Copyright
Nancy L Rose.

Labor rent-sharing and regulation : evidence from the trucking industry online

. (page 2 of 4)
Online LibraryNancy L RoseLabor rent-sharing and regulation : evidence from the trucking industry → online text (page 2 of 4)
Font size
QR-code for this ebook


they are excluded from the sample. I control for industry and occupation effects
by restricting the sample to full-time truck drivers employed in the for-hire
trucking industry. This excludes non-driver employees in the trucking industry,
as well as truck drivers who work in private carriage and self-employed drivers
(owner-operators). Data on worker location are used to control for geographical
wage variation via regional fixed effects.



13

Unfortunately, the CPS typically does not provide information on the firms

1 2
employing the responding individuals. The inability to control for firm

characteristics may bias the estimated union wage differential. For example, if

small firms typically pay lower wages, ceteris paribus, because nonunion drivers in

the trucking industry are more likely to work for smaller firms, omitting firm size

1 "^
may cause the estimates to overstate the union wage premium. This suggests

that one should use caution interpreting the level of the estimated premium,

although this potential bias should not invalidate tests nf changes in union premia

through time.

The CPS data are used to estimate a conventional semi-log wage equation of

the form:



1 9 •

Supplements occasionally gather data on workers establishments; see,

for example, the May 1979 Pension Plan Supplement.

Mellow (1983) investigates the firm size effect on wages, and finds that
estimated union wage differentials are smaller in the presence of controls for firm
characteristics, notably size. I used data from the May 1979 CPS Pension Plan
Supplement to test the significance of this effect in the trucking industry.
Although the point estimates suggest that larger firms tend to pay both higher
nonunion wages and higher union premia over nonunion wages, the small sample
size makes statistical inference difficult. Due to the large standard errors of the
estimates, statistical tests would not reject the restriction that wages levels and
union premia are constant across firm size in the trucking industry.

In addition to firm size, differences in firm markets may affects estimated
differentials. For example, the inability to distinguish workers employed in the
more profitable less-than-truckload (LTL) sector from those employed in the more
competitive, less unionized, and less profitable specialized commodity sector may
distort the results. The inability to exclude United Parcel (UPS) employees, who
negotiate a separate contract and who realized slightly better terms than the
NMFA in their 1982 contract, may bias estimates against a decline in the
deregulation union premium.

Freeman and Medoff (1984) discuss other potential sources of bias in
cross-sectional estimates. However, even if the level of the differential were
overstated by the use of cross-sectional data without adequate controls for all
worker and firm characteristics, there is no reason to expect the bias to change
in such a way as to substantially reduce estimated union wage differentials over
time. Indeed, the rapid growth of new, smaller nonunion firms in the 1980s, and
the maintenance of higher wages by drivers for UPS might exacerbate the
potential for overstatement of post-deregulation union wage differentials.



14



(1) LHWAGE = BO * BVUNION + B2-EDUC * B3-EXP ' RA'EXP'^ * 65NONWHITE
+ B6-SINGLE * B7NE * B8S0UTH * B9-WEST

where: LHWAGE = natural log of the hourly wage rate. Wages less than $1.00
per hour are assumed to be coding errors and the observations
are deleted.
UNION = 1 if a union member, otherwise.
EDUC = number of years of schooling completed
EXP = experience, defined as (Age - EDUC - 6)
NONWHITE = 1 if race is non-white, otherwise
SINGLE = 1 if marital status is other than married with spouse

present, if status is married, spouse present.
NE = 1 if region is Northeast, otherwise
SOUTH = 1 if region is Southern, otherwise
WEST = 1 if region is Western, otherwise

The North Central region dummy variable is omitted, so that estimates are
relative to this region. The results presented below are robust to variations in
this specification.

B. CPS Results

Table 2 presents sample mean union and nonunion wages, estimated union
wage coefficients from equation (1), and unionization rates for each year. The
estimates span four NMFA contract periods: 1973-76, 1976-79, 1979-82, and 1982-
85. The full set of estimated coefficients for each equation is reported in



15



Year



1973



1974



1975



1976



1977



1978



1979



1980



1981



1983



1984



TABLE 2
Estimated Wage Premiums and Sample Characteristics:

Annual CPS Data

Mean Wage



Union Nonunion



$5.65 $3.71



5.75



6.65



6.85



7.54



8.00



8.20



8.51



9.82



10.22



10.94



4.14



4.21



4.41



4.92



5.74



6.41



6.50



7.67



7.27



7.66



Union


Proportion


Sample


Coefficient


Union


Size


.440


.63


231


(.050)






.345


.62


258


(.052)






.448


.55


246


(.056)






.391


.61


236


(.050)






.441


.60


304


(.040)






.346


.57


291


(.046)






.212


.58


170


(.055)






.199


.58


91


(.089)






.208


.61


83


(.089)






.337


.54


85


(.083)






.321


.34


89


(.081)







Standard errors in parentheses.



