Globalization and Growth in the Twentieth Century
Author: Nicholas Crafts

Contributor Notes

Author’s E-Mail Address: N.Crafts@lse.ac.uk

This paper reviews the experience of economic growth during the twentieth century with a view to highlighting implications for both growth economists and policy-makers. The unprecedented divergence in income levels between the OECD economies and many developing countries is documented but so too is a more optimistic picture of widespread progress in terms of the Human Development Index. Various aspects of the changes in economic structure are explored in terms of their implications for growth performance both in retrospect and prospect. The possibility that the growth process will lead to another globalization backlash reminiscent of the 1930s is analyzed.

Abstract

This paper reviews the experience of economic growth during the twentieth century with a view to highlighting implications for both growth economists and policy-makers. The unprecedented divergence in income levels between the OECD economies and many developing countries is documented but so too is a more optimistic picture of widespread progress in terms of the Human Development Index. Various aspects of the changes in economic structure are explored in terms of their implications for growth performance both in retrospect and prospect. The possibility that the growth process will lead to another globalization backlash reminiscent of the 1930s is analyzed.

I. Introduction

This report reviews the experience of economic growth during the twentieth century in the context of trends in globalization and with a view to highlighting implications relevant to early twenty-first century policymakers. Chapter 1 provides a descriptive analysis of trends in growth and living standards during the century, which highlights differences between periods and across countries. Chapter 2 looks at the roller-coaster progress of globalization both in trade and capital flows since the late nineteenth century with particular emphasis on explanations for the prolonged reverse to globalization from the backlash of the interwar years. Chapter 3 uses the building blocks of the earlier chapters to provide an end of century perspective on growth prospects and the extent to which they are likely to be enhanced by improved policies, technological progress and globalization.

The main objective of chapter 1 is to present a basic quantification of growth during the twentieth century drawing on the economic history literature. The central themes that emerge include the massive and unprecedented divergence in income levels and growth performance across countries especially between the OECD and many developing countries in both the first and second halves of the century. The consequences in terms of income foregone of poor policy and/or inadequate institutions were much more serious than in earlier centuries. Nevertheless, a great deal of progress has been made everywhere in terms of the Human Development Index and the spectacular improvements in mortality that have taken place are a major reason to believe that historical national income accounts tend to underestimate growth in living standards. The contribution of technological progress to economic growth both in the advanced and the developing world has varied substantially over time for reasons that are not fully understood either by growth theorists or economic historians.

Chapter 2 spells out what is known about the dimensions of globalization over time. In many respects, the information is patchy but a clear pattern emerges of increases in trade flows and stocks of capital assets relative to levels of GDP before World War I and in recent decades, with the opposite characterizing the 1920s through the 1950s. Many aspects of the current situation are shown to be unprecedented, driven not only by changes in technology but also in economic policy. The retreat of globalization in the interwar period was, of course, driven by policy not technology and was, in large part, a response to the depression. A review of recent historical research concludes that a combination of macroeconomic policy errors and structural faults in banking systems and labor markets rather than globalization per se provoked the slump.

In Chapter 3 the historical experience is linked to insights from modern applied growth economics to examine growth potential after the liberalization and globalization of the recent past. It is argued that globalization itself could be subject to a renewed backlash either, as in the 1930s, in the event of a major world recession or in a new context of the impossibility of governments continuing to meet increased demands for economic security in a world of severe tax competition. In a less apocalyptic scenario, the impact of globalization on fiscal policy could address some of the growth inhibiting effects of the late twentieth century expansion of the public sector in OECD Europe. Prospects for catch-up growth and an escape from the divergence in incomes that has characterized the twentieth century are argued to depend very largely on policy/institutional reform in the developing world to reduce problems of agency costs, rent-seeking, asymmetric information, and opportunism thereby facilitating financial development, investment and innovation. The present policy consensus in favor of markets and outward orientation in the developing world potentially represents a significant improvement from the state led industrialization stance that was prevalent 25 years ago but the political economy of supply-side policy is central to further progress yet little understood. Despite the confidence of many stock market players, long run growth prospects for the United States are particularly hard to predict, both because theory suggests that they depend crucially on scale effects in innovation on which the jury is still out, and also because at this point of history it remains unclear whether the ICT revolution will eventually deliver a major boost to total factor productivity equivalent or superior to that of the era of electricity.

II. What Has Twentieth Century Economic Growth Delivered?

Since World War II, trend growth of real GDP per person has become a key policy objective in virtually all countries. This is not surprising since it is usually thought to be central to raising living standards and because, despite doubts raised by economic theorists, it is generally believed that government policy can influence long run growth outcomes. The twentieth century has seen unprecedented economic growth in many parts of the world but the benefits of growth have often been quite unequally distributed both between and within countries and growth rates have exhibited considerable variation over time. Moreover, it is now frequently suggested that GDP growth is a poor, or perhaps quite misleading, indicator of changing well being and, even within the national accounts framework, it has become widely accepted that important measurement issues have to be addressed, particularly when considering the long run.

In this section, experience of economic growth in the twentieth century is assessed both in terms of its impact on standards of living and also with a view to extracting the lessons that it can offer on why growth rates differ. Two questions are of central concern:

  • What has been the relationship between economic growth and changes in living standards ?

  • What explains the uneven pace and spread of economic growth ?

The renaissance of research in growth economics since the mid 1980s makes this a good time to conduct such a survey. So also does the proliferation of alternatives to national income accounting for the measurement of standards of living (Box 1.1). In addition, however, the end century vantage point suggests answers to these questions that are somewhat different from earlier conventional wisdom simply as a result of recent economic history.

A. Trends in the Human Development Index

One of the most widely discussed measures of the impact of economic development on well being is the Human Development Index (HDI). This has been designed to facilitate long run comparisons and is a measure of the distance traveled from minimum to maximum development in terms of three components, education, income and longevity (see Box 1.1). The focus of HDI is on the escape from poverty. Human development is regarded as a process of expanding people’s choices; income is assumed to impact on this primarily at low levels of material well-being and, above a threshold level, is considered to make a sharply diminishing contribution eventually tailing off to nothing. Life expectancy and education are taken to be central to the enhancement of human capabilities but not generally dependent on private income given the important role that public services usually play in these aspects.

While, as noted in Box 1.1, there are serious index number problems associated with HDI, it offers an important perspective on changing living standards to be considered together with the evidence on growth in real GDP per person. UNDP (1998) Provides estimates only for the period since 1960 but it is possible to extend this back to 1950 for many countries and earlier than that for a few cases. These long run HDI estimates, reported in Table 1.1, provide a comparative context in which to place recent Third World development and, taken together, Tables 1.1 and 1.2 offer a different angle on divergence between the First and Third Worlds from that which emerges from historical national accounting.

Table 1.1.

The Human Development Index and its Components: Long Run Estimates

article image
article image
Sources: HDI from Crafts (1997a), with coverage extended using the same underlying sources, and from UNDP (1998); GDP/head from Maddison (1995)(1997)(1998) measured in purchasing power parity adjusted 1990 dollars and updated where required using IMF (1999) and OECD (1998a); GDP data for Latin America are for 1995; Germany refers to West Germany.
Table 1.2.

The Human Development Index and Its Components in the Recent Past

article image
Sources: as for Table 1.1.

Using these tables, comparisons across countries can be made both of levels of HDI and also of the speed of reduction in the distance from maximum development in different eras. In addition, HDI gaps between advanced and developing countries can be considered. There are, of course, data problems but the broad outlines of the estimates in Tables 1.1 and 1.2 are robust enough for present purposes (Crafts, 1997a). The coverage of the tables is determined by data availability.

The most striking feature of these tables is that the 1995 HDI scores for poor developing countries in Table 1.2 are well ahead of the 1870 scores for the leading countries of that time shown in Table 1.1. Australia’s score of 0.539 in 1870 would rank 127th in the world in 1995. Conversely, Mozambique’s 1995 score of 0.281 (166th in the world) is distinctly above the levels achieved in some parts of Europe, for example, Italy and Spain, in 1870. At the same time, assuming that the HDI score of 0.055 for India in 1913 represents the lowest level, the absolute HDI gap of 0.775 in 1995 between the best and worst in the world (UNDP, 1998) is slightly bigger than in 1913. Since 1950, however, there has been a substantial fall in the gap between average HDI Africa and in the advanced countries of Western Europe, North America, and Oceania from 0.608 to 0.391.

Looking at the components of HDI, the low level of life expectancy at birth (e0) in leading countries in 1870 is apparent with the highest figure at only 49.3 years, a level which has now been exceeded by almost all countries. Research in historical demography long ago confirmed that, during the twentieth century, improvements in mortality resulting from advances in medical science and public health measures, have been largely independent of changes in real income (Preston, 1975). The levels of life expectancy (and HDI) now enjoyed by countries like Algeria and Tunisia were simply not attainable in 1870 for any country given the state of medical technology. By contrast, levels of literacy, which in 1950 were still very low in much of Africa and India, still compare unfavorably in many cases with the leading countries of 1870.

