Technological shifts and convergence in a European perspective since 1950.

Transcription

1 1 Technological shifts and convergence in a European perspective since Lennart Schön Department of Economic History Lund University Lennart.Schon@ekh.lu.se Paper for the 8 th Annual Conference on European Integration by the Swedish Network for European Studies in Economics and Business (SNEE) at Grand Hotel, Mölle, May 16-19, Introduction This is a discussion paper on some aspects of OECD and European growth and convergence from 1950 to the early 2000, with some perspectives also on the pre- First World War globalization era. The discussion will address the role of technological shifts (industrial revolutions) and the perspectives of traditional and new growth theory on convergence/divergence and in particular it will try to develop a measure of these forces as a continuous process. The analysis draws mainly upon data that have been available internationally (Maddison 1995, Williamson 1995, Groningen Growth and Development Centre 2006). It should be emphasised from the beginning, though, that theory has run ahead of measurement, especially with regard to examining differences within the advanced countries. (Crafts/Toniolo, 1996 p. 16) Three perspectives on growth Since the early 1990s, perspectives on growth from two different branches of economic theory have had a decisive impact upon economic history one perspective is focussed on the mechanisms of convergence and the other on growth as an endogenous process. These two strands of thought give largely very different perspectives on growth and a different understanding of the process. They will be compared and discussed in relation to a structural analysis of growth that has been followed in research at Economic History in Lund. Convergence in history Convergence of income levels and relative prices is a main characteristic in analyses of market integration or globalisation in history - notably by Williamson/O Rourke (1994) and Williamson (1995,1996). Their analysis is based upon standard neoclassical Economics with a Solow-model with diminishing returns to the accumulation of factors of production. Growth is not the direct object of study, but convergence and growth run largely parallel. There are mainly two sources of growth. One is technical change that is exogenous to the model. The other is the

2 2 reallocation of resources to more productive ends. That is the central mechanism in this analysis. According to Jeffrey Williamson the interaction between labour, capital and commodity markets should be in focus of studies in economic growth. Economic historians should attack these issues first before elevating international technological transfer to the status of prime mover, a thesis so ably argued by Gerschenkron that it has dominated the convergence debate ever since. (1995, p 162) When markets widen through integration of different national economies (in the process of globalisation), relative prices will change in a systematic way. In a given economy, prices of factors that prior to integration were relatively abundant will rise and consequently prices of scarce factors will fall. Thus, globalisation will increase the returns to labour in labour-abundant economies and to capital in capital-abundant economies. Through the flow of goods, labour and capital, prices and relative income levels will converge within a globalised economy. This is of course the mechanism of factor price equalisation put forward by Eli Heckscher and Bertil Ohlin, based upon experiences in the late 19 th century and early 20 th century. The First World War and the breakdown of international institutions put very effectively an end to that process of globalisation and convergence that was to recur after the Second World War. Thus, modern history has witnessed two periods of globalisation and convergence according to Williamson et.al. from 1870 up to the First World War and the period after the Second World War. Endogenous growth The new growth theory that was launched within Economics in the 1980s (notably Romer 1986, 1990; Lucas 1988) gives a very different perspective on the growth process and its outcome. Through the accumulation of broad capital that includes knowledge, human capital and innovations, growth was made endogenous to the process. Thus, knowledge was produced as any other investment good by the calculated use of labour and capital. The inclusion of knowledge as a key factor and asset had further consequences. While other factors had diminishing returns and were consumed in the production process, knowledge grew when it was used. Furthermore, through externalities including spillovers and complementarities the accumulation of knowledge was endowed with increasing returns. New complementarities arose through the advance of knowledge in different areas and new complementarities were created between human capital and technology. Convergence between economies at different income levels is not the logical outcome of a model with increasing returns, externalities and complementarities. Rather it will predict divergence since initial differences are enlarged in an endogenously determined growth process with such characteristics. Divergence is brought even further due to agglomeration forces (new economic geography) and these forces are rather strengthened through market integration. There are of course modifications and qualifications of the basic models that diminish the gulf between them. The concept of social capability (Abramovitz 1986) emphasises that there are decisive socially determined differences in the ability to integrate productivity-increasing methods and to transform structures and thus differences in the ability to join convergence clubs at certain historical points of time.

