Download A Tale of two Globalizations: World Trade and Openness 1800-2007

1
Giovanni Federico and Antonio Tena
[University of Pisa and University Carlos III, Madrid]
A tale of two globalizations: world trade and openness 1800-2010
Abstract
We compare the pre-1913 and the current levels of openness of the world economy, combining a new series of
world trade from 1800 to 1938 with a comprehensive collection of historical series of GDP. We argue that total
openness increased before 1870, fluctuated widely without any trend in the next century and grew beyond the
1913 (and 1870) level in the latest three decades. We also argue that in those years the rise of services (nontradables) relative to agriculture and manufacturing dampened the growth of openness. Finally, we show that
while globalization in 1830-1870 and 1973-2007 and the collapse during World War One and the Great
Depression affected all polities, and trends differed in the two other periods. From 1870 to 1913 openness
stagnated in most rich countries and grew in the rest of the world, while in the golden age the reverse was true.
Address for contact
Giovanni Federico
Department of Economics and Management
University of Pisa
Via Ridolfi 10 56126 Pisa
E-mail: [email protected]
2
1) Introduction
In 1999, R. Baldwin and P. Martin published a paper titled Two waves of globalization: superficial similarities,
fundamental differences (1999). The authors supply a long and detailed list of these latter, while they argue that
‘the chief similarities lie in the aggregate trade-to-GDP and capital-flows-to-GDP ratios. These stand today
approximately at the level that they attained at the end of the 19th century’ (1999 p.1). This conclusion is clearly
provisional, as it relies on few scattered data for UK, USA and other advanced countries until the mid-1990s,
when the latest phase of globalization was just beginning. The issue has been resumed recently, without
settling it. According to Klasing and Millionis (2014 Fig 5), world openness (imports plus exports divided by GDP)
increased fast between 1870 and 1913, up to 30%, collapsed in the interwar years, rebounded after World War
Two and boomed in the last decades, up to around 50% in 2005. Fouquin and Hugot (2014) publish several
series of ratio of exports to GDP for time-invariant samples, which show a fast increase before 1870, and wide
fluctuations thereafter. The ratio exceeded permanently the peak of the 1870s only in the mid-1990s.
Furthermore, the interpretation of these results is however difficult, as they depends on transportation costs and
in barriers to trade, which affect tradables only, and on the composition of total GDP. Ceteris, paribus an
increase in the share of non-tradables would cause openness to decrease without any change in underlying
trade costs. Arguably, openness should be measured by the ratio of trade to gross output of tradables.
The recent work on international trade seems to offer a way forward by suggesting a simple way to measure
bilateral trade costs, inclusive of transport, duties and other barriers to trade (Andersn and Van Wincoop 2003
and 2004, Head and Ries 2001). It factors out the effects of the changing size of the economy and thus one
would expect a strong inverse relation with openness. Unfortunately, the results of historical estimates are not
conclusive and seem to contrast with the strong growth in openness from the direct estimates. In their baseline
series, Jacks et al (2011 Figure 6) aggregate the series of costs in a world index with a fixed-effect regression
with time dummies and find evidence of a decline before 1913 and from 1950 to 1975, to recover after a sharp
increase during the Great Depression. However, trade costs seems to have remained stable or even increased
after 1980. Furthermore, other methods of aggregation yield somewhat different results (Figure 7). Likewise,
Hugot (2014) gets quite different results, according to the sample and the method of aggregation: trade costs fall
steadily for a pooled series of all pairs of countries (Figure 5), while for time- invariant samples, in 2012 they
were only marginally lower than in 1913 (Figure 4 and 6).
Summing up, we do not know what really happened to world openness in the last two centuries and why. In this
paper, we estimate new series of total openness, proxied by the ratio of world export to total GDP, and of
openness for tradable only, defined as the ratio of exports to Value Added in agriculture and manufacturing, from
1830 to present. We address four questions
i) is the world more open now than it was on the eve of World War One?.
ii) how did world openness change before the two peaks? When did it start to grow? How fast did it recover after
the shock of the Great Depression?
iii) why did world openness change? How large was the effect of change in trade costs as opposed to the
structural change within countries (changes in the share of not tradables in GDP) and across countries (changes
in the relative size of countries)
iv) did change affect all countries uniformly? If the pattern of change did differ across countries, can we detect
any regularity e.g. by location and/or level of development?
In a nutshell, we ask to borrow the title of a well known book – is this time different? Does the current
globalization differ from the pre-1913 one?
Section Two describe our data. We rely on standard international sources for the years after 1950, while for the
period before 1938 we have collected all available series of GDP and we have estimated new series of trade by
country. We sketch out the evolution of world trade in Section Three, while Section Four deals with changes in
openness. Figure 1 offers a flavor of our results: it shows two major periods of globalization, before 1870 and
3
after 1970, divided by a century of wide but trendless fluctuations, and a widening gap between total openness
and openness of tradables only during the latter period. However, the available data impose a trade-off between
length of the series and representativeness of the underlying sample of countries, and Figure 1 plots the longest
but also least representative ones. In Section Four we buttress our first impression with alternative samples. We
also discuss the proximate causes of change in total openness, estimating the contribution of structural change.
The data show wide difference in the changes in openness by country: in Section Five we try to find regularities
in total openness by continent and by level of development. Section Six concludes, sketching out a new view of
the two globalizations
Figure 1
World openness, 1830-2007
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
1830
1845
1860
1875
1890
1905
1920
Total openness
1935
1950
1965
1980
1995
Tradable only
2) The data: trade and GDP
In this Section, we remind only the main features of our historical trade data, which we describe in detail in a
companion working paper (Federico-Tena 2015).
i) our unit of analysis is the ‘trading polity’ – i.e. an independent country or a colony. For each of them, we
estimate four import and four export series – at current and constant (1913) prices and at current and constant
(1913) borders. All series start as soon as possible after 1800, or at the very latest in 1850 end in 1938, or when
the polity disappears, while they 1. If necessary, we guesstimate trade of a territory even before its
establishment as a trading polity, which in most cases coincided with Western colonization. For a number of
polities, we extend in time the series of exports at current and constant prices and current borders back before
1850.
iii) whenever possible, we use ‘modern’ series published by scholars, national statistical agencies and
international organizations such as the League of Nations. If such series are not available, we build current-price
series from trade statistics from different sources (national yearbooks, collection of colonial and imperial
1
During World War One, official trade series for sixteen polities are missing. We have substituted them with estimates of
exports only (see for details Federico-Tena 2015).
4
statistics and so on). When necessary we adjust the data to follow the current definition of (special) trade, which
excludes transit and re-exports. Whenever no data for a polity are available, we guesstimate the series by using
the trade statistics of trading partners, or by interpolating with series of similar countries.
iv) if ‘modern’ sources do not provide series at constant prices, we compute them by deflating the series at
current prices with a set of purposedly-built price indexes. These latter use a set of British prices, which we piece
together from Gayer-Rostow-Schwartz (1953) for 1800-1850 and Sauerbeck for 1846-1938, supplemented with
unit values from the trade statistics (Board of Trade ad annum) for 1853-1938. We adjust these prices for
transportation costs with a set of route-specific freights, mostly from Shah and Williamson (2004) and Jacks and
Pendakur (2010).
v) finally we convert all series from national currencies into US dollars, using whenever possible polity-specific
sources, rather than the handy but sometimes flawed series from GlobalFinancialData (www…)
The number of existing polities, according to our list, varied between 155 and 183 (Figure 2). In principle, we
should have series for all of them, after 1850 but in practice we omit large independent port cities, such as Hong
Kong, and few very small polities, with a population less than 0.1% of world total in 1913. The number of (export)
series in our data-base increases from 10 in 1800 to 125 in 1850 and thereafter it changes only to reflect
changes in political boundaries, up to a maximum of 142 in 1932-1937
Figure 2
Number of polities, 1800-1938
200
180
160
140
120
100
80
60
40
20
0
1800
1815
1830
Polities
1845
1860
Exports
1875
1890
1905
1920
Imports and exports
1935
5
Historical data have a reputation to be unreliable, and there is a substantial literature about the pitfalls of trade
statistics, starting from the seminal work by Morgestern (1965). He put forward a test of their accuracy, which is
arguably too demanding (Federico-Tena 1991), but we do not pretend that trade statistics are always perfect,
and anyway sometimes they are missing altogether. As the quality of the series may vary from one year to
another, we have assessed each observation with a classification ranging from very good (A) to mere conjecture
(E) 2. Before 1850, the A-rated polities, mostly from Europe and the Western Offshoots, accounted for between
a third and a half of world exports and this share rises up to about three quarters in the late 19th century (Figure
3).
