Wage Determination in the Union and Nonunion Sectors
Author(s): Farrell E. Bloch and Mark S. Kuskin
Source: ILR Review, Vol. 31, No. 2 (Jan., 1978), pp. 183-192
Published by: Sage Publications, Inc.
Stable URL: http://www.jstor.org/stable/2522386
Accessed: 29-03-2018 11:28 UTC
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.
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
http://about.jstor.org/terms
Sage Publications, Inc. is collaborating with JSTOR to digitize, preserve and extend access to
ILR Review
This content downloaded from 147.251.55.24 on Thu, 29 Mar 2018 11:28:45 UTC
All use subject to http://about.jstor.org/terms
WAGE DETERMINATION IN THE UNION
AND NONUNION SECTORS
FARRELL E. BLOCH and MARK S. KUSKIN
W AGE differences between the union and
nonunion sectors result from sectoral
variation in both worker characteristics,
such as educational attainment and job
experience, and labor market rewards to
these characteristics. Wage equations estimated
for union and nonunion workers
with only a dummy variable indicating
union membership explicitly constrain the
labor market rewards to all other worker
characteristics to be equal in the two sectors.
Such a procedure not only fails to address
the interesting question of differences
in wage determination across sectors but
also may yield poor estimates of unionnonunion
wage differentials for workers
with otherwise similar characteristics.
This study contains estimates of wage equations
for white male union and nonunion employees.
The authors find that nonunion wages are generally
more responsive than union wages to individuals'
education and experience and to regional
price-level variation. Despite those differences,
however, estimates of union-nonunion wage differentials
based on these separate equations do not
differ greatly from a differential obtained from a
union dummy variable in an equation based on
combined union and nonunion observations. Unionnonunion
differentials vary widely across occupational
groups and are generally larger in the lower
skilled and more highly unionized occupations. The
results for manufacturing, for which additional industry
data are available, indicate a negative impact
of high concentration ratios on the wages of
all workers and a greater impact of establishment
size on nonunion than on union wages. Data were
drawn from the May 1973 Current Population
Survey.
Farrell E. Bloch is Assistant Professor of Economics
at Princeton University. Mfark S. Kuskin
was a senior economics major at the time this study
was undertaken; he graduated from Princeton in
1976.-EDITOR
The objective of this paper is to examine
these issues by estimating separate wage
equations for union and nonunion white
male workers,' avoiding the common pitfalls
of using a union dummy variable or
permitting only partial interaction between
unionism and other determinants of wages.2
The first section of the paper contains
estimates of separate wage equations for
individuals in the union and nonunion
sectors. The second focuses only on individuals
in manufacturing industries for
whom more details on the industry in which
they work are available.
Basic Wage Equations
Our observations, taken from the May
1973 United States Current Population
Survey, are limited to white, non-Spanish
lThroughout this paper, all workers who are not
members of a union, including those covered hy
tunion contracts, are considered to he in the nonunion
sector. Orley Ashenfelter, "Union Relative
Wage Effects: New Evidence and a Survey of Their
Implications for Wage Inflation" (Working Paper
89, Industrial Relations Section, Princeton University),
cites a survey indicating about a 90%
overlap between "union members" and "workers
covered by a collective bargaining contract" for the
Current Population Survey, the data source for our
equations.
2For example, George E. Johnson and Kenwood
C. Youmans, "Union Relative Wage Effects by Age
and Education," Industrial and Labor Relations
Review, Vol. 24, No. 2 (January 1971), pp. 171-80,
examine union interaction only with education,
age, and race; and Michael J. Boskin, "Unions and
Relative Wages," Amnerican Economic Review, Vol.
62, No. 3 (June 1972), pp. 466-72, and Paul M.
RNyscavage, " Meastrifng IUnion -Nontunion Earnings
Differences," AMonthly Labor Review, Vol. 97, No.
12 (December 1974), pp. 3-10, examine union interaction
only with region and occupation.
Industrial and Labor Relations Review, Vol. 31, No. 2 January 1978). ? 1978 by Cornell University.
