I. INTRODUCTION
It has been accepted as a fact of history that, because of some fundamental errors of theory, much of the economics profession was once persuaded of the desirability of lowering unemployment by inflationary policy or, as it is sometimes put, “exploiting the Phillips curve.” The acceptance of this unflattering view is attributable more than anything to the account of the development of post-war macroeconomics by Milton Friedman (Reference Friedman1977)—that author’s Nobel Lecture. There, he went well beyond what he had said in Friedman (Reference Friedman1968)—his renowned Presidential Address to the American Economic Association. Whereas the earlier presentation developed the idea of the natural rate of unemployment and offered an expectations-based argument against attempts to lower unemployment by demand management, the Phillips curve, per se, played only a minor role. But in the Nobel Lecture, as in one slightly earlier presentation (Friedman Reference Friedman1975), he specifically offered an analysis and criticism of what he took to be the erstwhile consensus, and made the Phillips curve and its misinterpretation the centerpieces of his account of how it was that futile attempts to maintain low unemployment resulted in the rise of inflation in the 1960s and 1970s.
It might very well be argued that no policy of exploiting the Phillips curve in fact caused this inflation,Footnote 1 and it could also be said that, contrary to the impression given by parts of Friedman’s lecture, the Phillips curve has nothing to do with “Keynesian economics,” properly understood.Footnote 2 My objective, however, is to question the picture Friedman painted of the attitude of the economics profession generally to the Phillips curve and what it was taken to imply for policy. Part of his picture, considered in Part II, is that the profession welcomed A.W.H. Phillips’ research as a valuable addition to their apparatus. In fact, the profession at large was skeptical of Phillips’ findings, and some of the most prominent individuals were distinctly hostile to them. Another part suggests that there was a naïve theoretical understanding of the relationship depicted in the curve, and this failed to recognize the importance of real rather than nominal variables. But as I show in Part III, almost no one made the mistakes alleged by Friedman to be at the heart of the problem. And most importantly, of course, in Friedman’s picture, this enthusiasm and credulity led to the advocacy of inflation as the means of achieving full employment, until Friedman’s argument on expectations was accepted. Although, as I note in Part IV, inflation advocacy did occur, it was rather rare; usually restricted to those with little influence, just as often before Phillips as after him; and in other cases based on a far more effective theoretical argument than Friedman seems to contemplate. In the light of these arguments, I consider, finally, the lasting legacy of Friedman’s description of intellectual history.
II. THE RECEPTION OF THE PHILLIPS CURVE
A “Stable Relation”?
Friedman said that “professional analysis” had gone through two stages, of which the first was “the acceptance of a hypothesis … that there is a stable negative relation between the level of unemployment and the rate of change of wages—high levels of unemployment being accompanied by falling wages, low levels by rising wages,” and that it “was widely interpreted as a causal relation that offered a stable trade-off to policymakers” (1977, p. 454).
The emphasis on the “stability” of the relation is crucial, but Friedman’s treatment reveals a basic misunderstanding. Phillips’ intent had nothing to do with stability under a policy of inflation. His claim was that the relation had been stable for a long time. He estimated it for the period 1861–1913 and then noted (1958, pp. 293–295 and pp. 297–298) that later data points up to 1957 were either very close to the curve or could be explained away. This he took to suggest that the same underlying relation survived in the later period. He had no data that would have revealed the effects of sustained inflation, and made no reference to the possibility.
The significance of Phillips’ finding is that it amounts to a claim that the relation is independent of any of the extensive institutional or political changes that took place over the period. The pertinent consequence was that the Phillips curve seemed to offer a simple but powerful account of inflation. Acceptance of the analysis amounted to acceptance of the view that price stability could be achieved—but could only be achieved—by an appropriately high level of unemployment. The importance of “stability” was not that there was a “trade-off,” but—quite the contrary—that there was only one way to stop inflation.Footnote 3
A Keynesian Theory of Inflation?
This point leads to a second comment on Friedman. He interpreted Phillips’ finding as being welcome to Keynesians, whereas in fact it was most unwelcome. Friedman said that the Phillips curve “filled a gap in Keynes’s theoretical structure. It seemed to be the ‘one equation’ that Keynes himself had said ‘we are … short’” (1977, p. 469).
The use of this particular quotation is remarkable. It comes from the appendix to chapter 19 of J. M. Keynes (Reference Keynes1936) and is part of the discussion of the limitations of A. C. Pigou (Reference Pigou1933). It is Pigou, not Keynes, who is being said to be short the equation. In any case, Friedman’s implication that the Keynesians were lacking a theory of inflation is quite wrong.Footnote 4 The theory from Keynes (Reference Keynes1940) was a theory of excess-demand inflation. That could perhaps be seen as giving rise—under considerable simplification—to the theory of the “L-shaped” supply curve by which aggregate supply was perfectly elastic until full employment was reached and perfectly inelastic thereafter.Footnote 5 This suggested that good, Keynesian, management, by targeting full employment precisely, could deliver both it and price stability.
