(cid:1)(cid:2)(cid:3)(cid:4)(cid:5)(cid:6)(cid:3)(cid:3)(cid:7)(cid:8)(cid:9)(cid:8)(cid:10)(cid:6)(cid:3)(cid:11)(cid:7)(cid:10)(cid:12)(cid:5)(cid:13)(cid:7)(cid:14)(cid:15)(cid:16)(cid:6)(cid:3)(cid:7)(cid:15)(cid:5)(cid:17)(cid:7)(cid:18)(cid:19)(cid:15)(cid:3)(cid:6)(cid:3)(cid:7)(cid:12)(cid:20)(cid:7)(cid:8)(cid:15)(cid:18)(cid:4)(cid:21)(cid:15)(cid:10)(cid:4)(cid:3)(cid:21) (cid:17)(cid:6)(cid:16)(cid:6)(cid:10)(cid:12)(cid:18)(cid:22)(cid:6)(cid:5)(cid:21) (cid:23)(cid:24) (cid:15)(cid:25)(cid:26)(cid:27)(cid:28)(cid:7)(cid:22)(cid:29)(cid:30)(cid:30)(cid:31)(cid:28) (cid:25) (abbreviated version of chapter 4 of A. Maddison, Dynamic Forces in Capitalist Development, Oxford University Press, 1991) It is clear from the preceding analysis that the process of capitalist development has not been smooth. There have been distinct and important phases of development which are worthy of study, definition, and causal interpretation. I distinguish four phases, which I shall describe later, covering periods of unequal length; 1820-1913, 1913-50, 1950-73, and 1973 onwards. There have also been shorter-term fluctuations, usually called business cycles. My primary interest is not in these, but in major changes in trend which are distinguished from each other by changes in the institutional-policy mix and usually initiated by some sort of ’system shock’ which upsets established patterns of international intercourse. Before presenting my own diagnosis, it is useful to trace the history of cyclical or wave analysis, because my quantitative empirical approach is not the only one available. In the past there have been a number of theories concerning the nature of long waves in economic activity. These were revived and augmented in the 1970s after a period when even the business cycle was considered obsolete and the long-wave hypothesis was regarded as quaint.1 The unfortunate thing about revivalist approaches to new problems is that the adherents are often single-minded enthusiasts, so that the analytical apparatus of the old theories is rehabilitated in toto in spite of remediable weaknesses. (cid:8)(cid:24)!"#(cid:7)(cid:15)(cid:25)(cid:29)"(cid:24)(cid:28)(cid:31)(cid:28) Cyclical analysis for the capitalist period started with Clement Juglar in 1856. He emphasized periodicity in economic activity whereas most earlier writers had tended to interpret interruptions to growth as random financial crises. Juglar also believed that cycles were roughly synchronous in France, the UK, and USA.2 In his major work on cycles his attention was mainly concentrated on monetary phenomena - expansions or contractions in central bank activity, rates of interest, prices of key commodities, etc., plus narrative ’business annal’ material. Although it is frequently asserted that Juglar found cycles of a characteristic length of nine years, this is not in fact true. His cycles for France average seven years with a range from three to eighteen years, and for the UK six years with a range from two to ten years. 1 For several decades the quantitative indicators available to cyclical analysts were similar to those used by Juglar, though they were later augmented to include price indices, and data on output and foreign trade. A more sophisticated causal analysis was also developed, such as one finds in the study by the Russian economist Tugan-Baranowsky on the nineteenth-century cycle in the UK.3 The ultimate refinement in statistical analysis of business cycles was the massive effort of the National Bureau of Economic Research (NBER) in the USA. The first phase was a comprehensive collection of narrative data stretching back to the beginning of the nineteenth century with a cyclical periodization for seventeen countries. The second phase was publication of a series of reference cycles for four countries (France, Germany, Great Britain, and the USA) based mainly on monthly quantitative data, which start in 1854 for the last two countries, in 1865 for France, and in 1879 for Germany.4 The number of monthly series for the USA was nineteen for 1860 rising to 811 in 1942 (plus 161 annual indicators). The NBER derived its ’reference’ cycles by plotting most of this information in de-seasonalized form, and by iterative procedures of inspection, deriving a cluster of roughly concurrent fluctuations. Thus its central concept of economic activity was a somewhat fuzzy cocktail rather than a clearly defined measure of aggregate economic activity. Its main use was as a sensitive warning indicator