Technology and Geography in the Second Industrial Revolution:
New Evidence From the Margins of Trade
8/14/2014
Preliminary draft
Michael Huberman
Université de Montréal
Christopher Meissner
University of California, Davis & NBER
Kim Oosterlinck
Université Libre de Bruxelles
ABSTRACT: Between 1890 and 1910 Belgium experienced a trade boom. Productivity growth
showed a more modest acceleration. Using a new, highly disaggregated data set on exports and
imports, we introduce geography, or trade costs, as a factor in Belgium’s trade performance and
study its relation to industry-level productivity growth. The effect of trade costs was mediated by
the extensive and intensive margins of trade. In markets for differentiated goods, the extensive
margin dominated, the number of new goods and new markets expanding as trade costs fell. But
the same decline in trade costs enabled less productive concerns to become viable exporters,
average industry productivity stagnating as a result.
2 I. Workshop to the World: Round Two
Of the known unknowns about the dynamics of globalization, the relationship between
international trade and economic growth remains a vexing question. Consider Belgium’s trade
boom between 1870 and 1914. In the mid-nineteenth century, the country exported a narrow
range of products to a handful of destinations. By the eve of the war, the number of products
shipped abroad had more than doubled, and so did the number of export destinations. Like the
U.K. fifty years earlier, Belgium had emerged as a workshop to the world.1 The trouble is that
despite its strong trade performance growth in output failed to accelerate in step.
There are a number of competing explanations of the link between trade and growth in the
late nineteenth century. Exploiting a factor-endowment framework, O’Rourke and Williamson
(1999) claimed that interindustry trade, mainly between the labor-scarce and resource-abundant
New World and the labor-abundant and land-scarce Old World, precipitated real-wage
convergence. This explanation is incomplete, however, since in Europe, the core of the world
economy, intraindustry trade was as important, if not more so, than interindustry trade. Temin
(1997, 2000) invoked a strict Ricardian model of relative productivity differences to explain
British export superiority across a broad range of industries during the industrial revolution.2 The
application of this approach to the second industrial revolution has its drawbacks because exports
did not map onto productivity differences between countries. Whereas Belgium had few, if any,
world-class industries and the adoption of new technologies was lackluster, exporters shipped
goods to low and high-productivity countries alike.
Other leading trade models yield different predictions—and raise other problems.
Hummels and Klenow (2005) show that economic growth in a model incorporating increasing
returns and a “love of variety” (Krugman 1979) expands the range of goods produced and
exported. Because of home market effects and agglomeration economies, industrial activity
would have increasingly concentrated in the largest and richest domestic markets, the U.K.,
France, and Germany. This group would have exported a wide range of products to a great many
destinations, if not everywhere. Small countries would not have been so fortunate, even if they
did trade more relative to their income. The rub is that Belgium's degree of openness actually
tripled after 1870, that of the club of big and rich economies increasing between 10 and 25
percent. Taking the Krugman model as a baseline, the country exported a disproportionate
1
During this same period, Kindleberger (1956) reported comparable transitions in the Netherlands, Switzerland, and
to some extent in Scandinavia.
2
Temin believed that the British experience was unique:
I showed in my [Temin] 1997 article that the list of manufactures exported from England was very
long and very different from the list of imports. It seems unlikely that any other country was
exporting “other manufactures” in such diversity and at such low prices…There are trade data for
other countries in this period; did these other countries export the same astonishing variety of
manufactured goods as Britain? If not, and I predict not, then the suggestion of intra-industry trade
must be deemed ahistorical (Temin 2000).
3 amount of varieties to a remarkable number of outlets, surpassed only by the U.K. By 1900, it
was the fifth largest exporter in Europe (De Leener 1906). On this score, Belgium was clearly
punching above its weight.
This paper gives a new perspective on Belgium’s trade boom, and, more generally, the
relationship between globalization and growth before 1914. We do so in two ways. First, we
introduce explicitly geography as a determinant of Belgium's trade performance. A Krugman
type model assumes only variable trade costs. We posit that beachhead or fixed costs also
mattered. The intuition is that relative productivity and barriers to trade were two sides of the
same coin (Romer 1994). Below some threshold, unproductive firms did not export. Thus the
surge in exports could have represented an increase in the innate productivity advantages of new
and more firms to overcome the threshold. But it may have also been the case that more firms
participated in trade because the fall in fixed costs established a lower threshold of exporting.
Our second contribution derives from the level of analysis. In effect, the impact of trade
costs was mediated by the margins of trade: the intensive margin (average sales per product or
per country) and the extensive margin (the number of products exported or countries served).3
Using granular trade data, our decomposition of trade suggests a novel interpretation of the
supposed paradox of globalization and productivity in Belgium before 1914. The collapse in
fixed costs prompted the entry of new firms and new products (the extensive margin). But the
overall effect on average sales per firm (the intensive margin) was mixed, since the same fall in
fixed costs pulled into the international sector previously uncompetitive low-productivity firms.4
In this fashion, the expansion in the trade share between 1870 and 1913 went hand-in-hand with
the modest growth of output per capita.
Aside from data availability, our choice of Belgium merits comment. In many ways, it
was a microcosm of late nineteenth-century Europe. It shared a common institutional framework,
the legacy of longstanding commercial and political relations with its neighbors. It participated
in, if not initiated, international trade agreements on the Protection of Industrial Property (1883)
and the Brussels Convention on unfair competition (1900). Typical of other continental
economies, the proportion of intraindustry trade exceeded that of interindustry exchanges. And
like many of Europe’s small economies, it maintained low tariffs in the face of rising
protectionism on the continent. Nonetheless, compared to the average small country, Belgium’s
trade expansion was extraordinary. With the exception of the Netherlands, which operated mainly
as an entrepôt, the country’s trade boom far exceeded that of Denmark, Sweden, and Switzerland.
Still, relative to many countries of its size, Belgian productivity performance was anemic
(Maddison 2013).
3
A parallel decomposition holds for imports. For simplicity, we refer to export margins throughout this paper. More
loosely, the extensive margin refers to new products; the intensive margin to trade in established goods. See Cadot et
al. (2012) for a review of recent research.
4
Chaney (2008) presents a model of this type.
4 The Belgian riddle has wider implications for the study of European growth and how
economic historians have approached the wave of globalization before 1914. O’Rourke (2000)
has suggested that trade liberalization did not have the intended effects on growth that the
textbook model would have anticipated. Did Belgium rely too heavily on market forces and not
enough on state intervention? Our findings can be recast as well in Allen’s (2009) model of
endogenous technical change in which wages in continental Europe were too low relative to
Britain before 1850 to bring about the adoption of modern technology. Over the last quarter of
the century, the forces of international trade ought to have pushed wages up in open economies
like Belgium (O'Rourke and Williamson 1999; Meissner and Liu 2013).5 As more productive
firms congregated in the export sector, and uncompetitive firms closed down, the trade boom
would have nudged continental economies closer to the technological frontier, a defining feature
of the second industrial revolution. Positive selection of this type would have had spillover
effects on aggregate productivity (Melitz and Redding 2013). However, this is not the entire
story. As trade costs fell, firms did export a wider and improved range of products. But even as
wages rose, they still lagged British levels, because the same fall in trade costs induced the entry
of new unproductive firms.6 In this regard, the decline in trade costs had generated a race
between a core group of technological advanced firms and new uncompetitive entrants in export
activities. The final balance sheet was mixed, the trade boom having only a moderate effect on
growth.
The paper is organized as follows. In the next section, we give background information on
the decline in trade costs and introduce some alternative explanations of the weak association
between the Belgian trade boom and growth. Next, we present our dataset on trade. We then
proceed to evaluate the effect of trade costs on the extensive and intensive margins. In the last
section of the paper, we study the impact of the trade margins on productivity growth.
II. Trade and growth in Europe through the Belgian mirror: Evidence and theory
Measured by the share of exports and imports in GDP in Figure 1, Belgium's international
exposure doubled between 1870 and 1900, and then doubled again in the shorter period from
1900 to the eve of the war. On a per capita basis, Belgium’s trade exhibited the fastest growth in
Europe. Maritime freight rates fell by about 30 percent over this period, much of the decline
happening before 1900 (Jacks et al. 2010; Jacks and Pendakur 2011). All countries shared in the
decline in transport costs, but Belgium’s size had an added virtue. Because of its integrated rail
and canal network, the sharp contraction in shipping costs was matched by that in inland freight
5
Neither the Stolper-Samuelson effects explored in O’Rourke and Williamson (1999), nor the market potential
channel explored in Liu and Meissner (2013), depend on total factor productivity to bring about rising real wages and
convergence. Gains from trade accrue solely from the elimination of resource-wasting trade costs.
6
Between 1870 and 1913, wages remained at 75 percent of British levels (Williamson 1995). See Scholliers (1995)
on the persistence of the low-wage economy.
5 rates.7 While Antwerp was the major hub of traffic, the share of five other international ports (in
order of importance: Ostend, Ghent, Brussels, Bruges, and Nieuwpoort) increased over the
period. According to one assessment (Suykens 1986, p. 375), competition among ports forced
Antwerp to maintain the lowest pilotage and loading fees in Europe.8
Notwithstanding the decline in transport costs, the fixed costs of acquiring international
market presence were considerable. These costs included gaining familiarity with local market
conditions, establishing or accessing wholesale networks, and learning about shipping methods
and customs’ formalities. Unlike France and Great Britain, Belgium did not reap the advantages
of a large empire; nor could Belgian rely on emigrants to promote its goods. The country was
compelled to find other means to reduce transaction costs.9 Steadfast in its adherence to free
trade, beginning in the 1860s Belgium accorded most favored nation treatment with its close
neighbors and new trading partners.10 The number of treaties peaked when France and Germany
adopted restrictive trade policies. The country also signed general trade agreements, traité
d’amitiés, like that with Venezuela in 1884, and protocols to safeguard product design and trade
marks, such as that with Romania in 1881 (Traités de commerce 1900). As for financial risk, the
adoption of the gold standard (1878) progressively locked in exchange rates with its commercial
partners.11 Combined, the uncertainty of doing business abroad declined since trade policy
reversals became more predictable—but also less likely.
