CHRISTOPHER DOW ON MAJOR RECESSIONS [*].

Artis, Michael ; Flemming, John ; Matthews, Robin 等

The National Institute had planned a seminar to mark the launch of Christopher Dow's new book Major Recessions. The seminar had to be cancelled, because Christopher Dow died a few days before it was due to be held. As some kind of substitute, the CLARE Group decided to publish this review article. The sections of the article have different authors, but they have been designed to be complementary in scope and the authors have discussed them with each other and with other members of the Group, especially A.J.C. Britton, W.J. Carlin, C.A.E. Goodhart, P.M. Oppenheimer, M.V. Posner and J.R. Sargent. The Review is pleased to give hospitality to GLARE Group articles, but is not necessarily in agreement with the views expressed.

Robin Matthews starts with some comments on the style in which, and background from which, Dow wrote this work and which concentrates on five major British recessions of the 20th century. He presents five of Dow's more general conclusions with comments. This section is followed by a review by Mike Artis of Dow's analysis of the five major recessions (1920-1, 1929-33, 1973-5, 1979-82 and 1989-92) regarded as episodes in economic history. One of Dow's main theses is that major recessions permanently reduce productive capacity. Martin Weale examines statistically his suggestion that they reduce demand by consumers whose income expectations adjust in that way -- for which he finds little support. John Flemming addresses the 25 'golden' postwar years without a major recession presenting a critical analysis of Dow's method of distinguishing between major and minor recessions. Matthews and Flemming consider his treatment of the question whether major recessions are inevitable, how they might be avoided, and how thei r effects might be mitigated. While hankering for postwar controls Dow's Keynesianism is subtle. Finally Matthews suggests an analogy with Schumpeter's Business Cycles which survived a critical review by Kuznets, Dow's treatment of his data being much more thoughtful than Schumpeter's even if some parts of his model seem rather shaky.

Introduction: Robin Matthews

Some general characteristics of the book

Christopher Dow did not start work on the book until he was over 70, and he was approaching his 82nd birthday by the time it was completed. The style is entirely characteristic of the author and gives no hint of declining powers. Those who knew him will be able to hear his voice on every page. He is very conscious of the difficulty of the subject and the intractability of the material ("on some big questions I have changed my mind several times"). This does not prevent him from writing with serene self-assurance.

He is unconcerned by many prevailing fashions of thought. He is not ignorant of them, though perhaps he has not felt an obligation to acquaint himself comprehensively with writings that are based (in his opinion) on fundamentally erroneous premises. Real Business Cycle theory gets short shrift; so does Ricardian equivalence. As to econometrics, since there were only five major recessions in the UK to consider since World War I, "it is not possible to undertake the rigorous testing to which econometrics aspires ... It is not clear how great this loss is. Econometric procedures have proved much less valuable than most of us hoped a generation ago". His own preferred method is "quantitative-historical": the book contains a large number of well-designed charts and tables, and a strong admixture of theory.

Christopher Dow's career before he retired was predominantly in the public service -- in the Treasury, at the OECD, and in the Bank of England. He was close to the processes of public decision-making for much of the post-World War II period. Although he was one of the country's most eminent macroeconomists, he never held a university appointment. This background is reflected in the purposes that he sets himself in this book. He seeks to reach conclusions that are at the same time intellectually defensible and practically useful. This may sometimes mean making the best of a bad job, given the nature of the subject and the evidence available. He is not interested in deductive model-building as an end in itself.

A review article like this cannot cover all the rich contents of the book's 446 pages. We hope that we may stimulate readers to read the book itself. They must be prepared to find much greater emphasis than is customary in modern macroeconomics writing on demand-side explanations of the level and trend of real GDP, though the effects of demand are represented as working largely through their effect on the supply side. Readers must also be prepared for some other differences from the way that most economists are accustomed to address professional colleagues nowadays. Dow does not find it surprising that there should be large differences between individual major recessions and between countries. He allows himself a flexibility in interpretation that may be historically defensible but does sometimes come near to obscuring the central message.

It would be a poor tribute to a serious scholar to review his work entirely en couleur de rose. What follows includes a number of criticisms. These should not be taken to imply a negative view of the work as a whole.

The reference to the "five major recessions" indicates a couple of characteristics of the book.

World War I and World War II between them lasted ten years; but political historians may for many purposes find it convenient to treat them as two episodes, rather than ten. In the same way Dow is in general disposed to think of the five major recessions as five episodes, although he does have quite a lot to say about some individual years. He certainly does not view each year, let alone each quarter, as a sample drawn at random from some supposed universe of years or quarters. This partly explains why he finds econometrics unhelpful.

The reference to the five major recessions also indicates a limitation of the book. Its subtitle is "Britain and the World". But it is mainly a book about the UK. It does contain an extensive treatment of the Great Depression in America in the 1930s, and it also has some interesting summary charts and tables relating to other industrial countries. These countries are not neglected, but they are not treated on at all the same scale as the UK. There is not much discussion about the characteristics of major recessions in individual Continental countries or Japan. This is a little disappointing, in view of the author's experience as chief economist at the OECD. The treatment of the pre-World War I period is even more summary. The book's orientation towards the UK since World War I limits the generality of its conclusions on major recessions. Of course, it is easy to see that to have done otherwise would have increased the book's scope intolerably.

Five leading conclusions, and comments on them

In what follows, the passages in italics are intended to describe Dow's own views.

Conclusion 1. Major recessions, defined for practical purposes as occasions when there was a sizeable absolute fall in output measured from one calendar year to the next, have occurred at irregular intervals, not in any cyclical or other regular pattern. In the UK there have been five since World War I: 1920-1, 1929-33, 1973-5, 1979-82, 1989-92. Smaller recessions occurred between World War II and 1973 and also before World War I. In the period between 1856 (when UK national income statistics begin) and World War I, only the years 1891-93 and 1908 might just qualify as major recessions.

It is not only in severity that major recessions differ from smaller ones. In the first place, major recessions have no tendency to be self-correcting in the absence of upward shocks. Thus the level of unemployment remained much higher in 1929 than in 1920 or 1913; it likewise remained much higher in 1989 than in 1979. In the second place, the duration of the period between the bottom of a major recession and the peak of the next boom has typically been much longer than the duration of the preceding downswing, making the whole "cycle" from peak to peak much longer than in minor fluctuations.

Comments on Conclusion 1. Both the latter two points (non-self-correction and asymmetrical duration) will become clearer after we have discussed Conclusion 4.

It seems rather surprising that all recessions should be capable of being categorised without hesitation as major or minor. Such a sharp dichotomy has not been part of standard business cycle literature. It corresponds well to British experience since World War I, especially the period since World War II. It was obviously designed to fit it. It is not so obvious that it would hold more generally.

A certain amount depends on the definition of "major recession

We are accustomed nowadays to think of business cycles in terms of GDP at constant prices. It is true that 19th century business cycles were manifested more in the price level and in profits than in output. But there were large movements in output in at least some sectors. When Jevons was writing in the 1870s, the number of bricks produced was one of his criteria. In any case, it is difficult to believe that the fairly regular commercial fluctuations that are observable back to at least the 1820s were not a horse of the same colour.

