The Early Reception of the Phillips Curve in the United Kingdom: New Evidence from the Papers of the Council on Prices, Productivity, and Incomes

Abstract

As shown in a 2018 article by Cristiano and Paesani, the element of Phillips's 1958 article that immediately obtained attention in the policy debate was the estimate of the unemployment rate compatible with price stability. Building on this previous work and based on the unpublished papers of the Council on Prices, Productivity, and Incomes and other published sources, the present article throws fresh light on why the curve paper failed to influence policymaking. At least two reasons may account for this. First, by the late 1950s and early 1960s, works based on inferential statistics still failed to obtain widespread attention among professional economists in the United Kingdom. Second, Phillips's 1958 article came at the beginning of a process that led to the abandonment, for some time at least, of the idea of buying price stability at the cost of increasing the unemployment rate.

1. Introduction

In recent years, the early reception of Phillips 1958 has become a debated issue. After decades of tranquility, Forder (2014) provided massive evidence that the conventional textbook story of Phillips discovering a stable trade-off between wage inflation and unemployment, with the trade-off immediately adopted as a menu for policy, is only a “myth.” But Forder did not stop at this point. In his book of 2014 and many other writings (more recently in Forder 2019a, 2019b), he has insisted that Phillips's works are irrelevant in the history of economic thought and that Phillips 1958 in particular had no influence at all—a point of view that has been challenged by Laidler (2104), to which Forder (2018) forcefully replied, and more recently by Lipsey (2020).¹

However, despite the disagreements that remain, a couple of essential points seem to be settled. It is now agreed that nobody ever thought to use the Phillips curve as a “menu for policy,” as the myth has it.² And it is also commonly recognized that, immediately after its publication, and even though it eventually failed to exert any influence on current policymaking, Phillips's curve paper did elicit a (more or less) relevant amount of interest and discussion. As far as this latter point is concerned, the disagreement that remains relates to who immediately took interest in Phillips 1958, what of the curve paper attracted attention, and why it failed to have any influence on policy.

Leaving aside any general consideration of the relevance, originality, and influence of Phillips's works as an economist and focusing exclusively on the early reception of Phillips 1958 in Britain, it is on these latter questions that the present article has something new to propose. Building on Cristiano and Paesani 2018b and further archival research, it aims to present an episode that has been neglected so far and that, while confirming that the curve paper attracted attention, can also throw fresh light on the reasons why it failed to influence policymaking. As some answers to these questions have already been proposed, and as they point in different directions, a brief recap is in order.

Following a line of argument that can also be found in Laidler 2001 and that is largely based on de Marchi 1988, Forder (2014: 18–22) attributes the immediate interest aroused by Phillips 1958 to the “Popperianism” that was so much in vogue among Phillips's younger colleagues at the London School of Economics (LSE). The element of Popper's falsificationism that attracted the attention of the LSE youngsters was the idea that “science advances . . . by the framing of bold hypotheses, and subjecting them to tests” (Forder 2014: 19). The LSE youngsters who took part in the activities of the seminar titled Methodology, Measurement, and Testing (M2T) opposed this view to the more traditional methodology usually associated with the name of Lionel Robbins, a leading figure at LSE and the author of Robbins 1932.³ Whereas Robbins claimed that the intrinsic variability of economic circumstances would never permit economists to establish “universal laws” (Forder 2014: 19), it was precisely a law of this kind that Phillips (1958) claimed to have derived from the available data.

Among the young economists who were more attracted by Phillips's (1958) “Popperianism,” Forder (2014) highlights Lipsey, whose reminiscences as put forth in Lipsey 2000 confirm the point made by Forder. However, Lipsey has recently proposed the following account, in which a great emphasis is put on an element that, by contrast, Forder (2014: 12) dismissed as scarcely relevant:

While Forder (2014: 12) has argued that “that sort of numerical estimate of what could be achieved was nothing very new and nothing special either,” the point made by Lipsey has recently found some support in Cristiano and Paesani 2018b.⁴ Moreover, the new evidence presented in the present article confirms that Phillips 1958 attracted a good deal of attention in policy circles and that this happened for its measurement of the rate of unemployment compatible with price stability at “a little under 2 1/2 per cent.” In this respect, more than a new hypothesis about the early reception of Phillips 1958, the present article provides evidence in support of an old one for which, for various reasons, little documental proof has been available so far. But the hypothesis proposed here is not new. It was sometimes proposed even in the “pre-myth” literature, although only sporadically, and not without some contradictions.

Brittan (1964), for instance, placed Phillips within a larger group of academic economists and economic experts at the Treasury who had recently argued, in different ways, that unemployment had to rise to stabilize prices. However, Brittan (1964: 295–96) labeled this “family of ideas” as the “Paish school” (mainly on the basis of Paish 1962), and quoted Lipsey 1960 and Phillips 1962 as examples of the “great deal of technical discussion” about this idea, but gave no emphasis to Phillips 1958. In contrast, writing much earlier (and before the publication of Lipsey 1960), MacDougall (1959: 377) noted that while Paish 1958 (later included in Paish 1962) “gave no very elaborate statistical analysis,” Phillips 1958 was at an advantage, as it “analysed the experience of the last hundred years.” But, again, a look at Dow 1964, a landmark in the literature on British policymaking, reveals that Phillips 1958 is barely mentioned. This, in turn, is contradicted by Hutchison (1967: 144), who highlighted Phillips 1958 as a “much discussed pioneer contribution to the debate on the causation of inflation” and quoted the passage with the estimate of the unemployment rate compatible with price stability at “a little under 2 1/2 per cent.”

A fragment from Knowles 1962, Knowles himself being another economist who took part in the story reconstructed below, may be of some help in making sense of these dissonant accounts. This is how Knowles recalled Phillips 1958 a few years after its publication:

Knowles's brief account has something in common with the more recent one by Lipsey (2020). They both emphasize the deflationary implications that politicians and policymakers, more than economists, could have derived from Phillips 1958. The political dimension of the early reception of Phillips 1958, although sometimes reproduced in personal recollections at a later stage (along with Lipsey 2020; see, e.g., Brown 2000 and Holt 2000), is an element that has taken time to emerge. An endemic lack of evidence, and even some blind spots that will probably always remain, may account for this.

