The Phillips Curve says that there is an inverse relation between unemployment and inflation. Low unemployment is correlated with a rise in inflation. It’s an article of faith to economists of all stripes. It’s listed in the popular introductory economics textbook by N. Gregory Mankiw as one of the . It’s especially loved by the Fed, which raises or lowers interest rates depending in part on its predictions. Its critics point out that its predictions are poor.
In this post, I discuss the derivation of the Phillips Curve, its adaption by Samuleson and Solow to manage the economy, its breakdown in the 1970s, exploitation by neoliberals of that breakdown to replace Keynesian demand-based economics with monetarism and supply-side economics, its rejuvenation, and the evidence that it doesn’t make accurate predictions.
I conclude with some observations based on an important paper by Simcha Barkai that challenges the core beliefs of neoliberalism. It suggests we can raise wages substantially without causing inflation by lowering corporate profits.
This part is based on Sections I-III of Robert Gordon’s article, The History of the Phillips Curve: Consensus and Bifurcation, Economica (2011) 78, 10–50 (behind paywall, but you can find it online at your local library). Gordon is an economics professor at Northwestern and has worked on the Phillips Curve for decades.
William Phillips published a paper in 1958 showing a correlation between wage growth and inflation in the UK between 1861 and 1913. He fitted a curve to the data, and then compared that curve to UK data from two later periods. There was a remarkable similarity for most of the two periods, with exceptions Phillips explains away. Here is the curve Phillips derived:
1. wt = -.90 + 9.64U-1.39
Gordon says that “… the inflation rate would be expected to equal the growth rate of wages minus the long-term growth rate of productivity.” P. 12.
1a. p = w – k
Here p is inflation, w is wage growth, and k is productivity growth. Generally in these equations, lower case letters are rates of change and upper case letters are levels. We can substitute Equation 1a into Equation 1 to get the original Phillips Curve.
2. p = -.90 + 9.64U-1.39 – k.
Paul Samuelson and Robert Solow picked up on the Phillips paper with a paper of their own in 1960. Gordon says much of their paper is a discussion of pre-Phillips theory. They can’t find data on the US economy similar to that found by Phillips for the UK economy, so they work up some of their own data and make some calculations showing a result similar to that of Phillips. Whereas Phillips does not mention the possibility that the curve might shift, Samuelson and Solow find such shifts and offer possible explanations, such as strong labor unions.
Here’s a schematic drawing of the Phillips Curve from :
The standard curve might be the one on the left. It shows very high inflation at very low unemployment, but falls quickly as unemployment rises. That suggested to Samuelson and Solow that there is a trade-off if the economy is in specific parts of the Phillips Curve: by allowing a slightly higher level of inflation, you could get a big drop in unemployment. The US tried this idea in the 1960s. This policy was tied to Keynesianism, which was the predominate theory in the Kennedy and Johnson era, and into the Nixon Presidency.
When OPEC massively increased the price of oil in the early 70s, inflation soared far past the level suggested by the Phillips Curve. The neoliberals at the University of Chicago argued that the failure of the Phillips Curve proved that Keynesian economics was worthless, and pushed their solution: monetarism. They also had a formula to replace the Phillips Curve as a predictor of inflation.
Their explanation for the failure of the trade-off was something like this. Suppose the beginning rates of inflation and unemployment are at Point A on the above chart. The Fed lowers interest rates resulting in a small increase of inflation, so that the economy moves to Point B with lower unemployment. People believe that is unsustainable, and that the economy will revert to the natural rate of unemployment, the vertical line. As a result, the Phillips Curve shifts up and to the right over time, so that the economy moves to Point C, with the beginning unemployment rate but higher inflation.
The neoliberals won. Keynesianism lost out and was replaced by monetarism. This was probably not deserved, according to Gordon. He says that Samuelson and Solow were not talking about the situation that came about in the 70s, but rather the situation in the early 1960s. He also spends a good part of his paper showing that the formulas offered by Friedman and the neoliberals for predicting inflation were a total failure both on factual and theoretical grounds.
