Arnold Kling  

How Sticky Are Nominal Wages?

My Talk at Berkeley... My Back-Up Plan...

The broadest measure of employee compensation is called the employment cost index. It includes the cost of benefits, such as health insurance. It is available quarterly, starting in the first quarter of 2001. I used the private sector worker ECI.

The usual measure of prices is the consumer price index.

Here are data on the cumulative rate of change of these measures over two time periods:

Series2001 Q1 - 2006 Q12006 Q1 - 2011 Q1
Employment Cost Index18.6 %12.4 %
Consumer Price Index13.4 %11.4 %

If we compare the first period to the second, we see that wage growth decelerated much more sharply than prices. Which is exactly the opposite of what textbook models predict. In the current recession, wages were not sticky relative to prices.

Of course, you can always complain about measurement problems in the Consumer Price Index. Or you can say that the "true" (or New) Keynesian model does not depend on sticky wages, because it uses sticky prices. But if you pick the New Keynesian story, then using nominal GDP to measure aggregate demand becomes utterly vacuous. At that point, you are basically saying that "real GDP falls because real GDP falls." I give Scott Sumner credit for not stooping to that level.

Since the financial crisis, productivity growth has been pretty good, and it has exceeded real wage growth, which has been zero. Of course, you can always rescue the AD story by saying that, well, wage growth should have been even less, because of weak AD. Just as you can always rescue the stimulus story by saying that, well, job growth would have been even stronger if we had not had the stimulus.

I cannot argue against these sorts of hypotheticals. Macroeconomic data will never be enough to convince a motivated skeptic to change his mind. They can be read too many different ways.

UPDATE: Since a commenter asked, here are the figures using 2008 Q3 as the endpoint:

Series2001 Q1 - 2008 Q32008 Q3 - 2011 Q1
Employment Cost Index27.8 %5.5 %
Consumer Price Index24.5 %3.3 %

Using this endpoint, the recession and the preceding recovery look fairly similar to one another. During both, real wages crept up at an annual rate of less than one percent, which was far less than the increase in productivity. Again, one is welcome to dispute the CPI as a measure of price changes.

Comments and Sharing

COMMENTS (11 to date)
david writes:

When you cannot defend your hypothesis, attack someone else's...?

PSST is also full of handwaving, especially where it matters most (explaining the damned business cycle); what we have is a choice between more or less handwaving, ahem, "stylized facts", not a choice between certainty and epicyclic ignorance. And PSST hardly does well in that regard. Every alleged pattern of sustainable trade is just an arbitrary sunspot!

Alex Godofsky writes:

If you have sticky downward nominal wages of course you will still see wage increases, because that means it's a censored distribution.

How is this not just damning evidence in favor of nominal wages being sticky downward?

david writes:

@Alex Godofsky

No, Kling is alluding to the apparent anticyclicity of wages, whereas the standard Keynesian narrative implies procyclicity of wages to account for unemployment over the business cycle. Hence his reference to the New Keynesian reliance on sticky prices or prices+wages instead of wages alone, because this was indeed how NK responded to the empirical problem of non-procyclical wages.

I do not see the alleged vacuousness, myself: true, you then cannot use NGDP to meaningfully measure aggregate demand shortfalls, for even though they are the same, non-clearing prices means that the level of NGDP does not indicate the goods-market-clearing equilibrium anyway. An excess supply of goods and services doesn't show up as easily as an excess supply of labour. Hence Kling's invocation of ZMP: maybe NAIRU just went up really fast, how do you know it didn't? But of course one can easily observe the sloshing excess demand for safe assets.

Becky Hargrove writes:

It is hard for me to draw any real conclusions from this index because of the inclusion of health insurance. How much has that changed? My impression is that health benefits have declined in this period.

But more importantly, weak AD is not the best point to assume no real need for income growth. In that present day wealth generation comes strickly from credit, the (already committed to) debts are the first place to look for shortfalls. Because when incomes don't rise, no one is meeting the interest on the loans.

NGDP is about taking care of what already exists. PSST is about taking care of things that could exist in the future. Lest anyone think patterns of sustainable trade don't matter, only consider that their preservation makes all the difference when credit-based scenarios do break down.

James Oswald writes:

1. The primary AD disruption took place in 2008 Q3, not 2006 Q1, so that might result in a better outcome for wage stickiness.
2. Looking at overall wages is a poor way to determine if wages are sticky for individuals. Current workers could have extremely sticky wages and new hires could get lower wages resulting in an overall drop in wages.

Æternitatis writes:

I think @Becky is on to something. ECI may be the best measure of total employee compensation for most purposes, but if we are talking about the sticky *wage* hypothesis, it is the wrong measure.

If I understand it correctly, employees irrationally resist any cut in their nominal wage, even in a deflationary and high-unemployment environment, when that would be the best way to restore economic growth and employment. It is this feature of sticky wages that makes inflationary measure arguably a way out of the crisis.

But most employees, while dimly aware of their health benefits (at least if they are healthy, like most are) and only of the most obvious, but least important, cost features like co-pays (rather than the lion's share which is the employer contribution for most employees), are almost certainly completely oblivious to the cost of their benefits (health and otherwise), so there could be no irrational stickiness to them.

As ECI includes these benefits, I don't think we can draw any strong conclusion on the sticky wage hypothesis from it.

Lord writes:

The evidence of sticky prices is quite clear, but this doesn't imply NGDP is circular reasoning because the excess demand for safe assets is also clear.

Rick Hull writes:

How much of the wage stickiness is due to heavy government involvement in the labor market?

Tom writes:

Becky, though employees have been asked to pick up part of health insurance, I don't think this is from the company paying less of it's share (dollars not percentage). Costs have risen so quickly that employees have been asked to pick up part of the INCREASE, not of the total.

Say my benefits used to be 100, and are now 200 with my employer picking up 150 and me picking up 50. It appears to me that mt pay has decreased, while my real compensation has gone up by 50.

Becky Hargrove writes:

It could be that the above chart is more accurate for wage stickiness than it appears. When I thought in terms of health benefit totals, I was thinking primarily of companies opting out altogether, instead of paying less.

Æternitatis writes:

Two points:

@Becky: If employees are indeed picking up an increasing share of their medical insurance cost, their nominal wages would fall, something that sticky-wage theory would predict to be rare and create unproductive collateral effects. Not sure how this cuts with respect to Prof. Kling's argument.

@Rick: I'm as happy to blame the government for pretty much any stupidity and failure as the next hard-core libertarian, but how the government (rather than worker irrationality) would cause wage stickiness is unclear to me. All I can think of are some pretty trivial effects. Could you elucidate?

Comments for this entry have been closed
Return to top