Professional Documents
Culture Documents
Industrial Subsidies and Technology Adoption in General Equilibrium.
Industrial Subsidies and Technology Adoption in General Equilibrium.
Industrial Subsidies and Technology Adoption in General Equilibrium.
Abstract
Industrial subsidies to failing establishments are common across developed economies. The
paper constructs a dynamic general equilibrium model with a view to study the effects of this
policy. Interestingly, subsidies to failing plants may increase productivity and accelerate the
diffusion of new technologies. In spite of this, labor productivity, employment and income
decrease, as resources are devoted to maintaining and updating establishments that would
otherwise have closed.
r 2006 Elsevier B.V. All rights reserved.
Keywords: Industrial subsidies; Investment-specific technical change; Technology adoption; Plant lifecycle;
Failing plant
1. Introduction
0165-1889/$ - see front matter r 2006 Elsevier B.V. All rights reserved.
doi:10.1016/j.jedc.2005.09.016
ARTICLE IN PRESS
1590 R.M. Samaniego / Journal of Economic Dynamics & Control 30 (2006) 1589–1614
other innovative activities, more often than not they are directed towards establish-
ments that are unproductive.
The paper asks what are the quantitative effects of industrial subsidies to failing
establishments. This question is of interest for several reasons. First, such policies
may have a significant effect on aggregate productivity via changes to the
composition of the establishment pool. Second, an extensive literature examines
the aggregate effects of labor market regulations, particularly those that impose
penalties upon establishments that are reducing their workforce. This is partly
because differences in institutional firing costs have been pinpointed as a likely cause
of the divergent labor market outcomes of the US and Europe.1 However, several
authors have observed that Western European plants that were failing and needed to
reduce their payrolls in recent decades were less likely to be taxed than subsidized.2
Consequently, the effect of industrial support upon employment is of particular
interest. Third, a policy that directly affects the establishment lifecycle could have
significant implications for plant dynamics. Hazard rates, job flows and patterns of
technological adoption are just some features that could be affected.
In the paper, I develop a general equilibrium model of establishment dynamics.
Surviving units gradually fall behind the best practice technology, and may choose in
each period whether to upgrade, continue dropping behind, or shut down. Into this
environment I introduce industrial support to failing establishments.
Interestingly, subsidies have the effect of increasing the average productivity of
plants in operation. This is because an establishment’s optimal technology adoption
rule follows an ðS; sÞ policy, censored by the decision to exit in the face of sufficiently
adverse conditions. As a result, subsidies that enable plants to survive longer allow
more of them to enter the stage of their life at which renewing their technology
becomes optimal. Thus, the underlying determinants of technological adoption are
an important part of the response of the economy to industrial support. Nonetheless,
the economy spends a lot of resources on keeping alive plants that would otherwise
have shut down, and this results in a reduction in both output and employment on
the aggregate.
That the details of the plant lifecycle might be related to the aggregate effects of
public finance regimes has not been raised in the literature. Fuest and Huber (2000)
and Restuccia and Rogerson (2004) study the effects of industrial subsidies: however,
their models lack any lifecycle dynamics, and their subsidies are not directed towards
failing plants – indeed there is no notion of a failing plant in those models.
Samaniego (2006a) articulates such a notion to study the effects of firing costs upon
exit, but does not consider industrial subsidies nor technology adoption.
Section 2 provides an overview of industrial subsidies, and Section 3 introduces
the theoretical model. Section 4 characterizes the equilibrium behavior of the model,
and Section 5 outlines the calibration procedure. Section 6 studies the effects of
industrial support in the context of the model.
1
See Bentolila and Bertola (1990) and Lazear (1990), inter alia.
2
See Ford and Suyker (1990), Leonard and Van Audenrode (1993), OECD (1996), Murphy and
Pretschker (1998) and Fuest and Huber (2000).
ARTICLE IN PRESS
R.M. Samaniego / Journal of Economic Dynamics & Control 30 (2006) 1589–1614 1591
2. Industrial subsidies
Fig. 1. Industrial subsidies across the OECD, 1970–1988. Source: Ford and Suyker (1990).
8.0
7.0
6.0
5.0
% of GDP
4.0
3.0
2.0
1.0
0.0
ly
ay
nd
en d
en
es
he m
Be a
Po d
ey
al
nd
m
ria
s
Au n
ce
e
k
an
nd
an
an
ar
ai
ad
pa
li
an
ec
ur
Ita
Lu rtug
ew giu
do
ra
w
at
rk
an
rla
ed
st
la
Sp
bo
rla
m
re
al
nl
Ja
an
el
or
St
Tu
st
Au
Ire
ng
l
Fr
ite
Sw
Ze
Fi
er
Ic
G
m
N
C
Ki
d
Sw
xe
te
et
d
ni
N
te
N
U
ni
U
Fig. 2. Industrial subsidies by country, 1970–1988. Source: Ford and Suyker (1990).
