Trouble with the Phillips Curve

The Phillips curve is at the heart of recent debates within and outside the FOMC about the timing and pace of rate normalization. It is a staple of structural macro models and is the prism through which the debate about near-term inflation prospects takes place. For monetary policymakers, it is the relationship between inflation and labor market slack that provides the link by which monetary policy affects inflation, allowing them to achieve their inflation objective. But there are important questions about the properties and reliability of the Phillips curve as a guide for monetary policymakers on how to achieve their inflation objective, and its empirical properties can undermine the effectiveness of the FOMC and other central banks in achieving their respective inflation objectives. In this piece, we provide an overview of issues relating to the Phillips curve that is at the heart of the current monetary policymaking process. 

Yellen Relies on the Phillips Curve 

Chair Yellen, in a speech in 2015, emphasized the key role of the Phillips curve in her (and most participants’)  judgment that inflation would head to 2% as the unemployment rate continued to fall.1 More recently, in her  press conference following the March 2016 FOMC meeting, she reiterated that she finds the Phillips curve  framework useful: 

So the Phillips curve posits that there is a relationship between the degree of slack in the labor market and inflation, and it is an empirical relationship that, while not absolutely tight, has been a consistent relationship over time, and I believe that relationship still holds. 

The minutes of the July 2017 FOMC meeting described a discussion among participants about the Phillips curve, indicating that most agreed with Yellen’s sentiment. 

She presented in her 2015 speech what we take to be a conventional empirical Phillips curve model of short-run inflation dynamics, and we present it here to frame our discussion of issues surrounding the Phillips curve,  ones that have been debated within the Committee and have resulted in different perspectives about the timing of liftoff and now the pace of rate hikes. 

Her Phillips curve, specified in terms of core PCE inflation, is shown in equation (1) below: 

1Inflation Dynamics and Monetary Policy,” September 24, 2015, Chair Janet L. Yellen at the Philip Gamble Memorial  Lecture, University of Massachusetts, Amherst, Amherst, Massachusetts. 

Key Features 

The key factors driving core inflation in Phillips curve models like Yellen’s are past and expected inflation and labor market slack.2 The influence of past inflation introduces inertia into the relationship and long-term inflation expectations exert a gravitational pull on actual inflation, ideally toward the FOMC’s inflation objective. The coefficients on past and expected inflation sum to one, guaranteeing that there is a unique level of the unemployment rate, the NAIRU, that is consistent with stable inflation. But the Phillips curve is most associated with the short-run relationship between inflation and the unemployment rate: When slack increases, inflation slows. This is referred to as the short-run trade-off, because it holds as long as inflation expectations are well anchored, meaning they don’t move in response to deviations of actual inflation from the FOMC’s objective. That is usually the assumption in empirical Phillips curves and has been the case in recent years. Monetary policymakers rely on this relationship in their assessment of appropriate policy—that is, a policy that achieves the FOMC’s inflation objective over the policy horizon.  

Skepticism about Setting Monetary Policy Based on the Phillips Curve 

There are many questions within and outside the FOMC about the reliability of this relationship and specifically to what extent it should guide monetary policy in pursuit of the price stability objective. Governor Brainard,  for example, has said, “I am not confident we can count on the Phillips curve to restore inflation to target in  today’s economy.”3 And the July 2017 FOMC minutes said that “A few participants cited evidence suggesting  that [the Phillips curve framework] was not particularly useful in forecasting inflation.” 

The questions about the Phillips curve include: how to measure slack, how stable longer-term inflation expectations are, and what the implications of a very flat Phillips curve are. In addition, the most pressing unsettled question today—and the one most relevant to whether the Committee will hike again in December— is whether the recent slowing in inflation reflects the transitory effect of an idiosyncratic price level shock or instead suggests that underlying inflation may be lower than it had appeared to be.  

How to Measure Slack? 

An important and longstanding difference in views within and outside the Committee, both during the lead-up to the first-rate hike and currently as the pace of rate hikes is debated, has been whether there remains  “running room” or the economy is already close to or indeed beyond full employment. Those who believe we are already at or beyond full employment worry about “falling behind the curve” and the possibility that the  FOMC will need to tighten quickly to contain inflation at some point, which in turn would increase the risk of a recession. They tend to favor a steady pace of rate hikes, sometimes faster than the consensus, toward the neutral level of the fund’s rate. Those who believed, and perhaps still believe, that there remains running room have tended to question having another rate hike and favor a slower pace of rate hikes relative to the consensus. 

This disagreement, from the Phillips curve perspective, is centered on how to measure slack. In conventional  Phillips curve models like Yellen’s, slack is measured by the gap between the standard U-3 measure of the unemployment rate and its NAIRU. By that measure, and in the median projection of participants, we are already beyond full employment and expected to move further past it. The question is usually posed in terms  of whether the U-3 unemployment rate gap (relative to its NAIRU) fully or best captures the degree of slack  

2 Additional factors can be included as well. For example, Yellen’s specification also includes a term accounting for changes in the relative price of core goods imports. 

