and the population and the labour force (apart from the changes in the

number of days worked per year, the ratio between labour force and population

was assumed to be constant at 40%)

- the capital stock in the base period 1510/14 was estimated by assuming an

average rate of profit/interest on capital stock of 12%, which in combination

with a share of 33% of GDP accruing to capital resulted in an estimate of the

capital stock in the base year; it was furthermore assumed that depreciation

reduced the capital stock by 2.5% per year; additions to the capital stock were

the results of investment expressed as a certain percentage of GDP (so the

capital stock in year t+1 was 97.5% of the capital stock in year t plus gross

investment during year t).

The instrument variables are:

- the share of GDP that was invested annually (the starting point here is level of

5%); 6

5
The average values for 1510/14 were estimated at 7% for land, 67% for labour and 26% for capital, but
the share of labour tended to decline during the early modern period, so a slightly lower share was chosen
for the simulation
6
It was of course also possible to vary the rate of deprecation, which was now set at 2.5%; experimenting
with this share had a similar effect as experimenting with different investment shares

19
 - technological progress, which was Hicks-neutral, i.e. affected all factor

contributions equally; the starting point was the hypothesis was that there was

no technological progress;

- the number of working days per labourer; for Holland this variable was set at

200 before 1560, increasing gradually (under the impact of the Reformation and

declining real wages) to 250 in 1660 (according to Noordegraaf 1985 58-61; see

also De Vries and Van der Woude 1997: 616-7).

The model isolates two causes of long-term growth of real wages: capital accumulation

and productivity growth. As the simulations will show, the re-investment share has

generally only a limited effect on the long-term development of the economy (which is

restricted by the share of capital in GDP). By definition, productivity growth translates

itself directly in a proportionate increase in GDP and in real wages, and it therefore a

much stronger candidate for explaining long-term change. The simulation can therefore

help to answer the question when European countries began to experience sustained

productivity growth: was this only during the Industrial Revolution of the second half of

the18th century, or did periods of productivity growth occur already before 1750? Was,

for example, the Dutch Golden Age characterized by modern productivity growth as

De Vries and Van der Woude have maintained, or was growth simply the result of a

higher level of capital formation?

The third instrument variable (i.e. changing the relationship between labour

input and population) makes it possible to bridge the gap that might arise between the

development of real wages and the point estimates of GDP per capita (as will be

explained below). It is also an attempt to test the hypothesis formulated by Jan de Vries

(1994) that an industrious revolution preceded the industrial revolution of the 18th

century. He hypothesizes that labour input per capita increased strongly due to the fact

20
that men, women and children worked more days and hours per year than they did in the

late Middle Ages.

The procedure has been as follows: the real wage as simulated by the model was

compared to the real wage as estimated by Allen (2001), or rather the long-term pattern

of the wage series as estimated by a second-degree polynomial; the instrument variables

were manipulated in such a way that the development of the simulated real wage is

almost identical to this estimated second-degree polynomial (it is in practice impossible

to follow the ups and downs of the real wage curve itself, because the enormous

fluctuations are driven by many erratic supply and demand-shocks). Figure 1 pictures

the results.

Real wages decline between 1500 and the mid 1570s, thereafter there is a strong

increase of real wages continuing until about 1710; during the second half of the 18th

century real wages decline again. This pattern is simulated with different combinations

of the investment share and the rate of technological progress. The model easily

predicts the decline of real wages during the first three-quarter of the 16th century

when an investment share of 5% is used; declining capital stock and land per capita are

the explanations for the decline. After the 1570s population growth accelerates, but real

wages also start to increase, which clearly point to a (well known) break in economic

development in these years (which has for example been analyzed by De Vries and Van

der Woude as the Dutch take off). It is this combination of rapid growth of population

(and of the labour force which may have increased even faster because immigration of

young adults was a major factor contributing to the increase in population) and an

increase in real wages (the magnitude of which is still being disputed however, see Van

Zanden 2002a) which can given the slow growth of the cultivated land only be

21
explained by either a massive increase in the accumulation of physical capital and/or the

growth of total factor productivity. The different combinations of investment ratio and

the rate of productivity growth that produce almost identical results are: 12% and 0%,

10% and .1%, 7% and .25%, and 5% and .35%. Because it is likely that both the

investment ratio and the rate of technological progress went up simultaneously during

these years, a combination in the middle of this range (i.e. an investment ratio of 7%

and an increase of productivity of .25%) was selected to run the simulation. Moreover,

this hardly affects the growth path of GDP (per capita), because the wage income is a

fixed share of GDP (and the labour supply is exogenously given); the real wage curve

therefore directly determines the growth of GDP (see Figure 1).

The next turning point again well known from the more qualitative literature

is the 1660s and 1670s, when the increase in real wages begins to flatten, whereas at the

same time rapid population growth ends and the population of Holland even begins to

fall. This could be simulated by assuming that technological progress comes to an end

(in 1670), and that the investment ratio begins to fall from 7% in 1670 to 4% in 1750,

after which it remains constant.

The following step in this exercise is that we can now derive estimates for the

development of GDP (total and per capita) from the simulation. Figure 2 presents the

results by comparing the simulated estimates of the development of GDP per capita in

Holland with those of the Van Zanden 2001 dataset. According to this first round

simulation GDP per capita increased by only about 30% between 1510 and 1805, less

than the 50% which was the result of comparing the benchmark estimates. The long-

term development of GDP per capita shows the same pattern, however: the simulation

predicts a decline of GDP per capita during the 16th century (consistent with the

22
stagnation of the point estimates), followed by a strong increase during the Golden Age

(1580-1670) which is however less steep than the point estimates suggest.

