statistics and the german state

Introduction Today, statistics de®ne our knowledge of the economy. The countries of the world rank themselves in terms of their gross national product (GNP). Indicators such as the Retail Price Index (RPI) are used routinely in the regulation of everyday life. New numbers are news. The calendar of statistical publications provides grist to the mill of ®nancial speculation and business planning. Numerical representations shape our conception of the economy in subtle but profound ways. Statistics reinforce our sense of the economy as a realm apart from other spheres of life. The economy has assumed the status of a substantive entity, even an actor: an actor, however, who moves in one dimension. We speak of unemployment as `going up' or `down', `rising' or `falling'. In large part, this is surely because we think of unemployment as a statistic or a graph plotted over time. It would be eccentric to describe unemployment as `spreading'. Whereas an earlier language spoke of economic expansion, or progress, the master term in our vocabulary is `growth'. Statistics also shape our understanding of economic history. The reconceptualization of the economic past in terms of macroeconomic data has come close to obliterating the `industrial revolution'.1 The dramatic story of Arkwright and the dark satanic mills has been replaced by a narrative of undramatic growth in large statistical aggregates such as industrial production. This book is driven by the desire to understand how this peculiar structure of economic knowledge came into existence. In pursuit of this larger question, it explores the making of modern economic statistics in Germany in the ®rst half of the twentieth century. How to justify this narrow focus in time and place? When we scratch the surface we discover that modern economic statistics are of surprisingly recent origin. The ®rst recognizably modern statistical projects in Europe date to the birth of the modern state in the 1 D. Cannadine, `The Present and Past in the English Industrial Revolution', Past and Present, 103 (1984), pp. 131±172 and M. Berg and P. Hudson, `Rehabilitating the Industrial Revolution', Economic History Review, 45 (1992), pp. 24±50. 2 Statistics and the German State, 1900±1945 seventeenth century.2 The intrusive policies of absolutism made censuses a regular event in the eighteenth century. But it was the revolutions of the late eighteenth century, which gave shape to of®cial statistics in the form we know today. In 1787 the constitution of the newly independent United States called for a regular census to establish the membership of the House of Representatives. In 1800 revolutionary France established the ®rst `Bureau de statistique'.3 This was enough to persuade counter-revolutionary Britain. In 1753 the Houses of Parliament had rejected a census as an unwanted intrusion upon `English liberty'. In 1801 the ®rst modern census of population went ahead almost entirely unopposed. Statistical of®ces were established in Prussia in 1805, in Bavaria in 1806, in 1810 in Habsburg Vienna, in 1820 in WuÈrttemberg, in 1826 in the Netherlands and in 1831 in the newly independent Belgium. The British Board of Trade established its statistical department in 1832. Five years later demographic statistics were placed under the control of the Registrar General.4 Russian administrative statistics were put on an institutional footing in 1834. In 1833 Denmark set up a Central Statistical Commission, followed by Norway in 1837. Finland was the last of the Scandinavian countries to establish a statistical of®ce in 1865. The constitution of the `double-monarchy' was shortly followed in 1867 by the formation of an Hungarian statistical bureau. The provisional Republic put Spain on the statistical map in 1873. Inspired by the ideas of Saint-Simon the ¯edgling Greek state had set up a statistical section as early as 1834. In 1850 funds were ®nally appropriated to establish a semi-permanent of®ce of the census for the United States. By this time, no self-respecting state administration did without some kind of statistical equipment. This early history of statistics is a ®eld that has recently begun to attract historians.5 We have studies of social statistics, demography and the techniques of mathematical statistics. But, strangely enough, despite their obvious importance, the history of economic statistics remains 2 M. Rassem (ed.), Statistik und Staatsbeschreibung in der Neuzeit (Paderborn, 1980). For a general discussion see S.J. Woolf, `Statistics and the Modern State', Comparative Studies in Society and History, 31 (1989), pp. 588±604. 3 J.-C. Perrot and S.J. Woolf, State and Statistics in France, 1789±1915 (London, 1984); M.-N. Bourguet, DeÂchiffrer la France: la statistique deÂpartementale aÁ l'eÂpoque napoleÂonienne (Paris, 1988) and J. DupaÃquier and M. DupaÃquier, Histoire de la DeÂmographie (Paris, 1985), pp. 256±274. 4 P. Corrigan and D. Sayer, The Great Arch: English State Formation as Cultural Revolution (Oxford, 1985), pp. 124±125. 5 T.M. Porter, Trust in Numbers. The Pursuit of Objectivity in Science and Public Life (Princeton, 1995); S. Patriarca, Numbers and Nationhood: Writing Statistics in NineteenthCentury Italy (Cambridge, 1996); A. DesrosieÁres, La politique des grands nombres. Histoire de la raison statistique (Paris, 1993). Introduction 3 underexplored. We do have a number of excellent studies of econometrics.6 They trace the development of the mathematical techniques used to manipulate data and to test economic theories. These studies are fascinating in their own right. For the uninitiated they provide an excellent historical introduction to an arcane discipline. However, they are largely orientated towards the preoccupations of the discipline of econometrics itself. At their heart is the conversation between a series of canonical ®gures: Slutsky, Frisch, Tinbergen and Haavelmo. This book pursues an agenda which is different but complementary. It is concerned not with statistical techniques but with the production of factual economic knowledge. It makes a ®rst attempt to map out the development of the repertoire of modern economic statistics. Its subject matter is therefore more mundane in all senses of that word: more commonplace, but also more popular and widespread. The history of statistical facts cannot be written without reference to the history of statistical techniques, but the history of factual economic knowledge demands a wider approach. The statistical data discussed here are treated not as the sole property of academics but as an integral part of the economic and social world, which they seek to describe. As is suggested by the title, this book analyses how the German state set about making a modern form of economic knowledge. Statistics are not neutral re¯ections of social and economic reality. They are produced by particular social actors in an effort to make sense of the complex and unmanageable reality that surrounds them. The most fundamental aim of this book is to show that historical statistics should not therefore be relegated to footnotes, encapsulated within tables or consigned to appendices. They should be treated like other cultural artefacts, texts or images. Their history should be integrated within the wider history of the society that produces them. To historians of the medieval or early modern periods, in which the practices of quanti®cation ®rst became established, this need to treat statistics as cultural artefacts will be entirely obvious. For modern economists and economic historians the effect of contextualization may be somewhat more jarring: the independent status of our disciplines is founded to such a large extent on the authority of statistics. Perhaps it is therefore worth adding a few words of reassurance. By showing that statistical facts are produced by particular actors, in particular contexts, with particular interests, this book does not aim to `debunk' the efforts of economic statisticians; on the contrary. The attitude of this book is pragmatic. Over the last couple of centuries the usefulness of statistics 6 M. Morgan, The History of Economic Ideas (Cambridge, 1990) and J.L. Klein, Statistical Visions in Time. A History of Time Series Analysis 1662±1938 (Cambridge, 1997). 4 Statistics and the German State, 1900±1945 has surely been demonstrated beyond reasonable doubt. They are now so ubiquitous in the everyday practices of economic life that the idea of writing modern economic history without reference to statistics amounts to romantic nostalgia. Whether or not statistical facts can claim the status of `truth' or `objectivity' in some metaphysical sense is irrelevant for all practical purposes. This book, in any case, is not concerned with questions of philosophy. If there is a critical edge to the argument, it is political. The questions with which we will be concerned are about the relationship between practical knowledge and power ± and, in particular, the relationship between efforts to govern the economy and efforts to make the economy intelligible through systematic quanti®cation. I The systems of economic statistics, that we take for granted today, emerged across the industrialized world as the result of a dramatic burst of innovation. This began tentatively in the 1870s and gathered momentum around the turn of the century. The most intense phase of activity was unleashed by World War I. Three decades later, in the 1950s, the process culminated with the global standardization of the modern repertoire of macroeconomic statistics.7 In less than a century, the state of empirical economic knowledge was radically transformed. The result was a new empirical image of the economy. We can solidify this chronology with a brief comparative history of four key elements in this new matrix of statistical knowledge: the balance of payments, unemployment, prices and national income. Trade statistics are the oldest economic statistics. It might therefore be argued that they ®t least well with the modern chronology set out here. Records of goods crossing the borders of states and statelets go back to the early modern period. However, these data were compiled for administrative purposes. When did customs records become trade statistics as we know them today? In the British case, which may be taken as representative of the ®rst generation of nation-states, systematic records of trade began to be collected in the seventeenth century.8 But, at ®rst, no attempt was made to record the value of exports and imports in current terms. Throughout the eighteenth century the unit prices of 7 The term `macroeconomics' is used to distinguish aggregative economic analysis of all kinds from microeconomics, which focuses on individual economic agents and their interactions. 8 R. Davis, The Industrial Revolution and British Overseas Trade (Leicester, 1979), pp. 77±86 and A. Maizels, `The Overseas Trade Statistics of the United Kingdom', Journal of the Royal Statistical Society, 112 (1949) II, pp. 207±223. Introduction 5 imports and exports were ®xed at so-called `historic values' set in the 1690s. By the late eighteenth century these were wildly inaccurate. It was only in 1798 that export ®gures began to be compiled in current values. Of course, it was possible for economists and journalists to make their own estimates of the balance of trade, but these were unauthorized interpretations of the of®cial data. Of®cial import ®gures were not ®nally valued in current prices until the 1850s. The modern procedure for calculating the trade balance from customs declarations for both exports and imports was instituted only in 1869. From this point onwards one can de®nitely speak of an of®cial estimate of the balance of trade. On closer inspection, the history of British trade statistics thus falls into line with the chronology for late-developing European nationstates, such as Germany. The trade accounts of the Zollverein set up in the 1830s were really no more than spin-offs of the customs system.9 They covered only those goods on which duty was charged. It was only in 1879 that all goods crossing the borders of the new German Empire were systematically registered, classi®ed and valued. Between the 1840s and 1880 24 countries established reliable trade statistics. By 1913 this number had increased to 33 and by the 1920s the trade between 90 countries could be monitored in statistical terms.10 The inter-national economy was thus de®ned as a space of trading relationships between more or less clearly de®ned national economic units tied together by the well-monitored movements of goods. In statistical terms, however, the monitoring of this inter-national economy was still incomplete. The trade balance was only one part of an increasingly complex network of international economic transactions. Trade in services, earnings on foreign investments and international borrowing and lending matched visible trade ¯ows. The practice of international ®nance was, of course, well understood by contemporaries. However, it was not until the 1870s that economic theorists began to systematically integrate the balance of payments into their models, embracing both trade transactions and capital movements. It was the arguments over international ®nance and reparations in the 1920s that gave birth to international economics in the form that is still taught today.11 During that turbulent decade, the state of statistical information on the balance of payments remained, in the words of John Maynard Keynes `deplorably de®cient . . . in search of facts of vital 9 W. Heimer, Die Geschichte der deutschen Wirtschaftsstatistik von der GruÈndung des Deutschen Reichs bis zur Gegenwart (Frankfurt, 1928), pp. 10±33. 10 E. Wagemann, Wagen, WaÈgen und Wirtschaften. Erprobte Faustregeln ± Neue Wege (Hamburg, 1954), p. 72. 11 M.J. Flanders, International Monetary Economics, 1870±1960. Between the Classical and the New Classical (Cambridge, 1989). 