16
appendix table lA. Several aspects of Table 2 deserve mention. First, the

estimated union coefficients during the regulation period cluster around .4,

1 fi
implying a union premium of 50 percent above nonunion wages. This finding

comports well with Moore's (1978) estimate of a 48 percent premium in 1967 and

t 1 7

Hayden s (1977) estimated 50 percent premium in 1973-75. However, these

differentials are substantially above the average union differential for cross-
sections of industries estimated on similar data sets. For example. Freeman (1986)
reports an average union premium for all blue collar workers of .25, or 36
percent, during the 1973-78 period. Tliis suggests that the Teamsters may have

been able to capture higher rents for truck drivers than the average union

1 8
captured for its members.

The second interesting feature of this table is that the union differential

drops sharply between 1977 and 1979. From 1979 through 1981, the union

coefficient averages about .2] - half its level in 1973-1978- -implying premia of only

23 percent above nonunion wages. The estimated union coefficient increases



The remaining coefficients reported in the appendix table generally are
of the expected sign, although their effects are often quite small and
imprecisely estimated. This is not particularly surprising, given how narrowly
the sample is defined. One might expect that wages of full-time truck drivers
in the for-hire trucking industry would not vary much across workers with
different education or experience levels.

The 1974 coefficient is a notable exception. The low estimate in this
year is driven by a sharp decline in the estimated premium for the outgoing
rotation groups in the CPS. The union coefficient for these groups is .260
(.066 standard error) in 1974, as compared to .451 (.068) in 1973.

1 7

The data I use overlap with Hayden's data, although my specification of

the wage equation differs slightly from his. Because of this, the similarity

in results is to be expected for the 1973 through 1975 equations.

1 R

This conclusion holds up against industry-specific union differentials

estimated by Freeman and Medoff (1984) from 1973 CPS data. Freeman and Medoff

(p. 50) report that half of the 62 industries in their sample had union premia

greater than 15 percent, but only 8 of the 62 had union premia greater than 35

percent.



17

somewhat in 1983 and 1984, to an average of .33, although it remains substantially
below the average regulation level. The larger standard errors on the post-1979
estimates suggest that the 1979 through 1984 estimates are likely to be
statistically indistinguishable.

Finally, there may be a reduction in the proportion of drivers with union
affiliation at the end of the sample period. In 1984, only 34 percent of the
sample reported belonging to a union. Whether- this is an aberration of the 1984
sample or indicative of a decline in the union presence in trucking remains an
open question, and will not be addressed in this paper. Although much of the
growth in trucking firms has been through entry of nonunion carriers, additional
evidence is required to ascertain whether union representation in the industry has
diminished by such a substantial amount.

To test the statistical significance of the decrease in union wage differ-
entials, I estimate wage equations pooling data across years. These results are

reported in Table 3. Separate intercepts are estimated for each year, but all
other coefficients are constrained to be the same within the pooled sample. The



The higher estimated union premia in 1983 and 1984 could also be real
phenomena, linked to our inability to control for job tenure and firm charac-
teristics. The average seniority of unionized workers probably rose during the
1980s, as union unemployment and layoff rates rose. In addition, the a\erage size
of nonunion firms declined during the 1980s in response to substantial entry into
the industry. Both of these factors could tend to increase estimated union premia
in equations that omit controls for tenure and firm size. See notes 13 and 14,
supra .

The CPS survey design calls for half the respondents in a given year to
be reinterviewed the following year. This design creates an overlap in
respondents for adjacent years' surveys. Because of this, estimates from
contiguous years may not be statistically independent. I therefore have
estimated pooled equations omitting contiguous years (eliminating all even year
observations) to ensure independence of observations across time. The hypothe-
sis tests for homogeneity are not materially affected by excluding these years.
The coefficients' standard errors rise somewhat, but this is expected from the
smaller sample size.