Such was the basic outlook, for example of the thoroughly Keynesian Clark, Smithies, Kaldor, et al. (Reference Clark, Smithies, Kaldor, Uri and Ronald Walker1949, see especially pp. 43–46), writing for the United Nations. They, like many others, recognized what they took to be a subsidiary point, that inflation might, in fact, begin to rise before full employment was reached because of such things as supply-side bottlenecks and the development of pockets of monopoly power. Naturally, as low but persistent inflation continued throughout the 1950s, these “subsidiary” considerations attracted more attention and the outlook developed into the theory of cost-push inflation. A crucial aspect of that—argued powerfully from a Keynesian perspective by, for example, Thomas Balogh— (Reference Balogh1958) was that it held that cost developments were not, for practical purposes, controllable by the restriction of demand, and such restriction would, therefore, be ineffective in reducing inflation.Footnote 6
This had two consequences that are important for the current discussion. One is that those who took this view of inflation were unlikely to accept the Phillips curve, since it purported to show that the level of demand was the crucial, persisting determinant of inflation.Footnote 7 The second was that, from the Keynesian perspective, the problems of inflation and unemployment were essentially separate. Employment was determined by demand, while the tendency of unions to raise wages when employment was high could be controlled by measures such as direct price control, moral suasion, incomes policy, and perhaps even increasing the degree of product market competition and openness to international trade. On orthodox theory of the time, subject to the effectiveness of these measures, there was no impediment to achieving full employment and price stability.Footnote 8 Phillips’ analysis threatened this Elysian vision.
Adopted “with Alacrity”?
These points make it no surprise that it is hard to see much justification for Friedman’s further claim that the hypothesis of “a stable relation between the level of unemployment and the rate of inflation was adopted by the economics profession with alacrity” (1977, p. 469). In fact, the general tone of responses to Phillips’ work was rather hostile.Footnote 9 Some simply did not believe it. Joan Robinson (Reference Robinson1937, pp. 30–31) was one of the first to warn of a general tendency for full employment to promote rapid wage increases, but she never accepted the stability of a particular relation, and later she was happy to dismiss the “late-lamented Philips curve” (Robinson Reference Robinson and Robinson1973, p. 8).
Others produced analysis to dispute Phillips’ findings. Among them was Keith Griffin (Reference Griffin1962, p. 381), who found that the data showed that “The ‘Phillips curves’ shift systematically as percentage unemployment is reduced.” Guy Routh argued “his statistical material is not appropriate—different wage and unemployment series will yield significantly different results; and that, even if the imperfections of the material are disregarded, Professor Phillips reads more into his evidence than it can in fact sustain” (1959, p. 299).
One considerable group of Phillips curve critics believed that different variables were better predictors of wage change than were unemployment and changes in unemployment. For example, A.G. Hines (Reference Hines1964) found changes in trade-union membership to be more important than unemployment; Edwin Kuh (Reference Kuh1967) also found unemployment not to be a powerful predictor of wage changes and suggested that productivity changes performed better.
Others, such as Rattan Bhatia, tried to construct Phillips-type relations for the United States. He found:
The influence of the changing rates of unemployment is not clearly marked.… Although a relationship appears to exist between unemployment and the rate of change of money earnings in the period 1900–42, this relationship has varied from one period to another … from 1932 to 1942 earnings rose continuously despite the prevalence of high levels of unemployment (1961, p. 296).
The reaction of Nicholas Kaldor (Reference Kaldor1959a) perhaps shows most clearly the state of Keynesian thinking at the time. He gave a detailed and immediate response to Phillips, arguing that even if unemployment were raised, this would not prevent wage-push inflation. Wage increases were, he felt, determined more by profits than unemployment, so demand policy would control inflation only if it prevented profit growth, and this, in turn, could certainly not be done without compromising the growth of output.Footnote 10 Consequently, he said, “I have dwelt on Professor Phillips’ views at some length, simply because a misguided diagnosis of the causes of wage inflations can play such a vital role in policy decisions which are of the utmost importance to the economy” (p. 295).
This general approach to the problem was commonplace both at the end of the 1950s and later. Many economists, certainly including those who had most influence over policy, rejected the view that inflation could be controlled by raising unemployment and, with it, they rejected the Phillips curve.
A further indicator of how untenable is the supposition that the Phillips curve quickly came to dominate thinking is the number of authors who say little or nothing about it, even in the context of policy discussion. In Richard Caves (1968), only David Smith (Reference Smith, Caves and Krause1968) discusses the Phillips curve at any length and he comes to no particular conclusions about it. Wilfred Beckerman (Reference Beckerman1972), which aims to assess the Labour government of 1964–1970, is surely a place one would expect to find discussion of the curve if it had been important to policy makers. In fact, the book contains only one mention of it, in a footnote saying the curve receives too much attention. J.C.R. Dow (Reference Dow1964) hardly refers to the curve, and Herbert Stein (Reference Stein1969) says nothing at all about it. But how could it be, on Friedman’s account, that it played no role in “The Fiscal Revolution in America”?
