Arnold Kling  

Bryan and Scott respond, sort of

An 84-Word Reply to Arnold... Tattoos and the Labor Market...

Bryan responds to my question. Scott Sumner also responded in the comments on my original post.

My question is how to reconcile low employment with low unit labor costs. Presumably, low unit labor costs would cause labor demand to be higher rather than lower.

My translation of what they have to say is this:

Bryan: Notwithstanding remarkably low unit labor costs, if unit labor costs were even lower, we could have full employment.

Scott: Notwithstanding remarkably low unit labor costs, if we had higher nominal GDP, we could have full employment.

Each of their positions amounts to a non-falsifiable hypothetical. Not that we should be shocked by non-falsifiable statements. This is macro, of course.

Neither of their positions serves to explain the phenomenon I highlighted. Instead, they amount to saying, "Nothing to see here. Move along."

Tyler Cowen agrees with me that this is worth wondering about.

Even if you believe a sticky-price story that disconnects employment demand from real wages, there is still something to puzzle over. Historically, the price markup has been strongly pro-cylical. My guess is that the macroeconometric modelers at the Fed and private consulting firms are having to put gigantic fudge factors into their markup equations to get them to track recent behavior.

[update: Scott offers a longer answer that still is not an answer. He says that instead of looking at real unit labor cost we should look at wages divided by nominal GDP per capita. Here is what the two series measure:

real unit labor cost = wage rate divided by nominal output per worker. Simplifying the fraction, it is the wage rate times employment, divided by nominal output.

Sumner's measure = wage rate divided by nominal output per capita. Simplifying the fraction, it is the wage rate times population, divided by nominal output.

So, the ratio of Sumner's measure to real unit labor costs is population divided by employment, which goes up when the unemployment rate goes up as a matter of arithmetic.

The graph I copied shows the price markup rising, which means real unit labor costs have been falling. Sumner's measure has been rising. Arithmetically, this means that population divided by employment has been rising, i.e., that there has been higher unemployment.

But that is the mystery! Why has unemployment gone up, in spite of lower real unit labor costs? ]

Comments and Sharing

COMMENTS (25 to date)
Justin Irving writes:

My guess is that the macroeconometric modelers at the Fed and private consulting firms are having to put gigantic fudge factors into their markup equations to get them to track recent behavior.

You make it sound like this is something new for us...

brendan writes:

Scott's converted me to his views, but I agree this is a weird one. If current wages were high relative to NGDP, certainly that'd count as evidence in favor of market monetarists, right?

I assume high W/NGDP prevailed at the time of devaluation in 1933 when monetary policy was so potent and during other historical instances of negative demand-shock induced unemployment spikes.

Is there an historical example of an AD starved economy generating unusually high profit margins and unusually low W/NGDP rates? Or is this a first?

I understand that Bryan/Scott's argument don't depend on this. But its still weird isn't it?

Jacob AG writes:

I don't completely understand why Scott and Bryan's explanations don't answer your question (at least in part), but whatever. Maybe we can tell a PSST story instead.

Maybe what's missing from the graph in your original post is unit CAPITAL costs. It doesn't matter much that a unit of labor is very cheap, if a unit of capital required to produce the same output is getting even cheaper. If you look at manufacturing, IT, design, etc. etc., you'll see a lot of labor-saving technological advances out there. This puts folks out of work. PSST says that entrepreneurs are supposed to find ways to make use of all that excess labor (not to mention the excess labor from, you know, the bursting of a certain housing bubble), but entrepreneurs haven't done this yet, because of regulation, or uncertainty, or scarce credit, or whatever (pick your villain). In the meantime, labor costs are falling -- and maybe that's helping, see Bryan et al -- but not fast enough to keep up with the falling capital costs associated with technological advance.

I don't necessarily buy that story, but I thought I'd try adopting the PSST story to help come up with an answer. Please do tell me if I've mucked it up.

Jacob AG writes:

A flaw in the story I just told has now hit me in the face... labor-saving technological advances should lower prices, which according to that graph they have not.

Hmmm... *scratches head in confusion*

Bill Hocter writes:

Arnold-remember that we're not dealing with a closed system. Employment is higher-in China, Brazil, India, Viet Nam, etc. where wages are even lower and prospects for growth better. If you were looking at World employment from the vantage point of the Moon, the situation, overall, might look pretty good.

Airman Spry Shark writes:

The chart in your original post, as I understand it, shows average unit labor costs; hiring/wage decisions would/should be based on marginal unit labor cost.

Low average unit labor cost does not necessarily imply low marginal unit labor cost (e.g., Tyler Cowen's ZMP workers).

Lord writes:

It is quite what one would expect from sticky prices. Real wages may fall but higher profits don't result in hiring unless there is an increase in demand for product but with most of the rewards going to capital that has no need of more product business sits on their profits and the economy stagnates.

