Digitized by the Internet Archive
in
2011 with funding from
Boston Library Consortium Member Libraries
http://www.archive.org/details/dofirmboundariesOOmull
DEWEY
i~
Massachusetts Institute of Technology
Department of Economics
Working Paper Series
DO
FIRM BOUNDARIES MATTER?
Sendhil Mullainathan
David Scharfstein
Working Paper 01 -05
January 2001
Room
E52-251
50 Memorial Drive
Cambridge, MA 02142
This paper can be downloaded without charge from the
Social Science Research Network Paper Collection at
http://papers.ssrn.com/paper.taf7abstract id=259926
Massachusetts Institute of Technology
Department of Economics
Working Paper Series
DO
FIRM BOUNDARIES MATTER?
Sendhil Mullainathan
David Scharfstein
Working Paper 01 -05
January 2001
Room
E52-251
50 Memorial Drive
Cambridge,
MA 021 42
This paper can be downloaded without charge from the
Social Science Research Network Paper Collection at
http://papers.ssrn.com/paper.taf7abstract _id=259926
J
MASSACHUSfcl IS INSTITUTE
OF TECHNOLOGY
AUG 2 2
2001
LIBRARIES
Do Firm Boundaries Matter?
Sendhil Mullainathan and David Scharfstein*
In his
famous
article,
"The Nature of the Firm," Ronald Coase (1937) raised two
Do
fundamental questions that have spawned a large body of research:
affect the allocation
drawn? While the
firm boundaries
of resources? And, what determines where firm boundaries are
first
of these questions has received some theoretical attention
notably Oliver Williamson (1975, 1985), Benjamin Klein, Robert Crawford, and
Alchian (1978) and Sanford Grossman and Oliver Hart, (1986)
ignored empirically. Instead, the empirical
work
—
it
—
Armen
has largely been
in this area, discussed in the other
articles in this session,
has addressed the second question by analyzing the determinants
of vertical
Thus, while
integration.
firm boundaries,
behavior. This
we know
is
relatively
assess
how
how
these boundaries affect actual firm
we
analyze whether there
between integrated and non-integrated chemical manufacturers
the sole use of which
polyvinyl chloride (PVC).
liquid markets.
about
firm boundaries affect behavior,
investments in production capacity.
(VCM),
little
about the forces that determine
a major limitation in our understanding of the nature of the firm.
To begin to
are differences
we know something
is
is
We focus on producers of vinyl chloride monomer
in the production
of the widely used waterproof plastic,
VCM is a homogenous commodity and is traded in relatively
Moreover, there
other than that one
in their
is
no obvious production
link
an input into the other. For example,
between
PVC
is
VCM and PVC
not a by-product of
VCM production. Nevertheless, two thirds of VCM producers in our sample are
integrated
downstream
degree of integration.
into
We
PVC. The
existing literature
would ask why we observe
ask instead whether integrated and non-integrated
producers invest differently in production capacity.
VCM
this
I.
Data
Our
analysis
is
based on a global dataset of
provided to us by SRI Consulting,
Inc., a
firm based in
contains detailed plant-level information on
1974 to 1998 in 61 countries.
and consumption of
VCM and PVC manufacturers
Menlo
Park,
CA. The
VCM and PVC production capacities from
on production
also includes country-level information
It
dataset
VCM and PVC, but no information on prices and plant-level
We construct firm-level production capacities of VCM and PVC by
production.
aggregating this data across plant-level observations.
We restrict our analysis to firms
operating in consistently market-based economies with a
minimum
level
of GDP. This
leaves us with 1257 observations in 17 countries.
Table
1
provides
summary information on
firm accounts for 13.2% of country
VCM capacity (280.15/21 15.72).
account for two thirds of the observations.
The firms
their integration.
consume
internally,
we
that
label as
the firms in our sample.
1
have more
These firms
PVC
plants that they
own, we
to the
demand
label
ratio
— the
upstream supply of
merchant firms because they
internal
of
likely sell part
of their
VCM capacity than they need for
PVC
them captive producers.
