Digitized by the Internet Archive
in
2011 with funding from
Boston Library Consortium IVIember Libraries
http://www.archive.org/details/identifyingagglo00gree2
^$>3i
DEWEY ll
c7>'3i
Massachusetts
Institute of
Technology
Department of Econonnics
Working Paper Series
AGGLOMERATION SPILLOVERS:
EVIDENCE FROM MILLION DOLLAR PLANTS
IDENTIFYING
Michael Greenstone
Richard Hornbeck
Enrico Moretti
Working Paper 07-31
December 19, 2007
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
078027
http://ssrn.com/abstract=1
;1
5
at
Identifying Agglomeration Spillovers:
Evidence from Million Dollar Plants*
Michael Greenstone
Richard
Hombeck
Enrico Moretti
December 2007
*We
thank Daron Acemoglu, Jim Davis,
Vemon
Henderson, William Kerr, Jeffrey
Rosenthal, Christopher Rohlfs, Chad Syverson, and seminar participants
MIT,
NBER Summer
Institute,
at
KLling,
Jonathan Levin, Stuart
Berkeley, the Brookings Institution,
San Francisco Federal Reserve, Stanford, and Syracuse
for insightful
comments.
Greenwood provided valuable research assistance. The research in this paper was conducted while the
authors were Special Sworn Status researchers of the U.S. Census Bureau at the Boston Census Research Data
Center (BRDC). Support of the Census Research Data Center network from NSF grant no. 0427889 is gratefully
Elizabeth
acknowledged. Research results and conclusions expressed are those of the authors and do not necessarily
views of the Census Bureau. This paper has been screened to insure that no confidential data are revealed.
reflect the
Abstract
We
agglomeration spillovers by estimating the impact of the opening of a large
quantify
new
(TFP) of incumbent plants in the same county.
journal
Site Selection reveal the county where the "Million Dollar
Articles in the corporate real estate
Plant" ultimately chose to locate (the "winning county"), as well as the one or two runner-up counties (the
"losing counties"). The incumbent plants in the losing counties are used as a counterfactual for the TFP
manufacturing plant on the
total factor productivity
of incumbent plants in winning counties in the absence of the plant opening. Incumbent plants in winning
and losing counties have economically and statistically similar trends in TFP in the 7 years before the
opening, which supports the vahdity of the identifying assumption.
After the
new
plant opening, incumbent plants in winning counties experience a sharp relative increase in
TFP. Five years after the opening, TFP of incumbent plants in winning counties is 12% higher than TFP
of incumbent plants in losing counties. Consistent with some theories of agglomeration, this effect is
larger for incumbent plants that share similar labor and technology pools with the new plant. We also
find evidence of a relative increase in skill-adjusted labor costs in winning counties, indicating that the
ultimate effect on profits is smaller than the direct increase in productivity.
Michael Greenstone
MIT Department of Economics
50 Memorial Drive, E52-359
Cambridge,
02142-1347
and NBER
MA
mgreenst@mit.edu
Richard
Hombeck
MIT Department
of Economics
E52-391
77 Massachusetts Avenue
Cambridge
02142-1347
hombeck(a)mit.edu
MA
Enrico Moretti
University of California, Berkeley
Department of Economics
Berkeley, CA 94720-3880
and NBER
moretti@econ.berkeley.edu
Introduction
In
most
countries,
economic
activity
is
spatially concentrated.
While some of this concentration
is
explained by the presence of natural advantages that constrain specific productions to specific locations,
Ellison and Glaeser (1999) and others argue that natural advantages alone cannot account for the observed
degree of agglomeration. Spatial concentration
particularly remarkable for industries that
is
traded goods, because the areas where economic activity
nationally
characterized by high costs of labor and land.
that this concentration of
economic
activity
by firms when they locate near other
Since
may be
at least
concentrated are typically
Marshall (1890), economists have speculated
explained by cost or productivity advantages enjoyed
The
firms.
is
produce
list
of potential sources of these agglomerations
advantages includes: cheaper and faster supply of intermediate goods and services; proximity to workers
or
consumers; better quality of the worker-firm match
thicker
in
labor markets;
lower risk of
unemployment
for workers and lower risk of unfilled vacancies for firms following idiosyncratic shocks;
and knowledge
spillovers.
The possibihty of documenting
tantalizing,
because
it
that productivity
could provide insights into a series of important questions.
produce nationally traded goods willing to locate
characterized by extraordinary production costs?
historical
development?
Beside
spillovers has
its
advantages through agglomeration are real
Why
in cities like
In general,
New
why do
cities exist
and what explains
their
do income differences persist across regions and countries?
obvious interest for urban and growth economists, the existence of agglomeration
tremendous practical relevance. Increasingly,
for these incentives
govenmients compete by offering
local
The main economic
rationale
depends on whether the attraction of new businesses generates some form of
agglomeration externalities. In the absence of positive externalities,
money
are firms that
York, San Francisco or London,
substantial subsidies to industrial plants to locate within their jurisdictions.
taxpayer
Why
is
for subsidies based
it
is
difficult to justify the
use of
on efficiency grounds. The optimal magnitude of these incentives
depends on the magnitude of agglomerations
spillovers, if they exist.
^
Despite their enormous theoretical and practical relevance, the existence and exact magnitude of
agglomeration spillovers are considered open questions by many.
approaches for testing for spillovers.
country.
The
These "dartboard"
The
first tests
To
date, there are
two primary
for an unequal distribution of firms across the
style tests reveal that firms are spread
unevenly across the country and that
enormous, and can not be fully summarized here. Examples include, but are not
1991b), Henderson (2001a, 2001b, 2003), Davis and Henderson (2004),
Davis and Weinstein (2002), Henderson and Black (1999), Rosenthal and Stange (2001, 2004), Duranton and Puga
(2004), Audretsch and Feldman, (1996, 2004), Moretti (2004a, 2004b, 2004c), Dumais, Ellison and Glaeser (2002),
literature
on
this topic is
limited to, Lucas (1988),
Glaeser
^
We
( 1
Krugman (1991a,
999), Ottaviano and Thisse (2004).
discuss in
more
detail the policy implications
of local subsidies in Greenstone and Moretti (2004). See also
Card, Hallock, and Moretti (2007) and Glaeser (2001).
coagglomeration rates are higher between industries that are economically similar (Ellison, Glaeser and
This approach
Kerr, 2007).
based on equilibrium location decisions and does not provide a direct
is
measure of spillovers.
The second approach uses micro data
higher
when
to assess
A
similar firms are located nearby.
whether firms'
notable example
is
total factor productivity
(TFP)
is
Henderson (2003), which estimates
plant level production functions for machinery and high-tech industries as a function of the scale of other
same and
plants in the
different industries.^
on where
location decisions
their profits will
advantage, or other cost shifters.
this
The challenge
A
for both approaches
be highest, and
this
is
that firms
base
could be due to spillovers, natural
causal estimate of the magnitude of spillovers requires a solution to
problem of identification.
This paper
tests for
and quantifies agglomeration spillovers by estimating
incumbent manufacturing plants changes when a new, large plant opens
augmented Cobb-Douglass production functions
plant, using plant-level data
made
to
maximize
at the
that allow
chosen county
is
that affect the
time of opening and in future periods.
determinants of incumbent plants' TFP.
is
to rely
identify a valid counterfactual for
We
county.
estimate
depend on the presence of the new
from an average or randomly
Valid estimates of the plant opening's
where the plant decided
These determinants are
likely to include factors
on the reported location rankings of profit-maximizing firms
what would have happened
which includes a regular feature
where
to locate.
When
titled
to
incumbent plants' TFP
come from
in
2 or 3
the corporate real estate journal
"MiUion Dollar Plants"
that describes
how
a large
firms are considering where to open a large plant, they typically
finalists are selected.
plant ultimately chose
(i.e.,
losers are counties that
The
The "Million Dollar
the 'winner'), as well as the one or
is
that the
sites,
Plants" articles report the county that the
two runner-up counties
have survived a long selection process, but narrowly
identifying assumption
to
winning
begin by considering dozens of possible locations. They subsequently narrow the Hst to roughly 10
among which
is
identical to the county
is
counties in the absence of the plant opening. These rankings
plant decided
the productivity of
new plant's TFP.
This paper's solution
Site Selection,
to
in their
likely to differ substantially
spillover effect require the identification of a county that
to locate in the
TFP
how
from the Annual Survey of Manufacturers. Because the location decision
profits, the
chosen county both
The
their
incumbent plants
in
(i.e.,
lost the
the 'losers').
competition.
the losing county
form a valid
counterfactual for the incumbents in the winning counties, after conditioning on differences in preexisting trends, plant fixed effects, industry-by-year fixed effects, and other control variables.
to the rest
*
of the country, winning counties have higher rates of growth
in
income, population, and labor
Moretti (2004b) takes a similar approach to estimate agglomeration externalities generated
spillovers.
Compared
by human
capital
force participation.
But compared
to losing counties in the years before
tiie
opening of the
new
plant,
consistent with both our
winning counties have similar trends
in
most economic
presumption that the average county
is
not a credible counterfactual and our identifying assumption that
variables. This finding
is
the losers form a valid counterfactual for the winners.
We
first
measure the
effect of the
new
(TFP) of all incumbent manufacturing plants
we
for
incumbent plants
supports the validity of the identifying assumption. After the
counties experienced a sharp relative increase in TFP.
with a
12%
in
MDP
Five years
incumbent plants' TPP.""
relative increase in
(MDP) on
total factor productivity
winning counties. In the 7 years before the
in
TFP
find statistically equivalent trends in
Million Dollar Plant
opened, incumbent plants in winning
later, the
This effect
is
MDP
some
light
We
is
associated
and
$430 million higher
is
We
interpret this finding
by increased agglomeration.
of the existence of agglomeration spillovers,
in favor
on the possible mechanisms.
opening
statistically significant
5 years later (relative to incumbents in losing counties), holding constant inputs.
Having found evidence
opened,
winning and losing counties, which
economically substantial; on average incumbent plants' output in winning counties
as evidence of meaningful productivity spillovers generated
MDP
we
then try to shed
follow Moretti (2004b) and Ellison, Glaeser, and Kerr (2007)
and investigate whether the magnitude of the spillovers depends on economic linkages between the
incumbent plant and the
MDP.
economically
the
linked
to
Specifically,
MDP
we
experience
test
whether incumbents that are geographically and
larger
geographically close but economically distant from the
spillovers,
MDP. We
relative
to
incumbents
are
that
use several measures of economic links
including input and output flows, measures of the degree of sharing of labor pools, and measures of
technological linkages.^
We
the
MDP
find that spillovers are larger for
industry.
A
incumbent plants
one standard deviation increase
with a 7 percentage point increase
in
in industries that share
our measure of worker transitions
the magnitude of the spillover.
in
worker flows with
little
Surprisingly,
we
support for the importance of input and output flows in determining the magnitude of the
some support
spillover.
Overall, this evidence provides
that share
workers and use similar technologies.
To guide
model
associated
Similarly, the measures of
technological linkages indicate statistically meaningful increases in the spillover effect.
find
is
the analysis and interpret the results,
that incorporates spillovers
for the notion that spillovers occur
we
set out a straightforward
between firms
Roback (1982)
style
between producers and derives an equilibrium allocation of finns and
workers across locations. In the model, the entry of a new firm produces spillovers. This leads new firms
''
Notably, naive estimates that control for observables but do not use the
MDP
research design find negative
productivity effects.
'
We
are deeply indebted to
Glenn
Ellison,
these measures of economic distance.
Edward Glaeser, and William Kerr
for providing their data for five of
who
are interested in gaining access to the spillover to enter.
This process of entry leads to competition
for inputs so that incumbent firms face higher prices for labor, land,
and other
In the model,
local inputs.
firms produce nationally traded goods, so they cannot raise the price of their output in response to the
Thus, the long run equilibrium
higher input prices.
due
we
find increases in quality-adjusted
consistent with the
is
documented increase
a response to the spillovers).
in
if
higher input prices
First,
we
find positive net entry in
there are sufficiently large positive spillovers.
wages following
economic
the value of the increase in output
to the
data support two predictions derived from this model.
winning counties, which the model predicts will occur
Second,
when
obtained
of production due
to spillovers is equal to the increased costs
The
is
MDP
acfivity in the
openings.
These higher wages are
winning counties (which presumably
Furthermore, the higher wages support the model's prediction that the
productivity gains from agglomeration do not necessarily translate into higher profits for incumbents in
the long run.
The remainder of the paper
is
discusses the identification strategy.
econometric model.
Sections
V
organized as follows. Section
Section
III
I
presents a simple model. Section
presents the data sources.
IV presents
Section
and VI presents the empirical findings and interpret them.
II
the
Section VII
concludes.
I.
We
are
Theories of Agglomeration and Theoretical
interested in identifying
how
the opening of a
productivity, profits, and input use of existing plants in the
theories of agglomeration.
We
Framework
new
plant in a county affects the
same county.
We
begin by reviewing the
then present a simple theoretical fi-amework that guides the subsequent
empirical exercise and aids in interpreting the results.
A. Theories of Agglomeration
Economic
activity is geographically concentrated (Ellison
and Glaeser, 1997).
forces that can explain such agglomeration of economic activity? Here
we summarize
What
are the
five possible
reasons for agglomeration, and briefly discuss what each of them implies for the relationship between
productivity and the density of economic activity.
(1) First,
it is
possible that fimis (and workers) are attracted to areas with a high concentration of
other firms (and other workers) by the size of the labor market
why
larger labor markets
search frictions,
will
if jobs
may be
attractive.
First, a thick
.
There are
labor market
is
at least
two
different reasons
beneficial in the presence of
and workers are heterogeneous. In the presence of frictions, a worker-firm match
be on average more productive
in areas
where there are many firms offering jobs and many workers
looking for jobs.
Alternatively,
it
is
possible that large labor markets are
more
desirable because they provide
insurance against idiosyncratic shocks, either on the firm side or on the worker side
moving
is
costly for workers
and firms are subject
to idiosyncratic
(Krugman 1991a).
and unpredictable demand shocks
If
that
lead to lay-offs, workers will prefer to be in areas with thick labor markets to reduce the probability of
being unemployed. Similarly,
if finding
new workers
is
costly, firms will prefer to
be
in areas
with thick
labor markets to reduce the probability of having unfilled vacancies.^
These two hypotheses have different implications
economic
matches,
and productivity.
activity
we
If the size
for the relationship
between concentration of
of the labor market leads only
to better
worker-firm
should see that firms located in denser areas are more productive than otherwise identical
firms located in less dense areas. The exact form of this productivity gain depends on the shape of the
production function.*
On
the other hand, if the only effect of thickness in labor market
for workers
and a lower
risk
is
a
lower risk of unemployment
of unfilled vacancies for firms, there should not be differences
productivity between dense and less dense areas.
in
While productivity would not vary, wages would vary
across areas depending on the thickness of the labor market, although the exact effect of density on wages
a priori ambiguous.' This change
is
capital used. Unlike the case of
in relative factor prices will
change the relative amount of labor and
improved matching described above, the production function does not
change: for the same set of labor and capital inputs, the output of firms in denser areas should be similar
to the output
(2)
of firms
A
in less
dense areas.
second reason
transportation costs
why
the concentration of
economic
activity
may be
beneficial has to do with
(Krugman 1991a and 1991b, Glaeser and Kohlhase, 2003). Because
in this
paper
we
focus on firms that produce nationally traded goods, transportation costs of finished products are unlikely
to
^
be the relevant cost
For a related point
A third
in this
paper's setting. Only a small fraction of buyers of the final product
in a different context, see
is
likely
Petrongolo and Pissarides (2005).
that arise because of endogenous capital accumulation. For
example, in Acemoglu (1996), plants have more capital and better technology in areas where the number of skilled
workers is larger. If firms and workers find each other via random matching and breaking the match is costly,
externalities will arise naturally even without leaming or technological externalities. The intuition is simple. The
alternative hypothesis has to
do with spillovers
amount of skills depends on the amount of physical capital a worker expects to use. The privately
optimal amount of physical capital depends on the number of skilled workers. If the number of skilled workers in a
privately optimal
city increases, firms in that city, expecting to
employ these workers,
some of the workers end up working with more
physical capital and earn
For example,
denser areas.
productivity.
it is
It
is
will invest more.
more than
Because search
is
costly,
similar workers in other cities.
possible that the productivities of both capital and labor benefit from the improved match in
also possible that the
improved match caused by
a larger labor
market benefits only labor
This has different implications for the relative use of labor and capital, but
total factor productivity
be higher regardless.
' Its sign depends on the relative magnitude
of the compensating differentia! that workers are willing to pay for
lower risk of unemployment (generated by an increase in labor supply in denser areas) and the cost savings that
firms experience due to lower risk of unfilled vacancies (generated by an increase in labor demand in denser areas).
will
to
be located
in the
same area
as our manufacturing plants.
The
relevant costs are the transportation costs
Firms located in denser areas are likely to
of suppliers of local services and local intermediate goods.
enjoy cheaper and faster delivery of local services and local intermediate goods. For example, a high-tech
firm that needs a specialized technician to fix a machine
cost if
it is
located in Silicon Valley than in the
Nevada
is
likely to get service
more quickly and
at
lower
desert.
This type of agglomeration spillover does not imply that the production function varies as a
function of density of economic activity: for the same set of labor and capital inputs, the output of firms in
denser areas should be similar to the output of firms
should be lower
A
(3)
knowledge
in
in less
third reason
spillovers
may
that this type
are
more
why
the concentration of
There are
.
at least
two
economic
different versions of this hypothesis.
of spillover
likely to
may be
come from
the
important
same
in
some high-tech
skills
state or
industries.
First,
economists and
through formal and informal
For example, patent citations
metropolitan area as the originating patent (Jaffe
more
efficient
flow of
new
in Silicon
et al.
