European Review of Economic History, I, 153-190. Printed in the United Kingdom © 1997 Cambridge University Press
Around the European periphery
1870-1913: Globalization, schooling and
growth
KEVIN H. O'ROURKEf AND JEFFREY G. WILLIAMSON*
f Department of Economics, University College Dublin, Arts/Commerce
Building, Belfield, Dublin 4, Ireland
$ Department of Economics, Harvard University, 216 Littauer Center,
Cambridge MA 02138, USA
On average, the poor European periphery converged on the rich industrial
core in the four or five decades prior to the First World War. Some, like
the three Scandinavian economies, used industrialization to achieve a
spectacular convergence on the leaders, especially in real wages and living
standards. Some, like Ireland, seemed to do it without industrialization.
Some, like Italy, underwent a less spectacular catch-up, and it was limited
to the industrializing North. Some, like Iberia, actually fell back. What
•accounts for this variety? What role did trade and tariff policy play? What
about emigration and capitalflows?What about schooling? We offer a
tentative assessment of these contending explanations and conclude that
globalization was by far the dominant force accounting for convergence
(arid divergence) around the periphery. Some exploited it well, and some
badly.
1. Centre and periphery: the arguments
Does globalization produce convergence, or does it widen the gaps between
rich and poor nations? It turns out that theory is ambiguous about this
important question, so we need economic history to get the answers. The
late nineteenth century is one good place to look. There was economic
convergence in the Atlantic economy during this period, but the growth
experience around the backward European periphery was hardly uniform.
Some regions, like Ireland, converged on the economic leaders at about the
expected rate; others, like Scandinavia, converged at a ferocious rate; and
still others, like Iberia, failed to converge at all.
There^are, then, two questions that motivate this paper: First, why did
some nations around the periphery do so much better than others?
Second, is the explanation for success or failure always the same? Economists
involved in the convergence debate rarely ask such questions since they are
trained to stop with hypothesis testing. Economic historians demand more
since we want to know which forces mattered most, when, where and why.
154
European Review of Economic History
These take two forms: shared globalization forces, which have a potential
impact on everyone, and idiosyncratic forces, which are country-specific.
Education is a country-specific variable which has assumed great prominence in the historical debate on the European periphery and in the
empirical growth literature. Lars Sandberg has argued that good education
helps explain Swedish catch-up and Gabriel Tortella has argued that poor
education helps explain Latin fall-back. When European nations are
assessed collectively, does schooling persist as an important ingredient of
late nineteenth century catch-up and fall-back, or was it instead a peculiarity
of a few outlying countries?
The globalization forces at work between 1870 and 1913 are well-known
to economic historians: international labour, capital and commodity markets all became increasingly integrated up to the First World War. European
labour migrated to the labour-scarce New World and from low-wage to
high-wage European labour markets; Britain placed enormous amounts of
capital overseas while France and Germany invested large amounts around
the capital-scarce European periphery; transport costs plunged worldwide,
subjecting Europe to an invasion of new world grain and heightened
competition from each other. To what extent does participation in this
growing international economy explain late nineteenth century Scandinavian success? To what extent do Iberian efforts to insulate themselves
from these forces explain their failure? And what about Ireland and Italy,
lying in the middle of this range? How much of the catch-up and fall-back
performance around the periphery can be explained simply by each country's ability and willingness to emigrate?
Section 2 assesses the economic performance of the European periphery
in the late nineteenth century, documenting where convergence and
divergence took place. The important point is not so much whether there
was overall convergence, but rather why some countries did so much better
than others. Thus, Section 3 provides a test of the Sandberg and Tortella
hypothesis, estimating the impact of schooling on performance around the
periphery. Sections 4 and 5 then turn to the roles of emigration and
international capital flows in accounting for the variety around the European periphery, while Section 6 looks at the impact of declining transport
costs and commercial policy responses. A research agenda is offered in
Section 7.
2. Documenting performance around the periphery
This section exploits three sources of evidence, and four variables, in an
effort to gauge late nineteenth century performance around the periphery.
While each of these sources has flaws, they appear to support the same
conclusions.
The first contains purchasing-power-parity adjusted real wages for the
Around the European periphery 1870-1913
155
urban unskilled in sixteen countries, based mainly on the earlier work of one
of the present authors (Williamson 1995), but revised extensively since.
Williamson's sixteen countries include: four New World countries - Argentina, Australia, Canada and the United States; five European industrial
leaders forming the core - Belgium, France, Germany, Great Britain and the
Netherlands; and seven from the European periphery - Denmark, Ireland,
Italy, Norway, Portugal, Spain and Sweden. The data used here repair some
modest errors in the British series, and replace the older Spanish series with
a more recent one from Simpson (1995). Scholars are busy at work revising
the Argentine, Canadian and Dutch series, but they were not available at
time of writing. Finally, some have challenged the Scandinavian real wage
series as overstating the catch-up. This has been true especially of the
Norwegian data, which are based on unskilled wages in public works
construction. The reader is referred to Williamson's original 1995 article for
a defence, but note that we insist on using urban, male, unskilled wage rates
throughout, typically in the building trades. Only by doing so can we be
assured of job and skill comparability across countries.
The second source documents trends in the wage-rental ratio, a relative
factor price index that measures the scarcity of land compared with labour.
The ratio is based on the revised unskilled urban wages already discussed
and farm land values for ten of our countries (O'Rourke et al. 1996):
includes Argentina, Australia, Denmark, France, Germany, Great Britain,
Ireland, Spain, Sweden and the United States).
The third source is Angus Maddison, whose data have been used almost
exclusively by economic historians engaged in the convergence debate.
Maddison's GDP per worker-hour and GDP per capita estimates are for a
sample which overlaps with Williamson's sixteen, except that Maddison
excludes Argentina and Ireland, but includes Austria, Finland and Switzerland (Maddison 1994). However, we replace Maddison's estimates for
Portugal, Spain and Italy with those made more recently by Prados de la
Escosura et al. (1993) and Bardini et al. (1995).*
Before we proceed with our narrative, let us first define the members of
our European sample. Table 1 reports 1870 real wages and GDP per head
for fifteen European countries. The two indices reveal similar rankings. The
industrial core countries had levels of GDP per head 30 per cent higher than
the European average and 67 per cent higher than the periphery. Their real
wages were 86 per cent higher than the periphery. Alternatively, real wages
in the periphery were 46 per cent below the industrial core average and GDP
per head 40 per cent below. Austria and Denmark seem to have been on the
margin between core and periphery: without them, the periphery GDP per
1
Prados de la Escosura et al. (1993) replaced Maddison's data for the Mediterranean three
with alternative estimates (in 1990 US dollars), and their figures are used in Table 1.
Bardini et al. (1995) give even more recent estimates of national growth rates, and those
estimates are used in Table 2.
156
European Review of Economic History
Table 1. Who is in our European Periphery sample?
Country
(1)
(2)
Real wage per
urban worker, 1870
(GEt 1905 = 100)
Real GDP per head
Austria
Denmark
Finland
Ireland
Italy
Norway
Portugal
Spain
Sweden
na (na)
36 (85)
na (na)
49 (115)
1847
1836
1095
(IOI)
(101)
(60)
na
26
32
18
30
28
(61)
(75)
(42)
(70)
1568
1229
(66)
1388
1596
(na)
(86)
(67)
(44)
(76)
(87)
Average
31
(73)
1419
(78)
60
50
(141)
(117)
(136)
(157)
(134)
(na)
1870 (1990 US$)
The European Periphery
The European Industrial Core
Belgium
France
Germany
Great Britain
The Netherlands
Switzerland
Average
Europe
58
67
57
na
795
2572 (141)
1935 (106)
1619
(89)
3115 (171)
2490 (136)
2476 (136)
58 (136)
2368
(130)
43 (100)
1826
(100)
Notes and Sources: Col. (1) from sources underlying col. (1) in Table 2; col. (2) from Prados
et al (1993, Table 2, p. 5).
head would have been 46 per cent below the industrial core average.
Perhaps somewhat arbitrarily, we have thrown both Austria and Denmark
into the periphery group. Thus, the nine members of the European
periphery are: Austria, Denmark, Finland, Ireland, Italy, Norway, Portugal,
Spain and Sweden. The reader will note that the sample excludes east and
southeast Europe simply because the late nineteenth century data are
inadequate for those regions. We know, however, that they were relatively
poor. The evidence of Bairoch (1976a, Table 6, p. 286) suggests that none of
them had levels of GNP per head that were even half that of the core: e.g.,
Bulgaria, 42.3 per cent of the core; Greece, 48.1 per cent; Romania, 40.4 per
cent; Russia, 48.1 per cent; and Serbia, 44.2 per cent. Thus, with the
exception of Portugal, this paper ignores the poorest part of Europe.
Hopefully, as better historical data emerge from east and southeast Europe
they will confirm the assertions which follow.
Around the European periphery 1870-1913
157
Let us start the narrative with the success up North, the spectacular
Scandinavian catch-up on the leaders. Consistent with qualitative accounts,
the evidence in Table 2 confirms that Sweden and Denmark tended to
outperform Norway and Finland, but not by much. While real wages show
Sweden growing considerably faster than Norway, GDP per capita growth
and GDP per worker-hour growth reveal only modest differences between
the Scandinavians. Rapid growth seems to have been common to all four
Nordic countries.
Real wages in Scandinavia grew at rates almost three times those
prevailing in the European core; Swedish workers enjoyed real wage growth
about 2.7 times that of British workers; Danish workers enjoyed real wage
growth about 2.6 times that of German workers; and Norwegian workers
enjoyed real wage growth about 3.8 times that of Dutch workers. In fact, no
Table 2. Relative economic performance of the European periphery in
the late nineteenth century: growth per cent per annum.
Country
(1)
(4)
Real wage
Real GDP
per
worker-hour
1870-1913
(2)
(3)
Wage-rental Real GDP
ratio
per capita
per
urban worker 1870-1910 1870-1913
1870-1913
The European Periphery
Denmark
Finland
Norway
Sweden
Scandinavia
Italy
Portugal
Spain
Mediterranean Basin
with Italy
without Italy
Austria
Ireland
Other Periphery
Periphery
The European Industrial Core
Belgium
France
Germany
Great Britain
The Netherlands
Switzerland
Industrial Core
Europe
2.63
2.85
i-57
na
na
1.44
2.43
2.73
2.60
1.74
0.37
0.44
na
2.45
2.65
1.31
-0.43
1.46
1.45
1.28
0.69
1.11
1.90
1.80
1.65
1.74
1.77
1.33
1.10
1.52
0.85
0.41
na
1.79
1.79
1-73
-0.43
-0.43
na
4-39
4-39
2.32
1.03
0.90
1.46
na
1.46
1.29
1.32
1.31
1.76
na
1.76
1.60
0.92
0.91
1.02
1.03
0.64
na
1.80
0.87
2.54
na
na
1.74
2.07
1.05
1.30
1.63
1.01
1.01
1.20
1.20
1.25
1.24
1.58
1.88
1.23
1.34
1.46
1.46
1.54
na
0.90
1.39
na
na
158
European Review of Economic History
Table 2. Continued.
Country
The New World
Argentina
Australia
Canada
USA
New World
(1)
(2)
Real wage
per
urban worker
1870-1913
Wage-rental
ratio
1870-1910
(3)
Real GDP
per capita
1870-1913
(4)
Real GDP
per
worker-hour
1870-1913
i-74
0.14
1.65
1.04
1.14
-4.06
-3.30
na
-1.72
-3.03
na
0.87
2.29
1.81
1.66
na
1.08
2.31
1.93
1.77
Notes and Sources: All averages are unweighted.