ARTICLE IN PRESS
R.M. Samaniego / Journal of Economic Dynamics & Control 30 (2006) 1589–1614 1593
3. Model economy
In what follows, I describe the structure of the model economy. The model is
related to that of Hopenhayn (1992). The innovations are the inclusion of
investment-specific technical change, and the approach to plant exit, which is
similar to Samaniego (2006a). This endows the model with the notion of a ‘failing
plant’ which, as argued, is central to the question of industrial subsidies.
3.1. Production
3.1.1. Establishments
There is a continuum of heterogeneous plants of endogenous mass that live in
discrete time, discounting the future at rate i. At any point in time t a plant is
characterized by the vintage of its technology v, and by an idiosyncratic productivity
shock zt 2 ½z; z̄. It produces a quantity of a homogeneous good according to the
production function
a
gtz zt kt k nat n , (1)
where nt is labor input and kt is capital, for which it pays a wage wt and a rental rate
rv;t , respectively. gz is an exogenous productivity growth factor. An establishment’s
future shock ztþ1 is drawn from a distribution f ðztþ1 jzt Þ.
Vintage v does not directly enter the production function. Rather, v denotes the
type of capital that the establishment uses. Let iv;t and K v;t denote aggregate
investment and capital of type v, respectively. The aggregate stock of each type of
capital evolves as follows:
K v;tþ1 ¼ ð1 dÞK v;t þ qv iv;t . (2)
Here, qv ¼ gvq
is an index of investment-specific technical change that differs across
vintages. In equilibrium, different vintages of capital will command different interest
rates, and this is how vintage matters for establishment payoffs.
Again, a failing plant is one that has drawn a continuation shock f and for whom
X ðf; t; zt Þ ¼ 1.
Instead of closing, failing establishments receive a transfer Tðf; t; zt Þ that covers
their losses up to a certain point:
8
<0
> if ð1 yÞfwtþ1 oCðt; zÞ;
Tðf; t; zt Þ ¼ ð1 yÞfwtþ1 Cðt; zÞ if Cðt; zÞpð1 yÞfwtþ1 pð1 þ xÞCðt; zÞ;
>
:0 if ð1 yÞfwtþ1 4ð1 þ xÞCðt; zÞ:
ARTICLE IN PRESS
R.M. Samaniego / Journal of Economic Dynamics & Control 30 (2006) 1589–1614 1595
Any shortfall between the value of the firm and a large continuation shock is covered
by the subsidy, up to a proportion x of the continuation value. If a plant is faced with
a continuation shock ð1 yÞfwtþ1 4ð1 þ xÞCðt; zÞ, transfers would be insufficient to
avoid exit, so none are made.
Given x, the tax rate y is set to satisfy a balanced budget condition:
ZZ
Tðf; t; zt Þ dmt ðt; zt Þ dFðfÞ ¼ yPðmt Þ, (3)
where Pðmt Þ is total pre-tax profits when the measure is mt . That all transfers are
between plants in keeping with Leonard and Van Audenrode (1993) who note that, in
most Western European countries, net transfers to corporations were close to zero.
Section 6 explores the sensitivity of the results to this assumption.
3.2. Households
4. Model solution
4.1. Equilibrium
W ðt; zt Þ ¼ ð1 yÞPðt; zt Þ
Z
gy
þ maxf0; Cðt; zt Þ ð1 yÞfwtþ1 g dFðfÞ. ð9Þ
1þi
It will be useful to denote U as the optimal updating rule; t^ as the target technology
age; and X as the plants’ optimal shut-down rule:
Uðt; zt Þ ¼ arg max yUðzt Þ þ ð1 yÞEstþ1 W ðt þ 1; ztþ1 Þ,
y2f0;1g
0
t^ ðzt Þ ¼ arg max
0
fEztþ1 W ðt0 ; ztþ1 Þ ð1 yÞkgð1a
q
z Þt
g,
t X0
subject to the constraints in (5), the laws of motion for capital (2) and the law of
motion for the measure,5 where once more xt ¼ ðfK t;t g1 t¼0 ; mt Þ. Problem (10) is
standard except for the presence of entry and of multiple types of capital. Optimal
entry requires that the following equation holds, with equality if mt 40 for any given
t:
Z
W ðt; zÞ dcðzÞpzgqð1az Þt for all tX0. (11)
Let h equal labor supplied by the household, and let n ðt; zÞ be the employment
policy of a given type given the wage. Also let k ðt; zÞ be optimal capital use.
