The easiest mistake to correct is that Phillips was the first to suggest a negative relation between wages and unemployment. Work in this area started with Hume in 1752, with work intensifying in the first half of the 20th century, notably including Irving Fisher (1926) and Tinbergen (1937), who used work done by the Netherlands Statistical Office and won the Nobel Prize in 1969. A large number of other authors published work in this area in the 1950s such that the idea that Phillips' work was novel in any respect is not remotely credible. It was rushed into print to meet a publishing deadline, and so the extensive comments of at least one referee were ignored.

Nor was it well received, or considered important at the time. Philips himself described it as a “rush job”, “quick and dirty” and dismissed it as a “wet weekend's work”. The econometric techniques (to describe them favourably) were rudimentary, at times ad hoc, and widely criticised. Some authors reviewing his work much later see an ingenuity in his rudimentary approach that suggests they are not familiar with previously standard statistical techniques that were used to make quick estimations (that would not normally make their way into published journals) when computational power was still prohibitively expensive.

[O]f all the surveys and historical treatments… Santomero and Seater's is most notable. Determined that Phillips should be the inspiration of the literature, but unable to see why, they ask why it was that his 'competitors and their insights were ignored at the time'. The simple answer to that is that Phillips' so-called competitors and their insights were not ignored at the time. It was Phillips who was disparaged immediately and later, while others were much more highly regarded. ²⁾

It is puzzling why Lipsey's much stronger paper on the subject (1960) used Phillips' as a starting point, when so much better work had already been done. One explanation for this is the fashionable influence of Popper's work (1959) on Lipsey over the course of the late 1950s. Economics had been struggling with the question of whether any stable, underlying laws were discoverable, or whether economic enquiry would forever be limited to the discovery of short-term stable relationships through econometric work, that would always break down as unconsidered factors shifted over time. Popper's proposal of falsifiability as the sine qua non of science brought this challenge into clearer focus: economics needed to propose bold, powerful hypotheses for underlying laws, not 5-year relationships discovered through data-trawling. The one novelty that Phillips' 1958 paper did supply over previous work was that big idea: he had analysed a century of data and suggested that a relationship might exist that was consistent throughout. That suggestion, rather than any of Phillips' analysis, could understandably have appealed to Lipsey.

Lipsey misrepresents pre-Phillips Keynesianism as suggesting that there was a strict difference in the link between prices and unemployment depending on whether the economy is at full employment or not. He pretends that Phillips work softened this. In fact this is a mischaracterisation of earlier Keynesianism, which had implicitly said that there was no particular link between unemployment and prices regardless of whether the economy is at full employment (price levels being largely contingent on union behaviour, i.e. non-economic factors, for which there existed a wide range of different models).

[I]n the 1950s [there was] barely such a thing as 'the L-shaped aggregate supply curve.' That terminology is almost exclusively a later invention. It dates from the era when the Phillips curve was treated as a standard approach, and looking back on the 1950s the question was asked as to what economists had before they had the Phillips curve. Indeed, the theory of the 1940s and 1950s had no need of a 'curve' of any shape to link inflation and unemployment. They were simply seen as different problems. ³⁾

Phillips' 1958 paper did make an extraordinary claim, but in order to understand it it is necessary to briefly summarise the consensus on wage determination by the late 1950s. It was widely accepted that, while wage rates were limited in the extreme by market forces, the exact level of wages was essentially a non-economic problem: people negotiated, and the outcome of those negotiations rested on a long list of intangibles, concepts like “fairness” that economists had no place discussing. The outcome of negotiations was mainly determined by

1. changes in the cost of living,
2. profitability of the employer, and
3. the relative position of different groups of workers (i.e. maintaining established differentials).

Notably absent from this list was the overall level of unemployment in the economy.

In contrast, Phillips claimed that wages were determined by market forces, and that a century of data proved this to be true. “But Phillips, of course, was wrong.”⁴⁾ His claim was instantly rejected by more or less everyone, and subsequent empirical work failed to support it.

One consequence of [the 1950s consensus] is that it makes it easy to see the case for incomes policy… Later treatments have tended to regard prices and incomes policy as attempts to overrule market forces, and therefore as basically foolish. ⁵⁾

In contrast, in the 1950s prices and incomes policy was seen as a rational approach to influencing the outcome of wage negotiations that were known to determine wages within a wide range permitted by economic forces.

