The history of the Phillips curve

The wonky little chart on the right comes from Phillips’ 1958 paper, ‘The relation between Unemployment and the Rate of Chance of Money Wage Rates in the United Kingdom, 1861-1957’.

Phillips drew the heavy curved line as a representation of the pattern his analysis revealed in the dots. Guy Routh – an under-rated British economist of the 1950s and 1960s – thought the data points suggested an ostrich more than they did Phillips’ curve.

Something like Routh’s view was widely shared – Phillips was initially more mocked for this curve than he was respected. Quite apart from the question of whether there was really a curve there at all – the appearance of one really only emerges from Phillips’ statistical analysis – everything was wrong with his study. He had dubious data sources, very dubious data manipulation, some highly convenient assumptions, and an extraordinary willingness to extrapolate doubtful findings into bold conclusions.

The curve itself was derived from the data from 1861-1913. Phillips, though, said that later points, all the way up to 1957 more or less fitted the curve, if allowance was made for certain factors. That meant the same cure described this relationship for nearly 100 years.

The social changes over those 100 years were enormous – the rise of trade unionism and the creation of the welfare state are two points with an obvious connection to wage bargaining. That means that if Phillips’ suggestion were correct, or even nearly correct, it would be very big news. The link between wages and unemployment would have to be very deep indeed, if the same relationship prevailed through all those changes. That would make Phillips’ finding a ‘law’ of a kind and power that usually only featured in economists’ dreams.

Not quite an ostrich, but near enough: Rhea Darwinii, from John Gould, The Zoology of the Voyage of the Beagle, Part III: Birds. London 1838-41.

That idea did not last long. It was too easy to see flaws in Phillips analysis, and better attempts to replicate his findings failed to do so. While those studies were being conducted the label ‘Phillips curve’ started being applied to a much wider range of relationships involving wage change, unemployment, and other variables. None of these had anything at all to do with Phillips since the idea of these sorts of relationships was much older than his work. But still, his name was the one that came to be applied to them.

A little later, when inflation started to rise, the same label was applied to relationships between inflation and unemployment. Wage changes and inflation are surely connected, but they are also quite different things, so this really was an expansion of the idea of a ‘Phillips curve’ – and one owing even less to Phillips. Nevertheless, in the 1970s, when inflation and unemployment were both much higher than had been normal, it became natural to say ‘the Phillips curve has shifted’. That meant the sort of outcomes suggested by a curve drawn for data up to about 1965 were much more favourable than one for later years.

As things turned out, unemployment was high for much of the 1970s, but inflation came to be seen as the intractable problem in many countries. As a consequence, during the decade, policymakers gradually came to accept that inflation would only be controlled with a large reduction in demand. That meant, in practical terms, there would have to be deep recessions. And in due course, Paul Volcker at the Federal Reserve and Margaret Thatcher in the United Kingdom became famous for adopting that policy.

While the debate that led to those policies was going on, something entirely new started to be said about the ‘Phillips curve’. From about 1976 it was increasingly said that in the 1960s, policymakers had believed the curve offered them a ‘menu’ of inflation and unemployment outcomes. It had been believed – so said this story – that if policymakers would accept 5% inflation, they could have lower unemployment than if they were prepared to accept only 2%. Actually, practically no one ever believed that, but the story that it had been the normal view became very widely believed.

Another story went with that one. That was that the big mistake in the idea of the menu arose from failing to consider the effect of ongoing inflation on wage setting. Obviously, if there is steady inflation at, say, 5%, over a long period, wage bargainers will notice. And it is a piece of common sense that when they do, that will affect wage bargaining. It does not take much to see that if there are no other factors, bargainers will come to an agreement each year that makes wages 5% higher than they otherwise would have been. Once that starts to happen, it is no surprise if the supposedly beneficial effect of inflation in reducing unemployment will disappear. But the story that was told at the end of the 1970s was that the economists of the 1960s had not thought of any of that.

That is absurd of course. The argument is easy to find in the economics of the 1960s, as it is in the economics of the 1950s. It is not stated often because, of course, almost no one even contemplated the possibility that policy would allow inflation to persist. But it is stated often enough, and in terms which make it clear that the authors thought it would surprise no one, to make it clear that it was widely appreciated.

Nevertheless, the story that this was big news, and a ‘remarkable insight’ associated with Milton Friedman spread astonishingly quickly. It became one of the things that students learned from their textbooks. Economics textbooks tend to say practically nothing about the history of the subject, but the error over the Phillips curve, and Friedman’s great insight are two things repeated again and again.

Well, those stories are not true. Nothing like them is true either. The whole picture is a fantasy. That is remarkable and of course it is not easy to convince people. The story became common place in the undergraduate textbooks of the late 1970s. Ten years later, junior faculty members, who had learned it from those books, were telling the same stories in their lectures; and forty years later there was hardly an economist who had not been exposed to it in one way or another. Still, in Macroeconomics and the Phillips curve myth, I reckon I have presented enough evidence to convince anyone who can be convinced.