The levels of real GDP per person of many countries in 1990 reported in Table 1.2 were still well below those already attained by the leaders of 1870 shown in Table 1.1. Thus, the average of $1336 for Africa (Maddison, 1995, p. 228) is below the median of $1894 for the countries of Western Europe, North America, and Oceania in 1870. The level of real GDP per person in 1870 of $480 in Africa (Maddison, 1995, p. 228) was about an eighth of the leading country. In 1990, however, the African level was only about one sixteenth of the leading country. The gap between rich and poor measured in these terms has widened enormously or, as one author recently put it, the central feature of growth as measured by the national accounts is “divergence, big time” (Pritchett, 1997).

In order to fill out this point, Table 1.3 considers catching up and falling behind in real GDP/person in terms of regional aggregates. Over the long run since 1870, the most marked and sustained relative declines have been in Africa and in Eastern Europe, areas which account for about 20 per cent of the world’s population in most of this period. Latin America, then a very small fraction of total world population, improved its relative position during the globalization episode prior to World War I but has subsequently lost some ground relative to the United States. Prior to 1950, however, the most important aspect of divergence was the big decline in real GDP/person in Asia, which accounted for well over half the world’s population, relative to the leading country. According to the estimates in Maddison (1998, p. 158) real GDP/person in China in 1950 was only at the 1870 level while India experienced negative economic growth from 1920 to 1950 (Maddison, 1995). In recent decades China and East Asia have, of course, advanced very strongly, although China’s relative position at 11.2 percent of the United States in 1996 is still below its 13.2 percent of Britain in 1890. South Asia now is a similar proportion of the US level to that of 1950.

Table 1.3.

Catchine Up and Falling Behind

article image
Sources: area GDP and population levels based on definitions and data in Maddison (1995) updated using IMF (1999), except for East and South Asia which comprise the countries listed under these headings in Collins and Bosworth (1996); the leading country is defined as the UK in 1870 and the USA thereafter.

The estimates in Maddison (1995) permit a more detailed analysis of changes between 1950 and 1990, years for which he provides a much more detailed country breakdown of income levels. This reveals that 1123 million people (20.1 percent of the non-American world population) in 1990 lived in countries where the level of real GDP/person as a proportion of the US level had fallen since 1950, while another 3151 million (56.1 percent) lived in countries where the level relative to the US had risen by less than 5 percentage points. 16 countries

with a population of 165 million had a level of real GDP/person lower in absolute terms than in 1950 while in only 12 countries (of which only Hong Kong and Singapore were outside the OECD area) was the absolute gap in real GDP/person lower than in 1950.

The HDI measure shows much less divergence. This is partly because of its (controversial) discounting of higher incomes and partly through its taking into account mortality. It is clear, however, that any index of living standards that gives a substantial weight to life expectancy will make present day developing countries look much better relative to either past or present OECD countries than do comparisons based on real GDP/person. This might provoke one of two basic reactions, both of which are probably valid, either that this shows how important it is not to judge progress in development by GDP alone or that this underlines how important it is to pay more serious attention to the index number problems involved in measuring living standards.

Table 1.4 confirms that international comparisons of HDI generally exhibit long run convergence rather than divergence at the level of the large regional bloc and the contrast with Table 1.3 is striking. All regions, including South Asia and Africa, exhibit strong catch-up of

Table 1.4.

Weighted Averages of HDI.

article image
Sources: Western Europe defined as in Maddison (1995) and East and South Asia as in Collins and Bosworth (1996); weighted averages of countries for which data is available using sources as in Table 1.1.

the leading countries after 1950. The averages for both South Asia and Africa in 1995 are quite near to the North American level in 1870. Indeed, all developing countries for which an estimate of HDI is possible for 1950 reduced the gap with the leading country both proportionately and absolutely between 1950 and 1995. In the period 1913 to 1950 there is catch up in HDI for both Eastern Europe (markedly) and Latin America.

The HDI is most useful in looking at the escape from poverty or, in UNDP terminology, the move from a low to a medium level of human development marked by HDI exceeding 0.500. Taking the change in HDI as a measure of the speed of this transition, we find an interesting contrast between the late nineteenth and the twentieth centuries. The 16 countries observable in Table 1.1 in a state of low human development in 1870 posted an average HDI gain of 0.212 by 1913. The 48 countries in Tables 1.1 and 1.2 with an HDI below 0.500 in 1950 had achieved an average HDI gain of 0.350 by 1995. The pace of human development has been markedly higher in the more recent period.

B. Historical National Accounting Growth Estimates

Index number problems loom large in any long run economic growth estimates. National income accounting provides estimates of GDP in current prices which need to be deflated by a suitable price index to obtain inflation adjusted estimates in constant prices. The longer the period under investigation the more difficult this is to do well because the problem of allowing for new goods and services and of quality changes resulting from technological progress becomes much greater. In addition, if, as is often the case, the reason for making the growth estimate is to provide information on changes in living standards, it will be necessary to take account of aspects of well-being that are omitted from GDP such as changes in leisure or longevity and it may also be appropriate to consider the sustainability of consumption (see Box 1.1 and Chapter III).

It is generally accepted that, in advanced western economies in the recent past, inflation has been exaggerated by conventional measurement techniques and thus the growth rate of real GDP has been understated. This has been studied intensively for the United States, and in a recent survey of the evidence, Shapiro and Wilcox (1996) argue that the present day bias is probably somewhere between 0.6 and 1.5 percentage points per year. The problem is not new although it has probably become more serious during the last hundred years as the composition of GDP has moved away from run of the mill commodities towards more difficult to measure services, government activities and durable goods. By contrast, in Communist countries it is common for official statistics to underestimate inflation; for example, the recent study by Maddison (1998) finds it necessary to reduce official estimates of Chinese growth since 1978 by about 2 percent per year—a correction that is incorporated into the tables in this chapter.

Leaving aside the index number problems for the moment, Table 1.5 reports the available estimates and changes the focus from levels to growth rates. The table is based on the periodization in Maddison (1995) which is useful for OECD countries in particular for separating out the disturbed years around the depression and the two world wars (1913–50) and what is often termed the golden age of growth prior to the first OPEC shock (1950–73). Obviously, for some countries and some purposes, for example, an examination of the impact of communism on Eastern European growth or of reform on Chinese growth, this framework is not so suitable. Nor is this design set up to highlight shorter run fluctuations in growth rates such as the collapse of the early 1930s.

Table 1.5.

Growth Rates of Real GDP/Person

(Percent per year)

article image
Sources: as for Table 1.1.

The first point to note from Table 1.5 is that twentieth century growth has generally been much stronger than that prior to 1870. Thus, regions like India and Latin America where citizens are perhaps disappointed not to have matched the East Asian growth rates of the last four decades have nevertheless performed much better than in the mid nineteenth century and over the whole period since 1950 have grown faster than did the U.K. and the U.S. between 1820 and 1870. Under mid and even late nineteenth century conditions, 1.5 to 1.8 percent per year was about the maximum growth rate except in a few ‘regions of recent settlement’ such as Argentina and Canada.

This, however, represented a major breakthrough from pre-Industrial Revolution growth capabilities where long run growth at 0.2 percent per year was a good result. Recent research has established that, even during the First Industrial Revolution, Britain achieved a growth rate of real GDP per person no higher than about 0.4 percent per year between 1780 and 1830 (Crafts and Harley, 1992). Indeed, the experience of Industrial Revolution Britain, while representing a major breakthrough from the past in terms of technological advance and resulting in extremely rapid industrialization and urbanization, is also remarkable, viewed through a modern lens, for what it reveals about the limits to growth in the leading economy of the mid nineteenth century.

By modern standards, Industrial Revolution Britain had a very modest growth potential. Investment rates and formal schooling were very low relative to twentieth century levels and research and development spending was negligible. Market sizes were still small and traditional rent-seeking occupations absorbed much of the talent in the economy. The costs both of inventing and protecting the profits from invention were relatively high compared with later periods. Growth was based on quite different foundations from those characterizing twentieth century success stories (Crafts, 1998).

The next obvious feature of Table 1.5 is that, despite twentieth century acceleration, growth rates for real GDP per person above 2 to 3 percent per year have still been relatively unusual. For both Western and Eastern Europe, the early post World War II decades stand out as the episode of by far the most rapid growth followed by a marked slowdown. One way of reading the table might seem to be that in the late decades of the twentieth century OECD economies have returned to historically normal modern economic growth after periods of disruption followed by recovery. Indeed, the hypothesis that growth has returned to the pre-1914 trend rate cannot be rejected for six of the Maddison 16 countries, although in all but two of these cases at a higher level of GDP per person than would have resulted from simple extrapolation of the pre-1914 growth rate (Crafts and Mills, 1996). Nevertheless, as later sections will establish this would be a seriously misleading view of the evolution of the twentieth century growth process.