Questions of definition arise even if we confine ourselves to GDP at constant prices. Dow's definition cited above was in terms of an absolute fall in national output over at least a year. But this is intended by Dow as no more than a convenient approximation. The true measure of the magnitude of a recession must, as he recognises [p.2], [1] be the extent of the shortfall at the trough of actual output below its trend. It is therefore affected by what the trend rate of growth is. To take an example at random: it would be absurd to deny that it would have constituted a major setback if Japan had experienced two years in a row with stationary output in the 1960s, since its trend growth rate at that time was over 10 per cent per year. Moreover, it is not only between countries that there have been differences in trend rates of growth. There have also been differences between periods in individual countries, including the UK. In the light of the high average growth rate enjoyed by this country in 1950-73, maybe t he amplitude of the minor recessions then was not so trivial. (The measure of the amplitude of recessions in that period might be further increased if one measured the trend growth rate in the way preferred by Dow (see below), namely by the rate of growth of capacity during the upswing, instead of by more conventional measures.)

Conclusion 2. Major recessions are caused by unexpected shocks on the demand side, amplified to an important extent by resulting changes in the confidence of firms and consumers (Dow prefers the term "confidence" to "expectations", in recognition that the expectations of economic agents are vague and not single-valued). "Shocks" refers to identifiable major disturbances, capable of being given a name, not to a multiplicity of small random influences capable of sustaining fluctuations that would otherwise be antidamped or even non-existent, in the manner of Slutsky. Possible sources of identifiable major shocks are changes in the demand for exports or in the terms of trade and changes in government fiscal or monetary policy. Changes in confidence can even be a prime mover, not just a source of amplification. Wars are a special kind of shock: they have been important historically, but the level of capacity after a major war is difficult to judge.

Comments on Conclusion 2. Lack of confidence in the future level of demand would normally be thought of as amplification of the multiplier-accelerator interaction.

However Dow does not make such a distinction between the behaviour of firms and of households as the distinction between multiplier and acceleration principle would suggest. He is unsympathetic to the acceleration principle or to any reformulation of it that would make investment a function of the relation between demand and the capital stock. Confidence about the future, rather than memory of the past, is what matters, and the size of the stock of capital inherited from the past is not of prime importance, since much of it will be old and out of date. Hence investment and consumption both end up as functions simply of the level of confidence (though not necessarily with the same coefficients). The one exception allowed is that Dow agrees that investment in real property has on a number of occasions been depressed by the overhang of property inherited from a past over-optimistic construction boom, especially in the case of office buildings. It is not entirely clear why Dow repudiates the idea that similar eff ects might follow from past high investment in other sectors, e.g. the utilities sector in 1950-73. [2]

Major depressions, according to Dow, are more likely than other downturns to be characterised not only by severe shocks but also by severe loss of confidence, in the minds of both producers and consumers. The third section of this review offers one way of testing this hypothesis insofar as it relates to consumers. The conclusions are not very favourable, but the methods used are not exactly in the Dow spirit, and he might not have accepted them.

Conclusion 3. This conclusion is put forward more tentatively, but it is important. It provides the link between demand and supply.

Major recessions bring about not only an immediate fall in output but also a permanent downward shift in the level of capacity. Capacity output (equivalent to productive potential in other writers' language) may be understood as the maximum amount that can be produced in the Marshallian short period (my phrasing, nor Dow's). The growth of capacity incorporates the effects of both capital investment and innovation. It can be measured by the growth of output between two dates, minus any part that needs to be credited to increased employment. Major recessions reduce capacity through the premature scrapping of capital, the rusting of skills, and the destruction of the internal and external organisational capital of businesses, which is likely to have taken time and trouble to build up. [3]

Comment on Conclusion 3. Dow is not the first writer to postulate that demand affects supply.

The idea is obviously reminiscent of the literature on downward ratchets and hysteresis, though that literature has usually been more concerned with the labour market and inflation than with productive capacity. One peculiarity of Dow's treatment is his belief that in minor recessions the effect is not felt at all (as opposed to being felt to a smaller extent).

Another analogous belief, to which Dow does not subscribe, is that unusually strong demand has a permanent effect in raising the growth rate. This doctrine was put forward by Kaldor (1966, 1967) a generation ago, on the basis of the observation, dubbed by him Verdoorn's law, that there is a correlation between the rate of growth of production and the rate of growth of productivity. Kaldor regarded the rate of growth of production (determined by demand) as the independent variable in this relationship, the opposite of what would be the more usual view. Kaldor was thinking of a steady, perhaps linear, effect of demand pressure on the rate of growth of capacity, rather than of a once-for-all jump.

Dow prefers his own formulation to these and other al ternatives, not mainly on a priori grounds, but because he thinks it corresponds better to experience in the period he is studying.

Conclusion 4. Dow categorises years into three classes: they may show recession, trend growth, or fast growth (i.e. above trend). His assignment of historical years to these categories is shown in Table 1. The supposed sequence of events through which this comes about, and the effect of major recessions on long-term growth, may be seen from Chart 1. This is adapted from Dow (1998, p. 375) and it gives a schematic representation of the effects of two imaginary successive major recessions. The dashed line denotes actual output; the solid line capacity (productive potential), and the dotted lines the hypothetical trends of capacity in the absence respectively of the first major recession and of the second one. The vertical axis is on a log scale.

A demand-side shock and the associated blow to confidence cause output to fall (say, from A to B'). This leads to a fall in productive potential, for the reasons given under Conclusion 3. The fall in productive potential is rather smaller than the fall in actual output (the difference in Chart 1 being the distance from B to B'). There is a tendency for productive potential and also demand to rise at the trend rate (indicated in Chart 1 by the slope of BC). Once the recession is over, there will be a tendency for growth to be resumed at this rate. But there will be no tendency for the ground lost by the loss of productive potential to be made up. Nor will there be any automatic tendency for growth to take place at the higher rate required to eliminate the non-utilisation or underutilisation of resources created by the recession. However it is likely that there will gradually be a recovery in confidence, and when that happens it will permit output for a while to grow at faster than the trend rate. There is no u niformity about when this recovery of confidence will take place: early, late, continuously or not at all (compare Table 1). Moreover the economy may or may not reach full utilisation of potential before another major recession arrives. (Some of these different possibilities are illustrated by the differences between the aftermaths of the two recessions in Chart 1.)

The trend rate of growth (in Chart 1 the slope of BC or DE) is somewhat lower than the rate of growth of actual output over the cyclical upswings. On the other hand it is higher than conventional measures of the trend rate of growth, such as the peak-to-peak, because the latter is depressed by the once-for-all contraction in capacity (between A and B). The occurrence of a major recession has a permanently depressing effect on the level of capacity and hence on the rate of growth of output: demand affects supply. The effects of one single major recession on the long-run growth rate will be watered down by the passage of time. But repeated major recessions will have a permanently damaging effect on the long-run growth rate that is actually achieved.

