Category Archives: Economic Geography

What Chance Change? Driving Development through Transport Infrastructure

Locomotives of Local Growth: Short- and Long-Term Impact of Railroads in Sweden

By Thor Berger (Lund University) and Kerstin Enflo (Lund University)

Abstract: This paper uses city-level data to examine the impact of a first wave of railroad construction in Sweden, between 1855 and 1870, from the 19th century until today. We estimate that railroads accounted for 50% of urban growth, 1855-1870. In cities with access to the railroad network, property values were higher, manufacturing employment increased, establishments were larger, and more information was distributed through local post offices. Today, cities with early access to the network are 62% larger and to be found 11 steps higher in the urban hierarchy, compared to initially similar cities. We hypothesize that railroads set in motion a path dependent process that shapes the economic geography of Sweden today.


Review by Alexander Horkan (final-year PPE student, Queen’s University Belfast)

What impact did the introduction of railroads to Sweden have on town-level growth? This is the question being explored by Thor Berger and Kerstin Enflo, both of Lund University, in their EHES working paper circulated as part of NEP-HIS-2013-08-05. The paper focusses on the early development of the Swedish railroad network, between 1855 and 1870, and examines whether towns with early access to the network[1] experienced higher levels of expansion of economic activity, using population growth as a proxy measure for this. They expand the possibility of their have been effects beyond merely the initial shock and scrutinise whether there was a long-run impact on economic development over the 20th Century.

Berger and Enflo contribute to the discourse on the value of transport infrastructure to lowering trade costs, which frequently hypothesises that large infrastructure projects foster economic development ‘ahead of demand’. Although an intuitive suggestion serving as a core belief of policymakers regarding the localisation of growth and planning possibilities, it is historically troublesome to provide evidentiary credence that such growth is independent from endogenous, observable and unobservable preconditions. Modern transport infrastructure is rarely assigned randomly to locations, instead being focussed around connecting ‘hubs’ that inevitably possess advantageous biases towards growth. This builds on various works detailing how such biases plague neutral analysis of development, as infrastructure projects are seemingly inextricably linked with political interference at either end of the spectrum, whether promoting growth in areas of economic sterility, or those already growing through endogenous factors.

Berger and Enflo show how railroads affect the location, not the level, of growth

Does railroad access increase the overall level of growth, or just the location of growth?

This paper seems to be of extreme relevance to current debates surrounding the future of a high-speed rail network connecting Birmingham to London in the UK. Contemporary debates have been hazy, lacking clear focus on precise and demonstrable economic incentives, leading to many questioning the value brought to northern cities. This research can increase the scope of such debates, providing clear evidential support that early adoption of technological advancements in transport infrastructure ignites and fosters long-term economic growth, yet simultaneously causes large negative ‘spillover’ effects on nearby, unconnected towns. Such research seems valuable and relevant to both sides of the question and must only serve to enrich any subsequent discussion.[2]

Proof of their hypothesis is offered through the calculation of comparative populations of cities both connected and unconnected to the railroad network between 1855 and 1870. Through using a difference-in-difference framework, they show that those who gained early exposure to the rail network grew larger, with additional population growth of 26% on average. Such increases imply that levels of urbanisation in 1870, and the aggregate rate of growth by the same point, would have ‘decrease[d] by 15% and 50% respectively’ (p. 3) independent from rail infrastructure. These calculations prove correlation between the exposure to railways and subsequent growth, echoing work by Fishlow (1965).


Where Bergen and Enflo really contribute to expanding existing literature, however, is by providing robust justification to draw direct causal relationships between railroad placement and subsequent ‘ignition’ of economic development. This is achieved through a tripartite construct, initially matching observationally similar towns and their growth patterns before the railway introduction. These measures ensure that observable differences are not key to explaining growth of specific towns, i.e. they were not already growing faster than surrounding cities.