Samuelson and Solow (1960) is an important paper in the fictional history of the Phillips curve: this paper is frequently claimed to be the source of the idea that policymakers could choose from a set of stable combinations of unemployment and inflation, and that the authors advocated an inflationary choice that would reduce unemployment.

However, this is simply not an accurate reflection of the content of the paper. The conference paper, titled Analytical aspects of anti-inflationary policy discussed a wide range of inflationary theory, but when coming to discuss what they call the 'Fundamental Phillips schedule', they say

There may be no such relation for this country.⁶⁾

and

What is most interesting is the strong suggestion that the relation, such as it is, has shifted upward slightly but noticeably in the forties and fifties. ⁷⁾

The most interesting thing about the Phillips curve was that it did not describe a stable relation – it could be moved as a consequence of other factors.

The real content of the paper was to describe a variety of inflation theory and to comment that there was not yet enough data from the post-war period to distinguish between them (i.e. data since the advent of Keynesian policy).

Although later comments made by Solow have been misinterpreted or quoted out of context several times, there is no good evidence that there was a hidden intent in the paper to advocate inflationist policy. Solow rarely cites the paper when discussing similar issues in future, and gives the impression that he didn't see the paper as important.

Contemporary comments on the paper tend to conform to this reading: most cite the paper in support of the idea that the Phillips curve was not a stable phenomenon in America (e.g. Bronfenbrenner and Holzman, Shonfield, and Kuh). The few that seem to have understood the paper as suggesting a stable relationship reject that view. Many citations of the paper ignore its mention of the Phillips curve.

[I]t does not appear that there is any author in that period who could be said to have adopted the idea that inflation is beneficial because of Samuelson and Solow's paper, nor any who could be said to have found in the article inspiration for inflationist policy. ⁸⁾

Over time, the nature of comments on the paper gradually drifted, and a proportion of them became increasingly inaccurate either the true content of the paper, its effect on contemporary economists, or both. However, accurate representations of the paper continued in parallel.

This section is worth reading in full as a detailed illustration of the possibility and consequences of authors (and their journal referees), en masse, referring to historical papers that they either haven't read or haven't understood.

By the early 1970s, more authors are characterising the paper as offering a trade-off, but most of those still reject this imagined proposal.

In the late 1970s, authors such as Frisch, Nobay and Johnson credit Samuelson and Solow with convincing their peers of the feasibility of the trade-off (without mentioning who, in particular, they had convinced).

By the early 1980s, statements about Samuelson and Solow's advocacy of the trade-off and influence of their contemporaries become much clearer – and that their work led directly to inflationary policies. Bordo and Schwartz (1983) broke new ground by not only inverting Samuelson and Solow's conclusion but inventing new country data:

Phillips (1958), Samuelson and Solow (1960), and Lipsey (1960) reported evidence of a stable inverse relationship for the UK, the US, and other countries…⁹⁾

By the mid-1980s this had been repeated often enough that it was frequently assumed to be true in passing comments, and this view of historical fact was repeated through the 1990s.

By the late 1990s another couple of themes had emerged. One was for authors occasionally to read the original article carefully and conclude that it had not proposed inflationary policy, yet nevertheless to revert to the popular wisdom that most 1960s economists had been influenced by the paper in this direction, leading to the conclusion that the economics establishment must have misread Samuelson and Solow's work. Secondly, authors that believed in a stable Phillips curve relationship started to cite Samuelson and Solow positively as early proponents of their ideas.

[A]s time went on, the part of the literature which had come to see Samuelson and Solow as the source of damaging, inflationist policy proposals, that view came to be states more briefly, with the confidence of routine, and sometimes in more exaggerated forms. So James Galbraith (1997) remarked that in 1968 mainstream economists were committed to 'Samuelson and Solow's version' of the Phillips curve. He says nothing about what that was – presumably because he thought it apparent…

Kirshnew (2001) said, 'In the early 1960s, most Keynesian economists argued that governments could reduce unemployment indefinitely if they were willing to tolerate a higher rate of inflation', citing Samuelson and Solow as the only reference in support of this claim. ¹⁰⁾

Samuelson and Solow found no stable Phillips curve, quite the opposite, and did not really advocate inflationist policy.