Figure 1 Real wages in Holland: estimated and simulated, 1510-

1805 (indices 1510/14=100)

160

140

120

100

80

60

40

20

0
1510 1560 1610 1660 1710 1760

Simulation (second round) Real wages (Allen)

Simulation (first round) Poly. (Real wages (Allen))

23
 Figure 2 GDP per capita: point estimates and simulations,
1510-1805

45

40

35

30

25

20

15
1510 1535 1560 1585 1610 1635 1660 1685 1710 1735 1760 1785

simulation second round point estimates simulation first round

The difference between the point estimates and the simulation can be bridged by

manipulating another instrument variable, the number of days worked annually. This

instrument variable captures both quantitative and qualitative changes in the labour

force. The quantitative dimension is a real increase of the number of hours worked per

capita via the lengthening of the working week (through, for example, the abolishment

of holidays following the Reformation), or the increase of children and womens labour.

Jan de Vries (1994) has suggested that during this period an industrious revolution

occurred resulting in a strong increase in per capita working hours, and the fact that

simulated GDP per capita seems to grow much less than might be expected on the basis

of the development of the real wage in this period can be interpreted as confirmation of

his hypothesis. But qualitative changes much also be taken into account. The average

quality of the labour force increased substantially during the same period; the share of

24
services in the labour force for example went up from 22 % in 1510/14 to 42 % in

1807/08, and a large part of this increase consisted of highly skilled and paid work by

professionals, merchants, government employees etc. (but also the number of servants

and soldiers grew); moreover, the relative pay of these professionals did probably also

rise quite a lot. Also in industry the share of skilled workers may have increased,

whereas branches mainly using unskilled personnel (fisheries, peat digging and

agriculture) saw their relative size decline. These changes in the composition of the

labour force are one of the reasons why indices of real wages which are entirely based

on the wages of construction labourers (or of agricultural labourers) are not

completely reliable guides to the long-term development of the wage income. Ideally,

one would like to have a weighted average of wages of all members of the labour force.

Assuming that the average quality of the labour force increased by about 20-30% -

which is probably still an underestimate - will already bring the two curves (of

simulated GDP per capita and the point estimates of GDP per capita) quite close to each

other.

Another limitation of the model is that it assumes a closed economy, which

Holland clearly was not. Already at the beginning of the 16th century a large part of the

food supply consisted of imported grains, and this share increased strongly in the next

150 years. The imports from the Baltic are a good index of these changes: they went up

from 10.000 lasts (of about 2 tonne) representing about 25% of the total grain supply

during the first decade of the 16th century, to more than 50.000 lasts between 1610 and

1650 (or more than 70% of supply but part of its was re-exported). In return for these

land-intensive imports, Holland exported large amounts of labour-intensive and capital-

intensive products and services. It has been suggested that the economic development of

Western Europe in the early modern period was facilitated by the fact that Holland and

25
(from the 18th century onwards) Great Britain had access to the ghost acreages of

northern Europe/the Baltic and the ir colonies (see the discussion in Pomeranz 2000).

The simulation presented here can give an idea of the quantitative importance of these

ghost acreages by looking at the gap between the actual development of real rents

(which is affected by these international trade flows) and the simulated rent following

from the Cobb Douglas model assuming a closed economy. Unsurprisingly the

simulated rent increases much more during the Golden Age than real rents (see Figure

3); in a closed economy real rents would ha ve increased by almost 250% between 1500

and 1700, in reality they only grew by about 50%. Population decline and agricultural

productivity growth during the 18th century did much to reduce the gap between the two

series though (at about 1800 the Netherlands as a whole had become more or less self-

sufficient again, but Holland was still a net- importer of grains).

26
 Figure 3 Real rent in Holland: estimated and simulated
1510-1800 (1510/14 = 100)

400

350

300

250

200

150

100

50

0
1510 1535 1560 1585 1610 1635 1660 1685 1710 1735 1760 1785

estimated simulated

To find out how much this may have affected growth, a second simulation was carried

out with a recalculated series for the land input; the difference between the real rent and

the simulated rent from the first simulation was taken as a measure of the ghost

acreage, and another series of GDP and real wages was calculated with the new, larger

land input. The second round simulations in Figures 1 and 2 show that at their peak

the impact of the ghost acreages was about 10% of GDP per capita, and somewhat

more in terms of real wages.

Summing up, the model can replicate the long-term development of the

economy of Holland, identifying clearly the well-known turning points in the 1590s and

1660s quite well; between 1580 and 1670 there was a more or less constant rate of

productivity growth and investment was probably higher than before and after. To get

an increase in GDP per capita that matches the point estimates, one has to assume a

27
rather strong increase in labour input per capita stronger than the assumption used

here that it increased by 25% (from 200 to 250 working days per year). The model does

have some problems with the openness of the Dutch economy; real rents are not

simulated correctly, but this part of the experiment can give insight into the importance

of ghost acreages for economic growth.

5. England 1500-1800

The same procedure was applied to the English data. Obviously, the contribution of

different inputs to GDP was different from the Dutch case, and was estimated at 20%

for land (following Clark 2002), 30% for capital, and 50% for labour (rough estimates

based on a.o. the social tables by King, as reconstructed by Lindert & Williamson). 7

The real wage series of unskilled labourers by Allen was again taken as a starting point.

