Beruflich Dokumente
Kultur Dokumente
Author(s): Walter Y. Oi
Source: Journal of Political Economy, Vol. 70, No. 6 (Dec., 1962), pp. 538-555
Published by: The University of Chicago Press
Stable URL: http://www.jstor.org/stable/1828778 .
Accessed: 06/09/2011 11:43
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.
The University of Chicago Press is collaborating with JSTOR to digitize, preserve and extend access to Journal
of Political Economy.
http://www.jstor.org
LABOR AS A QUASI-FIXEI) FACTOR
WALTER Y. 011
UIfniversityof Washington
T HE cyclical behavior of labor mar- borne out by the available evidence. Thus
kets reveals a number of puzzling my theory provides a unified explanation
features for which there are no for various aspects of the cyclical func-
truly satisfying explanations. Included tioning of labor markets.
among these are (1) occupational differ-
ences in the stability of employment and I. A SHORT-RUN THEORY OF
employment, (3) the persistence of differ- According to the theory presented here
ential labor turnover rates, and (4) dis- cyclical changes in employment are ex-
criminatory hiring and firing policies. I plained by differential shifts in factor de-
believe that the major impediment to mands and supplies. The first two sec-
rational explanations for these phenom- tions develop a theory of factor demands
ena lies in the classical treatment of labor assuming rigid wage rates. In the next
as a purely variable factor. two sections this assumption is relaxed,
In this paper I propose a short-run the- allowing for variations in factor supplies.
ory of employment which rests on the
A. NATURE OF THE CLASSICAL SHORT-RUN
premise that labor is a quasi-fixed factor.
ADJUSTMENT PROCESS
The fixed employment costs arise from
investments by firms in hiring and train- In the classical short-run model certain
ing activities. The theory of labor as a paths of adjustment are barred to the
quasi-fixed factor is developed in Part I. firm. These barriers usually postulate
In Part II, the implications of this theory the presence of fixed factors, short-run
are subjected to various empirical tests. changes in output being effected by vary-
Finally, Part III turns to an examination ing only the remaining factors.
of alternative theories and an extension Changes in the amount demanded of
of my theory to a theory of occupational any factor are composed of two parts:
wage differentials. (a) response to changes in the rate of out-
The concept of labor as a quasi-fixed put--the scale effect and (b) response
factor is, in my opinion, the relevant one to variations in relative factor prices-
for a short-run theory of employment. the substitution effect. With an assump-
Its implications are amenable to empiri- tion of rigid wage rates, the substitution
cal verification and are, in the main, effects may be neglected and attention
focused on the scale effects.
1 This paper is taken from my unpublished doc-
toral dissertation "Labor as a Quasi-fixed Factor of Consider a firm faced by a decline ill
Production" (University of Chicago, 1961). I wish product demand. The adjustment proc-
to express my indebtedness to Professors A. C. Har- ess involves a reduction in output ac-
berger, H. G. Lewis, and A. E. Rees. A reading of
two unpublished articles by Professor Gary S. Becker companied by a decline in the demand
led me to revise the theory substantially. Financial for each variable factor. There is no rea-
and clerical assistance from the Social Science Re- son to expect that the demands for Lill
search Council and the Transportation Center,
Northwestern University are also gratefully ac- variable factors will be (lecreased by the
knowledged. same proportion. The reduced demands
538
LABOR AS A QUASI-FIXED FACTOR 539
for variable factors led to an increase in For analytic purposes fixed employ-
the relative employment of fixed factors. ment costs can be separated into two
In a sense, the firm now employs too categories called, for convenience, hiring
much of the fixed factors and would, and training costs. Hiring costs are de-
therefore, try to substitute fixed factors fined as those costs that have no effect
for variable factors. Consequently, those on a worker's productivity and include
variable factors that tend to be most sub- outlays for recruiting, for processing pay-
stitutable for, or least complementary roll records, and for supplements such
with, the fixed factors will experience the as unemployment compensation. These
greatest relative declines in demand due costs are closely related to the number of
to any given decrease in product demand. new workers and only indirectly related
The converse holds for an increase in to the flow of labor's services. Training
product demand. Thus the variable fac- expenses, on the other hand, are invest-
tors that are most substitutable with the ments in the human agent, specifically
fixed factors will exhibit the greatest rela- designed to improve a worker's produc-
tive shifts in factor demands. tivity. The effect of training on produc-
B. DEMAND FOR A QUASI-FIXED
tivity could be summarized by a produc-
FACTOR OF PRODUCTION tion function showing the increment to a
A quasi-fixed factor is defined as one worker's marginal value product in the
whose total employment cost is partially tth period, AMt, due to an investment in
variable and partially fixed. In the classi- training of K dollars per worker.2
cal short-run model all factors are classi- AMt = g(K). I
fied as either variable or fixed. Each fac-
tor may, however, possess a different de- The total discounted cost, C, of hiring an
gree of fixity along some continuum rath- additional worker is the sum of the pres-
er than lie at one extreme or the other. ent value of expected wage payments, the
From a firm's viewpoint labor is surely hiring cost, H, and training expense, K.
