1. Divergence Big Time: Geography Matters Nicholas Crafts The Maddison Lecture, October 9, 2013

2. Modern Economic Growth Post industrial revolution era Driven by technological progress that has substantial impact on productivity growth Need appropriate institutions and policies to take advantage of the opportunity Penalty for getting it wrong gets much bigger; income divergence is not new but increases dramatically

3. Real GDP/Person ($1990GK) 18701913195019732010 Asian Tigers 394 6031010 363123313 China 530 552 448 838 8032 India 533 673 619 853 3372 Africa 648 908 889 1387 2034 W. Europe2006348845171134620889 USA2445530195611668930491 Source: The Maddison Project (2013)

4. Divergence Big Time 20 th century growth unprecedented; GDP gap much greater than ever before Clearly not unconditional β-convergence so the pure neoclassical prediction does not work Conditional β-convergence is a viable hypothesis – but what are the key conditions?

5. The Solow Model in a Globalized World Y/L = A(K/L) a Diminishing returns to capital accumulation Technology universal Factors mobile, K/L equalized across countries Beta and sigma convergence

6. 20th vs. 21 st Century “The restoration of inter-society income equality will be one of the major economic events of the century to come” (Lucas, 2000) So divergence will be superseded by convergence and normal (neoclassical) service will be resumed

7. Lucas’s Underlying Argument Obstacles to growth removed through imitation of good policies, institutions In a globalized world, capital mobility and financial liberalization relax the savings constraint Speed of catch-up growth will increase markedly and K/L and TFP gaps will be rapidly reduced

8. Why Might the Solow Model Be Wrong? TFP is not the same across all countries because either efficiency or technology is not universal Obstacles to factor mobility prevent equalization of K/L Geography, institutions or economic policies differ

9. The North/Acemoglu View Institutions which affect investment and innovation are the underlying determinants of economic performance Institutions are formal and informal constraints that structure behaviour Property rights are the key to high incomes today and thus to divergence over time Institutions are persistent

10. Rule of Law Scores (-2.5 to +2.5) Kaufmann et al. (2013) 1996201219962012 Brazil-0.33-0.11Netherlands1.651.84 China-0.43-0.49Nigeria-1.26-1.18 India0.26-0.10Singapore1.281.77 Russia-0.87-0.82USA1.451.60

11. Institutions and Growth Important but surely not all that matters Institutional quality may not be well measured but growth regressions do not suggest it dominates recent differences in performance Policy plays a part and so too does geography

12. Growth of Real GDP/Person, 1960-2000 (% per year) Resource -Scarce &Coastal Resource- Scarce & Landlocked Resource -Rich Africa0.50 (33) -0.36 (33) 0.29 (33) Other Developing 3.79 (88) 1.40 (1) 2.89 (11) Source: Collier (2007); numbers in parentheses refer to percentages of population in each category

13. Divergence Big Time Persistent and widening income gaps characterize modern economic growth era Institutional/policy failures matter much more when growth opportunities increase BUT there is a strong spatial correlation of development outcomes Does this mean that geography undermines the mainstream assumption of a ‘level playing field’ for development ?

14. Geography and Income Geography may preclude full convergence Natural resources and market access; 1 st and 2 nd nature aspects Direct and indirect effects Indirect effects may work through institutions, e.g. ‘natural resource curse’

15. New Economic Geography: Key Ideas Agglomeration Benefits Market Potential Trade Costs Globalization may imply divergence

16. Transport Costs and the Location of Economic Activity Very High or Very Low: everything dispersed Intermediate: centralization of industry based on location in larger market with increasing returns and external economies of scale So New Economic Geography says that even with perfect institutions everywhere integration of markets may lead to divergence

17. Globalization and the Inequality of Nations (Krugman & Venables, 1995) Manufacturing goods are subject to increasing returns and are used both as final and as intermediate goods As transport costs fall, self-reinforcing advantage of larger market leads to country-specific external economies of scale and lower costs for manufacturing in core relative to periphery Eventually, if trade costs fall enough and/or wages in the core rise enough, manufacturing returns to (parts of) the periphery

18. Market Potential Market access matters for industrial location decisions; operationalized by ‘market potential’ which is distance (transport costs) -weighted GDP MP i = ∑GDP j d ij γ If data permit, can estimate γ using gravity model; traditionally assumed that γ = -1

19. Late 20 th Century Empirics (Redding & Venables, 2004) There is a high correlation between location and income so, following Acemoglu’s strategy, this also might explain divergence big time Market potential elasticity around 0.3 Location effects largely robust to including institutional quality

21. A Prediction If Zimbabwe were re-located to Hungary, real GDP per person would rise by 80 per cent Redding & Venables (2004)

22. Changes in 19 th -Century Economic Geography Industrialization and de-industrialization in globalizing world Concentration of world manufacturing production and, even more so, exports Changes in location influenced by transport costs; manufacturing cities proliferated in Europe and North America; mass production and mass distribution

