The Yield Curve and the Next Recession

John Cochrane:

The graph shows the 10 year government bond rate and the overnight Federal Funds rate, which the Fed controls. Notice how an “inverted” spread — when the funds rate is higher than the 10 year rate — is one of the most reliable indicators of a recession to come.

Notice that the yield spread is tightening now. So should we worry? Well, notice also the many times when the yield spread tightened… and then sat there for many years of growth. The late 1960s, the late 1980s, and the late 1990s are good examples.

"Notice also the many times when the yield spread tightened… and then sat there for many years of growth. The late 1960s, the late 1980s, and the late 1990s..."

Damn right. And this time is like the 1990s.

But is the spread really tightening now, as Cochrane says? To my eye, no.

To my eye, on Cochrane's graph the blue line hit bottom in mid-2012 and has been gradually rising since. Or, worst case, hit bottom in mid-2012, skidded along the bottom until mid-2016, and has been rising since. To my eye, the 10-year rate is keeping up with the Federal Funds rate. But that's why we do graphs: because we can't always believe our eye.

So I reconstructed John Cochrane's graph and added H-P trend lines to it:

Graph #2

Source data as thin lines, trends as heavy lines. Blue for the 10-year rate, red for the Federal Funds rate. Here's a close-up:

Graph #3

That blue line shows the "skidding along the bottom from 2013 to 2016. John Bull still can't stand two percent, apparently.

But it does seem that since 2015 or 2016, the red line is gaining on the blue. The gap between the H-P trend lines is definitely getting smaller.

Yeah, but there are known issues, known "end point" issues with the Hodrick-Prescott calculation. Since that last low in early 2016, for example, the blue data is going up faster than the blue trend. So I still think the blue data is climbing as fast as the red data.

Graph #4

Looking more closely, the red trend starts climbing by January 2015, while the  blue trend is still drifting down; running flat but drifting down. And yet, the red data is slower to rise than the red trend all thru 2015 and 2016.

And somehow, it looks like the red line rising is encouraging the blue to rise. Which is what's supposed to happen, I think.

I can see the heavy red line gaining on the heavy blue. But it is still not clear to me that the thin red line is gaining on the thin blue. Not since 2016.


If interest rates are rising now, rather than falling or running flat, then we are into a different trend. And if we're into a different trend, then I don't want to blur trends together by using H-P calcs that consider the whole dataset from the 1950s to 2018.

I want to see a close-up of this stuff, since 2016. 2016, because I think that's when the different trend starts. I subtracted the FedFunds data from the 10-year rate, to get the yield spread. And I re-figured the H-P numbers starting in January 2016, to eliminate all the influence of past trends. Here's what I got:

Graph #5

Oh you're right, you're right, mea culpa you're right. The yield spread is falling. The red line, sloping down, shows that the gap between the two interest rates is narrowing. I couldn't see it before.

When that red line gets to zero, the yield curve inverts. And we'll probably have a recession a year or two after that. Okay, so when will the yield curve invert?

I put a linear trend on the data and extended it out to see when it gets to zero:

Graph #6

Looks like a recession is due some time after 2026.

A non-linear trend line could predict an earlier recession, and let's call that realistic. But this year or next year or 2020? I don't see it.

As always, this is not investment advice.


Another look at the yield spread, based on Cochrane's data: The 10-year rate less the Federal Funds rate, since 1990.

Graph #7

Around 1995 there was a sharp drop almost to zero, and then five good years before there was a recession.

Around 2006, there was a sharp drop down to below zero, and a nasty recession a couple years later.

There has been no sharp drop this time around.

I eyeballed-in some red lines to show the trend of highs and lows for the last few years. Turns out that those trend lines meet some time around 2020.

Something has to change by then. But that doesn't mean recession is in the cards.

Only fear restrains the economy now. Jitters, Cochrane called it.

Arthurian, 7 April 2016:

I predict a boom of "golden age" vigor, beginning in 2016 and lasting eight to ten years. It has already begun. In two years everyone will be predicting it.

John Cochrane predicts it, 26 May 2018:

The economy has finally recovered from the 2008 recession.

It has recovered, except for the jitters.

Sometimes little things

Civilian Labor Force Participation Rate:

Graph #1

For about as long as anybody can remember, descriptions of the "Civilian Labor Force Participation Rate" have focused on that massive increase ("baby boom" ... "women in the workforce") or that massive decline ("baby boom") or both.

Yeah, okay. I'm not entirely comfortable with the standard explanations. (If the baby boom began in 1946 and people retire at 65, the downtrend should not begin until around 2011 if the baby boom explains it.) But I'm almost never entirely comfortable with standard explanations, at least until I work them out for myself. But I don't want to talk about that.

I want to talk about this:

Graph #2

The Participation Rate bottomed out in December 1954 at just over 58%. It peaked in January 1956 at just over 60%. Basically, there was an increase of two percentage points and the whole damn thing occurred in 1955.

The massive increase shown on Graph #1 begins around December 1964 and ends around July 1997. The increase was from about 58.5% to about 67%, for a total of 8.5 percentage points over a 31½-year period. That's a little over one quarter of one percent per year. Say instead that it ends in 1990, and it still only comes to one third of one percent each year.