a quasi-fixed factor. The largest part of T
total labor costs is the variable-wages bill C= J:We(l + r) t+H+K, (2)
representing payments for a flow of pro- t=O
ductive services. In addition the firm
where Wt is the expected wage in the tth
ordinarily incurs certain fixed employ-
period, r denotes the rate at which future
ment costs in hiring a specific stock of
costs are discounted, and T denotes the
workers. These fixed employment costs
expected period of employment. The to-
constitute an investment by the firm in
tal discounted revenue, Y, generated by
its labor force. As such, they introduce
the additional worker is similarly defined
an element of capital in the use of labor.
as the present value of his expected mar-
Decisions regarding the labor input can
no longer be based solely on the current ginal value products that, in each period,
relation between wages and marginal val- consist of his marginal product without
ue products but must also take cogni- training, MtI,and the increment due to
zance of the future course of these quan- his training, AMt.
tities. The theoretical implications of la- 2 The training activity typically entails direct
bor's fixity will be analyzed before turn- money outlays as well as numerousimplicit costs
such as the allocation of old workers to teaching
ing to the empirical magnitude of these skills and rejectionof unqualifiedworkersduringthe
fixed costs. trainingperiod.
540 WALTER Y. 01
competingemployments,as, for example,
=\M At)1 r-1- (3)s |-/ training of workers to operate com-
the
t=Oo
puters or to read railroadtariffs.
Profitswill be maximizedwhen the total Rationalbehaviorimpliesthat the bulk
discounted cost of an additional worker of a firm's investment in training must
is just equal to the total discounted be devoted to specific training. If train-
revenue. ing were completely general, all returns
would accrue to the worker and none to
T
the firm. Upon completion of general
11 +K= E (Mt+Alf t- WO (4) training, the workerwould find that his
t=O
marginal productivity to several firms
X (1 + r)t has been increased. He could now de-
Equation (4) yields the first implication. mand a higher wage, either from a com-
In equilibriuma worker'stotal marginal peting firm or from his presentemployer.
product, Mt + AMt, must exceed the In either case, the net value of the train-
wage rate, Wt, so long as the firm incurs ing to the firm would be reducedto zero.
any fixed employment costs. Even un- Indeed, a firm could capture these re-
der perfect competition wages would be turns only if there were impedimentsto
equated to marginal value products if competitionsuch as imperfectknowledge
and only if labor is a completely variable or binding labor contracts. Thus, no ra-
factor. tional firm would underwritecompletely
At this point, a digressionon the firm's generaltraining.3If, however, trainingis
investment in training is in order. The specificto a firm,then the worker'salter-
net value of trainingto the firm is simply native marginal product remains unaf-
the present value of the expected incre- fected. In this latter case, the firm could
ment in marginal value product, AMt, weigh the expected returns from this in-
due to training. vestment against the training cost. To
simplify the analysis, I shall assume that
T
the firm bears all specific training costs.
V = E AMt(1 +r- (5) As will be shown in section D below, the
t =O
implicationsof the theory are not serious-
An investment in training will prove ly affected by relaxing this assumption.
profitableif the net value to the firm, V, Returning to the equilibrium condi-
exceedsthe trainingexpense,K. Concep- tion, it is clear that some expectations
tually training may be categorized as model must be formulatedsince the vari-
eithergeneralor specific.Specifictraining ables refer to future quantities. Suppose
is definedas that which increasesa work- that the firm formulates the following
er'sproductivityto a particularfirmwith- single-valuedexpectations:
out affecting his productivity in alterna- Wt = W*, Mt = M*, AtJ = AM*
tive employments.The time requiredto (6)
adapt workers to the firm's particular (for allt= 0, 1, 2, ...T).
productionprocesses,or to its accounting 3 That a firm offers general training to its em-
and marketingprocesses,exemplifiesspe- ployees does not necessarily imply that the firm
cific training. General training, on the underwrites the general training expense. The work-
er may bear the training cost by accepting a lower
other hand, is defined as that which in- wage than that which he could obtain in some alter-
creasesa worker'sproductivity in several native employment.
LABOR AS A QUASI-FIXED FACTOR 541
Substituting these expected values into the equilibrium condition, equation (9),
equation (4), the equilibrium condition is satisfied by every factor or grade of
reduces to labor. For a competitive firm, a decline
+ H+K in product demand is equivalent to a fall
M* +AM*= T in product price, P*.4 The relevant com-
E
t=O
(1+ r)-- t ~7)(7 parison for short-run profit maximization
is that between total expected marginal
The concept of a periodic rent, R, may value product and the expected wage
be defined as rate, representing the variable compo-
nent of the total employment cost. Thus
T the short-run equilibrium condition ap-
(8))
1:( 1+ r)-t plicable to cyclical declines in product
t=O demands becomes
The periodic rent represents the fixed M* +AM* =W*. (10)
employment costs during each period. It
is the surplus that must be earned by The employment of a quasi-fixed factor
each worker in order to amortize the will only be reduced when M* + AM*
initial fixed employment costs over the falls below W*.
