It's baaack! The Phillips Curve Conversation

The Phillips Curve is back in the news with a vengeance.  And, I suspect it’ll be an even hotter topic in the coming months as everyone watches the Federal Reserve (Fed), European Central Bank (ECB) and other central banks to see if or when they slow rate hikes or even lower them again.

So I thought I’d take a minute to explain it and help put it in the right context.

What is the Phillips Curve?

The Phillips Curve is the statistical correlation between inflation and unemployment in an economy.  It was discovered by A.W. Phillips and published in 1958.  He was originally looking at wages inflation and unemployment.  Today, we look at overall inflation and unemployment.  In any case, here is his original graph:

That’s it. You can find that data for most economies around the world over a wide range of time horizons.

Why There Should Be a Phillips Curve

The reasons for a Phillips Curve are strong and straightforward.   To see it, we need to think about pizzas.

Friday is pizza night at our house.  If I suddenly doubled my pizza order on a random Friday evening, it’s no problem. They just make double for me.  Even if I decide to go on a binge and order that same amount every night for a week, it’s not an issue.

If all their customers double their orders on a Friday, it’s likely an issue. Tables run out, there are lines, service is understaffed, and they might run out of some items on the menu. If all their customers double orders every night, they will definitely start to run out of some supplies at first.

If the restaurant believes this strong demand will stay, they might raise their prices, they might not. It turns out that a single business in a relatively competitive market like that for pizzas is not eager to change prices too much for fear of losing customers. But they will have to reorder and now they too are doubling their orders for inputs like sauce, dough, cheese, peperoni and so on.

If all the customers in my town doubled their demand for pizzas at all the pizzerias we would likely see some additional effects. We might or might not see local pizza places raise their prices initially[1].  But each pizzeria would be buying more and more inputs since this is likely enough demand to push up prices on those inputs[2]. That’s especially true is pizzerias buy produce locally, for example. 

Additionally, we would soon see every local pizzeria trying to convince its employees to work over time and then paying overtime.  If the demand surge lasts and restaurants believe it will last longer, then they will also start hiring more employees.  At some point, the current staff can’t just work overtime all the time and the cost is much higher to pay people overtime than to hire new people as the business grows.  Now their costs will be rising for sure and their margins will shrink and they’ll have to consider raising prices.

One of two things happens at this point. Either the restaurants start to raise prices in earnest or new restaurants enter the market, or both.

How long does it take to open a restaurant?  Let’s say, at least a year.  So, one typical scenario would be that prices start to rise for pizzas in town. Some non-pizzerias begin offering more pizzas to capture some of the demand (this too is “a new competitor entering the market”), some labor starts entering the town to get the higher-paying service jobs (higher tips, consistent hours, more overtime) and some new restaurants go under construction.

A year later, the new restaurants are opening, there are more workers and now the supply of pizzas has increased to meet the sustained higher demand.  And, when that happens, wages settle a bit and the increased competition in the local pizza market means pizza prices likely fall some.  This is a simplified version of the story, but enough to make my points.

This process would have generated data for a Phillips Curve in this town.  First, unemployment would have fallen as restaurants move from just overtime to hiring more workers. Second, as costs rise for the businesses and, critically, as businesses come to believe the increase in demand is sustained, they raise prices.  Economic data would therefore show lower unemployment and a higher inflation (higher pizza prices) in this town.

This process would have unfolded slowly over that year.  We can imagine the following: Employment rises first (unemployment falls), then prices (inflation) starts to rise.  Inflation might then rise for a while as other restaurants become convinced the higher demand is here to stay.  Then unemployment falls further as pizza production expands and so on.

Then it will all peak and begin moving in the other direction (after a year or so). Unemployment could stay low, but if other people moved into town for the good pizza jobs, or to enjoy the abundance of good pizza itself, unemployment will slowly return toward the “natural rate of unemployment”[3] for this town.  It could be a little lower than it used to be, but it’ll be higher than it was during the early boom-growth period for sure.  And prices will eventually stabilize.  That means inflation will slow and be zero again (“stable prices” means they aren’t changing much, so inflation is zero).