Col. (1): Real wage rate for unskilled urban workers, from Williamson (1995, Table A1.1,
Great Britain revised), except for Spain which now uses Simpson (1995, Tables A.18 and
A. 19, pp. 250-2). In addition, the benchmark for The Netherlands, Portugal, Norway and
Spain has been shifted from 1905 to 1927, as the 1927 data to make the necessary PPP
adjustments are superior to the 1905 data, for these four countries. This change affects the
1870 levels in Table 1 but not the growth rates here. All real wage growth rates are
calculated from estimated log linear regressions 1870-1913 (raw wage data are annual).
Col. (2): Ratio of Williamson's revised wage in col. (1) to land values per unit of farmland.
Sources for the latter are detailed in O'Rourke et al. (1996, Appendix 1).
Col. (3): Gross domestic product in constant prices, per capita, from Maddison (1994,
Table 2-1), except for Italy, Portugal and Spain which use GDP from Bardini et al. (1995,
Appendix Table 1) and population from Mitchell (1978, Table Ai).
Col. (4): Gross domestic product in constant prices, per worker-hour, from Maddison
(1991, Table C.io and 1994, Table 2-1), except for Italy, Portugal and Spain which is
based on Bardini et al. (1995, Appendix Table 1), and on the assumption that worker-hours
per capita evolved the same way in Iberia as in Italy.
other country in our European sample underwent real wage growth even
close to that of Sweden, Denmark or Norway. What was true for real wages
was also true for the wage-rental ratio. While the ratio of wage rates per
worker to farm land values per acre fell everywhere in the New World, it rose
everywhere in Europe (with the exception of Spain). These events reflect the
invasion of grains from the New World (and Russia) which lowered farm
rents and land values in Europe and raised them in the American Midwest,
the Australian outback, the Argentine pampas and, we assume but do not
document, the Ukraine. While the Scandinavian wage-rental ratio seems to
have tracked the British ratio very closely (2.65 vs 2.54 per cent per annum
growth), the ratio rose half again faster in Scandinavia than in the European
core (2.65 vs 1.74 per cent per annum). Consistent with the predictions of
theory (O'Rourke and Williamson 1995b), product per worker-hour documents a less spectacular Scandinavian catch-up than factor prices, but even
these data confirm an impressive growth performance compared with the
Around the European periphery 1870-1913
159
European industrial core (1.77 vs 1.46 per cent per annum). Finally, the
superiority of Scandinavian GDP per capita growth over that of the
industrial core (1.45 vs 1.2 per cent per annum) is smaller than that of real
wages or GDP per worker-hour, but the Scandinavians still did better.
Scandinavia outperformed the rest of Europe (and probably the rest of
the world) in the late nineteenth century, of that there can be no doubt.
What about the rest of the periphery?
Based on Maddison's data, Austria seems to have done about as well as
Scandinavia: GDP per capita and GDP per worker-hour grew almost
exactly as fast (1.46 vs 1.45 and 1.76 vs 1.77). In contrast, while Ireland
certainly obeyed the laws of convergence, she was no over-achiever. Irish
real wages grew twice as fast as they did in the industrial core (1.79 vs 0.90
per annum), but they grew no faster than the periphery average, and at less
than 70 per cent of the Scandinavian rate. On the other hand, the wagerental ratio rose faster in Ireland than in Scandinavia.
As a unit, the western Mediterranean Basin did badly. Gabriel Tortella
(1994a) has recently surveyed performance in the Basin, so we can be brief.
The Iberian peninsula fell far behind the growth rates recorded in the rest of
the periphery. Real wages crawled upwards at about 0.4 per cent a year in
Iberia, while they surged five and a half times as quickly elsewhere around the
periphery: like some Third World countries since 1970, Spain and Portugal
seem to have missed out on a wave of global growth. While the wage-rental
ratio soared at 3.23 per cent a year elsewhere around the periphery, it fell by
0.43 per cent a year in Iberia. The same wide gap appears for GDP per
capita growth, 0.9 per cent per annum in Iberia and 1.42 per cent per annum
elsewhere around the periphery. Maddison's real GDP per worker-hour
data also confirm a poor Iberian performance, but the gap is not quite so
great: 1.31 per cent per annum in Iberia and 1.7 per cent per annum around
the rest of the periphery. According to all three living standards measures,
Iberia grew less rapidly than the core, confirming Iberian 'fall-back'. Italy
does somewhat better, but even she - except for real wages - falls below the
average for the periphery, and productivity grew less rapidly there than in
the core.
2.1 Peripheral performance in the light of convergence
Thus, we have four types of performers around the periphery: the overachieving catchers up (the Nordic four and Austria), the average catcher up
(Ireland), the slower catcher up (Italy), and the under-achieving fallers
behind (the western Mediterranean Basin). This paper is motivated by this
enormous variance, but note that on average there is evidence of convergence in Europe during the late nineteenth century. The periphery grew
faster than the industrial core. Real wages grew almost twice as fast (1.73 vs
0.90 per cent per annum). The wage-rental ratio grew a third faster (2.32 vs
i6o
European Review of Economic History
1.74 per cent per annum). GDP per capita and per worker-hour also grew
faster, although less so than was true for factor prices (1.29 vs 1.20 and 1.60
vs 1.46 per cent per annum). Thus, the periphery was catching up on the
leaders in Europe, especially in terms of workers' living standards. Furthermore, there was factor price convergence between the European periphery and the New World: real wages in the labour abundant periphery grew
about half as fast again as those in the labour scarce New World; and while
the wage-rental ratio around the periphery boomed at 2.32 per cent per
annum, it fell in the New World at about three per cent per annum.
Did Scandinavia grow as fast as economists' convergence models predict?
Or did it grow faster? Was Ireland really 'average'? And how far below the
prediction of a convergence model was the Mediterranean Basin? Figure 1
Swe
12
•
Den
Nor
•
•
1 <u
00
0.8 —-—
g
s
Can
•
0.6 -
2
Xi
2
O
Ger GET"~~-—--._
[Bel •
Fra
0.4 -
•
Por
0.2-
•
•
Net
•
Spa
•
Aus
•
i
i
0 -
2.5
USA
3.5
4
4.5
Log real wage (1870)
Figure 1. Unconditional real wage convergence (1870-1913).
supplies an answer using real wages,2 and Table 3 reports the underlying
unconditional convergence regressions for real wages as well as for GDP per
2
The /3 underlying Figure 1 is -0.177 (Table 3, row 1, entry 1). The rate of convergence is
A = (i/r)ln(j3 + 1)
where t is the time span (43 years) and jS is the coefficient for the log of initial real wages,
income per capita or labour productivity. Actually, we use the term 'speed of
convergence' too loosely in this context. Speed of convergence technically is A times the
initial gap. If A = 0.01, then it would take 70 years to cut the gaps in half. Eliminating big
initial gaps takes a long time, even when fast convergence is at work and even if it is not
interrupted. Previous history matters.
Around the European periphery i8yo-igi3
161
Table 3. Unconditional convergence regressions for the late nineteenth
century.
Sample
Coefficient
R2
N
0.08
16
0.005
O.OI
14
O.OOI
on log of
1870 value
(1) 1870-1913, real wage
Figure 1
(2) 1870-1913, GDP per
worker-hour
-0.177
(1.101)
-0.025
(0.255)
Source: See text. The Maddison GDP per worker-hour sample (N = 14) excludes Austria,
Finland and Switzerland; it includes everything in the Williamson real wage sample (N =
16) except Argentina and Ireland. From O'Rourke and Williamson (1995b, Table 2), except
for row (1) which is revised based on the real wage revisions reported in Table 2.
worker-hour.3 The equation estimated is widely used in the convergence
literature (Barro I99i3 Barro and Sala-i-Martin 1991, 1992; Mankiw et al.
1992; Prados de la Escosura et al. 1993). Thus, in the first row of Table 3 our
measure of late nineteenth century real wage growth is simply regressed on
the logarithm of the real wage in 1870. While the r-statistics are poor in both
cases, the results confirm unconditional convergence - i.e. the coefficient is
negative. Note also that Figure 1 suggests the source of the low r-statistics,
the poor growth performance of Portugal and Spain.4
3
4
The words 'conditional' and 'unconditional' come from the empirical growth literature,
and refer to whether or not convergence equations have been estimated controlling for
schooling and other forces.
The Maddison sample for the regressions in both Tables 3 and 4 exclude Austria,
Finland and Switzerland so as to make it as comparable with the real wage sample as
possible. However, when these three countries are thrown back in to the sample, the
unconditional and conditional convergence results are much poorer. To repeat: the real
wage sample includes Argentina, Australia, Belgium, Canada, Denmark, France,
Germany, Great Britain, Ireland, Italy, the Netherlands, Norway, Portugal, Spain,
Sweden and the United States; the 'overlapping' GDP per worker sample includes all of
these except Argentina and Ireland.
With each revision of Maddison's GDP per worker-hour data, the evidence of
unconditional convergence seems to slip further away. Maddison's 1982 data showed
strong convergence (Williamson 1995), but it did not include Latin American, Iberia or
Ireland. The rate of convergence (A) underlying Maddison's 1982 and 1994 GDP per
worker-hour data is very close to that estimated recently by Prados de la Escosura et al.
(1993, Table 4) for a pre-First World War European sample including Belgium,
Denmark, France, Germany, the Netherlands, Sweden, Switzerland, and the United
Kingdom. When, however, Maddison's revised 1994 sample uses more recent estimates
for Italy, Spain and Portugal (Bardini et al. 1995), unconditional convergence pretty
much disappears (row 2, Table 3). In a sense, the same seems to be true of real wages.
Before the underlying data set was revised, unconditional real wage convergence was
evident in the data (Williamson 1995, 1996). The unconditional convergence now
reported in Table 3 is much weaker.
162
European Review of Economic History
The question of whether there was convergence in the aggregate data is
not however the focus of this paper; rather, we are concerned with why some
countries did so much better than others. Throughout, we frame the
discussion by asking: what explains the catch-up or fall-back of individual
peripheral countries relative to the economic giants of the day. Great Britain
and the US? Was it schooling that mattered? Different rates of emigration
and capital inflows from abroad? The choice between free trade and
protection? Cultural-based mysteries?
3. Schooling, catch-up and fall-back
Sandberg (1979, p. 228) stated clearly that 'Sweden's pre-World War I
economic growth and industrialization was to a significant degree a result of
the country's disproportionately large initial stock of [human capital]'. The
human capital that Sandberg thought mattered most was schooling and
literacy. A decade before Sandberg's paper, Cipolla (1969) had already
offered plenty of evidence supporting the impoverished sophisticate thesis.
Cipolla (1969, Table 6) documented high literacy levels in Scandinavia
compared with the rest of Europe and based on such evidence, argued that
the 'more literate countries were the first to import the Industrial Revolution
(p. 87)'. By 1850, Sweden was the most literate country in Europe and was
the only European country that could measure up to the United States in
that dimension (Sandberg 1979, p. 230).5 Indeed, in a later paper Sandberg
used Cipolla's 1850 qualitative data on literacy to show that the 1850
educational ranking was highly correlated with the 1970 per capita income
ranking, and that up to 1913 'the poor, high literacy countries ... grew the
fastest... [and] the low literacy countries ... growth rate was clearly slower'
(Sandberg 1982, p. 689). Tortella (1994a) has recently elaborated on this
latter observation to find explanations for economic retardation in the
Mediterranean Basin.
How can we evaluate such claims? O'Rourke and Williamson (1995b)
followed Prados de la Escosura et al. (1993) in estimating conditional
convergence equations for the late nineteenth century; that is, they estimated convergence equations such as those reported in Table 3, conditional
on schooling. In this paper we revise the analysis, to take account of the new
real wage data documented above.
