The University of Chicago Press is collaborating with JSTOR to digitize, preserve and extend access to Journal of Political
Economy.
http://www.jstor.org
Sectoral Shifts and Cyclical Unemployment
Author(s): David M. Lilien
Source: Journal of Political Economy, Vol. 90, No. 4 (Aug., 1982), pp. 777-793
Published by: The University of Chicago Press
Stable URL: http://www.jstor.org/stable/1831352
Accessed: 18-03-2015 11:05 UTC
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
http://www.jstor.org/page/info/about/policies/terms.jsp
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content
in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship.
For more information about JSTOR, please contact support@jstor.org.
This content downloaded from 147.251.185.127 on Wed, 18 Mar 2015 11:05:45 UTC
All use subject to JSTOR Terms and Conditions
Sectoral Shifts and Cyclical Unemployment
David M. Lilien
Universityof California, San Diego
A substantial fraction of cyclical unemployment is better characterized
as fluctuations of the "frictional" or "natural" rate than as
deviations from some relatively stable natural rate. Shifts of ernployment
demand between sectors of the economy necessitate continuous
labor reallocation. Since it takes time for workers to find new
jobs, some unemployment is unavoidable. This paper presents evidence
that most of the unemployment fluctuations of the seventies
(unlike those in the sixties) were induced by unusual structural shifts
within the U.S. economy. Simple time-series models of layoffs and
unemployment are constructed that include a measure of structural
shifts within the labor market. These models are estimated and a
derived natural rate series is constructed.
I. Introduction
Some unemployment is unavoidable in free market economies. Variations
of factors, such as the demand for their products or the cost of
inputs to production, require firms continually to adjust the size of
their labor force. Even in periods of stable aggregate employment,
continuous labor reallocation within the United States results in almost
5 percent of employment leaving old jobs for new ones every
month. Because it takes time for separated workers to be matched to
jobs, some positive level of unemployment will always exist.
Economists have long recognized this fact and have labeled this necessary
quantity of unemployment the "frictional," "natural," or
"equilibrium" unemployment rate.
I would like to thank John Conlisk and Vincent Crawford for comments on the
current paper and Olivier Blanchard and Robert Lucas for comments on an earlier
draft. All remaining errors are mine.
[Journal of Political Economy, 1982, vol. 90, no. 4]
? 1982 by The University of Chicago. All rights reserved. 0022-3808/82/9004-0009$01.50
777
This content downloaded from 147.251.185.127 on Wed, 18 Mar 2015 11:05:45 UTC
All use subject to JSTOR Terms and Conditions
778 JOURNAL OF POLITICAL ECONOMY
Since the widespread acceptance of the natural rate hypothesis,
economists have come to view cyclical unemployment as deviations
from some relatively stable natural rate. In this paper, it will be
argued that, given a definition of the natural rate based on microeconomic
foundations, much of cyclical unemployment is better described
as fluctuations of the natural rate itself. As much as half of the
variance of unemployment over the postwar period can be attributed
to fluctuations of the natural rate brought about by the slow adjustment
of labor to shifts of employment demand between sectors of the
economy.
Several models describing the determinants of the equilibrium rate
of unemployment have been put forth in the microfoundations literature
(see, e.g., Phelps et al. 1970; Hall 1979). The Lucas and Prescott
(1974) model is perhaps the one most relevant to this paper. In their
paper, Lucas and Prescott explicitly derive the equilibrium unemployment
rate from the assumptions that labor is exchanged in many
spatially distinct markets and that labor mobility between markets is
time consuming. While aggregate demand is assumed to be constant
in their model, the product demand in individual markets is subject to
stochastic fluctuations. Random fluctuations of product demand
within markets induce fluctuations of labor demand and lead to
temporary wage differentials between markets. These wage differentials
encourage shifts of sectoral labor supply as workers leave lowwage
markets for high-wage markets. Since the process of moving
from one market to another is time consuming, a positive level of
unemployment exists in stationary equilibrium.
The stationary equilibrium described by Lucas and Prescott assumes
that the process generating market-specific demand fluctuations
has constant variance over time and therefore yields a constant
equilibrium unemployment rate. In the real world there is little reason
to believe that the variance of firm-specific or market-specific
demand is time invariant. In some periods, such as the mid-sixties,
product demand, and thus derived labor demand, may grow relatively
uniformly across labor market segments. In other periods, such as the
decade of the seventies, exogenous events such as the end of the war
in Viet Nam, oil boycotts and price increases, and changing import
competition in manufactured goods can induce dramatic shifts of
demand between labor market segments over relatively short intervals.