Table 1.1 shows that during the first half of the twentieth century the United States established a large lead in real GDP per person relative to Europe and East Asia. The American lead was based on successful exploitation of natural resources and a large domestic market together with prowess in high-technology based on tertiary education that other countries found hard to emulate. Also the World Wars did not damage the American economy unlike the experience elsewhere, for example in Germany and Japan. Since 1950, many Western European and East Asian countries have considerably reduced the percentage (if not the absolute) gap with the U.S. Thus, in 1950 levels of GDP per person in France and South Korea were 54 and 9 percent respectively of that in the United States but 77 and 54 percent respectively in 1996 and 69 and 21 percent respectively in 1900. While the gap between the richest and poorest has continued to widen, there has been a significant catch-up of the leader by OECD and Tiger economies.

The fastest growth has been achieved by economies that are successfully catching up from well behind the leader, such as high performaing East Asian economies in recent decades, while the United States has not exceeded 2.4 percent per year in any of these periods. The fast growth in postwar Europe also benefited from catch-up and seems to owe a good deal to the reduction in barriers to the emulation of American technology, as well as the reversal of earlier policy errors and the return to peacetime. Nelson and Wright (1992) stress the reduction in the advantage that America had gained from cheap natural resources and a large domestic market as transport costs fell and European integration and trade liberalization proceeded. Investments by Europeans in human capital and in research and development (R&D) facilitated the codification and spread of technological knowhow while high volumes of physical investment were achieved on either side of the Iron Curtain. It should also be recognized, however, that growth slowed down again well before the Europeans (or the Japanese) had completed their catch-up of the United States.

This is especially true of the communist countries which in the 1960s were sometimes thought likely to overtake the United States before the end of the twentieth century. Although they mobilized huge investment programs, capital accumulation ran into severely diminishing returns and incentive structures under communism were not conducive to sustaining high rates of innovation (Easterly and Fischer, 1995). Had catch-up been as successful as in Western Europe, countries like the Czech Republic and Hungary could have been expected to have GDP/person in 1996 at around the level of Austria or Italy, i.e., at least $10,000 (1990 international) higher (Fischer et al., 1998).

Among the now advanced economies, Table 1.5 reports striking differences in growth performance that have been reflected in relative advance and decline in levels of real GDP/person. The U.K., which in 1900 still had the highest real GDP per person, was by 1996 only 13th of the countries listed and had been overtaken by most European and several Asian countries including Hong Kong and Singapore. Conversely, Japan ranks 8th in 1996 but in 1900 had an income level below that of Russia. The damage done by communism is underlined by comparing the post-1950 performance of Czechoslovakia and Hungary (data refer to the area of those countries as in 1990) with that of Austria and Italy.

Prima facie, Table 1.5 may seem to suggest that there was more reason to be bullish about twentieth century growth in 1973 than now. The obstacles of colonialism, the World Wars and Depression seemed to be in the past and to many it seemed reasonable to suppose that rapid catch-up growth might spread much more widely. For the West, the power of technology, and the computer revolution in particular, seemed to promise sustained fast growth. Clearly, there has been a substantial slowdown in world growth in the last quarter century from which only Asia (until recently) has largely escaped.

So the growth experience of the last quarter century has produced several puzzles for growth economists. One is to produce an adequate account of the reasons for very strong growth in East Asia while in many other regions, including most of Africa, have had a dismal growth failure. It seems clear that a full explanation of these contrasting outcomes requires something more than can be found in conventional growth models and this has promoted investigations of what, following Abramovitz (1986), might be termed “social capability” for catch-up growth. Another new issue that has emerged following the development of endogenous growth models is to explain the failure of growth to accelerate in the United States despite increased investment in human capital and R&D (Jones, 1995).

C. Adjusted GDP as a Guide to Long Run Living Standards

During the twentieth century there have been big changes in hours spent on market work and in mortality in the countries for which we have long run estimates. Table 1.1 showed that average life expectancy at birth has roughly doubled since 1870 in leading OECD economies. In the same period, hours worked per person employed have roughly halved (Maddison, 1995). Both of these are aspects of improved living standards that would not have been captured by growth in real GDP. Usher (1980) argues that it is both possible and desirable to augment measures of economic growth to include these components of well-being (see Box 1.1). Although there is no consensus on exactly how best to accomplish imputations of this kind, it is useful to consider illustrative calculations along the lines proposed by Usher simply because the changes have been so great.

Crafts (1997a) explains in detail the method used to obtain the undeniably crude estimates in Table 1.6 which are almost certainly underestimates of the imputations that should be made for changing market work hours and mortality. Improvements in life expectancy that are unrelated to personal consumption expenditure are valued on a willingness to pay basis based on nineteenth century rather than twentieth century behavior, which would yield much bigger welfare gains (Nordhaus, 1998), while reductions in market work, valued using wages foregone, ignore the possibility of technical progress in non-market work. Even so, the results raise growth rates and, in some cases, by a substantial amount. Applications of this methodology would not always give this result—for example, in industrial revolution Britain there may well have been periods when the adjustment would tend to reduce growth (Crafts, 1999b).

Table 1.6.

Growth Rates Adjusted for Hours Worked and Mortality

(Percent per year)

article image
Source: updated from Crafts (1997a), see text.

Table 1.7 highlights the variability of hours worked per person over time and across countries. These differences result partly from demographic factors, partly from female labor force participation and partly from hours worked per employee. With regard to this last, Latin American and East Asian countries are now similar to Western Europe in the 1950s and the 1920s, respectively, but work years in both regions are well below the level of almost 3000 characteristic of late nineteenth century Europe.

Table 1.7.

Annual Hours Worked per Head of Population.

article image
Sources: Maddison (1995) updated for 1996 as in Crafts (1999a); estimates for South Asian and Latin American countries refer to 1992.

Labour inputs per person are reported in Table 1.7 for Latin America in 1992 to be much lower than for Western Europe in 1870. This has important implications for comparative productivity performance as well as for welfare comparisons. Thus, while the real GDP/person in Latin America in 1996 of $5155 reported in Table 1.3 is less than 60 per cent ahead of the UK level of $3263 in 1870, real output per hour worked was over three times that of the UK in 1870. The gap in real GDP/person clearly hugely understates the extent to which economic welfare in Latin America has outstripped the level attained in Britain in 1870, not only because life expectancy is so much higher but also because labor inputs are so much lower.

Differences in the age structure of the population and/or hours worked per person employed per year can also mean that comparative real GDP per person is a poor indicator of labor productivity in the present day. This turns out to matter most when comparisons are made between East Asia and Europe. Thus, whereas, by 1996, the leading Tiger economies, Hong Kong and Singapore, had overtaken most of Western Europe in real GDP per person there was still a substantial gap in terms of labor productivity. Their continued fast growth prior to the recent Asian crisis is less paradoxical when it is recognized that their scope for catch-up of the leading OECD economies continues to be quite substantial (Crafts, 1999a).

The adjustments considered thus far have tended to give reasons why conventionally measured real GDP growth per person may underestimate the improvement in living standards, at least for the OECD economies in the twentieth century. There are, of course, a number of other omissions from GDP which impinge on living standards and where a more negative view might be appropriate. The most often advanced reason why growth measured by historical national accounting might exaggerate growth in living standards is the neglect of environmental damage and depletion of non-renewal resources which might imply that NNP is an overestimate of sustainable consumption (Hicksian income).

This argument clearly has some validity. A recent calculation by Weitzman (1999) suggests that depletion of the most significant exhaustible minerals costs the world the equivalent of a little over 1 percent of average consumption each year compared with a counterfactual of a constant flow of resources at this year’s extraction cost. On the other hand, it is necessary to correct NNP not only for unmeasured natural resource capital depletion but also for additions to knowledge capital coming from technological advance to infer correctly sustainable consumption. Depending what is assumed about future technological progress, this may well imply a much bigger correction to NNP in a positive direction, as Nordhaus (1995) has argued.

It is, in fact, the possibility that returns to R&D may be falling, with an implication of reduced future productivity growth, rather than natural resource depletion that is by far the most important reason to suppose that in recent decades sustainable consumption has been growing much less fast than real GDP. Amendments to the correction to NNP for growth in knowledge capital as projections of future technological progress are revised are likely to dominate other adjustments to national accounts measures in assessing sustainability.