expected period of employment, realizing For a completely variable factor, the
a rate of return of r per cent on this in- long- and short-run equilibrium condi-
vestment. Thus, the equilibrium condi- tions, (9) and (10), are equivalent. Any
tion may be rewritten as decline in P* will reduce the demand for
a variable factor; with falling employ-
M*+AM*=W*+R. (9) ment, the variable factor's marginal phys-
In equilibrium, the total expected mar- ical product will be increased until the
ginal value product must exceed the ex- equilibrium conditions are again satis-
pected wage rate by the amount of the fied. The decline in P* may not be suf-
periodic rent. The degree of fixity, f, of ficient, however, to warrant a reduction
a factor will be defined as the ratio of the in the demand for some other factor with
periodic rent, R, to the total employment a higher degree of fixity. In fact, there is,
cost, W* + R. A value of zero corre- for each quasi-fixed factor, a critical price
sponds to a completely variable factor at which the firm will reduce its demand
while the degree of fixity,f, of a complete- for that factor. Furthermore, the critical
ly fixed factor is designated by a value price, in relation to the long-run equi-
of unity. librium price, will be lower for factors
The periodic rent drives a wedge be- with higher degrees of fixity. Thus, a
tween the wage rate and the marginal given decline in P* may induce a reduc-
value product, the relative magnitude of tion in demand for factors with low de-
the wedge being measured by the degree grees of fixity without affecting the de-
of fixity. In the short run, any fixed em- mand for factors with higher degrees of
ployment costs associated with the ac- fixity.
quisition of a labor force in prior periods Consider next, the case of an increase
are sunk costs; as such they should not 4The term, A* + 'AI*, in eq. (9) denotes the
affect a firm's short-run decisions. expected marginal value product. For a competitive
firm, this is simply the expected product price, P*,
Suppose that the firm is initially in a times the expected marginal physical product, X*
position of long-run equilibrium; that is, + AX*.
542 WALTER Y. 01
tive magnitude of this wedge, measured able. If all factor demands have the same
by the degree of fixity, differs among oc- wage-rate elasticity, and if all wages fall
cupations or grades of labor. In a sense, by the same proportion, the ultimate de-
the periodic rent forms a buffer absorbing cline in the employment of high fixity
short-run variations in a factor's margin- labor will still be less than the decline in
al value product. Thus, short-run changes the employment of labor with lower de-
in product demands lead to differential grees of fixity. On the other hand, if the
shifts in factor demands, depending on elasticity of demand and/or the relative
the degree of fixity. Factors with lower fall in wages are greater for labor with
degrees of fixity will experience relatively lower degrees of fixity, there is a tendency
greater shifts in demand as the result of in the opposite direction. This could off-
any given short-run change in product set or perhaps even reverse the differen-
demand. tial shifts in employment predicted by
the fixed-cost hypothesis.
C. SHORT-RUN FACTOR SUPPLIES WViththe exception of the 1929-33 re-
In the preceding sections all factor cession, the available evidence suggests
supplies were assumed to be infinitely a widening of occupational wage differen-
elastic at fixed market wage rates. This tials in the downswing and a narrowing
assumption is surely reasonable for a sin- in the upswing.7 Even in the 1929-33 re-,
gle competitive firm. The supply curve cession, the data are consistent with a
of labor as a whole or of certain skill hypothesis that there was no change in
categories may, however, be less than in- the occupational wage structure. As will
finitely elastic or may shift over time. be shown in Part II, the wage rate of an
This section deals with the effect of vari- occupation is highly correlated with its
ations in short-run factor supplies on
degree of fixity. Thus, over a cycle, the
relative factor employments.
high-wage occupations with greater de-
Consider the extreme case in which all
grees of fixity exhibit smaller relative
factor supplies have zero elasticity and
markets adjust rapidly. In a downswing, changes in wage rates. Finally, I suspect
the demand for labor shifts to the left, that the high-wage, skilled jobs tend to
the percentage shift being smaller for fac- be most complementary with the fixed
tors with higher degrees of fixity. The factor, capital. Ceteris paribus, greater
downward shifts in demand lead to re- complementarity with the fixed factor
ductions in wage rates only; factor em- implies a lower elasticity of demand. The
ployments are determined exogenously evidence thus indicates smaller relative
by the assumed supply conditions. The changes in wage rates and lower elastici-
relative decline in wage rates will be ties of demand for factors with higher
smaller for factors with higher degrees of degrees of fixity. The observed short-run
fixity. This case is, however, inconsistent variations in factor supplies therefore
with the emergence of involuntary un- predict differential shifts in employment
employment and may be dismissed as contrary to those implied by the fixed-
unrealistic. cost hypothesis.