3 3 Furthermore, in the augmented Solow-model the concept of capital is broadened to include knowledge. (Mankiw et.al. 1992) Basically, however, there are two very different propositions of the normal outcome of the growth process one of convergence and one of divergence. Analysis of structural change Analysis of historical growth performance may give ground for another position Convergence and divergence are components in that process in a quite logical manner; i.e. different periods are characterised by convergence and divergence respectively not because of shortcomings in the external conditions but because of qualities inherent in growth. Furthermore, this variation between convergence and divergence in growth rates and income levels may have followed a certain regularity that can be expressed in a model of the historical process that combines aspects from both bodies of standard economic theory. (Cf. a structural model presented in e.g. Krantz/Schön 1983, Schön 1994, 1998 and 2000). Inventions may appear and knowledge may accumulate at a rather constant pace, but the impacts upon the economy vary very much over time due to properties within the economic system such as complementarity and externality. Radical innovations that create important and far-reaching new complementarities as the steam-engine, the railway, the electrical motor and the combustion engine, the motor-car, the microprocessor, the Internet are rare and their diffusion spread over long periods. In the 1990s the ICT-revolution attracted increasing attention and such radical innovations were named General Purpose Technologies by some economists. (Bresnahan&Trajtenberg 1995) and applied also by some economic historians that had been sceptical to the concept of industrial revolutions (e.g. Crafts 2002). The importance of the phenomenon of new complementarities created around innovations was of course recognized much earlier by many economic historians. With a Swedish Schumpeterian concept the complementarities around innovations form development blocks that are at the centre of the growth process. (Dahmén 1950, 1988) The creation of new complementarities within a development block changes the relative price structure as does market integration but with another logic. In central areas of innovation relative output prices will generally fall that intensifies competition against old combinations of production factors. The expansion of activities in the innovating areas will, however, increase demand for inputs of goods and services and for complementary production that are supplied less elastically and for these, prices will rise. Within the development block remuneration to the factors of production will increase either as a consequence of increased productivity (supply push) or as a consequence of increased prices (demand push). Within old blocks remuneration will decrease. Furthermore, complementarities around radical innovations appear suddenly and unexpectedly, despite the fact that the breakthrough has been preceded by a long period of innovative activity. The wider repercussions of the innovation that forms the development block make a new turn of events. Such was the case with the electronic revolution (the third industrial revolution) following the advent of the microprocessor or with the second industrial revolution following upon the accumulation of engineering and scientific knowledge at the end of the 19 th century.

4 4 In such periods of more rapid transformation, regions and nations react differently. Evidently, there are leading regions and nations, since innovations appear in specific circumstances and from the existence of geographically confined complementarities and externalities. Diffusion is more rapid to regions and nations that are favoured by new demands due to their resource endowments, their institutional characteristics and their social capability. For the same reasons, the new turn of growth direction is unfavourable to other regions and nations. They may be firmly attached to old combinations and/or have endowments that are less advantageous under new circumstances. Under those conditions growth rates will diverge. Furthermore, it is reasonable that rich economies well endowed with knowledge and acting at the production/consumption frontier will be better situated to take the lead thus divergence will follow. Over time however (due to investments) competencies, infrastructures and institutions will be more generally adapted to the new complementarities. Hence the development blocks will be more widely diffused. Since further innovate activity (i.e. economic use of potentially available new combinations from the accumulation of knowledge) is restricted by the structure of complementarities and interests created (i.e. by path-dependency), the accumulation of broad capital will be captured within the confinements of diminishing returns. Further growth will be more determined by diffusion of the new technology and by the working of the market mechanism. In that diffusion process, the favourable position of the leaders is undermined while laggards will improve their position. Thus, divergence in growth rates and income levels will turn into convergence (among countries involved in this process). Diminishing returns will however shift relative profitability between established and emerging complementarities that pave the way for crisis and new structural transformation. Thus, from this perspective the growth process may be periodised in two parts. The first period is characterised by radical transformation of structures when development in a geographical context is uneven with growth accelerating in small nuclei. The second period is characterised by rationalisation when gaps are being levelled and the economy is made more homogenous with growth accelerating (to a certain point) in a wider context. Hence, convergence and divergence put their imprint on different phases and the traditional and the new growth theory evolve around properties that have characterised growth alternately. Two major discontinuities or technological shifts are indicated by the concepts of the Second and Third Industrial Revolutions both witnessing the early breakthroughs of innovations prone to become General Purpose Technologies. Thus, one may see the 1890s as a decisive period of the Second Industrial Revolution and the breakthrough of the modern industrial society. In the centre of development, there was a diffusion of new motors, electrical or combustion engines, with a widening scope for mechanisation within industry. In the interwar period, and still more marked after the Second World War, new development blocks were created around a widening of electrification and the diffusion of automobiles. In both cases, there were waves of investments around the development of an infrastructure from the basic innovations of the preceding cycle. In case of electricity, a much broader spectrum of applications was created ranging from the processing of electro steel to small motors in handicraft and household.