The increase in 1955 was two percentage points in one year. At that rate, you could do the whole 1964-1997 increase in just four years. The rapid growth of the Participation Rate occurred in 1955. It just didn't last long.

But that makes it all the more interesting. What was the reason for that sudden, sharp increase, when the trend was otherwise down from the early 1950s to the early 1960s, down except for 1955.

The reason? I don't know. I couldn't find anything on the internet except baby boom and women in the workforce.

So let me ask a different question: Does the 1955 increase in the Civilian Labor Force Participation Rate have anything to do with the inflation of 1955-1957? I recall what James Forder wrote:

The question [Samuelson and Solow] were addressing was that of the explanation of the inflation of the 1950s – particularly the period 1955-57 – and the implications it had for macroeconomics. Mild though that was later to seem, this 'creeping inflation' as it was called was, at the time, a source of much anxiety.

Maybe the sudden jump in labor force participation caused the inflation?

Similarities in Wage Growth: The 1990-91 Recession and the Great Recession

Mark Thoma links to the Kansas City Fed's Nominal Wage Rigidities and the Future Path of Wage Growth by José Mustre-del-Río and Emily Pollard. Here's the opening:

Although unemployment and other measures of labor underutilization have returned to their pre-crisis levels, wage growth has remained modest since the Great Recession. The modest pace of wage growth since the end of the Great Recession is at odds with its behavior during the previous recession, when wage growth rebounded more quickly and sharply. Chart 1 shows the year-over-year percentage change in average hourly earnings of production and nonsupervisory workers. By late 2005, roughly four years after the end of the 2001 recession, year-over-year wage growth had surpassed 3 percent, and it reached 4 percent shortly thereafter. In contrast, nearly nine years after the end of the Great Recession, year-over-year wage growth has still not reached 3 percent.

They identify their data and the units in which they display it, so the graph was easy to duplicate:

Graph #2: Average Hourly Earnings of Production and Nonsupervisory Employees: Total Private

The KC Fed article points out that after the 2001 recession, it took "roughly four years" for year-over-year wage growth to climb above 3 percent; but after the Great Recession, it has been nearly nine years now and "wage growth has still not reached 3 percent".

They assume, or maybe they want us to assume, that the time it takes to reach a 3% increase should be comparable for different recessions. Should be more comparable than nine years versus four. I don't think this time-comparability is a valid assumption. What about "long and variable lags" and all that?

Anyway, as I said near three years ago, the after-effects of the Great Recession follow the same pattern as that of the 1990-91 recession (note, not the 2001 recession) magnified by a time factor of 2.5 or so.

I find or expect to find similarity between the 1990-91 recession and the Great Recession in the following areas:

• The decline of accumulated debt, the increase of circulating money, the debt-per-dollar ratio, and household debt service data. (August 7, 2016)
• Productivity growth (August 21, 2016) (November 17, 2016)
• The ratio of private debt to public debt (February 11, 2017)
• Adequate growth of credit, with no excessive accumulation of debt for some time (March 27, 2017)
• Inflation (September 8, 2017) (September 13, 2017)
• Capacity Utilization (September 11, 2017)
• The recessions' "prolonged nature and the weakness of the subsequent recovery" (September 16, 2017)
• and also, now, nominal wage growth.

But up to now I have not looked into the comparability of nominal wage growth. Shall we?

Graph #3: Comparison of the 1990-91 Recession and the Great Recession

In red the graph shows the data from Graph #2, beginning in January 2006, a little before the start of the Great Recession. From January 2006, the line rises sharply to around 4.0%, then runs jiggy at the 4% level until some months after 2007-05 when it suddenly drops below the 4% level and runs jiggy at the lower level for a while. The small, sudden drop from the 4% level marks the start of the Great Recession.

Around 2008-09 (September) the red line appears to be climbing back toward 4.0%, but turns and makes a dramatic fall to near 2.5%. This sudden drop looks like it should mark the start of the recession; actually, the end of the recession occurs late in that decline, as the line falls below the 3% level.

In gray the graph shows the data from Graph #2 beginning in January 1990. I have delayed the gray data by 16 years so that we see it concurrent with the red (which begins in January 2006).

The gray rises quickly to the peak that marks the July 1990 start of the 1990-91 recession. Falling rapidly then, the gray looks similar to the red line's fall from 4% to 2.5%.

The recession ends just as the gray falls below 3%. Just like the red line! Perhaps it is no coincidence that the Kansas City Fed article considers how long it takes for the data to climb back up to the 3% level after a recession ends.

Both lines continue to fall after the recession ends, the red until 2012-10 and the gray until 2008-09 (September 1992).

The red line shows increase since the 2012-10 bottom. This improvement in average hourly earnings is concurrent with an increase in the gray line. But the gray increase is higher and runs for a longer time. The two increases peak together (around 2014-03), and neither line shows much increase thereafter. These observations seem to suggest the view that the recovery after the Great Recession was much weaker than the recovery after the 1990-91 recession. And they seem to suggest that growth is behind us and we now await the onset of the next recession.