Valley
ideas and ultimately causes faster innovation." Second,
also possible that proximity results in sharing of information on
first
be beneficial has to do with
generate positive production externalities across workers.'" Empirical evidence indicates
associated with a
faster
may
activity
Saxenian (1994) argues that geographical proximity of high-tech firms
1993).
costs
denser areas.
urban planners have long speculated that the sharing of knowledge and
interacdon
However, production
dense areas.
new
is
it is
technologies and therefore leads to
technology adoption. This type of social learning phenomenon applied to technology adoption was
proposed by Griliches (1958).
If density of
economic
agglomeration model
areas are
model,
is
activity results in intellectual externalities, the implication
higher productivity.
more productive than otherwise
this
In particular,
we
of
this
type of
should see that firms located in denser
identical firms located in less dense areas.
As with
the search
higher productivity could benefit both labor and capital, or only one of the two factors,
depending on the forni of the production function.
results in faster technology adoption
there should be no relationship
On
the other hand, if density of
and the price of new technologies
between productivity and density,
economic
activity only
reflects their higher productivity,
after properly controlling for quality
of
capital.
(4)
It is
possible that firms concentrate spatially not because of any technological spillover, but
because local amenities valued by workers are concentrated.
For example, skilled workers
may
prefer
'"
See, for example, Marshall (1890), Lucas (1988), Jovanovic and Rob (1989), Grossman and Helpman (1991),
Saxenian (1994), Glaeser (1999), and Moretti (2004a, 2004b and 2004c).
The entry decisions of new biotechnology firms in a city depend on the stock of outstanding scientists there, as
'
'
measured by the number of relevant academic publications (Zucker et al, 1998). Moretti (2004b) finds stronger
human capital spillovers between pairs of firms in the same city that are economically or technologically closer.
certain amenities
more than unskilled workers. This would
where these amenities
to concentrate in locations
workers
in productivity
any difference
differences in
wages
compensating
(5) Finally, spatial concentration
advantages. For example, the
states
have the most accessible
because that
is
oil
fields.
may
may be
skilled
should not see
we should
see
be explained by the presence of natural
number of
Similarly, the wine industry
where good weather and suitable land
producdons, the presence of a harbor
more
differential.
concentrated in a limited
is
we
dense areas, although
less
of some industries
industry
oil
relatively
In this case,
are available.
between dense areas and
that reflect the
employ
lead firms that
are
is
because those
concentrated in California
be found. For some manufacturing
to
The presence of
important.
states
natural advantages has the
implication that firms located in areas with high concentration of similar firms are
more productive, but of
course this correlation has nothing to do with agglomeration spillovers. Since most natural advantages are
fixed over time, this explanation
is
not particularly relevant for our empirical estimates, which exploit
variation over time in agglomeration.
A
B.
Simple Model
We
technology.
paper,
we
begin by considering the case where incumbent
Later
we
are
homogenous
firms are heterogeneous.
size
in
and
Throughout the
focus on the case of factor-neutral spillovers.
(a)
We
Homogeneous Incumbents.
technology that uses labor,
is
when incumbent
consider what happens
finTis
normalized to
1.
capital,
and land
Incumbent firms choose
to
assume
that
all
incumbent firms use a production
produce a nationally traded good whose price
their
amount of labor, L,
capital,
is
fixed and
K, and land, T, to maximize
the following expression:
Maxu,K,T{
where w,
r
and q are input prices and
A
is
fIA,L,K,T]-wL-rK-qT}
a productivity shifter (TFP). Specifically,
that affect the productivity of labor, capital,
spillovers, if they exist.
factors
all
we
allow
A
depend
to
activity in an area:
A = A(N)
(1)
where
includes
and land equally, such as technology and agglomeration
In particular, to explicitly allow for agglomeration effects,
on the density of economic
A
N
is
the
number of firms
that are active in a county,
factor-neutral agglomeration spillovers as the case
where
A
and
all
counties have equal size.
We
define
increases in N:
5A/5N>0
If instead
(5A /5N) =0, we say
Let
L
that there are
no factor-neutral agglomeration
spillovers.
(w,r,q) be the optimal level of labor inputs, given the prevailing
cost of industrial land.
wage, cost of
capital,
and
Similarly, let K*(w,r,q) and T*(w,r,q) be the optimal level of capital and land.
In equilibrium, L', K*,
respectively.
factors
is
equal to
We
that capital is internationally traded, so
However, we allow
supply conditions.
county
we
In particular,
conditions.
set so that the
marginal product of each of the three
price.
its
assume
and T* are
its
price does not
for the price of labor
and land
depend on
local
depend on
to
demand
or
economic
local
allow the supply of labor and land to be less than infinitely elastic
at the
level.
We
standard
upward sloping labor supply curve
attribute the
Roback (1982) model, we assume
housing, and that
to the existence
that workers' indirect utility
of moving
Roback (1982) model we allow
locations, but unlike the standard
for
ignore labor supply decisions within a given location and assume that
in the
Workers are mobile across
moving
all
Like
depends on wages and cost of
equilibrium workers are indifferent across locations.
in
costs.
costs.
For simplicity,
we
amount
residents provide a fixed
of labor.
To
plant.
illustrate this,
In particular,
worker
the marginal
new
m
is
in
consider that there are
such
c,
links the
level.
we
For example,
regulations.
indifferent
start rising,
is
it
When
county c and staying.
and some workers find
it
optimal to
therefore the slope of the labor supply ftinction,
active in a county to the local nominal
possible that the supply of land
Alternatively,
may
number of firms, N,
is
not be completely fixed, but
Irrespective of the reason,
that links the
to
move
to
c.
a
The
depend on the shape
wage
w.
level,
allow the supply of industrial land to be less than infinitely elastic
it
new
county c before the opening of the
between moving
has already been developed, so that the marginal land
develop.
in
Let w(N) be the inverse of the reduced-form labor supply function that
fiinction.
number of firms, N,
Similarly,
is
wages there
number of workers who move, and
of the mobility cost
workers
given the distribution of wages and the housing costs across localities,
that,
another county
plant opens in county
m
we
call
is
at the
county
fixed because of geography or land-use
it
is
possible that the best industrial land
of decreasing quality or more expensive to
q(N) the (inverse of the) reduced form land supply function
to the price
of land,
q.
We
can therefore write the equilibrium level of
profits, n*, as
n* =
-
f[
A(N), L*(w(N),
w(N) L'(w(N),
where we now make
r,
r,
q(N)), K*(w(N),
q(N))
- r K'(w(N),
explicit the fact that
q(N)), T*(w(N),
r,
r,
q(N))
r,
q(N))
- q(N) T*(w(N),
r,
]
-
q(N))
TFP, wages, and land prices depend on
number of firms
the
active in a county.
Consider the
total derivative
of incumbents' profits with respect to a change
in the
number of
firms:
(2)
dn7dN =
(6f/8A 8A/5N)
+ 5w/5N
{
[5L*/5w (5f/§L - w)
-
L*]
+ [5K*/5w (SfSK -
r)]
+ [5T*/5w
(5f/5T
-
q)]}
+ 5q/5N
{
[5L*/5q (5f/5L
and
If all firms are price takers
considerably and can be written
dn7dN =
(3)
- w)] +
[5K'/5q (5f/5K
(5f/5A 5A/5N)
represented by the
first
- [
5w/5N L' + 5q/5N
productivity of
more output using
if there are positive spillovers,
The second term,
- [
all
and 5q/5N >
N
in the local
demand
is
the
sum of two
is
is
opposite effects.
In equation (3), this effect
unambiguously
positive, because
represents the negative effect from increases in the
],
Formally, this term
an increase
for labor
and
in the level
is
negative because
elasticity
utihzations,
affected
its
compete
wages and land prices has two
increase in
change
to
of economic activity
it
on incumbents.
mechanically raises production costs.
use of the different production inputs.
by an increase
in
N, the firm
is
likely to
contrast, the effect
productivity of
all
Second,
On
costly for
entrant.
The
given level of input
leads the firm to re-optimize and to
is
not
end up using more capital than before:
=>0.
on the use of labor and land
factors increases.
it
First, for a
new
is
In particular, given that the price of capital
5K75N
By
county
in the
land.
for locally scare resources with the
effects
we
of the supply of
Unlike the beneficial effect of agglomeration spillovers, the increase in factor prices
now have
it
>0.
(The magnitudes depend on the
0.
N
labor and land.) Intuitively, an increase in
incumbent firms, because they
}
same amount of inputs. Formally, 5f /5A >0
cost of production, specifically the prices of labor and land.
and therefore an increase
in
factors increases.
the
5A/5N
5w/5N L + 5q/5N T
have assumed that 5w/5N >
T*]
T'].
term, (5f /5A 5A/5N). This effect
allows an incumbent firm to produce
by assumption and,
-
q)
as:
First, if there are positive spillovers, the
is
+ [5TV5q (5f75T -
r)]
factors are paid their marginal product, equation (2) simplifies
all
Equation (3) makes clear that the effect of an increase
on TFP
-
is
On
ambiguous.
the other hand, the price of labor
the one hand, the
and land might increase. The
net effect depends on the magnitude of the factor price increases, as well as on the exact shape of the
production function
(i.e.,
the strength
of technological complementarities between
labor, capital,
and
land).
It is
spillovers.
when
instructive to apply these derivations to the case of a
We
initially
MDP
consider the case where for incumbent firms dllVdN
the agglomeration spillover
is
the
MDP's
alternative case
is
that dll
/dN >
opening exceeds the increase
activity.
0,
0.
This would occur
due to the
one might expect
exit
MDP's
exit.'^
which occurs when the magnitude of the
in factor prices
Similarly, if the spillovers are zero or negative,
economic
<
smaller than the increase in production costs. In this case, the
opening would not lead to any entry and could cause some existing firms to
The
opening that causes positive
MDP's demand
spillover due to
for local inputs.
In
of incumbent firms and a reduction in local
the short run, profits will be positive for
new
the price of local factors, like land and possibly labor,
In the long run, there is an equilibrium
county where the
levels
new
of productivity are likely
may occur due
This
to
Since the amount of land
locales.
be capitalized into land prices.
to
moving
bid up.
is
such that firms and workers are indifferent between the
opened and other
plant has
also likely that
and
same
From
in other locations.
new
profits in the county with the
a practical perspective,
wages
will increase.
in factor prices
mean
plant (even in the presence of the spillovers)
impossible to
is
it
fixed, the higher
is
These adjustments make workers indifferent
'^
costs as noted above.
It is
between the county with the new plant and other counties. Similarly, the changes
that firms earn the
over time as
profits will disappear
These positive
entrants.
know when
the short run ends and
the long run begins.
There are two empirical predictions that apply when there are positive spillovers.
magnitude of the spillovers
is
large enough,
new
firms will enter the
have had sufficient time to respond. The second prediction
(b)
MDP and the new entrants bid for these inputs.
Heterogeneous Incumbents.
homogeneous? Consider
the case
extent from the type of workers
new
entrant
that the prices
entrants
of locally traded inputs will
if the
population of incumbent firms
non-
Assume
and low-tech.
firms: high-tech
is
of workers employed by high-tech firms, Lh, differs to some
employed by low-tech
a high-tech firm.
is
What happens
is
new
to the
'''
where there are two types of
that for technological reasons, the type
that the
county to gain access
This prediction of increased economic activity holds at any point after potential
spillover.
rise as the
MDP's
First, if the
firms, Ll, although there
some
is
Assume
overlap.
Equations (4) and (5) characterize the effect of the
new
high-tech
firm on high-tech and low-tech incumbents:
dnH'/dNH = (5fH/5AH 5Ah/5Nh)
(4)
" Even with zero moving
increases as workers will
-
[
5wh/5Nh L*h + 5q/5NH
costs and an infinitely elastic supply of labor,
demand higher wages
as
wages
T']
will increase if there are land price
compensation for the higher land rents for
their
homes (Roback
1982).
'*
This paper focuses on the case where the productivity benefits of the agglomeration spillovers are distributed
equally across
that
all factors.
What happens when agglomeration
spillovers are factor biased?
agglomeration spillovers raise the productivity of labor, but not the productivity of
technology
is
f[A, L, K, T], but
number of physical workers and
now L
6
is
represents units of effective labor.
a productivity shifter.
We
In particular,
Assume,
capital.
for
example,
Like before, the
L = GH, where H
is
the
define factor-biased agglomeration spillover as the
case where the productivity shifter 9 depends positively on the density of the economic activity in the county 6
=
0(N) and 50/5N >0. If 5A /5N =0 and factors are paid their marginal product, then the effect of an increase in the
density of the economic activity in a county on incumbent firms simplifies to dTlVdN = (5f /5H 59/5N) H - [ 6w/5N
H
+ 5q/5N
T
].
The
effect
on profits can be decomposed
in
two
parts.
The
first
term represents the increased
/5H >0), times the magnitude of the
agglomeration spillover (6G/5N >
by definition), times the number of workers. The second term is the same as in
equation (3), and represents the increase in the costs of locally supplied inputs. The increase in N changes the
optimal use of the production inputs. Labor is now more productive, and its equilibrium use increases; 5L /5N <=0.
Land is equally productive but its price increases. Its equilibrium use declines: 5T /5N <=0. Neither the price nor
the productivity of capital is affected by an increase in N. Its equilibrium use depends on technology. Specifically,
it depends on the elasticity of substitution between labor and capital.
productivity of labor.
It is
the product of the sensitivity of output to labor (6f
10
dOL'/dNH = (5f l/5Al 5Al/5Nh)
(5)
It is
-
[
5wl/5Nh
L\ +
5q/5NH T*]
plausible to expect that the beneficial effect of agglomeration spillovers generated by a
tech entrant
is
new
high-
larger for high-tech firms than for low-tech firms:
(5fH/5AH 5Ah/5Nh) >(5fL/5AL 5Al/5Nh)
(5')
At the same time, one might expect
incumbents, given that they are
that the increase in labor costs
now competing
for
is
also higher for the high-tech
workers with an additional high-tech firm:
5wh/6Nh > 5wl/5Nh
The
effect
on land prices should be similar for both firm types, since the assumption of a single land
market seems reasonable.
There are two takeaways here.
First,
it
may be
reasonable to expect larger spillovers to firms that are
economically "close" to the new plant. Second the relative impact of the new plant on profits
is
unclear,
because the economically "closer" plants are likely to have bigger spillovers and larger increases
in
production costs.
C. Empirical Predictions
The simple
theoretical fi-amework
above generates four predictions
that
we
bring to the data.
Specifically if there are positive spillovers, then:
1
the opening of a
2.
the increase in
3.
the density of
new
plant will increase the
TFP may be
economic
TFP of incumbent plants.
new
larger for firms that are economically "closer" to the
activity in the
county will increase as firms
move
in to
plant.
gain access to the
positive spillovers (if the spillovers are large enough).
4.
the price of factors of production that are traded locally will increase.
price of quality-adjusted labor,
which
is
We
test for
changes
in the
arguably the most important locally supplied factor of
production for manufacturing establishments.
II.
Plant Location Decisions and Research Design
In testing the four empirical predictions outlined above, the
fact that firms
where
do not choose
their expectation
their location
the
of the present discounted value of the stream of future profits
availability
of workers with particular
fi^equently unobserved. Further, they are likely to
Therefore, a naive comparison of the
opening with the
TFP of incumbents
is
the
randomly. Firms are profit maximizers and choose to locate
present value varies tremendously across locations, depending on
infrastructure,
main econometric challenge
many
skills,
TFP of incumbents
in counties that
11
This net
factors, including transportation
subsidies,
be correlated with the
is greatest.
TFP
etc.
These factors are
of existing plants.
in counties that
experience a plant
do not experience a plant opening
is likely to
yield
TFP of
biased estimates of productivity spillovers. Credible estimates of the impact of a plant opening on
incumbent plants require the identification of a location
decided
to locate in the determinants
BMW picked the location for one of
how
when
demonstrate the empirical difficulties that arise
is to
on the TFP of incumbent
circumvent these
similar to the location
is
where the plant
of incumbent plants' TFP.
This section provides a case study for
intent
that
Further,
plants.
illustrates
it
its
plants.
how
informally
later,
may
our research design
difficulties.
announced
months
The
estimating the effect of plant openings
After overseeing a worldwide competition and considering 250 potential sites for
BMW
'^
in
BMW
1991 that they had narrowed the
announced
two
that the
South Carolina, and Omaha, Nebraska.
list
of potential candidates
finalists in the
In 1992,
BMW
to
new
its
plant,
20 counties.
Six
competition were Greenville-Spartanburg,
announced
that they
would
site the
plant in
Greenville-Spartanburg and that they would receive a package of incentives worth approximately $115
million funded by the state and local governments.
Why
The
was
first
BMW
did
BMW's
Two
choose Greenville-Spartanburg?
expected
fiiture
of production
costs
factors
were important
250
to
BMW,
BMW's
production technology.
made Greenville-Spartanburg more
attractive than the other
considered were: low union density, a supply of qualified workers, the numerous global
sites initially
finns,
including 58
including
the characteristics that
air, rail,
German companies,
in
the area;
highway, and port access; and access
For our purposes, the important point
potential source of unobserved heterogeneity.
to
the high
quality transportation
infrastructure,
key local services.
to note here is that these
county characteristics are a
While these characteristics are well documented
BMW case, they are generally unknown and unobserved.
TFP
which are
Greenville-Spartanburg,
in
presumably a function of the county's expected supply of inputs and
According
in this decision.
of existing plants, a standard regression that compares Greenville-Spartanburg with the other 3,000
regression will overestimate the effect of plant openings on outcomes
more
improving transportation
attractive characteristics (e.g.,
growth.
A
TFP encouraged new
A
second important factor
expected economic benefits from
plant
is
in
standard
for example, counties that
iniirastructure) tend to
if,
have
faster
have
TFP
incumbent
for example,
entrants (e.g., cheaper availability of local inputs).
in
BMW's
Presumably Greenville-Spartanburg was willing
" This
if,
Conversely, a standard regression would underestimate the effect
plants' declining
we
growth of
If these characteristics also affect the
United States counties will yield biased estimates of the effect of the plant opening.
it
in the
BMW
to
decision
provide
was
the
value of the subsidy
it
received.
