Believe those who are seeking the truth. Doubt those who find it. Andre Gide

Friday, September 16, 2011

What Scott Sumner Knows for Certain

Love TheMoneyIllusion. But does Scott go a little too far with this proclamation?

Here’s what we know for certain about the US business cycle:
1.  If nominal wages are highly sticky, then NGDP slowdowns will raise unemployment.
2.  Nominal wages are highly sticky for at least some workers.
3.  The period after mid-2008 saw the largest NGDP growth collapse since the Great Depression.
4.  The period after mid-2008 saw a huge rise in unemployment.

Points 3 and 4 are empirical statements (and they are true).

Point two sounds like an empirical statement too, but it is not really. The reason is that there are many different theoretical (and empirical) notions of what exactly constitutes a "sticky" nominal wage. Most people have in mind the idea that nominal wage rates do not appear to adjust quickly to changing economic circumstances. But ideally, the concept should be defined more precisely than this (preferably within the context of an explicit economic model). At the very least, we could then be absolutely certain that we were talking about the same thing!

(The other thing I should like to point out here is that people often ignore the huge monthly gross flows of workers into and out of employment. The wages of these workers likely display much more flexibility than those workers employed for some time at a given establishment. I discuss turnover issues here.)

The first point is clearly a theoretical proposition: if X, then Y. As a theoretical proposition, it is likely to remain valid only under a set of specific conditions. Unfortunately, Scott does not provide us with the model he has in mind. And to make matters worse, he seems to want to make us believe that, whatever this model is, it "for certain" applies to the U.S. economy. (Most of what I am complaining about is probably the by-product of loose blogger language, but I think it's important for some things to be more precise.)

Now, I can certainly think of a model that might deliver something resembling the proposition in question. Think of a neoclassical labor market model, where money is somehow necessary, and were nominal price adjustment (or formulating contingent contracts) is costly. Then think of an exogenous decline in the price level (who knows what might have been responsible for that). The implication is that the real wage rises and that this reduces the demand for labor (though why this translates into an increase in unemployment, and not an increase in non-participation, is not usually discussed).

As I have argued elsewhere (The Sticky Price Hypothesis: A Critique), Marshall's scissors (static supply and demand curves) are probably not the best tool we have available to interpret the labor market. The labor market is a a market in relationships, much like the marriage market. The spot wage is irrelevant in enduring relationships; what matters is the time-path of wages and the division of the joint surplus. There are many different wage paths that cost the firm the same in net present value terms. A "sticky" wage (whether real or nominal) need not have any allocative consequences.

I therefore confess that I, for one, do not know "for certain" that if nominal wages are sticky, then NGDP slowdowns will raise unemployment.
It ain't what you don't know that gets you into trouble. It's what you know for sure that just ain't so. Mark Twain


  1. When you dont know what to do, often the best thing is to just do nothing. Sticky wages and prices could be a symptom of underlying uncertainty and coordination problems, rather than the source of the problem. Steve Williamson and his commenters have been discussing something like that this week. Keynesians or monetarists like Sumner seem stuck in some 1960s theoretical amber like in Jurassic Park.

  2. Anonymous,

    I don't think it's fair to make such a sweeping statement about groups of economists. Nor do I think it's fair to Sumner, who I think understands and appreciates competing viewpoints. I am just chiding him here for a sin that I think we all occasionally make (being too sure of what we think we know).

  3. I'm never sure how a macroeconomist should think about the US labor market, esp. these days.

    On the one hand, I have microeconomists reminding me, as David does, that the gross flows into and out of employment dwarf changes in aggregate employment. That makes me expect that, even if the wages paid in a given job-employee match are quite sticky (for example, because they are long-term contracts), aggregate wages should behave quite differently. It also makes me worry less about the social cost of unemployment because I think short spells between jobs are less serious than long-spells.

    But what I think we see in the aggregate data is that average nominal wages are quite sticky. I also keep hearing Keynesians worried about a sharp rise in long-duration unemployment in recent years; I think most economists would agree that's a bad sign.

    How do you think about movements in the duration of unemployment spells, David? And are sticky individual wages enough to explain sticky aggregate wages?

  4. Are aggregate wages really a meaningful concept? Doesn't wage aggregation require the labor to be homogeneous to have an economic interpretation?

  5. Simon,

    Is what you see in the aggregate wage data similar to what I report here:

    There does seem to be evidence suggesting that the deceleration in price level growth has been sharper than the deceleration in nominal wage growth. But really, as Prof J suggests above, I'm not sure that this level of aggregation tells us anything meaningful.

    Not only "Keynesians" are worried about the rise of long-duration unemployment. I would go further and say that the problem is understated if one just looks at reported unemployment because a lot of people (who likely should be working to build or maintain their human capital) have dropped out of the labor force.

    I'm not sure I understand your last question. I'm not even sure if nominal wages are "sticky" in a meaningful sense. Perhaps it has to do with time-varying bargaining power? (I've seen a paper recently, Ravenna and Walsh, I believe, that allows for stochastic variation in bargaining power that ends up looking a lot like a NK markup shock.

  6. 1. If nominal wages are highly sticky, then NGDP slowdowns will raise unemployment.

    Let me rephrase 1. If nominal wages are constant in the near-term and NGDP declines, the number of employed workers will decrease. Almost a tautology me thinks.

  7. Westslope:

    What you say may be true in some theoretical settings, but it most certainly does not hold true in all theoretical settings.

    Let me give you an example. Suppose that a firm and a worker are in a long-lasting relationship. They negotiate a contract to share in the NPV of their joint venture. They agree to do this by keeping the nominal wage fixed for discrete periods of time, with upward adjustments (think of an increasing step function.)

    Suppose now there is an unexpected decline in NGDP (maybe temporary, maybe persistent). Let's say this shock has no effect on the NPV of their joint venture. To keep the worker's share of the surplus unchanged, the two parties may agree to suspend the future increase in the nominal wage path.

    There is no change in the current nominal wage. The real value (NPV) of the wage cost has not increased for the firm. There is no reason to lay off the worker, or reduce hiring.

    This is a logical possibility. Whether it holds empirically, of course, is a different matter.