23. Source: Harley (1988) Real Cost of Ocean Shipping (1910=100)

24. Shares of World Manufacturing Output (%) 1750183018801913 Europe23.234.261.356.6 UK 1.9 9.522.913.6 USA 0.1 2.414.732.0 China32.829.812.5 3.6 India24.517.6 2.8 1.4 Source: Bairoch (1982)

25. Historiography (Rodrik, 2013) The explanations for 19 th century continental divergence are as follows: Imperialist exploitation (Mandel, 1975) Institutions (Acemoglu et al., 2002) Dutch Disease (Williamson, 2011) Directed technical change (Allen, 2012) But could NEG core-periphery have anything to do with it?

26. Market Access Then and Now (Redding & Venables, 2002; Liu & Meissner, 2013) 19101995 USA100North America100 UK88Western Europe92 India31South Asia40 Indonesia13Latin America35 Argentina7Africa34

27. Market Potential and GDP 100 Years Ago Has similar impact on real GDP/person to late 20 th century with elasticity of about 0.3 in whole world countries sample (Liu & Meissner, 2013) or in European regions sample (Caruana-Galizia, 2013) Core Europe has much greater market potential than peripheral Asia (and Southern Europe) at least in late 19 th century Liu & Meissner’s estimates suggest the following quote may not be entirely accurate

28. A Quotation “No deus ex machina translates endowments into political outcomes. If that were so, Argentina would be as rich as the United States” North et al. (2000)

29. Location of Manufacturing The ‘manufacturing belt’ in the United States is locked into place by market potential which interacts with scale and linkage effects (Klein & Crafts, 2012) Catalonia industrializes to a much greater extent than the rest of Spain as a result of favourable market size (Roses, 2003)

30. Location of British Cotton Textiles in 1838 (Crafts & Wolf, 2013) In 1850, UK had 69% world spindles (58% in 1900) In 1850 Lancashire had 66% of UK spindles (79% in 1903) and about 46% of world spindles (same in 1903) Lancashire is 1.3% of UK land mass and 0.002% of world land mass Cotton is classic example of core-periphery in the ‘1 st Unbundling’

31. 31

32. Why Lancashire?: Traditional “ The original advantages of Lancashire comprised its poverty, its climate, its water supply, its textile tradition and its mechanical inventions. The acquired advantages included its supply of coal [!], machinery and labour, its access to the markets of Liverpool and Manchester, its low transport cost, and its auxiliary industries ” (Farnie, 1979) 32

33. 33

34. 34

35. 35

36. 36

37. Why Lancashire?: Econometrics Celtic fringes have better 1 st -nature attributes apart from cheap coal; it’s definitely not humidity Adequate source of power is ‘necessary condition’ but market access has powerful effect Cotton industry concentrates on a subset of its original locations to which it becomes ‘locked in’

38. Results 38 PoissonCotton millsCotton employment Coefficient (robust SE) LnHumidity-34.917 (24.551)-27.197 (17.963) LnRugged1.249 (0.462)***0.580 (0.375) Lnq950.314 (0.035)***0.255 (0.043)*** LnCoalprice-0.426 (0.745)-1.081 (0.651)* LnDistweight-3.823 (0.962)***-3.292 (1.302)** LnMP014.405 (1.073)***4.561 (0.845)*** LnInnovations0.788 (0.032)***0.804 (0.030)*** Constant + Reg DumYes # of Obs148 Pseudo R20.7620.760 Robust SE adjusted for 10 regional clusters, *, **, *** ind. sign. at 10, 5, 1%

39. Lancashire Textiles and Globalization (Leunig, 2005) Lancashire a high wage industry: 6 x India and Japan in 1910 But continued to dominate world trade (60% world market share in cottons in 1910) Unit costs lower than India or Japan even before adjusting for output quality Lancashire flourished because of agglomeration benefits..... its productivity exceeded other British locations by 33%

40. Path Dependence Economic historians like the idea – ‘history matters’ – so less optimistic than neoclassical economists about future convergence Technological historians think of QWERTY The NIEH tradition sees institutions as the ‘carriers of history’ The NEG approach highlights 2 nd -nature geography as a source of potential lock-in

41. Death of Distance? Would have truly dramatic effect on world distribution of economic activity and income But “greatly exaggerated” ICT enables some things to go to the periphery (‘the 2 nd unbundling’) but enhances the strengths of the core at the same time Like steam, ICT rearranges geography but doesn’t abolish it

42. Concluding Questions Is the now-dominant institutional explanation for “divergence big time” over-sold ? Does market potential also play a key role – perhaps after a threshold level of institutional quality is reached ? Can we construct a new geography-based narrative to (partly) explain industrialization and de-industrialization since ‘globalization began’ ?