Such views are widely held today, but I believe them to be incorrect. I think the after-effects of the Great Recession follow the same pattern as that of the 1990-91 recession, magnified by a factor of 2.5 or so. The magnification delays the timing of events. Everything is happening in slow motion this time around.

To get a feel for the slow-motion pace of events since around 2006, I took the gray line from Graph #3 and multiplied its time increment values by 2.5. Thus on Graph #3, the gray line shows a peak around 2014-03, but on Graph #4 that same peak is delayed until well after 2025:

Graph #4: Comparison of Recessions with the 1990-91 Recession in Slow Motion

To my eye there is great similarity between red and gray on this graph, far more than on Graph #3. The red drops much lower than the gray by 2011 or 2012, indicating greater severity of recession. But there is a striking similarity of timing, from the 2006 peak to the 2012 trough, clearly visible on this graph.

One effect of the severity of the Great Recession was to slow things down more than a normal recession. With the gray data slowed by a factor of 2.5, the declines hit bottom concurrently and the slow-motion effect becomes evident to the naked eye. No comparable similarity is visible on Graph #3.

From its low point, the red data bounces up to the 2.5% level, where it meets the gray data. I don't see that as coincidence. I see it as part of the similarity of these two recessions.

The recession's severity gives the red line some waviness while the gray runs flat at the 2.5% level. But the red persistently returns to 2.5% and the waviness dissipates over time.

That brings us to the present moment and the end of the red line.

What does the future hold? If the similarity of red and gray persists, average hourly earnings growth will show increase from now to 2026. And wouldn't that be something.


In 2016, John Taylor wrote:

Because the economy has grown from the start of this recovery at a pace no greater than the prerecession trend, it has left a vulnerable gap of unrealized potential that can and should be closed with faster economic growth. In several key ways the US economy resembles an economy at the bottom of a recession, ready for a restart, even though the unemployment rate has reached 5 percent.

In the last two years the unemployment rate has dropped even more, and there has been some talk of improved growth. Warnings of another recession continue to arise from those who are thinking in terms of Graph #3.

I agree with John Taylor that in 2016 our economy resembled "an economy at the bottom of a recession, ready for a restart". In the last two years, our economy has been preparing for that restart. As Graph #4 shows, it is now ready. Prepare to be astounded by the vigor.


// EDIT, 16 June 2018: See also my follow-up post: Recessions happen when the economy slows down. Recoveries happen when the economy speeds up.

Wage Inflation

The equation of exchange. You know it:

Central to monetarism is the equation MV = PQ. M is the money supply; V is velocity -- the number of times per year the average dollar is spent; P is prices of goods and services; and Q is quantity of goods and services.

Hey, only P and Q today. Price and quantity. P is the price level. Q is the quantity of output. If Q by itself is Real GDP, then P times Q is Nominal GDP.

But that doesn't only work for GDP. It works for things with varying prices: To convert from "nominal" to "real" you divide by price. To convert from "real" to "nominal" you multiply by price. To see the prices, you divide "nominal" by "real". Mind your P's and Q's.

I was looking at my list of FRED data on productivity, and suddenly realized it includes both the real and nominal versions of "compensation per hour".

I could divide the nominal by the real and see wage inflation. That might be interesting.

I'm leaving output out of it. I'm not considering productivity here, only varying wages. I'm using business sector data, nominal divided by real compensation per hour. I show it in blue. For comparison, in red I'm showing the CPI. Indexing make the two series equal at the start:

Graph #1: Wage Inflation (blue) and the CPI (red), both indexed to 1948-08-01

Red and blue run tight together until around 1980. Since the 1980s, then, wages fall behind prices.

I thought that was pretty interesting. Since around 1980, wages don't keep up with prices. Maybe that's why the Fed has had such a hard time reaching its 2% inflation target. And maybe it's why wage earners have such hard times.

Wondering by how much the wages have fallen behind, I rearranged this data to show wage inflation relative to CPI inflation. What I found was pretty surprising:

Graph #2: Wage Inflation as a Percent of Consumer Price Inflation

Again, indexed to make them equal at the start.

The blue line runs at 100.0% consistently from 1947Q1 to 1977Q4, then suddenly falls beginning in 1978Q1. That's pretty weird. It almost looks like they were using business sector compensation costs to figure the CPI, and then suddenly they weren't.

// See also: Kitov's warning

The consensus is dangerously wrong, Palley says

Following up on Palley's latest:

NIRP has quickly become a consensus policy within the economics establishment. This paper argues that consensus is dangerously wrong, resting on flawed theory and flawed policy assessment...

If the consensus is wrong, it means our solution is part of the problem.

If the consensus is dangerously wrong, it means the problem is bigger than we think.


"In a growing civilization," Arnold J. Toynbee wrote, "a challenge meets with a successful response which proceeds to generate another and a different challenge which meets with another successful response."

"There is no term to this process of growth" -- there is no end to it, he says -- "unless and until a challenge arises which the civilization in question fails to meet--a tragic event which means a cessation of growth and what we have called a breakdown. Here the correlative rhythm begins. The challenge has not been met, but it nonetheless continues to present itself. A second convulsive effort is made to meet it, and, if this succeeds, growth will of course be resumed."