James A. Donald writes:

This looks perfectly straightforward to me, if we assume that price inflation is massively understated.

A Galt strike results in rising unemployment, falling employment, falling real wages, and rising prices - exactly what we are seeing.

It is Galt Strike.

AC writes:

Perhaps a post aimed at 3-year olds explaining the issue posed by this graph?

It's not directly showing a rise in labor costs, right?

Mark Michael writes:

Maybe some industries need worker skills that aren't available in adequate numbers, and the time to train them is substantial. Here in Ohio with the leasing of land for natural gas drilling, plans for a new steel mill to produce pipe, petrochemical plants on the drawing boards, there's a flurry of trying to find the right workers. Community colleges, universities are working w/ the companies to set up training courses, short courses, etc. to fill the need. But it doesn't happen overnight.

Natural gas is suddenly quite cheap. It's the "raw material" for a lot of manufacturing companies. Hence, those companies could be very competitive w/ other First World countries that produce those complex products. Their gas may be 3 or 4 times more expensive.

The housing boom & bust attracted large numbers of workers into the industries that support that. That boom lasted (maybe) 15 years. Suddenly ten's of thousands of those workers were out of jobs. Have to get retrained in jobs now available.

Another factor is simply the generosity and time that unemployment compensation is available. We've given unemployed workers very long times on unemployment compensation with this recession - 99 weeks. This will result in a certain percentage of workers not searching for a job as diligently as they might have otherwise.

There's also the unpredictable technology factors: How do you know when some big breakthru is going to come along that results in a large number of new jobs? Or even modest increments of new technology that results in lots of jobs? Within the last 20 years companies like Amazon, Google, Apple's big blossoming w/ the iPod, iPhone, iPad, eBay, Facebook, plus a lot more more lower-profile tech companies have come on the scene. They need workers with skills, knowledge not previously widely needed.

I'm not an economist but I assume those economists who model labor, unemployment, etc. build in the delays that match "real world" delays. If they don't, I'd think their models were "toy" models of little value.

Unless you do the extensive grunt work to gather up the statistics representing these industries - and see what's happening "on the ground" it's just so much academic noodling I'd think.

Andy Harless writes:

If you allow substitution of capital for labor, isn't the graph easily consistent with a sticky wage story? The graph tells you, for a unit of output, how much of value of that unit is from the value of the labor that went into it and how much of the value is the return to capital. Multiply the numerator and the denominator both by the number of units, and the graph is essentially giving you (wL+rK)/wL. If the effective real wage w rises (relative to the cost of capital r) because of some nominal shock, then producers will increase K relative to L. The net impact on (wL+rK)/wL depends on elasticities, but my guess is that it would be positive. Basically, when a recession hits, you fire a bunch of people, and as demand picks up afterward, you find new ways to produce that are less labor-intensive, because labor is overpriced. Less labor-intensive means more capital-intensive, which means a larger markup. Where's the mystery?

Arnold Kling writes:

The mystery is that this has not happened in any post-war recession/recovery. At the very least, one has to tell a story to explain a huge change in the elasticity of substitution between capital and labor.

Scott Sumner writes:

Arnold, I don't know exactly what data is being used, so I can't say what explains that pattern. For instance, is it the same hourly nominal wage series as I used? But I can imagine all sorts of possible explanations:

1. The price data is wrong. I've often argued that CPI data is basically meaningless. For instance, the CPI shows housing costs up 7.5% in the past 5 years, while Case Shiller has them down 32%, a 40% discrepancy. I'd argue the Case Shiller number is more meaningful for the question of "Why aren't construction companies building more houses?" The government numbers say it should be highly profitable, as housing prices are probably rising at least as fast as unit labor costs in construction. Do you think that is true? I don't, I think housing prices have plummeted. And housing is 1/3 of the CPI.

2. The productivity numbers may be wrong. Productivity is extremely hard to measure.

3. There could be composition bias, the unemployed workers are on average less productive than employed workers.

4. The wage data might just apply to a subset of the entire workforce. Does it include government workers, for instance.

5. There may be secular trends right now unrelated to the recession that are increasing the share of national income going to capital.

6. The wage data ignores the rising cost of non-wage benefits and regulations.

But what I would emphasize most strongly is that I'm not claiming that unemployment is high because real wages are too high. Especially real wages adjusted for productivity. So even if my comments are all wrong, and the data is 100% accurate, it has no bearing on my claim that unemployment is higher because the W/NGDP ratio soared in 2008-09.

Or my claim that unemployment is now falling because the W/NGDP ratio is now falling.

David Pearson writes:

Scott Sumner writes, "uemployment is higher because the W/NGDP ratio soared in 2008-09"

Here is a chart of the labor share of nominal GDP -- essentially W/NGDP. The profits share is overlaid on the labor share series. Its hard to see how one can get this chart to shout, "the 2008 increase in the labor share of income is the cause of our persistently high unemployment".