Definition. A stationary equilibrium is a value function W for the plant and V for
the household; exit and updating rules X , U and t ; employment and capital rental
rules n and k ; household decision rules regarding consumption c , labor h , and
entry fmt g; prices w and frt g; an aggregate state x ¼ ðfK t g; m Þ, a law of motion for
the state G and a tax rate y that satisfy:
t¼0 t¼0
Z
ðt;zÞ
þ kgqð1az Þt U ðt; zÞ dm . ð12Þ
5. b ¼ gy =ð1 þ iÞ.
6. Government budget balance: Eq. (3) is satisfied.
The following statements characterize the structure of exit and updating in the steady
state of this economy. They will prove useful in deriving the equilibrium transition
function G, and in forming an algorithm to compute the steady-state measure.6
6
Proofs, located in the Appendix, hinge upon a demonstration that the model of plant dynamics is
equivalent to a vintage capital model in equilibrium, net of a change of variables. While this result has been
ARTICLE IN PRESS
1598 R.M. Samaniego / Journal of Economic Dynamics & Control 30 (2006) 1589–1614
Fig. 3. Optimal decision rules in the calibrated economy. Contour plots are the hazard rates, whereas the
black line represents the updating threshold.
Proposition 1. If a plant of type ðt; zÞ updates, then one with ðt0 ; zÞ, t0 4t does also:
U ðt; zÞ is increasing in t.
Proposition 2. t ðt; zÞ ¼ 0 for all ðt; zÞ: all updating is to the frontier, and all entrants
adopt the frontier technology.
(footnote continued)
widely hinted at in the literature by the fact that the terms ‘capital-embodied technical change’ and
‘investment-specific technical change’ are sometimes used interchangeably, I am not aware of a
demonstration of their ‘equivalence.’
7
See Milgrom and Roberts (1990), Cooper et al. (1993), Cooley et al. (1997) and Sakellaris (2004).
ARTICLE IN PRESS
R.M. Samaniego / Journal of Economic Dynamics & Control 30 (2006) 1589–1614 1599
it updates its technology, returning to the vertical axis but still subject to f in terms of
the evolution of its idiosyncratic productivity.
The following assumption on f yields additional results.
RZ
Assumption 1. 1 f ðztþ1 jzt Þ dztþ1 is weakly decreasing in zt 8Z.
Proposition 3. U ðt; zÞ is increasing in z: if a plant of type ðt; zÞ prefers to update, one of
type ðt; z0 Þ does so too for z0 4z.
Proposition 4. The exit rule X ðf; t; zÞ is increasing in t, and decreasing in z.
5. Calibration
I calibrate the model to post-war US data on aggregates, as well as job flows and
plant dynamics, broadly following the procedure of Kydland and Prescott (1982). I
use US data for the following reasons. First, industrial subsidies have been
comparatively rare in the US, so that it serves a useful benchmark. Second, although
such data are available for some other European countries, the most comprehensive
data on job flows and investment patterns are for the US. Third, the related
literature on labor market regulation typically uses US data for purposes of
calibration. I set x ¼ y ¼ 0 in the benchmark economy.
Shocks z were taken over a grid of 20 points. Given a particular grid, multiplying
it by any factor affects only the size and not the relative composition of the economy.
I choose values evenly spaced over an interval ½z; 1.
The functional forms for the primitives of the model are as follows:
The process f is specified as a random walk over the grid of z values with normal
disturbances: ln ztþ1 ¼ ln zt þ tþ1 , where t Nð0; s2 Þ. Since there are boundaries
on zt , the normal distribution of t is truncated appropriately. This form of f
satisfies Assumption 1.
The distribution c over initial shocks is set to decline linearly starting at z until it
reaches zero at a point c̄. This choice reflects the fact that plants are generally
born smaller than average. As a result, in combination, f and c can be interpreted
as a stochastic learning process.
Following Hansen (1985) and Rogerson (1988), I set Lðl t Þ ¼ ll t . This structure
identifies total labor input with total employment.
Given F, one can solve the value function and obtain a probability Dðzt ; tÞ
that a given type of plant is faced with a shock that is too large for it to survive.
I proceed in reverse: I choose a simple functional form for D that depends
on a reduced number of parameters, compute value functions, and then
infer the form of F that yields the conjectured death function D. This greatly
simplifies the task of calibrating the model. In particular, I choose the simple
linear form
Dðzt ; tÞ ¼ minfmaxfo1 o2 log Cðzt ; tÞ; 0g; 1g.