of turning points in business activity, with indicators classified as leading, coincident, or lagging. The reference cycle has become part of the official statistical armoury of the USA for forecasting purposes, though it is of course supplemented by the more articulate short-term models on which other countries place main reliance. For the period 1857 to 1978 the NBER established twenty-eight successive peak-to-trough movements for the United States, giving a recession on average every four years, with a variation from two-and-a-half to nine-and-a-half years. For other countries the average duration was found to be longer: fifty-three months for France, sixty-two for the UK and sixty-four for Germany for prewar years. The NBER cycles are not adjusted to eliminate trend, so they are not measures of oscillation in economic actvity, and register recessions only when there is an absolute fall in the relevant indicators.5 However, the NBER technique of using monthly and rather volatile series does pick up more cycles than would a GDP index based on annual data, and those reference cycles that do correspond with GDP movements do not always have exactly the same dates.6 The NBER approach is a useful tool in interpreting quantitative economic history, but a major problem is that it yields no satisfactory measure of the amplitude of fluctuations because of the difficulty of producing a meaningful summary measure from such heterogeneous data. Thus one cannot use the reference cycle itself to distinguish major and minor cycles, in the same way that one can with simpler measures of industrial output or GDP fluctuations. 2 Hence my preference is for rather simple measures of annual movements in aggregate activity, which reveal clearly the big changes in the severity of recessions that have appeared systematically across our sixteen countries in the past century, as illustrated in Table 1. This table shows that peacetime business cycle history has been much milder since the Second World War than before, and that the 1920-38 period was generally much worse than 1870-1913. Except in 1929-33, when the Depression hit every country, the weighted average of cyclical movements for the sixteen countries as a group is dampened by the fact that individual country cycles are not synchronized. Before 1870 data on annual changes in GDP are not available for most countries, but it would seem that average cyclical experience was not too different from that of 1870- 1913.7 Table 2 shows the cyclical record for foreign trade. It confirms the pattern shown by GDP movements, with notably smaller cycles since the Second World War. Graphs 4.1 and 4.2 are intended to show both cyclical volatility and differences in growth trends in different phases of capitalist development. The striking thing in both graphs is the great volatility of the 1913-50 period, and the markedly faster growth since 1950. (cid:10) (cid:25)(cid:26)$(cid:14)(cid:29)%#(cid:7)(cid:15)(cid:25)(cid:29)"(cid:24)(cid:28)(cid:31)(cid:28) Although cyclical analysts had made distinctions between big and small recessions, and there had been some discussion of the Great Depression (in prices) in the last quarter of the nineteenth century, it is significant that the idea of recurrent long waves in capitalist development did not emerge until the First World War, i.e., about fifty years later than cycle analysis, and only after the rhythm of development had been very dramatically broken. The main figures in long-wave analysis are N.D. Kondratieff, S. Kuznets, and J.A. Schumpeter. All of them drew heavily on cyclical-type indicators to test their ideas quantitatively. 3 TABLE 1 (cid:15)&’"(cid:31)((cid:27)(cid:30)#(cid:7) )(cid:7)*#!#(cid:28)(cid:28)(cid:31) (cid:25)(cid:28)(cid:7)(cid:31)(cid:25)(cid:7)(cid:15)(cid:26)(cid:26)+#(cid:26)(cid:29)(#(cid:7)(cid:12)(cid:27)(’(cid:27)((cid:11)(cid:7),-./$,0-01 (cid:22)(cid:29)2(cid:31)&(cid:27)&(cid:7)(cid:18)#(cid:29)3$(cid:21)+ (cid:27)(cid:26)4(cid:7)(cid:20)(cid:29)""(cid:7)(cid:31)(cid:25)(cid:7)(cid:13)(cid:17)(cid:18)(cid:7) +(cid:7)(cid:10) 5#(cid:28)((cid:7)*(cid:31)(cid:28)# (annual data) 1870-1913 1920-38 1950-73 1973-89 Australia -17.1 -12.8 +0.9 +0.2 Austria - 2.3 -22.4 +0.1 -0.4 Belgium - 0.2 - 7.9 -0.8 -1.5 Canada - 7.7 -29.6 -0.7 -3.2 Denmark - 2.7 - 2.9 -0.7 -1.6 Finland - 4.2 - 4.0 +0.5 +0.1 France - 6.5 -14.7 +2.5 -0.3 Germany - 3.2 -16.9 -0.1 -1.6 Italy - 5.1 - 6.1 +1.6 -2.6 a Japan - 7.4 - 7.3 +4.3 -1.2 Netherlands (n.a.) - 9.5 -0.3 -2.1 Norway - 3.0 - 8.3 -0.9 +0.3 Sweden - 5.5 - 6.2 -0.2 -1.6 Switzerland (n.a.) - 8.0 -2.1 -8.6 UK - 4.1 - 8.1 -0.2 -3.5 USA - 8.2 -29.6 -1.4 -2.6 Arithmetic average - 5.5 -12.1 +0.2 -1.8 a) 1885-1913. Source: Appendix A of Maddison (1991). 