Estimates of population were supplied by Wrigley et al (1997), cultivated land was

assumed to grow by 10% per century (Campbell and Overton 1996); the capital stock in

1500 was simply estimated on the basis of a share of 30% in GDP going to capital, and

an average profit rate of 12%.

Figure 4 shows the degree of fit between the simulation and the estimated series

of real wages that results from the following manipulation of instrument variables: in

the sixteenth century the investment ratio was set at 4.5%, and there was no

technological progress. This results in a slow decline of real wages as the per capita

capital stock and in particular the per capita supply of land goes down. During the 1590s

7
Assuming 250 days per annum, a participation ratio of 40% and a wage rate of an unskilled labourer of
14 d., the wage sum is exactly 50% of the total national income as estimated by Lindert and Williamson
1982.

28
things start to shift and the real wage begins to develop more favorably. The decline of

real wages comes to an end, but population growth continues, a combination that is

inconsistent with the previous specification of the model. As in the case of Holland,

different combinations of an increased investment ratio and of productivity growth can

generate the pattern that emerges at about 1600/1610. In the simulation presented here,

we experimented with the variables that could be derived from growth accounting

studies for the second half of the 18th century, which show that 1/ the capital stock per

capita remained more or less constant and 2/ that the growth of total factor productivity

was about .25 to .30 per annum (Crafts 2004 for an overview). Surprisingly, assuming

the same variables (an investment ratio of 6.5% from 1590s onwards and a consistent

increase in total factor productivity of .25% from 1615 onwards) gives an almost prefect

simulation of the growth path of real wages during the 17th and 18th centuries. Keeping

these two variables constant for the rest of the period produces the turning point in the

early 1600s, an increase in the simulated real wage between the 1610 and the 1750s of

about 50%, and stabilization of real wages during in the second half of the 18th century

(when population growth accelerates again). After 1750 the Allen series is rather

pessimistic however (I have to assume a decline in investment ratio or in technological

progress to get the same result); Feinsteins more comprehensive series shows a

stabilization between 1750 and 1800 which is similar to the simulation results presented

here (Feinstein 1998).

In order to get a development of GDP per capita which is more or less consistent

with the available point estimates a very strong increase in labour input per capita has to

be assumed. Figure 5 presents the result of this part of the simulation, assuming a linear

29
increase of working days per year from 200 in 1500 to 350 in 1800. 8 It should again be

pointed out that this increase consists of a qualitative and a quantitative component.

England seems to be the heartland of Jan de Vries industrious revolution, where the

labour force grew much more rapidly than total population as a result of a strong

increase in (a.o.) the labour input by women and children. As a result, in the early 19th

century the ratio between labour force and population was in Great Britain was much

higher than elsewhere (ca 45% against 41% in the Netherlands and 35% in the United

States). Of course, at the same time the structure of the labour force changed

dramatically, the share of agriculture going down from more than 70% in 1500 via 55%

in 1700 to 35% in 1800 (Allen 2000), which must have had similar effects on the

quality of the labour force as we discussed for Holland.

These simulations clearly show that, if real wages are a reliable guide to macro-

economic performance, the transition towards modern economic growth of the English

economy did not occur during the 18th century, but at some point between the 1590s and

the 1620s, when it moved from what was basically a trajectory without technological

progress to one with a higher level of investment and a more or less constant rate of

productivity growth. The fact that this occurred during the same years as the take off

of the Dutch economy is striking; the big difference is, of course, that productivity

growth was sustained into the 18th century and beyond. What is perhaps equally striking

is that productivity growth did not accelerate during the 18th century. I added in Figure 5

a similar pattern of growth that emerged from another experiment in simulating growth

using a regression model with real wages and the degree of structural transformation of

8
This assumption has of course also consequences for the modeling of real wages; the development of
the instrument variables discussed in the previous paragraph (an increase of the investment ratio from
4.5% to 6.5% after 1590 and a continuous rate of productivity growth after 1615 of .25%) are from the
same simulation in which the increase of the number of working days from 200 to 350 is assumed; if one
drops this assumption and only assumes that is increased from 200 to 250 the other instrument variables
become 6% after 1590 and .2% technological change.

30
the economy as independent variables (described in Van Zanden 2004). Experimenting

with this model also pointed to a relatively early start of modern economic growth in

England at some point during the 1610s or 1620s. As I have argued elsewhere, there

can be no doubt at all that the English economy grew at an exceptional rate, precisely in

the centuries leading up to 1800 while its growth in the two centuries after 1800 has not

been so unique (see also Wrigley 2000). It is perhaps an irony that the country which

has provided the model for the classical `Indus trial Revolution' - the decisive break

between a stagnating agrarian society and a dynamic industrial economy - was in fact

characterized by such dynamism in the centuries preceding this `revolution'. From this

research it would appear that the `Industrial Revolution' of the second half of the 18th

century was no `accident', as Crafts would have us believe (Crafts 1977), but was the

logical continuation of the exceptionally dynamic development of the British economy

in previous centuries.