6 Statistics and the German State, 1900±1945 national importance, we . . . continue to grope in barbaric darkness'.12 In response to this situation, the League of Nations began to compile estimates of the balance of payments for the leading nations in the early 1920s. And in 1926 the United States Commerce Department issued the ®rst of®cial estimates of the US foreign account. Germany followed suit later in the decade. But Britain did not ®nally begin a regular series of of®cial estimates until after World War II.13 One of the ®rst tasks of the International Monetary Fund (IMF) in the late 1940s was to formulate global standards for the measurement of the balance of payments.14 Despite the early development of trade statistics, the modern system of international economic statistics took shape between the mid-nineteenth century and the 1950s. The statistical de®nition of unemployment falls in the same broad period. In the late nineteenth century, unemployment was viewed primarily as an issue of social policy. Analysts were primarily concerned with `the unemployed' rather than `unemployment'. Joblessness was attributed to the feckless character of the jobless, or to the peculiar problems of casual labour markets. By contrast, the concept of unemployment that emerged in the aftermath of World War I was de®ned primarily in economic terms. This has been a common ®nding of historical research on Britain, France, the United States and Germany.15 Unemployment was reconceived as a mismatch between the demand for and the supply of labour. Its primary explanation therefore lay not in individual behaviour but in wider economic problems. Of course, the social concern for the unemployed remained. Moral denunciations of the work-shy were never completely silenced. But the fundamental cause of unemployment was now seen as economic. This new understanding was underpinned by the increasing organization of the labour market, which in turn permitted the creation of new unemployment statistics. In Britain and Germany the innovations bunch around World War I. Uni®ed labour exchanges were established by Act of 12 J.M. Keynes, `The British Balance of Trade, 1925±27', Economic Journal, 37 (1927), pp. 551±565. 13 C.F. Carter and A.D. Roy, British Economic Statistics. A Report (Cambridge, 1954), pp. 79±93. 14 F. Machlup, `Three Concepts of the Balance of Payments and the So-Called Dollar Gap', Economic Journal, 40 (1950), pp. 46±68. 15 On Britain see J. Harris, Unemployment and Politics (Oxford, 1972) and W. Walters, `The Discovery of ``Unemployment'': New Forms for the Government of Poverty', Economy and Society, 23 (1994), pp. 265±290; on the United States see A. Keyssar, Out of Work. The First Century of Unemployment in Massachusetts (Cambridge, MA, 1986); on France R. Salais, N. Bavarez and B. Reynaud, L'invention du choÃmage (Paris, 1986); on Germany A. Faust, Arbeitsmarktpolitik im Deutschen Kaiserreich. Arbeitsvermittlung, Arbeitsbeschaffung und ArbeitslosenunterstuÈtzung 1890±1918 (Stuttgart, 1986). Introduction 7 Parliament in Britain in 1909 and unemployment insurance followed in 1911. In Germany a uni®ed national system of labour exchanges was established in 1915, which was brought fully under state control in 1922. Unemployment insurance followed in 1927. The insurance system generated data on the numbers in work and the numbers receiving bene®ts. The exchanges registered job seekers and vacancies. France and the United States monitored the labour market less closely but here too there was increasingly regular and intensive statistical measurement from the turn of the century. By the interwar period the masses of men and women unsuccessfully seeking work had been established as a phenomenon demanding economic analysis. The history of price statistics and national income accounting help further to solidify our chronology. In 1922, Irving Fisher, the foremost exponent of index numbers, wrote as follows: `index numbers are a very recent contrivance . . . although we may push back their invention a century and three quarters, their current use did not begin till 1869 at the earliest, and not in a general way till after 1900. In fact, it may be said that their use is only seriously beginning today.'16 In the 1860s and 1870s mathematicians and economists including Jevons, Paasche and Laspeyres experimented with the construction of index numbers. The weekly periodical The Economist published the ®rst regular price index in 1869.17 But the really dramatic upsurge in interest came in the late 1890s when generalized de¯ation of prices gave way to creeping in¯ation. The response, this time, came not just from private investigators and journalists but also from the state. The US Bureau of Labor published the ®rst of®cial index of wholesale prices in 1902. Retail price ®gures and a cost of living index followed in 1907 and 1919, respectively. The British Board of Trade was also a pioneer, producing the ®rst wholesale price index for Britain in 1903. The in¯ations of World War I triggered a boom in index numbers. By 1927 Fisher was able to list 120 price indices covering no less than 30 countries, published by of®cial statisticians, business periodicals, daily newspapers and large corporations. And it is not just the proliferation of numbers that should interest us. The new indices had a new economic content. The earliest indices were simple averages of commodity prices; by contrast, the wholesale indicators produced by the United States and Britain after the turn of the century were far more sophisticated. They were weighted averages, giving greater signi®cance to some prices rather than others. And there was a clear economic logic behind their construction. The weights 16 I. Fisher, The Making of Index Numbers. A Study of their Varieties, Tests and Reliability (Boston, 1927, 3rd edn), p. 460. 17 The Economist 1843±1943 (Oxford, 1943), pp. 138±154. 8 Statistics and the German State, 1900±1945 attached to the Board of Trade's wholesale price index were calculated to re¯ect the total consumption of key commodities by the economy as a whole. More speci®c indices covered the cost of living and retail prices. These were the ®rst attempts to make visible Adam Smith's `invisible hand'. National income statistics are the last and really conclusive piece of evidence.18 As with the other statistics, one can ®nd early attempts at national income estimation as far back as the seventeenth century. But the early twentieth century witnessed a sudden explosion in activity with new estimates being produced by academics and journalists and then increasingly by of®cial agencies. Studenski's pioneering study provides a truly extraordinary overview (see table 1). In 1900 estimates of national income had been prepared for no more than eight countries. By 1946 there were ®gures ± of®cial and unof®cial ± for 39 countries. Ten years later, there were more than 80. And here, too, the qualitative change in the data was dramatic. Estimates of national income produced up to the late nineteenth century tended to be crude extrapolations from fragmentary tax records. The questions they sought to answer were distribu18 They are also the one branch of economic statistics to have attracted sustained historical attention. See F. Fourquet, Les comptes de la puissance. Histoire de la comptabilite nationale et du plan (Paris, 1980) and M. Perlman, `Political Purpose and the National Accounts', in W. Alonson and P. Starr (eds.), The Politics of Numbers (New York, 1987). Table 1. Date of ®rst publication of of®cial estimate of national income Year Country 1886 Australia 1925 Soviet Union and Canada 1929 Germany 1931 Netherlands 1931 New Zealand 1934 United States 1935 Turkey 1937 Yugoslavia 1939 Switzerland and Mexico 1941 United Kingdom 1944 Sweden and Norway 1947 France Sources: P. Studenski, The Income of Nations (New York, revised edn. 1958), I, pp. 151±153; F. Fourquet, Les comptes de la puissance. Histoire de la comptabilite nationale et du plan (Paris, 1980). Introduction 9 tional. What share of income was attributable to the `unearned' rents of landowners? How was the remainder divided between capital and labour? Economic statistics were thus orientated towards `social' questions. The 1920s saw the emergence of a more purely `economic' interpretation of national income. With the advent of comprehensive censuses of production it became possible to match the ®gures for national income with estimates of national product. As a result, the interpretative focus began to shift away from issues of distribution towards primarily `economic' concerns, such as the comparative level of productivity in different sectors and the ¯uctuations over the businesscycle of total economic activity, as measured by national income or national product. This shift was completed in the late 1920s and early 1930s with the ®rst estimates of total expenditure, divided principally into consumption, investment and government expenditure. It now became possible to picture the economy, in statistical terms, as a selfcontained `circular ¯ow' of production, income and expenditure. This image, ®rst made real in the interwar years, has since occupied the ®rst pages of every textbook in macroeconomics. Taken together these interrelated statistical innovations constituted a new matrix of economic knowledge, which gave substance to a new conception of the economy.19 First of all `the economy' was envisioned as a separate system, distinct, for instance, from `the social', `the cultural', or `the political'. It was a measurable entity, a `thing'. This conception of `the economy' as an autonomous social system was more restricted than that embodied in eighteenth-century ideas of a commercial society, or Marx's totalizing conception of the mode of production. But it was also more concrete than those earlier formulations. Linguistic changes signal this shift to a more rei®ed idea of the economic world. In German it was already possible in the mid-nineteenth century to speak of the `Volkswirtschaft', or national economy.20 In England, true to its liberal heritage, `the economy' as a term with which to refer to the entire system of production and exchange did not come into common use until the 1930s.21 What de®ned this entity was the relationship between a limited number of key variables: national income, physical production, employment, the balance of payments, the volume of money in circula19 P. Miller and N. Rose, `Governing Economic Life', Economy and Society, 19 (1990), pp. 1±31. 20 J. Burckhardt, `Wirtschaft', in O. Brunner, W. Conze and R. Koselleck (eds.), Geschichtliche Grundbegriffe. Historisches Lexikon zur politisch-sozialen Sprache in Deutschland (Stuttgart, 1992), 7, pp. 511±594. 21 M. Emmison, ```The Economy'': Its Emergence in Media Discourse', in H. Davis and P. Walton (eds.), Language, Image, Media (Oxford, 1983), pp. 139±155. 10 Statistics and the German State, 1900±1945 tion and the aggregate price level. As we have seen, the measurement of each one of these variables had its separate history. The distinctively modern conception of the economy emerged when they began to be articulated with each other as an interconnected system. The interrelationships were established through the revival in the last decades of the nineteenth century of two of the founding metaphors of modern economics. The most fundamental of these was the conception of the economy as a self-reproducing, circular ¯ow of production and consumption, of expenditure and income. A second key metaphor was the so-called Quantity Theory of Money, which also enjoyed a major revival in the late nineteenth century.22 This expressed the value of money (the inverse of the aggregate level of prices) as a function of the quantity of money, the rate of its circulation and the level of real economic activity. This relationship was recast in the 1880s as an algebraic equation and acquired canonical status in 1911 with Irving Fisher's The Purchasing Power of Money. We shall have much more to say about both these representations of the economy. Suf®ce to say at this point that they allowed the key economic variables to be brought together as elements in a systematic, aggregative model of the economy. The result was a conception of the economy which since the 1930s has become known as `macroeconomic'. This new conception of the economy was emphatically national. This, too, was an option that had been left open by earlier theorizing. For liberals any boundaries imposed on the free operation of markets were arti®cial intrusions. Similarly, Marx's conception of the mode of production was potentially global in scope. By contrast, the new economic statistics measured the economy as a national unit. And in doing so they constituted it as an obvious ®eld of government action. The creation of the new economic statistics was inseparable from the appearance of a new set of practices known as `economic policy'. As Donald Winch has put it, our modern conception of `economic policy' emerged `out of elements that had previously been treated separately as questions of social administration on the one hand or narrow technical matters of banking and ®scal management on the other'. The new purpose of economic policy was precisely to manage the `connections between employment levels and monetary, exchange rate, and ®scal conditions'.23 Winch dates the emergence of this new ®eld of government to the interwar period, which of course coincides with the appearance of 22 D. Laidler, The Golden Age of the Quantity Theory (Princeton, 1991). 