The new growth empiricists use school enrolment rates as a proxy for
average educational achievement in analysing conditional convergence in
the post Second World War period. It is hardly a perfect index of average
schooling in the labour force (a stock) since it measures the enrolment rates
5
Others have pursued this connection between education and economic performance in
Sweden since Sandberg's paper appeared, and some are critical of the impoverished
sophisticate hypothesis (e.g. Nilsson and Pettersson 1990, 1992; Markussen 1990).
Around the European periphery 1870-1913
163
Table 4. School enrolment and literacy rates in the i8yosto 1890s.
(1)
Country
(2)
(3)
O'Rourke and Williamson
Literacy
Enrolment
rate
rate
estimates
estimates
Prados et al.
Enrolment
rate
estimates
0.70
0-99
0.89
0.98
0.98
0.51
0.15
0.47
0.48
0.96
O.98
0.47
0.38
0.42
0.40
O.49
The European Periphery
Denmark
Finland
Norway
Sweden
O.IO
0.64
0.65
Scandinavian Average
with Finland
without Finland
Italy
0.52
0.66
Spain
0.37
0.23
0.46
Mediterranean Basin
with Italy
without Italy
Austria
Ireland
Other Periphery
Periphery
0-35
0.35
0.59
0.45
0.52
0.47
0.42
0.40
0.66
0.24
0.42
0.91
na
o.79
o.74
0.42
0.56
0.86
0.96
0-97
0.96
0.97
0.99
0-95
0.83
0.41
0.55
0.51
0.40
0.47
0.57
0.49
0.41
0.46
0-97
na
0.84
0.80
0.90
0.93
0.88
na
na
na
0.69
0.86
0.80
0.92
na
na
Portugal
The European Industrial Core
Belgium
France
Germany
Great Britain
The Netherlands
Switzerland
Industrial Core
Europe
The New World
Argentina
Australia
Canada
USA
New World
with Argentina
without Argentina
0.80
0.73
0-53
0.65
0-77
0.67
0-55
0.20
0.26
0.16
0.32
0.25
O.35
Notes and Sources: Based on O'Rourke and Williamson (1995b Table 3).
of children (a flow). Table 4 also offers estimates of average literacy,
theoretically a far better index of average schooling even though in this case
164
European Review of Economic History
it applies only to young adults. Sad to say, there is reason to doubt the
quality of the literacy data. Indeed, this is exactly why the new growth
empiricists prefer school enrolment rates - quality and availability. Note also
that Table 4 offers two late nineteenth century enrolment rate estimates:
column (3) is taken from Leandro Prados and his collaborators; column (1)
reports our own estimates. While each of these three measures of schooling
is imperfect, they appear to tell roughly the same story.
Consider first our enrolment rate estimates in column (1) of Table 4. If
we exclude Finland from the Scandinavian average, then the impoverished
Scandinavian three measure up very well with the rest of Europe (0.66/0.55
or 20 per cent above the European average). So too does Austria. Ireland
does less well, falling a good bit below the rest of Europe (0.45/0.55 or 18 per
cent below). Italy does even worse, 33 per cent below the European average,
while the Iberian two bring up the rear at 36 per cent below average. Now
consider the literacy rate estimates in column (2): the European figures are
those reported for (mainly young adult) immigrants by United States
authorities in the 1890s (with the exception of Spain), while the New World
estimates are for adult residents. According to this measure, all four of the
Nordic countries are now well above the European average, 16 per cent
higher. Once again, Italy is below average, this time 43 per cent lower. The
Iberian two bring up the rear again, 52 per cent below average. So far, the
two measures of schooling are surprisingly consistent. Ireland is the deviant.6 Irish literacy was far better than her enrolment rates; indeed, the
former is ten per cent higher than the European average, while the latter is
18 per cent lower.
There are clearly huge potential problems with the literacy data, since
emigrants may have self-selected in terms of education. In fact, by this
measure Ireland was more literate than the US during the period, an absurd
result. Thus, while we perform all tests using both measures, in the text we
emphasise the results using enrolment data, which have the additional
advantage of being more directly comparable with the results of those
modern growth studies already referred to.
There was certainly variety in school enrolment and literacy rates around
the European periphery. The Scandinavian countries had a bigger education
endowment than they could, in some sense, afford.7 The Iberian countries
6
7
The same is true of Great Britain, which suggests it may have something to do with how
students were counted in the United Kingdom compared with the Continent;
alternatively, literacy may have been higher among UK emigrants, relative to the
population as a whole, than was true on the Continent.
Markussen (1990, p. 37) has stressed that the Nordic countries were unique in that there
was a long lag, perhaps 100-150 years, between development of reading and writing
skills. Indeed, while their reading skills and enrolment rates are well above what one
would expect for poor countries (Table 4), Markussen (1990, Table 1) shows that they
were well below in writing skills at least based on per capita letters and postcards sent.
Around the European periphery 18JO-1913
165
had a smaller education endowment than they could, in some sense, afford.
Ireland, Italy and Austria were somewhere in between. This schooling
endowment variety around the impoverished European periphery must have
been driven by non-economic forces, embedded in a path-dependent history
prior to the late nineteenth century. The interesting questions, however, are
these three: Does schooling explain much of late nineteenth century
convergence in the Atlantic economy and European periphery? Does
schooling explain much of the growth differentials around the European
periphery? Does schooling explain much of Scandinavian, Italian and Irish
catch-up on the European leaders, or of Iberian fall-back?
Some answers to the first question - does schooling account for much of
the catch-up in our sample of countries? - appear in Panel A of Table 5
where convergence equations are conditioned by schooling, the latter
proxied first by enrolment rates (the standard proxy) and second by literacy
rates. The conditional convergence equations were estimated on both real
wage and GDP per worker-hour data. The contribution of schooling to
GDP per worker-hour growth is statistically significant in both cases,
supporting the view that schooling was important to late nineteenth century
growth. Schooling also 'conditioned' real wage convergence in the late
nineteenth century, although literacy does better in terms of significance
than does enrolment. Poor countries well endowed with schooling caught
up faster than those poorly endowed, presumably because their 'social
capabilities' were better established (Abramovitz 1986). That is, they were
better able to exploit open economy and globalization effects or were better
able to absorb new technologies transferred from the leaders. Furthermore,
when conditioned by schooling, the rate of real wage convergence (X) rises
from 0.5 to 0.7 or 1 per cent per annum, and the rate of GDP per workerhour convergence rises from 0.1 per cent per annum to 0.4 or 0.8 per cent
per annum.
What we really want to know, however, is whether schooling played a
central role in accounting for Scandinavian, Italian and Irish convergence,
and for Iberian divergence. We can find out by asking a second question:
How much of each country's 'deviant' growth performance between 1870
and 1913 was due to each country's 'deviant' schooling performance? As the
notes to Panel B in Table 5 indicate, 'deviant' growth is defined as the
residual left over after controlling for initial real wage levels, while 'deviant'
schooling is simply how much it exceeded the average. Panel Bi answers the
question.
In fifteen of thirty-two cases schooling didn't matter at all. These were
almost always European industrial leaders or rich New World countries
who, presumably, had already fulfilled some minimum schooling precondition. But schooling did matter for about a tenth to a third of deviant
good growth among the Scandinavian three, the bigger numbers for literacy
and the smaller for enrolment: good schooling accounted for 7-31 per cent
166
European Review of Economic History
of deviant good growth in Sweden; for 9-29 per cent in Denmark; and for
8-39 per cent in Norway. Education did not account for a third of late
Table 5. Conditional convergence for the late nineteenth century: adding
schooling.
A. Convergence regressions:
Sample
Coefficients on:
log
1870
value
Ai. Using enrolment rate estimates
1870-1913, real wage
-O.258
(1.352)
1870-1913, GDP per worker
—0.277**
(3.361)
R2
NX
log
schooling
variable
O.192
(O.813)
0.446*
(4.537)
0.12
16
0.007
O.65
14
O,OQ,8
A2. Using literacy rate estimates:
0.534**
O.OIO
16
0.30
1870-1913, real wage
-0.350**
(2.026)
(2.076)
0.292**
O.27
-0.167
1870-1913, GDP per worker
0.004
14
(1.479)
(2.016)
Notes: ^-statistics in parentheses and * = significance at 1%, ** = 5%, and *** =
10%. See text on sample.
B. Convergence impact:
Country
Real wage growth using:
Enrolment
Bi. Share of'deviant' growth explained by schooling (per cent):
Argentina
none
Australia
none
Belgium
none
Canada
22
Denmark
9
France
none
Germany
none
Great Britain
14
Ireland
none
Italy
none
Netherlands
none
Norway
8
Portugal
26
Spain
6
Sweden
7
USA
all
Literacy
none
none
none
21
29
none
none
none
44
all
none
39
50
57
31
41
Around the European periphery 1870-1913
167
Table 5. Continued.
Country
Real wage growth using:
Enrolment
Literacy
B2.1. Share of growth gap, Periphery versus Britain, explained by schooling:
Denmark
8
5
Norway
6
5
Sweden
5
4
Ireland
none
none
Italy
none
none
Portugal
(58)
(all)
Spain
(10)
(all)
B2.2. Share of growth gap, Periphery versus USA, explained by schooling:
Denmark
none
9
Norway
none
9
Sweden
none
8
Ireland
none
5
Italy
none
none
Portugal
(94)
(all)
Spain
(51)
(all)
Notes: The following equation underlies the results given above in Panel Bi:
(y - y) - (3i(w - w) = @2(e - e) + e
where y = total real wage growth (43 years times growth rate)
w = log 1870 real wage
e = schooling variable (enrolment or literacy)
/3i is the coefficient on log 1870 wage in the regression of wage growth on initial wage and
enrolment (Panel Ai above) or literacy (Panel A2 above). @2 is the coefficient on the
schooling variable in the same regression. Variables with 'bars' refer to sample averages.
The list of countries in each regression sample is given in Panel Bi with the country-specific
results. The left side of the equation represents residual above or below average growth in
wages net of the initial wage. The right side is a calculation of the amount of wage growth
due to above or below average levels of the education variable (plus a residual term). By
dividing the schooling effect on the right side by the left side, we obtain the percentage of
above or below average 'residual' growth in wages attributable to above or below average
levels of schooling.
The results in Panel B2 are calculated by dividing the growth gap in question into /32 times
the difference in the log of education in the two countries being compared.
Parentheses around the Iberian entries in Panel B2 indicate that it is Iberian divergence^ not
convergence, that is being explained. Note that the results are based on unrounded growth
rates, whereas the growth rates in Table 2 have been rounded.
nineteenth century Scandinavian growth. Education didn't even account for
a third of Scandinavian catch-up. Education did account for between a tenth
and a third of the residual growth after controlling for initial 1870 conditions.
168
European Review of Economic History
For most of these sixteen countries, literacy and enrolment tell much the
same story. For some - like Ireland, Italy and Spain - the differences in
Panel Bi are marked, the Italian case being the most extreme. Such
differences alert us to problems of data quality; for reasons already stated,
we focus on the results using enrolment rates. Not surprisingly, education
cannot account for above-average growth in Ireland or Italy; while belowaverage education can account for 16 per cent of below-average Iberian
growth (using the suspect literacy data would boost this figure considerably,
just as is true in the Scandinavian case).
What about the third question? That is, how much of the gap in real wage
growth between each of these seven countries around the European periphery and two leaders - Britain and the USA - was due to schooling? The
answers are offered in Panels B2.1 and B2.2 of Table 5.