Granted the existence of involuntary
in cyclical downswings, 7 Phillip W. Bell, "Cyclical Variations and Trends
unemployment
in Occupational Wage Differentials in American In-
the relevance of the concept of a market dustries Since 1914," Reziez of Economics and Sta-
sup)p)lycurve becomes highly (question- lislics, XXXIII, No. 4 (November, 1954), 329-37.
544 WALTER Y. 01
mium upon completion of the specific To sum up, the specificity of labor to
training. If the present value of his ex- a firm favors the establishment of policies
pected wage premiums exceeds his share to lengthen expected periods of employ-
of the training cost, the arrangement ment. To achieve lower labor turnover,
would be acceptable to a rational worker. a firm might adopt a policy by which
Consider the extreme case in which all workers were required to share specific
specific training costs are paid by the training expenses and were rewarded by
worker.8 The worker clearly prefers to subsequent wage premiums. Under such
remain with the firm since his marginal a policy, both workers and firm would
product and wages in the specific firm benefit from lower labor turnover. Final-
are higher than his alternative marginal ly, the gains from lower labor turnover
product. In this case, however, all labor are greater for factors with higher degrees
costs become variable to the firm, and of fixity. Thus a higher degree of fixity
the firm has no motive for retaining these leads not only to greater stability of em-
workers during a cyclical downswing. ployment in terms of numbers or man-
A sensible policy is to have both par- hours employed but also to lower labor
ties share the specific training expense. turnover rates.
If both workers and firm have made in-
II. TESTS OF THE THEORY
vestments in specific training, both can
gain from longer expected periods of em- A. WAGE RATE AS INDEX OF DEGREE
OF FIXITY
ployment. The specifically trained work-
er could receive a wage above his alterna- A direct test of the theory would re-
tive marginal product but below his total quire estimates of the degree of fixity for
marginal product to the firm.9 At the different grades of labor. Since such data
same time, the firm could lower its total are not readily available,10 an auxiliary
employment cost by amortizing the fixed variable that is closely correlated with
employment costs over a longer period the degree of fixity must be employed as
of time. Short-run variations in factor a proxy. The use of an occupation's wage
supplies would tend to be smaller for rate as the proxy seems justifiable and
workers who had invested in specific derives support from the empirical evi-
training. Thus the payment of wage pre- dence presented in this section.
miums and the sharing of the specif- A study undertaken by the Interna-
ic training costs promote even greater tional Harvester Company (hereinafter
stability of employment for specifically referred to as "IH") in 1951 estimated
trained workers. fixed employment costs for three job
categories." The total fixed cost asso-
8 Firms and workers may differ in their evaluation
ciated with an annual labor turnover
of expected periods of employment and pertinent of 28,623 workers was $15.9 million, or
discount rates. As a result, the optimum investment
per worker in specific training may differ depending roughly 5.4 per cent of total wage pay-
on the incidence of these training costs. ments. The component cost items were
9 Where training is specific to a firm the situation placed in three categories. The IH esti-
closely resembles one of bilateral monopoly. From
10 The necessary data include (1) the initial fixed
the firm's viewpoint, the specifically trained worker
is differentiated from other workers. At the same employment cost, H + K, (2) the discount rate, r,
time, the worker finds that his value to his present (3) the expected period of employment, T, and (4)
employer is higher than his value to some competing the expected wage rate, W*.
employer. I wish to thank Professor Rees for point- 11International Harvester Company, "The Costs
ing out this analogue. of Labor Turnover" (mimeographed), 1951.
546 WALTERY. OI
mates of the amount invested in each is used to recoup any lost productive
new worker are presented in the first time; the result is a substantial reduction
four columns of Table 1; my revised esti- in this cost item. "Intrawork transfers"
mates appear in the last column. represents additional training expenses
Hiring costs comprised less than 5 per due to transfers of old workers to new
cent of total fixed costs. They include the jobs within the plant.
costs of terminating, laying off, and re- Finally, the cost of unemployment
calling since each worker has a positive compensation represents the difference
probability of passing through each stage between the firm's actual contributions
during his prospective tenure with the and the minimum legal contributions for
firm. The IH study weighted each cost unemployment compensation.13 Higher
by its relative frequency. labor turnover will, in general, increase a
TABLE 1
AMOUNT OF MONEY INVESTED PER NEW EMPLOYEE, INTERNATIONAL HARVESTER COMPANY, 1951
Hiring costs:
Recruiting................... $ 4.33 $ 86.38 . . . .. $ 5.48 $ 5.48
Hiring...................... 13.23 29.08 $ 28.89 13.23 13.19
Orientation.................. 1.56 1.56 1.56 1.56 1.56
Terminating ................. 3.77 3.77 3.77 3. 77 3. 77
Laying off ................... 1.21 1.21 1.21 1.21 1.21
Recalling .................... 1.30 1.30 1.30 1.30 1.30
Total ....................... 25.40 123.30 36.73 26.55 26.51
Training costs:
Training .................... 9.08 11,850.00 18,503.00 238.40 151.36
Tools and materials . . . ....... 164.76 41.19 41.19
Unfilled requisitions.......... 14.92 . ............. .............. 83. 12 24.66
Intrawork transfers........... 3.50 .............. .............. 94.14 64.49
Total ....................... 27.50 11,850.00 18, 667.76 456.85 281.70
Unemployment compensation.... 73.52 73.52 73.52 73.52 73.52
Total fixed employment cost..... 126.42 12,046.82 18, 778.01 556.92 381.73
Training cost is by far the major fixed firm's actual contributions.14 The differ-
employment cost. Two items under this ence between the actual and minimum
heading require further explanation. legal contributions resembles hiring costs
"Unfilled requisitions" refers to the im- and is therefore included in fixed em-
plicit cost of lost output, resulting from ployment costs.