Gordon himself proposed a version of the Phillips Curve designed to deal with the problem of supply and demand shocks like the Oil Shock:
3. p = Ept + b(Ut – UtN) + zt + et
In Equation 3, the second U term is the natural rate of unemployment, zt represents cost-push pressure, and et is apparently a constant. The natural rate of unemployment and the z term vary over time, and for some reason so does the e term. There is nothing left of the wage term. The Phillips Curve is now free from the bonds of factual data that gave Phillips his interesting result. It’s a curve-fitting exercise, using economic theories put together in a way that fits the data. It’s a complicated formula in which every term needs to be calculated from some other theory or data.
Gordon says that Equation 3 is the canonical version of the Phillips Curve. It is incorporated in most econometric models, modified by some other variables and terms, including levels of taxation, expectations of inflation, inflation inertia, which relates to price and wage rigidity in the short run, and a host of other terms. But as we shall see, it doesn’t work as a predictor.
Criticism of the Phillips Curve
The Phillips Curve has been controversial for a long time, as Mankiw in his introductory textbook. Ben Leubsdorf wrote a very readable criticism for the Wall Street Journal on August 14, 2014, just before the Fed started raising interest rates. His title is The Fed Has a Theory. Trouble is the Proof is Patchy (sadly behind a paywall; it’s available online at your local library).
Leubsdorf confirms that most economists believe that there is a short run trade-off between inflation and unemployment and also agree that this trade-off doesn’t hold in the long term, meaning that we can’t get permanently lower unemployment by accepting a bit more inflation. But the problem is that there is also no apparent connection in the short run either. Here’s a chart originally in Leubsdorf’s article and reprinted in a post discussing the article by .
Source: Wall Street Journal
To read this chart, select an expansion period from the list on the upper right, find the line that color, and locate the circle at one end of the line; that’s the starting point. Then follow the line to see how the relationship between unemployment (x-axis) and inflation (y-axis) changes over time. As you can see there is no apparent connection in any except the first expansion. The lack of connection to theory is especially obvious in the current expansion.
The Leubsdorf article has several quotes from Very Serious People to the effect we think there’s a relationship and we’re going to act like there is a relationship, and we can fine-tune the economy with our gut instincts.
It doesn’t look like the latest study will change minds either. That one comes from the Philadelphia Fed in August 2017, Do Phillips Curves Conditionally Help to Forecast Inflation? The conclusion is that the Phillips Curve does worse than something called a univariate model which I won’t discuss.
In this September 26 article there are more Very Serious People explaining they need to follow their instincts about the economy in deciding on interest rates and they are sure inflation is coming. Meanwhile, the economy continues to add jobs with no obvious increase in inflation as shown by the blue line on the above chart. Inflation is .
Damage From the Phillips Curve
We have already seen that the first notable failure of the Phillips Curve was used to undermine Keynesian economics in favor of monetarism. As a result, working people of all classes were doubly harmed, first by the abandonment of the Fed of any significant role in cutting unemployment, and second by the savage use of high rates to control inflation.
Here’s a chart showing the labor share in gross national product on the left axis (blue line) and the prime rate on the right axis (red line).
The grey bars are recessions. This chart shows that up to 2000 every time workers start to get a bigger share of the GDP, the Fed raises interest rates. The Phillips Curve was the justification for those rate hikes. There is some evidence wages are firming up today, and maybe even rising a fraction faster than inflation. Following tradition but not evidence, the Fed is raising rates. If that hurts workers, also in accordance with tradition, that’s just too bad.
Take another look at Equation 1a. If we set inflation at zero, Equation 1 says that wage growth equals productivity growth. That’s not true. Here’s a chart from the .
The wage line is for production and non-supervisory personnel, which the EPI says is about 80% of employed people. The average wage for all workers has grown somewhat faster, but is still well under the rate of increase of productivity over the long term.
Money produced in the economy goes either to capital or labor. So, the excess gains from productivity must be going to capital.
Actually, it seems strange to suggest that none of the gains from increased productivity go to capital, as Equation 1a does. Consider a company like Google. It can buy a few more computer blades and serve more customers with little or no increase in total wages. The gains from the productivity of the new capital all go to the company. Or consider a company that outsources its labor. Some of the gains might be used to cut prices, I suppose, but surely most of the gain stays with the company.