0019-7939/78/3102-0183$00.75
183
This content downloaded from 147.251.55.24 on Thu, 29 Mar 2018 11:28:45 UTC
All use subject to http://about.jstor.org/terms
184 INDUSTRIAL AND LABOR RELATIONS REVIEW
males between age 25 and 64 who are employed
in the private sector. Focusing on
white males circumvents complex interactions
among wage rates, unionism, and
discrimination. The age boundaries exclude
most white males in school or with
part-time jobs, for whom reported wage
rates often do not reflect market values of
accumulated human capital. Excluding
public sector employees avoids complications
due to the rather different wage
structures for private and government employees.4
However, although our restricted
sample reduces wage variation that cannot
be explained by our independent variables,
it should be noted that results based on
this sample may not be applicable to the
remaining sectors of the work force.
The dependent variable in our wage
equations is the natural logarithm of each
individual's usual weekly earnings divided
by his usual weekly hours. This variable
may include overtime work and thus
should not be confused with individual
marginal wage rates. The coefficients of
the independent variables in our wage
equations may be interpreted as the percentage
change in the wage rate effected
by unit changes in the explanatory vari-
ables.
Our most important independent variables
are probably education, experience,
and experience-squared. These variables
have been included in several previous
wage-and-earnings studies, with one rigorous
rationale for their inclusion being provided
by Mincer.5 Our education variable
indicates years of formal schooling between
zero and eighteen. One problem with this
measurement is that individuals with more
than eighteen years of schooling cannot be
differentiated from those with exactly
eighteen years of education. However, since
only .13 percent of the union and 3.24 percent
of the nonunion employees reported
3See Orley Ashenfelter, "Racial Discrimination
and Trade Unionism," Journal of Political Economy,
Vol. 80, No. 3, Part I (May/June 1972), pp.
435-64.
4See Sharon P. Smith, "Government Wage Differentials
by Sex," Journal of Human Resources, Vol.
11, No. 2 (Spring 1976), pp. 185-99.
5See Jacob Mincer, Schooling, Experience, and
Earnings (New York: National Bureau of Economic
Research, Columbia University Press, 1974).
this maximum level of schooling, we assume
that all such individuals have had
exactly eighteen years of education. Equations
estimated with a dummy variable
for eighteen years of schooling and an
integer variable for seventeen years or less
yield results with virtually identical coefficients
to the equations reported here.
The experience variable is defined as age
minus education minus 6. This variable
indicates potential work experience after
the completion of formal schooling. It
overstates work experience for those individuals
who have not held jobs for their
entire postschool careers, but this is much
less likely to be the case for our sample
than for a random sample of American
workers.
We expect that the coefficients of the
education and experience variables will be
positive in both equations and that the
coefficient of the experience-squared term
will be negative, reflecting diminishing returns
to work experience as experience
itself increases. We also expect that the
effect on wage rates of education and experience
will be greater in the nonunion
equation than in the union equation if
employers in the nonunion sector are more
responsive to market forces. This argument
is perhaps stronger for the education than
for the experience variable, given the importance
of seniority and built-in pay increases
for union members.6
Other independent variables in the equations
include marital status, perhaps regarded
by many employers as a proxy for
such personality traits as stability and responsibility,
and veteran status, indicating
either training in a given skill or time lost
in the civilian labor market. The effect on
wage rates of being married is expected to
be positive, that of veteran status ambiguous,
in both sectors.
Also included is a regional price index,
6Sherwin Rosen, "Trade Union Power, Threat
Effects, and the Extent of Organization," Review of
Economic Studies, Vol. 36(2), No. 106 (April 1969),
pp. 185-94, found that wages in the union sector
are affected more by age and experience than by
educational attainment. He explains this result by
the prevalence of seniority systems in which wage
rates are based on age and experience rather than
education and by the generally relatively low
formal educational attainment of older workers.
This content downloaded from 147.251.55.24 on Thu, 29 Mar 2018 11:28:45 UTC
All use subject to http://about.jstor.org/terms
UNION AND NONUNION SECTOR WAGE DETERMINATION 185
the construction of which is described in
the Appendix. Although we expect the
price index coefficient to be positive in
both the union and nonunion equations,
we expect it to.be larger in the nonunion
equation, given the tendency for unions to
attempt to eliminate wage differentials resulting
from geographic differences, especially
through centralized bargaining of
unions in national product markets. Finally,
we have included a set of (two-digit)
industry and occupation dummy variables
to correct for varying labor demand across
labor markets.