31
 Figure 4 England: real wage, simulated and estimated,
1500-1800 (1500=100)

130

120

110

100

90

80

70

60

50

40
1500 1525 1550 1575 1600 1625 1650 1675 1700 1725 1750 1775 1800

simulated estimated Poly. (estimated)

32
 Figure 5 English GDP per capita: point estimates and
simulated values, 1500-1800 (1500=100)

220

200

180

160

140

120

100

80

60

40
1500 1525 1550 1575 1600 1625 1650 1675 1700 1725 1750 1775 1800

cobb-douglas alternative model point estimates

Did England also profit from access to ghost-acreages? Figure 6 compares the series of

real rents as simulated by the Cobb-Douglas production function with the estimates

presented by Gregory Clark (2002). During much of the 16th and 17th centuries the two

series run more or less parallel; England was an exporter of agricultural commodities

until the middle of the 18th century, but these exports were a much smaller fraction of

supply or demand than in the case of Holland. This changed during the second half of

the 18th century, when it became a large net importer of grains (from the Baltic),

colonial products such as sugar and tea from the colonies, and, after 1780, of cotton

from the America. These developments are accurately reflected in the ratio between

simulated and real rents, and this comparison therefore adds credence to the approach

developed here.

Between the mid-18th century when agricultural trade was probably more or less

in equilibrium, and 1800 the ratio between simulated rents and real rents increased by

33
almost 40%, which is an index of the degree to which England profited from ghost

acreages. Given a share of 20% of land in GDP (which was assumed for this

simulation), this would add about 7% to GDP, but this 20% share is probably accurate

for the 16th and early 17th century and does not really reflect the changing structure of

English GDP thereafter (the share of agriculture in GDP was only 27% in 1800, and the

share of land cannot have been more than half that amount, say 10-12%, so the extra

GDP will have been less than 4.5%). Of course, as in the case of Holland, these are

upper-bound estimates, the result of the assumption that as a result of international trade

these countries acquired extra land as a free lunch, because we simply added ghost

acreages to the inputs in the production function. In reality, there are no free lunches,

and the services of this extra land were bought by selling capital- and labour-intensive

products to the land-abundant countries and colonies. This solved bottlenecks, and made

it possible to continue growing in spite of the Malthusian constraints, but the gains

must have been much smaller than the estimates presented here.

34
 Figure 6 English land rent: simulated and estimated,
1550/59-1790/99 (in constant prices, 1550/59=100)

500

450

400

350

300

250

200

150

100

50

0
1550 1575 1600 1625 1650 1675 1700 1725 1750 1775 1800

Simulated Estimated (Clark)

6. England 1300-1500

Allen also offers a real wage series for England 1300-1500 that can be used for

simulating growth in the late Middle Ages. The idea that the increase of real wages in

the period after the Black Death points to an increase in productivity and GDP per

capita is now new (see Epstein 2001), but has not been tested in this way. I changed the

shares in GDP somewhat by augmenting the share of land to 30% (consistent with

Campbell 2005), brought capital down to 20% and kept labour at 50%. Capital stock in

1300 was again estimated on the basis of the share of capital in GDP, and an assumed

profit rate/interest rate of 20% (higher than the 12% assumed for the post 1500 period).

A population series was taken from Clark (2004); it shows an almost continuous decline

35
of the population from the 1310s onwards; only during the second half of the 15th

century did English population stabilize at slightly more than 40% of the 1300 level. 9 It

was also assumed that the strong increase in real wages led to a reduction of working

hours and days (from 250 days per year before 1348 to 200 days after 1400). The

estimates of the cultivated land were derived from Overton and Campbell (1996),

according to whom it shrank by about 10% between 1300 and 1380 (a decline that was

assumed to have taken place between 1348 and 1380).

The simulation presented in Figure 7 starts with the assumption of a 4.5%

investment share, which basically remains uncha nged between 1300 and 1500; the

underlying idea was that because the ratio between capital stock (and land) and

population did improve much in these two centuries incentives for increasing the

investment ratio were not strong (although, at the same time, the interest rate also went

down, which may have favored investment). In order to get the strong increase in real

wages that is characteristic of this period, between 1348 and 1430 a constant rate of

productivity growth of .30 % has to be assumed; the increase in total factor productivity

during the whole period is almost 30%. This estimated growth total factor productivity

is obviously dependent on the exact specification of the parameters of the Cobb-

Douglas production function, and in particular of the share of labour in GDP. The

preferred simulation is based on a 50% labour share, which might be too low. 10

Increasing the share of labour in GDP results in even higher estimates of productivity

growth between 1350 and 1430; a 75% share, for example, is consistent with an .50%

9
Linking his series to an estimated total English population of 2,5 million in 1500 gives a population
estimate of 5,3 million in 1300, which is higher than the estimates by Campbell which I adopted for Table
1; perhaps the Clark series is overestimating the population decline, but it is not possible To Whom It
May Concern: correct for this; if the actual decline of the population was less than Clarks series suggests,
the model would predict an even milder increase in real wages than actually occurred, suggesting that
productivity growth in that scenario would have been even higher than in the simulation presented here.
10
Based on an average daily wage of 1.5 d., a 40% labour force participation rate and 250 working days
per annum, the total wage sum for the 1290s can estimated at 2,7 million pounds, or 74% of the estimated
GDP according to Table 1

36
rate of productivity growth. Experimenting with a lower share of labour in GDP gives

results with very low or even zero productivity growth the latter is consistent with a

share of labour in GDP of 40%. The most plausible scenario seems to be that the radical

changes in relative prices that occurred after 1348 and in particular the strong rise of

real wages resulted in the search for new technologies, a search that led to a

substantial increase in total factor productivity (see Mokyr 1990 for a discussion of the

technological dynamism of the late Medieval period).