3Navigating the Different Signals from Inflation and Unemployment,” May 30, 2017, Governor Lael Brainard at the New  York Association for Business Economics, New York, New York.

most relevant for short-run inflation dynamics. Some prefer a broader measure, like U-6.4 While this question lingers to a degree, the concern has diminished as U-3 has fallen to such a low level and the gap between the U-6 and U-3 measures, which increased sharply during and following the most recent recession, has declined to near its pre-recession level. A difference in views about the amount of slack remains to some degree, but this issue has faded in importance in the debate of near-term policy issues, for example, whether the FOMC should hike in December. It remains an issue concerning the pace of rate hikes thereafter. 

The Flat and Flatter Phillips Curve 

It is widely agreed that the slope of the Phillips curve—how much inflation rises for a given decline in the unemployment rate—has fallen over time. The Phillips curve is said to have become flatter and flatter,  diminishing the role of slack in explaining inflation. Indeed, as the slope becomes flattered and flatter, inflation begins to look more and more like a random walk, even if long-run inflation expectations remain reasonably well anchored. This is the theme in a paper prepared for the most recent Monetary Policy Forum: 

The central characteristic of our time‐series model is that inflation is heavily influenced by a slow-moving trend that we call the local mean. This model has no trouble explaining the post‐GFC inflation outcomes because it assigns a very limited role to slack in influencing the local mean. Although we do find a reliable statistical connection between slack and this highly persistent inflation trend, the quantitative impact is relatively small. This is consistent with the growing evidence that the Phillips curve has become flatter (IMF, 2013). From a practical perspective, given the range of labor market conditions observed in the last 30 years, this channel is weak enough that it is not likely to be a good predictor of near‐term inflation developments.5 

This paper takes the implications of the flat Phillips curve to its logical extreme, but unfortunately, that may not be so extreme in practice. Even if that is going too far, the flat slope makes the FOMC’s job in achieving its price stability mandate much more difficult. For example, if the underlying inflation rate today is in line with recent readings of 1½%, and that’s after the unemployment rate has already fallen more than five percentage points, how could the FOMC ever get back to the 2% objective?6 On the other hand, there is an upside as well. If the economy is at full employment and 2% inflation, then it’s hard to push the economy much away from 2%, short of a very severe recession. In this case, the FOMC wouldn’t have to worry as much about its inflation objective and instead could focus mostly on keeping the economy near full employment. 

Wage Inflation and Nonlinearity in Phillips Curves 

A couple of possible features of inflation dynamics would suggest that monetary policy nevertheless might be able to exert sufficient upward pressure on inflation to move inflation back to 2% in the medium term. First,  there might be a nonlinearity in the Phillips curve, such that after some point a further decline in the unemployment rate has a larger effect on inflation than suggested by a linear curve like Yellen’s. That is, after some point the slope of the Phillips curve might rise with further declines in the unemployment rate. There is  

4 Some economists prefer a narrower measure, short-duration unemployment. But that view has likely no advocates within the Committee. 

5Deflating Inflation Expectations: The Implications of Inflation’s Simple Dynamics,” Stephen G. Cecchetti, Michael E.  Feroli, Peter Hooper, Anil K Kashyap, Kermit L. Schoenholtz. U.S. Monetary Policy Forum, March 2017. 6 And if inflation were at 3%, a flat Phillips curve would mean that it might take a severe downturn to get inflation back to the 2% objective.

not much evidence of nonlinearity in the Phillips curve, but it might be hard to find because of the limited historical experience with very low levels of the unemployment rate.7 

On the other hand, there is evidence of nonlinearity in a Phillips curve for wage inflation.8 Of course, the  Yellen Phillips curve does not assign a causal relationship between wage inflation and price inflation—and we see that as the case in conventional Phillips curves in general—but some passthrough from wage inflation to price inflation does seem intuitive. Indeed, in a recent interview, Dudley said his expectation that inflation will  move to 2% is underpinned by his expectation that wage inflation will firm and pass through to higher price  inflation: “Inflation [is] somewhat below our objective — but we do expect as the labor market continues to  tighten to see firmer wage gains and that will ultimately filter into inflation moving up towards our 2%  objective.” We should add here that it is possible that the slope of even a linear Phillips curve may not really be as low as Yellen estimated, also improving the ability of monetary policymakers to achieve their inflation objective. 