Most remarkable, surely, is the point that although some have tried to paint things otherwise, Samuelson and Solow certainly did not convince anyone else of the desirability of inflationary policy on the basis of a stable Phillips curve, because before about 1970 there was no one who took such a lesson from them… no one both thought they took an inflationist view and agreed with them…

It is… well into the 1970s, when the idea of an exploitable relation had been raised and widely rejected – that things change. But it is still not the case that anyone takes an inflationist message from Samuelson and Solow. It is just that they come to believe that others previously had. ¹¹⁾

Prior to the impact of Friedman's address, the 90 or so papers that can be considered the “Phillips curve literature” have the following features:

• The “Phillips curve literature” does not begin with Phillips: the 1958 paper has a minimal impact on content and conclusions – whilst more papers were published from the late 1950s onwards, there was strong continuity with wage bargaining ideas discussed earlier
• There is almost no interest in calculating an optimal point on a tradeoff relation
• Factors such as profit, union power, rate of productivity increase and various proxies for demand feature often, and many papers do not give a prominent role to unemployment
• Phillips is often not mentioned, and usually when mentioned it is only to disparage the idea of unemployment being an important factor, or to dismiss the idea of a stable relation

These trends held true for the UK and US, although Canadian studies during the period place more emphasis on unemployment and some estimate Phillips curves.

It is notable that Phillips' influence increases from about 1966, including in papers published after Friedman's address but which have not been influenced by it; however, these papers are generally less thoughtful and of lower quality.

This is so true of some that they are not so much stragglers as perhaps the dying gasp of a literature in final decline. ¹²⁾

Price changes were included in all but a few of the relevant papers of the period and moreover, often significant emphasis is given to price expectations during the bargaining process that was then conceptualised as the principal price-setting mechanism.

Taking the literature as a whole, there would be a few – very few in fact – of which it could be said that the research is seriously flawed by the failure to consider price change…

The suggestion that somehow crept into Friedman's Nobel lecture – that this literature is generally characterised by a confusion of nominal and real variables – is, therefore, simply false, and the fact that it is widely believed is extraordinary. But the discussion of the question most demanding attention is that of Santomero and Seater (1978), for they, with extensive analysis, insist so firmly that the literature before Friedman (1968) was thoroughly senseless. ¹³⁾

Their complaint is not straightforward to discern but appears to claim either that authors should have used expected (rather than actual) inflation (which would have been harder to obtain credible data for and would contradict the authors' belief that negotiators used actual inflation figures), or that authors' incorporation of inflation was not explained to their liking – which was largely because most thought the point too obvious to require any explanation at all. In Santomero and Seater's (1978) words

the reasons given for including [price changes] are usually vague or nonexistent. ¹⁴⁾

The significant papers of the first decade after Phillips are not, on the whole, papers which lack theoretical rational[e] for the form of the equation, and are not critically flawed – or in most cases flawed at all – in their treatment of the difference between nominal and real variables. ¹⁵⁾

A possible source of modern confusion with this literature is that the coefficient on price changes in wage determination is rarely near 1, and often as low as 0.5 (which might appear to imply that real wages would consistently fall under persistent inflation). But there are three important explanations for this:

1. The model of wage determination they were implicitly using was based on bargaining (not market forces), so there was room for this figure to vary from 1 – they therefore calculated this figure by econometric means as part of their work.
2. In contrast to later economists' work (including Friedman), they were not attempting to construct a universal theory of wage determination, but rather to study a particular time and place in order to make near-term predictions, whilst acknowledging that this was the case. “For the modern reader to appreciate this work it is necessary altogether to shake off the idea that anyone thought of a wage equation as one component of an implicit general equilibrium model.” ¹⁶⁾
3. Most of the equations involved other terms, frequently a constant term, through which they were conscious that part of the influence of inflation could be transmitted. It is arguable that this was not optimal econometric technique, but not that they didn't understand the concepts they were dealing with.

The majority of papers during this period are hostile to inflation and do not consider the possibility of tolerating anything less than price stability in exchange for benefits such as higher employment. The general hostility is sufficient to be highly sceptical of any suggestion that authors were implying such conclusions through their silence. There are a minority that do intimate that they would tolerate moderate inflation to reduce unemployment, usually hesitantly, and one (Bodkin et al for the Economic Council of Canada) that goes as far as to calculate steady state equations to inform policymakers' choices (although they express scepticism about its applicability).