The depressions and great recessions of the capitalist era are the "rhythmical challenge" that we have failed to meet. Our economic troubles will defeat us unless we rise to the challenge. And, by "they will defeat us" I mean civilization dies, as Toynbee said. It's on us.

People who speak of "the end of the American century", people who express concern about the rise of totalitarian governments, they see it coming.

The Breakdown

"The nature of the breakdown can be summed up in three points," Toynbee wrote:

• "a failure of creative power in the creative minority, which henceforth becomes a merely 'dominant' minority;"
• "an answering withdrawal of allegiance and mimesis on the part of the majority;"
• "a consequent loss of social unity in the society as a whole."

Losing its creativity, the creative minority becomes a merely dominant minority. Government becomes more oppressive. A common complaint today, government is more oppressive.

Everybody complains about government, these days. "Withdrawal of allegiance", that is. But we can't solve our problems by focusing on government. What's missing is the "creativity". To fix the problem, we need the right solution to the right problem. Changing the government doesn't necessarily do it. Toynbee thought creativity solves the problem.

By "creativity" Toynbee meant the problem-solving ability.

If it sounds like circular argument when I say it, then I'm not saying it right.

Where we are now is late in the process. Things now are so bad that we see government as the problem. But that's a cascade effect. When "a challenge meets with a successful response", civilization continues to grow. But when there is no successful response, there are consequences. These consequences we call problems.

They are problems. But they are consequences. They are "cascade effect" problems. They are results, and you can't solve results. You have to solve the problem. Identify and solve the problem.

We have to solve the original problem, the one that keeps coming back. "Here the correlative rhythm begins," Toynbee said. "The challenge has not been met, but it nonetheless continues to present itself."

What causes depressions and great recessions? Thomas Palley identifies the problem:

... the flawed model of growth, based on debt and asset price inflation, which has already done such harm.

It is a problem of our own making. That's what makes it so difficult to solve. We think of it as our solution.

Our solution is part of the problem, and the problem is bigger than we think.

Tom Palley. Exactly.

Not sure what I did to get on Thomas Palley's mailing list. Oh, I contacted him by email one time, I'm sure of that. But I don't remember what economic detail I focused on, or what I might have said about it. And I don't know whether Palley liked what I said (and put me on his mailing list because he liked what I said) or hated what I said (and put me on his mailing list as a form of harassment) or none of the above.

But here's what I do know: About once a month, maybe less frequently, I get an email from Thomas Palley that opens with the words "Dear Friends & Colleagues," and, for some reason, that opening is just exactly right. I'm not a colleague, so I must be a friend.

And the once-a-month thing is nice. It doesn't feel like I'm being harassed. It doesn't feel like spam. It is something that, quite frankly, I'm starting to look forward to. (And I am a total curmudgeon.)

Opening my email early this morning, this is part of what I found, from Palley:

My own website has a new PERI working paper titled “Negative interest rate policy (NIRP) and the fallacy of the natural rate of interest: Why NIRP may worsen Keynesian unemployment” which I hope is of interest.

Which I hope is of interest, he says. Such understatement! Palley's ego is so small that you can't even see it. I like that. Then this:

Abstract: NIRP has quickly become a consensus policy within the economics establishment. This paper argues that consensus is dangerously wrong, resting on flawed theory and flawed policy assessment. Regarding theory, NIRP draws on fallacious pre-Keynesian classical economic logic that asserts there is a natural rate of interest which can ensure full employment. That pre-Keynesian logic has been augmented by ZLB economics which claims the natural rate may be negative in times of severe demand shortage, so that policy must deliver it since the market cannot. In contrast, Keynes argued investment could become saturated so lower interest rates cannot increase aggregate demand (AD) and no natural interest rate exists.

Okay. Here's what I think:

1. Not knowing anything, we should still know that "consensus" can be a problem. Ten years ago, the consensus was that the financial crisis was not a problem. Eleven years ago, the consensus was that the central problem of depression-prevention has been solved. Consensus is a dangerous thing.
2. Ooh look at that: Palley said "that consensus is dangerously wrong".
3. Palley's view:
• NIRP draws on fallacious pre-Keynesian classical economic logic that asserts there is a natural rate of interest which can ensure full employment.
• That pre-Keynesian logic has been augmented by ZLB economics which claims the natural rate may be negative in times of severe demand shortage, so that policy must deliver it since the market cannot.
• In contrast, Keynes argued investment could become saturated so lower interest rates cannot increase aggregate demand (AD) and no natural interest rate exists.

Ooh.

And then he says

Regarding policy assessment, NIRP turns a blind eye to the possibility that negative interest rates may reduce AD, cause financial fragility, create a macroeconomics of whiplash owing to contradictions between policy today and tomorrow, promote currency wars that undermine the international economy, and foster a political economy that spawns toxic politics.

Saving the best for last, he adds this thought:

Worst of all, NIRP maintains and encourages the flawed model of growth, based on debt and asset price inflation, which has already done such harm.