Start with the Scandinavian three, where the schooling thesis has had the
strongest following, and with the convergence on Britain; and let us, once
again, focus on the enrolment results. Here the figures are: about eight per
cent for Denmark, six per cent for Norway and five per cent for Sweden or
an overall Scandinavian average of six per cent. While these figures may
seem to confirm Sandberg's thesis, they are very small (and the figures are
smaller still if literacy rates are used). Furthermore, schooling explains even
less of the Scandinavian convergence on the United States, indeed none of
the convergence when education is measured by enrolment rates. Now
consider the Iberian Peninsula relative to Britain. Here the figures are ten
per cent for Spain and 58 per cent for Portugal; these imply an overall
average of 34 per cent for the Iberians. That is, one third of the Iberian fallback behind Britain was due to very poor schooling. Schooling accounts for
even more of the Iberian fall-back behind the USA, over 70 per cent. In
contrast, schooling explains none of the Italian real wage catch-up on Britain
or the USA. It looks like Tortella is only partly right: schooling mattered in
only one part of the Mediterranean Basin. Moreover, schooling explains
none of the Irish catch-up on Britain and the USA.
Schooling mattered to catch-up and fall-back around the European
periphery, but its impact was limited to Scandinavia and Iberia, and in only
one case, Iberia, did it matter much. Globalization forces mattered much
more.
4. The impact of emigration
Did emigration from the European periphery help create enough labour
scarcity at home to account for much of the catch-up?
We start with Scandinavia where emigration rates reached their peak in
the 1880s, at that time among the highest in Europe. The rate for the decade
was 95.2 per thousand of the population in Norway, 70.1 per thousand in
Sweden and 39.4 per thousand in Denmark (Hatton and Williamson 1994a,
Around the European periphery 1870-1913
169
Table 1.1); Sweden lay in the middle of the Scandinavian range. Emigration
went through booms and busts, but by 1910 the Danish population was 11
per cent below what it would have been in the absence of emigration over
the four decades following 1870, the Swedish population was 15 per cent
lower, and the Norwegian population 19 per cent lower (Taylor and
Williamson 1994, Table 1). Since emigration favoured young adults with
high labour force participation rates, the impact on the home labour force
was even bigger than on the home population: the Swedish labour force was
perhaps 18.1 per cent smaller in 1910 than it would have been in the absence
of emigration (O'Rourke and Williamson 1995a, Appendix Table 2.1).
With the estimated labour force impact in hand, a recent paper of ours
posed the following question (O'Rourke and Williamson 1995a): How much
of the Swedish catch-up could be assigned to mass migration, the latter
including both the emigrations from Sweden and the migrations experienced
by the leaders, Britain and the United States? Using computable general
equilibrium (CGE) models, we estimated that mass migration accounted for
more than ten per cent of the rapidly contracting wage gap between Britain
and Sweden (O'Rourke and Williamson, 1995a, Table i). 8 This estimate is
Table 6. The impact of mass migration on convergence around the
periphery, 1870-1910.
Country
(1)
(2a)
Net labour
migration rate
Cumulative
impact on
labour force
1870-1910,
per 000 per
(2b)
191O3 per cent
(3a)
Impact of
migration
(3b)
1870-1910
on real wages in
1910, per cent
annum
The Poor European Periphery
Denmark
Norway
Sweden
Italy
Portugal
Spain
Ireland
-3.67
-6.93
-5-55
-14
-24
-12
-22
-20
-18
-12.21
-1.40
i-53
-14.84
-39
-5
-6
-45
The Rich Old and New World
-2.97
Great Britain
United States
+5.3I
+ 12.4
+ 17.6
+ 13.6
+49-3
+5-3
+5-3
+56.3
+ 8.2
+ 15.0
+ 12.3
— 11
—10
+9.9
+6.6
+24
+21
-13.0
-15.1
Whereas partial equilibrium models look at single markets in isolation, general
equilibrium models explore the simultaneous interaction of all goods and factor markets
in an economy. Computable general equilibrium models are fitted to real-world data in a
benchmark year, and permit the quantification of those effects identified by theory. A
discursive treatment of CGE models can be found at http://www.gams.com/solvers/
mpsge/orourke .htm
170
European Review of Economic History
Table 6. Continued.
Country
Denmark
Norway
Sweden
Italy
Portugal
Spain
Ireland
Share of wage
convergence
1870-1910
on Britain
explained, per cent
(4a)
(4b)
Share of wage
convergence 1870-1910
on USA
explained, per cent
(5a)
(5b)
4-7
15-5
5-4
all
(58.6)
(23.3)
all
49-3
65.1
39-7
all
(none)
(none)
all
47.0
65.4
41.4
all
(none)
(none)
all
3-i
17-3
8.4
all
(16.6)
(6.6)
all
Notes and Sources: Cols. (1) and (2a) arefromTaylor and Williamson (1995, Table 1, p. 23).
Col. (2b) is from Taylor and Williamson (1994, Table I, p. 26).
Col. (3a) isfromTaylor and Williamson (1995, Table 4, p. 26), based on Col. (2a).
Col. (3b) is from O'Rourke and Williamson (1995a, pp. 178-80; I995d, Appendix Table
6.3) and the assumption that the Swedish wage-emigration elasticity applies to Denmark
and Norway as well.
Cols. (4) and (5) take actual wage convergence 1870-1910 from Williamson (1995, Table
A2.1, Great Britain, Italy, Norway, Portugal, Spain revised), while the migration impact is
based on cols. (3a) and (3b). Since there are no CGE results for the Mediterranean
countries and Ireland, the entries for these countries in (4b) and (5b) are based on column
(3a), and the British and US results on (3b). Parentheses around the Iberian entries in cols.
(4a)-(5b) indicate that it is Iberian divergence that is being explained. Note that cols. (4)
and (5) are based on unrounded real wages, while Table 1 and Williamson (1995) report
rounded real wages.
revised a bit downwards in Table 6 (col. 4b) to 8.4 per cent.9 That is, the
Anglo-Swedish wage gap fell by 144 percentage points between 1870 (138.3
per cent) and 1910 (-5.7 per cent), and migration accounted for 12.1
percentage points of it. The reason why the figure isn't a lot bigger is that
Britain recorded an impressive emigration rate too. But what about Swedish
catch-up with the United States, the country which absorbed most of the
Swedish emigrants? The immigration rate in the late nineteenth century
United States (of Swedes and non-Swedes combined) served to make the
1910 labour force there 21 per cent higher (Table 6, col. 2b) than it would
have been in its absence, making urban wages 15.1 per cent lower (Table 6,
col. 3b) than they would have been without the immigration. The AmericanSwedish wage gap fell by 243.4 percentage points between 1870 (312.7 per
9
The difference is due to the fact that Table 6 uses annual real wage data for 1870 and
1910, while our earlier estimate used three year averages centered on 1870 and 1910.
While we prefer the three-year averages, we cannot get them for all the countries around
the periphery.
Around the European periphery 1870-1913
171
cent) and 1910 (69.3 per cent), and mass migration accounted for 41.4 per
cent of it (100.7 percentage points).
So much for Sweden. What about the rest of Scandinavia? While Swedish
emigration after 1870 served to diminish the Swedish labour force in 1910 by
a little more than 18 per cent, it served to diminish the Danish labour force by
12 per cent and the Norwegian labour force by 22 per cent (Taylor and
Williamson 1994, Table 1). Given these migration-induced labour force
effects, and assuming that the wage-elasticity estimated for Sweden
(-0.6796) applies to the rest of Scandinavia as well, then we can estimate the
impact of mass migration on Danish and Norwegian convergence too - the
rest of Scandinavia 'viewed in the Swedish mirror' if you will (O'Rourke and
Williamson 1995b). Columns (4b) and (5b) in Table 6 suggest that mass
migration must have contributed less to Danish convergence on Britain (3.1
per cent) and America (47 per cent), but more to Norwegian convergence on
Britain (17.3 per cent) and America (65.4 per cent).
What about elsewhere around the European periphery? Since it seems
inappropriate to view the rest of the periphery in the Swedish mirror, and
since there are no CGE models yet available for the Mediterranean Basin,
we elect instead to use the work provided by Taylor and Williamson (1995)
who offered alternative estimates of the impact of the mass migrations
1870-1910 on the labour force in all nine countries listed in the first panel of
Table 6 (col. 2a) and assessed its impact econometrically on 1910 real wages
(col. 3a). As the reader can verify, the two estimates are quite close for the
five overlapping countries (col. 2a versus 2b, and col. 3a versus 3b). We shall
use the estimates labelled 'a' in everything that follows.
Note first that emigration rates from the poor periphery ranged enormously: they were huge for Ireland and Italy; they were very large for
Norway and Sweden; they were modest for Denmark; and they were tiny for
Portugal and Spain. The cumulative impact on the labour force at home also
varied enormously: it served to lower the Irish 1910 labour force by almost a
half (45 per cent) and the Italian 1910 labour force by more than a third (39
per cent); it served to lower the Scandinavian labour force by from 14 to 24
per cent; but it served to lower the Iberian labour force by onlyfiveto six per
cent. Note further that rich Britain also emigrated, so that her labour force
was diminished by 11 per cent, a much bigger impact than was true of Iberia.
Finally, immigration appears to have augmented the United States labour
force by 24 per cent.
Column (3a) records the impact of these mass-migration-induced labour
force changes on real wages, and the second panel of Table 6 reports the
bottom line: namely, the share of the observed real wage convergence of the
European periphery on Britain and America accounted for by mass migration.
The figures for Ireland and Italy are huge: mass migration accounted for
all of the convergence for those two countries. The amazing characteristic of
172
European Review of Economic History
these two countries is that they seem to have relied exclusively on emigration
to achieve some convergence on the leaders.10 We should be sensitive to the
possibility that it was unimpressive industrialization at home which helps
account for those high Irish and Italian emigration rates, but recall that
Table 5 (Panel B2) suggested that schooling could not account for any of the
convergence of these two countries on Britain or the USA.
Portugal and Spain, on the other hand, were unable to exploit emigration
possibilities. A significant share of their divergencefromBritain is explained
by 'under-emigration' (about 7-23 per cent for Spain and about 17-59 P e r
cent for Portugal). This is not to say that more fundamental problems
underlay Iberian performance. Even if the Spanish and Portuguese emigration rates had been large enough to induce the same labour force impact as
was true for Norway, these two countries still would have undergone
retardation relative to Britain. The 'none' in parentheses under cols. (5a)
and (5b) simply means that while mass migration should have helped
produce Iberian convergence on the United States, divergence factors
overwhelmed these forces. That is, mass migration cannot explain any of the
Iberian fall-back on the USA.
Scandinavia lies in between these extremes: on average, a little less than a
tenth of the Scandinavian convergence on Britain, and a little more than half
of the convergence on the United States, was due to mass migration. The
contribution of emigration to convergence was biggest for the poorest
country, Norway, and smallest for the richest country, Denmark.
Mass migration explains a very large share of the convergence and its
absence around the European periphery. Why did emigration rates vary so
much? Why did Iberia fail to exploit this powerful source of catch-up?
Hatton and Williamson (1994b) have shown that a culture-specific 'Latin'
explanation is not needed to account for those low rates in Iberia, but rather
that a common model of European emigration can do the job quite nicely.
The model is complex, but the key insight that applies here is the following:
since long distance moves are expensive, desperate poverty delays (free)
emigration. Labour in the poorest parts of the periphery couldn't finance the
move and thus had lower emigration rates (i.e. the Portuguese), while
labour in the less poor parts of the periphery could finance the move and
thus had higher emigration rates (i.e., the Irish). Furthermore, those who
sent out emigrants first in the 1840s and 1850s could use the remittances
from those pioneers to finance the moves of others following later. Thus, by
the 1880s it was the Irish and the Scandinavians who were best able to
exploit emigration as a convergence force, not the Iberians who had the
most to gain from mass migration.