The role played by Phillips 1958 in Hutchison's and Knowles's reconstructions, for instance, is limited to the isolated passages cited above. These sources provide no evidence supporting what they say about Phillips 1958. Nor does Lipsey (2020) add anything more than his recollections in support of the idea that the innovative measurement of the unemployment rate consistent with price stability was the element that attracted attention to Phillips 1958. Wulwick (1989: 173) and Laidler (1989: 35n9) provide more details, but the information contained in these works comes from personal communications. More specifically, A. J. Brown (in a letter to Wulwick) and Richard Sayers and Alec Cairncross (in private communications to Laidler) revealed that Phillips 1958 was largely circulated at LSE before publication. The same sources also reported that in their evidence before the Radcliffe Committee (of which Sayers and Cairncross were members and who wrote the Radcliffe Report), both A. J. Brown, an authority on inflation since the publication of Brown 1955, and Robert Hall, head of the Economic Section at the Treasury, had Phillips 1958 in mind (possibly among others) when they expressed their opinions that there was no dependable relation between unemployment and inflation (Hall) and that the level of unemployment compatible with price stability could be very high (Brown).⁵ However, the name of Phillips is never explicitly mentioned in the Radcliffe Committee's published documents. And even when written records may have existed, these have not resurfaced. In his biography of Hall, Kit Jones (1994: 149) reports that “[Phillips] sent Hall a draft of his article for Economica” and that “Hall took a good deal of trouble replying to him.” The “article for Economica” was Phillips 1958, so this correspondence between Hall and Phillips would be an extremely relevant piece of evidence, but Jones gave no references as to where it can be found. And the list may continue with the “series of lectures . . . attacking the [Phillips] curve on theoretical and empirical grounds” that, according to Lipsey (2000: 236–37), R. F. Khan “was reputed” to have given and of which, to the best of my knowledge, no evidence has ever been found.

More recently, Cristiano and Paesani (2018b) have provided a more systematic account, based both on published sources and archival material, confirming that the measurement of the unemployment rate consistent with price stability is the element of Phillips 1958 that immediately captured attention in the policy debate. Cristiano and Paesani (2018b) also show that, when Phillips 1958 appeared, numerous estimates of the rate of unemployment that could stabilize prices were being proposed, usually with little or no explanation as to how they were arrived at—a lack of empirical foundations that certainly made of Phillips's measurement a potential game changer. Cristiano and Paesani (2018b) also confirm that, as reported by Lipsey (2020), some opposition to Phillips 1958 came from Cambridge, but they also show that important criticisms came from LSE. This happened early in 1961, when the LSE economist Henry Phelps Brown drafted a paper for the Council on Prices, Productivity, and Incomes in which the curve paper was strongly criticized. As this recent finding suggests, and the present article confirms, the council occupies a significant place in the history of the early reception of Phillips 1958.

Phelps Brown's paper on Phillips 1958 mentioned in Cristiano and Paesani 2018b is dated February 1961. Shortly afterward, Phelps Brown's comments on Phillips 1958 became one of the specific topics for discussion in a series of meetings with academic economists that the council held in May and June 1961 at the universities of Oxford, Manchester, Glasgow, and Cambridge, after a previous meeting at LSE, in which Phillips took part but that was more focused on a presentation by Paish.

To the best of my knowledge, this episode has never emerged so far, maybe as a result of the minuscule attention that the history of the council has received. In most cases, the council is given no more than passing references in the literature on British policymaking in the period of Conservative governments from 1951 to 1964.⁶ A complete history of the council, based on the several hundreds of pages of documents conserved among the Treasury files at the National Archive at Kew, London, remains to be written.⁷

For reasons of space, extremely selective use of these materials is made here. The brief history of the council in the next section is no more than a sketch of how the council evolved in the four years of its existence. Its focus is on how the council stimulated the debate on the existence and measurement of the unemployment rate consistent with price stability. This happened in 1957 and 1958, when the council's view was that unemployment had to rise in order to stabilize prices.

2. A Brief History of the Council

The decision to establish the Council on Prices, Productivity, and Incomes was announced to the House of Commons by Peter Thorneycroft, chancellor of the exchequer in the first Macmillan government, on July 25, 1957. In August, Lord Lionel Cohen, a law lord, accepted the chairmanship, joined by the economist Dennis H. Robertson and the accountant Harold G. Howitt. The establishment of the council is usually mentioned as an episode in Thorneycroft's stormy tenure as chancellor of the exchequer between January 1957 and January 1958. The salient aspect of Thorneycroft's chancellorship was the attempt to impose price stability as the primary objective, even at the cost of breaking the taboo of full employment. As all reconstructions confirm, this deflationary turn created conflicts within the cabinet as well as within the Treasury, where Thorneycroft's ideas were opposed by Robert Hall, head of the Economic Section, the group of professional economists who worked full-time as government advisers.⁸

Disagreements were so strong that on some occasions Hall expressed his willingness to resign until it was Thorneycroft who eventually resigned as the cabinet rejected his proposals aimed at curbing public expenditure.⁹ However, a point upon which the minister and his chief economic adviser concurred was that unemployment had recently been too low—and that, in part at least, this explained inflation.

As the council reports would confirm, throughout the 1950s, a continuous tendency of wages to rise faster than productivity, which produced what was called “creeping inflation,” had been endemic in Britain. In the attempt to stabilize prices, since their return to office in 1951, the Conservatives relied mainly on aggregate demand control. Meanwhile, following the recommendations of the white paper titled Employment Policy (Cmd. 6527) of 1944, maintaining “high and stable” levels of employment remained the priority. This situation put policymakers in a strict corridor, in which aggregate demand sometimes had to be cut to reduce inflation but could not be cut too much, lest this could result in high levels of unemployment. As a viable solution to the problem of reconciling stable prices with high employment, incomes or wages policy was sometimes proposed, but the white paper of 1956 titled The Economic Implications of Full Employment (Cmd. 9725) put a firm limit on this option by excluding direct intervention in wage bargaining. Incomes policy was thus reduced to frequent exhortations to wage restraint that obtained hardly any results.

