Casey Ichniowski.

Ruling out productivity? : labor contract pages and plant performance online

. (page 1 of 2)
Online LibraryCasey IchniowskiRuling out productivity? : labor contract pages and plant performance → online text (page 1 of 2)
Font size
QR-code for this ebook


ft,., ■ . : . - - ^- —


Ruling Out Productivity?
Labor Contract Pages and Plant Performance


Casey Ichniowski
WP # 1568-84 J^Jne» 1984





Ruling Out Productivity?
Labor Contract Pages and Plant Performance

Casey Ichniowski
WP # 1568-84 J'JnS' 1984

The author would like to thank Richard Freeman and Thomas Kochan for their
comments on this paper.

June I98U

Ruling Out Productivity?:
Labor Contract Pages and Plant Performance


This study documents a strong inverse relationship between number of pages
of labor contracts in effect and the productivity observed in a sample of ten
unionized plants. It is argued that this relationship reflects the
productivity-inhibiting effects of increases in the number and complexity of
work rules. The study also argues that subsequent research should try to
improve the measurement of work, rules by considering the substance of the rules
and which parameters of a production function the rules are likely to affect.

Casey Ichniowski


I. Introduction

Despite a growing number of studies that document a significant positive
relationship between union status and product ivity,! a number of criticisms
remain unanswered. Among these complaints, one of the most frequently voiced by
managers is that unionized establishments operate under more restrictive work
rules and practices and that other potential gains of unionization, lower tur-
nover rates or formal grievance machinery, for example, could not outweigh the
inefficiency associated with these added restrictions. Despite the frequency
with which this complaint is raised, no direct empirical studies on the rela-
tionship between work rules and productivity have been made. The few studies
that have been conducted, address this issue indirectly by focusing on differen-
ces in substitution parameters obtained from equivalent union and nonunion pro-
duction functions. 2 This study takes a first step in analyzing the relationship
between work rules and productivity in a more direct fashion.

In this study, monthly data from January 19T6 to September 1982 on the
operations of eleven paper mills are analyzed. The number of pages in collec-
tive bargaining agreements is taken as a directly measurable proxy for the
number and complexity of work rules. This proxy for the extent of , work rule
regulation is then considered within the framework of a detailed, plant-level
production function to gauge the differences in productivity associated with
changes in the contract page measure.

The analysis is developed in the following five sections. The next sec-
tion presents the production function framework and describes the contract pages


variable. Section III presents estimates of the productivity-contract pages
relationship obtained from this production equation. Section IV considers
several other variables that might also indicate differences or changes in work
rules and practices within the production equation. The empirical analysis is
expanded in Section V by considering data for a nonunion mill. In this section,
an average productivity difference between this mill and the ten unionized mills
is calculated. While this mill has no collective bargaining aigreeraents,
contract pages values are predicted from the basic equation estimated in Section
III and input-output data for the nonunion mill. Finally, the conclusion sum-
marizes the results and argues for two improvements on the design of this study:
improvements in the measurement of work rules; and suggestions for improving the
model of how different types of work rules might affect the production process,
thereby altering the parameters of the production function.

By way of preview, a significant inverse relationship between plant pro-
ductivity and contract pages is documented. The estimated "output-contract
pages elasticity" is -.063. Additionally, several periods when work rules
are likely to be changing or in dispute are shown to be extremely unproductive
periods of plant operation. Specifically, around the time of contract nego-
tiations, and around the time that new machinery is installed, the plants pro-
duce significantly less output than one would expect given the stated level of
inputs. Finally, by using input-output data for a nonunion mill, an implicit
number of contract pages is predicted for a mill that has no collective
bargaining agreements. On average, this mill's productivity is 9.5*" below that
of the unionized mills in the sample.



With the aid of on-site investigations of each mill's production process,
the production fanction given by equation 1 was developed to account for
variations in productivity in these mills: 3

3 9 3 t

-r-j-r- In Q = Sq + Zi&n • KD^) + E (Bgi * ^^i) + ^iB^i ' PMDi) + Z (BUi *

i=l i=l i=l i=l

" ^' ■ PMV^) + (^5 • L) + (eg • E) + e (Equation l)

Where: Q = tons of physical output;

~~:i- KD]__3 = three plant dummies to control for two major product differences
-n;.2 (white paper vs. newsprint; sheeted vs. not sheeted) and one

;f - ' :~ major process difference (make vs. buy pulp) across the eleven

* :■- - mills;

E29_^r- 5^1_9 = total depreciated, deflated value of assets in nine distinct cate-
gories of aissets;
"" '- ?MD]__3 = a set of three related durany variables to describe whether a plant
is operating two, three, four, or five or more paper machines (the
~ two paper machine category is omitted);

PMV]__li = total depreciated, deflated value of the two, three, four or five
""''-'■" plus paper machines;