Exactly.

"the flawed model of growth, based on debt"

Exactly.

"debt and asset price inflation, which has already done such harm"

Exactly.

Exactly.

//

The link at PERI

Cost, income, and markets

As a rule it is a problem when cost is too high, but not when income is too high. To the recipient there is no such thing as income being too high. Excessive cost, however, can bring transactions to a dead stop.

Cost is a problem. Income is not. And yet, one person's income is another person's cost. If there is any limit to income, it is because every dollar of income is somebody's cost. In other words, the limit to income is a cost-side limit.

When cost is widely distributed, and income narrowly, the cost-side limit to income is less effective. The larger the market for your product, the less effective is the limit to your income. Markets allow inequality of income. Larger markets allow greater inequality.

Now you know why the wealthy favor globalization.

The 10-year average 10 years out

Springtime. Time for mowing the lawn every fourth day. Between that and binge-watching with the wife, who has time for blogging? Expect these posts to be intermittent ...

On the 10th I showed this graph of RGDP growth ("Gordon growth" I called it) and some other stuff:

Graph #1: Gordon Growth and Productivity -- Incremental 10-Year Periods

Today I keep the blue line and drop the rest.

In that earlier post I said

That last little tic it shows, from 2017 to 2018, is an uptick. Ooh ooh Trump.

But it isn't Trump, I said. The old data is gradually dropping out of the 10-year average. And the old data happens to be from 2007, 2008, 2009, the time of the Great Recession. The blue line shows an uptick at the end because 2007 dropped out of the 2018 average. And over the next couple years, the rest of the low numbers from the recession will drop out of the average as current data is added in. So, if current growth is at all higher than growth was during that recession, the blue line will go up more.

The RGDP data I've been using in these "Gordon growth" posts is Real Gross Domestic Product, Percent Change from Preceding Period, Quarterly, Seasonally Adjusted Annual Rate. The last data item is 2.3, for 2018Q1, preliminary and subject to change. I'll go with that number, a 2.3% annual rate of economic growth.

If the economy stays at the 2.3% growth rate, it is higher than growth during the recession. The blue line will go up. Assuming growth stays constant at 2.3% for ten years -- that's not a prediction, just a number to work with -- the blue line will follow the path shown here in red:

Graph #2: How 10 Years of 2.3% Growth Affects the Average

In the near term, at the start, the red line rises rapidly. But after a couple years it is already close to the 2.3% level, and after that there is not much change.

Point of interest: On this graph the sudden change in the red line (from mostly rising to mostly flat) occurs at 2020Q1, half a year or so before the next Presidential election. Of course, it also takes time for the data to be reported. So if the economy suddenly goes flat near 2.3% as shown here, we're not going to know about it until after the election.

In another hypothetical future, economic growth increases at 0.1% per quarter, from 2.3% in 2018Q1, to 2.4% in 2018Q2, to 2.5% in Q3 and like that, until it reaches a 3.1% annual rate, and then remains at 3.1%. Again, the red line shows the future path:

Graph #3: How 10 Years of Rising Growth Affects the Average

Here the red line rises rapidly at first, as the Great Recession falls out of the 10-year period. Then it rises more slowly, until the average approaches the 3.1% level. But there is a definite difference between this graph and the previous one.


I used the unfortunate term "prediction" to identify data on the graphs. In the spreadsheet, the word meant that some other data was the actual data. But on the graphs, it looks like a prediction. That was not my intent. I just wanted to see how the numbers affect the average as time passes.

And I missed my 4AM deadline.

No early warning? Of course there's an early warning.

Anticipating recession, John Mauldin writes:

Look around at all the great economic news. I’m aware of it. But the economy was hitting on all cylinders in early 2000 and late 2006, too. The numbers always look great right before a recession. Then it all rolls over at once.

I'll throw the challenge flag.

The economy was hitting on all cylinders in late 2006? Depends what numbers you look at.

Graph #1: Household Debt, Quarterly Change in Billions

The growth of Household Debt peaked in the first quarter of 2006. It was seriously downhill from there to the recession. And come to think of it, household debt peaked in the first quarter of 2000, a year or more before the 2001 recession. So I don't buy Mauldin's story that the numbers are all great and then suddenly there's a recession. I don't buy that.

By the way, the most recent data FRED provides at the moment is for the last quarter of 2017. It shows increase. It shows a trend of increase. No indication of recession.

As indicators go, the yield curve Mauldin discusses may be a little more "leading" than the change in household debt. But at the moment, the debt numbers give no indication of impending recession.


Nor do the Change in Employment numbers suggest impending recession:

Graph #2: Change in Employment

After running pretty flat (jiggy, but flat) from 1997 to 2000, the change in employment suddenly started dropping. A year or so later, recession.

After showing increase since the 2001 recession, the change in employment peaked in 2005 and started dropping. Two and a half years later, recession.

Since climbing out of the 2008-09 recession, the change in employment has run rather flat. You might see a hint of an S-curve in the trend. But there is definitely no dropping off. There is no indication of recession at this time.