6 (cid:25)(cid:30)+(cid:29)((cid:31)#)) Kondratieff was a Russian economist, whose work on long waves was done in the 1920s as director of the Businss Cycle Research Institute in Moscow. He distinguished three kinds of cycles: long ones of fifty years’ duration, middle ones of seven to ten years’, and short ones of three to four years’. He measured long waves by a double decomposition of time series - eliminating the trend and showing the deviations from it smoothed with a nine-year moving average. The nine-year average was enough to remove the influence of the two shorter types of cycle. His analysis covers the period 1770 to the 1920s and his long cycles fall into a range of forty to sixty years.8 4 TABLE 2 Amplitude of Recessions in Exports, 1870-1989 Maximum Peak-Trough Fall or Smallest Annual Rise in Export Volume (Annual Data) 1870-1913 1920-38 1950-73 1973-89 Australia -32.2 -19.7 - 7.6 - 3.6 Austria (n.a.) -48.7 - 7.3 - 5.5 Belgium -13.1 -31.8 - 9.6 - 6.8 Canada -13.9 -40.6 - 6.1 -10.3 Denmark -25.0 -20.3 - 6.7 - 4.1 Finland -20.9 -15.7 -13.4 -17.4 France -12.9 -47.3 -12.0 - 4.0 Germany -14.2 -50.1 + 2.3 -11.5 Italy -30.6 -69.1 - 9.1 - 8.5 Japan -23.7 -18.9 0.0 - 2.3 Netherlands (n.a.) -33.4 + 2.6 - 3.8 Norway - 7.7 -16.0 - 7.1 - 3.3 Sweden -11.0 -37.0 -10.7 -11.6 Switzerland (n.a.) -50.2 - 4.2 - 8.2 UK -12.5 -37.7 - 8.0 - 2.2 USA -18.9 -47.8 -14.3 -18.8 Arithmetic average -18.2 -36.5 - 7.0 - 7.6 Source: Appendix F of Maddison (1991). Kondratieff’s thesis was most clearly demonstrated by long-term movements in wholesale prices, where long waves were discernible without trend adjustment, though some of the long-term oscillation was obviously attributable to wars (e.g., the peaks in the Napoleonic wars and 1914-20). He analysed wholesale price developments for France, the UK, and the USA, and it is not surprising that in these relatively open economies he found that price trends were similar in the different countries, particularly as he adjusted the price indices to eliminated the effect of exchange rate changes which gives the series greater resemblance.9 On this basis Kondratrieff claimed his waves to be an international phenomenon. Most of Kondratieff’s other indicators contain a strong price element, because they are expressed in current values: e.g., wages, interest rates, the value of foreign trade, and bank deposits. Not surprisingly, the price component of these value series moves in the same way as the general price indices, so this evidence for his wave theory is not in fact independent of his first offering. 5 The only physical series in Kondratieff’s repertoire in his most famous article are those relating to per capita10 coal production in England (and coal consumption in France), and to pig iron and lead production in England. Here, as with his value indicators, he presents data from which the trend has been removed. There are some distinct oddities about Kondratieff’s presentation of the four physical indicators, which at first sight seem to contain long waves of large amplitude with a fair degree of synchronization. His charts for the physical indicators are shown as absolute deviations from trend. Thus he shows UK coal production 186 points above trend in 1869, 245 points below in 1894 and 164 points above in 1910. But in proportionate terms the deviations are much smaller: 5.4 per cent, -5.1 per cent, and 2.8 per cent respectively. He also follows the highly questionable practice of juxtaposing two series on the same graph to suggest that the amplitude of their movement is similar, this effect being secured by using quite different scales for each. From this graph it appears that UK coal output is more volatile than French coal consumption, whereas the proportionate swings in France were bigger than in the UK. Worse problems arise in his graph for British pig iron and lead production, because there is the further complication that he there compares two series with totally different trends. Pig iron output rose about four-fold over the period he covered, and lead output fell to less than a tenth of its original level.11 Kondratieff concluded tentatively that there had been three long waves in economic ’life’ (a rather vague term, but one that is clearly intended to include output as well as price movements). His