It would, of course, be wrong to suggest that Phillips attracted no supporters—Lipsey (Reference Lipsey1960) is the outstanding early example that proves otherwise. But the view that anything resembling Phillips’ formulation was either quickly or widely accepted—particularly that it was quickly and widely accepted by those most concerned with policy—is seriously mistaken. It was not a convenient finding for the orthodox of the day, it did not fill any gap in their theory, and it was not accepted with alacrity, or even, in many cases, ever.
III. EXPLAINING AND EXPLOITING THE INFLATION–UNEMPLOYMENT RELATION
The idea that the point of interest in Phillips’ research was its stability in the face of institutional change was soon lost. Hardly anyone showed much interest in estimating a single relation for long periods of time and those who did failed to find one. “The Phillips curve” became the label for estimated relations between inflation and unemployment, along with, invariably, other variables. Estimates of that kind dated back to Irving Fisher (Reference Fisher1926) and Netherlands Central Statistical Office (1933).Footnote 11 Researchers of the topic among Phillips’ contemporaries included Lawrence Klein and Arthur Goldberger (Reference Klein and Goldberger1955) and L.A. Dicks-Mireaux and Dow (Reference Dicks-Mireaux and Dow1959), to name only the most notable of the econometricians concerned. But Paul Samuelson and Robert Solow (Reference Samuelson and Solow1960) called the relation the “Phillips curve.” From then on, and in that sense, there was indeed a great deal of attention focused on “the Phillips curve,” but still not in the way that Friedman suggested.
Real and Nominal Variables
While summarizing his story of the Phillips curve, Friedman said Keynesian economics had had the effect that “the age-old confusion between absolute prices and relative prices gained a new lease on life” so that in considering the relation of wages and employment, “economists would analyze the relation between unemployment and nominal rather than real wages and would implicitly regard changes in anticipated nominal wages as equal to changes in anticipated real wages” (1977, p. 469).
Leaving aside, temporarily, the question of anticipations, on the treatment of the distinction between real and nominal variables, Friedman is wide of the mark. Admittedly, Phillips’ own treatment of this point now seems inadequate. He noted that cost of living increases might affect the wage bargain but modelled them as doing so only on occasions when that increase was greater than the increase in wages that would otherwise be indicated by the level of unemployment. K.G.J.C. Knowles and C.B. Winsten—more critics of Phillips—quickly pointed out the problem and said that his neglect of the “recent tendencies” of prices gave him “the handicap of appearing something of an anachronism” (1959, p. 120). Even Phillips, however, shows no sign of failing to understand the distinction between real and nominal variables.
In any case, it is extremely difficult to find others among the large number of people estimating “Phillips curves” who make any such mistake. On the contrary, it is thoroughly standard practice to include change in the price index as a variable explaining wages. If one looks to Phillips’ immediate successors, Lipsey (Reference Lipsey1960), Dicks-Mireaux and Dow (Reference Dicks-Mireaux and Dow1959), Klein and R.J. Ball (Reference Klein and Ball1959), and Dicks-Mireaux (Reference Dicks-Mireaux1961) all consider them. Even among his predecessors, Tinbergen (Reference Tinbergen1951) and Klein and Goldberger (Reference Klein and Goldberger1955) included them, and Klein and Ronald Bodkin (1964) evidently thought the inclusion of price changes in the wage equation perfectly routine.Footnote 12 Furthermore, both Klein and Goldberger, and Klein and Ball observe that although their estimations relate to nominal wages, their inclusion of prices as an explanatory variable removes any material concern arising from the theoretical proposition that bargains should be struck in real terms.
As time went on, and prices began to rise more quickly, the econometricians certainly did not cease including prices in their equations. John Vanderkamp (Reference Vanderkamp1966) is an example where the author thought the view that there was no money illusion was debatable but nevertheless thought that price change should certainly be included in the wage equation. And when Friedman (Reference Friedman1968) shifted attention onto expectations, almost everyone involved included some measure of expected price change in their equation. Clearly, it was never, in the research literature, presumed that bargains were struck without regard to real values.
It is true that in most Phillips curve estimates up to the mid-1970s, the coefficients on price changes or expected inflation were less than 1, meaning that the adjustment of wages to prices was, apparently, less than complete. That may have surprised or disappointed the monetarists but it is merely a fact about what the data said, not an elementary mistake by the analysts.Footnote 13 The fundamental point that, in one way or another, one must account for price changes was accepted by all the principal analysts after Phillips and almost all before him.