So there is Phillips curve inflationism. There is not much of it, but it is there. As Laidler (1997) noted, the first such inflationist was not Phillip (1958), nor Samuelson and Solow (1960), but Reuber (1962). Laidler could have gone further in noting how little followed. Amongst the British wore, there is nothing; amongst the American, there is a very small amount… In what is certainly a point that raises a further question, it could also be said that all the unhesitant, most clear-cut, and least equivocal cases of well-considered Phillips curve inflationism, of the 1960s were Canadian. ¹⁷⁾

Amongst the small minority of authors that contemplated inflation in these studies, most were discussing rates of 2 per cent or less; the highest rate calculated in any of the papers is 3.75 per cent.

Why would economists of the period have contemplated modest levels of inflation? There was a widely accepted argument that (a) there was a constant need for relative prices to shift between industries to ensure that the labour market responded to those growing and in decline, (b) wages were known to be downward-sticky, so that requiring downward adjustment would lead to real losses in output and employment, and (c) modest inflation (of the order of 1-2 per cent) could enable inter-industry adjustments of this type to happen more quickly, without downward adjustment. This argument is “astonishingly” evident in the literature,¹⁸⁾ such that it is inconceivable that it was not in the minds of those making these calculations.

On conventional accounts, before Friedman (1968) there developed a large body of work derived from Phillips, with far too little understanding of the importance of price change, and far too much inclination to pursue inflation as a means of lowering unemployment… [T]he reality is that we have rather little of any note derived from Phillips; a fine understanding of the role of price change; far more authors presuming price stability to be an essential than were inflationist; amongst the minority, usually only a very mild inflationism; and finally a good argument that there should have been much more of it. ¹⁹⁾

According to the traditional presentation of the Phillips curve history, Phelps and Friedman first presented a novel insight in the late 1960s that changed macroeconomics: the idea that inflation quickly leads the public to adjust their expectations of future prices, leading workers to demand higher wages, with the result that an increase in inflation can (at best) only temporarily increase employment until it is anticipated to continue.

In fact this is not the least bit original. Dozens of authors had presented the same idea in the preceding decades (even stretching back to Mill in 1844). It is usually presented as a commonplace observation. There is plenty of evidence that this was well understood and applied in practice by all unions. Many of the initial reactions to Friedman and Phelps indicate that this idea was recognised as being unoriginal, nor is it presented as original by Phelps. Moreover, both authors had themselves mentioned it previously, in earlier papers and less auspicious locations.

Significant changes occur in the literature around 1970, but they do not revolve around incorporating Friedman's supposed “insight”. The main changes were

• The expression “Phillips curve” was used to refer to relationships between unemployment and either prices or wages, as Friedman and Phelps had done.
• The general quality of the literature recovers after the late 1960s trough, because the Brookings Papers on Economic Activity appear, and the emergence of significant inflation alongside high unemployment in the US motivates more work.

This improved work does not revolve around expectations, however. Where the Phillips curve relation is discussed, it is generally in the context of using high unemployment to stabilise prices rather than the obverse. Many papers reference Friedman, but few put his expectations argument to the test, or defend the Phillips curve.

Just as it is not the case that the 1960s literature revolves around estimates of the tradeofff, so it is that in the 1970s it does not revolve around assessing Friedman's argument, whether it be thought new or old. ²⁰⁾

In the UK literature, reference to the Phillips curve increases markedly after 1970, and for the first time it is fair to say that Britain had an independent literature on the subject. Similarly to the US, there was some discussion of the “disappearance” of the curve, dated in 1966 or '67, but the overwhelming focus is on the effectiveness (or not) of incomes policy and regional policy.

An important reason that these papers are significant… is that… they go a long way to explaining the impression that the Phillips curve was at the heart of policy debate in the United Kingdom. Indeed, in a sense it was, although the debate was not about the effects of inflationary policy, or the optimal tradeoff point, or anything of that kind. ²¹⁾

A number of studies did test Friedman's hypothesis that rational expectations ensure that the Phillips curve is vertical. Results are mixed. Many econometric studies find that while inflation affects wage increases, at moderate levels of inflation the influence is less than perfect so that inflation does have real effects. Consciousness of the lubrication argument is generally evident or can be inferred in these cases. Other studies support Friedman's view. The prescriptive bias is still towards raising unemployment to reduce inflation, but it's fair to say that some authors are a little softer in tolerance of inflation than in the 1960s.