23 D. Winch, `Economic Knowledge and Government', in B. Supple and M. Furner, The State and Economic Knowledge. The American and British Experiences (Cambridge, 1990), pp. 62±63. Introduction 11 the new economic statistics. Economic policy presupposed the existence of a new object of government: an economy conceived of not as an amorphous mass of individuals and markets, but as a holistic entity constituted by the relationship between a limited number of highly aggregated variables. The task of the new economic statistics was to measure these variables and thus to make them governable. II How does this book relate to the existing literature on the history of economics? In particular, some readers may be wondering whether this is simply a retelling of the familiar story about the so-called `Keynesian revolution'. For a long time this has been the mainstay of the history of modern economic thought. The General Theory of Employment, Interest and Money published by John Maynard Keynes in 1936 is generally taken to be the founding text of modern macroeconomics. Its intellectual genesis through the 1920s and 1930s has been an obvious starting point for historians. It is impossible to adequately summarize this enormous literature in a few lines.24 However, the gist of most recent writing is that Keynes' central contribution was theoretical. The General Theory explained how an economy suffering from a shortfall in aggregate demand could ®nd itself in a state of heavy unemployment, from which it had no tendency to recover. The low level of activity brought on by a depressed level of investment became self-sustaining.25 It was this theoretical analysis of an unemployment equilibrium which gave Keynes' heretical policy prescriptions their originality and force. When the economy was seriously depressed monetary policy would be ineffective. Lowering interest rates would not be enough to raise investment. Government spending was essential to raise aggregate demand and to lift the economy out of recession. It was Keynes' long-running struggle with the Treasury that, in the British case, de®ned the new ®eld of economic policy. In the United States the battle for demand management was fought out within the New Deal administration. In the aftermath of World War II Keynes' ideas were carried across the globe, establishing the common sense of the postwar period.26 24 For two excellent summaries see G.C. Peden, Keynes, The Treasury and British Economic Policy (London, 1988) and P. Clarke, The Keynesian Revolution and its Economic Consequences (Cheltenham, 1998b). 25 Technically speaking this is the `theory of effective demand', see D. Patinkin, Anticipations of the General Theory. And other Essays on Keynes (Chicago, 1982), pp. 5±17. 26 P.A. Hall (ed.), The Political Power of Economic Ideas. Keynesianism across Nations (Princeton, 1989). 12 Statistics and the German State, 1900±1945 The almost obsessive focus on Keynes has produced a historiography of a remarkably high standard. But it has also served to obscure the wider context.27 As we have progressively sharpened our understanding of the speci®cities of Keynes' theoretical innovation and the complexities of his own intellectual biography, it has become ever more clear that Keynes' work must be situated within a broad sweep of new macroeconomic theorizing that can be traced back to the 1870s. It is more conventional to see the 1870s as the origin of modern marginalist economics. It was in this period that the building blocks of neo-classical economics were ®rst formulated by Jevons, Menger and Walras: the attribution of factor incomes to marginal productivity, the reformulation of demand theory in terms of consumer preferences and subjective utility, and the consistent linkage of the structure of production to the structure of demand through general equilibrium analysis. However, as David Laidler has shown in his study of neo-classical monetary theory, the 1870s can also be seen as the origin of twentieth-century macroeconomics.28 It was in the decades after 1870 that theorists such as Alfred Marshall, Irving Fisher and Knut Wicksell elaborated the Quantity Theory of Money into a consistent and powerful tool for understanding movements in the aggregate price level. Building on this analytical framework a second generation of theorists, notably in Britain and Austria, began, around the turn of the century, to elaborate what became known as monetary business-cycle theory. Their models were emphatically macroeconomic, their purpose being to explain the interaction between monetary ¯uctuations and movements in total production and employment.29 As Laidler has pointed out, it is quite possible to see even Keynes' General Theory as an extension of this tradition. Certainly, Keynes' earlier work can be seen as a linear development of Marshallian monetary macroeconomics. The development of mathematical techniques for analysing statistical data and testing theory ± the so-called econometric revolution ± was heavily in¯uenced by these early developments in monetary economics and business-cycle theory. It was interest in the ¯uctuations of prices that stimulated Jevons and Juglar to undertake the ®rst time-series analysis of price data in the 1860s. As has already been mentioned, it was the switchback of de¯ation in the 1870s and 1880s followed by in¯ation from the late 1890s that stimulated the development of index 27 For a powerful summary see R. Middleton, Charlatans or Saviours? Economists and the British Economy from Marshall to Meade (Cheltenham, 1998). 28 Laidler, The Golden Age, pp. 193±199. 29 In the British case the outstanding examples are A.C. Pigou, Wealth and Welfare (London, 1912), R. Hawtrey, Good and Bad Trade (London, 1913) and D.H. Robertson, A Study in Industrial Fluctuations (London, 1915). Introduction 13 numbers designed to accurately re¯ect the movements in the general purchasing power of money. By the 1920s the basic techniques for estimating trends and removing seasonal variations were well established. In the 1930s it was the desire to test monetary explanations of the business-cycle that stimulated Tinbergen and the Dutch statistical of®ce to construct the ®rst genuine mathematical model of an entire economy. And it was the econometrician, Ragnar Frisch, who ®rst introduced the term `macro-dynamics' into the literature, thus giving rise to the more familiar term macroeconomics.30 Again, this is a story which can be told almost entirely without reference to the Keynesian revolution. It is these twin `revolutions' in macroeconomic theory and in econometrics that provide the intellectual context for this book, not the Keynesian revolution per se. The explosion of new economic statistics between the 1870s and the 1950s deserves to be treated as a `statistical revolution' in its own right.31 As will be shown here, the development of the infrastructure of data-gathering had its own distinct history. But one gets a proper sense of the transformation of economic knowledge in this period only if one understands how the three revolutions were interrelated. The concepts that informed practical efforts at data collection were developed in dialogue with economic theory. And it was the new data produced by the statisticians that provided the material for the econometricians. The interweaving of these three separate strands constituted modern macroeconomic knowledge. To analyse this process of multiple innovation we need a new analytical model. The existing literature describes the Keynesian revolution in terms of a process of diffusion. At its core are Keynes and his intimates in the `Cambridge circus'. Out of this incestuous milieu sprang The General Theory. The central question for historians is to understand how economists and policy-makers across the world `reacted' to the provocation of this revolutionary book. In the language of the literary scholar Franco Moretti, this is a `tree model' of cultural development.32 Branches, stems and shoots sprout from the Cambridge trunk. By contrast, Moretti suggests that comparative cultural historians 30 See J.C. Andvig, `Ragnar Frisch and Business Cycle Research during the Interwar Years', History of Political Economy, 13 (1981), p. 713. Frisch applied the label `macrodynamics' to those studies of the business-cycle that focused on national aggregates, as opposed to those that focused on disequilibria in speci®c industries. The broader term, macroeconomics, was introduced in 1941, see Clarke, The Keynesian Revolution and its Economic Consequences, p. 213. 31 By contrast with Patinkin, Anticipations of the General Theory, pp. 223±260, who collapses the production of new economic statistics and the development of statistical techniques into a single `econometric revolution'. 32 F. Moretti, `Conjectures on World Literature', New Left Review, II, 1 (2000). 14 Statistics and the German State, 1900±1945 should adopt the metaphor of the wave. And this certainly seems a more appropriate concept on which to base the study of modern economic knowledge. In the ®rst half of the twentieth century, innovations in the conceptualization and measurement of the economy swept across the globe. Britain and the United States are well established as independent sites of theoretical and empirical innovation in the interwar years.33 But the war and the revolution of 1917 also ushered in a feverish period of innovation in Communist Russia.34 This included the construction of models of economic growth and elaborate and entirely unprecedented systems of national accounting.35 These experiments were terminated between 1928 and 1930 by the Stalinist crackdown. But, while they lasted, they formed an integral part of an international process of innovation. The Soviet economists followed developments in the West closely. Their work, in turn, had a considerable impact abroad. As will be discussed in chapters 3 and 5, the Weimar Republic appears to have played a strategic role in this transmission of ideas, through the rapid translation into German of Russian publications. Simultaneously, there emerged in Sweden a powerful line of macroeconomic business-cycle analysis ± known as the Stockholm school. Unfortunately, discussion of the Stockholm school has, until recently, centred around one question: did Wicksell, Lindahl, Myrdal and Ohlin anticipate Keynes' theory of effective demand, as set out in The General Theory?36 Probably not. But from the broader point of view adopted here, this matters little. The Stockholm school undoubtedly included some of the most sophisticated exponents of the new macroeconomics. Their concern was to account for ¯uctuations in overall economic activity. Like most of the early generation of macroeconomists their focus was on the aggregate price level. But during the 1930s the younger Swedish economists also turned to the questions of output and employment that were preoccupying Keynes.37 Accompanying this theoretical work was a parallel programme of empirical enquiry. Most notably a large grant from the 33 On the United States see G. Alchon, The Invisible Hand of Planning. Capitalism, Social Science, and the State in the 1920s (Princeton, 1985), W.J. Barber From New Era to New Deal. Herbert Hoover, The Economists, and American Economic Policy, 1921±1933 (Cambridge, 1985). 34 See L. Smolinski, `Planning Without Theory 1917±1967', Survey. A Journal of Soviet and East European Studies, 64 (1967), pp. 108±128. 35 N. Spulber (ed.), Foundations of Soviet Strategy for Economic Growth. Selected Soviet Essays, 1924±1930 (Bloomington, 1964) and V. Barnett, Kondratiev and the Dynamics of Economic Development: Long Cycles and Industrial Growth in Historical Context (London, 1998). 36 Patinkin, Anticipations of the General Theory, pp. 36±57. 37 L. Jonung (ed.), The Stockholm School of Economics Revisited (Cambridge, 1991). Introduction 15 Rockefeller Foundation enabled Swedish statisticians to compile one of the longest series for national income available in the 1930s.38 The steady pace of Swedish expansion, untroubled by major wars, provided the ideal case study of stable, long-run growth trend. The history of econometrics reveals two other sites of innovation. In Norway Ragnar Frisch was pivotal to the development of the theory of modern econometrics.39 He formed a crucial link between the theoretical work of the Russian statistician Slutsky, who was one of the few survivors of the Stalinist purges, and the Cowles Commission in Chicago that was to set the agenda for postwar econometrics. In the Netherlands, meanwhile, Jan Tinbergen and the Dutch Statistical Of®ce began work on the world's ®rst macroeconomic model.40 With the assistance of the League of Nations his techniques were later to be extended to modelling the US economy. This book aims to establish the existence in Germany of another major strand of `new economics'. Germany has hitherto played a shadowy role in debates about interwar economics. It has always been tempting to seek out precursors of Keynes amongst the German advocates of work-creation in the 1930s.41 However, this line of enquiry has proven an intellectual dead-end. The foundation for a more adequate understanding has now been provided by a number of important studies.42 These provide a panoramic reconstruction of the intellectual ®eld of German economic theory in the interwar period. The result, as in the United States, Sweden and Britain, has been to reveal a broadly based tradition of monetary macroeconomics. This originated in the decades before World War I, and by the 1920s had reached a considerable level of sophistication. This book hopes to consolidate this reassessment of economics in interwar Germany. It reveals how the new macroeconomic theory formed the basis for an innovative programme of 38 E. Lindahl, E. Dahlgren and K. Koch, National Income of Sweden 1861±1930 (Stockholm, 1937). 39 Andvig, `Ragnar Frisch'. 40 A. Wilts, `Changes in Dutch Economics in the 1930s', in P. Fontaine and A. Jolink (eds.), Historical Perspectives on Macroeconomics. Sixty Years after the General Theory (London, 1998), pp. 105±132. 41 G. Garvy, `Keynes and the Economic Activists of Pre-Hitler Germany', Journal of Political Economy, 83 (1975), pp. 391±405 and G. Bombach, K.-B. Netzband, H.-J. Ramser and M. Timmermann (eds.), Der Keynesianismus III. Die geld- und beschaÈftigungstheoretische Diskussion in Deutschland zur Zeit von Keynes (Berlin, 1981). 