the lag between the separation and sub- Each component of total fixed costs
sequent replacement of a worker.12In the appears to rise either at the same rate as
IH study the hourly cost of lost productive 13 Actual unemployment compensation tax rates
time was the gross profits per man-hour differ widely among firms, across states, and over
of $1.84. I have revised the estimate by time. For a discussion of these differences the reader
is referred to R. A. Lester, "Financing Unemploy-
assuming that standby, overtime labor ment Compensation," Industrial and LaborRelations
12 Supplementary data in the JH study showed an
Review, XIV, No. 1 (October, 1960), 52-67.
average lag between separation and replacement of 14 An exception to this generalization arises if all
39.1 hours. The lag was longer for jobs with higher separated workers immediately obtain new jobs.
skill requirements. This is, however, rarely the case.
LABOR AS A QUASI-FIXED FACTOR 547
the wage rate of a job or faster. The data ment." The alternative hypothesis is that
in Table 1 suggest that the ratio of total the high-wage occupations will experi-
fixed employment costs to wage rate in- ence smaller rates of change in employ-
creases as one moves to higher wage jobs. ment.
If the IH estimates of Table 1 are coupled The contingency table was chosen as
with some assumptions about wage rates the experimental design for testing the
and expected periods of employment, it hypothesis. The contingency table tests
is possible to estimate degrees of fixity for the independence of two variables of
for two overlapping groups of workers. classification. The x2 statistic is the cri-
Thus, degrees of fixity of 0.073 and 0.041 terion for the acceptance or rejection of
were obtained for "All employees" and the null hypothesis. The occupations in
"Common laborers."1 The high-wage, a firm are first divided into two equal
highly skilled jobs appear to be asso- groups according to the wage rate of the
ciated with higher degrees of fixity. In occupation. The same occupations are
the subsequent empirical tests I shall again separated, by reference to the ob-
assume that an occupation's wage rate served rates of change in employment
serves as an index of the degree of fixity. over a fixed time interval, into equal
groups. All occupations thus fall into the
B3. CYCLICAL BE'HAVIOR OF EMPLOYMENT
B3Y OCCUPATION
cells of a contingency table where the
cells identify groups of occupations with
During cyclical changes in aggregate
different wages and percentage changes in
employment different rates of change in
The marginal totals, which
employment are observed for different employment.
are fixed in advance, determine the num-
occupations even within the same firm.
ber of occupations in each cell to be ex-
This phenomenon is the subject of the
pected if in fact the null hypothesis is
first empirical test of the theory.
true. The major advantages of the con-
The basic proposition of Part I may
table over alternative test pro-
be stated as follows: those factors with tingency
cedures"6 include the following:
the highest degrees of fixity and the low-
est degrees of substitutability with the 1. No assumptions are required about the joint
fixed factor will experience the smallest probability distribution of the two variables
relative changes in employment due to of classification.
2. The design is fairly insensitive to extreme
any given change in product demand. values. Since the wage rate is an imperfect
For the empirical test occupational wage index, the possibility of a few extreme o1)-
differentials are assumed to reflect dif- servations must be acknowle(dge(d.
ferences in the degree of fixity. The hy-
pothesis submitted to a test is: "there Data for this test were taken from the
is no relation between wage rates and BES industry wage structure studies.
the observed rates of change in employ- 16 Two alternative test procedures are (1) a least-
" In arriving at these estimates I assumed ex- squares regression between the wage rate and the
pected periods of employment of 24 and 12 months percentage change in employment, and (2) a Stu-
for "All employees" and "Common labor." The dis- dent's "I" test for the difference in mean percentage
count rate was fixed at 1 per cent per month; varia- changes in employment for two groups of occupa-
tions of ? 0.5 per cent had very little effect on the tions classified by wage rates. The contingency table
degree of fixity. Finally, the IH study reported aver- design resembles a non-parametric correlation tech-
age hourly earnings for "All employees" of $1.952 nique known as tetrachoric r. This latter technique
in 1951. This figure was converted to a monthly is described in W. J. Dixon and F. J. Masey, Intro-
wage and applied to "All employees." For "Common duction to Statistical Analysis (New York: Mc Graw-
labor" I arbitrarily fixed an hourly wage of $1.50. Hill Book Co., 1951), p. 235.