During a meeting with the council held on October 22, 1957, Roger Makins, permanent secretary at the Treasury, presented this situation as “the background to the Council's appointment.”¹⁰ According to the record of the meeting, “Makins said that . . . the Government had been trying to operate a policy of persuasion inculcating restraint on all concerned in the inflation of incomes.” It was now becoming clear, however, that so far persuasion had failed and that new measures had to be adopted. In this context, the council's task should have been to convince the public that an increase in unemployment was necessary in order to reduce inflation. No consensus existed, however, on how much unemployment should have risen, nor was it clear whether an increase in unemployment was in itself sufficient to stop inflation. It was on these crucial issues that Hall and Thorneycroft were at odds.

Reduced to its operational terms, the disagreement between Thorneycroft and Hall pertained to the nature and measure of the rate of unemployment compatible with stable prices. Hall agreed that unemployment had recently been too low and that this corresponded to a situation of “over-full employment,” as he would also explain to the council on October 22. But while Hall was ready to support a limited aggregate demand reduction, he aimed at making price stability compatible with full (although not over-full) employment. The policy mix suggested by Hall included the adoption of a “guiding light” to prices, that is, an indication, based on the expected growth of productivity, of the maximum increase of wages compatible with stable prices. Thorneycroft, by contrast, put more emphasis on the need to reduce aggregate demand, even at the cost of higher unemployment rates. “We must be prepared to see the [unemployment] rate rise to around 3 per cent”—he wrote in a note dated September 7, 1957—“since unless we can get away from the idea that over-full employment must be supported at all costs we have no hope of curing inflation.” On his part, Hall feared that, unless some form of incomes policy was adopted, the necessary level of unemployment might be very high, probably higher than the 3 percent envisaged by the chancellor.¹¹

In terms of economic analysis, the disagreement between Hall and Thorneycroft reflected the underlying debate between the cost-push theory and the demand-pull theory of inflation. That of Thorneycroft was a typical demand-pull response to inflation. Hall's view, on the other hand, was more nuanced. It combined a demand-pull element, the pull on wages exerted by excess demand in the case of “over-full employment,” with a cost-push factor, the push on wages exerted by the unions in wage bargaining. This view implied that removing excess demand may not suffice if no restraint was put on the cost-push factor. During the council meeting of October 22, Hall suggested that “a position of balance might be reached between 1 1/2 and 2 per cent unemployment.” Up to this level, he saw a situation of over-full employment in which inflation was unavoidable, no matter how the unions would behave. But when the “position of balance” was reached, the inflation rate came to depend on the unions' behavior, making wage restraint necessary. Without this restraint, Hall thought that the unemployment rate necessary to stabilize prices was “indeterminate.”¹²

When he presented this analysis to the council, Hall probably knew that it might fall on deaf ears, as the council's First Report would confirm. Published in February 1958, the First Report fully reflected Robertson's ideas as they had already been expressed on other occasions.¹³

The report blamed the postwar inflation “on an abnormally high level of demand . . . maintained for an abnormally long time” (Council 1958a: sec. 78), explicitly rejected the cost-push argument (secs. 88–89) and the idea of a guiding light to prices (sec. 109), and took a firm stance in favor of the idea that it was necessary to increase unemployment in order to stabilize prices:

How large this “perceptible margin of unemployment” had to be, however, was not specified. The report admitted that there were “pessimists” who thought that the equilibrium level was much higher than the current level, while the council was taking “a more optimistic view” (Council 1958a: secs. 136–37), but not very much was offered in support of this optimism. In a context in which the cost-push argument had a large circulation, even within the Economic Section at the Treasury, the report assumed but did not demonstrate a demand-pull analysis of unemployment. No wonder, therefore, that much of the council's fame is due to the controversy caused by its First Report. It is on this episode that the abovementioned passing references to the council usually insist.

The archival resources shed some light on the events that followed. Soon after the publication of their First Report, Cohen and Robertson expressed their will to publish a “progress report within the next six months” and then to resign, and so did Howitt.¹⁴ The Second Report appeared in August 1958. Although it paid some lip service to full employment policies, it was a defense of the First Report and did not boost the council's popularity. Robertson eventually resigned in December 1958, to be replaced by Phelps Brown in February 1959. For a few months, Phelps Brown joined Cohen and Howitt, who had had to remain for lack of any viable alternative, until the publication of the Third Report in July 1959. Although admittedly in a purely interlocutory way, the Third Report opened the door to the various proposals on how to implement a “money incomes policy.” This novelty announced the U-turn eventually taken in the Fourth Report, published in July 1961, after the council had been suspended in autumn 1959 to be reactivated one year later with two new members who joined Phelps Brown to form a new council. Lord Heyworth took Cohen's place. Harold Emmerson (another accountant) replaced Howitt. After the publication of the Fourth Report, the council ceased to exist.

The Fourth Report was an innovative document. It positively denied that “appeals to restraint” could be of any use, and while admitting the need to get aggregate demand under control, it also denied that this might be enough “unless the restriction goes so far that the remedy is worse than the disease” (Council 1961: sec. 8). In its concluding sections, the Fourth Report insisted that a reform of the wage bargaining process was necessary and revived the idea of a guiding light to prices. It was well accepted in the press and government circles.¹⁵

In their history of the Economic Section, Cairncross and Watts (1989: 342) highlighted the report of the Organisation for European Economic Co-operation (OEEC) published as Fellner et al. 1961 and the council's Fourth Report as the two official documents that, for their resolute support to incomes policy, marked a turning point in economic advising. As far as the early reception of Phillips 1958 is concerned, these two documents also mark the defeat of the policy proposal that found empirical support in the curve paper. In the context in which it appeared, Phillips 1958 was a confutation of the cost-push argument and a piece of analysis supporting the idea that a moderate rise in unemployment would stabilize prices. However, most of Phillips's colleagues, and even some of those who were supposed to be more sympathetic to him on a policy ground, failed (or refused) to appreciate the methodological innovation that the curve paper was introducing. In part at least, this may have depended on a rather simplistic interpretation of the curve paper to whose creation Phillips himself perhaps inadvertently contributed.