Figure 3
Distribution of world export by reliability of estimates, trade weighted 1800-1938
100%
90%
D
80%
C
70%
B
60%
50%
A
40%
30%
20%
10%
0%
1800
1815
1830
1845
A
1860
B
1875
1890
1905
C
D
E
1920
1935
Given the assumed margin of error of each individual series, we compute a series of the standard error of their
sum following Feinstein and Thomas (2001). It exceeds 30% for Asia and Africa before 1850, but worldwide it
declines from about 10% at the beginning to about 5% in the 1850s and then it falls further to 2% or below.
2
We hypothesize that the range of error relative to the ‘true’ value is less than 5% for the As (i.e. the true value is within
a ± 2.5% of range from our observation), 5-10% for the Bs (±7.5%), 10-25% for the Cs (±17.5%), 25-40% for the Ds (±32.5%)
and over 40% for the Es. In this last case, we assume the interval to be ±25%. See for the criteria of classification FedericoTena 2015.
6
Thus, although the quality of data improves over time and some polity series are admittedly always very
tentative, we are confident that the data-base captures fairly well also the early trends.
As usual, we compute the world trade from the export side 3. After 1850, we can simply sum up all series, but
before 1850 this method would overstate trade, as the total would spuriously increase any time a new series
begins. We thus build three different (‘trade’) samples, with a constant number of polities, starting in 1800, 1823
and 1830. As Table 1 shows, even the smallest one, the 1800 trade sample, accounted for a high proportion of
world exports in 1850 although it is somewhat skewed towards advanced Atlantic countries 4.
Table 1
Trade samples
1800 sample
1823 sample
1830 sample
Full sample (1850)
Number of available
polities
10
62
89
125
Number of missing
polities
115
63
36
0
Trade as ratio to total
trade in 1850
55.9
81.0
95.1
100
We build our long-run index of world trade by extrapolating stepwise the total trade in 1850 according to ‘full
sample series, from 1850 to 1830 with the ‘1830 trade sample’, then to 1823 with the ‘1823 trade sample’ and
ultimately to 1800 with the ‘1800 trade sample’. Then we extend our series of world trade from 1938 to present
with the data from the United Nations web-site (See Federico-Tena 2014 Appendix D). The same source reports
data of exports at current prices by country since 1950 and volume indexes only since 1980. We have been
able to extend these series to 1938 for many countries with the data from the printed version (UN Statistical
Yearbook) and other sources, but there still are gaps for Socialist countries and for the Third World. Thus,
somewhat paradoxically, our data-base at constant prices is much more complete before 1938 than after the
war.
We use three different sets of GDP data for the denominator (see Appendix ?) - our own collection of series at
current prices from polity-specific sources, the series at ‘current’ prices from Klasing-Millionis (2014) and an
extended version of the Maddison data at constant prices. We have been able to find forty series of GDP at
current prices, which we link to series from UN, available since 19705. For 22 of these polities, we compute the
VA in tradables (agriculture, mining and manufacturing) with shares from Mitchell (on-line version), Groningen
data-base and the United Nations. Klasing-Millionis (2014) estimate synthetic series of GDP at current prices for
61 polities (45 since 1870) from 1870 to 1949 by using coefficient from a regression for post-1950 data 6.
Finally, we compute series of GDP at constant prices before 1938 by multiplying series of population at current
3
According to the UN standard rules of the United Nations, exports should be measured f.o.b. (free on board) and imports
c.i.f. (cost, insurance and freight). Thus, the sum of imports overstates the value of traded goods by amount of
transportation and related costs. Furthermore, the size of the bias would change in time if the ratio of costs to value of the
goods changes.
4
The 1800 sample consists of six countries from Europe (France, Netherlands, Portugal, Russia, Sweden and the United
Kingdom) one for Asia (India) and three from the Americas (Cuba, Mexico and the United States), but none from Oceania
or Africa. In 1850, the six European countries accounted for 70% of exports from the 1800 sample (vs 62% of all European
countries on world exports), India for 9% (vs 10% of Asia) and the American ones for 21% (vs. 23% of exports from all the
American continent).
5
We try to keep the reference area as homogeneous as possible with our pre-1938 series. Thus Germany from 1950 to
1991 is the sum of former Democratic and Federal republics of Germany, and India includes Pakistan, Bangladesh (after
1971) and Myammar, which was included in trade statistics of British India.
6
They run a panel regression for the period 1950-1990, with the ratio of nominal GDP per capita relative to the United
States as dependent variable, explained by the ratio of GDP per capita at 1990$ PPP from Penn World tables, plus
controls. Then they get series of GDP at current prices by multiplying the polity-specific ratio (the common coefficient plus
the coefficients for controls) by the GDP at PPP prices from Maddison. Cf. for a similar approach for benchmark years
Prados (2000).
7
borders (Federico-Tena ??) by estimates of GDP per capita at 1990 PPP dollars. We obtain most of these latter
from the latest version of Maddison project, supplemented by series for African countries from Prados (2013)
and for few Latin American countries by the MOXLAD data-base. After 1951, we simply use the Maddison data,
adjusting for changes relative to the pre-war boundaries whenever possible. The Maddison series, in 1990 dollar
at purchasing power parity, are not directly comparable to our series of trade at constant prices in 1913 dollars at
market exchange rates 7. We make the trade data comparable by replicating Maddison’s procedure – i.e. we
scale an index of exports at constant prices (1990=1) with the ratio of GDP at market prices and PPPs in 1990 8.
3) The growth of world trade
Before using our series of trade, we have to show that it is not biased before 1850 because of its partial
coverage. By definition, total trade (W) at time t is the sum of trade according to the largest sample available in
that year (W St) and of the missing polities (W Mt)
W=W St+W Mt
where W=∑Ti W St=∑Tj and W Mt=∑Tl
T refers to trade of a polity, and the subscripts respectively to all polities (i=1..n), to polities present in each
sample (j=1..m) and to the omitted polities (l=n-m). We test two different null hypotheses i) that the rates of
growth of trade are equal between each sample and its corresponding missing polities (i.e. wSt=wMt) or ii) that
rates of growth are equal between each sample and the full sample (w=wSt)9. Of course, we cannot compute w
or wMt before 1850, and thus we perform the test for the period 1850-1913 (Table 2)..
Table 2
The biases from missing polities before 1850
wSt
wMt
i)
ii)
1800 sample
3.02*** 3.52*** R*** FR
1823 sample
3.24*** 3.41*** R*** FR
1830 sample
3.21*** 4.07*** R*** FR
Full sample
3.22***
* significant at 10%; ** significant at 5%; *** significant at 1%; R rejected FR failed to reject
In all cases, the tests reject the null in its ‘strong’ version (i) but not the ‘weak’ one (ii). In other words, at least
after 1850, the missing polities grew faster than the available ones, but the difference is not large enough to bias
significantly the long-term rates 10.
One might however object that this test is not conclusive because trends after 1850 are not necessarily
representative of those before 1850. For instance, one might argue that exports of missing polities grew faster
after 1850 because they had been growing more slowly before 1850. In this case, the rate wSt would overstate
the true rate w. In the extreme cases that exports of the missing polities had not grown at all (wMt=0) total trade
would have grown by 95% over the whole period 1800-1850 rather than by 157%, and by 21% rather than by
7
The unadjusted ratio underestimates openness for a given country if its GDP is lower at market prices than at PPPs – i.e.
when the relative price of non tradables was lower than in the United States in 1990, as it was the rule in poor countries
(the Balassa-Samuelson effect). On the other hand, the unadjusted ratio would overestimate openness, if the level of
prices was higher than in the United States. This was the case for 17 advanced countries, with a maximum of 1.66 for
Switzerland.
8
Note that this procedure assumes a constant ratio between GPD at market and PPPs and thus does not take into account
the effect of economic growth on relative prices.