Comments on Conclusion 4. The notion that the economy would move along the trend line if left to itself is perhaps the most surprising feature of the model, in view of the emphasis that Dow elsewhere places on demand. It is a little redolent of Say's Law, qualified by the recognition of the periodical downward displacements of supply that arise from demand in major recessions.

Dow is not seeking to set up an integrated model of growth and fluctuations. He therefore does not commit himself on the question whether in the absence of major disturbances capacity and output could actually go on indefinitely growing along the trend line as he defines it (BC). Possibly it would ultimately slow down as the result of an encounter with some kind of ceiling [p. 100]. Nor does he postulate that a major recession would alter the trend rate of growth (BC or DE) one way or the other. It is sufficient for his purpose to establish that a major depression pushes the economy permanently off course.

Turning to some of Dow's data, it is at first sight rather surprising to find that, whereas all major industrial countries experienced trend retardation after the early 1970s, only the UK and Italy (of the eight countries listed in Table 2.10, p. 33) experienced larger deviations from trend in 1973-90 than in 1950-73. The tendency in the other six countries was actually in the opposite direction. However, this does not necessarily mean that they became less subject to major recessions in the sense of Dow. The deviations just referred to are deviations from a fitted trend, not downward displacements of the trend line. Germany is shown in Dow's chart as having had two substantial downward displacements of trend in the later period. On the other hand the displacements shown from France and Japan are very modest, yet they too underwent large retardations. Thus Dow's data do not support the hypothesis that the famous retardation after the early 1970s was everywhere due to increased incidence of major recessions. T o do Dow justice, he only hints at such a hypothesis. He says that major recessions cause retardation, not that all retardations are due to major recessions. It would nonetheless be a little disappointing if the same general explanation did not hold for a phenomenon that was so obviously worldwide.

Conclusion 5. It is a matter of high importance for governments to adopt policies that will prevent or at least mitigate major recessions. This will not be easy, and it would be wrong to be too optimistic about the prospects of success. In responding to external shocks, government faces many of the same difficulties as the private sector does. However government policy itself has been a significant source of shocks, and that at least is something that it should be possible to avoid.

In discussing Conclusion 5 later in this review, we shall pay particular attention to the periods when major recessions were avoided.

The above five conclusions are by no means Dow's only ones. Some of the others will be referred to in connection with his historical survey, to which we now proceed.

The case studies: Michael Artis

About one half of Dow's book is devoted to the five case studies of major recession that he detects between the end of World War I and the mid-1990s. Three of those recessions occur after the end of World War II, the first of them dated 1973-5. The other two are dated 1920-1 and 1929-32; in between there is the recovery of the mid- to late-1930s, World War II and its aftermath, a period in which no major recessions intervened.

The account that Dow gives of the first of these recessions, that of 1920-1, commences with a point of which too little -- arguably -- is subsequently made. This is that there was at that time a remarkable decline in hours of work. Matthews et al. (1982) discuss this remarkable phenomenon at some length (it was a phenomenon common to a number of European countries). Thus the observed decline in output was due in part to a straightforward decline in supply. Dow admits that this was probably to some extent still the case even in 1921 (and was certainly the driving force behind the earlier output falls in 1919 and 1920). But in that year there occurred a sharp fall in employment which Dow attributes to a demand decline. This might have been associated with the previous fall in working hours since the wage bill did not fall, and real wages and the share of wages in total factor income had consequently increased. An explanation hinging on "excessive real wages" might be a starter, but it is not Dow's explanation. Instead he points to a decline in world demand (there was a big drop in UK exports) and a negative fiscal policy impact as triggers for a decline in confidence. This recession is reasonably described as "very odd" - it certainly contains some unusual and confusing features, most notably the hours of work adjustment which left total output in 1920 some 5 per cent below its pre-World War I level.

When he comes to the second big interwar recession, Dow's argument points to the principal proximate cause as having been that of the Great Depression in the United States. That depression indirectly reduced the demand for UK exports (principally by depressing primary producers' demand). This effect in itself (amounting to a decline of some 4 per cent of GDP between 1929 and 1932) was, however, substantially modified by the accompanying favourable movement in the terms of trade. This movement supported consumers' expenditure which, Dow points out, actually rose throughout the recession, yielding a net negative impact of about 2 per cent. Relative to the conventional wisdom (new as well as old) that adherence to the Gold Standard had a lot to do with the recession, Dow argues robustly that the Gold Standard did not constrain the United States and that the fall in output there was extremely large relative both to the past experience of the United States itself and to the contemporary experience of other industr ialised countries. Unless it can be shown that a lack of gold was a key determinant of restrictive monetary policy in the United States itself, by far the largest gold holder in the world at the time and the centre country of the system, Dow's point seems a good one and the correction of perspective overdue. Even so, the subsequent contribution to the recovery in the UK of the devaluation and the turnaround in fiscal policy are testimony to the prior restrictiveness of adherence to the Gold Standard and the associated stringency in financial policy. Dow has little room for explanations based on "excessively high real wages", basically on the grounds that real wages are not the direct object of wage settlement and prices are fixed by manufacturers. This argument can be queried when wage negotiators explicitly aim at real (product) wages but this is unusual and was not the case at this time (a rare earlier exception is provided by the sliding scale that related miners' wages to the price of coal in the late 19t h century).

Given the centrality of the US depression as the leading and exogenous factor driving the UK into recession, Dow spends several pages discussing the causes of that depression. He attributes major importance to the amplification of initial causes due to the wave of bank failures. But the reader should not suppose that this is the same line of reasoning as that of Friedman and Schwarz who, famously, indict Federal Reserve policy for not seeing to it that bank reserves and money supply did not fall. Dow accuses them of ignoring the endogeneity of the bank reserves and the stock of money and, essentially, of getting the causation wrong. Bank failures reflected a lack of demand for bank loans and structural weaknesses in the US banking system of the time. All this is well argued, persuasive and important even if from the point of view of treating UK history it is necessarily something of a detour.

Dow's view about the two recessions 1973-5 and 1979-82 is relatively conventional. In both he ascribes a major role to the exogenous event of the oil price increase. The deflationary impact of the increases is described in terms of the Keynesian "indirect tax increase" analogy. As with an indirect tax increase there is a price-raising demand-reducing effect. In the case of the first oil price shock the revenue proceeds were initially largely saved by the OPEC countries so that a redistribution of world income was accompanied by a deflationary shock. In the second case - despite the fact that the UK was by then a net exporter of oil - the analysis is much the same, for the UK government, to which a large proportion of the increased revenues accrued, was bent on a tight fiscal policy. The reason why the shocks were so hard to deal with is also crisply acknowledged: the reduction in the disposable income equivalent of output means that real incomes must inevitably fall. Resistance to this implies inflationary pr essure. Dow remarks, without enthusiasm, that "The case for a restrictive policy is that it will encourage acceptance of the inevitable, and thus lessen and shorten the scale of inflation". He is sceptical that as much had been learnt about how to deal with oil price shocks by 1979 as some observers claim: mainly, it was just that the shock was smaller.