Secondly, they calculate a strong instrumental variable; this relies on proposed routes drawn up by Adolf van Rosen in 1845 and subsequently by Nils Ericson in 1856. As such routes were constructed in relative isolation of political and economic pressures; favouring conditions of topographical simplicity and military strategic importance (avoiding coastal areas traditionally predisposed to growth) such an instrument is robust in corroborating the evidence of the first measure. By estimating the pre-rail differences in population growth for towns included in these original plans, and calculating their relative differences as close to zero, further corroboration is given to assertions that there were no pre-existing conditions conducive to growth in these towns.

The final measure is the imagined construction of these proposed lines, and further ‘low cost routes’. By creating this strong counterfactual, the authors presuppose that these lines that were not built, due to political obstinacy and lassitude, and those proposed later, to link profitable hubs of commerce would show large increases in populations if the driving factor behind growth was some unobservable, predetermining factor. Conversely however, if growth failed to materialise, it would be clear that the most significant force at work was early exposure to railways.


What can policymakers today learn from the Swedish case?

In his 1964 paper Robert Fogel identified the aggregate contribution of railroads to the US economy through social savings, deeming it of very little significance to social savings against a comparable counterfactual canal system. The measures used by Berger and Enflo are inversely interested in the relative impact of the railroad on cities. The negative ‘spillovers’ to nearby, unconnected towns examined in this paper further confirm Fogel’s argument that, whilst railroads had little impact on aggregate economic activity, they had large effects on relative growth patterns.

The final key significance Berger and Enflo draw out is the persistency of the impact of early exposure to rail networks. There are a myriad of reasons for this: high value sunk investments provide large barriers to both entry to and exit from the market, prompting concentration of economic activity in specific places. Additionally emerging towns become identifiable with growth and development, thus almost gaining critical mass and organically attracting further growth by this virtue. This emergent path dependency mirrors that cited by Bleakley and Lin (2012) regarding US cities being focussed around portage sights, despite the increasing irrelevance of such a factor. The implications of this paper however shadow those of Redding, Sturm and Wolf (2011) and Jebwad and Moadi (2011), examining man-made advantages over natural ones, contributing more greatly to discourse on policy implications and growth strategy.


Throughout the paper, however, despite great lengths to isolate geographical preconditions for local growth, there was an absence of discussion regarding elasticity of demand for rail services across the country. It seems remiss to address reduction of trade costs, whilst ignoring the possibility for elasticity of demand for such services, for example during winter months where winter roads open new avenues of trade, significantly reducing goods transportation costs via substitutions. Such questions could raise insightful analysis of unexplored geographical factors in northerly cities not experiencing the same degree of negative ‘spillovers’ suffered by more central ones.

The scope of this rigorous analysis could be expanded beyond current high-speed rail debates explored above to varying fields. Pertinent could be investigation of whether such findings have significance surpassing large-scale travel infrastructure and technological advancements, to the increasingly relevant information and communication sector for example; examining whether early adoption of communications advancements and infrastructure lead growth in specific locations.


[1] Less than a third of towns were connected by the end of this period, and only around a tenth of the peak network size had been realised.

[2] For a wider discussion of the minutia of this debate please refer to:




Bleakley, H. and Lin, J. (2012). Portage and Path Dependence. The Quarterly Journal of Economics 127, 2, 587{644.

Fishlow, A. (1965). American Railroads and the Transformation of the Ante-bellum Economy. Vol. 127. Cambridge: Harvard University Press.

Fogel, R. (1964). Railroads and American Economic Growth. Baltimore: John Hopkins Press.

Jedwab, R. and Moradi, A. (2011). Transportation Infrastructure and Development in Ghana. Mimeo.

Redding, S. J., Sturm, D. M., and Wolf, N. (2011). History and Industry Location: Evidence from German Airports. Review of Economics and Statistics 93, 3, 814{831.


When did Globalization Actually Start?