548 WALTER Y. 01
Each study presents data on employ- low the median wage rate for all oc-
ment, hours, and earnings, classified by cupations. The occupations were then
occupation, sex, state, and industry, for reclassified by the observed percentage
a single year. Since two adjacent studies change in employment into two equal
are needed, attention was focused on the groups. Employment, in this context, re-
period 1928-31. For each of four indus- fers to the number of employees; the al-
tries a sample of states was selected. The ternative concept of man hours yielded
states included in the sample reported virtually identical results. The two-way
roughly the same number of establish- divisions fix the marginal totals which, in
TABLE 2
No. OF OCCUPATIONS IN SELECTED INDUSTRIES, CLASSIFIED BY
WAGE GROUP AND PER CENT CHANGE IN EMPLOYMENT*
Furniture:
High wage ........ 40 18 58 16.345
(29.26) (28.74) (3.841)
Low wage ......... 17 38 55
(27.74) (27.26)
All wages ......... 57 56 113
Foundries:
High wage ........ 20 16 36 0.889
(18) (18) (3.841)
Low wage ......... 16 20 36
(18) (18)
All wages.......... 36 36 72
Men's clothing:
High wage ........ 13 7 20 4.102
(9.76) (10.24) (3.841)
Low wage ......... 7 14 21
(10.24) (10.76)
All wages ......... 20 21 41
* Source of the data and detailed breakdownby states are shown in my thesis, Appendix C,
C-5, C-6, and C-7.
t Numbersin parenthesesdenote the 5 per cent critical values.
ments covered in the two adjacent sur- turn, determine the expected numbers of
veys. The states were also grouped by occupations under the null hypothesis.
geographic region.'7 For each state, the The contingency tables for four indus-
occupations were classified as high wage tries are presented in Tables 2 and 3,
or low wage, depending on whether an where the numbers in parentheses denote
occupation's wage rate was above or be- the expected numbers under the null hy-
" In these studies the BLS attempted wherever pothesis."8For furniture and men's cloth-
possible to cover the same establishments as those ing, the x2statistic indicates that the null
covered in the preceding study of the same industry.
Hence, inclusion of only those states showing ap-
hypothesis can be rejected at a 5 per cent
proximate equality in the number of establishments level of significance. The x2 statistic is
covered tends to minimize errors arising from a shift-
18 The procedure was slightly altered for lumber
ing establishment composition within states. Geo-
graphic grouping was incorporated to minimize dif- mills. The larger number of occupations in this in-
ferential shifts in product demands. dustry permitted a three-way division.
LABOR AS A QUASI-FIXED FACTOR 549
significant at only a 10 per cent level for than their production worker counter-
lumber mills. Finally, the data for found- parts. Thus, according to the fixed-cost
ries are consistent with the null hy- hypothesis, non-production workers
pothesis. The discrepancies between ac- should exhibit greater employment sta-
tual and expected numbers of occupa- bility.
tions in the cells of the contingency tables Consider a simple model in which the
conform to the pattern anticipated by desired rate of employment in the tth
the fixed-cost hypothesis. Given these year, Et*, is a linear function of the rate
systematic discrepancies,'9 the additive of output, Xt. That is, the desired de-
nature of x2 permits a test for the aggre- mand for some factor is solely a function
gate of all four industries. For the four of the scale effect, ignoring all substitu-
industries combined, the null hypothesis tion effects.
can be rejected at a 1 per cent level. E*=a+bXt. (12)
TABLE 3
No. OF OCCUPATIONS IN THE LUMBER INDUSTRY CLASSIFIED BY WAGE
GROUP AND PER CENT CHANGE IN EMPLOYMENT, 1928-30*
* Sourceof data and breakdownby states are shown in my thesis, Appendix C, Table C-8.
t Numbers in parentheses denote the 5 per cent critical values.
Thus, the evidence from the BLS studies Suppose the actual change in employ-
refutes the null hypothesis, and the alter- ment from the preceding period is some
native can be accepted. The wage rate proportion, k, of the desired change.
of an occupation is associated with its Et-Et1 = k(E*-E __). ( 13)
percentage change in employment. Low-
wage occupations, corresponding to low Substitution in (13) by the expression for
degrees of fixity, do experience relatively E* in (12) yields the following reduced
greater changes in employment. form, containing only observable quan-
C. EMPLOYMENT OF NON-PRODUCTION WORKERS tities:
Non-production workers in manufac- Et = ak+ bkXt + ( 1-k)Et-,. (14)
turing industries are more highly paid20
20 Data from the 1954 Census of Manufactures
and tend to be more specific to the firm
indicate that the average annual earnings of non-
19The contingency table design can only uncover production workers always exceeded those of produc-
an association between two variables of classifica- tion workers. The ratio of the earnings of non-
tion. The direction of this association is only dis- production to those of production workers ranged
cernible from an examination of discrepancies be- from a low of 1.47 in chemicals to a high of 2.11 in
tween actual and expected numbers. textile mill products.