BMW with $115 million in subsidies because
presence. According to local officials, the facility's ex-ante
Greenstone and Moretti's (2004) set of 82 MDP plants. Due to Census confidentiality
this plant is part of this paper's analysis.
cannot report whether
12
restrictions,
expected five-year economic impact on the region was $2 billion. As a part of
was expected
to create
new
agglomeration economies. (The empirical section
is
likely to
force,
plants or the expansion of existing plants caused by
tests
whether
this is
indeed the case on average). Thus,
is
relevant for this paper's identification strategy, because the magnitude of the
from a particular plant depends on the level and grovi^h of a county's
and a
of other unobserved variables. For
series
billion, the plant
be a function of the expected gains from agglomeration for the county.'*
This possibility
spillover
$2
2,000 jobs directly and another 2,000 jobs indirectly. In principle, these 2,000
additional jobs could reflect the entry of
the subsidy
this
industrial structure, labor
this reason, the factors that
determine the total size
of the potential spillover (and presumably the size of the subsidy) represent a second potential source of
unobserved heterogeneity.
If this
unobserved heterogeneity
is
correlated with incumbent plants' TFP,
standard regression equations will be misspecified due to omitted variables, just as described above.
In order to
make
valid inferences in the presence of the heterogeneity associated with the plant's
expected local production costs and the county's value of attracting the plant, knowledge of the exact
form of the selection rule that determines
example demonstrates, the two
plants' location decisions is generally necessary.
As
factors that determine plant location decisions are generally
the
BMW
unknown
to
researchers and, in the rare cases where they are known, are difficult to measure. Thus, the effect of a
plant opening on incumbents'
TFP
is
very likely to be confounded by differences in factors that determine
the plants' profitability at the chosen location.
As
a solution to this identification problem,
maximizing firms
winning counties
to identify a valid counterfactual for
in the
absence of the plant opening.
the corporate real estate journal Site Selection.
"Million Dollar Plants" that describes
how
report the county that the plant chose
counties
(i.e.,
the "losers")."
fi"om an initial
The
model
the
BMW)
on the reported location rankings of
what would have happened
We
to
profit-
incumbent plants
in
implement the research design using data from
Each issue of
this journal includes
where
to locate.
an article titled the
These
articles
always
the 'winner'), and usually report the runner-up county or
BMW case study indicates, the winner and losers are usually chosen
sites that in
many
cases
number more than
have survived a long selection process, but narrowly
fact that business organizations
a hundred.'^
The
lost the competition.
such as the Chambers of Commerce support these incentive plans (as was the
suggests that incumbent finns expect such increases. Greenstone and Moretti (2004) present a
determine local governments' bids for these plants and whether successfully
be welfare increasing or decreasing for the county.
that describes the factors that
attracting a plant will
In
As
rely
a large plant decided
(i.e.,
sample of "semi-finalisf
losers are counties that
case with
we
some
instances the "Million Dollar Plants" articles do not identify the runner-up county.
did a Lexis/Nexis search for other articles discussing the plant opening and in 4 cases,
were able
to identify the losing counties.
when
was unavailable
this
counties
is
For these cases,
the original 82,
we
we
The Lexis/Nexis searches were also used to identify the plant's industry
Comprehensive data on the subsidy offered by winning and losing
in Site Selection.
unavailable in the Site Selection articles.
The names of the
among
semi-finalists are rarely reported.
13
We
in the
use the losers to identify what would have happened to the productivity of incumbent plants
winning county
we assume
we
In practice,
adjust for covariates so our identifying assumption
subsequent analysis provides evidence that supports the validity of
assumption
that
incumbent firms'
trended identically in the absence of the plant opening in winning and losing counties
TFP would have
within a case.
absence of the plant opening. Specifically,
in the
hold,
fail to
we presume
adjustment to compare the
TFP of incumbent plants
is
in counties with
on observable
States counties or to using a matching procedure based
III.
approach
that this pairwise
Summary
Data Sources and
this
more
new
is
The
weaker.
Even
assumption.
if this
reliable than using regression
plants to the other 3,000 United
variables.
Statistics
A. Data Sources
The "Million Dollar
Plants" articles typically reveal the county where the
new
firm (the "Million
Dollar Plant") ultimately chose to locate (the "wirming county"), as well as the one or two runner-up
The
counties (the "losing counties").
government
focus on large plants that are the target of local
articles tend to
subsidies. Important limitations of these articles are that the
by the winning counties
is
in
many
cases unobserved and that the bid
magnitude of the subsidy offered
is
almost always unobserved for
losing counties.
We
which
identify the Million Dollar Plants in the Standard Statistical Establishment List
the
is
Census Bureau's "most complete,
establishment"""
- and matched
the plants to the
we
identified
47 genuine and useable
genuine and useable
a
new
MDP
Annual Survey of Manufactures {ASAf) and
be located
in the
county indicated
winning and losing counties
"
Propensity score matching
relative to our approach
observables.
score
is
As
it
is its
that
Census
openings
in the
manufacturing data. In order to qualify as a
owned by
in the
the
openings
years before and 3 years after the publication of the
to
business
MDP
MDP manufacturing opening, we
plant in the manufacturing sector,
and consistent data for U.S.
the 82
Of
of Manufactures (CM) from 1973-1998."'
current,
(SSEL) -
MDP
in
Greenstone and Moretti (2004),
imposed the following
the reported firm, appearing in the
MDP
article;
SSEL
article; 2) the plant identified in the
and 3) there had
were there for each of the previous
to
had
criterion: 1) there
Among
be
within 2
SSEL had
be incumbent plants
8 years.
to
in
the 35
both
MDP
an alternative approach (Rosenbaum and Rubin 1983). Its principal shortcoming
assumption that the treatment (i.e., winner status) is "ignorable" conditional on the
is
should be clear from the example, adjustment for observable variables through the propensity
unlikely to be sufficient.
^°
The SSEL is confidential and was accessed in a Census Data Research Center. The SSEL is updated continuously
and incorporates data from all Census Bureau economic and agriculture censuses and current business surveys,
quarterly and annual Federal income and payroll tax records, and other Departmental and Federal statistics and
administrative records programs.
"'
The sample is cut at 1998 because sampling methods in the
because of minor known inconsistencies with the 1972 CM.
14
ASM changed
for 1999.
The sample begins
in
1973
openings that did not qualify, roughly 20 were outside of the manufacturing sector.
exact
number because of the Census Bureau's
To
ASM and CM contain
4-digit
SIC code and county of
of the opening.
In this period, the
establishment that
sampled with
was
ASM in
in the
Additionally,
we
ASM
that
we
and these
use includes plants that
the 8 years preceding the year of the plant opening plus the year
drop observations on plants that have the same owner as the
company with manufacturing shipments exceeding $500
part of a
MDP
size.
Any
million
was
sampling scheme was positively related to firm and plant
were establishments with 250 or more employees.
certainty, as
There are a few noteworthy features of this sample of potentially affected
on existing plants allows for a
test
plants.
First,
the focus
of spillovers on a fixed sample of pre-existing plants, which eliminates
concerns related to the endogenous opening of
to
ASM
use the
location are also reported
The sample
possible to follow individual plants over time.""
were continuously present
we
Importantly, the manufacturing data contain a unique plant identifier,
play a key role in the analysis.
plants.
confidentiality rules).
information on employment, capital stock, total value of shipments,
The
plant age, and firm identifiers.
it
cannot report the
obtain information on incumbent establishments in winner and loser counties,
and CM. The
making
(We
form a genuine panel of manufacturing
new
plants.
plants and compositional bias.
Third, a disadvantage
is
Second,
it is
that the results
possible
may
not be
externally valid to smaller incumbent plants that are not sampled with certainty throughout this period.
Nevertheless,
it
shipments in the
is
relevant that this sample of plants accounts for
last
effects are larger in industries that are
distance.
of county-wide manufacturing
CM before the MDP opening.
Besides testing for an average spillover effect,
economic
54%
We
we
also test whether the estimated agglomeration
more closely linked
to the
MDP
based on some measure of
use six measures of economic distance in three categories.
supplier and customer linkages,
we
come from each
and the fraction of each industry's outputs sold
3-digit industry
rotation
move
we
file.
In particular,
manufacturers that are
Second, to measure the frequency of worker mobility between
we measure
each 2-digit industry.
fraction of patents
We
to
use data on labor market transitions from the Current Population Survey {CPS) outgoing
to firms in
industry.
measure
use data on the fraction of each industry's manufactured inputs that
purchased by each 3-digit industry.
industries,
First, to
manufactured
also use data on the
the fi-action of separating workers
from each
Third, to measure technological proximity,
in a 3-digit industry that cite patents
amount of
R&D
2-digit industry that
we
manufactured
expenditure in a 3-digit industry that
use data on the
in
is
each 3-digit
used in other
3-digit industries.
Finally,
one further data issue merits
of the plant opening. The
^^
See the appendix
in
first is
the
MDP
attention.
articles,
We
have two sources of information on the date
which often
are written
when ground
is
broken on the
Davis, Haltiwanger, and Schuh (1996) for a more thorough description of the^5'Mand
15
CM.
plant but other times are written wiien the location decision
second source
known
is
SSEL, which
the
made
SSEL
However,
year of operation.
it is
after their opening.
uncertainty about the date of the plant's opening.
is
The
or the plant begins operations.
in principle reports the plant's first
that plants occasionally enter the
Thus, there
is
Further, the date at
which
the
plant could affect the operations of existing plants depends on the channel for any agglomeration
economies.
If the
agglomeration economies are a consequence of supplier relationships, then they could
occur as soon as the plant
is
For example, the
announced.
plants and provide suggestions on operations.
not be evident until the plant
is
on when the new plant could
we emphasize
results using the earlier
that the
new
plant appears in the
the opening date.
Summary
B.
Table
The
labor.
affect other plants.
may be
driven
In this case, agglomeration economies
issues, there
is
not
Rather than take an unsupportable stand,
of the year of the publication of the magazine
We
visit existing
also report separate results based
article
and the year
on using these two years as
basic findings are robust to these alternatives.
Statistics
1
presents
summary
of the analysis. As discussed
we
SSEL.
management might
Based on these data and conceptual
operating.
clear guidance
plant's
Alternatively, the agglomeration spillovers
by the labor market and therefore may depend on sharing
may
new
on the sample of plant location decisions that forms the basis
statistics
previous subsection, there are 47 manufacturing
in the
MDP
openings that
can match to plant level data. There are plants in the same 2-digit SIC industry in both winning and
losing counties in the 8 years preceding the opening for just 16 of these openings.
The
table reveals
accompanying
some
We
other facts about the plant openings."'
loser(s) associated with
refer to the
winner and
each plant opening armouncement as a "case." There are two or
more
losers in 16 of the cases, so there are a total
Some
counties appear multiple times in the sample (as either a winner or loser), and the average county in
the
sample appears a
total
The
of 1.09 times.
publication and the year the plant appears in the
-1
years,
years,
and
1
For
to 3 years.
after the plant is identified in the
of 73 losing counties along with 47 winning counties.
difference between the year of the
SSEL
is
The
article's
roughly spread evenly across the categories -2 to
clarity, positive differences refer to
SSEL.
MDP
cases where the article appears
date of the plant openings ranges fi^om the early 1980s
through the early 1990s.
The remainder of Table
their assigned
opening date
to
1
provides
summary
statistics
on the new
provide a sense of their magnitude. These
MDP
MDPs
plants five years after
are quite large:
they are
more than twice
the size of the average incumbent plant and account for roughly nine percent of the
average county's
total
'^
A
number of
output one year prior to
the statistics in Table
1
its
opening.
are reported in broad categories to
comply with
confidentiality restrictions and to avoid disclosing the identities of any individual plants.
16
the Census Bureau's
Table 2 provides summary
of these variables. In
all
statistics
on the measures of industry linkages and further descriptions
cases, the proximity
For ease of interpretation
between industries
is
increasing in the value of the variable.
subsequent regressions, these variables are normalized to have a mean of
in the
zero and a standard deviation of one.
Table 3 presents the means of county-level and plant-level variables across counties.
means
are reported for wirmers, losers, and the entire United States in columns (1), (2),
These
respectively.
^"^
In the
in the
ASM in
the 8 years preceding the assigned opening date
and the assigned
All entries in the entire United States colurim are weighted across years to produce
date.
statistics for the
only calculated
(3),
winner and loser columns, the plant-level variables are calculated among the
incumbent plants present
opening
and
MDP
year of the average
among
opening
in
our sample.
ASMior
plants that appear in the
at least
Further, the plant characteristics are
9 consecutive years.
Column
the t-statistics from a test that the entries in (1) and (2) are equal, while
of equality between columns (1) and
(3).
where there are plants within the same
characteristics are calculated
among
Columns
2-digit
(6)
through (10) repeat
SIC industry
the plants in the
same
as the
Column
(5) repeats this for a test
this exercise
MDP plant.
(4) presents
among
the cases
In these columns, the plant
2-digit industry.
This exercise provides an opportunity to assess the validity of the research design, as measured by
pre-existing
observable
county
characteristics are balanced
analysis.
and plant
To
characteristics.
among wiiming and
the
extent
that
these
observable
losing counties, this should lend credibility to the
The comparison between wiimer counties and
the rest of the United States provides an
opportunity to assess the validity of the type of analysis that would be undertaken in the absence of a
quasi-experiment.
The top panel
reports on county-level characteristics
measured
plant opening and the percentage change between 7 years and
that
compared
population
to
manufacturing.
labor
Among
statistically significant at
wiimers are compared to
different at the
5%
force
'''''
The
participation
rates
level but
and
and
growth,
evident
and a higher share of labor
this is reflected in the fact that 3
none are
at
the
1%
where there
none of them would be judged
column
It is
to
would be judged
to
in
be
These differences are substantially mitigated when the
conventional levels.
losers within the subset of cases
smaller, and
year before the opening.
the 8 variables in this panel, 6 of the 8 differences
losers,
year before the assigned
country, winning counties have higher incomes, population and
the rest of the
growth,
1
in the
level.
statistically
Notably, the raw differences between wiimers and
are plants in the
be
of the 8 variables are
same 2
digit
SIC industry are generally
statistically significant.
weighted in the following manner. Losing counties are weighted by
Losing plants are weighted by the inverse of their number per-county,
multiplied by the inverse of the number of losing counties in their case. The result is that each county (and each
plant within each county) is given equal weight within the case and then all cases are given equal weight.
losing county entries in
the inverse of their
number
(2) are
in that case.
17
The second panel
their characteristics.
On
reports
on the number of sample plants and provides information on some of
In light of our
sample selection
criteria, the
number of plants
is
of special
interest.
average, there are 18.8 plants in the winner counties and 25.6 in the loser ones (and just 8.0 in the
United States). The covariates are well balanced between plants
there are
same
no
statistically significant differences either
2-digit industry.
all)
plants or
MDP
among
in fact,
the plants within the
winner-loser research design balances
observable county-level and plant-level covariates.
TFP were
demonstrate that trends in
MDP, which
all
winning and losing counties;
^^
Overall, Table 3 has demonstrated that the
(although not
among
in
In the subsequent analysis,
we
similar in winning and losing counties prior to the opening of the
lends further credibility to this design.
Of
course, this exercise does not guarantee that
unobserved variables are balanced across winner and loser counties or
outlines our full econometric
many
their plants.
The next
model and highlights the exact assumptions necessary
section
for consistent
estimation.
IV. Econometric
Building on the model in section
I,
we
start
Model
by assuming
that
incumbent plants use the following
Cobb-Douglas technology:
l,o;
I pijt
where p references
and we allow
equipment
output.
/^pijt
plant,
that uses earlier years
Jv
pijt js.
j
I,
two
Roughly
20%
t
p,jt
year;
Y,jct is
I,
the total value of shipments; Apy,
building capital stock Ik
and the dollar value of materials Im
.1
capital stock variables are calculated with the
I
Ap,j,.
of the winners were
OH, PA, NJ,
IL,
to
In particular,
in a
we
county: Apyt
=
A(Npijt).
is
TFP;
-I, machinery and
have separate impacts on
permanent inventory method
^^
allows for agglomeration spillovers by assuming that
of the number of firms that are active
defined as MI, IN,
ivi
of the data on book values and subsequent investment.
additional heterogeneity in
"'
and
case,
Recall that equation (1) in Section
a function
pij,
hours of production II •
stocklK
In practice, the
pijt
industry,
i
total labor
capital
L
TFP
is
Here we also allow for some
generalize equation (1) by allowing for permanent
in the Rust Belt, compared to roughly 25% of the losers (where the Rust Belt is
WI, NY). Roughly 65% of the winners were in the South, compared to roughly
45%
of the losers.
For the first date available, plants' historical capital stock book values are deflated to constant dollars using BEA
data by 2-digit industry. In all periods, plants' investment is deflated to the same constant dollars using Federal
Reserve data by 3-digit industry. Changes in the capital stock are constructed by depreciating the initial deflated
capital stock using Federal
Reserve depreciation rates and adding deflated investment.
In
each year, productive
capital stock
is
defined as the average over the beginning and ending values, plus the deflated level of capital rentals.
The analysis
by Becker et
is
performed separately for building capital and machinery capital. This procedure is described further
(2005), Chiang (2004), and Davis et al. (1996), from whose files we gratefully obtained deflators.
al.
differences in
TFP
across plants (Op), cases
stochastic error term
=
we need
to
+
Op
+
|a„
+
X,j
to estimate the causal effect
is
impose some structure on
8pij,
+
In particular,
A(Npijt).