Allowing the variance of individual market demands to vary over
time in the Lucas and Prescott model leads to an equilibrium unemployment
rate that itself varies as the quantity of required labor
reallocation within the economy changes.
Casual evidence suggests that part of the higher unemployment
rates of the last decade (the average adult male unemployment rate
This content downloaded from 147.251.185.127 on Wed, 18 Mar 2015 11:05:45 UTC
All use subject to JSTOR Terms and Conditions
CYCLICAL UNEMPLOYMENT 779
increased from 3.6 percent in the sixties to 4.5 percent in the seventies)
may be due to the unusual volatility of employment demand over
the period. Between 1969 and 1980, manufacturing's share of aggregate
employment fell from 28.7 percent to 22.4 percent, a decline of
22.8 percent. Over the same period, retail trade; finance, insurance,
and real estate (FIRE); and service industries' share of total employment
grew by 10.1 percent, 14.2 percent, and 23.3 percent, respectively.
To put these numbers into some perspective, note that manufacturing's
share of employment fell by only 6.1 percent between
1958 and 1969. The shifts of industry employment shares, while
significant in themselves, understate the full magnitude of employment
shifts by not accounting for the shifts of employment that took
place within broad industry groupings.
The cyclical pattern of unemployment over the decade provides
further supporting evidence for the hypothesis that unusually large
sectoral shifts contributed to unemployment increases. The shifts of
employment shares over the last decade did not take place smoothly
over the period. The shift out of durable manufacturing, in particular,
seems to be better described as the result of three distinct shocks
than as a secular trend: Durable manufacturing's share of total employment
fell by 12.6 percent in 1970-71, by 9.1 percent in 1975, and
by 5.3 percent in 1980. With the exception of a modest 3.1 percent
increase in 1973, its share remained relatively stable over the remainder
of the decade. These three periods of falling employment in
durable manufacturing coincided with the three cyclical increases in
unemployment over the decade: The annual unemployment rate
increased by 2.4 percentage points in 1970-7 1, by 2.9 points in 1975,
and by 1.3 points in 1980. Note that in all three downturns of the
economy, employment (as well as employment shares) actually rose in
retail trade, FIRE, and service industries. Further, with the exception
of the modest increase in 1973, the share of durables did not
significantly increase as unemployment abated. These last two facts
support the hypothesis that sectoral shifts played a role in inducing
the general economic downturn as opposed to the downturn inducing
temporary shifts of employment.
In the remainder of this paper, evidence will be presented to
support the hypothesis that a significant fraction of cyclical unemployment
over the postwar era can be explained by the slow adjustment
of labor to exogenous shifts of sectoral employment demand. In
Sections II and III we conduct an analysis of the effects of dispersion
in employment demand on aggregate layoffs and unemployment. A
measure of this dispersion is constructed and included in a flow model
of the unemployment rate. Finally, in Section IV the dispersion series
is combined with the parameter estimates from the unemployment
This content downloaded from 147.251.185.127 on Wed, 18 Mar 2015 11:05:45 UTC
All use subject to JSTOR Terms and Conditions
780 JOURNAL OF POLITICAL ECONOMY
model to construct an estimated equilibrium or natural unemployment
rate series.
II. Layoffs
In a typical month, between 4 and 5 percent of all workers flow
through the labor market. While over half of this flow is voluntary,
the overwhelming fraction of unemployment stems from involuntary
employment separations, layoffs initiated by firms in response to
economic conditions (see table 1). Cyclical fluctuations of the unemployment
rate are directly related to fluctuations of the aggregate
layoff rate. Therefore, this section will analyze the determinants of
the layoff rate as a prelude to our analysis of cyclical unemployment.
Even in periods of growing aggregate employment, a significant
fraction of the labor force is laid off every month. Positive layoffs in
periods of growing employment reflect the variance of employment
demand within the economy: Even when most firms are hiring new
workers, there are always some firms issuing layoffs. While this simple
observation is widely recognized, the importance of the dispersion in
hiring conditions in determining the aggregate layoff rate is not as
well understood. Far fewer layoffs would be generated in an economy
where employment was growing at 2 percent per year in all firms than
would be generated in an economy where employment was growing
at 8 percent per year in half of all firms and by -4 percent in the
remaining firms, despite the fact that both economies would have
identical aggregate growth rates.