3. Phillips's Curve Paper: A “Rush Job”?

The declared purpose of Phillips 1958 was to test the hypothesis that, exception being made for periods in which import prices had increased significantly, the level of unemployment and its rate of change could be used to predict the rate of change of money wage rates. Phillips took the unemployment rate as a proxy for excess demand, while the only cost-push element considered was the increase of import prices. As Forder (2019a: 591) points out, translated into the terms that were current in the debate, this corresponded to the hypothesis that the rate of change of wages depended on the interaction of demand and supply in the labor market, while the institutional factors related to wage bargaining had no effect on the dynamic of wages.

To verify his hypothesis, Phillips selected the period 1861–1913 as one in which, but for very few years, import prices had never risen significantly. If the starting hypothesis was correct, there should have been a stable relationship between unemployment and wage inflation in these years, which is what Phillips found in the first part of his paper.

Although the contents of Phillips 1958 are familiar, it is necessary to recall them before taking into consideration how the obtained results were presented in the curve paper and then received in the debate.¹⁶ On a scatter diagram in which each point corresponded to the combination of the unemployment rate and the rate of change of money wages for one of the years 1861–1913, Phillips connected the points in chronological order. In this way, the diagram points created “loops,” with each loop covering the period of one business cycle. The loops also showed a tendency for the rate of change of money wage rates at any given level of unemployment to be above (below) the average for that level when unemployment was decreasing (increasing) during the upswing (downswing) of the trade cycle. To obtain the average values around which the loops moved, Phillips divided the range of the unemployment rates covered in the diagram into six intervals. He then calculated the average unemployment rate and the average rate of change of money wages for each interval and indicated these average values with six crosses in the scatter diagram. Finally, assuming that each cross represented a fair approximation of the average rate of change of money wage rates for the corresponding average level of unemployment, he estimated the equation that fitted the six crosses.

Then came the second step. Phillips confronted the data on unemployment and the rate of change of money wages for the periods 1913–48 and 1948–57 with the values that could have been predicted using his equation for the years 1861–1913. For the interval 1913–48, the most interesting result was a confutation of the standard conjecture about the downward rigidity of wages determined by stronger trade unions after the return to gold in 1926. As Phillips (1958: 295) boldly pointed out,

For the years 1948–57, Phillips obtained similar results. This is shown in the table on page 298 of the curve paper, which is partially reproduced here as table 1. In the table in Phillips's curve paper, for each year, the first column reports the observed change in wage rates. In the second column, under “demand pull,” there is the change in wage rates predicted by the fitted equation for the years 1861–1913. This is followed in the third column by the “cost push” element, defined as the rate of change of wages that was necessary to keep money wages at pace with the cost of living as measured by an index of retail prices. Finally, in the fourth column, Phillips reproduced the change in import prices in the preceding year.¹⁷

Under Phillips's definition, the cost-push element could become effective only in those years in which the increase of wages due to the demand-pull factor was not sufficient to cover the increase in the cost of living. This had been the case in 1948, when wages had risen 3.9 percent instead of the predicted 3.5 percent. The figures for 1949 seemed to contradict Phillips's hypothesis, with an observed change of wage rates at 1.9 percent, well below the predicted 4.1 percent, in a year in which the cost-push element was at 2.9 percent. But, as Phillips noted, this could be explained as the result of the political support that the unions gave to Stafford Cripps, a Labour chancellor, and his policy of wage restraint. The observed data for 1950 (4.6) was in line with the predicted 4.4 percent. For 1951 and 1952, the rapid increase of import prices in 1950 and 1951 could explain observed data well above the predicted values. Finally, from 1953 to 1957, when the cost-push element was below the critical level, the table showed an almost perfect correspondence between the predicted “demand pull” increase of wage rates and the observed data. This latter part of Phillips's table is reproduced in table 1. For the period 1953–57, the correspondence was indeed remarkable.

Taken at face value, Phillips's analysis suggested at least two conclusions. The first one, although not explicitly stated, was that there was no need to resort to institutional elements connected with wage bargaining to explain inflation after 1952. The second and explicit one was that, assuming an increase in productivity of 2 percent per year, a rate of unemployment of “a little under 2 1/2 per cent” would have stabilized prices (Phillips 1958: 299). However, the way in which the curve was constructed, with the six crosses representing average values, suggested a less deterministic interpretation, in which any mechanical link connecting each level of unemployment to one level, and one only, of wage inflation would have to be excluded. Moreover, that the curve could not represent any mechanical relationship of cause and effect between unemployment and wage inflation seems to be altogether excluded when Phillips's previous works on stabilization, and especially Phillips 1954, are considered. Based on Phillips 1954, a rate of unemployment of “a little under 2 1/2 per cent” could never be a sufficient condition for price stability, but it may possibly be the target rate of a stabilization policy. Indeed, the only interpretation of the curve that seems to be consistent with Phillips's works on stabilization seems to be the one by Cliff Wymer, a friend of Phillips, quoted as a personal communication to the author in Bollard 2016: 170: “The relationship derived and estimated by Phillips was the locus of the long run equilibrium solution of a differential equation model. It was not a causal relationship and Phillips never suggested it.”¹⁸

Although reasonable, however, the point made by Wymer remains controversial. Desai 1975 was probably the first interpretation of the original Phillips curve as a long-run relationship, but it was rejected by Gilbert (1976), on the ground that it was more an application of Goodwin 1967 than a reconstruction of what Phillips had been doing while estimating the curve. In Gilbert's (1976: 57) perspective, Phillips's averaging procedure is taken to be no more than a “computational device,” deprived of any interesting “economic or statistical significance.” In their papers, however, Desai and Gilbert did not attempt to put Phillips 1958 in the context of Phillips's research agenda. When this has been done, for instance in Laidler 2001 and Hoover 2017, it has emerged that, if the curve paper is to be made consistent with Phillips's works on stabilization, Wymer's interpretation seems to be the one to be accepted.¹⁹ But, as Hoover (2017) has observed, the link connecting the curve paper with its theoretical underpinnings in Phillips 1954 is at best only “implicit” in Phillips 1958, which contributed to create the impression of the curve paper as a one-off endeavor, scarcely related to Phillips's main research agenda.