But where he parts company most notably with at least some other analysts is in downplaying the supply side effects of the oil shocks, especially the first and larger one. Here the main issue is why it was that after the first oil shock there was such a large fall in productivity and, afterwards, when growth was resumed, why the growth rate was so much lower than before. These phenomena have given rise to a number of competing explanations. One argument is that this productivity growth break was due to the exhaustion of opportunities for 'catchup' (of the US by other industrialised countries). This is only coincidentally related to the oil shock per se. Other arguments are more closely related to it. For example, one school of thought argued that the oil shocks left real wages too high and another that the effective capital stock was drastically lowered by the shock to energy prices (as if it had been bombed out of existence). The profession came to no compelling consensus on the merits of these rival exp lanations. Dow contributes two important points to the debate. First, he notes that much the same phenomena regarding productivity were found in the aftermath of the 1929 slump u and prefers, in this light, to attribute the productivity phenomena to the collapse in demand per se. Second, he makes some points on the possible scope of the supply side effect which seem quite telling, all the more so for being simple (e.g. he points out that two-thirds of the capital stock is embodied in dwellings, buildings and such whilst The annual return on the remaining third might ... be put at around 10 per cent of GDP ... Suppose the proportion of that third rendered smaller by dearer energy was somewhere between 1 and 10 percent... The extent to which potential total supply will ... have been reduced would then be between 0.1 and 1 per cent of GDP. Even the larger figure is only a fraction of the actual total loss of productivity [p. 287, italics supplied].

The most recent recession (1989u93) is clearly a different matter altogether. Most observers will be prepared to accept that the first two major postwar recessions can be traced in large part to the oil shocks, themselves reasonably treated as exogenous events. The third major postwar recession, on the other hand, equally unforeseen by the principal forecasting organisations, is another matter. There do not appear to be any clearcut exogenous 'events' that can be 'blamed' in the way that the oil price shocks can be blamed for the earlier recessions and no straightforward way to excuse the forecasters for what appears to be a major error.

An econometric analysis of consumption in recessions: Martin Weale

Dow's explanation is that the recession was largely due to the after-effects of a preceding "wave of optimism". This view in itself is eminently plausible, though Dow finds reason to contrast it with what he regards as the chief rival explanation, which is that the recession resulted from prior excesses due to "financial deregulation". He picks as the chief proponent of this view ex-Chancellor Lawson. It is nor too difficult to find fault with what seems to be an overly nit-picking approach to the main issues here. The central issue surely is whether a deregulated environment gives more scope for waves of optimism to build (and crash) or whether deregulation allows the economy's self-regulating mechanisms to come into play. Then there is the logically separate issue concerning learning about the new environment. Most of our ideas about how a financially deregulated economy can regulate itself presume a fully-learned, well-informed set of actors. Even then, on many accounts imperfect information makes herdin g behaviour endemic, leading to the likelihood that waves of optimism (and pessimism) will result from time to time. The extreme form of this is in speculative 'mania', on which a large literature exists. The problems of information are more acute, it seems reasonable to suggest, when an act of deregulation itself opens up new opportunities (note the word 'new'), and indeed there is a history of deregulation leading to over-lending and crisis. Had the western economies not already been greatly deregulated, it seems quite likely that the wave of optimism would not have had the scope to build.

Dow identifies a number of ways in which large recessions differ from small ones. This section looks at one of the points he makes about the influence of recessions on consumption.

"The reason why large recessions are large seems to be partly that the initial shock is larger; but also partly because the processes by which an initial shock gets amplified are then more powerful. The effect on consumption is larger because consumers' expectations of future income, and thus also their spending, fall more in response to a large shock." [p.3]

"A large recession may reduce consumers' life-time income expectations; the change in planned savings will then be more radical. If it lowers the rate of increase of expected income over their lifetime, that . . . could reduce spending by more than income has fallen." [p. 90]

"Consumers are likely to reduce spending in anticipation of any fall in income..." [p. 90]

These remarks have to be read in the context of his view that recessions lead to permanent falls of output, a basic proposition which is, unfortunately, very difficult to examine statistically because the various tests for structural breaks (Andrews, 1993) have low power. Looking at the period 1920-95, such tests might be able to identify one or two breaks, but it is very unlikely that they would be able to identify breaks associated with each of the five recessions which Dow discusses. On the other hand, one can examine the behaviour of consumption/ saving during the recessions, in order to examine how far consumption movements had the effect that Dow suggests.

The savings rote and the big recessions

An underlying framework which leads broadly to Dow's conclusion is an informal form of the life-cycle/permanent income model in conjunction with a neo-classical view of production. In this model, a representative forward-looking consumer will choose a consumption path on which the planned rate of growth of consumption is proportionate to the difference between the real rate of interest and the discount rate. If the real rate of interest is expected to remain constant at the discount rate, then each consumer will plan for constant future consumption. It follows that, if there is a fall in the level of output which is expected to be permanent, in that trend growth is resumed from a new, lower, level, then the savings ratio will be unaffected. On the other hand, the appropriate response to a temporary fall in income is to maintain consumption. If the fall of income is unanticipated, then, at the same time as the savings ratio declines, there will be modest permanent fall in consumption; the temporary fall in inc ome reduces the lifetime resources available to consumers.

Chart 2 shows the national savings ratio, calculated by adding together household and public consumption and national saving measured net of stock appreciation and depreciation of capital. Consumption is measured at market prices rather than factor cost and the underlying measure of national income (consumption plus saving) therefore differs from the accounting standard. [4] The conventional consumption figures are also redefined so that the depreciation of consumer durables is treated as consumption, while net increases in the stock of durables are regarded as financed by saving. Since the stock of durables tends to grow over time, this has the effect of raising saving and reducing consumption; it does not affect the measure of national income which remains the sum of the returns to labour and (net of depreciation) to the capital stock. [5]

The savings ratio is calculated from personal and public consumption aggregated and measured relative to national income rather than, as is usual in analyses of the consumption function, from personal saving measured relative to personal disposable income. This is because the intention of this study is to look specifically at the role of consumption movements in aggravating recessions; to do this it makes sense to look at movements in personal consumption after allowing for the offsetting effects of any movements in public consumption. Should personal consumption fall at the start of each recession, in the manner that Dow suggests, but that movement be offset by a rise in public consumption, it would be difficult to argue that movements in consumption are a factor intensifying the recession even if what is said in the quotation on the previous page is true.

It might be argued that a fall of personal consumption, even if offset by a rise in public consumption, was consistent with a view that households reduced their consumption in the light of a permanent expected reduction in income, as suggested by the second quotation at the start of this section. In order to study that question, as opposed to the more general one about consumption movements intensifying recessions, it would be necessary to estimate a full consumption function. Here we limit ourselves to the first issue which is better addressed with the consolidated consumption data.

Dow identifies recessions in 1920u21, 1929u32, 1973u 5, 1979u82 and 1989u93. The first year of each recession represents an output peak; thus it is not itself a recessionary year. Apart from World War II, the chart shows troughs in the savings ratio in 1921, 1931u3, 1975u7 and 1991u3, with a less obvious trough in 1981u2. There are also troughs associated with the General Strike and in 1952u4 with other less marked dips in other years.

