No More Easy Choices: The Death of the Phillips Curve?

“An economist is someone that will know tomorrow why the things he predicted yesterday didn’t happen today.” – Lawrence Peter

“There are no easy choices. Easy choices are long gone.” – Alan Greenspan

The issue of causation is among the most difficult to grapple with in the social sciences. That’s because all of us are inclined to see patterns and when we notice that two things are happening at the same time, human nature presumes the relationship will perpetually continue. Of course, it is possible that what we actually observed was a coincidence or that a third cause is generating the relationship being witnessed.

Steven Levitt and Stephen Dubner gave a terrific example of this human limitation in their popular book Freakonomics. Polio, an infectious disease, became a growing problem in the United States in the beginning of the twentieth century and the results were dire, including potentially permanent paralysis for those infected with the virus. Someone sifting through the data claimed to have found the answer as to why polio was spreading – ice cream since rates of infection and ice cream consumption showed a strong positive correlation. In reality, exposure rates simply rose in the summer time, precisely when ice cream consumption was highest. The cause of the sharply increased spread of polio in the early twentieth century was improved sanitation. Very young children who previously were exposed and immunized to the virus no longer were.

Another example frequently cited was the erroneous belief that hormone replacement therapy (HRT) lowered a woman’s risk for heart disease. That was because researchers noticed that women who received HRT had a lower risk of heart disease than women who did not receive the treatment. Throughout the 1980s and 1990s older women were often encouraged to participate in HRT for the sake of their hearts. It turns out that HRT actually causes a slight increase in a woman’s risk for heart disease. The erroneous conclusions that researchers had drawn in the past were actually caused by the fact that wealthier women were more likely to participate in HRT and are also less likely to have heart disease.

The point in drawing out these examples is to make clear that it is necessary to approach data and relationships with great humility. This is as true for economics as it is in any other discipline. The history of the Phillips Curve, a curve showing the relationship between inflation and the unemployment rate, is a great demonstration of this. It is also an important one since the Phillips Curve is among the most discussed ideas in macroeconomics and a long-time factor in setting Federal Reserve policy.

A.W. Phillips was a New Zealand born economist who published a paper in 1958 entitled “The Relation Between Unemployment and the Rate of Change of Money Wage Rates in the United Kingdom, 1861-1957,” that posited that inflation and the unemployment rate were inversely related to each other. This was not such a crazy suggestion – others had made it before and empirical evidence seemed to support it.

Scatter plot of the change in wages and the unemployment rate in the United Kingdom between 1861 and 1913 as it appeared in A.W. Philipps’ 1958 paper.

Further, there are obvious mechanisms by which causation could occur in the relationship. If employment levels are high, then workers should have greater bargaining power with employers and demand higher wages. This, should, then cause the overall price level in the economy to rise. That suggested mechanism was labeled “cost-push” at the time. Another mechanism, called “demand pull” suggested that at full employment the available productive capacity of the economy was already being put to use and so any increase in demand would necessitate a rise in prices in order for goods and services to be allocated without shortage.

Paul Samuelson and Robert Solow, two extremely well-respected economists, applied this reasoning to American data in a 1960 paper entitled “Analytical Aspects of Anti Inflation Policy.” Going further than Phillips did, many presumed that Samuelson and Solow argued that the Phillips Curve could be exploited as a policy tool in which policy makers had a set of inflation-unemployment choices to select. To be fair to the two Nobel Prize winners, a careful reading of their paper does not advocate for this approach, but that did not stop policy makers from reading it that way. If everyone has a job, wouldn’t they put up with a little higher price level? The politicians thought so.

And for the most part, the relationship held throughout the 1960s.

Most everyone knows what happened next – something some economists thought could never happen: high unemployment and high inflation at the same time. The relationship between inflation and the unemployment rate in the 1970s actually had a small, positive correlation.

So, what happened? Most explanations of the period of time owe their origins to Milton Friedman and Robert Lucas. The simple version of the explanation goes like this. So long as expectations in an economy for inflation are anchored at a particular level, a short-run trade-off between prices and employment can exist. If people believe that inflation in the long-run will average, say, 3% and they see prices rise in the short-term they falsely believe that this price increase is a signal of greater demand when in reality it is just excess money filtering through the economy. People are not perpetually stupid, though, and once their expectations of price stability have been dashed they stop making this false assumption. Further, high inflation expectations cause workers to demand higher wages and producers to raise prices, perpetuating a cycle.