9
Whenever possible (i.e. if the number of the observations exceed 25-30), we compute the rate of change of the i-th
series as w=- β/ψ, where β and ψ are coefficients from a regression (Razzaque et al 2007) Δ Ln Wt=α+β TIME+ψ lnWt-1+ φ
ln Δ Ln W t-1 +u. Otherwise we use a long-linear specification. Null hypotheses about rates (equal to zero or equal to rates
in other periods) are tested with a standard Wald restriction. We compute the cumulated change as Total= [exp(w)*n]-1.
10
The coefficients of correlation between each sample and the corresponding missing polities (i.e. between W St and WMt)
in 1850-1913 are extremely high (0.988 for the 1800 sample, 0.995 for the 1823 sample, 0.990 for the 1830 sample).
8
44% from 1800 to 1823 11. The difference is large but not huge and the counterfactual is rather implausible. In
fact, it implies that i) the whole increase in exports of polities in the sample was absorbed by other polities in the
sample and ii) the exports of the missing polities (including Italy, Germany and Austria-Hungary before 1823)
started to grow as soon as they entered the data-base12. These hypotheses contrast with most anecdotal
evidence about the growth of trade (cf. Federico and Tena 2014 for Italy) 13. Summing up, we cannot rule out that
the long run series overstates somewhat the growth in the first quarter of the 19th century but the difference is
likely to be small. Therefore, in the following analysis we will neglect this possible bias, which anyway does not
affect our measures of openness.
As a first step of the analysis, we explore the long-run stability of our series, with a Bai-Perron (2003) test for
structural breaks. It singles out the years 1817 and 1865 as breaks in period 1800-1913, and, less decisively, the
years 1970 or 1980 for the post-war years. Accordingly, we thus compute the rates of change for these periods
(Table 3).
Table 3
Rates of growth of world trade, 1800-2010
Rates (*100)
Cumulated change %
1800-1817 $
0.49
0.08
1817-1865
3.97***
5.98
1866-1913
3.07***
3.50
1817-1913
3.62***
4.89
1919-1938 §
0.10
0.01
1950-1973
8.08***
5.4
1973-1980 §
3.96***
0.3
1980-2007
5.86***
3.9
1950-2007
5.18***
18.2
* significant at 10%; ** significant at 5%; *** significant at 1%
§ log-linear estimate
As expected, trade remained roughly constant during the French wars, but at their end, it started to rise quite
fast. Part of this growth may reflect a return to pre-war levels, but this effect can account only for a small share of
the total increase. O’Rourke (2006) estimates that exports in 1815 were equal to the pre-war level in Sweden, a
third lower in the United Kingdom and in the United States and half in France. If in 1815 exports of all other
polities had been half their 1780 level, as in France, total trade would have been about 42% lower than before
the war and would have fully recovered by 1822. The recovery to 1780 level would account for less than 2% of
the overall increase in trade in the whole period 1815-1867. Contrary to the conventional wisdom, the growth of
world trade slowed down remarkably after 1865: if it had continued to grow as fast as in 1817-1866, world
exports in 1913 would have been 55% higher14. The outbreak of World War One caused world exports to drop
by about a quarter and trade remained depressed until 1918, in spite of a small rebound in 1916 and 1917.
Even if world exports increased immediately after the war, trade returned to its 1913 levels only in 1924 – i.e. six
years after the end of the conflict. It continued to grow steadily until 1929, when it was about a third higher than
in 1913. As it is well known, trade collapsed to a minimum below the 1913 level in 1933, and then in the next
five years they increased by a fourth, 20% above the pre-war peak. The recovery after World War Two was
11
We obtain our counterfactual trade by splicing together series for 1800-1822, 1823-1829 and 1830-1850, which we
compute as WC= Wt-nC*[πj1850*(1/exp(wSt*(t-n))+(1- πj1850)] where πj1850 is the share of the j-th sample from table 1 in 1850
and time t refers to the final year of each period.
12
The first condition assume that missing polities could not fund their deficit in the trade balance by importing capital.
13
One can add that the hypotheses of catching up from missing polities contrasts with the results of another statistical
test. From 1823 to 1850 the exports of the ten polities of the 1800 sample grew faster than the exports of the 47
additional polities which are present in the 1823 sample (3.68 vs. 3.04) and the difference is significant /C/.
14
This difference does not appear in the series by Lewis (1981), which grows almost as fast before and after 1866,
respectively at 3.18% (log-linear estimate) and 3.25%. The difference with rates from our series (4.20% in 1850-1866 and
3.04% in 1867-1913) are significant at 1% and 5% respectively.
9
somewhat faster than after WWI: in 1950, five years after the end of the war, world exports were already 25%
higher than in 1938. They grew at breakneck speed during the golden age, slowed down markedly in the 1970s
and accelerated again since 1980. The rates in the first and the last period and also the rate 1950-2007 exceed
significantly the pre-1913 maximum –i.e. 1817-1866. Furthermore, the effect of Great Recession on trade is
almost negligible if compared with the disaster of the Great Depression: world exports declined by 2.7% in 2009
but they rebounded the following years and continued to grow. Indeed, as figure 4 shows, world trade returned
close to its pre-1913 growth path (straight blue line) briefly in the 1970s, and exceeded it permanently since the
mid 1990s, with a growing difference, up to about a third at the end of the series. One could conclude that the
first globalization is not comparable with the second one, at least for the growth of trade.
Figure 4
The growth of world trade, 1800-2010 (log scale)
1800
1825
1850
1875
1900
1925
1950
1975
2000
This conclusion has to be qualified because the index refers to polities in their current borders. Before 1913,
changes in borders, such as the Unification of Italy and (for trade purposes) of Australia affected world export
only marginally: the difference between our series at current and 1913 borders reach a maximum of 0.57% in
1860. In contrast, the difference, as expected, is substantial after World War One and the collapse of the
Austrian-Hungarian and Ottoman empires. World exports were 3.1% higher at current borders than at constant
ones in 1924 and still 1.5% higher in 1938 – i.e. the series at current borders overvalues the level of trade in
interwar years in comparison with 1913, but it understates its growth 15. World trade was similarly affected by
the division of some British and French colonies in the 1950s and 1960s and by the fragmentation of former
USSR and Yugoslavia in the 1990s. According to Lavallée and Vicard (2010 tab 4), the boundary changes
accounted for 6.6% of the growth of exports during the Golden age (1950-1972) and for 17% of the overall rise
from 1950 to 2007 – equivalent to a third of a point of growth rate. Thus, border changes explain away part of
the gap in rate of growth the two periods, but by no means not all.
15
From 1924 to 1938 trade at constant prices increased by 9.7% if measured at current borders but by 11.1% if estimated
at 1913 borders.
10
Did the growth of trade spread evenly or did it benefit only a selected group of polities? An inspection of the
growth rates by polity fails to discover any obvious regularity and thus as a first approximation we settle on two
obvious groupings, by continent (Figure 5) and by initial level of development (Figure 6). We arbitrarily define
countries whose GDP per capita exceeded half the British one in 1870 as ‘advanced’ (or ‘old rich’), while we
distinguish all other polities between ‘newcomers’ (all the other OECD members as of 1971, including Italy and
Japan), ‘rest of Asia’ (including India and China) and ‘rest of the world’ 16. We use the series at current prices,
since 1830, with a small loss of precision before 1850 for the omission of some Asian and African polities (cf.