42 R. Vilk, Von der Konjunkturtheorie zur Theorie der Konjunkturpolitik (Wiesbaden, 1992), H. Janssen, NationaloÈkonomie und Nationalsozialismus. Die deutsche Volkswirtschaftslehre in den dreiûiger Jahren (Marburg, 1998) and H. Hagemann, `The Analysis of Wages and Unemployment Revisited: Keynes and Economic ``Activists'' in Pre-Hitler Germany', in L.C. Pasinetti and B. Schefold (eds.), The Impact of Keynes on Economics in the 20th Century (Cheltenham, 1999), pp. 117±130. 16 Statistics and the German State, 1900±1945 statistical investigation, heavily sponsored by the Weimar state. Germany must thus be counted alongside the United States, the Soviet Union, Britain, Sweden and the Netherlands as an important site in the development of modern macroeconomic statistics. This, in turn, should consolidate the more general shift in perspective being advocated here. The development of new forms of economic knowledge was too widespread to be described helpfully in terms of a process of diffusion. German-speaking economists read the theoretical and statistical work being published in Britain as a matter of course. But these in¯uences were seen as part of a more general move towards an aggregative conception of the economy. In the 1920s, when German macroeconomics began to be articulated most forcefully, Cambridge, England commanded no outstanding place in its intellectual universe. From a German perspective the emergence of the new economics appeared to be a truly global phenomenon. And the breadth of this development also implies that its end-point was uncertain. Teleology is one of the characteristic weaknesses of the literature on `proto-Keynesian' and `pre-Keynesian' economic thought. A case study of Germany provides a powerful antidote. The ®rst half of this book traces the development of Germany's precocious macroeconomic statistics to familiar intellectual origins in the quantity theory of money. After World War I this aggregative understanding of the economy was embodied in an innovative system of of®cial national accounts. Germany might therefore be seen as travelling along a path that led to the `Keynesian consensus' of the postwar period. However, after 1933 Germany took a radically different direction. The Reich's statisticians placed themselves at the service of the Nazi regime. As a result, the system of macroeconomic statistics created in the 1920s began to mutate into something quite different. By the ®nal stages of the war German statisticians were beginning to elaborate a system of comprehensive surveillance that resembled a full-blown system of Stalinist planning. The German case thus forms a bridge between the development of new techniques of economic governance in the capitalist West and that other great experiment in modern government in the East. Interwar modernity was multi-faceted, its ultimate destination uncertain. A case study of Germany provides a powerful reminder of this contingency. Understanding the development of modern economic knowledge as a wave of innovation rather than a process of diffusion requires us to stretch the chronological frame and expand our geographic range. It also requires us to problematize the postwar `Keynesian consensus' as the inevitable conclusion of our story. More generally it requires us to Introduction 17 rethink our analytical strategies.43 Local accounts need to be controlled by being set against a broader backdrop. Hitherto, the diffusionist story of the Keynesian revolution has tended to focus on the interactions between economists and state elites ± `experts', civil servants and politicians. This nexus must undoubtedly be central to each national story. But, given the ubiquity of new forms of economic knowledge, particular stories that focus on the interactions between small groups of individuals can hardly suf®ce. We must always bear in mind the role played by more general explanatory factors. The rest of this introduction will be concerned with four general in¯uences that have helped to frame this German case study: the transformation of the industrial economies themselves in the period between the 1880s and the 1930s; the crisisridden development of the `big state'; the development of new information technologies; and the cultural and intellectual tendencies encapsulated within the term `modernism'. III The decision to make real economic change the ®rst point on the list may require some justi®cation. In recent years it has been fashionable to adopt a constructivist approach to the study of knowledge. The literature has tended to emphasize the autonomy of the knowledge producers from the object they are observing. It has stressed the discursive construction of new forms of economic knowledge rather than the in¯uences of `real' economic changes.44 In the 1980s there was a parallel shift in the political sciences towards stressing the autonomy of the state. As a result, we have an important collection of essays on The State and Economic Knowledge, but no equivalent volume on `the economy and economic knowledge'.45 This is not to advocate a return to crude determinism. An engagement with the broader currents of economic history follows naturally from studying statistics like other forms of practical economic knowledge, such as accountancy. Statisticians, after all, generate their data not through abstract speculation but through interactions with economic actors themselves, through questionnaires returned by businesses or households, or indirectly by harvesting the data generated by other branches of the state in their dealings with the private sector. Changes in economic life thus have a direct impact on the 43 See the remarks in P. Hall `Introduction', in Hall, The Political Power of Economic Ideas, pp. 3±26. 44 A particularly radical example is P. Mirowski, More Heat than Light. Economics as Social Physics: Physics as Nature's Economics (Cambridge, 1989). 45 M.O. Furner and B. Supple (eds.), The State and Economic Knowledge. The American and British Experiences (Cambridge, 1990). 18 Statistics and the German State, 1900±1945 activity of data-gathering. In recent years this has been brought home very forcibly by discussions of the so-called `weightless economy'. How are statisticians to measure an economy that is increasingly driven by the rapid product cycle of microelectronics and the intangible products of the service sector?46 This book provides the historical backdrop to these present-day concerns. It shows how we learned to measure the `heavy economy'. By the turn of the twentieth century, economic and social development had transformed the conditions for economic data-gathering. For generations, statisticians and economists had dreamed of imposing an orderly scheme of measurement on the world. What distinguished early twentieth-century planners and social engineers from their predecessors was that they could actually do it! In the advanced economies of the world the vast bulk of productive activity was directed towards the market. Most production was separated from the domestic sphere and organized in businesses ± farms, industrial ®rms or commercial businesses ± with a clear-cut legal identity. Communication across the territory of nation-states was eased by the revolution in transport and communication technologies. These in turn encouraged the spread of literacy and numeracy, supported by the extension of formal education. These fundamental processes made the economy countable in a new way. In fact, as Marx pointed out, the economy was quantifying itself in an apparently unstoppable and profoundly alienating fashion. Liberals saw the same process as the progress of rationality; Max Weber, characteristically, saw both sides of the coin: the all-encompassing rationalization of the world and the disenchantment it inevitably entailed. Statisticians and accountants were part of the army of bureaucrats who were the agents of this process. Whereas eighteenth- and early nineteenth-century statistics had relied on impressionistic accounts provided by local notables, statisticians in the early twentieth century could hope to directly enumerate the entire economic process. Questions could be addressed to ®rms and businesses and they could be expected to provide a veri®able account of themselves in a mutually intelligible language. The fringes of the formal economy, such as the homeworkers who had so plagued nineteenth-century enquiries, were shrinking to extinction.47 Counting an economy dominated by a few thousand substantial ®rms presented immense new opportunities. The commanding heights could be surveyed with relative ease. By the 46 J. Madrick, `The Cost of Living', The New York Review of Books, 44, 4 (1997), pp. 19±23. 47 R. Meerwarth, `Die Erfassung der Hausindustrie durch die gewerbliche Betriebsstatistik', JahrbuÈcher fuÈr NationaloÈkonomie und Statistik, III, 42 (1911), pp. 313±330. Introduction 19 interwar period one could survey the vast majority of industrial activity by addressing questions only to ®rms with more than 10 employees, `reasonable' entities with at least a sense of modern managerial habits. Such ®rms could be asked to supply more information, more regularly and more quickly. Not, of course, that all was simple. Large ®rms posed their own problems of enumeration. As production became more complex and more integrated it became increasingly dif®cult to obtain information on separate processes. And this was not just a problem for statisticians looking in. Firms were far from transparent to themselves. Only through elaborate cost accounting systems was it possible for large corporations to `see inside' their own operations. The wave of new economic statistics rode in on the spring tide of modern bureaucracy and scienti®c management. The broad processes of economic and social change thus created new conditions for enumeration. But economic change, like every other aspect of reality, requires interpretation. What sense the statisticians made of economic change was not determined by the process itself. Spokesmen of various kinds played a creative role in naming and interpreting the profound economic transformations going on around them.48 Over the course of the nineteenth century of®cial statisticians came to occupy a particularly important role as interpreters of economic and social change. The German statisticians who are the subjects of this book were convinced that with the boom of the 1890s Germany had entered a new phase of corporate capitalism. Their ambition was to reorganize the system of economic statistics to match this challenge. And the descriptions they produced were not neutral. Statistics and economic research were a weapon of choice in the interest group struggle. Was the future agrarian or industrial? How might small-scale production survive alongside the giants of industry? Which were the industries of the future? People turned to statistics for answers. The statisticians did more than describe; they de®ned the parameters for interest group formation and political argument. In the process, they contributed to the shaping of social reality. IV The growth of modern economic statistics was thus linked to the emergence of the modern economy. But it also clearly belongs to the history of the state. The quantum leap in the production of economic knowledge that is the subject of this book was due largely to the 48 P. Bourdieu, Ce Que Parler veut Dire. L'eÂconomie des eÂchanges linguistiques (Paris, 1982), pp. 135±161. 20 Statistics and the German State, 1900±1945 involvement of the state. Modern economic statistics are an integral part of `big government'.49 The expansion of public sector activity was a general phenomenon observable across the globe from the late nineteenth century. Again, this can be interpreted as an incremental process driven by the demands of a complex civil society for services and regulation. The emergence of `labour statistics', for instance, was clearly a systematic response to the emergence of capitalist labour markets and organized forms of industrial relations. The revival of interest in national income estimation around the turn of the century can also be related to the pressures of `mass politics'. Imperialist nationalism was one factor. A national income estimate allowed one to compare one's position to that of other industrial powers. But the estimates also addressed common domestic concerns. The increasingly intense interest group struggles fermenting within industrializing societies made the language of productivism attractive to both intellectuals and politicians.50 Across the political spectrum the promotion of higher production and greater material welfare as ends in themselves was a characteristic feature of the early twentieth century. Productivism, as a politics of quantity, naturally spoke the language of statistics. How big was the cake that was to be divided? How fast was it expanding? How much did each group contribute? And how many people did it have to feed? These were crucial questions for the new democratic politics. And it was these questions that stimulated efforts to estimate national income in the United States, in the United Kingdom and in Germany.51 In due course it was the state that took responsibility for producing the ®gures that de®ned the parameters of social and economic policy. But the example of national income estimation also points to the independent momentum of state expansion. The production of new statistics was sustained by the growth in other branches of the state. It constituted a form of second-order state expansion. Statistical divisions sprang up to make the most of the paperwork accumulating within the of®cial bureaucracy. Tax records were the fundamental source for most of the early estimates of national income. As the share of national income going through the state coffers increased, it became easier to assemble the data from which to compile a national income estimate. The crucial threshold was the imposition of comprehensive income 49 For a very helpful discussion see R. Middleton, Government versus the Market. The Growth of the Public Sector, Economic Management and British Economic Performance, c. 1890±1979 (Cheltenham, 1996). 