ARTICLE IN PRESS
1600 R.M. Samaniego / Journal of Economic Dynamics & Control 30 (2006) 1589–1614
which does not depend on type and hence is not correlated with size. This is
consistent with the findings of Leonard and Van Audenrode (1993).9
Thus, the parameters to be matched are gq ; gz ; ak ; an ; l; d; z; o1 ; o2 ; z; c̄; s
and k.
Using data on the relative price of capital, Cummins and Violante (2002) find
that gq ¼ 1:026. On the other hand, post-war economic growth per head in
the US was about 1.02%. I choose gz so that the growth rate of output in the
economy matches this statistic given gq . This implies that investment-specific
technical change is responsible for about 60% of economic growth, as in Greenwood
et al. (1997).
The discount factor is chosen to be consistent with a 7% annual interest rate as is
common in the real business cycle literature. The income shares are set to ak ¼ 0:3
and an ¼ 0:6, which leaves 10% of GDP as profits. This is approximately the
figure obtained by adding equity, interest and proprietary income in the US National
Income and Product Accounts. The physical depreciation rate d can be derived
from the Euler equation: using the data in Greenwood et al. (1997), I find that
d ¼ 3:9%.
The disutility of labor l is chosen so that employment is 60%, which is roughly the
employment-to-population ratio reported in Rogerson (2002). In turn, I set the cost
of entry z to match the average plant size in the Longitudinal Research Database, as
reported in Hopenhayn and Rogerson (1993).
The remaining six parameters are o1 ; o2 ; z; c̄; s and k, which determine the
dynamics of surviving plants and the distribution of shut-downs. The six statistics
8
With this form for D, the corresponding form for F is
8
>
> 0 xpeðo1 1Þ=o2 ;
<R o2
x
FðxÞ ¼ eðo1 1Þ=o2
dz eðo1 1Þ=o2 oxpeo1 =o2 ;
>
> z
:
1 x4eo1 =o2 :
9
When x is very large, the most productive firms may be such that x is sufficient to insure that they never
exit. In this case, the probability of support will decline slightly with size above a certain threshold.
However, this only occurs for very large values of x.
ARTICLE IN PRESS
R.M. Samaniego / Journal of Economic Dynamics & Control 30 (2006) 1589–1614 1601
Table 1
Parameter values
Parameter Value
gq 1.026
gz 1.007
ak 0.3
an 0.6
o1 0.11
o2 0.044
l 1.254
b 0.954
d 0.039
z 0.42
c̄ 0.73
s 0.023
z 12.2
k 6
I match are:10
The rationale for matching these particular statistics is broadly as follows. o1 and
o2 should be closely related to the extent of shutdowns and to the magnitude of job
turnover that they account for. The size of the young and the amount of job turnover
they generate are connected the initial distribution of shocks, and hence to z and c̄ as
well as F. In turn, s affects the amount of overall job turnover, while s will also be
related to cross cohort differences in behavior (plant closures in particular) since it is
linked to the rate at which the idiosyncratic productivity of young establishments
spreads away from the initial distribution. Finally, k determines incentives to update,
and hence should be related to the lumpiness of capital.
Table 1 lists the resulting parameter values, and Table 2 displays some steady-state
statistics of the model economy. The matches are quite tight. Observe that k is about
10
The algorithm I use is simulated annealing: see Bertsimas and Tsitsilkis (1993). Data are drawn from
Davis and Haltiwanger (1992), Dunne et al. (1989), Evans (1987), Doms and Dunne (1998) and the US
National Income and Product Accounts.
ARTICLE IN PRESS
1602 R.M. Samaniego / Journal of Economic Dynamics & Control 30 (2006) 1589–1614
Table 2
Matched statistics
6. Policy experiments
In this Section 1 compare the behavior of the benchmark economy to one in which
y and x are non-zero. Given a choice of x, I select y so the balanced budget condition
is satisfied. I do this for a variety of levels of x, covering the empirically relevant
range of 0–3% of GDP. This corresponds to values of x between zero and 70%. In
turn, this involves corporate tax rates y ranging from zero up to 30%, which is
significant but not excessive.