Many factors determine the level of hiring done by individual firms.
Changes in product demand, changes in capital and raw material
costs, and changes in wage rates influence firms' hiring decisions.
Rather than model these factors explicitly, we will divide the factors
that affect firm hiring into two components: those that affect all firms
in the economy and those that are specific to individual firms. More
specifically, we will assume that the net hiring rate of a typical firm, ht,
is equal to the aggregate hiring rate, Ht, plus a random disturbance,
et:
ht=Ht + et. l
Here, Etis assumed to be distributed with mean zero and variance o-,
according to the density function f(E I 0ot); o-t is a measure of the
dispersion of employment demand conditions throughout the labor
market. Shocks to the economy that have differing impacts on firms,
such as a rise in oil prices, will lead to an increase in ot.
Net hiring, ht, which may be positive or negative, is equal to the
This content downloaded from 147.251.185.127 on Wed, 18 Mar 2015 11:05:45 UTC
All use subject to JSTOR Terms and Conditions
; c tS 0) a)v SvsotcO0
Cq CqG qInOt )S tot- t- x r) t-
0
su~~~~~~~CGMi
CMC >sOO 0>?
<0
n cq oXq0C:00
z
04
X C)o C=O C, n "t Ln vc t- x (0
78i
This content downloaded from 147.251.185.127 on Wed, 18 Mar 2015 11:05:45 UTC
All use subject to JSTOR Terms and Conditions
782 JOURNAL OF POLITICAL ECONOMY
difference between accessions to and layoffs from the firm:
t= (it - It (2)
If we further assume that all separations from the firm are either
layoffs or quits, we also have
ht = Aet + qt, (3)
where zAetis the rate of change of employment and qtis the quit rate.
Finally, we will assume that firms do not simultaneously hire and lay
off workers. While this assumption is not literally true (firms may wish
to change the composition of their employment, which would lead to
simultaneous hiring and layoffs), it is a reasonable approximation for
the majority of firms if we do not include discharges for cause in our
definition of layoffs. The layoff and accession rates of the firm can
now be expressed as:
It = max (0, -ht), (4)
at = max (0, ht). (5)
Given these assumptions, we can derive the aggregate layoff rate as
a function of Ht and o-t. Appealing to the law of large numbers, we
assume that the aggregate layoff rate equals the expected layoff rate
of a typical firm:
rHt
Lt = E (it IHt, 0Ut) - (Ht + E)f(E I ot)d e= g(Ht, ort), (6)
where
19glaHt=- Pt(E o-t)dE =-F(-Ht , t)>-) (7)
ag/art > 0, (8)
and F( ) is the cumulative distribution function associated with the
density functions( ). The aggregate accession rate,At, can be similarly
derived as At = Ht + g(Ht, Ut).
Figure 1 depicts a graphical derivation of A and L from the f()
distribution. The shaded area of theft distribution, to the left of zero,
is the truncated density function of firms having H + E <0 . The mean
of this truncated density function is -L. Similarly, A is the mean of the
unshaded positive portion off1. Increasing o-spreads the distribution
from ft to f2. Firms that had been laying off workers prior to the
increase in o-now increase their layoffs by an amount exactly equal to
the increase in accessions in firms that are hiring. Shifting to the left
(a decline of H) decreases all firms' net hiring by the amount of the
shift. Those firms laying off workers now increase their layoffs. Firms
This content downloaded from 147.251.185.127 on Wed, 18 Mar 2015 11:05:45 UTC
All use subject to JSTOR Terms and Conditions
CYCLICAL UNEMPLOYMENT 783
1 < 02
f~ 2
-L2 -Li 0 H Al A2
al = 0a2
> HI >H2
-L2 H2 -LI 0 A2 HI Al
FI(G. 1.-Derivation of L from theft distribution
having accessions reduce their hiring. Those firms that had accessions
less than the shift now lay off workers.
A heuristic explanation can be given for the magnitude of the first
derivative of the g( ) function given in equation (7). An increase of dH
in H causes all firms to increase their desired hiring by dH. For all
firms experiencing layoffs, there will be adH reduction of layoffs. For
firms that are hiring, there is no reduction in layoffs. Thus the
aggregate change in layoffs is just -dH multiplied by the fraction of
firms experiencing layoffs, F(-H C(T).
Table 2 presents estimates of several linearized versions of equation
(6), using the manufacturing layoff rate as the dependent variable.