50 C.S. Maier, In Search of Stability. Explorations in Historical Political Economy (Cambridge, 1987b), pp. 19±69. 51 J.A. Tooze, `Imagining National Economies: National and International Economic Statistics, 1900±1950', in G. Cubitt (ed.), Imagining Nations (Manchester, 1998). Introduction 21 taxes. But national income estimates were not the only branch of statistics to bene®t from an administrative free ride. Trade statistics, the earliest national economic statistics, were generated by the customs posts strung along the state's frontier. Many of the earliest industrial statistics were compiled from the reports of factory inspectors. Unemployment statistics were a spin-off of the labour administration. As the state expanded it generated within itself multiple re¯ections of economic life. But of®cial statisticians did not simply react to societal demands. The of®cial statisticians discussed in this book were self-conscious bureaucratic actors seeking to expand their administrative province, struggling to ®nd an advantageous position for themselves within the newly extended state. In this minimal sense they enjoyed autonomy from the demands of interested civil society. More generally, it was of®cial statisticians who in many areas took the initiative, acting in advance both of interested groups in civil society and of the rest of the state apparatus.52 Measuring a social or economic phenomenon, be it unemployment or the size of the Jewish population, was often the ®rst step towards de®ning a `problem'.53 This in turn might trigger interventionist activity from within the state or, by exposing the issue to the public gaze, it might generate public pressure for a remedy. As this book will show, the leaders of the Reich's Statistical Of®ce (SRA) pursued this role in a self-conscious fashion. Indeed, their historical self-consciousness blurs the distinction between structure and agency. The strategic action of the Reich's statisticians was motivated by their broader vision of economic development. According to the stage models that many of them espoused, the current phase of capitalism demanded a systematic role for the state in coordinating economic activity. The Reich's statisticians were not merely administrative empire-builders of the type beloved of analysts of `public choice'. They believed themselves to be accomplishing a historic mission in constructing an apparatus of expertise that would mediate between the state and the growing complexity of civil society. The accumulation of state expertise was thus multiply determined. On the one hand it can be traced to the actions of speci®c state elites. But their actions were inspired by a belief in the structural necessity of state expansion. And the existence of the activist elite itself was a product of the general proliferation of state bureaucracy that began in the late nineteenth century. The expansion of of®cial statistics as part of the general expansion of 52 M. Furner and B. Supple, `Ideas, Institutions, and State in the United States and Britain', in Furner and Supple (eds.), The State and Economic Knowledge, pp. 3±39. 53 I. Hacking, The Taming of Chance (Cambridge, 1990), pp. 189±199. 22 Statistics and the German State, 1900±1945 state activity thus had a degree of inescapability. But it was also an uneven process prone to crises and interruptions. Particularly in the aftermath of World War I, it was a contested and complex process.54 Conventionally this tends to be conceived of as a battle in which the remnants of private autonomy succumbed to public interference. Certainly, this is how economic historians describe the growth of the state.55 The share of private incomes taken in taxes increased. The scale of transfers and public procurement rose. Free markets were hedged around with regulations. But this is simplistic. `Big government' was not merely a threat to private freedom. It also called into question the identity of the state.56 The `big state' was a state without the clear contours and the undisputed sovereignty of the nineteenth-century model. Having involved itself in so many aspects of economic and social life the state could no longer claim to stand apart from, let alone above, civil society. To many, the new state appeared overextended, colonized by the con¯icts and tensions of civil society. Nowhere was this crisis more pronounced than in interwar Germany. The myth of the PrussoGerman State as the guiding light of the nation was shattered by defeat in World War I and revolution.57 In the aftermath, the Reich's civil service struggled to reinvent itself. The Weimar Republic experimented with a combination of parliamentary government and technocratic corporatism. Meanwhile, forces on the right demanded the restoration of a `strong state', which in practice actually meant a novel form of authoritarian, militarist dictatorship. And it was this vision that seemed to have triumphed with the `National Revolution' of 1933. But authoritarian conservatives were to be disappointed. The Nazis did not allow themselves to be corralled by the conservative political establishment, the army and the civil service. By the late 1930s they had overcome the resistance of traditional elites and had embarked on a radical project of political and social reconstruction. The Nazi answer to Germany's political crisis was not the creation of a `strong state'. Their goal was the 54 S. Skowronek, Building a New American State. The Expansion of National Administrative Capacities, 1877±1920 (Cambridge, 1982), P. Clarke, `The Twentieth-Century Revolution in Government: The Case of the British Treasury', in Clarke, The Keynesian Revolution and its Economic Consequences, pp. 175±189 and K. Burk (ed.), War and the State. The Transformation of British Government 1914±1919 (London, 1982). 55 H. James, The German Slump. Politics and Economics 1924±1936 (New York, 1986). 56 C.S. Maier, Recasting Bourgeois Europe. Stabilization in France, Germany, and Italy in the Decade after World War I (Princeton, 1988, reprint), M. Geyer, `The State in National Socialist Germany', in C. Bright and S. Harding (eds.), Statemaking and Social Movements: Essays in History and Theory (Ann Arbor, 1984), pp. 193±232, D. Melossi, The State of Social Control. A Sociological Study of Concepts of State and Social Control in the Making of Democracy (Cambridge, 1990). 57 J. Caplan, Government Without Administration. State and Civil Service in Weimar and Nazi Germany (Oxford, 1988). Introduction 23 puri®ed race-nation. In this struggle the state was no more than a means to an end. The existing structure of the German civil service was cast aside along with the conventions of ordinary government and the restraints of the law. When in 1941 the political theorist Franz Neumann spoke of an `Un-state' emerging in the Third Reich, he was not merely pointing to the administrative incoherence and disorder of Nazi politics. His point was that the order that was emerging in the Third Reich could no longer be described in terms of the categories of conventional political theory.58 German civil society had been destroyed but so had the fragile structure of the German state. The boundary between state and civil society that was so fundamental to modern conceptions of politics had dissolved. As this book will show, of®cial statistics in Germany were deeply affected by this crisis of the state. The expansion of of®cial economic statistics in the ®rst half of the twentieth century certainly involved the exploration of hitherto uncharted realms of private economic activity. As one might expect, this extension of of®cial enquiries encountered substantial resistance, particularly from within the German business community. But it also raised fundamental questions within the statistical establishment. The practice of of®cial statistics, as it developed over the course of the nineteenth century, was built around the liberal distinction between state and civil society.59 Unlike the extractive agencies of the state, such as the tax of®ce and the bureaucracy of military conscription, the statisticians did not resort to coercion. They did not employ threats to obtain responses to their questionnaires and they did not check the accuracy of the returns in an intrusive fashion. The `objectivity' of of®cial numbers was founded on a supposed bond of trust between the of®cial statisticians and the citizenry. This trust in turn was based on self-limitation on the part of the statisticians. They refrained from making enquiries on issues likely to provoke resistance. And they guaranteed respondents the protection of anonymity and con®dentiality. In particular, they promised never to reveal statistical returns to the tax of®ce. These guarantees, emblazoned on the of®cial questionnaires, implied an acknowledgement of the right to `privacy'. At 58 F. Neumann, Behemoth. The Structure and Practice of National Socialism 1933±1944 (New York, 1963, reprint) and K. Tribe, Strategies of Economic Order. German Economic Discourse, 1750±1950 (Cambridge, 1995), pp. 169±202. 59 See paradigmatically the vision of of®cial statistics outlined by Ernst Engel, chief statistician of ®rst Saxony and then Prussia in the mid-nineteenth century, E. Engel, `Die VolkszaÈhlung: ihre Stellung zur Wissenschaft und ihre Aufgabe in der Geschichte', Zeitschrift des koÈniglich preuûischen statistischen Bureaus (ZKPSB) 2 (1862), pp. 25±31 and E. Engel, `UÈ ber die Organisation der amtlichen Statistik mit besonderer Beziehung auf Preussen', ZKPSB 1 (1860), pp. 53±56. 24 Statistics and the German State, 1900±1945 the same time, the technology of large-scale censuses symbolically positioned `the state', represented by the bureau of of®cial statistics, outside and `above' civil society. The mass of particular, self-interested individuals were submerged in large aggregates. The accumulation of millions of individual returns would produce a true image of society as a whole. Of®cial statistics thus sustained the idealistic image of the civil service as a disinterested `general class' presiding wisely over the social consequences of industrialization and urbanization. This liberal order of power and knowledge did not survive the expansion of state activity. And it is the collapse of this nineteenthcentury understanding of of®cial statistics that is the starting point for this book. In the ®rst half of the twentieth century Germany's of®cial statisticians were forced to renegotiate their relationship with civil society. Private businesses had to be coaxed or coerced into answering more questions and returning questionnaires more frequently. This expanded the range of economic data, and enhanced the role of of®cial statistics in government. But it also implied a new relationship between the statisticians and civil society and thus a new identity for of®cial statistics. In the 1920s the Reich's Statistical Of®ce was reinvented as a clearing house for information, a corporatist centre of data-sharing between the state and the powers of civil society. In the 1930s, in the early years of the Nazi regime, this mutated in an authoritarian direction. The statisticians began to imagine themselves at the centre of statecontrolled economy. But, this statist fantasy was not to last. It was swept away in the late 1930s by the radicalization of the Nazi regime. Combining the Nazi ideal of the Volksgemeinschaft with visionary technology, a radical faction within the Statistical Of®ce sought to reconstruct German of®cial statistics as a seamless system of surveillance. They would record not the economic and social aggregates but every single ®rm, every single worker and every single machine tool, indeed the movement of every single product throughout the economy. This totalitarian vision clearly had profound implications for the `freedom' of the individual. But it also rede®ned the practice of of®cial statistics. An integrated system of surveillance would make no distinctions between statistical questionnaires, tax records or private accounts. They all fed into the common database. Statisticians would thus forfeit their separate identity as guardians of a unique form of of®cial knowledge. But as managers of the uni®ed information system of the Reich, their reach would be extended beyond the wildest dreams of conventional of®cial statistics. In pursuit of omniscience, statisticians in Nazi Germany extinguished the practice of of®cial statistics itself. Introduction 25 V Such fantasies were fuelled by new information technologies. The development of large-scale bureaucracy in the late nineteenth and early twentieth century propagated an astonishing range of new techniques for handling and processing data. Many of these were mundane, but as historians of business administration have made clear, they were crucial to managing the increased ¯ow of information and to the rationalization of administrative activity.60 The new technologies of the of®ce included the telegraph, the telephone and the typewriter. But there were also great breakthroughs in the handling of paper records. Carbon paper eliminated the copy book. Vertical ®ling allowed random access to material. The conventions of business correspondence were adapted to the new ®ling systems. The invention of the card index revolutionized the control of very large quantities of information. All of these technologies found their applications in the business of of®cial statistics, but the real buzz word of the early twentieth century was `mechanization'. The key date here is 1890 when the ®rst machines produced by the German± American inventor Hermann Hollerith were put to work on the US census. His target was to ®nish the count in three years rather than seven.61 From our present-day perspective it is hard to avoid seeing the introduction of this primitive form of digital data-processing as a breakthrough of fundamental importance. But we need to be cautious here. Hollerith's machines did not revolutionize the production of statistics. A prior division between different classes of human labour pre®gured mechanization. In the compilation of the ®rst of®cial statistics in the early nineteenth century, the basic operations ± observing, recording, classifying the individual observations, counting the totals ± were merged into a single process. Local notables simply returned an account of their surroundings to the new statistical bureaux.62 The only element of central control were the questionnaires, issued as templates for the local reports. The development of statistical technology over the course of the nineteenth century involved separating each one of these operations ± observing, recording, classifying and counting ± allowing the entire process to be 60 J. Yates, Control through Communication. The Rise of System in American Management (Baltimore, 1989). 