VCM producers are merchant or captive, we calculate the
ratio
of a firm's demand for
VCM from
its
own PVC
plants
VCM (assuming both types of plants operate at full capacity).
This calculation uses information on the units of
2
PVC. The
differ themselves in the nature
production. Because these firms likely only supply to downstream
To measure whether
internal
Integrated firms
VCM production capacity than they can
output to the outside market. Other firms have less
downstream
The average
demand
VCM required to produce one unit of
ratio is zero for non-integrated
VCM producers; positive but
less than
Table
1
one for "merchant"
indicates that the average internal
are merchant producers and
II.
VCM producers;
41.7%
and greater than one
demand
ratio is 1.10.
are captive producers
by our
for captive producers.
58.3% of integrated firms
definitions.
Analysis
Integration and the Sensitivity of Capacity to Consumption
We begin by examining the sensitivity of VCM capacity to
consumption of VCM. To do
this,
we
=
+
estimate the following simple equation:
(1)
log(Capacityn)
The dependent
variable in this equation
The key explanatory
variable,
a,
includes firm fixed effects
+ c*log(Consumptioni )
b,
t
is
the log of
\og(Consnmption it)
producers in the country where firm
(a,),
i
is
is
VCM capacity of firm
log consumption of
i
in year
/.
VCM by PVC
located at time t? Because the regression also
the coefficient c measures the extent to
changes in capacity correspond to changes in consumption.
-
downstream industry
which firm
We expect c to be positive —
that firms will increase capacity in line with consumption.
We are
of course,
interested,
non-integrated firms. Consequently,
two groups. Comparing the
integrated firms
is
more
in
how
we
these sensitivities differ for integrated and
will estimate this equation separately for these
resulting sensitivities will then tell us
whether the capacity of
or less sensitive than the capacity of non-integrated firms to
downstream consumption.
The
first
column
sensitive to consumption.
one percent increase
in
in
Table 2 indicates that non-integrated firms are extremely
The
coefficient of 0.977 implies roughly an elasticity of
consumption corresponds roughly
capacity for non-integrated firms.
to a
The second column shows,
1
1:
a
percent increase in
in contrast, that integrated
Not only
firms are completely insensitive to consumption.
statistically insignificant, the
magnitude
is
is
the estimated sensitivity
extremely small (0.038).
suggest that integrated firm capacity does not
move
among
The
producers.
The
firms.
firms, but
less in
integrated firms,
third
is
the regression coefficients for merchant
substantially smaller than that of the non-integrated
borderline statistically significant. This suggests that merchant firms
market than captive firms
of capacity to consumption
is
The
are.
-0.160. which
expect and statistically insignificant.
lie
firms. But, as noted
compare merchant and captive
tandem with markets than do non-integrated
step with the
firms
interesting to
column of Table 2 shows
coefficient of 0.205
it is
it is
results
with the market.
in step
These regressions compare integrated to non-integrated
earlier,
These
As
is
move
firms; however, they are less out of
fourth
column shows
that the sensitivity
both the opposite sign than one would
a whole, these results suggest that merchant
somewhere between stand-alone and captive firms
in their sensitivity to
market
conditions.
Do
Integrated Firms Hold Capacity at the Right Times?
While these
results tell us that integrated firms are less sensitive than
integrated firms to outside conditions, they
is
good or bad. For example, suppose
firms are overly sensitive to
that case, the
assess this,
lower
we now
to consumption.
(2)
demand
sensitivities
ask
how
do not
tell
non-
us whether this reduced sensitivity
that industries are stuck in
shocks, investing too
much
hog
cycles, in
which
during good times. In
by integrated firms would actually be a good
thing.
To
sensitive integrated firms are to industry capacity as well as
We estimate:
log(Capacityu)
=
a,
+ b + c*log(ConsnmptionjJ + d*log(VCMCapacityj,)
t
VCMCapacityu, denotes the
In this equation,
at
time
firm
<
is
0;
t
(excluding the firm's
to the capacity
when
capacity).