Results
Table 1 contains the results from our
estimated union and nonunion wage equations,
along with a t-statistic indicating the
difference between the corresponding coefficients
in the two equations. Although
not shown here because of space limitations,
we also estimated combined wage equations
for unionized and nonunionized employees,
both with and without a union membership
dummy variable." A comparison of
those and the Table 1 equations indicates
that in neither case can the null hypothesis
of equality of the coefficients across equations
be accepted. Using the equation with
the union dummy, testing the hypothesis
of equality for all coefficients except the
constant, F = 18.2 > 1.71 = F01 ( 30, 1000)
> F (34, 12505); using the equation without
the union dummy, testing the equality of all
coefficients including the constant, F -
27.1 > 1.71 = Fo01 (30, 1000) > F.01 (35,
12503).
Our predictions of signs and magnitudes
of coefficients are strongly supported by the
Table 1 regressions. The education coefficients,
the partial derivatives of the
logarithm of the wage rate with respect
7Let pB1,, /,, be the estimated coefficients of the
same explanatory variable in the union and nonunion
equation and aU and or, their respective estimated
standard errors. Then the t-statistic given in
Table 1 is computed as
j3u - On
V2TU + on
8Detailed results of all the combined equations
described in this paper, and of the equations on
which Table 2 below is based, can be obtained from
the authors on request.
to experience, and the price variable are
all significantly greater than zero at the
one percent level in each equation. Furthermore,
the corresponding constructs in each
nonunion equation than in the union
equation, again at the one percent level.9
In addition, the coefficients of the variables
equation are significantly greater in the
indicating marital status are positive in each
equation, with the effect stronger for married
men with wife absent or deceased.
The effect on the nonunion wage is significantly
greater than that on the union wage
at the one percent level only for those men
married with wife present. Veteran status
does not appear to affect wage rates significantly
in either sector.
In general, unions appear to have the
effect of flattening out the wage equation.
The union wage-experience profile peaks at
about 27 years and, as noted above, is flatter
than that of the nonunion sector, which
peaks at about 28.5 years. Thus, an additional
year of experience for a new worker
raises the wage of union workers by about
one percent and that of the nonunion
workers by about 2.3 percent. For each
additional 10 years of experience, the union
effect declines by roughly .4 percent and
the nonunion effect by .8 percent. The percentage
increases in wage rates resulting
from an additional year of formal education
are 1.8 percent for union workers and
5.1 percent for nonunion workers. The
price index coefficient is about one percent
in the nonunion equation and roughly
half that in the union equation.
Johnson and Youmans also observe relatively
flat union-education and union-age
profiles,'0 which they explain in terms of
9Hypothesis tests regarding the partial derivative
of wage rates with respect to experience require
estimates of covariances between coefficients of experience
and experience-squared, statistics unfortunately
not printed out by our computer program.
Our remarks in the text, however, will hold at most
experience levels unless the magnitudes of these
covariances are extraordinarily large. The nonunion
experience effect is greater than the union experience
effect for all positive experience levels. For the
corresponding equations dealing with workers in
manufacturing, however, the nonunion experience
effect exceeds the union experience effect for levels
of experience between 1.9 and 52.3 years.
l0johnson and Youmans, "Union Relative Wage
Effects by Age and Education."
This content downloaded from 147.251.55.24 on Thu, 29 Mar 2018 11:28:45 UTC
All use subject to http://about.jstor.org/terms
186 INDUSTRIAL AND LABOR RELATIONS REVIEW
Table 1. Separate Wage Equations for Union and Nonunion Employees.