Looking at Chart 2, one might be able to entertain the view that, in 1975u7 and 1981u2 the savings ratio remained at normal levels. The situation in the mid-i 970s is confused by the fact that the savings ratio was sharply lower after 1974 than before; it averaged 4.95 per cent between 1920 and 1938, 11.18 per cent between 1950 and 1974 and 7.10 per cent between 1975 and 1995. Thus the savings rate in 1975u7 can be regarded either as low by comparison with the period from 1950u74 or as normal for the period 1975u95. A similar point can be made for 1981u2. The trough exists because saving was abnormally high between 1978 and 1980 rather than because it was unusually low during the recession. On the other hand saving in 1991u3 was low in the context of the period in which the recession took place.

This chart then suggests that in 1921, and in the recessions of the 1930s and 1990s, saving was unusually low, failing to support the notion that consumption movements intensified other movements in demand. In the 1970s and 1980s, howevet savings behaviour is more compatible with the notion that the recessions were ex pected to have a permanent effect on output and thus with Dow's hypothesis. There is not very much evidence of saving rising in advance of a recession (saving up for a rainy day) although there are local peaks in the savings rate in 1920, 1979 and 1989; it could be argued that these were motivated in this way.

However, this visual analysis is not completely satisfactory. It is generally found that consumption is excessively sensitive to movements in income, meaning that consumption changes by more, and therefore the savings ratio changes by less than the life-cycle model predicts in response to a temporary change to income. Without trying, in any sense to fit a consumption function, we can shed more light on part of Dow's argument by looking at the relationship between income and the savings rate in greater detail.

This is done by estimating the statistical relationship between the national savings rate and the change in the logarithm of national incomeA The relationship is meant to summarise the data rather than to provide a coherent behavioural model (for which one would need to examine various forward-looking models). The savings rate is related to its own lagged values and current and lagged values of the growth of real income. The lag lengths are determined using the Schwarz Inequality Criterion (Schwarz, 1978). Using this method it is impossible to say anything about the 1921 recession; it comes hard on the heels of the Great War and there are no satisfactory lagged data with which to start the model. However, we can look at the other four recessions by putting dummy variables into the equation. If the data are to provide evidence that consumption movements intensified the recessions, then in these recessions the savings ratio should be higher than the summary of the rest of the data period suggests. In view of th e visual evidence that the effects of the recessions appeared only after a year into the recessions, and that saving remained low in the first recovery year, dummies were set for 1931u3, 1975 and 1976, 1981 and 1982 and 1991u3. A preliminary analysis also suggested that the savings rate was on average higher between 1950 and 1973 than between 1974 and 1995, and this difference could not be accounted for simply by the change in income growth. Accordingly another dummy was introduced for this period.

The results of the regression equations are shown in Table 2, for the prewar and in Table 3 postwar. Of the four recessions examined, the savings ratio is unusually low, not unusually high in three of them; in two of these (1931-3 and 1991-3) the difference from the basic model is statistically significant at a 5 per cent level, and in the third case (1975-6) the p-value is 8 per cent. In the recession of the early 1980s saving was very slightly higher than the basic structure suggests, but the difference is so small that it would be quite wrong to give any importance to this.

There remains the question whether the picture has been affected by the treatment of the first year of each recession. It has been assumed that people took some time to adjust to the recession, which was why saving held up. The alternative, of course, is that saving held up because they knew that the recession was going to have a permanent effect on output and there was no offsetting movement in public consumption. In 1930 saving was lower than the fitted value, by 0.44 per cent of national income; the standard error of the equation was 0.8 per cent of income. In 1974 and in 1980 the saving rate was higher, by 0.56 per cent and by 0.34 per cent than the model suggests. In 1990, on the other hand, it was lower by 0.95 per cent. The standard error of the equation was 0.58 per cent. The figures for 1974 and 1980 might give tentative support to the view that the output falls which occurred in these years were aggravated by falls of consumption, while the subsequent falls were perceived as temporary; but the evide nce in favour of this view could hardly be described as convincing.

Study of four of the five recessions considered by Dow suggests that, even if they did have a permanent effect on output, the savings rate does not reflect this. Far from the national savings rate holding up, or even rising during recessions, it tends to fall by an unusually large extent; movements in national saving have tended to offset rather than exaggerate recessionary forces.

Of course these arguments do not necessarily extend to the components of national saving. It is perfectly possible that household consumption has indeed behaved in the manner suggested by Dow, with the effect being offset by movements in public consumption. It is an open question whether these three should be regarded as completely independent or, as is implicitly assumed here, the outcome of a consolidated decision process. But even if the consumption decisions by the different sectors are considered independently, their aggregate outcome remains that described here: even if household consumption has tended to aggravate recessions in the manner Dow suggests, its effects have been offset by the behaviour of public sector consumption.

The long interval without major recessions 1945-73: John Flemming

The twenty-five years following World War II are now conventionally seen as a golden age. Keynesian policies delivered both full employment and unprecedented real growth. Admittedly other European countries grew even faster (having more wartime damage to recover from), policy was dominated by successive balance of payments crises and was characterised as stop-go. Matthews (1968) questioned the attribution of full employment to Keynesian policies.

The questions to ask here are whether this was truly a period free of the costs of recession, and, if so, what lessons one might learn as to how to avoid them -- the subject of the next section.

If a major recession is defined as involving high unemployment, it is clear that none occurred in the UK between 1945 and 1975. But one of Dow's reasons for focussing on major recessions is the cost they represent, particularly in foregone output, as, on his account, they permanently displace the path of capacity output.

Dow identifies five UK "mini cycles" between 1947 and 1973. They are certainly "mini" in their implications for unemployment, which, until 1970, oscillated between 1 and 2 per cent and never exceeded 3 per cent. But does this imply that there were no costs comparable to those associated with longer recessions?

Growth in upswings in this period consistently reached 4 per cent per annum, while the trend growth rate rose from 13/4 per cent per annum to 21/2 per cent per annum for an average of 21/4 per cent per annum. Although growth was rarely less than 1 per cent per annum, downswings were typically longer than upsprings (Chart 3) so that the average growth rate was less than 2 1/2 per cent per annum. Should the "mini cycles" be held responsible for the difference between 4 per cent and 21/2 per cent annual growth over 25 years - as large a cumulative total as that of the few later recessions? Dow clearly does not think so, for which he gives two reasons - one more explicit than the other.

In chapter 7 he argues that these mini cycles were less "stop-go" than "go-stop" as clumsy use of the fiscal accelerator encountered a balance of payments con-straint -- in an era of capital controls - and precipitated equally heavy footed braking. Thus the 4 per cent growth rate was artificially induced, unsustainable and did not represent true capacity growth. There have, however, been economists, such as Kaldor, cited above, who advocated just such policies as means of raising the trend rate of growth. This line of defence is not entirely convincing, although these mini-cycles, including their upswings, were dominated by stockbuilding.