The Expectations Issue and the Theory of the Phillips Curve
A second dimension to Friedman’s argument on this point is the implication that, as a result of lack of attention to expectations, there was no viable theory of the Phillips curve. The argument—by convention variously attributed to Friedman (Reference Friedman, Schultz and Aliber1966), Edmund Phelps (Reference Phelps1967), or Friedman (Reference Friedman1968)—was that wage bargains would be struck in the light of expected inflation so that reductions in unemployment resulting from inflation would persist only for as long as it took expectations to adjust to reality.Footnote 14 This led Friedman to say, “There is no stable trade-off between inflation and unemployment; there is a ‘natural rate of unemployment’ (UN) which is consistent with the real forces and with accurate perceptions; unemployment can be kept below that level only by an accelerating inflation” (1977, p. 458).
The implication—made perspicuous by the world “only”—is that once the importance of expectations is accepted, a long-run Phillips curve could not be rationalized.
A number of theoretical rationales for the Phillips curve are vulnerable to this kind of argument. Phillips himself followed Samuelson (1948) and Bent Hansen (Reference Hansen1951) in treating the speed of price adjustment as determined by the degree of excess demand. It seems reasonable to suppose that that relationship would change in a context of persistent inflation.
But a different line of thinking was deployed, for example by Tobin (Reference Tobin1972b). He suggested that at any point in time, some labor markets will be in excess demand and others in excess supply. If nominal wages rise more readily in markets in excess demand than they fall in those in excess supply, there will be some tendency to an increase in average wages, even when the sum of excess supplies is slightly greater than the sum of excess demands. A high enough level of unemployment will force wages down in the excess-supply sectors, but the achievement of wage stability requires unemployment. There is, then, clearly a trade-off at least at low rates of average wage increase. In this kind of case, as it is sometimes put, inflation “lubricates” the labor market.Footnote 15
This argument had been made many times before 1972, and indeed Friedman (Reference Friedman1958, pp. 251–252) had considered a version of it. There, he suggested that persistent inflation would transform downward nominal rigidity into downward real rigidity. In other words, even in labor markets in excess supply, nominal wages would rise. Whether this is correct clearly turns on what it is that explains the original nominal rigidity, and possibilities abound. Some may explain specifically nominal, but not real, rigidity. In that case the expectations argument does not show there can be no long-run trade-off. Versions of the lubrication argument continued to be presented after Friedman’s Nobel Lecture—for example, by William Baumol (Reference Baumol, Bergstrom, Catt, Peston and Silverstone1978)—and it forms a component of the argument of George Akerlof, William Dickens, and George Perry (Reference Akerlof, Dickens and Perry1996).
Therefore, not only were all the principal estimators of Phillips-type relations clearly conscious of the importance of changes in the price level, but the theorists of the Phillips curve also had readily available an account that responded to the question of expectations, and suggested the possibility of a persistent trade-off.
IV. THE ADVOCACY OF INFLATION
Following up his assertion that the Phillips curve had been understood to offer a stable trade-off, Friedman suggested that this mistake underlay a policy of seeking to reduce unemployment by means of inflation. He said, “in the circumstances of the post-World War II period, when governments everywhere were seeking to promote ‘full employment’, [the rate of inflation consistent with] it tended in any one country to rise over time” (1977, p. 455).
By now it is apparent that Friedman’s picture exaggerates the importance of the Phillips curve, and certainly of mistakes about it. That in itself suggests that his assertions about the policy consequences of Phillips’ research should be treated cautiously. But further, I suggest that little of the interest that was shown in the Phillips curve related to the advocacy of inflation. Some authors do express themselves in a way that suggests they are treating the curve as stable under inflation, but when they do so, very often they are either assuming that policy will aim at price stability or have something else altogether in mind. In those cases where it is presumed to be stable by an author who is advocating inflation, the lubrication argument is almost always visible. What is very rare indeed, but ought to be common on Friedman’s story, is actual advocacy of a policy of inflation on the basis of an account of the Phillips curve vulnerable to Friedman’s argument about expectations.
For example, a number of authors simply use estimates of the curve to advocate a policy of price stability. Typically, and understandably, they do not discuss its stability under a policy of inflation. One of these is Tsunehiko Watanabe, who estimated a Phillips curve for Japan and said, “The purpose of the present study is to provide a feasible answer to this problem of rising consumer prices from an economist’s (not from a politician’s) viewpoint” (1966, p. 32). He had no reason to consider the stability of the curve under a policy of inflation. Similarly, Otto Eckstein and Thomas Wilson revealed their attitude by the tone of their conclusion, saying, “These findings imply that there is no one critical level of unemployment which is consistent with ‘noninflationary’ wage increases. If profits are high, a much higher level of unemployment is necessary than if profits are low” (1962, p. 406).
Others made it clearer that they found the estimates of the required unemployment level alarmingly high, but, again, that is not the same thing as proposing a policy of inflation. For example, Klein and Bodkin (Reference Klein, Bodkin and Fox1964) remark that the indicated rate of unemployment is “more than we want to tolerate now” (p. 391), but they ended up contemplating compulsory savings as a means of restraining price increases (p. 428), and that, surely, does not reveal much tolerance of inflation.