Mass migration is one aspect of late nineteenth century globalization that
mattered a great deal. The problem, of course, is that the mass migrations
10
For more on the Irish experience, see Williamson (1994) and O'Rourke (1995).
Around the European periphery 1870-1913
173
must have been intimately related to international capital flows, trade and
industrialization.
5. The impact of international capital flows
From a global perspective, international capital flows were a force for
divergence in the late nineteenth century. After all, low wage Europe was a
net capital exporter, while the high wage New World was a net capital
importer. To explain this apparent paradox, we need not appeal to externalities (Lucas 1990), or to other new growth theory exotica; rather, we need
only appeal to the existence of a third factor, land, and the role of the
overseas frontier, pulling both labour and capital from poor to rich countries, as historians have long recognized.
Nevertheless, there were also substantial capital flows within Europe
during the late nineteenth century, with France and Germany taking the
lead in lending to the European periphery. Less than 6 per cent of British
investments were in Europe at the end of 1913 (Feis 1930, p. 23); but the
figure for France was 61 per cent (p. 51), and the figure for Germany was 53
per cent (p. 74). To what extent did these capital flows account for the
catch-up by the European periphery?
Since Sweden absorbed exceptionally heavy doses of foreign capital, we
begin the periphery assessment there. Foreign capital was directed into
Swedish cities and the railroads, and it was in response to government
demand. Most of these capital inflows were used to finance social overhead
construction and France was the main market for the Swedish bond issues.
A previous paper (O'Rourke and Williamson 1995a) showed how important
British capital exports, and US and Swedish capital imports, were for their
respective capital stocks. As Table 7 indicates, capital imports over the four
decades following 1870 served to make the 1910 Swedish capital stock 50.1
per cent bigger than it would have been in its absence. Capital exports
served to make the 1910 British capital stock 20.4 per cent smaller than it
would have been in its absence. The United States was a much more modest
capital importer than was Sweden (capital inflows only augmented its 1910
capital stock by 0.3 per cent), so global capital markets should have
contributed to Swedish catch-up on America, although much less than in
the Anglo-Swedish case.
Once again using CGE models, we estimated that international capital
flows accounted for more than a half of the decline in the Anglo-Swedish
wage gap, and more than four-tenths of the decline in the US-Swedish wage
gap (O'Rourke and Williamson 1995a, Table 1). The figures reported in
Table 7 are a bit smaller, 43 per cent and 34 per cent respectively.11 The
11
To repeat, the revision is due to the fact that we are not using three-year averages for
1870 and 1910, as we did in the earlier paper.
174
European Review of Economic History
results thus appear to support Heckscher's contention that the capital
import between i860 and 1910 'was a vital prerequisite for the country's
rapid economic upswing' (Heckscher 1954, p. 210).
What about the rest of Scandinavia? The contribution of foreign capital
imports to Danish convergence was likely to have been considerably smaller
since they financed a smaller share of domestic accumulation there (Johansen 1985, pp. 230-2; Hansen 1970, pp. 59-64; Jorberg 1970, pp. 478-9).
We have estimated that foreign capital imports served to make the 1910
Danish capital stock 16.3 per cent bigger than it would have been in its
absence (O'Rourke and Williamson 1995b). Norwegian capital imports
were also smaller than for Sweden (Riis and Thonstad 1989). Although
Norwegian capital imports were even larger than for Sweden after 1890,
Norway was actually a net capital exporter 1870-1890, so the net impact of
foreign capital on the 1910 Norwegian capital stock was to raise it by 'only'
17.4 per cent. Thus, the contribution of capital imports to Norwegian wage
convergence on Britain or America, while still big, was smaller (about 35 and
20 per cent) than to Swedish wage convergence (about 43 and 34 per cent).
The figures for Danish convergence are similar to those for Norway (about
30 and 16 per cent).
Capital flows thus made a substantial contribution to Scandinavian
convergence on the core during the late nineteenth century. What about the
Celtic and Mediterranean peripheries? Did international capital flows make
a powerful contribution to living standard improvement in these countries
too? It is difficult to know, for the simple reason that good balance of
payments data are only rarely available for this part of Europe. However,
such data as we have tell a surprising story.
We do not know for sure whether Ireland exported or imported capital
during the late nineteenth century since trade statistics were not collected
after the customs union with Britain in the mid-i82os. The indications,
however, are that post-Famine Ireland ran trade surpluses. Indeed, in his
evidence to a 1895 Royal Commission, Robert Giffen estimated that Ireland
ran a trade surplus of £5.5 m. in 1893.12 One cannot conclude from this fact
alone that Ireland was running current account surpluses, since there was
also an 'economic drain' consisting of rents remitted to absentee landlords
and excess taxation paid to the British Exchequer.13 However, Giffen
12
Royal Commission on the 'Financial Relation Between Great Britain and Ireland,'
Second Volume of Minutes of Evidence, London (HMSO), 1895 [C. 8008], p. 174
(Supplement to Table III).
When official Irish trade statistics began to be compiled in 1904, they showed a mixed
pattern but with deficits more common up to 1913. Report on the Trade in Imports and
Exports at Irish Ports During the Year Ended 31st December, 1914, Department of Agriculture
13
and Technical Instruction for Ireland, Dublin (1916) [Cd. 8208], p. vi.
See for example the discussion in the 1895 Royal Commission's minutes of evidence (op.
cit.: 3-4) or Solar (1979, p. 24).
Around the European periphery 1870-1913
175
concluded that Ireland was probably a net capital exporter and modern
Irish historians have speculated that 'Ireland's position on capital account
... moved from net debtor to net creditor status' after i860 (Kennedy 1995,
p. 108).
Thus, Ireland probably experienced net capital outflows during the late
nineteenth century. Since the US was a net capital importer, capital flows
were clearly a force for Irish-American divergence, rather than convergence;
since one can only assume that Britain exported even more capital than
Ireland, capital flows must have been a (small) force for Anglo-Irish
convergence. Capital flows may have implied both Irish catch-up and fallback, depending on which leader is being considered.
A similar picture emerges from official Italian statistics. True, the
merchandise trade account was substantially negative throughout the
period, but invisible earnings (tourism and shipping) helped offset this to a
considerable extent, and in some years were enough to lead to a trade
surplus. Net factor income from abroad was negative in the nineteenth
century, but positive after 1900, as emigrants' remittances more than offset
income earned by foreign capital in Italy (Zamagni 1993, pp. 126-7).
Official balance of payments statistics reveal substantial capital imports in
the 1860s and late 1880s, and substantial capital exports from 1894 to
1907.14 Over the entire period between 1870 and 1913, the official statistics
suggest that there were net capital exports from Italy, a force for divergence
rather than convergence.
However, the Italian official statistics have been questioned by Giovanni
Federico, among others. Certainly the qualitative literature emphasizes
capital imports, especially in the 1880s (when state bonds were sold) and the
turn of the century (when direct investments became more important).
These uncertainties are reflected in Table 7.
The conventional wisdom for Spain has been that the country ran large
deficits on current account; and Broder's (1976) estimates indicate capital
inflows throughout the period. However, Prados de la Escosura (1988, pp.
188-97) has challenged this view. His estimates suggest almost continuous
merchandise trade surpluses from 1875 to 1912 (Prados de la Escosura,
1988, pp. 252-4). More recent unpublished estimates suggest current
account deficits between 1870 and 1890, surpluses after the depreciation
and tariffs of 1891, deficits around the turn of the century, and surpluses in
the decade before the First World War. Net capital inflows must have been
very small; even Broder's figures suggest that gross inflows accounted only
for around seven per cent of gross domestic fixed capital formation between
1890 and 1913.15
Current account data are also lacking for Portugal. Lains (1992) has
14
15
The Italian capital imports figures are given in Fenoaltea (1988, Table 4, pp. 620-21).
Note that Tortella (1994b) has questioned Prados' trade figures.
176
European Review of Economic History
Table 7. The impact of international capital flows on convergence
around the periphery, 18J0-1910.
Country
(1)
(2)
Cumulative impact
on capital stock, per cent
Impact of capital flows,
1870-1910 on real wages
in 1910, per cent
The Poor European Periphery
Denmark
Norway
Sweden
Italy
Portugal
Spain
Ireland
+ 16.3
+ 17.4
+50.1
small negative?
small positive
small positive
negative
The Rich Old and New World
Great Britain
-20.4
United States
+0.3
Country
Denmark
Norway
Sweden
Italy
Portugal
Spain
Ireland
(3)
Share of wage convergence
1870-1910 on Britain
explained, per cent
30.0
35-i
43.0
positive
(none)
(none)
small positive
+ 8.2
+8.8
+25.2
small negative?
small positive
small positive
negative
-7-3
+0.1
(4)
Share of wage convergence
1870-1910 on USA
explained, per cent
16.3
20.0
34-0
none?
(none?)
(none?)
none
Notes and Sources: Cols. (1) and (2) are taken from the text.
Cols. (3) and (4) are derived from Williamson (1995, Table A2.1, Great Britain, Italy,
Norway, Portugal, Spain revised) and column 2. Three of the Italian entries, and two of the
Iberian entries, have question marks, indicating substantial uncertainty even as to the sign
of the effect. Capital flows explain none of the Iberian fall-back behind the US if capital
inflows raised Iberian capital stocks by more than 0.1 per cent. While the evidence for
Ireland is also weak, it is strong enough, we feel, to make unambiguous statements.
Parentheses around the Iberian entries in cols. (3) and (4) indicate that it is Iberian
divergence that is being explained. Note that cols. (3) and (4) are based on unrounded real
wages, while Table 1 and Williamson (1995) report rounded real wages.
revised the official trade statistics, making the balance of trade deficits much
smaller than official estimates had indicated. Moreover, emigrant remittances were also an important component of the Portuguese balance of
payments. Figures by Reis (1991) suggest capital inflows between 1865 and
1890. On the other hand, Salazar (1916) claimed that capital fled Portugal
after the 1891 financial crisis, that it returned after the 1902 agreement with
Around the European periphery 1870-1913
177
foreign creditors; and that it fled again in 19075 in response to the dictatorial
government ofJoao Franco.16 The safest assessment would seem to be that
capital imports can only have made a relatively small contribution to the
Portuguese capital stock in the late nineteenth century.17
There are two main conclusions to be drawn from this brief tour around
the European periphery.18 First, capital flows probably did not greatly
reduce wage gaps between the US and the European periphery, except in
the case of Scandinavia; but large British capital outflows meant that
international capital markets were reducing wage gaps between Britain and
the entire periphery. Second, the development of global capital markets did
not by itself guarantee that capital would seek out cheap labour. Capital
inflows may have made an important contribution to Scandinavian development, but they made no contribution at all to Irish (and possibly Italian)
catch-up, and only a tiny contribution to that of Iberia. Precisely why capital
did not flow to some poor countries remains an enduring puzzle: possible
explanations include insecurity of Irish property; the Iberian abandonment
of the Gold Standard; and cultural mysteries.