Friedman and Phelps cannot possibly have revolutionised the understanding of the long-term effects of inflation. Phelps even said it was an old idea, and Friedman as good as did the same by the way he presented it. Most of the literature, though, was not about responding to that idea at all, but about addressing the actual problems of the time – including analysing the consequences of sustained inflation. ²²⁾

There are serious difficulties in trying to summarise the 'general view' of inflation when selective quotation can do so much to achieve the desired results. Nevertheless, anti-inflationism dominated the wider literature beyond econometrics. The major international reports of the period (from the UN OECD) add credibility to this view. Although it is worth noting that “price stability” is normally taken to mean inflation of around 1-2 per cent, inflation proper being somewhat higher.²³⁾

Demand-pull inflation is inflation that results from excess aggregate demand; cost-push inflation results from cost increases which occur despite the fact that aggregate demand is not excessive. ²⁴⁾

Friedman (1966) argued that cost-push inflation was impossible, because unions and monopolies could force only a one-off increase in prices, not a continual process. This was widely criticised because the idea was not that a single union could cause the increase, it was that in a market where the maintenance of established differentials was central to perceived fairness, a collection of unions could repeatedly raise wages. A further response to this was the argument that restricting demand could always restrain inflation. This was not accepted by everyone, but even if true, the challenge of cost-push inflation was precisely the aim of achieving full employment and price stability together.

The primary issue was not whether it was possible for the policymaker to stop the inflation, but whether it is possible for the cost-push forces to exist so as to create for the policymaker the choice between inflation and unemployment. ²⁵⁾

Various complications with cost-push inflation were discussed.

1. It is hard to practically distinguish cost-push from demand-pull,
2. Shortages in a single market fall somewhere between the two: they're due to excessive demand but not generally excessive demand (sometimes called 'bottleneck' inflation),
3. It could be caused by the power of unions or oligopolistic firms (steel in the US gaining special mention), or both,
4. The 1970s introduced the sometime importance of commodity price booms, which required a different response,
5. It may not be possible to control cost-push using demand:
   1. Unemployment increases public sympathy for labour, strengthening unions' ability to strike,
   2. Contraction reduces real incomes, such that if militancy is a consequence of unmet expectations of real wage level, contraction could fuel militancy,
   3. Monopolistic firms may raise prices to protect their incomes in the face of falling demand.

These arguments are all “volitional” (rather than mechanical) and stem from the pre-game theory “bargaining theories”, and fairness was paramount.

[T]o recall a crucial point: it is not that analysis drifted into normative territory by extolling fair wages. It is that the authors believed that, as a matter of scientific truth, ideas of such things were causally important. ²⁶⁾

The general view of the 1960s was that even if restricting demand could control cost-push inflation, there were more efficient ways to do it. “Full employment and price stability could both be targets. In that sense, no tradeoff was perceived.” ²⁷⁾

There was a significant debate until the 1970s about whether a causal link existed between growth and inflation. Theoretical arguments were presented on both sides, and the empirical work remained inconclusive. Interest may have petered out partly because so many studies were inconclusive.

What is interesting, however, is not the outcomes of these debates, but that they occurred at all and were perfectly serious. And their relation to the issue of the Phillips curve should be clear: there was none. Yet it is here, if one were looking, that there would be [a] good case for saying that there was a genuine debate about a tradeoff involving inflation. ²⁸⁾

There were various sound reasons that “price stability” was generally interpreted to mean low, positive inflation rather than continuous zero inflation, and these were widely understood in the 1960s and 1970s (many of which need to be coupled with the idea of downward-sticky prices):

1. Multiple indices (retail prices, wholesale prices etc) diverged and there were no obvious grounds for selecting any of them.
2. The price level at full employment is higher than at lower levels because of lower marginal productivity.
3. Inflation provides a convenient way of reducing the real value of downward-sticky fixed claims (e.g. the minimum wage).
4. For statistical reasons, most economists agreed there was an upward bias in the price index.
5. If the most productive firms award wage increases based on productivity, comparison between firms and sectors would be expected to result in mild inflation.
6. Unions bargain individually, with each having a negligible impact on inflation; they therefore face a prisoner's dilemma.

