Chartbook Newsletter #22

How do you count inflation? Tracking Weimar's hyperinflation.

“We are inside.” - Mark Hulliung, Sartre and Clio. Encounters with History (2013), p. 10.

The threat of inflation makes daily headlines right now. In most advanced economies, only people over the age of 40 have any personal experience of serious inflation. But scary historical episodes are easily to hand. The 1970s ghost through policy-talk.

Whether or not such historical clichés are at all relevant to our current moment, is one line of debate. I am on the skeptical side of the fence. Fifty years on from the last great inflation in the West, the political economy, has radically changed. The bargaining power of labour is far weaker. Today, the risk of a wage-price spiral seems slim. The global headwinds that dampen inflationary risks continue to be strong. Indeed, so stark are these differences that it seems surprising that simple historical analogies are taken seriously at all. What is truly intriguing, however, is a deeper and more basic question. If there are underlying changes in the regime of price-setting, is our current system of inflation measurement set up to give us a good handle on 21st-century price dynamics? How do you deal with a shock like 2020? In its aftermath, how do you recognize inflation when it does arrive?

The simple answer is that inflation is happening when prices go up. Tbe tricky thing is that not all prices go up at the same rate. Some go up, others go down. And the prices that do go up, go up for very particular reasons. If microchip prices surge because of drought in Taiwan, can we speak of inflation? What, if a spike in demand for lumber in North America drives plywood prices sky high? Is that inflation? Or will it simply compress the price of other construction materials, as budgets have to be cut?

In discussing inflation, economics abstracts from idiosyncratic shocks. Inflation is defined as a general upward pressure on all prices, independent of idiosyncratic supply shocks. Inflation, in this sense, is a macroeconomic, aggregate concept.

The truly common denominator of economic activity in market societies is money. Goods exchange for money. So, as a pressure acting on the prices of all goods, it is with regard to money that inflation is defined. Indeed, as Rebecca L. Sprang reminded us in a typically brilliant op ed, when the term “inflation” began to be used in an economic context in the 1860s and 1870s it referred first and foremost to the expansion in the currency, not the price changes that might result from that monetary expansion. It was the currency that was inflating, not prices.

Milton Friedman drove home this essential connection with his famous remark that “inflation is always and everywhere a monetary phenomenon”. Read as a causal statement this rapidly unleashes fierce argument. What is less contentious is the interpretation of Friedman as laying out a definitional tautology. Inflation is always a monetary phenomenon because money is the denominator against which we judge a general movement of the price of “all goods”, as opposed to a relative movement in prices between goods. Inflation and deflation are shifts in the terms of trade between goods in general and money.

How do you measure a shift in the terms of trade between money and goods in general? One classic shortcut is to track the fluctuation in value of one currency against another, which by way of purchasing power parity ought to be governed by how many goods each currency buys. This, however, is no more than a shortcut and an imprecise one at that. Exchange rates are moved by a variety of forces other than purchase power parity.

To measure the movement of the terms of trade between money and goods directly we need some kind of statistical indicator of the general price level. How you construct that indicator will define how you see inflation and your ability to track it. There are narrow indicators based on consumption baskets for consumers. Measures of inflation based on raw materials or producer prices. Measure of inflation that apply to GDP as a whole, so-called GDP deflators. Thanks to digital technology which reduces the cost of calculation to close to zero, the WSJ now even offers personalized inflation numbers.

Inflation can thus be defined as a shift in the terms of trade between (1) money and (2) goods, as experienced (3) by a particular group of people and (4) captured by a particular statistical apparatus.

Picking a statistical indicator is not just a matter of design - choice of basket etc. Statistical indicators have to be produced. Their production involves capital and labour. It involves technology. Mobilizing those resources and putting them to work involves the exercise of power. To extract the raw material of data involves the exercise of authority, by way of law or other types of persuasion or coercion.

Data are business. Data are political. And that is particularly pertinent in the case of inflation, because inflations are contentious. They generate winners and losers. That is why we care about inflation. Winners may not want to have their gains documented. Losers will want to documenting their losses so as to seek redress. Inflation numbers are not merely descriptive. They are part of the political economy of the process they describe. When it comes to inflation there are, as Harold Garfinkel the great ethnomethodologist might have put it, good reasons for bad data.