One particular case that has attracted attention is that of Samuelson and Solow (Reference Samuelson and Solow1960). Leeson (Reference Leeson1997b), among others, suggests they had a special role in promoting the use of the Phillips curve to advocate inflationary policy. His case turns on the fact that, without any formal econometrics, they presented a roughly drawn “Phillips curve” for America, captioned the graph as “showing the menu” and offered their “best guesses” at the inflationary consequences of various levels of unemployment.
The issue of whether, in fact, the effect of Samuelson and Solow’s paper was to promote the acceptance of inflation is not under consideration here, but it would be easy to overstate the extent to which the actual content of the paper argues for that outlook. The title of the paper is “Analytical Aspects of Anti-inflation Policy” and most of it concerns the problem of distinguishing cost-push and demand-pull inflation. When the authors do address the issue of the Phillips curve’s offering a “trade-off,” they raise a number of doubts about it and, in particular, say “a past characterized by rising prices, high employment, and mild, short recessions is likely to breed an inflationary bias” (1960, p. 187). This leads them to suspect that high rates of unemployment would be only temporarily necessary to control inflation, and that, if a low-demand policy were followed, “it might be that the low-pressure demand would so act upon wage and other expectations as to shift the curve downward in the longer run—so that over a decade the economy might enjoy higher employment with price stability than our present-day estimate would indicate” (1960, p. 193). This clearly amounts to a position very much like that later taken by Friedman, and argues in favor of sound-money policy.
It is true that in Samuelson and Solow these are speculations, not conclusions. They also contemplate (p. 190) the possibility that attempts to control inflation by demand management might create adverse shifts in the Phillips curve, either because it would tend to lead to geographical concentrations of unemployment or (p. 193) because it would produce an increase in structural unemployment. They also suggest that low demand might lead to “class warfare and social conflict” and hence to slow technological progress. While these things argue for expansionary policy, they also clearly reveal substantial doubt as to whether the curve is stable. Even where they suggest (p. 192) that the “nonperfectionist’s goal” of 3% unemployment could be achieved by inflation of “as much as 4 to 5 percent,” this is immediately qualified not only by the words “in the years immediately ahead,” but also later by a warning that policy decisions might affect the curve in the longer run and that “it would be wrong” to imagine otherwise.
Overall, a balanced assessment of the actual content of the paper would say first and foremost that the Phillips curve plays a rather limited role in it. “Guesses” are presented as to the inflation costs of various employment targets, but it is specifically and clearly noted that these relate to the short run. The firmest position expressed about the Phillips curve is that there are a variety of things—on both the supply and demand sides—that might shift it. Indeed, Laidler (Reference Laidler, Johnson and Nobay1971), among many others, cited Samuelson and Solow as authority for the claim that there is no stable Phillips curve for the United States.
Laidler (Reference Laidler, Eaton and Harris1997) has also noted that it is no easy matter to find authentic, Phillips curve-based advocacy of inflation, even though by the end of the 1950s, discussion of various “trade-offs” of policy objectives was fairly widespread. He suggested that Grant Reuber (Reference Reuber1962) was the first to estimate the costs of inflation and unemployment, and deduce the optimal point on the trade-off. That study was a little-circulated paper for the (Canadian) Royal Commission on Banking and Finance (Dana Porter Reference Porter1964), but it led to Reuber (Reference Reuber1964), in which the author presents a diagram with a Phillips curve tangent to a community indifference curve at a positive inflation rate, and says that the Phillips curve offers “consistent and attainable combinations” of policy objectives. He ultimately concludes that for Canada, if international constraints were disregarded (and it is not clear that he thought they should be), the optimal combination would be 2.5% unemployment and 3.75% inflation. This is clearly something close to being a genuine piece of inflation-advocacy of the kind Friedman supposes was common, but it can hardly be said to have been influential. In fact, Laidler notes (fn 22) that the Royal Commission “seems to have been quite uninfluenced by Reuber’s study.” Indeed it was. It concluded (Porter Reference Porter1964, p. 419)—pre-empting Friedman on the point—that “if significant price increases come to be anticipated,” inflation would have no effect on output.
One group who clearly felt American unemployment had been too high in the 1950s was Kennedy’s Council of Economic Advisers. One might have expected them, perhaps following the lead of Samuelson and Solow, to deploy the Phillips curve in the early 1960s. In fact, they did no such thing.Footnote 16 Certainly, they argued for expansionary policy, but they did so on the basis that it could be achieved without inflation. Accordingly, John F. Kennedy’s first Economic Report of the President said, “the target for stabilization policy is to eliminate the unemployment which results from inadequate aggregate demand without creating a demand-induced inflation” (p. 46). It set a target of 4% unemployment as an interim goal and added, “If we move firmly to reduce the impact of structural unemployment, we will be able to move the unemployment target steadily from 4 percent to successively lower rates.”