6.1. Results
The results are quite striking. Fig. 4 shows that subsidies lead to a decrease in
employment over most of the empirically reasonable range of subsidies. This is in
spite of the fact that the direct effect of the subsidies – to cover more continuation
payments – is to increase employment. Only when subsidies are very large does this
ARTICLE IN PRESS
R.M. Samaniego / Journal of Economic Dynamics & Control 30 (2006) 1589–1614 1603
80 4
% Drop in Consumption
60 3
Subsidy ξ
40 2
20 1
0 0
0 1 2 3 0 1 2 3
61 6.5
Continuation Emp, %
60 6
Employment, %
59 5.5
58 5
57 4.5
0 1 2 3 0 1 2 3
6.5 15
Average z, % Increase
Average age τ
6 10
5.5 5
5 0
0 1 2 3 0 1 2 3
Subsidies, % of GDP Subsidies, % of GDP
direct effect begin to dominate. At the same time, consumption is eroded by over 3%
of its steady-state value.
These aggregate effects are due to a substantial decrease in overall labor
productivity. At the same time, it is not the case that plant productivity decreases. In
fact, Fig. 4 shows that both the quality of the average technology in use and the
average idiosyncratic shock are higher under industrial support. As a result, the
average plant size increases substantially, as seen in Fig. 5. However, the average
number of plants decreases commensurately. Resources that are devoted towards
ARTICLE IN PRESS
1604 R.M. Samaniego / Journal of Economic Dynamics & Control 30 (2006) 1589–1614
200 80
60
150
40
100
20
50 0
0 1 2 3 0 1 2 3
40 40
Failure Rate, %
Exit rate, %
30 30
20 20
10 10
0 1 2 3 0 1 2 3
Average Age upon Updating
10.5 4.5
Job creation rate, %
10 4
9.5 3.5
9 3
0 1 2 3 0 1 2 3
Subsidies, % of GDP Subsidies, % of GDP
Fig. 5. Effect of industrial subsidies on plant dynamics.
keeping plants in operation crowd out those used to create startups. Since startups
are typically small, this affects the size distribution.
That subsidies to failing firms might encourage the use of newer technologies may
appear surprising, considering that plant types have an equal probability of being
subsidized regardless of type. The reason has to do with the details of the lifecycle
dynamics. Recall that plants follow an ðS; sÞ updating policy, censored by their
optimal exit rules. The older a plant’s technology, the more likely it is to enter the
updating area of the type-space, while at the same time being more likely to fail.
ARTICLE IN PRESS
R.M. Samaniego / Journal of Economic Dynamics & Control 30 (2006) 1589–1614 1605
30 10.5
Job Creation from Birth, %
Job dest. from exit, %
20 10
10 9.5
0 9
0 1 2 3 0 1 2 3
Subsidies, % of GDP
16 32
Lumpy Investment, %
Lumpy investors, %
14
30
12
28
10
8 26
0 1 2 3 0 1 2 3
Subsidies, % of GDP Subsidies, % of GDP
there appears to be a trade-off between the duration of jobs and the total number of
jobs – see Hopenhayn and Rogerson (1993). This supports the proposition by
Leonard and Van Audenrode (1993) that industrial subsidies – and possibly other
forms of product market regulation – stand as a candidate explanation for the
phenomenon of ‘eurosclerosis’ as much as do firing costs and other forms of labor
market regulation.
Patterns of investment also change under the policy, with the proportion of
investors who undergo a comparatively large adjustment in each period increasing.
Again, this is because industrial support targets the old, who are closer to the region
of the type-space in which updating is optimal. However, the proportion of plant-
level investment accounted for by these episodes decreases over most of the range of
policies considered. The fact that plants are of lower t and higher z on average under
the policy implies that the average updating lag decreases under the policy, following
the same pattern as the proportion of investment that is lumpy. Plants update more
often and, as a result, each episode of updating involves a comparatively smaller
adjustment.
6.2. Robustness
Could it be that the financing scheme, rather than the subsidy scheme, is
responsible for these results? To answer this question, I repeated the experiments
under two alternative scenarios. Under one, the subsidy scheme is as before, but it is
financed via a lump sum tax on households ðx40; y ¼ 0Þ. In another, there are no
industrial subsidies: instead, profits are taxed as before, and the proceeds are
redistributed to households ðx ¼ 0; y40Þ. Fig. 7 reports the results. On their own,
profit taxes turn out to have only a weak effect on job turnover, plant turnover and
investment behavior. It is the subsidy scheme that has by far the greatest influence on
the structure of the model economy.
It is interesting to observe that both schemes contribute to the decrease in
employment. This is because the profit tax decreases the rewards to entry, which
suppresses employment. There is also a direct effect: because of decreasing returns to
scale in the production function, higher taxes decrease the optimal scale of
production at incumbents. While average size does increase with the tax in general
equilibrium, this direct effect is offsetting so that the increase in average size is
considerably less pronounced than in Fig. 5. At the same time, the effect of the
subsidy scheme on employment is stronger than that of the tax scheme over most of
the empirically relevant range.