chronology refers not to particular years but to spans, and he distinguishes only two phases, the rise and fall, in each wave. He does not discuss the amplitudes of these waves, which vary between series, but they are clearly considered large enough to exclude the need for discussion of growth trends. His dating is as in Table 3. There are several problems with Kondratieff’s approach. The first is his failure to establish that long waves exist as more than a monetary phenomenon. He fails to show the existence of broad movements in the volume of output that even remotely correspond to our present measures of aggregate economic activity. The second problem is that the trend is taken out and discarded as if it were irrelevant to the discussion. TABLE 3 6 (cid:25)(cid:30)+(cid:29)((cid:31)#))7(cid:28)(cid:7)(cid:10) (cid:25)(cid:26)$5(cid:29)%#(cid:7)(cid:8)4+ (cid:25) " (cid:26)(cid:24) Rise Decline 1. First long wave 1780s-90s to 1810-17 1810-17 to 1844-51 2. Second long wave 1844-51 to 1870-5 1870-5 to 1890-6 3. Third long wave 1890-6 to 1914-20 1914-20 to ? 6 . Thus, in comparing UK and US growth between 1820 and 1989, one finds British GDP has risen about twenty-seven-fold, and American by more than 450-fold. This fact is left out when the time series are decomposed for wave analysis, but such very different trends transform the nature and operational significance of any long waves that may be discerned. The third problem is that double decomposition of time series to eliminate trend and smooth out cycles blurs the impact of major historical events. Thus, Kondratieff’s chronology pays no attention to the impact of the First World War, and later long-wave analysts tend to brush off the catastrophic 1929-33 recession and the Second World War as well. Finally, Kondratieff failed to offset these empirical shortcomings by giving plausible causal explanations as to why capitalist development should involve long waves as a systematic phenomenon. In the USSR this problem involved Kondratieff in ideological difficulties because his wave theory seemed to conflict with the more fundamental Marxist expectation of the ultimate breakdown of capitalism.12 There is no doubt that Kondratieff’s contribution to long-wave analysis was fundamental in spite of its weakness,13 because he fully adumbrates the three-cycle schema later developed by Schumpeter, and his statistical technique was the same that Kuznets later used to distinguish ’secondary secular movements’. Furthermore, he pointed to the likelihood of poor terms of trade for agriculture in periods of decelerated development - a point given major stress later by Arthur Lewis. 6(cid:27)8(cid:25)#((cid:28)(cid:7)(cid:29)(cid:25)(cid:30)(cid:7)(cid:15)(cid:23)+(cid:29)& %(cid:31)(8 Chronologically, the next development in long-wave analysis was Kuznets’s work on ’secondary secular movements’, published in 1930.14 Kuznets’s basic technique for identifying long waves was the same as Kondratieff’s, i.e. to look at smoothed detrended series, though Kuznets made a special point of not eliminating population movements. His investigation was more detailed, involving careful analysis of fifty-nine series, most of which represented both physical output and the relevant price variance for particular commodities.15 Kuznets did not claim that these indicators could be added to provide a meaningful picture of aggregate economic activity, and he did not use aggregative indicators for sectors such as agricultural or industrial production, which were available when he wrote. His major conclusions are: (1) that ’secondary secular variations in production are in most cases similar to those in prices, the latter following a rather general course in agreement with the well-known historical periods of the rise and fall in the general price level’ (p. 197); (2) he found a much shorter periodicity than Kondratieff, ’about 22 years as the duration of a complete swing for production and 23 years for prices’ (p. 206); 7 (3) most fundamentally, he did not think there was enough evidence to conclude that these secondary secular variations were major cycles. They were ’rather specific, historical occurences’ (p. 258). There is ’an absence of factors that would explain the periodicity’ (p. 264). Kuznets did not attempt to cluster his individual series to present a global chronology of long waves in economic life, nor did he analyse the synchronization of the series.16 However, it is clear from other evidence that in the period Kuznets