Economic historians have rightly warned that economic growth as measured by historical national accounts is not always a good guide to the rate of improvement of average living standards. For most of the twentieth century, it seems likely that, on balance, GDP growth has been an underestimate in a world of falling mortality, increasing leisure, new goods and greater technological prowess.

D. The Sources of Economic Growth

A useful technique with which to examine long run growth is growth accounting. This approach, which is well explained in Barro (1998) and Maddison (1987), seeks to attribute growth to its proximate sources in terms of growth of factor inputs and of total factor productivity (TFP). TFP is the weighted average of the growth of productivity of the individual factor inputs. The basic formula used in growth accounting is the following:

ΔY/Y=αΔK/K+βΔL/L+ΔA/A

where the growth rate of output (Y) is accounted for in terms of the contribution of the growth of the capital stock (ΔK/K) times the elasticity of output with respect to capital (α), the contribution of the growth of the labor force (ΔL/L) times the elasticity of output with respect to labor (β) and the growth of TFP (ΔA/A).

In practice, α and β are approximated by the shares of profits and wages, respectively, in GDP and TFP growth is found as a residual when all the other components of the growth accounting equation have been entered. Capital stocks are estimated using the perpetual inventory method of adding up past investment flows and assuming a lifetime for capital assets, while labor inputs are usually measured in hours worked adjusted for the educational composition of the labor force based on human capital theory. This formula would be exactly right if, as in traditional neoclassical growth theory, the economy could be thought of as an aggregate Cobb-Douglas production function, Y = AKαLβ operating under conditions of perfect competition and constant returns to scale. The parameter A would reflect the state of technology and TFP growth would measure exogenous (Hicks-neutral) technological change.

Caution is required before assuming that residual TFP growth actually measures the contribution of technological change to economic growth. According to traditional analysis, the bias may go in either direction. First, technological change may be less than TFP growth if there are scale economies or improvements in the efficiency with which resources are used or if improvements in the quality of factors of production are underestimated, for example due to unmeasured human capital accumulation (Abramovitz, 1993). Second, if the elasticity of substitution between factors of production is less than 1 and technological change has a Hicks-labor saving bias, as many analysts think is the case, then conventional TFP growth underestimates the contribution of technological change and the mismeasurement increases with the growth in the capital to labour ratio, the degree of labor-saving bias and the inelasticity of substitution (Rodrik, 1997).

Since faster technological change raises the steady state rate of growth of the capital stock in a traditional neoclassical growth model and so part of its impact on growth compared with the counterfactual of no technological change shows up in capital’s measured contribution. The advent of endogenous growth theory strengthens this kind of reason to believe that the contribution of technological change exceeds TFP growth. Thus, in models which envisage endogenous innovation driving growth through expanded varieties of capital inputs a fraction of the contribution of the growth in varieties of capital facilitated by R&D accrues to capital and is not captured by the Solow residual. The undermeasurement will be greater the larger is the endogenous component in technological progress (Barro, 1998).

Table 1.8 reports a selection of growth accounting results which permit a long run view of the sources of growth in some G7 countries. The main trends are as follows. First, in the U.S. the contribution of capital seems to have been stronger in the nineteenth century than in the twentieth both absolutely and proportionately. In other countries, the strongest absolute contribution of capital came in the Golden Age investment boom after World War II. The late twentieth century decline in capital’s contribution is associated both with rising capital to output ratios and falling shares of investment in GDP but gross non-residential investment rates in most G7 countries are several percentage points above their late nineteenth century levels (Maddison, 1992).

Table 1.8.

Growth Accounting: Comparisons of Sources of Growth

(Percent per year)

article image
Sources: 1855–90 USA from Abramovitz (1993). 1873–1913 UK from Matthews et al. (1982). G7 countries: 1913–50 from Maddison (1991), 1950–92 from Maddison (1996). East Asia from Collins and Bosworth (1996) except for Hong Kong which is based on Young (1995) and China based on Maddison (1998) both with factor shares adjusted to match Collins and Bosworth’s assumptions. South Asia, Latin America, Africa, Middle East from Collins and Bosworth (1996).

Second, the contribution of labor inputs to growth has been strongest in the United States reflecting, in particular, higher population growth there than elsewhere in the G7 countries. Measured on this neoclassical basis education’s contribution has been steady but not spectacular—for example, contributing a little less than 0.5 percent per year in each period since 1913 in the United States where years of formal education rose from about 6 at the start of the century to 9.5 in 1950 and 13.5 in 1995 (Maddison, 1996; OECD, 1998b). In much of the OECD, declining hours worked has been a significant restraint on the growth of labor inputs during the twentieth century.

Third, the contribution from TFP growth has been highly variable ranging from 0.2 percent per year in the U.S. in 1973–92 to 3.6 percent per year in Japan in 1950–73. The interwar productivity surge in the United States appears to owe a good deal to electrification which raised TFP growth across the board in American manufacturing (David and Wright, 1999). By contrast, catching up, scale effects and improvements in resource allocation made strong contributions to TFP growth in Golden Age Europe and Japan (Maddison, 1996). The broad picture is that generally TFP growth rose from the late nineteenth century through the Golden Age and then declined sharply in the recent growth slowdown. The low TFP growth in both Britain and the U.S. in the nineteenth century is suggestive of the limits to growth at that time, discussed above. Rising TFP growth through the 1970s correlates with spending on R&D which was negligible in the nineteenth century, around 0.2 and 0.5 percent of GDP in the interwar U.K. and U.S. respectively (Edgerton and Horrocks, 1994), and rose to 2.3 and 2.9 percent respectively by the 1960s (United Nations, 1964). However, the decline in American (and G7) TFP growth in the last period occurred despite increased R&D spending.

The contrasts in TFP growth over time in the United States probably do reflect real changes in the contribution of technological change to growth but surely exaggerate the movements. The nineteenth century American economy experienced a rapid rise in the capital to labor ratio combined with strongly labor saving technological change such that the share of profits in national income rose. Early in the twentieth century, the bias in technological change seems to have ceased to be labor saving and to have been replaced by a capital saving (human capital using) bias and the share of profits fell (Abramovitz, 1993). In recent decades, this capital saving bias has been less apparent and the share of profits has fluctuated with no strong trend. In the light of our earlier discussion, it seems probable that conventionally measured TFP growth significantly understates the role of technological change in the nineteenth century and overstates it between the 1920s and the 1960s. The recent decline in TFP growth in the United States appears paradoxical in the light of the computer revolution and this has prompted many to suspect that it may be, at least to some extent, a statistical artifact (Griliches, 1994). We return to this issue in Chapter 3.

The developing world clearly offers some striking comparisons with the advanced world, although data limitations restrict measurement to the post-1960 period. The startling contrasts in this period are between the much faster growth in East Asia in a still to be completed process of catching up the OECD and the growth failure in Africa which has fallen further behind and has seen real GDP per person in the continent declining in the last quarter century.

Table 1.8 shows negative TFP growth for Africa in the period 1960–94, and contributions from factor inputs which look respectable until it is recalled that population growth was nearly 3 percent per year. Probing behind these numbers, the investment rate in Africa has been held down in real terms by the very high price of capital goods in highly protectionist economies and educational levels have remained low rising from 1.6 years in 1960 to only 3.5 years in 1994 (Collins and Bosworth, 1996). In East Asia the investment rate has been high relative to historical norms and has translated into a strong contribution from the capital stock because of unusually low capital to output ratios (Fukuda, 1999). The contribution from labor inputs has been boosted by a favourable demographic transition and rapid increases in schooling, from an average of 2.7 years in 1960 to 7.2 years in 1994 (Collins and Bosworth, 1996). The stronger TFP growth in East Asia may partly reflect effectiveness in technology transfer which compares very favorably with other Third World experience (Dahlman, 1994).

Comparisons of East Asian with European growth in the recent past are somewhat beside the point given the differences in capital to output ratios, scope for catch-up in TFP and demography. Examining differences in the sources of recent fast growth in East Asia and of rapid growth in Europe during catch-up in the 1950s and 1960s is instructive, however. Here, the obvious point to stand out from Table 1.8 is that East Asian growth has relied much more heavily on factor inputs, both labor and capital, and less on TFP growth than that of Golden Age West Germany. Although measurement error occasioned by factor-saving bias may exaggerate the TFP growth difference somewhat, the conclusion that East Asian TFP growth has been outstanding relative to that of Africa but not so impressive by earlier European standards seems robust (Crafts, 1999a).

E. Convergence and Divergence

The large divergence in income levels and growth rates that has emerged since the start of modern economic growth seems hard to square with the traditional neoclassical growth model which assumed constant returns to scale, diminishing returns to capital accumulation and universal technology which improved like manna from heaven. Although some early work in the growth regressions literature argued that differences in income levels across the world could largely be explained by human and physical capital per worker (Mankiw, Romer and Weil (1992) and tended to interpret β-convergence (i.e., the negative relationship between growth rates and initial labor productivity levels found in cross-section regressions when enough conditioning variables are included) as consistent with the Augmented-Solow growth model with convergence at 2 percent per year (Barro and Sala-i-Martin, 1991), these interpretations are now generally rejected (Temple, 1999).