5 5 From the mid-1970s the appearance of the microprocessor gave the spark to the technological shift of the Third Industrial Revolution and electronics became the centre in new development blocks. The use of knowledge and information in production of goods and services advanced with leaps, marking new directions of growth. There is still another long term pattern the First and Second Industrial Revolutions were followed by a wider diffusion of blocks around infrastructural development and institutional adaptation in the mid-19 th and mid-20 th centuries. One may hypothesise that forces of divergence are relatively strong in periods denounced as industrial revolutions while convergence would be more emphasised not only in the diffusion process of the technology in production but particularly in the ensuing period of infrastructural development. Phases of growth In a survey of post-war European growth Nicholas Crafts and Gianni Toniolo stated: It is useful to think in terms of epochs of growth. (1996, p. 32) There is a wellestablished periodisation of European economic development since 1870 that is as follows: the era of the classical Gold Standard the period of wars and reconstruction the post-war Golden Age of high growth rates deceleration and restructuring into a new economy This periodisation appears clearly in figure 1 and table 1 that depicts a measure of an aggregated European GDP per capita calculated from 14 countries and with a single PPP-benchmark in Although the construction of GDP series in individual European countries have progressed immensely in the last decades, there are still a number of problems and difficulties in these long-term cross country constructions that are far from solved and are being addressed by a number of European economic historians. (See for instance Prados 2000) In the present context, however, focus will be on some rather well established trends. Even if this broad periodisation fits rather well to overall trends in growth, it may obscure trend breaks that are vital for a discussion of the dynamics of growth.

6 6 Figure 1. GDP per capita in 14 European countries Gheary-Khamis Dollars Note and sources: See table 1. Table 1. Growth rates in GDP per capita in Europe. Average of 14 countries. Period Annual growth rate of GDP per capita , , , , ,8 Note: The countries are Austria, Belgium, Denmark, Finland, France, Germany, Italy, the Netherlands, Norway, Portugal, Spain, Sweden, Switzerland and the UK. Sources: Maddison (1995), (2006). The era of the classical Gold Standard In their work on globalisation, Williamson et.al. treat the decades of the Gold Standard from 1870 to the First World War as a homogenous period of convergence. The flows of capital and labour were particularly important in the process of global factor price equalisation. That equalisation occurred both within Europe and in the Atlantic Economy.

7 7 Overall convergence was, however, very much a result of growth and factor price development in the Scandinavian countries and the United States, which is apparent from Williamson s data and that is clearly shown in table 2. According to figures from Williamson and Maddison, the relative change of factor prices in Sweden was extraordinary but the change was almost as strong in Denmark and Norway. (On the importance of capital imports in Sweden, see Schön 1989, 1997) Increases in real wages clearly outstripped a fairly rapid growth in GDP per capita. In the highwage economy of the United States development was the opposite. Growth of wages was sluggish while GDP grew quite rapidly. In low-wage Latin Europe (the Iberian and Apennine peninsulas) GDP growth was weak but wage growth was even weaker. Thus, the Scandinavian countries clearly formed a convergence club at the end of the 19 th century that joined in modern economic growth, while Latin Europe generally did not. Table 2. Annual growth in indexes of real wages and in GDP per capita in European countries and the United States Country Annual growth of real wages Annual growth of GDP per capita Sweden 2,8 1,7 Denmark and Norway 2,6 1,3 France, Germany and UK 1,1 1,2 USA 1,1 1,6 Italy, Portugal and Spain 0,6 1,0 Sources: Wages from Williamson (1995); GDP, see table 1. However, the classical Gold Standard era was not a very homogenous period. There was a decisive turning point around In the period growth rates were much more even among countries also in a Scandinavian/Latin European comparison. (Table 3) That was a period characterised by the market integration that took place after the introduction of new institutions at mid-19 th century as well as the creation of a new infrastructure of railways, steam-ships and telegraph cables. It was an expansion built upon innovations of the first industrial revolution and the ensuing infrastructure development and upon the complementarity between an industrial centre and a largely raw-material producing periphery. From 1890 development changed character, however. While growth of GDP and wages accelerated in Scandinavia (as in the United States), it decelerated in Latin Europe (and stagnated in the United Kingdom). Development diverged both in the European periphery and in the Atlantic core countries. There was divergence also within Latin Europe. The economies of Spain and Portugal stagnated almost completely, while there was a strong industrial spurt in Italy from around the turn of the century that served to increase the diversity between northern and southern Italy even more.