Table 1). The two figures tell the same story. The first globalization did not feature big changes. In the early
1830s, Europe accounted for 62% of world exports and the advanced countries for about a half. This latter share
increased by ten points in the 1850s and then remained around 60% until the war, while the share of Europe
was drifting slightly downwards to 56%. The shares moved a lot in the following fifty years, but by the early
1970s they were almost back to their pre-1913 levels: Europe accounted for 52% of world exports and the ‘old
rich’ for 57%. In contrast, the distribution of world exports changed dramatically during the second globalization,
because of the rise of Asia. The share of the whole continent rose from about a sixth in the mid-1970s to a third
in the late 2000s, and the ‘rest of Asia’ (excluding Japan and Turkey, but including Korea) from a tenth to a
quarter. All other continents lost market shares, and Europe fared comparatively better than America and
Oceania. The fall in the share of ‘old rich’, from 56% to about 40% of world exports, was compensated until the
early 1990s by an increase in the of ‘other OECD’ (i.e. Italy and Japan), but in the last fifteen years has returned
to the level of the 1970s
Figure 5
Distribution of world exports, by continent, 1830-2010
100%
90%
80%
70%
60%
50%
40%
30%
20%
10%
0%
1830
1845
1860
1875
EUROPE
16
1890
1905
AMERICA
1920
ASIA
1935
1950
AFRICA
1965
1980
1995
2010
OCEANIA
Asia includes Turkey and Europe includes Russia. The advanced countries are Australia, Belgium, Canada, Denmark,
France Germany, Netherlands, New Zealand, Switzerland, United Kingdom and the United States. Using the GDP of the
United States in 1913 as the yardstick, the group of rich countries would include also Sweden and Argentina. The ‘other
OECD’ series includes Austria, Greece, Finland, Ireland, Iceland, Italy, Japan, Norway, Ottoman Empire (Turkey after
1918), Portugal, Spain and Sweden.
11
12
Figure 6
Distribution of world exports, by level of development, 1830-2010
100%
90%
80%
70%
60%
50%
40%
30%
20%
10%
0%
Rich 1870
Other OECD
Rest of Asia
Rest of World
Summing up, the analysis of changes in trade has highlighted some relevant differences between the two
globalizations, both in the rates of growth and in the distribution of exports. What about openness?
13
4) The change in world openness and its causes
The available data makes it possible to four historical indexes of openness
i) total openness at current prices, with GDP from country-specific sources
ii) total openness at current prices with estimates by Klasing-Millionis (2014)
iii) openness tradables at current prices, with GDP and share of tradables from country-specific sources
iv) total openness at constant prices
plus total openness and openness tradables for all countries since 1970 from UN sources.
In all these indexes, we prefer to have exports, rather than, as usual, total trade (imports plus exports) in the
numerator for two reasons. In practice, we can consider more polities (Section Two) and the export/GDP ratio is
more suitable than the standard measure to world-wide analysis. The standard measure is more precise for a
single country, as using only exports would underestimate (overestimate) its involvement in world trade if the
country runs a deficit (surplus) in its trade balance. This argument does not hold at world-wide level, where
surpluses and deficits in trade balances cancel each other out. As that level, the sum of world exports and
imports overstates total trade, and thus openness, by the amount of transportation and related costs, which are
included in imports. Furthermore, the bias would change whenever these cost change: for instance, a fall in
transportation costs would reduce openness, ceteris paribus. Thus the ultimately, the greater the geographical
coverage is, the stronger the argument for using exports only (Maddison 2001 p.127).
It is clearly highly desirable to keep the composition of the sample constant in time, but this implies a trade-off
between the length of the series and their coverage. Thus, for each index we build two series, a longer one with
fewer polities, since 1830 and a shorter one, with more polities, since 1870, plus – for total openness at current
prices only, an extended sample for 1913 and 2007 only (Table 4).
14
Table 4
Information on openness samples
i)
ii)
iv)
iii)
All available series
Number of polities
40
61
51
Share of world trade in 1913
91.6
96.1
86.8
Share of world trade in 2007
73.4
84.7
62.9
21
58.8
74.0
1830 sample
Number of polities
16
37
Share of world trade in 1913
51.1
82.8
Share of world trade in 2007
27.2
51.0
7
30.4
15.6
1870 sample
Number of polities
26
44
42
Share of world trade in 1913
77.5
92.6
86.5
Share of world trade in 2007
46.8
74.8
59.9
Extended sample
Number of polities
37
Share of world trade in 1913
91.6
Share of world trade in 2007
73.4
14
65.6
39.4
15
As a rule, the samples at constant prices (column iii) are more numerous than the corresponding samples at
current prices (column i), but the difference in terms of coverage is not so large, especially for the 1870 samples.
The tradable samples are decidedly less representative: the 1830 one includes only European countries and
Australia and the 1870 sample, although larger, remains heavily skewed towards the Atlantic economy, with only
India to represent poor countries.
The answer to our first question is very clear: the world economy was more open in 2007 than on the eve of
World War One. Exports accounted respectively for 22.5% and 12.5% of GDP at current prices(extended
sample) and for 24.3% and 11.8% of GDP at constant prices (1870 sample). The difference is so large and the
coverage is so high that it would remain even if all polities missing in 2007 did not export anything. This
hypothesis is clearly absurd: the decrease in coverage between the two periods implies that the exports of the
omitted polities grew faster than the world average. Indeed in 2007 the export/GDP ratio for all countries at
current prices, according to the UN data, was two points higher than our figure (24.7%).
How were the peaks of 1913 and 2007 achieved? Figure 7 answers by plotting all available series of total
openness
Figure 7
Total openness 1830-2007
0.3
0.25
0.2
0.15
0.1
0.05
0
1830
1850
1870
1890
1910
1930
1950
1970
1990
Constant prices 1830 sample
Constant prices 1870 sample
Current prices 1830 sample
Current Prices 1870 sample
Current Prices Klasing-Millionis
Current Prices UN
All series confirm our first result - the world was more open in 2007 than in 1913. However, they do not tally
with the conventional wisdom about the story of globalization. As posited, total openness collapsed during the
Great Depression and increased fast since the early 1980s, but the figure suggests that openness grew fast
before 1870 but slowly or not at all from 1870 to 1913 and from 1950 to 1972, the golden age of the world
economy. This impression is by and large buttressed by the corresponding rates of change (Table 5)
16
Table 5
Rates of change, total openness
Current prices
1830
1870
sample
sample
1870
sample
populationweighted
KlasingMillionis
United
Nations
Constant prices
1830
1870
sample
sample
1830-1870
1.56***
2.63***
1870-1913
-0.13
0.16
0.45*
0.81***
0.24
1924-1938§ -3.88**
-4.33***
-4.06***
-3.36*
-2.88**
1950-1972
0.14**
0.47**
-1.67***
2.12*
1972-2007
0.89***
1.00**
1.69***
1.61***
2.84***
1950-2007
1.52***
1.57***
1.81***
2.60***
*significant at 10%; ** significant at 5%; *** significant at 10% § log-linear specification
0.32*
-2.85**
1.83**
2.05***
2.77***
All the rates for the same measure are significantly different between periods at 5% or less, and a comparison
of rates across measures for the same period highlights some interesting patterns. First, the rates are higher if
total openness if measured at constant than at current prices, and the difference remains significant even if we
compare homogeneous samples. Thus, the gap must reflect movements in relative prices – i.e. a decrease in
prices of exports relative to domestic prices 17. This trend may be caused by the Baumol effect (a decrease in
prices of all tradables relative to prices of non tradables) and/or by a rise in protection (a decrease in prices of
exports relative to prices of importables). The changes in relative prices – i.e. the surge in oil prices in the 1970sexplain the spike in openness at current prices which does not show in the series at constant prices. Anyway, in
the following we prefer to use series at current prices, which are not subject to errors in deflation and reflect
more accurately prices faced by agents in each period. Second, there are differences between series at current
prices. Using the Klasing-Millionis (2014) GDP series as denominator yields a significant growth from 18701913, as some of their series differ widely from the estimates at current prices 18. There are significant differences
also if we weight the series of openness with the share of each polity on the cumulated population of the sample
(Figure 8).
Figure 8
Total openness, population weighted, 1870-2007
Let’s write openness at current prices as α=X/(T+NT) and at constant prices as
β=X/PX/[T/PT+NT/PNT]=X/T*PT/PX + X/NT*PNT/PX where X exports, T VA tradables, NT VA not tradables and P
price indexes. In any baseline year, all price indexes are 1 and thus α=β. Ceteris paribus, openness at constant
prices would rise more than at current prices (β> α at time t) if PT/PX>1 and/or PNT/PX>1
18
The null of equal rates is rejected for 1870-1913 both with the total series and for a comparable series of 25 polities
(equivalent to our 1870 sample without Cuba). The rates are not significantly different in 1922-1938.