=
+
ln(Ap,j,)
5 l(Winner)pj +
+K
+
(x
1
is
a
dummy
>=
+Y
0)j,
+
[x-a
+
Xj
equal to
trendj,
\\i
(trend *
(Winner) *
e, (1
+ ap+
is
and a
>=
(x
1
1
we
Q (trend
(i
0))pj,
>=
+
new
use a specification that allows for the
as well as
*
1
its
so,
growth over time:
(Winner))pj,
0))j,
62 (trend *
1
(Winner) *
1
>=
(x
0))pj,
epij,
if plant
1
To do
of winning a plant on incumbent plants' TFP.
TFP
where 1 (Winner)
()j.it),
A(Npij,).
plant in winning counties to affect both the level of
(7)
TFP
(8pijt):
ln(Apij,)
The goal
industry-specific time-varying shocks to
(kj),
p
is
located in a winner county; and x denotes year, but
it
=
a
normalized so that for each case the assigned year of the plant opening
x
is
0.
The
variable trendj,
is
simple time trend.
Combining equations
(6)
and
(7)
and taking
logs,
we
obtain the regression equation that forms the
basis of our empirical analysis:
^Ypur ) =
C8)
Equation
4hiJt y^2 4^p,Jt h^3 ^A^'^jt V Pa ^A^prjt )
+
5 l(Winner)pj
+
e, (1
+
ttp
(8) is
machinery
/?1
(Winner)
+ ^„ +
A-j
+
+
trendj,
v]/
>=
* l(x
and materials
capital,
to
have
and
effects.
which are the spillover
winning county
in the
among
the
same
after the
we
1
(Winner) *
differential
=
02
=
0,
l(x
>=
>=
0))j,
+y
(trend * l(x >=0))j,
0))pj,
The paper's focus
is
the
plant on incumbent plants' TFP, so the parameters of interest are
The former
opening of the
tests for a
MDP,
mean
while the
shift in
latter
TFP among incumbent
0]
plants
allows for a trend break in
tests for a
mean
In
(8).
which assumes
inputs, l(Winner)p,
E[(l (Winner) * l(x
and l(x >=
>=
0))pj,
we make
that differential trends are not relevant here.
a difference in differences estimator
ep,j,|
In other specifications,
0)j,,
and
refer to
it
as
Model
1
the consistency of 0) in this
Op, m,„ ^j
we
we
]
=
some
.
we
specifications,
In this model, any productivity effect
shift.
occur immediately and to remain constant over time. Specifically,
Y
l(x
impacts on output.
estimate two variants of Equation
parsimonious model that simply
=
02 (trend *
+k
TFP
plants.
In practice,
to
+
0))pj,
(Winner))pj,
1
an augmented Cobb-Douglass production function that allows labor, building capital,
new
,
D. (trend *
ep.jt.
estimation of the impact of the
02
+
is
is
a
assumed
the restrictions that
This specification
fit
\(/
=
Q
essentially
Formally, after adjustment for the
model requires
the assumption that
0.
estimate the
model without imposing such
19
restrictions
on the
trends. In
other words,
we
mean
in
shift
estimate the entire Equation
productivity,
productivity. In other words,
this
(8).
We
label this
is
.
specification allows both for a
Model 2 allows us
to investigate
immediately and whether the impact evolves over time.
because our sample
Model 2 While Model
only balanced through t
=
mean
only allows for a
1
and a trend break
shift
in
whether any productivity effect occurs
This specification
demanding of
is
the data,
5 so there are only 6 years per case to estimate 6i and
h.''
Equation
(8) allows for
unobserved determinants of
TFP
but could be confounded with the spillover effects if not properly accounted
intercept for
parameter
Q
all
observations from winning counties,
way
change and trend break
when
x
>=
also allows for a
It
includes a differential
common
time trend,
in
to assess the validity
TFP common
of
this
to plants in
The
v|/.
This will
research design. Finally, k and y capture the level
winning and losing counties
after the
MDP
opening
0).
In addition,
effects for
It
for.
allows the time trend to differ for winning counties prior to the plant opening.
serve as an important
(i.e.,
5.
opening
that are unrelated to the plant
all
each plant,
our models include three sets of fixed effects.
ttp,
so the comparisons are within a plant.
First,
Second,
they include separate fixed
la,,
represents the parameters
associated with a vector of 2-digit SIC industry by year fixed effects to account for industry-specific
shocks to TFP.
Third, the Xj's are separate fixed effects for each case that ensure that the impact of the
MDP's
is
opening
intuitive appeal
identified
from comparisons within a wiimer-loser
pair; they are a
way
to retain the
of pairwise differencing in a regression framework.
A few ftirther estimation details bear noting.
First,
unobserved demand shocks are
input utilization, and this raises the possibility that the estimated P's are inconsistent (see,
likely to affect
e.g.,
Grihches
and Mairesse 1995). This has been a topic of considerable research and we are unaware of a bullet-proof
solution.
forms
We
implement the standard fixes including modeling the inputs with alternative functional
(e.g., the translog),
using cost shares
at the plant
and industry-level rather than estimating the
P's,
and controlling for flexible functions of investment, capital and materials (Syverson 2004; van
Biesebroeck 2004; Olley and Takes 1996; Levinsohn and Petrin 2003). Additionally,
we
experiment with
adding fixed effects for region by year or region by industry by year, and allowing the effect of inputs
differ
by industry or by winner- and post-MDP
alterations in the specification.
We
status.
The
also note that unobserved
consistent estimation of the parameters of interest
plants in winning counties in the years after the
(i.e., 9)
MDP's
and
basic results are unchanged
demand shocks
82) if
by
these
are only a concern for the
they systematically affect incumbent
opening, after adjustment for the rich set of
covariates in equation (8).
"'
This specification allows for spillovers
to affect the level
(2001).
20
to
of TFP and to grow over time as
in
Glaeser and
Mare
Second, in some cases
most
specifications the sample
year from
x
=
-8 through t
=
is
0.
ASM fox at least
that are in the
equation
this
is fit
from winning and losing counties
limited to plants
When
on a sample of plants from the
data from the entire country
is
entire country, but in
in the
used, the sample
is
ASM for every
limited to plants
14 consecutive years. The smaller sample of plants from the wirming and
losing counties allows for the impact of the inputs and the industry shocks to differ in the winning-losing
county sample from the
sample
is
rest
of the country.
to observations in the years
the longest period for
we probe
Third,
specifications.
between
i
Finally, for
=
-7
which we have data from
the
and x =
5.
cases.
all
most of
Due
the analysis,
to the dates
we
of the
further restrict the
MDP
openings, this
"^
and robustness of our estimates with a number of alternative
validity
For example, we investigate whether changes
of output, capital
in the price
utilization,
public investment, or attrition influence the estimates.
Fourth,
all
correlation in
Fifth,
we
of the reported standard errors are clustered
outcomes among plants
in the
the county level to account for the
same county."^
focus on weighted versions of equation
the square root of the total value of shipments in x
differences in plant size.
at
(8).
=
Specifically, the specifications are weighted
-8 to account for heteroskedasticity associated with
This weighting also means that the results measure the change
for the average dollar of output,
which
in
our view
more meaningful than
is
by
in productivity
the impact of the
MDP on the
average plant.
The analysis
will also explore
two additional
issues.
It
will report
on the
fitting
of versions of
equation (8) that interact the spillover variables with measures of economic distance between the
and the incumbent
varies with
skill
plant.
economic
adjusted
MDP
These specifications assess whether the magnitude of the estimated spillovers
distance.
Finally, the paper will assess
wages and the entry and
exit decisions
whether the
MDP's
MDP's
county.
of plants in the
opening affects local
V. Results
This section
the opening of a
is
new
divided into four subsections.
The
first
reports baseline estimates of the effect of
Million Dollar Plant on the productivity of incumbent plants in the same county
through the estimation of equation
(8).
The second subsection discusses
the validity of our design and
explores the robustness of our estimates to a variety of different specifications.
The
third subsection
explores potential channels for the agglomeration effects by testing whether the estimated spillovers vary
as a function of
economic
distance.
The
final subsection discusses the implications
of our estimates for
Data from all cases is also available for x = -8, but output in this period is used to weight the regressions.
We experimented with clustering the standard errors at the 2-digit SIC by county level, but this occasionally
produced variance-covariance matrices that weren't positive definite. In instances where they were positive definite,
these standard errors were similar to those from clustering at the county level.
^*
^'
21
the profits of local firms.
Baseline Estimates
A.
Columns
and
(1)
version of equation
(8).
(2)
of Table 4 report estimated parameters and their standard errors from a
of output
Specifically, the natural log
is
regressed on the natural log of inputs,
year by 2-digit SIC industry fixed effects, plant fixed effects, and the event time indicators
that
is
=
restricted to the years t
indicators report
year before the
(losers)
mean TFP
MDP
opened
from the winner
the estimated
TFP
in
=
-7 through x
winning and losing counties, respectively,
(i.e.,
we have
subtracted off the x
(loser) estimates for
the event time
each event year relative to the
parameter estimates for the winners
-1
each event year). Column (3) reports the difference between
mean TFP
separately plots the
1
(taken from columns (1) and (2) of Table 4) against
Table
=
in
a sample
levels within each year.
The top panel of Figure
in the estimated
The parameters associated with
5.
in
The bottom panel of Figure
x.
winner and loser coefficients against
levels for wiraier
x.
Thus,
1
and loser counties
plots the difference
a graphical version of
it is
column
(3)
of
4.
Two
important findings are apparent in these figures.
First, the trends in
plants were very similar in the winning and losing counties in the years before the
were equal.
a statistical test fails to reject that the trends
TFP among incumbent
MDP opening.
In fact,
This finding supports the validity of our
identifying assumption that incumbent plants in losing counties provide a valid counterfactual for
incumbents
in
winning counties.
Second, there
is
upward break
a sharp
in the difference in
counties beginning with the year that the plant opened.
improvement
is
due
to the
trend in winning counties.
continued decline in
The
to the
increase in
TFP among incumbent
tests in the
remainder of the paper.
is
of counties in advance of the
plants in
much of
plant opening.
large manufacturing plants appears to be a general
For example,
we
winning counties
the paper's primary finding.
TFP of incumbent
productivity increases over time in the overall economy.
and losing counties.
in
This relative
winning counties will be confirmed throughout the battery of
worth noting that the
MDP
TFP
conclusion that the opening had a small negative impact on
Overall, these graphs reveal
it
reveals that this relative
and a flattening out of the TFP
in losing counties
For example, a naive comparison of mean
before and after the opening would lead
Before proceeding,
The top graph
the winning and losing
figures also serve to underscore the importance of the availabilify of
losing counties as a counterfactual.
incumbents' TFP.
TFP
TFP between
This finding
However,
phenomenon and
plants
may
this
is
was declining
in
both sets
appear surprising, because
downward
trend in
TFP among
not specific to plants in winning
estimated augmented Cobb-Douglas production fiinctions where
22
the constant dollar total value of shipments
is
the dependent variable
and the covariates are the
stock, labor, materials (also in constant dollars), 2-digit industry fixed effects,
The equations
ASM for at least
the total value of shipments and the
by
are weighted
The average 6 year change
statistically significant decline
winning and losing counties
restricted to plants in the
in
TFP
1
This decline
To
.
MDP
calculated over the years preceding the
of 4.7%.^°
in Figure
MDP analyses.
is
among incumbents
similar to the decline
the best of our
knowledge,
this
openings was a
finding of declining
in
TFP
large manufacturing plants has not been noted previously.^'
Now
turning to the statistical models, the
fitting different versions
reports the estimated
B
row. Panel
of equation
mean
shift
(8).
Models
parameter,
6|,
first
all
columns
(1)
and
(2),
and B, respectively. Panel
and
0|
TFP
evaluated
5, as the
the sample includes
in the pre-existing trends
restricted to include only plants in counties that
sample
all
for at least 14 consecutive years, excluding all plants
"Mean
is
at x
=
A
Shift"
5 in the
both reported. ^^ The row "Pre-
02 that are also
between plants
of these specifications, the estimated impact of the
determined during the period where -7 < x <
In
A
standard error (in parentheses) in the
its
Trend" contains the coefficient measuring the difference
winning and losing coimties. In
in Panels
MDP on incumbent plants'
determined by
is
four columns of Table 5 present the results from
and 2 are
1
and
reports the estimated impact of the
"Effect after 5 years" row, which
is
is
effects.
14 consecutive years in the period 1973-1998. Thus, the approach and sample selection
rule are similar in spirit to the ones used in the
among
sample
and plant fixed
capital
MDP's
in the
opening
is
balanced during these years.
manufacturing plants in the ^S'Mthat report data
owned by the MDP
won
or lost a
firm.
MDP.
In
column
(3), the
sample
This restriction means that the
impact of the inputs and the industry-year fixed effects are estimated solely fi-om plants in these counties.
Incumbent plants are now required
though
this
does not change the
to
be in the data only for -8 < x <
results).
Finally in
column
(4), the
(not also for
sample
is
4 consecutive years,
1
restricted further to include
only plant-year observations within the period of interest (where x ranges from -7 through
the input parameters and industry-year fixed effects to be estimated solely on plant
that identify the spillover parameters.
This sample
is
This forces
by year observations
used throughout the remainder of the paper.
Estimation details are noted at the bottom of the table and apply to both Models
The
5).
entries in Table 5 confirm the visual impression fi^om Figure
1
1
and
that the
2.
opening of the
MDP
The 6 year average is a weighted averaged calculated over the 6 year periods before each of the MDP openings
where the weights are the number of plant openings associated with each 6 year period. For example, if there are 2
plant openings in 1987 and
in 1988, then the average change between 1980 and 1986 receives twice the weight as
the average change calculated between 1981 and 1987.
^'
Since we are looking at large plants that have been active for a large number of years, we speculate that this
decline may have to do with aging. Additionally, many of the years preceding the MDP openings were in the late
1970s and early 1980s, which was a period of poor economic performance. Foster, Haltiwanger, and Krizan (2000)
have documented that within plant productivity growth is positively correlated with the economic cycle.
^°
1
•"^
TTiis is calculated as 9]
+
667,
because
we
allow the
MDP to affect outcomes from x =
23
through x
=
5.
is
with a
associated
Specifically,
Model
the impact on
TFP
substantial
TFP among incumbent
in
implies an increase in
1
TFP of roughly 4.8%. As
MDP's opening
winning counties.
in
the figure highlighted, however,
12%
associated with an approximately
is
The estimates from both models would be judged
later.
plants
appeared to be increasing over time so Model 2 seems more appropriate. This model's
results suggest that the
years
increase
to
be
increase in
TFP
five
by
statistically different fi:om zero
conventional criteria and are unaffected by the changes in the specifications. Furthermore, the entries in
the "Pre-trend"
row demonstrate
TFP among incumbents
of equal trends in
that the null hypothesis
in
winning and losing counties carmot be rejected.
The numbers
square brackets in column 4 measure the average size of the spillover fi-om a
in
MDP opening in millions
mean
of 2006$. This figure
Model
per year in
$429 million
level of
in
TFP
1
fi-om a
by multiplying
=
=
5.
is
These numbers are
the estimated
-1.
MDP was associated with an increase in total
The Model 2 estimate
.
year x
impact by the
This calculation indicates
output of about $170 million
even
larger, suggesting
large,
with the Model 2 effect
an increase in output of roughly
=
at t
5 nearly the
average
MDP output.
Column
(5) presents the results
without an explicit coimterfacUial.
ASM in the same years
manufacturing plants
by
inputs, year
ASM,
not also
SIC fixed
we
effects,
fit
the average
in
TFP
based on using plant openings
The remainder of the sample includes
randomly chosen firm,
and plant fixed
is
in a
effects.
In
winning county 7
Model
to
1
1,
on the natural log of
two additional
2, the
dummy
years before the randomly
shift is the difference in these
following the opening). In Model
all
that report data for at least
a regression of the natural log of output
The reported mean
to 5 years after.
change
the
is
47 plant openings were randomly chosen from
MDP openings.
owned by
variables are included for whether the plant
chosen opening or
a "naive" estimator that
and industries as the
in the
2-digit
from
Specifically, a set of
14 consecutive years. With these data,
(i.e.,
calculated
value of incumbent plants' total shipments in winning counties in i
that the increase in
the
is
same two
two
dummy
coefficients
variables are
included along with pre- and post-trend variables. The shift in level and trend are reported, along with the
pre-trend and the total effect evaluated after 5 years.
This naive "first-difference" style estimator indicates that the opening of a
with a
MDP
-6%
to -8%o effect
on incumbent plants' TFP, depending on the model.
plant
is
associated
If the estimates fi-om the
research design are correct, then this naive approach understates the extent of spillovers
(Model
1) to
18%) (Model
2).
Interestingly, the parameter
incumbent plants was on a downward trend
new
new
plants.
message
is
It
This
that the
is
is
similar to
what
is
in
on the "pre-trend" indicates
that the
wonder about
TFP
1
3%
of the
advance of the openings in the counties that attracted these
observed in our
MDP
sample of wiimers.
Overall, the primary
absence of a credible research design can lead to misleading inferences
natural to
by
in this setting.
the degree of heterogeneity in the treatment effects fi-om the
24
47
separate case studies that underlie the estimates presented thus
by
Specifically, the Figure plots results
>=
Model
plotting case-specific estimates of parameter 9i in
there
comply with
13 of the positive estimates
positive.
level,
to
would be judged
to
be
rules).
we
from zero
statistically different
a foreign company, and whether
it is
TFP
examine
impacts of a
is
MDP's
When
an auto company.
MDP
structural assumptions
to
shed
is
these multiple measures were
MDP's
opening.^'
As an
alternative
on incumbents' productivity, we have estimated directly the
opening on output (unadjusted for inputs) and inputs. The intent
changes in outputs and inputs
with less after the
MDP
MDP.