Thinking of economic statistics as forms of knowledge it is important not to mistake them for something they are not. They are not a radar gun pointed at a speeding car from the side of the road by uniformed police officer. They are more like an improvised speedometer attached to the inside of a ramshackle Max-Max vehicle. There is no fixed outside that any of us can access. You cant “get off” economic history or “outside it”. And inside the vehicle there is no undisputed authority. Who gets to attach the speedometer, is part of the battle to decide who is in charge. Not only are statistics on “the inside”, we were already in motion when we started rigging them up. We attached the speedometer to the wheels as they were already turning. We are always in medias res.

Inflation statistics are relatively recent. Through historical research we can retrospectively construct data that go back in a continuous series to the medieval period, to the origins of capitalism and the emergence of regularized trade in labour, land and other commodities. But, as a real time indicator, an accompaniment to economic and political events, delivered by authoritative agencies, inflation numbers are only a little more than a hundred fifty years old. They are part of the process through which the modern state and the modern economy were coproduced in the transition from the late 19th to the twentieth centuries.

The first hyperinflations of the early 20th century, in central Europe after World War I, were crash tests of the politics of economic measurement. The inflations resulted from radical social and political conflict that was carried into the finances of the state and out from there into the monetary system. The numbers that described the resulting inflations were themselves products of that struggle. Nowhere is this more true than in the Weimar Republic, the classic hyperinflation.

Weimar’s inflation statistics, as normally presented, show a surge that happened in three phases. Wartime inflation continued through to the summer of 1920, at which point the UK and US initiated a deflation that affected Weimar Germany as well. Then a acceleration began in the second half of 1921, culminating in the slide into hyperinflation towards the end of 1922 and into 1923. The familiar tables do a decent job of describing this trajectory. What they hide, however, is their own history, and thus the role of statistics in the politics of the inflation. The neat retrospective tabulations obscure the fact that the German state only began publishing an up-to-date index of the national cost of living in April 1921. By that time, according to the index, the Mad-Max jalopy was already traveling at speed. Consumer prices had risen 13 fold since 1914! Across the first wave of inflation and into the deflation of 1920, German society operated without real time national price data. Employers, trade unions and government Ministries were flying blind.

Contrary to preconceptions about the efficiency of German bureaucracy, this statistical failure did not come as a surprise. In compiling economic statistics, Imperial Germany was far from being an international leader. But even if we look beyond Germany’s borders, the technology of index numbers itself was of remarkably recent vintage. In 1860, the weekly journal the Economist became the first to publish a commodity price index.

It was followed in the United States by The Commercial and Financial Chronicle and in Britain by the Statist. From a technical point of view, these early journalistic experiments were extremely primitive. A collection of prices was combined by means of a simple average and expressed relative to a common base year. The idea of `weighting' the components of an index number was developed by German statisticians Laspeyres and Paasche in the 1860s and 1870s. Weighted indices did not become commonplace until the end of the century, when the modern era of index numbers began.

The surge in interest around the turn of the century is attributable to two factors. First of all, the universal adoption of the gold standard coincided with a revival in monetary theory and what would become known as the Quantity Theory of Money. This suggested the possibility of measuring the value of money by means of an index of the general price level. The most sophisticated exponent of index number techniques was the American economist and progressive Irving Fisher, who was also one of the foremost champions of the revived quantity theory of money. A novel index underpinned his effort to take monetary economics beyond tautologies into the realm of empirical science.

The second factor, which stimulated interest in index numbers, was the gradual increase in the cost of living that set in around the turn of the century. This prompted demands from trade unions that wages should be adjusted to prices. Again, index numbers were the key. Official statisticians began to take an interest. As Thomas Stapleford lays out in his history of US cost of living indices, in 1902 the US Bureau of Labour Statistics published an official index of wholesale prices, followed a year later by the British Board of Trade.