The point, obviously, was that removal of the “impact of structural unemployment” was necessary to the further reduction of unemployment without inflation. That report is also the one that introduced the “Guideposts” designed to encourage non-inflationary wage settlements.
The Guideposts themselves led to much controversy. One particularly interesting contribution in the current context is that of Samuelson, who said, “Any criticism of the Guideposts which does not explicitly take into account the Phillips curve concept I have to treat as having missed the fundamental point of all economic policy discussions” (1967, p. 54).
There is obviously some exaggeration here, but it is instructive to see what Samuelson was saying the fundamentals were. If wages and prices were perfectly flexible, he said, there would be no inflation below full employment. In that case, “the authorities would engineer fiscal and monetary expansion just up to the point of full employment.” This is clearly a statement of the “L-shape” theory. But, he said, in the actual economy, prices creep upwards before full employment is reached (ibid., pp. 52–53). The purpose of the Guideposts is to stop this happening.
The interesting point is that he characterized a “good” Phillips curve as one which, at high levels of employment, is steep but indicates inflation arising only at low levels of unemployment, whereas a “bad” one is more shallow but crosses the unemployment axis at a higher level of unemployment. (The full employment level is the same in both cases.) The “good” one is, obviously, more nearly L-shaped. But the fact that this is the steeper curve surely shows that a policy of moving up it and accepting inflation was not the real concern.Footnote 17
In other cases, the vocabulary of the Phillips curve is deployed in a manner seeming to condone the view that it is exploitable, and perhaps that it should be exploited but in fact the discussion concerns some other issue entirely. For example, Alan Peacock (Reference Peacock, Peston and Corry1972) used the same kind of diagram as Reuber but his purpose had nothing to do with advocating inflation. Rather, it was to illustrate a point about the coordination of policy tools. Gary Fromm and Paul Taubman (Reference Fromm and Taubman1968) used it and said that one would want to find the optimal point on a properly specified Phillips curve, but this was a vehicle for their discussion of the econometric difficulties in discovering that specification, not for policy advocacy. In something of a similar way, Maurice Peston (Reference Peston1974) used the diagram to illustrate the general idea of a trade-off, but also described the Phillips curve as being discredited and certainly showed no sign of drawing policy conclusions from this discussion. In any case, his generally skeptical view of it is evident from Peston (Reference Peston, Johnson and Nobay1971).
Lipsey also drew the curve, but in his case the point of the exercise was not to consider optimal policy but, as he put it, “to re-examine the controversy between the two competing explanations for higher rates of unemployment in the United States: the structuralist and the deficient-aggregate-demand theories” (1965, p. 210). The paper offered a clarification of the conceptual distinction between them with a view to preparing the ground for an empirical determination as to which is the predominant problem. There is no issue about advocating any policy. Curiously, Lipsey later is critical of his own treatment (1978), saying that he had forgotten the warning against presuming the curve stable; and again in Lipsey (Reference Lipsey1981) criticized himself for presuming the relation stable. But these comments seem harsh. For the purpose to which he was putting the curve, its stability was immaterial. In 1965 he had said, “problems arise, however, when the objective of reducing unemployment conflicts with other objectives such as maintaining a stable level of prices and a satisfactory balance of payments. The problems that concern us can be illustrated by [the diagram]” (Lipsey Reference Lipsey and Ross1965, p. 211). Obviously the Phillips curve diagram does not depict the balance of payments, so it can be only illustrative, and problems “that concern us” is a reference to the objective of clarifying the distinction between the two types of unemployment.
Clearly, the posing of this question presumes that some worthwhile sense can be given to the idea of demand-deficient unemployment, which, some would say, it cannot. But contemplating that possibility is not inflation-advocacy.Footnote 18 Lipsey notes that eliminating such unemployment may conflict with other objectives and ultimately defines structural unemployment as the amount that remains when demand is optimized and frictional unemployment subtracted. While the use of price stability to illustrate the conflict is suggestive of a stable Phillips curve trade-off, nothing actually turns on that, and the point is explicitly only illustrative. No policy advocacy occurs.
Admittedly, when Frank Brechling (Reference Brechling1968) lifted the diagram straight from Lipsey (Reference Lipsey and Ross1965), he did it for the purpose of considering a possible trade-off between inflation and unemployment, and ended up concluding that such a trade-off might exist. He clearly had it in mind that an inflationary policy would be appropriate. Even so, it was a tentative conclusion, in the light of various data limitations that concerned him, and he did not advocate any particular policy. Indeed, he ended his paper on a noncommittal note by saying, quite rightly, that more than inflation and unemployment would have to be considered in the optimization problem.