Economic historians reviewing the experience have tended to stress the importance of technological congruence and social capability (Abramovitz and David, 1996). The former relates to the profitability of using technology developed in the leader(s) in potential follower countries with different factor endowments and demand conditions. Thus, in the first half of the twentieth century, American technology which was natural resource intensive, physical capital-using and scale dependent was frequently not the optimal choice of technique in European conditions. Greater integration of world markets, reductions in the cost advantages of domestic natural resource endowments combined with increased importance of intangible capital (R&D and education) subsequently reduced the obstacles to catch-up first within the OECD and later elsewhere in East Asia. Social capability refers to a country’s culture, institutions and policy framework which influence the attractiveness of investment and innovative activity and the efficiency with which technological possibilities are exploited. The productivity gap with the leader informs the potential for rapid catch-up growth but catching-up is not automatic. This coheres with the more recent econometric evidence.

Prescott (1998) explores the possibilities of calibrating a neoclassical production function to account for cross-country labor productivity differences in terms of human and physical capital per worker with TFP common to all countries and concludes that the evidence is not remotely consistent with this hypothesis. While econometric evidence strongly supports diminishing returns to (broad) capital accumulation, estimates of rates of conditional convergence are now known to be highly sensitive to econometric specification and there appears to strong evidence both of differences in levels and/or rates of growth of TFP across countries. This strongly suggests that both policy and institutions matter for growth performance (Temple, 1999).

III. Globalization Then and Now

Many economists have discussed the twentieth century in terms of a phase of globalization ending with World War I followed by a phase of disintegration through until the postwar reconstruction after World War II. The last fifty years is then seen in terms of a gradual liberalization of trade and capital flows followed by a new and deeper version of globalization in the last quarter century or so. In both the OECD and the third world, perspectives on the advantage of outward orientation have clearly varied dramatically over time as has the impetus to globalization from technological change. The objectives of this section are to establish some of the key dimensions of globalization and to explain why the process was reversed in the interwar period as governments retreated to financial autarchy and trade protectionism.

A. Globalization in Trade and Factor Flows

It is widely known that for many countries the proportion of merchandise trade to GDP has now returned to levels quite similar to those at the start of the century. Prima facie, it might appear that the world economy has simply returned to its old level of economic integration and that international trade has merely resumed its earlier importance. As has increasingly been recognized, that view is seriously misleading. The nature of international trade is in several respects quite different now compared with one hundred years ago and it plays a much larger part in world economic activity (Bordo et al., 1999).

Table 2.1 reports ratios of world merchandise trade relative to world GDP and largely matches the stylized facts set out above. The trade ratio was increasingly rapidly before World War I, did not exceed the 1913 level until the late 1960s, and has risen sharply in recent decades. It should be noted, however, that the present level is unprecedently high and represents both a massive increase over the estimate of 1 percent for 1820 and a near doubling of the 7 percent in 1950 (Maddison, 1995, p. 233). On the other hand, for U.K., the leading player in the global economy of the late nineteenth century, merchandise trade is now a much smaller proportion of GDP then it was then and for the U.S. the increase in the trade ratio is quite modest.

Table 2.1.

Ratios of Merchandise Exports to GDP and to Merchandise Value-Added

(Percent)

article image
Sources: Feenstra (1998) except estimates for World which are derived from Maddison (1995).

The picture is rather different when changes in the structure of OECD economies are taken into account, in particular, the shift away from commodity to service sector production. Table 2.1 shows that when merchandise trade is expressed as a percentage of merchandise value added most countries now have a much higher ratio than in 1913. Most obviously, this is true of the United States where the trend since 1970 is a striking new development. The U.S. has also experienced a rapid growth in service sector exports which by the mid 1990s were around 40 percent of merchandise exports up from 30 percent in 1960 and dwarfing the 3 percent or so of 1900 (Bordo et al. 1999). This probably reflects both the increasing tradability of some services and also shifts in comparative advantage. It is not unprecedented, however, in the sense that service/merchandise proportions around 40 percent also characterized British exports in the period 1870–1913 (Imlah, 1958).

It is also important to recognize that the composition of world merchandise trade has change very substantially during the twentieth century, as Table 2.2 reports. The most obvious change is the huge relative decline of primary products and rise of manufactured goods especially since World War II. In many ways, perhaps the most significant development is the rise of the developing countries’ share of manufactured exports which was very low and showing no upward tendency through the 1960s but has grown dramatically in the last thirty years to around 25 percent by the mid 1990s.

Table 2.2.

Composition of World Merchandise Trade

(Percent, current prices)

article image
Sources: UNCTAD (1983)(1997) except for 1913: Yates (1959).

Also noteworthy in Table 2.2, is the steady increase throughout the century from a low initial level in the proportion of capital goods, represented here by machinery and transport equipment, in manufactured trade. It also seems clear that this has been accompanied in the recent past by a rapid increase in both outsourcing and vertically specialized trade which may now have reached 20 to 25 percent of world trade (Hummels et al., 1998). In the U.S., imports of SITC 7 manufactures (machinery and transport equipment) were still only 5 percent of total merchandise imports as recently as 1955 but are now around 50 percent (Yates, 1959;UNCTAD, 1997).

Another interesting development during the twentieth century has been the part played by multinationals in world trade and production. Multinational enterprise was already quite well established in the early twentieth century, as Table 2.3 reports and the book value of foreign direct investment relative to world GDP is probably only a couple of percentage points higher now than in 1914. Nevertheless, the market value of U.S. direct investment abroad has been estimated as 20 percent of GNP in 1996 compared with around 7 percent in 1914 (Bordo et al., 1999). Also, in recent decades, multinationals have become more important in technology transfer (Nelson and Wright, 1992) and in the proliferation of outsourcing (Lawrence, 1996).

Table 2.3.

Foreign Investment

article image
Sources: foreign assets/GDP from Obstfeld and Taylor (1999); stock of FDI from Jones and Schroter (1993) except for 1995 from OECD (1998); multinationals from Bostock and Jones (1994) except for 1995 from OECD (1997).

The story of foreign portfolio investment in the twentieth century is of flourishing growth through World War I followed by a pronounced retreat from the 1930s through the 1950s and then accelerating growth from the 1960s to the present. Table 2.3 reports that the stock of total foreign assets relative to world GDP regained the 1914 level around 1980 and has risen dramatically since that time.

Obstfeld and Taylor (1999) interpret these trends in terms of a macroeconomic policy trilemma, i.e., that a country can have at most two of a fixed exchange rate, free capital movements, and independent monetary policy. At the start of the century, the advanced country norm was to eschew the last of these in an era when politicians denied responsibility for the level of domestic economic activity and working class votes still did not matter in most countries. In the crisis of the 1930s, independent monetary policy ruled the roost, widespread capital controls were introduced and devaluations were commonplace. During the Bretton Woods period, fixed exchange rates returned and capital controls were retained in a world in which aggregate demand management was widely attempted while since the early 1970s the trend has been toward floating exchange rates, independent monetary policy and abandonment of capital controls.

This macroeconomic background is central to explaining trends in quantities of foreign investment but other factors have impinged on the composition of capital flows. It has long been recognized that portfolio investment before World War I was heavily concentrated in lending to the transportation and government sectors while relatively little went to banks or industry. According to estimates in Simon (1967), 69 percent of British portfolio investment between 1865 and 1914 was in social overhead capital, and 35 percent went to governments while only 4 percent was in manufacturing. Lending to emerging markets in the 1990s involves much more exposure to the financial sector and much less to infrastructure (Bordo et al., 1999). A plausible explanation for this may be that problems of asymmetric information and contract enforcement are, on average, somewhat reduced relative to the pre-1914 era.

The disintegration of world trade after World War 1 resulted from a surge of protectionism. Prior to World War I, the use of quantitative trade restrictions was negligible (Gordon, 1941) but during the 1930s they proliferated. Some estimates suggest that as much as 70 percent of world trade was affected, although Gordon herself suggests the figure was more like 50 percent. Trade liberalization after World War II centered initially in Europe on removing import controls but during the 1970s and 1980s there was a return to higher non-tariffbarriers to trade including new devices such as voluntary export restraints which evaded the GATT. Many of these NTBs embodied levels of protection equivalent to quite high tariffs. The Uruguay Round placed considerable emphasis on removing NTBs with some success, as Table 2.4 reports.

Table 2.4.