Belgium’s success in foreign markets was all the more remarkable because, relative to its
major competitors, exporters had to overcome the ‘natural’ disadvantages related to the country’s
size.12 Big countries like Germany and the United States tended to have more large firms that
established merchant houses and employed agents in the field to market goods and generally
reduce setup costs (De Leener 1906, p. 156). German exporters also offered customers favorable
credit terms. The average size of firms was noticeably smaller in Belgium—we return to this
feature below—and industry concentration considerably less. Belgian firms had few if no
commercial agents in the field. Often they relied on British and German services, but these were
not always dependable (De Leener 1906, p. 16).
7
On maritime and inland transport costs, see Jacks and Pendakur (2011). On the Belgian rail network, see Huberman
(2012).
8
On the port of Antwerp, see Loyen (2002).
9
From 1885 to 1908, the Congo Free State was the private enclave of the Emperor Léopold II, after which it became
a colony of Belgium. In 1900, the Congo accounted for less than 5 percent of manufacturing exports. On empire and
trade, see Mitchener and Weidenmeir (2008).
10
In the wake of the Cobden-Chevalier treaty of 1860, Belgium signed a MFN agreement with France in 1861,
followed rapidly by treaties with Britain (1862), the Netherlands (1863), and later with Prussia (1865). Subsequently,
the country extended its treaty network and signed new agreements with a range of countries, from Romania (1880)
to Venezuela (1882), Mexico (1895), Greece (1904), and Bolivia (1912). On Belgian commercial policy, see
Degrève (1982).
11
On the gold standard and trade, see Lopez-Cordova and Meissner (2003).
12
De Leener (1906) provides an exhaustive account of Belgium’s strengths and weaknesses in foreign markets.
6 To level the playing field, the state expended considerable resources in nurturing “related”
(Rauch 1999) trade networks. At home, the country hosted a series of international trade fairs
(Antwerp 1885 and 1894; Brussels, 1888, 1897, and 1910; and Liège 1905) to grow its customer
base. Abroad, Belgium established consular offices that sought to deepen commercial ties and
reduce transaction costs. It was the first foreign country to have a chamber of commerce in New
York (Bairoch 1989, p. 97). By 1900, 470 representatives in 84 destinations (independent
countries and colonies) sought to promote exports, many of whom were foreign nationals (De
Leener 1906, p. 209).13 Altogether, spending on political and consular agents rose from 0.6
million francs in 1860s to 1.5 million in 1890 (Bairoch 1989, p. 96). The more than doubling in
expenditures mirrored the increase in the degree of openness.
The annual reports, the Recueil consulaire (hereafter RC), of the trade consuls provided an
important source of information on barriers to entry and prospects in foreign markets. The
consuls observed that trade costs varied by product as well as by market. The résumé on Chile
(RC, vol. 63, 1888, pp. 325-53) concentrated on exchange rates and tariff policy, that for Japan
(RC, vol. 39, 1882 vol. 39, pp. 23-71) on consumer preferences, and the report on the Philippines
on packaging and labeling (RC, vol. 142, 1908, p. 492). Faced by these specific entry costs, the
consuls made recommendations on opportunities in individual markets.14 Did this type of
information have any effect? Although we give a more definite answer below, at first pass, the
answer would be in the affirmative. For instance, despite being a relatively new outlet, Japan had
become Belgium’s eighth most important export destination by 1910.
The collapse in trade costs ought to have coincided with a movement of resources to the
export sector and the shuttering of uncompetitive import-competing concerns. After all, there was
no shortage of capital, and, because of its dense rail network, Belgian workers were highly
mobile across sectors (Huberman 2012). The terms of trade would have improved. For the most
part, however, structural change was glacial, economic performance being hostage to a handful of
regionally based staple industries, such as textiles and glass making (De Brabander 1981). The
second industrial revolution had not left much of a footprint. There was no discernible trend in
patents delivered (Annuaire statistique, various years) and human capital formation was weak,
levels of education being well behind the European norm. Summarizing the dominant narrative,
Van der Wee and Goossens (1991) described an economy "ossified, rigid, and imprisoned in
traditional, unviable sectors, without economic future." Overall, Belgian per capita growth until
1890 was 0.65 percent per annum, and only slightly more, 0.89 percent, in the decades before the
13
14
Figures from Recueil Consulaire vol. 106 (1900, pp. 3-45).
In the Chilean market, the trade delegate (RC vol. 63, 1888, p. 374) reported these market opportunities:
Nous avons toutes facilités pour entrer en concurrence pour les objets de lingerie confectionnée pour
bommes, de toutes étoffes (chemises, caleçons, gilets, pantalons, cols, cravates, etc.), la lingerie
confectionnée pour dames (bonnets, chemises, jupons, peignoirs, cols, collerettes, manches,
mouchoirs, etc.), la lingerie de table, de toilette et de bain (nappes, serviettes, peignoirs, caleçons,
etc). Les corsets unis ou très-ornes, les bretelles et les jarretières ont un placement régulier.
7 war (Horlings and Smits 2002, p. 86).15 The ‘modernization’ of Belgian industry had to wait until
the 1920s (Boschma 1999).
A defining feature of the Belgian economy was the presence of small firms. In principle,
in dynamic export economies, firms in the tradable sector are supposed to be larger, more capital
intensive, and more productive than those in the domestic sector. This does not seem to have
been the case before 1914 (Cassiers 1989). The absolute size of the domestic or putting-out sector
actually increased after 1896 (Belgium, Recensement 1910). As for the modern sector, compared
to its European rivals, the Belgian economy had a larger share of small companies and a lower
degree of concentration (Van Meerten 2003). Capital investment lagged relative to the industrial
leaders. In textiles, the median firm had fewer machines per worker than its major European
competitors, and in the capital-intensive transport sector the degree of concentration actually
declined after 1900 (Van der Wee 1984).16 Cooperation between investment banks and industry
was restricted to iron and steel, and even in this sector cartelization was less pronounced than in
Germany.
There are a number of competing explanations why the exponential rise in international
trade did not unleash stronger economic growth. The puzzle is in fact not particular to Belgium.
The evidence on the gains from trade before 1914 is decidedly thin (O’Rourke 2000; Irwin 2002;
Liu and Meissner 2013). For skeptics, tariff protection seems to have been an ally as much as an
adversary of productivity growth. But limited to aggregate or sectoral data, these studies have
difficulty in pinning down the link, or its absence, between trade and growth.17
Drèze (1989), in a classic statement anticipating Krugman, offered an interpretation of the
modest effects of trade based on the nature of exports. Because of its small domestic market,
Belgian exports were restricted to “standardized goods,” by which Drèze meant semi-finished
and non-differentiated goods.18 High entry costs in distant markets compelled firms to sell to
customers close to home and ship more of the same goods, as opposed to diversifying into new
markets and new goods. The high degree of substitution between foreign and Belgian goods had
the same effect. As a result, the lack of diversification dampened the potential gains of trade. The
trouble is that Drèze’s explanation does not mesh with statements of foreign diplomats on the
range of goods being shipped to faraway destinations nor does it correspond to the enduring
presence of small firms with short production runs. Rauch (1999) made the more general point:
By their very nature, few manufactured commodities, even standardized goods, are actually
traded in organized markets.
15
In contrast, for the periods 1870 to 1890 and 1890 to 1914, the degree of openness increased by 1.1 and 1.5 percent
per annum.
16
See Delbeke (1982) for a study of the persistence of small firms in Antwerp.
17
Bernard et al.’s (2007, p. 121) remark that empirical and theoretical work using the gravity equation framework
"typically concentrates on the aggregate value of trade and ignores the roles of firms and products" underscores our
research strategy to collect ‘micro’ trade data.
18
Drèze (1989) is a translation of the original French article published in 1960.
8 In the remainder of this paper, we explore the possibility that highly differentiated exports
could themselves have moderated growth. To see this, we first decompose the value of trade into
the number of products or markets, the extensive margin, and average sales per product or per
market, the intensive margin.19 The breakdown is informative because the margins mediated the
effects of fixed and variable costs on trade. Since entry costs, like that of obtaining information
about local preferences and setting up marketing channels and credit facilities, varied by
destination and by product, not all firms/goods/industries shipped to all markets. To be clear, the
extensive margin existed because firms that could not cover fixed costs did not export at all.
In order to identify the relation between the margins and trade costs on average
productivity, consider Chaney’s (2008) model of international trade in which firms produce a
range of differentiated goods.20 For a given fall in trade costs, a high degree of differentiation, or
a low elasticity of substitution between home and foreign goods, implies a relatively larger
extensive margin, because competition is limited and new products and firms capture market
share. Average productivity will fall in the presence of new, previously uncompetitive firms.
With a high degree of substitution, the action is on the intensive side. Because of stiffer
competition in these export lines, average productivity rises because new products and firms
make only a small dent in the market.
The breakdown of trade into its components provides a reconciliation of Belgium’s trade
and growth puzzle. In the face of differential entry costs, firms would have pursued various
strategies. Initially, they might have restricted export lines to a handful of the lowest cost
destinations, and then expanded the breadth of their markets, either because trade costs declined
or because of productivity improvements. When goods were highly differentiated, exports found
a considerable number of destinations. Under these circumstances, the extensive trumped the
intensive margin, the emergence of new products and new destinations thereby magnifying the
trade boom. However, the same decline in (fixed) trade costs may have enabled low-productivity
entrants to begin exporting.21 The harm caused by the influx of low productive firms was in step
with the degree of product differentiation. If overall specialization increased, the trade boom
might not have shown up in the productivity numbers as strongly as standard models have
anticipated.
III. Boats, cars, and trams: Belgian trade, 1870-1914
Our main data source is the Tableau général du commerce extérieur. We recorded the
trade of manufactured goods at five-year intervals beginning in 1870 and until 1910.
19
The definition of the margins of trade varies in empirical research because of data availability (Lawless 2010a).
Chaney (2008) includes a parameter for firm heterogeneity. In industries when there are few large firms, exports
are sensitive to small changes in trade costs, because many firms exit and enter.
21
Some firms with very low productivity, often those producing for the domestic market, might also be eliminated
due to competition from imports. We focus only on industries that had positive exports.
20
9 Manufactured goods represented about 50 percent of all exports in 1900 and 20 percent of all
imports. The reliability of our source has been discussed elsewhere (Horlings 2002).22 The mode
of transport appears to have been relatively stable over the period. For exports, about 45 percent
of goods were shipped by boat, a similar share by land transport, and 10 percent by barge using
the canal and river system.