Only at a later stage, in his inaugural lecture as Tooke Professor at LSE on November 28, 1961 (published as Phillips 1962), did Phillips provide some indication as to the place that the results obtained in Phillips 1958 could have within his quantitative and dynamical approach to stabilization policy. In introducing the subject, Phillips argued that knowledge of the quantitative relations between economic variables that are either objectives or instruments of policy is necessary if the goals are to be obtained. Taking the unemployment-inflation relationship as an example, he assumed that, if any relation of this kind holds true, it will be valid “in given institutional conditions and on average over a period of years” (Phillips 1962: 2). Somehow distancing himself from the myth recently debunked by Forder, Phillips presented this relationship as a constraint rather than as a menu for policy: “Failure to take account of it”—he wrote—“may lead to the adoption of inconsistent objectives” (Phillips 1962: 2). Then, in a more specific section titled “Employment and Inflation,” he gave some afterthought to the curve paper, also in the light of other works that were now available.

In this section, Phillips (1962: 11) admitted that his attempt at measuring the relationship between unemployment and wage inflation in Phillips 1958 had been “very crude.” Comparing his assumption “that changes in the cost of living only affect wage changes in years when prices are rising rapidly” with the one made in Dicks-Mireaux and Dow 1959, Klein and Ball 1959, and Lipsey 1960, that “changes in the cost of living have a proportionate effect on wage rates in every year,” he also admitted that this latter assumption had proved to be “nearer the truth”—although he also claimed that, if this latter assumption proves to be true, it is necessary to take account that “changes in the cost of living are in turn mainly the result of earlier changes in wage rates and to a lesser extent of changes in import prices.” Finally, he suggested how his method could be combined with others' and employed for “prediction purposes”:

Phillips also pointed out that “the relation I obtained [in Phillips 1958] is best considered as a prediction relation of this sort,” and he concluded as follows:

Compared to Phillips 1962, Phillips 1958 made much bolder statements. It seemed to deny, rather than to assume, that the institutional element was important, and it put much more emphasis on the year-by-year correspondence between predicted and observed data, an eye-catching element in the context in which Phillips 1958 appeared. The council vicissitudes presented in section 2 clearly show that, when Phillips 1958 was published, the existence and measure of the unemployment rate consistent with price stability was a hotly debated issue even within the Treasury. As an observer who was on the spot would put it a few years later, “This question was the perennial, central operational issue of Britain's post-war domestic economic policy: the question as to the level of demand at which the economy should be run, as usually roughly judged by the criterion of the percentage of unemployment” (Hutchison 1967: 131). Put in this context, with no reference to its theoretical presuppositions (as in Phillips 1954, 1957) and no indication as to how the results should be interpreted (as in Phillips 1962), Phillips 1958 invited a simplified interpretation.

It is part of the Phillips curve oral tradition that Phillips considered the curve paper a “rush job” or a “wet weekend's work.” In most cases, these words are comments on the way the curve was constructed.²⁰ However, it was not only, and probably not so much, the econometrics of the curve that made Phillips 1958 scarcely plausible in the eyes of its earlier critics. More relevant was that these critics and its leading supporter (Lipsey) concurred in considering it a mere statistical regularity, thus excluding from view any link with the theoretical underpinning that Phillips's works on stabilization may have provided. And also very important was that Phillips's (1962) own indications as to how his empirical results should have been properly interpreted came only at a later stage. Meanwhile, the almost perfect correspondence between the observed and predicted rate of change of money wages for 1953–57 is what attracted attention. From the M2T seminar methodological standpoint, this result was remarkable. The fact that the post-1948 points on Phillips's graph lay closer to the curve than did the pre–World War I data to which it had been fitted was a startling and extremely attractive example of a hypothesis passing the test of predicting observations that had not been used in its formulation.²¹ But outside the M2T restricted group of enthusiasts, the result Phillips had obtained looked very much like a pure coincidence. It could not be denied, but it did not seem credible.²² As a result, the immediate reactions focused on the estimated level of the rate of unemployment compatible with stable prices and were, on the whole, rather harsh.

Knowles and Winsten (1959: 114) described Phillips's “a little under 2 1/2 per cent” as “a gloomy pronouncement by present-day standards.” Moreover, looking at the data presented in Phillips's paper, Knowles and Winsten noted that the same level of unemployment could correspond to different rates of change of money wage rates. Considering how the fitted curve had been obtained, this was quite obvious. Nonetheless, as the next section will show, the same observation that can be found in Knowles and Winsten 1959 would later be made also by Phelps Brown in his council paper of February 1961.

A more vigorous attack came from Routh (1959). Routh argued that Phillips had based his analysis on an utterly unreliable set of data, one prominent observation being that the definition of unemployment had undergone several modifications since 1861. Moreover, as Knowles and Winsten lamented the complete lack of attention to the historical, institutional, and social changes that had occurred over the last century, Routh (1959: 299) accused Phillips of painting the picture of an “institutionless world.”