The estimates of the development of GDP per capita that result from the

preferred experiment (with a 50% labour share) are presented in Figure 8; varying the

share of labour in GDP does not affect the increase in GDP per capita, though; also in

the other experiments does GDP per capita increase substantially between 1300 and

1450. Such an increase is quite striking, and is even faster than the point estimates

suggested. Characteristically, the increase of about 70% during these two centuries

more or less mirrors the decline of population numbers (by almost 60%). Had we

assumed that the number of working days per year remained constant between 1300 and

1500 (instead of a decline by 20%), the increase of GDP per capita would have been

even larger (in such a scenario one has to assume a constant rate of productivity growth

of .40% between 1349 and 1440 to get the best fit; GDP per capita increases by more

than 100% in this version of the model). All these simulations lead to the conclusion

that there was strong productivity growth in England between c. 1350 and c. 1450 but

also that after the middle of the 15th century this technological transition came to an

end. The net result was a much higher GDP per capita, and probably some reduction in

working hours.

37
 Figure 7 English real wages: simulated and real, 1301-
1500 (1300=100)

250

200

150

100

50

0
1301 1321 1341 1361 1381 1401 1421 1441 1461 1481

Simulated Estimated Poly. (Estimated)

38
 Figure 8 English GDP per capita (simulation and point
estimates) and English population, 1301-1500 (indices
1300=100)

200

180

160

140

120

100

80

60

40
1301 1321 1341 1361 1381 1401 1421 1441 1461 1481

Simulation Point estimates Population

39
7. Italy 1300-1800

The same experiment was carried out for Italy, for which we have the point estimates of

GDP and population by Malanima (1998, 2003). Because his estimates relate to

northern Italy, we are dealing with a relatively urban society here, in which land played

a smaller role than in England. Italy was also more or less self-sufficient in foodstuffs,

which means that we probably do not have the problems encountered in the case of

Holland. I assumed shares in GDP of 15% for land, 60% for labour, and 25% for

capital, but also experimented with a model with radically different shares (25%, 40%

and 35% respectively) to find out how sensitive the estimates are for this assumption.

Cultivated land remained basically unchanged between 1300 and 1800, the labour input

increased with population (and I experimented with a decline of working days from 250

to 200 between 1350 and 1450 followed by an increase from 200 to 250 between 1500

and 1600), and the capital stock in 1300 was determined by its share in GDP of 25% (or

35%) and an interest rate of 20%. For the first decades of the 14th century the Allen

(2001) wage series (which begin in 1326) was supplemented by the wage series

published by Malanima (2004).

Figure 9 presents the result of the simulation: real wages rise in the second half

of the 14th and first half of the 15th century by about 60% on average (less than in

England), and begin to fall after c 1450, a decline that is almost continuous but slow

until about 1750, and accelerates in the second half of the 18th century (according to

Malanima (2004) the Allen estimates overestimate the decline after 1750 however).

This pattern can be matched by the model assuming 1/ that the investment rate was

4,5% initially en 2/ that between 1348 and 1400 there was continuous growth of total

factor productivity of .40 % annually, the rate of which fell constantly to zero between
1400 and 1450, and 3/ that after 1450 the investment share also declines to a mere 1 %

of GDP, at which very low level is remains until the end of the 18th century (and, if the

simulation was to follow the sharp decline of real wages after 1750, investment would

have to fall to less than that, but in the results presented here investment remains at 1%).

The alternative experiment (with a much lower share of labour in GDP) started with the

same assumptions, but needed a more modest rate of technological progress after 1340

(of .20% between 1349 and 1420) and a slightly higher level of gross investment after

1450 (of 2% annually); the second experiment may therefore be somewhat more

realistic.

Figure 10 pictures the resulting estimates of GDP per capita (of the two

experiments). In the very long run both sets of estimates seem to concur, but the model

in both cases predicted a strong increase in the sixty to eighty years after the Black

Death, which is missing from the Malanima estimates. Between 1450 and 1550 the

Malanima estimates and the results of the Cobb Douglas simulation converge again, and

for the rest of the Early Modern period the two approaches give more or less similar

results (but note that if the simulation had followed the decline of real wages after 1750

according to the Allen estimate more closely, the fall in GDP per capita had been larger

than Malanima (2003) estimated). Differences with England in the same period are

striking: there the increase in real wages is higher (but the population decline is also

larger: almost 60% in England versus less than 30% in Italy), and real wages remained

on a relatively high level until the first quarter of the 16th century, whereas in Italy they

declined strongly after 1450. This is also reflected in different trajectories of GDP per

capita: in England the growth between 1350 and 1450 resulted in more or les

permanently higher income leve ls, whereas in Italy the gains were not consolidated after

1450. Apparently, during the century or so after the Black Death (northern) Italy was a
very dynamic economy, with a consistent rate of productivity growth for 70 to 100

years, but this dynamism completely disappeared after c 1450. By contrast, England

remained very dynamic in the long run, and per capita growth was only interrupted by

the population explosion of the 16th century. After about 1600 the same dynamism (in

terms of the growth of total factor productivity) returned in England (and the

Netherlands), whereas no such thing happened in Italy (and in other European

countries).
 Figure 9 Italy: real wages, simulated and estimated, 1300-
1800 (indices 1300=100)

300

250

200

150

100

50

0
1300 1350 1400 1450 1500 1550 1600 1650 1700 1750 1800

Simulated estimated Poly. (estimated)

 Figure 10 Italy: GDP per capita estimated and simulated,
1300-1800 (1310/19=100)

140

130

120

110

100

90

80

70

60

50
1300 1350 1400 1450 1500 1550 1600 1650 1700 1750 1800

simulated (1st) estimated (Malanima) simulated (2nd)