There is, however, another basis for distinguishing between major and minor recessions. Dow is rightly interested in some concept of capacity or productive potential. For reasons discussed earlier he is reluctant to put any weight on investment or capital stock data, preferring to use unemployment rates (despite the well known loose relationship between changes in employment and unemployment).

He defines "constant employment GDP" at time t as [GDP.sub.t] =(1 + [u.sub.t] - [u.sub.s,v]) where s [less than] t [less than] v are the ends of a subperiod over which [u.sub.s,v] is the average rate of unemployment. Deviations of GDP from "constant employment GDP", GDP, are plotted in Chart 4 [fig. 2.1] which also indicates by shading the five major recessions. It will be seen that different trend rates of growth of GDP are indicated on the chart for each inter-recessional period (also allowing a break for World War II).

As has been indicated in the Introduction the five recessions appear to have been selected initially by reference to a year-to-year drop in GDP and a plunge in employment raising unemployment to a (new) high level. The most interesting part of Dow's theories, however, requires that major recessions not merely mark changing slopes in trend growth of GDP but actual displacements of the trend line. Such displacements are indeed suggested by Chart 4 [fig. 2.1] and are illustrated more explicitly for the post-war period in Chart 5 [fig. 2.6].

It seems that Dow used informal methods to identify the major recessions and then fitted trends with breaks in level and slope at these points. Is this a legitimate procedure? It certainly looks somewhat casual. Could it be reformulated more formally and would that produce the same results?

Given a definition of constant employment GDP, which enables one to abstract from cyclical effects, there are well known, though not entirely definitive, techniques -- also discussed earlier by Weale -- for identifying intercept and slope changes at particular levels of significance. Such tests would also help to pin down major recessions, which could then be defined as downward displacements of trend constant employment GDP (and perhaps also of actual GDP from that trend).

A question to which Dow devotes more time is the appropriateness of his simple employment adjustment. He is well aware that Okun's law associates a rise in GDP relative to capacity with a much less than proportionate increase in employment (or the corresponding change in unemployment). He suggests [p. 22] that "at times of major recession ... employment (and unemployment) reacts more fully to output changes than in average year-to-year fluctuation as reflected in Okun's Law".

It seems that this is not a statement (only) about the reaction to recession-inducing shocks but to the whole of a period of a major recession and the subsequent recovery. Thus he regresses [delta]u on [delta]y separately for 1920- 38, 1950-73 and 1973-92 (when he substitutes [delta][y.sub.-1] for [delta]y). The first and last periods are characterised by major recessions (1920 and 1930; 1973, 1980, and 1990). In these periods the coefficient is about (-) 0.40 or 0.45 whereas the middle period without major recessions has a coefficient smaller than (-) 0.2.

Using the reciprocal of the relevant coefficient to make the constant employment adjustment to GDP (Method B) produces no downward displacements of the trend of GDP in the postwar period. Thus quite a lot of Dow's thesis rides on a defence of his choice of Method A - generating constant employment GDP as defined above.

He lists three reasons for preferring it on page 22:

(1) His first reason has already been quoted. He does give evidence of a significantly larger and more rapid (un)employment response in the periods of major recession, but the coefficient is still less than 0.5 (in absolute magnitude) and its reciprocal (at 2.25) is not much closer to the unity of Method A than the 5 of Method B (as applied to the period 1950-73). One is tempted to think that Dow might have argued that if 5.0 characterised a period without major recessions and 2.25 a long period with two or three such recessions, a major recession itself might be characterised by a coefficient of unity. But if that was his thought, to how long a sub-period could such a coefficient properly be applied? Not, presumably, the 20-year periods to which he applies Method A.

(2) Method B generates implausibly large (10 per cent or more) shortfalls of output from capacity output.

This is illustrated in Dow's figure 11.1 [p. 386] (reproduced here as Chart 6). Extrapolating the pre-1973 average GDP growth rate of 2.9 per cent to 1992 suggests a 23 per cent shortfall of actual from potential GDP. Dow reduces this to a 7 per cent shortfall of constant employment GDP from its capacity level in 1992 by adding back the 7 per cent point rise in unemployment and comparing it with a trend line displaced by 2.2 per cent in 1973 and 5.4 per cent in 1981.

(3) One should "expect that major recessions will damage economic capacity" so that a method revealing this is to be preferred.

All of these suggest that while Method A exemplifies a postulated mechanism it does nothing to substantiate its operation. In particular some of Dow's difficulties might be due to his reluctance to use any measure of the natural rate of unemployment (or the NAIRU) (see [p. 424]). If such indicators trended upwards over a period, and should be taken as reflecting a constraint on "full (or constant) employment GDP", then actual output is likely to diverge cumulatively from a trend of capacity GDP calculated without taking this factor into account. The procedure would then need downward displacement of the trend line in order to keep things on track as at (2) above. Thus Dow's procedure is particularly likely to locate "major recessions in such periods, and to attribute to them a retardation of growth more appropriately attributed to the determinants of the NAIRU.

Of course Dow's distaste for that concept is understandable, and on a hysteresis model (major) recessions may account for apparent increases in the NAIRU. Thus the debate is not closed -- but that may serve to reinforce the view that a rather more thorough statistical, if not econometric, approach would have been in order to pin down the mechanism and pattern of causation.

Given Dow's view that no major recessions occurred between 1945 and 1973 either in the sense of a large rise in unemployment or in the sense of a costly displacement of trend growth, how does he believe that that was achieved? A question given added point by the evident failure to avoid recession after World War I. He does compare the aftermath of the two world wars, and acknowledges that the later policy makers were influenced by their knowledge of the earlier episode in managing demand, but by fiscal and micro regulation. Demand was buoyant and Dow explicitly endorses a "bootstraps" argument that it stayed up because output income and demand remained so buoyant [[ss]7.2, p. 247].

All of this is fairly plausible -- with the post-World War I 1920/1 recession not recurring because the immediate postwar boom was contained both by formal rationing -- not only of consumer goods -- and, he suggests (or implies), the less formal rationing implied by stickier prices in the late 1940s than early 1920s.

One difficulty with this account is to square its emphasis on the consumer confidence supporting demand with the interpretation of subsequent labour market developments. Only after 1969 did unemployment persistently exceed 2 per cent and inflation was equally low as late as 1966. In 1976 I inferred that the natural rate of unemployment in the 1950-65 period seemed to lie between 13/4 and 2 per cent. It subsequently rose enormously -- on some accounts to a peak of about 10 per cent. As mentioned, Dow was never enamoured of NAIRU or natural rate concepts -- seeing the relevant relationship as being "of a sociological sort, reflecting many aspects of group behaviour ... it may ... easily change ... as a result of experience [including that] of inflation" [p. 424] -- or, one might add, recession. One account of the adverse shifts of the Phillips curve at the end of the golden age (and later) relates to the fading of memories of prewar unemployment -- and the passing of the generation that experienced it. As we ha ve seen, an upward trend of this kind might be thought to undermine Dow's approach unless it could be attributed to the experience of recession -- which is not plausible for this particular period.