6. Trade, tariffs, and economic convergence
What was the impact of trade on the European periphery? The late
nineteenth century was a period of dramatic commodity market integration:
railways and steamships lowered transport costs, and Europe moved
towards free trade in the wake of the i860 Cobden-Chevalier treaty. These
developments implied large trade-creating price shocks which affected every
European participant, the canonical case being the drop in European grain
prices. Eli Heckscher and Bertil Ohlin argued that such commodity market
integration should have led to international factor price convergence, as
countries everywhere expanded the production and export of commodities
which used their abundant (and cheap) factors relatively intensively. For
poor labour-abundant and land-scarce countries, this meant rising wages
and falling rents. In an earlier paper (O'Rourke and Williamson 1994) we
showed that the reduction in trans-Atlantic transport costs had a profound
impact on British factor prices, and explained a large fraction of that
country's real wage convergence on the US. Can this finding be generalized?
Did peripheral countries who actively participated in the development of a
global economy undergo more dramatic real wage and labour productivity
16
17
18
Cited in Lains (1992, pp. 215-16).
Mata (1995) suggests that capital inflows can only have had a small aggregate impact on
the Portuguese economy, but that they were important in particular sectors. His numbers
suggest that foreign capital accounted for 12 per cent of net investment between 1851 and
1890, and 42 per cent of net investment between 1891 and 1913. However, he uses
official trade statistics; using Lains' revisions would imply much smaller numbers.
It goes without saying, of course, that we urgently need balance of payments data for the
Celtic and Latin fringes to confront these issues more adequately.
178
European Review of Economic History
growth than those who tried to insulate themselves from international
market forces?
There was certainly a great diversity in trade policy around the periphery.
Table 8 summarizes such information as we have concerning tariff levels: it
includes seven of our peripheral countries together with France and Germany for comparison.19 These data come in several forms. First, there are
Bairoch's (1989) estimates of tariffs on wheat. Second, there are several
average tariffs, computed using a variety of weights, for both manufacturing
and the economy as a whole. These were computed by the League of
Nations in 1927, by Liepmann (1938), and by Bairoch (1989) himself.
Third, we report the estimates of sectoral and overall protection calculated
by Estevadeordal (1993). These represent the only application of Learner's
(1988) methodology to pre-1914 data. Table 8 indicates where individual
countries ranked among Estevadeordal's eighteen nations in terms of their
openness (the most open being ranked 1, and the most protected being
ranked 18). We prefer Estevadeordal's figures to the crude tariff averages,
but they seem to offer the same inference.
Ireland, of course, was a part of the United Kingdom's customs union,
and as such remained a free-trader throughout our period. This is clearly
Table 8. European tariffs 1875-1913.
Manufacturing
Country
1875
(per cent)
Denmark
Norway
Sweden
Italy
Portugal
Spain
Ireland
France
Germany
1913 (1)
(per cent)
1913 (2)
(per cent)
1913 (3)
(rank)
1913 (4)
(rank)
15-20
14
na
na
na
16
8
14
2-4
3-5
20
25
5
15
14
6
17
13
18
8-10
20-25
15-20
0
18
20
na
na
41
0
18
12-15
20
34
0
21
4-6
13
13
Agriculture
Country
Denmark
Norway
Sweden
Italy
Portugal
Spain
Ireland
France
Germany
19
Wheat
1913 (per cent)
All Agriculture
1913 (1) (rank)
All Agriculture
1913 (2) (rank)
0
1
1
4
16
7
13
8
12
16
18
28
40
Prohibitive
43
0
38
36
18
14
4
17
2
10
12
6
6
British tariff levels were of course identical to those in Ireland.
8
4
5
12
12
6
3
Around the European periphery 1870-1913
Table 8. Continued.
Overall
Country
Denmark
Norway
Sweden
Italy
Portugal
Spain
Ireland
France
Germany
Overall
1913 (5)
Overall
1913 (1)
(per cent)
Overall
1913 (2)
(per cent)
Overall
1913 (3)
(per cent)
Overall
1913 (4)
(rank)
5.8
11.4
9
na
na
na
2
4
11
12
9.0
9-7
16
28
7
7
17
25
16
23.7
14.3
5.6
8-7
7-9
na
33
na
37
17
18
17
15
18
(rank)
0
0
3
3
18
24
14
14
12
17
8
8
Notes: Manufacturing 1875: average levels of duties on manufactured products in 1875, from
Bairoch (1989), Table 5, p. 42.
Manufacturing 1913 (1): League of Nations estimate, as reported in Bairoch (1989), Table
9, p. 76.
Manufacturing 1913 (2): Liepmann (1938) estimate, as reported in Bairoch (1989), Table 9,
p. 76.
Manufacturing 1913 (3): rank among 18 countries (1 = least protectionist, 18 = most
protectionist), based on the adjusted trade intensity ratios in Estevadeordal (1993), Table
3-7, P- 149Manufacturing 1913 (4): rank among 18 countries (1 = least protectionist, 18 = most
protectionist), based on the openness measures in Estevadeordal (1993), Table 3.7, p. 150.
Wheat 1913: levels of duties on wheat, calculated by Bairoch (1989), Table 9, p. 76.
Agriculture 1913 (1): rank among 18 countries (1 = least protectionist, 18 = most
protectionist), based on the adjusted trade intensity ratios in Estevadeordal (1993), Table
3.7, p. 149.
Agriculture 1913 (2): rank among 18 countries (1 = least protectionist, 18 = most
protectionist), based on the openness measures in Estevadeordal (1993), Table 3.7, p. 150.
Overall 1913 (1): import duties as % of special total imports (1909-1913), calculated by
Bairoch (1989), Table 9, p. 76.
Overall 1913 (2): League of Nations estimate, as reported in Bairoch (1989), Table 9, p. 76.
Overall 1913 (3): Liepmann (1938) estimate, as reported in Bairoch (1989), Table 9, p. 76.
Overall 1913 (4): rank among 18 countries (1 = least protectionist, 18 = most
protectionist), based on the adjusted trade intensity ratios in Estevadeordal (1993), Table
3.8, p. 151.
Overall 1913 (5): rank among 18 countries (1 = least protectionist, 18 = most protectionist),
based on the openness measures in Estevadeordal (1993), Table 3.8, p. 151.
reflected in Table 8, which shows Ireland to be the most open country
overall in our sample according to nearly all the measures available.
Denmark, as is well known, adhered to free trade in agriculture throughout
the grain invasion, engaging in a radical structural adjustment in the process.
Table 8 also indicates that Denmark protected manufacturing to a greater
extent than is often appreciated. Overall, however, the conventional wisdom
appears to be borne out: Denmark was one of the most open countries
around the European periphery, second only to Ireland.
180
European Review of Economic History
Sweden, on the other hand, conforms well with the Continental model of
liberalisation followed by reversion to protection. Faced with the invasion of
New World grain, agricultural protection was adopted in 1888; Sweden
imposed moderately high tariffs on both agriculture and industry, although
she appears to have been rather more open than Italy and the Iberians. It
seems likely that Norway was less open than Sweden on the eve of the First
World War, but focusing on the end of the period masks most of the story: it
was only in 1895 that protectionist sentiment in Norway increased, and grain
was only protected in 1905. Over most of our period, Norway was less
protectionist than Sweden.
Italy is another country which conforms well with the Continental model.
A free trader in the wake of Unification, Italy introduced moderate tariffs in
1878, and rather more severe tariffs in 1887. The latter duties led to a trade
war with France, which lasted until 1892. By 1913 Italy was one of the most
highly protected economies in Europe, at least according to Estevadeordal.
Liberalization was both shorter and less dramatic in Spain. Prohibitions
were abolished in 1869 and replaced with tariffs of 30-35 per cent; 1892 saw
a return to very severe protection for cotton textiles, iron and steel, and
cereals. Finally, while Bairoch (1989) portrays Portuguese trade policy as
being fairly liberal until the adoption of a strict protectionist tariff in 1892,
Lains points to the average tariff evidence, which indicates that Portuguese
manufacturing enjoyed tariff protection of more than 20 per cent between
1843 and 1913.20 It is certainly clear from Table 8 that both Spain and
Portugal were highly protected in 1913: in our sample, Iberia appears to be
the region least open to trade on the eve of the First World War.
Can Scandinavian catch-up and Iberian fall-back be explained by relatively more liberal Nordic trade policies? Only history can supply the
answer, given that theory is ambiguous about the issue (see O'Rourke,
1996a, for an extended discussion). A recent paper by Sachs and Warner
(1995) indicated that convergence was a feature of open economies, but not
of closed economies, in the late twentieth century; but what has been true of
the late twentieth century need not have been true of the immediate postwar
period, the interwar years, or even the late nineteenth century. So, what was
the impact of trade policy on the European periphery prior to the First
World War?
The political economy literature on late nineteenth century European
trade is vast, but it focuses on the core industrial countries while tending to
ignore the periphery; it focuses on the causes of trade policies, rather than on
their consequences; and it focuses on individual country studies rather than
on comparative assessments.21 Obvious and notable exceptions include
'Portugal was never a free trade country' (Lains 1992, p. 50).
A key statement in the political economy literature remains Kindleberger (1951); more
recent contributions, by political scientists, include Rogowski (1989) and Verdier (1994).
A round the European periphery 1870-1913
181
Bairoch's (1976b) monumental book, and Berend and Ranki's (1980,1982)
work on the European periphery.22 Bairoch is a pessimist regarding the
impact of free trade in nineteenth century Europe. His figures show
aggregate growth on the Continent slowing during the free trade era while
accelerating during the succeeding protectionist phase up to the First World
War.23 Moreover, Bairoch (1972, pp. 224-26; 1976b, pp. 287-95) thought
the free trade era was associated with international divergence, while the
protectionist phase was associated with convergence. The more recent
evidence in Section 2 suggests the contrary, but in any case, Bairoch is
invoking post hoc ergopropter hoc logic. Observing correlations between trade
policy and GDP per worker-hour and real wage trends is not enough; we
need to isolate the size of the price shocks associated with trade policy
during this period, and we need economic models which can assess the
impact of those price shocks on living standards.
While neoclassical theory predicts that free trade improves aggregate
welfare, it is ambiguous about real wages. Consider a world in which
agriculture produces food using land and labour, and industry produces
manufactures using capital and labour. Let food be the import good and
manufactures the export good. When food prices decline due to a grain
invasion, agricultural labour demand falls, and nominal wages decline; on
the other hand, lower food prices imply a lower cost of living for workers.
The net impact on real wages is therefore ambiguous. If food is a sufficiently
important part of workers' budgets, and if agriculture is a sufficiently small
employer, then real wages increase; otherwise, they decline.
In countries such as Britain, where only a small share of the labour force
was in agriculture, one might surmise that the cost-of-living effect would
have dominated the labour demand effect, and that cheap grain would have
boosted real wages; whereas in peripheral countries, with much larger
agricultural sectors, the labour demand effect might have dominated, with
cheap grain lowering real wages. In that case, free trade in grain could have
led to real wage and living standard divergence within Europe, rather than
convergence, and O'Rourke (1996b) finds some evidence for this.
However, grain was not the only commodity which was traded in the late
22
23
According to Berend and Ranki (1980, p. 550), international commodity market
integration benefited Scandinavia and (in a less central way) Italy; it was much less
helpful to Spain, and no help at all to Portugal. The impact of foreign trade on the
periphery thus varied enormously: it depended on the strength of input-output linkages
and a host of geographical, political, economic and cultural factors.
For an English-language summary of the argument, see Bairoch (1972). Capie (1994)
takes issue with Bairoch's conclusions, arguing that protection was not as high as is
commonly thought in the late nineteenth century, and that in any case protection had
little or no effect on economic performance. To establish the former point, Capie shows
that in several cases, effective protection rates were much lower than nominal rates; to
establish the latter point, he regresses growth rates against average nominal tariffs, and
finds no significant relationship.
182
European Review of Economic History
nineteenth century, and cheap grain was not the only price shock to which
free-trading Europeans had to respond. To estimate the total impact of
globalization on the European periphery, we need to measure changing
international price gaps for animal products, 'Mediterranean' agricultural
products, primary commodities (such as iron ore and timber), manufactured goods, and grain itself.