Barriers to Trade: Average Tariffs on Manufactures and Import Coverage of NTBs

(In percent)

article image
Sources: Tariffs from Bairoch (1993) and Schott (1994); NTBs based on Gordon (1941) and Daly and Kuwahara (1998). For the 1930s, tariffs are for 1931, except UK for 1932, and NTBs are for 1937.

Trends in tariff protection are similar in some respects but differ in others. It is worth bearing in mind that the late nineteenth and early twentieth century were a period when tariff barriers to trade were generally increasing somewhat. Although the U.K. stood out as a committed free trader prior to World War I, the U.S. was a high tariff country throughout until World War II. Tariffs increased markedly during the trade wars triggered off by the Smoot-Hawley tariff of 1930 but were reduced steadily under the various multilateral GATT rounds especially from the Kennedy Round on. These reductions were not reversed during the macroeconomic turbulence of the 1970s. In sum, it seems quite probable that Bordo et al (1999) are right to claim that trade barriers today are quite likely lower than a century ago.

The early twentieth century was also a time of high international migration characterized in particular by emigration from Europe, increasingly from southern and eastern Europe, and immigration to the New World. This had a substantial impact in reducing wage gaps between sending and receiving regions and put downward pressure on wage rates of the unskilled in the United States (Williamson, 1996). In turn, the adverse implications for unskilled workers appear to have been a major reason for the tightening of restrictions from the 1860s onwards and ultimately the shutting of the door to immigrants in the United States and other countries in the 1920s (Timmer and Williamson, 1998). Thus, by the 1930s, tendencies to factor price equalization were much weakened by obstacles to factor flows and protectionism.

There has been no comparable relaxation of immigration controls to accompany the liberalization of trade and capital flows. As Table 2.5 reports, the proportion of foreign bora population in the U.S. has risen from the nadir reached in 1970 and immigrant flows have revived a little. In Western Europe the proportion of foreign born population has risen from 3.6 percent in 1965 to 6.1 percent in 1990 (Zlotnik, 1998), a record figure, but this period has also seen severe immigration restrictions imposed in a region previously accustomed to substantial net emigration. Indeed, absent these barriers, one might have anticipated very much larger total flows of migrants given lower transport costs and higher incomes in less-developed countries.

Table 2.5.

Immigration to the United States

article image
Source: U.S., Bureau of the Census

B. Explaining Trends in Protectionism

It is now widely accepted that openness is good for productivity growth (Edwards, 1998) and the relative decline of economies that adopted import-substituting industrialization strategies rather than outwardly orientated policies in recent decades has both strengthened this conventional wisdom and encouraged trade liberalization in the third world. These beliefs and policy stances were not widely shared until recently. And, in any event, big countries may seek to exploit optimal tariff policies even at some cost in productivity performance while in depressions considerations of external and internal balance may make expenditure switching policies an attractive short-term fix. The rise of protectionism through the middle of the century resulted from a combination of adverse macroeconomic shocks, the aggressive behavior of the new “hegemon”, the United States, and an absence of the link between openness and growth that emerged in more recent times.

It is well-known that downturns in economic activity are conducive to protectionism in societies where governments respond to the balance of interest group pressures (Gallarotti, 1985). Trends in protectionism in the United States during the twentieth century have repeatedly been shown broadly to conform with this generalization (Bohara and Kaempfer, 1991; Takacs, 1981). It is not therefore surprising that the world economic crisis of the 1930s saw a surge in trade restrictions of all kinds.

American tariff policy in the late 1920s was essentially the outcome of log-rolling by special interests in Congress rather than a carefully calculated foreign policy move by the administration (Irwin and Kroszner, 1996). Nevertheless, the trade wars of the 1930s can also be seen as equivalent to a move by the United States to exploit its position as a big country which backfired. The introduction of the Smoot-Hawley tariff in 1930 provoked retaliation on a wide scale notably by larger countries and, in retrospect, represents a serious miscalculation. The tactics adopted later in the decade following the Reciprocal Trade Agreements Amendment were a better way of seeking optimal tariffs which effectively distinguished between countries of relatively high and low bargaining power (Conybeare, 1987). During the Pax Americana after World War II the United States moved toward sub-ordinating trade policy to foreign policy and this aided considerably the shift back toward trade liberalization.

In recent decades protectionism seems to have reduced growth rates as a consequence of the distortions to which it gave rise. A detailed analysis of long-run Latin American growth performance by Taylor (1998) found that the key impact worked through a whole series of distortions that raised the price of capital goods rather than high tariffs per se. Similarly, Ades and di Telia (1997) have argued that there are serious adverse impacts of industrial policy from the corruption that it generally spawns which is fostered by protected markets.

In the early twentieth century, the inverse correlation between tariffs and growth was absent. A careful study by O’Rourke (1997) which uses all the obvious controls confirms the intuitions of earlier writers like Bairoch (1993). The reason for this result may well be that, in general, tariffs at this time tended to reduce the relative price of capital goods and thus to stimulate rather than to discourage real investment. This is suggested by the evidence in Collins and Williamson (1999) who found that prior to 1950 tariffs tended to lower the price of capital goods relative to consumer goods. More speculatively, it may also be that in an era when “industrial policy” was in its infancy there was less reason to associate protection with additional distortions.

C. The Origins and Nature of the Great Slump

The tendency to disintegration of the world economy in the interwar period was hugely exacerbated by the slump of the early 1930s. It seems highly probable that a return to global economic crisis of similar proportions would provoke similar antipathy to international capital mobility and free trade leading to widespread adoption of policies aimed at reimposing controls. Understanding the reasons for the depression, in particular of the parts played by structural faults and policy errors, is central to assessing the risk of a repeat and is thus a key building block in weighing up the likelihood of a future backlash against globalization. Also, understanding the propagation of the initial adverse shocks is fundamental to averting a repetition of a major slump and an associated retreat from openness.

The traditional literature on the Great Depression stressed catastrophic errors in American monetary policy which provoked a huge aggregate demand shock that reduced output and prices both domestically and then, through secondary effects, the rest of the world (Friedman and Schwartz, 1963). This is still a valid, but now clearly incomplete, picture as more recent research has brought out the importance of wage stickiness and aggregate supply in real output falls, of the operation of the newly restored Gold Standard in the inadequacy of the world macroeconomic policy response, and of the fragility of banking systems in the onset of financial crisis.

Those who argue for the fragility of the world economy prior to the depression are typically pointing to various legacies of World War I such as adjustment problems linked to world agricultural over-production and loss of export markets by European countries (Kindleberger, 1973). Their claims are not that the world economic crisis was inherent in early-twentieth century globalization but that the shock delivered by the war seriously impaired subsequent macroeconomic policymaking, led to increased pressures for protection and/or had adverse implications for the volatility of capital flows. Indeed, the main line of argument is now that the war shock was primarily responsible for the instability of the interwar gold standard but the significance of the changed economic environment was perceived too late (Temin, 1989).

Eichengreen (1992) established a picture of the pre-1914 Gold Standard as an era in which Britain was frequently able to operate as a Stackelberg leader because of sterling’s reserve currency role and as a world in which international cooperation sustained the system when under pressure. The priority given to gold convertibility co-existed with quite high volatility in domestic levels of economic activity but domestic politics permitted neglect of internal balance objectives. By the 1920s all this had changed with the rise of New York, the differing views of major players on the appropriate conduct of monetary policy, and the pursuit of working class votes. American monetary policy first provoked a general world tightening and then presided over a sensational collapse in the domestic money supply and output. This episode is discussed in detail in Eichengreen and Sussman, 2000.

A substantial rise in real interest rates and real wages ensued in the advanced world. Newell and Symons (1988) estimated that in the representative European country real wages rose by 13.8 percent and the real interest rate increased by 7.2 percentage points between 1928 and 1931 while the price of traded goods fell by 27.2 percent and domestic prices by 15.2 percent. The prices of primary goods fell dramatically from an index value of 28.01 in 1928 to 10.15 in 1932 compared with 48.31 in 1920 (Grilli and Yang, 1988). The terms of trade moved very sharply against primary producers—in Latin America by about 32 percent between 1929 and 1931. Capital flows from first to third world ceased and were then reversed; capital exports of $355mn in 1929 became inflows of $1.4bn and $1.7bn in 1931 and 1932 (Maddison, 1985). Currency and debt crises followed for many third world countries.

The depression was also notable for financial crises in many advanced countries featuring significant bank failures, switches from bank money into cash and drying up of the supply of bank loans (Bernanke and James, 1991). The most spectacular of these debacles was in the United States (see Box 2.1). Here research has established that a badly regulated banking system had engaged in excessive risk taking in the boom years preceding the depression and that the interruption to supplies of credit in the financial crisis of the early 1930s added extra deflationary pressure to that expected from a conventional negative money supply shock (Calomiris, 1993).