17
17
0.19
0.17
0.15
0.13
0.11
0.09
0.07
0.05
1870 1880 1890 1900 1910 1920 1930 1940 1950 1960 1970 1980 1990 2000
weighted GDP
weighted Population
The differences are modest before World War Two, but quite large afterwards: total openness weighted with
population fell during the golden age and grew faster in most recent years, without anyway catching up in full.
The difference reflects changes in India, which accounted for about half the population of the sample but for less
than 5% of the GDP.
Focusing on tradables only (Figure 9 and Table 6) does not alter the division of long-run change in five period,
nor the sign of the change: openness increased before 1870, stagnated until 1913, collapsed during the Great
Depression and boomed since the 1970s 19.
19
As said, the 1830 sample is not really representative. It is possible to increase its coverage by adding the United States,
under the assumption that their share of tradables on GDP remained constant from 1830 to 1870. The results are
very similar..
18
Figure 9
Openness tradables only, 1830-2007
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
1830
1850
1870
1890
1910
tradable sample 1830
1930
1950
1970
tradable sample 1870
1990
19
By definition, openness for tradables must be higher than total openness but the gap has been widening
(Figure 10). It was about 70% of the latter in 1870, about double in 1913 and 1938, 160% higher 1972 and three
times higher in 2007. The difference does not depend on the composition of the sample, as total openness for
the same 14 polities of the 1870 tradable sample traces perfectly openness for the full 1870 sample (red and
black lines Figure 10).
Figure 10
Total openness and openness tradables, 1870-2007
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
1870 1880 1890 1900 1910 1920 1930 1940 1950 1960 1970 1980 1990 2000
Current Total
Current consistent sample
Current tradable
Openness for tradables increased, rather than remaining constant, before 1913, declined less than total
openness in the interwar years and zoomed ahead after the war (Table 6). All the differences with the
corresponding rates of total openness are highly significant.
Table 6
Rates of change, openness tradables only
1830
1870
sample
sample
1830-1870
2.76***
1870-1913
0.28**
0.31
1924-1938§ -6.10***
-5.84***
1950-1972
1.86***
0.22
1972-2007
2.37***
0.73
1950-2007
2.83***
1.12***
*significant at 10%; ** significant at 5%; *** significant at 10% § log-linear specification
20
Admittedly, our computation might overstate the gap, to the extent that value added of tradables has grown
slower than gross output. We cannot test directly the size of the bias before 1970, because, as far as we know,
historical data on the gross output/Value Added ratios are not available
The growing gap between the two series of openness implies that the decrease in share of tradables (the
structural change within) has reduced the growth of total openness ceteris paribus. We estimate the size of this
effect by computing a counterfactual series, under the assumption that the share had remained constant since
1870 in all polities (Figure 11)
21
Figure 11
Total openness counterfactual estimates, 1870-2007
0.45
0.4
0.35
0.3
0.25
0.2
0.15
0.1
0.05
0
1870 1880 1890 1900 1910 1920 1930 1940 1950 1960 1970 1980 1990 2000
openness
share tradables
share tradables at constant GDP
The difference between the actual and the counterfactual series (i.e. the distance between the blue and the red
lines) remained very small during the first globalization and in 1913, it was only about 1%. The gap increased a
little in the interwar years but then declined again and as late as the mid-1960s it was still below two points.
Since then, it soared, up to over 15 points in 2007. Total openness would have been almost 40% if the
composition of GDP had remained constant since 1870, and almost 30% if had remained constant at its 1972
level.
The dotted line of Figure 11 adds a further counterfactual, hypothesizing that also the distribution of world GDP
remained constant since 1870 (no structural change across countries as well). The additional effect (the vertical
distance between the red and the black dotted lines) is fairly small in absolute terms but its sign is relevant. In
most periods, the changes in the location of ‘world’ GDP, as the decrease in share of tradables, have reduced
the growth of total openness, but there is an exception, the golden age 1950-1973, when the difference with the
baseline shrank. Table 7 shows that this conclusion holds true also for total openness, and also before 1870.
The column ‘change by polity’ reports the absolute changes in total openness between benchmarks, assuming
constant location of GDP since the first benchmark and the effect of changes in location, obtained as residual.
For instance, from 1830 to 1870 (first line) total openness rose by almost six points, but if the distribution of
world GDP had remained constant, openness would have increased by 7.5 points.
22
Table 7
A decomposition of the change in total openness
Initial
Total
Sample
openness change
6.16
1830-1870
1830
5.91
-1.60
2.50
3.28
-0.78
-5.50
-2.62
-2.88
1.62
-0.29
1.91
1870
10.05
9.35
9.10
0.25
1870
13.95
5.45
5.90
-0.45
1870
11.45
7.95
9.12
-1.11
12.46
10.02
11.80
-1.77
1870
1913-1950
1870
13.95
1950-1972
1870
8.44
1972-2007
1913-2007
1870-2007
1913-2007
Change distribution
world GDP
7.51
11.45
1870-1913
Change by
polity
extended
The changes in the location of GDP reduced total openness also from 1870 to 1950, and from 1913 to 2007,
while their contribution was positive from 1950 onwards. The effect was very small since 1972, but quite
substantial in the golden age. In the next Section we will argue that most of the openness-reducing effect of
structural change across countries reflects the rise of the GDP of the United States relative to other countries of
the sample.
Our results contrast quite clearly with the estimates of trade costs and structural change can be one of the key
to explain the difference. In fact, the bilateral trade costs are formula is
Τij=(Xii*Xjj)/(Xij*Xji)]-1/2ε
Where ε is the elasticity of trade to trade costs, Xij and Xji are trade flows between the two countries, Xii and Xjj
are trade flows within each country. These latter should refer to tradables only, but is practice, are proxied by
GDP less exports. Thus, the decline in the share of tradables would cause domestic trade flows to increase,
ceteris paribus, relative to foreign flows and thus the estimated trade costs to increase as well. Indeed Jacks et
al (2011 Appendix B ) and Hugot (2014 Figure 31) put forward alternative estimates of trade costs, for smaller
samples of countries, with gross output tradables only. As predictable, they show a steeper fall than their
baseline series. Furthermore, Hugot (2014 Figure 27) shows that an increase in the trade elasticity would cut the
estimated costs as well, relative to the (constant elasticity) baseline. Some combination of these two effects
would probably give a more realistic estimate of trade costs, which should be more consistent with our series of
openness.
5) How global were the two globalizations?
Although fairly large in terms of the number of polities, all our samples are dominated by a handful of large
countries: the big four countries (USA, UK, France and Germany) accounted for between a half and two thirds of
total GDP of the 1870 sample. Thus, the world-wide trends are bound to conceal movements in openness of
smaller polities, whenever they diverge from those of the big ones. This case is fairly frequent: nine out of 28
polities in the 1870 sample experienced a decline in their total openness ratios, from 1870 to 1913, and as many
out of the 37 polities of the extended sample from 1913 to 2007. The coefficients of variation of the rates of
change are quite high, respectively 1.39 and 1.93. The concentration is even greater, the top four accounting for
between two thirds and three quarters of total GDP, and the CV higher, for the openness tradables only.