18
TABLE 3
Pooled Wage Equation s



Variable
UNION



UNION-REG
(1973-78)

UNION-DEREG
(1979-84)

EDUC



EXP



EXP'



SINGLE



NONWHITE



NE



SOUTH



WEST



Mean Intercept



NOB

r2

SSR



1973-1978

.402
(.020)



1979-1984

.254
(.033)



1973-1984





.021




.005)




.021




.003)




.0004




.0001)




.025




.028)




.135




.033)




.053




.025)




.083




.023)




.035




.026)



1.010

1566

.39

182.79



.012
(.009)

.021
(.005)

-.0003
(.0001)

-.084
(.039)

-.051
(.048)

-.111
(.045)

-.066
(.041)

.119
(.045)

1.593

518
.25

62.24



.403
.019)

.251
.031)

.019
.004)

.020
.003)

.0004
.0001)

.015
.023)

.107
.027)

.067
.022)

.079
.020)

.058
.022)



1.226

2084

.43

246.89



These equations include time effects for each year. The
mean of the estimated time effects is reported as the mean
intercept. Standard errors are in parentheses.



19

first column pools data over the regulatory period, 1973 through 1978. Column 2
pools data over the deregulatory period, 1979 through 1984. These two samples
are combined in column 3, which reports results for the full sample, 1973-1984,
allowing the union coefficient to differ across regulatory regimes.

The constrained union coefficient for the regulatory period is estimated at
.402 (standard error, .020). By contrast, the union wage coefficient over the
pooled deregulation sample in column 2 is .254 (.0331, or roughly 60 percent of
the size of the union wage premium measured in column 1. Further, we cannot

reject the hypotheses that wages within each of these periods are generated by a

21
common process.

The full sample results reported in column 3 reject the restriction of equal

union coefficients across the regulation and deregulation periods at the .001

level. The null hypothesis of homogeneous coefficients across the entire sample

period, excepting the time intercepts and the union coefficients, cannot be

rejected at conventional levels of significance. This suggests that the dominant

change in the wage behavior of truck drivers over the period studied is a decline

in the ability of the union to maintain its relative wage advantage.



O 1

The F-statistic to test the restrictions implied by pooling the
regulatory years, 1973-1978, is 1.09, which is distributed as F(45,1506) under
the null hypothesis (HO) of homogeneous coefficients. The critical value at a
10 percent significance level is 1.24. The F-statistic to test the null
hypothesis of homogeneous coefficients for the deregulation period, 1979-1984,
is .75. This test statistic is distributed as F(36,46B) under HO, with a
critical value of 1.29 at the 10 percent significance level.

The t-statistic to test the hypothesis of equal union coefficients over
regulation and deregulation periods is 4.67.

90

The F-statistic to test this hypothesis is 1.038, which is distributed

as F(89, 1974) under the null hypothesis of homogeneous coefficients. The
critical value at a 10 percent significance level is 1.17. Note, however, that
the hypothesis that the coefficients in columns 1 and 2 are the same, excepting
the union coefficients, will be rejected at the 5 percent significant level.
The F-statistic for this test is 1.95, which is distributed as F(8,2055).



20
A puzzling difference between these results and the contract evidence is in
the timing of the union relative wage decline. Analysis of the Teamsters'
contracts in Section I suggested that major union concessions were first made in
the 1982 NMFA, while the CPS wage equations show an initial decline in the
union premium possibly as early as 1978, with a sharp reduction by 1979. The
aggregate wage evidence falls somewhere between the two, with average trucking
wages beginning to decline relative to the reference group wages after 1980.
While these are not necessarily contradictory (for example, non-contract wage
concessions, allegedly made by some Teamster locals for- many smaller firms, or an
increase in nonunion wages in the late 1970s are consistent with both the NMFA
and CPS findings), they do warrant further investigation.

Evaluating Wages For a "Representative" Driver
A first step toward understanding the nature of the decline in the union
differential is to look at wages over time for a workei- of given characteristics.
I use as a "representative" driver a married white male with 12 years of
education, 20 years experience, living in the North Central region. Predicted

union and nonunion wages for this worker are calculated from a single wage
equation pooling data from all sample years. Intercepts and union coefficients are
permitted to vary across years; all other coefficients are constrained to be the
same over time.

Table 4 reports predicted wages for the representative driver in both current



These characteristics are close to the 1979 sample means.

The complete set of results for this equation are reported in appendix
table 2A. The constant terms reported in the table are estimates of LHWAGE for
a nonunion driver with representative characteristics.