MDP
whether the
size,
a residual and residual labeling must be done cautiously.
the impact of the
5%
at the
is 9.
regressed the estimates against three measures of the
Ultimately,
to
figure reveals that
27 of the 45 estimates are
plants.
included jointly, none were significantly related to the estimated effect of the
way
The
assessed whether the estimated spillover effects are related to characteristics of the
Specifically,
owned by
confidence interval.
that interacts the variable (1 (Winner) * 1(t
TFP of incumbent
while the comparable figure for the negative estimates
We
95%
their
Census Bureau's confidentiality
the
heterogeneity in the estimated impacts on
is
1
and
1
each of the cases. There are 45 estimates of 6|, one for each case. (Results from
0)) with indicators for
two cases were omitted
from a version of Model
Figure 2 explores this heterogeneity
far.
light
contrast the
is to
on whether productivity increased without imposing the
of the production function. Put another way, are the incumbents producing more
MDP's
opening? Appendix Table
opening on incumbents' output and usage of
reports on estimates of the impact of a
1
The estimates
inputs.
are
from the Model
1
MDP
and Model 2
versions of equation (8) with the key difference that these equations do not include the inputs as
covariates. Again,
Column
we use
the
Model 2
(1) reports that the
results to estimate the
MDP
opening
is
impact of the opening 5 years afterwards.
associated with an
Columns
(2) through (5) report the results for the four inputs.
inputs
roughly equal to or less than the increase in output.
is
noteworthy, because the 8%i increase in output
Overall, the results in
produced more with
^^
Appendix Table
less; that is,
1
is
striking that the
The Model 2
accompanied by no increase
indicate that, after the
MDP's
increase in output.
change
in all of the
results are especially
in either
form of
capital.
opening, incumbent plants
they suggest that these plants became more productive, and this
Separate regressions of the case specific effects on the
statistically significant negative coefficients.
large incumbents are left to hire labor
It is
8-12%
This result
and other inputs
MDP's
is
total
output or the
MDP's
total labor force
consistent with the possibility that
that are inferior in
unobserved ways.
when
On
the
is
generated
MDP
is
very
the other hand,
we
any significant differences when separately testing whether the productivity effect varied by the ratio
of the MDP's output to county-wide manufacturing output, whether the MDP is owned by a foreign company, or
whether the MDP is an auto company.
failed to find
25
consistent with the
TFP
increases uncovered in Table
5.^''
Threats to Validity
B.
Estimates in Table 5 appear to be consistent with significant agglomeration spillovers generated
by
MDP
TFP
in
openings. Although the comparisons in Table 3 and the similarity of the pre-existing trends in
winning and losing counties support the validity of the research design,
that there
a
is
that accounts for the higher levels
form of unobserved heterogeneity
counties after the
MDP's
of different assumptions. Specifically,
of
TFP
we
and explores the robustness of the estimates
investigate
(i)
in
winning
(iv) the possible role
to a variety
the role of functional form assumptions and the
possible presence of unobserved industry and regional shocks;
of output;
of course, possible
opening. Consequently, this subsection investigates several possible alternative
interpretations of the estimated spillover effects
in the price
it is,
(ii)
the endogeneity of inputs;
(iii)
changes
of public investment; (v) changes in capital utilization; and
(vi) attrition.
Functional Form, Industry Shocivs and Regional Shocks.
(i)
specification checks.
column
Table 6 reports on a series of
For convenience, column (1) reports the results from the preferred specification
(4) specification
of Table
5.
These estimates are intended
to
in
serve as a basis of comparison for
the estimates in the remainder of the table.
We
begin by generalizing our assumption on technology.
Douglas technology. In column
Column
(3) is
(2)
of Table
6, the
SIC
level.
This model accounts for possible differences
industries, as well as for possible differences in the quality
it is
possible that even if technology
use more skilled labor than others.
counties and before/after the
Columns
(5)
and
(6)
digit industry fixed effects.
Column
of inputs used by different industries.
similar across different manufactiirers,
some
For
industries
allows the effect of the inputs to differ in winning/losing
add census division by year fixed effects and census division by year by 2-
These specifications aim
together, the results in
appear to be insensitive
Column
(4)
technology across
to
purge the spillover effects of unobserved region-
to productivity that
might be correlated with the probability of
MDP.
Taken
^'*
was
in
to
MDP opening.
wide shocks or region by industry shocks
winning a
modeled with the translog functional form.
inputs are
based on a Cobb-Douglas technology but allows the effect of each production input
differ at the 2-digit
example,
Estimates in Table 5 assume a Cobb-
to
columns
(2)
through (6) of Table 6 are striking. The estimates
the specific functional
form of the production function.
None of
the
Appendix Table 1 presents evidence on changes in the capital/labor ratio. The model suggests that
away from labor and toward capital. The estimated change in the capital/labor ratio is poorly
determined, making definitive conclusions unwarranted, but the point estimate is not supportive of this prediction.
(6) of
firms should substitute
26
specifications contradict
tiie
findings firom the baseline specification in Table
estimates are smaller than the baseline ones, the magnitude of the decline
are
all
fail to
within one standard error of the baseline estimate in both Models
undermine the conclusion from Table 5
TFP among
inctunbent plants and this
is
opening of a
and
Overall, these results
2.
MDP leads to a substantial increase in
consistent with theories of spillovers.
An
Endogeneity of Inputs.
(ii)
that the
of the
modest. For example, they
is
1
many
Although
5.
important conceptual concern
that capital
is
and labor inputs
should be treated as endogenous, because the same forces that determine output also determine a firm's
optimal choice of inputs (Griliches and Mairesse 1995).
functions, our
capital
aim
and labor
is
is
Unlike the usual estimation of production
the consistent estimation of the spillover parameters, 9|
only relevant to the extent that
it
We do this
by
to assess this issue's
in
two ways.
fixing the parameters
columns
(7)
and
(8)
at the relevant input's
Syverson 2004; Foster, Haltiwanger, and Syverson 2007).
techniques to control for the endogeneity
of Table
we
6,
at the plant level
calculate
TFP
for
each plant
share of total costs (van Biesebroeck 2004;
This method
and the
may
mitigate any bias in the
demand
estimation of the parameters on the inputs associated with unobserved
columns, the cost shares are calculated
so the endogeneity of
relevance in this paper's setting.
First, in
on the inputs
art
62,
of these parameters. This
results in biased estimates
subsection employs the productivity literature's state of the
of capital and labor
and
shocks.
SIC industry
3-digit
sample, respectively. The estimated spillover effects are largely insensitive
two
In these
level over the full
to this restriction.
Second, Table 7 presents estimates based on the widely-used methodologies proposed by Olley
and Pakes (1996) and Levinsohn and Petrin (2003).
These methods
are based
on the
result that,
certain conditions, adjustment for investment or intermediate inputs (e.g., materials) will
correlation
between input levels and unobserved shocks
specification
4*
adds
degree
machinery/equipment investment
degree polynomials
specification
is
even richer as
stock polynomial, but
(i.e.,
it
investment) to column
specification.
In the
^^
it
adds
Column
(3).
all
(5)
the
"own"
The column
column
interactions
is
investment
building
log
and
(3) specification
(2) equation.
log
adds 4*
The column
between polynomials
in
(4)
current
interacted with the building capital
adds a 4* degree function of log materials to the baseline
4"^ degree
(6) specification, a
in building capital
is
capital stocks to the
the building investment polynomial
in log materials
The exact measure used
of
not also interacted with the machinery/equipment polynomials for stocks or
column
4* degree polynomials
polynomials
is
functions
remove the
For example, the column (2)
output.
to the baseline specification.^^
two types of log
in the
investment and capital
polynomial
to
under
polynomial in materials
and machinery
and log investment and log
is
fully interacted
Column
(7) includes fourth-degree
capital stock for
both types of capital (not
capital.
log(l+investment), so zero values are not dropped. The results are very similar
including polynomial functions of the level of investment and a
.
27
with
dummy variable
for values equal to zero.
when
column
interacted). Finally,
The estimated
(8) includes the controls
that the possible endogeneity
variable
is
Changes
of labor and capital
in the Price of
Consequently,
quantity.
1
and
and (6) specifications.
2.
from the
Overall, this exercise fails to suggest
the source of the estimated productivity spillovers.
Output. Another concern
However, due
the quantity of output.
is
is
that the theoretically correct
to the data limitations faced
models
the productivity literature, the dependent variable in our
by
(4)
spillover effects in Table 7 are generally consistent with the findings
This finding holds in both Models
baseline specification.
(iii)
from the columns
is
by
dependent
of the
virtually all
rest
of
the value of output or price multiplied
possible that the estimated spillover effect reflects higher prices, instead
it is
of higher productivity.
We
do not expect
be a major factor
this to
our context.
in
sample
First, the
is
manufacturing establishments that generally produce nationally traded goods. Therefore,
in
many
cases the price of output
is
set at the national level,
and has
little
to
comprised of
likely that
is
it
do with what happens
in the
county where the goods are produced. In the extreme case of a perfectly competitive industry
that
produces a nationally traded good, there should be no effect on prices.
Second,
are
we
more regional
be larger
tested
or
whether the size of the estimated productivity
more concentrated. The
more
in industries that are
example, the case of an industry that
new
plant
effect
may
idea
is that if
effect
is
larger in industries that
price increases are possible, then they should
Consider for
local and/or in industries that are less competitive.
sells
mainly
ultimately increase incumbents'
on demand should be negative).
at the local level (e.g.,
demand by
If the industry
is
cement).
raising local
The opening of a
income (even though the
not very competitive, the increase in
large
initial
demand may
ultimately lead to price increases for the incumbents' output.
To implement
1(t
>=
0))|,,
and
(1
we
the test,
(Winner)
*
estimated a
1(t
>=
0))pj,
Model
1
version of equation (8) that interacts
These specifications
industries or
fail to
are interacted with a
produce evidence
more concentrated
We
also conducted a similar exercise
measure of the incumbent's industry concentration.^^
that the
estimated spillover effects are larger in more local
industries; in fact, there
plants that ship their products further.
(Winner)pj,
with incumbents' industry-specific measure of average
distance traveled by output between production and consumption.
where these same variables
1
Our conclusion
is
some evidence
is
for larger effects
that price increases
on incumbent
do not appear
to
be the
source of the estimated spillover effects.
(iv)
Public Investment.
new manufacturing
State
and
local
governments frequently offer substantial subsidies
plants to locate within their jurisdictions.
to
These incentives can include tax breaks,
^*
The information on distance is from Weiss (1972). Distance varies between 52 and 1337 miles, with a mean of
Examples of regional industries are: hydraulic cement, iron and steel products, metal scrap and waste tailings,
ice cream and related frozen desserts, and prefabricated wooden buildings.
The information on industry
498.
concentration
is
from the Bureau of Census ("Concentration Ratios", 2002).
28
worker training funds, the construction of roads, and other infrastructure investments.
public investment in infrastructure benefits firms other than the beneficiary of the incentive
some of the
some of
productivity of
public investment, then
To
incumbent firms.
the
Governments.
In
may
also benefit the
we have documented
are due to
we
estimated the effect of
MDP
openings on government
total
and government construction expenditures with data from the Annual Survey of
models similar
statistically
If the productivity gains
plant
inappropriate to interpret them as evidence of spillovers.
it is
investigate this possibility,
capital expenditures
with
MDP
For example, the construction of a new road intended for a
package.
possible that
It is
to
equation
(8),
we
find that the opening of a
MDP
opening
is
could generate a meaningful portion of the productivity gains
it
seems reasonable
to
associated
is
In fact, in
negative and statistically insignificant.
the specifications that produce positive insignificant estimates, there
measures of public investment,
plant
and construction expenditures.
insignificant increases in capital
specifications the estimated impact of a
MDP
is
no plausible
rate
wirming counties.
in
most
Even
in
of return that
Based on these
conclude that public investment cannot explain the
paper's results.
(v)
Changes
may respond
to the
in Capital Utilization.
MDPs
Another potential threat
by increasing the
to validity is that inciunbent plants
intensity of their capital usage.
This could happen
depressed counties where the existing capital stock had been used below capacity win the
by operating
increase production simply
possibility,
we
usage (which
Table
1
estimated whether the
is
their capital stock closer to capacity.
MDP's
opening affected the
utilization is unlikely to
we
Sample
Plants.
Differential attrition in the
winning and losing counties could contribute
MDP's
opening.
to the
measured
The
spillovers in
colunm
the
7 of
Appendix
conclude that greater capacity
sample of incumbent plants
This attrition could result from plants shutting
that are
down
in
among
operations or
surveyed with certainty
/15M"
available evidence suggests that differential attrition
winning counties.
(4) of
MDP
Column
differential in productivity trends
from plants continuing operations but dropping out of the group of plants
as part of the
indirect test of this
be the source of the findings of productivity spillovers.
Attrition of
survivors after the
Thus,
and
of the dollar value of energy
increasing in the use of the capital stock) to the capital stock.
reports small and insignificant changes in this measure.
(vi)
ratio
As an
MDPs
if
Table 5 and
opening were
in
still
First, in the
Tables 6 and
in the
sample
baseline sample
7),
72%
at its
(i.e.,
is
the one used to produce the results in
of the winning county plants operating
end
(i.e.,
x
=
5).
The analogous
is
68%. The
^'
Recall, establishments are sampled with certainty if they are part of a
slightly larger attrition rate in losing counties
exceeding $500 million or their
total
employment was
unlikely to explain the finding of
at least
29
250.
is
in the
year of
figure in losing counties
consistent with the paper's primary result.
company with manufacturing shipments
one seemingly reasonable interpretation of
Specifically,
plants to remain
some winning county
MDP
open
would have otherwise
that
MDP's opening
this result is that the
Thus
closed.
allowed
to the extent that a
TFP
opening keeps weaker plants operating, the above analysis will underestimate the overall
on the sample of plants
increase. Second, the estimation of equation (8)
7 to +5 yields results that are qualitatively similar
TFP among
hypothesis of equal trends in
MDP's
opening cannot be rejected; the
counties
was -0.0052
to
those from the
fiill
sample.
minus the TFP trend
trend in winning counties
-
Third, the null
winning and losing counties prior
attriting plants in
TFP
that is present for all years fi-om
to the
in losing
(0.0080).^^
Estimates of Spillovers by Economic Distance
C.
What can
mechanisms
light
explain the productivity gains uncovered above? Section
may
that
how
A
measured spillover
the
discussed some possible
Tables 8 and 9 attempt to shed some
be responsible for agglomeration spillovers.
on the possible mechanisms by investigating
I
effect varies as a ftinction
of economic distance.
By
Table 8 shows separate estimates from the baseline model for samples of
Industry.
incumbent plants
in the
MDP's
2-digit industry
the effects of spillovers decline with
does not shed direct
that spillovers
light
would be
all
other industries. In general, one might expect that
economic distance (equation
on which channel
is
5').
the source of the spillovers,
seems reasonable
it
MDP's
SIC
MDP
industry results,
2-digit industry.
We
it
is
also note that there can
research design and the available data do not permit an examination
Column
1
of Table 8 repeats the
all
intended to serve as a basis of comparison.
specification for incumbent plants in the
entries in these
to
presume
important to recall that just 16 of the 47
be substantial heterogeneity
and labor forces among the industries within a 2-digit SIC industry.
in technologies
The
Although looking within-industry
larger within an industry.
In examining the 2-digit
cases have plants in the
and
columns
are
industries estimates
Columns
MDP's
(2)
and
and
(4)
of Table 5 and
is
on estimates from the baseline
all
Just as in
this
industry definitions.
from column
(3) report
2-digit industry
from the same regression.
at finer
However,
other industries, respectively.
Table
5, the
numbers
in
square
brackets convert the parameter estimates into millions of 2006$.
we also tested whether the results are sensitive to
MDP's opening. The estimated spillovers are virtually unchanged when we use the year
observed in SSEL as the MDP's opening date. When the year of the MDP article in Site
In addition to the specification checks described in this section,
the choice of the date of the
that the plant
is
first
used as the plant's opening date, the Model
results are nearly identical to those in the Table 5 column
(4) specification and roughly 5% in the Model 2 specification. When the estimating equation is unweighted, the
evidence in favor of spillover effects is weaker indicating that the spillovers are concentrated among the larger
Selection
is
plants in the sample.
change
1
As discussed above, our view
in productivity for the
is
that the
economically relevant concept of spillover
average dollar of output, rather than the average plant.
30
is
the
The impacts
increase in
TFP
for plants in the
poorly determined
33%
at x
own
=
2-digit
same
5 in
3.3%
statistically insignificant
findings in the
own
are substantially larger in the
in
2-digit industry
Model
For example, the estimated
a statistically significant
is
and marginally significant 8.9%
1
17%
in
Model
1
and a
In contrast, the estimates for plants in other industries are a
2.
Model
2-digit industry.
and other industries are robust
in
Model
These basic
2.
to the different specifications in tables 6
and
7.^V
MDP
Figures 3 and 4 provide 2-digit
Importantly, there
statistical tests
is
industry and other industry analogues to Figure
not evidence of differential trends in the years before the
confirm
this visual
The
impression.
small sample size, which was also evident in the
2-digit
By
Direct Measure of Economic Proximity.
MDP
in the
proximity, and input-output flows.
are standardized to
we
Specifically,
1,
the estimated
and a cessation of the
To
economic proximity more
between the
In
all
cases, a positive value
estimate the following equation:
>=
l(Winner)pj * Proximity ij+
+
7t3
(I
+k
1(t
(Winner) * 1(t
>=
0))j,
0)pj,
*
+
6, (1
712
>=
(Winner) * 1(t
(1(t
>=
Proximity^
0)jt
)
0))pj,
* Proximityij)
+ Op +
)ijt
+
A,j
+
Spyt.
measure of economic proximity between the incumbent plant industry and the
industry. This equation
by
industries.