55( WALTER Y. OI
The parameter, k, may be interpreted as the 10 per cent level. Furthermore, an ex-
a short-run coefficient of adjustment. amination of seasonally adjusted month-
Higher values of k imply faster short-run ly employment data for the postwar pe-
adjustments. Indeed, a value of unity riod reveals lags in the turning points of
means that actual and desired rates of non-production worker employment. The
employment always coincide. If non-pro- turning points in non-production worker
duction workers are truly specific to the employment for the "All Manufactur-
firm, they should exhibit a smaller k rela- ing" sector lagged from two to six months
tive to production workers. behind the turning points in production
The slope parameters of the reduced worker employment. The smaller short-
form, equation (14), were estimated by run coefficient of adjustment and the lags
least squares using annual data for the in turning points are entirely consistent
"All Manufacturing" sector. The anal- with the higher degree of fixity of non-
ysis was confined to the prewar period, production workers.
1920-39, because of a strong secular trend
D. DIFFERENTIAL UNEMPLOYMENT AND
in employment of non-production work-
THE FIXED-COST HYPOTHESIS
ers during the postwar period2' that de-
stroyed the validity of the demand rela- Involuntary unemployment is never
tion. Output, Xt, was measured by the uniformly distributed among all workers.
Federal Reserve Board's Index of Indus- The fixed-cost hypothesis may help to
trial Production. The least-squares esti- explain some of these differences. Fixed
mates are given by equations (15) and employment costs constitute an invest-
(16), where N, and Pt, respectively, des- ment by the firm in its labor force. A
ignate non-production and production firm faced by a cyclical decline in product
worker employment. demand could protect this investment by
following a discriminatory layoff policy.
~t= a k + .0132 X +.561 1 N,-, To the extent that labor is a quasi-
(.0024) (.1100) (15) fixed factor, each new hire, whether an
(R2 = .7 79 3). addition or a replacement, entails an in-
vestment outlay by the firm for hiring
K
t = A + .0 722 Xt+.331 8Pt-, land training. Suppose that a firm is
(.0146) (.1340) (16) obliged to reduce the rate of employment
of some factor. At any point in time each
(RI= .72 19). firm observes a distribution of its workers
according to their expected periods of
The short-run coefficient of adjustment
employment. The average employment
for non-production workers is 0.44 while
tenure of the workers who remain em-
that for production workers is 0.67; the
ployed can be lengthened by laying off
difference is statistically significant at
the workers with the shortest expected
21 In the prewar period, the ratio of non-produc- periods of employment. Such a discrimi-
tion to production workers exhibited no secular natory layoff policy tends to reduce the
trend. From 1947 to 1958, virtually all the increase
in manufacturing employment is attributable to the -voluntary
quit rate facing the firm. The
growth in non-production worker employment. For policy tends to lower total fixed employ-
a discussion of these trends the reader is referred to rnent costs bly miniimizinigthe number of
lVintel States Bureau of Labor Statistics, ''"Not-
future neW hires tllht w(OUl(l 1)e re(ItUire(
p-o(lutiioll \Workers in I actories, 1919--56,,"AlfIt//l
lilb-r Reivie&, LXX.X, No. 4 (April, 1957), 435 -40. to replaicc o1(l workers who milight voluin-
LABOR AS A QUASI-FIXED FACTOR 551
tarily quit in subsequent periods. Thus, eral, be expected to have longer expected
an implication of the fixed-cost hypoth- periods of employment than either very
esis is that the incidence of unemploy- young or very old workers; this expecta-
ment should be highest for those workers tion has been corroborated by a study of
with the shortest expected periods of em- labor turnover in Swedish factories.22 At
ployment. Since a worker's prospective the same time, married males with spouse
employment tenure cannot be known present tend to be less mobile and should
with certainty and since labor contracts exhibit longer employment tenures than
TABLE 4
UNEMPLOYMEINT RATES BY COLOR, AGE, SEX, AND MARITAL STATUS, 1950
WHITE NON-WHITE
Male:
14 and over ...... 9.4 3.1 8.7 12.1 5.7 11.2
14-17 ........... 10.6 7.8 4.4 9.5 * *
18-19 ........... 11.2 6.3 17.1 14.1 8.6 18.7
20-24 ........... 9.9 4.5 12.0 14.1 7.2 15.8
25-29 ........... 8.4 3.1 10.2 12.5 6.8 13.9
30-34 ........... 7.6 2.5 8.8 11.2 5.8 11.4
35-44 ........... 7.9 2.6 9.0 9.9 5.1 11.3
45-54 ........... 8.5 3.0 9.0 10.5 5.2 10.0
55-64 ........... 9.5 3.7 8.8 10.3 6.0 10.3
65-74 ........... 8.3 4.3 6.7 7.0 5.7 8.9
75 and over ...... 3.1 2.8 3.8 * 2.7 6.5
Female:
14 and over ...... 4.7 3.2 5.6 10.7 6.6 8.1
14-17 ........... 12.6 13.9 16.1 13.8 18.7 *
18-19 ........... 6.9 6.6 17.5 15.8 12.0 22.8
20-24 ........... 4.3 4.0 9.5 12.3 10.1 15.0
25-29 ........... 3.5 3.5 7.1 8.8 7.8 11.8
30-34 ........... 3.4 3.3 6.5 6.0 6.5 9.5
35-44 ........... 2.8 2.7 5.6 6.5 5.7 6.6
45-54 ........... 2.3 2.8 5.0 5.5 5.0 6.4
55-64 ........... 3.0 2.8 5.0 4.7 4.6 5.5
65-74 ........... 3.1 2.3 3.9 6.9 7.8 4.9
75 and over ...... 3.0 2.8 2.8 * * 3.3
are rarely binding on employees, the firm either single males or males who are di-
can never exercise complete control over vorced, widowed, or separated.