The variable STtserves as a proxy for (otand is defined as the variance
of ZAet+ qt among the 21 two-digit manufacturing industries:
Tt = |
cj[zeit + qit (AEt + Qt)]2j,
where ct is the ith industry's 1968 share of total manufacturing employment.
Note that all variables are in rate form. The equations in
table 2 were estimated using both monthly and quarterly data by
Cochrane-Orcutt least squares.
The estimation results strongly support the hypothesized relationship
between Lt, Ht, and St given by equations (6)-(8). The estimated
t-statistics on the St coefficient ranged from 7.9 to 13.7, which indicates
a significantly positive relationship between the variance of
hiring conditions and the aggregate layoff rate. The Ht coefficients
This content downloaded from 147.251.185.127 on Wed, 18 Mar 2015 11:05:45 UTC
All use subject to JSTOR Terms and Conditions
(3 C)
C)O 5 5
0 00 00 ) C ()
03 00 00 00 t
o
n i
O4
G0 - 04 cq c) - Cq 04
0404040400i 0n 0
O (4) 00 - 04
<o. 00* (0 _ o ,:( _000 _ I>-ar
oo C(' (000-00 00 0 000?
._ Z ' O OOOO?
04 t
0H . I I I.I I I
00 00
'tooc :c< oc ot
C4 0t0'
.-
0 ,
CO (0 &0417.-P- 0r 04kCo Ocr) i
t_ t_~~~~~~~~~~~~~~~~~~c
0 oo- Z - -t -n 5)r
0 o04~4~0oo* o
C' b' o ~- ~o ~-~ 00
~~-~- ~- ~-
0 0 0 0 0 0 0 CO 0~~~~~~~~~~~~=~ 0 0 0 0 0 0 0 0
o~~ 0>?00 > ,szs;sz o o o o o~~~~~~~~~~~~~~*S S o< S a a a a
r~~~~~~~~~*
0ZZ
osro
784
This content downloaded from 147.251.185.127 on Wed, 18 Mar 2015 11:05:45 UTC
All use subject to JSTOR Terms and Conditions
CYCLICAL UNEMPLOYMENT 785
ranged from -.289 to -.391 (without the H2 term), which is consistent
with equation (7). A random increase of one manufacturing job
will lower manufacturing layoffs by one with approximate probability
'3 and will increase manufacturing accessions by one with approximate
probability 2/3.
The relatively good fit and high significance of Ht and St in estimation
is not particularly surprising. The g( ) function was derived
primarily from identities, the assumption of a relatively stable )
distribution over time, and the assumption that firms do not simultaneously
hire and lay off workers. No other behavioral assumptions
were required. The estimation also required St to be a reasonable
proxy for STt; a sufficient condition for this is that the firm-specific
hiring terms, E, be correlated within industries.
While the estimates in table 2 have little behavioral content, and
thus are better thought of as conditional expectations than as structural
estimation, they do point out the inadequacies of explaining
layoffs (and, as we shall show, unemployment) by purely aggregate
measures. Most empirical macro models have equations explaining
employment growth but pay little attention to the sectoral composition
of that growth. The results of this section suggest that an attempt
to explain layoffs by aggregate employment growth alone ignores an
important determinant of the cyclical pattern of layoffs.
III. Unemployment
Two types of factors determine the level of unemployment in the
economy: those that influence the size of the flow into unemployment
and those that determine the duration of individual unemployment
spells. An important feature of the model below is that sectoral shifts
of demand which would have no impact on the level of unemployment
or income if labor allocation were instantaneous will have an
impact when it takes time for laid-off workers to be reemployed.
Unemployment is simply the time it takes workers displaced from
contracting firms to find employment in expanding firms. The quantity
of unemployment generated by shifts of employment demand will
thus depend on the speed with which workers find new jobs. If
workers have strong firm or industry attachments, due in part to firmand
industry-specific skills and to wage premiums associated with
seniority, they are reluctant to seek employment in other sectors of
the economy. Thus the process of adjustment to sectoral shifts tends
to be slow and typically involves significant unemployment before
labor adjusts fully to new patterns of employment demand.
In this section, an attempt will be made to measure directly the
impact of dispersion in employment demand on aggregate unemThis
content downloaded from 147.251.185.127 on Wed, 18 Mar 2015 11:05:45 UTC
All use subject to JSTOR Terms and Conditions
786 JOURNAL OF POLITICAL ECONOMY
ployment. In the first part of this section, a reduced-form unemployment
equation is derived from a simple flow model of the unemployment
process. The equation is then estimated, and the impact of
sectoral shifts on cyclical unemployment is assessed.