61 H. Petzold, Rechnende Maschinen. Eine historische Untersuchung ihrer Herstellung und Anwendung vom Kaiserreich bis zur Bundesrepublik (DuÈsseldorf, 1985), pp. 195±290. 62 B. Curtis, `Administrative Infrastructure and Social Enquiry: Finding the Facts about Agriculture in Quebec, 1853±4', Journal of Social History, 32 (1998), pp. 308±327 and S. Woolf, `Statistics and the Modern State', Comparative Studies in Society and History, 31 (1989), pp. 588±604. 26 Statistics and the German State, 1900±1945 controlled from the centre while keeping costs to a minimum. The ®rst step was to dispense with the mediating role of local notables and to have the questionnaires ®lled out directly by households and businessmen. It was still important to have reliable local census-takers, but their task was now limited to checking the accuracy of the individual returns and compiling preliminary tables of local results. The next move was to centralize the entire process of tabulation. After checking by the census-takers, the original returns were despatched to the central statistical of®ce. This, for the ®rst time, gave the statisticians control over the process of classi®cation, but it also posed enormous practical problems. Coordinating the movement of millions of bulky statistical returns was dependent on a modern infrastructure of transport and communications. And once the returns were gathered in, how were the statisticians to cope? Initially, the returns were classi®ed and counted in a single process. A mark was made in a tabular form for every return of a particular type, the returns being discarded in the process of counting. For simple classi®cations, this was a quick procedure. But classifying and counting in a single operation made it impossible to check for errors or to make any retrospective changes to the system of classi®cation. Any second thoughts necessitated a complete recount. Furthermore, if errors were to be avoided, trained staff had to perform both the intellectual labour of classi®cation and the routine business of counting. It was therefore expensive. The ®nal stage in the development of nineteenthcentury enumeration was the separation of classi®cation, sorting and counting. This was achieved by numbering each return and matching it with a numbered counting card. In effect, the statisticians created a paper double of the census, an image of the image. As each return was classi®ed, a code was inscribed on a counting card, which was then added to the appropriate pile. If the ®nal results appeared dubious one could move back from the classi®ed counting cards to the numbered originals. Reclassi®cations could be performed simply by reviewing the relevant returns identi®ed by the counting cards. Finally, the counting operation itself was speeded up. The cards were far easier to handle than the bulky questionnaires and pre-sorting increased reliability. But the major saving came from the division of labour. The staff of expert statisticians were concentrated on the process of classi®cation while women and unskilled homeworkers were hired to count the batches of cards. The European censuses of the late nineteenth century thus came to resemble the putting-out operations of early industry. A core of trained assistants performed the business of classi®cation under the supervision of professional statisticians who were also responsible for preparing the ®nal volumes of tables and text. Meanwhile, a temporary Introduction 27 army of human `computers' accomplished the sorting and counting. The intelligent work of classi®cation was thus separated from the rest of the process. The path was cleared for mechanization. Hollerith's machines slotted neatly into this division of statistical labour. The codes marked on the counting cards by expert of®cials were transferred to a third set of cards, which recorded the information as a pattern of punched holes. These cards were then fed through the machines, which sorted and counted them mechanically or electromechanically. The Hollerith Company made bold claims for its machines. In practice, however, it took years to perfect a really useful technology. It proved very dif®cult to ®nd paper of suf®ciently high quality to withstand the beating in the machines. The early models lacked printing mechanisms. The entire operation had to be halted to allow results to be read off and recorded by hand. In the 1920s the fastest sorting machines were rated at 20,000 cards an hour. Experience in Germany suggested that they could in fact handle no more than 12,000 cards per hour. Tabulating machines with a nominal capacity of 7,500 cards were actually capable of producing the results from no more than 2,500 cards per hour.63 For enumerations of less than 5,000 cards the machines were actually less cost-effective than manual processing. The expensive equipment needed to be fed with very large batches of cards. The machines were useful for carrying out the rough classi®cation of large collections of data, but became increasingly inef®cient as the classi®cation was re®ned.64 Nevertheless, their use became increasingly common in both the public and business world before World War I and their spread was encouraged by the intense competition between Hollerith and its main rival, Powers. Classically, the machines were employed in the compilation of urgent trade returns or in the processing of large batches of ®nancial data. In these roles, they did deliver substantial savings of labour. The Reich's Statistical Of®ce estimated in the late 1920s that its extensive use of machines made possible a 25 per cent saving in its labour budget, which was the largest part of its costs.65 The conclusion is clear: mechanization of data-processing facilitated the statistical revolution. It enabled the accelerated production of large volumes of data at reduced cost. But it was not necessary in the strict sense of the word. However, such a rationalist appraisal may lead one to underestimate 63 GStA 1. HA Rep. 77 3893 no. 151 ff, Niederschrift der Verhandlungen der Statistiker des Reichs und der Bundesstaaten zu Bremen am 29. bis 31. Mai 1913, pp. 21±29. 64 P. Quante, `Die Erfahrung mit elektrischen ZaÈhlmaschinen in Preuûen bei der Volksund BerufszaÈhlung vom 16. Juni 1925', Allgemeines Statistisches Archiv (ASA), 20 (1930), pp. 82±112. 65 See chapter 3. 28 Statistics and the German State, 1900±1945 the importance of technical change. The vision of mechanical dataprocessing held a powerful fascination in its own right. One perceptive commentator spoke of the `suggestive power of the concept of ``centralization''' peddled by suppliers of business machines.66 Across the world, bureaucrats were inspired to dreams of omniscience. Arguably it was through these fantasies of total control that the technology of mechanical data-processing had its most profound impact on the history of statistics. For the ®rst time it became possible to conceive of an entire nation recorded in a single database instantly accessible by means of mechanical handling equipment. This fantasy seems to have occurred independently in at least three different European countries in the interwar decades. In France it began to be realized in the early 1940s under the Vichy regime.67 In Britain it dated back to an abortive scheme for national identity cards initiated during World War I.68 In Germany, too, it had its origin in police plans for personal identi®cation but took a sinister twist after 1933 in the form of an SS databank.69 As this book will demonstrate, the `romance of information technology' was also to have a dramatic impact on the history of economic statistics. New technologies of data-processing inspired contemporaries to believe that the economy might be most effectively controlled, not by manipulating national aggregates, but by a comprehensive system of individualized surveillance. VI As this discussion of information technology makes clear, it is not enough to understand the emergence of modern economic statistics simply as an effect of bureaucratization in the broadest sense of the word. It also needs to be understood as an intellectual and cultural phenomenon. To understand particular statistical projects one cannot avoid the detailed reconstruction of their particular intellectual context. But are there ways of generalizing about such cultural processes? As the sub-title claims, this book is about the emergence of a speci®cally 66 A. Busch, `Zur Frage der Verwendung von Lochkartenmaschinen', ASA, 20 (1930), pp. 260±265, in response to a puff from a representative of the Deutsche Hollerith Maschinen GmbH, K. Koch, `Die Verwendung von Speziallochkartenmaschinen bei der VolkszaÈhlung 1930', ASA, 19 (1930), pp. 560±568. 67 G. Chevry, `Un Nouvel Instrument de travail statistique: Le Fichier des eÂtablissements industriels et commerciaux?', Journal de la SocieÂte de Statistique de Paris ( JSSP), 89 (1948), pp. 245±262 and `NeÂcrologie: Rene Carmille', JSSP, 86 (1945), pp. 145±148. 68 J. Agar, `Modern Horrors: British Identity and Identity Cards', in J. Caplan and J. Torpey (eds.), Documenting Individual Identity (Princeton, 2001). 69 G. Aly and K.H. Roth, Die restlose Erfassung. VolkszaÈhlen, Identi®zieren, Aussondern im Nationalsozialismus (Berlin, 1984). Introduction 29 modern form of economic knowledge. Not that `modern' is a term to use without caution. Its meanings are contested and bewilderingly various. However, handled with care it can serve as a useful framework within which to generalize about intellectual and cultural trends. It helps us to place our particular stories of intellectual genesis in a broader context. A ®rst helpful distinction is between `modernity' and `modernism'. The term `modernity' is commonly used to refer to the processes of social, economic and technical change that have already been discussed: the development of modern industry and services, the ever-expanding role of bureaucracy in both corporate capitalism and the public sector, the development of new technologies of data handling, data-processing and communication. We can certainly describe the new economic statistics that began to emerge in the late nineteenth century as a form of economic knowledge peculiar to modernity. They were part of the deliberate effort by powerful social actors to exercise conscious control over the development of the national economy. In this sense they were an integral part of the institutionalized, social self-re¯exivity that numerous authors have identi®ed as characteristic of modernity.70 But what about `modernism' or `the modern'? The ®rst problem is to disentangle the multiple meanings of these words.71 On the one hand there is the usage favoured in debates surrounding postmodernism.72 In this context, `the modern' refers to the dominant continuity of Western rationalist, scienti®stic thought stretching back to the Enlightenment. This is the way in which Keith Tribe has used the term in his fascinating volume, Strategies of Economic Order. German Economic Discourse, 1750±1950. Tribe constructs a grand arch of continuity connecting the rationalist economic thought of the Enlightenment to the visionary projects of twentieth-century technocracy. According to Tribe they share the same `dream of reason', the compulsion to subject the world to the order of reason.73 The history of statistics certainly witnessed the compulsive recurrence of such dreams. Utopian rationalists of the eighteenth and early nineteenth century sketched numerous schemes for comprehensive statistical enumeration. In 1805, the ®rst head of the Prussian statistical of®ce, Leopold Krug, drew up an ambitious scheme for a complete enumeration of the incomes of the 70 A representative example is A. Giddens, The Nation-State and Violence, 2, A Contemporary Critique of Historical Materialism (Cambridge, 1985), pp. 180±181 and 308±310. 71 The following discussion owes much to Dorothy Ross (ed.), Modernist Impulses in the Human Sciences 1870±1930 (Baltimore, 1994). 72 A. Eysteinsson, The Concept of Modernism (Ithaca, 1990), pp. 3±4, 103±142. 73 Tribe, Strategies, p. 1. 