Capacity Utilization wanders a lot. But it goes downhill fast during recessions. It tends to fall or to run flat for six months or more before a recession. but sometimes the "downhill fast" starts even before the recession. Notice, though, that Capacity Utilization is never going uphill fast when a recession starts.

Graph #3: Capacity Utilization

At the moment, Capacity Utilization is going uphill fast.

My conclusions: No impending recession. And there are always early warnings.

The moment we've been waiting for!

What I'm now calling "Gordon Growth" is a comparison of time periods, where GDP growth is "measured between quarters with identical unemployment rates". I figure it by making a scatterplot showing the time periods with unemployment on the X-axis and RGDP growth on the Y-axis, then comparing the linear trend lines created by Excel.

The Gordon Growth number is the average of trend line growth rates.

At this point I couldn't tell you how I managed to turn the phrase "measured between quarters with identical unemployment rates" into the process I'm using. But the process works, far as I can tell. So here we are.

I want to go back to that first sentence which has occupied my attention for three or four days now. This time, I'm going to start thinking about the second half of the sentence, and productivity:

Measured between quarters with identical unemployment rates, U. S. economic growth slowed by more than half from 3.2 percent per year during 1970-2006 to only 1.4 percent during 2006-16, and only half of this GDP growth slowdown is accounted for diminished productivity growth.

Growth slowed from 3.2% to 1.4%, a difference of 1.8%. And only half of this slowdown can be attributed to diminished productivity growth. So I want to look for a slowdown in productivity growth of 0.9% or so. I should find this slowdown by comparing the time periods noted in the quote.

This is related to "growth accounting": The GDP growth rate can be calculated as the growth rate of the labor force plus the growth rate of labor productivity, as Menzie Chinn shows. So I think the productivity that Robert Gordon's talking about is labor productivity. Real output per hour.

Using PRS85006092 I get an average productivity growth rate of 2.1% for 1970-2016, and an average of 1.2% for 2006-2016. The one minus the other is 0.9%, exactly the number we're looking for.


I put a new spreadsheet together, formatted better, with the info for each time period all in one column on the sheet. I tested the sheet by duplicating my own previous results. It was good.

I checked the productivity numbers too. The sheet was good.

Satisfied that the new worksheet is good, I made a copy of the sheet and gave it different time periods to evaluate. For the three time series I'm looking at, my data goes from 1948Q1 to 2018Q1. Seventy years of data. I figured I'd look at ten-year periods:

Graph #1: Gordon Growth and Productivity -- 10-Year Periods

The blue dots show the Gordon Growth numbers by decade: numbers comparable to Robert Gordon's 3.2% for the 1970-2006 period. My dots for 1968-2008 are in that neighborhood. Gordon got 1.4% for the years 2006-2016. I get 1.5% (same as I got before), this time for the years 2008-2018.

There is a general downward trend in the Gordon Growth number on my graph.

The red dots show productivity by decade. There are high points at 1958-1968 and 1998-2008; these highs more or less agree with other histories of productivity that I have seen. The numbers I have are 2.1% for Robert Gordon's 1970-2006 and 1.2% for his 2006-2016. The latter is a good match to my 2008-2018 here, and the other looks about right as an average of the red dot values from 1968 to 2008. In the ballpark.

The green dots are blue minus red: Gordon Growth less Productivity Growth. To put it in the "growth accounting" framework, green shows growth attributable to things other than labor productivity.

Labor force size, maybe?

Maybe. But I don't like to close doors. If we attribute the green growth to labor force size, there is nothing left over that I can point to and say "it was caused by the growth of private debt". And I have to have something I can point to and talk about debt.


To me, breaking up 70 years into seven 10-year periods is much more informative than breaking it into two periods punctuated by a "great" recession. So I thought maybe I'd gain even more by breaking it into fourteen 5-year periods. But all I gained was messiness:

Graph #2: Gordon Growth and Productivity -- 5-Year Periods

First guess, the blue dots now show the effects of recessions. I count four definite low points:

• 1953-1958: (2) recessions
• 1978-1983: (2) recessions
• 1988-1993: (1) recession
• 2008-2013: (1) "great" recession

I remember reading that somebody, Kondratieff I think, used 9-year periods to minimize the effects of recession on his data. I can see that, now.

The red dots still show early and late high points, plus a smaller high in the 1980s.

The green is all up-and-down. It's indecipherable. But probably not closely tied to labor force size.


I want to go back to 10-year periods, to minimize the effects of recession. But this time I'll figure incremental, overlapping ten-year periods: 1948Q1-1958Q1, then 1949Q1-1959Q1, then 1950Q1-1960Q1, and like that. And instead of getting a "dot" for a ten year period, I'll have points on a line, where each point is the end of a ten-year period. I think that'll smooth out the picture, even better than the first graph did.

...

Here's the graph:

Graph #3: Gordon Growth and Productivity -- Incremental 10-Year Periods

First impression? Didn't smooth things out at all. There's a lot of agitation in those lines. Way more than I expected. Remember, consecutive points on the line are only a year apart.

See what I did here? I put the Legend off to the right rather than up top with the title. With the Legend off to the right, the years of data are compressed into a smaller space. The agitation would look less severe if the lines could stretch out into the space where the Legend is. But hey, I'm just showing you the graph as I first saw it, with the Legend at right and the lines compressed.