Another branch of the literature focuses specifically on the econometrics of the Phillips relation and in particular on the size of the coefficient on expectations. It should be remembered, however, that the discovery—frequently made until the mid-1970s—that it was less than 1 is a separate question from whether inflation is desirable.Footnote 19
Solow (Reference Solow and Rousseas1968) was one of those whose econometrics seemed to contradict Friedman. His and others’ research suggested the coefficient was between 0.3 and 0.4 and he concluded that for empirically relevant cases, there is a trade-off. Having made a collection of further numerical assumptions, he said, that
there is in the not-so-short-run, a trade-off locus between inflation and real output; and that its position is such that high employment and price stability may be incompatible.… What to do about it is a difficult and important question of policy. It is doubtful that there is any single, simple, dramatic solution (p. 16).
And he suggests no policy.
Solow (Reference Solow1969) presents similar econometrics and takes a similar line. This piece is presented specifically as a test of Friedman’s hypothesis from his Presidential Address, which Solow describes as “one of those things it is hard not to believe” (p. 2). Later he says that his estimates “offer no support whatever to the expectations hypothesis in its strict form. On the contrary, they imply the existence of a permanent trade-off surface …” (p. 14).
He then presents some evidence that high rates of inflation might be fully incorporated into expectations, suggesting that a steady 8.5% inflation would have this characteristic. And he subsequently concludes:
whatever may be true of Latin-American-size inflations or even smaller perfectly steady inflations, under the conditions that really matter—irregular price increases with an order of magnitude of a few percent a year—there is a trade-off between the speed of price increase and the real state of the economy. It is less favourable in the long run than it is at first. It may not be “permanent” but it is good enough for me (p. 17).
This is as near as he gets to a discussion of policy and, hence, as near as he gets to the advocacy of inflation. In particular, its being “good enough” for him did not lead to a policy proposal. Although the existence of the trade-off is clearly stated, it seems to be very much a conclusion of a scientific enquiry. The hypothesis has been tested and the results are reported.
In the same vein, a number of studies published by the Brookings Institution in the years following Friedman’s Presidential Address denied the expectations hypothesis. Robert J. Gordon’s (Reference Gordon1970) was one, and could be read as concluding that a policy of inflation would be desirable, although again he stops short of saying so. He is firmly convinced that there is a long-run trade-off, but offers no particular policy advice. Gordon (Reference Gordon1971a) and Gordon (Reference Gordon1971b) are much more inflationist pieces, but any of these are rather late to count as substantiating Friedman’s argument, and, in any case, much of their force relates to the issue of whether policy should seek to reduce inflation rather than whether it should have permitted it in the first place. Others who addressed the question were out-and-out opponents of inflationary policy: Arthur Okun (Reference Okun1971) might think that the trade-off exists, but is explicit that the policy of exploiting it could not be implemented.
There are also those who advocate inflation in one way or another, but not on the basis of the kind of Phillips curve that Friedman had in mind. There are, first of all, those who clearly adopt the “lubrication” argument associated with Tobin (Reference Tobin1972b) and are led by this to advocate moderate inflation. Indeed, Solow and Gordon may have had this argument in mind. That is surely the explanation of Tobin’s more inflationist contributions, including where he does not state the theory fully, such as Tobin and Leonard Ross (Reference Tobin and Ross1971) and Tobin (Reference Tobin1972c).
Another instance of a variety of inflation-advocacy comes from those who have a form of unemployment hysteresis in mind. Inflation is advocated not because some point along the Phillips curve is preferable to the point at price stability, but because the policy in question will move the Phillips curve in a desirable direction. As has already been noted, Samuelson and Solow (Reference Samuelson and Solow1960) speculated on that possibility, and it was mentioned by Brechling (Reference Brechling1968) and by Albert Rees (Reference Rees1970). It was advocated more forcefully by Franco Modigliani and Ezio Tarantelli (Reference Modigliani and Tarantelli1973). They suggested that in a “developing country,” expansionary policy would enlarge the pool of workers in the active labor force and that once engaged, these new workers would contribute to keeping wages down.Footnote 20 The long-run tendency would be towards stable prices at a higher level of employment than initially. Their discussion does presume that, except for their hysteresis effect, the Phillips curve is stable. But it is not the point of their argument to consider that, and their whole discussion is organized around the question of what it takes to achieve price stability.
There were, of course, true advocates of inflation basing their arguments on the Phillips curve, but, for the most part, they are more minor figures. Reuber was one—as already noted, not an influential one. William Bowen (Reference Bowen1960) was another, but his policy proposals were not repeated in Bowen and Albert Berry (Reference Bowen and Albert Berry1963). Bodkin (Reference Bodkin1966, pp. 279–280) estimated the rate of inflation that would result from a “full employment” policy, and was therefore clearly contemplating such a policy. Robert A. Gordon (Reference Gordon1967, p. 182) certainly espoused a moderate inflation as the price for high employment. But this hardly amounts to consensus.Footnote 21
If one casts the net a little wider, one could include Paul Sultan (Reference Sultan1957), a textbook treatment that advocates inflation. There is also Sumner Slichter (Reference Slichter1948, pp. 42–45), who certainly favored inflationary policy and continued to do so in popular writings in the 1950s, such as Slichter (Reference Slichter1952), Slichter (Reference Slichter1954), and Slichter (Reference Slichter1957). Then there are the theoretically more sophisticated contributions of William Vickrey (Reference Vickrey and Kurihara1955) and Kaldor (Reference Kaldor1959b), whose argument was based on the theory from Kaldor (Reference Kaldor1957). None of these, as will be apparent from their dates, found their motivation in Phillips (Reference Phillips1958), and the last two authors based their argument on altogether different ideas.
