[Note for bibliographic reference: Melberg, Hans O. (1996) The NAIRU approach to
unemployment: An application for a Nobel Prize?, http://www.oocities.org/hmelberg/papers/961123.htm]
The NAIRU approach to unemployment
An application for a Nobel Prize?
by Hans O. Melberg
The Unemployment Crisis
Richard Layard, Stephen Nickell, Richard Jackman
Oxford University Press, Oxford, 1994
ISBN: 0-19-877394-3, 157 pages
Some explanatory variables are considered off-limits for economists, either because
they are considered non-economic or non-quantifiable. Nevertheless, the authors of the
book under review - The Unemployment Crisis by Richard Layard, Stephen Nickell and
Richard Jackman [LNJ] - have made apparently non-economic and non-quantifiable variables
the cornerstone of their theory of unemployment. As they write, "In our view they
[unemployment fluctuations] "stem from two sources: long period changes in social
institutions [the benefit system and the system of wage determination]; and .. big shocks
to the system (such as oil price rises or major wars) have long lasting effects" (p.
4). In order to prove this argument they make two moves. First, they create a general
theoretical framework to explain unemployment in which unemployment is seen as the
variable which makes the competing claims (by workers and by capitalists) on national
output compatible. Second, they make imaginative use of the data to estimate the
apparently non-quantifiable variables. These results, in turn, are used to test the
theory. On both accounts I believe the book is valuable and stimulating. That is not to
say the book is without flaws. I will criticize its modeling of expectations; its failure
to test against some alternative explanatory variables; and its distinction between short
v- medium and the long run.
The frame
One reviewer claimed that LNJ's general theoretical approach simply was "recycled
Zeuthen-Nash" (Phelps, p. 26). Admittedly, there is an element of Nash bargaining
theory but it is unfair to claim it is recycled since the model is both extended, changed
and made concrete in the context of unemployment theory.
The best way to explain the frame is to consider an example (from p. 19). We start with
an economy of one worker and one factory owner. Suppose the worker produces 100 units of
output, each unit is initially priced at $1. He then demands a wage of $61 to continue his
work since he believes he is entitled to 61% of cake (the value of the output from the
factory). At the same time the owner of the factory demands $41 of the total output -
believing that the owner of capital is entitled to 41% of value of production. Clearly,
the two claims are not compatible since the cake to be shared is only worth $100 and $61 +
$41 (= $102) > $100. What are the consequences of this inconsistency? Assume that the
nominal wage is set first at $61. The owner of the factory, who controls the prices, will
then set the price of the product so as to gain what he consider is his "fair
share" (41%). If he sets the price to be $1, there is only 39% left after the wage to
the worker is deducted. Thus, to gain 41% he increases the price of the product to $1.034
which makes the value of total production 103.4. After paying the worker $61, the
capitalist is left with $42.4 which is 41% of the value of total output (42.4/104.4=41%).
However, when the price level increase from $1 to $1.034 the worker no longer gets 61% of
national output as he desired when he claimed a nominal wage of $61. He will therefore
demand a higher nominal wage at the next bargaining round. But, once again his hopes are
frustrated as the capitalist after conceding a higher nominal wage, simply raises the
price level again - thus lowering the workers nominal wage. Thus, the circle continues
with ever increasing inflation. Now, the question is what this story has to do with
unemployment.
Unemployment, in the words of LJN is "the mechanism that reconciles the struggle
for shares of the national cake, by cheating both price-setters (capitalists) and
wage-setters of what they intended" (p. 28). In order to avoid the ever increasing
level of inflation, one variable must adjust to make the claims on national output
compatible. One such variable is unemployment since high unemployment reduces the share
workers demand to get. Going back to our example we may find that if unemployment rises
(this, clearly, requires us to relax the assumption of only one worker and one
capitalist), the workers will accept a 59% share instead of the original claim of 61%. In
this way LNJ finds what they call the long-run NAIRU - the non-accelerating-inflation rate
of unemployment - with stable inflation and unemployment. In the long run LNJ believe the
economy converges to this point (although in the short-run they accept that hysteresis and
non-linearities may make the economy converge on a short-run NAIRU).
This, then, is the general theoretical approach which LNJ makes concrete in a short but
highly instructive mathematical model. Although I find the arguments convincing, the model
is not complete since LNJ does not provide the mechanism whereby the system will converge
to the NAIRU (at least not as far as I could see). As all economists know: To show that an
equilibrium exist is conceptually distinct from proving that the economy will actually
converge to this equilibrium. Nickell makes some comments on the existence, stability and
uniqueness of the NAIRU in a later article, but even there he largely throws his hands up
arguing that "As for stability, we have more or less nothing to say, in part because
economists know so little about the dynamics of aggregate demand ..." (Nickell, p.
137). I have to admit I did not understand why this prevents us from providing at least
suggesting some possible mechanisms whereby the economy would converge to the NAIRU - even
if we cannot give conclusive proofs as to whether these are the true mechanisms.
(Suggestions are welcome as I hope to return to this issue later)
The empirical evidence
Having presented the theory, LNJ goes test NAIRU against the empirical facts. For example,
we know that unemployment was low in the 1960s, higher in the 1970s and it remained high
in Europe, while it fell in the US. Moreover, we know that there are large differences
between the unemployment rate within and between countries. For example, the Scandinavian
countries are low unemployment countries, while Spain, Greece, Italy and Portugal are
traditionally high unemployment countries. Lastly, we know that unemployment is higher
among some groups (young manual workers) than among others (well educated managerial
workers). To prove this means that we have to quantify the determinants of the NARU and
see how they differ across time and space, as well as within and between, various
countries.