The starting point of our analysis is the flow identity
AUt = Olt-02t, (9)
where Oltis the flow into unemployment in period t, 02t is the flow out
of unemployment in period t, and AUt is the change in the unemployment
rate. All labor market variables will be expressed as rates
throughout this section.
The flow into unemployment consists of three components: layoffs,
quits that have not found employment prior to leaving their last job
(less than 40 percent of total quits according to Mattila [1974]), and
labor force entrants. We model the flow into unemployment as
Olt = Lt + ao + a1Qt + qlt, (10)
where eq is a random disturbance and the last three terms in (10) are
designed to capture the nonlayoff flow into unemployment.
Layoffs in the model will be determined by a linearized version of
equation (6):
Lt = bo- b1(Et + Qt) + b2ft + q2t- ( 1)
The level of voluntary labor turnover is given by
Qt = O- ClUt + 273t- (12)
The common assumption that quits vary inversely with labor market
tightness, as measured by Ut, is embodied in (12). A somewhat more
general version of (12) would make Qt a function of St and a measure
of unexpected monetary growth, DMR, but this generalization will
not affect the final specification of the unemployment equation that
will be estimated later in this section.
In an economy where prices and wages adjust rapidly, where little
uncertainty exists about the distribution of prices and wages in the
economy, and where firm-specific skills and worker-firm attachments
are relatively unimportant, the distribution of time it takes unemployed
workers to find employment would be relatively constant. One
possible way of modeling the flow out of unemployment under these
assumptions would be to assume a constant probability p of finding a
job in every period. In this case the flow out of unemployment would
be a constant fraction of last period's unemployment, 02t = pUt-IThese
assumptions, however, are not very realistic, and they run
counter to the popular practice of modeling unemployment duration
as a function of unanticipated inflation. There is a strong theoretical
This content downloaded from 147.251.185.127 on Wed, 18 Mar 2015 11:05:45 UTC
All use subject to JSTOR Terms and Conditions
CYCLICAL UNEMPLOYMENT 787
foundation in recent search and contract theories of unemployment,
as well as considerable empirical support, for the hypothesis that
unanticipated inflation or unanticipated monetary policy affects unemployment
duration. We will therefore make 02t a function of unanticipated
monetary policy, in addition to last period's unemployment:
k
02t = pUt-1 + E aiDMRt-i + 4t, (13)
i=O
where DMR is Barro's' measure of unanticipated monetary policy.
We chose the DMR series, rather than constructing an entirely new
series on expectational errors, because it works well in explaining
unemployment movements and because many readers are familiar
with its construction and properties, which facilitates comparison of
the model estimated in this section with existing studies.
The final equation necessary to close the model is the identity:
Et + Ut 1. (14)
Equations (9)-(14) can be solved to yield a dynamic reduced-form
equation for the unemployment rate in terms of the exogenous variables
of the model o-and DMR and the predetermined variable Ut-:
k
Ut= Bo + Blut - X X1DMRt-i+ B2Ut-1 + met- (15)
i=O
Equation (15) gives the specification of the unemployment equation
that will be estimated in this paper. It should be noted that, while an
attempt has been made in this section to derive (15) from structural
relationships, it is similar to more ad hoc unemployment equations
that have been estimated elsewhere in the literature (see, e.g., Sargent
1973 or Barro 1977), with the exception of the St term. It is the
inclusion of St that distinguishes (15) from previous models and,
given the simplified derivation of (15), a more ad hoc interpretation
of its origins may be appropriate.
Table 3 presents estimates of several versions of equation (15),
using annual data for the nonagricultural economy over the postwar
period. Again it was necessary to construct a proxy for St; however,
here the proxy had to measure sectoral shifts beyond those in manufacturing
industries. The variable (it was constructed using the
common 11-industry decomposition of aggregate employment according
to the following formula:
x (A logXat - A log Xt)2j
TIhe DMR series used in estimation was up(late(1 from Barro (1981). A description
of the series can be found in Barro (1977).