(Standard errors are in parentheses)
Dependent Variable
log (hourly wage)
t-Statistic
Independent Variables Union Nonunion of Difference
Education .01771** .05070** -10.771**
(.00222) (.00211)
Experience
Experience .00976** .02272** -5.587**
(.00167) (.00161)
Experience-squared -.00018** - .00040** 5.185**
(.00003) (.00003)
Price index .00523 .01115 -6.247**
(.00066) (.00068)
Marital status
Spouse present .09669** .21582** -4.454**
(.01939) (.01842)
Spouse absent .07578 .14859** -1.276
(.04037) (.04032)
Widowed .06159** .11524** -1.351
(.02781) (.02834)
Single, never married
Veteran status - .00603 - .01058 .339
(.00907) (.00987)
Occupation
Professional workers .29455** .36544** 1.633
(.03149) (.02986)
Managers .16377** .38286** -5.517**
(.02776) (.02842)
Sales workers .06551 .25837** -3.202**
(.05175) (.03080)
Clerical workers .05971* .14681** -2.115*
(.02588) (.03203)
Craftsmen .19981** .21318** - .414
(.01735) (.02728)
Operatives .07878** .05103** .790
(.01816) (.03007)
Transport equipment workers .05428** - .00369 1.463
(.02082) (.03371)
Service workers -.10337** -.00119 -2.202*
(.02974) (.03563)
Laborers
Industry
Mining .36746** .54119** -1.721*
(.08654) (.05200)
Construction .61119** .49830** 1.226
(.08308) (.03970)
Manufacturing-nondurables .23785** .46514** -2.475**
(.08314) (.03899)
Manufacturing-durables .23780** .49688** -2.842**
(.08278) (.03815)
Railroad .36207** .63607** -2.319*
(.08502) (.08201)
Other transportation .45511** .46091** - .061
(.08422) (.04525)
Other utilities .28670** .59577** -3.219**
(.08445) (.04566)
This content downloaded from 147.251.55.24 on Thu, 29 Mar 2018 11:28:45 UTC
All use subject to http://about.jstor.org/terms
UNION AND NONUNION SECTOR WAGE DETERMINATION 187
Table 1. (Continued).
Dependent Variable
log (hourly wage)
t-Statistic
Independent Variables Union Nonunion of Difference
Wholesale .28637** .41775** -1.374
(.08666) (.04039)
Retail .25799** .20195** .602
(.08465) (.03873)
Finance .29274** .43634** -1.341
(.09878) (.04144)
Business-repair .24169** .33518** - .929
(.09103) (.04302)
Personal services .26022** .11579* 1.154
(.11119) (.05617)
Entertainment .31364** .21384** .870
(.09925) (.06340)
Welfare -.11910 -.47953** 2.349**
(.14350) (.05440)
Hospitals .11821 .22501** - .814
(.11819) (.05520)
Medical, except hospitals .35951* .33770** .096
(.21678) (.07113)
Education .12998 .03052 .850
(.10548) (.05043)
Other professional services .43615** .46666** 2.271 *
(.10238) (.04716)
Agriculture
Constant 4.84777 3.31053
R2 .31042 . 39709
F 57.87279 157.52971
N 4406 8167
S.E.E. .28251 .41935
*Significant at the .05 level using a one-tailed test.
**Significant at the .01 level using a one-tailed test.
union seniority systems. Under these systems,
workers tend to be promoted if they remain
with a firm, so there is little incentive for
human capital investment. In addition, in
unionized firms with seniority systems, employers
are more likely to keep an older
worker in a responsible position than to
promote a younger worker.
The Table 1 regressions also indicate
that the union effect on laborers' wages is
high relative to that for most occupational
groups and that the effect on agricultural
workers' wages is high relative to that for
most industrial groups. No occupational
union-nonunion differential is significantly
higher at the 5 percent level than that for
the reference occupation, laborers; the
differentials for managers and sales, clerical,
and service workers are signifiantly
lower. Only workers in welfare and other
professional services industries have significantly
higher industrial union-nonunion
wage differentials than employees in the
reference industry, agriculture; the differentials
for workers in mining, manufacturing,
railroads, and other utilities are significantly
lower.
The above results must be qualified by
the failure of these equations to capture
possible effects of the explanatory variables
on forms of employee compensation other
than money wage rates. If such effects are
stronger for union members than others,
then the effect of such variables as education
and experience on total employee
compensation will not necessarily be
greater for nonunion workers.
This content downloaded from 147.251.55.24 on Thu, 29 Mar 2018 11:28:45 UTC
All use subject to http://about.jstor.org/terms
188 INDUSTRIAL AND LABOR RELATIONS REVIEW
Union-Nonunion Differential
The union-nonunion wage differential
can be computed in several ways. The simplest
is obtained from the coefficient of the
union dummy variable in a wage equation
identical, except for this dummy variable,
to those reported in Table 1 but based on
observations on both union and nonunion
employees. The antilog of this coefficient
(.14734 with standard error .00855) minus
one yields an estimated union-nonunion
wage differential of 15.87 percent. Alternatively,
using the Table 1 equations, we can
compute the logarithmic differential of the
wage rates assuming first that the union
wage structure applies to workers in both
sectors and second that the nonunion wage
structure applies to all workers. Letting W
represent logarithms of individual wage
rates, pl the vector of estimated coefficients
in the wage equations, and Z the vector
of explanatory variables, and using superscripts
for the union and nonunion wage
rates and bars to represent mean values, the
first of these measured differentials is
Wu - Wn + /3U(fl - Zu)
and the second is
Wu - Wn + g3,(Zn Zu).