[A]lmost anyone could accept that this family of arguments gave reason to be relaxed about mild inflation. ²⁹⁾

Having established “the proper context” for remarks relating to inflation, many authors' comments are analyzed, especially relating to incomes and manpower policy.³⁰⁾ In general their advocacy of or tolerance for higher inflation is found to be much weaker than depicted in more recent writing, and its connection to the Phillips curve (at least as a stable, exploitable relationship) weaker.

It's also worth mentioning that there is evidence in the 1960s that economists believed anti-inflation sentiment had hardened since the 1950s. The most explicit statement of this is in Samuelson's textbook, Economics, which in 1948 “said that 5 per cent steady inflation is not too serious”³¹⁾ – by 1955 this had fallen to 3 per cent and, by 1961, 2 per cent.

Much of this writing was in less formal contexts (lectures, journalism, representations to policymakers, books) and can be seen as an effort to “get the word out” to non-economists rather than arguments for a technical audience. The message, such as it existed, should be interpreted as

a counsel of calm – an acceptance that not everything could be perfect, and that policy, therefore, should not be set to react to every drift of the price index as if it were the only priority.³²⁾

There are a few minor comments to be made about discussion of the curve beyond econometrics, although “as regards the 1960s, the most important of these is just to say how little was said.”³³⁾ Phelps (1970) mentions that he and Friedman are trying to introduce the Phillips curve into established thinking, which he acknowledges saw money as neutral in the long term. The Phillips curve begins to be discussed in the context of the political business cycle – the idea that governments (especially in the UK in 1955 and 1959) can please voters by selecting a favourable short-run combination of unemployment and inflation just before elections that is not sustainable. The Phillips curve also featured in discussions of European Monetary Union³⁴⁾ In the process the concept of the Phillips curve changed again, becoming a simple relation with all of the variables included in the econometric studies stripped away.

Finally, there is a notable uptick in harsh criticism of the paper after the publication of Friedman (1968). Having been dismissed on publication and ignored in the intervening period, Friedman seemingly inspired more authors to dismiss it again (or reassert that “more perceptive economists” had dismissed it to start with).

1. Most of what was written about inflation was perfectly reasonable, including work on cost-push inflation.
2. There were good reasons to be relaxed about low inflation.
3. Even those authors who can be cherry-picked as favouring inflation understood real and nominal variables.
4. The change in attention to the Phillips curve is the opposite of the conventional story: in the 1970s the curve is both discussed and applied to policy more than prior to Friedman (1968).
5. The whole discussion is notably similar through the period, changing only due to events rather than important theoretical leaps.

There is… nothing in the policy record to suggest that 'the Phillips curve', or 'the tradeoff', or the same thing by any other name had any role in motivating or influencing policy. ³⁵⁾

Given that the academic literature had provided a range of viable justifications for modest inflation (see 5.4), it is somewhat surprising that policymakers remained so committed to price stability throughout the 1960s and 1970s.³⁶⁾

In the UK, the history of Chancellors' speeches emphasize the constant tension between unemployment (demand) and the balance of payments deficit. Costs and prices were controlled quite separately (in the minds of the policymakers) by incomes policy, which was uniformly perceived as a means of obtaining the prices and wages that would prevail under perfect competition (rather than an attempt to 'subvert the market').

Both the Nixon and Heath governments stated their intent to reduce state management of the economy, but when that led to inflation that was not constrained by high unemployment, both backtracked with incomes policy and reflation.

Not only was there no reason to describe policy in terms of the Phillips curve, almost no contemporary or near contemporary commentators did so.