L = labor input
■ ' E = energy input

The PCD variables provide a direct control for major product and process
differences observed in these mills. The more conventional method of

constructing a value added index is particularly difficult in these mills. ^ The
PMD variables provide some control over scale of operations. The KV and PMV
variables are fashioned to recognize the principles of input aggregation for a
heterogeneous capital stock. 5 For example, three categories involving energy
generation capital, certain land and buildings, and pollution and .recycling
capital are not a direct part of the machinery that acts upon the raw materials
flowing through the process. These categories of capital, then, are kept
separate from other categories of capital that is part of the production pro-
cess. The capital value variables are constructed from each mill's monthly
asset inventory which contains information on the current value of each asset.
In any month, there are some 15,000 assets that were allocated to these dif-
ferent categories of capital. L is defined as the natural logarithim of hourly
manhours. E is the natural logarithm of BTU's used in production.

This unconventional specification is developed to provide an accurate
model of the production processes in these mills. Equation 1 accounts for
over 95% of the total variation in production in this sample. More conven-
tional functional forms produce several nonsensical coefficients. For example,
in a Cobb-Douglas function, with capital inputs specified as one net investment
variable, the coefficient on capital is in fact negative for this set of plants
in which capital plays the central role in transforming raw materials into final
goods. More conventional forms explain a much smaller proportion of the total
variation in output. °

In this study, total pages of collective bargaining contracts in effect.


. _^ ^CONPAa, vn.ll serve as the proxy for the extent and complexity of vorknale reg^a-
, -(Ration in the mills. This is not an entirely satisfactory measure for several
reasons. It's most noticeable shortcoming is that it does not capture the

substance of the rules and practices in effect. Even setting aside this con-

cern, longer contracts may not necessarily mean more rules. Sidebar agreements
. . not incorporated in the contracts, rules set forth in arbitration decisions, or
unwritten shop floor practices will not be captured by this meausure. These
other sources of work rules may not necessarily increase with contract pages.
Still CONPAG is a direct measure and should provide information on work rules
since labor contracts are a major source of work rules. During 19T6 to I9S2,
each of the ten unionized had either three or four contract cycles. The average
number of pages in effect across all ten unionized mills over the seven year
period is 195 pages. The fewest number of pages ever in effect in these mills
is 68 pages; the maximum U65 pages. The lowest mill average is 78 pages; the
largest mill average is 379 pages.

While the format smd layout of the collective bargaining agreements is
similar from mill to mill, there are slight layout variations across mills. To
parcel out these plant-specific differences in the plants' contract formats, a
complete set of plant dummies, will replace the dummy variables described in
Equation 1, (KD]__3 and PMi_U ) that control for differences in products and
processes across plants. With this modification and the addition of the CONPAG
variable, the final specification becomes:

In Q = cSq + 1 (liii • PLANT) + I (tJ=i • KV^) + \ (631 • PMV^) +
i=l i=l i=l


(SU • L) + (65 • E) +e5(C0NPAG) {Equation 2)

where CONPAG will also be entered in logarithmic form.

Even if CONPAG does measure the extent and complexity of work rule regula-
tion, the relationship between productivtiy and the extent of work rule regula-
tion might not be described by the log-linear model of Equation 2, For example,
some degree of work rule regulation may promote efficient operations, while an
overabundance of rules might hinder efficiency. Therefore, the equation 2 model
will be expanded to include the square of the natural logarithim of CONPAG.


III. Empirical Results: The Contract Pages-Productivity Relationship

When the equation 2 specification is estimated, the results in Table 1 are
obtained. From the column 1 model, one sees that the overall relationship bet-
ween contract pages and output, controlling for variations in the levels of pro-
ductive inputs, is negative and significant. The estimated "output-contract
pages elasticity" is -.068. However, from the coefficients in column 2, one
observes that there appears to be some curvature in the contract pages-
productivity relationship. These coefficients suggest that output at first
decreases with additional contract pages; after some inflection point, output
increases with additional contract pages. To solve for the value of this
inflection point in units of contract pages, one can use the column 3 specifica-
tion and solve the following equation for CONPAG:


= (-3.3 E-3) + 2(5.1 E-6) • (CONPAG) =

The inflection point occurs at 32U contract pages. The vast majority of values

of the contract page variable are below 32U pages. Those that do exceed 32U
pages (only five of the thirty-five values of the contract page variable) tend
to cluster near the inflection point of 32U pages.