The following chart shows the sudden growth in top wealth. It demonstrates that the growth began at the same time as the productivity-wages gap began, more support for the idea that the gains from productivity are going to capital.
Capital can take many forms. It could be plant and equipment, commercial or agricultural land, personal residences, art, gold, and many other things believed to store value, whether or not they are actually producing anything, or even whether they actually store value. We know that top wealth is rising, the stock market is up, and the value of residential real property in all major cities is rising. All these and more suggest that the total amount of capital is increasing. All that increase is funded by the gains from productivity.
A recent paper by Simcha Barkai, , provides a convincing explanation. The labor share is declining he says. But so is what he calls the “capital share”, a defined term, calculated by multiplying the “required rate of return” by the capital stock deployed in the non-financial business sector. Capital stock includes plant and equipment, land, and intangibles such as patents and software, less depreciation. The required return on capital is approximately and sensibly defined as the cost of obtaining capital in the financial markets. He shows that the cost of capital has declined by 7% over the period of his study, 1984-2014. If the amount of capital deployed had increased as might be expected with this large drop in cost, the capital share might have remained the same. Instead, businesses did not deploy additional capital, and the capital share declined by some 30% over the period. During that period the labor share declined 10% from a larger starting point.
The combined losses were more than made up by increases in the profits share. Profits add to the value of the firm, and are distributed by the owners of firms as they see fit, which isn’t to lowly workers. This is from Barkai’s paper:
Across specifications, the profit share (equal to the ratio of profits to gross value added) has increased by more than 12 percentage points. To offer a sense of magnitude, the value of this increase in profits amounts to over $1.1 trillion in 2014, or $14 thousand for each of the approximately 81 million employees of the non-financial corporate sector. P. 3.
Barkai attributes this almost entirely to increased concentration of US industries, and most of the paper is devoted to proof of that conclusion. He links that increase in concentration to changes in anti-trust law and policy engineered by Robert Bork when he was at the University of Chicago.
Following Barkai, we should rewrite Equation 1a like this:
1b. p = w + γ + ct – k
where γ is the rate of growth of the profits share, ct is the rate of growth of the capital share, w is the rate of growth of wages, p is inflation, and k is productivity. Substituting the original Phillips equation, Equation 1 into Equation 1b gives us
4. p = -.90 + 9.64U-1.39 + γ + ct– k.
This equation calls attention to the role that profits play in the economy, something economists generally generally ignore. When people do discuss profits, it’s always in the context of the importance of capital and the need to coddle it. That view lies at the heart of neoliberalism, and at the heart of Fed policy. It is also at the heart of the Law and Economics movement also spawned at the University of Chicago, a movement that has changed the legal system to favor capital. If neoliberals were intellectually honest, they wouldn’t call it supply side economics, they’d call it philo-capital economics.
Equation 1 has been replaced by Equation 3 in the standard model of the Phillips Curve.
3. p = Ept + b(Ut – UtN) + zt + et
Making this work with Barkai’s analysis is harder. We get a clue from Gordon’s explanation of the z term: he call it cost push, meaning price shocks caused by labor unions and “bauxite barons”. This is where capital growth fits in. The ability to control markets gives firms the ability to cause price shocks, as when pharmaceutical companies drive up the price of epi-pens or other drugs, but also the ability to gradually increase prices above the rate of inflation. Therefore, I’d rewrite Equation 3 this way:
5. 3. p = Ept + b(Ut – UtN) + γ + ct + et
Gordon doesn’t explain the e term, so we’ll just let that pick up anything that used to be in the z term that is somehow missed by my addition. It would, for example, include demand-pull inflation, which hasn’t been a problem for some time.
In the current situation, with profits at very high levels, we can easily increase wages without increasing inflation if the rich were willing to accept lower profits, subject to the availability of sufficient resources to meet the new levels of demand substantially higher wages might cause.
Barkai says just distributing the historically high profits to workers would give every working person (other than those in the financial sector) a $14K raise. That dwarfs the make-believe the Republicans promise from their proposed tax cuts.
Unfortunately, the Phillips Curve isn’t the only thing blocking action to help the average citizen.