Interpreting these divergences would need a detailed investigation of the evolution of each polity, which is
beyond the scope of this paper. As a first approximation, we group polities by continent and level of income, as
in Figures 4 and 5. We report in Table 8 the absolute changes in points of openness (as in Table 7) and we plot
the yearly series for the 1870 sample in Figure 12
23
Table 8
Change in total openness, by continent and level of development
Africa
America
-0.6
Asia
Oceania
11.6
1830
1870-1913
1870
1.5
5.2
3.9
-1.9
1913-1950
1870
-3.9
-3.2
-1.9
10.6
1950-1972
1870
0.2
-4.8
0.6
-19.0
1972-2007
1870
5.4
10.3
13.2
1913-2007
1870
1.6
2.3
1870-2007
1870
3.1
7.5
1913-2007
extended
1.7
19.7
-14.6
3.4
Europe
10.2
1830-1870
rich
ROW
World
5.9
6.0
5.1
1.0
5.9
2.5
-6.1
-2.5
-5.5
2.4
-2.1
1.6
3.1
9.0
10.5
9.3
11.8
-5.3
5.3
5.9
5.5
15.7
-7.2
6.3
11.8
8.0
15.8
-5.3
5.3
16.3
10.1
24
Figure 12
Total openness, current prices, by continent and level of development 1870-2007
America
Asia
.4
.4
.3
.3
.2
.2
.1
.1
.0
1870
1895
1920
1945
1970
1995
.0
1870
1895
1920
Europe
.4
.3
.3
.2
.2
.1
.1
1895
1920
1945
1970
1995
.0
1870
1895
Rest of the world
.4
.3
.3
.2
.2
.1
.1
1895
1920
1945
1995
1920
1945
1970
1995
1970
1995
Advanced countries
.4
.0
1870
1970
Oceania
.4
.0
1870
1945
1970
1995
.0
1870
1895
1920
1945
25
The figure shows substantial and permanent differences in long run levels of openness by continent, but not by
level of development. The averages 1870-2007 for rich countries and the rest of the world were practically
identical (12.1% and 12%), while Oceania and Europe were much more open than the world (total ratios
respectively 17.9 and 17.6%), and Asia and America less open (7.7% and 7.4%) than the world (12%). Africa is
missing from the 1870 sample at current prices, while at constant price it results more open than Europe. Most
coefficients of correlation between total openness for the world and for each group are fairly high (0.98 for the
advanced countries, 0.86 for Americas and the rest of the world, 0.81 for Europe) but this is not true for Asia
(0.51 only) and especially Oceania. The coefficient is negative (-0.124), reflecting the peculiar boom and bust
pattern of the Australian economy (Mc Lean 2013). A look at patterns by group and at changes in openness of
the main polities shows quite distinctive patterns in each period
i) The first globalization, from 1830 to 1870 left behind very few polities. The ratio remained constant in the
United States, where the growth in GDP matched the fast increase in exports (the American share of world
exports increased from about 6.5% to around 8%) and declined in Brazil, where a national economy was starting
to develop (and the ratio fell from 40% to 25%). The export/GDP (at constant prices) declined in Jamaica, which,
as other sugar-exporting islands of the Caribbean was hit hard by the abolition of slavery (Federico and Tena
??). But these were exceptions. Total openness increased fast in exporters of primary products, such as
Argentina (from 7.5% to over 11%) or Cuba, where slavery had not been abolished (from 18% to 26%), but also
in industrializing countries such as France (from 4% to 12%) Belgium (from 8% to 14%) and above all the United
Kingdom (from 10 to 18.5%)
ii) the period 1870-1913 marks a massive divergence. Total openness continued to grow in the rest of the world,
including Argentina (a further increase to 18%), India (from 7% to 11.7%) and the Ottoman Empire (from 5.7% to
11.6%). The series at constant prices show a massive increase in Africa as well. In contrast, openness
increased very little or declined in advanced countries, with the exception of the substantial rise in Germany. The
ratio increased by one percentage point in the United States, but in the United Kingdom it remained below the
level of the early 1870s until 1910. The ratio to Value Added of tradables increased in almost all polities, but it
remained constant in the United States.
iii) world openness never fully recovered from the shock of world war One, and the collapse of world trade
during the Great Depression caused a further sharp fall. The changes in location of GDP did not help. It shifted
towards more closed countries, most notably the United States which accounted for 35% of the cumulated GDP
of the 1870 sample in 1913, for 45% in 1929 and for 59% in 1950. This change account for about two thirds of
the decrease in total openness from 1913 to 1929 and for more than half of the fall from 1913 to 1950 (Table 8).
Thus, the picture at country level is not so uniformly bleak: about a third of the polities, including Australia,
Canada, Belgium, Denmark and the United Kingdom were more open in 1950 than in 1913.
iv) as said in Table 9, total openness during the golden age increased only thanks to the change in the
distribution of world GDP. From 1950 to 1972, the United States declined from 59% to 42% of the cumulated
GDP of the countries of the 1870 sample, while the share of the European countries increased from 24% to
39%. Trends in rich countries were mixed: total openness collapsed in the United Kingdom (from 24 to 15%),
remained roughly constant in France, Japan and the United States, but it jumped in Germany from around 10%
to 18%. In contrast, openness fell in almost all poor countries of the sample. It halved in India (from about 9% to
less than 4%) and Turkey (from 7.5% to 3%) and collapsed in Argentina (from 28% to 5%). The structural
change explains much of the decline or, in other cases, the failed growth in total openness, but in some cases,
such as India or the United Kingdom also openness tradables declined, respectively from 40% to 33%, and from
13% to 6%.
v) the last period, from 1972 to 2007 featured a widespread increase in total openness and, a fortiori, in
openness tradables only. The export/GDP ratio declined in Cuba, hardly a typical country. However, there were
differences. For instance, total openness in India (around 13%) was only marginally higher in 2007 than in the
26
late 1880s, and in the United States (8%) only a couple of points higher than the pre-World War One peak. In
contrast, the ratio was 31% in Russia (up from 3% in 1972) and over 40% in China.
5) Conclusions
We can sum our results in three statements
i) world trade grew steadily in the century after Waterloo and in the sixty years since World War Two. Part of this
latter increase reflects the recovery after the disaster of the Great Depression, but arguably world exports moved
to a different and higher path since the 1990s at the latest.
ii) if measured by total openness, the view of the two globalizations changes. Openness rose in the first forty
years and in the last forty years of our period, roughly at the same rate. The total openness in 2007 was about
three point higher than the level of the early 1870s, if measured with the 1830 sample and six points higher if
measured with the 1870s sample. The century between these two periods featured two periods of stagnation or
very slow growth and the collapse of the Great Depression. This latter, as the growth of openness in 1830.1870
and since 1973, was common to (almost) all the world. In contrast, in the two other periods trends differed. From
1870 to 1913, openness continued to grow in the rest of the world and stalled in rich countries, while from 1950
to 1973 it grew in (most) rich countries and declined sharply in the Rest of the World
iii) the structural change played a minor role before 1950 but a substantial one afterwards. The rebalancing of
world GDP towards Europe increased somewhat world openness during the golden age and above all, the
increase of share of non tradables on total GDP dampened the effects of fast growth of openness tradables
during the second globalization
This last notation raises some interesting questions about the causes of globalization. The growth in openness
for tradables, although admittedly somewhat overvalued, seems too fast to be explained only by changes in
transportation costs and fall in barriers to trade for an invariant bundle of goods. It is likely that it reflects also the
growing exchange of varieties of the same consumer goods and the development of international supply chains
– unbundling (Baldwin). It is unclear whether these processes will go on or the level of openness of the late
2000s will prove to be a historical peak as it was the 1913 one, as suggested by a recent article of the
Economist (A troubling trajectory Dec 13th 2014). We will not speculate further on this. Suffice to say that the
answer to our basic question is, at least for trade, a resounding yes. The second globalization is different.
27
References
Bai and Perron (2003)
Baldwin Richard and Philippe Martin (1999) ‘Two waves of globalization: superficial similarities, fundamental
differences’, NBER WP 6904
Barbieri and Keschk (2009) accessed..
Board of trade (ad annum) Board of trade, Statistical Department Annual statement of the trade and navigation
of the United Kingdom with foreign countries and British possessions in the year … (later Annual statement of
the trade of the United Kingdom with foreign countries and British possessions in the year… London: His
Majesty Stationery Office
Chilosi David and Giovanni Federico (2013) ‘Asian Globalisations: Market Integration, trade and economic
growth, 1800-1938’ LSE Economic History WP 183
Federico Giovanni (2011) ‘When did the European market integrate?’ European Review of Economic History, 15,
2011, pp. 93-126
Federico Giovanni and Antonio Tena-Junguito (1991) ‘On the accuracy of foreign trade statistics (1909-35):
Morgestern revised’ in Explorations in Economic history 28, pp.259-273
Federico Giovanni and Antonio Tena (2013) ‘Independence, slave abolition and export performance in Latin
America and the Caribbean 1820-1870’ Paper to be presented in II Congreso Asociación de Historia Económica
del Caribe (AHEC) Santo Domingo 26-29 junio 2013.
Federico Giovanni and Antonio Tena (2014) ‘ A new series of world trade, 1800-1938’ mimeo
Feinstein Charles H. and Mark Thomas (2001) ‘A plea for errors’ University of Oxford, Discussion
papers in economic and social history number 41
Feenstra Robert (1998)
Findlay, Ronald, and Kevin H. O’Rourke (2007). Power and Plenty: Trade, War, and the World Economy in the
Second Millennium. Princeton, N.J.: Princeton University Press.