The logarithmic differential estimated by
the first expression is .0888, that of the
second .1461. The corresponding percentage
union-nonunion wage differentials are
9.29 percent and 15.73 percent. These estimates
almost bracket the 15.87 percent
differential obtained by simply using a
union dummy variable, despite strong evidence
of differing wage structures in the
union and nonunion sectors.
Equations similar to those reported in
Table 1 were estimated for each occupation
separately. The independent variables were
the same as in the Table 1 regressions except
for the omission of the occupation
dummies and mild aggregation of the industry
variables. Because of the small number
of observations in certain industryoccupation
cells, only mining, construction,
manufacturing (durable and nondurable),
railroad, transportation, utility, and wholesale
industry dummy variables were in-
cluded.
The results show that both the wageeducation
profile and the wage-experience
profile are flatter in the union than in the
nonunion equation, the former for all occupations
except sales and the latter for all
except sales, managers, and professional
workers. We can reject at the 5 percent
level the null hypothesis of equality of
coefficients across sectors for each equation
except for service, sales, and professional
workers."
The estimated union-nonunion differentials
for each occupation are presented in
Table 2. In all cases but "all occupations"
and "sales workers," estimated union-nonunion
wage differentials obtained from the
separate union and nonunion equations
bracket the estimated differential obtained
from the union dummy coefficient in a
combined equation for the corresponding
occupation. And for "all occupations" and
"sales workers," the combined equation differential
is only slightly greater in absolute
value than the differential based on the
separate equations and the nonunion structure.
The union-nonunion differentials are
generally greater in the lower skilled and
highly unionized occupations, consistent
with the occupation coefficients reported in
Table 1.
Workers in Manufacturing
We next estimate wage equations for
workers in manufacturing industries by
using the four-firm concentration ratio and
the percentage of establishments in the industry
with more than one hundred employees,
rather than industry dummy variables.12
The United States Census of Manufacturing
gives four-firm concentration ratios for
four-digit industries, but the Current Population
Survey reports three-digit rather
than four-digit industry classifications.
Therefore we constructed the three-digit
values by weighting the four-digit ratios by
sales. These national concentration ratios
"The hypothesis of equality was rejected at the
5% level for professional workers when the union
dummy was excluded, but not when it was in-
cluded.
12The means of these variables for the entire
manufacturing sample are 29.57 for concentration
ratio and 117.91 for establishment size. The range
for establishment size is 0 to 1000.
This content downloaded from 147.251.55.24 on Thu, 29 Mar 2018 11:28:45 UTC
All use subject to http://about.jstor.org/terms
UNION AND NONUNION SECTOR WAGE DETERMINATION 189
Table 2. Estimates of Union/Nonunion Wage Differentials.a
Percent of
Workers in t-Statistic
Occupation Union Nonunion Combined on Union
Occupation Unionized Structure Structure Equation Dummy
All occupations 35.04 9.29 15.73 15.87 17.23
Craftsmen 47.67 25.54 15.58 19.47 14.75
Operatives 61.84 22.2 15.95 18.64 2.65
Managers 8.70 -12.3 4.35 2.37 5.77
Professional workers 8.05 -6.92 2.86 .51 .11
Transport equipment operatives 57.82 43.99 32.93 38.37 12.14
Clerical workers 27.98 4.52 -2.08 2.31 .71
Laborers 44.98 61.56 33.38 42.21 10.55
Service workers 31.59 19.04 14.41 15.45 2.90
Sales workers 4.64 - .41 -2.96 -4.15 - .54
a The union (nonunion) structure column contains estimates based on the assumption that the union (nonunion)
equation applies to all workers. The combined equation estimates are based on the coefficient of the
union dummy variable in an equation computed from observations on both union and nonunion employees.