The Tableau distinguished between all goods shipped (commerce général), including reexports and goods in transit, and items produced exclusively in Belgium (commerce spécial). The
attraction of concealing goods either as re-exports or special goods from the authorities
diminished during the period under investigation, because custom duties came down. But the
same fall in duties would have encouraged “disguised transit”, the incentive for merchants to
declare their goods intended for re-export as imports for domestic consumption, and subsequently
claim the same goods as exports of domestic origin (in order to avoid red tape and the delays
associated with leaving goods in transit).23 Horlings (2002, p. 120) reported that most disguised
transit comprised low-taxed, low-value bulk and crude commodities, rather than processed and
manufactured goods that are our interest. Altogether, measuring the accuracy of trade statistics
using exports and imports from source and destination countries, Federico and Tena (1989)
scored Belgium only slightly below the figure of the average European country.
A more pressing concern is the use of official prices. This is a common dilemma of
nineteenth-century trade statistics. For Belgium, the problem is minimized somewhat because
prices were adjusted annually, although their reliability varied across commodities. For some of
the early years, many goods were declared in value only. That said, the available price data is
quite detailed by category and variety of imports and exports. An issue is that recorded prices do
not vary with destination or source, again a common feature of historical trade statistics
especially those collected on the European continent. The sources of the official prices and the
nature of their revisions are unclear. Horlings (2002, p. 114) presumed they were wholesale
prices rather than c.i.f. or f.o.b prices. Despite these drawbacks, his own recalculation of trade
values concentrated on an earlier period. Until 1845, Horlings (2002, p. 117) reduced export
prices by the order of 6 percent, but “during the remainder of the century the revision of price
data caused a change of between 1 and 3 percent of import and export values.”
Our classification of products follows that recorded in the Tableau. The level of detail
corresponds to the SITC four or five-digit level. For a sub-group of cotton textiles in 1890, we
have these categories:
1. Produits fabriqués
22
Degrève (1982) presents an exhaustive treatment of the Tableau général. He was mainly concerned with
estimating the value of trade. His method of recapitulation does not permit decomposing trade between incumbent
and new products. Also, he did not record export destinations and import sources.
23
To avoid delays, merchants were willing to pay customs duties. Unlike items in transit, re-exports passed through
the national trade framework.
10 10. Ouvrage en matières de textiles
100. Tissus de coton unis
1000. Tissus de coton unis, croisés et coutils pesant 3 kilogr et plus les 100 mètres carrés
1001. Tissus de coton unis ou croisés pesant moins de 3 kilogr les 100 mètres carrés
10001. Tissus de coton unis, croisés et coutils pesant 3 kilogr et plus les 100 mètres carrés, écrus
10002. Tissus de coton unis, croisés et coutils pesant 3 kilogr et plus les 100 mètres carrés, blanchis
10003. Tissus de coton unis, croisés et coutils pesant 3 kilogr et plus les 100 mètres carrés, teints
10004. Tissus de coton unis, croisés et coutils pesant 3 kilogr et plus les 100 mètres carrés, imprimés
10005. Tissus de coton unis, croisés et coutils pesant 3 kilogr et plus les 100 mètres carrés, fabriqués en tout ou en
partie avec des fils teints.
A recurrent problem with this type of exercise is the ability to distinguish genuinely new
products from new titles given to existing products, and other products listed in a new separate
entry but were in fact previously folded into another group. For clearly identified new products,
like aluminum streetcars, this concern was immaterial, but in broader classifications, like 'cotton
yarn' in 1870, this problem was unavoidable. To minimize double counting, beginning with the
1870 product listing, we labeled as existing products 'new' products with similar titles, and other
supposedly new goods whose export or import values were of the same order of magnitude as
previously recorded established products. For example, when the 1880 return subdivided Tissus
de coton unis (100) into categories 1000 and 1001, we counted these as one new product. This
procedure reduced the number of products by roughly ten percent. For 1910, we recorded 171
distinct products or categories exported.24 The appendix gives the complete listing of products
and SITC codes.
With regard to export destinations and import sources, we had two considerations. We
defined countries according to their 1910 borders. For instance, New South Wales was recorded
as Australia. But since we also needed information on GDP, we treated Newfoundland as part of
Canada. Altogether, we had 69 countries listed as export destinations in 1910 based on the
original sources. This was considerable achievement since there were about 60 independent
states, and altogether 200 sovereign states, colonies, and territories in 1900. To be clear, our
dataset on trade contains no information about firms. Nonetheless, if each firm is identified with a
particular good, or if productivity within an industry is much less dispersed than across
industries, then there is not too much harm in using products as a proxy.25
24
25
For all years, 202 different goods were reported.
Our approach follows standard practice. See Dutt et al. (2013).
11 Table 1 registers the numbers of export and import products and commercial partners. For
exports, the increase in shipping destinations was very much steady; the growth in products was a
post-1890 phenomenon. In a world of zero gravity (Eaton and Kortum 2002), firms would have
sold products to many destinations, and the new products that came on stream would have been
shipped to a comparable set of markets. The steady accretion in destinations is consistent with the
presence of significant entry costs in foreign markets. As in Albornoz et al.’s (2012), firms
learned about their export potential market-by-market. Sources of imports remained restricted,
although the arc in the number of import categories mirrored that of exports.
Figures 2 and 3 present a closer look at the number of destinations served per product by
time. Observations (periods) are for five-year intervals beginning in 1870 and ending in 1910. In
Figure 2, we draw a histogram of the number of destinations served after one, five, and nine
observations of positive exports. The number of countries varied as the duration in the market
was extended. After one period, the un-weighted mean number of destinations was 10; but a good
with non-zero exports in all nine years in our study, served 20 markets on average.26 In Figure 3,
the number of destinations is estimated from a weighted regression of the number of countries
served by each product on the number of periods Belgium had non-zero exports of this item, as
well as a full set of period and good level indicators. After the first period, the conditional
average number of destinations was 17.75. The average number of markets served peaked after
five periods, leveling off at about 26 markets. Firms, it appears, built their market base
sequentially, the ability to stay in each market being strongly correlated with the number of
trading partners. This correspondence sits well with Besedeš and Prusa (2006) who report that,
while the duration of trade is very low for many products, if firms survive the initial period or
two, and meet operating costs, they stay in these markets for a long time.
The types of goods traded (Figures 4A and 4B), their destinations, and sources (Figures
5A and 5B) evolved over the period. The relative share of textile exports declined, although it
still amounted to about 30 percent, whereas that of clothing, metal products, and vehicles
increased. The import pattern was not too dissimilar, exactly what we would expect of trade in
specialized goods. The shares of imports of clothing, machines, metals, and vehicles all
increased, and that of textiles declined. The rest of Europe bought more than 90 percent of
exports in 1870, but by 1910 their share had dropped to about 65 percent, replaced by the
development of new markets in Africa, the Americas, and Asia. Imports of manufactured goods
came increasingly from Germany, presumably of differentiated goods or specialized inputs.
Can the increase in trade be attributed to the exports and import of higher value goods? Yi
(2003) attributed the rise in trade in the second wave of globalization to the advance in vertical
26
The value-weighted averages are much larger at 17.4 and 33.8 respectively. There is also left-censoring since
goods active in at least one market as of 1870 might have been active in previous years. If we exclude 1870, the
unweighted average after one year drops to 9. For goods active 8 periods or less, in any year after 1870, the average
is 13.7.
12 specialization and the corresponding rise in unit values. Along these lines, average unit values for
Belgium recorded in Table 1 show a marked rise in the period before 1914, although the trend in
industry values in Table 2 gives a more nuanced picture. Until the mid-1890s, unit values of
manufactured imports actually exceeded that of exports; thereafter the trend was reversed.
Among other causes, the rise in prices after 1890 might have represented the entry of capitalintensive firms producing high-quality-high-value added items or the entry of poor performers
with high marginal costs, or both. For textiles, export prices in 1905 were only slightly higher
than they were twenty year earlier, and this was the case for paper and glass products, and pottery
as well. But machine and metal prices more than doubled. There was also a spike in average unit
values because of big-ticket items including the boats, cars, tramways, and railroad cars that
Belgium began exporting near the end of the century. The mixed picture relates to the presence of
entry costs. In textiles, entry costs become relatively negligible because of the spread of
commercial agents and other supporting infrastructure. Many of new firms were small and had
high marginal costs, and, consequently higher prices, productivity advance being muted (Van
Houtte 1949). In contrast, the manufacture of boats, cars, and other vehicles took place in large,
capital-intensive establishments paying premium wages. For this sector, the upward movement in
prices corresponded to higher value added production. Cassiers (1989) gives evidence along these
lines for the 1920s.
To summarize, the degree of product differentiation of Belgian exports and imports
challenges the presumption of homogenous goods typical of factor-endowment trade models and
the standardized good model of Drèze. The basic Krugman model of imperfect competition does
not fare any better because it ignores fixed costs of entry. The dataset we have assembled points
to the contrary. The decline in entry costs, which varied across markets and products, had an
important role in determining the depth and scope of the Belgian trade boom.
IV. The size and determinants of the extensive and intensive margins
The expansion in product categories and in destinations and sources was the defining
feature of Belgium's trade performance before 1914. Was this a result of technology, falling trade
barriers, or both, as new trade theory suggests? To begin, we derive an overall picture of the trend
in product diversity of exports using a Theil index.27 The Theil index has the added feature that it
can be decomposed into within and between components. The former is equivalent to changes in
concentration along the intensive margin (defined here as the set of commonly traded goods over
a certain period); the latter captures changes in the extensive margin (defined as newly traded or
disappearing goods and/or trade partners).28
27
This paragraph follows Cadot et al. (2012).
Export concentration measured at the intensive margin represents inequality between the shares of active export
lines. Conversely, diversification at the intensive margin means convergence in export shares.
28
13 For all possible combinations of export products and destinations, the index in the top
panel of Figure 6 declines in value, indicating exports were becoming more diversified, although
after 1905 there is a small movement toward concentration.29 Belgium appears to have been a
typical middle-income country in which technological advance seems to have been increasingly
widespread, perhaps along the lines Temin claimed for Britain during the heyday of the industrial
revolution.30 This cannot be the entire story, however. In a strict Ricardian model, with highly
disaggregated data, shocks in technology would have been revealed in abrupt changes in the
extensive margin. But, individually and combined, the extensive and intensive margins have a
gradual pattern consistent with the presence of differential trade costs by product and by country.