8 8 Table 3. Annual growth rates of GDP per capita in Europe and the United States Land Scandinavia 0,9 2,0 United Kingdom 0,9 0,9 Spain, Portugal 0,9 0,7 Italy 2,1 Continental Europe 1,0 1,3 United States 1,9 2,2 Note: Continental Europe = Austria, Belgium, France, Germany, the Netherlands and Switzerland. Sources: see table 1. The diverging growth rates from 1890 reflect different reactions to the new basis that was created for growth with the so-called second industrial revolution. Manufacturing industry became more sophisticated. Electricity and the combustion engine transformed the energy system, machinery became more complex, a new chemical industry arose all this gave a more central role to engineering, to education, to research and science. Abramovitz and David (1973) stated that this decade meant a new basis for growth in the United States. In the 19 th century, and particularly , American growth could by and large be explained by the accumulation of the traditional factors of production of land, labour and capital. Capital accumulation was of particular importance in enabling a growing labour force to take into use the abundant resources of land. But from 1890 the importance of that accumulation diminished. From this point of time human capital became all the more important to the increase in growth rates and to a sustained increase in the productivity of the traditional factors (i.e. in total factor productivity). Thus, the centre of industrialisation shifted from the building of railways, factories, and residential constructions to the development of competent labour and management. The returns from investments in knowledge and education increased. A similar new turn was recently pointed out by Goldin and Katz (1996) that found a new complementarity between capital and skills arising in the US in combination with the first wave of electrification. At the other extreme, United Kingdom failed to accelerate in the second industrial revolution. According to Crafts (1985) British capitalists behaved quite rationally when they accommodated to comparative advantages established from the first industrial revolution with comparatively high productivity within low-skilled manufacturing industries. The failure to adopt new technologies has also been attributed to the high degree of British integration in the world markets, particularly in the financial market that allegedly gave weaker linkages between risk capitalists and British industry than in e.g. Germany or Sweden. (Kennedy 1987) In continental Europe there was a western group consisting of France, Belgium and the Netherlands with close-to-uk growth rates and only weak acceleration from the 1890s and a central group of Germany, Austria and Switzerland with close-to- Scandinavian growth from Thus, we find a sphere of western Europe from UK down to Spain and Portugal with weak performance in the second industrial revolution , and a strong central axis from Scandinavia (particularly

9 9 Sweden with the new engineering industries) down to northern Italy with Sweden and Italy matching US growth rates. This paper will not go further into the background of the differing European reactions to this industrial revolution. It will only emphasise that the classical Gold Standard era holds two distinct periods that may not show up in aggregate growth rates. It also emphasises that the reactions to the second industrial revolution were strongly diverging among European countries, creating scope for a new convergence at a higher level than that of a convergence that was to be delayed due to war and disintegration from the 1910s to the 1950s. The two post-war periods It is widely recognised that the 1970s saw a structural break in the post-war period. In the mid-1970s there was not only a shift from high to low growth rates. There was also a shift from strong convergence to weak convergence or even divergence, at least within the industrial world. This was clearly so in the relation between the United States and Europe but there was also diverging trends at a regional basis. (Crafts/Toniolo, p 4 ff) Divergence was of course particularly strong in a wider European setting with the catastrophic development in Eastern Europe. While growth rates were quite acceptable in communist Europe in the post-war era of industrial growth up to the 1970s, the new conditions of the information revolution meant a catastrophe in terms of growth. Table 4. Growth rates in Western and Eastern Europe and in the USA Western Europe Eastern Europe USA Source: Groningen Growth and Development Centre, (2006). Lower growth rates and weaker convergence are problematic to theories of endogenous growth and of convergence respectively. The shift to weaker convergence or divergence is problematic to a theory on globalisation and convergence since there was no breakdown of international integration comparable to that of the interwar period. On the contrary, globalisation rather gained strength. The shift is not emphasised by Williamson who treats the period from 1950 to the late 1980s as a homogenous one from the perspective of convergence and globalisation. (1996) It ought to be a problem to the new growth theory that growth decelerated at a time when knowledge became more intensively used in most production processes. (Cf. Jones critique, Jones 1995, Josephson 2005) There is of course a reply to the latter objection by referring to the extraordinary character of the 1950s and 1960s in Europe. When compared with long run growth , the post-1973 record in Western Europe is not very disappointing. Eichengreen (1996) formulated an explanation to the trend-breaks in growth rates and convergence in Western Europe in the 1970s. The high European growth rates as well the strong convergence were driven by a double catch-up process that formed post-war European institutions but eventually, in the 1970s, the process eroded its

10 10 own basis. The double catch-up was a result of the gaps that had opened up vis-à-vis both the United States and Europe s own pre-war trend in the two decades of depression and war since (p. 38) With such possibilities to catch-up, institutions and social arrangements were constructed conducive to rapid and foreseeable growth; i.e. wage moderation enabled high levels of investment that were directed into clearly defined areas. When catch-up possibilities dwindled, the institutional framework of wage moderation broke down. These conclusions on the European post-war development can be put in a more general framework of the dynamics of uneven development. When new complementarities and development blocks have arisen in regions that have forged ahead, the divergence gives rise to differences in productivity and profitability that become the driving force both in market integration and in a catch-up process that produces convergence but eventually erode its basis. The long-run dynamics disappear either if new complementarities or development blocks fail to arise in the ensuing crisis (no imbalances created) or if market integration fails (no catch-up follows). In the history of European growth the first failure has not occurred as yet while the second failure marred the 1910s and the interwar period. The double catch up may be phrased in another way, in the framework of structural analysis. It was not only a catch-up in relation to USA and to some mystical growth trend but a catch-up in relation to two sets of complementarities that had arisen in earlier decades and that Europe to a large extent had been unable to implement and diffuse. One set was the new technologies that arose in the second industrial revolution from the 1890s including sophisticate engineering and chemistry and above all tayloristic methods of mass-production including the use of new motive power. These had developed up to 1910 primarily in the US and Germany and to some extent in Sweden, and could have been diffused more widely in Europe in the 1910s and 1920s but for the First World War and the following institutional shortcomings. The second set of complementarities involved a proper infrastructure built upon electrical motors and combustion engines as well as social arrangements for the industrial society widening participation in both economic and political terms. In this transformation, Sweden was a progressive country with new trends from the 1930s, whereas most of Europe was drawn into another long and violent conflict. From the early 1950s institutions were created both internationally and within Europe that facilitated the double catch-up process a diffusion that created strong growth and convergence within Europe as well as globally but with diminishing returns and profit squeezes that eventually ended in crisis in the 1970s. From that time the electronic revolution and the IT-revolution grew in momentum with an end to the strong convergence and with new elements of divergence infused. Rates of convergence in OECD and in Europe In this section a more systematic analysis of the convergence process within OECD and within Europe will be performed following the Barro and Sala-i-Martin (1992) formulation of the estimation of beta convergence as expressed in equation 1.