West versus East: Early Globalization and the Great Divergence’

By Rafael Dobado-González (Complutense), Alfredo Garcia-Hiernaux (Complutense), and David E. Guerrero-Burbano (CUNEF)

Abstract: This paper extends our previous work on grain market integration across Europe and the Americas in the eighteenth and nineteenth centuries (Dobado, García-Hiernaux and Guerrero, 2012). By using the same econometric methodology, we now present: 1) a search for statistical evidence in the East of an “Early Globalization” comparable to the one ongoing in the West by mid eighteenth century; 2) a study on the integration of grain markets in China and Japan and its functioning in comparison to Western countries; 3) a discussion of the relevance of our findings for the debate on the Great Divergence. Our main conclusions are: 1) substantial differences in the degree of integration and the functioning of grain markets are observed between East and West; 2) a certain degree of integration may be reached through different combinations of factors (agents, policies, etc.) and with dissimilar effects on long-run economic growth; 3) the absence of an “Early Globalization” in the East reveals the existence of some economic and institutional limitations in this part of the world and contributed to its “Great Divergence” with the West from at least the eighteenth century.


Circulated by NEP-HIS on: 2013-08-10

Guest Entry by Elizabeth Meagher (Bangor University)

NB: In January 2014, Chris Colvin and I started an experiment involving the use of Web 2.0 tools and third year undergraduate students. The aim was to incentivise active participation in academic exchange by reviewing recent additions to the broad literatures of business and economic history whilst following the same format as the NEP-HIS blog and limited to 1,000 words or less. The best entries are then published in the blog with minimal editing. This is the first of such entries.

We appreciate comments and inspiration from John Turner (Queen’s Belfast), Mar Rubio (Publica de Navarra) and Marcel Salles (ITESO). (BBL – Ed)


This paper examines the causes of early Globalization and the rise of the Industrial Revolution, analysing the divide between the East and West known as the ‘Great Divergence’. The literature suggests that globalization began steadily in the first half of the eighteenth century, taking off rapidly after the French Revolution and Napoleonic Wars and their aftermath. Once globalization ‘regained momentum’ it was supported by the rise in the Industrial Revolution across Europe and the US (Dobado-Gonzalez et al, 2013).

Source: The Economist "What was the Great Divergence"

Source: The Economist “What was the Great Divergence”

De Vries (2010) separates Globalization into two categories, ‘soft’ and ‘hard’. Flynn and Giraldez (2004) claim that ‘soft globalization’ began when the ‘old world’ merged with the US back in 1571. O’Rourke and Williamson (1999) defined ‘hard globalization’ as the integration of markets across a geographical location. An earlier study by Dobado and Guerrero (2009) claims that literature has often ignored earlier influences of globalisation on economic growth and focuses more on the results after the Industrial Revolution from 1700s onwards. It cannot be disregarded that integration between countries across Europe was evident between 1500 and 1800 before the first Industrial Revolution began. It is from these early stages of market and trade integration across Europe that industry and globalisation developed. The paper also suggests that in the first half of the eighteenth century, the rise in the West was due to the fact that China and the East had little to compete with, leaving the gates wide open for the West to expand its operations.

A panoramic view of London, c.1670 by Wenceslaus Hollar. Photograph: Heritage Image Partnership Ltd/Alamy Source:

A panoramic view of London, c.1670 by Wenceslaus Hollar. Photograph: Heritage Image Partnership Ltd/Alamy Source:

Additionally Dobado and colleagues focus on the impact of the dramatic expansion of foreign trade on economic growth within the West which has continued through to today. The paper also compares the differences and separation between the West and East during the first major economic growth in Europe between 1500 and 1800, known as the ‘First Great Divergence’. The authors suggest that the openness of the West resulted in dynamic trading across the Atlantic between the US with Britain and the Netherlands in particular (Dobado et al, 2013). The East however was not so far behind with the Opium Wars and the Silk Road increasing trade across Eastern countries moving towards the West which had never been done before. Nevertheless, before this point, the paper suggests that Eastern governments made one of the greatest economic mistakes in closing their economies, giving way to the ‘Great Divergence’ and ‘exceptionalism’ within the West (Dobado et al, 2013; Dobado and Guerrero, 2009). The restrictive trade policy practiced by the East is claimed to have prevented that part of the world from taking advantage of both direct and indirect benefits resulting from the expansion of trade during the Early Modern Era (Dobado et al, 2013). The paper recognises the absence of international grain market integration within the debate on the ‘Great Divergence’ and therefore seeks to examine the different markets within both the East and West.