Initially, the contributions of the two margins are about equal, but the intensive side increases
slightly half-way through the period. As for the decline in the extensive margin, sooner or later,
all trade folds into trade of existing product lines and with established markets. That said, the
tendency in concentration seems to be driven by the adding and dropping of products and
markets. The bottom panel in Figure 6 shows the same pattern for imports, although in this case
the share of the extensive trade is always larger.
An innovation of new trade models lies in the breakdown of the effects of trade costs on
the margins. In the baseline Melitz model (2003), the extensive margin depends positively on the
destination market’s GDP, and negatively on fixed and variable trade costs. The effect of a fall in
trade costs on the intensive side is more complex. The reduction in costs may increase or
decrease average sales depending on whether the market share of new entrants, who are always
less productive than incumbents, is relatively large.31 Average sales decline as market share of
new firms increases. Because of the uncertain effects of the fall in trade costs on the intensive
side, the contribution of the extensive margin to changes in total trade has generally been found
to be more important. The responsiveness of the extensive side is magnified when the degree of
substitution is low.
We use gravity models to assess the impact of trade costs on the margins. Following
convention (Lawless 2010a; Dutt et al. 2013), our dependent variables are: (1) bilateral exports or
imports; (2) the number of products exported to or imported from the trading partner (extensive
margin); (3) average sales to or from this country (intensive margin). Because of limited GDP
data, we use population of the trading partner as our measure of size. We include proxies for both
variable and fixed costs in our regression specification. In the absence of direct information on
freight rates, we use great circle distance between national capitals to approximate transport
29
We conduct a similar exercise for diversification based on destination and products separately. On both accounts,
the Theil declines in value.
30
In an influential paper, Imbs and Wacziarg (2003) found a U-shape pattern of diversification. At low levels of
income, countries' production is highly concentrated in a narrow range of products. As technology diffuses and skill
levels accumulate, product diversification intensifies. At high levels of income, countries revert to specialization.
31
More precisely, a decline in fixed costs lowers average sales because of the presence of new firms which are less
productive than incumbents. A reduction in variable costs has opposing effects. Incumbent firms sell more, but the
same decline in costs pulls in less productive concerns that previously were non-exporters. As a result, the effect of
lower variable costs on average sales per firm is ambiguous (Lawless 2010a).
14 costs. Other indicators of trade costs include whether or not the trading partner shared a common
border and language with Belgium, signed a trade agreement, and pegged its exchange rate. We
also exploit information on diplomatic representation both in Belgium and abroad, and include an
indicator variable for whether the trading partner is a colony of another country. To the extent
that these factors increased information and decreased uncertainty, they may be construed to be
good proxies for the fixed costs of trade, although we recognize that they could also plausibly be
associated with variable, ad valorem costs of trade.
Table 3 reports results for exports. For total exports, the coefficients have the expected
signs. Belgium traded more with larger countries that are geographically closer. It was relatively
easier to export to independent states than to colonies. Trade increased when the partner pegged
their exchange rate and had signed a trade agreement. Representation in a foreign market
increased the value of exports. There are subtle but important differences between the extensive
and intensive margins in columns two and three of the table. The former is less responsive to
distance than the latter. This is what we would expect if Belgium exported differentiated goods.
In line with the predictions of new trade models, the coefficients on variables that proxy for fixed
trade costs are generally larger in the regression of the extensive margin. A trade treaty does not
appear to have an effect on the extensive margin, perhaps because Belgium negotiated accords
with countries having previously established trade relationships. A complementary reading is that
because trade in differentiated goods was dependent on informal related networks, an MFN
agreement was less useful in opening doors than diplomatic ties.
(On the import side, not reported, the coefficients have the expected signs. Imports rise
with the size of the trading partner. The responsiveness of intensive and extensive margins to
distance is about the same, indicating, perhaps, that certain imports were basic (nondifferentiated) inputs into production or that the love of variety was limited to cheaper products
close to home. A trade treaty grows the number of imports products.)
Table 4 investigates the determinants of trade at varying levels of differentiation. If our
claims hold explanatory power, the association between trade costs and trade ought to vary by
level of differentiation.32 Standardized items, as Drèze maintained, would be shipped closer to
home, but highly differentiated goods could have found markets in far-flung destinations because
competition was slack. To see this more closely, we divided the products traded into four
categories of differentiation.33 The least differentiated items, category 1, include labor-intensive
manufactures and manufacturing inputs (for instance, threads, hides, parchment, and paving
stones) low-value metal products; category 2 consists of semi-skilled industrial goods and
textiles; and category 3 semi-skilled or high-skilled manufacturing goods with substantial capital
32
It would be interesting to study the differential impact of trade costs at different value to weight ratios. Low value,
high bulk goods should be more sensitive to distance and the opposite for high value goods. Informational proxies
should show up more strongly for industries with high fixed costs of production. We leave this exercise for future
work. It would also be profitable to classify goods by their elasticities of substitution. Unfortunately, the data do not
allow for estimation of these elasticities.
33
The appendix gives full details.
15 input (high-end textiles, crystal, and elaborate metal products). The most differentiated goods,
category 4, include high unit value capital-intensive manufactures, such as tramways, ships,
machines, and machine tools.
The effect of trade costs depended on the type of good exported. The difference between
category 1 and category 4 is telling in this regard. For the least differentiated goods, trade costs
strongly dampened trade. Exporters depended on proximity, a shared border, an MFN agreement,
and diplomatic representation to sell more goods. Distance was less of an obstacle for exporters
of the most specialized items. For these category of goods, better information and reductions in
uncertainty, such as a shared language and pegged exchange rate, abetted trade. The most highly
specialized goods seem to have sold by themselves, reputation being a more important
determinant of sales than MFN agreements or diplomatic representation. Interestingly, the
exposure of these goods in the colonies and protectorates of Britain, France, and Portugal was
delayed. This result supports Martínez López (2003) account of Belgian tramways. Investments
and sales in this sector were concentrated in the Middle East and Southern and Eastern Europe,
while U.K. manufacturers of similar goods were heavily invested in the British Empire.
The type of product exported also shows up in the relative size of the extensive and
intensive margins. Recall that in Chaney’s model the importance of the extensive margin varies
negatively with the elasticity of substitution between traded goods. New entrants can successfully
carve out market niches in highly differentiated goods (low elasticities) because they do not
directly compete with incumbents. In Figure 7, we compare the average size (weighted by value)
of margins across categories of differentiation. Our reference is category 1, the least
differentiated goods. Our results give comfort to the Chaney model. The value of intensive trade
is lower for categories 2, 3, and 4 than for the base, and the obverse holds for the extensive
margins. Category 3, composed of goods like high-end textiles, had more than 40 percent more
markets than the base category, and category 4, slightly less.
So far, our decomposition of trade has identified a large role for the extensive margin. The
introduction of new products and new destinations was sensitive to trade costs. The most highly
differentiated goods had the largest extensive margin because distance to market was not an
obstacle. In the next section, we look at the implications of these findings on growth in
productivity.
V. The margins of trade and productivity
New trade models provide conflicting predictions about the effects of trade costs on
aggregate productivity. A reduction in trade costs causes aggregate productivity to increase
because of positive selection effects. The exports sector benefits from the reallocation of
resources, the most productive firms within the set of exporters increasing their market share
(Melitz 2003); also, as revenues go up, exporters invest in new technologies (Bustos 2011). But a
decline in fixed costs also lowers the threshold level of productivity required for domestic firms
16 to begin exporting (Melitz 2003). This “negative” selection effect might be substantial in export
markets for differentiated goods. In these markets, firms do not compete head on against
commercial rivals and, as a result, they can succeed in capturing market share. Under these
conditions, the extensive (the number of goods or firms) trumps the intensive margin (sales per
good or sales per firm).34 In this section, we provide a simple test for the presence of these type of
selection effects during the Belgian trade boom. The collapse in the fixed costs of exporting
caused an expansion of the extensive margin. The growth of industry-level productivity
contracted as the extensive margin increased. This relationship was stronger in markets for
differentiated goods.
Our first step was to determine the association between openness and productivity at a
disaggregated level. Unfortunately, data on productivity is not as detailed as that on exports and
imports. We rely on Gaddiseur ‘s (1980; 1997) estimation of labor productivity growth for 20
industries for two subperiods, 1880-1886 and 1896-1910. The sectors studied comprise 80-90
percent of the products exported and imported. Average productivity growth of these industries in
the first period, 1880-1896, was 1.08 percent (1.26 on an export value-weighted basis); for 18961910, growth was 3.03 percent (2.45 percent on a weighted basis). We then proceeded to match
as many goods in our dataset as possible to the 20 industries. For 1880-1910, Figure 8 graphs the
relationship between the change in openness and average productivity for these years. The
‘circles’ represent industry size proportional to total exports in 1880. Overall, the trade boom did
not have much of an impact on productivity growth, if anything the relationship was negative.35
That said, the figure highlights considerable variation across industries.
The next step was to investigate the mechanism by which the degree of product
differentiation mediated the effect of trade on productivity. To see this, we broke down the
growth in the number of firms in the period by degree of specialization, our categories 1-4, where
group 4 included the most differentiated products. Recall that the growth in the extensive margin
was strongest in markets for more specialized goods. Because competition in these export lines
was less acute than for more standardized goods, the intensive margin in these sectors would
have been small or negative, and productivity growth weaker. Again, the evidence on firm entry
is not as detailed as our trade data, and it differs across sources. For this exercise, we use
evidence from the censuses (Belgium, Recensement, 1910) on firm entry and exit for 37 sectors
between 1896 and 1910. Figure 9 depicts that the entry of firms was most pronounced in highly
differentiated industries, categories 3 and 4. In the more competitive sectors, categories 1 and 2,
the number of firms actually declined. The flipside is that the intensive margin, and productivity
growth, was greater in non-differentiated goods sectors.
34
Of course, the intensive margin could be falling because variable trade costs have risen. This is unlikely to be the
case during our period.
35
In the regression underlying Figure 8, the coefficient of openness is -0.2360, with a t-statistic of -6.03; N = 17; Rsquared = 0.39; F = 36.38 (p = 0.00).