These were complemented over the following decade by numerous enquiries into the working-class cost of living. The British Board of Trade even ventured into the field of comparative statistics, undertaking a survey of living costs in the United Kingdom, Germany, France, Belgium and the United States. Germany lagged behind. No national system of price statistics was established before the war.

The Prussian military was the first Reich agency to request a national price survey. It wanted a benchmark against which to compare procurement and billeting costs. But, it proved impossible before World War I to devise a system agreeable to both the two most important member states, Prussia and Bavaria, In 1907, in conjunction with the international enquiry by the British Board of trade the newly established Labour Statistics Department of Germany’s Imperial Statistical Bureau did conduct a year-long survey of working-class budgets. But, this important source of information was not exploited for the purpose of constructing an index. Similarly, the member states published annual compendia of wholesale and retail prices but never constructed indices. The debate over the rise in living costs before the war was confined to academics and journalists. Conrad's Jahrbücher für Nationalökonomie und Statistik became a lively forum for debate about the rise in prices. Just before the outbreak of war, the famous Verein für Sozialpolitik completed a survey of living costs in Germany’s major cities. Influential private estimates were published by the renegade social democrat Robert Calwer, the head of the Frankfurt statistical office Moritz Elsas and the chief municipal statistician of the Berlin district of Schöneberg, Robert Kuczynski.

It took the advent of the Weimar Republic to break the deadlock. The Weimar Republic set out to complete the job left unfinished in the 1870s, to create a powerful national administrative apparatus for Germany. Central to this ambition were the new Reich's Ministries for Labour and for Economic Affairs. Establishing their authority meant overcoming the resistance of the member states, much enfeebled by the fiscal crisis of the war. But it also meant negotiating a new working relationship with organized business and labour, whose power was enormously enhanced by the war and its revolutionary aftermath.

For the fledgling Weimar Republic the joint organization of industry and labour, known as the Zentrale Arbeitsgemeinschaft or ZAG, was an indispensable corporatist partner in power. And it was with the ZAG's cooperation that planning began in the late summer of 1919 for nation-wide surveys both of the cost of living and the even more contentious topic of industrial earnings. The surveys would be carried out by the renamed Reich's Statistical Office (Statistisches Reichsamt, SRA) and the statistical offices of the member states. The SRA fell under the responsibility of the Ministry of Economic Affairs. However, the surveys were also of crucial interest to the Ministry of Labour, which was seeking to establish itself as the arbitrator in the emerging system of organized industrial relations. Reliable data on the cost of living and actual weekly earnings would allow its staff to intervene authoritatively in wage negotiations.

Public interest in the new statistics was intense. But Germany's official statisticians made a slow start. The survey of earnings was controversial from the start. It was postponed for three months and soon ran into insurmountable opposition from employers. Work on the price survey went ahead more smoothly. But, it was not until the late autumn of 1919 that the Reich's statisticians, the Ministries and the representatives of labour and business were able to agree on the mechanics of the new survey. The main problems were to choose a representative sample of reporting towns, to decide on the basket of goods to be included and to allocate weights to the contents. Finally, on 18 November 1919 the SRA despatched copies of the official questionnaires to the Statistical Offices of the Länder. Embodying the assumptions of a familial concept of welfare, the basket of goods was based roughly on the requirements of a family of two adults and three children. The survey was to capture a snapshot of the prices of the most important foodstuffs, the cost of fuel for heating and lighting, and rents for the month of December 1919. For lack of any more up-to-date information, the weighting scheme was derived from the survey of working-class budgets taken in 1907. Altogether, the survey was thought to cover ca. 87 percent of pre-war working-class expenditure. Reflecting the priorities of the industrial relations system, the questionnaires were to be completed by all towns with more than 10,000 inhabitants - slightly over 600 in total. This left a majority of German society, what one scholar has labelled the “nation of provincials”, excluded from the survey. Local magistrates were to report the prices as of 10 December 1919 to the Statistical Offices of their Land. To guard against accusations of fraud, the returns were countersigned by committees representing employers and labour. The aim was to plug the price reporting system into the local corporatist networks, which had grown up during demobilization, so as to obtain a consensual basis for making local wage adjustments. The Länder Statistical Offices were to check the returns and compile regional aggregates. The SRA would receive the national returns for December 1919, no later than 28 January 1920.