The paper proposes a model of plant dynamics to analyze the effects of policies
that affect establishments differently depending on their stage of their lifecycle. The
most empirically notable policy of this class is industrial subsides to failing
establishments. I find that the details of the lifecycle – and patterns of technology
adoption in particular – are important for the aggregate effects of such policies.
ARTICLE IN PRESS
R.M. Samaniego / Journal of Economic Dynamics & Control 30 (2006) 1589–1614 1607
Fig. 7. Comparison of the model with industrial subsidies to failing plants, financed by lump sum taxes
(full line), and the model with profit taxes that are redistributed to households (dotted line).
The results suggest that the quantitative analysis of policies that directly affect
establishment dynamics should be pursued in an environment in which these
dynamics are rich. Technology adoption is one factor that allows these dynamics to
be endogenous, and not simply the results of an exogenous stochastic process. The
endogenous response of plant dynamics to policy is shown here to be a potentially
important channel for their aggregate effects.
The present model is put forward as a benchmark for the aggregate effects of
industrial subsidies. The model economy does not contain an ‘upside’ for industrial
subsidies: however, there are reasons why the current results are likely to be robust to
extensions that do have this feature.
ARTICLE IN PRESS
1608 R.M. Samaniego / Journal of Economic Dynamics & Control 30 (2006) 1589–1614
First, the policy literature often cites credit constraints as a rationale for the
existence of industrial support. However, the empirical and theoretical literature on
credit constraints identifies the most constrained with young establishments, whereas
subsidies tend to be directed towards the old. As a result, the present framework
should capture the main effects of industrial subsidies of empirically relevant forms.11
Second, Fuest and Huber (2000) suggest that, if unions appropriate a large portion
of rents, entry may be suboptimally low, and that transfers to firms may ameliorate
this inefficiency. This channel too is unlikely to be of quantitative importance,
however. As noted, net transfers to industry seem to be about zero, so industrial
support does not constitute a net transfer from households. Consequently, any
effects of industrial support will not have to do with redistribution to industry but
rather between firms – which involves paying close attention to the determinants of
their heterogeneity, as here. Moreover, this redistribution would have to have a large
effects on payoffs close to birth to influence entry behavior. Although hazard rates
do decrease with age, empirically this difference is not large.12
In this paper I have concentrated on industrial subsidies to failing firms. An aspect
of industrial policy that the paper has not addressed, however, is the fact that a
significant portion of industrial support in certain countries may be directed towards
particular industries that may be ‘failing.’ In related work on labor market rigidities,
Samaniego (2006b) shows that, when the rate of technical change varies across
industries, policy may affect the equilibrium industry structure of the aggregate
economy, arguing that this channel has the unusual property that it may affect long
run rates of economic growth. Industrial subsidies that target particular industries
should clearly have this effect, and it would be interesting to extend the model in
order to quantify the potential effects of targeted industrial subsidies upon growth
rates – especially to the extent that political considerations direct subsidies towards
sectors that are ‘failing.’
Finally, the model is also suitable for thinking about other aspects of industrial
policy. In particular, an endogenous growth extension along the lines of Krusell
(1998) would yield a natural environment for the analysis of an aspect of industrial
policy that I have omitted here: subsidies to research and development.
Acknowledgment
Appendix A. Proofs
In this Appendix I report proofs of the propositions stated in the text and discuss
how they are used to compute the stationary equilibrium for a given parameter profile.
Without loss of generality in the following proofs I set y ¼ 0. This is because the tax
rate only affects the value function by multiplying all costs and benefits by ð1 yÞ.
First, observe that the steady-state Euler equation implies that
1
rt ¼ 1 þ d gtq
b
so that capital is relatively more expensive to rent for plants with older technology.
Solve for the optimal kt and define the following:
1=ð1ak Þ
s t ¼ zt ,
a =ð1ak Þ 1=ð1ak Þ
C ¼ ðakk ak Þrð0Þak =ð1ak Þ ,
an
a¼ ,
1 ak
gE ¼ gaqk =1ak .