One particular speech from the Labour Party Conference in 1976 was later used by Friedman (1977) to suggest otherwise:

We used to think that you could just spend your way out of a recession and increase employment by cutting taxes and boosting Government spending. I tell you, in all candour, that that option no longer exists, and that in so far as it ever did exist, it only worked on each occasion since the war by injecting a bigger dose of inflation into the economy followed by higher levels of unemployment as the next step.³⁷⁾

This section of the speech was written by Jay, and the intent seems to be consistent with the general thrust of his work at that time: that UK policy had veered too violently between expansion and contraction, leading ultimately to increased levels of inflation and unemployment. He blamed this on “irresponsible trade unionism and cost-push inflation”.³⁸⁾

Also notable are the contents (or lack thereof) of Friedman (1968). He doesn't claim the Phillips curve was ever the basis of policy, and doesn't say:

• that the curve or and understanding or misunderstanding of it was thought to offer a menu, or
• that it in fact guided policy, or
• was even used by policymakers to describe the results or intentions of their actions.³⁹⁾

The first antecedents of the Phillips curve myth can be seen emerging at the end of the 1960s, but it is fully established in the literature with “remarkable” suddenness between 1975 and 1977. From there it made the leap to textbooks by 1978, including to many that had previously contained accurate accounts. Even more oddly, alongside the new assertion that economists had believed in an obviously flawed model for a decade,

the lubrication argument disappears, even when it had been present in earlier editions. This gives the impression that the supposed inflationism had no explanation, except a failure to understand the expectations argument, but it also seems to show those authors as being overly anxious to display errors in the thinking of the past.⁴⁰⁾

Over the next few years it was cited routinely by a range of authors of “high standing” without citation, collectively giving the impression that it was universally accepted as fact. One of many examples is the opening lines of Feldstein (1979):

It was not so long ago that most economists regarded the Phillips curve as a stable menu of policy options. A permanent reduction in unemployment appeared to be possible if the nation were willing to pay the price if a permanently higher rate of inflation. Even rather pessimistic estimates of the slope of the Phillips curve suggested that the price was well worth paying.⁴¹⁾

In addition to the large number without any citation, a number cited a small range of sources that have already been discussed here: usually Samuelson and Solow (1960), less often the Council of Economic Affairs (1969) and the Callaghan party conference speech in 1976.

Throughout, and to the present, factually accurate accounts do persist, though in a tiny minority, among them work by Cairncross (1992), Perry (1976, 198) and Ulman (1998). Even more rarely do authors rebut some part of the mainstream narrative, such as in Nelson (2005, 2009) and Beggs (2010). “But such authors are very much the exceptions.”⁴²⁾

[T]here are occasional attempts to substantiate [the myth], but they are so flimsy [that] that fact itself emphasizes that the argument is wanting.⁴³⁾

The question that remains, then, is how it could be that a fiction, so suddenly introduced, could become established so quickly, with scarcely any protest being made.⁴⁴⁾

The “famous papers”, (meaning Phillips (1958), Samuelson and Solow (1960), Phelps (1968), Friedman (1968) and to a lesser extent Lipsey (1960)) were atypical of the wider literature:

1. They didn't consider price change (for justifiable reasons, but this was highly unusual)
2. Most used “Phillips curve” terminology to describe a price-change relation (unusual in the 1960s)
3. Samuelson and Solow used the out-of-place term “menu of choice”⁴⁵⁾

The proliferation of uses of the term (which became a “generonym, with… no special link to the original”⁴⁶⁾) and the division between econometric work where it was usually a complex relation and theoretical work that used the same term for “the simplest possible relation”⁴⁷⁾ could also have fostered confusion, especially with later readers. Moreover, the breadth that this church achieved came to construct an illusion of consensus: with everybody writing about their own “Phillips curve” (even if they were talking about different proposals with different theoretical underpinnings), there appeared to be nobody on the other side. The confused use of modifiers like “short-run”, “naive” and “simple” compounded matters. Unfortunately, certain inferences in Friedman (1977) seem to reinforce possible misreadings of earlier papers.

As time passed, the origin of three 'insights' came to be routinely misreported:

1. The “Phillips curve” came to be incorrectly attributed to Phillips (1958).
2. The expectations argument came to be attributed to Friedman (1968), implying it hadn't been routinely incorporated into econometric work in the 1960s.
3. The lubrication argument came to be attributed to Tobin.

To begin with, these attributions were quite innocent – these ideas were all so commonplace that nobody thought them worthy of attributing to an 'inventor', so they were attached to the name of a recent discussant for convenience.⁴⁸⁾ Only later readers came to form the impression that such straightforward ideas could have been unknown to authors prior to these respective papers.