By early February 1920, the SRA was in possession, for the first time, of a national figure for the cost of living. The survey was then repeated for the 10 February 1920 and monthly thereafter. However, despite enormous public interest, questions in the National Assembly and pressure from the interest groups, the Labour Ministry refused to allow the figures to be published. There were disturbing inconsistencies in the returns from different localities and even, in some cases, between the monthly reports from the same locality. Spot checks revealed why. There were wild variations in the kinds of prices being reported to Berlin. A minimal level of consistency could be enforced only by extensive checking. The release of the official cost of living statistics for February 1920 was not finally approved until the end of August 1920. With this kind of delay the official price statistics were of no practical value. In fact, given the switchback from inflation to deflation and back, the lagged data added to the confusion. The figure for August 1920 was published in early November 1920. At the time, it was clear from daily experience that prices were again rising. However, the official figures for August indicated that at that point prices had been falling by an annualized rate of 47 percent!

In the absence of good national figures from the SRA, alternative indicators of inflation proliferated. Many people looked to the daily foreign exchange quotation as an indicator of the value of the German currency. Commodity prices were readily available in the financial press. And in the spring of 1920, the SRA began to publish an index of wholesale prices. But there was also much activity at the municipal level with local index numbers appearing for Berlin, Magdeburg, Leipzig, Cologne, Nürnberg, Mannheim and other major industrial towns. By 1922 a municipal statistician commented: "not without justification one speaks of index number mania ... Such numbers, ... have shot out of the ground like mushrooms after a mild summer's rain, and the volume of comment on index numbers, their value, or otherwise, in the specialist press, in daily newspapers and in the associational literature, not to mention what is said at public meetings, has become quite unmanageable." Defenders of the purity of the German language railed against the inflationary use of the foreign term "index", yet another curse brought upon the German people by the infernal peace settlement.

The Labour Ministry had more practical concerns. The proliferation of competing statistics was confusing wage negotiators. What was required was an up-to-date official figure. Under pressure from the Labour Ministry, the SRA created the so-called Rapid Reporting Service (Eildienst). From November 1920, forty-seven carefully selected towns were instructed to make rapid reports to the Statistical Offices of their Länder. After spot checking the figures would be wired to the Reich's Office, allowing a national figure to be computed within five days of the end of each month. To create an index, each monthly figure had to be expressed in terms of a base period. The simplest solution would have been to set February 1920 as 100, the beginning of the official price series. However, this was politically unthinkable. In the aftermath of the disastrous war, Germans in all walks of life clung to pre-war standards. The base had to be 1914. But there were no national data for 1914. Publication was delayed to enable time consuming historical enquiries by the statistical offices of thirty-nine cities. The result was the so-called Reichsindex, the first practical index of the national cost of living to be produced by Germany's official statisticians. It finally appeared in the fortnightly journal of the Statistical Office, Wirtschaft und Statistik, on April 19 1921 and monthly thereafter. The Reich also continued to compile the slow moving price survey for the 600 large towns as well as figures for wholesale prices. However, it was the Reichsindex produced within a week of the end of each month that was to provide the authoritative running indicator of the hyperinflation. It was prescribed by the Ministry for Economic Affairs as the standard for contractual adjustments. And it was the basis on which the arbitrators of the Labour Ministry negotiated wage settlements.

What effect did the index have in practice? Was it used as its authors intended? Historians interested in the history of knowledge not just for its own sake but as part of a wider history cannot afford to dodge such questions. Unfortunately, however, it is generally far easier to study the production of texts and representations than it is to analyse the sense that is made of them and the uses to which they are put. In the case of statistics, these problems are somewhat eased by the fact that statistics are often produced with a specific use in mind and their audience tends to consist of powerful and well-documented actors. In the case of Weimar's Reichsindex there can be no doubt that it made an immediate impact. At a time of intense conflict between management, official unions and radical shopfloor militants, the principle that wages should be adjusted for inflation was one point of agreement. The problem was to establish the appropriate measure of inflation and to decide how to respond to any given change in prices. Was the inflation sufficiently large and sustained to warrant a permanent adjustment in wage levels?