The problem is that the determinants of both the short run and the long run NAIRU are
notoriously difficult to quantify. For example, a change in union militancy may indicate
that the workers demand a larger share of the national product, thus raising the
unemployment needed to keep inflation low. But, how do we quantify union militancy? One
may take the number of days lost in strikes as a proxy for union militancy, as LNJ does
(p. 5), but - as always with proxies - the relationship is imperfect since variables other
than the desire to earn more may motivate the workers to strike.
More generally the NAIRU is determined by the bargaining strength of the capitalists
vs. the workers. This, in turn, means that the benefit system is important since the
higher benefits the one gets as unemployed, the more inclined people may be to press for
high wages. The degree of centralization is also important since in centralized bargaining
the price level is no longer exogenous to the bargaining process and the workers cannot
exploit the "efficiency-wages" argument to gain higher wages. This means that
the unions are weaker and, in turn, they will settle for a lower wage and a resulting
higher level of employment. Finally, the effectiveness of those who are unemployed in
searching for a job is important. Some countries, LJN puts great emphasis on Sweden in
this respect, have very active labor-market policies in which agencies try to help the
unemployed to get work. Now, all these factors are difficult to quantify directly. The
benefit system differs not only in the amount of money they pay the unemployed, but also
in terms of how long these benefits last and at what conditions. The degree of
centralization is also difficult to measure since even in Scandinavia the centralized
bargaining is followed by local bargaining. Finally, it is difficult to put a number on
how active a country's labor market policies are. In short, the NAIRU is determined by
factors which are difficult to quantify.
This does not mean that it is impossible to test the theory. I have already mentioned
the use of data on strikes to measure union militancy. In this way LNJ create imaginative
proxies. For example, the degree of active labor market policies may be measured by the
case loads that a staff member at the public employment exchange service has to handle (35
in Sweden, more than 175 in Britain, p. 93). Moreover, although the degree of
centralization in the a wage-bargaining system cannot be summarized a single precise
number, it is possible to describe - and quantify - various aspects of it and rank the
countries on a coarse scale. LNJ efforts to quantify the non-quantifiable is a major
achievement of the book, although one may doubt whether it is "worthy of a Nobel
Prize" as one reviewer suggested in a fit of excessive enthusiasm (Cross, p. 121).
Criticism
Rational Expectations?
Inflation depends upon expectations since both price and wage setters base their demands
on estimates of the future price level. It is reasonable to require that these estimates
should not be systematically wrong. For example, if the workers every year receive a
smaller wage than they expect, it is reasonable to expect the workers to increase their
estimates of the future price level. A stronger requirement, but still quite reasonable,
is that the workers realize the potential for being cheated already before the wage
negotiations, since they know the incentives facing the capitalists (to raise the prices).
In this way expectations are forward looking, not just adaptive based on whether past
expectations were fulfilled or not. Surely, even if my expectations were accurate this
year, I will revise my estimate of inflation if the Minister of Finance presents an
expansive budget. On this account I object to LNJ's terminology of rational expectations
(see especially p. 21).
They may argue that if you start with the assumption that inflation is a random walk,
their conclusions are right. However, you cannot at the same time argue that inflation is
random walk, determined within the model and rational. If the expectations are rational
and if you have a model which determines inflation, the proper course for followers of
rational expectation adherents is to make the expectations of the agents the same as the
predictions of the model. (For more criticism of LNJ's expectations, see Phelps p. 20)
Alternative variables
Edmund Phelps has written about LNJ that "... their omission of whole categories of
factors are as striking as the stress they have placed on other factors" (page 7 of
his article). Some examples include the price of capital goods, internal and world
interest rates, exchange rates and wealth. Some of these variables are important because
they affect the bargaining power of the capitalist vs. the workers and labor demand, while
others have more indirect affects on unemployment for example through aggregate demand.
There is, however, no need to make this a strong criticism since Nickell has admitted that
more attention should be placed on these factors - especially interest rates - than they
did. (Nickell, p. 136)
The short-, medium-, and long-run
It may not be a major point, but it seems to me that there is a problem when one
distinguishes between contradictory effects over time, while not indicating exactly how
long time one uses. An example is LNJ discussion of hysteresis. According to LNJ
hysteresis affects the short rune and the medium run NAIRU, but in the long run it is not
important in determining the NAIRU. This may be the case, but as long one does not specify
the length of time one is considering it is difficult to test whether it is true. Then
again, it is probably impossible to require a theory to specify these lengths exactly
since we simply do not possess the knowledge that would make this possible.
Overall
In conclusion, I believe this is a very good book. The content is important, the arguments
largely convincing, and the style is superb - concise and clear. My only regret is that
there is nothing in between the short edition (157 pages) and the long one (618 pages)
since the I felt the short one was a bit too brief and the long one a bit too heavy. If we
broaden the criteria to include their overall work on unemployment - and hoping for some
future theoretical improvements - I might agree with the reviewer who thought they deserve
a Nobel Prize. At the very least they deserve to be read.
References:
All the non-LNJ references are from Journal of Economic Literature 20 (1993) no.
1/2, which contains a symposium of reviews on LJN theory of unemployment.
[Note for bibliographic reference: Melberg, Hans O. (1996) The NAIRU approach to
unemployment: An application for a Nobel Prize?, http://www.oocities.org/hmelberg/papers/961123.htm]