21

TABLE 4

Predicted Union and Nonunion Hourly Wages
for a Representative Driver
[Pooled Sample Estimates)

Nominal Dollars Constant 1984 Dollars





Predicted


Predicted


Predicted


Predicted


Keal


Y?9r


Union Waqe


Nonunion Wage


Union Wage


Nonunion Wage


Differential


1973


$6.26
(0.19)


$4.06
(0.16)


$14.64


$9.49


S5.15


1974


6.36
(0.19)


4.53
(0.18)


13.39


9.55


3.85


1975


7.09
(0.23)


4.56
(0.17)


13.68


6.81


4.87


1976


7.48
(0.24)


4.88
(0.19)


13.65


8.91


4.74


1977


8.38
(0.24)


5.45
(0.19)


14.37


9.34


5.03


1978


8.88
(0.27)


6.24
(0.22)


14.14


9.94


4.20


1979


9.00
(0.33)


7.23
(0.32)


12.88


10.34


2.54


1980


9.27
(0.46)


7.39
(0.43)


11.69


9.32


2.36


1981


10.62
(0.54)


6.64
(0.55)


12.13


9.86


2.26


19S3


10.88
(0.57)


8.03
(0.46)


11.34


8.37


2.97


1984


12.01


8.63


12.01


8.63


3.38




(0.76)


(0.41)









The representative driver- is a married, white male with 12 years of education
and 20 years of experience, living in the North Central region. Because the
wage equation assumes a Jognormal error distribution, the predicted wage is
calculated as exp(X& * o"/!), where XB is the predicted log wage and o' is the
error variance from the pooled equation reported in appendix table 2A. .^

Standard errors, in parentheses, are computed as exp(XB ' o*"/2) ' var(XB)



22
dollars (columns 1 and 2) and 1984 constant dollars (columns 3 and 4). The

union differential in 1984 dollars per hour is presented in column 5. Two aspects
of this table are noteworthy. First, the declining percentage union differential
observed in Tables 2 and 3 is associated with a decline in the growth of nominal
union wages from 1977 to 1979 relative to trend and higher growth rates for
nonunion wages between these years. This is illustrated in figures 2a and 2b
which plot predicted union and nonunion wages in current dollars and 1984
constant dollars, respectively. Predicted union wages aie lower in real terms
during the 1980s, witli real wages in 1983-84 falling 16 percent below the average
real wages in 1973-78. In contrast, real nonunion wages remain relatively high
through 1981 (compared to their early 1970s levels), and decline only 9 percent in
1983-84 relative to their average over 1973-78.

Second, the reduction in the dollar union wage premium is substantial. The
union wage differential declines in nominal dollars between 1977 and 1979; in real
terms the differential falls by almost one-half relative to its level under
regulation. Despite an increase in the real differential in 1983-84, it remains only
two-thirds of the 1973-78 level. The union premium in 1975 was $2.53 per hour
in nominal dollars, or about $5700 per year. In 1983, the current dollar premium
was $2.85 per hour, or $6400 per year. This represents a real decline of nearly
$4200 per year in 1984 dollars.

This wage pattern does not appear attributable to the NMFA's failure to
compensate adequately for inflation during the contract's life. The initial decline
in the union premium comes between 1977 and 1979, despite contract wage



9 fi

The constant dollar figures were obtained using the average urban
Consumer Price Index for each year.



23

FIGURE 2
PREDICTED HOURLV WAGES FOR UVIOK AND KOKuMON REPRESEKTATIVE DRIVER

2A: CURREKT dollars




UNION



NONJNIOK



1673 1S-«



1675 1676



1977



1660 16?!



i»e;



2£; CONSTA.VT 1964 DOU^^RS




UNION



^—-^ NONUNION



li7J li74 liTi 167t 1677



24
increases that basically matched CPI increases over this period. In addition,

the decline in the differential persists through the 1979 contract period and into
the 1982 contract period. This suggests a more fundamental cause for the
relative union decline than unanticipated inflation in the late 1970s.



Comparison of Trucking Wages with Alternative Benchmarks
Disentangling changes in union premia in the ti-ucking industry from general
economy-wide movements or idiosyncracies of the CPS data r.el is more prob-
lematic. To evaluate these possiblilities , I compare the CPS results for the
trucking industry to those for four alternative benchmark groups. Union
coefficients for these groups are presented in Table 5. For ease of reference,
union coefficients for the trucking industry are reproduced in column 1 . Column
2 reports Freeman's (1986) results for a cross-industry sample of all blue collar
workers. Columns 3-5 report my results for workers in the construction, motor
vehicle manufacturing, and printing industries." These industries were selected
because they employed large numbers of workers (necessary to obtain sufficient
sample sizes from the CPS tapes), were heavily unionized, and were unregulated.
Both the construction and motor vehicle manufacturing industries also underwent



77

Scheduled contract wage increases (including COLAs) between April 1977

and April 1979 amounted to roughly 18%; this compares to a 19.7% increase in

the CPI between May 1977 and May 1979.

op

The union coefficients are estimated from year-by-year equations
similar to (1), incorporating occupational fixed effects within each industry.
Roughly similar results are obtained by looking at particular occupational
categories within each of the three benchmark industries.