30 Statistics and the German State, 1900±1945 Prussian population.74 Needless to say, this plan never left the drawing board. Krug shortly lost his job to a more liberally inclined University Professor of Smithian persuasion. But, the dream returned. In 1870, Ernst Engel the director of the Prussian bureau launched a plan for a comprehensive system of national economic statistics.75 He, too, was to be disappointed. As we shall see, the Reich's Statistical Of®ce, established in 1872, was hedged around with restrictions. It was not until the 1920s that the dream was ®nally to be realized. The Statistical Of®ce of the Weimar Republic set about creating a comprehensive system of national accounts. By the 1940s Germany's statisticians were embarked on a minute enquiry into the structures of German industry, that would have satis®ed even the most ambitious advocates of absolutist `police' (Polizei). And these continuities were not lost on contemporaries. As will be discussed in chapter 3, the upsurge of interest in macroeconomic questions around the turn of the twentieth century was often described as the revival of an Enlightenment tradition. The metaphor of the circular ¯ow, which was so central to the understanding of the economy in the 1920s, was commonly attributed to the eighteenth century French physiocrat Quesnay. It was Quesnay's Tableau eÂconomique, which had inspired Krug in 1805. Maintaining the memory of Quesnay's work into the twentieth century was one of the great achievements credited to Karl Marx's Das Kapital. The other theoretical inspiration for twentieth century macroeconomics, the Quantity Theory of Money, was also understood as a theory of ancient pedigree. Its ®rst modern expression was commonly credited to David Hume, the philosopher of the Scottish Enlightenment, and its earliest origins could be traced to classical antiquity. Together these two metaphors provided the building blocks for an aggregative and macroeconomic conception of the economy. Whether or not there was a real continuity between the Enlightenment and the technocratic experiments of the early twentieth century, there was certainly an `invented tradition' of rationalist economics. But this is not the only meaning of the words `modern' and `modernist'. The term may also refer more speci®cally to the highly sceptical and corrosive intellectual currents of the turn of the century. H. Stuart Hughes in his classic, Consciousness and Society. The Reorientation of European Social Thought 1890±1930 showed how artistic modernism radiated powerfully onto the social sciences through such ®gures as Max 74 L. Krug, Ideen zur einer Staatswirthschaftlichen Statistik (Berlin, 1807). 75 E. Engel, `Die Notwendigkeit einer Reform der volkswirtschaftlichen Statistik', ZKPSB 10 (1870), pp. 141±408. Introduction 31 Weber and Sigmund Freud.76 More recently, Dorothy Ross has provided us with a helpful encapsulation. In the era of high Victorian optimism, she writes, `thinkers in Europe and the United States believed that scienti®c and historical knowledge would provide synthetic and normative foundations for modern life of the kind that philosophy and religion had traditionally provided. By the end of the century that hope could no longer be sustained.'77 The early twentieth century was thus marked by a new understanding of the subjectivity of knowledge, scepticism about rationality as a general guide to life, a turn to aesthetics and a radical stress on historicity and contingency. For the United States Ross has described how the crisis of the historical narrative of American exceptionalism led, in the last decades of the nineteenth century, to a search for new models for the social sciences. She uses the term `scientism' to describe the effort by American economists, sociologists and political scientists to refashion their disciplines in the image of the natural sciences. `Social science was to be an autonomous body of knowledge . . . directed at and constituted in accordance with the technological capacity for control.'78 It would thus legitimate itself in the face of modernist doubt. Germany is another country in which the development of twentiethcentury economics bears the scars of an encounter with modernism. The statisticians and economists that are discussed in this volume no longer enjoyed the self-con®dent authority commanded by the so-called `Historical School' of late nineteenth-century Germany.79 Figures such as Professor Gustav von Schmoller claimed an Olympian detachment from the fray of private interests. Schmoller's chosen audience was the civil service, the Hegelian guardians of the fate of the nation. Moral judgements were an indispensable component of his economic analysis. This self-con®dent model of academic politics (Gelehrtenpolitik) did not survive the turn of the century. Its intellectual foundations were subject to devastating methodological criticism by the younger generation, led by Max Weber. The impossibility of establishing a neutral vantage point from which to view society was brutally exposed by the bitter political divisions within the academy. Some professors were 76 H. Stuart Hughes, Consciousness and Society. The Reorientation of European Social Thought 1890±1930 (London, 1974). 77 Ross, `Introduction: Modernism Reconsidered', in Modernist Impulses, p. 1. 78 D. Ross, The Origins of American Social Science (Cambridge, 1991), p. 400. 79 Various stages in this process are described in D. Lindenlaub, RichtungskaÈmpfe im Verein fuÈr Sozialpolitik (Wiesbaden, 1967), D. KruÈger, NationaloÈkonomen im wilhelminischen Deutschland (GoÈ ttingen, 1983), F. Lenger, Werner Sombart 1863±1941 (Munich, 1994), R. vom Bruch, Wissenschaft, Politik und oÈffentliche Meinung. Gelehrtenpolitik im wilhelminischen Deutschland (1890±1914) (Husum, 1980). 32 Statistics and the German State, 1900±1945 drawn into party-political engagements. Others preached the need for support of the labour movement. And the reformist consensus was rejected altogether by an increasing number who embraced capitalism and sought to foster close links with industry and commerce. If the Historical School ever existed as a coherent body of knowledge, it was certainly dead by 1914. It was buried by war, revolution and defeat. This devastating experience destroyed the idealized image of the German state to which men like Schmoller had addressed their appeals for social reform. By contrast with their Wilhelmine forebears, the new experts of interwar Germany spoke a disillusioned language. The role models for Ernst Wagemann, the head of the Weimar Statistical Of®ce, were the immensely successful chemists and physicists. What economics had to offer was not moral leadership but speci®c expertise for the making of economic policy. The German statisticians discussed in this book were committed empiricists. But theirs was no longer the triumphant positivism of the mid-nineteenth century. They were acutely aware that they did not have access to reality in any fundamental sense of the world. Their ®rst ambition was to register symptoms, proxies for processes that were too complex to be grasped in the totality of their detail. They were cautious about making causal claims, conceiving their task more modestly as the sifting of evidence with a view to projecting stable patterns from the past into the future. In the early 1920s the leaders of the Reich's Statistical Of®ce hoped that this disillusioned empiricism might form the basis for a new relationship with academic economics. An attempt was made to turn the Statistical Of®ce into a coordinating centre around which economics in Germany could be reconstructed as an empirical, policy-orientated discipline. However, the strategy of scientism that was so hegemonic in the United States commanded no consensus in interwar Germany. It was unacceptable to those who sought to maintain a role for the economist as moral authority and spiritual guide. It was also unacceptable to the minority of mainly younger economists, who devoted themselves to the elaboration of abstract economic theory. The result was a widening gulf between the Statistical Of®ce and the increasingly divided ®eld of academic economics. Culture wars of this sort clearly have highly speci®c, national dynamics. However, the outcome appears to have been quite generic. In both the United States and Germany, economics was subject to a process of institutional and intellectual differentiation. Empirical, practical economic knowledge prospered by placing itself at the service of power. Theoretical economics in the Universities, on the other hand, developed as an ever-more arcane and mathematized discipline, driven Introduction 33 by its own intellectual logic. And both forms of economic knowledge insulated themselves institutionally and intellectually against the other social and human sciences, which were far more powerfully affected by modernism and its discontents. Statistics thus provided one of the foundations for a peculiarly self-contained, indeed isolated, form of modern economic knowledge. VII The development of modern economic statistics in the ®rst half of the twentieth century can thus be understood as the product of a number of quite general in¯uences. But what about the speci®cally German aspects of this story? Much more will be said about the German historiography in the following chapters. However, there are some basic interpretative issues that should be addressed from the start: ®rst of all the question of the Sonderweg. Since the late nineteenth century it has been common to accord the German state an exceptionally important role in guiding national economic development.80 It would not be surprising, therefore, to ®nd in Germany a strong tradition of of®cial economic statistics. The peculiarly dramatic expansion in the interwar period could thus be seen as an outgrowth of a deep German tradition.81 This is not the line followed in this book. The history of economic statistics, in fact, reveals the limits of the supposedly `strong' German state. The German nationstate established by Bismarck in 1871 had no national structures of economic intervention. And, over the following decades, despite the extraordinary development of the German economy, it did not acquire any. Of course, in the nineteenth century, states such as Prussia and Saxony played an active role in promoting local industrial and commercial development. Municipalities also performed a key function in providing economic infrastructure. However, it is fanciful to suppose that they pursued anything remotely akin to modern macroeconomic policy, in the sense described by Winch.82 It was World War I that ®nally forced the creation of national Ministries for the economy, for labour and for agriculture. The emergence of modern economic policy dates to the 1920s and it was this which set the stage for the sudden explosion of statistical innovation in interwar Germany. The new statistics of the 80 For a critical discussion see G. Herrigel, Industrial Constructions. The Sources of German Industrial Power (Cambridge, 1996). 81 For a history of Nazi economic policy constructed in these terms see A. Barkai, Das Wirtschaftssystem des Nationalsozialismus. Ideologie, Theorie, Politik 1933±1945 (Frankfurt, 1988, 2nd edn.). 82 V. Hentschel, Wirtschaft und Wirtschaftspolitik im wilhelminischen Deutschland. Organisierter Kapitalismus und Interventionsstaat (Stuttgart, 1978). 34 Statistics and the German State, 1900±1945 1920s and 1930s were the creation of the new national economic administration. They were established as a self-conscious break with tradition and they were opposed by those LaÈnder which still had pretensions to independence. As will be shown, the new statistics of the Weimar Republic were emphatically macroeconomic in orientation. The innovative activity of the Reich's Statistical Of®ce in the 1920s and 1930s was part of a wider effort to reinvent the German state. The threat of revolution in Germany may have been exaggerated, but the depth of the crisis affecting the German state is hard to exaggerate.83 The civil servants of the Weimar Republic struggled to create a new system of national economic administration in a peculiarly hostile environment. The statistical projects that are the subject matter of this book were a response to this crisis. Radical discontinuity rather than continuity is our theme. In stressing the genuinely innovative character of the Weimar Republic this book follows the ¯ow of recent historical writing.84 The revolution of 1918 was certainly incomplete. It left intact certain key elites such as the conservative judiciary. However, one should not make simplistic assumptions about continuities with the Wilhelmine era. As Michael Geyer has shown, even the German armed forces were far from conservative. The commanders of the Reichswehr may have been hostile to democracy but to think of them as reactionary hangovers is to ignore the modernity of militarism.85 In the 1920s and 1930s the German of®cer corps led a restless search for new means with which to conduct war. In non-traditional ®elds of government this tendency to innovate was even more pronounced. The state apparatus of the Weimar Republic was a laboratory of modern government, a conclusion strongly reinforced by a wealth of recent literature on areas of social work and welfare policy.86 One of the aims of this book is to incorporate the sphere of economic government into this new picture of the Republic. In practice, economic policy was of course conditioned by a variety of institutions and pressures. However, policy was not simply a re¯ex of interest group preserves. By sponsoring economic statistics and economic research the new national Ministry for Economic Affairs was 83 For a spectacular overview see G.D. Feldman, The Great Disorder. Politics, Economics, and Society in the German In¯ation, 1914±1924 (Oxford, 1993). 84 Led by the in¯uential work of Detlev Peukert above all. D. Peukert, Die Weimarer Republik. Krisenjahre der Klassischen Moderne (Frankfurt, 1987). For a critical appreciation of Peukert's contribution see D.F. Crew, `The Pathologies of Modernity: Detlev Peukert on Germany's Twentieth Century', Social History, 17 (1992), pp. 319±328. 85 M. Geyer, `Etudes in Political History: Reichswehr, NSDAP and the Seizure of Power', in P.D. Stachura (ed.), The Nazi Machtergreifung (London, 1983), pp. 101±123. 86 For a critical survey see Y.