This content downloaded from 147.251.185.127 on Wed, 18 Mar 2015 11:05:45 UTC
All use subject to JSTOR Terms and Conditions
(0o o~ to 0 01 ~
0 0 -010
O(C _ 00o - t
- Q0 in cq It in
01 mP- 0c -_
00j (0 00 00 0- t- C)
cqo ooo
(00100 ~00 1- 00 ' (000c
r; o oo on o _ o o
H * * _ _
Ct& 00_CJ?. 00. - 0
00l (o 04sr --o , N4
* ( 6 0* *. *_
0 '
qn 0xQ :)C )Q
- I _ C
_ X O o
o-r~0) 0 -4t - SxC
00 C S --- mi)-- 0- t00-I--
;~~~~~~~- C; 4 603 r- cr; C
S : I I "-I I I I I
mo~~ * -rc0ooo C
CC&00&(000~~~C) C
0~~~~~?0I - In ?
s0 jG l- - _ I - I
CO E~~~~~~~~~~~~
788
This content downloaded from 147.251.185.127 on Wed, 18 Mar 2015 11:05:45 UTC
All use subject to JSTOR Terms and Conditions
CYCLICAL UNEMPLOYMENT 789
where xitis employment in industry i and Xtis aggregate employment.
Figure 2 contains a plot of the constructed 5rtseries along with the
unemployment rate; actual values for the constructed series appear in
table 4. A time trend was also included in the equations to capture
demographic and other changes that may have occurred in the labor
market over the period.
All coefficient estimates are of the correct sign and are highly
significant with the exception of coefficients for some of the longer
lags of St and DMRt. Further, the estimates are quite robust to
changes in the sample period and changes in specification. Equation
(15) may be solved for Ut(with Ut-, eliminated) in terms of infinite lags
on o-t,DMRt, and m*t.Equation (4) of table 3 excludes Ut, and includes
longer lags on 0rtand DMRt. Although Ut, is highly significant and
clearly belongs in the model, its exclusion does not alter the signs or
statistical significance of the other included variables, as is so often the
case in simple time-series models. In addition, the estimated long-run
ot multiplier is not terribly sensitive to the inclusion of Ut-1: In
equations (1) and (3), with Ut-1 included, the long-run 5rtmultipliers
are 105 and 118, respectively. In equation (4), with Ut-1excluded, the
long-run multiplier is 88.
One characteristic of particular interest in the estimates is the effect
of excluding Sr from the equation. While the fit of the equation
worsens considerably when 5ris dropped, the magnitudes of the DMR
coefficients change relatively little. This result follows from the fact
that Sris virtually orthogonal to DMR. Consider the following regression
of 6rton its own lagged values and current and lagged values of
DMR:
rt .024 - .051Srt-l + .033t-2 - .108DMRt - .045DMRt-1,
(.005) (.193) (.106) (.171) (.170)
R2= .035, R2 -.103.
7
|- 00a0
.03
.02
1950 1955 1960 1965 1970 1975 1980
FIGJ. 2
This content downloaded from 147.251.185.127 on Wed, 18 Mar 2015 11:05:45 UTC
All use subject to JSTOR Terms and Conditions
79( JOURNAL OF POLITICAL ECONOMY
TABLE 4
UNEMPLOYMENT SERIES
DETREN DED
YEAR Ut U* Ut-U* Ut U* &t
1948 3.8 ... ... ... ... .0262
1949 5.9 5.3 .6 6.9 6.3 .0482
1950 5.3 5.0 .3 6.3 6.0 .0251
1951 3.3 5.3 -2.0 4.2 6.2 .0461
1952 3.0 4.1 -1.1 3.9 5.0 .0138
1953 2.9 3.9 - 1.0 3.7 4.7 .0283
1954 5.5 5.4 .1 6.2 6.1 .0508
1955 4.4 4.5 -.1 5.0 5.1 .0147
1956 4.1 3.8 .3 4.7 4.4 .0186
1957 4.3 4.0 .3 4.8 4.5 .0248
1958 6.8 5.5 1.3 7.2 5.9 .0489
1959 5.5 4.9 .6 5.8 5.2 .0160
1960 5.5 4.2 1.3 5.7 4.5 .0181
1961 6.7 4.6 2.1 6.9 4.8 .0282
1962 5.5 4.3 1.2 5.5 4.4 .0148
1963 5.7 4.0 1.6 5.7 4.1 .0156
1964 5.2 4.0 1.2 5.1 4.0 .0158
1965 4.5 4.1 .4 4.3 3.9 .0149
1966 3.8 4.3 -.5 3.6 4.1 .0194
1967 3.8 4.7 -.9 3.5 4.4 .0216
1968 3.6 4.7 - 1.1 3.2 4.3 .0157
1969 3.5 4.7 - 1.2 3.0 4.2 .0157
1970 4.9 5.7 -.8 4.3 5.2 .0347
1971 5.9 6.5 -.6 5.3 5.8 .0325
1972 5.6 5.8 -.2 4.9 5.1 .0133
1973 4.9 5.5 -.6 4.1 4.8 .0192
1974 5.6 5.6 -.0 4.7 4.8 .0196
1975 8.5 7.9 .6 7.5 6.9 .0583
1976 7.7 7.1 .6 6.7 6.1 .0138
1977 7.0 6.3 .7 5.9 5.2 .0171
1978 6.0 6.2 -.2 4.8 5.0 .0200
1979 5.8 6.1 -.3 4.5 4.9 .0165
1980 7.1 6.7 .4 5.7 5.4 .0284
Virtually none of the variance of ort is explained by its own lagged
values or by DMR. We can thus strongly reject Granger causality of 6rt
by DMR, which lends some support to our assumption that o-is truly
exogenous. Further support of the hypothesis that o- is exogenous is
provided by the low explanatory power of the other included variables
in the unemployment equation:
(rt = .042 - .013036t - .094DMRt - .128DMRt(.009)
(.198) (.156) (.165)
- .0032Ut1- 6.4E - 5T,
(.0022) (3.OE - 4)
R2= .173, R2 = .020.