Money wages generally proved to be very sticky downwards such that during 1931–1934 real wages were 20 to 40 percent above the 1929 level in most countries. Estimated wage adjustment equations for a panel of 22 countries confirm that wages generally reacted very sluggishly to price declines (Bernanke and Carey, 1996). This ensured that the deflationary shocks were translated into output reductions. In some cases, like that of Britain, the spread of collective bargaining arrangements and the implications of unemployment benefits in placing a floor under nominal wages may partly explain this (Crafts, 1989). The stickiness of money wages remains quite puzzling, however, especially in view of the much greater flexibility that they displayed both in Britain and the United States in the early 1920s. Calibrations of a Taylor overlapping contracts model for the United States indicate that a change to much slower wage adjustment in the early 1930s than a decade earlier is implied and the explanation of this is not yet clear (Bordo, et al., 1997).

Recent research has placed the operation of the non-cooperative interwar gold standard at the heart of the problems of the early 1930s. In the absence of cooperative monetary expansion, the fixed exchange rate regime buttressed by fiscal orthodoxy ensured that the representative country faced adverse aggregate demand shocks to which there was no effective policy remedy.

It has also become clear that leaving gold was the key to early recovery which was associated with the pursuit of independent monetary policies (which mitigated price declines and thus real wage increases) besides having more conventional implications for aggregate demand (Campa, 1990; Eichengreen and Sachs, 1985). Early devaluation also helped prevent currency crises which translated into banking crises (Grossman, 1994). At the same time, imposing capital controls, and erecting barriers to trade were seen as attractive ways to escape from conflicts of external and internal balance and, in effect, openness was increasingly seen as an obstacle to prosperity at least in the short term. Even in the hitherto staunchly free trade U.K. the macroeconomic downturn prompted the imposition of the General Tariff as long-standing protectionist interest groups seized their moment (Rooth, 1992).

If, ex post, the interwar Gold Standard turned out so badly, why ex ante did so many countries seek to return to gold after World War I? Obviously, a major reason was that the pre-1914 system had operated as an effective commitment technology for many countries (Bordo and Kydland, 1995) and the advantage of a rules based system for macroeconomic policy was highly valued—as Montague Norman said it was ‘knave proof. Recent research has also stressed the role of gold standard membership as the best available signal to capital markets of financial rectitude and thus as a means to borrowing from abroad at substantially lower interest rates (Bordo, Edelstein and Rockoff, 1999). In this way, globalization was tied to the gold standard and thus, given the international economic policy context of the early 1930s, carried with it the seeds of its own demise.

The upshot of this research is to highlight three key ingredients of the depression, namely, inappropriate international monetary and exchange rate policy, fragile banking systems, and inflexible labor markets. The chances of avoiding a replay of the 1930s will be better the less these apply in future. As noted above, the solution to the policy trilemma increasingly is to drop the fixed exchange rate and this removes a key aspect of the 1930s crisis. It might also be noted that in a world of floating exchange rates the G7 response to the stock market crash of 1987 indicates that policymakers both were alive to the lessons of history and able to take appropriate action. Inflexible labor markets seem to remain a problem, notably in Europe. The reasons for wage stickiness in the 1930s are not yet understood and are a priority for further research. At present, it is unclear whether there would now be greater downward wage flexibility in response to deflationary pressure but it would be unwise to rely on it (Akerlof et al., 1996).

The Asian crisis in 1997/8 is a reminder of the acute problems that can be created by the policy trilemma in the face of inadequately supervised and regulated banking systems. Attempts to use interest rate policy for domestic demand management with a pegged exchange rate were counterproductive in the face of capital inflows and the excessive risk taking which ensued threatened financial stability (Mishkin, 1999). In turn, as bank balance sheets deteriorated sharply, the fiscal implications of implicit government guarantees risked a currency crisis which itself would exacerbate the banking crisis (Burnside et al., 1999). Mutatis mutandis, the unholy combination of inappropriate monetary and exchange rate policy together with a fragile banking system is quite reminiscent of 1930s America. As with that episode, ex post, these errors seem to be widely recognised and improvements in the design of policy can be expected.

In sum, the message from the interwar period is that economic policymaking rather than globalization per se was the root of the Great Depression. Accordingly, there seem to be reasons to be optimistic both that a similar downturn can be averted in future and, if so, the probability of a reversal of globalization is lower.

IV. An End Of The Century Perspective

A. The Paradoxical OECD Growth Slowdown

One of the most discussed aspects of twentieth century economic growth is the productivity slowdown of the last quarter century in the OECD economies, and especially in the United States. In the context of the information and communications technology revolution, this has been seen by many as a considerable puzzle. The episode raises a number of important issues both for economic policymakers and growth economists. These include

  • Is structural change responsible for slower OECD growth?

  • Is there a post Golden Age hangover effect?

  • Is the growth slowdown sufficient to refute the endogenous growth hypothesis?

The most straightforward reason why fast growth comes to an end is that it is based on an investment boom that runs into diminishing returns and/or involves an episode of catch-up growth at the end of which productivity growth will naturally slow down. It is generally agreed that there are diminishing returns to routine physical investment and that as productivity gaps with leading countries narrow, ceteris paribus, TFP growth will decline (Temple, 1999). These are clearly facets of the growth slowdown of the 1970s in OECD countries. At that point, however, catch-up of the United States was clearly far from complete and subsequent slower TFP growth seems to reflect something more than this. And, of course, for the United States itself the catch-up effect carries no weight and greater investments in innovative activity might have led to faster growth according to some recent growth theorizing.

A striking feature of twentieth century economic growth in leading economies has been the associated structural change, as Table 3.1 reports. There has been a large decline in the proportion of employment devoted to agriculture and manufacturing and a substantial rise in the shares of both marketed and non-marketed services. This has made the measurement of economic growth more difficult and has probably exacerbated the underestimation of productivity growth. In addition, these structural shifts may also have reduced potential productivity improvement.

Table 3.1.

The Structure of Employment in Five Leading Economies, 1900–95

article image
Sources: 1950, 1973 and 1995 from O’Mahony(1999). For 1900, France: Carre et al. (1972); Germany and U.S.: Bairoch (1968); Japan: Ohkawa and Rosovsky (1973) U.K.: Feinstein (1972). Goods comprise agriculture, manufacturing, construction, mining and public utilities, market services include transport and communications, distributive trades, financial services, personal and domestic services and non-market services are government employment plus education and medical services. Except for 1900, Germany refers to West Germany.

The measurement problem should not, however, be exaggerated since the sectoral shift has been gradual and is unlikely to account for much of the recent productivity slowdown since the shift to hard to measure sectors was similar before and after 1973. Using the assumptions in Sichel (1997a), an upper bound estimate of the impact of the change in employment weights since 1973 in the economies of Table 3.1 is in no case more than 0.4 percent per year while the difference between 1900 and 1995 weights might add up to about 1 percent per year.

Table 3.2 confirms that there have been notable differences in sectoral productivity growth rates in G7 countries in the postwar period. Estimated growth of both labor productivity and TFP in manufacturing has been appreciably higher than in services. If these estimates are accurate, this appears to suggest that the marked de-industrialization of the last thirty years or so has reduced OECD growth potential. Again, however, a glance at Tables 3.1 and 3.2 suggests that the impact of de-industrialization cannot have been large compared with the within sector productivity slowdowns. An interesting question is whether technological change in the age of computers will eventually stimulate productivity growth in services and manufactures in fairly equal measure whereas in the ages of steam and electricity the impact on the latter was much greater.

Table 3.2.

Sectoral Productivity Growth Rates, 1950–95

(Percent per year)

article image
Source: Derived from O’Mahony (1999).

The person in the street might well suggest that macroeconomic shocks bring fast growth to an end. It is certainly true that severe recessions, currency and financial crises can have a severe short term impact on growth as occurred in the West both in the 1930s and 1970s. It is much less clear that there will be a long run effect and conventional macroeconomics would suggest that this should only be the case if the reaction to the crisis involves policy responses that damage the long run growth rate of productive potential. Clearly this does happen, as with the adoption of inwardly-orientated policies in Latin America in the 1930s. On the other hand, the United States in the 1930s rehabilitated its banking system, adopted an expansionary monetary policy and by the end of the 1930s had resumed its earlier strong trend growth which continued on through the 1960s (see Box 2.1).

Although these usual suspects have some relevance to the growth slowdown, a deeper underlying problem lay in the difficulty of adapting to new circumstances, i.e., in the political economy of growth. In early postwar Western Europe, many countries sought to create a “social contract” that sought to generate wage moderation in return for high investment and in which the expansion of welfare state provision was an integral part (Eichengreen, 1996). The downside of these postwar settlements emerged later on in the 1970s and 1980s, when catchup growth weakened and the macroeconomic environment turned sour. At that point, however, a form of status quo bias (Fernandez and Rodrik, 1991) seems to have meant that an implicit coalition of definite and possible losers was powerful enough to prevent reform. The bad news came in the form of growth of the public sector (de la Fuente, 1997), and inflexible labor and product markets (Koedijk and Kremers, 1996) which were detrimental to productivity performance.