This data deserves to be compressed. Look at that green line: It is up and down and up and down and up and down from start to finish. You don't want to hide that agitation.

The red line provides a pretty good display of productivity, I think, with peaks early and late, and listlessness between.

And the bright blue line, Gordon Growth, when I saw that line it reminded me of this graph from Fernando Martin of the St. Louis Fed:

Graph #4: From Why Does Economic Growth Keep Slowing Down? by Fernando Martin

Very similar to my blue line.


Well I'm almost out of things to say about Gordon Growth, at least for now. But first, look again at the blue line on Graph #3. That last little tic it shows, from 2017 to 2018, is an uptick.

Ooh ooh Trump. That was my first ball-bustin thought. It isn't, though. Isn't Trump. The graph shows ten-year averages.

The uptick from 2017Q1 to 2018Q1 is an uptick in 2008Q1-2018Q1 relative to 2007Q1-2017Q1. We see an uptick on Graph #3 because the year 2017 was okay and the year 2007 was horrible. It isn't Trump.

Here's FRED:

Graph #5: 2007Q1 to a Still-Preliminary 2018Q1, Quarterly (The Same Data I've Been Using)

All the values from 2008Q1 to 2017Q1 are included in both of the last two points on my bright blue. The difference between those last two points exists entirely because of the difference between the first few points and the last few points on this FRED graph.

The change from 2007Q1 (where the graph starts) is a jump up to over 2.5%, then a slight fall, then a bigger fall, then a sharp drop into negative territory and the start of the Great Recession. All in all, a big downer.

The change from 2017Q1 (the bottom of the last "V", at the right) is a jump up to over 2.5%, then a slight increase, then a slight fall, then another slight fall as the line heads to a tentative 2.3% rate. All in all, fairly flat.

The difference in the last two bright blue points on my #3 is this: The next-to-last point includes RGDP falling into recession. The last point doesn't, so the last point goes up. I'm trying to avoid tiresome detail here. But people are going to be saying Ooh ooh Trump.

During the next couple years, the bright blue line is going to go up and up significantly, as time goes by and the "Great Recession" gradually drops out of the calculation.

So in the next few years, if you hear anybody say "The economy is great again!" and they hold up a graph of 10-year average growth, well, now you have evidence to shoot them down.

I for one still expect the economy to improve for a number of reasons. But I expect to see it in current data, not in ten-year averages.

Acuppla things

Here's my graph based on Robert Gordon's description, from Sunday: The unemployment rate, sorted, on the X axis. Real GDP growth on the Y axis:

Graph #1: Data in Each Period Sorted by Unemployment Rate

Here's the same data, in chronological order:

Graph #2: Data in Chronological Order

The jiggy lines are different now. But the trend lines are unchanged, as are the numbers in the trend line equations.

That's what I expected to see. But I had to look.

//

Next, how does the graph look if I use the original unemployment values instead of rounding them to one decimal place? Like this:

Graph #3: Data not Rounded

Compare this graph to Graph #1. Or for that matter, you can compare the numbers in the trendline equations to those in Graph #2 (which has the same numbers as #1). With rounding omitted, the numbers are different.

No noticeable change in the trend lines, though. It's not a big change.

//

I wonder... Remember, at the end of Sunday's post, the average of my trend line values for 1970-2006 was 3.2%, the same number Robert Gordon got. But my average for 2006-2016 was a little off.  I got 1.5%. He got 1.4%. I wonder if that difference goes away when the rounding goes away.

Nope. I get 3.2% and 1.5%, same as before.

Notes on my Gordon's Way calcs

Yesterday I said

I want to use Robert Gordon's method of evaluating economic conditions... I'm not doing it by throwing away all the data where there is no identical unemployment rate in both time periods. The way I'm doing it is by putting linear trend lines on scatterplots, and compare the trends.

Proofreading that, looks like I should have said "comparing the trends." Also, I realized that since I'm not "doing it by throwing data away", I don't have to round the unemployment numbers to one decimal place.

Next time, no rounding.


As part of the process of creating my "Gordon" scatter plots, after I made subsets of the data for selected time periods, I sorted them. For each subset I sorted three columns (unemployment rate, RGDP growth rate, and date) on the unemployment column. The sort puts my X-axis values in sequence, lowest to highest, the way they would be if I was putting date values on that axis.

(I didn't need the "date" column for the graph. I needed it to improve my confidence in my work.)

What happens if I don't sort the data? I found out when I forgot to do the sort. It looked like a cluster or spiral or something. I knew right away when I saw it that I did something wrong, and I fixed it right away. I didn't stop to look at it. So I want to graph the unsorted data again -- and look at it this time:

Graph #1: Unemployment Rate (X-Axis) and RGDP Growth Rate (Y-Axis)

Not a spiral. Maybe a "scatter". If you follow along the blue line from dot to dot, you arrive at the dots in chronological order. Uh, the dots are in chronological order, not you.