Rethinking the NAIRU 

The NAIRU is an unobservable variable and therefore must be estimated. If it is stable, then movements in the unemployment rate capture changes in labor market slack without any ambiguity. But it is not always stable. Indeed, participants’ median estimate of the NAIRU has been trending lower over the last several years. We believe participants’ estimates of the NAIRU have declined as a result of the long period of lower 

than-expected inflation (see Inflation Surprises and the NAIRU).  

Lower-than-expected inflation is viewed as suggesting that there is more slack than reflected in the prevailing gap between the unemployment rate and the estimated NAIRU. For a given level of the unemployment rate,  a downward revision of the estimated NAIRU implies a greater degree of slack, therefore realigning the Phillips curve with the recent experience. The fact that the NAIRU is not stable and that we keep revising how much  

7 For example, a recent Federal Reserve Board staff paper finds evidence of a nonlinearity using metropolitan-level data,  though not in the aggregate data. The economic significance of the nonlinearity is small, however. See Babb, Nathan R.,  and Alan K. Detmeister (2017). “Nonlinearities in the Phillips Curve for the United States: Evidence Using Metropolitan  Data,” Finance and Economics Discussion Series 2017-070. Washington: Board of Governors of the Federal Reserve  System,

8 For example, see Fisher, Richard, and Evan Koenig (2014) “Are We There Yet? Assessing Progress Toward Full  Employment and Price Stability” Dallas Fed Economic Letter. Volume 9, Issue 13, October 2014; and Nalewaik, Jeremy  (2016). “Non-Linear Phillips Curves with Inflation Regime-Switching,” Finance and Economics Discussion Series 2016- 078. Washington: Board of Governors of the Federal Reserve System, 

slack remains undermine the reliability of the Phillips curve for inflation forecasts and hence as a guide for policymakers. 

Furthermore, if a ½-percentage-point change in the unemployment rate does not even change inflation by a  tenth—consistent with the Phillips curve that Yellen wrote down—how would one know that an unemployment rate ½ percentage point above or below the estimated NAIRU was not, in fact, the NAIRU? In this case, the NAIRU might better be thought of as a range, for example, one percentage point, encompassing  ½ percentage point on either side of the point estimate, or even wider! In terms of monetary policy,  policymakers should be very cautious in responding to movements of the unemployment rate within this range, then get more aggressive as it moves further and further away from the point estimate of the NAIRU. 

How Stable are Inflation Expectations? 

Another question that is much discussed by FOMC participants is how well anchored longer-run inflation expectations are at a level consistent with the FOMC’s longer-run inflation objective. This is complicated by the fact that there are many measures of longer-run inflation expectations, and they often send different signals.9 Currently, those who are concerned that inflation expectations are not firmly anchored at the inflation objective favor a slower pace of hikes to better ensure a quicker and more-assured rise in inflation to 2%, to help stabilize inflation expectations in the longer-run objective. The dominant view is that inflation expectations are sufficiently anchored at 2% and that this concern doesn’t stand in the way of rate hikes in the near term. But all participants agree, as reflected in the July minutes, that inflation expectations should be monitored carefully in light of the recent slowing in actual inflation, and that further signs of erosion should weigh on the pace of hikes. Indeed, if core inflation remains as subdued as it has been recently, the concern about such erosion in longer-term inflation expectations will move to the forefront for monetary policymakers. 

Is There Really a Stable Tradeoff Today? 

It is often said that the Phillips curve is vertical in the long run, meaning that there is no stable trade-off between inflation and unemployment over the longer run. In this case, if the unemployment rate is below the  NAIRU, inflation increases. If inflation expectations eventually respond by rising, then inflation will continue to increase as long as the unemployment rate remains below the NAIRU. This is called the accelerationist property of the long-run Phillips curve. The vertical long-run Phillips curve means there is only one unemployment rate consistent with stable inflation: the NAIRU! 

But is this really true over a period of a few years, for example over a three-year FOMC policy horizon?  Apparently not. Core PCE inflation has been below 2% for almost a decade and the Committee still sees longer-run inflation expectations as reasonably stable. In the context of policy today, monetary policymakers face a stable tradeoff: They can lower the unemployment rate below the NAIRU with only, at worst, a  temporary increase in inflation. And with such a flat Phillips curve, perhaps not even that! 

The Costs and Benefits of Undershooting the NAIRU 

We suspect that the flatness of the Phillips curve today is one reason why some FOMC participants believe  that policy should remain stimulative for longer to support a further decline in the unemployment rate, to  

9 Three measures get the most attention from participants: TIPS-implied five-year five-year forward break-even inflation  (BEI) rate; the Michigan survey of consumers median expectation of inflation five to ten years out; and the median 10- year PCE inflation forecast from the Philadelphia Fed’s Survey of Professional Forecasters (SPF). There are questions about each, but in practice, policymakers want to see confirmation in multiple measures, with a particular focus on survey measures. For Phillips curve models, the SPF measure, which has remained around 2%, is most often used as the measure of longer-term expected inflation.

further below the NAIRU. This is reflected in the median projections of participants as well as the public remarks of some participants. 