In the long-run, said Friedman and Lucas, there is no relationship between inflation and unemployment and that inflation itself is a consequence of excessive money printing.

The real Fed Funds Rate since 1954. Beginning in 1968, Fed policy became progressivey easier until 1979.

Again, there is empirical evidence to support this. A long range graph of inflation and the unemployment rate shows no strong correlation.

The Federal Reserve has been charged with the “dual mandate” of price stability and full employment since a 1977 law, the Federal Reserve Reform Act. But, if no relationship exists between the two, should the Federal Reserve really just be targeting price stability? And how does this relate to our current economic situation?

Certainly, many observers feel strongly that the Fed should lose the dual mandate. They claim that it only serves to encourage excess money creation and “boom-bust” cycles. Indeed, there is strong evidence that suggests that excess liquidity which at one time pushed wages higher and caused consumer inflation is now flowing in a way that is producing asset price inflation. The stock market reached unsustainable highs in both 2000 and 2008 while real estate prices crashed after peaking in 2006. Why has this shift happened?

Much, but not all, of the reason, is due to changes in how monetary policy is transmitted to the broader economy. In decades past, transmission happened through retail banks. When the Fed lowered interest rates, borrowing became cheaper and more loans were made which were invested in the broader economy, raising competition for goods and services. Today, enormous changes in the structure of the financial system mean that more and more frequently, lending is conducted through securitization markets. At the same time, firms themselves are more idea-intensive and less capital-intensive than they once were, weakening the transmission of monetary policy to the broader economy. Independent structural changes in the global economy that have resulted in increases in income inequality – and thus, a greater share of the wage pool being concentrated, also changed the ultimate effect of monetary policy on the real economy. Wealthier people save, poorer people consume.

Paul Krugman has dismissed concerns about the impact monetary policy can have on asset prices because he says the argument is circular in nature. If the real economy demands a 1% Federal Funds rate and low-interest rates on longer term bonds, then by definition the appropriate price for other assets, including equities, must be higher than in past interest rate environments. This is true, but it also ignores the peril that can come if the Federal Reserve would need to quickly shift direction and cause a major drop in equity and real estate markets.

Ultimately, we must return to our original question, ‘Is the Phillips Curve Dead?’ The original incantation of the Phillips Curve died a long time ago. But its death was not the result of it being “wrong” but incomplete. What policymakers saw as a relationship between inflation and a single variable is actually a multivariable relationship. Not only is the rate of inflation relevant, but also expectations of future inflation.

If inflation is running at a high level and a recession begins, then the unemployment rate would rise and inflation would likely fall. Not only would inflation fall, but expected inflation would fall with it. When the recession ends and the unemployment rate reverses lower, a particular unemployment rate would correspond to a lower rate of inflation from before the recession. Were this scenario to play out in the broader economy, one would expect the Phillips Curve to have a horseshoe or spiral shape to it, as one unemployment rate could give multiple inflation levels.

Looking at the Phillips Curve for the 1970s with monthly data…

…as well as the 1980s….

…do indeed have this shape to them.

The curious thing is that our experience since the recession has not been the same as the above Phillips Curves would suggest. The rate of inflation bottomed in 2014, but in recent months it has started to turn lower again. Year over year inflation in July was only 1.7%, even as the unemployment rate maintains a rate a little above 4%. The cause may well be a lower bound to inflation caused by stickiness in wages. Real wages are easy to cut when inflation runs high because nominal wages can simply rise less than prices do, but when there is no inflation nominal wages are less likely to be cut.

Meanwhile, asset prices stay enormously high in large measure thanks to Fed policy.

What, then, is the Federal Reserve to do? At the September meeting, observers expect some unwinding of past quantitative easing, but not an increase in the policy rate. Most are now split as to whether the Fed will hike in December.

The life and death of the Phillips Curve, along with ice cream and polio and HRT and heart disease, do teach powerful and complimentary lessons – namely, that assuming the world is simple is a sign of vast arrogance. The choices of more than three hundred million people in an economy are not easily reduceable and trying to engineer those choices almost always leads to unintended consequences just as seeing simple patterns in large data sets can lead to false conclusions.

The Phillips Curve lives on today, but in a highly modified form. Continuing to ignore asset prices in the equation is also a mistake and ignoring the effects of asset price inflation will not change the reality of monetary transmission.

Unfortunately, Alan Greenspan was correct when he said we no longer had easy choices. Of course, we never did, but we should realize it so much more acutely now. We can only hope that Janet Yellen does.


more recommended stories