17 The fates of Belgium’s iron, textile, glass, and transport industries are consistent with this
line of reasoning. With regard to the degree of product differentiation, these industries are
representative of our categories 1 through 4. A typical Belgian foundry produced basic metals
(category 1), such as crude to medium grade iron and steel (mainly used for construction) that
were indistinguishable from its German or British rivals. The share of steel output exported was
about 30 percent, producers serving a narrow customer base of 30 destinations in 1910 (Brooks
and La Croix 1920). Despite quota agreements with foreign makers of girders and semi-products
limiting access to domestic markets, competition was stiff (Burn 1940, pp. 345-46). Belgian labor
costs remained cheaper than those of the Germans, and, where the British makers failed, the
Belgians had made a success of Thomas steelmaking, enjoying as well the benefits of
technological integration. As a result, productivity growth in the industry was strong, about 3.38
percent per annum from 1896 until 1910, versus the average rate of 3.03 percent.
In the low-end branch of cotton textiles (category 2), the number of firms increased as
trade costs collapsed in the 1890s. According to one estimate, there were 20 new spinning
establishments started after 1896 (Van Houtte 1949, p. 101).36 These establishments were often
small, others were temporary sheds located in the countryside. For the entire industry, the uptake
in ring spinning frames lagged that of its European competitors in low and medium value goods
(Saxonhouse and Wright 2004). Nonetheless, between 1880-1900 and 1900-1910 the number of
destinations served by the textile industry doubled to 50. By the later date, total exports amounted
to about 50 percent of production. The source of success for many of the new small
establishments was enhanced product differentiation, even at the low-end of the market. The
trade consul (RC, vol. 53, 1885, p. 138) in Lisbon acknowledged that, while the productivity of
British workers was superior and that Lancashire’s hold on foreign markets for ‘sheetings’ and
shirtings’ impenetrable, Belgians firms had opportunities for “towlings”, embroidered textiles,
checked domestics, among other narrow varieties. In line with the imperfect nature of
competition, productivity in this sector increased by a meager 0.15 percent per annum between
1896 and 1910.
Belgium was a global leader in high-end glass and crystal manufacture (category 3).
Production was highly specialized, about 80 percent of which was exported. Because of the
industry’s reputation, competition in foreign markets was moderate. The growth in the extensive
margin was composed of an increase in the number of destinations. Manufacturers had developed
better methods of packaging, and by 1910 they served about 60 destinations, including far-flung
markets in Asia.37 Industrial organization remained relatively unchanged over the period. The
number of firms remained stable at about 20, and technological change was absent. One authority
(Douxchamps 1951, p. 492) labeled the industry “artisanal”. All told, productivity advance
36
37
Van Houtte (1949, p. 103) estimated 142 establishments in the cotton-textile sector in 1896, and 306 in 1910.
On the packaging of crystal shipments to Persia, see Destree (1988).
18 between 1896 and 1901 was 1.03 percent per annum, less than half the average rate of
Gaddiseur’s 20 industries.
The transport industry (category 4), a leading sector of the second industrial revolution,
was a latecomer in Belgian economic history (Boschma 1999; Martínez López 2003). Initially
there were a handful of large companies producing tramways, entry in the industry being
naturally limited. But the sector did have access to a plentiful source of capital due to its
preferred relationship with banking, and the number of manufacturers of tramways increased
rapidly, from 31 to 51 in the five-year period ending in 1900 (Belgium, Annuaire statistique,
various years).38 Every manufacturer was also an exporter, foreign sales comprising about 90
percent of production (Baudhuin 1924, p. 132). The Tableau du commerce recorded 40
destinations for tramways in 1910. To fend off French and German rivals, foreign diplomats
heavily promoted sales that were often joined to complementary foreign investments. In the case
of Syria, one manufacturer received a concession of twenty years to operate the tramway system
and to construct of a hydro-electric plant (RC vol. 156, 1912, pp. 485-86). The overall scorecard
on the relationship between the margins and productivity is mixed. Because of the nature of
fixed-capital investments and the degree of competition, the negative selection effect was weaker
in this sector. Still, even as productivity surged by about 3.5 percent per annum, average sales per
firm in the decade before the war contracted (Van der Wee 1984, p. 209).
Finally, we estimate the relationship between productivity and firm entry for all 20
industries between 1880 and 1910, controlling for the degree of openness. The dependent
variable is the 30-year (cumulative) change of productivity from Gaddiseur (1980; 1997). The
independent variables are the 30-year change in (ln) firms, the extensive margin, and the
corresponding change in exports per firm, the intensive margin, and the 30-year change in output.
(In this instance, the firm evidence is taken from the Annuaire statistique [various years]). We
weight observations by the total value of exports within an industry. Our regression results are:39
∆
𝑦
𝑙
!
=
𝐸𝑥𝑝𝑜𝑟𝑡𝑠
0.37
0.37
−
Δ ln
(1.59)
1.52
𝐹𝑖𝑟𝑚
!
−
0.61
Δ ln 𝐹𝑖𝑟𝑚
2.45
! +
0.50
Δ ln 𝑂𝑢𝑡𝑝𝑢𝑡 ! .
3.36
Absolute values t-statistics are in parentheses. As anticipated, the growth in the number of firms
appears to have been the main driver behind the decline in productivity. The coefficient on the
intensive margins is not significant at the 10 percent level, the positive effect of the fall in trade
costs on exports per firm in standardized goods being offset by the opposite effect in
differentiated products. Our conclusion is that negative selection was a feature of globalization
before 1914. In this regard, the textile industry, as Van der Wee and Goossens (1991) described,
was still very much the typical Belgian industry. The full development of the transport and
38
Van der Wee (1984, p. 202) gives another measure of growth in the extensive margin. In 1900, he classified 19
firms in the transport sector among the largest fifty enterprises in the country; in 1910 the figure was 26.
39
N = 17; R-squared = 0.36; F = 16.18 (p = 0.00).
19 engineering sectors, the products of the second industrial revolution, had to wait until the
interwar years.
VI. Conclusion: Technology vs geography in the second industrial revolution
O’Rourke and Williamson (1999) have popularized the textbook factor-endowment model
to track the dynamics of trade and growth in the first era of globalization. Others, like Temin
(1997), have posited a Ricardian structure to back out the forces of early industrialization. These
competing paradigms dominate our understanding of nineteenth-century international trade, the
outcome of an inordinate focus on British trade and an over-reliance on aggregate trade statistics.
To a large extent, both approaches ignore the presence of intraindustry trade in differentiated
goods and a “love of variety”.
Exploiting granular data, we shift attention to Belgium, a small open economy in the
midst of the second industrial revolution during the pre-1914 trade boom. In the spirit of modern
trade theories, we find that trade costs explain as much, if not more, of the patterns of trade as
factor endowments. Fixed and variable costs mattered for the aggregate value of trade, the
direction of trade, and the composition of goods traded.
We have made a preliminary attempt to connect changes in trade costs and productivity.
The expansion of new trade partners, new goods, and new firms did not guarantee faster growth
in per capita income. Indeed, the opposite may have held. As the fixed costs of establishing new
goods in new markets came down, new entrants with lower productivity than incumbents
expanded market share. This process underlies the seemingly paradoxical acceleration in
openness beginning in the 1890s and Belgium’s lackluster productivity performance in the same
period.
Can we generalize from these results? It has become a cliché to claim that deeper
integration reallocates resources more efficiently. But the pathways underlying this process
cannot be taken for granted. In the presence of entry costs, a type of race can ensue between the
positive selection effects caused by shifting resources and the negative effects due to declining
trade costs, or, more broadly, technology vs geography. The bottom line is that globalization may
not always show up in the growth numbers. Our findings echo Rodrik’s (2011) observation that,
sooner or later, the process of international integration comes up against diminishing marginal
returns. If so, we have uncovered another way in which the second globalization echoes the first.
20 References
Albornoz, Facundo, Hector F. Calvo Pardo, Gregory Corcos, and Emanuel Ornelas (2012). “Sequential Exporting.”
Journal of International Economics 88: 17-31.
Allen, Robert C. (2009). The British Industrial Revolution in Global Perspective. New York: Cambridge University
Press.
Almanach de Gotha (1939). Annuaire généalogique, diplomatique et statistique. Gotha: Justus Perthes.
Bairoch, Paul (1989). “European Trade Policy, 1815-1914.” In The Cambridge Economic History of Europe, Volume
VIII, The Industrial Economies: The Development of Economic and Social Policies, edited by Peter Mathias and
Sidney Pollard. Cambridge: Cambridge University Press. Pp. 1-160.
Baudhuin, Fernand (1924). Le capital de la Belgique et le rendement de son industrie. Louvain: René Fonteyn. Belgium (various years). Annuaire statistique. Brussels.
Belgium (various years). Recueil consulaire: les rapports commerciaux des agents belges à l'étranger. Brussels.
Belgium (various years). Recensement de l’industrie et du commerce. Brussels.
Belgium (various years). Tableau général du commerce avec les pays étrangers, 1870-1914. Brussels: Ministre des
Finances.
Bernard, Andrew B., J. Bradford Jensen, and Peter K. Schott (2006). “Trade Costs, Firms and Productivity.” Journal
of Monetary Economics 53: 917-37.
Bernard, Andrew B., J. Bradford Jensen, Stephen J. Redding, and Peter K. Schott (2007). “Firms in International
Trade.” Journal of Economic Perspectives 21: 105-30.
Besedeš, Tibor, and Thomas J. Prusa (2006). “Ins, Outs, and the Duration of Trade.” Canadian Journal of
Economics/Revue canadienne d'économique 39: 266-295.
Boschma, Ron A. (1999). “The Rise of Clusters of Innovative Industries in Belgium During the Industrial Epoch.”
Research Policy 28: 853-71.
Brooks, Alfred H. and Morris F. La Croix (1920). The Iron and Associated Industries of Lorraine, the Sarre District,
Luxemburg, and Belgium. United States Department of Interior, Geological Survey. Washington: GPO.
Burn, Duncan Lyall (1940). The Economic History of Steelmaking 1867-1939: A Study in Competition. Cambridge,
U.K.: Cambridge University Press.
Bustos, Paula (2011). "Trade Liberalization, Exports, and Technology Upgrading: Evidence on the Impact of
MERCOSUR on Argentinian Firms." American Economic Review 101: 304-40.
Cadot, Olivier, Céline Carrère, and Vanessa Strauss-Kahn (2012). “Trade Diversification, Income, and Growth:
What Do We Know?” Journal of Economic Surveys 26: 1-23.