The questions that this sort of (implicit?) approach raises for his account is the coherence of an explanation of strong consumer demand based on apparently high (and growing) confidence at the same time as wage restraint was based on apparently high (but waning) insecurity. The combination should have led to an earlier and more rapid shift in the Phillips curve than in fact occurred. Dow's emphasis on confidence rather than expectation is very much in the spirit of Keynes. It may be that consumer demand and confidence was not on its own so consistently strong as a first reading suggests that Dow maintains. Both in discussing the early post-World War II period, and in discussing remedies for recession, Dow displays a now unfashionable yearning for credit and consumer controls. If consumer demand, based on a moderate level of confidence, was pressing against these limits, and thus deferred into periods in which it might otherwise have slackened, its persistent strength relative to supply should not be mistaken for spontaneous and stable strength in any absolute sense. On this view there are other problems, if consumer demand was being restrained why was investment demand so strong? Here the role of reconstruction (and exports) must be crucial with consequences for employment and income that contribute to maintaining consumer demand pressure on capacity and control constraints.

A third surprise is the ease with which demobilised labour could be absorbed into civil production at the same time as workers at home were released from wartime production, although the continuation of National Service for fifteen years meant that demobilisation was less complete than after World War I. This may have its parallel in fighting a recession following a sustained boom and bubble such as characterised Japan in the 1980s. In the boom-and-bubble-economy labour is disproportionately employed in investment, construction and luxury services. If unemployment were to be avoided in its aftermath many of the workers in these sectors would have to find jobs in cooperation with the limited capital stock devoted to producing more basic consumer goods (or exports) which would depress the (market clearing) real consumption wage -- possibly quite sharply. If it is hard to imagine this working after a bubble, so that redeployment could not avert unemployment, how did it work so well after the war? This is not a q uestion addressed by Dow in this work; nor indeed by authors such as Wilson (1952) and Wright (1979), whose accounts of demand pressing against administrative constraints parallel Dow's.

Are recurrent major recessions inevitable? John Flemming and Robin Matthews

This is the title of the book's concluding chapter. It is a more tentative chapter than the earlier ones. "Full answers to these questions would require another book.. what follows is a semi-reasoned statement of belief, or a series of essays intended to give an idea of the considerations that led to my conclusions" [p. 414].

Dow's general approach implies that the answer to the question in the title must depend on the answers to two sub-questions:

(1) are recurrent unforeseen major shocks inevitable?

(2) is it inevitable that government policy will be unable to prevent those shocks from creating major recessions?

It is obvious what the answer will be to the first of these questions. Historically, shocks have come from a variety of sources and have varied in their timing and in their severity to a major degree. They have not been well foreseen by private agents, nor by governments. There is every reason to suppose that they will persist.

So we come to the second question. Not surprisingly, in view of Dow's emphasis on the diversity of past experience, he finds it difficult to make contributions that are both new and of major general importance. And like authors of all books on practical macroeconomic policy, it is in a few places overtaken by events in consequence of unavoidable publication lag. However, he does make some interesting points that are not commonplace.

It is easy to make fun of Panglossian models in which perfect knowledge of present and future are imputed to private economic agents. A more serious question, as Dow recognises, is whether governments are on balance better informed and/or better motivated than the sum of individuals. Here is one possible example of democratic malfunction, different from the ones that are usually cited. Dow suggests that democracy (rather than the sinister machinations of financiers) may be what causes macroeconomic policy to be biased in a deflationary direction. The reason is that inflation affects nearly all voters, whereas unemployment affects only a small minority (though it affects those more severely than inflation affects the generality). Politicians in search of votes favour the large battalions.

That is a general point. Dow also makes a number of suggestions about faults in policy in past periods that may well be valid in the light of hindsight but are not very helpful for the future (his views on the causes of the five major recessions are conveniently summarised in the box on [pp. 411-2]). Another example already briefly touched on relates to the years immediately after World War II. The experience of the post-World War I recession, as well as the memories of the 1930s unemployment, made the government then lean towards excess demand. In retrospect Dow thinks this was pushed too far. At the time it was possible to pursue such a policy without causing undue inflation, because many direct controls inherited from the war were still in place -- things were not so different, in that respect, from the centrally planned economies of Eastern and Central Europe at that time. However, it would hardly be feasible technically or psychologically to reinstate those controls now.

Dow does not generally favour the various attempts that have been made to shortcut the task of the policymaker by seeking stability in a single variable, such as the money supply or the exchange rate. He frowns on the more fashionable policy of targeting the inflation rate, not only because price stability is not the only objective of policy but also because experience shows that the relationship between trends in the price level and trends in other variables is far from regular. His book went to press too soon for him to be able to comment on the recent bouleversement of the task placed on the Bank of England's Monetary Policy Committee, by which the inflation target has been made in principle a minimum as well as a maximum, thus treating unusually low inflation as a proxy for demand deficiency. One may conjecture that he might have thought it an improvement on what went before, but he would have regarded it as subject to the same objection as that just stated, namely that historically there has been a quite irregular relationship between price trends and trends in activity (in the golden age 1950-73, for example, prices generally rose less rapidly in cyclica l upswings than in cyclical downswings).

Anyone who thinks of Dow as an unregenerate Keynesian is in for some surprises. In an interesting section on the national debt, he points out first that historically there have been very large changes in the ratio of national debt to GDP unconnected with budget deficits and surpluses as usually defined. At the same time he does take the size of the national debt as a serious constraint on fiscal policy, at least in some circumstances. In particular, sudden increases in the national debt will cause financial markets, reasonably, to suspect that the government is letting things get out of control. Apart from that [p. 433] "countries with debt-GDP ratios much above the average are, I think, bound to take steps to limit any further rise in the ratio, and probably to reduce it. Conversely, governments not in this position need to take care not to get into it, but otherwise need not be so actively concerned about the ratio" -- which rather alters things for the British reader. This is perhaps one of a number of pla ces in the book where Dow's many years spent drafting the Bank of England Quarterly Bulletin have influenced his style. "He writes as a banker does, saying only the needful", said Sir Walter Scott. One might nowadays substitute "he writes as a central banker does, for reading between the lines".

Although Dow is thus not really too worried in the British context about increases in the national debt (which are in any case likely to be only temporary as recovery resumes), he does have sympathy with the conventional hostility of Finance Ministries to the use of fiscal policy for stabilisation purposes, because of the danger that it will undermine discipline in spending departments. This should nor apply to the automatic stabilisers, which are of limited extent and are well understood; these should be kept in place. For the rest, he suggests various ways in which public policy might affect private spending demand. Some of these are reminiscent of Swedish schemes fashionable in the 1960s and 1970s -- but Dow concentrates on the government as leasing facilities commissioned from private contractors, as under the British 'Private Finance Initiative', or stimulating now privatised, but regulated, public utilities into countercyclical patterns of investment. This would have the advantage of lying outside the b udget and so not causing the threat mentioned to financial discipline. Dow recognises but perhaps does not sufficiently emphasise the timing problems and the dangers of counterproductive effects on confidence in the private sector.