But looking at it, I get the impression that the dots are grouped, with empty space between the groups. Look at the dot closest to the upper-left corner: There are no dots below it! Three dots off to the left, maybe five to the right, but there is a big space with no dots below that upper-left one.

And look at the two highest dots, the ones above the 15.0 level. Below each of them is a broad white strip with few scattered dots in it. A little off to the left, a little off to the right, the dots are more tightly packed together. Odd, isn't it?

Maybe that's "random": unexpected groupings separated by unexpected empty space.

Then I remembered that I rounded the values. I ran into that problem before, where the dots got packed into groups by the rounding. I did the graph over right away, using the original, unrounded data:

Graph #2

Looks almost the same.

... ?

Oh, of course: Rounded to one decimal place, I could have nine dots side-by-side between 4.0 and 5.0 on the X-axis, with the dots separated by gaps only about the size of a dot. The grouping isn't from the rounding.

Then you get these strange ideas, like maybe there is some relation between unemployment and growth which favors certain places on the graph over other places.

... Nah.

But you have to think about those things, you know? Because one of those ideas could lead to the light bulb that works.

Using Robert Gordon's method

Since last we spoke, I figured out how to use the SLOPE() and INTERCEPT() functions in Excel. Better late than never, huh?

I want to use Robert Gordon's method of evaluating economic conditions, the method we looked at yesterday. I want to pick time periods that interest me this time, and compare economic growth in those periods his way: for "quarters with identical unemployment rates".

I'm not doing it by throwing away all the data where there is no identical unemployment rate in both time periods. The way I'm doing it is by putting linear trend lines on scatterplots, and compare the trends. Like I did yesterday. It's the method that makes sense to me. I don't know how Robert Gordon did it. But I figure he must have done it the same way, as it is the only way to do it that I thought of. :)

Anyway, now I can calculate the slope and intercept values. I don't have to make the graph and add the trend line and copy the values from the trend line equation. That'll save a lot of work.


Since we crawled out of the last recession, for the longest time people were talking about how economic growth was so much slower than before.

• Julian Brookes in Rolling Stone, 2012: "Four years after the start of the Great Recession, nobody would mistake U.S. economy for a thrumming engine of growth, prosperity, and human flourishing."
• Chris Matthews in Fortune, 2014: "GDP growth has been tepid since 2009 (just 2.1% per year, below the post-war average and far below the average for previous recoveries)..."
• James Hamilton at EconBrowser, 2017: "The Bureau of Economic Analysis announced yesterday that U.S. real GDP grew at a 1.9% annual rate in the fourth quarter, well below the historical average of 3.1% per year, but close to the 2.1% average since the recovery from the Great Recession began in 2009:Q3."

Now, somehow, slow growth has become "normal". I think it's a dangerous way of thinking, to lie to ourselves like that. Hey, I don't want to dwell on how bad things are. I just want to understand the economy. So I think we have to accept the facts for what they are.

Does this mean we have  to accept slow growth as normal? No, because that's a prediction. Predictions are not facts. The fact is that the economy has been slower since 2009 than it was before.


Now, about that "before" time, and the growth then: "It´s more or less recognized that US RGDP is trend stationary," Marcus Nunes told me, "with real growth averaging about 3.3% from the early 50s to 2007."

At Trading Economics I read that "GDP Growth Rate in the United States averaged 3.21 percent from 1947 until 2018". I know they want to simplify and have just one number, and that can be useful. But it's not useful if you want to see how growth trends have changed.

Marcus's method is better. He figures an average only thru 2007. There is data left over, so we can figure an average for the more recent years. Then we can compare the two averages and see how growth trends have changed.

Trading Economics figures the average from a 1947 start. That's what I would do. It uses all of the commonly available quarterly data. Marcus figures the average from 1952 "to avoid the post war adjustment". That makes sense, too.

What doesn't make sense to me is to lump all the years together and get one average value, if there was a change somewhere in the middle. Scott Sumner says "growth in US living standards slowed after 1973". Somewhere in the middle.

Ross Perot showed the same:

Graph #1 Source: Ross Perot, United We Stand
(from when Perot was running for President in 1992)

So maybe we want to look at economic growth from 1947 to 1973, and from 1973 to 2007, and since 2007. Three numbers. Three growth trends. Then we can look at all three, and compare them. That's the kind of thing that makes sense to me.

So I made a graph like the last one I did yesterday, using Robert Gordon's method, but showing the three growth trends:

Graph #2: Progressive Decline in Trend Growth
Unemployment on the X Axis. RGDP Growth on the Y Axis.

Forget the thin, jiggy lines. Forget all those dots. Focus on the three thick trend lines. The highest one, the blue one, is for the years 1948 thru 1973. (1947 got lost somewhere.)

The red line is lower. The economic growth is lower. This line shows the trend for the years 1973 thru 2007.

The green line is lower yet. Economic growth is lower yet. This line shows the trend for 2007 thru 2017.

The average growth rate of the 1948-1973 line is 4.1%. The average for the 1973-2007 line is 3.1%. The average for 2007-2017 is 1.5%.

Graph #3: Average Growth Rate by Period

Graph #4: Average Unemployment Rate by Period