The
VCM capacity in the country
coefficient
higher, for the
is
column of Table 3
first
same
level
We see,
sensitive the
We would expect d
as before, that the capacity of non-integrated firms
we
also see that their capacity
For the same level of consumption, a
1
3 indicates that integrated firms
behave quite
consumption (0.074), but
it
coefficient of industry capacity
is
0.
also sensitive to
industry capacity
is
high
The second column of Table
18%.
differently.
As
before, their capacity
is
also quite insensitive to industry capacity.
is statistically
magnitude of the coefficient for non-integrated
if integrated firms are better at
when
is
comoves
percent increase in industry capacity
reduces non-integrated firm's capacity by roughly
third the
how
of consumption, the firm has
industry capacity; they are less likely to increase capacity
While the
us
significant,
firms.
it
is less
than one
In other words,
it is
"timing" the market and simply holding the capacity
not as
at the
right time.
firms.
The
last
two columns of Table 3 again separately analyze merchant and captive
The
tird
column shows,
sensitive to
firms.
It
as before, that the capacity of merchant firms
consumption (0.301), though
also
i
reports the results of estimating this equation for non-
closely with consumption (0.913). But,
insensitive to
tells
of firm
hold capacity.
integrated firms.
(-0.184).
d
of competitors, holding constant consumption.
industry capacity
less incentive to
The
own
total
shows
less so than the capacity
is
in fact
of non- integrated
that their capacity is negatively related to industry capacity,
not statistically significant. In the fourth column,
the coefficient
is
firms not only
show no
sensitivity to
consumption, but also very
we
little
though
see that captive
sensitivity to
Their sensitivity to capacity
industry capacity.
merchant firms are somewhere
Is the
in
is
half that of merchant firms.
between stand alones and captive
Capacity of Integrated Firms Sensitive to Internal
If capacity
what does
it
possibility
product from their downstream
is
PVC
that they
plants
To
before,
firms.
Demand?
of captive firms does not comove with external market
move? One
As
factors, with
respond to internal demand for their
assess whether this
the case
is
we now
estimate the following regression for integrated firms:
(2) log(Capacityii)
=
a,
+
b,
+ c*log(Consumptioriit) + d*log(VCM Capacitya) +
e *log(InternalDemandi,)
In this equation
PVC
InternalDemand measures the implied demand
it
capacity at time
t
(assuming that the downstream
for
PVC plants
VCM from firm i's
operate at full
capacity).
The
first
that the capacity
column of Table 4 estimates
of these firms
is
this equation for
merchant firms.
very sensitive to internal demand.
4
One
column
sensitivity to internal
demand. Thus, both merchant and captive firms
capacity, but only to internal
III.
demand not
see
finds a similar
though captive firms show greater
pattern in the second
for captive firms,
We
in fact adjust their
external demand.
Conclusion
Our
findings suggest that there are differences in the
integrated firms allocate resources.
to be insensitive to
producers
is
way
The capacity of integrated
integrated
and non-
VCM producers appears
downstream consumption, while the capacity of non-integrated
very sensitive to consumption. The capacity of integrated producers
depends only on the internal demand for
VCM from
its
own downstream PVC
units.
These empirical findings suggest
capacity for internal
their product.
Our
of capacity relative
What might
us.
demand only and
largely ignoring changes in external
seem
to
to hold capacity
first
it
is
least
needed,
i.e.,
explain this insulation from the external market?
two
possibilities
might be broadly called organizational focus. Managers
different markets.
is
essentially
used for
PVC
when
PVC
market. This focus on
failing to notice a profitable opportunity in
not focused in the past,
it
may
PVC may
VCM production.
lack the contacts or specific
Our
there
is
a lot
investigation
seem important
in a firm that
may be
production
These managers may view themselves as primarily
focus their energies on the
the outside
for
consumption.