-S. Hong, Welfare, Modernity, and the Weimar State, 1919±1933 (Princeton, 1998). Introduction 35 seeking to lay the foundations for an innovative programme of government founded on empirical data and systematic analysis. By 1927 the Ministry had developed concrete plans for economic stabilization that gave a key role to the Statistical Of®ce and the state-sponsored Institute for Business-Cycle Research (IfK). What destroyed this optimistic scenario was the cataclysm of the Great Depression. The Weimar Republic thus makes an ideal case study for a history of modern economic knowledge. But what about the Third Reich? Surely, Hitler's regime was an atavistic dictatorship with little or no interest in questions of rational economic management. Its barbarity surely makes it a case so special as to be irrelevant to any wider discussion. One response to such objections is to make a distinction between means and ends. The ends of Nazi ideology may have been genocidal and reactionary. Hitler may have dreamed of returning to an agrarian, Aryan utopia. However, as was argued long ago by David Schoenbaum, to achieve its goals, Hitler's regime had to make use of the technology and managerial resources provided by industrial modernity.87 At the very least the Third Reich required a supply of up-todate military technology with which to ®ght its war. And Hitler certainly had no dif®culty in rallying professional and technical elites to the National Revolution.88 Indeed, as historical research has revealed the full extent of elite collaboration our perception of the Nazi regime has changed.89 This has been ampli®ed by our increasing awareness of the ambiguities of modernity and modernism. Is it, in fact, true that the regime's aims were atavistic or reactionary? Hitler's conception of the Volksgemeinschaft certainly contained many elements normally associated with programmes of social modernization and modernity.90 He was a bitter opponent of old-fashioned class divisions and provincialism. Even his anti-Semitism was of a modern variety. Hitler's hatred of the Jews was couched in terms of race not religion and the SS was generous in its sponsorship of `racial science'. Furthermore, Nazi ideology more generally can clearly be related to currents in modernist culture. The movement's anti-materialism, its 87 D. Schoenbaum, Hitler's Social Revolution: Class and Status in Nazi Germany 1933±1939 (New York, 1966). 88 For German economics in National Socialism see C. Kruse, Die Volkswirtschaftslehre im Nationalsozialismus (Freiburg, 1988) and W. Krause, Wirtschaftstheorie unter dem Hakenkreuz. Die buÈrgerliche politische OÈ konomie in Deutschland waÈhrend der faschistischen Herrschaft (Berlin, 1969). 89 For a survey see M. Roseman, `National Socialism and Modernization', in R. Bessel (ed.), Fascist Italy and Nazi Germany. Comparisons and Contrasts (Cambridge, 1996), pp. 197±229. 90 R. Zitelmann, Hitler. SelbstverstaÈndnis eines RevolutionaÈrs (Stuttgart, 1990, 2nd edn.). 36 Statistics and the German State, 1900±1945 obsessions with the idea of national rebirth, its emphasis on subjective will, all had a wider resonance.91 But, it is not the re-evaluation of Nazi ideology that most concerns us here. Far more important from the point of view of this book is the parallel re-evaluation of the relationship between means and ends, between technology, expertise and Nazism. Schoenbaum's formulation pivoted on a distinction between the atavism of Nazi ideology and the modern technology that provided its weapons. But this comforting distinction was never stable. In the 1940s it had already been called into question by Horkheimer and Adorno.92 In the Dialectic of the Enlightenment they presented fascism not as a counter-reaction to modernity but as an outgrowth of alienated, objectifying reason. In the 1960s Hannah Arendt's commentary on the Eichmann trial provided an explanation of how bureaucratic banality made mass murder possible.93 These early interpretations have had a profound effect on the recent literature. Far from being counterposed to Nazi racial ideology, modern bureaucracy and technology have come to be seen as conducive to genocide.94 Bureaucracy enabled mass murder in a technical sense. It also had moral and political effects. Far removed from the consequences of their decisions, the Nazi elite and their servants could coolly order the destruction of Europe's Jews. At a deeper level, it was the rationalist dream of remodelling society that was, itself, the root of evil. The hubristic fantasies of social engineers inspired many of the disasters of the twentieth century.95 It was modern medicine and genetic theory that gave rise to the idea of the national body and the associated policies of eugenics and racial cleansing. In a Nietzschean twist, Detlev Peukert located the spirit of science at the origin of the genocidal murder of Europe's Jews.96 Far from there being a con¯ict between Nazi ideology and modernity, there was a natural complicity between the two. This kind of interpretation has also been applied to our subject. In the 1980s GoÈ tz Aly and Karl-Heinz Roth published a short book on the history of German statistics.97 As an intervention in the popular protest against the German census of 1983, their aim was to cast doubt on the claim that statistics were a harmless technology of social administration. 91 R. Griffen, The Nature of Fascism (London, 1991). 92 T. Adorno and M. Horkheimer, Dialektic der AufklaÈrung (Frankfurt, 1980, 7th edn.). 93 H. Arendt, Eichmann in Jerusalem. A Report on the Banality of Evil (New York, 1963). 94 Most notably Z. Bauman, Modernity and the Holocaust (Cambridge, 1991). 95 J.C. Scott, Seeing Like a State. How Certain Schemes to Improve the Human Condition have Failed (New Haven, 1998). 96 D. Peukert, `The Genesis of the ``Final Solution'' from the Spirit of Science', in T. Childers and J. Caplan (eds.), Reevaluating the Third Reich (New York, 1993), pp. 234±252. 97 Aly and Roth, Die restlose Erfassung. Introduction 37 To make their point, Aly and Roth exposed the willing collaboration of German of®cial statisticians in the system of Nazi tyranny. As active proponents of eugenics, Directors of the Statistical Of®ce such as Dr Friedrich BurgdoÈ rfer helped to provide the intellectual foundations for policies of sterilization, euthanasia and ultimately genocide. Other statisticians collaborated with the SS in creating systems for registering the entire population in card®le databases. Aly and Roth even went so far as to suggest that it was administrative technologies of this kind that allowed the regime to maintain its grip so effectively for so long. Statisticians are thus presented as key agents of the regime. And this collaboration was no accident. It followed from the objectifying, dehumanizing logic of the statistical technology itself. Once people were reduced to numbers it was a short step to cancelling them out.98 In answer to the question raised above, Aly and Roth would contend that it is precisely by studying Nazism that we can comprehend the real potential of statistics as a tool of oppressive, dehumanizing reason. Though it is not a study of racial policy or genocide this book is indebted to these recent reinterpretations of the Third Reich. The Third Reich is discussed here in a contextualized way, situated both in relation to the Weimar Republic and contemporary developments outside Germany. In this fundamental respect Nazism is treated as an outgrowth of modernity rather than a sudden departure from historical continuity. This is an important advance in our understanding, made possible by the latest generation of research. As has been stressed in this introduction, the structures of national government inherited from the Weimar Republic were a work in progress. For many bureaucrats the early years of the Nazi regime, at least, were an exhilarating period of liberation from the fetters of parliamentarianism. The Third Reich appeared as an opportunity to realize the experiments in national government initiated by the Weimar Republic. This book, however, is also the result of learning in a negative sense. It rejects the simplistic equation drawn by authors such as Aly and Roth between Nazism and technocracy. Studying statistics as an infrastructure of modern power has its attractions. It opens a new window on the business of government. It also allows one to uncover new parallels and continuities across time and space between super®cially dissimilar regimes. But this approach is also associated with certain intellectual risks, risks that are starkly revealed by the work of Aly and Roth. First, there is the problem of critical distance. Government tends to 98 This line of argument has subsequently been extended into a general analysis of the genesis of the Holocaust in G. Aly and S. Heim, Vordenker der Vernichtung. Auschwitz und die deutschen PlaÈne fuÈr eine neue europaÈische Ordnung (Hamburg, 1991). 38 Statistics and the German State, 1900±1945 present a monolithic visage to the outside world. And it is tempting to take this at face value, particularly if one is seeking to establish the importance of administrative and technical history for our more general historical understanding. Aly and Roth, for instance, are remarkably uncritical of the technical achievements of Nazi technocracy. And they have little incentive to be otherwise. Their purpose, after all, is to warn us of the dangers posed by statistics in an age of electronic dataprocessing. Why not, therefore, let the Nazi technocrats indict themselves? Their bold claims to have delivered a functioning system of totalitarian surveillance provide precisely the evidence that Aly and Roth are looking for. The result, however, is a shallow and instrumentalized interpretation of the Third Reich. This book, by contrast, attempts to provide a realistic assessment of the contribution made by economic statistics to the functioning of the Nazi dictatorship. The facade of the Reich's Statistical Of®ce may have been monolithic. However, it hid a multiplicity of competing visions of the statistical future. The result, in practice, was a shambles. Successive waves of bureaucratic initiative produced an information system fundamentally incapable of satisfying the needs of wartime economic government. The grandiose schemes for individualized personal surveillance, on which Aly and Roth concentrate their attention, were the least successful of all. By focusing on the more ominous technocratic initiatives, they misunderstand the role played by Germany's statisticians in the functioning of the Nazi regime. By the later stages of the war the ideologues of totalitarian surveillance had been sidelined. It was the statistical apparatus created by the Weimar Republic that really provided the underpinnings for the Nazi war effort. Aly and Roth fail to understand this because they take the propaganda of technocratic totalitarianism at face value. Aly and Roth's work however suffers from more than just a lack of critical distance. Its most fundamental problem is its technological determinism. The alliance between the Reich's Statistical Of®ce and the Third Reich is not seen as the contingent product of the crisis of the German state. Rather, it is presented as the necessary result of the dehumanizing logic of statistical technology. This in turn underpins Aly and Roth's reductive account of the Federal Republic as a direct descendant of the Third Reich. If any form of quantifying social discourse is inherently dehumanizing and oppressive, then the political and legal framework in which it is situated makes little difference.99 Statisticians in the Weimar Republic, the Third Reich and the Federal 99 For a powerful critique of overly deterministic approaches to social policy in the Third Reich see D. Peukert `Zur Erforschung der Sozialpolitik im Dritten Reich', in H.-U. Otto and H. SuÈnker (eds.), Soziale Arbeit und Faschismus (Frankfurt, 1989), p. 39. Introduction 39 Republic were all pursuing the same basic objective. The Statistisches Bundesamt in Wiesbaden, the statistical of®ce of the Federal Republic, may appear to be an innocuous centre of social administration. But according to Aly and Roth it in fact served as a sinister haven for the racial scientists of the Third Reich and their intellectual descendants. In their portrayal of the Federal Republic, Aly and Roth push the logic of their argument to absurd extremes. However, this tendency towards determinism is inherent in any study which addresses itself to technologies of government such as statistics. After all, this introduction began in a similar vein, blurring conventional distinctions and insisting that we should view the development of modern economic knowledge as an interconnected process, stretching across more than half a century and a surprising variety of countries. Unlike Aly and Roth, however, this book insists on the need to differentiate. The statistical systems developed in Nazi Germany, the Soviet Union, Scandinavia, the Netherlands, Great Britain and the United States were indeed related. They shared certain common intellectual origins, certain technical preconditions and they were marked by their simultaneous appearance at a particular moment in time. But they were not identical. They were differentiated in technical terms. But, more fundamentally, they were distinguished by their relationship to politics. The most serious side-effect of technological determinism is that it makes it impossible to take politics seriously. One ends up, like Aly and Roth, equating the Third Reich and the Federal Republic simply because they shared certain administrative practices. This is a reductionism reminiscent of the crudest forms of Marxism. By contrast, this book shows how the development of new forms of economic knowledge was combined in the interwar decades with an intense argument over the appropriate political and legal framework for economic government. In Germany, this argument was taken to extremes. It is this which makes the study of the Third Reich so important. It reveals the potential inherent in common technologies of economic knowledge when combined with a peculiarly racist brand of collectivism.