11 11 1/T * ln(yfinal/yinitial) = alfa -((1-e^beta*T)/T) * ln(yinitial) + E (1) where Y is GDP per capita in the respective sample countries and T is the number of years. Furthermore, convergence in Europe is estimated in two groups, one for Western Europe and one for all Europe. Of OECD countries Turkey is excluded as an outlier, while all Europe includes the former communist economies and Turkey. In table 5 the rates of convergence are estimated both for the whole period and for the two sub-periods and Thus, for the two periods and the convergence rates are regressed upon the initial year 1950, while the estimate of convergence between 1975 and 2004 is regressed upon the initial year 1975 as base year. Table 5. Annual convergence rates in OECD and in Europe , fixed base years b P-value b P-value b P-alue OECD (pre 1994) 1,37 0,000 2,42 0,000 1,61 0,004 Western Europe 1,30 0,000 1,96 0,000 2,02 0,001 All Europe 0,15 0,512 0,90 0,001-0,70 0,082 N: OECD 23; Western Europe 17; All Europe 26. The result in table 5 is in one sense very clear. From the first two columns one can conclude that convergence within OECD and Western Europe was strong until 1975 with beta-coefficients above or around 2 percent but that convergence was considerably weaker when 2004 is regressed upon In both cases, though, convergence is highly significant statistically. For All Europe there is a rather weak convergence (slightly below 1 percent) until 1975 that almost disappears with the longer time period. These results all indicate that convergence, when estimated from the 1950 basis, came to a halt or was considerably weakened in the period after This conclusion is further strengthened by the measure of sigma convergence (the variance of income per capita within the group of countries, measured annually). That is demonstrated in figure 2 by the sigma convergence of the OECD countries This measure indicates a strong convergence from 1950 up to the early 1970s when income variance was roughly halved and then there were fluctuations of about 7-8 years (Juglars) on a stable level all through the 1980s, 1990s and early 2000s. Graph 2 about here When the estimation of beta convergence is made separately for the period , and consequently with 1975 as basis, the result is rather different, though, as is shown in the third column of table 5. In the OECD group of countries there is still convergence and it is only slightly weaker than prior to 1975, while convergence within Western Europe becomes even stronger than in the post-war decades. These seemingly ambiguous results, for Western Europe in particular, is a version of the

12 12 traditional index problem with the shifting of base years in the estimation of beta convergence. It can easily be demonstrated by the relation between income levels and growth rates in Germany and Great Britain respectively. Germany was on a very low level and Britain on a high European level in 1950 but to be surpassed by Germany until In the following decades however Britain once again overtook Germany and was on a substantially higher level in Thus, in both sub periods there was convergence in relation to the base years 1950 and 1975, while the last period meant a reversal of trends in relation to the basis of 1950 and thus weakened convergence in the perspective. Furthermore, the conflicting results also demonstrate the differences and limitations of the concepts of beta and sigma convergence. While sigma only measures the variance each year irrespective of the position of the particular countries any other year, the beta just measures the latter, namely the change of the position of each country in relation to its prior base year position. (Cf. Quah s critique of these concepts, 1993). Thus, if there is beta convergence, there may or may not be sigma convergence, while if there is sigma convergence there must also be beta convergence of some magnitude. The estimation of beta convergence is usually made over longer periods, regressing long term growth rates upon the cross section of income levels of a base year. Such estimates leave, however, the historical process of convergence out of sight. The sigma measure on the other hand may be made annually but is an incomplete measure of convergence. It is possible, however, to estimate the rate of beta convergence within any time period, e.g. from one year to the next and with a consequent annual shift of the base year. The resulting estimates of consecutive beta coefficients will then form a time series of a convergence process. The merit of such a series is e.g. that it may be the starting point for further analysis of impacts upon the process. There are draw backs as well of course. One is that the statistical significance for each estimate will be lower (see the confidence intervals in the graphs of the time series) since there is more room for influences of a short term character that are unrelated to differences in income levels. In the present paper three time series of beta convergence with annually shifted base years have been estimated for the OECD countries, for Western Europe and for All Europe. They are reported in the graphs 3-5 with confidence intervals. Graphs 3-5 about here The graphs of the beta convergence time series show considerable annual fluctuations but on a rather high positive level within OECD and Western Europe in the post-war period but falling to a low positive level from the early 1970s to the early 1980s. From then onwards there are great fluctuations between recurrent divergence and convergence. The All European pattern (including Eastern Europe and Turkey) is somewhat similar but on a lower level of convergence up to the mid- 1970s and with a shift to consecutive divergence ending in the catastrophic development of the early 1990s followed by indication of more persistent convergence at the end of the period.