VCM plant whose output
sell to
when
too preliminary to single out one answer. But
The
demand
findings also suggest a potential inefficient allocation of resources in
that integrated firms
much
that integrated firms are inwardly focused, holding
is
to
has a
focusing on
PVC
producers and
in turn result in
them
And, because the firm has
human
capital necessary to
VCM market.
A second possible explanation of our findings would stem from differences in the
way
transfer prices
that differs
and market prices are
from the market price
— say
set.
at
If
VCM producers sell internally at a price
average cost
— then they may not respond as
strongly as non-integrated producers to changes in market demand.
merchant producers
somewhat
like
— those
that appear to sell internally
non- integrated producers suggests that
this
The
and externally
may be
allocation of resources.
tell
— behave
part of the explanation
of our findings. Separating out these (and possibly other) explanations
avenue of future research that will help to
fact that
is
an interesting
us whether firm boundaries affect the
References
Coase, Ronald, "The Nature of the Firm," Economica, 1937, 4, 386-405.
Grossman, Sanford, and Oliver Hart, "The Costs and Benefits of Ownership: A
Theory of Vertical and Lateral Integration," Journal of Political Economy, 1986, 94, 691719.
Klein, Benjamin, Robert Crawford, and
Armen
Alchian, "Vertical Integration,
Approbriable Rents, and the Comeptitive Contracting Process," Journal of Law and
Economics, 1978, 21, 297-326.
Williamson, Oliver, Markets and Hierarchies: Analysis and Antitrust
Implications
.
New York:
Free Press, 1975.
Williamson, Oliver, The Economic Institutions of Capitalism
Press, 1985.
New York:
Free
Table
Observations are by firm, chemical
PVC
and
firms only. It is the ratio of
capacity.
Country
firm produces both
1:
Summary
(VCM) and
VCM.
Internal
Statistics
year.
Integrattion
Demand
Ratio
Dummy
is
equals
1
if the
defined for integrated
VCM consumed to produce PVC at full capacity to total
VCM Capacity Utilization
the ratio of country VCM
VCM
production to
is
VCM Capacity.
No. of
Mean
Std.
Dev.
Min.
Max
Obs.
Firm VCM Capacity (0001b)
Country VCM Capacity (0001b)
1257
280.15
226.31
8.00
1634
1257
2115.72
1719.71
8.00
7902.00
Dummy (0 or
PVC Capacity (0001b)
1257
0.67
0.47
Internal
842
235.10
208.24
16.00
1622.00
Internal
Demand
842
1.10
0.81
0.083
10.20
1257
1861.40
1543.67
28.00
6.97
1257
0.82
0.17
0.00
1.25
Integration
1
Ratio
Country
PVC
Country
VCM Capacity Util.
Capacity (0001b)
1
Table 2
Integration and Sensitivity to
Downstream Demand
(VCM) and year. Dependent variable is log VCM capacity. Integrated
VCM and PVC capacity; non-integrated firms are those that have only VCM
capacity and no PVC capacity. Integrated merchant firms are integrated firms with VCM capacity that
exceeds their internal demand for VCM from their PVC plants. Integrated captive firms are integrated firms
with VCM capacity is less than their internal demand for VCM from their PVC plants. The regressions
Observations are by firm, chemical
firms are those that have both
include firm and year fixed effects, t-statistics, which are based on robust standard errors clustered by
country and year, are in parentheses.
Non-
Integrated
Integrated
Log Country
VCM
Integrated/
Integrated/
Merchant
Captive
0.977
0.038
0.205
-0.160
Consumption
(6.24)
(0.35)
(1.95)
(1.14)
R^
0.922
0.942
0.949
0.966
415
842
492
351
Number of Observations
Table 3
Integration and Sensitivity to
Observations are by firm, chemical
firms are those that have both
capacity and no
PVC
(VCM) and
VCM
capacity.
and
PVC
year.