Significantly enough, even Robertson attacked Phillips on this ground. Phillips (1958) indirectly supported Robertson's demand-pull argument, but the former member of the council had little confidence in econometric methods. The proceedings of a conference of the International Economic Association held in 1959, published as Hague 1962, report this summary of Robertson's views:

Comments of this kind, so much in tune with old “Robbinsonian” methodology, betray a certain reluctance to follow Phillips on his new “Popperian” ground. One aspect, in particular, is revealing. To say that Phillips had ignored the institutional, political, and historical factors is questionable. He positively admitted that a political factor could explain the behavior of prices in 1949. Only, he found that, based on his empirical conjecture, there was no need to call into question these factors, or some change in any of them, to explain the behavior of wages after 1952 (and immediately after 1926). The reliability of his data and the accuracy of his measurement method could undoubtedly be put into question, but to say that Phillips had ignored the noneconomic factors seems to be a further misrepresentation of his work.²³

On the whole, the early critics failed to appreciate what Phillips, for right or wrong, had been trying to do. Instead of using data to confirm some a priori argument, Phillips was letting the data speak for themselves. This aspect is strongly emphasized in Lipsey 1962, a defense of Phillips 1958 and a manifesto against “a priori arguments,” described as “inconclusive,” and in favor of using quantitative methods for the discovery of “stable, hence predictable, patterns of human behavior in the economic sphere [that] must be sought from empirical data” (Lipsey 1962: 109). Lipsey sided with Phillips, not because his analysis had proved watertight, as Lipsey (1960) himself had recently shown, but because he saw no other way of providing a “valid theory of wages” than starting from the data. From Phillips's (1958) and Lipsey's (1962) standpoint, Knowles and Winsten, as well as Routh and Robertson, were shifting the burden of proof. Reasoning on a priori ground, they were assuming, but not demonstrating, that there was some correlation between institutional, political, and psychological change on the one hand and the behavior of wages on the other.

Another way of criticizing the curve paper was proposed in MacDougall 1959. MacDougall gave due emphasis to Phillips's effort to come to a conclusion that was empirically based but noted that other works in which alternative hypotheses were empirically tested were now available, and, contrary to what Phillips (1962) would argue, according to MacDougall they pointed in a different direction. The first to appear after Phillips 1958 was Dicks-Mireaux and Dow 1959. While confirming that a correlation existed between unemployment and the rate of change of money wages, Dicks-Mireaux and Dow noted that it had become less evident after the war, giving more emphasis to the “pushfulness” of trade unions. Along with the already mentioned Klein and Ball 1959, another paper that appeared before the council tour of UK universities is Klein, Ball, and Hazlewood 1959. Both these papers were spin-offs of the preparatory work for Klein et al. 1961. To take account of the implications for wages of political change that had happened in 1952, Klein and his coauthors employed a dummy variable called F, representing the “political factor in wage bargaining,” this factor being that “the trade unions became more aggressive or militant in pressing their claims after 1952 under the Conservative Government” (Klein, Ball, and Hazlewood 1959: 5; Klein and Ball 1959: 467).

As MacDougall (1959) observed in his survey, taken together, these works suggested that Phillips was wrong: the rate of unemployment was not enough to explain the trend of prices in Britain over the last few years. And then came Lipsey 1960. As pointed out in Cristiano and Paesani 2018b, although it confirmed that a significant relation had been discovered, Lipsey's more detailed and accurate study put into question the possibility of reaching any definitive estimate of the rate of unemployment compatible with price stability, thus strongly weakening the most relevant policy implication that could be derived from the simplistic reading of Phillips's curve paper that was prevailing.

As this brief survey has shown, strong arguments were being put forth in the econometric literature that could be used against Phillips 1958. However, arguments of this kind were barely mentioned during the meetings with academic economists in 1961. The council seminars almost continuously remained on the ground covered by Routh (1959) and Knowles and Winsten (1959). With very few exceptions, the available econometric works remained in the background, and analogous considerations apply to Phelps Brown's “Council paper n. 126,” upon which the discussions were apparently based.

4. Phelps Brown, the Council, and Phillips 1958

As shown in Cristiano and Paesani 2018b, among the Phelps Brown papers conserved in the LSE archives, there is the typescript of a paper, “Notes on Causes of the Movement of Money Wages in Recent Years,” dated February 24, 1961, that served the purpose of dismantling Phillips's (1958) argument. A copy of the same document is extant among the council papers at the National Archives, cataloged as “Council paper n. 126” (in T 178/5). Documents in another file (T 178/14) seem to suggest that Council paper n. 126 (or, maybe, a redraft of this document) was included as “Paper 3” in a set of six papers that were circulated as the basis for discussion during the meetings that the council held at Oxford, Glasgow, Manchester, and Cambridge in May and June 1961.²⁴ If these meetings had any effect, this was to reinforce Phelps Brown's conclusion that Phillips's estimate of the equilibrium rate of unemployment was unreliable; a previous meeting held at LSE, in which Phillips was present, could not do much to change the situation.

In Council paper n. 126, Phelps Brown used the median rate of unemployment and the median rate of change of earnings in 130 industries to show that, even though there were significant differences between industries, a Phillips-like relation between unemployment and the rate of change in earnings was, on the whole, roughly confirmed.²⁵ The problem was that, replicating the view already expressed in Phelps Brown 1958, such a relation was not “dependable” enough to make it a guide for policy. It is clear from the following and other passages from Council paper n. 126 that are quoted below that the idea of Phillips's fitted curve as the locus of long-run equilibria was completely alien to Phelps Brown, nor did he consider the possibility that Phillips's results could be valid “on average over a period of years,” as Phillips (1962) would later suggest:

To show that no “continuous quantitative relation, on which the authorities could rely,” existed, Phelps Brown made no mention of the existing econometric literature, even though it is implausible that he did not know at least the few papers mentioned above. His way of reasoning, in any case, was much more akin to Routh (1959) and Knowles and Winsten (1959) than to, say, Klein and Ball (1959) or Lipsey (1960). Just as Knowles and Winsten (1959) had done, he noted that just looking at “Prof. Phillips's materials,” one could observe that “for example, an unemployment rate of about 1 1/2 [per cent] has been associated since the war with six different rates of rise in wage-rates, varying from under 2 to over 10 per cent a year.”