71]
a
directly
ease the interpretation, the economic proximity or linkage variables
+
is
simply an augmented version of Model
linkage variables with l(Winner)pj, 1(t
the coefficient
is
that the spillover is larger for
of economic proximity that capture worker flows, technological
5 l(Winner)pj
is
Having found
investigate the role of
+
where Proximityy
"after,"
in losing counties
have a mean of zero and standard deviation of one.
indicates a "closer" relationship
which
we
industry,
exphcit measures
using several
,
Just as in Figure
trend in wirming counties.
incumbent plants
is 713
TFP
MDP's opening and
industry estimates are noisy due to the
statistical results.
impact reflects the continuation of a downward trend in
downward
MDP
1.
on the
>=
0))jt
,
and
(1
(Winner)
triple interaction
1
that
* 1(t
between the
>=
MDP
adds interactions of the industry
0))pjt.
dummy
The
coefficient of interest
for winner, the
dummy
for
and the measure of proximity. This coefficient assesses whether "closer" industries benefit more
from the
MDP's
after the
MDP
opening.
A positive
coefficient
means
that the estimated productivity spillover is larger
opening for incumbents that are geographically and economically close
to the
new
plant.
Within the same 2-digit SIC, 71% of incumbents in winning counties and 69% of incumbents in losing counties
were still in the sample 5 years after the opening. Additionally, attriting plants within the same 2-digit SIC were
also on statistically indistinguishable trends prior to the MDP opening. Thus, differential attrition seems unlikely to
^'
explain the 2-digit results.
31
relative to
to the
incumbents that are geographically close but economically distant from the new plant (relative
same comparison among incumbents
productivity spillover
proximity to the
same
the
is
in loser counties).
new plant.
example, column (1) suggests
between incumbent
This finding
the spillover.
across firms.
One
information on
new
that a
6 columns include the interactions in one
one standard deviation increase
plants' industry
is
first
and the
MDP's
industry
is
in the
at a time.
CPS Worker
7%
associated with a
For
Transitions
increase in
consistent with the theory that spillovers occur through the flow of workers
possibility
is
new workers
that
share ideas on
how
organize production or
to
technologies that they learned with their previous employer.
be especially high within 2-digit industries, so
Table
economic
for all the incumbents in a county, regardless of their
Table 9 reports estimates of 713. The
variable
A zero coefficient means that the estimated
this finding
was foreshadowed by
This measure tends to
the
own
2-digit results in
8.
The measures of
spillover.
The
intellectual
or technological linkages indicate meaningful increases in the
mechanism by which
precise
these ideas are shared
is
unclear, although both the flow of
workers across firms and the mythical exchange of ideas over beers between workers fi-om different firms
are possibilities. Notably, there
CPS labor
more
variation in these measures within 2-digit industries than in the
transitions measure.
Columns
magnitude of
(5)
and
spillovers.
(6) provide little support for the
Thus, the data
encourages (or even forces)
owned by
is
the
MDP's
its
support the types of stories where an auto manufacturer
suppliers to adopt
more
efficient production techniques.
Recall, all plants
firm are dropped from the analysis, so this finding does not rule out this channel
The finding on
within firms.
fail to
flow of goods and services in determining the
the importance of labor flows
is
consistent with the results in Ellison,
Glaeser and Kerr (2007) and Dumais, Ellison, and Glaeser (2002), while the finding on input and output
flows stands
in contrast
In the
column
the citation pattern,
be
with these papers' findings.
(7) specification,
we
include
all
the interactions simultaneously.
and the technology input interactions
statistically insignificant.
The
all
remain positive but
interactions with proximity to customers
The labor
flow,
now would be judged
now
and suppliers are
to
both
negative.
Overall, this analysis provides
that share
is
some support
for the notion that spillovers occur
between firms
workers and between firms that use similar technologies. In terms of Section IC,
consistent with intellectual extemahties, to the extent that they are
from firm
to firm,
similar.
Table 9 seems
and
to the extent that they
less
embodied
in
this
workers
evidence
who move
occur among firms that use technologies that are reasonably
consistent with the hypothesis that agglomeration occurs because of
proximity to customers and suppliers.
We
caution against definitive conclusions, because the utilized
32
measures are
between workers and firms could not be
Entry and Labor Costs
D.
directly tested with these data.
as Indirect Tests of Spillovers
The paper has uncovered economically
TFP
of the
digit industry
This effect
five years later.
new
is
MDP
a
plant opening
MDP's
is
The
entries in Panel
1)
and the log of
In both columns, all plants
respectively.
The sample
12%
same
that are in the
2-
1
this
tests.
they are larger than the increase in costs
(i.e.,
new firms
come from
subsection
(relative to the losing counties).
regressions that use data from the Census
conducted every five years. The dependent variables are the log of the number
of establishments (column
variable.
of a sufficient magnitude
county should experience entry by
tests this prediction.
of Manufactures, which
associated with a
plant and for plants that tend to share workers and technologies. In the presence
First, if the spillovers are
Table 10
is
even larger for incumbent plants
of positive spillovers, the model has two empirical predictions which
in the short run), the
incumbent establishments
sizable productivity gains for
new MDP. For example,
following the opening of the
increase in
Further, the possibility of better matches
imperfect proxies for the potential channels.
all
total
owned by
manufacturing output (column 2)
the
MDP's
in the
county,
firm are excluded from the dependent
comprised of observations from winning and losing counties only.
is
The
covariates include a full set of county fixed effects, year fixed effects, case fixed effects, and an indicator
for
whether the observation
is
from
after the
MDP's
opening. The parameter of interest
associated with
is
the interaction of indicators for an observation from a winning county and the post-opening indicator, so
it is
a difference in differences estimator of the impact of the
Column
(1) reports that the
winning counties
MDP
after the
plants are of an equal size.
an increase
of a
MDP
in output at
plant
although this
is
is
The
number of manufacturing
plant's opening.
total
A
value of output
an existing plant and a
new
MDP
manufacturing
opening.
limitation of this
is
"*"
plants increased
measure
is
by roughly 12.5%
that
it
assumes
economically more meaningful, because
plant equally.
As column
in
that all
it
treats
opening
(2) highlights, the
associated with a roughly 14.5% increase in total output in the manufacturing sector
not estimated precisely.
Overall, these results are consistent with the
that the
MDP's
attracted
sector.'"
new economic
Presumably,
TFP
activity to the
this
new
results
of substantial spillovers in that
winning counties
activity located in the
it
appears
(relative to losing ones) in the
winning counties
to gain access to
the spillovers.
"''
Because data
is
available every 5 years, depending on the Census year relative to the
MDP
opening, the sample
opening and 4-8 years after the MDP opening. Thus, each
associated with one earlier date and one later date. Models are weighted by the number of plants
years are
1
-
5 years before the
years -6 to -10 and
*'
It is
column 4
possible that the
expect increased entry
is
MDP
weighted by the county's
MDP's
spillovers extended
total
manufacturing output
beyond manufacturing.
in other sectors too.
33
in
MDP opening is
in the
county in
years -6 to -10.
In this case,
it
might be reasonable to
The second prediction
as
is
The most important
firms compete for these factors of production.
manufacturing plants
the log
is
Column
labor.
and losing
These data are preferable
counties,''"
Manufacturers
wage
the aggregate
(i.e.,
(3) in Panel 2
bill for
to the
supplied input for
2000 Censuses of Population from the winning
measure of labor costs reported
in the
Census of
production and non-production workers), which does not
provide information on the quality of the labor force
estimate models for In(wage) and control for
locally
of Table 10 reports the results from regressions of
the 1970, 1980, 1990, and
wage using data from
of local inputs will increase
that if the spillovers are positive, the prices
(e.g.,
dummies
education and experience).
Specifically,
worker age and year, age-
for interactions of
squared and year, education and year, sex and race and Hispanic and citizen, and case fixed effects.
also include indicators for whether the observation
opening, and the interaction of these two
is
an adjusted difference
equation
is
analogous
The
in
indicators.''^
in differences estimator
to the
Model
1
is
from a wirming county, occurs
This interaction
of the impact of the
version of equation (8) that
MDP's
was used
after the
MDP's
the estimated
2.7% wage
earnings in winning counties implies that the quality-adjusted annual
industries increased
by roughly $151 million
positive spillovers and an
(as in Section
is
upward sloping
after the
MDP's
opening.
and
This
analyze TFP.
to
opening. This effect appears quantitatively sizable and
The multiphcation of
statistically significant.
We
MDP's
opening on wages.
estimate indicates that after adjusting for observable heterogeneity, wages increase
wirming counties
It
after the
the focus of the regression
is
we
increase
wage
is
by 2.7%
marginally
by the average labor
bill for
This finding
employers
in all
consistent with
is
labor supply curve, perhaps due to imperfect mobility of labor
I).
wages
possible to use the estimated increase in
calculations of the
MDP's
indicated an increase in
impact on incumbent plants'
TFP
profits.
to
make some back of
Recall, the
of approximately 4.8% (we focus on Model
1
Model
because
1
the envelope
result in
Table 5
impossible to
it is
estimate a version of Model 2 with the decennial population Census data). If we assume that workers are
homogenous
or that high and low skill workers are perfectly substitutable in production, then the labor
market-wide increase
in
accounts for roughly
23%
wages
of
applies throughout the manufacturing sector.
total costs, so the
that manufacturers' costs increased
estimated
2.7%
by approximately 0.62%.
In our sample, labor
increase in skill adjusted
wages implies
The increased production
costs due to
The sample is limited to individuals who worked last year, worked more than 26 weeks, usually work more than
20 hours per week, are not in school, are at work, and who work for wages in the private sector. One important
limitation of the Census data is that they lack exact county identifiers for counties with populations below 100,000.
''^
Instead,
it
is
possible to identify
PUMAs
Census, which
in the
introduces significant measurement error, which
is
in rural areas
can include several counties.
This
partly responsible for the imprecision of the estimate.
The pre-period is defined as the most recent census before the MDP opening. The post-period is defined as the
most recent census 3 or more years after the MDP opening. Thus, the sample years are 1 - 10 years before the MDP
opening and 3-12 years after the MDP opening.
34
13% of the
higher wages are therefore
gain in TFP.
These calculations demonstrate
that the gains in
TFP do
Further, the observations on
possible to determine the total increase in production costs.
expands
to
due
Since the prices and quality of other inputs are not observable,
higher costs of local inputs.
only a few years after the
not translate directly to profits
MDP's
opening and the impact on wages
may
increase
more
it
to the
is
not
wages occurs
as production
gain access to the spillover. For these reasons, this back of the envelope calculation should be
interpreted as a lower
the total impact
on
bound of
profits
is
the increase in input costs.
In the long run, an equilibrium requires that
zero.
Summary and
VI.
Interpretation
This paper makes three principal contributions. This section summarizes them and places them in
some
context.
The
and most robust
first
result is that the successful attraction
TFP
average associated with increased
was about 12%
counties
years after the
it
is
Model
In the preferred
specification tests.
MDP's
higher.
for
incumbent
2, the
of a "Million Dollar Plant"
This finding
plants.
is
on
is
robust to a battery of
estimates suggest that incumbents'
TFP
in
winning
This translates into an additional $430 million in armual output five
opening. This
is
an economically large number and
it
is
natural to
wonder whether
"too large" to be plausible.
There are several related issues worth noting when considering
be an unobserved factor correlated with the
invalidate the identifying assumption.
MDP's
TFP documented
characteristics of
winning and losing counties and
in Figure
1
First, there
may
opening that can explain these results and would
The likelihood of
the pre-trends in
this possibility.
diminished by the similarity of
this possibility is
and the balancing of many of the ex-ante observable
incumbent
their
plants.
Nevertheless, this possibility
caimot be dismissed as would be the case in a randomized experiment.
Second,
it is
possible that the estimated impacts on
TFP
of the workforce employed by incumbent plants. The sign of
hand, the
MDP
may
attract
quality of their workforce.
total
this
change
is
by changes
in the quality
a priori unclear.
On
the one
higher quality workers to the county, allowing the incumbents to upgrade the
On
the other hand, the
Regardless of which force prevails,
measured as
are influenced
this
MDP
could hire away the incumbents' best workers.
could affect the estimated impact on
TFP
because labor
hours by production workers (recall that education or other measures of
skill are
is
not
included in the ASM).
Third, the external validity of the results
sample
differ
is
from the average manufacturing plant
plants generated bidding
from
local governments,
unknown.
in
many
which
35
is
In particular, the
respects.
47
MDP
plants in the
Perhaps most importantly, these
a first indication that there
may
be an ex-ante
Further, these plants are substantially larger than the typical
expectation of significant spillovers.
The point
manufacturing plant.
opening of more
that these results are unlikely to generalize to the
is
typical plants.
TFP
Fourth, the estimated impact on
As Table 10
on the TFP of incumbents.
opening
new
in the
same county. Thus,
Consequently,
plants.
effect of the
MDP's
it
more appropriate
opening and everything
The second contribution
underscored that increases
in
is
MDP's
indicated, the
may
the estimated spillover
likely
is
TFP do
one
and
economic
activity also
county that leads
The
documented
to higher local prices
Table 10
TFP
Specifically, our tentative conclusion
This
technologies that are reasonably similar.
Additionally,
will be
accompanied by increases
in
Further, the increased levels of
demand
winning
to locate in the
is
light
on the charmels
that underlie the
that the spillovers are larger
between firms
it
firm to finn and occur
is
among
firms that use
consistent with higher rates of
TFP due
to
efficiencies of worker-firm matches.
Finally, these results
may have some
governments providing subsidies
for the locality but
it
is
not always transparent,
production
(e.g.,
to
new
believe that
it
In contrast, the finding of spillovers
is
In this case, tax competitions
MDP
and consequently might choose a location where
spillovers are greatest.
that
Although the economics of
the basis of an
even from a national perspective. Specifically, the
outcome
is
it
may be
this
rational
argument are
generally refers to cases of disequilibrium where local factors of
labor or buildings) are unemployed.
the socially efficient
standard critique of local
plants to locate within their jurisdictions
welfare decreasing for the nation.
we
A
surprising policy implications.
locally but suboptimal nationally (at least in the absence of significant
that
model
consistent with intellectual externalities, to the
is
who move from
However,
In particular, the
an equilibrium. The finding of higher prices
some
extent that they are embodied in workers
improved
data.
MDP alone.
and the new equilibrium.
workers and use similar technologies.
that share
in
reflect the increased
third conti'ibution is that the paper has shed
estimated spillovers.
by the
and these
reduced form
effect as the
rather than the impact of the
consistent with this prediction.
is
in
MDP
not translate directly into higher profits available for other
that these increases are necessary for
for quality-adjusted labor in Table 10
it
TFP
impact
to lead to other plants
from the
reflect spillovers
that is supported
manufacturing plants. The model predicts that the increases
local input prices
opening appears
to interpret the
accompanies
that
a theoretical
MDP's
unlikely to be a structural estimate of the
is
is
its
economic
produce the spillovers so that they internalize
In thinking about the policy implications,
costs are
is
36
costs).
justification for local subsidies,
its
own
low but the spillovers are minimal.
where
the
sum of their
maximized when payments
this externality in
it is
be beneficial
plant cannot capture these spillovers on
for these plants to locate
In this setting, national welfare
worker moving
may
making
are
profits
made
and the
to plants
their location decision.
important to bear in mind that the estimated
12%
gain in
TFP
is
an average
the estimated impact
is
As
effect.
negative in
Figiire 2 demonstrated, there
40%
of the cases. The point
be unwilling to provide tax incentives with
this distribution
plants to incumbent plants.
Thus,
this
For example,
governments
may
flowing from large
new
that risk averse local
of outcomes.
that there are substantial spillovers
Further, these spillovers are larger
use similar technologies.
agglomerate in certain
documented
substantial variability.
Conclusion
VII.
Overall, this paper has
is
is
between plants
that share labor pools
and
paper has provided evidence consistent with the idea that firms
localities, at least in part,
because they are more productive for being close to other
firms.
There are several implications for
fiiture research.
First, the
paper has demonstrated the value of
quasi-experiments that plausibly avoid the confounding of spillovers with differences in the determinants
of
TFP
across locations.
Second, the paper highlights that
conducted by directly measuring TFP. These
of rates of coagglomeration
it is
that
may
tests
the heterogeneity in the estimated spillovers
an especially
is still
presence of spillovers can be
can serve as an important complement to measurement
reflect spillovers, cost shifters, or natural advantages.
important to determine whether impacts on
Table 9 underscore that there
tests for the
much
TFP
In this spirit,
are evident outside the manufacturing sector.
documented
in
Figure 2 and the results on the mechanism in
to learn about the structural source
fruitful area for future research.
37
Third,
of these spillovers. This
is
References
Acemoglu, Daron.
1996.
"A Microfoundation
Social
for
Increasing
returns
in
Human
Capital
Accumulation." Quarterly Journal of Economics 11 (August): 779-804.
Audretsch, David and Maryann Feldman. 1996. "R&D Spillovers and the Geography of Innovation and
Production." American Economic Review 86 (June): 630-640.
Audretsch, David and Maryann Feldman. 2004. "Knowledge Spillovers and the Geography of
Innovation." In Handbook of Urban and Regional Economics 4, edited by J. Vernon Henderson and
Jacques-Fran9ois Thisse. Amsterdam: North-Holland.
Becker, Randy, Jolin Haltiwanger, Ron Jarmin, Shawn Klimek, and Daniel Wilson. 2005. "Micro and
Macro Data Integration: The Case of Capital." Bureau of the Census: Center for Economic Studies
Working Paper 05-02.
Bloom, Nick, Mark Schankerman, and John Van Reenen. 2007. "Identifying Technology Spillovers and
Product Market Rivalry." National Bureau of Economic Research Working Paper No. 13060.
Branstetter, Lee G. 2001. "Are Knowledge Spillovers International or Intranational in Scope?
Microeconometric evidence from US and Japan." Journal of International Economics 53
,
(February): 53-79.
Card, David, Kevin F. Hallock, and Enrico Moretti. 2007. "The Geography of Giving: The Effect of
Corporate Headquarters on Local Charities." University of California, Berkeley Mimeograph.
Hyowook. 2004. "Learning By Doing, Worker Turnover,
Econometric Society 2004 Far Eastern Meeting Paper No. 593.