Fouquin Michel and Jules Hugot (2014) Trade globalization: 1827-2012 When did trade costs start to fall? CEPII
WP n.?
Frankel Jeffrey (2000) Globalization of the economy, NBER WP 7858
Friedman The world is flat
Gayer Arthur D., Walt W Rostow and Anna Jacobson Schwartz (1953) The growth and fluctuation of the British
economy 1790-1850, reprint Hassock: Harvester Press 1976
Hugot Jules (2014) ‘Trade Costs and the Two Globalizations: 1827-2012’, mimeo
Lavallée Emmanuelle and Vincent Vicard (2010) ‘National Borders Matter...Where One draws The Lines Too’
Banque de France Document de travail DT 271 January 2010
Lewis, W.A. (1981) ‘The Rate of Growth of World Trade, 1830-1973’, in L.S. Grassman and E. Lundberg (eds.),
The World Economic Order, London: Macmillan pp. 11-74
28
Jacks David, Christopher Meissner and Dennis Novy (2009) ‘Trade booms, trade busts and trade costs’ NBER
WP 15267
Jacks David S. and Krishna Pendakur (2010) ‘Global trade and the maritime transport revolution’ Review
Economics and Statistics 92 pp. 745-755
Klasing Mariko and Petros Millionis (2014) ‘Quantifying the evolution of world trade, 1870-1949’ Journal of
international economics 92 pp.185-197
Mac Lean Ian (2013) Why Australia prospered. The shifting sources of economic growth Princeton: Princeton
University Press
Maddison Angus (1962) ‘Growth and fluctuations in the world economic 1870-1960’ Banca Nazionale del Lavoro
Quarterly review 15 pp.150-195
Maddison Angus (2001)
Morgestern Oskar (1965) On the accuracy of economic observations Princeton University Press: Princeton [1st
ed. 1950]
O’Rourke Kevin (2006) ‘The worldwide economic impact of the French revolutionary and Napoleonic wars, 17921815’ Journal of global history 1 pp.123-149
Persson Karl Gunnar An economic history of Europe
Prados de la Escosura Leandro (2000) International comparisons of real product, 1820-1990: an alternative
data-set’ Explorations in economic history 37 pp. 1-41
Prados de la Escosura Leandro (2012) ‘Output per head in pre-independence Africa: quantitative conjectures’
Economic history of Developing regions 27 pp. 1-36
Razzaque, Mohammad, Philip Osafa-Kwaako and Roman Grynberg (2007) ‘Long-run trend in the relative price:
empirical estimation for individual commodities’ in Roman Grynberg and Samantha Newton (eds) Commodity
prices and development, Oxford: Oxford University Press 2007 pp. 35-67
Shah Mohammed, Saif I. and Jeffrey G. Williamson (2004) ‘Freight rates and productivity gains in British tramp
shipping 1869-1950’ Explorations in economic history 41 pp.174-203
Sharp Paul and Jacob Weisdorf (2013) ‘Globalization revisited: market integration and the wheat trade between
North America and Britain from the eighteenth century’ Explorations in Economic History, 50 pp.88-98
UN Yearbook (ad annum) Yearbook of international trade statistics
29
Appendix A
Sources of GDP data, polity series
A.1 Current prices, 1800-2010
After 1970, we use UN data (GDP 1970-2010 http://unstats.un.org/unsd/snaama/resQuery.asp accessed
January 2014), in million dollars. Before 1969 we estimate GDP at current prices for each polity as GDP in local
currency times the exchange rates If not otherwise stated, we obtain these latter before 1938 from our database (i.e. consistent with export data) while after 1950 from GFD (when necessary using official rates). Sources
for series of GDP local currency series 20
Argentina (1820) 1820-1869 we extrapolate the 1870 GDP with Bulmer Thomas (2014) on line Statistical
Appendix tab A.3.4 times population (Federico and Tena ??); 1870-1969 Ferreres, O., Dos siglos de economía
argentina: edición bicentenario, Buenos Aires: El Ateneo (norte y sur fundacion), 2010
Australia (1800-2007) Mitchell J1,
Austria: (1924) Mitchell J1
Austria-Hungary (1913): Schulze (2000) Table A1+A2
Belgium (1835) Smits et al 21
Bulgaria (1887-1938) Ivanov (2012) 1887-1924 and Chakalov (1946) 1925-1939
Brazil (1820) IPEA
Canada (1870) 1870-1926 Urquhart 1993 table 1.1 (accessed via HNA Smits et al 2009) 1927-1969 Mitchell J1
Chile (1810) Braun et al 2000 Reflating data at 1995 prices (tab 1.1) with implicit deflator (Tab 4.2)
China (1840) 1840-1912 Ma et al 2014 per capita GDP in silver taels times population from (Federico and Tena
?)
Colombia (1820) 1820-1904 Bulmer Thomas (2014) on line Statistical Appendix tab A.3.4 times population
(Federico and Tena ??); 1905-1969 GRECO 1999
Cuba (1820) 1820-1902 Bulmer Thomas (2014) on line Statistical Appendix tab A.3.4 times population (Federico
and Tena ??) extrapolated to 1969 GDP with the series by Mitchell J1
Denmark (1818) Mitchell tab J1
Egypt (1886) 1886-1945 Youssef 2002 Tab. A.1 1950-1969 Mitchell
Finland (1860) Hjerrpe 1989
France (1815) Toutain, J.C. 1997 V41
French Indochina (1890): Bassino 1870 ff in http://gpih.ucdavis.edu/GDP.htm (accessed Nov 2014); scaled up
with share Vietnam on the cumulated population of French Indochina in 1950-1852 (ca 80%). Vietnam piaster
assumed=Mexican peso to 1930, as suggested by the author, afterwards GFD, official rate
Germany (1850) 1850-1938 Hoffmann (1965) Tab. 248; 1950-1969 Mitchell Table J1 (sum of East and West
Germany)
Greece (1833) 1833-1938 Kostelenos 2003 tab 2a 1950-1969 Mitchell Table J1
20
21
Mitchell J1 refers to the on line version, accessed June 2014
We assume 1830-1834 constant level 1835-1837
30
Korea (1911) 1911-1938 HNA; 1953-1969 Mitchell Table J1 (South Korea only)
India (1870) 1870-1899 Goldsmith 1983 1900-1946 Sivasubramonian 2000 tab. 6.9 (GDP) 1950-1969 Mitchell
Tab J1
Italy (1861) Baffigi et al 2013 (available at https://www.bancaditalia.it/statistiche/storiche accessed February
2014)
Japan (1885) 1885-1938 Okhawa and Shinohara (1979) tab
A7; 1950-1969 Mitchell tab J1.
Netherlands (1815) 1815-1938 HNA and 1950-1969 Mitchell J1
New Zealand (1860) Statistics New Zealand table E1.1 column Z (consolidated)
Norway (1830) Grytten O. (2003) tab 4 and 5
Ottoman Empire and Turkey (1830) personal communication by S. Pamuk. He has provided a series of Turkish
GDP after 1923 and export/GDP estimates for the Ottoman Empire 1820, 1840, 1860, 1880, 1900 and 19111913. We have interpolated these latter to get a continuous series and we have computed the GDP in dollars
with our estimates of export
Peru (1820-1913) Bulmer Thomas Bulmer Thomas (2014) on line Statistical Appendix tab A.3.4 times population
(Federico and Tena ??)