That’s the normal market process.  And it should generate a local Phillips Curve.

This was just one town and one market (pizza).  When the same thing happens in the economy overall, it raises demand for all goods and services and generates the same effects. Higher aggregate demand raises all prices on average (called “the price level”) and lowers unemployment in general. And that generates an aggregate Phillips Curve.

The Phillips Curve in the USA

Notice that the Phillips Curve is a “short-run” phenomenon in the sense that it would exist for about 1-2 years in my pizza example. After that, everything is stable again. Inflation would bounce around a bit, as would unemployment, but they wouldn’t necessarily be connected in any way.  So, inflation and unemployment will likely look random during stable times.  You need a sustained increase in aggregate demand for the above “pizza boom process” to play out and to generate a clear Phillips Curve in data.

There are a few periods in US history that make for very clear Phillips Curves.  I’ve chosen the 1960s to show a classic curve. I’ll discuss other periods and further issues in another column. I include the Phillip Curves for the 1990s and 2010s at the end of this column.

There’s a very clear Phillips curve in the 1960s. You can understand why everyone was excited by his 1958 publication. Phillips found it in the UK for 1861-1913 and now everyone could find it in every country for almost any time period they looked.

It follows logically from a basic understanding of economics and should hold at the micro and the macro level, as the pizza story hopefully illustrates.  As a result, nearly all theoretical macroeconomic models will also generate a Phillips Curve when you simulate them. By all measures, this is a pretty fundamental curve.

The Policy Discussion in 2023

The Phillips Curve is back in the news because every central bank in the world is struggling to balance unemployment with inflation.  (See my “Less Liquid New Year”[4] for a recent discussion.)  They are all raising interest rates to fight inflation but they know this will generate a recession and that will generate higher unemployment.

In other words, they are pushing us from the upper left of the Phillips Curve (high inflation, low unemployment) down and to the right (low inflation, high unemployment).  And everyone wonders the same thing: how much unemployment is “enough” for the central bankers to be satisfied they lowered inflation? 

If the curve looks like it did in the 1960s, people reason, then maybe 2% inflation can be achieved with 4-5% unemployment. Today, US unemployment is around 3.7%, so many observers reason that when it rises another percentage point, the US Fed will stop raising interest rates.  There’s also a belief that the US’s “natural rate of unemployment” is around 4-5%. So, all that makes sense and sounds pretty good to market participants and talking heads on TV.

This thinking – and Phillips Curve reasoning – can be seen in US Fed Chair Powell’s December 5, 2022, press conference[5] response to a question from Howard Schneider at Reuters. Howard asked whether the new interest rate hike of .5% was likely to generate a recession.  Powell’s answer was that they currently project slower growth, but not a recession.  Then he said,

“in that condition, labor market conditions are softening a bit. Unemployment does go up a bit. I would say that many analysts believe that the natural rate of unemployment is actually elevated at this moment. So it's not clear that those forecasts or inflation are really much above the natural rate of unemployment. We can never identify its location with great precision. But that 4.7 percent [“natural rate of unemployment”] is still a strong labor market.” 

Similar statements and discussions can be found in the ECB’s and Bank of England’s press conferences too.

The concern is this, however. Monetary policy works through “long and varied lags” meaning that, empirically speaking, increasing the interest rates today might not have an effect for 6-18 months.  The worry then is that we raised rates enough in, say, October 2022 to get us to that 2% inflation and 4.5% unemployment rate in March or April 2023. We just don’t know it yet due to the lags.

The Fed kept raising rates after October and that means in the Spring, as unemployment starts to rise in response to rate hikes last summer even, unemployment will hit 4%, 5% and keep going to 7% as the rate increases from Fall and even Winter continue to have their lagged effects.  That would be an “overshoot” and a serious recession, what they are calling a “hard landing”.