Chiang,
Davis, James and
J.
and Productivity Dynamics."
Vernon Henderson. 2004. "The Agglomeration of Headquarters." Bureau of
the
Census: Center for Economic Studies Working Paper 04-02.
Davis, Steven
J.,
MA: MIT
John C. Haltiwanger, and Scott Schuh. 1996. Job Creation and Destruction. Cambridge,
Press.
Davis, Donald and David E. Weinstein. 2002. "Bones, Bombs, and Break Points: The Geography of
Economic
American Economic Review 92 (December).
and Glaeser, Edward L. 2002. "Geographic Concentration as a Dynamic
Process." Review of Economics and Statistics 84 (May): 193-204.
Duranton, Giles and Diego Puga. 2004 "Micro-Foundations of Urban Agglomeration Economies." In
Handbook of Urban and Regional Economics 4, edited by J. Vernon Henderson and JacquesFranfois Thisse. Amsterdam: North-Holland.
Ellison, Glenn and Edward L. Glaeser. 1997. "Geographic Concentration in U.S. Manufacturing
Dumais, Guy,
Activity."
Ellison, Glenn,
A 'Darihoard A.^p'^rodich." Journal of Political Economy 105 (October): 889-927.
Glenn and Edward L. Glaeser. 1999. "The Geographic Concentration of Industry: Does Natural
Advantage Explain AggXamcraXxouT American Economic Review 89 (May): 311-316.
Ellison, Glenn, Edward L. Glaeser, and William Kerr. 2007. "What Causes Industry Agglomeration?
Evidence from Coagglomeration Patterns." National Bureau of Economic Research Working Paper
No. 13068.
Foster,
Lucia,
John
Haltiwanger,
and
C.
Chad
Syverson.
2007.
"Reallocation,
Firm
Turnover,
and
Efficiency:
Selection
Profitability?"
on
Productivity
or
University of Chicago Mimeograph..
Foster, Lucia, John C. Haltiwanger, and C.J. Krizan. 2000. "Aggregate Productivity Growth: Lessons
from Microeconomic Evidence." Mimeograph.
Glaeser, Edward L. 1999. "Learning in Cities." Journal of Urban Economics 46(September): 254-77.
Glaeser, Edward L. 2001. "The Economics of Location-Based Tax Incentives." Harvard Institute of
Economic Research, Discussion Paper Number 1932.
Glaeser, Edward L. and Janet E. Kohlhase. 2003. "Cities, Regions and the Decline of Transport Costs."
Papers in Regional Science 83 (January): 197-228.
Glaeser, Edward L. and David C. Mare. 2001. "Cities and Skill." Journal of Labor Economics 19(2):
Industries:
Ellison,
316-342.
38
Greenstone, Michael and Enrico Moretti. 2004. "Bidding for Industrial Plants: Does Winning a 'Million
Dollar Plant' Increase Welfare?" National Bureau of Economic Research Working Paper No. 9844.
Griliches, Zvi. 1958. "Research Cost and Social Returns:
of Political
Economy 66
Hybrid
Com
and Related Innovations." Journal
(October): 419-431.
Zvi and Jacques Mairesse. 1995. "Production Functions: the Search for Identification."
National Bureau of Economic Research Working Paper No. 5067.
Griliches,
Grossman, Gene M. and Elhanan Helpman. 1991. Innovation and Growth in the Global Economy.
Cambridge, MA: MIT Press.
Henderson, J. Vernon. 2001. "Urban Scale Economies." In Handbook of Urban Studies, edited by Ronan
Paddison. London: SAGE Publications.
Henderson, J. Vernon. 2003. "Marshall's Scale Economies." Journal of Urban Economics 55 (January):
1-28.
J. Vernon and Duncan Black. 1999. "A Theory of Urban Growth." Journal of Political
Economy, 107 (Apnl): 252-284
Adam B. 1986. "Technological Opportunities and Spillovers of R&D: Evidence from Firm's
Patents, Profit and Market Value." American Economic Review 76 (December): 984-1001.
Adam B. and Manuel Trajtenberg. 2002. Patents, Citations and Innovations: A Window on the
Knowledge Economy. Cambridge, MA: MIT Press.
Adam B., Manuel Trajtenberg, and Michael S. Fogarty. 2000. "Knowledge Spillovers and Patent
Citations: Evidence from a Survey of Inventors." American Economic Review 90 (May): 215-218.
Adam B., Manuel Trajtenberg and Rebecca M. Henderson. 1993. "Geographic Locahzation of
Knowledge Spillovers as Evidenced by Patent Citation." Quarterly Journal of Economics 108
Henderson,
Jaffe,
Jaffe,
Jaffe,
Jaffe,
(August): 577-598.
Jovanovic, Boyan and Rafael Rob.
1989. "The Growth and Diffusion of Knowledge." Review of
Economic Studies 56 (October): 569-582.
Krugman, Paul. 1991a. Geography and Trade. Cambridge, MA: MIT Press.
Krugman, Paul. 1991b. "Increasing Returns and Economic Geography," Journal of Political Economy 99
(June): 483-499.
Levinsohn, James and Amil Petrin. 2003. "Estimating Production Functions Using Inputs
to Control for
Unobservables." Review of Economic Studies 70 (April): 3 1 7-342.
Lucas, Robert E. Jr. 1988. "On the Mechanics of Economic Development." Journal of Monetary
Economics, 22 (July): 3-42.
Marshall, Alfred. 1890. Principles of Economics. New York: Macmillan and Co.
McGahan, Anita W. and Brian S. Silverman. 2001. "How Does Irmovative Activity
Change
Industries Mature?" International Journal of Industrial Organization 19 (July):
1
as
141-1 160.
Moretti, Enrico. 2004a. "Estimating the external return to higher education: Evidence from crosssectional and longitudinal daia." Journal of Econometrics 120 (July-August): 175-212.
Moretti, Enrico. 2004b. "Workers' Education, Spillovers and Productivity: Evidence from Plant-
Level Production Functions." American Economic Review 94 (June): 656-690.
Moretti, Enrico. 2004c.
"Human
Regional Economics
Handbook of Urban and
Vernon Henderson and Jacques-Fran9ois Thisse. Amsterdam:
Capital Externalities in Cities." In
4, edited
by
J.
North-Holland.
Mowery, David C. and Arvids A.
Ziedonis. 2001. "The Geographic Reach of Market and Non-market
Channels of Technology Transfer: Comparing Citations and Licenses of University Patents."
National Bureau of Economic Research Working Paper No. 8568.
Olley, Steven G. and Ariel Pakes. 1996. "The Dynamics of Productivity in the
Telecommunications Equipment Industry." Econometrica 64 (November): 1263-98.
Ottaviano, Gianmarco and Jacques-Franfois Thisse. 2004. "Agglomeration and Economic Geography." In
Handbook of Urban and Regional Economics 4, edited by J. Vernon Henderson and JacquesFran9ois Thisse. Amsterdam: North-Holland.
Petrongolo, Barbara and Christopher A. Pissarides. 2005. "Scale Effects in Markets with Search." The
39
Economic Journal, 116 (January): 21-44.
Rauch, James E. 1993. "Productivity Gains from Geographic Concentration of Human Capital: Evidence
from the Cities." Jowma/ of Urban Economics 34 (November): 380-400.
Roback, Jennifer. 1982. "Wages, Rents and the Quality of LifQ." Journal of Political Economy 90
(December): 1257-1278.
"Wages, rents and amenities: Differences among workers and regions."
Roback, Jennifer. 1988.
Economic Inquiry 26 (January): 23-41
Rosenbaum, Paul R. and Donald B. Rubin. 1983. "The Central Role of the Propensity Score
in
Observational Studies for Causal Effects." Biometrika 70 (April): 41-55.
Rosenthal, Stuart S. and William C. Strange. 2001. "The Determinants of Agglomeration."
Jowma/ of
Urban Economics 50 (September): 191-229.
Rosenthal, Stuart S. and William C. Strange. 2004. "Evidence on the Nature and Sources of
Agglomeration." In Handbook of Urban and Regional Economics 4, edited by
Henderson and Jacques-Francois Thisse. Amsterdam: North-Holland.
Saxenian,
Anna
Lee. 1994. Regional Advantage: Culture
128. Cambridge,
and Competition
J.
Vernon
in Silicon Valley
and Route
MA:
Harvard University Press.
Syverson, Chad. 2004a. "Market Structure and Productivity:
Economy 6 (December): 1181-1 222.
A Concrete Example." Journal of Political
Syverson, Chad. 2004b. "Product Substitutability and Productivity Dispersion." Review of Economics and
Statistics 2 (May): 534-550.
van Biesebroeck, Johannes. 2004. "Robustness of Productivity Estimates." National Bureau of Economic
Research Working Paper No. 10303.
Venable, Tim. 1992.
"BMW Drives into South CaroHna with $300 Million Auto Plant." Site Selection
August: 630-631.
W. 1972. "The Geographic Size of Markets in Manufacturing." The Review of Economics
and Statistics 54 (August): 245-257.
Zucker, Lyrme G., Michael R. Darby, and Marilynn B. Brewer. 1998. "Intellectual Human Capital and the
Birth of U.S. Biotechnology Entciprises." American Economic Review 88 (March): 290-306.
Weiss, Leonard
40
Figure
1:
The
Effect of a "Million Dollar Plant"
Opening on TFP of All Manufacturing Plants
in
Winner
and Loser Counties.
All Industries: Winners Vs. Losers
0.1
0.05
n
•
A-..
"
•
-A
-
-7
-0.05
-0.1
-0.15
-6
::t^.
^i.
-5
-4
-3
-2
"^•-,..2.
-1
3
~:^
-
''A
-
-
Year, relative to opening
——
Winning Counties
Difference: Winners
Notes: These figures accompany Table
4.
41
---a--- Losing Counties
- Losers
Figure
2.
Distribution of Case-Specific
Mean
Shift Effects
from the Opening of a "Milhon Dollar Plant"
0.5
0.4
0.3
0.2
'
I
0.1
f+
U iU HllJlH^'
45
-0.2
-0.3
-0.4
-0.5
Notes: The figure reports results from a version of Mode!
cases.
The
figure reports only 45 estimates because
1
that estimates the
two cases were dropped
42
6i for each of the 47 MDP
Census confidentiality reasons.
parameter
for
The Effect of a "Million Dollar Plant" Opening on TFP of Manufacturing Plants
2-Digit Industry in Winner and Loser Counties.
Figure
3:
2-digit
MDP Industry:
Winners Vs. Losers
Difference (Winners
Notes: These figures accompany Table
8,
Column 2 (MDP's
43
- Losers)
2-digit Industry).
in the
MDP's
Figure
4:
The
Industries,
Effect of a "Million Dollar Plant"
Except the
MDP's
Opening on TFP of Manufacturing Plants
Winner and Loser Counties.
in All
2-Digit Industry in
Other Industries: Winners Vs. Losers
0.1
.^
0.05
A
""*""
!>>.
v.;*;^^----
A
-7
-6
-5
-4
-3
-
•
-A
-2
-1
-0.05
-0.1
-
-0.15
Year, relative to opening
-Winning Counties
Difference (Winners
Notes: These figures accompany Table
8,
Column
---a--- Losing Counties
-
Losers)
3 (All 2-digit Industries,
44
except the
MDP's
2-digit Industry).
Table
1.
The "Million Dollar Plant" Sample
(1)
Sample
MDP Openings':
47
Across All Industries
Within Same 2-digit SIC
16
Across All Industries:
Number of Loser Counties per Winner County:
31
1
2+
16
Reported Year - Matched Year:
^
20
-2 to -1
15
to
1
12
3
Reported Year of MDP Location:
1981-1985
1986-1989
1990-1993
11
18
18
MDP Characteristics,
5 years after opening:'
452801
Output ($1000)
(901690)
Output, relative to county output
1
0.086
year prior
(0.109)
2986
Hours of Labor (1000)
(6789)
'
Million Dollar Plant openings that were matched to the Census data and for which there were incumbent plants in
both winning and losing counties that are observed
date
is
in
each of the eight years prior to the opening date (the opening
defined as the earliest of the magazine reported year and the year observed in the SSEL.) This sample
restricted to include
matches for which there were incumbent plants
in the
is
then
Million Dollar Plant's 2-digit SIC in both
locations.
few of these differences are 3. Census confidentiality rules prevent being more specific.
original 47 cases, these statistics represent 28 cases. A few very large outlier plants were dropped so that
the mean would be more representative of the entire distribution (those dropped had output greater than half of their
county's previous output and sometimes much more). Of the remaining cases: most SSEL matches were found in
the ASM or CM but not exactly 5 years after the opening date; a couple of SSEL matches in the 2xxx-3xxx SICs
were never found in the ASM or CM\ and a couple of SSEL matches not found were in the 4xxx SICs. The MDP
characteristics are similar for cases identifying the effect within same 2-digit SIC. Standard deviations are reported
in parentheses. All monetary amounts are in 2006 US dollars.
^
Only
'
Of the
a
45
Table
2.
Summary'
Statistics for
Measures of Industry Linkages
Mean
1''
Only 4*
Standard
All Plants
Quartile
Quartile
Deviation
0.119
0.002
0.317
0.249
0.022
0.001
0.057
0.033
0.022
0.000
0.106
0.084
0.011
0.000
0.042
0.035
0.017
0.000
0.075
0.061
0.042
0.000
0.163
0.139
Only
Measure of
Description
Industry Linkage
Labor Market Pooling:
CPS Worker
Proportion of workers leaving a job
Transitions
in this
industry that
move
to the
MDP industry (15 months
Intellectual or
Technology
Citation pattern
later)
Spillovers:
Percentage of manufactured industry
patents that cite patents manufactured
Technology
MDP industry
R&D flows from MDP industry, as a
Input
percentage of all private sector
in
technological expenditures
Technology
Output
R&D flows to MDP industry, as a
percentage of all original research
expenditures
Proximity to Customers and Suppliers:
Manufacturing
Industry inputs from
Input
as a percentage
of
its
MDP industry,
manufacturing
inputs
MDP
Manufacturing
Industry output used by
Output
industry, as a percentage of
to
CPS Worker
its
output
manufacturers
was calculated from the frequency of worker industry movements in the rotating
is by Census Industry codes, matched to 2-digit SIC.
The last 5 measures of
cross-industry relationships were provided by Ellison, Glaeser, and Kerr (NBER Working Paper 13068). These
measures are defined in a 3-digit SIC by 3-digit SIC matrix, though much of the variation is at the 2-digit level. In
all cases, more positive values indicate a closer relationship between industries. Column
reports the mean value of
the measure for all incumbent plants matched to their respective MDP. Column 2 reports the mean for the lowest
25% and column 3 reports the mean for the highest 25%. Column 4 reports the standard deviation across all
observations. The sample of plants is all incumbent plants, as described for Table 1, for which each industry linkage
measure is available for the incumbent plant and its associated MDP. These statistics are calculated when weighting
by the incumbent plant's total value of shipments eight years prior to the MDP opening.
Notes:
CPS
survey groups.
Transitions
This variation
1
46
oo
1
^
-K
fS
o
r-^
d
VO
m
<>
<*1
SO
r-~
i/^
r-
t-~
-;
rn
o
d
m
00
CN
m
m
CO
d
weight
>
ox '5)
Co
a
'a
*^
r--
O
^
oo
-ef
d
OS
d
o
so
c-~
"*
5^
CO
d
m
cs
d
_:
(N
OS
d
in
oo
d
^S
en
CO
r-
•35
So"
-
^ U
o
d
o
m
oo
oo
O
d
so
d
O
d
m
d
r-
o
o
d
00
in
oo
d
(N
1
o
in
q.
^^
*
^
d
>>
(N
s
i>
i2
c C
en
a>
'S
oo
p
0\
in
a 1 ^U
c
o
d
O
o
d
d
oo
m
o
d
d
o
o
d
00
m
OS
(N
oo"
r<-i
r-
r-
OO
OS
d
JS
tca
3
T3
ca
o
en
en
3
O
T3
O
§
o"
>0
VO
o
d
in
o
o
d
OJ
o
so
d
in
o
d
so
OS
d
o
m
O
d
o
in
lO
On,
r-"
00
o
d
es
C/3
t?
'C
CU
1
^
S
wn
r-~
i/i
_;
^
in
rn
-*
r)
(S
rn
OS
Tf
rn
^
"1
<s
»s
T-*
rn
'er
so
d
M
q
tT
r-H
(S
rii
OS
d
m
Tl;
ri
—
en
ca
_
—
O
O
1
„
-35
1/)
o
rj-
rs
^
SB
s
«
*-
00
d
o
—
o
fS
5
—
oo
so
d
1/1
in
ri
^
m
d
in
rn
'-•
o
d
in
r-^
^
.^
D
^
^H
d
oo
d
^
c
^
iL
en
->
i-
^
=u -s
V3
S3
S
.
"^
-3 en
K O
r4
a
a
00
CN
11
o
d
oo
m
<N
oo
oo
rr,
O
d
d
O
d
>n
d
O
o
d
00
00
OS
rn
oo
in
>^
d
o
^00
r--
d
oo
g
-a
c g ^
o ° S
"
t"
"
^
o
00
~
^
1
vO
ON
o
d
oo__
>n
o
d
OS
r-
m
d
r-
o
d
in
d
O
d
in
so
in
rs|
oo
(N
OO
o
d
oo
so
in
-*
q
d
CN
(N
m
O
d
tn
s
C8
o!