13 13 The confidence intervals show clearly that most estimates of beta convergence at an annual basis are statistically weak. Quite naturally that goes for estimates of coefficients close to zero. One way of evaluating the result is however to investigate the correspondence between the process of convergence/divergence and other indicators of economic behaviour in the major economies. One such indicator is the investment ratio. It is reasonable to assume that convergence has a negative relation to investment growth in the rich US economy, while it has a positive relation to investment growth in European economies. Thus a model is formulated (equation 2) where the convergence is determined by the investment ratios in the US and Western Europe respectively. Convergence = a C + b1 USinvrat + b2 WEinvrat(-1)+AR(1) (2) Equations and graphs 6-7 of estimates here The model gives a good fit to the convergence process within the OECD, in particular, and in Western Europe (only weak significance though for US investments in the latter case). The signs are also as expected investment increases in the US contributes to divergence and investment increases in Western Europe to convergence. One can also notice that there is a time lag from investments to convergence in Europe but an immediate relation between investments in the US and divergence, maybe indicating that growth was more investment driven in catching up countries while in the leading country growth stimulated investments. (Note: The convergence series is stationary, while there is a weak trend in the investment series. An estimate on differentiated series gives the same result, however, but with a lower statistical significance.) The model is furthermore especially strong in the period with high significance for all variables. In the last two decades variations and deviations from the estimates clearly increase. Not too surprisingly the model does not give very satisfactory result for the all European process and certainly not for a period encompassing the divergence of the early 1990s. As a general remark, however, it is fair to say that the regression results support the economic significance of the annual estimates of the beta convergence process. The averages of the annual convergence rates with annually shifted base years (table 6) give a somewhat different picture than the convergence estimates with fixed base years, in table 5. For the OECD countries and Western Europe the long term convergence rates become higher. Furthermore, the periodic pattern before and after 1975 becomes even more distinct for the OECD and Western Europe increasing the convergence rates before 1975 and decreasing them after 1975, compared to the fixed base year computations in table 5. Table 6. Annual convergence rates in OECD and in Europe , annually shifted base year OECD (pre 1994) 1,91 3,10 0,88 Western Europe 1,97 2,42 1,57 All Europe -0,06 0,88-0,89 Note: Since the periodic rates are means of annually estimated beta coefficients, no P-values are given.

14 14 The convergence process is more clearly visualized in graphs with the annual convergence rates calculated into a time series from an initial index number of 100, according to the formula: Conv (t) = Conv (t-1) * (1-beta(t)) Graph 8 about here. The All European pattern is very pregnant weak convergence up to mid-1970s and then accelerating divergence until 1992, followed by stagnation and recovery. For OECD and Western Europe the accumulated rates show the acceleration of convergence in the early 1960s and the slow down from 1972 up to the first years of the 1980s. Then two divergences followed, in the first half of the 1980s and 1990s respectively, both giving way markedly to convergence. One can notice that these latter processes fit to the business cycle pattern in relation both to economic activity and to major technological shifts. From low activity in the early 1980s, a business cycle upswing saw the diffusion of the PC and the mobile phone as components of new developments blocks and organisational change, while the upswing from the crisis in the early 1990s was connected to the diffusion of Internet and a more general use of ICT. One can also notice that the shifts to convergence were decisively stronger in Western Europe than within the OECD group in these decades. The variation between divergence/convergence over these decades indicates that the early phase of these upswings had a positive relation between income levels and growth rates while the ensuing diffusion processes saw the recurrence of convergence. Confronting the series based on annual estimates of the beta and sigma coefficients may give further insights into the convergence process. In graph 10, a ratio between the two index series of convergence among the OECD countries shows in what period convergence is driven mainly by decreased variance (sigma) and in what period it is mainly driven by growth rates relative to income levels. Clearly in the 1960s and in the period the latter force contributed much to convergence. Furthermore, the stronger convergence in Western Europe compared to OECD since the 1980s indicates that this was mainly a force in Europe, within a more invariant OECD distribution. Graph 9 about here Thus, both sigma and beta coefficients indicate a slow-down of convergence after 1975 as well as cyclical shifts between divergence and convergence - as noted above there are indications of divergence in the early phase of upswings in the 1980s and 1990s followed by convergence over the later phase of the swings, particularly strongly in the 1980s. These spurts in convergence were more pronounced within Europe than within OECD. Furthermore, there are both technological breakthroughs (the PC and the ICT) and institutional changes with a deepened European integration and single market reforms. One explanation to the pattern could be that rich and knowledge intensive countries were favoured in the early phase of each breakthrough (endogenous growth) but that convergence followed quite rapidly with market integration substantially reducing obstacles to diffusion.