The t-statistics in the last column are associated with these dummy variable coefficients.
understate the market power of firms in
industries with clearly regional or local
markets. Employers in concentrated industries
may pay higher wages than others if
workers in these industries share the
monopoly or oligopoly rents obtained by
these firms, a situation especially likely to
occur when these workers are unionized
and therefore have a relatively strong bargaining
position. On the other hand, firms
in concentrated industries may have substantial
financial reserves with which to
resist union wage demands, despite the fact
that these firms originally may have been
easier for unions to organize because of
economics of scale.
The Census of Manufacturing also
provides establishment-size information.
Because concentration ratios and the incidence
of large firms are positively correlated,
some arguments supporting relationships
between wage rates and concentration
ratios have also been used in positing
relationships between wage rates and establishment
size. To the extent that establishment
size indicates the ability to achieve
economics of scale in providing fringe
benefits, establishment size may have negative
coefficients in regressions explaining
the monetary wage, assuming a zero effect
on total employee compensation. To the
extent that unions achieve relatively high
percentages of fringe benefits in total employee
compensation, the establishment size
variable may be more strongly negative in
the union equation than in the nonunion
equation.
The manufacturing wage equations are
reported in Table 3. The last column gives
the t-statistic for the difference in corresponding
coefficients in the union and
nonunion equations.
Generally, the results are rather similar
to those reported in Table 1. The signs
and relative magnitudes of the education,
experience, and price coefficients are all the
same as in Table 1. The veteran-status coefficients
again do not differ significantly
from zero and the marital-status variables
are again positive, although not always
significantly greater than zero at conventional
test levels and in no case significantly
different across the two equations. Again
comparing these two equations with an
equation estimated for all manufacturing
employees with and without a union
dummy variable yields the conclusion that
the equations differ significantly across sectors.
The null hypothesis of equality of the
coefficients is strongly rejected at the one
percent level: F = 4.8 > 2.01 = F01 (16,
1000) > F01 (18, 2038), using the equation
with the union dummy variable; F = 5.3
> 2.01 = F01 (16, 1000) > F01 (19, 2036),
using the equation without the union
dummy variable.
This content downloaded from 147.251.55.24 on Thu, 29 Mar 2018 11:28:45 UTC
All use subject to http://about.jstor.org/terms
190 INDUSTRIAL AND LABOR RELATIONS REVIEW
Table 3. Separate Wage Equations for Employees in Manufacturing Industries.
(Standard errors are in parentheses)
t-Statistic
Independent Variables Union Nonunion of Difference
Education 0.02908** 0.06859** -6.09850
(0.00468) (0.00448)
Experience 0.00422 0.02427** -3.99929**
(0.00354) (0.00355)
Experience squared -0.00002 -0.00039** 3.73756**
(0.00007) (0.00007)
Married-spouse present 0.11986** 0.14002** -0.34225
(0.04030) (0.04296)
Married-spouse absent 0.05647 0.07210 -0.12672
(0.08377) (0.09053)
Widowed 0.12439* 0.03251 1.07524
(0.05761) (0.06311)
Veteran status 0.01666 -0.01839 1.23868
(0.01926) (0.02073)
Concentration ratio -0.00160* -0.00091 -0.58047
(0.00081) (0.00087)
Establishment size -0.00012 0.00016 -1.97007*
(0.00009) (0.00011)
Price index 0.00313** 0.00740** -2.26851*
(0.00128) (0.00141)
Occupation
Professional workers 0.026995 0.43751 ** - 1.71829*
(0.06352) (0.07399)
Managers 0.47615** 0.49199** -0.13548
(0.09013) (0.07447)
Sales workers 0.34333** 0.35711** -0.08892
(0.13132) (0.08229)
Clerical workers 0.12868* 0.19387** -0.65510
(0.06055) (0.07897)
Craftsmen 0.22454** 0.27889** -0.63472
(0.04652) (0.07189)
Operatives 0.12574** 0.12740* -0.01916
(0.04615) (0.07335)
Transport equipment operatives 0.17043** 0.11710 0.50224
(0.05627) (0.09005)
Service workers -0.02488 -0.08240 0.46783
(0.06853) (0.10208)
Constant 5.12560 3.88667
R2 .19 .47
F 10.54 65.76
N 778 1296
S.E.E. .244 .350
*Significant at the .05 level using a one-tailed test.