To illustrate the nature of the relationship between contract pages and
productivity in this sample, the results in Table 1 are presented graphically.
Figure 1 uses the coefficients in the column 3 specification to evaluate
3 Q/3 CONPAG for each value of the contract page variable in the sample (as indi-
cated by the open dots). Except for the two largest contract pages values
(CONPAG = 390, U65), the inverse relationship between contract pages and produc-

Table 1: Contract Pages Coeffiecients in Production Function^
[Dependent Variable: In Tons of Paper; N = 626]

1. In contract pages

2. (in contract pages) 2

3. contract pages

h. (contract pages) 2

5 . plant dumnies smd other
input controls in
equation 2 Specification








-3.3 E-3***
(0.6 E-3)

5.1 E-6***
(0.9 E-6)







a -

»»* _

»* _

* _

standard errors in parentheses

significant at the .01-level, one-tailed test

significant at the .05-level, one-tailed test

significant at the .10-level, one-tailed test

«•*«• ,_



tivity holds for virtually the entire range of contract page values, liven the
small nunber of observations to the right of the inflection point in Figure 1,
and the imperfect nature of contract pages as a measure of work rules, it would
seem -onreasonable to extend the positive contract page-productivity relationship
beyond the range or observed values. Again, the simple linear "output-contract
pages elasticity" is significant and negative (-.068). Probably the most reaso-
nable conclusion one should draw from these data is that contract pages provides
some information on the extent of work rule regulations and that there exists an
inverse relationship between work rules and productivity, holding fixed the
levels of other productive inputs.











IV. Changing Work Rules: New Contracts and New Machinery

While the results of the previous section provide some support for the
claim that increased work rule regulation inhibits productivity, it may he
that the process of changing work rules also affects a plant's efficiency.
Two sets of variables are constructed to measure periods when work rules come
under pressure to change. First, a set of four dummy variables is created to
describe each plant's contract cycle: the quarter before negotiations without a
strike; the quarter after negotiations without a strike; the quarter before a
strike; the quarter after a strike; and the omitted group of months from the
central period of contract administration. Second, there is a set of variables
to describe the periods when four major machines were installed in these mills
during this seven year period. One variable is created for the six month
installation period; a second for the first six months of operating the new


A number of different forces may be operating during these periods to
influence how efficiently the plants are operating. Before negotiations, mana-
gers may try to speed up production to build inventories of standard products
thereby dampening the economic power of the union during a possible strike;
employees may try to reduce output for the opposite reasons. After nego-
tiations, particularly if a strike occurred, the plant may not be operating
efficiently as it tries to adapt to the new agreements put in place during nego-
tiations. Conversely, after negotiations without a strike, there may be a
honeymoon period in which employees operate the plant at higher efficiency


levels .

For the new machinery variables, the installation of a major piece of
capital may disrapt existing plant operations. After the capitalization date,
the plant's capital variables are increased by the full purchase price. This
may overstate the value of the capital in its early stages of operations for
several reasons. First, a learning curve effect may exist as operators learn
how to operate the machine. Second, and directly related to the work practices
in the plant, the Jurisdiction over any new work may be in dispute as old work
rules fail to cover the new situation in the plant. As these issues are sorted
out, the machinery may not be operating fully. Finally, it may simply take
several months to integrate new machinery into the existing production process.

When the new variables are added to the previously estimated equations,
the results in Table 2 are obtained. The addition of these new sets of
variables leave the contract pages coefficients virtually unaffected. The
simple output — contract pages elasticity is still -.068 (column l). The addi-
tion of the squared contract pages term improves the specification and the same
pattern of curvature in the output-contract pages relationship seen in Table 1
still exists (column 2).

The coefficients in lines 3a-3d, describe the relative levels of produc-
tivity during the periods when the number and substance of the contract pages
are changing. These coefficients reveal that the periods around the time
of negotiations are relatively productive periods given the stated levels of
inputs. The most significant among these positive coefficients is the period

Tatle 2: Contract Page, Contract Cycle, and
Hew Machinery Coefficients in Production Equationa
Lable: In Tons of Paper; N = 62b 1

[Dependent Vari?

1. In contract pages

2. (In contract pages )2

3. Contract cycle variables

a. quarter before

b. quarter after

c. quarter before a strike

d. quarter after a strike

U. New machinery variables

a. six-month installation

b. six-month after

5. Plant dummies and other

input controls in Equation
2 specification




























a -

««» _

«« _

» _

standard errors in parentheses

significant at the .01-level, one-tailed test
significant at the .05-level, one-tailed test
significant at the .10-level, one-tailed test


after negotiations without a strike (line 3^) - possibly signalling some
honeymoon period; and the quarter before strike - possibly signalling the
efforts of management to build up inventories before the strike.

The coefficients in lines Ua and Ub, on the other hand, show the periods
around the introduction of new machinery to be quite unproductive periods. The
installation period (line Ua) is from 8.3fo to IO.S'^'j less productive than normal
given the stated level of inputs. After installations (line hb) , productivity
is 5.8% to 6.6% below what would be expected given the level of inputs in
place. While the installation of major pieces of machinery is likely to
disrupt the existing production process, the post-capitalization period may be
relatively unproductive for a combination of reasons. Work practices may need
to be adjusted in the face of the new operating needs of the plant; employees
may need several months to learn how to operate the equipment effectively; the
capital itself may not be well integrated with the production process.