15 15 Convergence or divergence is basically driven by the ability in different economies to adapt to new conditions in technology and markets. One may assume that convergence means a relatively stronger increase in those sectors that are intensive in using knowledge and new technology and that the rate of increase is negatively related to the relative size of that sector. Thus one may measure the structural convergence in modern economies from the relative size and growth rates of knowledge intensive industries (including services). The Barro/Sala-y-Martin equation is accordingly reformulated into 1/T * ln(sfinal/sinitial) = alfa -((1-e^beta*T)/T) * ln(sinitial) + E (3) where S is the share of knowledge intensive production of goods and services in each economy. The process of structural beta convergence has been estimated from the set of 60 industries table at the Groningen data base, including 21 OECD countries for the period The structural process is depicted in an indexed annual series in graph 11. The result is quite unambiguous. In the first half of the 1980s the index fluctuated with some divergence in the trough years of followed by a pronounced and strong structural convergence during the second half of the 1980s. Around 1990 convergence came to a halt and structural divergence followed in the 1990s. Thus, from the 1990s knowledge intensive industries grew in relative size in countries with relatively large such sectors. This may indicate that the ICT revolution, which really gained strength in the 1990s, at this stage favoured relatively rich and knowledge intensive countries. Graph 10 about here There are also shifts in the regional patterns of growth within Europe since 1950 that may be explained by these recurrent shifts in the basis of growth. Western Europe may be divided into three regions both geographically and, with some modifications, also economically: that is the rich Northwest (including Ireland, UK, Holland and the Scandinavian countries), the middle-income Continent (including Germany, Belgium, France, Switzerland, Austria and Italy) and the poor Southwest (Portugal and Spain). The economic categorization is not fully water-proof for the whole period and for every region but it gives a generalized view. Table 7 gives the aggregated growth rates for these regions. Table 7. Annual growth rates in GDP per capita in regions of Western Europe Northwest Continental Southwest Source: See table 4.

16 16 The overall convergence within Western Europe was certainly very strong up to 1973 and it continued at a slower pace until From 1990 there is a clear shift however. While the relatively rich Northwest accelerated somewhat, the poor southwest decelerated further - though still converging. In Continental Europe deceleration was drastic however. It did not only affect Germany but also the rest of the Continent with Switzerland at the bottom in terms of growth rates. It is clear that the growth pattern from 1990 involves elements of both convergence and divergence. But it is also striking that the acceleration in the Northwest, mainly working for divergence, fits to a broader global pattern - accelerating growth in English-speaking or relatively English-speaking environments coincides with the ICT-revolution from the 1990s. This may involve a crucial element in the diffusion and adaptation process to a new communication standard in a globalising knowledge intensive economy. It is also an indication of rather complex investment patterns for catch-up processes to come. Conclusion The idea of this paper is mainly two-folded. Firstly, processes of convergence and divergence, that logically follow from neo-classical theory and new-growth theory respectively, characterise different phases of the growth process. Their alternations may constitute patterns related to structures behind growth. Secondly, those processes could form a periodisation of growth. While growth rates apparently were very even (built upon the new institutions and infrastructures developed in the 1860s and 1870s), the second industrial revolution introduced new forces of diverging rates from the 1890s. The outbreak of war in the 1910s and the dismantling of market integration obstructed a possible convergence with the diffusion of these new development blocks in the following decades. In the interwar period a second set of complementarities evolved very unevenly and were diffused together with technologies from the second industrial revolution in the Golden Ages of growth and convergence With the third industrial revolution, gaining momentum in the second half of the 1970s, forces of divergence and transformation of structures were released once more. In particular, the ICT revolution of the 1990s released forces of divergence comparable to the technology of electricity a hundred years earlier but in a different regional and cultural setting. If electricity put its imprint on regional growth patterns then, it is rather English today. It may also be the case that technological shifts and forces of divergence are particularly strong in periods of industrial revolutions, while convergence gains strength during the subsequent century culminating with the construction and diffusion of infrastructural development blocks that serve to diffuse technological change, increase institutional integration and underpin modern economic growth more efficiently.