With this change of variables, the value function can be specified as follows:
Cðt; st Þ ¼ maxfUðst Þ; Estþ1 W ðt þ 1; stþ1 Þg, (14)
0
Uðst Þ ¼ max fEstþ1 W ðt0 ; stþ1 Þ kgt
E g, (15)
0 t pt
a
W ðt; st Þ ¼ max Cst gt
E n wt n
n
Z
þb maxf0; Cðt; st Þ fwtþ1 g dFðfÞ . ð16Þ
Let U be the optimal updating rule and X be the plants’ optimal shut-down rule,
defined as follows:
Uðt; st Þ ¼ arg max yUðst Þ þ ð1 yÞEstþ1 W ðt þ 1; stþ1 Þ,
y2f0;1g
0
Uðst ; vÞ ¼ max fEztþ1 vðt0 ; stþ1 Þ kgt
E g, (18)
0t pt
a
Bvðt; st ; vÞ ¼ max Cgt
E st nt wn
n
Z
þb max 0; Cðt; st ; vÞ fwtþ1 dFðfÞ . ð19Þ
Let U be the optimal updating rule and X be the plants’ optimal exit rule. Then,
Uðt; st ; vÞ ¼ arg max yUðst ; vÞ þ ð1 yÞEstþ1 vðt þ 1; st Þ,
y2f0;1g
Assume that w and p are positive and finite. The following results are useful:
Lemma 5. The plant’s value function exists, is unique, is strictly decreasing and strictly
convex in plant age t.
Lemma 6. Under Assumption 1, W is strictly increasing in z.
Proof of Lemma 5. It is straightforward to show that the recursive system described
above satisfies Blackwell’s sufficient conditions for being the unique fixed point of
the Bellman operator defined over the appropriate function. By the contraction
mapping theorem, for any O that is compact under the uniform norm, if v 2 O and
Bv 2 O then W 2 O. This applies to the claim that W is weakly decreasing, as well as
to the claim of convexity.13 In addition, considering an open set O O, if Bv 2 O for
any v 2 O =O (which is closed), then W 2 O (as the limit cannot be in O =O). This
applies to the claims of strictness. &
Proof of Lemma 6. Assume that v is increasing in z. Then, Estþ1 vð; st Þ – the value of
not updating – is increasing in s also – as is U – because f ðjzÞ is increasing in s.
Consequently, Bv is also, as the instantaneous payoff is strictly increasing in s.
Arguments identical to those in Lemma 1 yield strictness. &
Proof of Proposition 1. Observe that U does not depend on t. Hence, since the
alternative to updating is decreasing in t, the updating rule must be increasing. &
13
Consider v decreasing in t. Then, U and C are either constant in or decreasing in t, as is the rest of Bv.
The same applies to the claim of convexity, since the supremum of any set of convex functions must itself
be convex.
ARTICLE IN PRESS
R.M. Samaniego / Journal of Economic Dynamics & Control 30 (2006) 1589–1614 1611
This defines a contraction as before. Let W ~ be the fixed point. It is easily shown
that
R the fixed point is increasing and convex in x, since 1=ð1 aÞ41. Thus,
W~ ðx0 ; e
sÞ df ðe s kx0 must be strictly convex in x0 , as W
sjsÞ de ~ ðx0 ; e
sÞ df ðe
sjsÞ de
s is the
weighted sum of strictly convex functions. Consequently,
Z
max W ðx0 ; e
sÞ df ðe s kx0
sjsÞ de
0px p10
Taking W as given, this expression can be rewritten for D in the form of Eq. (19) to
define a contraction that satisfies Blackwell’s conditions and which can be
shown to be increasing in s because both W and the static labor maximization
problem are. &
ARTICLE IN PRESS
1612 R.M. Samaniego / Journal of Economic Dynamics & Control 30 (2006) 1589–1614
Now I describe the definition and computation of the steady-state measure m. Let
Ið:Þ be the indicator function that equals one if the argument is true and zero
otherwise. The transition function G that describes how the measure mt evolves over
time must satisfy the condition that
Z
mtþ1 ðT; ZÞ ¼ M t Iðztþ1 2 ZÞIð0 2 TÞ dcðztþ1 Þ
ZZ
þ Iðt þ 1 2 TÞIðztþ1 2 ZÞð1 Uðt; zÞÞ
where c1 ¼ c. This is the measure if all plants die after one period. Let m12 be the
measure computed according to X and U on the assumption that all updating results
in death and that all plants die after two periods. Similarly, define m1i . Define
m1 ¼ limi!1 m1i .
Now, let cjþ1 to be the mass of establishments using technology of age zero in mj ,
so that
ZZZ
1
cjþ1 ðZÞ ¼ cj ðZÞ þ Iðztþ1 2 ZÞUðt; zÞð1 X ðf; t; zÞÞf ðztþ1 jzt Þ dmt dFðfÞ,
M
for i ¼ j ¼ T 0 oT is equivalent to tracking the activities of all plants that are younger
than T 0 , making it simple to isolate different age groups.
