Lessons from the Phillips Curve
Note: An earlier version of this article appeared at Economic Forces.
Anyone who has followed macroeconomic debates of the last half-century is surely familiar with the Phillips Curve, a name given to the supposed negative relationship between the unemployment rate and the inflation rate. The Phillips Curve started out as an empirical relationship. Yet, the stability of the statistical relationship convinced some that the relationship might have been more than that. It might be something to be explained by theory. And, perhaps more importantly, it might be a relationship that could be exploited by policymakers to achieve desired policy outcomes.
Unfortunately, for macroeconomists and the world, the exploration of the Phillips Curve has taken many different paths and yet somehow ended up in the same place. The world could have been spared much of this meandering with a more firm grounding in what theory does and does not tell us.
Mistake #1: The “Natural” Rate of Unemployment
One of the most influential attacks on the Phillips Curve was levied by Milton Friedman in his presidential address to the American Economic Association in December 1967. In his talk, Friedman argued that the ability to bring down the unemployment rate through expansionary monetary policy would be limited.
In making this argument, he used the concept of a “natural” rate of unemployment. The basic idea is that there is a particular unemployment rate that is consistent with the basic structure of the labor market (e.g., real wages, the informational costs associated with search for jobs and posting vacancies, the volatility of supply and demand for both goods and labor). Put differently, if one were to model an economy with all of these labor market characteristics, the “natural” rate might be thought of as the equilibrium unemployment rate in that model.
In reality, at any given point in time, the actual unemployment rate might not be equal to the natural rate. Nonetheless, there would be a tendency for real wages to adjust to push unemployment toward this natural rate.
Friedman’s use of the term “natural rate” was largely due to the fact that he was trying to discuss the role of monetary policy in targeting a particular level of employment (or a particular unemployment rate). For example, suppose that the unemployment rate is currently equal to the “natural” rate. Monetary policymakers might try to reduce the unemployment rate by conducting expansionary monetary policy. An increase in the money supply would increase prices. Since Friedman assumed that output prices adjusted faster than input prices, this would mean that firms’ profits would increase and they would want to produce more. Furthermore, real wages would decline and firms would want to hire more workers. The unemployment rate would decline. However, input prices would ultimately begin to rise as the demand for labor increased. This adjustment of real wages would continue until the economy returned to the natural rate of unemployment.
The fundamental point here is that when the unemployment rate is above the natural rate, policymakers can reduce the unemployment rate with expansionary policy. However, according to Friedman’s theory, once unemployment got to the natural rate, monetary policymakers could only keep unemployment below the natural rate with ever-higher inflation.
Friedman’s “natural” rate was a modeling device. It was used to illustrate the idea that the unemployment rate tends to be pinned down by the structure of labor market. And, while policymakers could cause unemployment to fall, they wouldn’t necessarily know what the “natural” rate was and therefore risked creating costly, inflationary policy errors by trying to target a particular rate of unemployment that was “too low,” given the conditions of the labor market.
Unfortunately, people drew the wrong lessons from Friedman’s discussion. For Friedman, the natural rate was a theoretical concept to illustrate the limitations of trying to fine-tune the economy with monetary policy. Yet, in the years since then, policymakers and researchers have spent an inordinate amount of time trying to use economic data to estimate the natural rate of unemployment so that they might better be able to fine-tune the economy.
Mistake #2: What Role Do Expectations Play?
A second mistake that people make when thinking about the Phillips Curve is a failure to take into account expectations when thinking about empirical relationships. In fact, Milton Friedman made this point in his speech. His argument is that people would adapt their expectations of inflation based on their recent observations of inflation. If the inflation rate increased from 2 percent to 4 percent, this would temporarily reduce the unemployment rate. Real wages, however, would adjust to the new inflation rate and so would consumers’ inflation expectations. Thus, to get the unemployment rate to decline, the central bank would have to raise the inflation rate above 4 percent. What this meant was that unemployment would decline only when there was unexpected inflation.
Robert Lucas took this one step farther. He used a theoretical model to demonstrate that one would get different outcomes depending on whether changes in the money supply were anticipated or unanticipated. If the growth rate of the money supply was entirely random, then firms would see an increase in the demand for their product. Unsure of whether this increase in demand was caused by an increase in the demand for their particular product or due to an increase in the money supply, firms would increase both output and prices. This would generate a negative relationship between inflation and unemployment. If the change in the money supply was anticipated,however, the policy would not be effective at reducing unemployment at all. In fact, the higher expected rate of inflation would be counterproductive.