As pointed out in section 3 above, this contradiction between the data employed in Phillips 1958 and the straightforward policy prescription that seems to derive from the curve becomes more apparent than real once the way the curve was constructed or its relation to Phillips's previous work on stabilization is taken into consideration. Moreover, it is unlikely that Phelps Brown, who is credited (see Lipsey 2000: 234–35; 1978: 50) with providing the data employed in Phillips 1958 and who was in close contact with Phillips at LSE, was unaware of Phillips's research on stabilization. On the other hand, it must be observed that when Phillips was given the opportunity to make the link connecting the curve to his own research on stabilization more explicit, he apparently failed to do so. As a prologue to the tour outside London, the council took the opportunity to discuss these topics directly with Phillips, among others. This happened in the sixty-seventh meeting with witnesses, which the council held at LSE on March 21, 1961.²⁶ However, it seems that not even on this occasion did the link connecting Phillips 1958 to Phillips 1954 emerge.²⁷

The record reports that the meeting began with a “seminar” by Paish and that “Professor Phillips then followed suit.” The topic of Paish's seminar was a memorandum on the proposal, which is analogous to Phillips's equilibrium rate of unemployment, that in order to stop inflation, it was necessary to accept “a margin of unused capacity.” It is not clear with which topic Phillips “followed suit.” Judging from the “conclusions” at which he is reported to have arrived, this included those institutional aspects that he allegedly ignored in the curve paper and that would find renewed consideration in Phillips 1962:

Nevertheless, the relation between wage inflation and unemployment was certainly discussed:

That Phillips's results had to be considered valid on average is at least implicit in Lipsey's remark. However, based on the record of the meeting, all that Phillips would have said in reply to Phelps Brown is what follows: “Professor Phillips said he had done some work on the past relationship of unemployment and wage increases in the United States and Australia, and the connection seemed fairly good.”

As far as Australia is concerned, this probably refers to Phillips (1959) 2000, a replica of the curve paper based on data on Australia in which more refined methods of analysis were employed. Phillips presented this work at the economics department of the University of Adelaide but never submitted it for publication. As pointed out by Bollard (2016: 172–73), in part at least, this reluctance to publish the Australian paper may have resulted from the severe criticism received at Adelaide. Moreover, Bollard argues that Phillips's interest in this kind of research faded away after 1959. However, the record of the council meeting at LSE in 1961, in which Phillips also referred to similar work he had been doing on US data, seems to show that, at this date, Phillips may have not yet completely abandoned this line of research. Moreover, the same evidence apparently confirms that Phillips was reluctant to connect the line of empirical research pursued in Phillips 1958 to his theoretical work on stabilization. In the event, this link never emerged, neither during the meeting at LSE nor in the other meetings with academic economists.

Another thing that these meetings reveal has more to do with Phelps Brown. Unlike Robertson before him, Phelps Brown was extremely careful in collecting the opinions of his fellow economists, at least on the most relevant topics, before putting his signature on the council's Fourth Report. An important decision the new council had to make concerned the proposal, made in the First Report, of raising unemployment to stabilize prices. In the end, undoubtedly influenced by Phelps Brown, the council positively rejected this policy. Furthermore, the meetings with economists seem to show that, while Phillips 1958 had become representative, at least in the council proceedings, of the idea of buying price stability at the cost of more unemployment, Phelps Brown received considerable support in rejecting that idea.

At Nuffield College, Oxford, on May 18, 1961, the participants were the warden (Norman Chester), P. W. S. Andrews, T. Balogh, J. R. Hicks, E. F. Jackson, K. G. J. C. Knowles (of Knowles 1962 and Knowles and Winsten 1959), I. M. D. Little, Sir Donald MacDougall (of MacDougall 1959), C. R. Ross, and G. D. N. Worswick. The record reports that when “the meeting . . . turned to the paper on ‘Causes of the Movement of Money Wages in Recent Years’ (Paper 3),” “Mr Jackson said that the theories of Professor Phillips on the relationship between unemployment and wage increases were not worth considering. Mr. Balogh said he thought there was no systematic relationship between earnings and rates; one could not say that movements in the one caused movements in the other in any predictable way.” After these remarks, the discussion momentarily turned to an exchange of opinions between Worswick, Knowles, and Phelps Brown on the relationship between rates and earnings in different industries. Then, one of the council members (Emmerson) brought the discussion back on track:

Unfortunately, it has not been possible to identify “the last sentence on page 5” mentioned by MacDougall. Even if the contents of Paper 3 mentioned in the record corresponds, as it seems plausible, to those of Phelps Brown's Council paper n. 126 (or some variation of it), we do not know in which form this paper was transmitted to the members of Nuffield College.²⁸ Further doubts emerge from the record of a restricted meeting that the council held the morning after, before leaving from Nuffield College. In the summary of this meeting (in T 178/16), there is this sentence: “It was felt that the reactions to Professor Phillips's ideas had been interesting, and it was noted, in view of the way the sentence about a 3 per cent unemployment rate had been picked on in the seminar, that in its Report the Council would have to phrase very carefully any reference to the advantages or otherwise of using unemployment as a weapon to fight inflation.”

Once again, it is not clear where this “sentence about a 3 per cent unemployment rate” comes from. Neither Phelps Brown's paper nor the record of the meeting with the Oxford economists includes such a sentence. Nonetheless, even though a complete reconstruction of the events is not possible, the message that the council received at Oxford is clear, nor did it change at Glasgow University, where T. Johnston and I. Stewart from Edinburgh University joined A. K. Cairncross, D. J. Robertson, T. Wilson, and J. R. Parkinson to meet the council on May 23, 1961.