8. Europe 1500-1800

The three countries this paper concentrated on so far, are all rather exceptional: Italy

because of its strong long-term decline, Holland and England because of their dynamic

development. In between these two extremes were the other European countries, of

which the GDP per capita may not have changed very much between 1500 and 1800. In

principle the experiment can be carried out for all countries for which we have long-

term series of real wages and population estimates, but in order to make a long story

short, I experimented with Europe as a whole only. Allen 2001 has estimates for the

development of real wages in (capital cities of) Italy, Spain, France, Austria, Germany,

Poland, the Low Countries and Great Britain, which can be weighted with their

respective populations (taken from Allen 2000) to get a series for real wages in Europe.
The following standard assumptions are used in this version: share of labour 60%, of

capital and land both 20%; cultivated land increases by 5% per century; capital stock is

estimated on the basis of an interest rate in 1500 of 12%; working days per annum

increases from 200 in 1500 to 300 in 1800.

The results are presented in Figure 11; this simulation assumes a very low

investment ratio (of only 2%), and no technological change. The general pattern of the

simulation is rather similar to that of the estimated wage series: rapid decline during the

16th century, stabilization between 1600 and 1750, followed by another decline after

about 1750. The long term development of European wages seems to follow a path that

can be simulated by a Cobb Douglas production function assuming no technological

change and a very low level of new capital formation; the different phases of population

growth (in the 16th century and after 1750) then produce the pattern estimated by Allen.

It is also clear that the divergent development of real wages in England and Holland

where wages began to rise again during the 17th century as a result of productivity

growth has a very limited impact on the European average. The reason for this was

that their combined population was quite small, although as a share of European

population it grew from slightly less than 6% in 1500 to 9% in 1800.

Estimates of the development of European GDP per capita that can be derived

from this experiment all give a rather bleak picture: assuming a 50% increase in

working days per annum over the whole 1500-1800 period gives an in the long run

almost flat curve (Figure 12), with some decline during the 16th century and a small

recovery in the next 150 years. In order to get a little per capita growth (consistent with

the estimates of Table 3), one has to assume an even larger increase in labour input per

head of the population, which is unlikely to have occurred however.

 Figure 11 Europe, real wage: estimated and simulated, 1500-1800
(indices 1500=100)

120

110

100

90

80

70

60

50

40
1500 1550 1600 1650 1700 1750 1800

Simulated Estimated Poly. (Estimated)

9. Conclusion

This paper presented a synthesis of recent research into the long-term development of a

number of European economies in the centuries before the Industrial Revolution, and an

attempt to find out if these estimates are consistent with the (more detailed) evidence

that we have on the evolution of real wages, population, labour input, land input and

real rents, given the assumptions of the Cobb Douglas production function. The overall

conclusion is that these estimates are indeed consistent with this evidence, although a
few times we had to assume quite substantial increases in the quality and per capita

quantity of the labour input to get reconciliation between the real wages series and the

GDP estimates. This would be much more difficult with the estimates published by

Maddison, which generally show much higher rates of economic growth. For Europe as

a whole, for example, Maddison estimates a 56% increase in per capita GDP between

1500 and 1820, which is not consistent with the decline of real wages occurring in the

same period (even my much more modest estimates of 10-25% growth are at the limit

of what might be consistent with the real wage data). The same applies to his

assumption that in the same period income levels in the Netherlands increased by 140%

(against my 50% estimate); especially the near doubling of per capita GDP between

1500 and 1600 is difficult to reconcile with the decline of real wages (and the near

constancy of real rents) during this century.

 Figure 12 Simulated GDP per capita: Europe, Italy and
England, 1300-1800 (England in 1800=100)

120

100

80

60

40

20

0
1300 1350 1400 1450 1500 1550 1600 1650 1700 1750 1800

Euope Italy England

The results of the experiment with the Cobb Douglas production function can be

summarized in Figure 12 and Table 4 (I added guestimates for Europe as a whole for the

1300-1500 period, based on an average of the wages in Italy and England). The

contrasting long-term trajectories of Italy and England are quite clear from this

comparison (which is consistent with the estimates presented in Table 3). Before 1550

Italy has a (much) higher income level than the rest of Europe, and even at about 1600 it

is still consid erably richer than England. The latter country has its transition to modern

economic growth during the first half of the 17th century, when the rest of Europe (with

the exception of the Netherlands) stagnated after the decline due to Malthusian forces

during the 16th century.

Table 4 Estimates of the growth of GDP and GDP per capita in Europe, 1300-1800 (in

average annual growth rates)

1300-1350 1350-1450 1450-1600 1600-1700 1700-1800

Italy, GDP 0.07 0.06 0.01 0.00 0.23

Italy, GDP p.c. 0.09 0.26 -0.25 -0.01 -0.09

England, GDP -0,25 -0,20 0.29 0.71 0.91

England, GDP p.c. 0.37 0.34 -0.11 0.49 0.41

Europe, GDP -0.04 -0.14 0.14 0.09 0.28

Europe, GDP p.c. 0.27 0.18 -0.06 0.00 -0.10

Source: see text.

Potentially even more interesting are the new insights that this experiment has

produced. It was possible to identify a number of periods during which there was a more

or less sustained increase in total factor productivity: Italy and England during the

century or so after the Black Death, England again after the 1610s (and, strikingly

enough, unchanging in tempo between the 1620s and 1800), and Holland between the

1590s and 1670s (and it is probable that a similar process of productivity growth

occurred in Holland between 1350 and 1450, see Van Bavel and Van Zanden 2004). In

the rest of Europe there was no similar increase in productivity between 1500 and 1800;

in fact, income levels could only be maintained because of increases in labour input per

capita.