Despite the significant effects of these sets of variables on production,
the contract pages - productivity relationship remains unaffected: for most
of the range of values for the contract page variable, output declines with
more contract pages. The simple output-contract pages elasticity remains
significantly negative at -.068.


Y J .

V. Estimating Contract Pages for a Non-Union Firm.

Monthly data for one nonunion mill were also available in this sample.
However, without negotiated agreements, it could not be incorporated in the
preceding analysis. Still, these data can be incorporated in another way.
First, by estimating the equation 1 specification without plant dummies, and
only one additional dummy for the nonunion mill, the average difference between
this mill and the ten unionized mills can be gauged. This analysis reveals that
the nonunion mill is on average 9.5% less productive than the union mills for
the stated level of productive inputs. The coefficient is significant at con-
ventional levels and in keeping with the union productivity effects estimated in
other studies.

Second, the coefficients obtained in the model in column 1 of Table 1 can
be used to estimate an implicit number of contract pages for each month for the
nonunion mill."^ Simply, by evaluating that equation for the input-output values
observed in the nonunion mill, the predicted contract page values are obtained:


In Qnu - (a -t- ginu)
CPNPAGn,, = • (Equation 3)


where CONFAGfj^ = predicted nonunion contract pages
Qnu = output data from nonunion mill

Inu = input data from nonunion mill

2 '^CONPAG ~ estimated coefficients obtained in Table 1, column 1 specifi-
cation using data from sample of unionized firms.
Since the nonunion mill is 9.5% less productive than the union mills, the


estimated contract page values for the nonunion mill will tend to exceed the
contract page values observed in the union mills. That is, this -9.5% nonunion
differential will be attributed entirely to increased numbers of contract
pages. While it may be more reasonable to suspect that a number of other fac-
tors omitted from the basic production model are partly responsible for the
observed productivity differential between the nonunion mill and its unionized
counterparts, the analysis provides a useful illustration. It illustrates how
much one would have to exaggerate the effect of contract pages on productivity
to account for the entire union productivity differential.

When this analysis is performed, and the nonunion productivity differen-
tial is attributed solely to a greater number of contract pages, an implicit
number of contract pages for each month is estimated for the nonunion mill. The
average value for the estimated In contract pages variable is 15.27. The
average value for In contract pages among the union mills is only 5.08. For the
entire nonunion productivity differential to be accounted for by the effect of
contract pages, the implicit number of "nonunion contract pages" would have to
far exceed the number of contract pages in the unionized mills.

The size of this differential in contract pages indicates that other fac-
tors would also seem to be responsible for the observed union productivity dif-
ferential. Still, it may not be unreasonable to consider the nonunion mill
as having a much less efficient set of work rules than the union mills.
Specifically, by not having rules well specified in explicit negotiated
agreements, more confusion about operations may in fact exist in the plant.


Furthermore, there is some evidence that work rules proliferate as a plant
ages. 3 The nonunion mill is one of the oldest establishments in the sample.
The combination of proliferation of informal work practices as the mill ages
coupled with an inability to use one principal source (a labor contract) as an
official statement of those practices may be a serious obstacle to efficient


VI, Conclusion

This study documents a significant inverse relationship between contract
pages and productivity for a sample of unionized plants. While the introduc-
tion of a squared contract pages term in the production equation indicates a
positive output-contract pages relationship among the few large contract page
values, the negative relationship holds for most of the range of contract page
values .

The absence of a uniformly negative relationship between contract pages
and productivity is likely due to several shortcomings of this study that future
research on this topic should attempt to remedy. First, contract pages, while a
directly measureable proxy for work rules, is far from the ideal measure. Other
sources of work rules (arbitration rulings, sidebar agreements, and informal
practices) need to be considered.

More importantly, the substance of work rules and how they interact with
other inputs in the production function need to be modelled directly. Many work
rules probably alter substitution parameters; for example; limitations on
introducing labor-saving devices, limit capital-labor substitution. Other rules
may increase the number of inputs that need to be specif iced. For example, if
there are jurisdictional boundaries between operating and maintenance jobs,
one would want to specify the labor input as two components with substitution
limited according to how often the jurisdictional boundaries can be crossed.

Finally, it seems reasonable to expect that management supervision is a
critical variable omitted from the above analysis. With low levels of super-


Jos- -


vision, more work rule regulation would probably be beneficial to efficient
plant operations. For increased -levels of supervision, rules would inhibit
managers ability to deploy resources as it sees fit.

While providing some direct evidence of a inverse relationship between


Online LibraryCasey IchniowskiRuling out productivity? : labor contract pages and plant performance → online text (page 1 of 2)