During the lull of 1920-1921, German employers argued that the inflationary wave had broken and would soon be followed by the kind of deflation wracking American and British industry. Rather than permanent wage adjustments, they therefore argued for one-off compensation payments. All too often they opposed any adjustment whatsoever. In some areas joint commissions were set up to monitor local prices. However, the arguments on these committees often drove the parties further apart. The new Reichsindex served as a check on both sides. Having agreed to use the official index, the negotiators could establish trigger clauses. For instance, negotiations would only be reopened if the Reichsindex moved by at least twenty percentage points. A statistical presentation became the staple opening move in every session of wage bargaining. The employers, who were at first reluctant to engage in collective bargaining, soon became adept at the numbers game. A favourite ploy was to counter trade union demands by selective quotation from the statistics produced by left-wing municipal statisticians such as Kuczynski. Anticipating this strategy, the Labour Ministry supplied its arbitrators with a compilation of all the latest figures. Technical notes were attached to explain the discrepancies between official and unofficial series. The production and interpretation of price statistics thus became an integral part of the day to day business of labour relations.

Trade unions and business representatives were involved in the production and interpretation of the Reichsindex at every level. In the municipalities, employers and labour representatives were required to counter sign the returns. Some Länder statistical offices established consultative committees to watch over the reports forwarded to the Reich's Office. And trade unions and employers frequently checked the Reich's figures against regional data of their own. Only at the level of the Reich was there no outside input into the overall design and preparation of the index. The details of the weighting scheme and the final process of arithmetic assembly were hidden from view. This was important in giving the statisticians a free hand. It allowed them to respond rapidly and flexibly to the teething troubles of the data-gathering operation. But there was no hiding from the fact that Germany’s national cost of living was in many ways a fragile `object'.

The idea of a national working class food basket was a fiction. Despite large scale migration and urbanization, striking differences in the consumption of meat and fish, milk products and the staple carbohydrate persisted across Germany. The German food trade, compared to that of Britain or the United States was small-scale and fragmented. There were very few standardized items of consumption. That made statistical enumeration extremely difficult. The solution adopted by the Reich's statisticians was to establish a system of `substitutions'. Fish and shellfish, for instance, replaced cheap cuts of meat in the North German figures. Further ad hoc adjustments might be required to take account of specific shortages in isolated areas of the Reich. This balancing act was performed behind closed doors, which was both administratively efficient and helped to preserve the credibility of the Reichsindex as a truly national indicator. In the long run, however, the exclusion of the interest groups from the final assembly of the index was to expose the statisticians to powerful challenges.

The involvement of trade unions and employers in both the production and reception of the Reichsindex had ambiguous implications. On the one hand, their involvement provided the official figures with a minimum level of legitimacy. On the other hand, the widespread adoption of the Reichsindex by the industrial relations system exposed a fragile instrument to intense pressures. The statisticians faced a dilemma. The more widely their index was put to use, the less control they could assert over its interpretation. The meaning of the index was disputed. What exactly did the Reichsindex measure? The statisticians were cautious. It did not define an absolute standard of living, nor did it cover all goods that were essential to a working-class household. For the SRA, the index was properly interpreted as a rough guide to the monthly movements, across the country, of certain important consumer prices. Since it approximated to a working-class budget it was an indicator suitable for making wage adjustments. At a pinch, it might be applied to commercial contracts, though one might dispute its relevance outside the area of household consumption. But, this was all that the statisticians thought was technically possible. Certainly, it was all that their political masters in the Ministry for Economic Affairs and the Labour Ministry were prepared to support. However, this restrictive definition was not accepted by its users. They sought to impose on the index much wider meanings and they criticized its failure to live up to these higher expectations.