Downstream Demand
Dependent variable
is
log
VCM
capacity. Integrated
capacity; non-integrated firms are those that have only
Integrated merchant firms are integrated firms with
VCM
VCM
capacity that
VCM
exceeds their internal demand for
from their PVC plants. Integrated captive firms are integrated firms
with
capacity is less than their internal demand for
from their PVC plants. The regressions
include firm and year fixed effects, t-statistics, which are based on robust standard errors clustered by
VCM
VCM
country and year, are in parentheses.
Non-
Integrated
Integrated
Integrated/
Integrated/
Merchant
Captive
Log(Country VCM
Consumption)
0.913
0.074
0.301
-0.137
(5.59)
(0.65)
(2.64)
(0.88
Log(Country Capacity)
-0.184
-0.050
-0.074
-0.037
(3.01)
(2.49)
(1.73)
(0.70)
0.924
0.943
0.952
0.967
415
842
492
351
R2
Number of Observations
10
Table 4
Integration and Sensitivity to
Downstream Demand
(VCM) and year. Dependent variable is log VCM capacity. Integrated
VCM and PVC capacity; non-integrated firms are those that have only VCM
capacity and no PVC capacity. Integrated merchant firms are integrated firms with VCM capacity that
exceeds their internal demand for VCM from their PVC plants. Integrated captive firms are integrated firms
with VCM capacity is less than their internal demand for VCM from their PVC plants. The regressions
Observations are by firm, chemical
firms are those that have both
include firm and year fixed effects, t-statistics, which are based on robust standard errors clustered by
country and year, are in parentheses.
VCM
Integrated/
Integrated/
Merchant
Captive
Log(Country
Consumption)
0.051
-0.545
(0.52)
(3.89)
Log(Country Capacity)
-0.017
0.028
(0.89)
(0.94)
Log(Internal
VCM Demand)
R2
Number of Observations
0.476
0.681
(4.49)
(5.86)
0.971
0.976
491
351
11
Footnotes
*MIT and NBER; MIT and NBER. We
support, to
Oghuzan Ozbas
are grateful to the National Science Foundation for research
for exceptional research assistance, Paul Bjacek of
SRI Consulting
for
supplying data and insights and to Jeffrey Zwiebel for helpful comments.
Note
that firms that
further upstream
we
consider non-integrated for the purposes of this analysis may, in
from ethyl dichloride
into
VCM. Our focus,
however,
is
on the
fact,
VCM-PVC
be integrated
integration
link.
"
Private
*
The consumption measure
is
communication with SRI,
defined
at the
understated.
To
regressions.
It is
is
Inc.
defined
at
the country-year level, whereas the dependent variable (capacity)
firm-year level. One, therefore, naturally worries that the standard errors
deal with this problem,
we have allowed
also worth noting that in this equation
we
abstracting from changes in capacity utilization. Ideally,
level data
One
on
artifact
utilization
we
for firm-year
will
random
may be
effects (clustering) in all the
be focusing on capacity changes and
we would
like to look at both but lacking firm
can only examine one of the two dimensions for adjustment.
of this regression
is
worth noting. The addition of internal demand makes the effect of
country consumption insignificant. This, however,
is
because downstream industry capacity
is
a better
predictor of upstream capacity than downstream industry consumption. If we use downstream capacity
instead,
it
remains significant.
12
Filename:
Mullain0105
Directory:
Template:
C:\Program Files\Eudora\Attach
C:\Program Files\Microsoft Office\Templates\Normal.dot
Title:
3
Subject:
Author:
Sendhil Mullainathan
Keywords:
Comments:
Creation Date:
02/13/01 2:53
Change Number:
Last Saved On:
Last Saved By:
2
Total Editing Time:
Last Printed On:
02/13/01 2:53
linda
PM
PM
woodbury
Minutes
02/13/01 3:23
As of Last Complete Printing
Number of Pages:
13
Number of Words:
2,481
Number of Characters:
PM
(approx.)
14,145 (approx.)
6*f5
36
Date Due
Lib-26-67
MIT LIBRARIES
3 9080 02246 0809