The Problem with the Phillips Curve: Causation vs Correlation

The problem with the Phillips Curve - and Powell knows this and is careful in his language - is that it is just a statistical correlation at a specific period in time.  There is nothing in our pizza story that tells us that “higher unemployment causes inflation to fall” or that “lower inflation causes unemployment”.  Both are the result of a common driver: demand and supply conditions.  Specifically, as demand increases, prices (inflation) rise and unemployment falls, then as supply increases prices stabilize (inflation falls) and unemployment rises back to normal.

The relationship between inflation and unemployment in an economy is a statistical correlation, nothing more. It is not a causal relationship.

The bad Phillips Curve reasoning – and I hear this a lot from financial commentators in the news – is to think that the Fed can pick points on the Phillips curve and target them.  You often hear things like this, “well, historically we have to push unemployment above 5% to get inflation back to the 2% range so if the Fed can push unemployment to 6% we will get below 2% and the Fed can ease interest rate pressure…”.  Wrong. The Fed doesn’t control unemployment and use it to drive inflation.

The Phillips Curve is just a statistical relationship and not a menu of policy options we can just choose between.  And, the next time you hear “historically the Phillips Curve tells us…” just ignore what follows and think of the next paragraph and pictures I’m about to share of US Phillips Curves.

Recall that in relatively normal times (after the pizza boom settles, in our story), you don’t even see a relationship as both inflation and unemployment bounce around randomly.  Here’s what the Phillips Curve looks like for the 1990s and the 2010s, both periods of relative economic calm.

I tried to run a trend line through both of those and, if anything, the Phillips Curve slopes upward!  It is just a statistical relationship.  Powell knows this and is careful. That’s why he says we don’t truly know the natural rate of unemployment that the economy “should” have in normal times. And he ends that same quote from above with this,

“you know, there are channels through which the labor market can come back into balance with relatively modest increases in unemployment, we believe. None of that is guaranteed, but that is what their forecasts reflects.”[5] 

So, he’s saying that its possible we get inflation down without much change in unemployment at all.

The Fed is betting that they need to raise interest rates to fight inflation.  They know unemployment will rise some because higher interest rates contract demand. But they know that any precise forecast in the unemployment-inflation relationship is dangerous. 

For today, I wanted everyone to understand the discussion, the Phillips Curve, and understand how to correctly read it. It is a statistical correlation for a specific period in time. That’s all.  It is important but it is not causal.  There is no sense in which the Fed can just decide to raise unemployment by 1% and that will, itself, lower inflation by x%.  Or vice versa. In the short run, the Fed can change interest rates or liquidity. That will encourage or discourage demand in aggregate. The change in demand should influence both prices and unemployment as well.

It’s a subtle but important distinction that matters in understanding monetary policy and the discussions around it. You’re going to hear more about the Phillips Curve in the news so I hope this helps.

---

[1]  Although, I would not be surprised to see that people wait to get tables and need to tip more to get the same service they always expect. Tipping more for the same food and same service you got when the restaurant was less busy is indeed a price increase to an economist as is the time you spend waiting to be seated.

[2]   Similarly, the time spent ordering, dealing with suppliers, stocking and re-stocking shelves are also all an “increase in the price” of those goods.

[3]   This is the concept of the “natural rate of unemployment”. It’s the unemployment rate a stable, steady economy will have due to the natural churn of life. People graduate from college, look for and eventually find.  They are unemployed during their job search.  People change careers. Business close and new ones open. All these things lead to people switching jobs, retiring and being replaced, etc. and generate a stead rate of employment when everything is running “naturally”. And it tends to be short-term unemployment that we actually want in a free society.  If you have governmental programs or a culture that encourage people not to work, that can raise an economy’s natural rate.  So the natural rate is not fixed, it evolves overtime with culture, laws, etc.

[4]  

A Less Liquid New Year

[5]   Powell’s December 14th press conference. My quotes from pages 8 and 9. https://www.federalreserve.gov/mediacenter/files/FOMCpresconf20221214.pdf