In this context public works run two risks. They may be seen as too little too late and thus as a signal from above that things are worse than has been realised -- thus stimulating private saving. This has happened in Japan, with successive unsuccessful attempts to prime the pump of domestic demand. The problem is the impossibility of judging what will prove 'enough' to resuscitate a depressed economy. If one cannot credibly say one has done enough, what reaction should one expect from private agents? On the other hand a tax cut to be held until GDP recovers to a pre-set level does provide a boost to disposable income, bridging until earnings recover, which consumers should be willing to spend; in principle it is no more threatening to fiscal discipline than the automatic stabilisers.

Dow's treatment of monetary and fiscal policy is contrary to what is widely regarded as Keynesian orthodoxy in another respect. He emphasises the prime importance of avoiding excessive booms, as a means of avoiding major recessions -- just as Marshall and other 19th century authors emphasised the importance of preventing 'overtrading'. The Barber boom and the Lawson boom figure largely in Dow's list of past avoidable errors.

In discussing monetary policy, Dow recognises the dilemma created by the influence of interest rate on exchange rates. At the same time there is here what seems to us to be the greatest weakness in his chapter on policy. Globalisation is acknowledged, but it is nor put at the centre of the stage, as many observers would be inclined to think it ought to be. How far are the limitations of government stabilisation policy the limitations of the powers of the government of a single stare, operating in a world where the private sector is not subject to a similar limitation? There is an interesting brief discussion and footnote [on p. 436] about blocs of nations. But, not surprisingly, he does not address the basic dilemma of how to reconcile the need for transnational stabilisation policy with national democracy. The experience of Japan (and more recently, perhaps, of Eire) suggest that national stabilisation policy may still have its uses.

Dow's own account of the causes of major recessions suggests another aspect of policymaking which it might have been interesting to pursue. Suppose it is true, as he postulates, that major recessions are characterised by once-for-all loss of capacity. Is this loss of capacity in all cases really irrevocable? If not, policy can be directed towards the retrieval of the capacity that has been lost. The extent to which this is feasible is likely to depend on which of the various possible causes enumerated by Dow was responsible on any particular occasion for the loss of capacity. It is also likely to be affected by the length of time that has elapsed since the capacity was lost.

Concluding comment: Robin Matthews

When Joseph Schumpeter's Business Cycles was published in 1939, it received an adverse review from archempiricist Simon Kuznets. Kuznets's criticism was that, although the book contained dozens of charts and hundreds of pages of narrative, all this empirical material did not really establish much of a case for Schumpeter's model.

Yet what is now remembered and come to form part of the lore of economics is Schumpeter's model, not Kuznets's criticism.

Can this be taken as a likely precedent for Dow's Major Recessions? The two cases differ in important ways.

On the one hand, Dow's stated objectives have a much narrower scope than the "magnificent dynamics" of Schumpeter. On the other hand, Dow's thoughtful analysis of the data is quite unlike Schumpeter's undoubtedly slapdash procedures. So historians of the particular episodes studied by Dow will learn much from his book, whereas the historical portions of Business Cycles are not now much read -- and to tell the truth never have been. It was thinking about the data that led Dow to his conclusions, instead of the model coming first and the examination of the data second, as with Schumpeter (and many more recent economists too).

Yet there remain certain parallels. Dow's model is broadly compatible with the course of these historical events. But by its nature, his way of analysing the data is not likely to establish sharp general conclusions convincingly. He would probably have replied that this is an unavoidable feature of the subject and that more ambitious claims are bogus. Not all will agree, and some parts of his model may prove shaky in the light of further empirical analysis. Whether this is so or not, we are not likely to forget his general picture of the irregularly occurring major recessions, greatly affected by confidence, which lead to permanent downward shifts in the growth path of the economy, though not to its irretrievable collapse.

Dow, J.C.R. (1998), Major Recessions: Britain and the World, 1920-1995. Oxford, Oxford University Press (paperback forthcoming Autumn 2000). Charts 3-6 reproduced by kind permission of Oxford University Press.

NOTES

(1.) References in square brackets refer to page nos in Major Recessions.

(2.) Perhaps the reason is that office blocks and dwellings are more likely to have become objects of speculation in the boom than power stations. Alternatively the reason is perhaps that they have a longer life than most forms of capital. It has long been recognised, of course, that building is subject to fluctuations that may be longer and have a different timing from fluctuations in general activity. In the UK, building booms before World War II occurred in the 1870s, in the late 1890s and early 1900s, and in the 1930s, and their effects are still visible in our towns. In more recent times there have been notoriously violent recessions in the building industry in the mid 1970s and the early 1990s.

(3.) The last point is another very Marshallian idea (though Dow does not credit it to Marshall). It is exemplified in the unusual title of Book IV of the Principles: "The Agents of Production: Land, Labour Capital, and Organization" (italics added).

(4.) The savings and consumption figures are taken from Sefton and Weale (1995), extended using the 1996 Blue Book beyond 1990. The estimates of durable goods' depreciation are derived from the calculations performed by Solomou and Weale (1997) to calculate holdings of durable goods shown in the personal sector balance sheet.

(5.) Although, if an income were imputed to the stock of consumer durables, then the measure of aggregate income would differ further from its conventional value. Income is either consumed or saved and we calculate it by adding together consumption (public and private) and saving as conventionally measured (net of depreciation and stock appreciation). However, we then add to the consumption estimate our estimate of the depreciation of the stock of capital goods and deduct from it the value of purchases of capital goods, making an offsetting adjustment to saving so as to leave income unchanged.

(6.) Having established that the savings rate is 1(0) while the log of real income appears to be 1(1).

REFERENCES

Andrews, D.W.K. (1993), Tests for parameter instability and structural change with unknown break point', Econometrica, 59, pp. 817-58.

Kaldor, N. (1966), Causes of the Slow Rate of Growth of the United Kingdom, Inaugural Lecture, Cambridge University.

-----(1967), Strategic Factors in Economic Development, Frank Pierce Memorial Lecture, Cornell University.

Matthews, R.C.O. (1968), 'Why has Britain had full employment since the war?', Economic Journal, 78, 311, September, pp. 555-69.

Matthews, R.C.O., Feinstein, C.H. and Odling-Smee, J.E. (1982), British Economic Growth 1856-1973, Oxford, Clarendon Press.

Schwarz, G. (1978), 'Estimating the dimension of a model', Annals of Statistics, pp. 461-4.

Sefton, J. and Weale, M. (1995), Reconciliation of National Income and Expenditure: Balanced Estimates of U.K. National Accounts, 1920-1990, Cambridge, Cambridge University Press.

Solomou, S. and Weale, M. (1997), 'Personal sector wealth in the United Kingdom, 1920-1965', Review of Income and Wealth, 43, pp. 297-318.

Wilson, T. (1952), 'Manpower', in Worswick, G.D.N. and Ady, P.H. (eds). The British Economy 1945-50, Oxford, Oxford University Press, pp. 223-52.

Wright, J.F. (1979), Britain in the Age of Economic Management an Economic History since 1939, Oxford, Oxford University Press.