25
TABLE 5
Union Coefficients for Benchmark Industry Group s



Year Truckin g
1973



1974



1975



1976



1977



1978



1979



1980



1981



1983



1984



0.440
(.050)

0.345
(.052)

0.448
(.056)

0.391
(.050)

0.441
(.040)

0.346
(.046)

0.212
(.055)

0.199
(.089)

0.208
(.089)

0.337
(.063)

0.321
(.081)



Average
Blue
Colla r

0.23
(.01)

0.22
(.01)

0.22
(.01)

0.25
(.01)

0.28
(.01)

0.27
(.01)

0.22
(.01)

0.21
(.01)

0.21
(.01)

0.22
(.01)

0.25
(.01)



Construe-
tion

0.420
(.016)

0.391

(.016)

0.410
(.016)

0.416
(.017)

0.438
(.015)

0.417
(.015)

0.319
(.023)

0.335
(.026)

0.352
(.029)

0.291
(.041)

0.369
(.034)



Motor

Vehicle

Mfg.

0.124
(.042)

0.122
(.037)

0.067
(.036)

0.063
(.043)

0.200
(.037)

0.201
(.039)

0.253
( .046)

0.268
(.079)

0.292
( .082)

0.148
(.085)

0.204
(.099)



Printing

0.289
(.036)

0.207
(.037)

0.243
(.039)

0.239
(.037)

0.263
(.038)

0.288
(.034)

0.156
( .048)

0.280
(.095)

0.192
(.077)

0.258
(.096)

0.219
(.075)



Standard errors in parentheses. Blue collar estimates
are from Freeman (1986). Remaining estimates are
from annual CPS wage equations for workers within
each industry, controlling for occupational fixed
effects, sex, and the variables described in equation
(1) above.



26

on

changes that potentially reduced industry rents. Use of these benchmarks may

tend to understate the extent of regulation-induced changes in the trucking
industry .

These results are somewhat mixed. Several aspects of the benchmark results
are qualitatively similar to those I obtain for the trucking industry. For example,
the union coefficient declines between 1978 and 1979 in three of the four
benchmark groups; the exception is motor vehicle manufacturing. This decline is
reversed in 1980 in the printing industry, but the average union premia for all
blue collar and construction workers remain near their 1979 levels for several
years, before increasing in 1984. Despite these appar-ent similarities, however,
important distinctions between the benchmark results and those for the trucking
industry suggest that movements in trucking union wage premia are not merely
reflective of general trends in the economy as a whole or in the CPS data.

Consider first the average blue collar union premium, as estimated by
Freeman (1986). This union coefficient rises from .23 in 1973 to a peak of .28 in
1977, then falls back to .22 in 1979, a decline in 10 percent from its average
level over 1973-77. In contrast, the differential in trucking hovers around .4
from 1973-77, then drops to one-half of its 1973-77 average level by 1979. The
decline in the trucking differential brings it into the range of the average blue
collar differential - a significant change from its pre-1979 relative position.
Moreover, Freeman notes that by 1984 the blue collar union differential had
returned to its 1973-78 average level. For the trucking industry, the differential
remained 20 percent below its average 1973-78 level. The patterns of union
premia for the other benchmark industries are varied. The premia in the



29

For example, the oil price shock and subsequent demand shift to

imported cars adversely affected motor vehicle manufacturers during this

period.



27

construction industry exhibit behavior similar to that of the trucking premia.
The 1973-77 and 1983-84 average premia are essentially identical for the two
industries, although the average construction premium during 1979-84 is
substantially higher than the trucking premium during this time period. This
similarity may be a result of the decline in union influence in the construction
industry during the 1970s, as discussed by Allen (1985). This pattern is not
repeated in the other benchmark industries. The union premium in motor vehicle
manufacturing appears to have increased during the late 1970s and early 1980s and


2 4

Online LibraryNancy L RoseLabor rent-sharing and regulation : evidence from the trucking industry → online text (page 2 of 4)