These results imply that the following simple algorithm can be used to compute
the stationary equilibrium. Given a wage w, value functions and decision rules can be
computed using standard recursive methods. w is selected so that the free entry
condition is satisfied. With the resulting decision rules, the measure m can be
computed net of a constant which is the level of entry M. This is chosen so as to
equalize labor demand and labor supply. &
References
Bentolila, S., Bertola, G., 1990. Firing costs and labor demand: how bad is eurosclerosis? Review of
Economic Studies 57, 381–402.
Bertsimas, D., Tsitsilkis, J., 1993. Simulated annealing. Statistical Science 8, 10–15.
Cooley, T., Greenwood, J., Yorukoglu, M., 1997. The replacement problem. Journal of Monetary
Economics 40, 457–499.
Cooper, R., Haltiwanger, J., Power, L., 1993. The aggregate implications of machine replacement: theory
and evidence. American Economic Review 83, 360–382.
Cummins, J.G., Violante, G.L., 2002. Equipment-embodied technical change in the US (1947–2000):
measurement and macroeconomic consequences. Review of Economic Dynamics 5, 243–284.
Davis, S.J., Haltiwanger, J., 1992. Gross job creation, gross job destruction, and employment reallocation.
Quarterly Journal of Economics 107, 819–863.
Doms, M., Dunne, T., 1998. Capital adjustment patterns in manufacturing plants. Review of Economic
Dynamics 1, 409–429.
Dunne, T., Roberts, M.J., Samuelson, L., 1989. The growth and failure of U.S. manufacturing plants.
Quarterly Journal of Economics 104, 671–698.
Evans, D.S., 1987. The relationship between firm growth, size, and age: estimates for 100 manufacturing
industries. Journal of Industrial Economics 35, 567–581.
Ford, R., Suyker, W., 1990. Industrial subsidies in the OECD Economies, OECD Economics Department
Working Papers 74.
Fuest, C., Huber, B., 2000. Why do countries subsidize investment and not employment? Journal of Public
Economics 78, 171–192.
Greenwood, J., Hercowitz, Z., Krusell, P., 1997. Long-run implications of investment-specific
technological change. American Economic Review 87, 342–362.
Hansen, G.D., 1985. Indivisible labor and the business cycle. Journal of Monetary Economics 16,
309–327.
Hopenhayn, H.A., 1992. Entry, exit, and firm dynamics in long run equilibrium. Econometrica 60 (5),
1127–1150.
Hopenhayn, H.A., Prescott, E.C., 1992. Stochastic monotonicity and stationary distributions for dynamic
economies. Econometrica 60, 1387–1406.
Hopenhayn, H.A., Rogerson, R., 1993. Job turnover and policy evaluation: a general equilibrium analysis.
Journal of Political Economy 101, 915–938.
King, R.G., Plosser, C.I., Rebelo, S.T., 2002. Production, growth and business cycles: technical appendix.
Computational Economics 20, 87–116.
Krusell, P., 1998. Investment-specific R&D and the decline in the relative price of capital. Journal of
Economic Growth 3, 131–141.
Kydland, F.E., Prescott, E.C., 1982. Time to build and aggregate fluctuations. Econometrica 50,
1345–1370.
Lazear, E.P., 1990. Job security provisions and employment. Quarterly Journal of Economics 105,
699–726.
ARTICLE IN PRESS
1614 R.M. Samaniego / Journal of Economic Dynamics & Control 30 (2006) 1589–1614
Leonard, J.S., Van Audenrode, M., 1993. Corporatism run amok: job stability and industrial policy in
Belgium and the United States. Economic Policy 3, 356–400.
Milgrom, P., Roberts, J., 1990. The economics of modern manufacturing: technology, strategy and
growth. American Economic Review 80, 511–528.
Murphy, M., Pretschker, U., 1998. Spotlight on Public Support to Industry. OECD, Paris.
OECD, 1996. Public Support to Industry: Report to the Industry Committee to the Council at the
Ministerial Level. OECD, Paris.
Restuccia, D., Rogerson, R., 2004. Policy Distortions and Aggregate Productivity with Heterogeneous
Plants. University of Toronto, Manuscript.
Rogerson, R., 1988. Indivisible labor, lotteries, and equilibrium. Journal of Monetary Economics 21, 3–16.
Rogerson, R., 2002. The Employment of Nations. Arizona State University, Manuscript.
Sakellaris, P., 2004. Patterns of plant adjustment. Journal of Monetary Economics 51, 425–450.
Samaniego, R.M., 2006a. Do firing costs affect the incidence of firm bankruptcy? Macroeconomic
Dynamics, 10 (4), forthcoming.
Samaniego, R.M., 2006b. Employment protection and high-tech aversion. Review of Economic Dynamics
9 (2), 224–241.