**Significant at the .01 level using a one-tailed test
The union-nonunion wage differentials
based on the separate equations, -3.43
percent, assuming the union structure applies
to all employees, and 6.97 percent,
assuming the nonunion structure applies to
all employees, bracket the estimate obtained
from the coefficient of the union
dummy variable in the combined equation,
5.31 percent. These relatively small differentials
are consistent with the greater
coefficients for manufacturing industry
dummy variables in the nonunion as compared
with the union equation in Table 1.
The concentration ratio variable is negaThis
content downloaded from 147.251.55.24 on Thu, 29 Mar 2018 11:28:45 UTC
All use subject to http://about.jstor.org/terms
UNION AND NONUNION SECTOR WAGE DETERMINATION 191
tive in each equation, significantly so at the
5 percent level only in the union equation
and not significantly different across equations.13
These results provide mild support
for the ability of firms in concentrated
industries to withstand union wage demands
and are also consistent with the
hypothesis of concentrated firms' monopsony
power. The establishment-size variable
is negative in the union equation and
positive in the nonunion equation, although
in neither case significantly different
from zero at conventional test levels.
However, the nonunion establishment-size
coefficient is significantly greater than the
union coefficient at the 5 percent level. This
is consistent with our hypothesis that both
large firms and unions have relatively
strong preferences for fringe benefits as
compared with monetary wage payments.14
Conclusions
The results of this paper clearly indicate
differing structures of wage determination
in the union and nonunion sectors. Nonunion
sector wages are generally more responsive
to individual worker levels of
education and experience and to regional
price-level variation. Despite the greater
labor market rewards to these characteristics
in the nonunion sector, union-nonunion
wage differentials are positive for most oc'3Sherwin
Rosen, "Trade Union Power, Threat
Effects, and the Extent of Organization," and Leonard
WV. Weiss, "Concentration and Labor Earnings,"
American Economic Review, Vol. 56, No. 1 (March
1966), pp. 96-117 also found negative coefficients
for union concentration ratio interaction terms in
their (combined) wage equations. Our mild negative
relationship between concentration ratios and wages
differs from the findings of Weiss, who found no
relationship, and James A. Dalton and E. J. Ford,
Jr., "Concentration and Labor Earnings in Manufacturing
and Utilities," Industrial and Labor Relations
Review, Vol. 31, No.1 (October 1977), pp. 45-
60, who found a statistically significant positive
relationship between concentration ratios and individual
wages. Rosen found a positive but not statistically
significant relationship between concentration
ratios and mean industry wages. These studies
are more compatible with our nonunion than with
our union results, especially considering the presumed
positive correlation between concentration
ratio and establishment size, a variable not included
by Dalton and Ford.
14Albert Rees suggested this interpretation.
cupations, especially so in those that are
more highly unionized and less skilled.
Despite the different wage structures in
the two sectors, estimated union-nonunion
wage differentials obtained from separate
union and nonunion equations do not
differ greatly from differentials obtained
from coefficients of union dummy variables
in wage equations incorporating observations
on both union and nonunion employees.
Consequently, the simpler methodology
does not appear to produce seriously
biased estimates of union-nonunion differentials
despite its masking of sectoral differences
in wage determination.
Our estimates of these differentials from
both separate and combined equations are
generally slightly lower than those obtained
by other investigators. Although
there are important differences in the
analysis and data used in other studies that
render comparisons difficult, the most
likely explanation for our relatively low
estimates is the high unemployment and
unanticipated inflation in 1973 relative to
the late sixties, when the studies cited
above were undertaken. One would expect
union bargaining power to be weaker in
periods of higher unemployment. In addition,
contract-determined union wages are
generally not as responsive to unanticipated
inflation as are wages in the nonunion
sector, although the increased use of
cost-of-living escalators should reverse this
trend.
As many authors have pointed out, these
estimated union-nonunion differentials are
imperfect measures of the extent to which
unions have raised wages over levels that
would have prevailed in the absence of
unionism because nonunion wages themselves
are affected by the presence of unionism.15
Nonunion employers may set higher
wages in an attempt to prevent their firms
from being unionized or simply to compete
with union employers in recruiting labor.