We did precisely this for Britain, the United States and Sweden in
previous papers (O'Rourke and Williamson 1994, 1995a). There was
certainly trans-Atlantic commodity price convergence between 1870 and
1913, affecting grain, beef, pork, bacon, mutton, butter, bar iron, cotton
textiles, coal, copper, hides, wool, tin, cotton and many other tradables.
What of Anglo-Swedish price gaps? While we found price convergence for
vegetable products, animal products, and forestry products, the price gap
between Britain and Sweden in Swedish home-market-oriented industries
fell only modestly, perhaps reflecting the effects of rising tariffs, while there
was no evidence of price convergence affecting Swedish export industries.
What impact did this Swedish commodity market integration into the
global economy have on catch-up? As Table 9 indicates, when a Swedish
CGE model is used to estimate the effects of Anglo-Swedish price convergence, the results were hardly dramatic. Anglo-Swedish commodity price
convergence served to raise Swedish wages by only 1.9 per cent, accounting
for only about 3 per cent of the decline in the Anglo-Swedish wage gap.24
US-Swedish commodity price convergence had a little bigger impact,
although still small. Commodity price convergence between the US and
Sweden increased Swedish real wages by 6.2 per cent, and raised US real
wages by 0.3 per cent (O'Rourke and Williamson, 1995c, revised Table 3, p.
922), accounting for a little less than one-tenth of the Swedish catch-up on
the US.
However, the really important point about these results is that while
cheap grain on its own might have lowered Swedish real wages (O'Rourke
1996b), commodity price convergence in general increased Swedish real
wages. We suspect that what was true of Sweden was true of the rest of
Scandinavia as well. From what we know about the trade policies of our
three countries, Heckscher-Ohlin effects were probably larger in Norway
than in Sweden, and a lot larger in Denmark. This is reflected in the entries
for the two countries in Table 9.25
24
25
Calculated from O'Rourke and Williamson (1995a, Appendix Table 2.4), and the wages
in Williamson (1995, Appendix Table A2.1, Great Britain revised). Again, the share of
convergence explained is a bit lower than the figure given in O'Rourke and Williamson
(1995a, Table 1), as the earlier paper used three-year averages to calculate 1870 and 1910
real wages.
The Danish and Norwegian numbers are not always identical to the Swedish numbers,
since the same increase in the domestic wage will imply different percentage changes in
wage gaps for different countries.
Around the European periphery 1870-1913
183
Table 9. The impact of commodity market integration on convergence
around the periphery, 1870-1910.
Country
Denmark
Norway
Sweden
Italy
Portugal
Spain
Ireland
Country
Denmark
Norway
Sweden
Italy
Portugal
Spain
Ireland
Impact of commodity
market integration,
trans-Atlantic,
1870-1910, on real
wages in 1910, per cent
Impact of commodity
market integration,
intra-European,
1870-1910, on real
wages in 1910, per cent
(1)
(2)
>6.2
>6.2
6.2
>i. 9
>i. 9
1-9
negative?
negative?
negative?
-8.8
positive?
positive?
positive?
negative?
Share of wage
convergence 1870-1910
on USA explained, per cent
Share of wage
convergence 1870-1910
on Britain explained, per cent
(3)
(4)
>I2.I
>3-9
>i3-9
>44
9-4
none?
(positive?)
(positive?)
none
3-i
positive?
(none?)
(none?)
none?
Notes and Sources: See text for cols. (1) and (2). Cols. (3) and (4) derived from cols. (1) and
(2), O'Rourke and Williamson (1993d, Revised Table 3) which gives the impact of transAtlantic price shocks on US real wages, and Williamson (1995: Table A2.1), Great Britain,
Italy, Norway, Portugal, Spain revised. Parentheses around the Iberian entries in cols. (3)
and (4) indicate that it is Iberian divergence that is being explained. Note that cols. (3) and
(4) are based on unrounded real wages, while Table 1 and Williamson (1995) report
rounded real wages.
What about the rest of the European periphery? Boyer et at (1994)
constructed a model of the Irish economy for 1907-8. While the model was
originally constructed in order to assess the impact of emigration on Irish
living standards, it can also be used to calculate the impact of declining
trans-Atlantic price gaps on the Irish economy. We ask: what would Irish
real wages have been in 1908 if trans-Atlantic price gaps had remained
constant in the four decades after 1870, rather than declining as they
actually did? When these counterfactual price shocks are imposed on the
Irish model, real Irish wages increase by 9.6 per cent, implying that
declining trans-Atlantic price gaps lowered Irish real wages by 8.8 per
184
European Review of Economic History
cent.26 Real wages in the United States would only have been a fraction
lower in the absence of trans-Atlantic price convergence. Table 9 thus
indicates that Heckscher-Ohlin forces actually increased the US-Irish wage
gap; they did not contribute to US-Irish convergence at all.
We have not estimated the evolution of Anglo-Irish price gaps in the late
nineteenth century, but suspect that commodity market integration across
the Irish Sea led above all to an increase in Irish animal product prices, as
rail and steamships helped Irish farmers meet the growing British urban
demand for breakfast foods. To the extent that Irish animal husbandry was
land-intensive, this may have further reduced the demand for labour in
Ireland. To evaluate the hypothesis more carefully, we would need an Irish
model which distinguished between tillage and pasture (the current version
has only one agricultural sector); and we would need price information
which is currently unavailable. Table 9 reflects our uncertainty.
The Mediterranean countries generate even more uncertainty. As before,
we would like to distinguish between trans-Atlantic and intra-European
commodity market integration. If the Irish experience is any guide, then the
trans-Atlantic effect, by lowering grain prices, may have been to lower
Mediterranean wages. On the other hand, intra-European integration may
have had the opposite effect, by increasing the output of labour-intensive
Mediterranean products (such as olives and wine), and of labour-intensive
industrial and mining activities in Italy and Spain. Table 9 thus suggests that
while trans-Atlantic integration may have contributed to the Iberian divergence observed, intra-European integration probably contributed nothing to
the observed divergence since it should have led to convergence. We admit
that this is purely speculative.
Measuring commodity market integration in various parts of the European periphery, and calculating the impact of this integration on individual
countries, should be a major research priority. But what we know so far
suggests that these forces did not play a consistent role in contributing to
catch-up and falling behind around the periphery. While it made a significant contribution to Scandinavian real wage catch-up, the contribution
was far more modest, and often negative, elsewhere.
7. The challenge of comparative history
Table 10 summarizes our initial efforts to isolate the sources of late nineteenth
century real wage and living standard catch-up and fall-back around the
European periphery. The comparative history suggests an explicit agenda.
First, it suggests that it might be of some value to think a little less like an
26
This counterfactual implies that Irish agricultural prices would have been 21.4 per cent
higher than they actually were in 1908, while imported manufactured goods would have
been 9.8 per cent cheaper than they actually were.
Around the European periphery 1870-1913
185
Table 10. The sources of catch-up and fall back around the European
periphery, 1870-1910 (in per cent).
Schooling
Mass
migration
Capital
flows
Trade
A. How much of real wage convergence (or divergence) on Britain explained?
Denmark
30.0
5-8
3-1- 4-7
>3-9
Norway
>44
5-6
I5-5-I7-3
35-1
Sweden
5.4- 8.4
43.0
4-5
3-i
0?
0
Ireland
all
small positive
Italy
all
0
positive?
positive
(0?)
(16.6-58.6)
Portugal
(58-all)
(0)
(0?)
(10-all)
Spain
(0)
(6.6-23.3)
B. How much of real wage convergence (or divergence) on America explained?
0-9
>I2.7
47.0-49.3
Denmark
16.3
20.0
65.1-65.4
0-9
Norway
>I4.6
0-8
34.O
39.7-41.4
Sweden
9-9
0
Ireland
0
all
0-5
0?
0
0?
Italy
all
(positive?)
(0?)
Portugal
(94-all)
(0)
(positive?)
(51-all)
Spain
(0?)
(0)
Residual
<534-58-O
<37.2-40.0
40.5-44.5
0
0
(0-25.4)
(0-834)
<I2.7-24.O
< o - 0.3
6.7-16.4
0
0
(0-6)
(0-49)
Sources: Taken from Tables 5 (Panels B2.1 and B2.2), Table 6 (cols. 4a-5b), Table 7 (cols.
3-4), and Table 9 (cols. 3-4). Parentheses around the Iberian entries indicate that it is
Iberian divergence that is being explained.
economist and a little more like an historian. That is, it would be a mistake to
try to force that experience into one tidy explanation, whether it comes from
the fertile mind of Heckscher, Ohlin, Sandberg, or Tortella. Consider:
Bad schooling explains an enormous share of the Iberian fall-back, but it
explains none of the Irish and Italian catch-up. Good schooling certainly
helps explain some of the Scandinavian success, but is five or ten per cent
worth all the shouting? Oddly enough, the Scandinavian schooling thesis
seems to work best in Iberia.
The workings of international factor markets on capital deepening was
profound everywhere around the periphery. Mass migrations and international capital flows together served to explain a third to a half of the
spectacular Scandinavian catch-up on Britain. They served to explain all of
the Irish and Italian catch-up. And their relatively small numbers served to
explain an important part of Iberian failure. Iberian isolation did contribute
to its late nineteenth century failure, but it was factor market isolation that
mattered most. Oddly enough, comparative debates over performance
around the periphery have said little about factor market integration. In
particular, it has said little about the inability of some poor countries to
exploit emigration while others exploited it so well. These questions warrant
more attention.
The libraries are full of debates over late nineteenth century tariff policy
and the questions: What were the implications of the policy choice between
free trade and autarky? What were the implications of world commodity
186
European Review of Economic History
market integration induced by declining transport costs? Given the amount
of ink spilt on the question, we were surprised by the tentative answer
emerging from Table 10. The figures, where we can calculate them, are
uniformly small! They are even small (or negative) in the presence of free
trade, as in Denmark and Ireland. They are unlikely to have explained much
more than five per cent of Scandinavian catch-up on Britain, and while
autarky may help explain Iberian failure, we suspect that it doesn't explain
much. These are striking inferences that warrant more attention.
Second, note that the residual is missing in some parts of the periphery.
The last column in Table 10 reports the residual after the first four columns
are added up. The entries for Scandinavia seem plausible: about half of the
catch-up on Britain can be explained by globalization and schooling;
technological mysteries must explain the other half. Reasonably enough, the
residual is much smaller in the case of Scandinavian catch-up on the
dynamic US economy. The entries, however, are zero for Ireland and Italy,
and almost zero for Portugal. A zero implies that none of the Irish or Italian
catch-up on Britain was due to higher rates of Irish or Italian total factor
productivity growth. It also implies that almost none of the Portuguese fallback can be attributed to slower rates of Portuguese total factor productivity
growth relative to Britain. If true, these are striking inferences that warrant
more attention.
Third, our tour around the periphery excluded eastern Europe, southeastern Europe, the eastern Mediterranean Basin, and the southern Mediterranean Basin. Is there reason to believe that the sources of convergence or
divergence were different there?
Acknowledgements
This paper was first presented to the Congress of the European Association of
Historical Economics, Universita Ca Foscari di Venezia, Venice, Italy (January
19-20, 1996). It extends to the rest of the European periphery the arguments and
evidence offered for Scandinavia in two of our papers in the Scandinavian Economic
History Review. 'Open economy forces and late 19th century Swedish catch-up: a
quantitative accounting' (1995a), 43(2), pp. 171-203 and 'Education, globalization
and catch-up: Scandinavia in the Swedish mirror' (1995b), 43(3), pp. 287-309. We
are grateful for the excellent research assistance of Bill Collins, Josh Greenfield and
Tien Quek; for the generous help of Giovanni Federico, Pedro Lains, Leandro
Prados and Jaime Reis; for the useful comments of the participants at the Venice
EAHE Conference; and for editorial guidance from Gunnar Persson and his
referees.