St Paul's Cathedral viewed from Southwark, across the River Thames, in 1859. Photograph: William England/Getty Images Source:

St Paul’s Cathedral viewed from Southwark, across the River Thames, in 1859. Photograph: William England/Getty Images Source:

The paper found that despite geographical proximity and the easiness of transportation between China and Japan in the East, no statistical evidence of grain market integration is found between the two countries. This finding contrasts with the increasing expansion of trade in the West both before 1792 and after the 1840s (Dobado et al, 2013). As a result the East and West were dissimilar in terms of market integration both before and after the Industrial Revolution. Therefore this supports the concept of the ‘Great Divergence’ between the East and West and early Globalization presented direct and indirect benefits for the West which left the East to fall behind. In closing their economies in the Early Modern Era, governments within the East had committed what might be considered one of the biggest economic policy mistakes ever made, losing out on the greater benefits of early Globalization (Dobado et al, 2013).


One aspect of the research which could be criticised is that the samples used for comparing the different grain markets in both the East and West are unequal, with twelve studied within the West, and eleven in the East (Dobado et al, 2013). This could suggest the results may be biased towards the wheat market within the West. In addition to this, the study works with “[…] long yearly data series covering most of the eighteenth and the nineteenth centuries for most markets” (Dobado et al, 2013, p.6). This could also alter the results as using the same data is critical for a fair analysis. Another valuable point to consider is that the only countries which gained from ‘Early Globalization’ in Latin America were Peru and Chile who traded wheat.

It may also be beneficial to take into account cultural differences between the East and West at this time. The demand for wheat across Europe and the US would have been greater than the demand for rice from the East, causing further integration within the East.

Based on the points discussed above, it is apparent that there is a strong divide between the economic activity within both the East and West before the Industrial Revolution gained momentum. Additionally, it is evident that Globalization was present within Europe between 1500 and 1800 with the integration of markets within the West.


De Vries, J. (2010) ‘The Limits of Globalization in the Early Modern World’, The Economic History Review, Vol.63, pp 671-707.

Dobado-Gonzalez, R., Garcia-Hiernaux, A., and Guerrero-Burbano, D. (2013) ‘West verus East: Early Globalization and the Great Divergence’, accessed 6 March 2014.

Dobado, R. and Guerrero, D. (2009) ‘The Integration of Western Hemisphere Grain Markets in the Eighteenth Century: Early Progress and Decline of Globalization’, Accessed 9 March 2014.

Flynn, D.O. and Giraldez, A. (2004) ‘Path Dependence, Time Lags and the Birth of Globalisation: A Critique of O’Rourke and Williamson”, European Review of Economic History, Vol.8 No.1, pp 81-108

O’Rourke, K.H. and Williamson, J.G. (1999) Globalization and History, Cambridge, MA: The MIT Press.

The challenges of updating the contours of the world economy (1AD – today)

The First Update of the Maddison Project: Re-estimating Growth Before 1820

by Jutta Bolt (University of Groningen) and Jan Luiten van Zanden (Utrecht University)

Abstract: The Maddison Project, initiated in March 2010 by a group of close colleagues of Angus Maddison, aims to develop an effective way of cooperation between scholars to continue Maddison’s work on measuring economic performance in the world economy. This paper is a first product of the project. Its goal is to inventory recent research on historical national accounts, to briefly discuss some of the problems related to these historical statistics and to extend and where necessary revise the estimates published by Maddison in his recent overviews (2001; 2003; 2007) (also made available on his website at


Review by Emanuele Felice

Angus Maddison (1926-2010) left an impressive heritage in the form of his GDP estimates. These consider almost all of the world, from Roman times until our days, and are regularly cited by both specialists and non-specialists for long-run comparisons of economic performance. The Maddison project was launched in March 2010 with the aim of expanding and improving Maddison’s work. One of the first products is the paper by Jutta Bolt and Jan Luiten van Zanden, which aims to provide an inventory while also critically review the available research on historical national accounts. It also aims “to extend and where necessary revise” Maddison’s estimates. This paper was circulated by NEP-HIS on 2014-01-26.