On the other hand, high union wages and
possibly resultant high product prices will
tend to reduce employment in the union
sectors (and in those nonunion establishl5See
especially H. Gregg Lewis, Unionism and
Relative Wages in the United States (Chicago: The
University of Chicago Press, 1963).
This content downloaded from 147.251.55.24 on Thu, 29 Mar 2018 11:28:45 UTC
All use subject to http://about.jstor.org/terms
192 INDUSTRIAL AND LABOR RELATIONS REVIEW
ments where the threats of union organization
and competing recruitment are effective)
and to increase the supply of labor to
the nonunion sector, thus depressing nonunion
wages. Union-nonunion wage differentials
also allow union employers to ration
the presumed excess supply of labor to union
jobs by selecting especially well-qualified
workers. Part of our estimated union-nonunion
differentials may reflect sectoral differences
in labor quality not accounted for
by our independent variables.
Finally, although the positive correlation
generally observed between unionization
and wage rates is usually attributed to the
positive effect of union membership on
wages, it may be the case that the probability
of union organization is a positive
function of preexisting wage rates. Ashenfelter
and Johnson and also Schmidt and
Strauss have examined this question with
industry and individual observations, respectively,
by estimating a simultaneous
model including a union membership equation
as well as a wage equation.16 Wages
used in the membership equations are current
and postunionization rather than the
more appropriate preunionization wages.
In both cases, the authors found a much
lower positive effect of unions on wages
than would be obtained from a corresponding
single equation model.
Appendix
This appendix contains a discussion of
the price variable used in the regressions in
the main text. The May 1973 Current Population
Survey lists the specific SMSA in
which each individual resides for the largest
99 SMSAs in the United States, and
otherwise the individual's region (Northeast,
North Central, South, or West) and
whether or not he resides in an SMSA.17
'lOOrley Ashenfelter and George E. Johnson,
"Unionism, Relative Wages and Labor Quality in
U.S. Manufacturing Industries," International Economic
Review, Vol. 13, No. 3 (October 1972), pp.
488-508, and Peter Schmidt and Robert P. Strauss,
"The Effect of Unions on Earnings and Earnings
on Unions: A Mixed Logit Approach," International
Economic Review, Vol. 17, No. 1 (February
1976), pp. 204-12.
l7Except in New England, a standard metropolitan
statistical area is a county or group of contiguous
counties that contains at least one city of
The U.S. Department of Labor Handbook
of Labor Statistics 1975 (Washington, D.C.:
G.P.O., 1976), Table 145, p. 375, provides
annual budget indices at an intermediate
living standard for four-person families for
39 SMSAs for autumn 1973. These budget
indices may be better than consumer price
indices as indicators of the buying power of
wages, since they take into consideration
regional variation in consumer market baskets.
In order to assign budget indices for
the remaining SMSAs, we estimated an
equation for these 39 SMSAs in which the
budget index is the dependent variable and
regional dummy variables and SMSA population
are the independent variables. We
expect the coefficient on SMSA population
to be positive, since higher population may
indicate increased demand for land and
ultimately higher prices for final goods and
services. The results are reported in Table 4.
Table 4. Variations in Budget Indices.a
Independent Standard
Variables Coefficients Error
Constant 102.68298** 1.9445
North Central -4.58795* 2.1039
West -6.54861 ** 2.5214
South -10.81971** 2.2971
Population .000000979 0.00034
RI = .5567 S.S.R. = 714.612
N = 39 S.E.E. = 4.65348
*Significant at the .05 level usi
**Significant at the .01 level u
aThe price indices for our observations are equal
to the original budget indices for the 39 SMSAs
incorporated in the above equation, the predicted
value of the dependent variable for the remaining
SMSAs that were specifically noted in the data set,
and the predicted value of the dependent variable
for SMSAs not specifically cited when the population
observation is set equal to 160,000. The 160,000
figure is roughly the mean of SMSAs ranked 100 to
250 in order of population.
50,000 inhabitants or more. In addition to the
county, or counties, containing such a city or cities,
contiguous counties are included in a standard
metropolitan statistical area if according to certain
criteria they are essentially metropolitan in character
and socially and economically integrated with
the central city. In New England, standard metropolitan
statistical areas have been defined on a
town rather than a county basis.
This content downloaded from 147.251.55.24 on Thu, 29 Mar 2018 11:28:45 UTC
All use subject to http://about.jstor.org/terms