This experiment also threw some light on the importance of ghost acreages for

economic growth in the core countries around the North Sea. For England, a net

exporter of agricultural commodities before 1750, this was not an important source of
growth. Only after 1750 did imports of agricultural goods (foodstuffs and cotton)

contribute (a bit) to solving the land constraint, but the overall effect on GDP was

limited to a few percent. For Holland the story is clearly different, and international

trade access to foodstuffs and building materials from the Baltic in particular was

essential for its development. A very rough simulation of this effect leads to the

conclusion that these ghost acreages may have increased the level of GDP by perhaps

as much as 10% at the economic peak of the region in the second half of the 17th

century.

The contrast between the North Sea region (Holland and England) and the rest

of Europe that is well known from the literature and from the real wage study by Allen

(2001), who even used the term great divergence to describe it, is also very clear from

this attempt to measure GDP growth. The consistent strong performance of England is

perhaps the most striking result, but also the near stagnation of the rest of Europe

demands explanation. Why were most Europeans, in view of the absence of productivity

growth, unable to improve their institutions, technologies and the quality of their inputs

in a systematic way between 1500 and 1800? Why did the incentives created by a

market economy not produce a systematic search for more efficient solutions or why

was this search so relatively fruitless? Or is such a pattern normal, similar to what can

be observed in other highly developed societies such as China, India, or the Ottoman

Empire during the Early modern period. This normal pattern has been coined Smithian

growth, characterized by an increase of population and income, driven by the expansion

of markets and related processes of specialization and urbanization, but without the

qualitative leaps of sustained productivity growth which can be found in parts of Europe

after the Black Death and during the 17th and 18th centuries (see Bin Wong 1997).
 Another series of questions is related to what happened in other European

countries before 1500. Was, for example, the increase in productivity growth found in

Italy and England (and possibly in Holland) a pan-European phenomenon? Our brief

excursus into the Medieval period already suggested that the GDP growth found in

England an increase of per capita GDP of as much as 50-60% between 1086 and 1500

was probably normal; in fact, large parts of Europe may have developed more rapidly

during the high Middle Ages. Why did growth stop? What happened in northern Italy

during the 15th century that its economy almost came to a standstill, after it had been

perhaps the most dynamic region in the previous centuries (see Epstein 1991 for an

interpretation)?

References

Allen, R.C., 2000. Economic structure and agricultural productivity in Europe, 1300-

1800. European Review of Economic History, 4, pp. 1-27.

Allen, R.C., 2001. The great divergence in European wages and prices, Explorations in

economic history, 38, pp. 411-47.

Bavel, J.P.B. van, and J.L. van Zanden, 2004. The Jump-start of the Holland economy

during the late Medieval crisis, c. 1350 c. 1500. Economic History Review, LVII.

Blomme, J., and Van der Wee, H., 1994. 'The Belgian economy in a Long-Term

perspective', in A. Maddison and H. van der Wee, Economic growth and structural

change. Comparative approaches over the long run on the basis of reconstructed
natinoal accounts. Eleventh International Economic History Congress Milan 1994,

session B 13.

Campbell, B.M.S., 2000. English Seigniorial Agriculture 1250-1450. Cambridge

University Press, Cambridge.

Campbell, B.M.S., 2005. England circa 1290. A view from the Periphery. Unpublished

paper, Belfast.

Carreras, A., 2003. Modern Spain. In J. Mokyr et.al. (eds.) The Oxford Encyclopedia of

Economic History. Oxford U.P., Vol 4, pp. 546-53.

Clark, G., 2002. Land rental values and the agr arian economy: England and Wales,

1500-1914. European Review of Economic History, 6, pp. 281-309.

Clark, G., 2004. The long march of history: farm laborers wages in England, 1209-

1869. Unpublished paper, Davis.

Crafts, N.F.R., 1977. Industrial Revolution in Britain and France: some thoughts on the

question Why was England First. Economic History Review 30, pp. 429-41.

Crafts, N.F.R., 1985. British Economic Growth during the Industrial Revolution.

Clarendon Press, Oxford.

Crafts, N.F.R. and C.K. Harley, 1992. Output Growth and the British Industrial

Revolution: a restatement of the Crafts-Harley view. Economic History Review 45, pp.

703-730.

Deane, P. and W.A. Cole, 1962. British Economic Growth 1688-1959. Trends and

Structures. Cambridge University Press, Cambridge.

Epstein, S.R., 1991. Cities, Regions and the Late Medieval Crisis: Sicily and Tuscany

compared. Past and Present, 130, pp. 3-50.

Epstein, S.R., 2001. The late medieval crisis as an integration crisis. In M. Prak (ed.)

Early Modern Capitalism. London: Routledge, pp. 25-51.

Federico, G., 2002. The world economy 0-2000 AD: a review article. European Review

of Economic History, 6, pp.111-121.

Feinstein, Charles H., 1998. Pessimism perpetuated: real wages and the standard of

living in Britain during and after the industrial revolution, Journal of Economic History,

58, pp. 625-58.

Harley, C.K., 1993. Reassessing the Industrial Revolution. In J. Mokyr (ed.) The British

Industrial Revolution. Westview, Boulder, pp. 171-226.