The paper starts by presenting, in a concise but clear way, the reasons that motivated the Maddison’s project and its main goals. It also tells that some issues are left to be the subject of future work, particularly thorny issues left out include the use of 2005 purchasing power parities rather than Maddison’s (1990) ones; and the consistency of benchmarks and time series estimates over countries and ages.

Jutta Bolt

Firstly (and fairly enough, from a ontological perspective) Bolt  and van Zanden deal with the possibility of providing greater transparency in the estimates. Instead of presenting the margins of errors of each estimate (which in turn would be based “on rather subjective estimates of the possible margins of error of the underlying data”), the authors, following an advice by Steve Broadberry, choose to declare explicitly the provenance of the estimates and the ways in which they have been produced. This leads to classifying Maddison’s estimates in four groups: a) official estimates of GDP, released by national statistical offices or by international agencies; b) historical estimates (that is, estimates produced by economic historians) which roughly follow the same method as the official ones and are based on a broad range of data and information; c) historical estimates based on indirect proxies for GDP (such as wages, the share of urban population, etc.); d) “guess estimates”.

Jan Luiten van Zanden

Then the article moves on to review and discuss new estimates: although revisions for the nineteenth and twentieth century (mostly falling under the “b” category) are also incorporated, the most important changes come from the pre-industrial era (“c” kind estimates). For Europe, we now have a considerable amount of new work, for several countries including England, Holland, Italy, Spain and Germany (but not for France). The main result is that, from 1000 to 1800 AD, growth was probably more gradual than what proposed by Maddison; that is, European GDP was significantly higher in the Renaissance (above 1000 PPP 1990 dollars in 1500, against 771 proposed by Maddison); hence, growth was slower in the following three centuries (1500-1800), while faster in the late middle ages (1000-1500). For Asia, the new (and in some cases very detailed) estimates available for some regions of India (Bengal) and China (the Yangzi Delta), for Indonesia and Java, and for Japan, confirm Maddison’s view of the great divergence, against Pomeranz revisionist approach: in the late eighteenth and early nineteenth century, a significant gap between Europe and Asia was already present (for instance GDP per capita in the whole of China was 600 PPP 1990 dollars in 1820, as in Maddison; against 1455 of Western Europe, instead of 1194 proposed by Maddison).

New estimates are also included for some parts of Africa and for the Americas, with marginal changes on the overall picture (for the whole of Latin America, per capita GDP in 1820 is set to 628 PPP 1990 dollars, instead of 691). For Africa, however, there are competing estimates for the years 1870 to 1950, by Leandro Prados de la Escosura (based on the theoretical relationship between income terms of trade per head and GDP per capita) on the one side, and Van Leeuwen, Van Leeuwen-Li and Foldvari (mostly based on real wage data, deflated with indigenous’ crops prices) on the other. The general trends of these differ substantially: the authors admit that they “are still working on ways to integrate this new research into the Maddison framework” and thus at the present no choice is made between the two, although both are included in the data appendix.

New long-run estimates are presented also for the Near East, as well as for the Roman world, in this latter case with some differences (smaller imbalances between Italy and the rest of the empire) as compared to Maddison’s picture. The authors also signal the presence of estimates for ancient Mesopotamia, produced by Foldvari and Van Leeuwen, which set the level of average GDP a bit below that of the Roman empire (600 PPP 1990 dollars per year, versus 700), but they are not included in the dataset.

Per capita GDP in Roman times, according to Maddison (1990 PPP dollars)

Per capita GDP in Roman times, according to Maddison (1990 PPP dollars)

What can we say about this impressive work? First, that it is truly impressive and daring. But then come the problems. Needless to say Maddison’s guessed estimates is one of the main issues or limitations, and this looks kind of downplayed by Bolt and van Zanden. As pointed out by Gregory Clark, in his 2009 Review of Maddison’s famous Contours of the World Economy:

“All the numbers Maddison estimates for the years before 1820 are fictions, as real as the relics peddled around Europe in the Middle Ages (…) Just as in the Middle Ages, there was a ready market for holy relics to lend prestige to the cathedrals and shrines of Europe (…), so among modern economists there is a hunger by the credulous for numbers, any numbers however dubious their provenance, to lend support to the model of the moment. Maddison supplies that market” (Clark 2009, pp. 1156-1157).