One common practice was to use the local components of the Reichsindex to make regional comparisons of price movements. Workers and civil servants in regions with high index values could use these numbers to demand compensating cost of living adjustments. The SRA resisted this use of the index for both political and technical reasons. Regions with low index values were bound to query the accuracy of the official figures. The statisticians strongly suspected that the municipal officials, who were charged with collecting returns, did their best to inflate the figures in the hope of triggering cost of living supplements. There were also technical reasons for resisting such transverse regional comparisons. The Reichsindex may have been advertised as a national measure of the cost of living. However, as we have seen, the uniformity this implied was a convenient fiction. In practice, it was simply not possible to ensure that all the reporting towns noted prices in the same way for exactly the same qualities or even types of food. The black market added another element of uncertainty. The extent of unregulated trading and the prices on black markets varied across the country. And the reporting cities differed in the extent to which they recorded such prices. The effects on local indices could be very dramatic. Since absolute uniformity was unobtainable, the statisticians devoted themselves to ensuring that there was a consistent pattern to the local reports. Spot checks aimed to ensure not consistency between towns, but consistency for each reporting town from month to month. It was in this sense that the index served as a rough guide to national price movements over time. Comparisons between regional sub-units were unsafe. Of course, in the long run, if large regional differences in reporting persisted, the stable pattern of relativities was unlikely to survive. The index would then lose its value even as a measure of movement.

Apart from regional lobby groups, the trade unions were by far the most active interpreters of the index. But, their reading differed fundamentally from that of the official statisticians. By focussing on the role of the Reichsindex as an indicator of movement, the Reich's statisticians sought to deflect attention from the thing that was actually being measured. The precise content of the various baskets of goods was less important than the general direction of price movements. By contrast, the trade unions saw the index as a way of securing official recognition for a certain basic standard of living. For them, the consumption basket represented an actual collection of goods they aimed to secure for their members. Their interpretation swung between the optimistic and the desperate. On the one hand, they demanded that the Reichsindex should cement the pre-war standard. The Reich should take responsibility for measuring how much more it now cost to maintain the pre-war standard of living of the German worker. The trade unions bitterly criticized any attempt by the statisticians to adjust the basket to a more realistic post-war content.

As successive bursts of inflation ate into real wages, the trade unions retreated to another line of defence. The Reich should at least take responsibility for measuring the cost of subsistence for a working class family. This would provide a minimum safety net, indicating the level below which wages could not be allowed to fall. For the statisticians, of course, neither of these interpretations was acceptable. `Their' Reichsindex was no more and no less than a measure of relative movement. But this was not an easy line to defend. After all, the index did claim to represent the movement of the national cost of living. To admit the technical inadequacies that made the Reichsindex incapable of serving the trade union's purpose threatened its entire legitimacy. The bottom line, however, was stark. Neither the Reich's Labour Ministry nor the employers would countenance any official endorsement of a particular standard of living, whether it be the pre-war level or a minimum safety net.

The most widely criticized aspect of the Reichsindex was its failure to include clothing. For the trade unions, this was part of a consistent strategy of immiseration. The comprehensive local indices published by municipal statisticians in Kiel, Hannover and other cities revealed that the cost of clothing, manufactured from imported cotton and wool, had far outstripped the prices of all other items of household expenditure. No coincidence, therefore, that it had been excluded from the official measure of the cost of living! On this point the statisticians were, in fact, in agreement with their critics. The exclusion of clothing was indefensible. The problem here was not one of conflicting interpretation. The problem was to obtain the cooperation of the garment trade. Clothing was the most variegated item in the working-class budget. Given the many regional differences across Germany and the relative underdevelopment of ready-to-wear clothing it was impossible to define a standard wardrobe. Here, the useful fiction of a national cost of living was stretched to its limit. The only way to allow for regional variations was to describe the items of clothing in the most general terms: a “man's suit”, a “pair of heavy shoes” etc. Imposing absolute uniformity would have yielded meaningless results. But, this lack of precision imposed a heavy burden of responsibility on those filling out the local returns. One might reasonably expect municipal administrators to make consistent returns about foodstuffs and coal, but there was no hope of them competently assessing the vast variety of textiles and clothing. Close cooperation with the garment trade was essential. In June 1920, therefore, the SRA entered into negotiations with the business associations of the clothing retailers. They, however, proved recalcitrant. They feared that the publication of regional price figures would unleash ruinous competition between towns with different price levels. The official price figures that began to appear in the autumn of 1920 therefore excluded clothing. Negotiations with the textiles trade dragged on fruitlessly into 1921. Finally, in February 1922, under intense pressure from the trade unions, the SRA initiated its own survey of garment prices without trade cooperation. To obtain consistent reports the statisticians employed the device of physical matching, the last resort of social measurement. A cloth sampler was attached to the questionnaires and the local committees were instructed to ask for the price of clothing made of cloth `of this kind'. Words had failed Germany's statisticians. They could do no better than simply to point to `the things themselves'. This was the logical consequence of the garment trade's attempt to deny the statisticians access to economic reality by denying them access to the specialized language necessary to describe it. Not surprisingly, progress with the sampler-questionnaire was slow. The selection of materials was criticized by those experts who were willing to help. An unrealistically fine grade of cloth had been specified for the men's suits. And the textile association continued to put up opposition, even calling upon its members to make false returns. As usual, there was no hope of imposing absolute uniformity, or of actually checking a significant number of returns. But, by April 1922 a stable pattern seemed to have emerged across the 47 municipalities. The first official figures for the cost of clothing were released in June 1922. They confirmed the trade union's suspicions. With the cost of clothing included, the Reichsindex was almost 10 percent higher.