An acute version of this phenomenon has emerged in Japan which has suffered a severe decline in TFP growth and has failed to fulfil the projections of catch-up made by authoritative observers in the 1970s (Denison and Chung, 1976). Disappointing Japanese TFP growth reflects excessive investment and industrial policies that have diverted resources away from high growth sectors towards declining industries and did not have a positive effect on TFP growth during 1960–1990 (Beason and Weinstein, 1996). The Japanese economy has also been heavily regulated which has been costly in terms of productivity, and has continued to have high hidden unemployment in non-tradables. The scope for TFP gains from deregulation in Japan appears to be about 6 times larger than in the United States (Blondal and Pilat, 1997).

Fast growth in Golden Age Japan was predicated on a number of institutional innovations made in the 1940s including the main bank system, the keiretsu and lifetime employment based on deferred compensation (Noguchi, 1998) geared to deliver rapid mobilization of resources based on low cost rather than efficient use of capital (Ide, 1996) and productivity growth concentrated on manufacturing while sheltered/non-tradable sectors of the economy sustained employment based on low labor productivity. By the 1990s, these features of the Japanese economy were increasingly seen as obstacles to further catch-up of the United States but very difficult politically to reform (Ito, 1996). Japan is perhaps a good example of both of the need for and the difficulties of transition from the institutions designed to achieve a Gerschenkronian escape from backwardness, see below, section 3C.

The issue of endogenous growth continues to excite economists. It seems increasingly clear that the study of (conditional) convergence based on the Summers-Heston dataset cannot provide answers. The common finding that growth is inversely related to the initial income level or labor productivity gap does not discriminate well between rival growth models even though this clearly is a prediction of the neoclassical Solow or Augmented-Solow growth models. For example, if there are obstacles to technological diffusion, then should the costs of technology transfer fall, there may be periods of catch-up even when the underlying growth process is endogenous and long run growth rates show no tendency to equalize across countries (Sala-i-Martin, 1996). In fact, this seems a possible interpretation of recent OECD experience given that time-series analysis of the Maddison dataset rejects the hypotheses that long run forecasts of differences in output per person tend to zero or are proportional with a single long term trend (Mills and Crafts, 1999).

Jones (1995) argues that long run time series evidence is indeed the key to evaluating the endogenous growth hypothesis. He stresses that the failure of OECD, and especially the United States, trend growth rates to accelerate in recent decades appears to contradict the predictions of endogenous growth models that increased investments in human capital and R&D will permanently raise the growth rate and thus also calls into question the central claim of endogenous growth, namely that good policy can raise the growth rate rather than merely have a levels effect as in the Solow model.

The growth accounting estimates discussed earlier provide a useful input to this discussion. It has already been noted that the changes over time in measured TFP growth which inform the discussion in Jones (1995) are not necessarily a good guide to the rate of technological change and so a fortiori need not be a good guide to the rate of endogenous innovation. Indeed, the TFP growth exhibited by the United States at mid-century is “too good to be true” in the sense that it would imply a phenomenal rate of return on the R&D expenditure of the time. This opens the possibility that TFP growth may have slowed down for reasons other than declining returns to innovative effort. Econometric investigation of the TFP growth slowdown in American manufacturing in the 1960s and 1970s strengthens this suspicion as it finds that the underlying rate of technological change (which accounted for less than half of conventionally measured TFP growth) did not decline but productivity improvement from economies of scale fell substantially (Morrison, 1992).

If there are diminishing returns to research effort, then increases in the share of national income devoted to R&D will raise the level of GDP per person and will have a transitory but not a permanent positive effect on the growth rate which in the long run will be determined by the exogenous natural rate, i.e., we are back in an Augmented-Solow world. This can be illustrated by the following simplified model (Jones, 1999):

Y=AσLy where A is the stock of ideas and Ly is labor used in goods production

ΔA=δLAAϕ where ϕ < 1.

In the steady state, ΔA/A and thus the growth of income per head is n/(l − ϕ) where n is the population rate and is independent of the allocation of labor to research.

Modellers have also discussed the impact of a proliferation of product varieties. Suppose ΔA/A = δLA/V where V is the variety of products available then the nature of scale effects will depend on whether the number of varieties grows more or less rapidly than population. The growth rate will depend only on allocation of labor to research and not on scale when population and variety grow at the same rate.

Jones (1997) claimed that not only is this permutation on the Augmented-Solow model the best available interpretation of the long run data on productivity growth but also, since investment in R&D and human capital have been rising relative to American GDP, the economy must have been growing above the steady state rate. He therefore projected a substantial slowdown when this transition phase is complete. A long run perspective suggests, however, that the historical experience has been variable and would indicate some caution before accepting Jones’s projection.

In particular, the impact of innovative effort on overall TFP growth seems to depend on whether it results in the invention of new general purpose technologies such as electricity in the early twentieth century (David and Wright, 1999) or unconnected real costs reductions in individual sectors as in recent decades (Harberger, 1998). In Harberger” striking metaphor the former is a yeast type growth process and the latter is a mushrooms type. Moreover, general purpose technologies are also thought by some economists to be likely to involve a productivity slowdown prior to a subsequent acceleration as a result of the adjustment costs of the new opportunities.

The information and communications technology revolution is clearly the big technological development of the late twentieth century. At least until quite recently, however, it has not had dramatic effects on either labor productivity or TFP growth and economists’ assessments of its likely impact have tended to be sceptical (Gordon, 1999). Growth accounting estimates suggest that this is because computers remain a relatively small, although fast-growing, component of the capital stock and because there appears to be relatively little evidence of strong spillover effects (Jorgenson and Stiroh, 1999)(see Box 3.1). This will not surprise new economic historians since a staple finding in the cliometric literature is that even dramatic changes in technology like railways or the steam engine have modest effects on overall GDP growth, especially in the early days (Fogel, 1964; von Tunzelmann, 1978).

Nevertheless, the long run TFP implications of changes in technology are often unclear initially. Thus, while American productivity surged in the 1920s when electrification was developed as a general purpose technology this was fully 40 years after electricity was first commercially generated (David, 1991). On this analogy, it is too soon to assess the productivity implications of the computer revolution. Moreover, it seems likely that well-informed observers at the end of either the eighteenth or nineteenth centuries would have failed to predict subsequent productivity improvement given the very limited implications of steam and electricity at those junctures. If, however, substantial yeast-like implications for TFP growth are to come from the computer revolution, it will need to progress to a new phase such as a massive increase in teleworking that has substantial implications for capital savings in terms of transport infrastructure and commercial buildings (David and Wright, 1999).

The bottom line of this review of the OECD productivity slowdown is that only a small part seems to be due to the impact of structural change including associated increases in measurement error. The end of the Golden Age left a legacy of much expanded public sectors which probably has reduced growth rates somewhat, see Section B below. The modest impact of the ICT revolution on economic growth thus far is not surprising from a growth accounting perspective. The growth slowdown is paradoxical in the light of some endogenous growth theorizing and it remains unclear whether the long run rate of productivity improvment from technological progress will respond positively to more innovative effort. Depending on one’s prior about scale effects in innovation, a case can be made that OECD growth in the first quarter of the next century is about to speed up as the delayed effects of greater spending on R&D feed through or slow down as transitional dynamics unwind.

B. The Expansion of the Public Sector in the OECD

To the Victorians perhaps the most surprising feature of late twentieth century economies would be the very large share of government spending, especially on transfer payments, in national income in the rich countries. Two questions need to be addressed:

  • Has the rise of the public sector now come to an end?

  • Is the growth of public spending an important reason for slower OECD growth?

Table 3.3 reports the rise of public spending in a number of OECD economies since 1870. This shows that a major difference before and after World War II was that in the latter period government expenditure as a share of GDP expanded primarily due to increased social transfers which had been negligible in late Victorian times. The period of the most rapid growth in this aggregate was generally from the 1940s through the 1970s and in some countries there has been a noticeable retreat from a peak reached in the early 1980s. Understanding why public expenditure has grown over the long run may be helpful in projecting its future trajectory.

Table 3.3.

Government Expenditures/GDP, 1870-1998

(Percent)

article image
Sources: total outlays from Tanzi and Schuknecht (1997) updated using OECD (1999); social transfers, defined to include spending on pensions, welfare, unemployment compensation and health by national and local governments, from Lindert (1994)(1996) updated using Oxlcy and MacFarlan (1995) and OECD (1997).

Central government only through 1937.