covered there were rather large depressions in the USA at intervals of fifteen to twenty years. This is directly observable in indices of industrial production (including construction), which Arthur Lewis has prepared (see Table 4). It is also clear that the recession/depression sequence was different in France, Germany, and the UK, which is the major reason why the aggregate performance of these four countries (which Lewis calls ’the core’, to distinguish them from ’the periphery’ - the rest of the world) is more stable than they are individually. In comparing the cyclical performance of these countries, it is useful to keep in mind the differences in their long-run growth performance. A country like France or the UK, with slow growth, is likely to have more small recessions than Germany or the USA, which had much higher growth. A rough measure of how far recessions fell below the potential growth path is to combine the trend and the cyclical amplitude: e.g., the average French recession involved a fall of 6.7 per cent from trend (4.1 + 2.6 per cent), and the average German recession a fall of 7.5 per cent from trend (3.2 + 4.3 per cent). After his early study of secondary secular movements, Kuznets moved on to fundamental definitional work on the rationale (scope, valuation, and net-ness) for GDP as an aggregate economic indicator within a system of national accounts, and produced historical estimates of US economic development which made it possible to analyse long-term movements in economic life on a much more satisfactory conceptual basis than the cocktail approach that virtually all economic analysts had previously been forced to use. Furthermore, Kuznets successfully stimulated and inspired replication of his work by scholars in many other countries. This accounting approach still has some drawbacks for cyclical analysis, because until recently data were available only on an annual basis, but it has revolutionized the study of growth and greatly facilitates testing of long-wave analysis. From time to time after 1930 Kuznets returned to long-swing analysis in a rather tentative way. Unlike his disciples, he himself never called them ’cycles’, as the word implies greater certainty about such phenomena and their periodicity than Kuznets concedes. In 1956 he did advance a tentative chronology of long swings in GDP, for eight countries,17 but the periodization looks very odd, because the logic of the analysis calls for a declining phase in the decades following 1946, and Kuznets later dropped this one attempt to suggest a general chronology for long swings. 8 TABLE 4 Amplitude and Duration of Cycles in Industrial Production (including Construction), 1870-1913 Peak year Trough year Percentage Duration of amplitude of recession peak trough (years movement below peak) France 1878 1879 - 0.7 1 1882 1885 -10.6 6 1892 1893 - 3.4 1 1894 1895 - 4.4 1 1899 1902 - 7.9 5 1907 1908 - 1.4 1 1909 1910 - 2.5 1 1912 1913 - 2.1 1 Average amplitude -4.1 Percentage of years 39.5 below peak Trend growth rate 2.6 Germany 1873 1874 - 0.5 1 1876 1877 - 5.7 2 Average amplitude - 3.2 Percentage of years 7.0 below peak Trend growth rate 4.3 UK 1876 1879 - 4.1 3 1883 1886 - 9.7 4 1891 1893 - 6.3 3 1902 1903 - 2.1 2 1907 1908 - 8.0 4 Average amplitude - 6.0 Percentage of years 37.2 below peak Trend growth rate 2.1 9 TABLE 4 (continued) Amplitude and Duration of Cycles in Industrial Production (including Construction), 1870-1913 Peak year Trough year Percentage Duration of amplitude of recession peak trough (years movement below peak) USA 1872 1876 -14.8 6 1883 1885 - 6.0 2 1890 1891 - 1.5 1 1892 1894 -15.9 2 1885 1896 - 7.1 1 1903 1904 - 6.3 1 1906 1908 -16.7 2 1910 1911 - 4.2 1 Average amplitude 9.1 Percentage of years 37.2 below peak Trend growth rate 4.7 Four 1873 1874 - 1.2 2 countries combined = 1876 1877 - 0.4 1 core 1883 1885 - 5.0 2 1892 1893 - 5.2 2 1899 1900 - 0.3 1 1903 1904 - 2.1 1 1907 1908 - 9.5 1 Average amplitude - 3.4 Percentage of years 23.3 below peak Trend growth rate 3.6 Source: Maddison (1991) Kuznets’s work on long swings was only a small part of his output and was concentrated on US experience. His most affirmative article was a 1958 study on population growth,18 which found the ’long-swing’ hypothesis most plausible in relation to US population growth and to ’population-sensitive’ components of capital formation such as housing and railway construction. It was applicable in weaker and sometimes inverse form in other national accounting aggregates. 10