The working paper by Bolt and Van Zanden makes significant progress in substituting some fictitious numbers (d), with indirect estimates of GDP (c), but then in discussing the results it leaves unclear which numbers are reliable, which not, thus still leaving some ground for the “market for holy relics”.


This is all the more problematic if we think that nominally all the estimates have been produced at 1990 international dollars. It is true that there is another part of the Maddison project specifically aiming at substituting 1990 purchasing power parities with 2005 ones. But this is not the point. The real point is that even 2005 PPPs would not change the fact that we are comparing economies of distant times under the assumption that differences in the cost of living remained unchanged over centuries, or even over millennia. This problem, not at all a minor neither a new one − e.g. Prados de la Escosura (2000) − is here practically ignored. One indeed may have the feeling that the authors (and Maddison before them) simply don’t care about the parities they use, de facto treating them as if they were at current prices. For example, they discuss the evidence emerging from real wages, saying that they confirm the gaps in per capita GDP: but the gaps in real wages are usually at the current parities of the time, historical parities, while those in GDP are at constant 1990 parities. If we assume, as reasonably should be, that differences in the cost of living changed over the centuries, following the different timing of economic growth, then the evidence from real wages (at current prices) may actually not confirm the GDP figures (at constant 1990 PPPs). Let’s take, for instance, China. It could be argued that differences in the cost of living, as compared to Europe, were before the industrial revolution, say in 1820, lower than in 1990, given that also the differences in per capita GDP were lower in 1820 than in 1990; hence, prices in 1820 China were relatively higher. The same is true for China when compared to Renaissance or Roman Italy (since prices in 1990 China were arguably significantly lower than prices in 1990 Italy, in comparison with the differences in the sixteenth century or in ancient times). This would mean that real GDP at current PPPs would be in 1820 even lower, as compared to Europe; or that 1820 China would have a per capita GDP remarkably lower than that of the Roman empire, maybe even lower than that of ancient Mesopotamia. Is this plausible?


Clark, G. (2009). Review essay: Angus Maddison, Contours of the world economy, 1-2030 AD: essays in macro-economic history. Journal of Economic History 69(4): 1156−1161.

Prados de la Escosura, L. (2000). International comparisons of real product, 1820–1990: an alternative data set. Explorations in Economic History 37(1):1–41.

On the many failures of (southern) Italy to catch up

Regional income inequality in Italy in the long run (1871–2001). Patterns and determinants


Emanuele FELICE ( Departament d’Economia i d’Història Econòmica, Universitat Autònoma de Barcelona


The chapter presents up-to-date estimates of Italy’s regional GDP, with the present borders, in ten-year benchmarks from 1871 to 2001, and proposes a new interpretative hypothesis based on long-lasting socio-institutional differences. The inverted U-shape of income inequality is confirmed: rising divergence until the midtwentieth century, then convergence. However, the latter was limited to the centrenorth: Italy was divided into three parts by the time regional inequality peaked, in 1951, and appears to have been split into two halves by 2001. As a consequence of the falling back of the south, from 1871 to 2001 we record s-divergence across Italy’s regions, i.e. an increase in dispersion, and sluggish ß-convergence. Geographical factors and the market size played a minor role: against them are both the evidence that most of the differences in GDP are due to employment rather than to productivity and the observed GDP patterns of many regions. The gradual converging of regional GDPs towards two equilibria instead follows social and institutional differences – in the political and economic institutions and in the levels of human and social capital – which originated in pre-unification states and did not die (but in part even increased) in postunification Italy.


Review by Anna Missiaia

This paper was distributed by NEP-HIS on 2013-12-29. The author, Emanuele Felice, engages  with the mother of all research questions in the economic history of post-Unification Italy, which is “why did the south fall behind?”. The large and widening economic gap between the north and south of Italy remains one of the “big topics” in Italian economic history and one upon which consensus is far from being reached. The paper by Felice aims at providing both new quantitative data to assess this gap and a discussion on what caused and, equally important, what did not cause the formation and persistence of the north/south divide. 