[tinyurl replaced with full url. Please do not use shortened urls on EconLog. Our readers want to know where they are going before they click a link. Our readers are interested in the sources and credibility of links and charts. If a chart is one that you, your company, or your associates have prepared, please say so when you provide a link.--Econlib Ed.]

RPLong writes:

I agree with Bill Hocter. I think US labor is extremely over-priced and under-productive. The market is asking for a major labor price correction, and people are resisting it. But you can't fight the invisible hand. It will happen, it just requires reality to hit a few more people in the face.

And if governments would stop screwing with barriers to trade and inflating the money supply, that would be useful, too.

DPG writes:


I would appreciate a post from you on Modern Monetary Theory/Chartalism. It has been getting some attention in the mainstream media and I'm having a hard time wrapping my head around its assumptions and implications.

Norman Pfyster writes:

Arnold, you are wrong about the data not being consistent with past recessions. All of them show a decline with a sharp increase on the exit from the recession. The big difference is that there was no decline heading into the recession and the recession itself barely slowed the long-term increase. It looks like there has been an increase since the mid to late 1980's, which is consistent with Karl Smith's interpretation that there is a long-term shift in the nature of the economy away from manufacturing (more equality in labor pricing), and has nothing to do with monetary phenomena at all.

Andy Harless writes:

Arnold: "The mystery is that this has not happened in any post-war recession/recovery."

Who says it hasn't? The only difference is that wages came back to equilibrium fairly quickly in the past (presumably because recessions were shallower and/or the inflation rate was higher), so the substitution was reversed. (This didn't happen after the 2001 recession, but there has to be a role for secular trends here too. For one thing, labor force growth slowed after 2000, so the equilibrium would presumably involve more capital-intensive production than in the past.)

Of course, more generally, one could expect, certainly at lower-than-business-cycle frequencies, that there would be shifts in the US labor demand curve. One example that comes to mind is the increased availability of labor abroad with the liberalization in emerging markets. Presumably this means that markups abroad are higher (i.e. productivity has risen faster than wages), so the equilibrium markup in the US is also higher. (In other words, sure, maybe real wages are low now in the US, but they're even lower in China, so a historically low US real wage could still be above equilibrium.) This trend could easily obscure changes in the markup related to the business cycle.

Becky Hargrove writes:

You said, "Not that we should be shocked by non-falsifiable statements. This is macro of course." LOL. No wonder nutcases like me have so much fun in this playground and I'm glad you 'came out to play' this week. What's interesting is that the way we respond to this situation says some pretty basic things about our belief system.

So I'll throw out my own idea of causality. We get five primary areas of economic activity at levels large and small: maintenance - building - creating - healing (literal and figurative) - understanding. Teaching isn't in here because it is all of these. Which of these might we get a bank loan for? Building, even though the product could be from the other areas. And the proceeds from building basically cover the rest. This is why we need knowledge prior systems to assist with what money cannot always provide.

Becky Hargrove writes:

Missing from the above is the idea of retail. When I got a loan for my bookstore, the loan was in response to a product created by knowledge priors. I took a rather large pile of books and arranged them into a recognizable construct that was different from the already existing bookstores of the area. That different definition was the product.

CKE writes:

It may be that for some tasks, there is no value of wages that would motivate employers to substitute labor for capital.

For example, one could substitute labor (with picks and shovels) for heavy equipment to build a road. But it may be that no wage (even free) would make it cost effective in the United States to do so. The costs to manage "free" labor, the cost in schedule, the cost in lives lost or injury, the cost in quality of the work, there are other costs associated with substitution and these may outway trading labor for capital no matter what the wage -- given the technically feasible alternatives.

And per Sumner, "There may be secular trends ... increasing the share of income going to capital". Such as a shifting of more income toward consumption and away from capital investment, thereby making capital (on a per capita basis) more scarce, and labor (or a per capital basis) less scarce.

Charlie writes:

This chart is supposed to represent the inverse of wage share of GDP. But if you do something simple, like take national income / employee compensation, this interesting pattern goes away. It just bobs up and down between 1.55 and 1.60 between 1997 and 2010.

It's weird stuff going on with the specific data and adjustment for these calculations. This paper has more detail about those challenges.

Becky Hargrove writes:

Regarding the above paper, the inclusion of government labor will distort measurements of all kinds because of the random productivity involved. However it is probably best to keep this knowledge service labor as a part of measurement for now, because it allows governments of the developed world to maintain GDP measures in general. Plus, it is not just government employees that have random productivity (compared to the building discipline or production/manufacture) but also the knowledge based services that are largely made possible by government in the present.

Patrick R. Sullivan writes:
Here is a chart of the labor share of nominal GDP -- essentially W/NGDP. The profits share is overlaid on the labor share series.

Please! That old visual trick went out with. 'How to Lie with Statistics'.

Lord writes:

This isn't any post war recession, this is a depression.

Comments for this entry have been closed
Return to top