17 17 Figure 2. Estimate of sigma convergence in the OECD countries Fig 3. Annual convergence rates in All Europe Estimated with moving base years. Confidence intervals of the estimations

18 18 Fig 4. Annual convergence rates in the OECD countries Estimated with moving base years. Confidence intervals of the estimations Fig 5. Annual convergence rates in Western Europe Estimated with moving base years. Confidence intervals of the estimations

19 19 Equation (2): Convergence OECD Dependent Variable: CONVERGOECD Method: Least Squares Date: 01/16/06 Time: 13:04 Sample(adjusted): Included observations: 49 after adjusting endpoints Convergence achieved after 7 iterations Variable Coefficient Std. Error t-statistic Prob. C USINVRAT WEINVRAT(-1) AR(1) R-squared Mean dependent var Adjusted R-squared S.D. dependent var S.E. of regression Akaike info criterion Sum squared resid Schwarz criterion Log likelihood F-statistic Durbin-Watson stat Prob(F-statistic) Inverted AR Roots.17 Fig 6. Annual convergence rates in the OECD countries with moving base years (full line) and estimated rates by equation 2 (dotted line) CONVERGOECD ESTIMATE

20 20 Equation (2): Convergence Western Europe Dependent Variable: CONVERGWEST Method: Least Squares Date: 01/16/06 Time: 13:06 Sample(adjusted): Included observations: 49 after adjusting endpoints Convergence achieved after 8 iterations Variable Coefficient Std. Error t-statistic Prob. C USINVRAT WEINVRAT(-1) AR(1) R-squared Mean dependent var Adjusted R-squared S.D. dependent var S.E. of regression Akaike info criterion Sum squared resid Schwarz criterion Log likelihood F-statistic Durbin-Watson stat Prob(F-statistic) Inverted AR Roots.36 Fig 7. Annual convergence rates in Western Europe with moving base years (full line) and estimated rates by equation 2 (dotted line) CONVERGWEST ESTIMATE

21 21 Figure 8. Convergence estimated with annually shifted base in OECD, Western Europe and All Europe Index 1950= OECD WE AllE Figure 9. The ratio between annually estimated beta convergence and sigma convergence in the OECD

22 22 Figure 10. Structural convergence in OECD countries Index 1950=

23 23 References Abramovitz, M. (1986), Catching Up, Forging Ahead and Falling Behind, The Journal of Economic History, vol. 46. Abramovitz, M. and David, P. (1973), Reinterpreting Economic Growth: Parables and Realities, American Economic Review, vol. LXIII, no 2. Barro, R.J. & Sala-i-Martin, X. (1992), Convergence, The Journal of Political Economy, Vol. 100, no 2, pp Bresnahan, T.F. &Trajtenberg, M. (1995), General Purpose Technologies: Engines of Growth?, Journal of Econometrics vol. LXV, pp Crafts, N.F.R. (1985), British Economic Growth during the Industrial Revolution. Clarendon Press, Oxford. Crafts, N.F.R. and Toniolo, G. (1996), Postwar growth: an overview, in Crafts and Toniolo (ed.), Economic Growth in Europe since Cambridge University Press. Crafts, N. (2002), The Solow Productivity Paradox in Historical Perspective, CEPR Discussion Paper Dahmén, E. (1950), Svensk industriell företagarverksamhet. Kausalanalys av den industriella utvecklingen Part I II. IUI, Stockholm. Dahmén, E. (1988), Development Blocks in Industrial Economics, Scandinavian Economic History Review, vol. 36. Eichengreen, B. (1996), Institutions and economic growth: Europe after World War II, in Crafts and Toniolo (ed.), Economic Growth in Europe since Cambridge University Press. Goldin, C. and Katz, L. (1996) The Origins of Technology-Skill Complementarity, NBER Working Paper 5657, Cambridge, MA. Jones, C.I. (1995), R&D-based models of economic growth, The Journal of Political Economy, Vol 103, pp Josephson, C. (2005), Growth and Business Cycles. Swedish Manufacturing Industry Lund Studies in Economic History, Almqvist&Wiksell.. Kennedy, W.P. (1987), Industrial Structure, Capital Markets and the Origins of British Economic Decline. Cambridge. Krantz, O. and Schön, L. (1983a), Den svenska krisen i långsiktigt perspektiv, Ekonomisk Debatt, nr 7. Lucas Jr, R.E. (1988), On the mechanics of economic development, Journal of Monetary Economics, vol 22, pp Maddison, A. (1995), Monitoring the World Economy , OECD. Mankiw, N.G. et.al. (1992), A Contribution to the Empirics of Economic Growth, The Quarterly Journal of Economics, May. O Rourke, K.H. and Williamson, J.G. (1994), Late-Nineteenth Century Anglo- American Factor Price Convergence: Were Heckscher and Ohlin Right?, The Journal of Economic History 1994:4. Prados de la Escosura, L. (2000), International Comparisons of Real Product, : An Alternative Data Set, Explorations in Economic History, vol. 37. Quah, D. (1993), Galton s fallacy and test of the convergence hypothesis, The Scandinavian Journal of Economics, vol 95 no 4, pp