This point was especially important given the evidence. The early, robust, negative empirical relationship between inflation and unemployment was based on sample periods in which the monetary regime was the gold standard. Under the gold standard, the price level tends to be relatively constant over time. Fluctuations in the price level (and therefore the rate of inflation) are caused by changes in the supply and demand for gold. Large changes in the supply of gold tend to come about through gold discoveries. Large changes in the demand for gold tended to come from countries entering or leaving the gold standard. These sorts of fluctuations are fairly random. Once the gold standard was abandoned and central banks started using the money supply to try to achieve desired policy outcomes, however, the statistical relationship broke down.
Finn Kydland and Ed Prescott extended the insights of both Milton Friedman and Lucas. They demonstrated that if unexpected inflation could generate lower unemployment, and the central bank’s belief about the natural rate of unemployment was below the “true” natural rate, the central bank would have an incentive to promise low inflation and then produce a higher rate of inflation in order to reduce unemployment. What was most important, however, was that they showed that if the central bank had this incentive, people would recognize this incentive and expect that the central bank to act accordingly. Given the central bank’s incentives and given how those incentives shape the expectations of the public, any attempt by the central bank to produce unexpected inflation to lower unemployment would be unsuccessful. In fact, all this sort of policy would produce is higher inflation.
All four of these economists won the Nobel Prize in Economics, in part for these insights. It is therefore worth exploring the implications here.
The first is that whether a Phillips Curve would ever be evident in the data depends on both policy and expectations about policy. The statistical relationships between economic variables will depend on the policy regime and how people behave (and how people expect policymakers to behave) under that monetary regime.
The second important implication is that policymakers cannot simply manipulate a statistical relationship between two economic variables to generate a desired outcome. Expectations matter.
Mistake #3: The Phillips Curve and the Data Generating Process
One issue for the Phillips Curve is as follows. If a central bank is conducting what most economists would consider “good” policy, there should be no evidence of the Phillips-Curve-type relationship between the inflation rate and the unemployment rate in the data. To understand why, consider the thermostat in your house (another Milton Friedman analogy).
Suppose that you set the thermostat in your home at 70 degrees, and suppose you have one of those fancy thermostats that automatically uses the air conditioning when the temperature gets too high and the heat when it gets too cold. If you were to keep track of the temperature in your house over time, it wouldn’t change much. The temperature would always stay close to 70 degrees. But now suppose that you also collected data on the amount of time that the heating and air conditioning in your house were running, and you tried to explain fluctuations in the temperature by the amount of time that your heating and air conditioning systems were running. Since the temperature is roughly constant, you would be unlikely to find any statistical relationship between the amount of time the systems were running and the temperature. But only a fool would conclude from this statistical relationship that the system wasn’t working!
The same idea is important for understanding issues with the Phillips Curve. For example, suppose that demand is rising in the economy. All else equal, this will tend to cause prices and production to increase. Since production is rising, it will tend to cause unemployment to fall. This generates a negative relationship between unemployment and inflation. However, all else isn’t equal. If policymakers are tasked with keeping the price level relatively constant, then when prices rise, they will tend to contract monetary policy to offset the increase in demand.
All else equal, when there are supply shocks to the economy (as in the case of supply chain disruptions), this will tend to increase prices, but reduce output. Thus, prices will rise, but so will unemployment. Ideally, since monetary policymakers can’t fix the supply chain by increasing the money supply, monetary policy wouldn’t respond to these sorts of changes. But think about what this means for the data generating process. When there are demand shocks, these are offset by monetary policymakers, and inflation and unemployment remain largely unchanged. When there are supply-side disruptions, policymakers will tend to ignore them, and there will be a positive relationship between unemployment and inflation.
What this tells us is that if there is evidence in the data of a Phillips Curve, that doesn’t mean that this is some structural relationship that policymakers can exploit. On the contrary, this would mean that policymakers are doing a bad job!
Price theory can be an antidote to bad arguments. It is not only important to understand what theory tells us, but it is also important to understand what theory doesn’t tell us. Friedman’s “natural” rate of unemployment tells us that there are limitations to policy based on real resource constraints. It does not tell us that we should try to estimate this natural rate in order to try to manipulate policy around that rate. In fact, subsequent contributions to economic theory tell us that this is a fool’s errand.
The post Lessons from the Phillips Curve was first published by the American Institute for Economic Research (AIER), and is republished here with permission. Please support their efforts.