There are also a few who accept that changing expectations will change the relation (or some equivalent argument) and nevertheless contemplate inflationary policy. Brechling (Reference Brechling1969) speculated on this. There is no reason, he said,
why slowly accelerating inflation and high output should not be preferable to constant (or zero) inflation and low output. Moreover, a stabilization program might conceivably involve the maintenance of low levels of unemployment interrupted by sharp increases in unemployment to destroy inflationary expectations (p. 161).
Obviously, a more developed argument about expectations might put paid to that, but, of course, he is not advocating the policy. Abba Lerner (Reference Lerner1960) is in the category of inflation advocate, if one takes seriously his idea for alternating accelerating inflations with currency reform. Perhaps most notably, there is Phelps (Reference Phelps1967) himself. His point was that if there is a positive rate of time discount and adaptive expectations, there would be some optimal rate at which inflation would increase. It is a kind of inflation advocacy, but deeply theoretical, and it is something of a curiosity that he is one of the people persistently quoted as having destroyed the rationale of the Phillips curve.
As to the Phillips curve itself, however, its role in practical policy making would seem to be the opposite of what Friedman suggested. The first appearance of anything like it in Economic Reports of the President comes in 1969 where it appears in the context of a discussion of how to reduce inflation, and the clear implication is that higher unemployment will do the job.Footnote 22 Such a policy is not espoused, but rather the point is made that it would be preferable for the Guideposts to work. Similarly, Michael Stewart (Reference Stewart1977) says, with some sarcasm, of the Conservative government elected in the United Kingdom in 1970, that they promised to control inflation, and that “Underlying the Government’s confidence in the policy this time lay one of the most dazzling magic keys in the history of economics: the Phillips Curve” (p. 157).
In other words, they were going to raise unemployment. In both countries, it would seem, the prominence of the Phillips curve (or a like relation) in the discussion of inflation policy emerges when it is deployed to describe the requirements of price stability.
V. SOME CONSEQUENCES OF THE ACCEPTANCE OF FRIEDMAN’S ACCOUNT
That the economics profession ever made such foolish mistakes as Friedman claimed is hardly a recommendation of it; but nor is the fact that it quiescently accepted the view that they were so often made. For both these reasons, and simply as a matter of proper scholarship, the received account of the orthodox approach of the 1960s to the problems of inflation and unemployment should be corrected. As far as the history of economics is concerned, Friedman’s Nobel Lecture is no more than mythologizing.
It should be noted, however, that although it is only a myth, Friedman’s depiction of a naïve and, indeed, rather stupid professional consensus of the 1960s may not be entirely harmless. Two consequences in particular should be considered. The first is that Friedman subjected that orthodoxy not merely to criticism, but to derision. He suggested that the greatest economists of the time had, to what can only be the deepest professional discredit, committed the most foolish of intellectual blunders. Once that is accepted, it hardly needs to be added that, shameful as it was ever to have made those inflationary mistakes, it would be more shameful still ever to make them again. Thus this romance of a Dark Age of Economics does much, I suggest, to promote the orientation of policy around the idea that the pre-eminent goal must be stable prices and thereby to shore up the support for approaches to policy making based on central bank independence, inflation targeting, and fiscal conservatism.
Secondly, while he is remembered for dismissing the possibility of a long-run trade-off, Friedman’s greater achievement was to put the short-run trade-off at the center of almost everyone’s macroeconomic analysis. The historically attested theoretical alternative to the vertical Phillips curve is not a trade-off in the long run, but a denial of it in the short run. On this view the problem of unemployment was, over a very wide range, simply a different problem from that of inflation. It is as much the triumph of Friedman’s history as of his theory that it has become almost impossible to articulate the view that macroeconomic policy should be concerned with unemployment without seeming merely to plan a day-trip up the Phillips curve. But this is not the only view one might take.
It may be that the experience in the United States and the United Kingdom around the turn of the century will revive interest in a macroeconomics without the Phillips curve. In those countries, policy evidently did not—whatever the accepted theory says—feel itself governed by the view that falling unemployment must be inflationary, and nor did efforts to make durable estimates of the NAIRU prove altogether successful. The recognition that the economics of the Keynesian era was not just a matter of considering the ramification of elementary errors, and that the Dark Age is fictitious, would, perhaps, facilitate the development of thinking in this area.