References
M. (1986). Catching up, forging ahead and falling behind. Journal of
Economic History 46, pp. 385-406.
ABRAMOVITZ,
Around the European periphery 18JO-1913 187
P. (1972). Free trade and European economic development in the 19th
century. European Economic Review 3, pp. 211-45.
BAIROCH, P. (1976a). Europe's gross national product: 1800-1975. Journal of
European Economic History 5, pp. 273-340.
BAIROCH, P. (1976b). Commerce Exterieur et Developpement Economique de I'Europe
au XIX Siecle. Paris: Mouton.
BAIROCHJ P. (1989). European trade policy, 1815-1914. In P. Mathias and S.
Pollard (eds), The Cambridge Economic History of Europe 8. Cambridge, UK:
Cambridge University Press.
BARDINI, C , CARRERAS, A. and LAINS, P. (1995). The national accounts for
Italy, Spain and Portugal. Scandinavian Economic History Review 43, pp. 11546.
BARRO, R. J. (1991). Economic growth in a cross section of countries. Quarterly
Journal of Economics 106, pp. 407-43.
BARRO, R. J. and SALA-I-MARTIN, X. (1991). Convergence across states and
regions. Brookings Papers on Economic Activity 1, pp. 107-82.
BARRO, R. J. and SALA-I-MARTIN, X. (1992). Convergence. Journal of Political
BAIROCH,
Economy 100, pp. 223-51.
BERENDJ I. T. and RANKI, G. (1980). Foreign trade and the industrialization of the
European periphery in the XlXth century. Journal of European Economic History 9,
pp. 539-84BEREND, I. T. and RANKI, G. (1982). The European Periphery and Industrialization
1780-1914. Cambridge, UK: Cambridge University Press.
BOYER, G. R., HATTON, T. J. and O'ROURKE, K. H. (1994). Emigration and
economic growth in Ireland, 1850-1914. In T. J. Hatton and J. G. Williamson
(eds), International Migration and World Development. London: Routledge.
e
BRODER, A. (1976). Les investissements etrangers en Espagne au XLX siecle:
methodologie et quantification. Revue d'Histoire Economique et Sociale 54, pp.
29-63.
CAPIE, F. (1994). Tariffs and Growth: Some Insightsfromthe World Economy,
1850-1940. Manchester, UK: Manchester University Press.
CIPOLLA, C. M. (1969). Literacy and Development in the West. London: Penguin.
ESTEVADEORDAL, A. (1993). Historical essays on comparative advantage, 1913-38.
PhD thesis, Harvard University.
FEIS, H. (1930). Europe, The World's Banker, 18/0-1914. New Haven, CT: Yale
University Press.
FENOALTEA, S. (1988). International resource flows and construction movements
in the Atlantic economy: the Kuznets cycle in Italy, 1861-1913. Journal of Economic
History 48, pp. 605-37.
HANSEN, S. A. (1970). Early Industrialization in Denmark. Copenhagen: Akademisk
Forlag.
HATTON, T. J. and WILLIAMSON, J. G. (1994a). International migration
1850-1939: an economic survey. In T. J. Hatton and J. G. Williamson (eds),
Migration and the International Labor Market 1850-1939. London: Routledge.
HATTON, T. J. and WILLIAMSON, J. G. (1994b). Late-comers to mass emigration:
the Latin experience. In T. J. Hatton and J. G. Williamson (eds), Migration and the
International Labor Market, 1850-1939. London: Routledge.
188
European Review of Economic History
HECKSCHER, E. F. (1954). An Economic History of Sweden. Cambridge, MA:
Harvard University Press.
JOHANSEN, H. C. (1985). Danish Historical Statistics 1814-1980. Copenhagen:
Gyldendal.
JORBERG, L. (1970). The industrial revolution in the Nordic countries. In C. M.
Cipolla (ed.), The Emergence of Industrial Societies: Part 2. London: Harvester Press.
K. A. (1995). The national accounts for Ireland in the nineteenth and
twentieth centuries. Scandinavian Economic History Review 43, pp. 101-14.
KENNEDY,
KiNDLEBERGER, C. P. (1951). Group behavior and international trade. Journal of
Political Economy 59, pp. 30-46.
P. (1992). Foreign trade and economic growth in the European periphery:
Portugal, 1851-1913. PhD thesis, European University Institute, Florence.
LEAMER, E. E. (1988). Measures of openness. In R. E. Baldwin (ed.), Trade Policy
Issues and Empirical Analysis. Chicago, IL: University of Chicago Press.
LAINS,
LJEPMANN,
H. (1938). Tariff Levels and the Economic Unity of Europe. London:
Allen and Unwin.
LUCAS, R. (1990). Why doesn't capital flow from rich to poor countries? American
Economic Review 80, pp. 92-96.
MADDISON, A. (1991). Dynamic Forces in Capitalist Development. Oxford, UK:
Oxford University Press.
MADDISON, A. (1994). Explaining the economic performance of nations. In W. J.
Baumol, R. Nelson and E. N. Wolff (eds), Convergence of Productivity: CrossNational Studies and Historical Evidence. New York: Oxford University Press.
MANKIW, N. G., ROMER, D. and WEIL, D. N. (1992). A contribution to the
empirics of economic growth. Quarterly Journal of Economics 107, pp. 407-37.
MARKUSSEN, I. (1990). The development of writing ability in the Nordic countries
in the eighteenth and nineteenth centuries. Scandinavian Journal of History 15, pp.
37-63MATA, E. (1995). Foreign investments in the Portugese economy from the middle
nineteenth century till the first World War. Paper presented to XVEncontro da
Associagao Portuguesa de Historia Economica e Social^ Evora.
MITCHELL, B. R. (1978). European Historical Statistics 1750-1975- New York:
Columbia University Press.
A. and PETTERSSON, L. (1990). Some hypotheses regarding education
and economic growth in Sweden during the first half of the 19th century. In G.
NILSSON,
Tortella (ed.), Education and Economic Development since the Industrial Revolution.
Valencia: Generalitat Valenciana.
NILSSON, A. and PETTERSSON, L. (1992). Education, knowledge, and economic
transformation: the case of Swedish agriculture 1800-1870. Lund Papers in
Economic History 13, Department of Economic History, Lund University.
O'ROURKE, K. H. (1995). Emigration and living standards in Ireland since the
Famine. Journal of Population Economics 8, pp. 407-21.
O'ROURKE, K. H. (1996a). Trade, migration and convergence: an historical
perspective. CEPR Discussion Paper 13193 Centre for Economic Policy Research,
London.
O'ROURKE, K. H. (1996b). The European grain invasion, 1870-1913. Mimeo,
Department of Economics, University College, Dublin.
Around the European periphery 1870-1913
189
O'ROURKE, K. H., TAYLOR, A. M. and WILLIAMSON, J. G. (1996). Factor price
convergence in the late 19th century. International Economic Review 37, pp.
499-530.
O'ROURKE, K. H. and WILLIAMSON, J. G. (1994). Late 19th century Anglo-
American factor price convergence: were Heckscher and Ohlin right? Journal of
Economic History 54, pp. 892-916.
O'ROURKE, K. H. and WILLIAMSON, J. G. (1995a). Open economy forces and late
19th century Swedish catch-up. Scandinavian Economic History Review 43, pp.
171-203.
O'ROURKE, K. H. and WILLIAMSON, J. G. (1995b). Education, globalization and
catch-up: Scandinavia in the Swedish mirror. Scandinavian Economic History
Review 43, pp. 287-309.
O'ROURKE, K. H. and WILLIAMSON, J. G. (1995c). Erratum. Journal of Economic
History 55, pp. 921-2.
O'ROURKE, K. H. and WILLIAMSON, J. G. (i995d). Open economy forces and late
19th century Scandinavian catch-up. HIER Discussion Paper 1709, Harvard
University, Cambridge, MA.
PRADOS DE LA ESCOSURA, L. (1988). De Imperio a Nation: Crecimiento y Atraso
Economico en Espana (1780-1930). Madrid: Alianza.
PRADOS DE LA ESCOSURA, L., SANCHEZ, T. and OLJVA, J. (1993). De te fabula
narratur? Growth, structural change and convergence in Europe, 19th and 20th
centuries. Working Paper D-93009, Ministerio de Economia y Hacienda, Madrid.
REIS, J. (1991). The gold standard in Portugal, 1854-1891. Paper presented to the
Conference on the Gold Standard in the Periphery, 1854-1939, Universidade Nova de
Lisboa.
RIIS, C. and THONSTAD, T. (1989). A counterfactual study of economic impacts of
Norwegian emigration and capital imports. In I. Gordon and A. P. Thirlwall
(eds), European Factor Mobility: Trend and Consequences. London: Macmillan.
ROGOWSKI, R. (1989). Commerce and Coalitions: How Trade Affects Domestic Political
Arrangements. Princeton, NJ: Princeton University Press.
SACHS, J. D. and WARNER, A. (1995). Economic reform and the process of global
integration. Brookings Papers on Economic Activity 1, pp. 1-118.
SALAZAR, A. D E O. (1916). O Agio do Ouro: Sua Natureza e Suas causas, 1891-191$.
Coimbra: Imprensa da Universidade.
SANDBERG, L. G. (1979). The case of the impoverished sophisticate: human
capital and Swedish economic growth before World War I. Journal of Economic
History 39, pp. 225-41.
SANDBERG, L. G. (1982). Ignorance, poverty and economic backwardness in the
early stages of European industrialization. Journal of European Economic History 11,
pp. 675-97.
SIMPSON, J. (1995). Real wages and labour mobility in Spain, i860-1936. In P.
Scholliers and V. Zamagni (eds), Labour's Reward. Hants, UK: Elgar.
SOLAR, P. M. (1979). The agricultural trade statistics in the Irish Railway
Commissioners' Report. Irish Economic and Social History 6, pp. 24-40.
TAYLOR, A. M. and WILLIAMSON, J. G. (1994). Convergence in the age of mass
migration. NBER Working Paper 4711, National Bureau of Economic Research,
Cambridge, MA.
190
European Review of Economic History
TAYLOR, A. M. and WILLIAMSON, J. G. (1995). Convergence in the age of mass
migration. Harvard University, Cambridge, MA.
TORTELLA, G. (1994a). Patterns of economic retardation and recovery in southwestern Europe in the nineteenth and twentieth centuries. Economic History Review
47, pp. 1-21.
TORTELLA, G. (1994b). El De'sarrollo de la Espana Contemporanea: Historia
Economica de los Sighs XIXy XX. Madrid: Alianza.
VERDIER, D. (1994). Democracy and International Trade: Britain, France, and the
United States, 1860-1990. Princeton, NJ: Princeton University Press.
WILLIAMSON, J. G. (1994). Economic convergence: placing post-Famine Ireland in
comparative perspective. Irish Economic and Social History 21, pp. 1-27.
WILLIAMSON, J. G. (1995). The evolution of global labor markets since 1850:
background evidence and hypotheses. Explorations in Economic History 32, pp.
1-54.
WILLIAMSON, J. G. (1996). Globalization, convergence and history. Journal of
Economic History 56, pp. 277-306.
ZAMAGNI, V. (1993). The Economic History of Italy 1860-1990. Oxford, UK:
Clarendon Press.