Not surprisingly, at least Parkinson expressed some support for Phillips: “Mr. Parkinson said that he disagreed with the scepticism expressed about Professor Phillips's basic premise; he thought there was a relationship between unemployment and the rate of increase of wages.” Parkinson was a convinced supporter of the demand-pull school and the author of the leading contribution in the symposium on inflation edited by Cairncross and published in the Scottish Journal of Political Economy in 1958. Even more than in the symposium, however, his voice remained isolated. As far as the summary of the meeting can show, a reply to Parkinson came from Phelps Brown, who said that “he agreed that very high or very low unemployment had an effect on the rate of increase of wages; but it was much less obvious that there was a relationship over the intervening range of unemployment.” As it seems, while no one replied to this argument, the discussion continued with some remarks on a hypothesis, introduced by Phelps Brown, that “a given level of unemployment might have a different effect on the rate of increase of wages according to whether it had just been reached, or had been in operation for some time.” This was a rare occasion in which the idea that Phillips's “a little under 2 1/2 per cent” might become the target of a stabilization policy was at least adumbrated. But after a few remarks on this point by D. J. Robertson, Heyworth, and Cairncross, the meeting moved on to another topic.

The day after, May 24, the council was at Manchester University to meet C. F. Carter, J. Johnston, B. R. Williams, H. A. F. Turner, R. J. Ball (of Klein and Ball 1959), E. M. Hugh Jones (from Keele), A. J. Brown (from Leeds), and (from Liverpool) G. L. S. Shackle and F. E. Hyde.

Not surprisingly, when the participants began to discuss “the paper on Causes of Movements of Money Wages in recent years (Paper 3),” Ball had something to say:

Stimulated by this remark, Phelps Brown this time decided to follow the econometrician on his ground:

In this way, based on the most advanced econometric analyses (which, as it seems, did not include Phillips 1958), Ball was expressing a critical point that would also recur in the Fourth Report. If full employment and price stability were to coexist, the control of demand was a necessary but not a sufficient condition, and some change of “institutions and habits” (i.e., of wage bargaining) was in order. However, as the following extract from the summary of the discussion confirms, it was still too soon for these choices to be based on econometric studies:

Econometric studies were barely beginning to enter the debate. At Cambridge, during the meeting held on June 8, there was the following exchange between A. Robinson, Kahn, and W. A. B. Reddaway: “Professor Robinson asked what determined the point, in terms of unemployment, at which wage-push inflation started. Professor Kahn said that the Dow and Dicks-Mireaux analysis suggested the degree of trade union pushfulness as one factor that affected this, but Mr. Reddaway thought that one should be very careful in interpreting the Dow-Dicks analysis.”

However, even on this occasion, the discussion was soon brought back onto a more “realistic” (or descriptive) track, in this case by Emmerson, who “said that one very important factor was the personality of the leading figures involved in wage negotiations at different times.”

Along with A. Robinson, Kahn, and Reddaway, the other Cambridge University members who were present were J. E. Meade and J. R. Stone. On Paper 3, Kahn had no doubts. He “said he certainly agreed with the paper's rejection of Professor Phillips's theories, which were based on data relating to cyclical movements, and provided no information about what might happen if a higher rate of unemployment was permanent.” To this, Robinson is reported to have added that “one should also consider how very low the rate of growth would probably be if one had to damp down demand every time unemployment fell below the relatively high level that Professor Phillips's model suggested.”

The questions that emerged at Cambridge, raised by Kahn and A. Robinson, pertained exactly to the stabilization problems that Phillips was investigating in the works that he did not mention in the curve paper and that, at least in the evidence considered in the present reconstruction, never resurfaced in the early debate on Phillips 1958.

5. Concluding Remarks

Eighteen years after the publication of Phillips 1958, and one year after Phillips's death in 1975, Blackaby (1976) observed that, over the prior two decades, the use of quantitative methods in macroeconomic policy had increased dramatically. In the same paper, Blackaby reconstructed the slow but continuous rise of the maximum level of unemployment that was considered tolerable from a political point of view. Indirectly, Blackaby suggests that Phillips 1958 appeared at too early a stage of both processes. The “a little under 2 1/2 per cent” equilibrium rate of unemployment was still too high for the standards of 1958. In the same vein, between 1958 and 1961, the use of inferential statistics and quantitative methods in the policy debate was still in its infancy. Both elements put Phillips at a disadvantage. Nonetheless, combined with the evidence presented in Cristiano and Paesani 2018b, the present article has shown that the curve paper certainly played a significant role in the policy debate.

A further tentative conclusion that can be derived from the reconstruction presented in this article is that the role played by Phillips in the policy debate was perhaps the unintended consequence of a rather simplistic interpretation of his curve that Phillips himself had inadvertently favored. Maybe by mere coincidence, but certainly with an innovative method, Phillips had produced an equation, based on data from a distant period, that was apparently able to predict the rate of change of wages forty years hence. This may have created a kind of “magic formula effect,” something that gave him the feeling that he had put his finger on something but that particularly irritated many of his early readers. Building on Cristiano and Paesani 2018b, the present article has investigated the reactions of these readers and their main consequences, which were to give Phillips a certain amount of visibility in the policy debate while at the same time helping to propagate a rather simplistic interpretation of the curve paper. On the other hand, that Phillips had put his finger on something was probably true. At least, he had indicated a possible direction that empirical research could take. And he had also shown, unintentionally, that there was an increasing demand for empirically based works on policy issues, as Blackaby would later confirm.

If there was an obstacle in the way of empirically based economics, this might have been on the supply side. In this respect, what the council seminars on the whole convey is the existence of a bias against econometrics, a negative attitude that Phillips's (1958) too bold and “crude” statements may certainly have reinforced. In most cases, what attracted attention was either the data employed by Phillips, the conclusions inferred from these data, or both. But almost no one among the economists who took part in the council seminars ever entered into the econometrics of the curve paper. Borrowing from Forder's interpretation, it was as if the “Popperian” element of Phillips 1958, which was certainly to be found in the econometrics of the curve, was looked at through the lens of Robbins's (1932) methodology. The result was that, on the whole, the innovative methodological element that had attracted attention among the LSE youngsters was not even actually rejected but, more simply, bypassed and therefore ignored by the other economists.

The author wishes to thank Massimo Di Matteo, Paolo Paesani, two anonymous referees, and the editor of this journal for their useful comments and suggestions on earlier drafts of the article. The usual disclaimer applies.

Notes

References