Portugal (1837) Valerio et al 2001 tab 6.6B and 6.6C
Russia (1885) 1885-1913 Gregory 1982 tab 3.2 and 1928-1969 Mitchell J1 exchange rate paper rubles to 1913
GFD
South Africa (1911) Mitchell J1 to 1938 exchange rate pound sterling (1 pound=2 rand) 1950 ff GFD
Spain (1850) Prados 2003 cuadro A.2.7
Sweden (1800) Krantz, O. and L. Schön (2012) table V (GDP market prices)
Switzerland (1851) Stohr 2014
United Kingdom (1830) Mitchell 1988 National Accounts5 (GDP at factor costs)
United States (1800) Mitchell J1 (GNP)
Uruguay (1870-1969) Bonino et al 2012
Taiwan (1903-2007) Mitchell tab J1 22
A.2
Constant prices
Given the method computation exports constant prices, need constinuous series – thus here only list polities we
have computed ratio
/C Portugal Norway Venezuela Ceylon 1922-1938/
We use maddison’s figures after 1951 GDP at 1990 $ as Maddison project times population (Federico and Tena
2014) – when necessary interpolating linearly between benchmark years
+Prados for Africa
22
The UN does not report data for Taiwan
31
China de Jong (2014) 1840-1912 1800-1840 Broadberry et al 2014 Wu?
Bulgaria (Ivanovic 2012) tab.51 1870-1945
Switzerland Stohr 2014, extrapolated backwards to 1830 with the Maddison rate of change 1820-1830as stated
by Stohr trends plausible
Vietnam Bassino 1870 ff in http://gpih.ucdavis.edu/GDP.htm (accessed Nov 2014) . do we have trade data?
UK extrapolating backwards the 1850 GDP per capita (inclusive of Ireland, Scotland) with the series by
Broadberry 1820-1850, which refers to England and Wales only /C original/
India Broadberry et al (EEH forthcoming) 1820 1871
Population after 1950: United Nations, Department of Economic and Social Affairs, Population Division (2011). World Population Prospects:
The 2010 Revision, CD-ROM
A.3 Share of services
As a rule, share current prices..
/Antonio is preparing a file/
32
References
Bonino, N.; Román, C. y Willebald, H. (2012): “PIB y estructura productiva en Uruguay (1870-2011): Revisión de
series históricas y discusión metodológica”, Series Documento de Trabajo, 05/12, Instituto de Economía (FCEAUniversidad de la Repubblica) Montevideo.
Stephen Broadberry, Hanhui Guan, Peking University, David Daokui Li
China, Europe and the great divergence: a study in historical national accounting, 980-1850
Memo 2014
Braun, J., M. Braun, I. Briones, and J. Díaz, "Economía Chilena, 1810-1995: Estadísticas Históricas," Instituto de
Economía - Pontifica Universidad Católica de Chile, Documento de Trabajo No. 187, 2000
Gregory Paul (1982) Russian national income 1885-1913 Cambridge: Cambridge University Press
Hjerrpe, Riita (1989). The Finnish Economy, 1860-1985. Growth and Structural Change. Helsinki:
Bank of Finland.
Ivanov Martin (2012) The gross domestic product of Bulgaria 1870-1945 Ciela Sofia
J.P. Smits, P.J. Woltjer and D. Ma (2009), 'A Dataset on Comparative Historical National Accounts, ca. 1870-1950: A TimeSeries Perspective', Groningen Growth and Development Centre Research Memorandum GD-107, Groningen:
University of Groningen
K. Okhawa,
K. and Shinohara, M. (1979). Patterns of Japanese Economic Development. New Haven
and London: Yale University Press
Stohr Christian (2014) Let’s Get This Right: Swiss GDP and Value Added by Industry from 1851 to 2008 Geneva
School of Economics and management WPS 14-09-01
Grytten, O., “The gross domestic product for Norway 1830–2003,” in Eitrheim, Ø., J. Klovland, and J. Qvigstad,
Historical Monetary Statistics for Norway 1819–2003, Oslo: Norges Bank, 2004, pp. 241–288, table 5
Ma Ye, Herman de Jong and Tianshu Chu (2014) ‘Living standards in China between 1840 and 1912:
A new estimate of Gross Domestic Product per Capita’ Groningen Growth and development Centre GGDC
research memorandum 147 14July 2014
Yousef Tarik M. (2002) ‘Under economic liberalism: a reassessment with new GDP estimates, 1885-1945’ ERF
Working Paper 0211 Cairo 2002
Urquhart, M., Gross National Product, Canada, 1870-1926: The Derivation of the Estimates, McGill-Queen’s
University Press, Canada, 1993, table 1.1
Valerio (ed) Portuguese Historical Statistics 2001
Schulze Max-Stephen (2000) ‘Patterns of growth and stagnation in the late nineteenth century Habsburg empire’
European Review of Economic history 4 pp.311-340
Toutain, J. C. 1997. “La Croissance Française 1789-1990. Nouvelles Estimations.” Economies et
sociétés. Cahiers de l’ISMEA. Serie Histoire quantitative de l’economie francaise. Serie HEQ n.1.
Paris
33
Kostelenos G. (2003) ‘Historical estimates of national accounts magnitudes in Greece: 1830-1939’ Spoudai Vol.
53 pp.37-64
Sivasubramonian, S. (2000). The National Income of India in the Twentieth Century. New Delhi: Oxford
University Press.
Schön & Krantz (2012), “Swedish Historical National Accounts 1560―2010” Lund Papers in Economic History
123, Lund University on line version accessed November 2014
Gregory, Paul (1982). Russian National Income 1885-1913. Cambridge: Cambridge University Press.
Hoffmann, W. (1965). Das Wachstum der Deutschen Wirtschaft Seit der Mitte des 19. Jahrhunderts.
Berlin: Springer.
Statistics New Zealand Long-Term Data Series,
(http://www.stats.govt.nz/browse_for_stats/economic_indicators/NationalAccounts/long-term-dataseries/national-income.aspx) accessed June 2014
Prados de la Escosura Leandro (2003) El progreso economico de Espana 1850-2000 Fundacion BBVA Bilbao
Mitchell B.R. (1988) British historical statistics Cambridge: Cambridge University Press
GRECO (1999) El desempeño macroeconómico colombiano, series estadísticas. Documento de Trabajo del
Banco Central de Colombia. www.banrep.gov.co/docum/ftp/borra134.pdf
IPEA Instituto Brasileiro de Geografia e Estatística, Sistema de Contas Nacionais Referência 2000
(IBGE/SCN 2000 Anual) - SCN_PIBN
Appendix B
Composition of samples for openness measures
Current prices
Current prices
Constant prices
Tradables
1830 sample
Argentina
Australia
Belgium
Brasil
Canada
Ceylon (Sri Lanka)
Chile
Colombia
Cuba
1830 sample
Klasing-Millionis
1830 sample
Argentina
Australia
Belgium
Brasil
Chile
Colombia
Cuba
Denmark
France
1870 sample
Algeria
Argentina
Australia
Belgium
Brasil
British Malaya
Canada
Ceylon (Sri Lanka)
Chile
Australia
Belgium
Denmark
France
Netherlands
Sweden
United Kingdom
34
Netherlands
Norway
Ottoman Empire/Turkey
Portugal
Sweden
United Kingdom
United States
1870 sample (additional)
Canada
Finland
Germany
Greece
India
Italy
New Zealand
Spain
Switzerland
Uruguay
Extended sample (1913 and
2007)
Austria-Hungary *
Bulgaria
China
Egypt
French Indochina **
Japan
Korea
Peru
Russia
South Africa
Taiwan
China
Denmark
Dutch East Indies (Indonesia)
Egypt
Finland
France
Germany
Ghana
Greece
India
Italy
Jamaica
Japan
Morocco
Mexico
Netherlands
New Zealand
Norway
Ottoman Empire/Turkey
Persia (Iran)
Philippines
Denmark
Dutch East Indies (Indonesia)
Ecuador
Finland
France
Germany
India
Italy
Jamaica
Mauritius
Mexico
Morocco
Netherlands
New Zealand
Norway
Ottoman Empire/Turkey
Peru
Philippines
Portugal
South Africa
Spain
Portugal
Romania
Russia/USSR
Serbia/Yugoslavia
Siam (Thailand)
South Africa
Spain
Sweden
Switzerland
Tunisia
United Kingdom
United States
Uruguay
Venezuela
Sweden
Switzerland
Tunisia
United Kingdom
United States
Uruguay
Venezuela
1870 sample (additional)
Egypt
Cameroon
British Malaya
Japan
Siam
* in 2007 sum of Austria, Czech Republic, Croatia, Hungary, Slovenia and Slovakia
** in 2007 sum of Cambodia, Laos and Vietnam
1870 sample
(additional)
Finland
Germany
India
Italy
Norway
Spain
United States
35