Cassiers, Isabelle (1989). Crosssance, crise et regulation en économie ouverte: La belgique entre les deux guerres.
Brussels: De Boeck-Wesmael.
Chaney, Thomas (2008). “Distorted Gravity: The Intensive and Extensive Margins of International Trade.” American
Economic Review 98: 1707-21.
21 De Brabander, Guido L. (1981). Regional Specialization, Employment and Economic Growth in Belgium Between
1846 and 1970. New York: Arno Press.
Degrève, Daniel (1982). Le commerce extérieur de la Belgique, 1830-1913-1939: Présentation critique des données
statistiques. Brussels.
Delbeke, Jos (1982). “The Mechanization of Flemish Industry, 1812-1930. The Case of Antwerp.” Katholieke
Universiteit Leuven.” Workshop on Quantitative Economic History. Discussion Paper 8103.
De Leener, Georges (1906). Ce qui manque au commerce Belge d’exportation. Reprinted by Forgotten Books 2013.
De Leener, Georges (1926). “Le commerce extérieur.” In La Belgique Restaurée, edited by Ernest Mahaim.
Brussels: Maurice Lamertin. Pp. 251-96.
Destree, Annette (1988). “Quelques tentatives Belges d’implantation industrielle et commercial en Perse dans les
dernières années du XIXe siècle.” Civilisations 38: 82-100.
Douxchamps, Yves (1951). L’évolution séculaire de l’indsutrie et du verre àvitres et de glacerie en Belgique du 1823
à 1913.” Bulletin de l’Institut de Recherche Économiques et Sociales, no. 5: 451-517.
Drèze, Jacques H. (1980). “The Standrd Goods Hypothesis.” In The European Internal Market, edited by Alexis
Jacquemin and André Sapir. Oxford: Oxford University Press. Pp. 13-32.
Dutt, Pushan, Ilian Mihov, and Timothy Van Zandt (2013). "The Effect of WTO on the Extensive and Intensive
Margins of Trade. INSEAD, working paper, 2013/38/EPS.
Eaton, Jonathan and Samuel Kortum (2002). "Technology, Geography, and Trade." Econometrica 70: 1741-79.
Eckel, Carsten and J. Peter Neary (2010). "Multi-product Firms and Flexible Manufacturing in the Global
Economy." Review of Economic Studies 77: 188–217.
Gaddiseur, Jean (1980). “La produit physique de l’économie belge, 1831-1913.” Ph.D. thesis, Université de Liège.
Gaddiseur, Jean (1997). “Output per Worker and its Evolution in Belgian Industry, 1846-1910.” In The Economic
Development of Belgium Since 1870, edited by Herman Van der Wee and Jan Blomme. Lyme, NH: Edward Elgar,.
Horlings, Edwin (1997). “The International Trade of a Small and Open Economy. Revised Estimates of the Imports
and Exports of Belgium, 1835-1990.” NEHA–Jaarboek 65: 110-42.
Horlings, Edwin and Jan-Pieter Smits (2002). “A Comparison of the Pattern of Growth and Structural Change in the
Netherlands and Belgium, 1800-1913.” Jahrbuch fur Wirtschaftsgeshichte 2: 83-106.
Huberman, Michael (2012). Odd Couple: International Trade and Labor Standards in History. New Haven: Yale
University Press.
Hummels, David and Peter J. Klenow (2005). “The Variety and Quality of a Nation’s Exports.” American Economic
Review 95: 704-23.
Imbs, Jean and Romain Wacziarg (2003). "Stages of Diversification." American Economic Review 93: 63-86. Irwin, Douglas A. (2002). "Interpreting the Tariff-Growth Correlation of the Late 19th Century." American
Economic Review 92: 165-69. Jacks, David, Christopher Meissner, and Denis Novy (2010). "Trade Costs in the First Wave of
Globalization." Explorations in Economic History 47: 127-41.
22 Jacks, David and Krishna Pendakur (2011). "Global Trade and the Maritime Transport Revolution." Review of
Economics and Statistics 92: 745-55.
Kindleberger, Charles P. (1956). The Terms of Trade: A European Case Study. New York: Wiley.
Krugman, Paul (1979). "Increasing Returns, Monopolistic Competition, and International Trade." Journal of
International Economics 9: 469-79.
Lawless, Martina (2010a). "Deconstructing Gravity: Trade Costs and Extensive and Intensive Margins." Canadian
Journal of Economics 43: 1149-72.
Lawless, Martina (2010b). "Geography and Firm Exports: New Evidence on the Nature of Sunk Costs." Review of
World Economics (Weltwirtschaftliches Archiv) 146: 691-707.
Liu, Dan and Christopher M. Meissner (2013). "Market Potential and the Rise of US Productivity Leadership."
NBER working paper no 18819. Loyen, Reginald (2002). “Throughput in the Port of Antwerp (1901-2000): An Integrated Functional Approach.” In
Struggling for Leadership: Antwerp-Rotterdam Port Competition, edited by Reginald Loyen, Erik Buyst, and Greta
Devos. Heidelberg: Physica-Verlag Publishers. Pp. 29-86.
López-Córdova , J. Ernesto and Christopher M. Meissner (2000). "Exchange-Rate Regimes and International Trade:
Evidence from the Classical Gold Standard Era." American Economic Review 93: 344-53.
Maddison, Angus (2013). Historical Statistics of the World Economy (webpage).
Martínez López, Alberte (2003). “Belgian Investment in Tramways and Light Railways: An International Approach,
1892-1935.” The Journal of Transport History 24: 59-77.
Melitz, Marc (2003). “The Impact of Trade on Intra-Industry Reallocations and Aggregate Industry Productivity.”
Econometrica 71: 1695-1725.
Melitz, Marc and Stephen J. Redding (2013). “New Trade Models, New Welfare Implications.” NBER working
paper no. 18919.
Mitchener, Kris James and Marc Weidenmier (2008). "Trade and Empire." Economic Journal 118: 1805-34.
O'Rourke, Kevin (2000). "Tariffs and Growth in the Late 19th Century." Economic Journal 110: 456-83.
O'Rourke, Kevin H. and Jeffrey G. Williamson (1999). Globalization and History: The Evolution of a NineteenthCentury Atlantic Economy. Cambridge, Mass: MIT Press.
Pahre, Robert (2007). Politics and Trade Cooperation in the Nineteenth Century. The ‘Agreeable Customs’ of 1815–
1914. New York: Cambridge University Press.
Rauch, J. E. “Networks Versus Markets in International Trade.” Journal of International Economics 48: 7-35.
Rodrik, Dani (2011). The Globalization Paradox: Democracy and the Future of the World Economy. New York:
Norton.
Romer, Paul (1994). “New Goods, Old theory, and the Welfare Costs of Trade Restrictions.” Journal of
Development Economics 4: 5-38.
23 Saxonhouse, Gary and Gavin Wright (2004). “Technological Evolution in Cotton Spinning, 1878-1933.” In The
Fibre That Changed the World: The Cotton Industry in International Perspective, 1600-1990s, edited by Douglas A.
Farnie and David J. Jeremy. Oxford: Oxford University Press. Pp. 129-52.
Scholliers, Peter (1995). “A Century of Real Industrial Wages in Belgium, 1840-1939.” In Labour's Reward: Real
Wages and Economic Change in 19th- and 20th-century Europe, edited by Peter Scholliers and Vera Zamagni.
Aldershot U.K. : Edward Elgar.
Suykens, F. (1986). Antwerp: A Port For All Seasons. Antwerp: MIM Publishing Company.
Temin, Peter (1997). “Two Views of the British Industrial Revolution.” Journal of Economic History 57: 63-82.
Temin, Peter (2000). “A Response to Harley and Crafts.” Journal of Economic History 60: 842-46.
Traités de commerce et de navigation. Conventions relatives aux attributions consulaires (1900). Braine-le-comte:
Zech et fils.
Van der Wee, Herman (1984). “Large Firms in Belgium, 1892-1974: An Analysis of their Structure and Growth.” In
Enterprise and History: Essays in Honour of Charles Wilson, edited by D.C. Coleman and Peter Mathias.
Cambridge: Cambridge University Press. Pp. 199-212.
Van der Wee, Herman, and Marine Goossens (1991). “Belgium.” In International Banking, 1870-1914, edited by
V.I. Bovykin and Rondo Cameron. New York: Oxford University Press. Pp. 113-28.
Van Houtte, F-X. (1949). L’évolution de l’industrie textile en Belgique et dans le monde de 1800 à 1939. Leuven.
Van Meerten, Michelangelo (2003). Capital Formation in Belgium, 1900-1995. Leuven: Leuven University Press.
Yi, Kei-Mu (2003). "Can Vertical Specialization Explain the Growth of World Trade?" Journal of Political Economy
111: 52-102.
24 Appendix 1: Comprehensive List of Goods Exported
and Degree of Differentiation
0 - semi-raw goods, 1 - processed goods (raw),
2 - semi-skilled manufacture, 3 - semi/high skilled manufacture,
4 - capital intensive highly differentiatied goods
SITC
code
613
613
851
831
831
831
651
651
651
651
651
651
651
651
651
651
651
651
651
651
651
651
Item
No.
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
651
654
654
654
654
654
654
654
654
654
654
654
654
654
652
652
23
24
25
26
27
28
30
31
32
33
34
35
36
37
38
39
Item
Peaux tannées, corroyées et autrement préparées
Peaux. Parchemin.
Peaux ouvrées. Chaussures
Peaux. Ouvrages de cuir et de peau. Ganterie
Maroquinerie
Peaux. Ouvrages de cuir et de peau. Autres.
Fils de laine non tors et non teints
Fils de laine tors et teints
Fils de laine peignée, non tors non teints
Fils de laine peignée, simples teints
Fils de laine peignée, retors non teints
Fils de laine peignée, tors ou teints
Fils de poils du chèvre, etc., non tors et non teints
Fils de poils du chèvre, etc., tors ou teinte
Fils de poils
Fils de soie
Fils de coton écrus et blanchis
Fis de coton teints et ourdis
Fils de coton mélangé d'au moins 20% de laine, coton dominant
Fils de lin, de chanvre et de jute non tors et non teints
Fils de lin, de chanvre et de jute tors et teints
Fils préparés pour la vente en détail. Fils de coton mesurant plus de 65,000 mètre
Fils préparés pour la vente en détail. Autre de toute espèce à l'exception du fil de
soie
Tissus de laine. Châles, écharpes et cachemires de l'inde
Tissus de laine. Châles et écharpes de laine
Tissus de laine. Coatings, duffels, calmoucks et autres tissus lourdes
Tissus de laine. Draps, Casimirs et tissus similaires
Tissus de laine; Tapis et tapisseries de laine
Tissus de laine pesant moins de 200 grammes par mètre carrée
Tissus de laine. Passementerie
Tissus de laine. Rubanerie
Tissus de laine. Tous autres.