In the third quarter of 1922, arguments over the interpretation and composition of the index were replaced by one overriding concern: the frequency of its publication. Annualized inflation leapt to 467 percent and then shot up to 1030 percent by February 1923. Germany was entering hyperinflation. Trade unions' efforts to protect their membership increasingly centered on the demand for rapid wage adjustments and ultimately on a call for official indexation. This would make the Reichsindex into the linchpin of economic adjustment and it would necessitate fortnightly or weekly publication. The Labour Ministry had different interests. The acceleration of wage negotiations would, it was feared, stoke the rise of prices. And since the Reichsindex had come to assume such a central role in wage bargaining it was tempting to retard its publication so as to slow down the wage-price spiral. The power of statistics was to be put to the test. Could they be used not only to regularize and harmonize economic transactions but also to alter economic decision-making?

Some larger cities had begun to calculate weekly indices as early as June 1921. Since March 1922 the monthly Reichsindex had itself been based on two samples of prices taken at roughly fortnightly intervals. However, these fortnightly figures were not published separately. Towards the end of October 1922, the Reich's Ministry of Economic Affairs demanded that the SRA should release the fortnightly numbers. Maintaining a monthly rhythm of publication would reduce the Reichsindex to irrelevance. However, the Department responsible for wage control within the Reich's Ministry of Labour, with the strong backing of the Minister himself, opposed any increase in the speed of publication. The Reichsindex continued to appear only once a month. The consequence was disintegration. Given the ferocity of the inflation, negotiators would not wait for the official numbers. Early in 1923 many of the Länder broke ranks and began to publish their own weekly indices. Pay negotiators began to ignore the national statistics in favour of more up-to-date local figures. The unity of the national system was fragmenting. Finally, in May 1923 the matter was taken up by the Reich’s cabinet. Statistics were now too important to be left to the statisticians. Unfortunately, however, the involvement of the cabinet coincided with a temporary lull in the inflation. The Labour Ministry won out. The SRA was ordered to stick with the monthly publication of the Reichsindex. The stage was set for a fiasco.