Emanuele Felice

Emanuele Felice

Let us start with the quantitative assessment. Felice provides new estimates of regional GDP at present borders. Given the long-run perspective adopted, it is necessary to make sure that we are comparing the same regions through time. This is not straightforward for Italy as it experienced several changes in its borders between 1871 and 2001. Felice collected detailed data from foreign (mostly Austrian) sources on territories that eventually become part of northern Italy. This data enables him to produce regional GDP per capita estimates for 10 year benchmarks from 1871 to 2001.

Felice then measures convergence and divergence across regions. The bottom line is that Italian regions diverged during most of the period under study. This divergence exacerbated the most between World War I (WWI) and the late 1950s. Then during the so called “Economic Miracle” of the 1960s, Italian regions experienced a degree of convergence. This convergence took place during a period of very high economic growth in the north and Felice attributed this convergence to the heavy subsidising of the southern economy. Felice also observes that while the south failed to catch up with the rich north, the northeast and centre succeeded in the task, reaching similar GDP per capita levels to those of the original Industrial Triangle towards the end of the 20th century. 

After the number crunching, Felice moves on to tackle the determinants of the income inequality. Following the path of a debate almost as old as Italy, he focuses on some well known hypothesis. The first one is that the south had a geographical disadvantage either in terms of factor endowment or market access. Felice discards the first hypothesis noting that differences between the north and south were not as marked and that the macro-areas were more different within than between them. Are a result the endowment argument is not a good candidate to explain the north-south divide. On market access, Felice notes that the south had a fairly high access to markets in the period before WWI compared to the north and the situation reversed gradually. Also, after WWI regions with a quite low access to markets (Trentino Alto-Adige and Valle d’Aosta) managed to reach high levels of GDP and regions in the south with a good access to markets performed poorly in GDP growth. 

Trentino Alto-Adig

After excluding geographical factors, Felice discusses the human element to explain divergence. He looks at human capital, social capital and institutions. At the time of unification, the south was lagging behind in both human and social capital (for a more detailed discussion and some numbers see Felice (2012)). Felice’s thesis is that economic development in the south was highly affected by its low human capital until WWII. In spite of the catch up in literacy rates after WWII, measures of social capital show that the south has never reached the level of the north. The persistence of the gap has therefore to be attributed to persistence of low levels of social capital that allowed the consolidation of poor institutional settings as well as the flourishing of organized crime.  

Reading Felice’s paper, one’s impression is that the author managed to convey several years of quantitative research into a nice narrative on how the south fell behind. He uses a mix of hard data and qualitative reasoning to guide the reader through. In particular, he takes timing of turning points (i.e in market access, state intervention or catch up in literacy rates) to explain how different elements could or could not explain the divide. He also uses the case of the northeastern regions to explain how path dependence can be overcome (the northeast had very low levels of human capital at the time of unification but managed to catch up with the rest of the north).  

For the Italian readers, Emanuele Felice, 2014, "Perche' il Sud e' rimasto indietro", Il Mulino, Bologna.

For the Italian readers, Emanuele Felice, 2014, “Perche’ il Sud e’ rimasto indietro”, Il Mulino, Bologna.

To conclude, it is often the case that this narrative argues that the south was not disadvantaged in all the factors and that different periods were driving economic growth in the country. However, it seems like it was advantaged in a given factor of growth only when that factor was not important. For example, it had a good market access before WWII, when human capital was more important; it had cough up in terms of human capital after WWII but at that time social capital started being more important. The picture that emerges from this work is that the south suffered from a mix of poor starting conditions, bad timing and unfortunate development strategies that trapped it into the gap that we still observe today.



Emanuele Felice, 2012. Regional convergence in Italy, 1891–2001: testing human and social capital, Cliometrica, Journal of Historical Economics and Econometric History, Association Française de Cliométrie (AFC), vol. 6(3), pages 267-306, October.