Krantz, O., 2004. An estimate of Swedish GDP in 1571, in S. Heikkinen and J.L. van

Zanden (eds.) The experience of economic growth. Aksant, Amsterdam, pp. 105-131.

Lindert, P.H. and J.G. Williamson, 1982. Revising Englands social tables 1688-1812.

Explorations in Economic History, 19, pp. 385-409.

Maddison, A., 2001. The world economy: a millennial perspective. OECD, Paris.

Maddison, A., 2003. The world economy: historical statistics. OECD, Paris.

Maddison, A., 2004. Macromeasurement before and after Colin Clark. Working paper at

http://www.eco.rug.nl/~Maddison/articles/colin_clark.pdf

Malanima, P., 1994. Italian Economic Performance, 1600-1800, in A. Maddison and H.

van der Wee (eds.), Economic growth and structural change. Comparative approaches

over the long run on the basis of reconstructed national accounts. Eleventh International

Economic History Congress Milan 1994, session B 13.

Malanima, P., 1998. Italian cities 1300-1800. A quantitative approach. Rivista di storia

economica, XIV, 2, pp. 91-126.

Malanima, P., 2003. Measuring the Italian economy, 1300-1861. Rivista di storia

economica, XIX.

Malanima, P., 2004. Labour, Productivity, Wages in Italy 1270-1913. Paper for

conference Towards a global history of prices and wages Utrecht 2004.

Mayhew, N., 1995. Modelling medieval monetisation. In Richard H. Britnell and Bruce

M.S. Campbell (eds.) A commercialising economy. England 1086 to c. 1300.

Manchester University Press, pp. 55-78.

Milanovic, B., 2004. An estimate of average income and inequality in Byzantium

around the year 1000. Working paper; Economics working paper archive, available at

http://ideas.repec.org/p/wpa/wuwpeh/0501003.html#download

Mokyr, Joel, 1990. The lever of riches. Technological creativity and economic progress.

New York, Oxford U.P.

Officer, L.H.. n.d.. What was the UK GDP then? A data study. Working paper available

at http://www.eh.net/hmit/ukgdp/ukgdpstudy.pdf

Overton, M. and Campbell, B.M.L., 1997. Production et productivit dans lagriculture

anglaise, 1086-1871. Histoire et Msure, 11, pp. 255-97.

Pomeranz, K., 2000. The Great Divergence. China, Europe and the Making of the

Modern World Economy. Princeton U.P.

Snooks, G.D., 1990. Economic Growth During the Last Millenium: A Quantitative

Perspective for the British Industrial Revolution, Working Paper in Economic History

no. 140, Australian National University.

Snooks, G.D., 1994. Great waves of economic change, in Snooks (ed.) Was the

Industrial Revolution necessary? London, pp.43-79.

Snooks, G.D., 1995. The dynamic role of the market in the Anglo-Norman economy

and beyond, 1086-1300. In Richard H. Britnell and Bruce M.S. Campbell (eds.) A

commercialising economy. England 1086 to c. 1300. Manchester University Press, pp.

27-55.
Vries, J. de, 1984.The decline and rise of the Dutch economy 1675-1900' in G.

Saxonhouse and G Wright (eds.) Forms and methods in economic history: essays in

honor of William N. Parker, Greenwich, pp. 149-189.

Vries, J. de, 1994. The Industrial Revolution and the Industrious Revolution. Journal of

Economic History 54, pp. 249-70.

Vries, J. de and Van der Woude, A., 1997. The first modern economy: success, failure,

and perseverance of the Dutch economy, 1500-1815. Cambridge 1997.

Wong, R. Bin, 1997. China transformed: historical change and the limits of European

experience. Ithaca.

Wrigley, E.A., et.al., 1997. English population history from family reconstitution, 1580-

1837. Cambridge.

Wrigley, E.A., 2000. The Divergence of England: the growth of the English econo my in

the seventeenth and eighteenth centuries. Transactions of the Royal Historical Society,

6th series, 10, pp.117-41.

Yun, B., 1994. Proposals to Quantify Long Term Performance in the Kingdom of

Castile, 1550-1800, paper Session 'Economic Growth and Structural Change' (Milan,

1994)

Zanden, J.L. van, 1987. De economie van Holland in de periode 1650-1805: groei of

achteruitgang? Bijdragen en mededelingen betreffende de geschiedenis der Nederlanden

102, pp. 562-609.

Zanden, J.L. van, 1992. Economic growth in the Golden Age, Economic and social

history in the Netherlands IV, pp. 5-26.

Zanden, J.L. van, 1999. Wages and the standard of living in Europe, 1500-1800.

European Review of Economic History, 2, pp. 175-197.

Zanden, J.L. van, 2001. Early modern economic growth: a survey of the European

economy 1500-1800, in M. Prak (ed.) Early Modern Capitalism. London: Routledge,

pp. 69-87.

Zanden, J.L. van, 2002a. The revolt of the early modernists: an assessment', Economic

History Review, LV, pp. 595-623.

Zanden, J.L. van, 2002b. Taking the measure of the early modern economy. Historical

national accounts for Holland in 1510/14, European Review of Economic History 6, pp.

3-36.

Zanden, J.L. van, 2004. Simulating early modern economic growth. Working paper

IISH, available at http://www.iisg.nl/research/jvz-simulating.pdf

Zanden, J.L. van, and A. van Riel, 2004. The strictures of inheritance. State, economy

and institutional change in the Netherlands 1780-1914. Princeton U.P.