A lack of quality in the responses to Friedman's arguments certainly enabled a sudden drift in the literature, as well-established ideas were attacked without response.

1. The lubrication argument was poorly defended (by Tobin and Ross) in a New York Times article that mis-paraphrased it as requiring “that you can fool all of the people all of the time”.⁴⁹⁾
2. Friedman challenged his peers to provide a sturdy theoretical model of cost-push inflation – critical to much of the foregoing work – and nobody did.
3. The expectations argument was refuted again and again in subsequent econometric work, but the authors “all allow the econometrics to speak for itself in showing Friedman to be wrong, giving no explanation of what economic behaviour might underlie such a result.”⁵⁰⁾
4. In presenting his thesis, Friedman (1968) simply assumes that the economy is at the natural rate of unemployment, whereas the accepted Keynesian methods for using demand to reduce unemployment without raising inflation applied when this was not the case. The “amazing”⁵¹⁾ lack of response to Friedman on this point allowed the actual techniques of demand management to be overshadowed by a straw man.

In the event, inflation rose, due to the way in which the Vietnam War was financed by the US. This has sometimes been seen as a vindication of Friedman, though he never predicted it – his arguments were all hypothetical. More importantly, the size and exclusively demand-pull nature of this inflation forced consideration of previously important concepts (cost-push mechanisms, lubrication, incomes policy) out of the limelight – and it became harder for 1970s economists to appreciate their importance in earlier discussion. The oddly sudden emergence of the myth meant that authors of papers just before it emerged could not anticipate it nor defend their work from it, further confusing subsequent readers that could not imagine good reasons for not taking the time to refute the myth. On the econometric side, there was a discontinuity between the 1960s and 70s, with only Perry doing significant work across both – and standing out as later providing a rare source of accurate history.

[T]here is surely relevance in the point that dismissing earlier work is often easier than studying it. Johnson said Keynesianism was just difficult enough since it allowed the young to leave their elders behind without having to read what they had written.⁵²⁾

To summarise the 'new story' of the Phillips curve:

• In 1958, it was well understood that wage bargaining was influenced by ideas of fairness, so factors such as profit, inflation, productivity and other workers' situations were studied.
• Market forces were not ignored in setting limits to the bargain.
• The idea of a relation between wages and employment was commonplace but uninteresting and of lesser importance.
• Therefore, full employment with price stability was the natural objective.
• Cost-push inflation was possible, so unions could create inflation without full employment, although the perception that this was unfair could be used to limit its effects.
• The “Phillips curve” was used for four different things in the 1960s:
  1. As used by Phillips, to challenge conventional theory by suggesting that factors other than unemployment (and its rate of change) were irrelevant to wage determination.
  2. The curve could be used to subject traditional ideas to econometric analysis (as computing power made ever more possible). The descriptor of “Phillips curve” in this case has been mostly retrospective.
  3. Investigation of whether price stability and full employment really were compatible after all (and if not, how incompatible).
  4. Some time later, to test the effect of incomes policy, regional policy, and union bargaining (the basic Phillips curve relation being assumed in the process).
• In the 1970s, the “Phillips curve” could again be selected as the category under which various different lines of enquiry could be pursued (and the term was used more frequently than before, often very informally):
  1. To investigate the causes of the new inflation (including expectations).
  2. Theoretical work asking why the relation should exist, and giving answers such as lubrication and search theories.
  3. Theoretical work on the political business cycle and optimal currency areas.
• Finally, the concepts of the “short-run” and “long-run” Phillips curves emerge.

[T]he idea of the 'short-run' Phillips curve makes no appearance at all until after Friedman (1968). It is a peculiar consequence, but whereas Friedman (and Phelps) are believed to have killed off the idea of a long-run exploitable relationship, it is nearer to the truth to say they invented the idea of a short-run exploitable relationship.⁵³⁾

Further, the myth now lies at the heart of many important contemporary debates, such that it is impossible to understand any of the following without a corrected understanding of the history:

• demand management
• full employment policy
• incomes policy
• stagflation
• fixed exchange rates
• central bank independence, or
• the creation of the EMU.

Finally, the existence of the short-run Phillips curve is now taken for granted, but even today there is no reliable evidence that places its existence beyond question – “nor even a reliable measure of the price-stability measure of unemployment.”⁵⁴⁾