B
S
O
lO
m
in
d
_
o
d
<t
'^
o
d
d
o
d
OS
00
od
q,
d"
OS
1
<N
OO
o
d
CO
o
in
'-2
-3
<N
.a
g
g
en
1
_«
1)
>
O
>
o
|u£
"o 3 H
,°
o
*U
(lT
00
00
ca
"3
o.
o
u
c
ea
U
n.
o
CL,
en
so
o
en
so
o
00
c
3
U
eS
3
3
C3
2
t_
c
>
_ca
o
oo
c
ca
U
Ui
L>
£3
>
1
en
en
_ca
Oh
—
C3
,
U
O
2S
1
CO
e^
O
%
u
S 3
J2
CL,
O
en
O
O
O
>
o
oT
en
S g
!i
13 ^
—
-S
§
U >
-a
S
O "S
"__
_3 u
>^ U O u
"
«
X
=
>
o
O
(U
00
ca
ea
c
U
^
— -o
U _
en
i|
3
e^
00
^
^
0-,
Q
en
=5
^
^
1-1
ID
en
(U
en
i2
u
-Q
li;
=5
en
CS
S
r-
(U
1-
ca
(U
3
^en E
c
O
S2
o;3 5 «
ID
oo
J2
*"*
O 1>
w O
- 1
so
00
S^
u
ca
u
ca
aj
S
a.
ca
en
o
c
en
(-1
SO
1
oT
c
2
c
o
tn
TJ
d
en
.2
C3
SO
"5
U
3
H
en
E-S
CO
-C
s
-4-*
1
00
U
^
2
g
3 V
't:;
c
S
C 3
ca
J= -3
V^l it
1!1
.s -^
-a
u
^
r-1
.a
D
.S
C
^C ^
en
CU
ll
^
3 " M
ca
en
.,
<u
u
^
o ^
jn f^
0)
r-
r7
en
ca
^
a
e
c
«
00-
I-
o
OV
r~
rt
o
so
d
1
iS
t:
o
d
IT)
g^S
1
oo
OS
so
in
qual the
-a
C 3 V
« E S
" § en
00^
^
E
00
im
<U
r
ca
u
tN
3
0^
oo
-digit
3
a.
O
„
o
winning
of
i_
(U
o^
county
between
se
y
K
SIC
their
S
gy
1)
c
^S
cr
i
-*
c3
(/:
«
3 a M
§.S S
O 3 ^ 8 g §
u u ca
-g 1 s
u
3
O
K
^1
^1^1
Table
Incumbent Plant Productivity, Relative
4.
Event Year
T
=
-7
T
=
-6
T
T
T
T
=
=
=
=
T
=
T
=
T=
T
Winning
In
T
(2)
(3)
0.067
0.040
0.027
(0.058)
(0.053)
(0.032)
0.047
0.028
0.018
(0.044)
(0.046)
(0.023)
0.041
0.021
0.020
(0.036)
(0.040)
(0.025)
-0.003
0.012
-0.015
(0.030)
(0.030)
(0.024)
-
0.011
-0.013
0.024
(0.022)
(0.022)
(0.021)
-3
-0.003
0.001
-0.005
(0.027)
(0.01 1)
(0.028)
-2
Year of a
MDP Opening
(l)-(2)
Counties
-l
0.013
-0.010
0.023
(0.018)
(0.01 1)
(0.019)
0.023
-0.028
0,051*
(0.026)
(0.024)
(0.023)
1
=2
0.004
-0.046
0.050
(0.036)
(0.046)
(0.033)
0.003
-0.073
0.076+
(0.047)
(0.057)
(0.043)
=4
=
to the
Difference
(1)
T=3
T
Losing
Counties
-5
-4
In
0.004
-0.072
0.076*
(0.053)
(0.062)
(0.033)
-0.023
-0.100
0.077*
(0.069)
(0.067)
(0.035)
5
R-squared
0.9861
Observations
28732
Notes; Standard errors are clustered
at
the natural log of output
is
regression:
machinery
the county level.
Columns
SIC fixed effects, plant fixed
in a winning or losing county
capital, materials), year x 2-digit
dummy variables for whether
When a plant is a winner or
the plant
loser
1
and 2 report coefficients from the same
regressed on the natural log of inputs
is
more than once,
it
receives a
(all
worker hours, building capital,
and the reported
effects, case fixed effects,
in
dummy
each year relative
to the
MDP
variable for each incident.
opening.
Plant-year
MDP opening. Data on
opening and 5 years after. Capital stocks were
calculated using the pennanent inventory method from early book values and subsequent investment. The sample of
observations are weighted by the plant's
plants in
all
cases
incumbent plants
significance at
5%
is
is
total
value of shipments eight years prior to the
only available 8 years prior to the
the
level,
MDP
- 2 of Table
same as in columns
+ denotes significance at 10% level.
1
48
3.
** denotes significance at
1%
level, *
denotes
Table
The
5.
Opening of a
Effect of the
MDP Plant on the Productivity of Incumbent Plants
MDP Winners MDP Losers
MDP Counties
MDPWiinners MDP Losers
(1)
(2)
(3)
0.0442+
0.0435+
(0.0235)
All Counties
Model
All Counties
Random
Winners
(4)
(5)
0.0524*
0.0477*
-0.0824**
(0.0225)
(0.0231)
(0.0177)
1
Mean
Shift
(0.0233)
[$170m]
0.9811
0.9812
0.9812
0.9860
0.9828
418064
418064
50842
28732
426853
0.1301*
0.1324*
0.1355**
0.1203*
-0.0559+
(0.0533)
(0.0529)
(0.0477)
(0.0517)
(0.0299)
R-squared
Observations
(plant X year)
Model
2:
Effect after 5 years
[$429m]
0.0277
0.0251
0.0255
0.0290
-0.0197
(0.0241)
(0.0221)
(0.0186)
(0.0210)
(0.0312)
Level Change
Trend Break
0.0171 +
0.0179*
0.0183*
0.0152+
-0.0060
(0.0091)
(0.0088)
(0.0078)
(0.0079)
(0.0072)
-0.0057
-0.0058
-0.0048
-0.0044
-0.0057**
(0.0046)
(0.0046)
(0.0046)
(0.0044)
(0.0029)
Pre-trend
R-squared
Observations
0.9811
0.9812
0.9813
0.9861
0.9828
418064
418064
50842
28732
426853
YES
YES
YES
YES
YES
YES
YES
YES
NO
All
All
All
(plant X year)
Plant
&
Ind-Year FEs
Case FEs
Years Included
Notes: The table reports results from the
fitting
-7
<
T
<
N/A
All
5
of several versions of equation
of the natural log of output on the natural log of inputs, year x 2-digit SIC fixed
effects.
In
before the
Model
MDP
1,
two additional
opening or
the average change in
TFP
dummy
Specifically, entries are
(8).
from a regression
and case fixed
effects, plant fixed effects,
variables are included for whether the plant
is in
a winning county 7 to
1
years
The reported mean shift indicates the difference in these two coefficients, i.e.,
opening. In Model 2, the same two dummy variables are included along with pre-
to 5 years after.
following the
and post-trend variables. The shift in level and trend are reported, along with the pre-trend and the total effect evaluated after
5 years. In columns (1), (2), and (5), the sample is composed of all manufacturing plants in the ASM that report data for 14
consecutive years, excluding
all
plants
owned by the MDP firm. In these models, additional control variables are included
= -7 through x = 5 (i.e., -20 to -8 and 6 to 17). Column (2) adds the case fixed
for the event years outside the range fi-om t
effects that equal
1
during the period that t ranges from -7 through
include only plants in counties that
fi"om plants in these counties.
won
5.
In
columns
(3)
and
(4),
the sample
is
restricted to
MDP. This forces the industry-year fixed effects to be estimated solely
are now required to be in the data only when the MDP opens and all 8 years
or lost a
Incumbent plants
prior (not also for 14 consecutive years, though this does not change the results).
For column
further to include only plant-year observations within the period of interest (where i ranges
(4), the
from
sample
-7 to 5).
is
restricted
This forces the
industry-year fixed effects to be estimated solely on plant by year observations that identify the parameters of interest.
In
were randomly chosen from the ASM in the same years and
industries as the MDP openings. For all regressions, plant-year observations are weighted by the plant's total value of
shipments eight years prior to the opening. Plants not in a winning or losing county are weighted by their total value of
shipments in that year. All plants fi-om two uncommon 2-digit SIC values were excluded so that estimated clustered
variance-covariance matrices would always be positive definite. In brackets is the value in 2006 US$ from the estimated
increase in productivity: the percent increase is multiplied by the total value of output for the affected incumbent plants in
column
(5),
a set of 47 plant openings in the entire country
the winning counties.
denotes significance
at
Standard error clustered
5%
level,
+
at
the county level in parenthesis.
denotes significance
at
1
0%
level.
49
** denotes significance
at
1%
level,
*
-^ "" c:
CO ro 02.
ts
O
U
—
o o
o o
I
a
cd
u
P.
1-
>
£^J
^
•a <^ -S !^
a-
^ oo
o o
o o
>-i
O
3
>> P.
OS ^O
O O
JD
C
crt
JS
C o
O ca
u
u:-°
a
tl-
I
J3
o
ii
c3
+1!
^
5- 12
'
'
^
l^
a
__
c
+
CM
a
OB
o
rt
a
3
u
8
O
I
ci
(-.
a
^
o
"
Z-
"^
^
o
S-
o
S
o
O
3
c/D
QJ
-C
a
o
a
«
ft
^^
is
oS
(—
01
ro
+j
00
as
O C
"^ .2
Oh
Q
ss°
o
T3
o
o
*
V
c^
6i3
§S
a
'a
03
(u
o
S o
2.
a.
O
g
(U
?
— 5 ^00 ->
^
O
00 -S
'^
u
u
c
SC
»:>
O)
^
H
a
_o
t3
O
1:5
_4J
3« ou
H a
(/3
Z H
C « -^
,
c/1
T3
^
C C
c
-5 .2
II
^u£
o
S
.S
i
:
I
'
:
,
cs
!
I
c
u
E
ON
as
tn
c3
u^
CD
"3
2^
B.
rt
U
>
d
£
3^s
1
c
CO
^
C
<"
e
"co
-g
c
2
—^
§.
t«
>
-S
QO
S
e
«
"
t3
=i
«
1=:
S
g
^
S
"
•-"
:
is
d
-a
>
S c
go
§
•
Oh
B
«
-2
0^
_
It
CO rj
u
"^ ^^ -^
d
S
S
5
a
sS
s
s
u
d
"^^
o «
I^
B I
S
Si
C
>
c
rt
.2
go
d £
u a
c
—;
c
a3o
2 o
u
.A
—
CO
>
u
go
d
c
o
s
CM
g P
o
II
o
0-1
Q
S
U
®
o
o
Da
.
^«
S
K «
Da
'
c -S
nog,
•ft
U
o o
c n
1)
I;-
-ffl
M
1^
d
8-
*
c^
o
S dS
s s
c
o ^
J3
00
C
CO
U
E
<
H
c
O
o
o
Id
Of
H
E .S
S
C
"O
a
a
u
O
it*
a
U
a
s
a
o
c
in
3
TS
U
-*
fL^
£ •« _
O £ ^
(U
,e
-«-t
^
o qS
2
o o
o^ 10
r<-l
IN
g§§S
O H (N
o
fee
a
SS2,
O
o
o
o
"
r—
o
o
rt
^
2
2
^
—
00
o
o
o
p'
^^'
S >^
g
o
H o
2 '^
j:J
r<-i
b
<^^
s
iS
&^
TD
(»
OJ
0^ 'C
Q
§
;3
C/3
'^
3 Q
C
« N^
iH
D£f
c
u
t:
^
'5)
-3
*
ON "^
00 00
10 ON
-Sf
S^
r^
&
ON
—
00 o ro
00 o o o R
o 000
r-~
'S-
'^j-
;:r
.^1
1
t-:
_c
t-?
aj
e
«
i>
r-
Z3
<
-*-*
§^
o
o S r^E
^ o2
2 sS
c rr
•^-^
p.
JS
C
On CO
—
s
O
,_ On
r-;
O
C\ CO
o
<~^
&0
O
ON
(N
o
O
+
O
— [^ ON
CIj O
o3 o
O O
CNl
(_;
3
^
S
P
P
^
^
O
o
O
^
00
<
u
te
U
3
C
T3
lu
C 3
03
-4^
|3c
H
00
-3
OD
C
eg
1
Si
c«
JU 0*
3Q
§
—
a
i
^
U
rt
r^
13
c
OJ
(u
Qi
O
^
cd
(U
^
CL
T3
oa
c
T3
^
>
c:
lU
«
(D
H
D^
J
;—
(U
cd
O o
<-)
«
Qri
rN)
en
U
U
S
en
a
D
—
'
^
-.2
"-'
J-^
o
^
r— VO
m
o (N
o
oo
VO 00
o
O O
o o
u-t
<N CN
,
o
1
V~l
,
1
cs
-*
oo
d d
'^
o
o
o
d
^
^
O
d
r<-)
rn ON
£r oo
Po
d
d
o
Ro
I—
s
*^
W-1
do
d
<N
O <D
oo
dd
c 3
a
Xi
E
3
u
S
O rn
—
o
oo
dd
o
w^
"ni
S
"H.
-^
t^^)
o
'>
-^^
—
O
°. d
o ^-
u
3
o
u
\0
ON CN
U-)
-O
OO
<u
JS
3
O
s
S o
d
t^
_2
Oh
Q
*
i/^
(N
On
—
o o
^
.•s
•nT
1/1
u
T3 N=
OX)
3
'3
t:
a,
*
— m
O
r- o
o
-;
O
CSl
v
u
r~-
e^
3
£^
tS
Is
3
-*
S
ex
a,
o
«°
OS
S
^
o
c
o
U
00
CI
t
o
00
c
o
"o
3
o
u
H
=^
«
O
ir;
~
t;
"^
o
e
—o -c ^
- o
c/1
OJ
S B 6
I
.-t:^
-a
cd
+
,o
C3
C
O
O
c
ca
o
>OJ
"=
(U
-
Fi
"W
c/:
M^^
a
S
O
o S
cd
-w
"^
^
'
PQ
S-7
S2
•
u
^g-^
KJ
(N
03
^
« " 5
D.
o
o
>
T3
Dh
U- _«
Q
^
ca
m ID
o o
3
R
„
K>
K
3
«
1^
.2
75
t/3
S «
ca
en
>;
« b '^^
<1J
a
to
a
TO
J:^
i>
.
E
'
-n
3
>O
ca
D.
03
o
Q H00
o
J
C
o
^
in o
^
— ON
o
o d
rs
^
-
xi
g 5
O C
u
^ ^
(N
^
—
5!
O
p
3 g
P
ca
'S
3
-5
^
f- -S
.S
J,
ca
-a
t>
o
^
-^
00
en
S
o
o o
Cu
P g-.s
-5
2
o =
(D
'"^
"r;
.a -a
^
*
ci.j=
gx«(up-ca>.5
!"
>.
?
I-
£
-a -a
-^g-'^iis?
w
"
--
U
_
J3
J3^
00
c/i
o a « " ^
Q
v_>
ca
o
E.^5
§^° g^^cgv s
u
e
>
5
'S
OJ
«
<u
3*^ §
oos £
J£ o u S S -o g
-§
li5-^-C
^^ KH3
ca
§ §
en
S£
ca
S
,,
ca
S
^^
H
Jg'g
g
I
•S
0)
«.
ca
o
-
ca
OJ)
1^
B
g
D.
a.
g
"^
O
o
ON cs
1)
c-uCopg.u.S>u_
<u
l~^
>
(L>
S
-
0)
Ph
"^
Sf
.a
-^
"^
^X!
i2
t«
•a
^-i
?^
G « °
,tC « ^' C/2
S
'^
g u o
>
c
o
c
K!
^-5
OJ
1-
-t3
OJ
J3
1-'
o
a
o
a
w
„ ^
o
O
U5
a
O .« —
o o
o 3 a
P o
S 00 c
D-
,
° -^ .2P
cc„^.™^
6
g
g^ o2^u ^ig
12
0)
g5iio||ef
^^a.?5Ptap
m C ^
c *- t; u
fa
_ Q
Qcag-Jgu*:s^g-3^
o
^ en — q K-Ct3
(U
00. > "S
'
a C "T^
O O 00
g "
g-op.S oS2
tc d 5 to o
is
'TO
"^
f—
.
'
.023-^3
?P^
a- D-.ci
<u
rt
-t-*
o
^
o
L=
c2 c^
ta
,ca
ca
(D
o
caiLi5'^5ca'^n=^
Xi
I
Cci
-
en
O Zo S
2
OJ
-S
"
.-on *
S
'S
uh
Q
r-."
i*
ea
°°S
5 8
C ca
'^
(U
^
§^^ 2
v^
^
C
^;
.5
3
en
.^
g-~
£
=2
3 o
t
-^
ca
q3
(u
>
-o
ca
-d
uej
u -a
00:5
co^J
00
S
-2 .S
^ .S
2
1^
iS
ij
^
.2
~
B
u
s
o
a
a
OO (N
o
o m
o
o
o o
r-~
5u
au
OO ON
oo
^^
S
'S -^
B
^
c3^
H o
9 d
3
*
s
(N
—
— O
m
ON o
O
t3 en
5
> U (U
U .S O
E -2 C
a
C3
O E
O o
C
e
C
o
a
«
^
I—
C3
G.-*
^^
c
^u
CL,
Q
ca
in
—
o OO
^
o o
d d
'
i3
Cl
«
r-i
"
U
•0
c V
E u
_3
^
(N ON
0£
S
B
u~,
o o
d
a.
S
d
^
^
"2
>
H.
C3
^^
01
3
E-
3
O
+
C
—
rn
o
00
5
u.
_3
ca
0
-3
>^
<
es
H
S£
>
e s
e
l-H
s
c
u
£^ <£
ra
ivo
E?
c u
00
q
d d
cd
J3
00
^
u 1)
u
u c C/2
(U
i-
d_
o
'3
p
^
t;
l4-
N >
u
.si
w
c/^
'e
T3
03
B M
9
o
-a
s
^
0)
a
c
•a
ra
> c
-a
.a
=^
^
i^
30*
•a
0
g
-a
'^
In
3 E 3 c
Z
-a
/
If
HT^iiUiliHHhMlhiwihhakttimhtliUti^trdiiA^t