This content downloaded from 147.251.185.127 on Wed, 18 Mar 2015 11:05:45 UTC
All use subject to JSTOR Terms and Conditions
CYC(LICAL UNEMPLOYMENT 791
IV. The Natural Rate
The regression results of Sections II and III suggest that the process
of labor reallocation in response to the shifting pattern of employment
demand is a significant source of cyclical unemployment. In this
section, a measure of the unemployment induced by the fluctuations
of o-t will be constructed and discussed.
Since the widespread acceptance of the natural rate hypothesis,
economists have come to view cyclical unemployment as deviations of
unemployment from some relatively stable natural rate. The natural
rate is thought of as the level of frictional unemployment (necessary
to carry out the continuous process of labor allocation) that would
occur in a steady state when agents correctly perceive the distribution
of prices and wages throughout the economy. The view that the
natural rate is relatively constant and that it is the unemployment
necessary to accommodate the process of labor reallocation seems to
conflict with the regression results. The latter would seem to indicate
strong cyclical patterns in the volume of required labor allocation.
Part of the source of this contradiction stems from the variety of
meanings given to the term "natural rate." In much of the microfoundation
literature (see, e.g., Lucas and Prescott 1974), equilibrium
unemployment arises because firms face continuous fluctuations in
demand for their products even when aggregate demand is stable.
There is nothing in this literature to suggest that the equilibrium
unemployment will be time invariant. Rather, we would expect periods
of rapidly shifting demand-whether induced by changes in
taste, import prices, technological change, or whatever-to be associated
with increases in the natural, or equilibrium, rate of unem-
ployment.
In much of the recent equilibrium business cycle literature, these
events, which might be thought of as leading to shifts of demand and
increases of the natural rate, are modeled as aggregate supply shocks.
Unemployment is viewed as consisting of three components: a relatively
constant natural rate, unemployment associated with expectational
errors, and unemployment associated with supply shocks.
Given this characterization, it may not be unreasonable to view the
natural rate as constant or evolving slowly over time as a result of
changes in the demographic and institutional characteristics of the
labor market. Here, however, the term "natural rate" must be given a
somewhat different interpretation, that is, the level of unemployment
associated with the average or typical quantity of labor reallocation
required within the economy. Events that require unusual labor reallocation
lead to deviations from, rather than shifts of, the natural
rate.
In what follows we will construct a measure of the natural rate that
This content downloaded from 147.251.185.127 on Wed, 18 Mar 2015 11:05:45 UTC
All use subject to JSTOR Terms and Conditions
792 JOURNAL OF POLITICAL ECONOMY
is more closely related to the microfoundations concept than the
business cycle model concept. In terms of the unemployment equation
(15), we will define the natural rate as that level of unemployment
that would have existed if current and past values of DMRt and mthad
been identically zero. It is given by the equation
ZB2(Bo + Bl-t-i + B3Tt-), (16)
i=O
where B3 is the coefficient of the time trend variable T.