the expected period of employment. Fur- Unemployment rates classified by race,
thermore, seniority rights or other union age, sex, and marital status are presented
practices may prevent a firm from follow- in Table 4 for the census year 1950.23 For
ing a discriminatory layoff policy. 2" Magnus Hedberg, "Labor Turnover, the Flow
Certain personal characteristics such of Personnel through the Factory" (Stockholm:
as age, sex, afn(1marital status are al- Swedish Council for Personnel Administration).
(Mimeographed.)
lege(1ly associate(1 with employment teni- 23 This fromiPhillip M. Hlau-
table is reprlodLuced
ure. 1\fi(Il(e--gaT(Iworkers wvou1(l,in geni- scr's article with his kinidpermission. The rea(ler is
552 WALTER Y. 01
both races, and in almost all age groups, tion, but the data indicate, in the main,
married persons with spouse present re- higher unemployment rates for younger
vealed the lowest unemployment rates. and older workers than for middle-aged
The only exception for males was in the workers.
"14-17" age group, while single women The sex differentials in unemployment
had lower unemployment rates in two rates cannot be explained by differences
cases. The observed differentials in male in expected periods of employment. The
unemployment rates thus conform to the volatility of female labor-force participa-
pattern implied by differences in expected tion rates suggests that unemployment
periods of employment. rates are not the appropriate variable.
The distribution by age of unemploy- Furthermore, the evidence that there are
ment rates is roughly the same, except differences in employment tenure by sex
for level, for each race-marital status is not at all convincing."6Finally, the oc-
group of the male labor force; similar pat- cupational differentials in unemployment
terns are evident in data for other years. rates are also consistent with the impli-
Unemployment rates fall over the range cations of the fixed cost hypothesis.
from fourteen to thirty-five years of age
E. ADDITIONAL EVIDENCE
even when marital status is held con-
stant. Beyond thirty-five years of age in- The analysis of subsection ID implies
creasing age is associated with higher un- that labor of higher degrees of fixity
employment rates although the trend is should enjoy lower labor turnover rates.
reversed in the oldest age group ("65 and Manufacturing industries vary widely in
over"). The problems inherent in defin- the skill composition of their labor forces.
ing the labor force for older workers make The high-wage industries typically em-
the interpretation of their unemployment ploy larger proportions of skilled workers
rates particularly difficult. Hauser specu- with higher degrees of fixity. Low labor
lates that this reversal in the age pattern turnover rates should, therefore, be ob-
might be attributable to a withdrawal of served for high-wage industries. Other
older workers from the labor force when factors that affect labor turnover rates
they become unemployed.24 A logical im- were held constant by a multivariate re-
plication of the withdrawal hypothesis is gression model that cannot be fully re-
that the labor-force participation rate of ported here.27 Based on annual data for
older workers will be negatively corre- sixty-four manufacturing industries from
lated with cyclical changes in employ- 1951 to 1958, the regression model re-
ment. The latter implication is not sup-
26The BLS turnover data by sex for two-digit
ported by data presented in a study by manufacturing industries reveal slightly higher turn-
Long.25 I cannot explain this contradic- over rates for females. Hauser finds (p. 244) that
females in a previous state of employment are less
vulnerable to unemployment than males. This same
referred to "Differential Unemployment and Charac- stability of employment of female workers was also
teristics of the Unemployed in the United States, found in the Philadelphia study; see United States
1940-1954," The Measurement and Behavior of Un- Department of Labor, Bureau of Labor Statistics,
employment (Princeton, N.J.: Princeton University The Social and Economic Characterof Unemployment
Press, 1957), pp. 243-80. in Philadelphia (Bulletin No. 520, April, 1929).
24 Ibid., p. 251.
27
This portion of the study derived from a paper
Clarence D. Long, The LaborForce under Chang-
25 I read at the winter, 1960, meetings of the Econo-
ing Income and Employment (Princeton, N.J.: Prince- metrics Society. I wish to thank Stanley Lebergott
ton University Press, 1958), pp. 323-25, Table B-2. and Jacob Mincer for their helpful comments.
LABOR AS A QUASI-FIXED FACTOR 553
University Press, 1956), p. 364. cago: University of Chicago Press, 1923), pp. 357-85.
LABOR AS A QUASI-FIXED FACTOR 555