At the end of May 1923, the Mark suddenly collapsed on the foreign exchanges and a jump in wholesale prices signaled an impending inflationary surge. Anyone who indexed wages on the lagging Reichsindex would suffer a crippling loss in income. Public criticism of the belated publication of the monthly price index mounted to a crescendo. The Labour Ministry was forced into an embarrassing U-turn. In mid-June it ordered the SRA to prepare a weekly index based on a small group of large cities. This figure was to provide the basis for an official system of weekly wage indexation. On past experience, it would take time to achieve even a minimal level of consistency in a weekly reporting system. But, in the summer of 1923 the SRA was forced into a desperate improvisation. To produce a weekly index at short notice they married two distinctly ill-matched price surveys. From July 1923 onwards, the fortnightly returns from the 71 cities now included in the Reichsindex were published separately. In addition, a new weekly price reporting service was established, relying on returns from 29 of the larger cities, all of which also participated in the fortnightly surveys. Apparently, it was thought impossible to demand weekly returns from all 71 cities. To obtain a consistent weekly index, the two series had to be spliced. It was this, which led to disaster. The weekly survey of 29 cities was used to calculate the rate of price changes in the period intervening between the more comprehensive fortnightly polls. The weekly Reichsindex was calculated by applying the week-on-week change measured in the 29 cities to the index level established fortnightly by the survey of 71 cities. In effect, the statisticians were gambling that the movement of prices in the 29 cities, when cumulated over two weeks, would coincide with the inflation rate as measured fortnightly in the larger sample of 71 cities. This was a bet at long odds. It was well known that the rate of inflation tended to be more moderate and more stable in larger, better supplied cities. To make matters worse, the weekly and fortnightly surveys were actually taken on different days of the week - the weekly surveys on Mondays and the fortnightly surveys on Wednesdays - a small difference of timing that was highly significant under conditions of hyperinflation.

Not surprisingly, therefore, the new weekly Reichsindex soon encountered problems. The first of the fortnightly surveys in the 71 towns revealed that the rate of inflation in these towns was significantly higher than the figure calculated from the weekly returns of the 29 towns. However, the lower figure had already been published for the second week in August. On 20 August 1923 the Reich's statisticians unleashed a confused public outcry when they released an estimate for the rate of inflation in the third week of August based not on the figure that had been published a week earlier, but on the higher price level indicated by the sample of 71 cities. The week-on-week change in the 29 cities was only 54 percent, whereas the rate of inflation implied by comparing the composite Reichsindex for the second and third weeks was 72.5 percent. In effect, the SRA had published two vastly discrepant estimates of inflation in a single week. In the fourth week of August the procedure, which had been proved flawed only a week before was simply repeated. The public was blithely instructed that since the results from the fortnightly survey of 71 cities were not yet available, the weekly figure from the 29 cities should be taken at face value. The wage negotiators, who been promised a reliable weekly figure on which to base their efforts at indexation, were confused and angry. The press comment was scandalous. The Reichsindex, which had established itself so rapidly as a national institution during 1921, had forfeited its authority. The blame clearly lay with the Labour Ministry, but it was the statisticians who were in the firing line. To retrieve the situation the pyramid of corporatist representation was capped by the creation of a national price statistical commission on which the key interest groups would oversee the production of this crucial national indicator.

With the stabilization of prices in 1924, the Reichsindex was restored to a more modest role. The basket of goods was changed to satisfy the demands of the labour movement. The rhythm of publication was slowed down, which helped to calm the nerves of employers. A thoroughly revised index was presented secretly to the tripartite Index Commission on 4 February 1925 and released to the public in March. By then, it could count on solid support in the political mainstream. When the official index came under attack in the Communist press, the Social Democratic Vorwärts sprang to its defence. With the full backing of the trade unions, critics of the Reichsindex could now be dismissed as ignorant troublemakers. After six years of struggle, the official index of the German cost of living was finally established as a national institution.

Establishing a regime to measure inflation was thus an elaborate exercise in political argument and social engineering. In the early 20th century, that struggle was defined by the corporatist triangle of the state and the organized interests of employers and labour. Their debates defined the concept of a standard of living, which became one of the anchors of the age of the “great compression” in income inequality. The same period saw the institutionalization of unemployment statistics, sophisticated balance of payments numbers and GDP. The history of price statistics in Weimar’s hyperinflation should interest us at this point not because we are at any risk of repeating that inflation, but because it reveals the politics of economic measurement in the raw. Who counts what, for whom, on the basis of which conceptual grid? When are numbers published? Who has the authority to interpret them? Today, a hundred years on, we may be deep in the weeds debating the relative merits of core inflation numbers. Electronic data may enable us to scrape billions of prices at a time from the internet and compress that data into personalized inflation indices. But these remain the questions that define the politics of economic measurement.