A series for the natural rate was calculated using the estimates of
equation (3) in table 3. They are presented along with the actual
values of Ut and (rtin table 4 and figure 2. While equation (16) makes
U* a function of the infinite history of (it, the weights die off so
quickly that only a few years of lagged data are necessary to construct
reasonable approximations. Nevertheless, the requirement of previous
observations of (rtprecludes estimating U* over the entire sample.
The U* series varies significantly over the 1949 and 1980 period,
tracking movements in Ut reasonably well. Over half of the variance in
Ut can be explained by U*; the simple correlation between Ut and U*
is .74. Of course, part of the reason for the high correlation between
Ut and U* is that they both have trended upward over the postwar
period. Column 5 of table 3 presents a constructed natural rate series
using the formula given in equation (16), but replacing the value of
the time trend with its average value over the period. It is thus a
detrended measure of the natural rate that reflects only fluctuations
due to current and lagged values of 6rt.The correlation of this series
with the detrended unemployment series is .60.
The pattern of Ut and U*, revealed in figure 2, points out the
variety of sources of cyclical unemployment. Unemployment over the
seventies is particularly well explained by U*, which suggests that real
factors influencing the natural rate were the major source of cyclical
unemployment. This observation is consistent with the popular view
that supply shocks were responsible for much of the cyclical activity of
the last decade. In marked contrast, unemployment over the sixties is
characterized by significant deviations from U*. In 7 of the 8 years
between 1956 and 1963, monetary growth was below its expected
level, as indicated by DMR, which resulted in unemployment
significantly above the natural rate in the late fifties to early sixties. In
5 of the 6 years between 1964 and 1969, monetary growth was above
its expected level, which resulted in unemployment below the natural
rate in the late sixties.
V. Concluding Remarks
While explicit consideration of stabilization policy is beyond the scope
of this paper, there are policy implications in its findings. Our analyThis
content downloaded from 147.251.185.127 on Wed, 18 Mar 2015 11:05:45 UTC
All use subject to JSTOR Terms and Conditions
CYCLICAL UNEMPLOYMENT 793
sis suggests that much of the unemployment of the seventies could not
have been avoided through aggregate monetary and fiscal policies.
Such policies may have been successful in delaying or smoothing the
cyclical pattern of unemployment, but since inadequate demand was
not the source of unemployment, aggregate demand policies were not
an appropriate cure. This is in marked contrast to the early sixties,
where our estimates suggest that unemployment was well above the
natural rate, due in part to several years of lower-than-anticipated
monetary growth (as measured by DMR) or, more generally, to a
contractionary set of aggregate demand policies. The inability of
aggregate demand policies to eliminate unemployment induced by
"real"factors does not rule out a role for all policy. Targeted demand
policies, or supply-side policies aimed at easing the transition of workers
from declining to growing sectors of the economy, or policies to
stimulate productivity in declining sectors may be appropriate. The
nontrivial task of analyzing stabilization policy within a disaggregated
framework would seem to be a productive line for future research.
The model developed in this paper does not claim to give a total
description of the operation of labor markets or the process of unemployment.
It does, however, suggest some of the limitations of aggregate
models that do not explicitly account for the multisectoral
character of production and employment and the imperfect shortrun
mobility of resources between sectors. It also suggests that nonmonetary
factors, whether viewed as supply shocks or as shifts of the
natural rate, have been an important source of cyclical unemployment
and deserve greater attention in the business cycle literature.
References
Barro, Robert J. "Unanticipated Money Growth and Unemployment in the
United States." A.E.R. 67 (March 1977): 101-15.
. "Intertemporal Substitution and the Business Cycle." J. Monetary
Econ. 14 (suppl., Spring 1981): 237-68.
Hall, Robert E. "A Theory of the Natural Unemployment Rate and the
Duration of Employment."J. MonetaryEcon. 5 (April 1979): 153-69.
Lucas, Robert E., Jr., and Prescott, Edward C. "Equilibrium Search and
Unemployment." J. Econ. Theory7 (February 1974): 188-209.
Mattila, J. Peter. "Job Quitting and Frictional Unemployment." A.E.R. 64
(March 1974): 235-39.
Phelps, Edmund S., et al. Microeconomic Foundations of Employment and Inflation
Theory.New York: Norton, 1970.
Sargent, Thomas J. "Rational Expectations, the Real Rate of Interest and the
'Natural' Rate of Unemployment." Brookings Papers Econ. Activity, no. 2
(1973), pp. 429-72.
This content downloaded from 147.251.185.127 on Wed, 18 Mar 2015 11:05:45 UTC
All use subject to JSTOR Terms and Conditions