Tissus de soie. Passementerie
Tissus de soie. Rubanerie
Tissus. Tulles, dentelles et blondes de soie
Tissus de soie
Tissus de coton, écrus
Tissus de coton blanchis
Level of
differentiation
1
1
2
1
2
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
25 652
652
40
41
652
42
652
652
652
652
652
652
652
652
652
652
652
654
654
654
654
654
654
652
654
655
656
654
656
858
848
848
848
842
841
845
845
845
845
845
845
43
44
45
47
48
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
845
621
635
635
635
80
81
82
83
84
Tissus de coton teints
Tissus de coton imprimés
Tissus de coton unis, croisés et coutils pesant 3 kilogr, plus les 100 mètres carrées,
fabriqués en tout ou en partie avec des fils teints
Tissus de coton unis ou croisés pesant moins de 3 kilogr les 100 mètres carrés
croisés et coutils pesant 3 kilogr, plus les 100 mètres carrées, écrus
Tissus de coton. Velours de coton. Façon de soie, écrus
Tissus de coton. Velours de coton. Façon de soie, teints
Tissus de coton. Velours de coton autres, écrus
Tissus de coton. Velours de coton autres, teints
Tissus de coton. Piqués, basins, façonnés, damassés et brillantes
Tissus de coton mélangé de soie, le coton dominant en poids
Tissus. Tulles, dentelles et blondes de coton
Tissus de coton. Passementerie
Tissus de coton. Rubanerie
Tissus de coton non dénommés
Tissus de lin, de chanvre et de jute. Toiles unies et croisées. Écrues
Tissus. Tulles, dentelles et blondes de lin
Tissus de lin, de chanvre et de jute blanchis et imprimés
Tissus de lin, de chanvre et de jute teintes
Tissus de lin, de chanvre, de jute autres que toile unies et croisées: Passementerie
Tissus de lin, de chanvre, de jute autres que toile unies et croisées: Rubanerie
Tissus. Tulles et dentelles de lin
Tissus de lin, de chanvre et de jute autres
Tissus. Broderies à la main
Tissus. Toiles cirées de toute espèce
Tissus. Tresses de paille de toute espèce
Tissus. Toutes sortes, non dénommées
Habillements, lingerie et confections. Lingerie de toute espèce
Habillements, lingerie et confections. Chapeaux non garnis
Habillements, lingerie et confections. Chapeaux garnis pour femmes
Habillements, lingerie et confections. Chapeaux garnis pour hommes
Habillements
Habillements, lingerie et confections. Vêtements pour hommes
Habillements, lingerie et confections. Bonneterie de coton
Tissus de laine. Bonneterie
Tissus de lin, de chanvre, de jute autres que toile unies et croisées: Bonneterie
Habillements, lingerie et confections. Bonneterie de soie
Habillements, lingerie et confections. Bonneterie autre
Habillements, lingerie et confections. Cols et manchettes en tissus de laine
Habillements, lingerie et confections. Objets confectionnés en tout ou en partie non
compris parmi ceux désignés ci-dessus
Caoutchouc ouvré
Balais communs
Futaies montées ou démontées
Bois ouvrés
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
2
2
3
3
3
3
1
2
1
1
26 642
642
642
892
892
892
892
661
666
666
666
666
666
85
86
87
88
89
90
91
92
93
94
95
96
97
666
666
98
99
666
666
666
664
664
664
664
664
664
664
664
665
665
665
665
664
664
665
664
684
684
682
682
687
687
672
672
672
672
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
Papiers à meubler
Papiers autres
Papiers. Autres
Produits typographiques. Livres en feuilles et en brochés
Produits typographiques cartonnés et reliés
Produits typographiques. Journaux et publications périodiques
Produits typographiques autres
Pierres polies et sculptées
Potteries. Faiences
Potteries. Porcelains
Potteries. Porcelaines
Potteries. Terre cuite, tuiles vernissés ou émaillées et tuiles à emboitement
Potteries. Terre cuite, tuiles autres
Potteries. Carreaux pour pavement et constructions, carreaux et pavés ceramiques
en terre fine
Potteries. Carreaux pour pavement et constructions, carreaux en ciment comprimé
Potteries. Carreaux pour pavement et constructions, carreaux en faience ou en
porcelaine
Potteries. Carreaux pour pavement et constructions, de tout espèce: autres
Potteries communes
Verreries. Glaces brutes
Verreries. Glaces
Verreries. Glaces étamées
Verreries. Ordinaires
Verreries. Fines
Verreries. Verres de vitrage mats sans dessins
Verreries. Verre de vitrage
Verreries. Verres de vitrage autres
Bouteilles, fioles, boubonnes, etc. en verre blanc ou demi blanc
Verreries. Communes. Bouteilles ordinaires
Boubonnes, dames-jeannes ou touries
Verreries. Communes
Verreries ordinaires. Bouteilles
Verreries ordinaires autres
Gobeleterie
Verrerie non dénommée
Métaux. Aluminium brut
Métaux. Aluminium battu, étiré ou laminé
Cuivre et nickel battus, étirés et laminés
Cuivre et nickel ouvrés
Métaux. Étain battus, étirés ou laminés
Étain ouvré
Fer. Ouvrages de fonte
Fer. Fer battu, étiré et laminé. Fils
Métaux. Fer battu, étiré ou laminé. Poutrelles
Fer. Fer battu, étiré et laminé. Rails
2
1
1
3
3
4
3
4
4
4
4
2
2
1
1
1
1
2
2
2
2
2
3
2
3
3
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
27 676
677
676
677
672
672
672
672
129
130
131
132
133
134
135
136
678
679
679
672
694
677
679
679
672
672
672
683
683
685
685
686
728
728
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
157
158
728
728
728
728
728
791
159
160
161
162
163
164
791
791
791
783
784
781
785
165
166
167
168
169
170
171
781
172
781
173
Métaux. Acier en barres, feuilles ou fils. Poutrelles
Acier en barres, feuilles et fils
Métaux. Acier en barres, feuilles ou fils. Tôles
Métaux. Acier fondu brut en barres feuilles ou fils. Autres
Fer. Fer battu, étiré et laminé. Tôles
Fer. Fer battu, étiré et laminé. Autres
Fer de fer. Clous
Fer. Ouvrages de fer. Ancres et chaînes pour la marine
Fer et acier. Fils d'acier clairs ou galvanisés, pour la fabrication des câbles et des
cordes
Métaux. Acier fondu brut
Fer et acier. Fils ou verges de fer ou d'acier
Fer et acier. Ouvrages spécialement dénommés au tarif officiel
Métaux. Acier ouvré. Clous
Acier ouvré
Fer et acier. Autres ouvrages. Ronces artificielles
Fer et acier. Autres ouvrages. Ronces artificielles
Fer de fer. Autres.
Métaux. Fer cuivré, nickelé, plombé ou zingué non ouvré
Métaux. Fer blanc ouvré
Fer-blanc. Ouvré
Métaux. Nickel ouvré
Métaux. Plomb battu, étiré ou laminé
Plomb ouvré
Zinc ouvré
Machines et mécaniques et outils. Courroies pour machines en cuir en caoutchouc
Machines et mécaniques et outils. Courroies pour machines en toute autre matière
Machines, mécaniques et outils: machines mécaniques non dénommées en
aluminium
Machines et mécaniques de fonte
Machines et mécaniques de fer et d'acier
Machines et mécaniques de fer et de bois
Machines et mécaniques de cuivre et de toute autre matière
Machines, mécaniques et outils: voitures pour chemins de fer et tramways en fonte
Machines, mécaniques et outils: voitures pour chemins de fer et tramways en fer et
acier
Machines, mécaniques et outils: voitures pour chemins de fer et tramways en bois
Machines, mécaniques et outils: voitures pour chemins de fer et tramways en cuivre
Voitures. Vélocipèdes
Voitures. Vélocipèdes. Parties et pièces détachées
Voitures
Voitures automobiles. Parties
Voitures. Motocycles et tous véhicules de l'espèce, autres que les voitures
automobiles. Complets
Voitures. Motocycles et tous véhicules de l'espèce, autres que les voitures
automobiles, parties
3
3
3
3
3
3
1
3
2
1
2
3
1
3
1
1
1
0
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
4
4
4
4
4
4
4
28 781
174
781
793
793
793
793
793
695
695
695
695
898
872
885
885
885
891
891
891
891
891
891
891
726
657
657
695
821
896
891
695
175
176
177
178
179
180
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
Voitures autres de toute espèce excepté les voitures pour les chemins de fer et
tramways: complets
Voitures autres de toute espèce excepté les voitures pour les chemins de fer et
tramways, parties
Navires et bateaux
Navires et bateaux. Toiles à voiles
Navires et bateaux. Ancres et chaines
Navires et bateaux. Bois pour mâts, vergues et espars
Navires et bateaux. Autres agrès et apparaux
Machines, mécaniques et outils en fonte
Machines, mécaniques. Outils en fer et en acier
Machines, mécaniques. Outils en bois
Machines, mécaniques. Outils en cuivre
Instruments de musique
Instruments de chirurgie, de précision
Montres et fournitures pou montres
Montres. Boîtes de montres
Fournitures pour montres
Armes à feu portatives. Armes de guerre
Armes
Armes à feu portatives. Pistolets, revolvers
Armes blanches
Armes Bois de fusils, finis ou non finis
Armes. Bouches à feu, mortiers, obusiers, etc.