There are two reasons why an unemployment rate below the NAIRU might be appropriate, and the flatness of the Phillips curve is extremely relevant to both. The first is to bring about a faster and more assured return to price stability. That is less likely to be successful the flatter the Phillips curve is or at least would take a  much sharper decline in the unemployment rate than implied by participants’ median projection. The second reason is that a flatter Phillips curve makes the tradeoff between inflation and the unemployment rate more favorable. That is, there is a smaller cost in terms of higher inflation of lowering the unemployment rate below the NAIRU. And there may be longer-term benefits, some say, if a tight labor market encourages some workers to return to the labor force or if more workers who are involuntarily working only part-time find full-time work.  With such a favorable trade-off, this strategy seems more compelling. 

Why Has Inflation Been Lower than Expected? A Price Level Shock? 

Today the most pressing question concerning hiking rates for a third time this year, in December, is how to explain the recent unexpected and sharp slowing in the 12-month core PCE inflation rate, from a cyclical high of 1.9% during most of the period from August 2016 through February 2017—suggesting the FOMC  had virtually met its inflation objective—to 1.5% in June 2017. The question confronting the FOMC today is whether this slowing in core inflation reflects a transitory decline in core inflation relative to its underlying rate or the underlying inflation rate is lower than earlier indicated. Giving more weight to the latter view would shed greater doubt on the prudence of a rate hike in December.

The Chair has specifically emphasized the transitory effect of a one-time decline in the price of wireless communication services, and she has also mentioned prescription drug prices. If this effect is indeed a one time shock to the price level, then inflation will be lower only temporarily, though the effect on year-over-year rates, which the FOMC generally prefers to look at, will be more persistent, lasting until the month(s)  where the price level shock occurred moves out of the 12-month window. Note that the arithmetic contribution of the decline in the price level of wireless communications was to lower the 12-month core PCE  inflation rate by a couple of tenths, perhaps suggesting that the underlying inflation rate is little changed. But nearly all participants are at least somewhat uneasy about attributing the sharply slower inflation rate entirely to transitory price level shocks. Even Yellen said that this price level shock only “partly” explains the  slowdown in core inflation and that “there may be more going on.” 

Why Has Inflation Been Lower than Expected? An Inflation Shock? 

Evans recently wondered whether the persistently lower-than-expected inflation could reflect, in part, more structural forces, including the effect of technology, for example, price competition unleashed by opportunities to purchase goods on the internet. Dudley recently made the same point about globalization. How does such a more persistent, secular process affect inflation dynamics and what are the implications for monetary policymakers?

Without incorporating this term, the equation would systematically overpredict inflation, leading to persistently negative residuals. How should policymakers respond to a shock like this? Just as they have been, by revising down the NAIRU in response to systematic over-prediction of inflation! So this doesn’t mean that the FOMC has to tolerate inflation systematically below 2%. The FOMC has to work harder to achieve its inflation objective by offsetting the inflation shock with a more stimulative policy that drives the unemployment rate even lower relative to the estimated NAIRU. Dudley emphasized this in a recent interview:  “If there are these secular forces that are pushing inflation lower, perhaps we can actually go to a somewhat  lower unemployment rate.” Following that logic, we can rewrite equation (3) so that the inflation shock is  incorporated into the slack term:

The Phillips Curve and Monetary Policy Today 

So we end with the question of whether the FOMC is on the way toward meeting its 2% objective. If we assume it is a price level shock that has depressed inflation, then we are in good shape. Indeed, we might even say that the FOMC has achieved its dual mandate—with core PCE inflation virtually at 2% for more than six months before the recent softening and the economy at least at full employment, if not beyond. This shouldn’t be surprising, because the unemployment rate has fallen by more than five percentage points, inertia is working through, shocks other than the most recent one have faded, and inflation expectations are reasonably stable at the inflation objective. But the softer monthly readings for core inflation have persisted beyond the time of the March price level shock. So, as Dudley has said, and as is widely agreed, the jury is still out on the source of the slower inflation, putting a December hike more in doubt. While we do expect a  firming in monthly readings for core inflation that would be sufficient for a December hike, the probability we assign to a December hike has declined with each new soft monthly reading, to closer to 50%. 

10 This exercise is simply illustrative. We disregard here all issues relating to an estimation that these modifications would present.

Like this article?

Share on linkedin
Share on Linkdin
Share on facebook
Share on Facebook
Share on twitter
Share on Twitter
Share on pinterest
Share on Pinterest

Our Newest Content

Would you like a preview of our newest, members-only commentaries? Simply click the button below to see some of our newest commentaries and request a free trial today.


Please fill out the form below for a trial of our services.