Industrial Location and Path Dependency during the British Industrial Revolution

The Location of the UK Cotton Textiles Industry in 1838: a Quantitative Analysis


Nicholas CRAFTS (  University of Warwick

Nikolaus WOLF ( Humboldt University


We examine the geography of cotton textiles in Britain in 1838 to test claims about why the industry came to be so heavily concentrated in Lancashire. Our analysis considers both first and second nature aspects of geography including the availability of water power, humidity, coal prices, market access and sunk costs. We show that some of these characteristics have substantial explanatory power. Moreover, we exploit the change from water to steam power to show that the persistent effect of first nature characteristics on industry location can be explained by a combination of sunk costs and agglomeration effects.


Review by Anna Missiaia

This paper was distributed by NEP-HIS on 2013-09-13. Nick Crafts and Nikolaus Wolf, who have both provided significant contributions to the literature on industrial location, engage here in the analysis of the UK cotton textile industry. In particular cotton production during the Industrial Revolution was heavily concentrated in Lancashire, the region just north of Manchester. Moreover, this concentration persisted over the 19th century. The two authors are therefore interested in explaining both the original concentration and its persistence throughout time.

The paper presents a solid statistical work. The dataset comprises 1823 cotton mills and covers 148 locations in all of the UK. To explain the employment in textiles across these locations, the authors use information on coal prices, geography, climate and access to markets. All these measures are specific to each location and region fixed effects are included to avoid the omitted variable bias. Firms are assumed to be profit maximizers in their location decisions. The factors that influence the location decisions are considered into two broad groups: the “first nature” characteristics, which are considered exogenous to earlier location choices (i.e. climate) and the “second nature” characteristics which are endogenous (i.e. access to markets). Crafts and Wolf separate these two elements because they are interested in identifying the case of location choices that eventually modify the characteristic of the location itself (in particular market access).

They reckon that access to market can be so important that the cotton industry remained in a location in spite of higher variable costs because these were outweighed by better access to markets. Another way in which past choices can affect current choices is through sunk costs: once an investment in energy production was made in one location, it could hardly be moved to another location. However, it could often be adapted to new technology. The example provided is the switch from water to steam power, during which waterwheels were adapted to steam. This allowed some location that at that point did not have a fist nature advantage, to maintain their industries through path dependence.


Hibert, Platt & Son’s cotton machines.  Illustrated London News, 23 August 1851. 

On the empirical side, the paper uses a Poisson model to estimate the expected number of cotton mills and employed persons for each location as a function of the characteristics of the locations. The main findings are that water power production and number of patents registered increase the likelihood of location; coal prices had a surprisingly weak effect; agglomeration forces had a positive effect on the number of persons employed but a negative effect of the average size of the mills, suggesting that cotton industry was organized in a network of small specialized mills. This is confirmed by anecdotal evidence on Lancashire’s cotton industry. The authors also provide several robustness checks on their data to support their claims.

The paper then moves on to discussing why Lancashire achieved such a high concentration of cotton industries. The two authors explain that the high concentration was the result of a combination of first and second nature geography. To prove this statistically, Crafts and Wolf perform a counterfactual analysis in which they replace each characteristic with the average value of the UK and then impose a 10% change in the variables to compare their effect individually. Doing so, they come up with a “conversion table” that tells us what variation of the x variable is needed to offset a 10% variation of the y variable. The main results are that the location choices were driven both by first nature characteristics such as water power and second nature characteristics such as market access. The persistence of the location is liked to sunk costs and agglomeration economies, which allow some regions to maintain their industries in spite of the original advantage being vanished.


An image of the Lewis Textile Museum in Blackburn, Lancashire.

To conclude, the contributions of this paper are several. First, it makes for the first time use of statistical techniques to explain the location of cotton industries, which were crucial during the British Industrial revolution. Doing so, it contributes to the wider debate about the determinants of the location of industries in general, proposing a methodology based on counterfactuals which allows to compare the relative strenghts of the different factors. Finally, the paper adresses the always ‘hot issue’ of path dependency in location choices, which is faced by any researcher in this particular field. The next step in the research, to which we look forward,  will be to estimate the model as a panel in order to cast more light on the persistence of location through time.