Armes non dénommées
Caractères typographiques
Cordages
Filets et autres ustensiles pour la pêche maritime
Mercerie et quincaillerie
Meubles
Objets d'art et de collection
Poudre à tirer
Ustensiles et objets de ménage en fonte, en fer ou en acier émaillés
4
4
4
4
4
3
3
4
4
4
4
4
4
4
4
4
4
4
4
4
4
4
4
4
2
2
3
4
4
2
4
Table 1 Exports and Imports in Belgium, 1870-1910, Summary Statistics
29 Table 2 Unit Values of Exports, Belgium, 1870-1910
30 31 Table 3 Determinants of Total Exports, Intensive Margin, and Extensive Margin of Belgian Exports, 1870-1910
ln(distance)
Shared border
Pegged exchange rate
Shared Language
MFN treaty
Colony of another country
Partner has diplomatic rep. in Belgium
Belgium has diplomatic rep. in country
ln (population) of trade partner
Number of Observations
Total Exports
Extensive Margin (N)
Intensive Margin (X/N)
-0.63***
-0.26***
-0.41***
[0.077]
[0.033]
[0.055]
-0.05
0.08
-0.03
[0.137]
[0.093]
[0.149]
0.77***
0.45***
0.29**
[0.151]
[0.079]
[0.130]
-0.22
0.06
-0.19
[0.143]
[0.090]
[0.130]
0.41**
0.12
0.23*
[0.163]
[0.073]
[0.119]
-0.43*
-0.63***
-0.13
[0.247]
[0.141]
[0.171]
0.68***
0.44***
0.30**
[0.225]
[0.105]
[0.138]
0.84**
0.90***
0.74***
[0.333]
[0.219]
[0.205]
0.47***
0.28***
0.33***
[0.036]
[0.017]
[0.027]
639
639
639
Psuedo R-squared
Method of estimation
32 0.8
0.64
0.61
PPML
PPML
PPML
Notes: Dependent variable in column 1 is total exports to country d. Dependent variable in column 2 is the number of items in which Belgium
exports to country d---a measure of the extensive margin of exports. Dependent variable in column 3 is the total value of exports divided the
number of items in which Belgium exports to country d---a measure of the intensive margin of exports. Time dummies included but not
reported. Sample includes 1870, 1875,…,1910. Robust standard errors are reported in brackets.*** p-value<0.01, ** p-value < 0.05, * pvalue < 0.1. Sources: Distance is great circle distance in km between Brussels and capital of export destination or import source from LópezCórdova and Meissner (2000); exchange rate - López-Córdova and Meissner (2000): population - Maddison (2013); MFN from Pahre
(2007); diplomatic representation from Almanach de Gotha (1939) and Recueil Consulaire (various years).
33 Table 4 Determinants of Total Value of Belgian Exports by “Level of Differentiation”, 1870-1910
ln(distance)
Shared border
Pegged exchange rate
Shared Language
MFN treaty
Colony of another country
Partner has diplomatic rep. in Belgium
Belgium has diplomatic rep. in country
Category 1
Category 2
Category 3
Category 4
-1.07***
-0.54***
-0.41***
-0.32**
[0.088]
[0.055]
[0.056]
[0.138]
0.56***
1.02***
0.80***
0.40*
[0.153]
[0.102]
[0.098]
[0.226]
0.18
-0.51***
-0.27**
0.75***
[0.159]
[0.134]
[0.130]
[0.166]
-1.16***
0.42***
0.1
0.35**
[0.170]
[0.127]
[0.118]
[0.156]
0.56***
0.66***
0.31***
-0.2
[0.152]
[0.093]
[0.103]
[0.202]
-0.05
-0.70***
-0.45**
-1.26***
[0.220]
[0.183]
[0.197]
[0.417]
0.58***
0.56***
0.53***
0.82**
[0.193]
[0.146]
[0.142]
[0.336]
0.68**
1.50***
0.83***
0.34
[0.289]
[0.194]
[0.194]
[0.282]
ln (population) of trade partner
34 0.47***
0.44***
0.51***
0.45***
[0.027]
[0.022]
[0.024]
[0.041]
Number of Observations
13,560
23,818
16,617
17,318
Method of estimation
PPML
PPML
PPML
PPML
Notes: Dependent variable in each column is exports of product k to country d at a given level of differentiation. Method of estimation is
Poisson PML. Category 1 consists of least differentiated goods, such as labor-intensive manufactures and manufacturing inputs; category 2,
semi-skilled industrial goods and textiles; category 3, semi-skilled or high skilled manufacturing goods with substantial capital intensity;
category 4 consists of the most differentiated goods including high unit value capital intensive manufactures such as tramways, ships,
machines and machine tools.. Time dummies included but not reported. Sample includes 1870, 1875,…,1910. Robust standard errors
clustered over destinations are reported in brackets. *** p-value<0.01, ** p-value < 0.05, * p-value < 0.1. See Table 3 for sources.
35 7,1
70
7
60
6,9
50
6,8
40
6,7
30
6,6
ln (GDP per capita)
1910 BF
20
6,5
Openness
10
6,4
0
1870
1875
1880
1885
1890
1895
1900
Notes: Openness is defined as nominal exports plus imports divided by nominal GDP.
1905
1910
(EX + IM)/GDP
ln (GDP/pop.) M 1910 BF
Figure 1 Trade Openness and Real GDP per capita, 1870-1913
36 0
.02
density
.04
.06
Figure 2 Kernel Densities of Number of Destinations by Time in Market
0
10
1 period
20
30
Number of Destinations
5 periods
40
50
9 periods
Notes: Unweighted kernel density plots of the number of export destinations for each good exported. Number of periods refers to how long the
good has had a “spell” of non-zero exports to at least one country. Data are observed every five years. We count each of these as a period.
37 Figure 3 Number of Countries Served, by Time with Positive Exports
Number of Countries Served
40
35
30
25
20
15
10
5
0
1
2
3
4
5
6
7
Number of periods of non-zero exports
8
9
Notes: The line plot shows the average number countries served for a good which has had a “spell” of the given length of non-zero exports. The
averages are from a weighted regression of the number of countries served on a set of year indicators and good dummies. Weights are the total
value of exports of the good. Robust confidence intervals were used to form the error bounds.
38 Figure 4A Export Shares by Industry, 1870-1910
100% Other Percentage share of industry exports 90% Guns, rifles, weapons 80% Scien?fic instruments 70% Tools Boats 60% Automobiles 50% Machines 40% Metal products Paper, glass, poKery 30% Wood products 20% Rubber products 10% 0% 1870 Clothing Tex?les 1875 1880 1885 1890 1895 1900 1905 1910 Leather products 39 Figure 4B Import Share by Industry, 1870-1910
100% Other Percentage share of industry imports 90% Guns, rilfes, weapons Scien?fic Instruments 80% Tools 70% Boats 60% Automobiles 50% Machines 40% Metal products Paper, glass, poKery 30% Wood prodcuts 20% Rubber products 10% Clothing 0% 1870 Tex?les 1875 1880 1885 1890 1895 1900 1905 1910 Leather products 40 Figure 5A Export Shares by Destination and Source, 1870-1910 Percentage share of geographical region 100% 90% Germany & Lux 80% Northern EU 70% UK 60% France 50% Southern EU 40% Central Eastern EU 30% Americas 20% Africa & Asia & Australia 10% 0% 1870 1875 1880 1885 1890 1895 1900 1905 1910 41 Figure 5B Import Shares by Destination and Source, 1870-1910
100% Percentage share of geographical region 90% Germany & Lux 80% Northern EU 70% UK 60% France 50% Southern EU 40% Central Eastern EU 30% Americas 20% Africa & Asia & Australia 10% 0% 1870 1875 1880 1885 1890 1895 1900 1905 1910 42 Figure 6 Theil Indexes for Exports and Imports by Classification
1870
1875
1880
1885
1890
1895
1900
1905
1910
1870
1875
1880
1885
1890
1895
1900
1905
1910
1,5
theilB_imp
_prod
1
theilW_imp
_prod
0,5
0
1870
1875
1880
1885
1890
1895
1900
1905
1910
1910
1905
1900
1895
0
1890
theilW_imp_dest
1
0
1885
theilB_imp_dest
1,5
0,5
1880
2
2
1
1875
2,5
1870
1875
1880
1885
1890
1895
1900
1905
1910
theilW_imp
Theil Index Products Imports
Theil Index
2,5
Theil Index
4
1870
Theil Index
Theil Index Sources Imports
3
theilB_exp_prod
theilB_exp_prod
theilW_exp_prod
theilW_exp_prod
1
0
5
2
0,5
1,5
0
3,5
theilB_imp
1
2
0 0,5
6
3
1,5
Theil Index Products Exports
Theil Index Products Exports
2,5
00,5
Theil Index Prod-Sources Imports
.
2
Theil Index
0,5 1
theilB_exp_dest
theilB_exp_dest
theilW_exp_dest
theilW_exp_dest
2,5
Theil Index
2 2
1,5
1,5
1
1910
1905
1900
1895
1890
1885
1880
theilW_exp
Theil Index Destinations Exports
Theil Index Destinations Exports
2,5
2,5
Theil Index
Theil Index
theilB_exp
1875
5
4,5
4
3,5
3
2,5
2
1,5
1
0,5
0
1870
Theil Index
Theil Index Prod-Destination Exports
43 Figure 7 Relative Extensive and Intensive Margins by Level of Differentiation
0,6
0,4
ln point difference
0,2
0
-0,2
-0,4
-0,6
-0,8
-1
-1,2
-1,4
2
3
4
level of differentiation
Extensive Margin (N)
Intensive Margin (Value/N)
Notes: Values plotted are the estimated dummies for each class of differentiation in a regression of the extensive margin (number of countries) on
a set of indicators for level of differentiation and time dummies. Data are pooled, 1870-1910. The omitted category (category 1) is the least
differentiated product group. Category 4 contains the most differentiated products. See appendix for details.
Figure 8 Average Productivity Growth and Average Openness, 1880-1910.
Notes and sources: For regression details see text and footnotes. The ‘circles’ represent industry size proportional to total exports in 1880.
Labor productivity is from Gaddiseur (1980, 1997). Openness is calculated by industry from Tableau général du commerce.
44 45 Figure 9 Entry and Exit of Firms by Category of Product Differentiation, 1896-1910
category 4,
45,99
% cahnge in number of firms
50,00
40,00
30,00
category 3,
13,03
20,00
10,00
% change in
number of firms
0,00
-10,00
category 2,
-5,93
-20,00
-30,00
-40,00 category 1,
-33,02
Category of differentiation
Notes and sources: Category 1 contains least differentiated products. See appendix for details. Number of
firms by category taken from Belgium, Rencensement (1910).